xref: /freebsd/sys/kern/vfs_syscalls.c (revision a2aef24aa3c8458e4036735dd6928b4ef77294e5)
1 /*-
2  * Copyright (c) 1989, 1993
3  *	The Regents of the University of California.  All rights reserved.
4  * (c) UNIX System Laboratories, Inc.
5  * All or some portions of this file are derived from material licensed
6  * to the University of California by American Telephone and Telegraph
7  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
8  * the permission of UNIX System Laboratories, Inc.
9  *
10  * Redistribution and use in source and binary forms, with or without
11  * modification, are permitted provided that the following conditions
12  * are met:
13  * 1. Redistributions of source code must retain the above copyright
14  *    notice, this list of conditions and the following disclaimer.
15  * 2. Redistributions in binary form must reproduce the above copyright
16  *    notice, this list of conditions and the following disclaimer in the
17  *    documentation and/or other materials provided with the distribution.
18  * 3. 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 kern_chflagsat(struct thread *td, int fd, const char *path,
98     enum uio_seg pathseg, u_long flags, int atflag);
99 static int setfflags(struct thread *td, struct vnode *, u_long);
100 static int getutimes(const struct timeval *, enum uio_seg, struct timespec *);
101 static int getutimens(const struct timespec *, enum uio_seg,
102     struct timespec *, int *);
103 static int setutimes(struct thread *td, struct vnode *,
104     const struct timespec *, int, int);
105 static int vn_access(struct vnode *vp, int user_flags, struct ucred *cred,
106     struct thread *td);
107 
108 /*
109  * Sync each mounted filesystem.
110  */
111 #ifndef _SYS_SYSPROTO_H_
112 struct sync_args {
113 	int     dummy;
114 };
115 #endif
116 /* ARGSUSED */
117 int
118 sys_sync(struct thread *td, struct sync_args *uap)
119 {
120 	struct mount *mp, *nmp;
121 	int save;
122 
123 	mtx_lock(&mountlist_mtx);
124 	for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
125 		if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK)) {
126 			nmp = TAILQ_NEXT(mp, mnt_list);
127 			continue;
128 		}
129 		if ((mp->mnt_flag & MNT_RDONLY) == 0 &&
130 		    vn_start_write(NULL, &mp, V_NOWAIT) == 0) {
131 			save = curthread_pflags_set(TDP_SYNCIO);
132 			vfs_msync(mp, MNT_NOWAIT);
133 			VFS_SYNC(mp, MNT_NOWAIT);
134 			curthread_pflags_restore(save);
135 			vn_finished_write(mp);
136 		}
137 		mtx_lock(&mountlist_mtx);
138 		nmp = TAILQ_NEXT(mp, mnt_list);
139 		vfs_unbusy(mp);
140 	}
141 	mtx_unlock(&mountlist_mtx);
142 	return (0);
143 }
144 
145 /*
146  * Change filesystem quotas.
147  */
148 #ifndef _SYS_SYSPROTO_H_
149 struct quotactl_args {
150 	char *path;
151 	int cmd;
152 	int uid;
153 	caddr_t arg;
154 };
155 #endif
156 int
157 sys_quotactl(struct thread *td, struct quotactl_args *uap)
158 {
159 	struct mount *mp;
160 	struct nameidata nd;
161 	int error;
162 
163 	AUDIT_ARG_CMD(uap->cmd);
164 	AUDIT_ARG_UID(uap->uid);
165 	if (!prison_allow(td->td_ucred, PR_ALLOW_QUOTAS))
166 		return (EPERM);
167 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
168 	    uap->path, td);
169 	if ((error = namei(&nd)) != 0)
170 		return (error);
171 	NDFREE(&nd, NDF_ONLY_PNBUF);
172 	mp = nd.ni_vp->v_mount;
173 	vfs_ref(mp);
174 	vput(nd.ni_vp);
175 	error = vfs_busy(mp, 0);
176 	vfs_rel(mp);
177 	if (error != 0)
178 		return (error);
179 	error = VFS_QUOTACTL(mp, uap->cmd, uap->uid, uap->arg);
180 
181 	/*
182 	 * Since quota on operation typically needs to open quota
183 	 * file, the Q_QUOTAON handler needs to unbusy the mount point
184 	 * before calling into namei.  Otherwise, unmount might be
185 	 * started between two vfs_busy() invocations (first is our,
186 	 * second is from mount point cross-walk code in lookup()),
187 	 * causing deadlock.
188 	 *
189 	 * Require that Q_QUOTAON handles the vfs_busy() reference on
190 	 * its own, always returning with ubusied mount point.
191 	 */
192 	if ((uap->cmd >> SUBCMDSHIFT) != Q_QUOTAON)
193 		vfs_unbusy(mp);
194 	return (error);
195 }
196 
197 /*
198  * Used by statfs conversion routines to scale the block size up if
199  * necessary so that all of the block counts are <= 'max_size'.  Note
200  * that 'max_size' should be a bitmask, i.e. 2^n - 1 for some non-zero
201  * value of 'n'.
202  */
203 void
204 statfs_scale_blocks(struct statfs *sf, long max_size)
205 {
206 	uint64_t count;
207 	int shift;
208 
209 	KASSERT(powerof2(max_size + 1), ("%s: invalid max_size", __func__));
210 
211 	/*
212 	 * Attempt to scale the block counts to give a more accurate
213 	 * overview to userland of the ratio of free space to used
214 	 * space.  To do this, find the largest block count and compute
215 	 * a divisor that lets it fit into a signed integer <= max_size.
216 	 */
217 	if (sf->f_bavail < 0)
218 		count = -sf->f_bavail;
219 	else
220 		count = sf->f_bavail;
221 	count = MAX(sf->f_blocks, MAX(sf->f_bfree, count));
222 	if (count <= max_size)
223 		return;
224 
225 	count >>= flsl(max_size);
226 	shift = 0;
227 	while (count > 0) {
228 		shift++;
229 		count >>=1;
230 	}
231 
232 	sf->f_bsize <<= shift;
233 	sf->f_blocks >>= shift;
234 	sf->f_bfree >>= shift;
235 	sf->f_bavail >>= shift;
236 }
237 
238 static int
239 kern_do_statfs(struct thread *td, struct mount *mp, struct statfs *buf)
240 {
241 	struct statfs *sp;
242 	int error;
243 
244 	if (mp == NULL)
245 		return (EBADF);
246 	error = vfs_busy(mp, 0);
247 	vfs_rel(mp);
248 	if (error != 0)
249 		return (error);
250 #ifdef MAC
251 	error = mac_mount_check_stat(td->td_ucred, mp);
252 	if (error != 0)
253 		goto out;
254 #endif
255 	/*
256 	 * Set these in case the underlying filesystem fails to do so.
257 	 */
258 	sp = &mp->mnt_stat;
259 	sp->f_version = STATFS_VERSION;
260 	sp->f_namemax = NAME_MAX;
261 	sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
262 	error = VFS_STATFS(mp, sp);
263 	if (error != 0)
264 		goto out;
265 	*buf = *sp;
266 	if (priv_check(td, PRIV_VFS_GENERATION)) {
267 		buf->f_fsid.val[0] = buf->f_fsid.val[1] = 0;
268 		prison_enforce_statfs(td->td_ucred, mp, buf);
269 	}
270 out:
271 	vfs_unbusy(mp);
272 	return (error);
273 }
274 
275 /*
276  * Get filesystem statistics.
277  */
278 #ifndef _SYS_SYSPROTO_H_
279 struct statfs_args {
280 	char *path;
281 	struct statfs *buf;
282 };
283 #endif
284 int
285 sys_statfs(struct thread *td, struct statfs_args *uap)
286 {
287 	struct statfs *sfp;
288 	int error;
289 
290 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
291 	error = kern_statfs(td, uap->path, UIO_USERSPACE, sfp);
292 	if (error == 0)
293 		error = copyout(sfp, uap->buf, sizeof(struct statfs));
294 	free(sfp, M_STATFS);
295 	return (error);
296 }
297 
298 int
299 kern_statfs(struct thread *td, char *path, enum uio_seg pathseg,
300     struct statfs *buf)
301 {
302 	struct mount *mp;
303 	struct nameidata nd;
304 	int error;
305 
306 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
307 	    pathseg, path, td);
308 	error = namei(&nd);
309 	if (error != 0)
310 		return (error);
311 	mp = nd.ni_vp->v_mount;
312 	vfs_ref(mp);
313 	NDFREE(&nd, NDF_ONLY_PNBUF);
314 	vput(nd.ni_vp);
315 	return (kern_do_statfs(td, mp, buf));
316 }
317 
318 /*
319  * Get filesystem statistics.
320  */
321 #ifndef _SYS_SYSPROTO_H_
322 struct fstatfs_args {
323 	int fd;
324 	struct statfs *buf;
325 };
326 #endif
327 int
328 sys_fstatfs(struct thread *td, struct fstatfs_args *uap)
329 {
330 	struct statfs *sfp;
331 	int error;
332 
333 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
334 	error = kern_fstatfs(td, uap->fd, sfp);
335 	if (error == 0)
336 		error = copyout(sfp, uap->buf, sizeof(struct statfs));
337 	free(sfp, M_STATFS);
338 	return (error);
339 }
340 
341 int
342 kern_fstatfs(struct thread *td, int fd, struct statfs *buf)
343 {
344 	struct file *fp;
345 	struct mount *mp;
346 	struct vnode *vp;
347 	cap_rights_t rights;
348 	int error;
349 
350 	AUDIT_ARG_FD(fd);
351 	error = getvnode(td, fd, cap_rights_init(&rights, CAP_FSTATFS), &fp);
352 	if (error != 0)
353 		return (error);
354 	vp = fp->f_vnode;
355 	vn_lock(vp, LK_SHARED | LK_RETRY);
356 #ifdef AUDIT
357 	AUDIT_ARG_VNODE1(vp);
358 #endif
359 	mp = vp->v_mount;
360 	if (mp != NULL)
361 		vfs_ref(mp);
362 	VOP_UNLOCK(vp, 0);
363 	fdrop(fp, td);
364 	return (kern_do_statfs(td, mp, buf));
365 }
366 
367 /*
368  * Get statistics on all filesystems.
369  */
370 #ifndef _SYS_SYSPROTO_H_
371 struct getfsstat_args {
372 	struct statfs *buf;
373 	long bufsize;
374 	int mode;
375 };
376 #endif
377 int
378 sys_getfsstat(struct thread *td, struct getfsstat_args *uap)
379 {
380 	size_t count;
381 	int error;
382 
383 	if (uap->bufsize < 0 || uap->bufsize > SIZE_MAX)
384 		return (EINVAL);
385 	error = kern_getfsstat(td, &uap->buf, uap->bufsize, &count,
386 	    UIO_USERSPACE, uap->mode);
387 	if (error == 0)
388 		td->td_retval[0] = count;
389 	return (error);
390 }
391 
392 /*
393  * If (bufsize > 0 && bufseg == UIO_SYSSPACE)
394  *	The caller is responsible for freeing memory which will be allocated
395  *	in '*buf'.
396  */
397 int
398 kern_getfsstat(struct thread *td, struct statfs **buf, size_t bufsize,
399     size_t *countp, enum uio_seg bufseg, int mode)
400 {
401 	struct mount *mp, *nmp;
402 	struct statfs *sfsp, *sp, *sptmp, *tofree;
403 	size_t count, maxcount;
404 	int error;
405 
406 	switch (mode) {
407 	case MNT_WAIT:
408 	case MNT_NOWAIT:
409 		break;
410 	default:
411 		if (bufseg == UIO_SYSSPACE)
412 			*buf = NULL;
413 		return (EINVAL);
414 	}
415 restart:
416 	maxcount = bufsize / sizeof(struct statfs);
417 	if (bufsize == 0) {
418 		sfsp = NULL;
419 		tofree = NULL;
420 	} else if (bufseg == UIO_USERSPACE) {
421 		sfsp = *buf;
422 		tofree = NULL;
423 	} else /* if (bufseg == UIO_SYSSPACE) */ {
424 		count = 0;
425 		mtx_lock(&mountlist_mtx);
426 		TAILQ_FOREACH(mp, &mountlist, mnt_list) {
427 			count++;
428 		}
429 		mtx_unlock(&mountlist_mtx);
430 		if (maxcount > count)
431 			maxcount = count;
432 		tofree = sfsp = *buf = malloc(maxcount * sizeof(struct statfs),
433 		    M_STATFS, M_WAITOK);
434 	}
435 	count = 0;
436 	mtx_lock(&mountlist_mtx);
437 	for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
438 		if (prison_canseemount(td->td_ucred, mp) != 0) {
439 			nmp = TAILQ_NEXT(mp, mnt_list);
440 			continue;
441 		}
442 #ifdef MAC
443 		if (mac_mount_check_stat(td->td_ucred, mp) != 0) {
444 			nmp = TAILQ_NEXT(mp, mnt_list);
445 			continue;
446 		}
447 #endif
448 		if (mode == MNT_WAIT) {
449 			if (vfs_busy(mp, MBF_MNTLSTLOCK) != 0) {
450 				/*
451 				 * If vfs_busy() failed, and MBF_NOWAIT
452 				 * wasn't passed, then the mp is gone.
453 				 * Furthermore, because of MBF_MNTLSTLOCK,
454 				 * the mountlist_mtx was dropped.  We have
455 				 * no other choice than to start over.
456 				 */
457 				mtx_unlock(&mountlist_mtx);
458 				free(tofree, M_STATFS);
459 				goto restart;
460 			}
461 		} else {
462 			if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK) != 0) {
463 				nmp = TAILQ_NEXT(mp, mnt_list);
464 				continue;
465 			}
466 		}
467 		if (sfsp != NULL && count < maxcount) {
468 			sp = &mp->mnt_stat;
469 			/*
470 			 * Set these in case the underlying filesystem
471 			 * fails to do so.
472 			 */
473 			sp->f_version = STATFS_VERSION;
474 			sp->f_namemax = NAME_MAX;
475 			sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
476 			/*
477 			 * If MNT_NOWAIT is specified, do not refresh
478 			 * the fsstat cache.
479 			 */
480 			if (mode != MNT_NOWAIT) {
481 				error = VFS_STATFS(mp, sp);
482 				if (error != 0) {
483 					mtx_lock(&mountlist_mtx);
484 					nmp = TAILQ_NEXT(mp, mnt_list);
485 					vfs_unbusy(mp);
486 					continue;
487 				}
488 			}
489 			if (priv_check(td, PRIV_VFS_GENERATION)) {
490 				sptmp = malloc(sizeof(struct statfs), M_STATFS,
491 				    M_WAITOK);
492 				*sptmp = *sp;
493 				sptmp->f_fsid.val[0] = sptmp->f_fsid.val[1] = 0;
494 				prison_enforce_statfs(td->td_ucred, mp, sptmp);
495 				sp = sptmp;
496 			} else
497 				sptmp = NULL;
498 			if (bufseg == UIO_SYSSPACE) {
499 				bcopy(sp, sfsp, sizeof(*sp));
500 				free(sptmp, M_STATFS);
501 			} else /* if (bufseg == UIO_USERSPACE) */ {
502 				error = copyout(sp, sfsp, sizeof(*sp));
503 				free(sptmp, M_STATFS);
504 				if (error != 0) {
505 					vfs_unbusy(mp);
506 					return (error);
507 				}
508 			}
509 			sfsp++;
510 		}
511 		count++;
512 		mtx_lock(&mountlist_mtx);
513 		nmp = TAILQ_NEXT(mp, mnt_list);
514 		vfs_unbusy(mp);
515 	}
516 	mtx_unlock(&mountlist_mtx);
517 	if (sfsp != NULL && count > maxcount)
518 		*countp = maxcount;
519 	else
520 		*countp = count;
521 	return (0);
522 }
523 
524 #ifdef COMPAT_FREEBSD4
525 /*
526  * Get old format filesystem statistics.
527  */
528 static void freebsd4_cvtstatfs(struct statfs *, struct ostatfs *);
529 
530 #ifndef _SYS_SYSPROTO_H_
531 struct freebsd4_statfs_args {
532 	char *path;
533 	struct ostatfs *buf;
534 };
535 #endif
536 int
537 freebsd4_statfs(struct thread *td, struct freebsd4_statfs_args *uap)
538 {
539 	struct ostatfs osb;
540 	struct statfs *sfp;
541 	int error;
542 
543 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
544 	error = kern_statfs(td, uap->path, UIO_USERSPACE, sfp);
545 	if (error == 0) {
546 		freebsd4_cvtstatfs(sfp, &osb);
547 		error = copyout(&osb, uap->buf, sizeof(osb));
548 	}
549 	free(sfp, M_STATFS);
550 	return (error);
551 }
552 
553 /*
554  * Get filesystem statistics.
555  */
556 #ifndef _SYS_SYSPROTO_H_
557 struct freebsd4_fstatfs_args {
558 	int fd;
559 	struct ostatfs *buf;
560 };
561 #endif
562 int
563 freebsd4_fstatfs(struct thread *td, struct freebsd4_fstatfs_args *uap)
564 {
565 	struct ostatfs osb;
566 	struct statfs *sfp;
567 	int error;
568 
569 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
570 	error = kern_fstatfs(td, uap->fd, sfp);
571 	if (error == 0) {
572 		freebsd4_cvtstatfs(sfp, &osb);
573 		error = copyout(&osb, uap->buf, sizeof(osb));
574 	}
575 	free(sfp, M_STATFS);
576 	return (error);
577 }
578 
579 /*
580  * Get statistics on all filesystems.
581  */
582 #ifndef _SYS_SYSPROTO_H_
583 struct freebsd4_getfsstat_args {
584 	struct ostatfs *buf;
585 	long bufsize;
586 	int mode;
587 };
588 #endif
589 int
590 freebsd4_getfsstat(struct thread *td, struct freebsd4_getfsstat_args *uap)
591 {
592 	struct statfs *buf, *sp;
593 	struct ostatfs osb;
594 	size_t count, size;
595 	int error;
596 
597 	if (uap->bufsize < 0)
598 		return (EINVAL);
599 	count = uap->bufsize / sizeof(struct ostatfs);
600 	if (count > SIZE_MAX / sizeof(struct statfs))
601 		return (EINVAL);
602 	size = count * sizeof(struct statfs);
603 	error = kern_getfsstat(td, &buf, size, &count, UIO_SYSSPACE,
604 	    uap->mode);
605 	if (error == 0)
606 		td->td_retval[0] = count;
607 	if (size != 0) {
608 		sp = buf;
609 		while (count != 0 && error == 0) {
610 			freebsd4_cvtstatfs(sp, &osb);
611 			error = copyout(&osb, uap->buf, sizeof(osb));
612 			sp++;
613 			uap->buf++;
614 			count--;
615 		}
616 		free(buf, M_STATFS);
617 	}
618 	return (error);
619 }
620 
621 /*
622  * Implement fstatfs() for (NFS) file handles.
623  */
624 #ifndef _SYS_SYSPROTO_H_
625 struct freebsd4_fhstatfs_args {
626 	struct fhandle *u_fhp;
627 	struct ostatfs *buf;
628 };
629 #endif
630 int
631 freebsd4_fhstatfs(struct thread *td, struct freebsd4_fhstatfs_args *uap)
632 {
633 	struct ostatfs osb;
634 	struct statfs *sfp;
635 	fhandle_t fh;
636 	int error;
637 
638 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
639 	if (error != 0)
640 		return (error);
641 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
642 	error = kern_fhstatfs(td, fh, sfp);
643 	if (error == 0) {
644 		freebsd4_cvtstatfs(sfp, &osb);
645 		error = copyout(&osb, uap->buf, sizeof(osb));
646 	}
647 	free(sfp, M_STATFS);
648 	return (error);
649 }
650 
651 /*
652  * Convert a new format statfs structure to an old format statfs structure.
