xref: /freebsd/sys/kern/vfs_syscalls.c (revision 0495ed398c4f64013bab2327eb13a303e1f90c13)
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 #include <ufs/ufs/quota.h>
93 
94 MALLOC_DEFINE(M_FADVISE, "fadvise", "posix_fadvise(2) information");
95 
96 static int kern_chflagsat(struct thread *td, int fd, const char *path,
97     enum uio_seg pathseg, u_long flags, int atflag);
98 static int setfflags(struct thread *td, struct vnode *, u_long);
99 static int getutimes(const struct timeval *, enum uio_seg, struct timespec *);
100 static int getutimens(const struct timespec *, enum uio_seg,
101     struct timespec *, int *);
102 static int setutimes(struct thread *td, struct vnode *,
103     const struct timespec *, int, int);
104 static int vn_access(struct vnode *vp, int user_flags, struct ucred *cred,
105     struct thread *td);
106 static int kern_fhlinkat(struct thread *td, int fd, const char *path,
107     enum uio_seg pathseg, fhandle_t *fhp);
108 static int kern_readlink_vp(struct vnode *vp, char *buf, enum uio_seg bufseg,
109     size_t count, struct thread *td);
110 static int kern_linkat_vp(struct thread *td, struct vnode *vp, int fd,
111     const char *path, enum uio_seg segflag);
112 
113 static uint64_t
114 at2cnpflags(u_int at_flags, u_int mask)
115 {
116 	u_int64_t res;
117 
118 	MPASS((at_flags & (AT_SYMLINK_FOLLOW | AT_SYMLINK_NOFOLLOW)) !=
119 	    (AT_SYMLINK_FOLLOW | AT_SYMLINK_NOFOLLOW));
120 
121 	res = 0;
122 	at_flags &= mask;
123 	if ((at_flags & AT_BENEATH) != 0)
124 		res |= BENEATH;
125 	if ((at_flags & AT_RESOLVE_BENEATH) != 0)
126 		res |= RBENEATH;
127 	if ((at_flags & AT_SYMLINK_FOLLOW) != 0)
128 		res |= FOLLOW;
129 	/* NOFOLLOW is pseudo flag */
130 	if ((mask & AT_SYMLINK_NOFOLLOW) != 0) {
131 		res |= (at_flags & AT_SYMLINK_NOFOLLOW) != 0 ? NOFOLLOW :
132 		    FOLLOW;
133 	}
134 	return (res);
135 }
136 
137 int
138 kern_sync(struct thread *td)
139 {
140 	struct mount *mp, *nmp;
141 	int save;
142 
143 	mtx_lock(&mountlist_mtx);
144 	for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
145 		if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK)) {
146 			nmp = TAILQ_NEXT(mp, mnt_list);
147 			continue;
148 		}
149 		if ((mp->mnt_flag & MNT_RDONLY) == 0 &&
150 		    vn_start_write(NULL, &mp, V_NOWAIT) == 0) {
151 			save = curthread_pflags_set(TDP_SYNCIO);
152 			vfs_periodic(mp, MNT_NOWAIT);
153 			VFS_SYNC(mp, MNT_NOWAIT);
154 			curthread_pflags_restore(save);
155 			vn_finished_write(mp);
156 		}
157 		mtx_lock(&mountlist_mtx);
158 		nmp = TAILQ_NEXT(mp, mnt_list);
159 		vfs_unbusy(mp);
160 	}
161 	mtx_unlock(&mountlist_mtx);
162 	return (0);
163 }
164 
165 /*
166  * Sync each mounted filesystem.
167  */
168 #ifndef _SYS_SYSPROTO_H_
169 struct sync_args {
170 	int     dummy;
171 };
172 #endif
173 /* ARGSUSED */
174 int
175 sys_sync(struct thread *td, struct sync_args *uap)
176 {
177 
178 	return (kern_sync(td));
179 }
180 
181 /*
182  * Change filesystem quotas.
183  */
184 #ifndef _SYS_SYSPROTO_H_
185 struct quotactl_args {
186 	char *path;
187 	int cmd;
188 	int uid;
189 	caddr_t arg;
190 };
191 #endif
192 int
193 sys_quotactl(struct thread *td, struct quotactl_args *uap)
194 {
195 	struct mount *mp;
196 	struct nameidata nd;
197 	int error;
198 
199 	AUDIT_ARG_CMD(uap->cmd);
200 	AUDIT_ARG_UID(uap->uid);
201 	if (!prison_allow(td->td_ucred, PR_ALLOW_QUOTAS))
202 		return (EPERM);
203 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
204 	    uap->path, td);
205 	if ((error = namei(&nd)) != 0)
206 		return (error);
207 	NDFREE(&nd, NDF_ONLY_PNBUF);
208 	mp = nd.ni_vp->v_mount;
209 	vfs_ref(mp);
210 	vput(nd.ni_vp);
211 	error = vfs_busy(mp, 0);
212 	if (error != 0) {
213 		vfs_rel(mp);
214 		return (error);
215 	}
216 	error = VFS_QUOTACTL(mp, uap->cmd, uap->uid, uap->arg);
217 
218 	/*
219 	 * Since quota on operation typically needs to open quota
220 	 * file, the Q_QUOTAON handler needs to unbusy the mount point
221 	 * before calling into namei.  Otherwise, unmount might be
222 	 * started between two vfs_busy() invocations (first is our,
223 	 * second is from mount point cross-walk code in lookup()),
224 	 * causing deadlock.
225 	 *
226 	 * Require that Q_QUOTAON handles the vfs_busy() reference on
227 	 * its own, always returning with ubusied mount point.
228 	 */
229 	if ((uap->cmd >> SUBCMDSHIFT) != Q_QUOTAON &&
230 	    (uap->cmd >> SUBCMDSHIFT) != Q_QUOTAOFF)
231 		vfs_unbusy(mp);
232 	vfs_rel(mp);
233 	return (error);
234 }
235 
236 /*
237  * Used by statfs conversion routines to scale the block size up if
238  * necessary so that all of the block counts are <= 'max_size'.  Note
239  * that 'max_size' should be a bitmask, i.e. 2^n - 1 for some non-zero
240  * value of 'n'.
241  */
242 void
243 statfs_scale_blocks(struct statfs *sf, long max_size)
244 {
245 	uint64_t count;
246 	int shift;
247 
248 	KASSERT(powerof2(max_size + 1), ("%s: invalid max_size", __func__));
249 
250 	/*
251 	 * Attempt to scale the block counts to give a more accurate
252 	 * overview to userland of the ratio of free space to used
253 	 * space.  To do this, find the largest block count and compute
254 	 * a divisor that lets it fit into a signed integer <= max_size.
255 	 */
256 	if (sf->f_bavail < 0)
257 		count = -sf->f_bavail;
258 	else
259 		count = sf->f_bavail;
260 	count = MAX(sf->f_blocks, MAX(sf->f_bfree, count));
261 	if (count <= max_size)
262 		return;
263 
264 	count >>= flsl(max_size);
265 	shift = 0;
266 	while (count > 0) {
267 		shift++;
268 		count >>=1;
269 	}
270 
271 	sf->f_bsize <<= shift;
272 	sf->f_blocks >>= shift;
273 	sf->f_bfree >>= shift;
274 	sf->f_bavail >>= shift;
275 }
276 
277 static int
278 kern_do_statfs(struct thread *td, struct mount *mp, struct statfs *buf)
279 {
280 	int error;
281 
282 	if (mp == NULL)
283 		return (EBADF);
284 	error = vfs_busy(mp, 0);
285 	vfs_rel(mp);
286 	if (error != 0)
287 		return (error);
288 #ifdef MAC
289 	error = mac_mount_check_stat(td->td_ucred, mp);
290 	if (error != 0)
291 		goto out;
292 #endif
293 	error = VFS_STATFS(mp, buf);
294 	if (error != 0)
295 		goto out;
296 	if (priv_check_cred_vfs_generation(td->td_ucred)) {
297 		buf->f_fsid.val[0] = buf->f_fsid.val[1] = 0;
298 		prison_enforce_statfs(td->td_ucred, mp, buf);
299 	}
300 out:
301 	vfs_unbusy(mp);
302 	return (error);
303 }
304 
305 /*
306  * Get filesystem statistics.
307  */
308 #ifndef _SYS_SYSPROTO_H_
309 struct statfs_args {
310 	char *path;
311 	struct statfs *buf;
312 };
313 #endif
314 int
315 sys_statfs(struct thread *td, struct statfs_args *uap)
316 {
317 	struct statfs *sfp;
318 	int error;
319 
320 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
321 	error = kern_statfs(td, uap->path, UIO_USERSPACE, sfp);
322 	if (error == 0)
323 		error = copyout(sfp, uap->buf, sizeof(struct statfs));
324 	free(sfp, M_STATFS);
325 	return (error);
326 }
327 
328 int
329 kern_statfs(struct thread *td, const char *path, enum uio_seg pathseg,
330     struct statfs *buf)
331 {
332 	struct mount *mp;
333 	struct nameidata nd;
334 	int error;
335 
336 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
337 	    pathseg, path, td);
338 	error = namei(&nd);
339 	if (error != 0)
340 		return (error);
341 	mp = nd.ni_vp->v_mount;
342 	vfs_ref(mp);
343 	NDFREE_NOTHING(&nd);
344 	vput(nd.ni_vp);
345 	return (kern_do_statfs(td, mp, buf));
346 }
347 
348 /*
349  * Get filesystem statistics.
350  */
351 #ifndef _SYS_SYSPROTO_H_
352 struct fstatfs_args {
353 	int fd;
354 	struct statfs *buf;
355 };
356 #endif
357 int
358 sys_fstatfs(struct thread *td, struct fstatfs_args *uap)
359 {
360 	struct statfs *sfp;
361 	int error;
362 
363 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
364 	error = kern_fstatfs(td, uap->fd, sfp);
365 	if (error == 0)
366 		error = copyout(sfp, uap->buf, sizeof(struct statfs));
367 	free(sfp, M_STATFS);
368 	return (error);
369 }
370 
371 int
372 kern_fstatfs(struct thread *td, int fd, struct statfs *buf)
373 {
374 	struct file *fp;
375 	struct mount *mp;
376 	struct vnode *vp;
377 	int error;
378 
379 	AUDIT_ARG_FD(fd);
380 	error = getvnode(td, fd, &cap_fstatfs_rights, &fp);
381 	if (error != 0)
382 		return (error);
383 	vp = fp->f_vnode;
384 	vn_lock(vp, LK_SHARED | LK_RETRY);
385 #ifdef AUDIT
386 	AUDIT_ARG_VNODE1(vp);
387 #endif
388 	mp = vp->v_mount;
389 	if (mp != NULL)
390 		vfs_ref(mp);
391 	VOP_UNLOCK(vp);
392 	fdrop(fp, td);
393 	return (kern_do_statfs(td, mp, buf));
394 }
395 
396 /*
397  * Get statistics on all filesystems.
398  */
399 #ifndef _SYS_SYSPROTO_H_
400 struct getfsstat_args {
401 	struct statfs *buf;
402 	long bufsize;
403 	int mode;
404 };
405 #endif
406 int
407 sys_getfsstat(struct thread *td, struct getfsstat_args *uap)
408 {
409 	size_t count;
410 	int error;
411 
412 	if (uap->bufsize < 0 || uap->bufsize > SIZE_MAX)
413 		return (EINVAL);
414 	error = kern_getfsstat(td, &uap->buf, uap->bufsize, &count,
415 	    UIO_USERSPACE, uap->mode);
416 	if (error == 0)
417 		td->td_retval[0] = count;
418 	return (error);
419 }
420 
421 /*
422  * If (bufsize > 0 && bufseg == UIO_SYSSPACE)
423  *	The caller is responsible for freeing memory which will be allocated
424  *	in '*buf'.
425  */
426 int
427 kern_getfsstat(struct thread *td, struct statfs **buf, size_t bufsize,
428     size_t *countp, enum uio_seg bufseg, int mode)
429 {
430 	struct mount *mp, *nmp;
431 	struct statfs *sfsp, *sp, *sptmp, *tofree;
432 	size_t count, maxcount;
433 	int error;
434 
435 	switch (mode) {
436 	case MNT_WAIT:
437 	case MNT_NOWAIT:
438 		break;
439 	default:
440 		if (bufseg == UIO_SYSSPACE)
441 			*buf = NULL;
442 		return (EINVAL);
443 	}
444 restart:
445 	maxcount = bufsize / sizeof(struct statfs);
446 	if (bufsize == 0) {
447 		sfsp = NULL;
448 		tofree = NULL;
449 	} else if (bufseg == UIO_USERSPACE) {
450 		sfsp = *buf;
451 		tofree = NULL;
452 	} else /* if (bufseg == UIO_SYSSPACE) */ {
453 		count = 0;
454 		mtx_lock(&mountlist_mtx);
455 		TAILQ_FOREACH(mp, &mountlist, mnt_list) {
456 			count++;
457 		}
458 		mtx_unlock(&mountlist_mtx);
459 		if (maxcount > count)
460 			maxcount = count;
461 		tofree = sfsp = *buf = malloc(maxcount * sizeof(struct statfs),
462 		    M_STATFS, M_WAITOK);
463 	}
464 
465 	count = 0;
466 
467 	/*
468 	 * If there is no target buffer they only want the count.
469 	 *
470 	 * This could be TAILQ_FOREACH but it is open-coded to match the original
471 	 * code below.
472 	 */
473 	if (sfsp == NULL) {
474 		mtx_lock(&mountlist_mtx);
475 		for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
476 			if (prison_canseemount(td->td_ucred, mp) != 0) {
477 				nmp = TAILQ_NEXT(mp, mnt_list);
478 				continue;
479 			}
480 #ifdef MAC
481 			if (mac_mount_check_stat(td->td_ucred, mp) != 0) {
482 				nmp = TAILQ_NEXT(mp, mnt_list);
483 				continue;
484 			}
485 #endif
486 			count++;
487 			nmp = TAILQ_NEXT(mp, mnt_list);
488 		}
489 		mtx_unlock(&mountlist_mtx);
490 		*countp = count;
491 		return (0);
492 	}
493 
494 	/*
495 	 * They want the entire thing.
496 	 *
497 	 * Short-circuit the corner case of no room for anything, avoids
498 	 * relocking below.
499 	 */
500 	if (maxcount < 1) {
501 		goto out;
502 	}
503 
504 	mtx_lock(&mountlist_mtx);
505 	for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
506 		if (prison_canseemount(td->td_ucred, mp) != 0) {
507 			nmp = TAILQ_NEXT(mp, mnt_list);
508 			continue;
509 		}
510 #ifdef MAC
511 		if (mac_mount_check_stat(td->td_ucred, mp) != 0) {
512 			nmp = TAILQ_NEXT(mp, mnt_list);
513 			continue;
514 		}
515 #endif
516 		if (mode == MNT_WAIT) {
517 			if (vfs_busy(mp, MBF_MNTLSTLOCK) != 0) {
518 				/*
519 				 * If vfs_busy() failed, and MBF_NOWAIT
520 				 * wasn't passed, then the mp is gone.
521 				 * Furthermore, because of MBF_MNTLSTLOCK,
522 				 * the mountlist_mtx was dropped.  We have
523 				 * no other choice than to start over.
524 				 */
525 				mtx_unlock(&mountlist_mtx);
526 				free(tofree, M_STATFS);
527 				goto restart;
528 			}
529 		} else {
530 			if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK) != 0) {
531 				nmp = TAILQ_NEXT(mp, mnt_list);
532 				continue;
533 			}
534 		}
535 		sp = &mp->mnt_stat;
536 		/*
537 		 * If MNT_NOWAIT is specified, do not refresh
538 		 * the fsstat cache.
539 		 */
540 		if (mode != MNT_NOWAIT) {
541 			error = VFS_STATFS(mp, sp);
542 			if (error != 0) {
543 				mtx_lock(&mountlist_mtx);
544 				nmp = TAILQ_NEXT(mp, mnt_list);
545 				vfs_unbusy(mp);
546 				continue;
547 			}
548 		}
549 		if (priv_check_cred_vfs_generation(td->td_ucred)) {
550 			sptmp = malloc(sizeof(struct statfs), M_STATFS,
551 			    M_WAITOK);
552 			*sptmp = *sp;
553 			sptmp->f_fsid.val[0] = sptmp->f_fsid.val[1] = 0;
554 			prison_enforce_statfs(td->td_ucred, mp, sptmp);
555 			sp = sptmp;
556 		} else
557 			sptmp = NULL;
558 		if (bufseg == UIO_SYSSPACE) {
559 			bcopy(sp, sfsp, sizeof(*sp));
560 			free(sptmp, M_STATFS);
561 		} else /* if (bufseg == UIO_USERSPACE) */ {
562 			error = copyout(sp, sfsp, sizeof(*sp));
563 			free(sptmp, M_STATFS);
564 			if (error != 0) {
565 				vfs_unbusy(mp);
566 				return (error);
567 			}
568 		}
569 		sfsp++;
570 		count++;
571 
572 		if (count == maxcount) {
573 			vfs_unbusy(mp);
574 			goto out;
575 		}
576 
577 		mtx_lock(&mountlist_mtx);
578 		nmp = TAILQ_NEXT(mp, mnt_list);
579 		vfs_unbusy(mp);
580 	}
581 	mtx_unlock(&mountlist_mtx);
582 out:
583 	*countp = count;
584 	return (0);
585 }
586 
587 #ifdef COMPAT_FREEBSD4
588 /*
589  * Get old format filesystem statistics.
590  */
591 static void freebsd4_cvtstatfs(struct statfs *, struct ostatfs *);
592 
593 #ifndef _SYS_SYSPROTO_H_
594 struct freebsd4_statfs_args {
595 	char *path;
596 	struct ostatfs *buf;
597 };
598 #endif
599 int
600 freebsd4_statfs(struct thread *td, struct freebsd4_statfs_args *uap)
601 {
602 	struct ostatfs osb;
603 	struct statfs *sfp;
604 	int error;
605 
606 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
607 	error = kern_statfs(td, uap->path, UIO_USERSPACE, sfp);
608 	if (error == 0) {
609 		freebsd4_cvtstatfs(sfp, &osb);
610 		error = copyout(&osb, uap->buf, sizeof(osb));
611 	}
612 	free(sfp, M_STATFS);
613 	return (error);
614 }
615 
616 /*
617  * Get filesystem statistics.
618  */
619 #ifndef _SYS_SYSPROTO_H_
620 struct freebsd4_fstatfs_args {
621 	int fd;
622 	struct ostatfs *buf;
623 };
624 #endif
625 int
626 freebsd4_fstatfs(struct thread *td, struct freebsd4_fstatfs_args *uap)
627 {
628 	struct ostatfs osb;
629 	struct statfs *sfp;
630 	int error;
631 
632 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
633 	error = kern_fstatfs(td, uap->fd, sfp);
634 	if (error == 0) {
635 		freebsd4_cvtstatfs(sfp, &osb);
636 		error = copyout(&osb, uap->buf, sizeof(osb));
637 	}
638 	free(sfp, M_STATFS);
639 	return (error);
640 }
641 
642 /*
643  * Get statistics on all filesystems.
