xref: /freebsd/sys/kern/vfs_syscalls.c (revision 77013d11e6483b970af25e13c9b892075742f7e5)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (c) 1989, 1993
5  *	The Regents of the University of California.  All rights reserved.
6  * (c) UNIX System Laboratories, Inc.
7  * All or some portions of this file are derived from material licensed
8  * to the University of California by American Telephone and Telegraph
9  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
10  * the permission of UNIX System Laboratories, Inc.
11  *
12  * Redistribution and use in source and binary forms, with or without
13  * modification, are permitted provided that the following conditions
14  * are met:
15  * 1. Redistributions of source code must retain the above copyright
16  *    notice, this list of conditions and the following disclaimer.
17  * 2. Redistributions in binary form must reproduce the above copyright
18  *    notice, this list of conditions and the following disclaimer in the
19  *    documentation and/or other materials provided with the distribution.
20  * 3. Neither the name of the University nor the names of its contributors
21  *    may be used to endorse or promote products derived from this software
22  *    without specific prior written permission.
23  *
24  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
25  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
28  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
29  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
30  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
31  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
32  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
33  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
34  * SUCH DAMAGE.
35  *
36  *	@(#)vfs_syscalls.c	8.13 (Berkeley) 4/15/94
37  */
38 
39 #include <sys/cdefs.h>
40 __FBSDID("$FreeBSD$");
41 
42 #include "opt_capsicum.h"
43 #include "opt_ktrace.h"
44 
45 #include <sys/param.h>
46 #include <sys/systm.h>
47 #include <sys/bio.h>
48 #include <sys/buf.h>
49 #include <sys/capsicum.h>
50 #include <sys/disk.h>
51 #include <sys/sysent.h>
52 #include <sys/malloc.h>
53 #include <sys/mount.h>
54 #include <sys/mutex.h>
55 #include <sys/sysproto.h>
56 #include <sys/namei.h>
57 #include <sys/filedesc.h>
58 #include <sys/kernel.h>
59 #include <sys/fcntl.h>
60 #include <sys/file.h>
61 #include <sys/filio.h>
62 #include <sys/limits.h>
63 #include <sys/linker.h>
64 #include <sys/rwlock.h>
65 #include <sys/sdt.h>
66 #include <sys/stat.h>
67 #include <sys/sx.h>
68 #include <sys/unistd.h>
69 #include <sys/vnode.h>
70 #include <sys/priv.h>
71 #include <sys/proc.h>
72 #include <sys/dirent.h>
73 #include <sys/jail.h>
74 #include <sys/syscallsubr.h>
75 #include <sys/sysctl.h>
76 #ifdef KTRACE
77 #include <sys/ktrace.h>
78 #endif
79 
80 #include <machine/stdarg.h>
81 
82 #include <security/audit/audit.h>
83 #include <security/mac/mac_framework.h>
84 
85 #include <vm/vm.h>
86 #include <vm/vm_object.h>
87 #include <vm/vm_page.h>
88 #include <vm/uma.h>
89 
90 #include <fs/devfs/devfs.h>
91 
92 MALLOC_DEFINE(M_FADVISE, "fadvise", "posix_fadvise(2) information");
93 
94 static int kern_chflagsat(struct thread *td, int fd, const char *path,
95     enum uio_seg pathseg, u_long flags, int atflag);
96 static int setfflags(struct thread *td, struct vnode *, u_long);
97 static int getutimes(const struct timeval *, enum uio_seg, struct timespec *);
98 static int getutimens(const struct timespec *, enum uio_seg,
99     struct timespec *, int *);
100 static int setutimes(struct thread *td, struct vnode *,
101     const struct timespec *, int, int);
102 static int vn_access(struct vnode *vp, int user_flags, struct ucred *cred,
103     struct thread *td);
104 static int kern_fhlinkat(struct thread *td, int fd, const char *path,
105     enum uio_seg pathseg, fhandle_t *fhp);
106 static int kern_readlink_vp(struct vnode *vp, char *buf, enum uio_seg bufseg,
107     size_t count, struct thread *td);
108 static int kern_linkat_vp(struct thread *td, struct vnode *vp, int fd,
109     const char *path, enum uio_seg segflag);
110 
111 static uint64_t
112 at2cnpflags(u_int at_flags, u_int mask)
113 {
114 	u_int64_t res;
115 
116 	MPASS((at_flags & (AT_SYMLINK_FOLLOW | AT_SYMLINK_NOFOLLOW)) !=
117 	    (AT_SYMLINK_FOLLOW | AT_SYMLINK_NOFOLLOW));
118 
119 	res = 0;
120 	at_flags &= mask;
121 	if ((at_flags & AT_RESOLVE_BENEATH) != 0)
122 		res |= RBENEATH;
123 	if ((at_flags & AT_SYMLINK_FOLLOW) != 0)
124 		res |= FOLLOW;
125 	/* NOFOLLOW is pseudo flag */
126 	if ((mask & AT_SYMLINK_NOFOLLOW) != 0) {
127 		res |= (at_flags & AT_SYMLINK_NOFOLLOW) != 0 ? NOFOLLOW :
128 		    FOLLOW;
129 	}
130 	if ((mask & AT_EMPTY_PATH) != 0 && (at_flags & AT_EMPTY_PATH) != 0)
131 		res |= EMPTYPATH;
132 	return (res);
133 }
134 
135 int
136 kern_sync(struct thread *td)
137 {
138 	struct mount *mp, *nmp;
139 	int save;
140 
141 	mtx_lock(&mountlist_mtx);
142 	for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
143 		if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK)) {
144 			nmp = TAILQ_NEXT(mp, mnt_list);
145 			continue;
146 		}
147 		if ((mp->mnt_flag & MNT_RDONLY) == 0 &&
148 		    vn_start_write(NULL, &mp, V_NOWAIT) == 0) {
149 			save = curthread_pflags_set(TDP_SYNCIO);
150 			vfs_periodic(mp, MNT_NOWAIT);
151 			VFS_SYNC(mp, MNT_NOWAIT);
152 			curthread_pflags_restore(save);
153 			vn_finished_write(mp);
154 		}
155 		mtx_lock(&mountlist_mtx);
156 		nmp = TAILQ_NEXT(mp, mnt_list);
157 		vfs_unbusy(mp);
158 	}
159 	mtx_unlock(&mountlist_mtx);
160 	return (0);
161 }
162 
163 /*
164  * Sync each mounted filesystem.
165  */
166 #ifndef _SYS_SYSPROTO_H_
167 struct sync_args {
168 	int     dummy;
169 };
170 #endif
171 /* ARGSUSED */
172 int
173 sys_sync(struct thread *td, struct sync_args *uap)
174 {
175 
176 	return (kern_sync(td));
177 }
178 
179 /*
180  * Change filesystem quotas.
181  */
182 #ifndef _SYS_SYSPROTO_H_
183 struct quotactl_args {
184 	char *path;
185 	int cmd;
186 	int uid;
187 	caddr_t arg;
188 };
189 #endif
190 int
191 sys_quotactl(struct thread *td, struct quotactl_args *uap)
192 {
193 	struct mount *mp;
194 	struct nameidata nd;
195 	int error;
196 	bool mp_busy;
197 
198 	AUDIT_ARG_CMD(uap->cmd);
199 	AUDIT_ARG_UID(uap->uid);
200 	if (!prison_allow(td->td_ucred, PR_ALLOW_QUOTAS))
201 		return (EPERM);
202 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
203 	    uap->path, td);
204 	if ((error = namei(&nd)) != 0)
205 		return (error);
206 	NDFREE(&nd, NDF_ONLY_PNBUF);
207 	mp = nd.ni_vp->v_mount;
208 	vfs_ref(mp);
209 	vput(nd.ni_vp);
210 	error = vfs_busy(mp, 0);
211 	if (error != 0) {
212 		vfs_rel(mp);
213 		return (error);
214 	}
215 	mp_busy = true;
216 	error = VFS_QUOTACTL(mp, uap->cmd, uap->uid, uap->arg, &mp_busy);
217 
218 	/*
219 	 * Since quota on/off operations typically need to open quota
220 	 * files, the implementation may need to unbusy the mount point
221 	 * before calling into namei.  Otherwise, unmount might be
222 	 * started between two vfs_busy() invocations (first is ours,
223 	 * second is from mount point cross-walk code in lookup()),
224 	 * causing deadlock.
225 	 *
226 	 * Avoid unbusying mp if the implementation indicates it has
227 	 * already done so.
228 	 */
229 	if (mp_busy)
230 		vfs_unbusy(mp);
231 	vfs_rel(mp);
232 	return (error);
233 }
234 
235 /*
236  * Used by statfs conversion routines to scale the block size up if
237  * necessary so that all of the block counts are <= 'max_size'.  Note
238  * that 'max_size' should be a bitmask, i.e. 2^n - 1 for some non-zero
239  * value of 'n'.
240  */
241 void
242 statfs_scale_blocks(struct statfs *sf, long max_size)
243 {
244 	uint64_t count;
245 	int shift;
246 
247 	KASSERT(powerof2(max_size + 1), ("%s: invalid max_size", __func__));
248 
249 	/*
250 	 * Attempt to scale the block counts to give a more accurate
251 	 * overview to userland of the ratio of free space to used
252 	 * space.  To do this, find the largest block count and compute
253 	 * a divisor that lets it fit into a signed integer <= max_size.
254 	 */
255 	if (sf->f_bavail < 0)
256 		count = -sf->f_bavail;
257 	else
258 		count = sf->f_bavail;
259 	count = MAX(sf->f_blocks, MAX(sf->f_bfree, count));
260 	if (count <= max_size)
261 		return;
262 
263 	count >>= flsl(max_size);
264 	shift = 0;
265 	while (count > 0) {
266 		shift++;
267 		count >>=1;
268 	}
269 
270 	sf->f_bsize <<= shift;
271 	sf->f_blocks >>= shift;
272 	sf->f_bfree >>= shift;
273 	sf->f_bavail >>= shift;
274 }
275 
276 static int
277 kern_do_statfs(struct thread *td, struct mount *mp, struct statfs *buf)
278 {
279 	int error;
280 
281 	if (mp == NULL)
282 		return (EBADF);
283 	error = vfs_busy(mp, 0);
284 	vfs_rel(mp);
285 	if (error != 0)
286 		return (error);
287 #ifdef MAC
288 	error = mac_mount_check_stat(td->td_ucred, mp);
289 	if (error != 0)
290 		goto out;
291 #endif
292 	error = VFS_STATFS(mp, buf);
293 	if (error != 0)
294 		goto out;
295 	if (priv_check_cred_vfs_generation(td->td_ucred)) {
296 		buf->f_fsid.val[0] = buf->f_fsid.val[1] = 0;
297 		prison_enforce_statfs(td->td_ucred, mp, buf);
298 	}
299 out:
300 	vfs_unbusy(mp);
301 	return (error);
302 }
303 
304 /*
305  * Get filesystem statistics.
306  */
307 #ifndef _SYS_SYSPROTO_H_
308 struct statfs_args {
309 	char *path;
310 	struct statfs *buf;
311 };
312 #endif
313 int
314 sys_statfs(struct thread *td, struct statfs_args *uap)
315 {
316 	struct statfs *sfp;
317 	int error;
318 
319 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
320 	error = kern_statfs(td, uap->path, UIO_USERSPACE, sfp);
321 	if (error == 0)
322 		error = copyout(sfp, uap->buf, sizeof(struct statfs));
323 	free(sfp, M_STATFS);
324 	return (error);
325 }
326 
327 int
328 kern_statfs(struct thread *td, const char *path, enum uio_seg pathseg,
329     struct statfs *buf)
330 {
331 	struct mount *mp;
332 	struct nameidata nd;
333 	int error;
334 
335 	NDINIT(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, td);
336 	error = namei(&nd);
337 	if (error != 0)
338 		return (error);
339 	mp = vfs_ref_from_vp(nd.ni_vp);
340 	NDFREE_NOTHING(&nd);
341 	vrele(nd.ni_vp);
342 	return (kern_do_statfs(td, mp, buf));
343 }
344 
345 /*
346  * Get filesystem statistics.
347  */
348 #ifndef _SYS_SYSPROTO_H_
349 struct fstatfs_args {
350 	int fd;
351 	struct statfs *buf;
352 };
353 #endif
354 int
355 sys_fstatfs(struct thread *td, struct fstatfs_args *uap)
356 {
357 	struct statfs *sfp;
358 	int error;
359 
360 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
361 	error = kern_fstatfs(td, uap->fd, sfp);
362 	if (error == 0)
363 		error = copyout(sfp, uap->buf, sizeof(struct statfs));
364 	free(sfp, M_STATFS);
365 	return (error);
366 }
367 
368 int
369 kern_fstatfs(struct thread *td, int fd, struct statfs *buf)
370 {
371 	struct file *fp;
372 	struct mount *mp;
373 	struct vnode *vp;
374 	int error;
375 
376 	AUDIT_ARG_FD(fd);
377 	error = getvnode_path(td, fd, &cap_fstatfs_rights, &fp);
378 	if (error != 0)
379 		return (error);
380 	vp = fp->f_vnode;
381 #ifdef AUDIT
382 	if (AUDITING_TD(td)) {
383 		vn_lock(vp, LK_SHARED | LK_RETRY);
384 		AUDIT_ARG_VNODE1(vp);
385 		VOP_UNLOCK(vp);
386 	}
387 #endif
388 	mp = vfs_ref_from_vp(vp);
389 	fdrop(fp, td);
390 	return (kern_do_statfs(td, mp, buf));
391 }
392 
393 /*
394  * Get statistics on all filesystems.
395  */
396 #ifndef _SYS_SYSPROTO_H_
397 struct getfsstat_args {
398 	struct statfs *buf;
399 	long bufsize;
400 	int mode;
401 };
402 #endif
403 int
404 sys_getfsstat(struct thread *td, struct getfsstat_args *uap)
405 {
406 	size_t count;
407 	int error;
408 
409 	if (uap->bufsize < 0 || uap->bufsize > SIZE_MAX)
410 		return (EINVAL);
411 	error = kern_getfsstat(td, &uap->buf, uap->bufsize, &count,
412 	    UIO_USERSPACE, uap->mode);
413 	if (error == 0)
414 		td->td_retval[0] = count;
415 	return (error);
416 }
417 
418 /*
419  * If (bufsize > 0 && bufseg == UIO_SYSSPACE)
420  *	The caller is responsible for freeing memory which will be allocated
421  *	in '*buf'.
422  */
423 int
424 kern_getfsstat(struct thread *td, struct statfs **buf, size_t bufsize,
425     size_t *countp, enum uio_seg bufseg, int mode)
426 {
427 	struct mount *mp, *nmp;
428 	struct statfs *sfsp, *sp, *sptmp, *tofree;
429 	size_t count, maxcount;
430 	int error;
431 
432 	switch (mode) {
433 	case MNT_WAIT:
434 	case MNT_NOWAIT:
435 		break;
436 	default:
437 		if (bufseg == UIO_SYSSPACE)
438 			*buf = NULL;
439 		return (EINVAL);
440 	}
441 restart:
442 	maxcount = bufsize / sizeof(struct statfs);
443 	if (bufsize == 0) {
444 		sfsp = NULL;
445 		tofree = NULL;
446 	} else if (bufseg == UIO_USERSPACE) {
447 		sfsp = *buf;
448 		tofree = NULL;
449 	} else /* if (bufseg == UIO_SYSSPACE) */ {
450 		count = 0;
451 		mtx_lock(&mountlist_mtx);
452 		TAILQ_FOREACH(mp, &mountlist, mnt_list) {
453 			count++;
454 		}
455 		mtx_unlock(&mountlist_mtx);
456 		if (maxcount > count)
457 			maxcount = count;
458 		tofree = sfsp = *buf = malloc(maxcount * sizeof(struct statfs),
459 		    M_STATFS, M_WAITOK);
460 	}
461 
462 	count = 0;
463 
464 	/*
465 	 * If there is no target buffer they only want the count.
466 	 *
467 	 * This could be TAILQ_FOREACH but it is open-coded to match the original
468 	 * code below.
469 	 */
470 	if (sfsp == NULL) {
471 		mtx_lock(&mountlist_mtx);
472 		for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
473 			if (prison_canseemount(td->td_ucred, mp) != 0) {
474 				nmp = TAILQ_NEXT(mp, mnt_list);
475 				continue;
476 			}
477 #ifdef MAC
478 			if (mac_mount_check_stat(td->td_ucred, mp) != 0) {
479 				nmp = TAILQ_NEXT(mp, mnt_list);
480 				continue;
481 			}
482 #endif
483 			count++;
484 			nmp = TAILQ_NEXT(mp, mnt_list);
485 		}
486 		mtx_unlock(&mountlist_mtx);
487 		*countp = count;
488 		return (0);
489 	}
490 
491 	/*
492 	 * They want the entire thing.
493 	 *
494 	 * Short-circuit the corner case of no room for anything, avoids
495 	 * relocking below.
496 	 */
497 	if (maxcount < 1) {
498 		goto out;
499 	}
500 
501 	mtx_lock(&mountlist_mtx);
502 	for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
503 		if (prison_canseemount(td->td_ucred, mp) != 0) {
504 			nmp = TAILQ_NEXT(mp, mnt_list);
505 			continue;
506 		}
507 #ifdef MAC
508 		if (mac_mount_check_stat(td->td_ucred, mp) != 0) {
509 			nmp = TAILQ_NEXT(mp, mnt_list);
510 			continue;
511 		}
512 #endif
513 		if (mode == MNT_WAIT) {
514 			if (vfs_busy(mp, MBF_MNTLSTLOCK) != 0) {
515 				/*
516 				 * If vfs_busy() failed, and MBF_NOWAIT
517 				 * wasn't passed, then the mp is gone.
518 				 * Furthermore, because of MBF_MNTLSTLOCK,
519 				 * the mountlist_mtx was dropped.  We have
520 				 * no other choice than to start over.
521 				 */
522 				mtx_unlock(&mountlist_mtx);
523 				free(tofree, M_STATFS);
524 				goto restart;
525 			}
526 		} else {
527 			if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK) != 0) {
528 				nmp = TAILQ_NEXT(mp, mnt_list);
529 				continue;
530 			}
531 		}
532 		sp = &mp->mnt_stat;
533 		/*
534 		 * If MNT_NOWAIT is specified, do not refresh
535 		 * the fsstat cache.
536 		 */
537 		if (mode != MNT_NOWAIT) {
538 			error = VFS_STATFS(mp, sp);
539 			if (error != 0) {
540 				mtx_lock(&mountlist_mtx);
541 				nmp = TAILQ_NEXT(mp, mnt_list);
542 				vfs_unbusy(mp);
543 				continue;
544 			}
545 		}
546 		if (priv_check_cred_vfs_generation(td->td_ucred)) {
547 			sptmp = malloc(sizeof(struct statfs), M_STATFS,
548 			    M_WAITOK);
549 			*sptmp = *sp;
550 			sptmp->f_fsid.val[0] = sptmp->f_fsid.val[1] = 0;
551 			prison_enforce_statfs(td->td_ucred, mp, sptmp);
552 			sp = sptmp;
553 		} else
554 			sptmp = NULL;
555 		if (bufseg == UIO_SYSSPACE) {
556 			bcopy(sp, sfsp, sizeof(*sp));
557 			free(sptmp, M_STATFS);
558 		} else /* if (bufseg == UIO_USERSPACE) */ {
559 			error = copyout(sp, sfsp, sizeof(*sp));
560 			free(sptmp, M_STATFS);
561 			if (error != 0) {
562 				vfs_unbusy(mp);
563 				return (error);
564 			}
565 		}
566 		sfsp++;
567 		count++;
568 
569 		if (count == maxcount) {
570 			vfs_unbusy(mp);
571 			goto out;
572 		}
573 
574 		mtx_lock(&mountlist_mtx);
575 		nmp = TAILQ_NEXT(mp, mnt_list);
576 		vfs_unbusy(mp);
577 	}
578 	mtx_unlock(&mountlist_mtx);
579 out:
580 	*countp = count;
581 	return (0);
582 }
583 
584 #ifdef COMPAT_FREEBSD4
585 /*
586  * Get old format filesystem statistics.
587  */
588 static void freebsd4_cvtstatfs(struct statfs *, struct ostatfs *);
589 
590 #ifndef _SYS_SYSPROTO_H_
591 struct freebsd4_statfs_args {
592 	char *path;
593 	struct ostatfs *buf;
594 };
595 #endif
596 int
597 freebsd4_statfs(struct thread *td, struct freebsd4_statfs_args *uap)
598 {
599 	struct ostatfs osb;
600 	struct statfs *sfp;
601 	int error;
602 
603 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
604 	error = kern_statfs(td, uap->path, UIO_USERSPACE, sfp);
605 	if (error == 0) {
606 		freebsd4_cvtstatfs(sfp, &osb);
607 		error = copyout(&osb, uap->buf, sizeof(osb));
608 	}
609 	free(sfp, M_STATFS);
610 	return (error);
611 }
612 
613 /*
614  * Get filesystem statistics.
615  */
616 #ifndef _SYS_SYSPROTO_H_
617 struct freebsd4_fstatfs_args {
618 	int fd;
619 	struct ostatfs *buf;
620 };
621 #endif
622 int
623 freebsd4_fstatfs(struct thread *td, struct freebsd4_fstatfs_args *uap)
624 {
625 	struct ostatfs osb;
626 	struct statfs *sfp;
627 	int error;
628 
629 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
630 	error = kern_fstatfs(td, uap->fd, sfp);
631 	if (error == 0) {
632 		freebsd4_cvtstatfs(sfp, &osb);
633 		error = copyout(&osb, uap->buf, sizeof(osb));
634 	}
635 	free(sfp, M_STATFS);
636 	return (error);
637 }
638 
639 /*
640  * Get statistics on all filesystems.