653  */
654 static void
655 freebsd4_cvtstatfs(struct statfs *nsp, struct ostatfs *osp)
656 {
657 
658 	statfs_scale_blocks(nsp, LONG_MAX);
659 	bzero(osp, sizeof(*osp));
660 	osp->f_bsize = nsp->f_bsize;
661 	osp->f_iosize = MIN(nsp->f_iosize, LONG_MAX);
662 	osp->f_blocks = nsp->f_blocks;
663 	osp->f_bfree = nsp->f_bfree;
664 	osp->f_bavail = nsp->f_bavail;
665 	osp->f_files = MIN(nsp->f_files, LONG_MAX);
666 	osp->f_ffree = MIN(nsp->f_ffree, LONG_MAX);
667 	osp->f_owner = nsp->f_owner;
668 	osp->f_type = nsp->f_type;
669 	osp->f_flags = nsp->f_flags;
670 	osp->f_syncwrites = MIN(nsp->f_syncwrites, LONG_MAX);
671 	osp->f_asyncwrites = MIN(nsp->f_asyncwrites, LONG_MAX);
672 	osp->f_syncreads = MIN(nsp->f_syncreads, LONG_MAX);
673 	osp->f_asyncreads = MIN(nsp->f_asyncreads, LONG_MAX);
674 	strlcpy(osp->f_fstypename, nsp->f_fstypename,
675 	    MIN(MFSNAMELEN, OMFSNAMELEN));
676 	strlcpy(osp->f_mntonname, nsp->f_mntonname,
677 	    MIN(MNAMELEN, OMNAMELEN));
678 	strlcpy(osp->f_mntfromname, nsp->f_mntfromname,
679 	    MIN(MNAMELEN, OMNAMELEN));
680 	osp->f_fsid = nsp->f_fsid;
681 }
682 #endif /* COMPAT_FREEBSD4 */
683 
684 #if defined(COMPAT_FREEBSD11)
685 /*
686  * Get old format filesystem statistics.
687  */
688 static void freebsd11_cvtstatfs(struct statfs *, struct freebsd11_statfs *);
689 
690 int
691 freebsd11_statfs(struct thread *td, struct freebsd11_statfs_args *uap)
692 {
693 	struct freebsd11_statfs osb;
694 	struct statfs *sfp;
695 	int error;
696 
697 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
698 	error = kern_statfs(td, uap->path, UIO_USERSPACE, sfp);
699 	if (error == 0) {
700 		freebsd11_cvtstatfs(sfp, &osb);
701 		error = copyout(&osb, uap->buf, sizeof(osb));
702 	}
703 	free(sfp, M_STATFS);
704 	return (error);
705 }
706 
707 /*
708  * Get filesystem statistics.
709  */
710 int
711 freebsd11_fstatfs(struct thread *td, struct freebsd11_fstatfs_args *uap)
712 {
713 	struct freebsd11_statfs osb;
714 	struct statfs *sfp;
715 	int error;
716 
717 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
718 	error = kern_fstatfs(td, uap->fd, sfp);
719 	if (error == 0) {
720 		freebsd11_cvtstatfs(sfp, &osb);
721 		error = copyout(&osb, uap->buf, sizeof(osb));
722 	}
723 	free(sfp, M_STATFS);
724 	return (error);
725 }
726 
727 /*
728  * Get statistics on all filesystems.
729  */
730 int
731 freebsd11_getfsstat(struct thread *td, struct freebsd11_getfsstat_args *uap)
732 {
733 	struct freebsd11_statfs osb;
734 	struct statfs *buf, *sp;
735 	size_t count, size;
736 	int error;
737 
738 	count = uap->bufsize / sizeof(struct ostatfs);
739 	size = count * sizeof(struct statfs);
740 	error = kern_getfsstat(td, &buf, size, &count, UIO_SYSSPACE,
741 	    uap->mode);
742 	if (error == 0)
743 		td->td_retval[0] = count;
744 	if (size > 0) {
745 		sp = buf;
746 		while (count > 0 && error == 0) {
747 			freebsd11_cvtstatfs(sp, &osb);
748 			error = copyout(&osb, uap->buf, sizeof(osb));
749 			sp++;
750 			uap->buf++;
751 			count--;
752 		}
753 		free(buf, M_STATFS);
754 	}
755 	return (error);
756 }
757 
758 /*
759  * Implement fstatfs() for (NFS) file handles.
760  */
761 int
762 freebsd11_fhstatfs(struct thread *td, struct freebsd11_fhstatfs_args *uap)
763 {
764 	struct freebsd11_statfs osb;
765 	struct statfs *sfp;
766 	fhandle_t fh;
767 	int error;
768 
769 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
770 	if (error)
771 		return (error);
772 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
773 	error = kern_fhstatfs(td, fh, sfp);
774 	if (error == 0) {
775 		freebsd11_cvtstatfs(sfp, &osb);
776 		error = copyout(&osb, uap->buf, sizeof(osb));
777 	}
778 	free(sfp, M_STATFS);
779 	return (error);
780 }
781 
782 /*
783  * Convert a new format statfs structure to an old format statfs structure.
784  */
785 static void
786 freebsd11_cvtstatfs(struct statfs *nsp, struct freebsd11_statfs *osp)
787 {
788 
789 	bzero(osp, sizeof(*osp));
790 	osp->f_version = FREEBSD11_STATFS_VERSION;
791 	osp->f_type = nsp->f_type;
792 	osp->f_flags = nsp->f_flags;
793 	osp->f_bsize = nsp->f_bsize;
794 	osp->f_iosize = nsp->f_iosize;
795 	osp->f_blocks = nsp->f_blocks;
796 	osp->f_bfree = nsp->f_bfree;
797 	osp->f_bavail = nsp->f_bavail;
798 	osp->f_files = nsp->f_files;
799 	osp->f_ffree = nsp->f_ffree;
800 	osp->f_syncwrites = nsp->f_syncwrites;
801 	osp->f_asyncwrites = nsp->f_asyncwrites;
802 	osp->f_syncreads = nsp->f_syncreads;
803 	osp->f_asyncreads = nsp->f_asyncreads;
804 	osp->f_namemax = nsp->f_namemax;
805 	osp->f_owner = nsp->f_owner;
806 	osp->f_fsid = nsp->f_fsid;
807 	strlcpy(osp->f_fstypename, nsp->f_fstypename,
808 	    MIN(MFSNAMELEN, sizeof(osp->f_fstypename)));
809 	strlcpy(osp->f_mntonname, nsp->f_mntonname,
810 	    MIN(MNAMELEN, sizeof(osp->f_mntonname)));
811 	strlcpy(osp->f_mntfromname, nsp->f_mntfromname,
812 	    MIN(MNAMELEN, sizeof(osp->f_mntfromname)));
813 }
814 #endif /* COMPAT_FREEBSD11 */
815 
816 /*
817  * Change current working directory to a given file descriptor.
818  */
819 #ifndef _SYS_SYSPROTO_H_
820 struct fchdir_args {
821 	int	fd;
822 };
823 #endif
824 int
825 sys_fchdir(struct thread *td, struct fchdir_args *uap)
826 {
827 	struct vnode *vp, *tdp;
828 	struct mount *mp;
829 	struct file *fp;
830 	cap_rights_t rights;
831 	int error;
832 
833 	AUDIT_ARG_FD(uap->fd);
834 	error = getvnode(td, uap->fd, cap_rights_init(&rights, CAP_FCHDIR),
835 	    &fp);
836 	if (error != 0)
837 		return (error);
838 	vp = fp->f_vnode;
839 	vrefact(vp);
840 	fdrop(fp, td);
841 	vn_lock(vp, LK_SHARED | LK_RETRY);
842 	AUDIT_ARG_VNODE1(vp);
843 	error = change_dir(vp, td);
844 	while (!error && (mp = vp->v_mountedhere) != NULL) {
845 		if (vfs_busy(mp, 0))
846 			continue;
847 		error = VFS_ROOT(mp, LK_SHARED, &tdp);
848 		vfs_unbusy(mp);
849 		if (error != 0)
850 			break;
851 		vput(vp);
852 		vp = tdp;
853 	}
854 	if (error != 0) {
855 		vput(vp);
856 		return (error);
857 	}
858 	VOP_UNLOCK(vp, 0);
859 	pwd_chdir(td, vp);
860 	return (0);
861 }
862 
863 /*
864  * Change current working directory (``.'').
865  */
866 #ifndef _SYS_SYSPROTO_H_
867 struct chdir_args {
868 	char	*path;
869 };
870 #endif
871 int
872 sys_chdir(struct thread *td, struct chdir_args *uap)
873 {
874 
875 	return (kern_chdir(td, uap->path, UIO_USERSPACE));
876 }
877 
878 int
879 kern_chdir(struct thread *td, char *path, enum uio_seg pathseg)
880 {
881 	struct nameidata nd;
882 	int error;
883 
884 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
885 	    pathseg, path, td);
886 	if ((error = namei(&nd)) != 0)
887 		return (error);
888 	if ((error = change_dir(nd.ni_vp, td)) != 0) {
889 		vput(nd.ni_vp);
890 		NDFREE(&nd, NDF_ONLY_PNBUF);
891 		return (error);
892 	}
893 	VOP_UNLOCK(nd.ni_vp, 0);
894 	NDFREE(&nd, NDF_ONLY_PNBUF);
895 	pwd_chdir(td, nd.ni_vp);
896 	return (0);
897 }
898 
899 /*
900  * Change notion of root (``/'') directory.
901  */
902 #ifndef _SYS_SYSPROTO_H_
903 struct chroot_args {
904 	char	*path;
905 };
906 #endif
907 int
908 sys_chroot(struct thread *td, struct chroot_args *uap)
909 {
910 	struct nameidata nd;
911 	int error;
912 
913 	error = priv_check(td, PRIV_VFS_CHROOT);
914 	if (error != 0)
915 		return (error);
916 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
917 	    UIO_USERSPACE, uap->path, td);
918 	error = namei(&nd);
919 	if (error != 0)
920 		goto error;
921 	error = change_dir(nd.ni_vp, td);
922 	if (error != 0)
923 		goto e_vunlock;
924 #ifdef MAC
925 	error = mac_vnode_check_chroot(td->td_ucred, nd.ni_vp);
926 	if (error != 0)
927 		goto e_vunlock;
928 #endif
929 	VOP_UNLOCK(nd.ni_vp, 0);
930 	error = pwd_chroot(td, nd.ni_vp);
931 	vrele(nd.ni_vp);
932 	NDFREE(&nd, NDF_ONLY_PNBUF);
933 	return (error);
934 e_vunlock:
935 	vput(nd.ni_vp);
936 error:
937 	NDFREE(&nd, NDF_ONLY_PNBUF);
938 	return (error);
939 }
940 
941 /*
942  * Common routine for chroot and chdir.  Callers must provide a locked vnode
943  * instance.
944  */
945 int
946 change_dir(struct vnode *vp, struct thread *td)
947 {
948 #ifdef MAC
949 	int error;
950 #endif
951 
952 	ASSERT_VOP_LOCKED(vp, "change_dir(): vp not locked");
953 	if (vp->v_type != VDIR)
954 		return (ENOTDIR);
955 #ifdef MAC
956 	error = mac_vnode_check_chdir(td->td_ucred, vp);
957 	if (error != 0)
958 		return (error);
959 #endif
960 	return (VOP_ACCESS(vp, VEXEC, td->td_ucred, td));
961 }
962 
963 static __inline void
964 flags_to_rights(int flags, cap_rights_t *rightsp)
965 {
966 
967 	if (flags & O_EXEC) {
968 		cap_rights_set(rightsp, CAP_FEXECVE);
969 	} else {
970 		switch ((flags & O_ACCMODE)) {
971 		case O_RDONLY:
972 			cap_rights_set(rightsp, CAP_READ);
973 			break;
974 		case O_RDWR:
975 			cap_rights_set(rightsp, CAP_READ);
976 			/* FALLTHROUGH */
977 		case O_WRONLY:
978 			cap_rights_set(rightsp, CAP_WRITE);
979 			if (!(flags & (O_APPEND | O_TRUNC)))
980 				cap_rights_set(rightsp, CAP_SEEK);
981 			break;
982 		}
983 	}
984 
985 	if (flags & O_CREAT)
986 		cap_rights_set(rightsp, CAP_CREATE);
987 
988 	if (flags & O_TRUNC)
989 		cap_rights_set(rightsp, CAP_FTRUNCATE);
990 
991 	if (flags & (O_SYNC | O_FSYNC))
992 		cap_rights_set(rightsp, CAP_FSYNC);
993 
994 	if (flags & (O_EXLOCK | O_SHLOCK))
995 		cap_rights_set(rightsp, CAP_FLOCK);
996 }
997 
998 /*
999  * Check permissions, allocate an open file structure, and call the device
1000  * open routine if any.
1001  */
1002 #ifndef _SYS_SYSPROTO_H_
1003 struct open_args {
1004 	char	*path;
1005 	int	flags;
1006 	int	mode;
1007 };
1008 #endif
1009 int
1010 sys_open(struct thread *td, struct open_args *uap)
1011 {
1012 
1013 	return (kern_openat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1014 	    uap->flags, uap->mode));
1015 }
1016 
1017 #ifndef _SYS_SYSPROTO_H_
1018 struct openat_args {
1019 	int	fd;
1020 	char	*path;
1021 	int	flag;
1022 	int	mode;
1023 };
1024 #endif
1025 int
1026 sys_openat(struct thread *td, struct openat_args *uap)
1027 {
1028 
1029 	AUDIT_ARG_FD(uap->fd);
1030 	return (kern_openat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
1031 	    uap->mode));
1032 }
1033 
1034 int
1035 kern_openat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1036     int flags, int mode)
1037 {
1038 	struct proc *p = td->td_proc;
1039 	struct filedesc *fdp = p->p_fd;
1040 	struct file *fp;
1041 	struct vnode *vp;
1042 	struct nameidata nd;
1043 	cap_rights_t rights;
1044 	int cmode, error, indx;
1045 
1046 	indx = -1;
1047 
1048 	AUDIT_ARG_FFLAGS(flags);
1049 	AUDIT_ARG_MODE(mode);
1050 	cap_rights_init(&rights, CAP_LOOKUP);
1051 	flags_to_rights(flags, &rights);
1052 	/*
1053 	 * Only one of the O_EXEC, O_RDONLY, O_WRONLY and O_RDWR flags
1054 	 * may be specified.
1055 	 */
1056 	if (flags & O_EXEC) {
1057 		if (flags & O_ACCMODE)
1058 			return (EINVAL);
1059 	} else if ((flags & O_ACCMODE) == O_ACCMODE) {
1060 		return (EINVAL);
1061 	} else {
1062 		flags = FFLAGS(flags);
1063 	}
1064 
1065 	/*
1066 	 * Allocate a file structure. The descriptor to reference it
1067 	 * is allocated and set by finstall() below.
1068 	 */
1069 	error = falloc_noinstall(td, &fp);
1070 	if (error != 0)
1071 		return (error);
1072 	/*
1073 	 * An extra reference on `fp' has been held for us by
1074 	 * falloc_noinstall().
1075 	 */
1076 	/* Set the flags early so the finit in devfs can pick them up. */
1077 	fp->f_flag = flags & FMASK;
1078 	cmode = ((mode & ~fdp->fd_cmask) & ALLPERMS) & ~S_ISTXT;
1079 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, fd,
1080 	    &rights, td);
1081 	td->td_dupfd = -1;		/* XXX check for fdopen */
1082 	error = vn_open(&nd, &flags, cmode, fp);
1083 	if (error != 0) {
1084 		/*
1085 		 * If the vn_open replaced the method vector, something
1086 		 * wonderous happened deep below and we just pass it up
1087 		 * pretending we know what we do.
1088 		 */
1089 		if (error == ENXIO && fp->f_ops != &badfileops)
1090 			goto success;
1091 
1092 		/*
1093 		 * Handle special fdopen() case. bleh.
1094 		 *
1095 		 * Don't do this for relative (capability) lookups; we don't
1096 		 * understand exactly what would happen, and we don't think
1097 		 * that it ever should.
1098 		 */
1099 		if ((nd.ni_lcf & NI_LCF_STRICTRELATIVE) == 0 &&
1100 		    (error == ENODEV || error == ENXIO) &&
1101 		    td->td_dupfd >= 0) {
1102 			error = dupfdopen(td, fdp, td->td_dupfd, flags, error,
1103 			    &indx);
1104 			if (error == 0)
1105 				goto success;
1106 		}
1107 
1108 		goto bad;
1109 	}
1110 	td->td_dupfd = 0;
1111 	NDFREE(&nd, NDF_ONLY_PNBUF);
1112 	vp = nd.ni_vp;
1113 
1114 	/*
1115 	 * Store the vnode, for any f_type. Typically, the vnode use
1116 	 * count is decremented by direct call to vn_closefile() for
1117 	 * files that switched type in the cdevsw fdopen() method.
1118 	 */
1119 	fp->f_vnode = vp;
1120 	/*
1121 	 * If the file wasn't claimed by devfs bind it to the normal
1122 	 * vnode operations here.
1123 	 */
1124 	if (fp->f_ops == &badfileops) {
1125 		KASSERT(vp->v_type != VFIFO, ("Unexpected fifo."));
1126 		fp->f_seqcount = 1;
1127 		finit(fp, (flags & FMASK) | (fp->f_flag & FHASLOCK),
1128 		    DTYPE_VNODE, vp, &vnops);
1129 	}
1130 
1131 	VOP_UNLOCK(vp, 0);
1132 	if (flags & O_TRUNC) {
1133 		error = fo_truncate(fp, 0, td->td_ucred, td);
1134 		if (error != 0)
1135 			goto bad;
1136 	}
1137 success:
1138 	/*
1139 	 * If we haven't already installed the FD (for dupfdopen), do so now.
1140 	 */
1141 	if (indx == -1) {
1142 		struct filecaps *fcaps;
1143 
1144 #ifdef CAPABILITIES
1145 		if ((nd.ni_lcf & NI_LCF_STRICTRELATIVE) != 0)
1146 			fcaps = &nd.ni_filecaps;
1147 		else
1148 #endif
1149 			fcaps = NULL;
1150 		error = finstall(td, fp, &indx, flags, fcaps);
1151 		/* On success finstall() consumes fcaps. */
1152 		if (error != 0) {
1153 			filecaps_free(&nd.ni_filecaps);
1154 			goto bad;
1155 		}
1156 	} else {
1157 		filecaps_free(&nd.ni_filecaps);
1158 	}
1159 
1160 	/*
1161 	 * Release our private reference, leaving the one associated with
1162 	 * the descriptor table intact.
1163 	 */
1164 	fdrop(fp, td);
1165 	td->td_retval[0] = indx;
1166 	return (0);
1167 bad:
1168 	KASSERT(indx == -1, ("indx=%d, should be -1", indx));
1169 	fdrop(fp, td);
1170 	return (error);
1171 }
1172 
1173 #ifdef COMPAT_43
1174 /*
1175  * Create a file.
1176  */
1177 #ifndef _SYS_SYSPROTO_H_
1178 struct ocreat_args {
1179 	char	*path;
1180 	int	mode;
1181 };
1182 #endif
1183 int
1184 ocreat(struct thread *td, struct ocreat_args *uap)
1185 {
1186 
1187 	return (kern_openat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1188 	    O_WRONLY | O_CREAT | O_TRUNC, uap->mode));
1189 }
1190 #endif /* COMPAT_43 */
1191 
1192 /*
1193  * Create a special file.