644  */
645 #ifndef _SYS_SYSPROTO_H_
646 struct freebsd4_getfsstat_args {
647 	struct ostatfs *buf;
648 	long bufsize;
649 	int mode;
650 };
651 #endif
652 int
653 freebsd4_getfsstat(struct thread *td, struct freebsd4_getfsstat_args *uap)
654 {
655 	struct statfs *buf, *sp;
656 	struct ostatfs osb;
657 	size_t count, size;
658 	int error;
659 
660 	if (uap->bufsize < 0)
661 		return (EINVAL);
662 	count = uap->bufsize / sizeof(struct ostatfs);
663 	if (count > SIZE_MAX / sizeof(struct statfs))
664 		return (EINVAL);
665 	size = count * sizeof(struct statfs);
666 	error = kern_getfsstat(td, &buf, size, &count, UIO_SYSSPACE,
667 	    uap->mode);
668 	if (error == 0)
669 		td->td_retval[0] = count;
670 	if (size != 0) {
671 		sp = buf;
672 		while (count != 0 && error == 0) {
673 			freebsd4_cvtstatfs(sp, &osb);
674 			error = copyout(&osb, uap->buf, sizeof(osb));
675 			sp++;
676 			uap->buf++;
677 			count--;
678 		}
679 		free(buf, M_STATFS);
680 	}
681 	return (error);
682 }
683 
684 /*
685  * Implement fstatfs() for (NFS) file handles.
686  */
687 #ifndef _SYS_SYSPROTO_H_
688 struct freebsd4_fhstatfs_args {
689 	struct fhandle *u_fhp;
690 	struct ostatfs *buf;
691 };
692 #endif
693 int
694 freebsd4_fhstatfs(struct thread *td, struct freebsd4_fhstatfs_args *uap)
695 {
696 	struct ostatfs osb;
697 	struct statfs *sfp;
698 	fhandle_t fh;
699 	int error;
700 
701 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
702 	if (error != 0)
703 		return (error);
704 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
705 	error = kern_fhstatfs(td, fh, sfp);
706 	if (error == 0) {
707 		freebsd4_cvtstatfs(sfp, &osb);
708 		error = copyout(&osb, uap->buf, sizeof(osb));
709 	}
710 	free(sfp, M_STATFS);
711 	return (error);
712 }
713 
714 /*
715  * Convert a new format statfs structure to an old format statfs structure.
716  */
717 static void
718 freebsd4_cvtstatfs(struct statfs *nsp, struct ostatfs *osp)
719 {
720 
721 	statfs_scale_blocks(nsp, LONG_MAX);
722 	bzero(osp, sizeof(*osp));
723 	osp->f_bsize = nsp->f_bsize;
724 	osp->f_iosize = MIN(nsp->f_iosize, LONG_MAX);
725 	osp->f_blocks = nsp->f_blocks;
726 	osp->f_bfree = nsp->f_bfree;
727 	osp->f_bavail = nsp->f_bavail;
728 	osp->f_files = MIN(nsp->f_files, LONG_MAX);
729 	osp->f_ffree = MIN(nsp->f_ffree, LONG_MAX);
730 	osp->f_owner = nsp->f_owner;
731 	osp->f_type = nsp->f_type;
732 	osp->f_flags = nsp->f_flags;
733 	osp->f_syncwrites = MIN(nsp->f_syncwrites, LONG_MAX);
734 	osp->f_asyncwrites = MIN(nsp->f_asyncwrites, LONG_MAX);
735 	osp->f_syncreads = MIN(nsp->f_syncreads, LONG_MAX);
736 	osp->f_asyncreads = MIN(nsp->f_asyncreads, LONG_MAX);
737 	strlcpy(osp->f_fstypename, nsp->f_fstypename,
738 	    MIN(MFSNAMELEN, OMFSNAMELEN));
739 	strlcpy(osp->f_mntonname, nsp->f_mntonname,
740 	    MIN(MNAMELEN, OMNAMELEN));
741 	strlcpy(osp->f_mntfromname, nsp->f_mntfromname,
742 	    MIN(MNAMELEN, OMNAMELEN));
743 	osp->f_fsid = nsp->f_fsid;
744 }
745 #endif /* COMPAT_FREEBSD4 */
746 
747 #if defined(COMPAT_FREEBSD11)
748 /*
749  * Get old format filesystem statistics.
750  */
751 static void freebsd11_cvtstatfs(struct statfs *, struct freebsd11_statfs *);
752 
753 int
754 freebsd11_statfs(struct thread *td, struct freebsd11_statfs_args *uap)
755 {
756 	struct freebsd11_statfs osb;
757 	struct statfs *sfp;
758 	int error;
759 
760 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
761 	error = kern_statfs(td, uap->path, UIO_USERSPACE, sfp);
762 	if (error == 0) {
763 		freebsd11_cvtstatfs(sfp, &osb);
764 		error = copyout(&osb, uap->buf, sizeof(osb));
765 	}
766 	free(sfp, M_STATFS);
767 	return (error);
768 }
769 
770 /*
771  * Get filesystem statistics.
772  */
773 int
774 freebsd11_fstatfs(struct thread *td, struct freebsd11_fstatfs_args *uap)
775 {
776 	struct freebsd11_statfs osb;
777 	struct statfs *sfp;
778 	int error;
779 
780 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
781 	error = kern_fstatfs(td, uap->fd, sfp);
782 	if (error == 0) {
783 		freebsd11_cvtstatfs(sfp, &osb);
784 		error = copyout(&osb, uap->buf, sizeof(osb));
785 	}
786 	free(sfp, M_STATFS);
787 	return (error);
788 }
789 
790 /*
791  * Get statistics on all filesystems.
792  */
793 int
794 freebsd11_getfsstat(struct thread *td, struct freebsd11_getfsstat_args *uap)
795 {
796 	struct freebsd11_statfs osb;
797 	struct statfs *buf, *sp;
798 	size_t count, size;
799 	int error;
800 
801 	count = uap->bufsize / sizeof(struct ostatfs);
802 	size = count * sizeof(struct statfs);
803 	error = kern_getfsstat(td, &buf, size, &count, UIO_SYSSPACE,
804 	    uap->mode);
805 	if (error == 0)
806 		td->td_retval[0] = count;
807 	if (size > 0) {
808 		sp = buf;
809 		while (count > 0 && error == 0) {
810 			freebsd11_cvtstatfs(sp, &osb);
811 			error = copyout(&osb, uap->buf, sizeof(osb));
812 			sp++;
813 			uap->buf++;
814 			count--;
815 		}
816 		free(buf, M_STATFS);
817 	}
818 	return (error);
819 }
820 
821 /*
822  * Implement fstatfs() for (NFS) file handles.
823  */
824 int
825 freebsd11_fhstatfs(struct thread *td, struct freebsd11_fhstatfs_args *uap)
826 {
827 	struct freebsd11_statfs osb;
828 	struct statfs *sfp;
829 	fhandle_t fh;
830 	int error;
831 
832 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
833 	if (error)
834 		return (error);
835 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
836 	error = kern_fhstatfs(td, fh, sfp);
837 	if (error == 0) {
838 		freebsd11_cvtstatfs(sfp, &osb);
839 		error = copyout(&osb, uap->buf, sizeof(osb));
840 	}
841 	free(sfp, M_STATFS);
842 	return (error);
843 }
844 
845 /*
846  * Convert a new format statfs structure to an old format statfs structure.
847  */
848 static void
849 freebsd11_cvtstatfs(struct statfs *nsp, struct freebsd11_statfs *osp)
850 {
851 
852 	bzero(osp, sizeof(*osp));
853 	osp->f_version = FREEBSD11_STATFS_VERSION;
854 	osp->f_type = nsp->f_type;
855 	osp->f_flags = nsp->f_flags;
856 	osp->f_bsize = nsp->f_bsize;
857 	osp->f_iosize = nsp->f_iosize;
858 	osp->f_blocks = nsp->f_blocks;
859 	osp->f_bfree = nsp->f_bfree;
860 	osp->f_bavail = nsp->f_bavail;
861 	osp->f_files = nsp->f_files;
862 	osp->f_ffree = nsp->f_ffree;
863 	osp->f_syncwrites = nsp->f_syncwrites;
864 	osp->f_asyncwrites = nsp->f_asyncwrites;
865 	osp->f_syncreads = nsp->f_syncreads;
866 	osp->f_asyncreads = nsp->f_asyncreads;
867 	osp->f_namemax = nsp->f_namemax;
868 	osp->f_owner = nsp->f_owner;
869 	osp->f_fsid = nsp->f_fsid;
870 	strlcpy(osp->f_fstypename, nsp->f_fstypename,
871 	    MIN(MFSNAMELEN, sizeof(osp->f_fstypename)));
872 	strlcpy(osp->f_mntonname, nsp->f_mntonname,
873 	    MIN(MNAMELEN, sizeof(osp->f_mntonname)));
874 	strlcpy(osp->f_mntfromname, nsp->f_mntfromname,
875 	    MIN(MNAMELEN, sizeof(osp->f_mntfromname)));
876 }
877 #endif /* COMPAT_FREEBSD11 */
878 
879 /*
880  * Change current working directory to a given file descriptor.
881  */
882 #ifndef _SYS_SYSPROTO_H_
883 struct fchdir_args {
884 	int	fd;
885 };
886 #endif
887 int
888 sys_fchdir(struct thread *td, struct fchdir_args *uap)
889 {
890 	struct vnode *vp, *tdp;
891 	struct mount *mp;
892 	struct file *fp;
893 	int error;
894 
895 	AUDIT_ARG_FD(uap->fd);
896 	error = getvnode(td, uap->fd, &cap_fchdir_rights,
897 	    &fp);
898 	if (error != 0)
899 		return (error);
900 	vp = fp->f_vnode;
901 	vrefact(vp);
902 	fdrop(fp, td);
903 	vn_lock(vp, LK_SHARED | LK_RETRY);
904 	AUDIT_ARG_VNODE1(vp);
905 	error = change_dir(vp, td);
906 	while (!error && (mp = vp->v_mountedhere) != NULL) {
907 		if (vfs_busy(mp, 0))
908 			continue;
909 		error = VFS_ROOT(mp, LK_SHARED, &tdp);
910 		vfs_unbusy(mp);
911 		if (error != 0)
912 			break;
913 		vput(vp);
914 		vp = tdp;
915 	}
916 	if (error != 0) {
917 		vput(vp);
918 		return (error);
919 	}
920 	VOP_UNLOCK(vp);
921 	pwd_chdir(td, vp);
922 	return (0);
923 }
924 
925 /*
926  * Change current working directory (``.'').
927  */
928 #ifndef _SYS_SYSPROTO_H_
929 struct chdir_args {
930 	char	*path;
931 };
932 #endif
933 int
934 sys_chdir(struct thread *td, struct chdir_args *uap)
935 {
936 
937 	return (kern_chdir(td, uap->path, UIO_USERSPACE));
938 }
939 
940 int
941 kern_chdir(struct thread *td, const char *path, enum uio_seg pathseg)
942 {
943 	struct nameidata nd;
944 	int error;
945 
946 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
947 	    pathseg, path, td);
948 	if ((error = namei(&nd)) != 0)
949 		return (error);
950 	if ((error = change_dir(nd.ni_vp, td)) != 0) {
951 		vput(nd.ni_vp);
952 		NDFREE_NOTHING(&nd);
953 		return (error);
954 	}
955 	VOP_UNLOCK(nd.ni_vp);
956 	NDFREE_NOTHING(&nd);
957 	pwd_chdir(td, nd.ni_vp);
958 	return (0);
959 }
960 
961 /*
962  * Change notion of root (``/'') directory.
963  */
964 #ifndef _SYS_SYSPROTO_H_
965 struct chroot_args {
966 	char	*path;
967 };
968 #endif
969 int
970 sys_chroot(struct thread *td, struct chroot_args *uap)
971 {
972 	struct nameidata nd;
973 	int error;
974 
975 	error = priv_check(td, PRIV_VFS_CHROOT);
976 	if (error != 0)
977 		return (error);
978 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
979 	    UIO_USERSPACE, uap->path, td);
980 	error = namei(&nd);
981 	if (error != 0)
982 		goto error;
983 	error = change_dir(nd.ni_vp, td);
984 	if (error != 0)
985 		goto e_vunlock;
986 #ifdef MAC
987 	error = mac_vnode_check_chroot(td->td_ucred, nd.ni_vp);
988 	if (error != 0)
989 		goto e_vunlock;
990 #endif
991 	VOP_UNLOCK(nd.ni_vp);
992 	error = pwd_chroot(td, nd.ni_vp);
993 	vrele(nd.ni_vp);
994 	NDFREE_NOTHING(&nd);
995 	return (error);
996 e_vunlock:
997 	vput(nd.ni_vp);
998 error:
999 	NDFREE_NOTHING(&nd);
1000 	return (error);
1001 }
1002 
1003 /*
1004  * Common routine for chroot and chdir.  Callers must provide a locked vnode
1005  * instance.
1006  */
1007 int
1008 change_dir(struct vnode *vp, struct thread *td)
1009 {
1010 #ifdef MAC
1011 	int error;
1012 #endif
1013 
1014 	ASSERT_VOP_LOCKED(vp, "change_dir(): vp not locked");
1015 	if (vp->v_type != VDIR)
1016 		return (ENOTDIR);
1017 #ifdef MAC
1018 	error = mac_vnode_check_chdir(td->td_ucred, vp);
1019 	if (error != 0)
1020 		return (error);
1021 #endif
1022 	return (VOP_ACCESS(vp, VEXEC, td->td_ucred, td));
1023 }
1024 
1025 static __inline void
1026 flags_to_rights(int flags, cap_rights_t *rightsp)
1027 {
1028 
1029 	if (flags & O_EXEC) {
1030 		cap_rights_set_one(rightsp, CAP_FEXECVE);
1031 	} else {
1032 		switch ((flags & O_ACCMODE)) {
1033 		case O_RDONLY:
1034 			cap_rights_set_one(rightsp, CAP_READ);
1035 			break;
1036 		case O_RDWR:
1037 			cap_rights_set_one(rightsp, CAP_READ);
1038 			/* FALLTHROUGH */
1039 		case O_WRONLY:
1040 			cap_rights_set_one(rightsp, CAP_WRITE);
1041 			if (!(flags & (O_APPEND | O_TRUNC)))
1042 				cap_rights_set_one(rightsp, CAP_SEEK);
1043 			break;
1044 		}
1045 	}
1046 
1047 	if (flags & O_CREAT)
1048 		cap_rights_set_one(rightsp, CAP_CREATE);
1049 
1050 	if (flags & O_TRUNC)
1051 		cap_rights_set_one(rightsp, CAP_FTRUNCATE);
1052 
1053 	if (flags & (O_SYNC | O_FSYNC))
1054 		cap_rights_set_one(rightsp, CAP_FSYNC);
1055 
1056 	if (flags & (O_EXLOCK | O_SHLOCK))
1057 		cap_rights_set_one(rightsp, CAP_FLOCK);
1058 }
1059 
1060 /*
1061  * Check permissions, allocate an open file structure, and call the device
1062  * open routine if any.
1063  */
1064 #ifndef _SYS_SYSPROTO_H_
1065 struct open_args {
1066 	char	*path;
1067 	int	flags;
1068 	int	mode;
1069 };
1070 #endif
1071 int
1072 sys_open(struct thread *td, struct open_args *uap)
1073 {
1074 
1075 	return (kern_openat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1076 	    uap->flags, uap->mode));
1077 }
1078 
1079 #ifndef _SYS_SYSPROTO_H_
1080 struct openat_args {
1081 	int	fd;
1082 	char	*path;
1083 	int	flag;
1084 	int	mode;
1085 };
1086 #endif
1087 int
1088 sys_openat(struct thread *td, struct openat_args *uap)
1089 {
1090 
1091 	AUDIT_ARG_FD(uap->fd);
1092 	return (kern_openat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
1093 	    uap->mode));
1094 }
1095 
1096 int
1097 kern_openat(struct thread *td, int fd, const char *path, enum uio_seg pathseg,
1098     int flags, int mode)
1099 {
1100 	struct proc *p = td->td_proc;
1101 	struct filedesc *fdp;
1102 	struct pwddesc *pdp;
1103 	struct file *fp;
1104 	struct vnode *vp;
1105 	struct nameidata nd;
1106 	cap_rights_t rights;
1107 	int cmode, error, indx;
1108 
1109 	indx = -1;
1110 	fdp = p->p_fd;
1111 	pdp = p->p_pd;
1112 
1113 	AUDIT_ARG_FFLAGS(flags);
1114 	AUDIT_ARG_MODE(mode);
1115 	cap_rights_init_one(&rights, CAP_LOOKUP);
1116 	flags_to_rights(flags, &rights);
1117 	/*
1118 	 * Only one of the O_EXEC, O_RDONLY, O_WRONLY and O_RDWR flags
1119 	 * may be specified.
1120 	 */
1121 	if (flags & O_EXEC) {
1122 		if (flags & O_ACCMODE)
1123 			return (EINVAL);
1124 	} else if ((flags & O_ACCMODE) == O_ACCMODE) {
1125 		return (EINVAL);
1126 	} else {
1127 		flags = FFLAGS(flags);
1128 	}
1129 
1130 	/*
1131 	 * Allocate a file structure. The descriptor to reference it
1132 	 * is allocated and used by finstall_refed() below.
1133 	 */
1134 	error = falloc_noinstall(td, &fp);
1135 	if (error != 0)
1136 		return (error);
1137 	/* Set the flags early so the finit in devfs can pick them up. */
1138 	fp->f_flag = flags & FMASK;
1139 	cmode = ((mode & ~pdp->pd_cmask) & ALLPERMS) & ~S_ISTXT;
1140 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, fd,
1141 	    &rights, td);
1142 	td->td_dupfd = -1;		/* XXX check for fdopen */
1143 	error = vn_open(&nd, &flags, cmode, fp);
1144 	if (error != 0) {
1145 		/*
1146 		 * If the vn_open replaced the method vector, something
1147 		 * wonderous happened deep below and we just pass it up
1148 		 * pretending we know what we do.
1149 		 */
1150 		if (error == ENXIO && fp->f_ops != &badfileops)
1151 			goto success;
1152 
1153 		/*
1154 		 * Handle special fdopen() case. bleh.
1155 		 *
1156 		 * Don't do this for relative (capability) lookups; we don't
1157 		 * understand exactly what would happen, and we don't think
1158 		 * that it ever should.