641  */
642 #ifndef _SYS_SYSPROTO_H_
643 struct freebsd4_getfsstat_args {
644 	struct ostatfs *buf;
645 	long bufsize;
646 	int mode;
647 };
648 #endif
649 int
650 freebsd4_getfsstat(struct thread *td, struct freebsd4_getfsstat_args *uap)
651 {
652 	struct statfs *buf, *sp;
653 	struct ostatfs osb;
654 	size_t count, size;
655 	int error;
656 
657 	if (uap->bufsize < 0)
658 		return (EINVAL);
659 	count = uap->bufsize / sizeof(struct ostatfs);
660 	if (count > SIZE_MAX / sizeof(struct statfs))
661 		return (EINVAL);
662 	size = count * sizeof(struct statfs);
663 	error = kern_getfsstat(td, &buf, size, &count, UIO_SYSSPACE,
664 	    uap->mode);
665 	if (error == 0)
666 		td->td_retval[0] = count;
667 	if (size != 0) {
668 		sp = buf;
669 		while (count != 0 && error == 0) {
670 			freebsd4_cvtstatfs(sp, &osb);
671 			error = copyout(&osb, uap->buf, sizeof(osb));
672 			sp++;
673 			uap->buf++;
674 			count--;
675 		}
676 		free(buf, M_STATFS);
677 	}
678 	return (error);
679 }
680 
681 /*
682  * Implement fstatfs() for (NFS) file handles.
683  */
684 #ifndef _SYS_SYSPROTO_H_
685 struct freebsd4_fhstatfs_args {
686 	struct fhandle *u_fhp;
687 	struct ostatfs *buf;
688 };
689 #endif
690 int
691 freebsd4_fhstatfs(struct thread *td, struct freebsd4_fhstatfs_args *uap)
692 {
693 	struct ostatfs osb;
694 	struct statfs *sfp;
695 	fhandle_t fh;
696 	int error;
697 
698 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
699 	if (error != 0)
700 		return (error);
701 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
702 	error = kern_fhstatfs(td, fh, sfp);
703 	if (error == 0) {
704 		freebsd4_cvtstatfs(sfp, &osb);
705 		error = copyout(&osb, uap->buf, sizeof(osb));
706 	}
707 	free(sfp, M_STATFS);
708 	return (error);
709 }
710 
711 /*
712  * Convert a new format statfs structure to an old format statfs structure.
713  */
714 static void
715 freebsd4_cvtstatfs(struct statfs *nsp, struct ostatfs *osp)
716 {
717 
718 	statfs_scale_blocks(nsp, LONG_MAX);
719 	bzero(osp, sizeof(*osp));
720 	osp->f_bsize = nsp->f_bsize;
721 	osp->f_iosize = MIN(nsp->f_iosize, LONG_MAX);
722 	osp->f_blocks = nsp->f_blocks;
723 	osp->f_bfree = nsp->f_bfree;
724 	osp->f_bavail = nsp->f_bavail;
725 	osp->f_files = MIN(nsp->f_files, LONG_MAX);
726 	osp->f_ffree = MIN(nsp->f_ffree, LONG_MAX);
727 	osp->f_owner = nsp->f_owner;
728 	osp->f_type = nsp->f_type;
729 	osp->f_flags = nsp->f_flags;
730 	osp->f_syncwrites = MIN(nsp->f_syncwrites, LONG_MAX);
731 	osp->f_asyncwrites = MIN(nsp->f_asyncwrites, LONG_MAX);
732 	osp->f_syncreads = MIN(nsp->f_syncreads, LONG_MAX);
733 	osp->f_asyncreads = MIN(nsp->f_asyncreads, LONG_MAX);
734 	strlcpy(osp->f_fstypename, nsp->f_fstypename,
735 	    MIN(MFSNAMELEN, OMFSNAMELEN));
736 	strlcpy(osp->f_mntonname, nsp->f_mntonname,
737 	    MIN(MNAMELEN, OMNAMELEN));
738 	strlcpy(osp->f_mntfromname, nsp->f_mntfromname,
739 	    MIN(MNAMELEN, OMNAMELEN));
740 	osp->f_fsid = nsp->f_fsid;
741 }
742 #endif /* COMPAT_FREEBSD4 */
743 
744 #if defined(COMPAT_FREEBSD11)
745 /*
746  * Get old format filesystem statistics.
747  */
748 static void freebsd11_cvtstatfs(struct statfs *, struct freebsd11_statfs *);
749 
750 int
751 freebsd11_statfs(struct thread *td, struct freebsd11_statfs_args *uap)
752 {
753 	struct freebsd11_statfs osb;
754 	struct statfs *sfp;
755 	int error;
756 
757 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
758 	error = kern_statfs(td, uap->path, UIO_USERSPACE, sfp);
759 	if (error == 0) {
760 		freebsd11_cvtstatfs(sfp, &osb);
761 		error = copyout(&osb, uap->buf, sizeof(osb));
762 	}
763 	free(sfp, M_STATFS);
764 	return (error);
765 }
766 
767 /*
768  * Get filesystem statistics.
769  */
770 int
771 freebsd11_fstatfs(struct thread *td, struct freebsd11_fstatfs_args *uap)
772 {
773 	struct freebsd11_statfs osb;
774 	struct statfs *sfp;
775 	int error;
776 
777 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
778 	error = kern_fstatfs(td, uap->fd, sfp);
779 	if (error == 0) {
780 		freebsd11_cvtstatfs(sfp, &osb);
781 		error = copyout(&osb, uap->buf, sizeof(osb));
782 	}
783 	free(sfp, M_STATFS);
784 	return (error);
785 }
786 
787 /*
788  * Get statistics on all filesystems.
789  */
790 int
791 freebsd11_getfsstat(struct thread *td, struct freebsd11_getfsstat_args *uap)
792 {
793 	struct freebsd11_statfs osb;
794 	struct statfs *buf, *sp;
795 	size_t count, size;
796 	int error;
797 
798 	count = uap->bufsize / sizeof(struct ostatfs);
799 	size = count * sizeof(struct statfs);
800 	error = kern_getfsstat(td, &buf, size, &count, UIO_SYSSPACE,
801 	    uap->mode);
802 	if (error == 0)
803 		td->td_retval[0] = count;
804 	if (size > 0) {
805 		sp = buf;
806 		while (count > 0 && error == 0) {
807 			freebsd11_cvtstatfs(sp, &osb);
808 			error = copyout(&osb, uap->buf, sizeof(osb));
809 			sp++;
810 			uap->buf++;
811 			count--;
812 		}
813 		free(buf, M_STATFS);
814 	}
815 	return (error);
816 }
817 
818 /*
819  * Implement fstatfs() for (NFS) file handles.
820  */
821 int
822 freebsd11_fhstatfs(struct thread *td, struct freebsd11_fhstatfs_args *uap)
823 {
824 	struct freebsd11_statfs osb;
825 	struct statfs *sfp;
826 	fhandle_t fh;
827 	int error;
828 
829 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
830 	if (error)
831 		return (error);
832 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
833 	error = kern_fhstatfs(td, fh, sfp);
834 	if (error == 0) {
835 		freebsd11_cvtstatfs(sfp, &osb);
836 		error = copyout(&osb, uap->buf, sizeof(osb));
837 	}
838 	free(sfp, M_STATFS);
839 	return (error);
840 }
841 
842 /*
843  * Convert a new format statfs structure to an old format statfs structure.
844  */
845 static void
846 freebsd11_cvtstatfs(struct statfs *nsp, struct freebsd11_statfs *osp)
847 {
848 
849 	bzero(osp, sizeof(*osp));
850 	osp->f_version = FREEBSD11_STATFS_VERSION;
851 	osp->f_type = nsp->f_type;
852 	osp->f_flags = nsp->f_flags;
853 	osp->f_bsize = nsp->f_bsize;
854 	osp->f_iosize = nsp->f_iosize;
855 	osp->f_blocks = nsp->f_blocks;
856 	osp->f_bfree = nsp->f_bfree;
857 	osp->f_bavail = nsp->f_bavail;
858 	osp->f_files = nsp->f_files;
859 	osp->f_ffree = nsp->f_ffree;
860 	osp->f_syncwrites = nsp->f_syncwrites;
861 	osp->f_asyncwrites = nsp->f_asyncwrites;
862 	osp->f_syncreads = nsp->f_syncreads;
863 	osp->f_asyncreads = nsp->f_asyncreads;
864 	osp->f_namemax = nsp->f_namemax;
865 	osp->f_owner = nsp->f_owner;
866 	osp->f_fsid = nsp->f_fsid;
867 	strlcpy(osp->f_fstypename, nsp->f_fstypename,
868 	    MIN(MFSNAMELEN, sizeof(osp->f_fstypename)));
869 	strlcpy(osp->f_mntonname, nsp->f_mntonname,
870 	    MIN(MNAMELEN, sizeof(osp->f_mntonname)));
871 	strlcpy(osp->f_mntfromname, nsp->f_mntfromname,
872 	    MIN(MNAMELEN, sizeof(osp->f_mntfromname)));
873 }
874 #endif /* COMPAT_FREEBSD11 */
875 
876 /*
877  * Change current working directory to a given file descriptor.
878  */
879 #ifndef _SYS_SYSPROTO_H_
880 struct fchdir_args {
881 	int	fd;
882 };
883 #endif
884 int
885 sys_fchdir(struct thread *td, struct fchdir_args *uap)
886 {
887 	struct vnode *vp, *tdp;
888 	struct mount *mp;
889 	struct file *fp;
890 	int error;
891 
892 	AUDIT_ARG_FD(uap->fd);
893 	error = getvnode_path(td, uap->fd, &cap_fchdir_rights,
894 	    &fp);
895 	if (error != 0)
896 		return (error);
897 	vp = fp->f_vnode;
898 	vref(vp);
899 	fdrop(fp, td);
900 	vn_lock(vp, LK_SHARED | LK_RETRY);
901 	AUDIT_ARG_VNODE1(vp);
902 	error = change_dir(vp, td);
903 	while (!error && (mp = vp->v_mountedhere) != NULL) {
904 		if (vfs_busy(mp, 0))
905 			continue;
906 		error = VFS_ROOT(mp, LK_SHARED, &tdp);
907 		vfs_unbusy(mp);
908 		if (error != 0)
909 			break;
910 		vput(vp);
911 		vp = tdp;
912 	}
913 	if (error != 0) {
914 		vput(vp);
915 		return (error);
916 	}
917 	VOP_UNLOCK(vp);
918 	pwd_chdir(td, vp);
919 	return (0);
920 }
921 
922 /*
923  * Change current working directory (``.'').
924  */
925 #ifndef _SYS_SYSPROTO_H_
926 struct chdir_args {
927 	char	*path;
928 };
929 #endif
930 int
931 sys_chdir(struct thread *td, struct chdir_args *uap)
932 {
933 
934 	return (kern_chdir(td, uap->path, UIO_USERSPACE));
935 }
936 
937 int
938 kern_chdir(struct thread *td, const char *path, enum uio_seg pathseg)
939 {
940 	struct nameidata nd;
941 	int error;
942 
943 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
944 	    pathseg, path, td);
945 	if ((error = namei(&nd)) != 0)
946 		return (error);
947 	if ((error = change_dir(nd.ni_vp, td)) != 0) {
948 		vput(nd.ni_vp);
949 		NDFREE_NOTHING(&nd);
950 		return (error);
951 	}
952 	VOP_UNLOCK(nd.ni_vp);
953 	NDFREE_NOTHING(&nd);
954 	pwd_chdir(td, nd.ni_vp);
955 	return (0);
956 }
957 
958 static int unprivileged_chroot = 0;
959 SYSCTL_INT(_security_bsd, OID_AUTO, unprivileged_chroot, CTLFLAG_RW,
960     &unprivileged_chroot, 0,
961     "Unprivileged processes can use chroot(2)");
962 /*
963  * Change notion of root (``/'') directory.
964  */
965 #ifndef _SYS_SYSPROTO_H_
966 struct chroot_args {
967 	char	*path;
968 };
969 #endif
970 int
971 sys_chroot(struct thread *td, struct chroot_args *uap)
972 {
973 	struct nameidata nd;
974 	struct proc *p;
975 	int error;
976 
977 	error = priv_check(td, PRIV_VFS_CHROOT);
978 	if (error != 0) {
979 		p = td->td_proc;
980 		PROC_LOCK(p);
981 		if (unprivileged_chroot == 0 ||
982 		    (p->p_flag2 & P2_NO_NEW_PRIVS) == 0) {
983 			PROC_UNLOCK(p);
984 			return (error);
985 		}
986 		PROC_UNLOCK(p);
987 	}
988 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
989 	    UIO_USERSPACE, uap->path, td);
990 	error = namei(&nd);
991 	if (error != 0)
992 		goto error;
993 	error = change_dir(nd.ni_vp, td);
994 	if (error != 0)
995 		goto e_vunlock;
996 #ifdef MAC
997 	error = mac_vnode_check_chroot(td->td_ucred, nd.ni_vp);
998 	if (error != 0)
999 		goto e_vunlock;
1000 #endif
1001 	VOP_UNLOCK(nd.ni_vp);
1002 	error = pwd_chroot(td, nd.ni_vp);
1003 	vrele(nd.ni_vp);
1004 	NDFREE_NOTHING(&nd);
1005 	return (error);
1006 e_vunlock:
1007 	vput(nd.ni_vp);
1008 error:
1009 	NDFREE_NOTHING(&nd);
1010 	return (error);
1011 }
1012 
1013 /*
1014  * Common routine for chroot and chdir.  Callers must provide a locked vnode
1015  * instance.
1016  */
1017 int
1018 change_dir(struct vnode *vp, struct thread *td)
1019 {
1020 #ifdef MAC
1021 	int error;
1022 #endif
1023 
1024 	ASSERT_VOP_LOCKED(vp, "change_dir(): vp not locked");
1025 	if (vp->v_type != VDIR)
1026 		return (ENOTDIR);
1027 #ifdef MAC
1028 	error = mac_vnode_check_chdir(td->td_ucred, vp);
1029 	if (error != 0)
1030 		return (error);
1031 #endif
1032 	return (VOP_ACCESS(vp, VEXEC, td->td_ucred, td));
1033 }
1034 
1035 static __inline void
1036 flags_to_rights(int flags, cap_rights_t *rightsp)
1037 {
1038 	if (flags & O_EXEC) {
1039 		cap_rights_set_one(rightsp, CAP_FEXECVE);
1040 		if (flags & O_PATH)
1041 			return;
1042 	} else {
1043 		switch ((flags & O_ACCMODE)) {
1044 		case O_RDONLY:
1045 			cap_rights_set_one(rightsp, CAP_READ);
1046 			break;
1047 		case O_RDWR:
1048 			cap_rights_set_one(rightsp, CAP_READ);
1049 			/* FALLTHROUGH */
1050 		case O_WRONLY:
1051 			cap_rights_set_one(rightsp, CAP_WRITE);
1052 			if (!(flags & (O_APPEND | O_TRUNC)))
1053 				cap_rights_set_one(rightsp, CAP_SEEK);
1054 			break;
1055 		}
1056 	}
1057 
1058 	if (flags & O_CREAT)
1059 		cap_rights_set_one(rightsp, CAP_CREATE);
1060 
1061 	if (flags & O_TRUNC)
1062 		cap_rights_set_one(rightsp, CAP_FTRUNCATE);
1063 
1064 	if (flags & (O_SYNC | O_FSYNC))
1065 		cap_rights_set_one(rightsp, CAP_FSYNC);
1066 
1067 	if (flags & (O_EXLOCK | O_SHLOCK))
1068 		cap_rights_set_one(rightsp, CAP_FLOCK);
1069 }
1070 
1071 /*
1072  * Check permissions, allocate an open file structure, and call the device
1073  * open routine if any.
1074  */
1075 #ifndef _SYS_SYSPROTO_H_
1076 struct open_args {
1077 	char	*path;
1078 	int	flags;
1079 	int	mode;
1080 };
1081 #endif
1082 int
1083 sys_open(struct thread *td, struct open_args *uap)
1084 {
1085 
1086 	return (kern_openat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1087 	    uap->flags, uap->mode));
1088 }
1089 
1090 #ifndef _SYS_SYSPROTO_H_
1091 struct openat_args {
1092 	int	fd;
1093 	char	*path;
1094 	int	flag;
1095 	int	mode;
1096 };
1097 #endif
1098 int
1099 sys_openat(struct thread *td, struct openat_args *uap)
1100 {
1101 
1102 	AUDIT_ARG_FD(uap->fd);
1103 	return (kern_openat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
1104 	    uap->mode));
1105 }
1106 
1107 int
1108 kern_openat(struct thread *td, int fd, const char *path, enum uio_seg pathseg,
1109     int flags, int mode)
1110 {
1111 	struct proc *p = td->td_proc;
1112 	struct filedesc *fdp;
1113 	struct pwddesc *pdp;
1114 	struct file *fp;
1115 	struct vnode *vp;
1116 	struct nameidata nd;
1117 	cap_rights_t rights;
1118 	int cmode, error, indx;
1119 
1120 	indx = -1;
1121 	fdp = p->p_fd;
1122 	pdp = p->p_pd;
1123 
1124 	AUDIT_ARG_FFLAGS(flags);
1125 	AUDIT_ARG_MODE(mode);
1126 	cap_rights_init_one(&rights, CAP_LOOKUP);
1127 	flags_to_rights(flags, &rights);
1128 
1129 	/*
1130 	 * Only one of the O_EXEC, O_RDONLY, O_WRONLY and O_RDWR flags
1131 	 * may be specified.  On the other hand, for O_PATH any mode
1132 	 * except O_EXEC is ignored.
1133 	 */
1134 	if ((flags & O_PATH) != 0) {
1135 		flags &= ~(O_CREAT | O_ACCMODE);
1136 	} else if ((flags & O_EXEC) != 0) {
1137 		if (flags & O_ACCMODE)
1138 			return (EINVAL);
1139 	} else if ((flags & O_ACCMODE) == O_ACCMODE) {
1140 		return (EINVAL);
1141 	} else {
1142 		flags = FFLAGS(flags);
1143 	}
1144 
1145 	/*
1146 	 * Allocate a file structure. The descriptor to reference it
1147 	 * is allocated and used by finstall_refed() below.
1148 	 */
1149 	error = falloc_noinstall(td, &fp);
1150 	if (error != 0)
1151 		return (error);
1152 	/* Set the flags early so the finit in devfs can pick them up. */
1153 	fp->f_flag = flags & FMASK;
1154 	cmode = ((mode & ~pdp->pd_cmask) & ALLPERMS) & ~S_ISTXT;
1155 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, fd,
1156 	    &rights, td);
1157 	td->td_dupfd = -1;		/* XXX check for fdopen */
1158 	error = vn_open(&nd, &flags, cmode, fp);
1159 	if (error != 0) {
1160 		/*
1161 		 * If the vn_open replaced the method vector, something
1162 		 * wonderous happened deep below and we just pass it up
1163 		 * pretending we know what we do.
1164 		 */
1165 		if (error == ENXIO && fp->f_ops != &badfileops) {
1166 			MPASS((flags & O_PATH) == 0);
1167 			goto success;
1168 		}
1169 
1170 		/*
1171 		 * Handle special fdopen() case. bleh.
1172 		 *
1173 		 * Don't do this for relative (capability) lookups; we don't
1174 		 * understand exactly what would happen, and we don't think
1175 		 * that it ever should.
1176 		 */
1177 		if ((nd.ni_resflags & NIRES_STRICTREL) == 0 &&
1178 		    (error == ENODEV || error == ENXIO) &&
1179 		    td->td_dupfd >= 0) {
1180 			error = dupfdopen(td, fdp, td->td_dupfd, flags, error,
1181 			    &indx);
1182 			if (error == 0)
1183 				goto success;
1184 		}
1185 
1186 		goto bad;
1187 	}
1188 	td->td_dupfd = 0;
1189 	NDFREE(&nd, NDF_ONLY_PNBUF);
1190 	vp = nd.ni_vp;
1191 
1192 	/*
1193 	 * Store the vnode, for any f_type. Typically, the vnode use
1194 	 * count is decremented by direct call to vn_closefile() for
1195 	 * files that switched type in the cdevsw fdopen() method.
1196 	 */
1197 	fp->f_vnode = vp;
1198 
1199 	/*
1200 	 * If the file wasn't claimed by devfs bind it to the normal
1201 	 * vnode operations here.
1202 	 */
1203 	if (fp->f_ops == &badfileops) {
1204 		KASSERT(vp->v_type != VFIFO || (flags & O_PATH) != 0,
1205 		    ("Unexpected fifo fp %p vp %p", fp, vp));
1206 		if ((flags & O_PATH) != 0) {
1207 			finit(fp, (flags & FMASK) | (fp->f_flag & FKQALLOWED),
1208 			    DTYPE_VNODE, NULL, &path_fileops);
1209 			vhold(vp);
1210 			vunref(vp);
1211 		} else {
1212 			finit_vnode(fp, flags, NULL, &vnops);
1213 		}
1214 	}
1215 
1216 	VOP_UNLOCK(vp);
1217 	if (flags & O_TRUNC) {
1218 		error = fo_truncate(fp, 0, td->td_ucred, td);
1219 		if (error != 0)
1220 			goto bad;
1221 	}
1222 success:
1223 	/*
1224 	 * If we haven't already installed the FD (for dupfdopen), do so now.
1225 	 */
1226 	if (indx == -1) {
1227 		struct filecaps *fcaps;
1228 
1229 #ifdef CAPABILITIES
1230 		if ((nd.ni_resflags & NIRES_STRICTREL) != 0)
1231 			fcaps = &nd.ni_filecaps;
1232 		else
1233 #endif
1234 			fcaps = NULL;
1235 		error = finstall_refed(td, fp, &indx, flags, fcaps);
1236 		/* On success finstall_refed() consumes fcaps. */
1237 		if (error != 0) {
1238 			filecaps_free(&nd.ni_filecaps);
1239 			goto bad;
1240 		}
1241 	} else {
1242 		filecaps_free(&nd.ni_filecaps);
1243 		falloc_abort(td, fp);
1244 	}
1245 
1246 	td->td_retval[0] = indx;
1247 	return (0);
1248 bad:
1249 	KASSERT(indx == -1, ("indx=%d, should be -1", indx));
1250 	falloc_abort(td, fp);
1251 	return (error);
1252 }
1253 
1254 #ifdef COMPAT_43
1255 /*
1256  * Create a file.