1194  */
1195 #ifndef _SYS_SYSPROTO_H_
1196 struct mknodat_args {
1197 	int	fd;
1198 	char	*path;
1199 	mode_t	mode;
1200 	dev_t	dev;
1201 };
1202 #endif
1203 int
1204 sys_mknodat(struct thread *td, struct mknodat_args *uap)
1205 {
1206 
1207 	return (kern_mknodat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode,
1208 	    uap->dev));
1209 }
1210 
1211 #if defined(COMPAT_FREEBSD11)
1212 int
1213 freebsd11_mknod(struct thread *td,
1214     struct freebsd11_mknod_args *uap)
1215 {
1216 
1217 	return (kern_mknodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1218 	    uap->mode, uap->dev));
1219 }
1220 
1221 int
1222 freebsd11_mknodat(struct thread *td,
1223     struct freebsd11_mknodat_args *uap)
1224 {
1225 
1226 	return (kern_mknodat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode,
1227 	    uap->dev));
1228 }
1229 #endif /* COMPAT_FREEBSD11 */
1230 
1231 int
1232 kern_mknodat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1233     int mode, dev_t dev)
1234 {
1235 	struct vnode *vp;
1236 	struct mount *mp;
1237 	struct vattr vattr;
1238 	struct nameidata nd;
1239 	cap_rights_t rights;
1240 	int error, whiteout = 0;
1241 
1242 	AUDIT_ARG_MODE(mode);
1243 	AUDIT_ARG_DEV(dev);
1244 	switch (mode & S_IFMT) {
1245 	case S_IFCHR:
1246 	case S_IFBLK:
1247 		error = priv_check(td, PRIV_VFS_MKNOD_DEV);
1248 		if (error == 0 && dev == VNOVAL)
1249 			error = EINVAL;
1250 		break;
1251 	case S_IFWHT:
1252 		error = priv_check(td, PRIV_VFS_MKNOD_WHT);
1253 		break;
1254 	case S_IFIFO:
1255 		if (dev == 0)
1256 			return (kern_mkfifoat(td, fd, path, pathseg, mode));
1257 		/* FALLTHROUGH */
1258 	default:
1259 		error = EINVAL;
1260 		break;
1261 	}
1262 	if (error != 0)
1263 		return (error);
1264 restart:
1265 	bwillwrite();
1266 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
1267 	    NOCACHE, pathseg, path, fd, cap_rights_init(&rights, CAP_MKNODAT),
1268 	    td);
1269 	if ((error = namei(&nd)) != 0)
1270 		return (error);
1271 	vp = nd.ni_vp;
1272 	if (vp != NULL) {
1273 		NDFREE(&nd, NDF_ONLY_PNBUF);
1274 		if (vp == nd.ni_dvp)
1275 			vrele(nd.ni_dvp);
1276 		else
1277 			vput(nd.ni_dvp);
1278 		vrele(vp);
1279 		return (EEXIST);
1280 	} else {
1281 		VATTR_NULL(&vattr);
1282 		vattr.va_mode = (mode & ALLPERMS) &
1283 		    ~td->td_proc->p_fd->fd_cmask;
1284 		vattr.va_rdev = dev;
1285 		whiteout = 0;
1286 
1287 		switch (mode & S_IFMT) {
1288 		case S_IFCHR:
1289 			vattr.va_type = VCHR;
1290 			break;
1291 		case S_IFBLK:
1292 			vattr.va_type = VBLK;
1293 			break;
1294 		case S_IFWHT:
1295 			whiteout = 1;
1296 			break;
1297 		default:
1298 			panic("kern_mknod: invalid mode");
1299 		}
1300 	}
1301 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1302 		NDFREE(&nd, NDF_ONLY_PNBUF);
1303 		vput(nd.ni_dvp);
1304 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1305 			return (error);
1306 		goto restart;
1307 	}
1308 #ifdef MAC
1309 	if (error == 0 && !whiteout)
1310 		error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp,
1311 		    &nd.ni_cnd, &vattr);
1312 #endif
1313 	if (error == 0) {
1314 		if (whiteout)
1315 			error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE);
1316 		else {
1317 			error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp,
1318 						&nd.ni_cnd, &vattr);
1319 			if (error == 0)
1320 				vput(nd.ni_vp);
1321 		}
1322 	}
1323 	NDFREE(&nd, NDF_ONLY_PNBUF);
1324 	vput(nd.ni_dvp);
1325 	vn_finished_write(mp);
1326 	return (error);
1327 }
1328 
1329 /*
1330  * Create a named pipe.
1331  */
1332 #ifndef _SYS_SYSPROTO_H_
1333 struct mkfifo_args {
1334 	char	*path;
1335 	int	mode;
1336 };
1337 #endif
1338 int
1339 sys_mkfifo(struct thread *td, struct mkfifo_args *uap)
1340 {
1341 
1342 	return (kern_mkfifoat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1343 	    uap->mode));
1344 }
1345 
1346 #ifndef _SYS_SYSPROTO_H_
1347 struct mkfifoat_args {
1348 	int	fd;
1349 	char	*path;
1350 	mode_t	mode;
1351 };
1352 #endif
1353 int
1354 sys_mkfifoat(struct thread *td, struct mkfifoat_args *uap)
1355 {
1356 
1357 	return (kern_mkfifoat(td, uap->fd, uap->path, UIO_USERSPACE,
1358 	    uap->mode));
1359 }
1360 
1361 int
1362 kern_mkfifoat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1363     int mode)
1364 {
1365 	struct mount *mp;
1366 	struct vattr vattr;
1367 	struct nameidata nd;
1368 	cap_rights_t rights;
1369 	int error;
1370 
1371 	AUDIT_ARG_MODE(mode);
1372 restart:
1373 	bwillwrite();
1374 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
1375 	    NOCACHE, pathseg, path, fd, cap_rights_init(&rights, CAP_MKFIFOAT),
1376 	    td);
1377 	if ((error = namei(&nd)) != 0)
1378 		return (error);
1379 	if (nd.ni_vp != NULL) {
1380 		NDFREE(&nd, NDF_ONLY_PNBUF);
1381 		if (nd.ni_vp == nd.ni_dvp)
1382 			vrele(nd.ni_dvp);
1383 		else
1384 			vput(nd.ni_dvp);
1385 		vrele(nd.ni_vp);
1386 		return (EEXIST);
1387 	}
1388 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1389 		NDFREE(&nd, NDF_ONLY_PNBUF);
1390 		vput(nd.ni_dvp);
1391 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1392 			return (error);
1393 		goto restart;
1394 	}
1395 	VATTR_NULL(&vattr);
1396 	vattr.va_type = VFIFO;
1397 	vattr.va_mode = (mode & ALLPERMS) & ~td->td_proc->p_fd->fd_cmask;
1398 #ifdef MAC
1399 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1400 	    &vattr);
1401 	if (error != 0)
1402 		goto out;
1403 #endif
1404 	error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
1405 	if (error == 0)
1406 		vput(nd.ni_vp);
1407 #ifdef MAC
1408 out:
1409 #endif
1410 	vput(nd.ni_dvp);
1411 	vn_finished_write(mp);
1412 	NDFREE(&nd, NDF_ONLY_PNBUF);
1413 	return (error);
1414 }
1415 
1416 /*
1417  * Make a hard file link.
1418  */
1419 #ifndef _SYS_SYSPROTO_H_
1420 struct link_args {
1421 	char	*path;
1422 	char	*link;
1423 };
1424 #endif
1425 int
1426 sys_link(struct thread *td, struct link_args *uap)
1427 {
1428 
1429 	return (kern_linkat(td, AT_FDCWD, AT_FDCWD, uap->path, uap->link,
1430 	    UIO_USERSPACE, FOLLOW));
1431 }
1432 
1433 #ifndef _SYS_SYSPROTO_H_
1434 struct linkat_args {
1435 	int	fd1;
1436 	char	*path1;
1437 	int	fd2;
1438 	char	*path2;
1439 	int	flag;
1440 };
1441 #endif
1442 int
1443 sys_linkat(struct thread *td, struct linkat_args *uap)
1444 {
1445 	int flag;
1446 
1447 	flag = uap->flag;
1448 	if (flag & ~AT_SYMLINK_FOLLOW)
1449 		return (EINVAL);
1450 
1451 	return (kern_linkat(td, uap->fd1, uap->fd2, uap->path1, uap->path2,
1452 	    UIO_USERSPACE, (flag & AT_SYMLINK_FOLLOW) ? FOLLOW : NOFOLLOW));
1453 }
1454 
1455 int hardlink_check_uid = 0;
1456 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_uid, CTLFLAG_RW,
1457     &hardlink_check_uid, 0,
1458     "Unprivileged processes cannot create hard links to files owned by other "
1459     "users");
1460 static int hardlink_check_gid = 0;
1461 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_gid, CTLFLAG_RW,
1462     &hardlink_check_gid, 0,
1463     "Unprivileged processes cannot create hard links to files owned by other "
1464     "groups");
1465 
1466 static int
1467 can_hardlink(struct vnode *vp, struct ucred *cred)
1468 {
1469 	struct vattr va;
1470 	int error;
1471 
1472 	if (!hardlink_check_uid && !hardlink_check_gid)
1473 		return (0);
1474 
1475 	error = VOP_GETATTR(vp, &va, cred);
1476 	if (error != 0)
1477 		return (error);
1478 
1479 	if (hardlink_check_uid && cred->cr_uid != va.va_uid) {
1480 		error = priv_check_cred(cred, PRIV_VFS_LINK, 0);
1481 		if (error != 0)
1482 			return (error);
1483 	}
1484 
1485 	if (hardlink_check_gid && !groupmember(va.va_gid, cred)) {
1486 		error = priv_check_cred(cred, PRIV_VFS_LINK, 0);
1487 		if (error != 0)
1488 			return (error);
1489 	}
1490 
1491 	return (0);
1492 }
1493 
1494 int
1495 kern_linkat(struct thread *td, int fd1, int fd2, char *path1, char *path2,
1496     enum uio_seg segflg, int follow)
1497 {
1498 	struct vnode *vp;
1499 	struct mount *mp;
1500 	struct nameidata nd;
1501 	cap_rights_t rights;
1502 	int error;
1503 
1504 again:
1505 	bwillwrite();
1506 	NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, segflg, path1, fd1,
1507 	    cap_rights_init(&rights, CAP_LINKAT_SOURCE), td);
1508 
1509 	if ((error = namei(&nd)) != 0)
1510 		return (error);
1511 	NDFREE(&nd, NDF_ONLY_PNBUF);
1512 	vp = nd.ni_vp;
1513 	if (vp->v_type == VDIR) {
1514 		vrele(vp);
1515 		return (EPERM);		/* POSIX */
1516 	}
1517 	NDINIT_ATRIGHTS(&nd, CREATE,
1518 	    LOCKPARENT | SAVENAME | AUDITVNODE2 | NOCACHE, segflg, path2, fd2,
1519 	    cap_rights_init(&rights, CAP_LINKAT_TARGET), td);
1520 	if ((error = namei(&nd)) == 0) {
1521 		if (nd.ni_vp != NULL) {
1522 			NDFREE(&nd, NDF_ONLY_PNBUF);
1523 			if (nd.ni_dvp == nd.ni_vp)
1524 				vrele(nd.ni_dvp);
1525 			else
1526 				vput(nd.ni_dvp);
1527 			vrele(nd.ni_vp);
1528 			vrele(vp);
1529 			return (EEXIST);
1530 		} else if (nd.ni_dvp->v_mount != vp->v_mount) {
1531 			/*
1532 			 * Cross-device link.  No need to recheck
1533 			 * vp->v_type, since it cannot change, except
1534 			 * to VBAD.
1535 			 */
1536 			NDFREE(&nd, NDF_ONLY_PNBUF);
1537 			vput(nd.ni_dvp);
1538 			vrele(vp);
1539 			return (EXDEV);
1540 		} else if ((error = vn_lock(vp, LK_EXCLUSIVE)) == 0) {
1541 			error = can_hardlink(vp, td->td_ucred);
1542 #ifdef MAC
1543 			if (error == 0)
1544 				error = mac_vnode_check_link(td->td_ucred,
1545 				    nd.ni_dvp, vp, &nd.ni_cnd);
1546 #endif
1547 			if (error != 0) {
1548 				vput(vp);
1549 				vput(nd.ni_dvp);
1550 				NDFREE(&nd, NDF_ONLY_PNBUF);
1551 				return (error);
1552 			}
1553 			error = vn_start_write(vp, &mp, V_NOWAIT);
1554 			if (error != 0) {
1555 				vput(vp);
1556 				vput(nd.ni_dvp);
1557 				NDFREE(&nd, NDF_ONLY_PNBUF);
1558 				error = vn_start_write(NULL, &mp,
1559 				    V_XSLEEP | PCATCH);
1560 				if (error != 0)
1561 					return (error);
1562 				goto again;
1563 			}
1564 			error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
1565 			VOP_UNLOCK(vp, 0);
1566 			vput(nd.ni_dvp);
1567 			vn_finished_write(mp);
1568 			NDFREE(&nd, NDF_ONLY_PNBUF);
1569 		} else {
1570 			vput(nd.ni_dvp);
1571 			NDFREE(&nd, NDF_ONLY_PNBUF);
1572 			vrele(vp);
1573 			goto again;
1574 		}
1575 	}
1576 	vrele(vp);
1577 	return (error);
1578 }
1579 
1580 /*
1581  * Make a symbolic link.
1582  */
1583 #ifndef _SYS_SYSPROTO_H_
1584 struct symlink_args {
1585 	char	*path;
1586 	char	*link;
1587 };
1588 #endif
1589 int
1590 sys_symlink(struct thread *td, struct symlink_args *uap)
1591 {
1592 
1593 	return (kern_symlinkat(td, uap->path, AT_FDCWD, uap->link,
1594 	    UIO_USERSPACE));
1595 }
1596 
1597 #ifndef _SYS_SYSPROTO_H_
1598 struct symlinkat_args {
1599 	char	*path;
1600 	int	fd;
1601 	char	*path2;
1602 };
1603 #endif
1604 int
1605 sys_symlinkat(struct thread *td, struct symlinkat_args *uap)
1606 {
1607 
1608 	return (kern_symlinkat(td, uap->path1, uap->fd, uap->path2,
1609 	    UIO_USERSPACE));
1610 }
1611 
1612 int
1613 kern_symlinkat(struct thread *td, char *path1, int fd, char *path2,
1614     enum uio_seg segflg)
1615 {
1616 	struct mount *mp;
1617 	struct vattr vattr;
1618 	char *syspath;
1619 	struct nameidata nd;
1620 	int error;
1621 	cap_rights_t rights;
1622 
1623 	if (segflg == UIO_SYSSPACE) {
1624 		syspath = path1;
1625 	} else {
1626 		syspath = uma_zalloc(namei_zone, M_WAITOK);
1627 		if ((error = copyinstr(path1, syspath, MAXPATHLEN, NULL)) != 0)
1628 			goto out;
1629 	}
1630 	AUDIT_ARG_TEXT(syspath);
1631 restart:
1632 	bwillwrite();
1633 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
1634 	    NOCACHE, segflg, path2, fd, cap_rights_init(&rights, CAP_SYMLINKAT),
1635 	    td);
1636 	if ((error = namei(&nd)) != 0)
1637 		goto out;
1638 	if (nd.ni_vp) {
1639 		NDFREE(&nd, NDF_ONLY_PNBUF);
1640 		if (nd.ni_vp == nd.ni_dvp)
1641 			vrele(nd.ni_dvp);
1642 		else
1643 			vput(nd.ni_dvp);
1644 		vrele(nd.ni_vp);
1645 		error = EEXIST;
1646 		goto out;
1647 	}
1648 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1649 		NDFREE(&nd, NDF_ONLY_PNBUF);
1650 		vput(nd.ni_dvp);
1651 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1652 			goto out;
1653 		goto restart;
1654 	}
1655 	VATTR_NULL(&vattr);
1656 	vattr.va_mode = ACCESSPERMS &~ td->td_proc->p_fd->fd_cmask;
1657 #ifdef MAC
1658 	vattr.va_type = VLNK;
1659 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1660 	    &vattr);
1661 	if (error != 0)
1662 		goto out2;
1663 #endif
1664 	error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, syspath);
1665 	if (error == 0)
1666 		vput(nd.ni_vp);
1667 #ifdef MAC
1668 out2:
1669 #endif
1670 	NDFREE(&nd, NDF_ONLY_PNBUF);
1671 	vput(nd.ni_dvp);
1672 	vn_finished_write(mp);
1673 out:
1674 	if (segflg != UIO_SYSSPACE)
1675 		uma_zfree(namei_zone, syspath);
1676 	return (error);
1677 }
1678 
1679 /*
1680  * Delete a whiteout from the filesystem.
1681  */
1682 #ifndef _SYS_SYSPROTO_H_
1683 struct undelete_args {
1684 	char *path;
1685 };
1686 #endif
1687 int
1688 sys_undelete(struct thread *td, struct undelete_args *uap)
1689 {
1690 	struct mount *mp;
1691 	struct nameidata nd;
1692 	int error;
1693 
1694 restart:
1695 	bwillwrite();
1696 	NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | AUDITVNODE1,
1697 	    UIO_USERSPACE, uap->path, td);
1698 	error = namei(&nd);
1699 	if (error != 0)
1700 		return (error);
1701 
1702 	if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) {
1703 		NDFREE(&nd, NDF_ONLY_PNBUF);
1704 		if (nd.ni_vp == nd.ni_dvp)
1705 			vrele(nd.ni_dvp);
1706 		else
1707 			vput(nd.ni_dvp);
1708 		if (nd.ni_vp)
1709 			vrele(nd.ni_vp);
1710 		return (EEXIST);
1711 	}
1712 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1713 		NDFREE(&nd, NDF_ONLY_PNBUF);
1714 		vput(nd.ni_dvp);
1715 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1716 			return (error);
1717 		goto restart;
1718 	}
1719 	error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE);
1720 	NDFREE(&nd, NDF_ONLY_PNBUF);
1721 	vput(nd.ni_dvp);
1722 	vn_finished_write(mp);
1723 	return (error);
1724 }
1725 
1726 /*
1727  * Delete a name from the filesystem.
1728  */
1729 #ifndef _SYS_SYSPROTO_H_
1730 struct unlink_args {
1731 	char	*path;
1732 };
1733 #endif
1734 int
1735 sys_unlink(struct thread *td, struct unlink_args *uap)
1736 {
1737 
1738 	return (kern_unlinkat(td, AT_FDCWD, uap->path, UIO_USERSPACE, 0));
1739 }
1740 
1741 #ifndef _SYS_SYSPROTO_H_
1742 struct unlinkat_args {
1743 	int	fd;
1744 	char	*path;
1745 	int	flag;
1746 };
1747 #endif
1748 int
1749 sys_unlinkat(struct thread *td, struct unlinkat_args *uap)
1750 {
1751 	int flag = uap->flag;
1752 	int fd = uap->fd;
1753 	char *path = uap->path;
1754 
1755 	if (flag & ~AT_REMOVEDIR)
1756 		return (EINVAL);
1757 
1758 	if (flag & AT_REMOVEDIR)
1759 		return (kern_rmdirat(td, fd, path, UIO_USERSPACE));
1760 	else
1761 		return (kern_unlinkat(td, fd, path, UIO_USERSPACE, 0));
1762 }
1763 
1764 int
1765 kern_unlinkat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1766     ino_t oldinum)
1767 {
1768 	struct mount *mp;
1769 	struct vnode *vp;
1770 	struct nameidata nd;
1771 	struct stat sb;
1772 	cap_rights_t rights;
1773 	int error;
1774 
1775 restart:
1776 	bwillwrite();
1777 	NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1,
1778 	    pathseg, path, fd, cap_rights_init(&rights, CAP_UNLINKAT), td);
1779 	if ((error = namei(&nd)) != 0)
1780 		return (error == EINVAL ? EPERM : error);
1781 	vp = nd.ni_vp;
1782 	if (vp->v_type == VDIR && oldinum == 0) {
1783 		error = EPERM;		/* POSIX */
1784 	} else if (oldinum != 0 &&
1785 		  ((error = vn_stat(vp, &sb, td->td_ucred, NOCRED, td)) == 0) &&
1786 		  sb.st_ino != oldinum) {
1787 			error = EIDRM;	/* Identifier removed */
1788 	} else {
1789 		/*
1790 		 * The root of a mounted filesystem cannot be deleted.