1159 		 */
1160 		if ((nd.ni_resflags & NIRES_STRICTREL) == 0 &&
1161 		    (error == ENODEV || error == ENXIO) &&
1162 		    td->td_dupfd >= 0) {
1163 			error = dupfdopen(td, fdp, td->td_dupfd, flags, error,
1164 			    &indx);
1165 			if (error == 0)
1166 				goto success;
1167 		}
1168 
1169 		goto bad;
1170 	}
1171 	td->td_dupfd = 0;
1172 	NDFREE(&nd, NDF_ONLY_PNBUF);
1173 	vp = nd.ni_vp;
1174 
1175 	/*
1176 	 * Store the vnode, for any f_type. Typically, the vnode use
1177 	 * count is decremented by direct call to vn_closefile() for
1178 	 * files that switched type in the cdevsw fdopen() method.
1179 	 */
1180 	fp->f_vnode = vp;
1181 	/*
1182 	 * If the file wasn't claimed by devfs bind it to the normal
1183 	 * vnode operations here.
1184 	 */
1185 	if (fp->f_ops == &badfileops) {
1186 		KASSERT(vp->v_type != VFIFO, ("Unexpected fifo."));
1187 		finit_vnode(fp, flags, NULL, &vnops);
1188 	}
1189 
1190 	VOP_UNLOCK(vp);
1191 	if (flags & O_TRUNC) {
1192 		error = fo_truncate(fp, 0, td->td_ucred, td);
1193 		if (error != 0)
1194 			goto bad;
1195 	}
1196 success:
1197 	/*
1198 	 * If we haven't already installed the FD (for dupfdopen), do so now.
1199 	 */
1200 	if (indx == -1) {
1201 		struct filecaps *fcaps;
1202 
1203 #ifdef CAPABILITIES
1204 		if ((nd.ni_resflags & NIRES_STRICTREL) != 0)
1205 			fcaps = &nd.ni_filecaps;
1206 		else
1207 #endif
1208 			fcaps = NULL;
1209 		error = finstall_refed(td, fp, &indx, flags, fcaps);
1210 		/* On success finstall_refed() consumes fcaps. */
1211 		if (error != 0) {
1212 			filecaps_free(&nd.ni_filecaps);
1213 			goto bad;
1214 		}
1215 	} else {
1216 		filecaps_free(&nd.ni_filecaps);
1217 		falloc_abort(td, fp);
1218 	}
1219 
1220 	td->td_retval[0] = indx;
1221 	return (0);
1222 bad:
1223 	KASSERT(indx == -1, ("indx=%d, should be -1", indx));
1224 	falloc_abort(td, fp);
1225 	return (error);
1226 }
1227 
1228 #ifdef COMPAT_43
1229 /*
1230  * Create a file.
1231  */
1232 #ifndef _SYS_SYSPROTO_H_
1233 struct ocreat_args {
1234 	char	*path;
1235 	int	mode;
1236 };
1237 #endif
1238 int
1239 ocreat(struct thread *td, struct ocreat_args *uap)
1240 {
1241 
1242 	return (kern_openat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1243 	    O_WRONLY | O_CREAT | O_TRUNC, uap->mode));
1244 }
1245 #endif /* COMPAT_43 */
1246 
1247 /*
1248  * Create a special file.
1249  */
1250 #ifndef _SYS_SYSPROTO_H_
1251 struct mknodat_args {
1252 	int	fd;
1253 	char	*path;
1254 	mode_t	mode;
1255 	dev_t	dev;
1256 };
1257 #endif
1258 int
1259 sys_mknodat(struct thread *td, struct mknodat_args *uap)
1260 {
1261 
1262 	return (kern_mknodat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode,
1263 	    uap->dev));
1264 }
1265 
1266 #if defined(COMPAT_FREEBSD11)
1267 int
1268 freebsd11_mknod(struct thread *td,
1269     struct freebsd11_mknod_args *uap)
1270 {
1271 
1272 	return (kern_mknodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1273 	    uap->mode, uap->dev));
1274 }
1275 
1276 int
1277 freebsd11_mknodat(struct thread *td,
1278     struct freebsd11_mknodat_args *uap)
1279 {
1280 
1281 	return (kern_mknodat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode,
1282 	    uap->dev));
1283 }
1284 #endif /* COMPAT_FREEBSD11 */
1285 
1286 int
1287 kern_mknodat(struct thread *td, int fd, const char *path, enum uio_seg pathseg,
1288     int mode, dev_t dev)
1289 {
1290 	struct vnode *vp;
1291 	struct mount *mp;
1292 	struct vattr vattr;
1293 	struct nameidata nd;
1294 	int error, whiteout = 0;
1295 
1296 	AUDIT_ARG_MODE(mode);
1297 	AUDIT_ARG_DEV(dev);
1298 	switch (mode & S_IFMT) {
1299 	case S_IFCHR:
1300 	case S_IFBLK:
1301 		error = priv_check(td, PRIV_VFS_MKNOD_DEV);
1302 		if (error == 0 && dev == VNOVAL)
1303 			error = EINVAL;
1304 		break;
1305 	case S_IFWHT:
1306 		error = priv_check(td, PRIV_VFS_MKNOD_WHT);
1307 		break;
1308 	case S_IFIFO:
1309 		if (dev == 0)
1310 			return (kern_mkfifoat(td, fd, path, pathseg, mode));
1311 		/* FALLTHROUGH */
1312 	default:
1313 		error = EINVAL;
1314 		break;
1315 	}
1316 	if (error != 0)
1317 		return (error);
1318 restart:
1319 	bwillwrite();
1320 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
1321 	    NOCACHE, pathseg, path, fd, &cap_mknodat_rights,
1322 	    td);
1323 	if ((error = namei(&nd)) != 0)
1324 		return (error);
1325 	vp = nd.ni_vp;
1326 	if (vp != NULL) {
1327 		NDFREE(&nd, NDF_ONLY_PNBUF);
1328 		if (vp == nd.ni_dvp)
1329 			vrele(nd.ni_dvp);
1330 		else
1331 			vput(nd.ni_dvp);
1332 		vrele(vp);
1333 		return (EEXIST);
1334 	} else {
1335 		VATTR_NULL(&vattr);
1336 		vattr.va_mode = (mode & ALLPERMS) &
1337 		    ~td->td_proc->p_pd->pd_cmask;
1338 		vattr.va_rdev = dev;
1339 		whiteout = 0;
1340 
1341 		switch (mode & S_IFMT) {
1342 		case S_IFCHR:
1343 			vattr.va_type = VCHR;
1344 			break;
1345 		case S_IFBLK:
1346 			vattr.va_type = VBLK;
1347 			break;
1348 		case S_IFWHT:
1349 			whiteout = 1;
1350 			break;
1351 		default:
1352 			panic("kern_mknod: invalid mode");
1353 		}
1354 	}
1355 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1356 		NDFREE(&nd, NDF_ONLY_PNBUF);
1357 		vput(nd.ni_dvp);
1358 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1359 			return (error);
1360 		goto restart;
1361 	}
1362 #ifdef MAC
1363 	if (error == 0 && !whiteout)
1364 		error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp,
1365 		    &nd.ni_cnd, &vattr);
1366 #endif
1367 	if (error == 0) {
1368 		if (whiteout)
1369 			error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE);
1370 		else {
1371 			error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp,
1372 						&nd.ni_cnd, &vattr);
1373 			if (error == 0)
1374 				vput(nd.ni_vp);
1375 		}
1376 	}
1377 	NDFREE(&nd, NDF_ONLY_PNBUF);
1378 	vput(nd.ni_dvp);
1379 	vn_finished_write(mp);
1380 	if (error == ERELOOKUP)
1381 		goto restart;
1382 	return (error);
1383 }
1384 
1385 /*
1386  * Create a named pipe.
1387  */
1388 #ifndef _SYS_SYSPROTO_H_
1389 struct mkfifo_args {
1390 	char	*path;
1391 	int	mode;
1392 };
1393 #endif
1394 int
1395 sys_mkfifo(struct thread *td, struct mkfifo_args *uap)
1396 {
1397 
1398 	return (kern_mkfifoat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1399 	    uap->mode));
1400 }
1401 
1402 #ifndef _SYS_SYSPROTO_H_
1403 struct mkfifoat_args {
1404 	int	fd;
1405 	char	*path;
1406 	mode_t	mode;
1407 };
1408 #endif
1409 int
1410 sys_mkfifoat(struct thread *td, struct mkfifoat_args *uap)
1411 {
1412 
1413 	return (kern_mkfifoat(td, uap->fd, uap->path, UIO_USERSPACE,
1414 	    uap->mode));
1415 }
1416 
1417 int
1418 kern_mkfifoat(struct thread *td, int fd, const char *path,
1419     enum uio_seg pathseg, int mode)
1420 {
1421 	struct mount *mp;
1422 	struct vattr vattr;
1423 	struct nameidata nd;
1424 	int error;
1425 
1426 	AUDIT_ARG_MODE(mode);
1427 restart:
1428 	bwillwrite();
1429 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
1430 	    NOCACHE, pathseg, path, fd, &cap_mkfifoat_rights,
1431 	    td);
1432 	if ((error = namei(&nd)) != 0)
1433 		return (error);
1434 	if (nd.ni_vp != NULL) {
1435 		NDFREE(&nd, NDF_ONLY_PNBUF);
1436 		if (nd.ni_vp == nd.ni_dvp)
1437 			vrele(nd.ni_dvp);
1438 		else
1439 			vput(nd.ni_dvp);
1440 		vrele(nd.ni_vp);
1441 		return (EEXIST);
1442 	}
1443 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1444 		NDFREE(&nd, NDF_ONLY_PNBUF);
1445 		vput(nd.ni_dvp);
1446 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1447 			return (error);
1448 		goto restart;
1449 	}
1450 	VATTR_NULL(&vattr);
1451 	vattr.va_type = VFIFO;
1452 	vattr.va_mode = (mode & ALLPERMS) & ~td->td_proc->p_pd->pd_cmask;
1453 #ifdef MAC
1454 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1455 	    &vattr);
1456 	if (error != 0)
1457 		goto out;
1458 #endif
1459 	error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
1460 	if (error == 0)
1461 		vput(nd.ni_vp);
1462 #ifdef MAC
1463 out:
1464 #endif
1465 	vput(nd.ni_dvp);
1466 	vn_finished_write(mp);
1467 	NDFREE(&nd, NDF_ONLY_PNBUF);
1468 	if (error == ERELOOKUP)
1469 		goto restart;
1470 	return (error);
1471 }
1472 
1473 /*
1474  * Make a hard file link.
1475  */
1476 #ifndef _SYS_SYSPROTO_H_
1477 struct link_args {
1478 	char	*path;
1479 	char	*link;
1480 };
1481 #endif
1482 int
1483 sys_link(struct thread *td, struct link_args *uap)
1484 {
1485 
1486 	return (kern_linkat(td, AT_FDCWD, AT_FDCWD, uap->path, uap->link,
1487 	    UIO_USERSPACE, FOLLOW));
1488 }
1489 
1490 #ifndef _SYS_SYSPROTO_H_
1491 struct linkat_args {
1492 	int	fd1;
1493 	char	*path1;
1494 	int	fd2;
1495 	char	*path2;
1496 	int	flag;
1497 };
1498 #endif
1499 int
1500 sys_linkat(struct thread *td, struct linkat_args *uap)
1501 {
1502 	int flag;
1503 
1504 	flag = uap->flag;
1505 	if ((flag & ~(AT_SYMLINK_FOLLOW | AT_BENEATH |
1506 	    AT_RESOLVE_BENEATH)) != 0)
1507 		return (EINVAL);
1508 
1509 	return (kern_linkat(td, uap->fd1, uap->fd2, uap->path1, uap->path2,
1510 	    UIO_USERSPACE, at2cnpflags(flag, AT_SYMLINK_FOLLOW | AT_BENEATH |
1511 	    AT_RESOLVE_BENEATH)));
1512 }
1513 
1514 int hardlink_check_uid = 0;
1515 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_uid, CTLFLAG_RW,
1516     &hardlink_check_uid, 0,
1517     "Unprivileged processes cannot create hard links to files owned by other "
1518     "users");
1519 static int hardlink_check_gid = 0;
1520 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_gid, CTLFLAG_RW,
1521     &hardlink_check_gid, 0,
1522     "Unprivileged processes cannot create hard links to files owned by other "
1523     "groups");
1524 
1525 static int
1526 can_hardlink(struct vnode *vp, struct ucred *cred)
1527 {
1528 	struct vattr va;
1529 	int error;
1530 
1531 	if (!hardlink_check_uid && !hardlink_check_gid)
1532 		return (0);
1533 
1534 	error = VOP_GETATTR(vp, &va, cred);
1535 	if (error != 0)
1536 		return (error);
1537 
1538 	if (hardlink_check_uid && cred->cr_uid != va.va_uid) {
1539 		error = priv_check_cred(cred, PRIV_VFS_LINK);
1540 		if (error != 0)
1541 			return (error);
1542 	}
1543 
1544 	if (hardlink_check_gid && !groupmember(va.va_gid, cred)) {
1545 		error = priv_check_cred(cred, PRIV_VFS_LINK);
1546 		if (error != 0)
1547 			return (error);
1548 	}
1549 
1550 	return (0);
1551 }
1552 
1553 int
1554 kern_linkat(struct thread *td, int fd1, int fd2, const char *path1,
1555     const char *path2, enum uio_seg segflag, int follow)
1556 {
1557 	struct nameidata nd;
1558 	int error;
1559 
1560 	do {
1561 		bwillwrite();
1562 		NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, segflag,
1563 		    path1, fd1, &cap_linkat_source_rights, td);
1564 		if ((error = namei(&nd)) != 0)
1565 			return (error);
1566 		NDFREE(&nd, NDF_ONLY_PNBUF);
1567 		error = kern_linkat_vp(td, nd.ni_vp, fd2, path2, segflag);
1568 	} while (error ==  EAGAIN || error == ERELOOKUP);
1569 	return (error);
1570 }
1571 
1572 static int
1573 kern_linkat_vp(struct thread *td, struct vnode *vp, int fd, const char *path,
1574     enum uio_seg segflag)
1575 {
1576 	struct nameidata nd;
1577 	struct mount *mp;
1578 	int error;
1579 
1580 	if (vp->v_type == VDIR) {
1581 		vrele(vp);
1582 		return (EPERM);		/* POSIX */
1583 	}
1584 	NDINIT_ATRIGHTS(&nd, CREATE,
1585 	    LOCKPARENT | SAVENAME | AUDITVNODE2 | NOCACHE, segflag, path, fd,
1586 	    &cap_linkat_target_rights, td);
1587 	if ((error = namei(&nd)) == 0) {
1588 		if (nd.ni_vp != NULL) {
1589 			NDFREE(&nd, NDF_ONLY_PNBUF);
1590 			if (nd.ni_dvp == nd.ni_vp)
1591 				vrele(nd.ni_dvp);
1592 			else
1593 				vput(nd.ni_dvp);
1594 			vrele(nd.ni_vp);
1595 			vrele(vp);
1596 			return (EEXIST);
1597 		} else if (nd.ni_dvp->v_mount != vp->v_mount) {
1598 			/*
1599 			 * Cross-device link.  No need to recheck
1600 			 * vp->v_type, since it cannot change, except
1601 			 * to VBAD.
1602 			 */
1603 			NDFREE(&nd, NDF_ONLY_PNBUF);
1604 			vput(nd.ni_dvp);
1605 			vrele(vp);
1606 			return (EXDEV);
1607 		} else if ((error = vn_lock(vp, LK_EXCLUSIVE)) == 0) {
1608 			error = can_hardlink(vp, td->td_ucred);
1609 #ifdef MAC
1610 			if (error == 0)
1611 				error = mac_vnode_check_link(td->td_ucred,
1612 				    nd.ni_dvp, vp, &nd.ni_cnd);
1613 #endif
1614 			if (error != 0) {
1615 				vput(vp);
1616 				vput(nd.ni_dvp);
1617 				NDFREE(&nd, NDF_ONLY_PNBUF);
1618 				return (error);
1619 			}
1620 			error = vn_start_write(vp, &mp, V_NOWAIT);
1621 			if (error != 0) {
1622 				vput(vp);
1623 				vput(nd.ni_dvp);
1624 				NDFREE(&nd, NDF_ONLY_PNBUF);
1625 				error = vn_start_write(NULL, &mp,
1626 				    V_XSLEEP | PCATCH);
1627 				if (error != 0)
1628 					return (error);
1629 				return (EAGAIN);
1630 			}
1631 			error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
1632 			VOP_UNLOCK(vp);
1633 			vput(nd.ni_dvp);
1634 			vn_finished_write(mp);
1635 			NDFREE(&nd, NDF_ONLY_PNBUF);
1636 		} else {
1637 			vput(nd.ni_dvp);
1638 			NDFREE(&nd, NDF_ONLY_PNBUF);
1639 			vrele(vp);
1640 			return (EAGAIN);
1641 		}
1642 	}
1643 	vrele(vp);
1644 	return (error);
1645 }
1646 
1647 /*
1648  * Make a symbolic link.
1649  */
1650 #ifndef _SYS_SYSPROTO_H_
1651 struct symlink_args {
1652 	char	*path;
1653 	char	*link;
1654 };
1655 #endif
1656 int
1657 sys_symlink(struct thread *td, struct symlink_args *uap)
1658 {
1659 
1660 	return (kern_symlinkat(td, uap->path, AT_FDCWD, uap->link,
1661 	    UIO_USERSPACE));
1662 }
1663 
1664 #ifndef _SYS_SYSPROTO_H_
1665 struct symlinkat_args {
1666 	char	*path;
1667 	int	fd;
1668 	char	*path2;
1669 };
1670 #endif
1671 int
1672 sys_symlinkat(struct thread *td, struct symlinkat_args *uap)
1673 {
1674 
1675 	return (kern_symlinkat(td, uap->path1, uap->fd, uap->path2,
1676 	    UIO_USERSPACE));
1677 }
1678 
1679 int
1680 kern_symlinkat(struct thread *td, const char *path1, int fd, const char *path2,
1681     enum uio_seg segflg)
1682 {
1683 	struct mount *mp;
1684 	struct vattr vattr;
1685 	const char *syspath;
1686 	char *tmppath;
1687 	struct nameidata nd;
1688 	int error;
1689 
1690 	if (segflg == UIO_SYSSPACE) {
1691 		syspath = path1;
1692 	} else {
1693 		tmppath = uma_zalloc(namei_zone, M_WAITOK);
1694 		if ((error = copyinstr(path1, tmppath, MAXPATHLEN, NULL)) != 0)
1695 			goto out;
1696 		syspath = tmppath;
1697 	}
1698 	AUDIT_ARG_TEXT(syspath);
1699 restart:
1700 	bwillwrite();
1701 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
1702 	    NOCACHE, segflg, path2, fd, &cap_symlinkat_rights,
1703 	    td);
1704 	if ((error = namei(&nd)) != 0)
1705 		goto out;
1706 	if (nd.ni_vp) {
1707 		NDFREE(&nd, NDF_ONLY_PNBUF);
1708 		if (nd.ni_vp == nd.ni_dvp)
1709 			vrele(nd.ni_dvp);
1710 		else
1711 			vput(nd.ni_dvp);
1712 		vrele(nd.ni_vp);
1713 		error = EEXIST;
1714 		goto out;
1715 	}
1716 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1717 		NDFREE(&nd, NDF_ONLY_PNBUF);
1718 		vput(nd.ni_dvp);
1719 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1720 			goto out;
1721 		goto restart;
1722 	}
1723 	VATTR_NULL(&vattr);
1724 	vattr.va_mode = ACCESSPERMS &~ td->td_proc->p_pd->pd_cmask;
1725 #ifdef MAC
1726 	vattr.va_type = VLNK;
1727 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1728 	    &vattr);
1729 	if (error != 0)
1730 		goto out2;
1731 #endif
1732 	error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, syspath);
1733 	if (error == 0)
1734 		vput(nd.ni_vp);
1735 #ifdef MAC
1736 out2:
1737 #endif
1738 	NDFREE(&nd, NDF_ONLY_PNBUF);
1739 	vput(nd.ni_dvp);
1740 	vn_finished_write(mp);
1741 	if (error == ERELOOKUP)
1742 		goto restart;
1743 out:
1744 	if (segflg != UIO_SYSSPACE)
1745 		uma_zfree(namei_zone, tmppath);
1746 	return (error);
1747 }
1748 
1749 /*
1750  * Delete a whiteout from the filesystem.