1257  */
1258 #ifndef _SYS_SYSPROTO_H_
1259 struct ocreat_args {
1260 	char	*path;
1261 	int	mode;
1262 };
1263 #endif
1264 int
1265 ocreat(struct thread *td, struct ocreat_args *uap)
1266 {
1267 
1268 	return (kern_openat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1269 	    O_WRONLY | O_CREAT | O_TRUNC, uap->mode));
1270 }
1271 #endif /* COMPAT_43 */
1272 
1273 /*
1274  * Create a special file.
1275  */
1276 #ifndef _SYS_SYSPROTO_H_
1277 struct mknodat_args {
1278 	int	fd;
1279 	char	*path;
1280 	mode_t	mode;
1281 	dev_t	dev;
1282 };
1283 #endif
1284 int
1285 sys_mknodat(struct thread *td, struct mknodat_args *uap)
1286 {
1287 
1288 	return (kern_mknodat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode,
1289 	    uap->dev));
1290 }
1291 
1292 #if defined(COMPAT_FREEBSD11)
1293 int
1294 freebsd11_mknod(struct thread *td,
1295     struct freebsd11_mknod_args *uap)
1296 {
1297 
1298 	return (kern_mknodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1299 	    uap->mode, uap->dev));
1300 }
1301 
1302 int
1303 freebsd11_mknodat(struct thread *td,
1304     struct freebsd11_mknodat_args *uap)
1305 {
1306 
1307 	return (kern_mknodat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode,
1308 	    uap->dev));
1309 }
1310 #endif /* COMPAT_FREEBSD11 */
1311 
1312 int
1313 kern_mknodat(struct thread *td, int fd, const char *path, enum uio_seg pathseg,
1314     int mode, dev_t dev)
1315 {
1316 	struct vnode *vp;
1317 	struct mount *mp;
1318 	struct vattr vattr;
1319 	struct nameidata nd;
1320 	int error, whiteout = 0;
1321 
1322 	AUDIT_ARG_MODE(mode);
1323 	AUDIT_ARG_DEV(dev);
1324 	switch (mode & S_IFMT) {
1325 	case S_IFCHR:
1326 	case S_IFBLK:
1327 		error = priv_check(td, PRIV_VFS_MKNOD_DEV);
1328 		if (error == 0 && dev == VNOVAL)
1329 			error = EINVAL;
1330 		break;
1331 	case S_IFWHT:
1332 		error = priv_check(td, PRIV_VFS_MKNOD_WHT);
1333 		break;
1334 	case S_IFIFO:
1335 		if (dev == 0)
1336 			return (kern_mkfifoat(td, fd, path, pathseg, mode));
1337 		/* FALLTHROUGH */
1338 	default:
1339 		error = EINVAL;
1340 		break;
1341 	}
1342 	if (error != 0)
1343 		return (error);
1344 	NDPREINIT(&nd);
1345 restart:
1346 	bwillwrite();
1347 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
1348 	    NOCACHE, pathseg, path, fd, &cap_mknodat_rights,
1349 	    td);
1350 	if ((error = namei(&nd)) != 0)
1351 		return (error);
1352 	vp = nd.ni_vp;
1353 	if (vp != NULL) {
1354 		NDFREE(&nd, NDF_ONLY_PNBUF);
1355 		if (vp == nd.ni_dvp)
1356 			vrele(nd.ni_dvp);
1357 		else
1358 			vput(nd.ni_dvp);
1359 		vrele(vp);
1360 		return (EEXIST);
1361 	} else {
1362 		VATTR_NULL(&vattr);
1363 		vattr.va_mode = (mode & ALLPERMS) &
1364 		    ~td->td_proc->p_pd->pd_cmask;
1365 		vattr.va_rdev = dev;
1366 		whiteout = 0;
1367 
1368 		switch (mode & S_IFMT) {
1369 		case S_IFCHR:
1370 			vattr.va_type = VCHR;
1371 			break;
1372 		case S_IFBLK:
1373 			vattr.va_type = VBLK;
1374 			break;
1375 		case S_IFWHT:
1376 			whiteout = 1;
1377 			break;
1378 		default:
1379 			panic("kern_mknod: invalid mode");
1380 		}
1381 	}
1382 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1383 		NDFREE(&nd, NDF_ONLY_PNBUF);
1384 		vput(nd.ni_dvp);
1385 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1386 			return (error);
1387 		goto restart;
1388 	}
1389 #ifdef MAC
1390 	if (error == 0 && !whiteout)
1391 		error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp,
1392 		    &nd.ni_cnd, &vattr);
1393 #endif
1394 	if (error == 0) {
1395 		if (whiteout)
1396 			error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE);
1397 		else {
1398 			error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp,
1399 						&nd.ni_cnd, &vattr);
1400 		}
1401 	}
1402 	VOP_VPUT_PAIR(nd.ni_dvp, error == 0 && !whiteout ? &nd.ni_vp : NULL,
1403 	    true);
1404 	vn_finished_write(mp);
1405 	NDFREE(&nd, NDF_ONLY_PNBUF);
1406 	if (error == ERELOOKUP)
1407 		goto restart;
1408 	return (error);
1409 }
1410 
1411 /*
1412  * Create a named pipe.
1413  */
1414 #ifndef _SYS_SYSPROTO_H_
1415 struct mkfifo_args {
1416 	char	*path;
1417 	int	mode;
1418 };
1419 #endif
1420 int
1421 sys_mkfifo(struct thread *td, struct mkfifo_args *uap)
1422 {
1423 
1424 	return (kern_mkfifoat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1425 	    uap->mode));
1426 }
1427 
1428 #ifndef _SYS_SYSPROTO_H_
1429 struct mkfifoat_args {
1430 	int	fd;
1431 	char	*path;
1432 	mode_t	mode;
1433 };
1434 #endif
1435 int
1436 sys_mkfifoat(struct thread *td, struct mkfifoat_args *uap)
1437 {
1438 
1439 	return (kern_mkfifoat(td, uap->fd, uap->path, UIO_USERSPACE,
1440 	    uap->mode));
1441 }
1442 
1443 int
1444 kern_mkfifoat(struct thread *td, int fd, const char *path,
1445     enum uio_seg pathseg, int mode)
1446 {
1447 	struct mount *mp;
1448 	struct vattr vattr;
1449 	struct nameidata nd;
1450 	int error;
1451 
1452 	AUDIT_ARG_MODE(mode);
1453 	NDPREINIT(&nd);
1454 restart:
1455 	bwillwrite();
1456 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
1457 	    NOCACHE, pathseg, path, fd, &cap_mkfifoat_rights,
1458 	    td);
1459 	if ((error = namei(&nd)) != 0)
1460 		return (error);
1461 	if (nd.ni_vp != NULL) {
1462 		NDFREE(&nd, NDF_ONLY_PNBUF);
1463 		if (nd.ni_vp == nd.ni_dvp)
1464 			vrele(nd.ni_dvp);
1465 		else
1466 			vput(nd.ni_dvp);
1467 		vrele(nd.ni_vp);
1468 		return (EEXIST);
1469 	}
1470 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1471 		NDFREE(&nd, NDF_ONLY_PNBUF);
1472 		vput(nd.ni_dvp);
1473 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1474 			return (error);
1475 		goto restart;
1476 	}
1477 	VATTR_NULL(&vattr);
1478 	vattr.va_type = VFIFO;
1479 	vattr.va_mode = (mode & ALLPERMS) & ~td->td_proc->p_pd->pd_cmask;
1480 #ifdef MAC
1481 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1482 	    &vattr);
1483 	if (error != 0)
1484 		goto out;
1485 #endif
1486 	error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
1487 #ifdef MAC
1488 out:
1489 #endif
1490 	VOP_VPUT_PAIR(nd.ni_dvp, error == 0 ? &nd.ni_vp : NULL, true);
1491 	vn_finished_write(mp);
1492 	NDFREE(&nd, NDF_ONLY_PNBUF);
1493 	if (error == ERELOOKUP)
1494 		goto restart;
1495 	return (error);
1496 }
1497 
1498 /*
1499  * Make a hard file link.
1500  */
1501 #ifndef _SYS_SYSPROTO_H_
1502 struct link_args {
1503 	char	*path;
1504 	char	*link;
1505 };
1506 #endif
1507 int
1508 sys_link(struct thread *td, struct link_args *uap)
1509 {
1510 
1511 	return (kern_linkat(td, AT_FDCWD, AT_FDCWD, uap->path, uap->link,
1512 	    UIO_USERSPACE, AT_SYMLINK_FOLLOW));
1513 }
1514 
1515 #ifndef _SYS_SYSPROTO_H_
1516 struct linkat_args {
1517 	int	fd1;
1518 	char	*path1;
1519 	int	fd2;
1520 	char	*path2;
1521 	int	flag;
1522 };
1523 #endif
1524 int
1525 sys_linkat(struct thread *td, struct linkat_args *uap)
1526 {
1527 	int flag;
1528 
1529 	flag = uap->flag;
1530 	if ((flag & ~(AT_SYMLINK_FOLLOW | AT_RESOLVE_BENEATH |
1531 	    AT_EMPTY_PATH)) != 0)
1532 		return (EINVAL);
1533 
1534 	return (kern_linkat(td, uap->fd1, uap->fd2, uap->path1, uap->path2,
1535 	    UIO_USERSPACE, flag));
1536 }
1537 
1538 int hardlink_check_uid = 0;
1539 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_uid, CTLFLAG_RW,
1540     &hardlink_check_uid, 0,
1541     "Unprivileged processes cannot create hard links to files owned by other "
1542     "users");
1543 static int hardlink_check_gid = 0;
1544 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_gid, CTLFLAG_RW,
1545     &hardlink_check_gid, 0,
1546     "Unprivileged processes cannot create hard links to files owned by other "
1547     "groups");
1548 
1549 static int
1550 can_hardlink(struct vnode *vp, struct ucred *cred)
1551 {
1552 	struct vattr va;
1553 	int error;
1554 
1555 	if (!hardlink_check_uid && !hardlink_check_gid)
1556 		return (0);
1557 
1558 	error = VOP_GETATTR(vp, &va, cred);
1559 	if (error != 0)
1560 		return (error);
1561 
1562 	if (hardlink_check_uid && cred->cr_uid != va.va_uid) {
1563 		error = priv_check_cred(cred, PRIV_VFS_LINK);
1564 		if (error != 0)
1565 			return (error);
1566 	}
1567 
1568 	if (hardlink_check_gid && !groupmember(va.va_gid, cred)) {
1569 		error = priv_check_cred(cred, PRIV_VFS_LINK);
1570 		if (error != 0)
1571 			return (error);
1572 	}
1573 
1574 	return (0);
1575 }
1576 
1577 int
1578 kern_linkat(struct thread *td, int fd1, int fd2, const char *path1,
1579     const char *path2, enum uio_seg segflag, int flag)
1580 {
1581 	struct nameidata nd;
1582 	int error;
1583 
1584 	NDPREINIT(&nd);
1585 	do {
1586 		bwillwrite();
1587 		NDINIT_ATRIGHTS(&nd, LOOKUP, AUDITVNODE1 | at2cnpflags(flag,
1588 		    AT_SYMLINK_FOLLOW | AT_RESOLVE_BENEATH | AT_EMPTY_PATH),
1589 		    segflag, path1, fd1, &cap_linkat_source_rights, td);
1590 		if ((error = namei(&nd)) != 0)
1591 			return (error);
1592 		NDFREE(&nd, NDF_ONLY_PNBUF);
1593 		if ((nd.ni_resflags & NIRES_EMPTYPATH) != 0) {
1594 			error = priv_check(td, PRIV_VFS_FHOPEN);
1595 			if (error != 0) {
1596 				vrele(nd.ni_vp);
1597 				return (error);
1598 			}
1599 		}
1600 		error = kern_linkat_vp(td, nd.ni_vp, fd2, path2, segflag);
1601 	} while (error ==  EAGAIN || error == ERELOOKUP);
1602 	return (error);
1603 }
1604 
1605 static int
1606 kern_linkat_vp(struct thread *td, struct vnode *vp, int fd, const char *path,
1607     enum uio_seg segflag)
1608 {
1609 	struct nameidata nd;
1610 	struct mount *mp;
1611 	int error;
1612 
1613 	if (vp->v_type == VDIR) {
1614 		vrele(vp);
1615 		return (EPERM);		/* POSIX */
1616 	}
1617 	NDINIT_ATRIGHTS(&nd, CREATE,
1618 	    LOCKPARENT | SAVENAME | AUDITVNODE2 | NOCACHE, segflag, path, fd,
1619 	    &cap_linkat_target_rights, td);
1620 	if ((error = namei(&nd)) == 0) {
1621 		if (nd.ni_vp != NULL) {
1622 			NDFREE(&nd, NDF_ONLY_PNBUF);
1623 			if (nd.ni_dvp == nd.ni_vp)
1624 				vrele(nd.ni_dvp);
1625 			else
1626 				vput(nd.ni_dvp);
1627 			vrele(nd.ni_vp);
1628 			vrele(vp);
1629 			return (EEXIST);
1630 		} else if (nd.ni_dvp->v_mount != vp->v_mount) {
1631 			/*
1632 			 * Cross-device link.  No need to recheck
1633 			 * vp->v_type, since it cannot change, except
1634 			 * to VBAD.
1635 			 */
1636 			NDFREE(&nd, NDF_ONLY_PNBUF);
1637 			vput(nd.ni_dvp);
1638 			vrele(vp);
1639 			return (EXDEV);
1640 		} else if ((error = vn_lock(vp, LK_EXCLUSIVE)) == 0) {
1641 			error = can_hardlink(vp, td->td_ucred);
1642 #ifdef MAC
1643 			if (error == 0)
1644 				error = mac_vnode_check_link(td->td_ucred,
1645 				    nd.ni_dvp, vp, &nd.ni_cnd);
1646 #endif
1647 			if (error != 0) {
1648 				vput(vp);
1649 				vput(nd.ni_dvp);
1650 				NDFREE(&nd, NDF_ONLY_PNBUF);
1651 				return (error);
1652 			}
1653 			error = vn_start_write(vp, &mp, V_NOWAIT);
1654 			if (error != 0) {
1655 				vput(vp);
1656 				vput(nd.ni_dvp);
1657 				NDFREE(&nd, NDF_ONLY_PNBUF);
1658 				error = vn_start_write(NULL, &mp,
1659 				    V_XSLEEP | PCATCH);
1660 				if (error != 0)
1661 					return (error);
1662 				return (EAGAIN);
1663 			}
1664 			error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
1665 			VOP_VPUT_PAIR(nd.ni_dvp, &vp, true);
1666 			vn_finished_write(mp);
1667 			NDFREE(&nd, NDF_ONLY_PNBUF);
1668 			vp = NULL;
1669 		} else {
1670 			vput(nd.ni_dvp);
1671 			NDFREE(&nd, NDF_ONLY_PNBUF);
1672 			vrele(vp);
1673 			return (EAGAIN);
1674 		}
1675 	}
1676 	if (vp != NULL)
1677 		vrele(vp);
1678 	return (error);
1679 }
1680 
1681 /*
1682  * Make a symbolic link.
1683  */
1684 #ifndef _SYS_SYSPROTO_H_
1685 struct symlink_args {
1686 	char	*path;
1687 	char	*link;
1688 };
1689 #endif
1690 int
1691 sys_symlink(struct thread *td, struct symlink_args *uap)
1692 {
1693 
1694 	return (kern_symlinkat(td, uap->path, AT_FDCWD, uap->link,
1695 	    UIO_USERSPACE));
1696 }
1697 
1698 #ifndef _SYS_SYSPROTO_H_
1699 struct symlinkat_args {
1700 	char	*path;
1701 	int	fd;
1702 	char	*path2;
1703 };
1704 #endif
1705 int
1706 sys_symlinkat(struct thread *td, struct symlinkat_args *uap)
1707 {
1708 
1709 	return (kern_symlinkat(td, uap->path1, uap->fd, uap->path2,
1710 	    UIO_USERSPACE));
1711 }
1712 
1713 int
1714 kern_symlinkat(struct thread *td, const char *path1, int fd, const char *path2,
1715     enum uio_seg segflg)
1716 {
1717 	struct mount *mp;
1718 	struct vattr vattr;
1719 	const char *syspath;
1720 	char *tmppath;
1721 	struct nameidata nd;
1722 	int error;
1723 
1724 	if (segflg == UIO_SYSSPACE) {
1725 		syspath = path1;
1726 	} else {
1727 		tmppath = uma_zalloc(namei_zone, M_WAITOK);
1728 		if ((error = copyinstr(path1, tmppath, MAXPATHLEN, NULL)) != 0)
1729 			goto out;
1730 		syspath = tmppath;
1731 	}
1732 	AUDIT_ARG_TEXT(syspath);
1733 	NDPREINIT(&nd);
1734 restart:
1735 	bwillwrite();
1736 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
1737 	    NOCACHE, segflg, path2, fd, &cap_symlinkat_rights,
1738 	    td);
1739 	if ((error = namei(&nd)) != 0)
1740 		goto out;
1741 	if (nd.ni_vp) {
1742 		NDFREE(&nd, NDF_ONLY_PNBUF);
1743 		if (nd.ni_vp == nd.ni_dvp)
1744 			vrele(nd.ni_dvp);
1745 		else
1746 			vput(nd.ni_dvp);
1747 		vrele(nd.ni_vp);
1748 		nd.ni_vp = NULL;
1749 		error = EEXIST;
1750 		goto out;
1751 	}
1752 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1753 		NDFREE(&nd, NDF_ONLY_PNBUF);
1754 		vput(nd.ni_dvp);
1755 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1756 			goto out;
1757 		goto restart;
1758 	}
1759 	VATTR_NULL(&vattr);
1760 	vattr.va_mode = ACCESSPERMS &~ td->td_proc->p_pd->pd_cmask;
1761 #ifdef MAC
1762 	vattr.va_type = VLNK;
1763 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1764 	    &vattr);
1765 	if (error != 0)
1766 		goto out2;
1767 #endif
1768 	error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, syspath);
1769 #ifdef MAC
1770 out2:
1771 #endif
1772 	VOP_VPUT_PAIR(nd.ni_dvp, error == 0 ? &nd.ni_vp : NULL, true);
1773 	vn_finished_write(mp);
1774 	NDFREE(&nd, NDF_ONLY_PNBUF);
1775 	if (error == ERELOOKUP)
1776 		goto restart;
1777 out:
1778 	if (segflg != UIO_SYSSPACE)
1779 		uma_zfree(namei_zone, tmppath);
1780 	return (error);
1781 }
1782 
1783 /*
1784  * Delete a whiteout from the filesystem.
1785  */
1786 #ifndef _SYS_SYSPROTO_H_
1787 struct undelete_args {
1788 	char *path;
1789 };
1790 #endif
1791 int
1792 sys_undelete(struct thread *td, struct undelete_args *uap)
1793 {
1794 	struct mount *mp;
1795 	struct nameidata nd;
1796 	int error;
1797 
1798 	NDPREINIT(&nd);
1799 restart:
1800 	bwillwrite();
1801 	NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | AUDITVNODE1,
1802 	    UIO_USERSPACE, uap->path, td);
1803 	error = namei(&nd);
1804 	if (error != 0)
1805 		return (error);
1806 
1807 	if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) {
1808 		NDFREE(&nd, NDF_ONLY_PNBUF);
1809 		if (nd.ni_vp == nd.ni_dvp)
1810 			vrele(nd.ni_dvp);
1811 		else
1812 			vput(nd.ni_dvp);
1813 		if (nd.ni_vp)
1814 			vrele(nd.ni_vp);
1815 		return (EEXIST);
1816 	}
1817 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1818 		NDFREE(&nd, NDF_ONLY_PNBUF);
1819 		vput(nd.ni_dvp);
1820 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1821 			return (error);
1822 		goto restart;
1823 	}
1824 	error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE);
1825 	NDFREE(&nd, NDF_ONLY_PNBUF);
1826 	vput(nd.ni_dvp);
1827 	vn_finished_write(mp);
1828 	if (error == ERELOOKUP)
1829 		goto restart;
1830 	return (error);
1831 }
1832 
1833 /*
1834  * Delete a name from the filesystem.