1791 		 *
1792 		 * XXX: can this only be a VDIR case?
1793 		 */
1794 		if (vp->v_vflag & VV_ROOT)
1795 			error = EBUSY;
1796 	}
1797 	if (error == 0) {
1798 		if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1799 			NDFREE(&nd, NDF_ONLY_PNBUF);
1800 			vput(nd.ni_dvp);
1801 			if (vp == nd.ni_dvp)
1802 				vrele(vp);
1803 			else
1804 				vput(vp);
1805 			if ((error = vn_start_write(NULL, &mp,
1806 			    V_XSLEEP | PCATCH)) != 0)
1807 				return (error);
1808 			goto restart;
1809 		}
1810 #ifdef MAC
1811 		error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
1812 		    &nd.ni_cnd);
1813 		if (error != 0)
1814 			goto out;
1815 #endif
1816 		vfs_notify_upper(vp, VFS_NOTIFY_UPPER_UNLINK);
1817 		error = VOP_REMOVE(nd.ni_dvp, vp, &nd.ni_cnd);
1818 #ifdef MAC
1819 out:
1820 #endif
1821 		vn_finished_write(mp);
1822 	}
1823 	NDFREE(&nd, NDF_ONLY_PNBUF);
1824 	vput(nd.ni_dvp);
1825 	if (vp == nd.ni_dvp)
1826 		vrele(vp);
1827 	else
1828 		vput(vp);
1829 	return (error);
1830 }
1831 
1832 /*
1833  * Reposition read/write file offset.
1834  */
1835 #ifndef _SYS_SYSPROTO_H_
1836 struct lseek_args {
1837 	int	fd;
1838 	int	pad;
1839 	off_t	offset;
1840 	int	whence;
1841 };
1842 #endif
1843 int
1844 sys_lseek(struct thread *td, struct lseek_args *uap)
1845 {
1846 
1847 	return (kern_lseek(td, uap->fd, uap->offset, uap->whence));
1848 }
1849 
1850 int
1851 kern_lseek(struct thread *td, int fd, off_t offset, int whence)
1852 {
1853 	struct file *fp;
1854 	cap_rights_t rights;
1855 	int error;
1856 
1857 	AUDIT_ARG_FD(fd);
1858 	error = fget(td, fd, cap_rights_init(&rights, CAP_SEEK), &fp);
1859 	if (error != 0)
1860 		return (error);
1861 	error = (fp->f_ops->fo_flags & DFLAG_SEEKABLE) != 0 ?
1862 	    fo_seek(fp, offset, whence, td) : ESPIPE;
1863 	fdrop(fp, td);
1864 	return (error);
1865 }
1866 
1867 #if defined(COMPAT_43)
1868 /*
1869  * Reposition read/write file offset.
1870  */
1871 #ifndef _SYS_SYSPROTO_H_
1872 struct olseek_args {
1873 	int	fd;
1874 	long	offset;
1875 	int	whence;
1876 };
1877 #endif
1878 int
1879 olseek(struct thread *td, struct olseek_args *uap)
1880 {
1881 
1882 	return (kern_lseek(td, uap->fd, uap->offset, uap->whence));
1883 }
1884 #endif /* COMPAT_43 */
1885 
1886 #if defined(COMPAT_FREEBSD6)
1887 /* Version with the 'pad' argument */
1888 int
1889 freebsd6_lseek(struct thread *td, struct freebsd6_lseek_args *uap)
1890 {
1891 
1892 	return (kern_lseek(td, uap->fd, uap->offset, uap->whence));
1893 }
1894 #endif
1895 
1896 /*
1897  * Check access permissions using passed credentials.
1898  */
1899 static int
1900 vn_access(struct vnode *vp, int user_flags, struct ucred *cred,
1901      struct thread *td)
1902 {
1903 	accmode_t accmode;
1904 	int error;
1905 
1906 	/* Flags == 0 means only check for existence. */
1907 	if (user_flags == 0)
1908 		return (0);
1909 
1910 	accmode = 0;
1911 	if (user_flags & R_OK)
1912 		accmode |= VREAD;
1913 	if (user_flags & W_OK)
1914 		accmode |= VWRITE;
1915 	if (user_flags & X_OK)
1916 		accmode |= VEXEC;
1917 #ifdef MAC
1918 	error = mac_vnode_check_access(cred, vp, accmode);
1919 	if (error != 0)
1920 		return (error);
1921 #endif
1922 	if ((accmode & VWRITE) == 0 || (error = vn_writechk(vp)) == 0)
1923 		error = VOP_ACCESS(vp, accmode, cred, td);
1924 	return (error);
1925 }
1926 
1927 /*
1928  * Check access permissions using "real" credentials.
1929  */
1930 #ifndef _SYS_SYSPROTO_H_
1931 struct access_args {
1932 	char	*path;
1933 	int	amode;
1934 };
1935 #endif
1936 int
1937 sys_access(struct thread *td, struct access_args *uap)
1938 {
1939 
1940 	return (kern_accessat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1941 	    0, uap->amode));
1942 }
1943 
1944 #ifndef _SYS_SYSPROTO_H_
1945 struct faccessat_args {
1946 	int	dirfd;
1947 	char	*path;
1948 	int	amode;
1949 	int	flag;
1950 }
1951 #endif
1952 int
1953 sys_faccessat(struct thread *td, struct faccessat_args *uap)
1954 {
1955 
1956 	return (kern_accessat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
1957 	    uap->amode));
1958 }
1959 
1960 int
1961 kern_accessat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1962     int flag, int amode)
1963 {
1964 	struct ucred *cred, *usecred;
1965 	struct vnode *vp;
1966 	struct nameidata nd;
1967 	cap_rights_t rights;
1968 	int error;
1969 
1970 	if (flag & ~AT_EACCESS)
1971 		return (EINVAL);
1972 	if (amode != F_OK && (amode & ~(R_OK | W_OK | X_OK)) != 0)
1973 		return (EINVAL);
1974 
1975 	/*
1976 	 * Create and modify a temporary credential instead of one that
1977 	 * is potentially shared (if we need one).
1978 	 */
1979 	cred = td->td_ucred;
1980 	if ((flag & AT_EACCESS) == 0 &&
1981 	    ((cred->cr_uid != cred->cr_ruid ||
1982 	    cred->cr_rgid != cred->cr_groups[0]))) {
1983 		usecred = crdup(cred);
1984 		usecred->cr_uid = cred->cr_ruid;
1985 		usecred->cr_groups[0] = cred->cr_rgid;
1986 		td->td_ucred = usecred;
1987 	} else
1988 		usecred = cred;
1989 	AUDIT_ARG_VALUE(amode);
1990 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF |
1991 	    AUDITVNODE1, pathseg, path, fd, cap_rights_init(&rights, CAP_FSTAT),
1992 	    td);
1993 	if ((error = namei(&nd)) != 0)
1994 		goto out;
1995 	vp = nd.ni_vp;
1996 
1997 	error = vn_access(vp, amode, usecred, td);
1998 	NDFREE(&nd, NDF_ONLY_PNBUF);
1999 	vput(vp);
2000 out:
2001 	if (usecred != cred) {
2002 		td->td_ucred = cred;
2003 		crfree(usecred);
2004 	}
2005 	return (error);
2006 }
2007 
2008 /*
2009  * Check access permissions using "effective" credentials.
2010  */
2011 #ifndef _SYS_SYSPROTO_H_
2012 struct eaccess_args {
2013 	char	*path;
2014 	int	amode;
2015 };
2016 #endif
2017 int
2018 sys_eaccess(struct thread *td, struct eaccess_args *uap)
2019 {
2020 
2021 	return (kern_accessat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2022 	    AT_EACCESS, uap->amode));
2023 }
2024 
2025 #if defined(COMPAT_43)
2026 /*
2027  * Get file status; this version follows links.
2028  */
2029 #ifndef _SYS_SYSPROTO_H_
2030 struct ostat_args {
2031 	char	*path;
2032 	struct ostat *ub;
2033 };
2034 #endif
2035 int
2036 ostat(struct thread *td, struct ostat_args *uap)
2037 {
2038 	struct stat sb;
2039 	struct ostat osb;
2040 	int error;
2041 
2042 	error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE,
2043 	    &sb, NULL);
2044 	if (error != 0)
2045 		return (error);
2046 	cvtstat(&sb, &osb);
2047 	return (copyout(&osb, uap->ub, sizeof (osb)));
2048 }
2049 
2050 /*
2051  * Get file status; this version does not follow links.
2052  */
2053 #ifndef _SYS_SYSPROTO_H_
2054 struct olstat_args {
2055 	char	*path;
2056 	struct ostat *ub;
2057 };
2058 #endif
2059 int
2060 olstat(struct thread *td, struct olstat_args *uap)
2061 {
2062 	struct stat sb;
2063 	struct ostat osb;
2064 	int error;
2065 
2066 	error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2067 	    UIO_USERSPACE, &sb, NULL);
2068 	if (error != 0)
2069 		return (error);
2070 	cvtstat(&sb, &osb);
2071 	return (copyout(&osb, uap->ub, sizeof (osb)));
2072 }
2073 
2074 /*
2075  * Convert from an old to a new stat structure.
2076  */
2077 void
2078 cvtstat(struct stat *st, struct ostat *ost)
2079 {
2080 
2081 	bzero(ost, sizeof(*ost));
2082 	ost->st_dev = st->st_dev;
2083 	ost->st_ino = st->st_ino;
2084 	ost->st_mode = st->st_mode;
2085 	ost->st_nlink = st->st_nlink;
2086 	ost->st_uid = st->st_uid;
2087 	ost->st_gid = st->st_gid;
2088 	ost->st_rdev = st->st_rdev;
2089 	if (st->st_size < (quad_t)1 << 32)
2090 		ost->st_size = st->st_size;
2091 	else
2092 		ost->st_size = -2;
2093 	ost->st_atim = st->st_atim;
2094 	ost->st_mtim = st->st_mtim;
2095 	ost->st_ctim = st->st_ctim;
2096 	ost->st_blksize = st->st_blksize;
2097 	ost->st_blocks = st->st_blocks;
2098 	ost->st_flags = st->st_flags;
2099 	ost->st_gen = st->st_gen;
2100 }
2101 #endif /* COMPAT_43 */
2102 
2103 #if defined(COMPAT_43) || defined(COMPAT_FREEBSD11)
2104 int ino64_trunc_error;
2105 SYSCTL_INT(_vfs, OID_AUTO, ino64_trunc_error, CTLFLAG_RW,
2106     &ino64_trunc_error, 0,
2107     "Error on truncation of inode number, device id or link count");
2108 int
2109 freebsd11_cvtstat(struct stat *st, struct freebsd11_stat *ost)
2110 {
2111 
2112 	ost->st_dev = st->st_dev;
2113 	ost->st_ino = st->st_ino;
2114 	if (ost->st_ino != st->st_ino) {
2115 		switch (ino64_trunc_error) {
2116 		default:
2117 		case 0:
2118 			break;
2119 		case 1:
2120 			return (EOVERFLOW);
2121 		case 2:
2122 			ost->st_ino = UINT32_MAX;
2123 			break;
2124 		}
2125 	}
2126 	ost->st_mode = st->st_mode;
2127 	ost->st_nlink = st->st_nlink;
2128 	if (ost->st_nlink != st->st_nlink) {
2129 		switch (ino64_trunc_error) {
2130 		default:
2131 		case 0:
2132 			break;
2133 		case 1:
2134 			return (EOVERFLOW);
2135 		case 2:
2136 			ost->st_nlink = UINT16_MAX;
2137 			break;
2138 		}
2139 	}
2140 	ost->st_uid = st->st_uid;
2141 	ost->st_gid = st->st_gid;
2142 	ost->st_rdev = st->st_rdev;
2143 	ost->st_atim = st->st_atim;
2144 	ost->st_mtim = st->st_mtim;
2145 	ost->st_ctim = st->st_ctim;
2146 	ost->st_size = st->st_size;
2147 	ost->st_blocks = st->st_blocks;
2148 	ost->st_blksize = st->st_blksize;
2149 	ost->st_flags = st->st_flags;
2150 	ost->st_gen = st->st_gen;
2151 	ost->st_lspare = 0;
2152 	ost->st_birthtim = st->st_birthtim;
2153 	bzero((char *)&ost->st_birthtim + sizeof(ost->st_birthtim),
2154 	    sizeof(*ost) - offsetof(struct freebsd11_stat,
2155 	    st_birthtim) - sizeof(ost->st_birthtim));
2156 	return (0);
2157 }
2158 
2159 int
2160 freebsd11_stat(struct thread *td, struct freebsd11_stat_args* uap)
2161 {
2162 	struct stat sb;
2163 	struct freebsd11_stat osb;
2164 	int error;
2165 
2166 	error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE,
2167 	    &sb, NULL);
2168 	if (error != 0)
2169 		return (error);
2170 	error = freebsd11_cvtstat(&sb, &osb);
2171 	if (error == 0)
2172 		error = copyout(&osb, uap->ub, sizeof(osb));
2173 	return (error);
2174 }
2175 
2176 int
2177 freebsd11_lstat(struct thread *td, struct freebsd11_lstat_args* uap)
2178 {
2179 	struct stat sb;
2180 	struct freebsd11_stat osb;
2181 	int error;
2182 
2183 	error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2184 	    UIO_USERSPACE, &sb, NULL);
2185 	if (error != 0)
2186 		return (error);
2187 	error = freebsd11_cvtstat(&sb, &osb);
2188 	if (error == 0)
2189 		error = copyout(&osb, uap->ub, sizeof(osb));
2190 	return (error);
2191 }
2192 
2193 int
2194 freebsd11_fhstat(struct thread *td, struct freebsd11_fhstat_args* uap)
2195 {
2196 	struct fhandle fh;
2197 	struct stat sb;
2198 	struct freebsd11_stat osb;
2199 	int error;
2200 
2201 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
2202 	if (error != 0)
2203 		return (error);
2204 	error = kern_fhstat(td, fh, &sb);
2205 	if (error != 0)
2206 		return (error);
2207 	error = freebsd11_cvtstat(&sb, &osb);
2208 	if (error == 0)
2209 		error = copyout(&osb, uap->sb, sizeof(osb));
2210 	return (error);
2211 }
2212 
2213 int
2214 freebsd11_fstatat(struct thread *td, struct freebsd11_fstatat_args* uap)
2215 {
2216 	struct stat sb;
2217 	struct freebsd11_stat osb;
2218 	int error;
2219 
2220 	error = kern_statat(td, uap->flag, uap->fd, uap->path,
2221 	    UIO_USERSPACE, &sb, NULL);
2222 	if (error != 0)
2223 		return (error);
2224 	error = freebsd11_cvtstat(&sb, &osb);
2225 	if (error == 0)
2226 		error = copyout(&osb, uap->buf, sizeof(osb));
2227 	return (error);
2228 }
2229 #endif	/* COMPAT_FREEBSD11 */
2230 
2231 /*
2232  * Get file status
2233  */
2234 #ifndef _SYS_SYSPROTO_H_
2235 struct fstatat_args {
2236 	int	fd;
2237 	char	*path;
2238 	struct stat	*buf;
2239 	int	flag;
2240 }
2241 #endif
2242 int
2243 sys_fstatat(struct thread *td, struct fstatat_args *uap)
2244 {
2245 	struct stat sb;
2246 	int error;
2247 
2248 	error = kern_statat(td, uap->flag, uap->fd, uap->path,
2249 	    UIO_USERSPACE, &sb, NULL);
2250 	if (error == 0)
2251 		error = copyout(&sb, uap->buf, sizeof (sb));
2252 	return (error);
2253 }
2254 
2255 int
2256 kern_statat(struct thread *td, int flag, int fd, char *path,
2257     enum uio_seg pathseg, struct stat *sbp,
2258     void (*hook)(struct vnode *vp, struct stat *sbp))
2259 {
2260 	struct nameidata nd;
2261 	struct stat sb;
2262 	cap_rights_t rights;
2263 	int error;
2264 
2265 	if (flag & ~AT_SYMLINK_NOFOLLOW)
2266 		return (EINVAL);
2267 
2268 	NDINIT_ATRIGHTS(&nd, LOOKUP, ((flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW :
2269 	    FOLLOW) | LOCKSHARED | LOCKLEAF | AUDITVNODE1, pathseg, path, fd,
2270 	    cap_rights_init(&rights, CAP_FSTAT), td);
2271 
2272 	if ((error = namei(&nd)) != 0)
2273 		return (error);
2274 	error = vn_stat(nd.ni_vp, &sb, td->td_ucred, NOCRED, td);
2275 	if (error == 0) {
2276 		SDT_PROBE2(vfs, , stat, mode, path, sb.st_mode);
2277 		if (S_ISREG(sb.st_mode))
2278 			SDT_PROBE2(vfs, , stat, reg, path, pathseg);
2279 		if (__predict_false(hook != NULL))
2280 			hook(nd.ni_vp, &sb);
2281 	}
2282 	NDFREE(&nd, NDF_ONLY_PNBUF);
2283 	vput(nd.ni_vp);
2284 	if (error != 0)
2285 		return (error);
2286 #ifdef __STAT_TIME_T_EXT
2287 	sb.st_atim_ext = 0;
2288 	sb.st_mtim_ext = 0;
2289 	sb.st_ctim_ext = 0;
2290 	sb.st_btim_ext = 0;
2291 #endif
2292 	*sbp = sb;
2293 #ifdef KTRACE
2294 	if (KTRPOINT(td, KTR_STRUCT))
2295 		ktrstat(&sb);
2296 #endif
2297 	return (0);
2298 }
2299 
2300 #if defined(COMPAT_FREEBSD11)
2301 /*
2302  * Implementation of the NetBSD [l]stat() functions.
2303  */
2304 void
2305 freebsd11_cvtnstat(struct stat *sb, struct nstat *nsb)
2306 {
2307 
2308 	bzero(nsb, sizeof(*nsb));
2309 	nsb->st_dev = sb->st_dev;
2310 	nsb->st_ino = sb->st_ino;
2311 	nsb->st_mode = sb->st_mode;
2312 	nsb->st_nlink = sb->st_nlink;
2313 	nsb->st_uid = sb->st_uid;
2314 	nsb->st_gid = sb->st_gid;
2315 	nsb->st_rdev = sb->st_rdev;
2316 	nsb->st_atim = sb->st_atim;
2317 	nsb->st_mtim = sb->st_mtim;
2318 	nsb->st_ctim = sb->st_ctim;
2319 	nsb->st_size = sb->st_size;
2320 	nsb->st_blocks = sb->st_blocks;
2321 	nsb->st_blksize = sb->st_blksize;
2322 	nsb->st_flags = sb->st_flags;
2323 	nsb->st_gen = sb->st_gen;
2324 	nsb->st_birthtim = sb->st_birthtim;
2325 }
2326 
2327 #ifndef _SYS_SYSPROTO_H_
2328 struct freebsd11_nstat_args {
2329 	char	*path;
2330 	struct nstat *ub;
2331 };
2332 #endif
2333 int
2334 freebsd11_nstat(struct thread *td, struct freebsd11_nstat_args *uap)
2335 {
2336 	struct stat sb;
2337 	struct nstat nsb;
2338 	int error;
2339 
2340 	error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE,
2341 	    &sb, NULL);
2342 	if (error != 0)
2343 		return (error);
2344 	freebsd11_cvtnstat(&sb, &nsb);
2345 	return (copyout(&nsb, uap->ub, sizeof (nsb)));
2346 }
2347 
2348 /*
2349  * NetBSD lstat.  Get file status; this version does not follow links.