1751  */
1752 #ifndef _SYS_SYSPROTO_H_
1753 struct undelete_args {
1754 	char *path;
1755 };
1756 #endif
1757 int
1758 sys_undelete(struct thread *td, struct undelete_args *uap)
1759 {
1760 	struct mount *mp;
1761 	struct nameidata nd;
1762 	int error;
1763 
1764 restart:
1765 	bwillwrite();
1766 	NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | AUDITVNODE1,
1767 	    UIO_USERSPACE, uap->path, td);
1768 	error = namei(&nd);
1769 	if (error != 0)
1770 		return (error);
1771 
1772 	if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) {
1773 		NDFREE(&nd, NDF_ONLY_PNBUF);
1774 		if (nd.ni_vp == nd.ni_dvp)
1775 			vrele(nd.ni_dvp);
1776 		else
1777 			vput(nd.ni_dvp);
1778 		if (nd.ni_vp)
1779 			vrele(nd.ni_vp);
1780 		return (EEXIST);
1781 	}
1782 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1783 		NDFREE(&nd, NDF_ONLY_PNBUF);
1784 		vput(nd.ni_dvp);
1785 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1786 			return (error);
1787 		goto restart;
1788 	}
1789 	error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE);
1790 	NDFREE(&nd, NDF_ONLY_PNBUF);
1791 	vput(nd.ni_dvp);
1792 	vn_finished_write(mp);
1793 	if (error == ERELOOKUP)
1794 		goto restart;
1795 	return (error);
1796 }
1797 
1798 /*
1799  * Delete a name from the filesystem.
1800  */
1801 #ifndef _SYS_SYSPROTO_H_
1802 struct unlink_args {
1803 	char	*path;
1804 };
1805 #endif
1806 int
1807 sys_unlink(struct thread *td, struct unlink_args *uap)
1808 {
1809 
1810 	return (kern_funlinkat(td, AT_FDCWD, uap->path, FD_NONE, UIO_USERSPACE,
1811 	    0, 0));
1812 }
1813 
1814 static int
1815 kern_funlinkat_ex(struct thread *td, int dfd, const char *path, int fd,
1816     int flag, enum uio_seg pathseg, ino_t oldinum)
1817 {
1818 
1819 	if ((flag & ~AT_REMOVEDIR) != 0)
1820 		return (EINVAL);
1821 
1822 	if ((flag & AT_REMOVEDIR) != 0)
1823 		return (kern_frmdirat(td, dfd, path, fd, UIO_USERSPACE, 0));
1824 
1825 	return (kern_funlinkat(td, dfd, path, fd, UIO_USERSPACE, 0, 0));
1826 }
1827 
1828 #ifndef _SYS_SYSPROTO_H_
1829 struct unlinkat_args {
1830 	int	fd;
1831 	char	*path;
1832 	int	flag;
1833 };
1834 #endif
1835 int
1836 sys_unlinkat(struct thread *td, struct unlinkat_args *uap)
1837 {
1838 
1839 	return (kern_funlinkat_ex(td, uap->fd, uap->path, FD_NONE, uap->flag,
1840 	    UIO_USERSPACE, 0));
1841 }
1842 
1843 #ifndef _SYS_SYSPROTO_H_
1844 struct funlinkat_args {
1845 	int		dfd;
1846 	const char	*path;
1847 	int		fd;
1848 	int		flag;
1849 };
1850 #endif
1851 int
1852 sys_funlinkat(struct thread *td, struct funlinkat_args *uap)
1853 {
1854 
1855 	return (kern_funlinkat_ex(td, uap->dfd, uap->path, uap->fd, uap->flag,
1856 	    UIO_USERSPACE, 0));
1857 }
1858 
1859 int
1860 kern_funlinkat(struct thread *td, int dfd, const char *path, int fd,
1861     enum uio_seg pathseg, int flag, ino_t oldinum)
1862 {
1863 	struct mount *mp;
1864 	struct file *fp;
1865 	struct vnode *vp;
1866 	struct nameidata nd;
1867 	struct stat sb;
1868 	int error;
1869 
1870 	fp = NULL;
1871 	if (fd != FD_NONE) {
1872 		error = getvnode(td, fd, &cap_no_rights, &fp);
1873 		if (error != 0)
1874 			return (error);
1875 	}
1876 
1877 restart:
1878 	bwillwrite();
1879 	NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1 |
1880 	    at2cnpflags(flag, AT_BENEATH | AT_RESOLVE_BENEATH),
1881 	    pathseg, path, dfd, &cap_unlinkat_rights, td);
1882 	if ((error = namei(&nd)) != 0) {
1883 		if (error == EINVAL)
1884 			error = EPERM;
1885 		goto fdout;
1886 	}
1887 	vp = nd.ni_vp;
1888 	if (vp->v_type == VDIR && oldinum == 0) {
1889 		error = EPERM;		/* POSIX */
1890 	} else if (oldinum != 0 &&
1891 		  ((error = VOP_STAT(vp, &sb, td->td_ucred, NOCRED, td)) == 0) &&
1892 		  sb.st_ino != oldinum) {
1893 		error = EIDRM;	/* Identifier removed */
1894 	} else if (fp != NULL && fp->f_vnode != vp) {
1895 		if (VN_IS_DOOMED(fp->f_vnode))
1896 			error = EBADF;
1897 		else
1898 			error = EDEADLK;
1899 	} else {
1900 		/*
1901 		 * The root of a mounted filesystem cannot be deleted.
1902 		 *
1903 		 * XXX: can this only be a VDIR case?
1904 		 */
1905 		if (vp->v_vflag & VV_ROOT)
1906 			error = EBUSY;
1907 	}
1908 	if (error == 0) {
1909 		if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1910 			NDFREE(&nd, NDF_ONLY_PNBUF);
1911 			vput(nd.ni_dvp);
1912 			if (vp == nd.ni_dvp)
1913 				vrele(vp);
1914 			else
1915 				vput(vp);
1916 			if ((error = vn_start_write(NULL, &mp,
1917 			    V_XSLEEP | PCATCH)) != 0) {
1918 				goto fdout;
1919 			}
1920 			goto restart;
1921 		}
1922 #ifdef MAC
1923 		error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
1924 		    &nd.ni_cnd);
1925 		if (error != 0)
1926 			goto out;
1927 #endif
1928 		vfs_notify_upper(vp, VFS_NOTIFY_UPPER_UNLINK);
1929 		error = VOP_REMOVE(nd.ni_dvp, vp, &nd.ni_cnd);
1930 #ifdef MAC
1931 out:
1932 #endif
1933 		vn_finished_write(mp);
1934 	}
1935 	NDFREE(&nd, NDF_ONLY_PNBUF);
1936 	vput(nd.ni_dvp);
1937 	if (vp == nd.ni_dvp)
1938 		vrele(vp);
1939 	else
1940 		vput(vp);
1941 	if (error == ERELOOKUP)
1942 		goto restart;
1943 fdout:
1944 	if (fp != NULL)
1945 		fdrop(fp, td);
1946 	return (error);
1947 }
1948 
1949 /*
1950  * Reposition read/write file offset.
1951  */
1952 #ifndef _SYS_SYSPROTO_H_
1953 struct lseek_args {
1954 	int	fd;
1955 	int	pad;
1956 	off_t	offset;
1957 	int	whence;
1958 };
1959 #endif
1960 int
1961 sys_lseek(struct thread *td, struct lseek_args *uap)
1962 {
1963 
1964 	return (kern_lseek(td, uap->fd, uap->offset, uap->whence));
1965 }
1966 
1967 int
1968 kern_lseek(struct thread *td, int fd, off_t offset, int whence)
1969 {
1970 	struct file *fp;
1971 	int error;
1972 
1973 	AUDIT_ARG_FD(fd);
1974 	error = fget(td, fd, &cap_seek_rights, &fp);
1975 	if (error != 0)
1976 		return (error);
1977 	error = (fp->f_ops->fo_flags & DFLAG_SEEKABLE) != 0 ?
1978 	    fo_seek(fp, offset, whence, td) : ESPIPE;
1979 	fdrop(fp, td);
1980 	return (error);
1981 }
1982 
1983 #if defined(COMPAT_43)
1984 /*
1985  * Reposition read/write file offset.
1986  */
1987 #ifndef _SYS_SYSPROTO_H_
1988 struct olseek_args {
1989 	int	fd;
1990 	long	offset;
1991 	int	whence;
1992 };
1993 #endif
1994 int
1995 olseek(struct thread *td, struct olseek_args *uap)
1996 {
1997 
1998 	return (kern_lseek(td, uap->fd, uap->offset, uap->whence));
1999 }
2000 #endif /* COMPAT_43 */
2001 
2002 #if defined(COMPAT_FREEBSD6)
2003 /* Version with the 'pad' argument */
2004 int
2005 freebsd6_lseek(struct thread *td, struct freebsd6_lseek_args *uap)
2006 {
2007 
2008 	return (kern_lseek(td, uap->fd, uap->offset, uap->whence));
2009 }
2010 #endif
2011 
2012 /*
2013  * Check access permissions using passed credentials.
2014  */
2015 static int
2016 vn_access(struct vnode *vp, int user_flags, struct ucred *cred,
2017      struct thread *td)
2018 {
2019 	accmode_t accmode;
2020 	int error;
2021 
2022 	/* Flags == 0 means only check for existence. */
2023 	if (user_flags == 0)
2024 		return (0);
2025 
2026 	accmode = 0;
2027 	if (user_flags & R_OK)
2028 		accmode |= VREAD;
2029 	if (user_flags & W_OK)
2030 		accmode |= VWRITE;
2031 	if (user_flags & X_OK)
2032 		accmode |= VEXEC;
2033 #ifdef MAC
2034 	error = mac_vnode_check_access(cred, vp, accmode);
2035 	if (error != 0)
2036 		return (error);
2037 #endif
2038 	if ((accmode & VWRITE) == 0 || (error = vn_writechk(vp)) == 0)
2039 		error = VOP_ACCESS(vp, accmode, cred, td);
2040 	return (error);
2041 }
2042 
2043 /*
2044  * Check access permissions using "real" credentials.
2045  */
2046 #ifndef _SYS_SYSPROTO_H_
2047 struct access_args {
2048 	char	*path;
2049 	int	amode;
2050 };
2051 #endif
2052 int
2053 sys_access(struct thread *td, struct access_args *uap)
2054 {
2055 
2056 	return (kern_accessat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2057 	    0, uap->amode));
2058 }
2059 
2060 #ifndef _SYS_SYSPROTO_H_
2061 struct faccessat_args {
2062 	int	dirfd;
2063 	char	*path;
2064 	int	amode;
2065 	int	flag;
2066 }
2067 #endif
2068 int
2069 sys_faccessat(struct thread *td, struct faccessat_args *uap)
2070 {
2071 
2072 	return (kern_accessat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
2073 	    uap->amode));
2074 }
2075 
2076 int
2077 kern_accessat(struct thread *td, int fd, const char *path,
2078     enum uio_seg pathseg, int flag, int amode)
2079 {
2080 	struct ucred *cred, *usecred;
2081 	struct vnode *vp;
2082 	struct nameidata nd;
2083 	int error;
2084 
2085 	if ((flag & ~(AT_EACCESS | AT_BENEATH | AT_RESOLVE_BENEATH)) != 0)
2086 		return (EINVAL);
2087 	if (amode != F_OK && (amode & ~(R_OK | W_OK | X_OK)) != 0)
2088 		return (EINVAL);
2089 
2090 	/*
2091 	 * Create and modify a temporary credential instead of one that
2092 	 * is potentially shared (if we need one).
2093 	 */
2094 	cred = td->td_ucred;
2095 	if ((flag & AT_EACCESS) == 0 &&
2096 	    ((cred->cr_uid != cred->cr_ruid ||
2097 	    cred->cr_rgid != cred->cr_groups[0]))) {
2098 		usecred = crdup(cred);
2099 		usecred->cr_uid = cred->cr_ruid;
2100 		usecred->cr_groups[0] = cred->cr_rgid;
2101 		td->td_ucred = usecred;
2102 	} else
2103 		usecred = cred;
2104 	AUDIT_ARG_VALUE(amode);
2105 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF |
2106 	    AUDITVNODE1 | at2cnpflags(flag, AT_BENEATH | AT_RESOLVE_BENEATH),
2107 	    pathseg, path, fd, &cap_fstat_rights, td);
2108 	if ((error = namei(&nd)) != 0)
2109 		goto out;
2110 	vp = nd.ni_vp;
2111 
2112 	error = vn_access(vp, amode, usecred, td);
2113 	NDFREE_NOTHING(&nd);
2114 	vput(vp);
2115 out:
2116 	if (usecred != cred) {
2117 		td->td_ucred = cred;
2118 		crfree(usecred);
2119 	}
2120 	return (error);
2121 }
2122 
2123 /*
2124  * Check access permissions using "effective" credentials.
2125  */
2126 #ifndef _SYS_SYSPROTO_H_
2127 struct eaccess_args {
2128 	char	*path;
2129 	int	amode;
2130 };
2131 #endif
2132 int
2133 sys_eaccess(struct thread *td, struct eaccess_args *uap)
2134 {
2135 
2136 	return (kern_accessat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2137 	    AT_EACCESS, uap->amode));
2138 }
2139 
2140 #if defined(COMPAT_43)
2141 /*
2142  * Get file status; this version follows links.
2143  */
2144 #ifndef _SYS_SYSPROTO_H_
2145 struct ostat_args {
2146 	char	*path;
2147 	struct ostat *ub;
2148 };
2149 #endif
2150 int
2151 ostat(struct thread *td, struct ostat_args *uap)
2152 {
2153 	struct stat sb;
2154 	struct ostat osb;
2155 	int error;
2156 
2157 	error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE,
2158 	    &sb, NULL);
2159 	if (error != 0)
2160 		return (error);
2161 	cvtstat(&sb, &osb);
2162 	return (copyout(&osb, uap->ub, sizeof (osb)));
2163 }
2164 
2165 /*
2166  * Get file status; this version does not follow links.
2167  */
2168 #ifndef _SYS_SYSPROTO_H_
2169 struct olstat_args {
2170 	char	*path;
2171 	struct ostat *ub;
2172 };
2173 #endif
2174 int
2175 olstat(struct thread *td, struct olstat_args *uap)
2176 {
2177 	struct stat sb;
2178 	struct ostat osb;
2179 	int error;
2180 
2181 	error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2182 	    UIO_USERSPACE, &sb, NULL);
2183 	if (error != 0)
2184 		return (error);
2185 	cvtstat(&sb, &osb);
2186 	return (copyout(&osb, uap->ub, sizeof (osb)));
2187 }
2188 
2189 /*
2190  * Convert from an old to a new stat structure.
2191  * XXX: many values are blindly truncated.
2192  */
2193 void
2194 cvtstat(struct stat *st, struct ostat *ost)
2195 {
2196 
2197 	bzero(ost, sizeof(*ost));
2198 	ost->st_dev = st->st_dev;
2199 	ost->st_ino = st->st_ino;
2200 	ost->st_mode = st->st_mode;
2201 	ost->st_nlink = st->st_nlink;
2202 	ost->st_uid = st->st_uid;
2203 	ost->st_gid = st->st_gid;
2204 	ost->st_rdev = st->st_rdev;
2205 	ost->st_size = MIN(st->st_size, INT32_MAX);
2206 	ost->st_atim = st->st_atim;
2207 	ost->st_mtim = st->st_mtim;
2208 	ost->st_ctim = st->st_ctim;
2209 	ost->st_blksize = st->st_blksize;
2210 	ost->st_blocks = st->st_blocks;
2211 	ost->st_flags = st->st_flags;
2212 	ost->st_gen = st->st_gen;
2213 }
2214 #endif /* COMPAT_43 */
2215 
2216 #if defined(COMPAT_43) || defined(COMPAT_FREEBSD11)
2217 int ino64_trunc_error;
2218 SYSCTL_INT(_vfs, OID_AUTO, ino64_trunc_error, CTLFLAG_RW,
2219     &ino64_trunc_error, 0,
2220     "Error on truncation of device, file or inode number, or link count");
2221 
2222 int
2223 freebsd11_cvtstat(struct stat *st, struct freebsd11_stat *ost)
2224 {
2225 
2226 	ost->st_dev = st->st_dev;
2227 	if (ost->st_dev != st->st_dev) {
2228 		switch (ino64_trunc_error) {
2229 		default:
2230 			/*
2231 			 * Since dev_t is almost raw, don't clamp to the
2232 			 * maximum for case 2, but ignore the error.