1835  */
1836 #ifndef _SYS_SYSPROTO_H_
1837 struct unlink_args {
1838 	char	*path;
1839 };
1840 #endif
1841 int
1842 sys_unlink(struct thread *td, struct unlink_args *uap)
1843 {
1844 
1845 	return (kern_funlinkat(td, AT_FDCWD, uap->path, FD_NONE, UIO_USERSPACE,
1846 	    0, 0));
1847 }
1848 
1849 static int
1850 kern_funlinkat_ex(struct thread *td, int dfd, const char *path, int fd,
1851     int flag, enum uio_seg pathseg, ino_t oldinum)
1852 {
1853 
1854 	if ((flag & ~(AT_REMOVEDIR | AT_RESOLVE_BENEATH)) != 0)
1855 		return (EINVAL);
1856 
1857 	if ((flag & AT_REMOVEDIR) != 0)
1858 		return (kern_frmdirat(td, dfd, path, fd, UIO_USERSPACE, 0));
1859 
1860 	return (kern_funlinkat(td, dfd, path, fd, UIO_USERSPACE, 0, 0));
1861 }
1862 
1863 #ifndef _SYS_SYSPROTO_H_
1864 struct unlinkat_args {
1865 	int	fd;
1866 	char	*path;
1867 	int	flag;
1868 };
1869 #endif
1870 int
1871 sys_unlinkat(struct thread *td, struct unlinkat_args *uap)
1872 {
1873 
1874 	return (kern_funlinkat_ex(td, uap->fd, uap->path, FD_NONE, uap->flag,
1875 	    UIO_USERSPACE, 0));
1876 }
1877 
1878 #ifndef _SYS_SYSPROTO_H_
1879 struct funlinkat_args {
1880 	int		dfd;
1881 	const char	*path;
1882 	int		fd;
1883 	int		flag;
1884 };
1885 #endif
1886 int
1887 sys_funlinkat(struct thread *td, struct funlinkat_args *uap)
1888 {
1889 
1890 	return (kern_funlinkat_ex(td, uap->dfd, uap->path, uap->fd, uap->flag,
1891 	    UIO_USERSPACE, 0));
1892 }
1893 
1894 int
1895 kern_funlinkat(struct thread *td, int dfd, const char *path, int fd,
1896     enum uio_seg pathseg, int flag, ino_t oldinum)
1897 {
1898 	struct mount *mp;
1899 	struct file *fp;
1900 	struct vnode *vp;
1901 	struct nameidata nd;
1902 	struct stat sb;
1903 	int error;
1904 
1905 	fp = NULL;
1906 	if (fd != FD_NONE) {
1907 		error = getvnode_path(td, fd, &cap_no_rights, &fp);
1908 		if (error != 0)
1909 			return (error);
1910 	}
1911 
1912 	NDPREINIT(&nd);
1913 restart:
1914 	bwillwrite();
1915 	NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1 |
1916 	    at2cnpflags(flag, AT_RESOLVE_BENEATH),
1917 	    pathseg, path, dfd, &cap_unlinkat_rights, td);
1918 	if ((error = namei(&nd)) != 0) {
1919 		if (error == EINVAL)
1920 			error = EPERM;
1921 		goto fdout;
1922 	}
1923 	vp = nd.ni_vp;
1924 	if (vp->v_type == VDIR && oldinum == 0) {
1925 		error = EPERM;		/* POSIX */
1926 	} else if (oldinum != 0 &&
1927 	    ((error = VOP_STAT(vp, &sb, td->td_ucred, NOCRED, td)) == 0) &&
1928 	    sb.st_ino != oldinum) {
1929 		error = EIDRM;	/* Identifier removed */
1930 	} else if (fp != NULL && fp->f_vnode != vp) {
1931 		if (VN_IS_DOOMED(fp->f_vnode))
1932 			error = EBADF;
1933 		else
1934 			error = EDEADLK;
1935 	} else {
1936 		/*
1937 		 * The root of a mounted filesystem cannot be deleted.
1938 		 *
1939 		 * XXX: can this only be a VDIR case?
1940 		 */
1941 		if (vp->v_vflag & VV_ROOT)
1942 			error = EBUSY;
1943 	}
1944 	if (error == 0) {
1945 		if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1946 			NDFREE(&nd, NDF_ONLY_PNBUF);
1947 			vput(nd.ni_dvp);
1948 			if (vp == nd.ni_dvp)
1949 				vrele(vp);
1950 			else
1951 				vput(vp);
1952 			if ((error = vn_start_write(NULL, &mp,
1953 			    V_XSLEEP | PCATCH)) != 0) {
1954 				goto fdout;
1955 			}
1956 			goto restart;
1957 		}
1958 #ifdef MAC
1959 		error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
1960 		    &nd.ni_cnd);
1961 		if (error != 0)
1962 			goto out;
1963 #endif
1964 		vfs_notify_upper(vp, VFS_NOTIFY_UPPER_UNLINK);
1965 		error = VOP_REMOVE(nd.ni_dvp, vp, &nd.ni_cnd);
1966 #ifdef MAC
1967 out:
1968 #endif
1969 		vn_finished_write(mp);
1970 	}
1971 	NDFREE(&nd, NDF_ONLY_PNBUF);
1972 	vput(nd.ni_dvp);
1973 	if (vp == nd.ni_dvp)
1974 		vrele(vp);
1975 	else
1976 		vput(vp);
1977 	if (error == ERELOOKUP)
1978 		goto restart;
1979 fdout:
1980 	if (fp != NULL)
1981 		fdrop(fp, td);
1982 	return (error);
1983 }
1984 
1985 /*
1986  * Reposition read/write file offset.
1987  */
1988 #ifndef _SYS_SYSPROTO_H_
1989 struct lseek_args {
1990 	int	fd;
1991 	int	pad;
1992 	off_t	offset;
1993 	int	whence;
1994 };
1995 #endif
1996 int
1997 sys_lseek(struct thread *td, struct lseek_args *uap)
1998 {
1999 
2000 	return (kern_lseek(td, uap->fd, uap->offset, uap->whence));
2001 }
2002 
2003 int
2004 kern_lseek(struct thread *td, int fd, off_t offset, int whence)
2005 {
2006 	struct file *fp;
2007 	int error;
2008 
2009 	AUDIT_ARG_FD(fd);
2010 	error = fget(td, fd, &cap_seek_rights, &fp);
2011 	if (error != 0)
2012 		return (error);
2013 	error = (fp->f_ops->fo_flags & DFLAG_SEEKABLE) != 0 ?
2014 	    fo_seek(fp, offset, whence, td) : ESPIPE;
2015 	fdrop(fp, td);
2016 	return (error);
2017 }
2018 
2019 #if defined(COMPAT_43)
2020 /*
2021  * Reposition read/write file offset.
2022  */
2023 #ifndef _SYS_SYSPROTO_H_
2024 struct olseek_args {
2025 	int	fd;
2026 	long	offset;
2027 	int	whence;
2028 };
2029 #endif
2030 int
2031 olseek(struct thread *td, struct olseek_args *uap)
2032 {
2033 
2034 	return (kern_lseek(td, uap->fd, uap->offset, uap->whence));
2035 }
2036 #endif /* COMPAT_43 */
2037 
2038 #if defined(COMPAT_FREEBSD6)
2039 /* Version with the 'pad' argument */
2040 int
2041 freebsd6_lseek(struct thread *td, struct freebsd6_lseek_args *uap)
2042 {
2043 
2044 	return (kern_lseek(td, uap->fd, uap->offset, uap->whence));
2045 }
2046 #endif
2047 
2048 /*
2049  * Check access permissions using passed credentials.
2050  */
2051 static int
2052 vn_access(struct vnode *vp, int user_flags, struct ucred *cred,
2053      struct thread *td)
2054 {
2055 	accmode_t accmode;
2056 	int error;
2057 
2058 	/* Flags == 0 means only check for existence. */
2059 	if (user_flags == 0)
2060 		return (0);
2061 
2062 	accmode = 0;
2063 	if (user_flags & R_OK)
2064 		accmode |= VREAD;
2065 	if (user_flags & W_OK)
2066 		accmode |= VWRITE;
2067 	if (user_flags & X_OK)
2068 		accmode |= VEXEC;
2069 #ifdef MAC
2070 	error = mac_vnode_check_access(cred, vp, accmode);
2071 	if (error != 0)
2072 		return (error);
2073 #endif
2074 	if ((accmode & VWRITE) == 0 || (error = vn_writechk(vp)) == 0)
2075 		error = VOP_ACCESS(vp, accmode, cred, td);
2076 	return (error);
2077 }
2078 
2079 /*
2080  * Check access permissions using "real" credentials.
2081  */
2082 #ifndef _SYS_SYSPROTO_H_
2083 struct access_args {
2084 	char	*path;
2085 	int	amode;
2086 };
2087 #endif
2088 int
2089 sys_access(struct thread *td, struct access_args *uap)
2090 {
2091 
2092 	return (kern_accessat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2093 	    0, uap->amode));
2094 }
2095 
2096 #ifndef _SYS_SYSPROTO_H_
2097 struct faccessat_args {
2098 	int	dirfd;
2099 	char	*path;
2100 	int	amode;
2101 	int	flag;
2102 }
2103 #endif
2104 int
2105 sys_faccessat(struct thread *td, struct faccessat_args *uap)
2106 {
2107 
2108 	return (kern_accessat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
2109 	    uap->amode));
2110 }
2111 
2112 int
2113 kern_accessat(struct thread *td, int fd, const char *path,
2114     enum uio_seg pathseg, int flag, int amode)
2115 {
2116 	struct ucred *cred, *usecred;
2117 	struct vnode *vp;
2118 	struct nameidata nd;
2119 	int error;
2120 
2121 	if ((flag & ~(AT_EACCESS | AT_RESOLVE_BENEATH | AT_EMPTY_PATH)) != 0)
2122 		return (EINVAL);
2123 	if (amode != F_OK && (amode & ~(R_OK | W_OK | X_OK)) != 0)
2124 		return (EINVAL);
2125 
2126 	/*
2127 	 * Create and modify a temporary credential instead of one that
2128 	 * is potentially shared (if we need one).
2129 	 */
2130 	cred = td->td_ucred;
2131 	if ((flag & AT_EACCESS) == 0 &&
2132 	    ((cred->cr_uid != cred->cr_ruid ||
2133 	    cred->cr_rgid != cred->cr_groups[0]))) {
2134 		usecred = crdup(cred);
2135 		usecred->cr_uid = cred->cr_ruid;
2136 		usecred->cr_groups[0] = cred->cr_rgid;
2137 		td->td_ucred = usecred;
2138 	} else
2139 		usecred = cred;
2140 	AUDIT_ARG_VALUE(amode);
2141 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF |
2142 	    AUDITVNODE1 | at2cnpflags(flag, AT_RESOLVE_BENEATH |
2143 	    AT_EMPTY_PATH), pathseg, path, fd, &cap_fstat_rights, td);
2144 	if ((error = namei(&nd)) != 0)
2145 		goto out;
2146 	vp = nd.ni_vp;
2147 
2148 	error = vn_access(vp, amode, usecred, td);
2149 	NDFREE_NOTHING(&nd);
2150 	vput(vp);
2151 out:
2152 	if (usecred != cred) {
2153 		td->td_ucred = cred;
2154 		crfree(usecred);
2155 	}
2156 	return (error);
2157 }
2158 
2159 /*
2160  * Check access permissions using "effective" credentials.
2161  */
2162 #ifndef _SYS_SYSPROTO_H_
2163 struct eaccess_args {
2164 	char	*path;
2165 	int	amode;
2166 };
2167 #endif
2168 int
2169 sys_eaccess(struct thread *td, struct eaccess_args *uap)
2170 {
2171 
2172 	return (kern_accessat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2173 	    AT_EACCESS, uap->amode));
2174 }
2175 
2176 #if defined(COMPAT_43)
2177 /*
2178  * Get file status; this version follows links.
2179  */
2180 #ifndef _SYS_SYSPROTO_H_
2181 struct ostat_args {
2182 	char	*path;
2183 	struct ostat *ub;
2184 };
2185 #endif
2186 int
2187 ostat(struct thread *td, struct ostat_args *uap)
2188 {
2189 	struct stat sb;
2190 	struct ostat osb;
2191 	int error;
2192 
2193 	error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE,
2194 	    &sb, NULL);
2195 	if (error != 0)
2196 		return (error);
2197 	cvtstat(&sb, &osb);
2198 	return (copyout(&osb, uap->ub, sizeof (osb)));
2199 }
2200 
2201 /*
2202  * Get file status; this version does not follow links.
2203  */
2204 #ifndef _SYS_SYSPROTO_H_
2205 struct olstat_args {
2206 	char	*path;
2207 	struct ostat *ub;
2208 };
2209 #endif
2210 int
2211 olstat(struct thread *td, struct olstat_args *uap)
2212 {
2213 	struct stat sb;
2214 	struct ostat osb;
2215 	int error;
2216 
2217 	error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2218 	    UIO_USERSPACE, &sb, NULL);
2219 	if (error != 0)
2220 		return (error);
2221 	cvtstat(&sb, &osb);
2222 	return (copyout(&osb, uap->ub, sizeof (osb)));
2223 }
2224 
2225 /*
2226  * Convert from an old to a new stat structure.
2227  * XXX: many values are blindly truncated.
2228  */
2229 void
2230 cvtstat(struct stat *st, struct ostat *ost)
2231 {
2232 
2233 	bzero(ost, sizeof(*ost));
2234 	ost->st_dev = st->st_dev;
2235 	ost->st_ino = st->st_ino;
2236 	ost->st_mode = st->st_mode;
2237 	ost->st_nlink = st->st_nlink;
2238 	ost->st_uid = st->st_uid;
2239 	ost->st_gid = st->st_gid;
2240 	ost->st_rdev = st->st_rdev;
2241 	ost->st_size = MIN(st->st_size, INT32_MAX);
2242 	ost->st_atim = st->st_atim;
2243 	ost->st_mtim = st->st_mtim;
2244 	ost->st_ctim = st->st_ctim;
2245 	ost->st_blksize = st->st_blksize;
2246 	ost->st_blocks = st->st_blocks;
2247 	ost->st_flags = st->st_flags;
2248 	ost->st_gen = st->st_gen;
2249 }
2250 #endif /* COMPAT_43 */
2251 
2252 #if defined(COMPAT_43) || defined(COMPAT_FREEBSD11)
2253 int ino64_trunc_error;
2254 SYSCTL_INT(_vfs, OID_AUTO, ino64_trunc_error, CTLFLAG_RW,
2255     &ino64_trunc_error, 0,
2256     "Error on truncation of device, file or inode number, or link count");
2257 
2258 int
2259 freebsd11_cvtstat(struct stat *st, struct freebsd11_stat *ost)
2260 {
2261 
2262 	ost->st_dev = st->st_dev;
2263 	if (ost->st_dev != st->st_dev) {
2264 		switch (ino64_trunc_error) {
2265 		default:
2266 			/*
2267 			 * Since dev_t is almost raw, don't clamp to the
2268 			 * maximum for case 2, but ignore the error.
2269 			 */
2270 			break;
2271 		case 1:
2272 			return (EOVERFLOW);
2273 		}
2274 	}
2275 	ost->st_ino = st->st_ino;
2276 	if (ost->st_ino != st->st_ino) {
2277 		switch (ino64_trunc_error) {
2278 		default:
2279 		case 0:
2280 			break;
2281 		case 1:
2282 			return (EOVERFLOW);
2283 		case 2:
2284 			ost->st_ino = UINT32_MAX;
2285 			break;
2286 		}
2287 	}
2288 	ost->st_mode = st->st_mode;
2289 	ost->st_nlink = st->st_nlink;
2290 	if (ost->st_nlink != st->st_nlink) {
2291 		switch (ino64_trunc_error) {
2292 		default:
2293 		case 0:
2294 			break;
2295 		case 1:
2296 			return (EOVERFLOW);
2297 		case 2:
2298 			ost->st_nlink = UINT16_MAX;
2299 			break;
2300 		}
2301 	}
2302 	ost->st_uid = st->st_uid;
2303 	ost->st_gid = st->st_gid;
2304 	ost->st_rdev = st->st_rdev;
2305 	if (ost->st_rdev != st->st_rdev) {
2306 		switch (ino64_trunc_error) {
2307 		default:
2308 			break;
2309 		case 1:
2310 			return (EOVERFLOW);
2311 		}
2312 	}
2313 	ost->st_atim = st->st_atim;
2314 	ost->st_mtim = st->st_mtim;
2315 	ost->st_ctim = st->st_ctim;
2316 	ost->st_size = st->st_size;
2317 	ost->st_blocks = st->st_blocks;
2318 	ost->st_blksize = st->st_blksize;
2319 	ost->st_flags = st->st_flags;
2320 	ost->st_gen = st->st_gen;
2321 	ost->st_lspare = 0;
2322 	ost->st_birthtim = st->st_birthtim;
2323 	bzero((char *)&ost->st_birthtim + sizeof(ost->st_birthtim),
2324 	    sizeof(*ost) - offsetof(struct freebsd11_stat,
2325 	    st_birthtim) - sizeof(ost->st_birthtim));
2326 	return (0);
2327 }
2328 
2329 int
2330 freebsd11_stat(struct thread *td, struct freebsd11_stat_args* uap)
2331 {
2332 	struct stat sb;
2333 	struct freebsd11_stat osb;
2334 	int error;
2335 
2336 	error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE,
2337 	    &sb, NULL);
2338 	if (error != 0)
2339 		return (error);
2340 	error = freebsd11_cvtstat(&sb, &osb);
2341 	if (error == 0)
2342 		error = copyout(&osb, uap->ub, sizeof(osb));
2343 	return (error);
2344 }
2345 
2346 int
2347 freebsd11_lstat(struct thread *td, struct freebsd11_lstat_args* uap)
2348 {
2349 	struct stat sb;
2350 	struct freebsd11_stat osb;
2351 	int error;
2352 
2353 	error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2354 	    UIO_USERSPACE, &sb, NULL);
2355 	if (error != 0)
2356 		return (error);
2357 	error = freebsd11_cvtstat(&sb, &osb);
2358 	if (error == 0)
2359 		error = copyout(&osb, uap->ub, sizeof(osb));
2360 	return (error);
2361 }
2362 
2363 int
2364 freebsd11_fhstat(struct thread *td, struct freebsd11_fhstat_args* uap)
2365 {
2366 	struct fhandle fh;
2367 	struct stat sb;
2368 	struct freebsd11_stat osb;
2369 	int error;
2370 
2371 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
2372 	if (error != 0)
2373 		return (error);
2374 	error = kern_fhstat(td, fh, &sb);
2375 	if (error != 0)
2376 		return (error);
2377 	error = freebsd11_cvtstat(&sb, &osb);
2378 	if (error == 0)
2379 		error = copyout(&osb, uap->sb, sizeof(osb));
2380 	return (error);
2381 }
2382 
2383 int
2384 freebsd11_fstatat(struct thread *td, struct freebsd11_fstatat_args* uap)
2385 {
2386 	struct stat sb;
2387 	struct freebsd11_stat osb;
2388 	int error;
2389 
2390 	error = kern_statat(td, uap->flag, uap->fd, uap->path,
2391 	    UIO_USERSPACE, &sb, NULL);
2392 	if (error != 0)
2393 		return (error);
2394 	error = freebsd11_cvtstat(&sb, &osb);
2395 	if (error == 0)
2396 		error = copyout(&osb, uap->buf, sizeof(osb));
2397 	return (error);
2398 }
2399 #endif	/* COMPAT_FREEBSD11 */
2400 
2401 /*
2402  * Get file status
2403  */
2404 #ifndef _SYS_SYSPROTO_H_
2405 struct fstatat_args {
2406 	int	fd;
2407 	char	*path;
2408 	struct stat	*buf;
2409 	int	flag;
2410 }
2411 #endif
2412 int
2413 sys_fstatat(struct thread *td, struct fstatat_args *uap)
2414 {
2415 	struct stat sb;
2416 	int error;
2417 
2418 	error = kern_statat(td, uap->flag, uap->fd, uap->path,
2419 	    UIO_USERSPACE, &sb, NULL);
2420 	if (error == 0)
2421 		error = copyout(&sb, uap->buf, sizeof (sb));
2422 	return (error);
2423 }
2424 
2425 int
2426 kern_statat(struct thread *td, int flag, int fd, const char *path,
2427     enum uio_seg pathseg, struct stat *sbp,
2428     void (*hook)(struct vnode *vp, struct stat *sbp))
2429 {
2430 	struct nameidata nd;
2431 	int error;
2432 
2433 	if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH |
2434 	    AT_EMPTY_PATH)) != 0)
2435 		return (EINVAL);
2436 
2437 	NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(flag, AT_RESOLVE_BENEATH |
2438 	    AT_SYMLINK_NOFOLLOW | AT_EMPTY_PATH) | LOCKSHARED | LOCKLEAF |
2439 	    AUDITVNODE1, pathseg, path, fd, &cap_fstat_rights, td);
2440 
2441 	if ((error = namei(&nd)) != 0)
2442 		return (error);
2443 	error = VOP_STAT(nd.ni_vp, sbp, td->td_ucred, NOCRED, td);
2444 	if (error == 0) {
2445 		if (__predict_false(hook != NULL))
2446 			hook(nd.ni_vp, sbp);
2447 	}
2448 	NDFREE_NOTHING(&nd);
2449 	vput(nd.ni_vp);
2450 #ifdef __STAT_TIME_T_EXT
2451 	sbp->st_atim_ext = 0;
2452 	sbp->st_mtim_ext = 0;
2453 	sbp->st_ctim_ext = 0;
2454 	sbp->st_btim_ext = 0;
2455 #endif
2456 #ifdef KTRACE
2457 	if (KTRPOINT(td, KTR_STRUCT))
2458 		ktrstat_error(sbp, error);
2459 #endif
2460 	return (error);
2461 }
2462 
2463 #if defined(COMPAT_FREEBSD11)
2464 /*
2465  * Implementation of the NetBSD [l]stat() functions.
2466  */
2467 void
2468 freebsd11_cvtnstat(struct stat *sb, struct nstat *nsb)
2469 {
2470 
2471 	bzero(nsb, sizeof(*nsb));
2472 	nsb->st_dev = sb->st_dev;
2473 	nsb->st_ino = sb->st_ino;
2474 	nsb->st_mode = sb->st_mode;
2475 	nsb->st_nlink = sb->st_nlink;
2476 	nsb->st_uid = sb->st_uid;
2477 	nsb->st_gid = sb->st_gid;
2478 	nsb->st_rdev = sb->st_rdev;
2479 	nsb->st_atim = sb->st_atim;
2480 	nsb->st_mtim = sb->st_mtim;
2481 	nsb->st_ctim = sb->st_ctim;
2482 	nsb->st_size = sb->st_size;
2483 	nsb->st_blocks = sb->st_blocks;
2484 	nsb->st_blksize = sb->st_blksize;
2485 	nsb->st_flags = sb->st_flags;
2486 	nsb->st_gen = sb->st_gen;
2487 	nsb->st_birthtim = sb->st_birthtim;
2488 }
2489 
2490 #ifndef _SYS_SYSPROTO_H_
2491 struct freebsd11_nstat_args {
2492 	char	*path;
2493 	struct nstat *ub;
2494 };
2495 #endif
2496 int
2497 freebsd11_nstat(struct thread *td, struct freebsd11_nstat_args *uap)
2498 {
2499 	struct stat sb;
2500 	struct nstat nsb;
2501 	int error;
2502 
2503 	error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE,
2504 	    &sb, NULL);
2505 	if (error != 0)
2506 		return (error);
2507 	freebsd11_cvtnstat(&sb, &nsb);
2508 	return (copyout(&nsb, uap->ub, sizeof (nsb)));
2509 }
2510 
2511 /*
2512  * NetBSD lstat.  Get file status; this version does not follow links.