2350  */
2351 #ifndef _SYS_SYSPROTO_H_
2352 struct freebsd11_nlstat_args {
2353 	char	*path;
2354 	struct nstat *ub;
2355 };
2356 #endif
2357 int
2358 freebsd11_nlstat(struct thread *td, struct freebsd11_nlstat_args *uap)
2359 {
2360 	struct stat sb;
2361 	struct nstat nsb;
2362 	int error;
2363 
2364 	error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2365 	    UIO_USERSPACE, &sb, NULL);
2366 	if (error != 0)
2367 		return (error);
2368 	freebsd11_cvtnstat(&sb, &nsb);
2369 	return (copyout(&nsb, uap->ub, sizeof (nsb)));
2370 }
2371 #endif /* COMPAT_FREEBSD11 */
2372 
2373 /*
2374  * Get configurable pathname variables.
2375  */
2376 #ifndef _SYS_SYSPROTO_H_
2377 struct pathconf_args {
2378 	char	*path;
2379 	int	name;
2380 };
2381 #endif
2382 int
2383 sys_pathconf(struct thread *td, struct pathconf_args *uap)
2384 {
2385 
2386 	return (kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name, FOLLOW));
2387 }
2388 
2389 #ifndef _SYS_SYSPROTO_H_
2390 struct lpathconf_args {
2391 	char	*path;
2392 	int	name;
2393 };
2394 #endif
2395 int
2396 sys_lpathconf(struct thread *td, struct lpathconf_args *uap)
2397 {
2398 
2399 	return (kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name,
2400 	    NOFOLLOW));
2401 }
2402 
2403 int
2404 kern_pathconf(struct thread *td, char *path, enum uio_seg pathseg, int name,
2405     u_long flags)
2406 {
2407 	struct nameidata nd;
2408 	int error;
2409 
2410 	NDINIT(&nd, LOOKUP, LOCKSHARED | LOCKLEAF | AUDITVNODE1 | flags,
2411 	    pathseg, path, td);
2412 	if ((error = namei(&nd)) != 0)
2413 		return (error);
2414 	NDFREE(&nd, NDF_ONLY_PNBUF);
2415 
2416 	error = VOP_PATHCONF(nd.ni_vp, name, td->td_retval);
2417 	vput(nd.ni_vp);
2418 	return (error);
2419 }
2420 
2421 /*
2422  * Return target name of a symbolic link.
2423  */
2424 #ifndef _SYS_SYSPROTO_H_
2425 struct readlink_args {
2426 	char	*path;
2427 	char	*buf;
2428 	size_t	count;
2429 };
2430 #endif
2431 int
2432 sys_readlink(struct thread *td, struct readlink_args *uap)
2433 {
2434 
2435 	return (kern_readlinkat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2436 	    uap->buf, UIO_USERSPACE, uap->count));
2437 }
2438 #ifndef _SYS_SYSPROTO_H_
2439 struct readlinkat_args {
2440 	int	fd;
2441 	char	*path;
2442 	char	*buf;
2443 	size_t	bufsize;
2444 };
2445 #endif
2446 int
2447 sys_readlinkat(struct thread *td, struct readlinkat_args *uap)
2448 {
2449 
2450 	return (kern_readlinkat(td, uap->fd, uap->path, UIO_USERSPACE,
2451 	    uap->buf, UIO_USERSPACE, uap->bufsize));
2452 }
2453 
2454 int
2455 kern_readlinkat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2456     char *buf, enum uio_seg bufseg, size_t count)
2457 {
2458 	struct vnode *vp;
2459 	struct iovec aiov;
2460 	struct uio auio;
2461 	struct nameidata nd;
2462 	int error;
2463 
2464 	if (count > IOSIZE_MAX)
2465 		return (EINVAL);
2466 
2467 	NDINIT_AT(&nd, LOOKUP, NOFOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
2468 	    pathseg, path, fd, td);
2469 
2470 	if ((error = namei(&nd)) != 0)
2471 		return (error);
2472 	NDFREE(&nd, NDF_ONLY_PNBUF);
2473 	vp = nd.ni_vp;
2474 #ifdef MAC
2475 	error = mac_vnode_check_readlink(td->td_ucred, vp);
2476 	if (error != 0) {
2477 		vput(vp);
2478 		return (error);
2479 	}
2480 #endif
2481 	if (vp->v_type != VLNK && (vp->v_vflag & VV_READLINK) == 0)
2482 		error = EINVAL;
2483 	else {
2484 		aiov.iov_base = buf;
2485 		aiov.iov_len = count;
2486 		auio.uio_iov = &aiov;
2487 		auio.uio_iovcnt = 1;
2488 		auio.uio_offset = 0;
2489 		auio.uio_rw = UIO_READ;
2490 		auio.uio_segflg = bufseg;
2491 		auio.uio_td = td;
2492 		auio.uio_resid = count;
2493 		error = VOP_READLINK(vp, &auio, td->td_ucred);
2494 		td->td_retval[0] = count - auio.uio_resid;
2495 	}
2496 	vput(vp);
2497 	return (error);
2498 }
2499 
2500 /*
2501  * Common implementation code for chflags() and fchflags().
2502  */
2503 static int
2504 setfflags(struct thread *td, struct vnode *vp, u_long flags)
2505 {
2506 	struct mount *mp;
2507 	struct vattr vattr;
2508 	int error;
2509 
2510 	/* We can't support the value matching VNOVAL. */
2511 	if (flags == VNOVAL)
2512 		return (EOPNOTSUPP);
2513 
2514 	/*
2515 	 * Prevent non-root users from setting flags on devices.  When
2516 	 * a device is reused, users can retain ownership of the device
2517 	 * if they are allowed to set flags and programs assume that
2518 	 * chown can't fail when done as root.
2519 	 */
2520 	if (vp->v_type == VCHR || vp->v_type == VBLK) {
2521 		error = priv_check(td, PRIV_VFS_CHFLAGS_DEV);
2522 		if (error != 0)
2523 			return (error);
2524 	}
2525 
2526 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2527 		return (error);
2528 	VATTR_NULL(&vattr);
2529 	vattr.va_flags = flags;
2530 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2531 #ifdef MAC
2532 	error = mac_vnode_check_setflags(td->td_ucred, vp, vattr.va_flags);
2533 	if (error == 0)
2534 #endif
2535 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
2536 	VOP_UNLOCK(vp, 0);
2537 	vn_finished_write(mp);
2538 	return (error);
2539 }
2540 
2541 /*
2542  * Change flags of a file given a path name.
2543  */
2544 #ifndef _SYS_SYSPROTO_H_
2545 struct chflags_args {
2546 	const char *path;
2547 	u_long	flags;
2548 };
2549 #endif
2550 int
2551 sys_chflags(struct thread *td, struct chflags_args *uap)
2552 {
2553 
2554 	return (kern_chflagsat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2555 	    uap->flags, 0));
2556 }
2557 
2558 #ifndef _SYS_SYSPROTO_H_
2559 struct chflagsat_args {
2560 	int	fd;
2561 	const char *path;
2562 	u_long	flags;
2563 	int	atflag;
2564 }
2565 #endif
2566 int
2567 sys_chflagsat(struct thread *td, struct chflagsat_args *uap)
2568 {
2569 	int fd = uap->fd;
2570 	const char *path = uap->path;
2571 	u_long flags = uap->flags;
2572 	int atflag = uap->atflag;
2573 
2574 	if (atflag & ~AT_SYMLINK_NOFOLLOW)
2575 		return (EINVAL);
2576 
2577 	return (kern_chflagsat(td, fd, path, UIO_USERSPACE, flags, atflag));
2578 }
2579 
2580 /*
2581  * Same as chflags() but doesn't follow symlinks.
2582  */
2583 #ifndef _SYS_SYSPROTO_H_
2584 struct lchflags_args {
2585 	const char *path;
2586 	u_long flags;
2587 };
2588 #endif
2589 int
2590 sys_lchflags(struct thread *td, struct lchflags_args *uap)
2591 {
2592 
2593 	return (kern_chflagsat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2594 	    uap->flags, AT_SYMLINK_NOFOLLOW));
2595 }
2596 
2597 static int
2598 kern_chflagsat(struct thread *td, int fd, const char *path,
2599     enum uio_seg pathseg, u_long flags, int atflag)
2600 {
2601 	struct nameidata nd;
2602 	cap_rights_t rights;
2603 	int error, follow;
2604 
2605 	AUDIT_ARG_FFLAGS(flags);
2606 	follow = (atflag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
2607 	NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, pathseg, path, fd,
2608 	    cap_rights_init(&rights, CAP_FCHFLAGS), td);
2609 	if ((error = namei(&nd)) != 0)
2610 		return (error);
2611 	NDFREE(&nd, NDF_ONLY_PNBUF);
2612 	error = setfflags(td, nd.ni_vp, flags);
2613 	vrele(nd.ni_vp);
2614 	return (error);
2615 }
2616 
2617 /*
2618  * Change flags of a file given a file descriptor.
2619  */
2620 #ifndef _SYS_SYSPROTO_H_
2621 struct fchflags_args {
2622 	int	fd;
2623 	u_long	flags;
2624 };
2625 #endif
2626 int
2627 sys_fchflags(struct thread *td, struct fchflags_args *uap)
2628 {
2629 	struct file *fp;
2630 	cap_rights_t rights;
2631 	int error;
2632 
2633 	AUDIT_ARG_FD(uap->fd);
2634 	AUDIT_ARG_FFLAGS(uap->flags);
2635 	error = getvnode(td, uap->fd, cap_rights_init(&rights, CAP_FCHFLAGS),
2636 	    &fp);
2637 	if (error != 0)
2638 		return (error);
2639 #ifdef AUDIT
2640 	vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
2641 	AUDIT_ARG_VNODE1(fp->f_vnode);
2642 	VOP_UNLOCK(fp->f_vnode, 0);
2643 #endif
2644 	error = setfflags(td, fp->f_vnode, uap->flags);
2645 	fdrop(fp, td);
2646 	return (error);
2647 }
2648 
2649 /*
2650  * Common implementation code for chmod(), lchmod() and fchmod().
2651  */
2652 int
2653 setfmode(struct thread *td, struct ucred *cred, struct vnode *vp, int mode)
2654 {
2655 	struct mount *mp;
2656 	struct vattr vattr;
2657 	int error;
2658 
2659 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2660 		return (error);
2661 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2662 	VATTR_NULL(&vattr);
2663 	vattr.va_mode = mode & ALLPERMS;
2664 #ifdef MAC
2665 	error = mac_vnode_check_setmode(cred, vp, vattr.va_mode);
2666 	if (error == 0)
2667 #endif
2668 		error = VOP_SETATTR(vp, &vattr, cred);
2669 	VOP_UNLOCK(vp, 0);
2670 	vn_finished_write(mp);
2671 	return (error);
2672 }
2673 
2674 /*
2675  * Change mode of a file given path name.
2676  */
2677 #ifndef _SYS_SYSPROTO_H_
2678 struct chmod_args {
2679 	char	*path;
2680 	int	mode;
2681 };
2682 #endif
2683 int
2684 sys_chmod(struct thread *td, struct chmod_args *uap)
2685 {
2686 
2687 	return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2688 	    uap->mode, 0));
2689 }
2690 
2691 #ifndef _SYS_SYSPROTO_H_
2692 struct fchmodat_args {
2693 	int	dirfd;
2694 	char	*path;
2695 	mode_t	mode;
2696 	int	flag;
2697 }
2698 #endif
2699 int
2700 sys_fchmodat(struct thread *td, struct fchmodat_args *uap)
2701 {
2702 	int flag = uap->flag;
2703 	int fd = uap->fd;
2704 	char *path = uap->path;
2705 	mode_t mode = uap->mode;
2706 
2707 	if (flag & ~AT_SYMLINK_NOFOLLOW)
2708 		return (EINVAL);
2709 
2710 	return (kern_fchmodat(td, fd, path, UIO_USERSPACE, mode, flag));
2711 }
2712 
2713 /*
2714  * Change mode of a file given path name (don't follow links.)
2715  */
2716 #ifndef _SYS_SYSPROTO_H_
2717 struct lchmod_args {
2718 	char	*path;
2719 	int	mode;
2720 };
2721 #endif
2722 int
2723 sys_lchmod(struct thread *td, struct lchmod_args *uap)
2724 {
2725 
2726 	return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2727 	    uap->mode, AT_SYMLINK_NOFOLLOW));
2728 }
2729 
2730 int
2731 kern_fchmodat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2732     mode_t mode, int flag)
2733 {
2734 	struct nameidata nd;
2735 	cap_rights_t rights;
2736 	int error, follow;
2737 
2738 	AUDIT_ARG_MODE(mode);
2739 	follow = (flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
2740 	NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, pathseg, path, fd,
2741 	    cap_rights_init(&rights, CAP_FCHMOD), td);
2742 	if ((error = namei(&nd)) != 0)
2743 		return (error);
2744 	NDFREE(&nd, NDF_ONLY_PNBUF);
2745 	error = setfmode(td, td->td_ucred, nd.ni_vp, mode);
2746 	vrele(nd.ni_vp);
2747 	return (error);
2748 }
2749 
2750 /*
2751  * Change mode of a file given a file descriptor.
2752  */
2753 #ifndef _SYS_SYSPROTO_H_
2754 struct fchmod_args {
2755 	int	fd;
2756 	int	mode;
2757 };
2758 #endif
2759 int
2760 sys_fchmod(struct thread *td, struct fchmod_args *uap)
2761 {
2762 	struct file *fp;
2763 	cap_rights_t rights;
2764 	int error;
2765 
2766 	AUDIT_ARG_FD(uap->fd);
2767 	AUDIT_ARG_MODE(uap->mode);
2768 
2769 	error = fget(td, uap->fd, cap_rights_init(&rights, CAP_FCHMOD), &fp);
2770 	if (error != 0)
2771 		return (error);
2772 	error = fo_chmod(fp, uap->mode, td->td_ucred, td);
2773 	fdrop(fp, td);
2774 	return (error);
2775 }
2776 
2777 /*
2778  * Common implementation for chown(), lchown(), and fchown()
2779  */
2780 int
2781 setfown(struct thread *td, struct ucred *cred, struct vnode *vp, uid_t uid,
2782     gid_t gid)
2783 {
2784 	struct mount *mp;
2785 	struct vattr vattr;
2786 	int error;
2787 
2788 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2789 		return (error);
2790 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2791 	VATTR_NULL(&vattr);
2792 	vattr.va_uid = uid;
2793 	vattr.va_gid = gid;
2794 #ifdef MAC
2795 	error = mac_vnode_check_setowner(cred, vp, vattr.va_uid,
2796 	    vattr.va_gid);
2797 	if (error == 0)
2798 #endif
2799 		error = VOP_SETATTR(vp, &vattr, cred);
2800 	VOP_UNLOCK(vp, 0);
2801 	vn_finished_write(mp);
2802 	return (error);
2803 }
2804 
2805 /*
2806  * Set ownership given a path name.
2807  */
2808 #ifndef _SYS_SYSPROTO_H_
2809 struct chown_args {
2810 	char	*path;
2811 	int	uid;
2812 	int	gid;
2813 };
2814 #endif
2815 int
2816 sys_chown(struct thread *td, struct chown_args *uap)
2817 {
2818 
2819 	return (kern_fchownat(td, AT_FDCWD, uap->path, UIO_USERSPACE, uap->uid,
2820 	    uap->gid, 0));
2821 }
2822 
2823 #ifndef _SYS_SYSPROTO_H_
2824 struct fchownat_args {
2825 	int fd;
2826 	const char * path;
2827 	uid_t uid;
2828 	gid_t gid;
2829 	int flag;
2830 };
2831 #endif
2832 int
2833 sys_fchownat(struct thread *td, struct fchownat_args *uap)
2834 {
2835 	int flag;
2836 
2837 	flag = uap->flag;
2838 	if (flag & ~AT_SYMLINK_NOFOLLOW)
2839 		return (EINVAL);
2840 
2841 	return (kern_fchownat(td, uap->fd, uap->path, UIO_USERSPACE, uap->uid,
2842 	    uap->gid, uap->flag));
2843 }
2844 
2845 int
2846 kern_fchownat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2847     int uid, int gid, int flag)
2848 {
2849 	struct nameidata nd;
2850 	cap_rights_t rights;
2851 	int error, follow;
2852 
2853 	AUDIT_ARG_OWNER(uid, gid);
2854 	follow = (flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
2855 	NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, pathseg, path, fd,
2856 	    cap_rights_init(&rights, CAP_FCHOWN), td);
2857 
2858 	if ((error = namei(&nd)) != 0)
2859 		return (error);
2860 	NDFREE(&nd, NDF_ONLY_PNBUF);
2861 	error = setfown(td, td->td_ucred, nd.ni_vp, uid, gid);
2862 	vrele(nd.ni_vp);
2863 	return (error);
2864 }
2865 
2866 /*
2867  * Set ownership given a path name, do not cross symlinks.
2868  */
2869 #ifndef _SYS_SYSPROTO_H_
2870 struct lchown_args {
2871 	char	*path;
2872 	int	uid;
2873 	int	gid;
2874 };
2875 #endif
2876 int
2877 sys_lchown(struct thread *td, struct lchown_args *uap)
2878 {
2879 
2880 	return (kern_fchownat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2881 	    uap->uid, uap->gid, AT_SYMLINK_NOFOLLOW));
2882 }
2883 
2884 /*
2885  * Set ownership given a file descriptor.
2886  */
2887 #ifndef _SYS_SYSPROTO_H_
2888 struct fchown_args {
2889 	int	fd;
2890 	int	uid;
2891 	int	gid;
2892 };
2893 #endif
2894 int
2895 sys_fchown(struct thread *td, struct fchown_args *uap)
2896 {
2897 	struct file *fp;
2898 	cap_rights_t rights;
2899 	int error;
2900 
2901 	AUDIT_ARG_FD(uap->fd);
2902 	AUDIT_ARG_OWNER(uap->uid, uap->gid);
2903 	error = fget(td, uap->fd, cap_rights_init(&rights, CAP_FCHOWN), &fp);
2904 	if (error != 0)
2905 		return (error);
2906 	error = fo_chown(fp, uap->uid, uap->gid, td->td_ucred, td);
2907 	fdrop(fp, td);
2908 	return (error);
2909 }
2910 
2911 /*
2912  * Common implementation code for utimes(), lutimes(), and futimes().
2913  */
2914 static int
2915 getutimes(const struct timeval *usrtvp, enum uio_seg tvpseg,
2916     struct timespec *tsp)
2917 {
2918 	struct timeval tv[2];
2919 	const struct timeval *tvp;
2920 	int error;
2921 
2922 	if (usrtvp == NULL) {
2923 		vfs_timestamp(&tsp[0]);
2924 		tsp[1] = tsp[0];
2925 	} else {
2926 		if (tvpseg == UIO_SYSSPACE) {
2927 			tvp = usrtvp;
2928 		} else {
2929 			if ((error = copyin(usrtvp, tv, sizeof(tv))) != 0)
2930 				return (error);
2931 			tvp = tv;
2932 		}
2933 
2934 		if (tvp[0].tv_usec < 0 || tvp[0].tv_usec >= 1000000 ||
2935 		    tvp[1].tv_usec < 0 || tvp[1].tv_usec >= 1000000)
2936 			return (EINVAL);
2937 		TIMEVAL_TO_TIMESPEC(&tvp[0], &tsp[0]);
2938 		TIMEVAL_TO_TIMESPEC(&tvp[1], &tsp[1]);
2939 	}
2940 	return (0);
2941 }
2942 
2943 /*
2944  * Common implementation code for futimens(), utimensat().