2233 			 */
2234 			break;
2235 		case 1:
2236 			return (EOVERFLOW);
2237 		}
2238 	}
2239 	ost->st_ino = st->st_ino;
2240 	if (ost->st_ino != st->st_ino) {
2241 		switch (ino64_trunc_error) {
2242 		default:
2243 		case 0:
2244 			break;
2245 		case 1:
2246 			return (EOVERFLOW);
2247 		case 2:
2248 			ost->st_ino = UINT32_MAX;
2249 			break;
2250 		}
2251 	}
2252 	ost->st_mode = st->st_mode;
2253 	ost->st_nlink = st->st_nlink;
2254 	if (ost->st_nlink != st->st_nlink) {
2255 		switch (ino64_trunc_error) {
2256 		default:
2257 		case 0:
2258 			break;
2259 		case 1:
2260 			return (EOVERFLOW);
2261 		case 2:
2262 			ost->st_nlink = UINT16_MAX;
2263 			break;
2264 		}
2265 	}
2266 	ost->st_uid = st->st_uid;
2267 	ost->st_gid = st->st_gid;
2268 	ost->st_rdev = st->st_rdev;
2269 	if (ost->st_rdev != st->st_rdev) {
2270 		switch (ino64_trunc_error) {
2271 		default:
2272 			break;
2273 		case 1:
2274 			return (EOVERFLOW);
2275 		}
2276 	}
2277 	ost->st_atim = st->st_atim;
2278 	ost->st_mtim = st->st_mtim;
2279 	ost->st_ctim = st->st_ctim;
2280 	ost->st_size = st->st_size;
2281 	ost->st_blocks = st->st_blocks;
2282 	ost->st_blksize = st->st_blksize;
2283 	ost->st_flags = st->st_flags;
2284 	ost->st_gen = st->st_gen;
2285 	ost->st_lspare = 0;
2286 	ost->st_birthtim = st->st_birthtim;
2287 	bzero((char *)&ost->st_birthtim + sizeof(ost->st_birthtim),
2288 	    sizeof(*ost) - offsetof(struct freebsd11_stat,
2289 	    st_birthtim) - sizeof(ost->st_birthtim));
2290 	return (0);
2291 }
2292 
2293 int
2294 freebsd11_stat(struct thread *td, struct freebsd11_stat_args* uap)
2295 {
2296 	struct stat sb;
2297 	struct freebsd11_stat osb;
2298 	int error;
2299 
2300 	error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE,
2301 	    &sb, NULL);
2302 	if (error != 0)
2303 		return (error);
2304 	error = freebsd11_cvtstat(&sb, &osb);
2305 	if (error == 0)
2306 		error = copyout(&osb, uap->ub, sizeof(osb));
2307 	return (error);
2308 }
2309 
2310 int
2311 freebsd11_lstat(struct thread *td, struct freebsd11_lstat_args* uap)
2312 {
2313 	struct stat sb;
2314 	struct freebsd11_stat osb;
2315 	int error;
2316 
2317 	error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2318 	    UIO_USERSPACE, &sb, NULL);
2319 	if (error != 0)
2320 		return (error);
2321 	error = freebsd11_cvtstat(&sb, &osb);
2322 	if (error == 0)
2323 		error = copyout(&osb, uap->ub, sizeof(osb));
2324 	return (error);
2325 }
2326 
2327 int
2328 freebsd11_fhstat(struct thread *td, struct freebsd11_fhstat_args* uap)
2329 {
2330 	struct fhandle fh;
2331 	struct stat sb;
2332 	struct freebsd11_stat osb;
2333 	int error;
2334 
2335 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
2336 	if (error != 0)
2337 		return (error);
2338 	error = kern_fhstat(td, fh, &sb);
2339 	if (error != 0)
2340 		return (error);
2341 	error = freebsd11_cvtstat(&sb, &osb);
2342 	if (error == 0)
2343 		error = copyout(&osb, uap->sb, sizeof(osb));
2344 	return (error);
2345 }
2346 
2347 int
2348 freebsd11_fstatat(struct thread *td, struct freebsd11_fstatat_args* uap)
2349 {
2350 	struct stat sb;
2351 	struct freebsd11_stat osb;
2352 	int error;
2353 
2354 	error = kern_statat(td, uap->flag, uap->fd, uap->path,
2355 	    UIO_USERSPACE, &sb, NULL);
2356 	if (error != 0)
2357 		return (error);
2358 	error = freebsd11_cvtstat(&sb, &osb);
2359 	if (error == 0)
2360 		error = copyout(&osb, uap->buf, sizeof(osb));
2361 	return (error);
2362 }
2363 #endif	/* COMPAT_FREEBSD11 */
2364 
2365 /*
2366  * Get file status
2367  */
2368 #ifndef _SYS_SYSPROTO_H_
2369 struct fstatat_args {
2370 	int	fd;
2371 	char	*path;
2372 	struct stat	*buf;
2373 	int	flag;
2374 }
2375 #endif
2376 int
2377 sys_fstatat(struct thread *td, struct fstatat_args *uap)
2378 {
2379 	struct stat sb;
2380 	int error;
2381 
2382 	error = kern_statat(td, uap->flag, uap->fd, uap->path,
2383 	    UIO_USERSPACE, &sb, NULL);
2384 	if (error == 0)
2385 		error = copyout(&sb, uap->buf, sizeof (sb));
2386 	return (error);
2387 }
2388 
2389 int
2390 kern_statat(struct thread *td, int flag, int fd, const char *path,
2391     enum uio_seg pathseg, struct stat *sbp,
2392     void (*hook)(struct vnode *vp, struct stat *sbp))
2393 {
2394 	struct nameidata nd;
2395 	int error;
2396 
2397 	if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_BENEATH |
2398 	    AT_RESOLVE_BENEATH)) != 0)
2399 		return (EINVAL);
2400 
2401 	NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(flag, AT_BENEATH |
2402 	    AT_RESOLVE_BENEATH | AT_SYMLINK_NOFOLLOW) | LOCKSHARED | LOCKLEAF |
2403 	    AUDITVNODE1, pathseg, path, fd, &cap_fstat_rights, td);
2404 
2405 	if ((error = namei(&nd)) != 0)
2406 		return (error);
2407 	error = VOP_STAT(nd.ni_vp, sbp, td->td_ucred, NOCRED, td);
2408 	if (error == 0) {
2409 		if (__predict_false(hook != NULL))
2410 			hook(nd.ni_vp, sbp);
2411 	}
2412 	NDFREE_NOTHING(&nd);
2413 	vput(nd.ni_vp);
2414 #ifdef __STAT_TIME_T_EXT
2415 	sbp->st_atim_ext = 0;
2416 	sbp->st_mtim_ext = 0;
2417 	sbp->st_ctim_ext = 0;
2418 	sbp->st_btim_ext = 0;
2419 #endif
2420 #ifdef KTRACE
2421 	if (KTRPOINT(td, KTR_STRUCT))
2422 		ktrstat_error(sbp, error);
2423 #endif
2424 	return (error);
2425 }
2426 
2427 #if defined(COMPAT_FREEBSD11)
2428 /*
2429  * Implementation of the NetBSD [l]stat() functions.
2430  */
2431 void
2432 freebsd11_cvtnstat(struct stat *sb, struct nstat *nsb)
2433 {
2434 
2435 	bzero(nsb, sizeof(*nsb));
2436 	nsb->st_dev = sb->st_dev;
2437 	nsb->st_ino = sb->st_ino;
2438 	nsb->st_mode = sb->st_mode;
2439 	nsb->st_nlink = sb->st_nlink;
2440 	nsb->st_uid = sb->st_uid;
2441 	nsb->st_gid = sb->st_gid;
2442 	nsb->st_rdev = sb->st_rdev;
2443 	nsb->st_atim = sb->st_atim;
2444 	nsb->st_mtim = sb->st_mtim;
2445 	nsb->st_ctim = sb->st_ctim;
2446 	nsb->st_size = sb->st_size;
2447 	nsb->st_blocks = sb->st_blocks;
2448 	nsb->st_blksize = sb->st_blksize;
2449 	nsb->st_flags = sb->st_flags;
2450 	nsb->st_gen = sb->st_gen;
2451 	nsb->st_birthtim = sb->st_birthtim;
2452 }
2453 
2454 #ifndef _SYS_SYSPROTO_H_
2455 struct freebsd11_nstat_args {
2456 	char	*path;
2457 	struct nstat *ub;
2458 };
2459 #endif
2460 int
2461 freebsd11_nstat(struct thread *td, struct freebsd11_nstat_args *uap)
2462 {
2463 	struct stat sb;
2464 	struct nstat nsb;
2465 	int error;
2466 
2467 	error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE,
2468 	    &sb, NULL);
2469 	if (error != 0)
2470 		return (error);
2471 	freebsd11_cvtnstat(&sb, &nsb);
2472 	return (copyout(&nsb, uap->ub, sizeof (nsb)));
2473 }
2474 
2475 /*
2476  * NetBSD lstat.  Get file status; this version does not follow links.
2477  */
2478 #ifndef _SYS_SYSPROTO_H_
2479 struct freebsd11_nlstat_args {
2480 	char	*path;
2481 	struct nstat *ub;
2482 };
2483 #endif
2484 int
2485 freebsd11_nlstat(struct thread *td, struct freebsd11_nlstat_args *uap)
2486 {
2487 	struct stat sb;
2488 	struct nstat nsb;
2489 	int error;
2490 
2491 	error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2492 	    UIO_USERSPACE, &sb, NULL);
2493 	if (error != 0)
2494 		return (error);
2495 	freebsd11_cvtnstat(&sb, &nsb);
2496 	return (copyout(&nsb, uap->ub, sizeof (nsb)));
2497 }
2498 #endif /* COMPAT_FREEBSD11 */
2499 
2500 /*
2501  * Get configurable pathname variables.
2502  */
2503 #ifndef _SYS_SYSPROTO_H_
2504 struct pathconf_args {
2505 	char	*path;
2506 	int	name;
2507 };
2508 #endif
2509 int
2510 sys_pathconf(struct thread *td, struct pathconf_args *uap)
2511 {
2512 	long value;
2513 	int error;
2514 
2515 	error = kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name, FOLLOW,
2516 	    &value);
2517 	if (error == 0)
2518 		td->td_retval[0] = value;
2519 	return (error);
2520 }
2521 
2522 #ifndef _SYS_SYSPROTO_H_
2523 struct lpathconf_args {
2524 	char	*path;
2525 	int	name;
2526 };
2527 #endif
2528 int
2529 sys_lpathconf(struct thread *td, struct lpathconf_args *uap)
2530 {
2531 	long value;
2532 	int error;
2533 
2534 	error = kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name,
2535 	    NOFOLLOW, &value);
2536 	if (error == 0)
2537 		td->td_retval[0] = value;
2538 	return (error);
2539 }
2540 
2541 int
2542 kern_pathconf(struct thread *td, const char *path, enum uio_seg pathseg,
2543     int name, u_long flags, long *valuep)
2544 {
2545 	struct nameidata nd;
2546 	int error;
2547 
2548 	NDINIT(&nd, LOOKUP, LOCKSHARED | LOCKLEAF | AUDITVNODE1 | flags,
2549 	    pathseg, path, td);
2550 	if ((error = namei(&nd)) != 0)
2551 		return (error);
2552 	NDFREE_NOTHING(&nd);
2553 
2554 	error = VOP_PATHCONF(nd.ni_vp, name, valuep);
2555 	vput(nd.ni_vp);
2556 	return (error);
2557 }
2558 
2559 /*
2560  * Return target name of a symbolic link.
2561  */
2562 #ifndef _SYS_SYSPROTO_H_
2563 struct readlink_args {
2564 	char	*path;
2565 	char	*buf;
2566 	size_t	count;
2567 };
2568 #endif
2569 int
2570 sys_readlink(struct thread *td, struct readlink_args *uap)
2571 {
2572 
2573 	return (kern_readlinkat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2574 	    uap->buf, UIO_USERSPACE, uap->count));
2575 }
2576 #ifndef _SYS_SYSPROTO_H_
2577 struct readlinkat_args {
2578 	int	fd;
2579 	char	*path;
2580 	char	*buf;
2581 	size_t	bufsize;
2582 };
2583 #endif
2584 int
2585 sys_readlinkat(struct thread *td, struct readlinkat_args *uap)
2586 {
2587 
2588 	return (kern_readlinkat(td, uap->fd, uap->path, UIO_USERSPACE,
2589 	    uap->buf, UIO_USERSPACE, uap->bufsize));
2590 }
2591 
2592 int
2593 kern_readlinkat(struct thread *td, int fd, const char *path,
2594     enum uio_seg pathseg, char *buf, enum uio_seg bufseg, size_t count)
2595 {
2596 	struct vnode *vp;
2597 	struct nameidata nd;
2598 	int error;
2599 
2600 	if (count > IOSIZE_MAX)
2601 		return (EINVAL);
2602 
2603 	NDINIT_AT(&nd, LOOKUP, NOFOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
2604 	    pathseg, path, fd, td);
2605 
2606 	if ((error = namei(&nd)) != 0)
2607 		return (error);
2608 	NDFREE_NOTHING(&nd);
2609 	vp = nd.ni_vp;
2610 
2611 	error = kern_readlink_vp(vp, buf, bufseg, count, td);
2612 	vput(vp);
2613 
2614 	return (error);
2615 }
2616 
2617 /*
2618  * Helper function to readlink from a vnode
2619  */
2620 static int
2621 kern_readlink_vp(struct vnode *vp, char *buf, enum uio_seg bufseg, size_t count,
2622     struct thread *td)
2623 {
2624 	struct iovec aiov;
2625 	struct uio auio;
2626 	int error;
2627 
2628 	ASSERT_VOP_LOCKED(vp, "kern_readlink_vp(): vp not locked");
2629 #ifdef MAC
2630 	error = mac_vnode_check_readlink(td->td_ucred, vp);
2631 	if (error != 0)
2632 		return (error);
2633 #endif
2634 	if (vp->v_type != VLNK && (vp->v_vflag & VV_READLINK) == 0)
2635 		return (EINVAL);
2636 
2637 	aiov.iov_base = buf;
2638 	aiov.iov_len = count;
2639 	auio.uio_iov = &aiov;
2640 	auio.uio_iovcnt = 1;
2641 	auio.uio_offset = 0;
2642 	auio.uio_rw = UIO_READ;
2643 	auio.uio_segflg = bufseg;
2644 	auio.uio_td = td;
2645 	auio.uio_resid = count;
2646 	error = VOP_READLINK(vp, &auio, td->td_ucred);
2647 	td->td_retval[0] = count - auio.uio_resid;
2648 	return (error);
2649 }
2650 
2651 /*
2652  * Common implementation code for chflags() and fchflags().
2653  */
2654 static int
2655 setfflags(struct thread *td, struct vnode *vp, u_long flags)
2656 {
2657 	struct mount *mp;
2658 	struct vattr vattr;
2659 	int error;
2660 
2661 	/* We can't support the value matching VNOVAL. */
2662 	if (flags == VNOVAL)
2663 		return (EOPNOTSUPP);
2664 
2665 	/*
2666 	 * Prevent non-root users from setting flags on devices.  When
2667 	 * a device is reused, users can retain ownership of the device
2668 	 * if they are allowed to set flags and programs assume that
2669 	 * chown can't fail when done as root.
2670 	 */
2671 	if (vp->v_type == VCHR || vp->v_type == VBLK) {
2672 		error = priv_check(td, PRIV_VFS_CHFLAGS_DEV);
2673 		if (error != 0)
2674 			return (error);
2675 	}
2676 
2677 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2678 		return (error);
2679 	VATTR_NULL(&vattr);
2680 	vattr.va_flags = flags;
2681 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2682 #ifdef MAC
2683 	error = mac_vnode_check_setflags(td->td_ucred, vp, vattr.va_flags);
2684 	if (error == 0)
2685 #endif
2686 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
2687 	VOP_UNLOCK(vp);
2688 	vn_finished_write(mp);
2689 	return (error);
2690 }
2691 
2692 /*
2693  * Change flags of a file given a path name.
2694  */
2695 #ifndef _SYS_SYSPROTO_H_
2696 struct chflags_args {
2697 	const char *path;
2698 	u_long	flags;
2699 };
2700 #endif
2701 int
2702 sys_chflags(struct thread *td, struct chflags_args *uap)
2703 {
2704 
2705 	return (kern_chflagsat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2706 	    uap->flags, 0));
2707 }
2708 
2709 #ifndef _SYS_SYSPROTO_H_
2710 struct chflagsat_args {
2711 	int	fd;
2712 	const char *path;
2713 	u_long	flags;
2714 	int	atflag;
2715 }
2716 #endif
2717 int
2718 sys_chflagsat(struct thread *td, struct chflagsat_args *uap)
2719 {
2720 
2721 	if ((uap->atflag & ~(AT_SYMLINK_NOFOLLOW | AT_BENEATH |
2722 	    AT_RESOLVE_BENEATH)) != 0)
2723 		return (EINVAL);
2724 
2725 	return (kern_chflagsat(td, uap->fd, uap->path, UIO_USERSPACE,
2726 	    uap->flags, uap->atflag));
2727 }
2728 
2729 /*
2730  * Same as chflags() but doesn't follow symlinks.
2731  */
2732 #ifndef _SYS_SYSPROTO_H_
2733 struct lchflags_args {
2734 	const char *path;
2735 	u_long flags;
2736 };
2737 #endif
2738 int
2739 sys_lchflags(struct thread *td, struct lchflags_args *uap)
2740 {
2741 
2742 	return (kern_chflagsat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2743 	    uap->flags, AT_SYMLINK_NOFOLLOW));
2744 }
2745 
2746 static int
2747 kern_chflagsat(struct thread *td, int fd, const char *path,
2748     enum uio_seg pathseg, u_long flags, int atflag)
2749 {
2750 	struct nameidata nd;
2751 	int error;
2752 
2753 	AUDIT_ARG_FFLAGS(flags);
2754 	NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(atflag, AT_SYMLINK_NOFOLLOW |
2755 	    AT_BENEATH | AT_RESOLVE_BENEATH) | AUDITVNODE1, pathseg, path, fd,
2756 	    &cap_fchflags_rights, td);
2757 	if ((error = namei(&nd)) != 0)
2758 		return (error);
2759 	NDFREE_NOTHING(&nd);
2760 	error = setfflags(td, nd.ni_vp, flags);
2761 	vrele(nd.ni_vp);
2762 	return (error);
2763 }
2764 
2765 /*
2766  * Change flags of a file given a file descriptor.
2767  */
2768 #ifndef _SYS_SYSPROTO_H_
2769 struct fchflags_args {
2770 	int	fd;
2771 	u_long	flags;
2772 };
2773 #endif
2774 int
2775 sys_fchflags(struct thread *td, struct fchflags_args *uap)
2776 {
2777 	struct file *fp;
2778 	int error;
2779 
2780 	AUDIT_ARG_FD(uap->fd);
2781 	AUDIT_ARG_FFLAGS(uap->flags);
2782 	error = getvnode(td, uap->fd, &cap_fchflags_rights,
2783 	    &fp);
2784 	if (error != 0)
2785 		return (error);
2786 #ifdef AUDIT
2787 	vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
2788 	AUDIT_ARG_VNODE1(fp->f_vnode);
2789 	VOP_UNLOCK(fp->f_vnode);
2790 #endif
2791 	error = setfflags(td, fp->f_vnode, uap->flags);
2792 	fdrop(fp, td);
2793 	return (error);
2794 }
2795 
2796 /*
2797  * Common implementation code for chmod(), lchmod() and fchmod().