2513  */
2514 #ifndef _SYS_SYSPROTO_H_
2515 struct freebsd11_nlstat_args {
2516 	char	*path;
2517 	struct nstat *ub;
2518 };
2519 #endif
2520 int
2521 freebsd11_nlstat(struct thread *td, struct freebsd11_nlstat_args *uap)
2522 {
2523 	struct stat sb;
2524 	struct nstat nsb;
2525 	int error;
2526 
2527 	error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2528 	    UIO_USERSPACE, &sb, NULL);
2529 	if (error != 0)
2530 		return (error);
2531 	freebsd11_cvtnstat(&sb, &nsb);
2532 	return (copyout(&nsb, uap->ub, sizeof (nsb)));
2533 }
2534 #endif /* COMPAT_FREEBSD11 */
2535 
2536 /*
2537  * Get configurable pathname variables.
2538  */
2539 #ifndef _SYS_SYSPROTO_H_
2540 struct pathconf_args {
2541 	char	*path;
2542 	int	name;
2543 };
2544 #endif
2545 int
2546 sys_pathconf(struct thread *td, struct pathconf_args *uap)
2547 {
2548 	long value;
2549 	int error;
2550 
2551 	error = kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name, FOLLOW,
2552 	    &value);
2553 	if (error == 0)
2554 		td->td_retval[0] = value;
2555 	return (error);
2556 }
2557 
2558 #ifndef _SYS_SYSPROTO_H_
2559 struct lpathconf_args {
2560 	char	*path;
2561 	int	name;
2562 };
2563 #endif
2564 int
2565 sys_lpathconf(struct thread *td, struct lpathconf_args *uap)
2566 {
2567 	long value;
2568 	int error;
2569 
2570 	error = kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name,
2571 	    NOFOLLOW, &value);
2572 	if (error == 0)
2573 		td->td_retval[0] = value;
2574 	return (error);
2575 }
2576 
2577 int
2578 kern_pathconf(struct thread *td, const char *path, enum uio_seg pathseg,
2579     int name, u_long flags, long *valuep)
2580 {
2581 	struct nameidata nd;
2582 	int error;
2583 
2584 	NDINIT(&nd, LOOKUP, LOCKSHARED | LOCKLEAF | AUDITVNODE1 | flags,
2585 	    pathseg, path, td);
2586 	if ((error = namei(&nd)) != 0)
2587 		return (error);
2588 	NDFREE_NOTHING(&nd);
2589 
2590 	error = VOP_PATHCONF(nd.ni_vp, name, valuep);
2591 	vput(nd.ni_vp);
2592 	return (error);
2593 }
2594 
2595 /*
2596  * Return target name of a symbolic link.
2597  */
2598 #ifndef _SYS_SYSPROTO_H_
2599 struct readlink_args {
2600 	char	*path;
2601 	char	*buf;
2602 	size_t	count;
2603 };
2604 #endif
2605 int
2606 sys_readlink(struct thread *td, struct readlink_args *uap)
2607 {
2608 
2609 	return (kern_readlinkat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2610 	    uap->buf, UIO_USERSPACE, uap->count));
2611 }
2612 #ifndef _SYS_SYSPROTO_H_
2613 struct readlinkat_args {
2614 	int	fd;
2615 	char	*path;
2616 	char	*buf;
2617 	size_t	bufsize;
2618 };
2619 #endif
2620 int
2621 sys_readlinkat(struct thread *td, struct readlinkat_args *uap)
2622 {
2623 
2624 	return (kern_readlinkat(td, uap->fd, uap->path, UIO_USERSPACE,
2625 	    uap->buf, UIO_USERSPACE, uap->bufsize));
2626 }
2627 
2628 int
2629 kern_readlinkat(struct thread *td, int fd, const char *path,
2630     enum uio_seg pathseg, char *buf, enum uio_seg bufseg, size_t count)
2631 {
2632 	struct vnode *vp;
2633 	struct nameidata nd;
2634 	int error;
2635 
2636 	if (count > IOSIZE_MAX)
2637 		return (EINVAL);
2638 
2639 	NDINIT_AT(&nd, LOOKUP, NOFOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
2640 	    pathseg, path, fd, td);
2641 
2642 	if ((error = namei(&nd)) != 0)
2643 		return (error);
2644 	NDFREE_NOTHING(&nd);
2645 	vp = nd.ni_vp;
2646 
2647 	error = kern_readlink_vp(vp, buf, bufseg, count, td);
2648 	vput(vp);
2649 
2650 	return (error);
2651 }
2652 
2653 /*
2654  * Helper function to readlink from a vnode
2655  */
2656 static int
2657 kern_readlink_vp(struct vnode *vp, char *buf, enum uio_seg bufseg, size_t count,
2658     struct thread *td)
2659 {
2660 	struct iovec aiov;
2661 	struct uio auio;
2662 	int error;
2663 
2664 	ASSERT_VOP_LOCKED(vp, "kern_readlink_vp(): vp not locked");
2665 #ifdef MAC
2666 	error = mac_vnode_check_readlink(td->td_ucred, vp);
2667 	if (error != 0)
2668 		return (error);
2669 #endif
2670 	if (vp->v_type != VLNK && (vp->v_vflag & VV_READLINK) == 0)
2671 		return (EINVAL);
2672 
2673 	aiov.iov_base = buf;
2674 	aiov.iov_len = count;
2675 	auio.uio_iov = &aiov;
2676 	auio.uio_iovcnt = 1;
2677 	auio.uio_offset = 0;
2678 	auio.uio_rw = UIO_READ;
2679 	auio.uio_segflg = bufseg;
2680 	auio.uio_td = td;
2681 	auio.uio_resid = count;
2682 	error = VOP_READLINK(vp, &auio, td->td_ucred);
2683 	td->td_retval[0] = count - auio.uio_resid;
2684 	return (error);
2685 }
2686 
2687 /*
2688  * Common implementation code for chflags() and fchflags().
2689  */
2690 static int
2691 setfflags(struct thread *td, struct vnode *vp, u_long flags)
2692 {
2693 	struct mount *mp;
2694 	struct vattr vattr;
2695 	int error;
2696 
2697 	/* We can't support the value matching VNOVAL. */
2698 	if (flags == VNOVAL)
2699 		return (EOPNOTSUPP);
2700 
2701 	/*
2702 	 * Prevent non-root users from setting flags on devices.  When
2703 	 * a device is reused, users can retain ownership of the device
2704 	 * if they are allowed to set flags and programs assume that
2705 	 * chown can't fail when done as root.
2706 	 */
2707 	if (vp->v_type == VCHR || vp->v_type == VBLK) {
2708 		error = priv_check(td, PRIV_VFS_CHFLAGS_DEV);
2709 		if (error != 0)
2710 			return (error);
2711 	}
2712 
2713 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2714 		return (error);
2715 	VATTR_NULL(&vattr);
2716 	vattr.va_flags = flags;
2717 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2718 #ifdef MAC
2719 	error = mac_vnode_check_setflags(td->td_ucred, vp, vattr.va_flags);
2720 	if (error == 0)
2721 #endif
2722 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
2723 	VOP_UNLOCK(vp);
2724 	vn_finished_write(mp);
2725 	return (error);
2726 }
2727 
2728 /*
2729  * Change flags of a file given a path name.
2730  */
2731 #ifndef _SYS_SYSPROTO_H_
2732 struct chflags_args {
2733 	const char *path;
2734 	u_long	flags;
2735 };
2736 #endif
2737 int
2738 sys_chflags(struct thread *td, struct chflags_args *uap)
2739 {
2740 
2741 	return (kern_chflagsat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2742 	    uap->flags, 0));
2743 }
2744 
2745 #ifndef _SYS_SYSPROTO_H_
2746 struct chflagsat_args {
2747 	int	fd;
2748 	const char *path;
2749 	u_long	flags;
2750 	int	atflag;
2751 }
2752 #endif
2753 int
2754 sys_chflagsat(struct thread *td, struct chflagsat_args *uap)
2755 {
2756 
2757 	if ((uap->atflag & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH |
2758 	    AT_EMPTY_PATH)) != 0)
2759 		return (EINVAL);
2760 
2761 	return (kern_chflagsat(td, uap->fd, uap->path, UIO_USERSPACE,
2762 	    uap->flags, uap->atflag));
2763 }
2764 
2765 /*
2766  * Same as chflags() but doesn't follow symlinks.
2767  */
2768 #ifndef _SYS_SYSPROTO_H_
2769 struct lchflags_args {
2770 	const char *path;
2771 	u_long flags;
2772 };
2773 #endif
2774 int
2775 sys_lchflags(struct thread *td, struct lchflags_args *uap)
2776 {
2777 
2778 	return (kern_chflagsat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2779 	    uap->flags, AT_SYMLINK_NOFOLLOW));
2780 }
2781 
2782 static int
2783 kern_chflagsat(struct thread *td, int fd, const char *path,
2784     enum uio_seg pathseg, u_long flags, int atflag)
2785 {
2786 	struct nameidata nd;
2787 	int error;
2788 
2789 	AUDIT_ARG_FFLAGS(flags);
2790 	NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(atflag, AT_SYMLINK_NOFOLLOW |
2791 	    AT_RESOLVE_BENEATH | AT_EMPTY_PATH) | AUDITVNODE1, pathseg, path,
2792 	    fd, &cap_fchflags_rights, td);
2793 	if ((error = namei(&nd)) != 0)
2794 		return (error);
2795 	NDFREE_NOTHING(&nd);
2796 	error = setfflags(td, nd.ni_vp, flags);
2797 	vrele(nd.ni_vp);
2798 	return (error);
2799 }
2800 
2801 /*
2802  * Change flags of a file given a file descriptor.
2803  */
2804 #ifndef _SYS_SYSPROTO_H_
2805 struct fchflags_args {
2806 	int	fd;
2807 	u_long	flags;
2808 };
2809 #endif
2810 int
2811 sys_fchflags(struct thread *td, struct fchflags_args *uap)
2812 {
2813 	struct file *fp;
2814 	int error;
2815 
2816 	AUDIT_ARG_FD(uap->fd);
2817 	AUDIT_ARG_FFLAGS(uap->flags);
2818 	error = getvnode(td, uap->fd, &cap_fchflags_rights,
2819 	    &fp);
2820 	if (error != 0)
2821 		return (error);
2822 #ifdef AUDIT
2823 	if (AUDITING_TD(td)) {
2824 		vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
2825 		AUDIT_ARG_VNODE1(fp->f_vnode);
2826 		VOP_UNLOCK(fp->f_vnode);
2827 	}
2828 #endif
2829 	error = setfflags(td, fp->f_vnode, uap->flags);
2830 	fdrop(fp, td);
2831 	return (error);
2832 }
2833 
2834 /*
2835  * Common implementation code for chmod(), lchmod() and fchmod().
2836  */
2837 int
2838 setfmode(struct thread *td, struct ucred *cred, struct vnode *vp, int mode)
2839 {
2840 	struct mount *mp;
2841 	struct vattr vattr;
2842 	int error;
2843 
2844 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2845 		return (error);
2846 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2847 	VATTR_NULL(&vattr);
2848 	vattr.va_mode = mode & ALLPERMS;
2849 #ifdef MAC
2850 	error = mac_vnode_check_setmode(cred, vp, vattr.va_mode);
2851 	if (error == 0)
2852 #endif
2853 		error = VOP_SETATTR(vp, &vattr, cred);
2854 	VOP_UNLOCK(vp);
2855 	vn_finished_write(mp);
2856 	return (error);
2857 }
2858 
2859 /*
2860  * Change mode of a file given path name.
2861  */
2862 #ifndef _SYS_SYSPROTO_H_
2863 struct chmod_args {
2864 	char	*path;
2865 	int	mode;
2866 };
2867 #endif
2868 int
2869 sys_chmod(struct thread *td, struct chmod_args *uap)
2870 {
2871 
2872 	return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2873 	    uap->mode, 0));
2874 }
2875 
2876 #ifndef _SYS_SYSPROTO_H_
2877 struct fchmodat_args {
2878 	int	dirfd;
2879 	char	*path;
2880 	mode_t	mode;
2881 	int	flag;
2882 }
2883 #endif
2884 int
2885 sys_fchmodat(struct thread *td, struct fchmodat_args *uap)
2886 {
2887 
2888 	if ((uap->flag & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH |
2889 	    AT_EMPTY_PATH)) != 0)
2890 		return (EINVAL);
2891 
2892 	return (kern_fchmodat(td, uap->fd, uap->path, UIO_USERSPACE,
2893 	    uap->mode, uap->flag));
2894 }
2895 
2896 /*
2897  * Change mode of a file given path name (don't follow links.)
2898  */
2899 #ifndef _SYS_SYSPROTO_H_
2900 struct lchmod_args {
2901 	char	*path;
2902 	int	mode;
2903 };
2904 #endif
2905 int
2906 sys_lchmod(struct thread *td, struct lchmod_args *uap)
2907 {
2908 
2909 	return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2910 	    uap->mode, AT_SYMLINK_NOFOLLOW));
2911 }
2912 
2913 int
2914 kern_fchmodat(struct thread *td, int fd, const char *path,
2915     enum uio_seg pathseg, mode_t mode, int flag)
2916 {
2917 	struct nameidata nd;
2918 	int error;
2919 
2920 	AUDIT_ARG_MODE(mode);
2921 	NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(flag, AT_SYMLINK_NOFOLLOW |
2922 	    AT_RESOLVE_BENEATH | AT_EMPTY_PATH) | AUDITVNODE1, pathseg, path,
2923 	    fd, &cap_fchmod_rights, td);
2924 	if ((error = namei(&nd)) != 0)
2925 		return (error);
2926 	NDFREE_NOTHING(&nd);
2927 	error = setfmode(td, td->td_ucred, nd.ni_vp, mode);
2928 	vrele(nd.ni_vp);
2929 	return (error);
2930 }
2931 
2932 /*
2933  * Change mode of a file given a file descriptor.
2934  */
2935 #ifndef _SYS_SYSPROTO_H_
2936 struct fchmod_args {
2937 	int	fd;
2938 	int	mode;
2939 };
2940 #endif
2941 int
2942 sys_fchmod(struct thread *td, struct fchmod_args *uap)
2943 {
2944 	struct file *fp;
2945 	int error;
2946 
2947 	AUDIT_ARG_FD(uap->fd);
2948 	AUDIT_ARG_MODE(uap->mode);
2949 
2950 	error = fget(td, uap->fd, &cap_fchmod_rights, &fp);
2951 	if (error != 0)
2952 		return (error);
2953 	error = fo_chmod(fp, uap->mode, td->td_ucred, td);
2954 	fdrop(fp, td);
2955 	return (error);
2956 }
2957 
2958 /*
2959  * Common implementation for chown(), lchown(), and fchown()
2960  */
2961 int
2962 setfown(struct thread *td, struct ucred *cred, struct vnode *vp, uid_t uid,
2963     gid_t gid)
2964 {
2965 	struct mount *mp;
2966 	struct vattr vattr;
2967 	int error;
2968 
2969 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2970 		return (error);
2971 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2972 	VATTR_NULL(&vattr);
2973 	vattr.va_uid = uid;
2974 	vattr.va_gid = gid;
2975 #ifdef MAC
2976 	error = mac_vnode_check_setowner(cred, vp, vattr.va_uid,
2977 	    vattr.va_gid);
2978 	if (error == 0)
2979 #endif
2980 		error = VOP_SETATTR(vp, &vattr, cred);
2981 	VOP_UNLOCK(vp);
2982 	vn_finished_write(mp);
2983 	return (error);
2984 }
2985 
2986 /*
2987  * Set ownership given a path name.
2988  */
2989 #ifndef _SYS_SYSPROTO_H_
2990 struct chown_args {
2991 	char	*path;
2992 	int	uid;
2993 	int	gid;
2994 };
2995 #endif
2996 int
2997 sys_chown(struct thread *td, struct chown_args *uap)
2998 {
2999 
3000 	return (kern_fchownat(td, AT_FDCWD, uap->path, UIO_USERSPACE, uap->uid,
3001 	    uap->gid, 0));
3002 }
3003 
3004 #ifndef _SYS_SYSPROTO_H_
3005 struct fchownat_args {
3006 	int fd;
3007 	const char * path;
3008 	uid_t uid;
3009 	gid_t gid;
3010 	int flag;
3011 };
3012 #endif
3013 int
3014 sys_fchownat(struct thread *td, struct fchownat_args *uap)
3015 {
3016 
3017 	if ((uap->flag & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH |
3018 	    AT_EMPTY_PATH)) != 0)
3019 		return (EINVAL);
3020 
3021 	return (kern_fchownat(td, uap->fd, uap->path, UIO_USERSPACE, uap->uid,
3022 	    uap->gid, uap->flag));
3023 }
3024 
3025 int
3026 kern_fchownat(struct thread *td, int fd, const char *path,
3027     enum uio_seg pathseg, int uid, int gid, int flag)
3028 {
3029 	struct nameidata nd;
3030 	int error;
3031 
3032 	AUDIT_ARG_OWNER(uid, gid);
3033 	NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(flag, AT_SYMLINK_NOFOLLOW |
3034 	    AT_RESOLVE_BENEATH | AT_EMPTY_PATH) | AUDITVNODE1, pathseg, path,
3035 	    fd, &cap_fchown_rights, td);
3036 
3037 	if ((error = namei(&nd)) != 0)
3038 		return (error);
3039 	NDFREE_NOTHING(&nd);
3040 	error = setfown(td, td->td_ucred, nd.ni_vp, uid, gid);
3041 	vrele(nd.ni_vp);
3042 	return (error);
3043 }
3044 
3045 /*
3046  * Set ownership given a path name, do not cross symlinks.
3047  */
3048 #ifndef _SYS_SYSPROTO_H_
3049 struct lchown_args {
3050 	char	*path;
3051 	int	uid;
3052 	int	gid;
3053 };
3054 #endif
3055 int
3056 sys_lchown(struct thread *td, struct lchown_args *uap)
3057 {
3058 
3059 	return (kern_fchownat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
3060 	    uap->uid, uap->gid, AT_SYMLINK_NOFOLLOW));
3061 }
3062 
3063 /*
3064  * Set ownership given a file descriptor.
3065  */
3066 #ifndef _SYS_SYSPROTO_H_
3067 struct fchown_args {
3068 	int	fd;
3069 	int	uid;
3070 	int	gid;
3071 };
3072 #endif
3073 int
3074 sys_fchown(struct thread *td, struct fchown_args *uap)
3075 {
3076 	struct file *fp;
3077 	int error;
3078 
3079 	AUDIT_ARG_FD(uap->fd);
3080 	AUDIT_ARG_OWNER(uap->uid, uap->gid);
3081 	error = fget(td, uap->fd, &cap_fchown_rights, &fp);
3082 	if (error != 0)
3083 		return (error);
3084 	error = fo_chown(fp, uap->uid, uap->gid, td->td_ucred, td);
3085 	fdrop(fp, td);
3086 	return (error);
3087 }
3088 
3089 /*
3090  * Common implementation code for utimes(), lutimes(), and futimes().
3091  */
3092 static int
3093 getutimes(const struct timeval *usrtvp, enum uio_seg tvpseg,
3094     struct timespec *tsp)
3095 {
3096 	struct timeval tv[2];
3097 	const struct timeval *tvp;
3098 	int error;
3099 
3100 	if (usrtvp == NULL) {
3101 		vfs_timestamp(&tsp[0]);
3102 		tsp[1] = tsp[0];
3103 	} else {
3104 		if (tvpseg == UIO_SYSSPACE) {
3105 			tvp = usrtvp;
3106 		} else {
3107 			if ((error = copyin(usrtvp, tv, sizeof(tv))) != 0)
3108 				return (error);
3109 			tvp = tv;
3110 		}
3111 
3112 		if (tvp[0].tv_usec < 0 || tvp[0].tv_usec >= 1000000 ||
3113 		    tvp[1].tv_usec < 0 || tvp[1].tv_usec >= 1000000)
3114 			return (EINVAL);
3115 		TIMEVAL_TO_TIMESPEC(&tvp[0], &tsp[0]);
3116 		TIMEVAL_TO_TIMESPEC(&tvp[1], &tsp[1]);
3117 	}
3118 	return (0);
3119 }
3120 
3121 /*
3122  * Common implementation code for futimens(), utimensat().
3123  */
3124 #define	UTIMENS_NULL	0x1
3125 #define	UTIMENS_EXIT	0x2
3126 static int
3127 getutimens(const struct timespec *usrtsp, enum uio_seg tspseg,
3128     struct timespec *tsp, int *retflags)
3129 {
3130 	struct timespec tsnow;
3131 	int error;
3132 
3133 	vfs_timestamp(&tsnow);
3134 	*retflags = 0;
3135 	if (usrtsp == NULL) {
3136 		tsp[0] = tsnow;
3137 		tsp[1] = tsnow;
3138 		*retflags |= UTIMENS_NULL;
3139 		return (0);
3140 	}
3141 	if (tspseg == UIO_SYSSPACE) {
3142 		tsp[0] = usrtsp[0];
3143 		tsp[1] = usrtsp[1];
3144 	} else if ((error = copyin(usrtsp, tsp, sizeof(*tsp) * 2)) != 0)
3145 		return (error);
3146 	if (tsp[0].tv_nsec == UTIME_OMIT && tsp[1].tv_nsec == UTIME_OMIT)
3147 		*retflags |= UTIMENS_EXIT;
3148 	if (tsp[0].tv_nsec == UTIME_NOW && tsp[1].tv_nsec == UTIME_NOW)
3149 		*retflags |= UTIMENS_NULL;
3150 	if (tsp[0].tv_nsec == UTIME_OMIT)
3151 		tsp[0].tv_sec = VNOVAL;
3152 	else if (tsp[0].tv_nsec == UTIME_NOW)
3153 		tsp[0] = tsnow;
3154 	else if (tsp[0].tv_nsec < 0 || tsp[0].tv_nsec >= 1000000000L)
3155 		return (EINVAL);
3156 	if (tsp[1].tv_nsec == UTIME_OMIT)
3157 		tsp[1].tv_sec = VNOVAL;
3158 	else if (tsp[1].tv_nsec == UTIME_NOW)
3159 		tsp[1] = tsnow;
3160 	else if (tsp[1].tv_nsec < 0 || tsp[1].tv_nsec >= 1000000000L)
3161 		return (EINVAL);
3162 
3163 	return (0);
3164 }
3165 
3166 /*
3167  * Common implementation code for utimes(), lutimes(), futimes(), futimens(),
3168  * and utimensat().