2945  */
2946 #define	UTIMENS_NULL	0x1
2947 #define	UTIMENS_EXIT	0x2
2948 static int
2949 getutimens(const struct timespec *usrtsp, enum uio_seg tspseg,
2950     struct timespec *tsp, int *retflags)
2951 {
2952 	struct timespec tsnow;
2953 	int error;
2954 
2955 	vfs_timestamp(&tsnow);
2956 	*retflags = 0;
2957 	if (usrtsp == NULL) {
2958 		tsp[0] = tsnow;
2959 		tsp[1] = tsnow;
2960 		*retflags |= UTIMENS_NULL;
2961 		return (0);
2962 	}
2963 	if (tspseg == UIO_SYSSPACE) {
2964 		tsp[0] = usrtsp[0];
2965 		tsp[1] = usrtsp[1];
2966 	} else if ((error = copyin(usrtsp, tsp, sizeof(*tsp) * 2)) != 0)
2967 		return (error);
2968 	if (tsp[0].tv_nsec == UTIME_OMIT && tsp[1].tv_nsec == UTIME_OMIT)
2969 		*retflags |= UTIMENS_EXIT;
2970 	if (tsp[0].tv_nsec == UTIME_NOW && tsp[1].tv_nsec == UTIME_NOW)
2971 		*retflags |= UTIMENS_NULL;
2972 	if (tsp[0].tv_nsec == UTIME_OMIT)
2973 		tsp[0].tv_sec = VNOVAL;
2974 	else if (tsp[0].tv_nsec == UTIME_NOW)
2975 		tsp[0] = tsnow;
2976 	else if (tsp[0].tv_nsec < 0 || tsp[0].tv_nsec >= 1000000000L)
2977 		return (EINVAL);
2978 	if (tsp[1].tv_nsec == UTIME_OMIT)
2979 		tsp[1].tv_sec = VNOVAL;
2980 	else if (tsp[1].tv_nsec == UTIME_NOW)
2981 		tsp[1] = tsnow;
2982 	else if (tsp[1].tv_nsec < 0 || tsp[1].tv_nsec >= 1000000000L)
2983 		return (EINVAL);
2984 
2985 	return (0);
2986 }
2987 
2988 /*
2989  * Common implementation code for utimes(), lutimes(), futimes(), futimens(),
2990  * and utimensat().
2991  */
2992 static int
2993 setutimes(struct thread *td, struct vnode *vp, const struct timespec *ts,
2994     int numtimes, int nullflag)
2995 {
2996 	struct mount *mp;
2997 	struct vattr vattr;
2998 	int error, setbirthtime;
2999 
3000 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
3001 		return (error);
3002 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3003 	setbirthtime = 0;
3004 	if (numtimes < 3 && !VOP_GETATTR(vp, &vattr, td->td_ucred) &&
3005 	    timespeccmp(&ts[1], &vattr.va_birthtime, < ))
3006 		setbirthtime = 1;
3007 	VATTR_NULL(&vattr);
3008 	vattr.va_atime = ts[0];
3009 	vattr.va_mtime = ts[1];
3010 	if (setbirthtime)
3011 		vattr.va_birthtime = ts[1];
3012 	if (numtimes > 2)
3013 		vattr.va_birthtime = ts[2];
3014 	if (nullflag)
3015 		vattr.va_vaflags |= VA_UTIMES_NULL;
3016 #ifdef MAC
3017 	error = mac_vnode_check_setutimes(td->td_ucred, vp, vattr.va_atime,
3018 	    vattr.va_mtime);
3019 #endif
3020 	if (error == 0)
3021 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3022 	VOP_UNLOCK(vp, 0);
3023 	vn_finished_write(mp);
3024 	return (error);
3025 }
3026 
3027 /*
3028  * Set the access and modification times of a file.
3029  */
3030 #ifndef _SYS_SYSPROTO_H_
3031 struct utimes_args {
3032 	char	*path;
3033 	struct	timeval *tptr;
3034 };
3035 #endif
3036 int
3037 sys_utimes(struct thread *td, struct utimes_args *uap)
3038 {
3039 
3040 	return (kern_utimesat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
3041 	    uap->tptr, UIO_USERSPACE));
3042 }
3043 
3044 #ifndef _SYS_SYSPROTO_H_
3045 struct futimesat_args {
3046 	int fd;
3047 	const char * path;
3048 	const struct timeval * times;
3049 };
3050 #endif
3051 int
3052 sys_futimesat(struct thread *td, struct futimesat_args *uap)
3053 {
3054 
3055 	return (kern_utimesat(td, uap->fd, uap->path, UIO_USERSPACE,
3056 	    uap->times, UIO_USERSPACE));
3057 }
3058 
3059 int
3060 kern_utimesat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
3061     struct timeval *tptr, enum uio_seg tptrseg)
3062 {
3063 	struct nameidata nd;
3064 	struct timespec ts[2];
3065 	cap_rights_t rights;
3066 	int error;
3067 
3068 	if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3069 		return (error);
3070 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, fd,
3071 	    cap_rights_init(&rights, CAP_FUTIMES), td);
3072 
3073 	if ((error = namei(&nd)) != 0)
3074 		return (error);
3075 	NDFREE(&nd, NDF_ONLY_PNBUF);
3076 	error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3077 	vrele(nd.ni_vp);
3078 	return (error);
3079 }
3080 
3081 /*
3082  * Set the access and modification times of a file.
3083  */
3084 #ifndef _SYS_SYSPROTO_H_
3085 struct lutimes_args {
3086 	char	*path;
3087 	struct	timeval *tptr;
3088 };
3089 #endif
3090 int
3091 sys_lutimes(struct thread *td, struct lutimes_args *uap)
3092 {
3093 
3094 	return (kern_lutimes(td, uap->path, UIO_USERSPACE, uap->tptr,
3095 	    UIO_USERSPACE));
3096 }
3097 
3098 int
3099 kern_lutimes(struct thread *td, char *path, enum uio_seg pathseg,
3100     struct timeval *tptr, enum uio_seg tptrseg)
3101 {
3102 	struct timespec ts[2];
3103 	struct nameidata nd;
3104 	int error;
3105 
3106 	if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3107 		return (error);
3108 	NDINIT(&nd, LOOKUP, NOFOLLOW | AUDITVNODE1, pathseg, path, td);
3109 	if ((error = namei(&nd)) != 0)
3110 		return (error);
3111 	NDFREE(&nd, NDF_ONLY_PNBUF);
3112 	error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3113 	vrele(nd.ni_vp);
3114 	return (error);
3115 }
3116 
3117 /*
3118  * Set the access and modification times of a file.
3119  */
3120 #ifndef _SYS_SYSPROTO_H_
3121 struct futimes_args {
3122 	int	fd;
3123 	struct	timeval *tptr;
3124 };
3125 #endif
3126 int
3127 sys_futimes(struct thread *td, struct futimes_args *uap)
3128 {
3129 
3130 	return (kern_futimes(td, uap->fd, uap->tptr, UIO_USERSPACE));
3131 }
3132 
3133 int
3134 kern_futimes(struct thread *td, int fd, struct timeval *tptr,
3135     enum uio_seg tptrseg)
3136 {
3137 	struct timespec ts[2];
3138 	struct file *fp;
3139 	cap_rights_t rights;
3140 	int error;
3141 
3142 	AUDIT_ARG_FD(fd);
3143 	error = getutimes(tptr, tptrseg, ts);
3144 	if (error != 0)
3145 		return (error);
3146 	error = getvnode(td, fd, cap_rights_init(&rights, CAP_FUTIMES), &fp);
3147 	if (error != 0)
3148 		return (error);
3149 #ifdef AUDIT
3150 	vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
3151 	AUDIT_ARG_VNODE1(fp->f_vnode);
3152 	VOP_UNLOCK(fp->f_vnode, 0);
3153 #endif
3154 	error = setutimes(td, fp->f_vnode, ts, 2, tptr == NULL);
3155 	fdrop(fp, td);
3156 	return (error);
3157 }
3158 
3159 int
3160 sys_futimens(struct thread *td, struct futimens_args *uap)
3161 {
3162 
3163 	return (kern_futimens(td, uap->fd, uap->times, UIO_USERSPACE));
3164 }
3165 
3166 int
3167 kern_futimens(struct thread *td, int fd, struct timespec *tptr,
3168     enum uio_seg tptrseg)
3169 {
3170 	struct timespec ts[2];
3171 	struct file *fp;
3172 	cap_rights_t rights;
3173 	int error, flags;
3174 
3175 	AUDIT_ARG_FD(fd);
3176 	error = getutimens(tptr, tptrseg, ts, &flags);
3177 	if (error != 0)
3178 		return (error);
3179 	if (flags & UTIMENS_EXIT)
3180 		return (0);
3181 	error = getvnode(td, fd, cap_rights_init(&rights, CAP_FUTIMES), &fp);
3182 	if (error != 0)
3183 		return (error);
3184 #ifdef AUDIT
3185 	vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
3186 	AUDIT_ARG_VNODE1(fp->f_vnode);
3187 	VOP_UNLOCK(fp->f_vnode, 0);
3188 #endif
3189 	error = setutimes(td, fp->f_vnode, ts, 2, flags & UTIMENS_NULL);
3190 	fdrop(fp, td);
3191 	return (error);
3192 }
3193 
3194 int
3195 sys_utimensat(struct thread *td, struct utimensat_args *uap)
3196 {
3197 
3198 	return (kern_utimensat(td, uap->fd, uap->path, UIO_USERSPACE,
3199 	    uap->times, UIO_USERSPACE, uap->flag));
3200 }
3201 
3202 int
3203 kern_utimensat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
3204     struct timespec *tptr, enum uio_seg tptrseg, int flag)
3205 {
3206 	struct nameidata nd;
3207 	struct timespec ts[2];
3208 	cap_rights_t rights;
3209 	int error, flags;
3210 
3211 	if (flag & ~AT_SYMLINK_NOFOLLOW)
3212 		return (EINVAL);
3213 
3214 	if ((error = getutimens(tptr, tptrseg, ts, &flags)) != 0)
3215 		return (error);
3216 	NDINIT_ATRIGHTS(&nd, LOOKUP, ((flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW :
3217 	    FOLLOW) | AUDITVNODE1, pathseg, path, fd,
3218 	    cap_rights_init(&rights, CAP_FUTIMES), td);
3219 	if ((error = namei(&nd)) != 0)
3220 		return (error);
3221 	/*
3222 	 * We are allowed to call namei() regardless of 2xUTIME_OMIT.
3223 	 * POSIX states:
3224 	 * "If both tv_nsec fields are UTIME_OMIT... EACCESS may be detected."
3225 	 * "Search permission is denied by a component of the path prefix."
3226 	 */
3227 	NDFREE(&nd, NDF_ONLY_PNBUF);
3228 	if ((flags & UTIMENS_EXIT) == 0)
3229 		error = setutimes(td, nd.ni_vp, ts, 2, flags & UTIMENS_NULL);
3230 	vrele(nd.ni_vp);
3231 	return (error);
3232 }
3233 
3234 /*
3235  * Truncate a file given its path name.
3236  */
3237 #ifndef _SYS_SYSPROTO_H_
3238 struct truncate_args {
3239 	char	*path;
3240 	int	pad;
3241 	off_t	length;
3242 };
3243 #endif
3244 int
3245 sys_truncate(struct thread *td, struct truncate_args *uap)
3246 {
3247 
3248 	return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3249 }
3250 
3251 int
3252 kern_truncate(struct thread *td, char *path, enum uio_seg pathseg, off_t length)
3253 {
3254 	struct mount *mp;
3255 	struct vnode *vp;
3256 	void *rl_cookie;
3257 	struct vattr vattr;
3258 	struct nameidata nd;
3259 	int error;
3260 
3261 	if (length < 0)
3262 		return(EINVAL);
3263 	NDINIT(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, td);
3264 	if ((error = namei(&nd)) != 0)
3265 		return (error);
3266 	vp = nd.ni_vp;
3267 	rl_cookie = vn_rangelock_wlock(vp, 0, OFF_MAX);
3268 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) {
3269 		vn_rangelock_unlock(vp, rl_cookie);
3270 		vrele(vp);
3271 		return (error);
3272 	}
3273 	NDFREE(&nd, NDF_ONLY_PNBUF);
3274 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3275 	if (vp->v_type == VDIR)
3276 		error = EISDIR;
3277 #ifdef MAC
3278 	else if ((error = mac_vnode_check_write(td->td_ucred, NOCRED, vp))) {
3279 	}
3280 #endif
3281 	else if ((error = vn_writechk(vp)) == 0 &&
3282 	    (error = VOP_ACCESS(vp, VWRITE, td->td_ucred, td)) == 0) {
3283 		VATTR_NULL(&vattr);
3284 		vattr.va_size = length;
3285 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3286 	}
3287 	VOP_UNLOCK(vp, 0);
3288 	vn_finished_write(mp);
3289 	vn_rangelock_unlock(vp, rl_cookie);
3290 	vrele(vp);
3291 	return (error);
3292 }
3293 
3294 #if defined(COMPAT_43)
3295 /*
3296  * Truncate a file given its path name.
3297  */
3298 #ifndef _SYS_SYSPROTO_H_
3299 struct otruncate_args {
3300 	char	*path;
3301 	long	length;
3302 };
3303 #endif
3304 int
3305 otruncate(struct thread *td, struct otruncate_args *uap)
3306 {
3307 
3308 	return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3309 }
3310 #endif /* COMPAT_43 */
3311 
3312 #if defined(COMPAT_FREEBSD6)
3313 /* Versions with the pad argument */
3314 int
3315 freebsd6_truncate(struct thread *td, struct freebsd6_truncate_args *uap)
3316 {
3317 
3318 	return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3319 }
3320 
3321 int
3322 freebsd6_ftruncate(struct thread *td, struct freebsd6_ftruncate_args *uap)
3323 {
3324 
3325 	return (kern_ftruncate(td, uap->fd, uap->length));
3326 }
3327 #endif
3328 
3329 int
3330 kern_fsync(struct thread *td, int fd, bool fullsync)
3331 {
3332 	struct vnode *vp;
3333 	struct mount *mp;
3334 	struct file *fp;
3335 	cap_rights_t rights;
3336 	int error, lock_flags;
3337 
3338 	AUDIT_ARG_FD(fd);
3339 	error = getvnode(td, fd, cap_rights_init(&rights, CAP_FSYNC), &fp);
3340 	if (error != 0)
3341 		return (error);
3342 	vp = fp->f_vnode;
3343 #if 0
3344 	if (!fullsync)
3345 		/* XXXKIB: compete outstanding aio writes */;
3346 #endif
3347 	error = vn_start_write(vp, &mp, V_WAIT | PCATCH);
3348 	if (error != 0)
3349 		goto drop;
3350 	if (MNT_SHARED_WRITES(mp) ||
3351 	    ((mp == NULL) && MNT_SHARED_WRITES(vp->v_mount))) {
3352 		lock_flags = LK_SHARED;
3353 	} else {
3354 		lock_flags = LK_EXCLUSIVE;
3355 	}
3356 	vn_lock(vp, lock_flags | LK_RETRY);
3357 	AUDIT_ARG_VNODE1(vp);
3358 	if (vp->v_object != NULL) {
3359 		VM_OBJECT_WLOCK(vp->v_object);
3360 		vm_object_page_clean(vp->v_object, 0, 0, 0);
3361 		VM_OBJECT_WUNLOCK(vp->v_object);
3362 	}
3363 	error = fullsync ? VOP_FSYNC(vp, MNT_WAIT, td) : VOP_FDATASYNC(vp, td);
3364 	VOP_UNLOCK(vp, 0);
3365 	vn_finished_write(mp);
3366 drop:
3367 	fdrop(fp, td);
3368 	return (error);
3369 }
3370 
3371 /*
3372  * Sync an open file.
3373  */
3374 #ifndef _SYS_SYSPROTO_H_
3375 struct fsync_args {
3376 	int	fd;
3377 };
3378 #endif
3379 int
3380 sys_fsync(struct thread *td, struct fsync_args *uap)
3381 {
3382 
3383 	return (kern_fsync(td, uap->fd, true));
3384 }
3385 
3386 int
3387 sys_fdatasync(struct thread *td, struct fdatasync_args *uap)
3388 {
3389 
3390 	return (kern_fsync(td, uap->fd, false));
3391 }
3392 
3393 /*
3394  * Rename files.  Source and destination must either both be directories, or
3395  * both not be directories.  If target is a directory, it must be empty.
3396  */
3397 #ifndef _SYS_SYSPROTO_H_
3398 struct rename_args {
3399 	char	*from;
3400 	char	*to;
3401 };
3402 #endif
3403 int
3404 sys_rename(struct thread *td, struct rename_args *uap)
3405 {
3406 
3407 	return (kern_renameat(td, AT_FDCWD, uap->from, AT_FDCWD,
3408 	    uap->to, UIO_USERSPACE));
3409 }
3410 
3411 #ifndef _SYS_SYSPROTO_H_
3412 struct renameat_args {
3413 	int	oldfd;
3414 	char	*old;
3415 	int	newfd;
3416 	char	*new;
3417 };
3418 #endif
3419 int
3420 sys_renameat(struct thread *td, struct renameat_args *uap)
3421 {
3422 
3423 	return (kern_renameat(td, uap->oldfd, uap->old, uap->newfd, uap->new,
3424 	    UIO_USERSPACE));
3425 }
3426 
3427 int
3428 kern_renameat(struct thread *td, int oldfd, char *old, int newfd, char *new,
3429     enum uio_seg pathseg)
3430 {
3431 	struct mount *mp = NULL;
3432 	struct vnode *tvp, *fvp, *tdvp;
3433 	struct nameidata fromnd, tond;
3434 	cap_rights_t rights;
3435 	int error;
3436 
3437 again:
3438 	bwillwrite();
3439 #ifdef MAC
3440 	NDINIT_ATRIGHTS(&fromnd, DELETE, LOCKPARENT | LOCKLEAF | SAVESTART |
3441 	    AUDITVNODE1, pathseg, old, oldfd,
3442 	    cap_rights_init(&rights, CAP_RENAMEAT_SOURCE), td);
3443 #else
3444 	NDINIT_ATRIGHTS(&fromnd, DELETE, WANTPARENT | SAVESTART | AUDITVNODE1,
3445 	    pathseg, old, oldfd,
3446 	    cap_rights_init(&rights, CAP_RENAMEAT_SOURCE), td);
3447 #endif
3448 
3449 	if ((error = namei(&fromnd)) != 0)
3450 		return (error);
3451 #ifdef MAC
3452 	error = mac_vnode_check_rename_from(td->td_ucred, fromnd.ni_dvp,
3453 	    fromnd.ni_vp, &fromnd.ni_cnd);
3454 	VOP_UNLOCK(fromnd.ni_dvp, 0);
3455 	if (fromnd.ni_dvp != fromnd.ni_vp)
3456 		VOP_UNLOCK(fromnd.ni_vp, 0);
3457 #endif
3458 	fvp = fromnd.ni_vp;
3459 	NDINIT_ATRIGHTS(&tond, RENAME, LOCKPARENT | LOCKLEAF | NOCACHE |
3460 	    SAVESTART | AUDITVNODE2, pathseg, new, newfd,
3461 	    cap_rights_init(&rights, CAP_RENAMEAT_TARGET), td);
3462 	if (fromnd.ni_vp->v_type == VDIR)
3463 		tond.ni_cnd.cn_flags |= WILLBEDIR;
3464 	if ((error = namei(&tond)) != 0) {
3465 		/* Translate error code for rename("dir1", "dir2/."). */
3466 		if (error == EISDIR && fvp->v_type == VDIR)
3467 			error = EINVAL;
3468 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3469 		vrele(fromnd.ni_dvp);
3470 		vrele(fvp);
3471 		goto out1;
3472 	}
3473 	tdvp = tond.ni_dvp;
3474 	tvp = tond.ni_vp;
3475 	error = vn_start_write(fvp, &mp, V_NOWAIT);
3476 	if (error != 0) {
3477 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3478 		NDFREE(&tond, NDF_ONLY_PNBUF);
3479 		if (tvp != NULL)
3480 			vput(tvp);
3481 		if (tdvp == tvp)
3482 			vrele(tdvp);
3483 		else
3484 			vput(tdvp);
3485 		vrele(fromnd.ni_dvp);
3486 		vrele(fvp);
3487 		vrele(tond.ni_startdir);
3488 		if (fromnd.ni_startdir != NULL)
3489 			vrele(fromnd.ni_startdir);
3490 		error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH);
3491 		if (error != 0)
3492 			return (error);
3493 		goto again;
3494 	}
3495 	if (tvp != NULL) {
3496 		if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
3497 			error = ENOTDIR;
3498 			goto out;
3499 		} else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
3500 			error = EISDIR;
3501 			goto out;
3502 		}
3503 #ifdef CAPABILITIES
3504 		if (newfd != AT_FDCWD) {
3505 			/*
3506 			 * If the target already exists we require CAP_UNLINKAT
3507 			 * from 'newfd'.