2798  */
2799 int
2800 setfmode(struct thread *td, struct ucred *cred, struct vnode *vp, int mode)
2801 {
2802 	struct mount *mp;
2803 	struct vattr vattr;
2804 	int error;
2805 
2806 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2807 		return (error);
2808 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2809 	VATTR_NULL(&vattr);
2810 	vattr.va_mode = mode & ALLPERMS;
2811 #ifdef MAC
2812 	error = mac_vnode_check_setmode(cred, vp, vattr.va_mode);
2813 	if (error == 0)
2814 #endif
2815 		error = VOP_SETATTR(vp, &vattr, cred);
2816 	VOP_UNLOCK(vp);
2817 	vn_finished_write(mp);
2818 	return (error);
2819 }
2820 
2821 /*
2822  * Change mode of a file given path name.
2823  */
2824 #ifndef _SYS_SYSPROTO_H_
2825 struct chmod_args {
2826 	char	*path;
2827 	int	mode;
2828 };
2829 #endif
2830 int
2831 sys_chmod(struct thread *td, struct chmod_args *uap)
2832 {
2833 
2834 	return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2835 	    uap->mode, 0));
2836 }
2837 
2838 #ifndef _SYS_SYSPROTO_H_
2839 struct fchmodat_args {
2840 	int	dirfd;
2841 	char	*path;
2842 	mode_t	mode;
2843 	int	flag;
2844 }
2845 #endif
2846 int
2847 sys_fchmodat(struct thread *td, struct fchmodat_args *uap)
2848 {
2849 
2850 	if ((uap->flag & ~(AT_SYMLINK_NOFOLLOW | AT_BENEATH |
2851 	    AT_RESOLVE_BENEATH)) != 0)
2852 		return (EINVAL);
2853 
2854 	return (kern_fchmodat(td, uap->fd, uap->path, UIO_USERSPACE,
2855 	    uap->mode, uap->flag));
2856 }
2857 
2858 /*
2859  * Change mode of a file given path name (don't follow links.)
2860  */
2861 #ifndef _SYS_SYSPROTO_H_
2862 struct lchmod_args {
2863 	char	*path;
2864 	int	mode;
2865 };
2866 #endif
2867 int
2868 sys_lchmod(struct thread *td, struct lchmod_args *uap)
2869 {
2870 
2871 	return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2872 	    uap->mode, AT_SYMLINK_NOFOLLOW));
2873 }
2874 
2875 int
2876 kern_fchmodat(struct thread *td, int fd, const char *path,
2877     enum uio_seg pathseg, mode_t mode, int flag)
2878 {
2879 	struct nameidata nd;
2880 	int error;
2881 
2882 	AUDIT_ARG_MODE(mode);
2883 	NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(flag, AT_SYMLINK_NOFOLLOW |
2884 	    AT_BENEATH | AT_RESOLVE_BENEATH) | AUDITVNODE1, pathseg, path, fd,
2885 	    &cap_fchmod_rights, td);
2886 	if ((error = namei(&nd)) != 0)
2887 		return (error);
2888 	NDFREE_NOTHING(&nd);
2889 	error = setfmode(td, td->td_ucred, nd.ni_vp, mode);
2890 	vrele(nd.ni_vp);
2891 	return (error);
2892 }
2893 
2894 /*
2895  * Change mode of a file given a file descriptor.
2896  */
2897 #ifndef _SYS_SYSPROTO_H_
2898 struct fchmod_args {
2899 	int	fd;
2900 	int	mode;
2901 };
2902 #endif
2903 int
2904 sys_fchmod(struct thread *td, struct fchmod_args *uap)
2905 {
2906 	struct file *fp;
2907 	int error;
2908 
2909 	AUDIT_ARG_FD(uap->fd);
2910 	AUDIT_ARG_MODE(uap->mode);
2911 
2912 	error = fget(td, uap->fd, &cap_fchmod_rights, &fp);
2913 	if (error != 0)
2914 		return (error);
2915 	error = fo_chmod(fp, uap->mode, td->td_ucred, td);
2916 	fdrop(fp, td);
2917 	return (error);
2918 }
2919 
2920 /*
2921  * Common implementation for chown(), lchown(), and fchown()
2922  */
2923 int
2924 setfown(struct thread *td, struct ucred *cred, struct vnode *vp, uid_t uid,
2925     gid_t gid)
2926 {
2927 	struct mount *mp;
2928 	struct vattr vattr;
2929 	int error;
2930 
2931 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2932 		return (error);
2933 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2934 	VATTR_NULL(&vattr);
2935 	vattr.va_uid = uid;
2936 	vattr.va_gid = gid;
2937 #ifdef MAC
2938 	error = mac_vnode_check_setowner(cred, vp, vattr.va_uid,
2939 	    vattr.va_gid);
2940 	if (error == 0)
2941 #endif
2942 		error = VOP_SETATTR(vp, &vattr, cred);
2943 	VOP_UNLOCK(vp);
2944 	vn_finished_write(mp);
2945 	return (error);
2946 }
2947 
2948 /*
2949  * Set ownership given a path name.
2950  */
2951 #ifndef _SYS_SYSPROTO_H_
2952 struct chown_args {
2953 	char	*path;
2954 	int	uid;
2955 	int	gid;
2956 };
2957 #endif
2958 int
2959 sys_chown(struct thread *td, struct chown_args *uap)
2960 {
2961 
2962 	return (kern_fchownat(td, AT_FDCWD, uap->path, UIO_USERSPACE, uap->uid,
2963 	    uap->gid, 0));
2964 }
2965 
2966 #ifndef _SYS_SYSPROTO_H_
2967 struct fchownat_args {
2968 	int fd;
2969 	const char * path;
2970 	uid_t uid;
2971 	gid_t gid;
2972 	int flag;
2973 };
2974 #endif
2975 int
2976 sys_fchownat(struct thread *td, struct fchownat_args *uap)
2977 {
2978 
2979 	if ((uap->flag & ~(AT_SYMLINK_NOFOLLOW | AT_BENEATH |
2980 	    AT_RESOLVE_BENEATH)) != 0)
2981 		return (EINVAL);
2982 
2983 	return (kern_fchownat(td, uap->fd, uap->path, UIO_USERSPACE, uap->uid,
2984 	    uap->gid, uap->flag));
2985 }
2986 
2987 int
2988 kern_fchownat(struct thread *td, int fd, const char *path,
2989     enum uio_seg pathseg, int uid, int gid, int flag)
2990 {
2991 	struct nameidata nd;
2992 	int error;
2993 
2994 	AUDIT_ARG_OWNER(uid, gid);
2995 	NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(flag, AT_SYMLINK_NOFOLLOW |
2996 	    AT_BENEATH | AT_RESOLVE_BENEATH) | AUDITVNODE1, pathseg, path, fd,
2997 	    &cap_fchown_rights, td);
2998 
2999 	if ((error = namei(&nd)) != 0)
3000 		return (error);
3001 	NDFREE_NOTHING(&nd);
3002 	error = setfown(td, td->td_ucred, nd.ni_vp, uid, gid);
3003 	vrele(nd.ni_vp);
3004 	return (error);
3005 }
3006 
3007 /*
3008  * Set ownership given a path name, do not cross symlinks.
3009  */
3010 #ifndef _SYS_SYSPROTO_H_
3011 struct lchown_args {
3012 	char	*path;
3013 	int	uid;
3014 	int	gid;
3015 };
3016 #endif
3017 int
3018 sys_lchown(struct thread *td, struct lchown_args *uap)
3019 {
3020 
3021 	return (kern_fchownat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
3022 	    uap->uid, uap->gid, AT_SYMLINK_NOFOLLOW));
3023 }
3024 
3025 /*
3026  * Set ownership given a file descriptor.
3027  */
3028 #ifndef _SYS_SYSPROTO_H_
3029 struct fchown_args {
3030 	int	fd;
3031 	int	uid;
3032 	int	gid;
3033 };
3034 #endif
3035 int
3036 sys_fchown(struct thread *td, struct fchown_args *uap)
3037 {
3038 	struct file *fp;
3039 	int error;
3040 
3041 	AUDIT_ARG_FD(uap->fd);
3042 	AUDIT_ARG_OWNER(uap->uid, uap->gid);
3043 	error = fget(td, uap->fd, &cap_fchown_rights, &fp);
3044 	if (error != 0)
3045 		return (error);
3046 	error = fo_chown(fp, uap->uid, uap->gid, td->td_ucred, td);
3047 	fdrop(fp, td);
3048 	return (error);
3049 }
3050 
3051 /*
3052  * Common implementation code for utimes(), lutimes(), and futimes().
3053  */
3054 static int
3055 getutimes(const struct timeval *usrtvp, enum uio_seg tvpseg,
3056     struct timespec *tsp)
3057 {
3058 	struct timeval tv[2];
3059 	const struct timeval *tvp;
3060 	int error;
3061 
3062 	if (usrtvp == NULL) {
3063 		vfs_timestamp(&tsp[0]);
3064 		tsp[1] = tsp[0];
3065 	} else {
3066 		if (tvpseg == UIO_SYSSPACE) {
3067 			tvp = usrtvp;
3068 		} else {
3069 			if ((error = copyin(usrtvp, tv, sizeof(tv))) != 0)
3070 				return (error);
3071 			tvp = tv;
3072 		}
3073 
3074 		if (tvp[0].tv_usec < 0 || tvp[0].tv_usec >= 1000000 ||
3075 		    tvp[1].tv_usec < 0 || tvp[1].tv_usec >= 1000000)
3076 			return (EINVAL);
3077 		TIMEVAL_TO_TIMESPEC(&tvp[0], &tsp[0]);
3078 		TIMEVAL_TO_TIMESPEC(&tvp[1], &tsp[1]);
3079 	}
3080 	return (0);
3081 }
3082 
3083 /*
3084  * Common implementation code for futimens(), utimensat().
3085  */
3086 #define	UTIMENS_NULL	0x1
3087 #define	UTIMENS_EXIT	0x2
3088 static int
3089 getutimens(const struct timespec *usrtsp, enum uio_seg tspseg,
3090     struct timespec *tsp, int *retflags)
3091 {
3092 	struct timespec tsnow;
3093 	int error;
3094 
3095 	vfs_timestamp(&tsnow);
3096 	*retflags = 0;
3097 	if (usrtsp == NULL) {
3098 		tsp[0] = tsnow;
3099 		tsp[1] = tsnow;
3100 		*retflags |= UTIMENS_NULL;
3101 		return (0);
3102 	}
3103 	if (tspseg == UIO_SYSSPACE) {
3104 		tsp[0] = usrtsp[0];
3105 		tsp[1] = usrtsp[1];
3106 	} else if ((error = copyin(usrtsp, tsp, sizeof(*tsp) * 2)) != 0)
3107 		return (error);
3108 	if (tsp[0].tv_nsec == UTIME_OMIT && tsp[1].tv_nsec == UTIME_OMIT)
3109 		*retflags |= UTIMENS_EXIT;
3110 	if (tsp[0].tv_nsec == UTIME_NOW && tsp[1].tv_nsec == UTIME_NOW)
3111 		*retflags |= UTIMENS_NULL;
3112 	if (tsp[0].tv_nsec == UTIME_OMIT)
3113 		tsp[0].tv_sec = VNOVAL;
3114 	else if (tsp[0].tv_nsec == UTIME_NOW)
3115 		tsp[0] = tsnow;
3116 	else if (tsp[0].tv_nsec < 0 || tsp[0].tv_nsec >= 1000000000L)
3117 		return (EINVAL);
3118 	if (tsp[1].tv_nsec == UTIME_OMIT)
3119 		tsp[1].tv_sec = VNOVAL;
3120 	else if (tsp[1].tv_nsec == UTIME_NOW)
3121 		tsp[1] = tsnow;
3122 	else if (tsp[1].tv_nsec < 0 || tsp[1].tv_nsec >= 1000000000L)
3123 		return (EINVAL);
3124 
3125 	return (0);
3126 }
3127 
3128 /*
3129  * Common implementation code for utimes(), lutimes(), futimes(), futimens(),
3130  * and utimensat().
3131  */
3132 static int
3133 setutimes(struct thread *td, struct vnode *vp, const struct timespec *ts,
3134     int numtimes, int nullflag)
3135 {
3136 	struct mount *mp;
3137 	struct vattr vattr;
3138 	int error, setbirthtime;
3139 
3140 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
3141 		return (error);
3142 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3143 	setbirthtime = 0;
3144 	if (numtimes < 3 && !VOP_GETATTR(vp, &vattr, td->td_ucred) &&
3145 	    timespeccmp(&ts[1], &vattr.va_birthtime, < ))
3146 		setbirthtime = 1;
3147 	VATTR_NULL(&vattr);
3148 	vattr.va_atime = ts[0];
3149 	vattr.va_mtime = ts[1];
3150 	if (setbirthtime)
3151 		vattr.va_birthtime = ts[1];
3152 	if (numtimes > 2)
3153 		vattr.va_birthtime = ts[2];
3154 	if (nullflag)
3155 		vattr.va_vaflags |= VA_UTIMES_NULL;
3156 #ifdef MAC
3157 	error = mac_vnode_check_setutimes(td->td_ucred, vp, vattr.va_atime,
3158 	    vattr.va_mtime);
3159 #endif
3160 	if (error == 0)
3161 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3162 	VOP_UNLOCK(vp);
3163 	vn_finished_write(mp);
3164 	return (error);
3165 }
3166 
3167 /*
3168  * Set the access and modification times of a file.
3169  */
3170 #ifndef _SYS_SYSPROTO_H_
3171 struct utimes_args {
3172 	char	*path;
3173 	struct	timeval *tptr;
3174 };
3175 #endif
3176 int
3177 sys_utimes(struct thread *td, struct utimes_args *uap)
3178 {
3179 
3180 	return (kern_utimesat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
3181 	    uap->tptr, UIO_USERSPACE));
3182 }
3183 
3184 #ifndef _SYS_SYSPROTO_H_
3185 struct futimesat_args {
3186 	int fd;
3187 	const char * path;
3188 	const struct timeval * times;
3189 };
3190 #endif
3191 int
3192 sys_futimesat(struct thread *td, struct futimesat_args *uap)
3193 {
3194 
3195 	return (kern_utimesat(td, uap->fd, uap->path, UIO_USERSPACE,
3196 	    uap->times, UIO_USERSPACE));
3197 }
3198 
3199 int
3200 kern_utimesat(struct thread *td, int fd, const char *path,
3201     enum uio_seg pathseg, struct timeval *tptr, enum uio_seg tptrseg)
3202 {
3203 	struct nameidata nd;
3204 	struct timespec ts[2];
3205 	int error;
3206 
3207 	if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3208 		return (error);
3209 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, fd,
3210 	    &cap_futimes_rights, td);
3211 
3212 	if ((error = namei(&nd)) != 0)
3213 		return (error);
3214 	NDFREE_NOTHING(&nd);
3215 	error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3216 	vrele(nd.ni_vp);
3217 	return (error);
3218 }
3219 
3220 /*
3221  * Set the access and modification times of a file.
3222  */
3223 #ifndef _SYS_SYSPROTO_H_
3224 struct lutimes_args {
3225 	char	*path;
3226 	struct	timeval *tptr;
3227 };
3228 #endif
3229 int
3230 sys_lutimes(struct thread *td, struct lutimes_args *uap)
3231 {
3232 
3233 	return (kern_lutimes(td, uap->path, UIO_USERSPACE, uap->tptr,
3234 	    UIO_USERSPACE));
3235 }
3236 
3237 int
3238 kern_lutimes(struct thread *td, const char *path, enum uio_seg pathseg,
3239     struct timeval *tptr, enum uio_seg tptrseg)
3240 {
3241 	struct timespec ts[2];
3242 	struct nameidata nd;
3243 	int error;
3244 
3245 	if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3246 		return (error);
3247 	NDINIT(&nd, LOOKUP, NOFOLLOW | AUDITVNODE1, pathseg, path, td);
3248 	if ((error = namei(&nd)) != 0)
3249 		return (error);
3250 	NDFREE_NOTHING(&nd);
3251 	error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3252 	vrele(nd.ni_vp);
3253 	return (error);
3254 }
3255 
3256 /*
3257  * Set the access and modification times of a file.
3258  */
3259 #ifndef _SYS_SYSPROTO_H_
3260 struct futimes_args {
3261 	int	fd;
3262 	struct	timeval *tptr;
3263 };
3264 #endif
3265 int
3266 sys_futimes(struct thread *td, struct futimes_args *uap)
3267 {
3268 
3269 	return (kern_futimes(td, uap->fd, uap->tptr, UIO_USERSPACE));
3270 }
3271 
3272 int
3273 kern_futimes(struct thread *td, int fd, struct timeval *tptr,
3274     enum uio_seg tptrseg)
3275 {
3276 	struct timespec ts[2];
3277 	struct file *fp;
3278 	int error;
3279 
3280 	AUDIT_ARG_FD(fd);
3281 	error = getutimes(tptr, tptrseg, ts);
3282 	if (error != 0)
3283 		return (error);
3284 	error = getvnode(td, fd, &cap_futimes_rights, &fp);
3285 	if (error != 0)
3286 		return (error);
3287 #ifdef AUDIT
3288 	vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
3289 	AUDIT_ARG_VNODE1(fp->f_vnode);
3290 	VOP_UNLOCK(fp->f_vnode);
3291 #endif
3292 	error = setutimes(td, fp->f_vnode, ts, 2, tptr == NULL);
3293 	fdrop(fp, td);
3294 	return (error);
3295 }
3296 
3297 int
3298 sys_futimens(struct thread *td, struct futimens_args *uap)
3299 {
3300 
3301 	return (kern_futimens(td, uap->fd, uap->times, UIO_USERSPACE));
3302 }
3303 
3304 int
3305 kern_futimens(struct thread *td, int fd, struct timespec *tptr,
3306     enum uio_seg tptrseg)
3307 {
3308 	struct timespec ts[2];
3309 	struct file *fp;
3310 	int error, flags;
3311 
3312 	AUDIT_ARG_FD(fd);
3313 	error = getutimens(tptr, tptrseg, ts, &flags);
3314 	if (error != 0)
3315 		return (error);
3316 	if (flags & UTIMENS_EXIT)
3317 		return (0);
3318 	error = getvnode(td, fd, &cap_futimes_rights, &fp);
3319 	if (error != 0)
3320 		return (error);
3321 #ifdef AUDIT
3322 	vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
3323 	AUDIT_ARG_VNODE1(fp->f_vnode);
3324 	VOP_UNLOCK(fp->f_vnode);
3325 #endif
3326 	error = setutimes(td, fp->f_vnode, ts, 2, flags & UTIMENS_NULL);
3327 	fdrop(fp, td);
3328 	return (error);
3329 }
3330 
3331 int
3332 sys_utimensat(struct thread *td, struct utimensat_args *uap)
3333 {
3334 
3335 	return (kern_utimensat(td, uap->fd, uap->path, UIO_USERSPACE,
3336 	    uap->times, UIO_USERSPACE, uap->flag));
3337 }
3338 
3339 int
3340 kern_utimensat(struct thread *td, int fd, const char *path,
3341     enum uio_seg pathseg, struct timespec *tptr, enum uio_seg tptrseg,
3342     int flag)
3343 {
3344 	struct nameidata nd;
3345 	struct timespec ts[2];
3346 	int error, flags;
3347 
3348 	if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_BENEATH |
3349 	    AT_RESOLVE_BENEATH)) != 0)
3350 		return (EINVAL);
3351 
3352 	if ((error = getutimens(tptr, tptrseg, ts, &flags)) != 0)
3353 		return (error);
3354 	NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(flag, AT_SYMLINK_NOFOLLOW |
3355 	    AT_BENEATH | AT_RESOLVE_BENEATH) | AUDITVNODE1,
3356 	    pathseg, path, fd, &cap_futimes_rights, td);
3357 	if ((error = namei(&nd)) != 0)
3358 		return (error);
3359 	/*
3360 	 * We are allowed to call namei() regardless of 2xUTIME_OMIT.