3169  */
3170 static int
3171 setutimes(struct thread *td, struct vnode *vp, const struct timespec *ts,
3172     int numtimes, int nullflag)
3173 {
3174 	struct mount *mp;
3175 	struct vattr vattr;
3176 	int error, setbirthtime;
3177 
3178 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
3179 		return (error);
3180 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3181 	setbirthtime = 0;
3182 	if (numtimes < 3 && !VOP_GETATTR(vp, &vattr, td->td_ucred) &&
3183 	    timespeccmp(&ts[1], &vattr.va_birthtime, < ))
3184 		setbirthtime = 1;
3185 	VATTR_NULL(&vattr);
3186 	vattr.va_atime = ts[0];
3187 	vattr.va_mtime = ts[1];
3188 	if (setbirthtime)
3189 		vattr.va_birthtime = ts[1];
3190 	if (numtimes > 2)
3191 		vattr.va_birthtime = ts[2];
3192 	if (nullflag)
3193 		vattr.va_vaflags |= VA_UTIMES_NULL;
3194 #ifdef MAC
3195 	error = mac_vnode_check_setutimes(td->td_ucred, vp, vattr.va_atime,
3196 	    vattr.va_mtime);
3197 #endif
3198 	if (error == 0)
3199 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3200 	VOP_UNLOCK(vp);
3201 	vn_finished_write(mp);
3202 	return (error);
3203 }
3204 
3205 /*
3206  * Set the access and modification times of a file.
3207  */
3208 #ifndef _SYS_SYSPROTO_H_
3209 struct utimes_args {
3210 	char	*path;
3211 	struct	timeval *tptr;
3212 };
3213 #endif
3214 int
3215 sys_utimes(struct thread *td, struct utimes_args *uap)
3216 {
3217 
3218 	return (kern_utimesat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
3219 	    uap->tptr, UIO_USERSPACE));
3220 }
3221 
3222 #ifndef _SYS_SYSPROTO_H_
3223 struct futimesat_args {
3224 	int fd;
3225 	const char * path;
3226 	const struct timeval * times;
3227 };
3228 #endif
3229 int
3230 sys_futimesat(struct thread *td, struct futimesat_args *uap)
3231 {
3232 
3233 	return (kern_utimesat(td, uap->fd, uap->path, UIO_USERSPACE,
3234 	    uap->times, UIO_USERSPACE));
3235 }
3236 
3237 int
3238 kern_utimesat(struct thread *td, int fd, const char *path,
3239     enum uio_seg pathseg, struct timeval *tptr, enum uio_seg tptrseg)
3240 {
3241 	struct nameidata nd;
3242 	struct timespec ts[2];
3243 	int error;
3244 
3245 	if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3246 		return (error);
3247 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, fd,
3248 	    &cap_futimes_rights, td);
3249 
3250 	if ((error = namei(&nd)) != 0)
3251 		return (error);
3252 	NDFREE_NOTHING(&nd);
3253 	error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3254 	vrele(nd.ni_vp);
3255 	return (error);
3256 }
3257 
3258 /*
3259  * Set the access and modification times of a file.
3260  */
3261 #ifndef _SYS_SYSPROTO_H_
3262 struct lutimes_args {
3263 	char	*path;
3264 	struct	timeval *tptr;
3265 };
3266 #endif
3267 int
3268 sys_lutimes(struct thread *td, struct lutimes_args *uap)
3269 {
3270 
3271 	return (kern_lutimes(td, uap->path, UIO_USERSPACE, uap->tptr,
3272 	    UIO_USERSPACE));
3273 }
3274 
3275 int
3276 kern_lutimes(struct thread *td, const char *path, enum uio_seg pathseg,
3277     struct timeval *tptr, enum uio_seg tptrseg)
3278 {
3279 	struct timespec ts[2];
3280 	struct nameidata nd;
3281 	int error;
3282 
3283 	if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3284 		return (error);
3285 	NDINIT(&nd, LOOKUP, NOFOLLOW | AUDITVNODE1, pathseg, path, td);
3286 	if ((error = namei(&nd)) != 0)
3287 		return (error);
3288 	NDFREE_NOTHING(&nd);
3289 	error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3290 	vrele(nd.ni_vp);
3291 	return (error);
3292 }
3293 
3294 /*
3295  * Set the access and modification times of a file.
3296  */
3297 #ifndef _SYS_SYSPROTO_H_
3298 struct futimes_args {
3299 	int	fd;
3300 	struct	timeval *tptr;
3301 };
3302 #endif
3303 int
3304 sys_futimes(struct thread *td, struct futimes_args *uap)
3305 {
3306 
3307 	return (kern_futimes(td, uap->fd, uap->tptr, UIO_USERSPACE));
3308 }
3309 
3310 int
3311 kern_futimes(struct thread *td, int fd, struct timeval *tptr,
3312     enum uio_seg tptrseg)
3313 {
3314 	struct timespec ts[2];
3315 	struct file *fp;
3316 	int error;
3317 
3318 	AUDIT_ARG_FD(fd);
3319 	error = getutimes(tptr, tptrseg, ts);
3320 	if (error != 0)
3321 		return (error);
3322 	error = getvnode(td, fd, &cap_futimes_rights, &fp);
3323 	if (error != 0)
3324 		return (error);
3325 #ifdef AUDIT
3326 	if (AUDITING_TD(td)) {
3327 		vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
3328 		AUDIT_ARG_VNODE1(fp->f_vnode);
3329 		VOP_UNLOCK(fp->f_vnode);
3330 	}
3331 #endif
3332 	error = setutimes(td, fp->f_vnode, ts, 2, tptr == NULL);
3333 	fdrop(fp, td);
3334 	return (error);
3335 }
3336 
3337 int
3338 sys_futimens(struct thread *td, struct futimens_args *uap)
3339 {
3340 
3341 	return (kern_futimens(td, uap->fd, uap->times, UIO_USERSPACE));
3342 }
3343 
3344 int
3345 kern_futimens(struct thread *td, int fd, struct timespec *tptr,
3346     enum uio_seg tptrseg)
3347 {
3348 	struct timespec ts[2];
3349 	struct file *fp;
3350 	int error, flags;
3351 
3352 	AUDIT_ARG_FD(fd);
3353 	error = getutimens(tptr, tptrseg, ts, &flags);
3354 	if (error != 0)
3355 		return (error);
3356 	if (flags & UTIMENS_EXIT)
3357 		return (0);
3358 	error = getvnode(td, fd, &cap_futimes_rights, &fp);
3359 	if (error != 0)
3360 		return (error);
3361 #ifdef AUDIT
3362 	if (AUDITING_TD(td)) {
3363 		vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
3364 		AUDIT_ARG_VNODE1(fp->f_vnode);
3365 		VOP_UNLOCK(fp->f_vnode);
3366 	}
3367 #endif
3368 	error = setutimes(td, fp->f_vnode, ts, 2, flags & UTIMENS_NULL);
3369 	fdrop(fp, td);
3370 	return (error);
3371 }
3372 
3373 int
3374 sys_utimensat(struct thread *td, struct utimensat_args *uap)
3375 {
3376 
3377 	return (kern_utimensat(td, uap->fd, uap->path, UIO_USERSPACE,
3378 	    uap->times, UIO_USERSPACE, uap->flag));
3379 }
3380 
3381 int
3382 kern_utimensat(struct thread *td, int fd, const char *path,
3383     enum uio_seg pathseg, struct timespec *tptr, enum uio_seg tptrseg,
3384     int flag)
3385 {
3386 	struct nameidata nd;
3387 	struct timespec ts[2];
3388 	int error, flags;
3389 
3390 	if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH |
3391 	    AT_EMPTY_PATH)) != 0)
3392 		return (EINVAL);
3393 
3394 	if ((error = getutimens(tptr, tptrseg, ts, &flags)) != 0)
3395 		return (error);
3396 	NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(flag, AT_SYMLINK_NOFOLLOW |
3397 	    AT_RESOLVE_BENEATH | AT_EMPTY_PATH) | AUDITVNODE1,
3398 	    pathseg, path, fd, &cap_futimes_rights, td);
3399 	if ((error = namei(&nd)) != 0)
3400 		return (error);
3401 	/*
3402 	 * We are allowed to call namei() regardless of 2xUTIME_OMIT.
3403 	 * POSIX states:
3404 	 * "If both tv_nsec fields are UTIME_OMIT... EACCESS may be detected."
3405 	 * "Search permission is denied by a component of the path prefix."
3406 	 */
3407 	NDFREE_NOTHING(&nd);
3408 	if ((flags & UTIMENS_EXIT) == 0)
3409 		error = setutimes(td, nd.ni_vp, ts, 2, flags & UTIMENS_NULL);
3410 	vrele(nd.ni_vp);
3411 	return (error);
3412 }
3413 
3414 /*
3415  * Truncate a file given its path name.
3416  */
3417 #ifndef _SYS_SYSPROTO_H_
3418 struct truncate_args {
3419 	char	*path;
3420 	int	pad;
3421 	off_t	length;
3422 };
3423 #endif
3424 int
3425 sys_truncate(struct thread *td, struct truncate_args *uap)
3426 {
3427 
3428 	return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3429 }
3430 
3431 int
3432 kern_truncate(struct thread *td, const char *path, enum uio_seg pathseg,
3433     off_t length)
3434 {
3435 	struct mount *mp;
3436 	struct vnode *vp;
3437 	void *rl_cookie;
3438 	struct vattr vattr;
3439 	struct nameidata nd;
3440 	int error;
3441 
3442 	if (length < 0)
3443 		return (EINVAL);
3444 	NDPREINIT(&nd);
3445 retry:
3446 	NDINIT(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, td);
3447 	if ((error = namei(&nd)) != 0)
3448 		return (error);
3449 	vp = nd.ni_vp;
3450 	rl_cookie = vn_rangelock_wlock(vp, 0, OFF_MAX);
3451 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) {
3452 		vn_rangelock_unlock(vp, rl_cookie);
3453 		vrele(vp);
3454 		return (error);
3455 	}
3456 	NDFREE(&nd, NDF_ONLY_PNBUF);
3457 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3458 	if (vp->v_type == VDIR)
3459 		error = EISDIR;
3460 #ifdef MAC
3461 	else if ((error = mac_vnode_check_write(td->td_ucred, NOCRED, vp))) {
3462 	}
3463 #endif
3464 	else if ((error = vn_writechk(vp)) == 0 &&
3465 	    (error = VOP_ACCESS(vp, VWRITE, td->td_ucred, td)) == 0) {
3466 		VATTR_NULL(&vattr);
3467 		vattr.va_size = length;
3468 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3469 	}
3470 	VOP_UNLOCK(vp);
3471 	vn_finished_write(mp);
3472 	vn_rangelock_unlock(vp, rl_cookie);
3473 	vrele(vp);
3474 	if (error == ERELOOKUP)
3475 		goto retry;
3476 	return (error);
3477 }
3478 
3479 #if defined(COMPAT_43)
3480 /*
3481  * Truncate a file given its path name.
3482  */
3483 #ifndef _SYS_SYSPROTO_H_
3484 struct otruncate_args {
3485 	char	*path;
3486 	long	length;
3487 };
3488 #endif
3489 int
3490 otruncate(struct thread *td, struct otruncate_args *uap)
3491 {
3492 
3493 	return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3494 }
3495 #endif /* COMPAT_43 */
3496 
3497 #if defined(COMPAT_FREEBSD6)
3498 /* Versions with the pad argument */
3499 int
3500 freebsd6_truncate(struct thread *td, struct freebsd6_truncate_args *uap)
3501 {
3502 
3503 	return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3504 }
3505 
3506 int
3507 freebsd6_ftruncate(struct thread *td, struct freebsd6_ftruncate_args *uap)
3508 {
3509 
3510 	return (kern_ftruncate(td, uap->fd, uap->length));
3511 }
3512 #endif
3513 
3514 int
3515 kern_fsync(struct thread *td, int fd, bool fullsync)
3516 {
3517 	struct vnode *vp;
3518 	struct mount *mp;
3519 	struct file *fp;
3520 	int error;
3521 
3522 	AUDIT_ARG_FD(fd);
3523 	error = getvnode(td, fd, &cap_fsync_rights, &fp);
3524 	if (error != 0)
3525 		return (error);
3526 	vp = fp->f_vnode;
3527 #if 0
3528 	if (!fullsync)
3529 		/* XXXKIB: compete outstanding aio writes */;
3530 #endif
3531 retry:
3532 	error = vn_start_write(vp, &mp, V_WAIT | PCATCH);
3533 	if (error != 0)
3534 		goto drop;
3535 	vn_lock(vp, vn_lktype_write(mp, vp) | LK_RETRY);
3536 	AUDIT_ARG_VNODE1(vp);
3537 	if (vp->v_object != NULL) {
3538 		VM_OBJECT_WLOCK(vp->v_object);
3539 		vm_object_page_clean(vp->v_object, 0, 0, 0);
3540 		VM_OBJECT_WUNLOCK(vp->v_object);
3541 	}
3542 	error = fullsync ? VOP_FSYNC(vp, MNT_WAIT, td) : VOP_FDATASYNC(vp, td);
3543 	VOP_UNLOCK(vp);
3544 	vn_finished_write(mp);
3545 	if (error == ERELOOKUP)
3546 		goto retry;
3547 drop:
3548 	fdrop(fp, td);
3549 	return (error);
3550 }
3551 
3552 /*
3553  * Sync an open file.
3554  */
3555 #ifndef _SYS_SYSPROTO_H_
3556 struct fsync_args {
3557 	int	fd;
3558 };
3559 #endif
3560 int
3561 sys_fsync(struct thread *td, struct fsync_args *uap)
3562 {
3563 
3564 	return (kern_fsync(td, uap->fd, true));
3565 }
3566 
3567 int
3568 sys_fdatasync(struct thread *td, struct fdatasync_args *uap)
3569 {
3570 
3571 	return (kern_fsync(td, uap->fd, false));
3572 }
3573 
3574 /*
3575  * Rename files.  Source and destination must either both be directories, or
3576  * both not be directories.  If target is a directory, it must be empty.
3577  */
3578 #ifndef _SYS_SYSPROTO_H_
3579 struct rename_args {
3580 	char	*from;
3581 	char	*to;
3582 };
3583 #endif
3584 int
3585 sys_rename(struct thread *td, struct rename_args *uap)
3586 {
3587 
3588 	return (kern_renameat(td, AT_FDCWD, uap->from, AT_FDCWD,
3589 	    uap->to, UIO_USERSPACE));
3590 }
3591 
3592 #ifndef _SYS_SYSPROTO_H_
3593 struct renameat_args {
3594 	int	oldfd;
3595 	char	*old;
3596 	int	newfd;
3597 	char	*new;
3598 };
3599 #endif
3600 int
3601 sys_renameat(struct thread *td, struct renameat_args *uap)
3602 {
3603 
3604 	return (kern_renameat(td, uap->oldfd, uap->old, uap->newfd, uap->new,
3605 	    UIO_USERSPACE));
3606 }
3607 
3608 #ifdef MAC
3609 static int
3610 kern_renameat_mac(struct thread *td, int oldfd, const char *old, int newfd,
3611     const char *new, enum uio_seg pathseg, struct nameidata *fromnd)
3612 {
3613 	int error;
3614 
3615 	NDINIT_ATRIGHTS(fromnd, DELETE, LOCKPARENT | LOCKLEAF | SAVESTART |
3616 	    AUDITVNODE1, pathseg, old, oldfd, &cap_renameat_source_rights, td);
3617 	if ((error = namei(fromnd)) != 0)
3618 		return (error);
3619 	error = mac_vnode_check_rename_from(td->td_ucred, fromnd->ni_dvp,
3620 	    fromnd->ni_vp, &fromnd->ni_cnd);
3621 	VOP_UNLOCK(fromnd->ni_dvp);
3622 	if (fromnd->ni_dvp != fromnd->ni_vp)
3623 		VOP_UNLOCK(fromnd->ni_vp);
3624 	if (error != 0) {
3625 		NDFREE(fromnd, NDF_ONLY_PNBUF);
3626 		vrele(fromnd->ni_dvp);
3627 		vrele(fromnd->ni_vp);
3628 		if (fromnd->ni_startdir)
3629 			vrele(fromnd->ni_startdir);
3630 	}
3631 	return (error);
3632 }
3633 #endif
3634 
3635 int
3636 kern_renameat(struct thread *td, int oldfd, const char *old, int newfd,
3637     const char *new, enum uio_seg pathseg)
3638 {
3639 	struct mount *mp = NULL;
3640 	struct vnode *tvp, *fvp, *tdvp;
3641 	struct nameidata fromnd, tond;
3642 	u_int64_t tondflags;
3643 	int error;
3644 
3645 again:
3646 	bwillwrite();
3647 #ifdef MAC
3648 	if (mac_vnode_check_rename_from_enabled()) {
3649 		error = kern_renameat_mac(td, oldfd, old, newfd, new, pathseg,
3650 		    &fromnd);
3651 		if (error != 0)
3652 			return (error);
3653 	} else {
3654 #endif
3655 	NDINIT_ATRIGHTS(&fromnd, DELETE, WANTPARENT | SAVESTART | AUDITVNODE1,
3656 	    pathseg, old, oldfd, &cap_renameat_source_rights, td);
3657 	if ((error = namei(&fromnd)) != 0)
3658 		return (error);
3659 #ifdef MAC
3660 	}
3661 #endif
3662 	fvp = fromnd.ni_vp;
3663 	tondflags = LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART | AUDITVNODE2;
3664 	if (fromnd.ni_vp->v_type == VDIR)
3665 		tondflags |= WILLBEDIR;
3666 	NDINIT_ATRIGHTS(&tond, RENAME, tondflags, pathseg, new, newfd,
3667 	    &cap_renameat_target_rights, td);
3668 	if ((error = namei(&tond)) != 0) {
3669 		/* Translate error code for rename("dir1", "dir2/."). */
3670 		if (error == EISDIR && fvp->v_type == VDIR)
3671 			error = EINVAL;
3672 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3673 		vrele(fromnd.ni_dvp);
3674 		vrele(fvp);
3675 		goto out1;
3676 	}
3677 	tdvp = tond.ni_dvp;
3678 	tvp = tond.ni_vp;
3679 	error = vn_start_write(fvp, &mp, V_NOWAIT);
3680 	if (error != 0) {
3681 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3682 		NDFREE(&tond, NDF_ONLY_PNBUF);
3683 		if (tvp != NULL)
3684 			vput(tvp);
3685 		if (tdvp == tvp)
3686 			vrele(tdvp);
3687 		else
3688 			vput(tdvp);
3689 		vrele(fromnd.ni_dvp);
3690 		vrele(fvp);
3691 		vrele(tond.ni_startdir);
3692 		if (fromnd.ni_startdir != NULL)
3693 			vrele(fromnd.ni_startdir);
3694 		error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH);
3695 		if (error != 0)
3696 			return (error);
3697 		goto again;
3698 	}
3699 	if (tvp != NULL) {
3700 		if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
3701 			error = ENOTDIR;
3702 			goto out;
3703 		} else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
3704 			error = EISDIR;
3705 			goto out;
3706 		}
3707 #ifdef CAPABILITIES
3708 		if (newfd != AT_FDCWD && (tond.ni_resflags & NIRES_ABS) == 0) {
3709 			/*
3710 			 * If the target already exists we require CAP_UNLINKAT
3711 			 * from 'newfd', when newfd was used for the lookup.
3712 			 */
3713 			error = cap_check(&tond.ni_filecaps.fc_rights,
3714 			    &cap_unlinkat_rights);
3715 			if (error != 0)
3716 				goto out;
3717 		}
3718 #endif
3719 	}
3720 	if (fvp == tdvp) {
3721 		error = EINVAL;
3722 		goto out;
3723 	}
3724 	/*
3725 	 * If the source is the same as the destination (that is, if they
3726 	 * are links to the same vnode), then there is nothing to do.
3727 	 */
3728 	if (fvp == tvp)
3729 		error = ERESTART;
3730 #ifdef MAC
3731 	else
3732 		error = mac_vnode_check_rename_to(td->td_ucred, tdvp,
3733 		    tond.ni_vp, fromnd.ni_dvp == tdvp, &tond.ni_cnd);
3734 #endif
3735 out:
3736 	if (error == 0) {
3737 		error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
3738 		    tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
3739 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3740 		NDFREE(&tond, NDF_ONLY_PNBUF);
3741 	} else {
3742 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3743 		NDFREE(&tond, NDF_ONLY_PNBUF);
3744 		if (tvp != NULL)
3745 			vput(tvp);
3746 		if (tdvp == tvp)
3747 			vrele(tdvp);
3748 		else
3749 			vput(tdvp);
3750 		vrele(fromnd.ni_dvp);
3751 		vrele(fvp);
3752 	}
3753 	vrele(tond.ni_startdir);
3754 	vn_finished_write(mp);
3755 out1:
3756 	if (fromnd.ni_startdir)
3757 		vrele(fromnd.ni_startdir);
3758 	if (error == ERESTART)
3759 		return (0);
3760 	if (error == ERELOOKUP)
3761 		goto again;
3762 	return (error);
3763 }
3764 
3765 /*
3766  * Make a directory file.