3508 			 */
3509 			error = cap_check(&tond.ni_filecaps.fc_rights,
3510 			    cap_rights_init(&rights, CAP_UNLINKAT));
3511 			if (error != 0)
3512 				goto out;
3513 		}
3514 #endif
3515 	}
3516 	if (fvp == tdvp) {
3517 		error = EINVAL;
3518 		goto out;
3519 	}
3520 	/*
3521 	 * If the source is the same as the destination (that is, if they
3522 	 * are links to the same vnode), then there is nothing to do.
3523 	 */
3524 	if (fvp == tvp)
3525 		error = -1;
3526 #ifdef MAC
3527 	else
3528 		error = mac_vnode_check_rename_to(td->td_ucred, tdvp,
3529 		    tond.ni_vp, fromnd.ni_dvp == tdvp, &tond.ni_cnd);
3530 #endif
3531 out:
3532 	if (error == 0) {
3533 		error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
3534 		    tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
3535 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3536 		NDFREE(&tond, NDF_ONLY_PNBUF);
3537 	} else {
3538 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3539 		NDFREE(&tond, NDF_ONLY_PNBUF);
3540 		if (tvp != NULL)
3541 			vput(tvp);
3542 		if (tdvp == tvp)
3543 			vrele(tdvp);
3544 		else
3545 			vput(tdvp);
3546 		vrele(fromnd.ni_dvp);
3547 		vrele(fvp);
3548 	}
3549 	vrele(tond.ni_startdir);
3550 	vn_finished_write(mp);
3551 out1:
3552 	if (fromnd.ni_startdir)
3553 		vrele(fromnd.ni_startdir);
3554 	if (error == -1)
3555 		return (0);
3556 	return (error);
3557 }
3558 
3559 /*
3560  * Make a directory file.
3561  */
3562 #ifndef _SYS_SYSPROTO_H_
3563 struct mkdir_args {
3564 	char	*path;
3565 	int	mode;
3566 };
3567 #endif
3568 int
3569 sys_mkdir(struct thread *td, struct mkdir_args *uap)
3570 {
3571 
3572 	return (kern_mkdirat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
3573 	    uap->mode));
3574 }
3575 
3576 #ifndef _SYS_SYSPROTO_H_
3577 struct mkdirat_args {
3578 	int	fd;
3579 	char	*path;
3580 	mode_t	mode;
3581 };
3582 #endif
3583 int
3584 sys_mkdirat(struct thread *td, struct mkdirat_args *uap)
3585 {
3586 
3587 	return (kern_mkdirat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode));
3588 }
3589 
3590 int
3591 kern_mkdirat(struct thread *td, int fd, char *path, enum uio_seg segflg,
3592     int mode)
3593 {
3594 	struct mount *mp;
3595 	struct vnode *vp;
3596 	struct vattr vattr;
3597 	struct nameidata nd;
3598 	cap_rights_t rights;
3599 	int error;
3600 
3601 	AUDIT_ARG_MODE(mode);
3602 restart:
3603 	bwillwrite();
3604 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
3605 	    NOCACHE, segflg, path, fd, cap_rights_init(&rights, CAP_MKDIRAT),
3606 	    td);
3607 	nd.ni_cnd.cn_flags |= WILLBEDIR;
3608 	if ((error = namei(&nd)) != 0)
3609 		return (error);
3610 	vp = nd.ni_vp;
3611 	if (vp != NULL) {
3612 		NDFREE(&nd, NDF_ONLY_PNBUF);
3613 		/*
3614 		 * XXX namei called with LOCKPARENT but not LOCKLEAF has
3615 		 * the strange behaviour of leaving the vnode unlocked
3616 		 * if the target is the same vnode as the parent.
3617 		 */
3618 		if (vp == nd.ni_dvp)
3619 			vrele(nd.ni_dvp);
3620 		else
3621 			vput(nd.ni_dvp);
3622 		vrele(vp);
3623 		return (EEXIST);
3624 	}
3625 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3626 		NDFREE(&nd, NDF_ONLY_PNBUF);
3627 		vput(nd.ni_dvp);
3628 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3629 			return (error);
3630 		goto restart;
3631 	}
3632 	VATTR_NULL(&vattr);
3633 	vattr.va_type = VDIR;
3634 	vattr.va_mode = (mode & ACCESSPERMS) &~ td->td_proc->p_fd->fd_cmask;
3635 #ifdef MAC
3636 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
3637 	    &vattr);
3638 	if (error != 0)
3639 		goto out;
3640 #endif
3641 	error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
3642 #ifdef MAC
3643 out:
3644 #endif
3645 	NDFREE(&nd, NDF_ONLY_PNBUF);
3646 	vput(nd.ni_dvp);
3647 	if (error == 0)
3648 		vput(nd.ni_vp);
3649 	vn_finished_write(mp);
3650 	return (error);
3651 }
3652 
3653 /*
3654  * Remove a directory file.
3655  */
3656 #ifndef _SYS_SYSPROTO_H_
3657 struct rmdir_args {
3658 	char	*path;
3659 };
3660 #endif
3661 int
3662 sys_rmdir(struct thread *td, struct rmdir_args *uap)
3663 {
3664 
3665 	return (kern_rmdirat(td, AT_FDCWD, uap->path, UIO_USERSPACE));
3666 }
3667 
3668 int
3669 kern_rmdirat(struct thread *td, int fd, char *path, enum uio_seg pathseg)
3670 {
3671 	struct mount *mp;
3672 	struct vnode *vp;
3673 	struct nameidata nd;
3674 	cap_rights_t rights;
3675 	int error;
3676 
3677 restart:
3678 	bwillwrite();
3679 	NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1,
3680 	    pathseg, path, fd, cap_rights_init(&rights, CAP_UNLINKAT), td);
3681 	if ((error = namei(&nd)) != 0)
3682 		return (error);
3683 	vp = nd.ni_vp;
3684 	if (vp->v_type != VDIR) {
3685 		error = ENOTDIR;
3686 		goto out;
3687 	}
3688 	/*
3689 	 * No rmdir "." please.
3690 	 */
3691 	if (nd.ni_dvp == vp) {
3692 		error = EINVAL;
3693 		goto out;
3694 	}
3695 	/*
3696 	 * The root of a mounted filesystem cannot be deleted.
3697 	 */
3698 	if (vp->v_vflag & VV_ROOT) {
3699 		error = EBUSY;
3700 		goto out;
3701 	}
3702 #ifdef MAC
3703 	error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
3704 	    &nd.ni_cnd);
3705 	if (error != 0)
3706 		goto out;
3707 #endif
3708 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3709 		NDFREE(&nd, NDF_ONLY_PNBUF);
3710 		vput(vp);
3711 		if (nd.ni_dvp == vp)
3712 			vrele(nd.ni_dvp);
3713 		else
3714 			vput(nd.ni_dvp);
3715 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3716 			return (error);
3717 		goto restart;
3718 	}
3719 	vfs_notify_upper(vp, VFS_NOTIFY_UPPER_UNLINK);
3720 	error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
3721 	vn_finished_write(mp);
3722 out:
3723 	NDFREE(&nd, NDF_ONLY_PNBUF);
3724 	vput(vp);
3725 	if (nd.ni_dvp == vp)
3726 		vrele(nd.ni_dvp);
3727 	else
3728 		vput(nd.ni_dvp);
3729 	return (error);
3730 }
3731 
3732 #if defined(COMPAT_43) || defined(COMPAT_FREEBSD11)
3733 int
3734 freebsd11_kern_getdirentries(struct thread *td, int fd, char *ubuf, u_int count,
3735     long *basep, void (*func)(struct freebsd11_dirent *))
3736 {
3737 	struct freebsd11_dirent dstdp;
3738 	struct dirent *dp, *edp;
3739 	char *dirbuf;
3740 	off_t base;
3741 	ssize_t resid, ucount;
3742 	int error;
3743 
3744 	/* XXX arbitrary sanity limit on `count'. */
3745 	count = min(count, 64 * 1024);
3746 
3747 	dirbuf = malloc(count, M_TEMP, M_WAITOK);
3748 
3749 	error = kern_getdirentries(td, fd, dirbuf, count, &base, &resid,
3750 	    UIO_SYSSPACE);
3751 	if (error != 0)
3752 		goto done;
3753 	if (basep != NULL)
3754 		*basep = base;
3755 
3756 	ucount = 0;
3757 	for (dp = (struct dirent *)dirbuf,
3758 	    edp = (struct dirent *)&dirbuf[count - resid];
3759 	    ucount < count && dp < edp; ) {
3760 		if (dp->d_reclen == 0)
3761 			break;
3762 		MPASS(dp->d_reclen >= _GENERIC_DIRLEN(0));
3763 		if (dp->d_namlen >= sizeof(dstdp.d_name))
3764 			continue;
3765 		dstdp.d_type = dp->d_type;
3766 		dstdp.d_namlen = dp->d_namlen;
3767 		dstdp.d_fileno = dp->d_fileno;		/* truncate */
3768 		if (dstdp.d_fileno != dp->d_fileno) {
3769 			switch (ino64_trunc_error) {
3770 			default:
3771 			case 0:
3772 				break;
3773 			case 1:
3774 				error = EOVERFLOW;
3775 				goto done;
3776 			case 2:
3777 				dstdp.d_fileno = UINT32_MAX;
3778 				break;
3779 			}
3780 		}
3781 		dstdp.d_reclen = sizeof(dstdp) - sizeof(dstdp.d_name) +
3782 		    ((dp->d_namlen + 1 + 3) &~ 3);
3783 		bcopy(dp->d_name, dstdp.d_name, dstdp.d_namlen);
3784 		bzero(dstdp.d_name + dstdp.d_namlen,
3785 		    dstdp.d_reclen - offsetof(struct freebsd11_dirent, d_name) -
3786 		    dstdp.d_namlen);
3787 		MPASS(dstdp.d_reclen <= dp->d_reclen);
3788 		MPASS(ucount + dstdp.d_reclen <= count);
3789 		if (func != NULL)
3790 			func(&dstdp);
3791 		error = copyout(&dstdp, ubuf + ucount, dstdp.d_reclen);
3792 		if (error != 0)
3793 			break;
3794 		dp = (struct dirent *)((char *)dp + dp->d_reclen);
3795 		ucount += dstdp.d_reclen;
3796 	}
3797 
3798 done:
3799 	free(dirbuf, M_TEMP);
3800 	if (error == 0)
3801 		td->td_retval[0] = ucount;
3802 	return (error);
3803 }
3804 #endif /* COMPAT */
3805 
3806 #ifdef COMPAT_43
3807 static void
3808 ogetdirentries_cvt(struct freebsd11_dirent *dp)
3809 {
3810 #if (BYTE_ORDER == LITTLE_ENDIAN)
3811 	/*
3812 	 * The expected low byte of dp->d_namlen is our dp->d_type.
3813 	 * The high MBZ byte of dp->d_namlen is our dp->d_namlen.
3814 	 */
3815 	dp->d_type = dp->d_namlen;
3816 	dp->d_namlen = 0;
3817 #else
3818 	/*
3819 	 * The dp->d_type is the high byte of the expected dp->d_namlen,
3820 	 * so must be zero'ed.
3821 	 */
3822 	dp->d_type = 0;
3823 #endif
3824 }
3825 
3826 /*
3827  * Read a block of directory entries in a filesystem independent format.
3828  */
3829 #ifndef _SYS_SYSPROTO_H_
3830 struct ogetdirentries_args {
3831 	int	fd;
3832 	char	*buf;
3833 	u_int	count;
3834 	long	*basep;
3835 };
3836 #endif
3837 int
3838 ogetdirentries(struct thread *td, struct ogetdirentries_args *uap)
3839 {
3840 	long loff;
3841 	int error;
3842 
3843 	error = kern_ogetdirentries(td, uap, &loff);
3844 	if (error == 0)
3845 		error = copyout(&loff, uap->basep, sizeof(long));
3846 	return (error);
3847 }
3848 
3849 int
3850 kern_ogetdirentries(struct thread *td, struct ogetdirentries_args *uap,
3851     long *ploff)
3852 {
3853 	long base;
3854 	int error;
3855 
3856 	/* XXX arbitrary sanity limit on `count'. */
3857 	if (uap->count > 64 * 1024)
3858 		return (EINVAL);
3859 
3860 	error = freebsd11_kern_getdirentries(td, uap->fd, uap->buf, uap->count,
3861 	    &base, ogetdirentries_cvt);
3862 
3863 	if (error == 0 && uap->basep != NULL)
3864 		error = copyout(&base, uap->basep, sizeof(long));
3865 
3866 	return (error);
3867 }
3868 #endif /* COMPAT_43 */
3869 
3870 #if defined(COMPAT_FREEBSD11)
3871 #ifndef _SYS_SYSPROTO_H_
3872 struct freebsd11_getdirentries_args {
3873 	int	fd;
3874 	char	*buf;
3875 	u_int	count;
3876 	long	*basep;
3877 };
3878 #endif
3879 int
3880 freebsd11_getdirentries(struct thread *td,
3881     struct freebsd11_getdirentries_args *uap)
3882 {
3883 	long base;
3884 	int error;
3885 
3886 	error = freebsd11_kern_getdirentries(td, uap->fd, uap->buf, uap->count,
3887 	    &base, NULL);
3888 
3889 	if (error == 0 && uap->basep != NULL)
3890 		error = copyout(&base, uap->basep, sizeof(long));
3891 	return (error);
3892 }
3893 
3894 int
3895 freebsd11_getdents(struct thread *td, struct freebsd11_getdents_args *uap)
3896 {
3897 	struct freebsd11_getdirentries_args ap;
3898 
3899 	ap.fd = uap->fd;
3900 	ap.buf = uap->buf;
3901 	ap.count = uap->count;
3902 	ap.basep = NULL;
3903 	return (freebsd11_getdirentries(td, &ap));
3904 }
3905 #endif /* COMPAT_FREEBSD11 */
3906 
3907 /*
3908  * Read a block of directory entries in a filesystem independent format.
3909  */
3910 int
3911 sys_getdirentries(struct thread *td, struct getdirentries_args *uap)
3912 {
3913 	off_t base;
3914 	int error;
3915 
3916 	error = kern_getdirentries(td, uap->fd, uap->buf, uap->count, &base,
3917 	    NULL, UIO_USERSPACE);
3918 	if (error != 0)
3919 		return (error);
3920 	if (uap->basep != NULL)
3921 		error = copyout(&base, uap->basep, sizeof(off_t));
3922 	return (error);
3923 }
3924 
3925 int
3926 kern_getdirentries(struct thread *td, int fd, char *buf, size_t count,
3927     off_t *basep, ssize_t *residp, enum uio_seg bufseg)
3928 {
3929 	struct vnode *vp;
3930 	struct file *fp;
3931 	struct uio auio;
3932 	struct iovec aiov;
3933 	cap_rights_t rights;
3934 	off_t loff;
3935 	int error, eofflag;
3936 	off_t foffset;
3937 
3938 	AUDIT_ARG_FD(fd);
3939 	if (count > IOSIZE_MAX)
3940 		return (EINVAL);
3941 	auio.uio_resid = count;
3942 	error = getvnode(td, fd, cap_rights_init(&rights, CAP_READ), &fp);
3943 	if (error != 0)
3944 		return (error);
3945 	if ((fp->f_flag & FREAD) == 0) {
3946 		fdrop(fp, td);
3947 		return (EBADF);
3948 	}
3949 	vp = fp->f_vnode;
3950 	foffset = foffset_lock(fp, 0);
3951 unionread:
3952 	if (vp->v_type != VDIR) {
3953 		error = EINVAL;
3954 		goto fail;
3955 	}
3956 	aiov.iov_base = buf;
3957 	aiov.iov_len = count;
3958 	auio.uio_iov = &aiov;
3959 	auio.uio_iovcnt = 1;
3960 	auio.uio_rw = UIO_READ;
3961 	auio.uio_segflg = bufseg;
3962 	auio.uio_td = td;
3963 	vn_lock(vp, LK_SHARED | LK_RETRY);
3964 	AUDIT_ARG_VNODE1(vp);
3965 	loff = auio.uio_offset = foffset;
3966 #ifdef MAC
3967 	error = mac_vnode_check_readdir(td->td_ucred, vp);
3968 	if (error == 0)
3969 #endif
3970 		error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, NULL,
3971 		    NULL);
3972 	foffset = auio.uio_offset;
3973 	if (error != 0) {
3974 		VOP_UNLOCK(vp, 0);
3975 		goto fail;
3976 	}
3977 	if (count == auio.uio_resid &&
3978 	    (vp->v_vflag & VV_ROOT) &&
3979 	    (vp->v_mount->mnt_flag & MNT_UNION)) {
3980 		struct vnode *tvp = vp;
3981 
3982 		vp = vp->v_mount->mnt_vnodecovered;
3983 		VREF(vp);
3984 		fp->f_vnode = vp;
3985 		fp->f_data = vp;
3986 		foffset = 0;
3987 		vput(tvp);
3988 		goto unionread;
3989 	}
3990 	VOP_UNLOCK(vp, 0);
3991 	*basep = loff;
3992 	if (residp != NULL)
3993 		*residp = auio.uio_resid;
3994 	td->td_retval[0] = count - auio.uio_resid;
3995 fail:
3996 	foffset_unlock(fp, foffset, 0);
3997 	fdrop(fp, td);
3998 	return (error);
3999 }
4000 
4001 /*
4002  * Set the mode mask for creation of filesystem nodes.
4003  */
4004 #ifndef _SYS_SYSPROTO_H_
4005 struct umask_args {
4006 	int	newmask;
4007 };
4008 #endif
4009 int
4010 sys_umask(struct thread *td, struct umask_args *uap)
4011 {
4012 	struct filedesc *fdp;
4013 
4014 	fdp = td->td_proc->p_fd;
4015 	FILEDESC_XLOCK(fdp);
4016 	td->td_retval[0] = fdp->fd_cmask;
4017 	fdp->fd_cmask = uap->newmask & ALLPERMS;
4018 	FILEDESC_XUNLOCK(fdp);
4019 	return (0);
4020 }
4021 
4022 /*
4023  * Void all references to file by ripping underlying filesystem away from
4024  * vnode.
4025  */
4026 #ifndef _SYS_SYSPROTO_H_
4027 struct revoke_args {
4028 	char	*path;
4029 };
4030 #endif
4031 int
4032 sys_revoke(struct thread *td, struct revoke_args *uap)
4033 {
4034 	struct vnode *vp;
4035 	struct vattr vattr;
4036 	struct nameidata nd;
4037 	int error;
4038 
4039 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
4040 	    uap->path, td);
4041 	if ((error = namei(&nd)) != 0)
4042 		return (error);
4043 	vp = nd.ni_vp;
4044 	NDFREE(&nd, NDF_ONLY_PNBUF);
4045 	if (vp->v_type != VCHR || vp->v_rdev == NULL) {
4046 		error = EINVAL;
4047 		goto out;
4048 	}
4049 #ifdef MAC
4050 	error = mac_vnode_check_revoke(td->td_ucred, vp);
4051 	if (error != 0)
4052 		goto out;
4053 #endif
4054 	error = VOP_GETATTR(vp, &vattr, td->td_ucred);
4055 	if (error != 0)
4056 		goto out;
4057 	if (td->td_ucred->cr_uid != vattr.va_uid) {
4058 		error = priv_check(td, PRIV_VFS_ADMIN);
4059 		if (error != 0)
4060 			goto out;
4061 	}
4062 	if (vcount(vp) > 1)
4063 		VOP_REVOKE(vp, REVOKEALL);
4064 out:
4065 	vput(vp);
4066 	return (error);
4067 }
4068 
4069 /*
4070  * Convert a user file descriptor to a kernel file entry and check that, if it
4071  * is a capability, the correct rights are present. A reference on the file
4072  * entry is held upon returning.