3361 	 * POSIX states:
3362 	 * "If both tv_nsec fields are UTIME_OMIT... EACCESS may be detected."
3363 	 * "Search permission is denied by a component of the path prefix."
3364 	 */
3365 	NDFREE_NOTHING(&nd);
3366 	if ((flags & UTIMENS_EXIT) == 0)
3367 		error = setutimes(td, nd.ni_vp, ts, 2, flags & UTIMENS_NULL);
3368 	vrele(nd.ni_vp);
3369 	return (error);
3370 }
3371 
3372 /*
3373  * Truncate a file given its path name.
3374  */
3375 #ifndef _SYS_SYSPROTO_H_
3376 struct truncate_args {
3377 	char	*path;
3378 	int	pad;
3379 	off_t	length;
3380 };
3381 #endif
3382 int
3383 sys_truncate(struct thread *td, struct truncate_args *uap)
3384 {
3385 
3386 	return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3387 }
3388 
3389 int
3390 kern_truncate(struct thread *td, const char *path, enum uio_seg pathseg,
3391     off_t length)
3392 {
3393 	struct mount *mp;
3394 	struct vnode *vp;
3395 	void *rl_cookie;
3396 	struct vattr vattr;
3397 	struct nameidata nd;
3398 	int error;
3399 
3400 	if (length < 0)
3401 		return (EINVAL);
3402 retry:
3403 	NDINIT(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, td);
3404 	if ((error = namei(&nd)) != 0)
3405 		return (error);
3406 	vp = nd.ni_vp;
3407 	rl_cookie = vn_rangelock_wlock(vp, 0, OFF_MAX);
3408 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) {
3409 		vn_rangelock_unlock(vp, rl_cookie);
3410 		vrele(vp);
3411 		return (error);
3412 	}
3413 	NDFREE(&nd, NDF_ONLY_PNBUF);
3414 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3415 	if (vp->v_type == VDIR)
3416 		error = EISDIR;
3417 #ifdef MAC
3418 	else if ((error = mac_vnode_check_write(td->td_ucred, NOCRED, vp))) {
3419 	}
3420 #endif
3421 	else if ((error = vn_writechk(vp)) == 0 &&
3422 	    (error = VOP_ACCESS(vp, VWRITE, td->td_ucred, td)) == 0) {
3423 		VATTR_NULL(&vattr);
3424 		vattr.va_size = length;
3425 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3426 	}
3427 	VOP_UNLOCK(vp);
3428 	vn_finished_write(mp);
3429 	vn_rangelock_unlock(vp, rl_cookie);
3430 	vrele(vp);
3431 	if (error == ERELOOKUP)
3432 		goto retry;
3433 	return (error);
3434 }
3435 
3436 #if defined(COMPAT_43)
3437 /*
3438  * Truncate a file given its path name.
3439  */
3440 #ifndef _SYS_SYSPROTO_H_
3441 struct otruncate_args {
3442 	char	*path;
3443 	long	length;
3444 };
3445 #endif
3446 int
3447 otruncate(struct thread *td, struct otruncate_args *uap)
3448 {
3449 
3450 	return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3451 }
3452 #endif /* COMPAT_43 */
3453 
3454 #if defined(COMPAT_FREEBSD6)
3455 /* Versions with the pad argument */
3456 int
3457 freebsd6_truncate(struct thread *td, struct freebsd6_truncate_args *uap)
3458 {
3459 
3460 	return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3461 }
3462 
3463 int
3464 freebsd6_ftruncate(struct thread *td, struct freebsd6_ftruncate_args *uap)
3465 {
3466 
3467 	return (kern_ftruncate(td, uap->fd, uap->length));
3468 }
3469 #endif
3470 
3471 int
3472 kern_fsync(struct thread *td, int fd, bool fullsync)
3473 {
3474 	struct vnode *vp;
3475 	struct mount *mp;
3476 	struct file *fp;
3477 	int error, lock_flags;
3478 
3479 	AUDIT_ARG_FD(fd);
3480 	error = getvnode(td, fd, &cap_fsync_rights, &fp);
3481 	if (error != 0)
3482 		return (error);
3483 	vp = fp->f_vnode;
3484 #if 0
3485 	if (!fullsync)
3486 		/* XXXKIB: compete outstanding aio writes */;
3487 #endif
3488 retry:
3489 	error = vn_start_write(vp, &mp, V_WAIT | PCATCH);
3490 	if (error != 0)
3491 		goto drop;
3492 	if (MNT_SHARED_WRITES(mp) ||
3493 	    ((mp == NULL) && MNT_SHARED_WRITES(vp->v_mount))) {
3494 		lock_flags = LK_SHARED;
3495 	} else {
3496 		lock_flags = LK_EXCLUSIVE;
3497 	}
3498 	vn_lock(vp, lock_flags | LK_RETRY);
3499 	AUDIT_ARG_VNODE1(vp);
3500 	if (vp->v_object != NULL) {
3501 		VM_OBJECT_WLOCK(vp->v_object);
3502 		vm_object_page_clean(vp->v_object, 0, 0, 0);
3503 		VM_OBJECT_WUNLOCK(vp->v_object);
3504 	}
3505 	error = fullsync ? VOP_FSYNC(vp, MNT_WAIT, td) : VOP_FDATASYNC(vp, td);
3506 	VOP_UNLOCK(vp);
3507 	vn_finished_write(mp);
3508 	if (error == ERELOOKUP)
3509 		goto retry;
3510 drop:
3511 	fdrop(fp, td);
3512 	return (error);
3513 }
3514 
3515 /*
3516  * Sync an open file.
3517  */
3518 #ifndef _SYS_SYSPROTO_H_
3519 struct fsync_args {
3520 	int	fd;
3521 };
3522 #endif
3523 int
3524 sys_fsync(struct thread *td, struct fsync_args *uap)
3525 {
3526 
3527 	return (kern_fsync(td, uap->fd, true));
3528 }
3529 
3530 int
3531 sys_fdatasync(struct thread *td, struct fdatasync_args *uap)
3532 {
3533 
3534 	return (kern_fsync(td, uap->fd, false));
3535 }
3536 
3537 /*
3538  * Rename files.  Source and destination must either both be directories, or
3539  * both not be directories.  If target is a directory, it must be empty.
3540  */
3541 #ifndef _SYS_SYSPROTO_H_
3542 struct rename_args {
3543 	char	*from;
3544 	char	*to;
3545 };
3546 #endif
3547 int
3548 sys_rename(struct thread *td, struct rename_args *uap)
3549 {
3550 
3551 	return (kern_renameat(td, AT_FDCWD, uap->from, AT_FDCWD,
3552 	    uap->to, UIO_USERSPACE));
3553 }
3554 
3555 #ifndef _SYS_SYSPROTO_H_
3556 struct renameat_args {
3557 	int	oldfd;
3558 	char	*old;
3559 	int	newfd;
3560 	char	*new;
3561 };
3562 #endif
3563 int
3564 sys_renameat(struct thread *td, struct renameat_args *uap)
3565 {
3566 
3567 	return (kern_renameat(td, uap->oldfd, uap->old, uap->newfd, uap->new,
3568 	    UIO_USERSPACE));
3569 }
3570 
3571 #ifdef MAC
3572 static int
3573 kern_renameat_mac(struct thread *td, int oldfd, const char *old, int newfd,
3574     const char *new, enum uio_seg pathseg, struct nameidata *fromnd)
3575 {
3576 	int error;
3577 
3578 	NDINIT_ATRIGHTS(fromnd, DELETE, LOCKPARENT | LOCKLEAF | SAVESTART |
3579 	    AUDITVNODE1, pathseg, old, oldfd, &cap_renameat_source_rights, td);
3580 	if ((error = namei(fromnd)) != 0)
3581 		return (error);
3582 	error = mac_vnode_check_rename_from(td->td_ucred, fromnd->ni_dvp,
3583 	    fromnd->ni_vp, &fromnd->ni_cnd);
3584 	VOP_UNLOCK(fromnd->ni_dvp);
3585 	if (fromnd->ni_dvp != fromnd->ni_vp)
3586 		VOP_UNLOCK(fromnd->ni_vp);
3587 	if (error != 0) {
3588 		NDFREE(fromnd, NDF_ONLY_PNBUF);
3589 		vrele(fromnd->ni_dvp);
3590 		vrele(fromnd->ni_vp);
3591 		if (fromnd->ni_startdir)
3592 			vrele(fromnd->ni_startdir);
3593 	}
3594 	return (error);
3595 }
3596 #endif
3597 
3598 int
3599 kern_renameat(struct thread *td, int oldfd, const char *old, int newfd,
3600     const char *new, enum uio_seg pathseg)
3601 {
3602 	struct mount *mp = NULL;
3603 	struct vnode *tvp, *fvp, *tdvp;
3604 	struct nameidata fromnd, tond;
3605 	u_int64_t tondflags;
3606 	int error;
3607 
3608 again:
3609 	bwillwrite();
3610 #ifdef MAC
3611 	if (mac_vnode_check_rename_from_enabled()) {
3612 		error = kern_renameat_mac(td, oldfd, old, newfd, new, pathseg,
3613 		    &fromnd);
3614 		if (error != 0)
3615 			return (error);
3616 	} else {
3617 #endif
3618 	NDINIT_ATRIGHTS(&fromnd, DELETE, WANTPARENT | SAVESTART | AUDITVNODE1,
3619 	    pathseg, old, oldfd, &cap_renameat_source_rights, td);
3620 	if ((error = namei(&fromnd)) != 0)
3621 		return (error);
3622 #ifdef MAC
3623 	}
3624 #endif
3625 	fvp = fromnd.ni_vp;
3626 	tondflags = LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART | AUDITVNODE2;
3627 	if (fromnd.ni_vp->v_type == VDIR)
3628 		tondflags |= WILLBEDIR;
3629 	NDINIT_ATRIGHTS(&tond, RENAME, tondflags, pathseg, new, newfd,
3630 	    &cap_renameat_target_rights, td);
3631 	if ((error = namei(&tond)) != 0) {
3632 		/* Translate error code for rename("dir1", "dir2/."). */
3633 		if (error == EISDIR && fvp->v_type == VDIR)
3634 			error = EINVAL;
3635 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3636 		vrele(fromnd.ni_dvp);
3637 		vrele(fvp);
3638 		goto out1;
3639 	}
3640 	tdvp = tond.ni_dvp;
3641 	tvp = tond.ni_vp;
3642 	error = vn_start_write(fvp, &mp, V_NOWAIT);
3643 	if (error != 0) {
3644 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3645 		NDFREE(&tond, NDF_ONLY_PNBUF);
3646 		if (tvp != NULL)
3647 			vput(tvp);
3648 		if (tdvp == tvp)
3649 			vrele(tdvp);
3650 		else
3651 			vput(tdvp);
3652 		vrele(fromnd.ni_dvp);
3653 		vrele(fvp);
3654 		vrele(tond.ni_startdir);
3655 		if (fromnd.ni_startdir != NULL)
3656 			vrele(fromnd.ni_startdir);
3657 		error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH);
3658 		if (error != 0)
3659 			return (error);
3660 		goto again;
3661 	}
3662 	if (tvp != NULL) {
3663 		if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
3664 			error = ENOTDIR;
3665 			goto out;
3666 		} else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
3667 			error = EISDIR;
3668 			goto out;
3669 		}
3670 #ifdef CAPABILITIES
3671 		if (newfd != AT_FDCWD && (tond.ni_resflags & NIRES_ABS) == 0) {
3672 			/*
3673 			 * If the target already exists we require CAP_UNLINKAT
3674 			 * from 'newfd', when newfd was used for the lookup.
3675 			 */
3676 			error = cap_check(&tond.ni_filecaps.fc_rights,
3677 			    &cap_unlinkat_rights);
3678 			if (error != 0)
3679 				goto out;
3680 		}
3681 #endif
3682 	}
3683 	if (fvp == tdvp) {
3684 		error = EINVAL;
3685 		goto out;
3686 	}
3687 	/*
3688 	 * If the source is the same as the destination (that is, if they
3689 	 * are links to the same vnode), then there is nothing to do.
3690 	 */
3691 	if (fvp == tvp)
3692 		error = ERESTART;
3693 #ifdef MAC
3694 	else
3695 		error = mac_vnode_check_rename_to(td->td_ucred, tdvp,
3696 		    tond.ni_vp, fromnd.ni_dvp == tdvp, &tond.ni_cnd);
3697 #endif
3698 out:
3699 	if (error == 0) {
3700 		error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
3701 		    tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
3702 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3703 		NDFREE(&tond, NDF_ONLY_PNBUF);
3704 	} else {
3705 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3706 		NDFREE(&tond, NDF_ONLY_PNBUF);
3707 		if (tvp != NULL)
3708 			vput(tvp);
3709 		if (tdvp == tvp)
3710 			vrele(tdvp);
3711 		else
3712 			vput(tdvp);
3713 		vrele(fromnd.ni_dvp);
3714 		vrele(fvp);
3715 	}
3716 	vrele(tond.ni_startdir);
3717 	vn_finished_write(mp);
3718 out1:
3719 	if (fromnd.ni_startdir)
3720 		vrele(fromnd.ni_startdir);
3721 	if (error == ERESTART)
3722 		return (0);
3723 	if (error == ERELOOKUP)
3724 		goto again;
3725 	return (error);
3726 }
3727 
3728 /*
3729  * Make a directory file.
3730  */
3731 #ifndef _SYS_SYSPROTO_H_
3732 struct mkdir_args {
3733 	char	*path;
3734 	int	mode;
3735 };
3736 #endif
3737 int
3738 sys_mkdir(struct thread *td, struct mkdir_args *uap)
3739 {
3740 
3741 	return (kern_mkdirat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
3742 	    uap->mode));
3743 }
3744 
3745 #ifndef _SYS_SYSPROTO_H_
3746 struct mkdirat_args {
3747 	int	fd;
3748 	char	*path;
3749 	mode_t	mode;
3750 };
3751 #endif
3752 int
3753 sys_mkdirat(struct thread *td, struct mkdirat_args *uap)
3754 {
3755 
3756 	return (kern_mkdirat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode));
3757 }
3758 
3759 int
3760 kern_mkdirat(struct thread *td, int fd, const char *path, enum uio_seg segflg,
3761     int mode)
3762 {
3763 	struct mount *mp;
3764 	struct vattr vattr;
3765 	struct nameidata nd;
3766 	int error;
3767 
3768 	AUDIT_ARG_MODE(mode);
3769 restart:
3770 	bwillwrite();
3771 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
3772 	    NC_NOMAKEENTRY | NC_KEEPPOSENTRY | FAILIFEXISTS | WILLBEDIR,
3773 	    segflg, path, fd, &cap_mkdirat_rights, td);
3774 	if ((error = namei(&nd)) != 0)
3775 		return (error);
3776 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3777 		NDFREE(&nd, NDF_ONLY_PNBUF);
3778 		vput(nd.ni_dvp);
3779 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3780 			return (error);
3781 		goto restart;
3782 	}
3783 	VATTR_NULL(&vattr);
3784 	vattr.va_type = VDIR;
3785 	vattr.va_mode = (mode & ACCESSPERMS) &~ td->td_proc->p_pd->pd_cmask;
3786 #ifdef MAC
3787 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
3788 	    &vattr);
3789 	if (error != 0)
3790 		goto out;
3791 #endif
3792 	error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
3793 #ifdef MAC
3794 out:
3795 #endif
3796 	NDFREE(&nd, NDF_ONLY_PNBUF);
3797 	vput(nd.ni_dvp);
3798 	if (error == 0)
3799 		vput(nd.ni_vp);
3800 	vn_finished_write(mp);
3801 	if (error == ERELOOKUP)
3802 		goto restart;
3803 	return (error);
3804 }
3805 
3806 /*
3807  * Remove a directory file.
3808  */
3809 #ifndef _SYS_SYSPROTO_H_
3810 struct rmdir_args {
3811 	char	*path;
3812 };
3813 #endif
3814 int
3815 sys_rmdir(struct thread *td, struct rmdir_args *uap)
3816 {
3817 
3818 	return (kern_frmdirat(td, AT_FDCWD, uap->path, FD_NONE, UIO_USERSPACE,
3819 	    0));
3820 }
3821 
3822 int
3823 kern_frmdirat(struct thread *td, int dfd, const char *path, int fd,
3824     enum uio_seg pathseg, int flag)
3825 {
3826 	struct mount *mp;
3827 	struct vnode *vp;
3828 	struct file *fp;
3829 	struct nameidata nd;
3830 	cap_rights_t rights;
3831 	int error;
3832 
3833 	fp = NULL;
3834 	if (fd != FD_NONE) {
3835 		error = getvnode(td, fd, cap_rights_init_one(&rights, CAP_LOOKUP),
3836 		    &fp);
3837 		if (error != 0)
3838 			return (error);
3839 	}
3840 
3841 restart:
3842 	bwillwrite();
3843 	NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1 |
3844 	    at2cnpflags(flag, AT_BENEATH | AT_RESOLVE_BENEATH),
3845 	    pathseg, path, dfd, &cap_unlinkat_rights, td);
3846 	if ((error = namei(&nd)) != 0)
3847 		goto fdout;
3848 	vp = nd.ni_vp;
3849 	if (vp->v_type != VDIR) {
3850 		error = ENOTDIR;
3851 		goto out;
3852 	}
3853 	/*
3854 	 * No rmdir "." please.
3855 	 */
3856 	if (nd.ni_dvp == vp) {
3857 		error = EINVAL;
3858 		goto out;
3859 	}
3860 	/*
3861 	 * The root of a mounted filesystem cannot be deleted.