3767  */
3768 #ifndef _SYS_SYSPROTO_H_
3769 struct mkdir_args {
3770 	char	*path;
3771 	int	mode;
3772 };
3773 #endif
3774 int
3775 sys_mkdir(struct thread *td, struct mkdir_args *uap)
3776 {
3777 
3778 	return (kern_mkdirat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
3779 	    uap->mode));
3780 }
3781 
3782 #ifndef _SYS_SYSPROTO_H_
3783 struct mkdirat_args {
3784 	int	fd;
3785 	char	*path;
3786 	mode_t	mode;
3787 };
3788 #endif
3789 int
3790 sys_mkdirat(struct thread *td, struct mkdirat_args *uap)
3791 {
3792 
3793 	return (kern_mkdirat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode));
3794 }
3795 
3796 int
3797 kern_mkdirat(struct thread *td, int fd, const char *path, enum uio_seg segflg,
3798     int mode)
3799 {
3800 	struct mount *mp;
3801 	struct vattr vattr;
3802 	struct nameidata nd;
3803 	int error;
3804 
3805 	AUDIT_ARG_MODE(mode);
3806 	NDPREINIT(&nd);
3807 restart:
3808 	bwillwrite();
3809 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
3810 	    NC_NOMAKEENTRY | NC_KEEPPOSENTRY | FAILIFEXISTS | WILLBEDIR,
3811 	    segflg, path, fd, &cap_mkdirat_rights, td);
3812 	if ((error = namei(&nd)) != 0)
3813 		return (error);
3814 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3815 		NDFREE(&nd, NDF_ONLY_PNBUF);
3816 		vput(nd.ni_dvp);
3817 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3818 			return (error);
3819 		goto restart;
3820 	}
3821 	VATTR_NULL(&vattr);
3822 	vattr.va_type = VDIR;
3823 	vattr.va_mode = (mode & ACCESSPERMS) &~ td->td_proc->p_pd->pd_cmask;
3824 #ifdef MAC
3825 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
3826 	    &vattr);
3827 	if (error != 0)
3828 		goto out;
3829 #endif
3830 	error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
3831 #ifdef MAC
3832 out:
3833 #endif
3834 	NDFREE(&nd, NDF_ONLY_PNBUF);
3835 	VOP_VPUT_PAIR(nd.ni_dvp, error == 0 ? &nd.ni_vp : NULL, true);
3836 	vn_finished_write(mp);
3837 	if (error == ERELOOKUP)
3838 		goto restart;
3839 	return (error);
3840 }
3841 
3842 /*
3843  * Remove a directory file.
3844  */
3845 #ifndef _SYS_SYSPROTO_H_
3846 struct rmdir_args {
3847 	char	*path;
3848 };
3849 #endif
3850 int
3851 sys_rmdir(struct thread *td, struct rmdir_args *uap)
3852 {
3853 
3854 	return (kern_frmdirat(td, AT_FDCWD, uap->path, FD_NONE, UIO_USERSPACE,
3855 	    0));
3856 }
3857 
3858 int
3859 kern_frmdirat(struct thread *td, int dfd, const char *path, int fd,
3860     enum uio_seg pathseg, int flag)
3861 {
3862 	struct mount *mp;
3863 	struct vnode *vp;
3864 	struct file *fp;
3865 	struct nameidata nd;
3866 	cap_rights_t rights;
3867 	int error;
3868 
3869 	fp = NULL;
3870 	if (fd != FD_NONE) {
3871 		error = getvnode(td, fd, cap_rights_init_one(&rights,
3872 		    CAP_LOOKUP), &fp);
3873 		if (error != 0)
3874 			return (error);
3875 	}
3876 
3877 	NDPREINIT(&nd);
3878 restart:
3879 	bwillwrite();
3880 	NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1 |
3881 	    at2cnpflags(flag, AT_RESOLVE_BENEATH),
3882 	    pathseg, path, dfd, &cap_unlinkat_rights, td);
3883 	if ((error = namei(&nd)) != 0)
3884 		goto fdout;
3885 	vp = nd.ni_vp;
3886 	if (vp->v_type != VDIR) {
3887 		error = ENOTDIR;
3888 		goto out;
3889 	}
3890 	/*
3891 	 * No rmdir "." please.
3892 	 */
3893 	if (nd.ni_dvp == vp) {
3894 		error = EINVAL;
3895 		goto out;
3896 	}
3897 	/*
3898 	 * The root of a mounted filesystem cannot be deleted.
3899 	 */
3900 	if (vp->v_vflag & VV_ROOT) {
3901 		error = EBUSY;
3902 		goto out;
3903 	}
3904 
3905 	if (fp != NULL && fp->f_vnode != vp) {
3906 		if (VN_IS_DOOMED(fp->f_vnode))
3907 			error = EBADF;
3908 		else
3909 			error = EDEADLK;
3910 		goto out;
3911 	}
3912 
3913 #ifdef MAC
3914 	error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
3915 	    &nd.ni_cnd);
3916 	if (error != 0)
3917 		goto out;
3918 #endif
3919 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3920 		NDFREE(&nd, NDF_ONLY_PNBUF);
3921 		vput(vp);
3922 		if (nd.ni_dvp == vp)
3923 			vrele(nd.ni_dvp);
3924 		else
3925 			vput(nd.ni_dvp);
3926 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3927 			goto fdout;
3928 		goto restart;
3929 	}
3930 	vfs_notify_upper(vp, VFS_NOTIFY_UPPER_UNLINK);
3931 	error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
3932 	vn_finished_write(mp);
3933 out:
3934 	NDFREE(&nd, NDF_ONLY_PNBUF);
3935 	vput(vp);
3936 	if (nd.ni_dvp == vp)
3937 		vrele(nd.ni_dvp);
3938 	else
3939 		vput(nd.ni_dvp);
3940 	if (error == ERELOOKUP)
3941 		goto restart;
3942 fdout:
3943 	if (fp != NULL)
3944 		fdrop(fp, td);
3945 	return (error);
3946 }
3947 
3948 #if defined(COMPAT_43) || defined(COMPAT_FREEBSD11)
3949 int
3950 freebsd11_kern_getdirentries(struct thread *td, int fd, char *ubuf, u_int count,
3951     long *basep, void (*func)(struct freebsd11_dirent *))
3952 {
3953 	struct freebsd11_dirent dstdp;
3954 	struct dirent *dp, *edp;
3955 	char *dirbuf;
3956 	off_t base;
3957 	ssize_t resid, ucount;
3958 	int error;
3959 
3960 	/* XXX arbitrary sanity limit on `count'. */
3961 	count = min(count, 64 * 1024);
3962 
3963 	dirbuf = malloc(count, M_TEMP, M_WAITOK);
3964 
3965 	error = kern_getdirentries(td, fd, dirbuf, count, &base, &resid,
3966 	    UIO_SYSSPACE);
3967 	if (error != 0)
3968 		goto done;
3969 	if (basep != NULL)
3970 		*basep = base;
3971 
3972 	ucount = 0;
3973 	for (dp = (struct dirent *)dirbuf,
3974 	    edp = (struct dirent *)&dirbuf[count - resid];
3975 	    ucount < count && dp < edp; ) {
3976 		if (dp->d_reclen == 0)
3977 			break;
3978 		MPASS(dp->d_reclen >= _GENERIC_DIRLEN(0));
3979 		if (dp->d_namlen >= sizeof(dstdp.d_name))
3980 			continue;
3981 		dstdp.d_type = dp->d_type;
3982 		dstdp.d_namlen = dp->d_namlen;
3983 		dstdp.d_fileno = dp->d_fileno;		/* truncate */
3984 		if (dstdp.d_fileno != dp->d_fileno) {
3985 			switch (ino64_trunc_error) {
3986 			default:
3987 			case 0:
3988 				break;
3989 			case 1:
3990 				error = EOVERFLOW;
3991 				goto done;
3992 			case 2:
3993 				dstdp.d_fileno = UINT32_MAX;
3994 				break;
3995 			}
3996 		}
3997 		dstdp.d_reclen = sizeof(dstdp) - sizeof(dstdp.d_name) +
3998 		    ((dp->d_namlen + 1 + 3) &~ 3);
3999 		bcopy(dp->d_name, dstdp.d_name, dstdp.d_namlen);
4000 		bzero(dstdp.d_name + dstdp.d_namlen,
4001 		    dstdp.d_reclen - offsetof(struct freebsd11_dirent, d_name) -
4002 		    dstdp.d_namlen);
4003 		MPASS(dstdp.d_reclen <= dp->d_reclen);
4004 		MPASS(ucount + dstdp.d_reclen <= count);
4005 		if (func != NULL)
4006 			func(&dstdp);
4007 		error = copyout(&dstdp, ubuf + ucount, dstdp.d_reclen);
4008 		if (error != 0)
4009 			break;
4010 		dp = (struct dirent *)((char *)dp + dp->d_reclen);
4011 		ucount += dstdp.d_reclen;
4012 	}
4013 
4014 done:
4015 	free(dirbuf, M_TEMP);
4016 	if (error == 0)
4017 		td->td_retval[0] = ucount;
4018 	return (error);
4019 }
4020 #endif /* COMPAT */
4021 
4022 #ifdef COMPAT_43
4023 static void
4024 ogetdirentries_cvt(struct freebsd11_dirent *dp)
4025 {
4026 #if (BYTE_ORDER == LITTLE_ENDIAN)
4027 	/*
4028 	 * The expected low byte of dp->d_namlen is our dp->d_type.
4029 	 * The high MBZ byte of dp->d_namlen is our dp->d_namlen.
4030 	 */
4031 	dp->d_type = dp->d_namlen;
4032 	dp->d_namlen = 0;
4033 #else
4034 	/*
4035 	 * The dp->d_type is the high byte of the expected dp->d_namlen,
4036 	 * so must be zero'ed.
4037 	 */
4038 	dp->d_type = 0;
4039 #endif
4040 }
4041 
4042 /*
4043  * Read a block of directory entries in a filesystem independent format.
4044  */
4045 #ifndef _SYS_SYSPROTO_H_
4046 struct ogetdirentries_args {
4047 	int	fd;
4048 	char	*buf;
4049 	u_int	count;
4050 	long	*basep;
4051 };
4052 #endif
4053 int
4054 ogetdirentries(struct thread *td, struct ogetdirentries_args *uap)
4055 {
4056 	long loff;
4057 	int error;
4058 
4059 	error = kern_ogetdirentries(td, uap, &loff);
4060 	if (error == 0)
4061 		error = copyout(&loff, uap->basep, sizeof(long));
4062 	return (error);
4063 }
4064 
4065 int
4066 kern_ogetdirentries(struct thread *td, struct ogetdirentries_args *uap,
4067     long *ploff)
4068 {
4069 	long base;
4070 	int error;
4071 
4072 	/* XXX arbitrary sanity limit on `count'. */
4073 	if (uap->count > 64 * 1024)
4074 		return (EINVAL);
4075 
4076 	error = freebsd11_kern_getdirentries(td, uap->fd, uap->buf, uap->count,
4077 	    &base, ogetdirentries_cvt);
4078 
4079 	if (error == 0 && uap->basep != NULL)
4080 		error = copyout(&base, uap->basep, sizeof(long));
4081 
4082 	return (error);
4083 }
4084 #endif /* COMPAT_43 */
4085 
4086 #if defined(COMPAT_FREEBSD11)
4087 #ifndef _SYS_SYSPROTO_H_
4088 struct freebsd11_getdirentries_args {
4089 	int	fd;
4090 	char	*buf;
4091 	u_int	count;
4092 	long	*basep;
4093 };
4094 #endif
4095 int
4096 freebsd11_getdirentries(struct thread *td,
4097     struct freebsd11_getdirentries_args *uap)
4098 {
4099 	long base;
4100 	int error;
4101 
4102 	error = freebsd11_kern_getdirentries(td, uap->fd, uap->buf, uap->count,
4103 	    &base, NULL);
4104 
4105 	if (error == 0 && uap->basep != NULL)
4106 		error = copyout(&base, uap->basep, sizeof(long));
4107 	return (error);
4108 }
4109 
4110 int
4111 freebsd11_getdents(struct thread *td, struct freebsd11_getdents_args *uap)
4112 {
4113 	struct freebsd11_getdirentries_args ap;
4114 
4115 	ap.fd = uap->fd;
4116 	ap.buf = uap->buf;
4117 	ap.count = uap->count;
4118 	ap.basep = NULL;
4119 	return (freebsd11_getdirentries(td, &ap));
4120 }
4121 #endif /* COMPAT_FREEBSD11 */
4122 
4123 /*
4124  * Read a block of directory entries in a filesystem independent format.
4125  */
4126 int
4127 sys_getdirentries(struct thread *td, struct getdirentries_args *uap)
4128 {
4129 	off_t base;
4130 	int error;
4131 
4132 	error = kern_getdirentries(td, uap->fd, uap->buf, uap->count, &base,
4133 	    NULL, UIO_USERSPACE);
4134 	if (error != 0)
4135 		return (error);
4136 	if (uap->basep != NULL)
4137 		error = copyout(&base, uap->basep, sizeof(off_t));
4138 	return (error);
4139 }
4140 
4141 int
4142 kern_getdirentries(struct thread *td, int fd, char *buf, size_t count,
4143     off_t *basep, ssize_t *residp, enum uio_seg bufseg)
4144 {
4145 	struct vnode *vp;
4146 	struct file *fp;
4147 	struct uio auio;
4148 	struct iovec aiov;
4149 	off_t loff;
4150 	int error, eofflag;
4151 	off_t foffset;
4152 
4153 	AUDIT_ARG_FD(fd);
4154 	if (count > IOSIZE_MAX)
4155 		return (EINVAL);
4156 	auio.uio_resid = count;
4157 	error = getvnode(td, fd, &cap_read_rights, &fp);
4158 	if (error != 0)
4159 		return (error);
4160 	if ((fp->f_flag & FREAD) == 0) {
4161 		fdrop(fp, td);
4162 		return (EBADF);
4163 	}
4164 	vp = fp->f_vnode;
4165 	foffset = foffset_lock(fp, 0);
4166 unionread:
4167 	if (vp->v_type != VDIR) {
4168 		error = EINVAL;
4169 		goto fail;
4170 	}
4171 	aiov.iov_base = buf;
4172 	aiov.iov_len = count;
4173 	auio.uio_iov = &aiov;
4174 	auio.uio_iovcnt = 1;
4175 	auio.uio_rw = UIO_READ;
4176 	auio.uio_segflg = bufseg;
4177 	auio.uio_td = td;
4178 	vn_lock(vp, LK_SHARED | LK_RETRY);
4179 	AUDIT_ARG_VNODE1(vp);
4180 	loff = auio.uio_offset = foffset;
4181 #ifdef MAC
4182 	error = mac_vnode_check_readdir(td->td_ucred, vp);
4183 	if (error == 0)
4184 #endif
4185 		error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, NULL,
4186 		    NULL);
4187 	foffset = auio.uio_offset;
4188 	if (error != 0) {
4189 		VOP_UNLOCK(vp);
4190 		goto fail;
4191 	}
4192 	if (count == auio.uio_resid &&
4193 	    (vp->v_vflag & VV_ROOT) &&
4194 	    (vp->v_mount->mnt_flag & MNT_UNION)) {
4195 		struct vnode *tvp = vp;
4196 
4197 		vp = vp->v_mount->mnt_vnodecovered;
4198 		VREF(vp);
4199 		fp->f_vnode = vp;
4200 		foffset = 0;
4201 		vput(tvp);
4202 		goto unionread;
4203 	}
4204 	VOP_UNLOCK(vp);
4205 	*basep = loff;
4206 	if (residp != NULL)
4207 		*residp = auio.uio_resid;
4208 	td->td_retval[0] = count - auio.uio_resid;
4209 fail:
4210 	foffset_unlock(fp, foffset, 0);
4211 	fdrop(fp, td);
4212 	return (error);
4213 }
4214 
4215 /*
4216  * Set the mode mask for creation of filesystem nodes.
4217  */
4218 #ifndef _SYS_SYSPROTO_H_
4219 struct umask_args {
4220 	int	newmask;
4221 };
4222 #endif
4223 int
4224 sys_umask(struct thread *td, struct umask_args *uap)
4225 {
4226 	struct pwddesc *pdp;
4227 
4228 	pdp = td->td_proc->p_pd;
4229 	PWDDESC_XLOCK(pdp);
4230 	td->td_retval[0] = pdp->pd_cmask;
4231 	pdp->pd_cmask = uap->newmask & ALLPERMS;
4232 	PWDDESC_XUNLOCK(pdp);
4233 	return (0);
4234 }
4235 
4236 /*
4237  * Void all references to file by ripping underlying filesystem away from
4238  * vnode.
4239  */
4240 #ifndef _SYS_SYSPROTO_H_
4241 struct revoke_args {
4242 	char	*path;
4243 };
4244 #endif
4245 int
4246 sys_revoke(struct thread *td, struct revoke_args *uap)
4247 {
4248 	struct vnode *vp;
4249 	struct vattr vattr;
4250 	struct nameidata nd;
4251 	int error;
4252 
4253 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
4254 	    uap->path, td);
4255 	if ((error = namei(&nd)) != 0)
4256 		return (error);
4257 	vp = nd.ni_vp;
4258 	NDFREE_NOTHING(&nd);
4259 	if (vp->v_type != VCHR || vp->v_rdev == NULL) {
4260 		error = EINVAL;
4261 		goto out;
4262 	}
4263 #ifdef MAC
4264 	error = mac_vnode_check_revoke(td->td_ucred, vp);
4265 	if (error != 0)
4266 		goto out;
4267 #endif
4268 	error = VOP_GETATTR(vp, &vattr, td->td_ucred);
4269 	if (error != 0)
4270 		goto out;
4271 	if (td->td_ucred->cr_uid != vattr.va_uid) {
4272 		error = priv_check(td, PRIV_VFS_ADMIN);
4273 		if (error != 0)
4274 			goto out;
4275 	}
4276 	if (devfs_usecount(vp) > 0)
4277 		VOP_REVOKE(vp, REVOKEALL);
4278 out:
4279 	vput(vp);
4280 	return (error);
4281 }
4282 
4283 /*
4284  * This variant of getvnode() allows O_PATH files.  Caller should
4285  * ensure that returned file and vnode are only used for compatible
4286  * semantics.
4287  */
4288 int
4289 getvnode_path(struct thread *td, int fd, cap_rights_t *rightsp,
4290     struct file **fpp)
4291 {
4292 	struct file *fp;
4293 	int error;
4294 
4295 	error = fget_unlocked(td->td_proc->p_fd, fd, rightsp, &fp);
4296 	if (error != 0)
4297 		return (error);
4298 
4299 	/*
4300 	 * The file could be not of the vnode type, or it may be not
4301 	 * yet fully initialized, in which case the f_vnode pointer
4302 	 * may be set, but f_ops is still badfileops.  E.g.,
4303 	 * devfs_open() transiently create such situation to
4304 	 * facilitate csw d_fdopen().
4305 	 *
4306 	 * Dupfdopen() handling in kern_openat() installs the
4307 	 * half-baked file into the process descriptor table, allowing
4308 	 * other thread to dereference it. Guard against the race by
4309 	 * checking f_ops.
4310 	 */
4311 	if (fp->f_vnode == NULL || fp->f_ops == &badfileops) {
4312 		fdrop(fp, td);
4313 		return (EINVAL);
4314 	}
4315 
4316 	*fpp = fp;
4317 	return (0);
4318 }
4319 
4320 /*
4321  * Convert a user file descriptor to a kernel file entry and check
4322  * that, if it is a capability, the correct rights are present.
4323  * A reference on the file entry is held upon returning.
4324  */
4325 int
4326 getvnode(struct thread *td, int fd, cap_rights_t *rightsp, struct file **fpp)
4327 {
4328 	int error;
4329 
4330 	error = getvnode_path(td, fd, rightsp, fpp);
4331 
4332 	/*
4333 	 * Filter out O_PATH file descriptors, most getvnode() callers
4334 	 * do not call fo_ methods.
4335 	 */
4336 	if (error == 0 && (*fpp)->f_ops == &path_fileops) {
4337 		fdrop(*fpp, td);
4338 		error = EBADF;
4339 	}
4340 
4341 	return (error);
4342 }
4343 
4344 /*
4345  * Get an (NFS) file handle.
4346  */
4347 #ifndef _SYS_SYSPROTO_H_
4348 struct lgetfh_args {
4349 	char *fname;
4350 	fhandle_t *fhp;
4351 };
4352 #endif
4353 int
4354 sys_lgetfh(struct thread *td, struct lgetfh_args *uap)
4355 {
4356 
4357 	return (kern_getfhat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->fname,
4358 	    UIO_USERSPACE, uap->fhp, UIO_USERSPACE));
4359 }
4360 
4361 #ifndef _SYS_SYSPROTO_H_
4362 struct getfh_args {
4363 	char *fname;
4364 	fhandle_t *fhp;
4365 };
4366 #endif
4367 int
4368 sys_getfh(struct thread *td, struct getfh_args *uap)
4369 {
4370 
4371 	return (kern_getfhat(td, 0, AT_FDCWD, uap->fname, UIO_USERSPACE,
4372 	    uap->fhp, UIO_USERSPACE));
4373 }
4374 
4375 /*
4376  * syscall for the rpc.lockd to use to translate an open descriptor into
4377  * a NFS file handle.
4378  *
4379  * warning: do not remove the priv_check() call or this becomes one giant
4380  * security hole.