4073  */
4074 int
4075 getvnode(struct thread *td, int fd, cap_rights_t *rightsp, struct file **fpp)
4076 {
4077 	struct file *fp;
4078 	int error;
4079 
4080 	error = fget_unlocked(td->td_proc->p_fd, fd, rightsp, &fp, NULL);
4081 	if (error != 0)
4082 		return (error);
4083 
4084 	/*
4085 	 * The file could be not of the vnode type, or it may be not
4086 	 * yet fully initialized, in which case the f_vnode pointer
4087 	 * may be set, but f_ops is still badfileops.  E.g.,
4088 	 * devfs_open() transiently create such situation to
4089 	 * facilitate csw d_fdopen().
4090 	 *
4091 	 * Dupfdopen() handling in kern_openat() installs the
4092 	 * half-baked file into the process descriptor table, allowing
4093 	 * other thread to dereference it. Guard against the race by
4094 	 * checking f_ops.
4095 	 */
4096 	if (fp->f_vnode == NULL || fp->f_ops == &badfileops) {
4097 		fdrop(fp, td);
4098 		return (EINVAL);
4099 	}
4100 	*fpp = fp;
4101 	return (0);
4102 }
4103 
4104 
4105 /*
4106  * Get an (NFS) file handle.
4107  */
4108 #ifndef _SYS_SYSPROTO_H_
4109 struct lgetfh_args {
4110 	char	*fname;
4111 	fhandle_t *fhp;
4112 };
4113 #endif
4114 int
4115 sys_lgetfh(struct thread *td, struct lgetfh_args *uap)
4116 {
4117 	struct nameidata nd;
4118 	fhandle_t fh;
4119 	struct vnode *vp;
4120 	int error;
4121 
4122 	error = priv_check(td, PRIV_VFS_GETFH);
4123 	if (error != 0)
4124 		return (error);
4125 	NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
4126 	    uap->fname, td);
4127 	error = namei(&nd);
4128 	if (error != 0)
4129 		return (error);
4130 	NDFREE(&nd, NDF_ONLY_PNBUF);
4131 	vp = nd.ni_vp;
4132 	bzero(&fh, sizeof(fh));
4133 	fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
4134 	error = VOP_VPTOFH(vp, &fh.fh_fid);
4135 	vput(vp);
4136 	if (error == 0)
4137 		error = copyout(&fh, uap->fhp, sizeof (fh));
4138 	return (error);
4139 }
4140 
4141 #ifndef _SYS_SYSPROTO_H_
4142 struct getfh_args {
4143 	char	*fname;
4144 	fhandle_t *fhp;
4145 };
4146 #endif
4147 int
4148 sys_getfh(struct thread *td, struct getfh_args *uap)
4149 {
4150 	struct nameidata nd;
4151 	fhandle_t fh;
4152 	struct vnode *vp;
4153 	int error;
4154 
4155 	error = priv_check(td, PRIV_VFS_GETFH);
4156 	if (error != 0)
4157 		return (error);
4158 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
4159 	    uap->fname, td);
4160 	error = namei(&nd);
4161 	if (error != 0)
4162 		return (error);
4163 	NDFREE(&nd, NDF_ONLY_PNBUF);
4164 	vp = nd.ni_vp;
4165 	bzero(&fh, sizeof(fh));
4166 	fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
4167 	error = VOP_VPTOFH(vp, &fh.fh_fid);
4168 	vput(vp);
4169 	if (error == 0)
4170 		error = copyout(&fh, uap->fhp, sizeof (fh));
4171 	return (error);
4172 }
4173 
4174 /*
4175  * syscall for the rpc.lockd to use to translate a NFS file handle into an
4176  * open descriptor.
4177  *
4178  * warning: do not remove the priv_check() call or this becomes one giant
4179  * security hole.
4180  */
4181 #ifndef _SYS_SYSPROTO_H_
4182 struct fhopen_args {
4183 	const struct fhandle *u_fhp;
4184 	int flags;
4185 };
4186 #endif
4187 int
4188 sys_fhopen(struct thread *td, struct fhopen_args *uap)
4189 {
4190 	struct mount *mp;
4191 	struct vnode *vp;
4192 	struct fhandle fhp;
4193 	struct file *fp;
4194 	int fmode, error;
4195 	int indx;
4196 
4197 	error = priv_check(td, PRIV_VFS_FHOPEN);
4198 	if (error != 0)
4199 		return (error);
4200 	indx = -1;
4201 	fmode = FFLAGS(uap->flags);
4202 	/* why not allow a non-read/write open for our lockd? */
4203 	if (((fmode & (FREAD | FWRITE)) == 0) || (fmode & O_CREAT))
4204 		return (EINVAL);
4205 	error = copyin(uap->u_fhp, &fhp, sizeof(fhp));
4206 	if (error != 0)
4207 		return(error);
4208 	/* find the mount point */
4209 	mp = vfs_busyfs(&fhp.fh_fsid);
4210 	if (mp == NULL)
4211 		return (ESTALE);
4212 	/* now give me my vnode, it gets returned to me locked */
4213 	error = VFS_FHTOVP(mp, &fhp.fh_fid, LK_EXCLUSIVE, &vp);
4214 	vfs_unbusy(mp);
4215 	if (error != 0)
4216 		return (error);
4217 
4218 	error = falloc_noinstall(td, &fp);
4219 	if (error != 0) {
4220 		vput(vp);
4221 		return (error);
4222 	}
4223 	/*
4224 	 * An extra reference on `fp' has been held for us by
4225 	 * falloc_noinstall().
4226 	 */
4227 
4228 #ifdef INVARIANTS
4229 	td->td_dupfd = -1;
4230 #endif
4231 	error = vn_open_vnode(vp, fmode, td->td_ucred, td, fp);
4232 	if (error != 0) {
4233 		KASSERT(fp->f_ops == &badfileops,
4234 		    ("VOP_OPEN in fhopen() set f_ops"));
4235 		KASSERT(td->td_dupfd < 0,
4236 		    ("fhopen() encountered fdopen()"));
4237 
4238 		vput(vp);
4239 		goto bad;
4240 	}
4241 #ifdef INVARIANTS
4242 	td->td_dupfd = 0;
4243 #endif
4244 	fp->f_vnode = vp;
4245 	fp->f_seqcount = 1;
4246 	finit(fp, (fmode & FMASK) | (fp->f_flag & FHASLOCK), DTYPE_VNODE, vp,
4247 	    &vnops);
4248 	VOP_UNLOCK(vp, 0);
4249 	if ((fmode & O_TRUNC) != 0) {
4250 		error = fo_truncate(fp, 0, td->td_ucred, td);
4251 		if (error != 0)
4252 			goto bad;
4253 	}
4254 
4255 	error = finstall(td, fp, &indx, fmode, NULL);
4256 bad:
4257 	fdrop(fp, td);
4258 	td->td_retval[0] = indx;
4259 	return (error);
4260 }
4261 
4262 /*
4263  * Stat an (NFS) file handle.
4264  */
4265 #ifndef _SYS_SYSPROTO_H_
4266 struct fhstat_args {
4267 	struct fhandle *u_fhp;
4268 	struct stat *sb;
4269 };
4270 #endif
4271 int
4272 sys_fhstat(struct thread *td, struct fhstat_args *uap)
4273 {
4274 	struct stat sb;
4275 	struct fhandle fh;
4276 	int error;
4277 
4278 	error = copyin(uap->u_fhp, &fh, sizeof(fh));
4279 	if (error != 0)
4280 		return (error);
4281 	error = kern_fhstat(td, fh, &sb);
4282 	if (error == 0)
4283 		error = copyout(&sb, uap->sb, sizeof(sb));
4284 	return (error);
4285 }
4286 
4287 int
4288 kern_fhstat(struct thread *td, struct fhandle fh, struct stat *sb)
4289 {
4290 	struct mount *mp;
4291 	struct vnode *vp;
4292 	int error;
4293 
4294 	error = priv_check(td, PRIV_VFS_FHSTAT);
4295 	if (error != 0)
4296 		return (error);
4297 	if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4298 		return (ESTALE);
4299 	error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4300 	vfs_unbusy(mp);
4301 	if (error != 0)
4302 		return (error);
4303 	error = vn_stat(vp, sb, td->td_ucred, NOCRED, td);
4304 	vput(vp);
4305 	return (error);
4306 }
4307 
4308 /*
4309  * Implement fstatfs() for (NFS) file handles.
4310  */
4311 #ifndef _SYS_SYSPROTO_H_
4312 struct fhstatfs_args {
4313 	struct fhandle *u_fhp;
4314 	struct statfs *buf;
4315 };
4316 #endif
4317 int
4318 sys_fhstatfs(struct thread *td, struct fhstatfs_args *uap)
4319 {
4320 	struct statfs *sfp;
4321 	fhandle_t fh;
4322 	int error;
4323 
4324 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
4325 	if (error != 0)
4326 		return (error);
4327 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
4328 	error = kern_fhstatfs(td, fh, sfp);
4329 	if (error == 0)
4330 		error = copyout(sfp, uap->buf, sizeof(*sfp));
4331 	free(sfp, M_STATFS);
4332 	return (error);
4333 }
4334 
4335 int
4336 kern_fhstatfs(struct thread *td, fhandle_t fh, struct statfs *buf)
4337 {
4338 	struct statfs *sp;
4339 	struct mount *mp;
4340 	struct vnode *vp;
4341 	int error;
4342 
4343 	error = priv_check(td, PRIV_VFS_FHSTATFS);
4344 	if (error != 0)
4345 		return (error);
4346 	if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4347 		return (ESTALE);
4348 	error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4349 	if (error != 0) {
4350 		vfs_unbusy(mp);
4351 		return (error);
4352 	}
4353 	vput(vp);
4354 	error = prison_canseemount(td->td_ucred, mp);
4355 	if (error != 0)
4356 		goto out;
4357 #ifdef MAC
4358 	error = mac_mount_check_stat(td->td_ucred, mp);
4359 	if (error != 0)
4360 		goto out;
4361 #endif
4362 	/*
4363 	 * Set these in case the underlying filesystem fails to do so.
4364 	 */
4365 	sp = &mp->mnt_stat;
4366 	sp->f_version = STATFS_VERSION;
4367 	sp->f_namemax = NAME_MAX;
4368 	sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
4369 	error = VFS_STATFS(mp, sp);
4370 	if (error == 0)
4371 		*buf = *sp;
4372 out:
4373 	vfs_unbusy(mp);
4374 	return (error);
4375 }
4376 
4377 int
4378 kern_posix_fallocate(struct thread *td, int fd, off_t offset, off_t len)
4379 {
4380 	struct file *fp;
4381 	struct mount *mp;
4382 	struct vnode *vp;
4383 	cap_rights_t rights;
4384 	off_t olen, ooffset;
4385 	int error;
4386 #ifdef AUDIT
4387 	int audited_vnode1 = 0;
4388 #endif
4389 
4390 	AUDIT_ARG_FD(fd);
4391 	if (offset < 0 || len <= 0)
4392 		return (EINVAL);
4393 	/* Check for wrap. */
4394 	if (offset > OFF_MAX - len)
4395 		return (EFBIG);
4396 	AUDIT_ARG_FD(fd);
4397 	error = fget(td, fd, cap_rights_init(&rights, CAP_PWRITE), &fp);
4398 	if (error != 0)
4399 		return (error);
4400 	AUDIT_ARG_FILE(td->td_proc, fp);
4401 	if ((fp->f_ops->fo_flags & DFLAG_SEEKABLE) == 0) {
4402 		error = ESPIPE;
4403 		goto out;
4404 	}
4405 	if ((fp->f_flag & FWRITE) == 0) {
4406 		error = EBADF;
4407 		goto out;
4408 	}
4409 	if (fp->f_type != DTYPE_VNODE) {
4410 		error = ENODEV;
4411 		goto out;
4412 	}
4413 	vp = fp->f_vnode;
4414 	if (vp->v_type != VREG) {
4415 		error = ENODEV;
4416 		goto out;
4417 	}
4418 
4419 	/* Allocating blocks may take a long time, so iterate. */
4420 	for (;;) {
4421 		olen = len;
4422 		ooffset = offset;
4423 
4424 		bwillwrite();
4425 		mp = NULL;
4426 		error = vn_start_write(vp, &mp, V_WAIT | PCATCH);
4427 		if (error != 0)
4428 			break;
4429 		error = vn_lock(vp, LK_EXCLUSIVE);
4430 		if (error != 0) {
4431 			vn_finished_write(mp);
4432 			break;
4433 		}
4434 #ifdef AUDIT
4435 		if (!audited_vnode1) {
4436 			AUDIT_ARG_VNODE1(vp);
4437 			audited_vnode1 = 1;
4438 		}
4439 #endif
4440 #ifdef MAC
4441 		error = mac_vnode_check_write(td->td_ucred, fp->f_cred, vp);
4442 		if (error == 0)
4443 #endif
4444 			error = VOP_ALLOCATE(vp, &offset, &len);
4445 		VOP_UNLOCK(vp, 0);
4446 		vn_finished_write(mp);
4447 
4448 		if (olen + ooffset != offset + len) {
4449 			panic("offset + len changed from %jx/%jx to %jx/%jx",
4450 			    ooffset, olen, offset, len);
4451 		}
4452 		if (error != 0 || len == 0)
4453 			break;
4454 		KASSERT(olen > len, ("Iteration did not make progress?"));
4455 		maybe_yield();
4456 	}
4457  out:
4458 	fdrop(fp, td);
4459 	return (error);
4460 }
4461 
4462 int
4463 sys_posix_fallocate(struct thread *td, struct posix_fallocate_args *uap)
4464 {
4465 	int error;
4466 
4467 	error = kern_posix_fallocate(td, uap->fd, uap->offset, uap->len);
4468 	return (kern_posix_error(td, error));
4469 }
4470 
4471 /*
4472  * Unlike madvise(2), we do not make a best effort to remember every
4473  * possible caching hint.  Instead, we remember the last setting with
4474  * the exception that we will allow POSIX_FADV_NORMAL to adjust the
4475  * region of any current setting.
4476  */
4477 int
4478 kern_posix_fadvise(struct thread *td, int fd, off_t offset, off_t len,
4479     int advice)
4480 {
4481 	struct fadvise_info *fa, *new;
4482 	struct file *fp;
4483 	struct vnode *vp;
4484 	cap_rights_t rights;
4485 	off_t end;
4486 	int error;
4487 
4488 	if (offset < 0 || len < 0 || offset > OFF_MAX - len)
4489 		return (EINVAL);
4490 	AUDIT_ARG_VALUE(advice);
4491 	switch (advice) {
4492 	case POSIX_FADV_SEQUENTIAL:
4493 	case POSIX_FADV_RANDOM:
4494 	case POSIX_FADV_NOREUSE:
4495 		new = malloc(sizeof(*fa), M_FADVISE, M_WAITOK);
4496 		break;
4497 	case POSIX_FADV_NORMAL:
4498 	case POSIX_FADV_WILLNEED:
4499 	case POSIX_FADV_DONTNEED:
4500 		new = NULL;
4501 		break;
4502 	default:
4503 		return (EINVAL);
4504 	}
4505 	/* XXX: CAP_POSIX_FADVISE? */
4506 	AUDIT_ARG_FD(fd);
4507 	error = fget(td, fd, cap_rights_init(&rights), &fp);
4508 	if (error != 0)
4509 		goto out;
4510 	AUDIT_ARG_FILE(td->td_proc, fp);
4511 	if ((fp->f_ops->fo_flags & DFLAG_SEEKABLE) == 0) {
4512 		error = ESPIPE;
4513 		goto out;
4514 	}
4515 	if (fp->f_type != DTYPE_VNODE) {
4516 		error = ENODEV;
4517 		goto out;
4518 	}
4519 	vp = fp->f_vnode;
4520 	if (vp->v_type != VREG) {
4521 		error = ENODEV;
4522 		goto out;
4523 	}
4524 	if (len == 0)
4525 		end = OFF_MAX;
4526 	else
4527 		end = offset + len - 1;
4528 	switch (advice) {
4529 	case POSIX_FADV_SEQUENTIAL:
4530 	case POSIX_FADV_RANDOM:
4531 	case POSIX_FADV_NOREUSE:
4532 		/*
4533 		 * Try to merge any existing non-standard region with
4534 		 * this new region if possible, otherwise create a new
4535 		 * non-standard region for this request.
4536 		 */
4537 		mtx_pool_lock(mtxpool_sleep, fp);
4538 		fa = fp->f_advice;
4539 		if (fa != NULL && fa->fa_advice == advice &&
4540 		    ((fa->fa_start <= end && fa->fa_end >= offset) ||
4541 		    (end != OFF_MAX && fa->fa_start == end + 1) ||
4542 		    (fa->fa_end != OFF_MAX && fa->fa_end + 1 == offset))) {
4543 			if (offset < fa->fa_start)
4544 				fa->fa_start = offset;
4545 			if (end > fa->fa_end)
4546 				fa->fa_end = end;
4547 		} else {
4548 			new->fa_advice = advice;
4549 			new->fa_start = offset;
4550 			new->fa_end = end;
4551 			fp->f_advice = new;
4552 			new = fa;
4553 		}
4554 		mtx_pool_unlock(mtxpool_sleep, fp);
4555 		break;
4556 	case POSIX_FADV_NORMAL:
4557 		/*
4558 		 * If a the "normal" region overlaps with an existing
4559 		 * non-standard region, trim or remove the
4560 		 * non-standard region.
4561 		 */
4562 		mtx_pool_lock(mtxpool_sleep, fp);
4563 		fa = fp->f_advice;
4564 		if (fa != NULL) {
4565 			if (offset <= fa->fa_start && end >= fa->fa_end) {
4566 				new = fa;
4567 				fp->f_advice = NULL;
4568 			} else if (offset <= fa->fa_start &&
4569 			    end >= fa->fa_start)
4570 				fa->fa_start = end + 1;
4571 			else if (offset <= fa->fa_end && end >= fa->fa_end)
4572 				fa->fa_end = offset - 1;
4573 			else if (offset >= fa->fa_start && end <= fa->fa_end) {
4574 				/*
4575 				 * If the "normal" region is a middle
4576 				 * portion of the existing
4577 				 * non-standard region, just remove
4578 				 * the whole thing rather than picking
4579 				 * one side or the other to
4580 				 * preserve.
4581 				 */
4582 				new = fa;
4583 				fp->f_advice = NULL;
4584 			}
4585 		}
4586 		mtx_pool_unlock(mtxpool_sleep, fp);
4587 		break;
4588 	case POSIX_FADV_WILLNEED:
4589 	case POSIX_FADV_DONTNEED:
4590 		error = VOP_ADVISE(vp, offset, end, advice);
4591 		break;
4592 	}
4593 out:
4594 	if (fp != NULL)
4595 		fdrop(fp, td);
4596 	free(new, M_FADVISE);
4597 	return (error);
4598 }
4599 
4600 int
4601 sys_posix_fadvise(struct thread *td, struct posix_fadvise_args *uap)
4602 {
4603 	int error;
4604 
4605 	error = kern_posix_fadvise(td, uap->fd, uap->offset, uap->len,
4606 	    uap->advice);
4607 	return (kern_posix_error(td, error));
4608 }
4609