4381  */
4382 #ifndef _SYS_SYSPROTO_H_
4383 struct getfhat_args {
4384 	int fd;
4385 	char *path;
4386 	fhandle_t *fhp;
4387 	int flags;
4388 };
4389 #endif
4390 int
4391 sys_getfhat(struct thread *td, struct getfhat_args *uap)
4392 {
4393 
4394 	if ((uap->flags & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH)) != 0)
4395 		return (EINVAL);
4396 	return (kern_getfhat(td, uap->flags, uap->fd, uap->path, UIO_USERSPACE,
4397 	    uap->fhp, UIO_USERSPACE));
4398 }
4399 
4400 int
4401 kern_getfhat(struct thread *td, int flags, int fd, const char *path,
4402     enum uio_seg pathseg, fhandle_t *fhp, enum uio_seg fhseg)
4403 {
4404 	struct nameidata nd;
4405 	fhandle_t fh;
4406 	struct vnode *vp;
4407 	int error;
4408 
4409 	error = priv_check(td, PRIV_VFS_GETFH);
4410 	if (error != 0)
4411 		return (error);
4412 	NDINIT_AT(&nd, LOOKUP, at2cnpflags(flags, AT_SYMLINK_NOFOLLOW |
4413 	    AT_RESOLVE_BENEATH) | LOCKLEAF | AUDITVNODE1, pathseg, path,
4414 	    fd, td);
4415 	error = namei(&nd);
4416 	if (error != 0)
4417 		return (error);
4418 	NDFREE_NOTHING(&nd);
4419 	vp = nd.ni_vp;
4420 	bzero(&fh, sizeof(fh));
4421 	fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
4422 	error = VOP_VPTOFH(vp, &fh.fh_fid);
4423 	vput(vp);
4424 	if (error == 0) {
4425 		if (fhseg == UIO_USERSPACE)
4426 			error = copyout(&fh, fhp, sizeof (fh));
4427 		else
4428 			memcpy(fhp, &fh, sizeof(fh));
4429 	}
4430 	return (error);
4431 }
4432 
4433 #ifndef _SYS_SYSPROTO_H_
4434 struct fhlink_args {
4435 	fhandle_t *fhp;
4436 	const char *to;
4437 };
4438 #endif
4439 int
4440 sys_fhlink(struct thread *td, struct fhlink_args *uap)
4441 {
4442 
4443 	return (kern_fhlinkat(td, AT_FDCWD, uap->to, UIO_USERSPACE, uap->fhp));
4444 }
4445 
4446 #ifndef _SYS_SYSPROTO_H_
4447 struct fhlinkat_args {
4448 	fhandle_t *fhp;
4449 	int tofd;
4450 	const char *to;
4451 };
4452 #endif
4453 int
4454 sys_fhlinkat(struct thread *td, struct fhlinkat_args *uap)
4455 {
4456 
4457 	return (kern_fhlinkat(td, uap->tofd, uap->to, UIO_USERSPACE, uap->fhp));
4458 }
4459 
4460 static int
4461 kern_fhlinkat(struct thread *td, int fd, const char *path,
4462     enum uio_seg pathseg, fhandle_t *fhp)
4463 {
4464 	fhandle_t fh;
4465 	struct mount *mp;
4466 	struct vnode *vp;
4467 	int error;
4468 
4469 	error = priv_check(td, PRIV_VFS_GETFH);
4470 	if (error != 0)
4471 		return (error);
4472 	error = copyin(fhp, &fh, sizeof(fh));
4473 	if (error != 0)
4474 		return (error);
4475 	do {
4476 		bwillwrite();
4477 		if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4478 			return (ESTALE);
4479 		error = VFS_FHTOVP(mp, &fh.fh_fid, LK_SHARED, &vp);
4480 		vfs_unbusy(mp);
4481 		if (error != 0)
4482 			return (error);
4483 		VOP_UNLOCK(vp);
4484 		error = kern_linkat_vp(td, vp, fd, path, pathseg);
4485 	} while (error == EAGAIN || error == ERELOOKUP);
4486 	return (error);
4487 }
4488 
4489 #ifndef _SYS_SYSPROTO_H_
4490 struct fhreadlink_args {
4491 	fhandle_t *fhp;
4492 	char *buf;
4493 	size_t bufsize;
4494 };
4495 #endif
4496 int
4497 sys_fhreadlink(struct thread *td, struct fhreadlink_args *uap)
4498 {
4499 	fhandle_t fh;
4500 	struct mount *mp;
4501 	struct vnode *vp;
4502 	int error;
4503 
4504 	error = priv_check(td, PRIV_VFS_GETFH);
4505 	if (error != 0)
4506 		return (error);
4507 	if (uap->bufsize > IOSIZE_MAX)
4508 		return (EINVAL);
4509 	error = copyin(uap->fhp, &fh, sizeof(fh));
4510 	if (error != 0)
4511 		return (error);
4512 	if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4513 		return (ESTALE);
4514 	error = VFS_FHTOVP(mp, &fh.fh_fid, LK_SHARED, &vp);
4515 	vfs_unbusy(mp);
4516 	if (error != 0)
4517 		return (error);
4518 	error = kern_readlink_vp(vp, uap->buf, UIO_USERSPACE, uap->bufsize, td);
4519 	vput(vp);
4520 	return (error);
4521 }
4522 
4523 /*
4524  * syscall for the rpc.lockd to use to translate a NFS file handle into an
4525  * open descriptor.
4526  *
4527  * warning: do not remove the priv_check() call or this becomes one giant
4528  * security hole.
4529  */
4530 #ifndef _SYS_SYSPROTO_H_
4531 struct fhopen_args {
4532 	const struct fhandle *u_fhp;
4533 	int flags;
4534 };
4535 #endif
4536 int
4537 sys_fhopen(struct thread *td, struct fhopen_args *uap)
4538 {
4539 	return (kern_fhopen(td, uap->u_fhp, uap->flags));
4540 }
4541 
4542 int
4543 kern_fhopen(struct thread *td, const struct fhandle *u_fhp, int flags)
4544 {
4545 	struct mount *mp;
4546 	struct vnode *vp;
4547 	struct fhandle fhp;
4548 	struct file *fp;
4549 	int fmode, error;
4550 	int indx;
4551 
4552 	error = priv_check(td, PRIV_VFS_FHOPEN);
4553 	if (error != 0)
4554 		return (error);
4555 	indx = -1;
4556 	fmode = FFLAGS(flags);
4557 	/* why not allow a non-read/write open for our lockd? */
4558 	if (((fmode & (FREAD | FWRITE)) == 0) || (fmode & O_CREAT))
4559 		return (EINVAL);
4560 	error = copyin(u_fhp, &fhp, sizeof(fhp));
4561 	if (error != 0)
4562 		return(error);
4563 	/* find the mount point */
4564 	mp = vfs_busyfs(&fhp.fh_fsid);
4565 	if (mp == NULL)
4566 		return (ESTALE);
4567 	/* now give me my vnode, it gets returned to me locked */
4568 	error = VFS_FHTOVP(mp, &fhp.fh_fid, LK_EXCLUSIVE, &vp);
4569 	vfs_unbusy(mp);
4570 	if (error != 0)
4571 		return (error);
4572 
4573 	error = falloc_noinstall(td, &fp);
4574 	if (error != 0) {
4575 		vput(vp);
4576 		return (error);
4577 	}
4578 	/*
4579 	 * An extra reference on `fp' has been held for us by
4580 	 * falloc_noinstall().
4581 	 */
4582 
4583 #ifdef INVARIANTS
4584 	td->td_dupfd = -1;
4585 #endif
4586 	error = vn_open_vnode(vp, fmode, td->td_ucred, td, fp);
4587 	if (error != 0) {
4588 		KASSERT(fp->f_ops == &badfileops,
4589 		    ("VOP_OPEN in fhopen() set f_ops"));
4590 		KASSERT(td->td_dupfd < 0,
4591 		    ("fhopen() encountered fdopen()"));
4592 
4593 		vput(vp);
4594 		goto bad;
4595 	}
4596 #ifdef INVARIANTS
4597 	td->td_dupfd = 0;
4598 #endif
4599 	fp->f_vnode = vp;
4600 	finit_vnode(fp, fmode, NULL, &vnops);
4601 	VOP_UNLOCK(vp);
4602 	if ((fmode & O_TRUNC) != 0) {
4603 		error = fo_truncate(fp, 0, td->td_ucred, td);
4604 		if (error != 0)
4605 			goto bad;
4606 	}
4607 
4608 	error = finstall(td, fp, &indx, fmode, NULL);
4609 bad:
4610 	fdrop(fp, td);
4611 	td->td_retval[0] = indx;
4612 	return (error);
4613 }
4614 
4615 /*
4616  * Stat an (NFS) file handle.
4617  */
4618 #ifndef _SYS_SYSPROTO_H_
4619 struct fhstat_args {
4620 	struct fhandle *u_fhp;
4621 	struct stat *sb;
4622 };
4623 #endif
4624 int
4625 sys_fhstat(struct thread *td, struct fhstat_args *uap)
4626 {
4627 	struct stat sb;
4628 	struct fhandle fh;
4629 	int error;
4630 
4631 	error = copyin(uap->u_fhp, &fh, sizeof(fh));
4632 	if (error != 0)
4633 		return (error);
4634 	error = kern_fhstat(td, fh, &sb);
4635 	if (error == 0)
4636 		error = copyout(&sb, uap->sb, sizeof(sb));
4637 	return (error);
4638 }
4639 
4640 int
4641 kern_fhstat(struct thread *td, struct fhandle fh, struct stat *sb)
4642 {
4643 	struct mount *mp;
4644 	struct vnode *vp;
4645 	int error;
4646 
4647 	error = priv_check(td, PRIV_VFS_FHSTAT);
4648 	if (error != 0)
4649 		return (error);
4650 	if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4651 		return (ESTALE);
4652 	error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4653 	vfs_unbusy(mp);
4654 	if (error != 0)
4655 		return (error);
4656 	error = VOP_STAT(vp, sb, td->td_ucred, NOCRED, td);
4657 	vput(vp);
4658 	return (error);
4659 }
4660 
4661 /*
4662  * Implement fstatfs() for (NFS) file handles.
4663  */
4664 #ifndef _SYS_SYSPROTO_H_
4665 struct fhstatfs_args {
4666 	struct fhandle *u_fhp;
4667 	struct statfs *buf;
4668 };
4669 #endif
4670 int
4671 sys_fhstatfs(struct thread *td, struct fhstatfs_args *uap)
4672 {
4673 	struct statfs *sfp;
4674 	fhandle_t fh;
4675 	int error;
4676 
4677 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
4678 	if (error != 0)
4679 		return (error);
4680 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
4681 	error = kern_fhstatfs(td, fh, sfp);
4682 	if (error == 0)
4683 		error = copyout(sfp, uap->buf, sizeof(*sfp));
4684 	free(sfp, M_STATFS);
4685 	return (error);
4686 }
4687 
4688 int
4689 kern_fhstatfs(struct thread *td, fhandle_t fh, struct statfs *buf)
4690 {
4691 	struct mount *mp;
4692 	struct vnode *vp;
4693 	int error;
4694 
4695 	error = priv_check(td, PRIV_VFS_FHSTATFS);
4696 	if (error != 0)
4697 		return (error);
4698 	if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4699 		return (ESTALE);
4700 	error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4701 	if (error != 0) {
4702 		vfs_unbusy(mp);
4703 		return (error);
4704 	}
4705 	vput(vp);
4706 	error = prison_canseemount(td->td_ucred, mp);
4707 	if (error != 0)
4708 		goto out;
4709 #ifdef MAC
4710 	error = mac_mount_check_stat(td->td_ucred, mp);
4711 	if (error != 0)
4712 		goto out;
4713 #endif
4714 	error = VFS_STATFS(mp, buf);
4715 out:
4716 	vfs_unbusy(mp);
4717 	return (error);
4718 }
4719 
4720 /*
4721  * Unlike madvise(2), we do not make a best effort to remember every
4722  * possible caching hint.  Instead, we remember the last setting with
4723  * the exception that we will allow POSIX_FADV_NORMAL to adjust the
4724  * region of any current setting.
4725  */
4726 int
4727 kern_posix_fadvise(struct thread *td, int fd, off_t offset, off_t len,
4728     int advice)
4729 {
4730 	struct fadvise_info *fa, *new;
4731 	struct file *fp;
4732 	struct vnode *vp;
4733 	off_t end;
4734 	int error;
4735 
4736 	if (offset < 0 || len < 0 || offset > OFF_MAX - len)
4737 		return (EINVAL);
4738 	AUDIT_ARG_VALUE(advice);
4739 	switch (advice) {
4740 	case POSIX_FADV_SEQUENTIAL:
4741 	case POSIX_FADV_RANDOM:
4742 	case POSIX_FADV_NOREUSE:
4743 		new = malloc(sizeof(*fa), M_FADVISE, M_WAITOK);
4744 		break;
4745 	case POSIX_FADV_NORMAL:
4746 	case POSIX_FADV_WILLNEED:
4747 	case POSIX_FADV_DONTNEED:
4748 		new = NULL;
4749 		break;
4750 	default:
4751 		return (EINVAL);
4752 	}
4753 	/* XXX: CAP_POSIX_FADVISE? */
4754 	AUDIT_ARG_FD(fd);
4755 	error = fget(td, fd, &cap_no_rights, &fp);
4756 	if (error != 0)
4757 		goto out;
4758 	AUDIT_ARG_FILE(td->td_proc, fp);
4759 	if ((fp->f_ops->fo_flags & DFLAG_SEEKABLE) == 0) {
4760 		error = ESPIPE;
4761 		goto out;
4762 	}
4763 	if (fp->f_type != DTYPE_VNODE) {
4764 		error = ENODEV;
4765 		goto out;
4766 	}
4767 	vp = fp->f_vnode;
4768 	if (vp->v_type != VREG) {
4769 		error = ENODEV;
4770 		goto out;
4771 	}
4772 	if (len == 0)
4773 		end = OFF_MAX;
4774 	else
4775 		end = offset + len - 1;
4776 	switch (advice) {
4777 	case POSIX_FADV_SEQUENTIAL:
4778 	case POSIX_FADV_RANDOM:
4779 	case POSIX_FADV_NOREUSE:
4780 		/*
4781 		 * Try to merge any existing non-standard region with
4782 		 * this new region if possible, otherwise create a new
4783 		 * non-standard region for this request.
4784 		 */
4785 		mtx_pool_lock(mtxpool_sleep, fp);
4786 		fa = fp->f_advice;
4787 		if (fa != NULL && fa->fa_advice == advice &&
4788 		    ((fa->fa_start <= end && fa->fa_end >= offset) ||
4789 		    (end != OFF_MAX && fa->fa_start == end + 1) ||
4790 		    (fa->fa_end != OFF_MAX && fa->fa_end + 1 == offset))) {
4791 			if (offset < fa->fa_start)
4792 				fa->fa_start = offset;
4793 			if (end > fa->fa_end)
4794 				fa->fa_end = end;
4795 		} else {
4796 			new->fa_advice = advice;
4797 			new->fa_start = offset;
4798 			new->fa_end = end;
4799 			fp->f_advice = new;
4800 			new = fa;
4801 		}
4802 		mtx_pool_unlock(mtxpool_sleep, fp);
4803 		break;
4804 	case POSIX_FADV_NORMAL:
4805 		/*
4806 		 * If a the "normal" region overlaps with an existing
4807 		 * non-standard region, trim or remove the
4808 		 * non-standard region.
4809 		 */
4810 		mtx_pool_lock(mtxpool_sleep, fp);
4811 		fa = fp->f_advice;
4812 		if (fa != NULL) {
4813 			if (offset <= fa->fa_start && end >= fa->fa_end) {
4814 				new = fa;
4815 				fp->f_advice = NULL;
4816 			} else if (offset <= fa->fa_start &&
4817 			    end >= fa->fa_start)
4818 				fa->fa_start = end + 1;
4819 			else if (offset <= fa->fa_end && end >= fa->fa_end)
4820 				fa->fa_end = offset - 1;
4821 			else if (offset >= fa->fa_start && end <= fa->fa_end) {
4822 				/*
4823 				 * If the "normal" region is a middle
4824 				 * portion of the existing
4825 				 * non-standard region, just remove
4826 				 * the whole thing rather than picking
4827 				 * one side or the other to
4828 				 * preserve.
4829 				 */
4830 				new = fa;
4831 				fp->f_advice = NULL;
4832 			}
4833 		}
4834 		mtx_pool_unlock(mtxpool_sleep, fp);
4835 		break;
4836 	case POSIX_FADV_WILLNEED:
4837 	case POSIX_FADV_DONTNEED:
4838 		error = VOP_ADVISE(vp, offset, end, advice);
4839 		break;
4840 	}
4841 out:
4842 	if (fp != NULL)
4843 		fdrop(fp, td);
4844 	free(new, M_FADVISE);
4845 	return (error);
4846 }
4847 
4848 int
4849 sys_posix_fadvise(struct thread *td, struct posix_fadvise_args *uap)
4850 {
4851 	int error;
4852 
4853 	error = kern_posix_fadvise(td, uap->fd, uap->offset, uap->len,
4854 	    uap->advice);
4855 	return (kern_posix_error(td, error));
4856 }
4857 
4858 int
4859 kern_copy_file_range(struct thread *td, int infd, off_t *inoffp, int outfd,
4860     off_t *outoffp, size_t len, unsigned int flags)
4861 {
4862 	struct file *infp, *outfp;
4863 	struct vnode *invp, *outvp;
4864 	int error;
4865 	size_t retlen;
4866 	void *rl_rcookie, *rl_wcookie;
4867 	off_t savinoff, savoutoff;
4868 
4869 	infp = outfp = NULL;
4870 	rl_rcookie = rl_wcookie = NULL;
4871 	savinoff = -1;
4872 	error = 0;
4873 	retlen = 0;
4874 
4875 	if (flags != 0) {
4876 		error = EINVAL;
4877 		goto out;
4878 	}
4879 	if (len > SSIZE_MAX)
4880 		/*
4881 		 * Although the len argument is size_t, the return argument
4882 		 * is ssize_t (which is signed).  Therefore a size that won't
4883 		 * fit in ssize_t can't be returned.
4884 		 */
4885 		len = SSIZE_MAX;
4886 
4887 	/* Get the file structures for the file descriptors. */
4888 	error = fget_read(td, infd, &cap_read_rights, &infp);
4889 	if (error != 0)
4890 		goto out;
4891 	if (infp->f_ops == &badfileops) {
4892 		error = EBADF;
4893 		goto out;
4894 	}
4895 	if (infp->f_vnode == NULL) {
4896 		error = EINVAL;
4897 		goto out;
4898 	}
4899 	error = fget_write(td, outfd, &cap_write_rights, &outfp);
4900 	if (error != 0)
4901 		goto out;
4902 	if (outfp->f_ops == &badfileops) {
4903 		error = EBADF;
4904 		goto out;
4905 	}
4906 	if (outfp->f_vnode == NULL) {
4907 		error = EINVAL;
4908 		goto out;
4909 	}
4910 
4911 	/* Set the offset pointers to the correct place. */
4912 	if (inoffp == NULL)
4913 		inoffp = &infp->f_offset;
4914 	if (outoffp == NULL)
4915 		outoffp = &outfp->f_offset;
4916 	savinoff = *inoffp;
4917 	savoutoff = *outoffp;
4918 
4919 	invp = infp->f_vnode;
4920 	outvp = outfp->f_vnode;
4921 	/* Sanity check the f_flag bits. */
4922 	if ((outfp->f_flag & (FWRITE | FAPPEND)) != FWRITE ||
4923 	    (infp->f_flag & FREAD) == 0) {
4924 		error = EBADF;
4925 		goto out;
4926 	}
4927 
4928 	/* If len == 0, just return 0. */
4929 	if (len == 0)
4930 		goto out;
4931 
4932 	/*
4933 	 * If infp and outfp refer to the same file, the byte ranges cannot
4934 	 * overlap.
4935 	 */
4936 	if (invp == outvp && ((savinoff <= savoutoff && savinoff + len >
4937 	    savoutoff) || (savinoff > savoutoff && savoutoff + len >
4938 	    savinoff))) {
4939 		error = EINVAL;
4940 		goto out;
4941 	}
4942 
4943 	/* Range lock the byte ranges for both invp and outvp. */
4944 	for (;;) {
4945 		rl_wcookie = vn_rangelock_wlock(outvp, *outoffp, *outoffp +
4946 		    len);
4947 		rl_rcookie = vn_rangelock_tryrlock(invp, *inoffp, *inoffp +
4948 		    len);
4949 		if (rl_rcookie != NULL)
4950 			break;
4951 		vn_rangelock_unlock(outvp, rl_wcookie);
4952 		rl_rcookie = vn_rangelock_rlock(invp, *inoffp, *inoffp + len);
4953 		vn_rangelock_unlock(invp, rl_rcookie);
4954 	}
4955 
4956 	retlen = len;
4957 	error = vn_copy_file_range(invp, inoffp, outvp, outoffp, &retlen,
4958 	    flags, infp->f_cred, outfp->f_cred, td);
4959 out:
4960 	if (rl_rcookie != NULL)
4961 		vn_rangelock_unlock(invp, rl_rcookie);
4962 	if (rl_wcookie != NULL)
4963 		vn_rangelock_unlock(outvp, rl_wcookie);
4964 	if (savinoff != -1 && (error == EINTR || error == ERESTART)) {
4965 		*inoffp = savinoff;
4966 		*outoffp = savoutoff;
4967 	}
4968 	if (outfp != NULL)
4969 		fdrop(outfp, td);
4970 	if (infp != NULL)
4971 		fdrop(infp, td);
4972 	td->td_retval[0] = retlen;
4973 	return (error);
4974 }
4975 
4976 int
4977 sys_copy_file_range(struct thread *td, struct copy_file_range_args *uap)
4978 {
4979 	off_t inoff, outoff, *inoffp, *outoffp;
4980 	int error;
4981 
4982 	inoffp = outoffp = NULL;
4983 	if (uap->inoffp != NULL) {
4984 		error = copyin(uap->inoffp, &inoff, sizeof(off_t));
4985 		if (error != 0)
4986 			return (error);
4987 		inoffp = &inoff;
4988 	}
4989 	if (uap->outoffp != NULL) {
4990 		error = copyin(uap->outoffp, &outoff, sizeof(off_t));
4991 		if (error != 0)
4992 			return (error);
4993 		outoffp = &outoff;
4994 	}
4995 	error = kern_copy_file_range(td, uap->infd, inoffp, uap->outfd,
4996 	    outoffp, uap->len, uap->flags);
4997 	if (error == 0 && uap->inoffp != NULL)
4998 		error = copyout(inoffp, uap->inoffp, sizeof(off_t));
4999 	if (error == 0 && uap->outoffp != NULL)
5000 		error = copyout(outoffp, uap->outoffp, sizeof(off_t));
5001 	return (error);
5002 }
5003