3862 	 */
3863 	if (vp->v_vflag & VV_ROOT) {
3864 		error = EBUSY;
3865 		goto out;
3866 	}
3867 
3868 	if (fp != NULL && fp->f_vnode != vp) {
3869 		if (VN_IS_DOOMED(fp->f_vnode))
3870 			error = EBADF;
3871 		else
3872 			error = EDEADLK;
3873 		goto out;
3874 	}
3875 
3876 #ifdef MAC
3877 	error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
3878 	    &nd.ni_cnd);
3879 	if (error != 0)
3880 		goto out;
3881 #endif
3882 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3883 		NDFREE(&nd, NDF_ONLY_PNBUF);
3884 		vput(vp);
3885 		if (nd.ni_dvp == vp)
3886 			vrele(nd.ni_dvp);
3887 		else
3888 			vput(nd.ni_dvp);
3889 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3890 			goto fdout;
3891 		goto restart;
3892 	}
3893 	vfs_notify_upper(vp, VFS_NOTIFY_UPPER_UNLINK);
3894 	error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
3895 	vn_finished_write(mp);
3896 out:
3897 	NDFREE(&nd, NDF_ONLY_PNBUF);
3898 	vput(vp);
3899 	if (nd.ni_dvp == vp)
3900 		vrele(nd.ni_dvp);
3901 	else
3902 		vput(nd.ni_dvp);
3903 	if (error == ERELOOKUP)
3904 		goto restart;
3905 fdout:
3906 	if (fp != NULL)
3907 		fdrop(fp, td);
3908 	return (error);
3909 }
3910 
3911 #if defined(COMPAT_43) || defined(COMPAT_FREEBSD11)
3912 int
3913 freebsd11_kern_getdirentries(struct thread *td, int fd, char *ubuf, u_int count,
3914     long *basep, void (*func)(struct freebsd11_dirent *))
3915 {
3916 	struct freebsd11_dirent dstdp;
3917 	struct dirent *dp, *edp;
3918 	char *dirbuf;
3919 	off_t base;
3920 	ssize_t resid, ucount;
3921 	int error;
3922 
3923 	/* XXX arbitrary sanity limit on `count'. */
3924 	count = min(count, 64 * 1024);
3925 
3926 	dirbuf = malloc(count, M_TEMP, M_WAITOK);
3927 
3928 	error = kern_getdirentries(td, fd, dirbuf, count, &base, &resid,
3929 	    UIO_SYSSPACE);
3930 	if (error != 0)
3931 		goto done;
3932 	if (basep != NULL)
3933 		*basep = base;
3934 
3935 	ucount = 0;
3936 	for (dp = (struct dirent *)dirbuf,
3937 	    edp = (struct dirent *)&dirbuf[count - resid];
3938 	    ucount < count && dp < edp; ) {
3939 		if (dp->d_reclen == 0)
3940 			break;
3941 		MPASS(dp->d_reclen >= _GENERIC_DIRLEN(0));
3942 		if (dp->d_namlen >= sizeof(dstdp.d_name))
3943 			continue;
3944 		dstdp.d_type = dp->d_type;
3945 		dstdp.d_namlen = dp->d_namlen;
3946 		dstdp.d_fileno = dp->d_fileno;		/* truncate */
3947 		if (dstdp.d_fileno != dp->d_fileno) {
3948 			switch (ino64_trunc_error) {
3949 			default:
3950 			case 0:
3951 				break;
3952 			case 1:
3953 				error = EOVERFLOW;
3954 				goto done;
3955 			case 2:
3956 				dstdp.d_fileno = UINT32_MAX;
3957 				break;
3958 			}
3959 		}
3960 		dstdp.d_reclen = sizeof(dstdp) - sizeof(dstdp.d_name) +
3961 		    ((dp->d_namlen + 1 + 3) &~ 3);
3962 		bcopy(dp->d_name, dstdp.d_name, dstdp.d_namlen);
3963 		bzero(dstdp.d_name + dstdp.d_namlen,
3964 		    dstdp.d_reclen - offsetof(struct freebsd11_dirent, d_name) -
3965 		    dstdp.d_namlen);
3966 		MPASS(dstdp.d_reclen <= dp->d_reclen);
3967 		MPASS(ucount + dstdp.d_reclen <= count);
3968 		if (func != NULL)
3969 			func(&dstdp);
3970 		error = copyout(&dstdp, ubuf + ucount, dstdp.d_reclen);
3971 		if (error != 0)
3972 			break;
3973 		dp = (struct dirent *)((char *)dp + dp->d_reclen);
3974 		ucount += dstdp.d_reclen;
3975 	}
3976 
3977 done:
3978 	free(dirbuf, M_TEMP);
3979 	if (error == 0)
3980 		td->td_retval[0] = ucount;
3981 	return (error);
3982 }
3983 #endif /* COMPAT */
3984 
3985 #ifdef COMPAT_43
3986 static void
3987 ogetdirentries_cvt(struct freebsd11_dirent *dp)
3988 {
3989 #if (BYTE_ORDER == LITTLE_ENDIAN)
3990 	/*
3991 	 * The expected low byte of dp->d_namlen is our dp->d_type.
3992 	 * The high MBZ byte of dp->d_namlen is our dp->d_namlen.
3993 	 */
3994 	dp->d_type = dp->d_namlen;
3995 	dp->d_namlen = 0;
3996 #else
3997 	/*
3998 	 * The dp->d_type is the high byte of the expected dp->d_namlen,
3999 	 * so must be zero'ed.
4000 	 */
4001 	dp->d_type = 0;
4002 #endif
4003 }
4004 
4005 /*
4006  * Read a block of directory entries in a filesystem independent format.
4007  */
4008 #ifndef _SYS_SYSPROTO_H_
4009 struct ogetdirentries_args {
4010 	int	fd;
4011 	char	*buf;
4012 	u_int	count;
4013 	long	*basep;
4014 };
4015 #endif
4016 int
4017 ogetdirentries(struct thread *td, struct ogetdirentries_args *uap)
4018 {
4019 	long loff;
4020 	int error;
4021 
4022 	error = kern_ogetdirentries(td, uap, &loff);
4023 	if (error == 0)
4024 		error = copyout(&loff, uap->basep, sizeof(long));
4025 	return (error);
4026 }
4027 
4028 int
4029 kern_ogetdirentries(struct thread *td, struct ogetdirentries_args *uap,
4030     long *ploff)
4031 {
4032 	long base;
4033 	int error;
4034 
4035 	/* XXX arbitrary sanity limit on `count'. */
4036 	if (uap->count > 64 * 1024)
4037 		return (EINVAL);
4038 
4039 	error = freebsd11_kern_getdirentries(td, uap->fd, uap->buf, uap->count,
4040 	    &base, ogetdirentries_cvt);
4041 
4042 	if (error == 0 && uap->basep != NULL)
4043 		error = copyout(&base, uap->basep, sizeof(long));
4044 
4045 	return (error);
4046 }
4047 #endif /* COMPAT_43 */
4048 
4049 #if defined(COMPAT_FREEBSD11)
4050 #ifndef _SYS_SYSPROTO_H_
4051 struct freebsd11_getdirentries_args {
4052 	int	fd;
4053 	char	*buf;
4054 	u_int	count;
4055 	long	*basep;
4056 };
4057 #endif
4058 int
4059 freebsd11_getdirentries(struct thread *td,
4060     struct freebsd11_getdirentries_args *uap)
4061 {
4062 	long base;
4063 	int error;
4064 
4065 	error = freebsd11_kern_getdirentries(td, uap->fd, uap->buf, uap->count,
4066 	    &base, NULL);
4067 
4068 	if (error == 0 && uap->basep != NULL)
4069 		error = copyout(&base, uap->basep, sizeof(long));
4070 	return (error);
4071 }
4072 
4073 int
4074 freebsd11_getdents(struct thread *td, struct freebsd11_getdents_args *uap)
4075 {
4076 	struct freebsd11_getdirentries_args ap;
4077 
4078 	ap.fd = uap->fd;
4079 	ap.buf = uap->buf;
4080 	ap.count = uap->count;
4081 	ap.basep = NULL;
4082 	return (freebsd11_getdirentries(td, &ap));
4083 }
4084 #endif /* COMPAT_FREEBSD11 */
4085 
4086 /*
4087  * Read a block of directory entries in a filesystem independent format.
4088  */
4089 int
4090 sys_getdirentries(struct thread *td, struct getdirentries_args *uap)
4091 {
4092 	off_t base;
4093 	int error;
4094 
4095 	error = kern_getdirentries(td, uap->fd, uap->buf, uap->count, &base,
4096 	    NULL, UIO_USERSPACE);
4097 	if (error != 0)
4098 		return (error);
4099 	if (uap->basep != NULL)
4100 		error = copyout(&base, uap->basep, sizeof(off_t));
4101 	return (error);
4102 }
4103 
4104 int
4105 kern_getdirentries(struct thread *td, int fd, char *buf, size_t count,
4106     off_t *basep, ssize_t *residp, enum uio_seg bufseg)
4107 {
4108 	struct vnode *vp;
4109 	struct file *fp;
4110 	struct uio auio;
4111 	struct iovec aiov;
4112 	off_t loff;
4113 	int error, eofflag;
4114 	off_t foffset;
4115 
4116 	AUDIT_ARG_FD(fd);
4117 	if (count > IOSIZE_MAX)
4118 		return (EINVAL);
4119 	auio.uio_resid = count;
4120 	error = getvnode(td, fd, &cap_read_rights, &fp);
4121 	if (error != 0)
4122 		return (error);
4123 	if ((fp->f_flag & FREAD) == 0) {
4124 		fdrop(fp, td);
4125 		return (EBADF);
4126 	}
4127 	vp = fp->f_vnode;
4128 	foffset = foffset_lock(fp, 0);
4129 unionread:
4130 	if (vp->v_type != VDIR) {
4131 		error = EINVAL;
4132 		goto fail;
4133 	}
4134 	aiov.iov_base = buf;
4135 	aiov.iov_len = count;
4136 	auio.uio_iov = &aiov;
4137 	auio.uio_iovcnt = 1;
4138 	auio.uio_rw = UIO_READ;
4139 	auio.uio_segflg = bufseg;
4140 	auio.uio_td = td;
4141 	vn_lock(vp, LK_SHARED | LK_RETRY);
4142 	AUDIT_ARG_VNODE1(vp);
4143 	loff = auio.uio_offset = foffset;
4144 #ifdef MAC
4145 	error = mac_vnode_check_readdir(td->td_ucred, vp);
4146 	if (error == 0)
4147 #endif
4148 		error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, NULL,
4149 		    NULL);
4150 	foffset = auio.uio_offset;
4151 	if (error != 0) {
4152 		VOP_UNLOCK(vp);
4153 		goto fail;
4154 	}
4155 	if (count == auio.uio_resid &&
4156 	    (vp->v_vflag & VV_ROOT) &&
4157 	    (vp->v_mount->mnt_flag & MNT_UNION)) {
4158 		struct vnode *tvp = vp;
4159 
4160 		vp = vp->v_mount->mnt_vnodecovered;
4161 		VREF(vp);
4162 		fp->f_vnode = vp;
4163 		foffset = 0;
4164 		vput(tvp);
4165 		goto unionread;
4166 	}
4167 	VOP_UNLOCK(vp);
4168 	*basep = loff;
4169 	if (residp != NULL)
4170 		*residp = auio.uio_resid;
4171 	td->td_retval[0] = count - auio.uio_resid;
4172 fail:
4173 	foffset_unlock(fp, foffset, 0);
4174 	fdrop(fp, td);
4175 	return (error);
4176 }
4177 
4178 /*
4179  * Set the mode mask for creation of filesystem nodes.
4180  */
4181 #ifndef _SYS_SYSPROTO_H_
4182 struct umask_args {
4183 	int	newmask;
4184 };
4185 #endif
4186 int
4187 sys_umask(struct thread *td, struct umask_args *uap)
4188 {
4189 	struct pwddesc *pdp;
4190 
4191 	pdp = td->td_proc->p_pd;
4192 	PWDDESC_XLOCK(pdp);
4193 	td->td_retval[0] = pdp->pd_cmask;
4194 	pdp->pd_cmask = uap->newmask & ALLPERMS;
4195 	PWDDESC_XUNLOCK(pdp);
4196 	return (0);
4197 }
4198 
4199 /*
4200  * Void all references to file by ripping underlying filesystem away from
4201  * vnode.
4202  */
4203 #ifndef _SYS_SYSPROTO_H_
4204 struct revoke_args {
4205 	char	*path;
4206 };
4207 #endif
4208 int
4209 sys_revoke(struct thread *td, struct revoke_args *uap)
4210 {
4211 	struct vnode *vp;
4212 	struct vattr vattr;
4213 	struct nameidata nd;
4214 	int error;
4215 
4216 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
4217 	    uap->path, td);
4218 	if ((error = namei(&nd)) != 0)
4219 		return (error);
4220 	vp = nd.ni_vp;
4221 	NDFREE_NOTHING(&nd);
4222 	if (vp->v_type != VCHR || vp->v_rdev == NULL) {
4223 		error = EINVAL;
4224 		goto out;
4225 	}
4226 #ifdef MAC
4227 	error = mac_vnode_check_revoke(td->td_ucred, vp);
4228 	if (error != 0)
4229 		goto out;
4230 #endif
4231 	error = VOP_GETATTR(vp, &vattr, td->td_ucred);
4232 	if (error != 0)
4233 		goto out;
4234 	if (td->td_ucred->cr_uid != vattr.va_uid) {
4235 		error = priv_check(td, PRIV_VFS_ADMIN);
4236 		if (error != 0)
4237 			goto out;
4238 	}
4239 	if (devfs_usecount(vp) > 0)
4240 		VOP_REVOKE(vp, REVOKEALL);
4241 out:
4242 	vput(vp);
4243 	return (error);
4244 }
4245 
4246 /*
4247  * Convert a user file descriptor to a kernel file entry and check that, if it
4248  * is a capability, the correct rights are present. A reference on the file
4249  * entry is held upon returning.
4250  */
4251 int
4252 getvnode(struct thread *td, int fd, cap_rights_t *rightsp, struct file **fpp)
4253 {
4254 	struct file *fp;
4255 	int error;
4256 
4257 	error = fget_unlocked(td->td_proc->p_fd, fd, rightsp, &fp);
4258 	if (error != 0)
4259 		return (error);
4260 
4261 	/*
4262 	 * The file could be not of the vnode type, or it may be not
4263 	 * yet fully initialized, in which case the f_vnode pointer
4264 	 * may be set, but f_ops is still badfileops.  E.g.,
4265 	 * devfs_open() transiently create such situation to
4266 	 * facilitate csw d_fdopen().
4267 	 *
4268 	 * Dupfdopen() handling in kern_openat() installs the
4269 	 * half-baked file into the process descriptor table, allowing
4270 	 * other thread to dereference it. Guard against the race by
4271 	 * checking f_ops.
4272 	 */
4273 	if (fp->f_vnode == NULL || fp->f_ops == &badfileops) {
4274 		fdrop(fp, td);
4275 		return (EINVAL);
4276 	}
4277 	*fpp = fp;
4278 	return (0);
4279 }
4280 
4281 /*
4282  * Get an (NFS) file handle.
4283  */
4284 #ifndef _SYS_SYSPROTO_H_
4285 struct lgetfh_args {
4286 	char *fname;
4287 	fhandle_t *fhp;
4288 };
4289 #endif
4290 int
4291 sys_lgetfh(struct thread *td, struct lgetfh_args *uap)
4292 {
4293 
4294 	return (kern_getfhat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->fname,
4295 	    UIO_USERSPACE, uap->fhp, UIO_USERSPACE));
4296 }
4297 
4298 #ifndef _SYS_SYSPROTO_H_
4299 struct getfh_args {
4300 	char *fname;
4301 	fhandle_t *fhp;
4302 };
4303 #endif
4304 int
4305 sys_getfh(struct thread *td, struct getfh_args *uap)
4306 {
4307 
4308 	return (kern_getfhat(td, 0, AT_FDCWD, uap->fname, UIO_USERSPACE,
4309 	    uap->fhp, UIO_USERSPACE));
4310 }
4311 
4312 /*
4313  * syscall for the rpc.lockd to use to translate an open descriptor into
4314  * a NFS file handle.
4315  *
4316  * warning: do not remove the priv_check() call or this becomes one giant
4317  * security hole.
4318  */
4319 #ifndef _SYS_SYSPROTO_H_
4320 struct getfhat_args {
4321 	int fd;
4322 	char *path;
4323 	fhandle_t *fhp;
4324 	int flags;
4325 };
4326 #endif
4327 int
4328 sys_getfhat(struct thread *td, struct getfhat_args *uap)
4329 {
4330 
4331 	if ((uap->flags & ~(AT_SYMLINK_NOFOLLOW | AT_BENEATH |
4332 	    AT_RESOLVE_BENEATH)) != 0)
4333 		return (EINVAL);
4334 	return (kern_getfhat(td, uap->flags, uap->fd, uap->path, UIO_USERSPACE,
4335 	    uap->fhp, UIO_USERSPACE));
4336 }
4337 
4338 int
4339 kern_getfhat(struct thread *td, int flags, int fd, const char *path,
4340     enum uio_seg pathseg, fhandle_t *fhp, enum uio_seg fhseg)
4341 {
4342 	struct nameidata nd;
4343 	fhandle_t fh;
4344 	struct vnode *vp;
4345 	int error;
4346 
4347 	error = priv_check(td, PRIV_VFS_GETFH);
4348 	if (error != 0)
4349 		return (error);
4350 	NDINIT_AT(&nd, LOOKUP, at2cnpflags(flags, AT_SYMLINK_NOFOLLOW |
4351 	    AT_BENEATH | AT_RESOLVE_BENEATH) | LOCKLEAF | AUDITVNODE1,
4352 	    pathseg, path, fd, td);
4353 	error = namei(&nd);
4354 	if (error != 0)
4355 		return (error);
4356 	NDFREE_NOTHING(&nd);
4357 	vp = nd.ni_vp;
4358 	bzero(&fh, sizeof(fh));
4359 	fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
4360 	error = VOP_VPTOFH(vp, &fh.fh_fid);
4361 	vput(vp);
4362 	if (error == 0) {
4363 		if (fhseg == UIO_USERSPACE)
4364 			error = copyout(&fh, fhp, sizeof (fh));
4365 		else
4366 			memcpy(fhp, &fh, sizeof(fh));
4367 	}
4368 	return (error);
4369 }
4370 
4371 #ifndef _SYS_SYSPROTO_H_
4372 struct fhlink_args {
4373 	fhandle_t *fhp;
4374 	const char *to;
4375 };
4376 #endif
4377 int
4378 sys_fhlink(struct thread *td, struct fhlink_args *uap)
4379 {
4380 
4381 	return (kern_fhlinkat(td, AT_FDCWD, uap->to, UIO_USERSPACE, uap->fhp));
4382 }
4383 
4384 #ifndef _SYS_SYSPROTO_H_
4385 struct fhlinkat_args {
4386 	fhandle_t *fhp;
4387 	int tofd;
4388 	const char *to;
4389 };
4390 #endif
4391 int
4392 sys_fhlinkat(struct thread *td, struct fhlinkat_args *uap)
4393 {
4394 
4395 	return (kern_fhlinkat(td, uap->tofd, uap->to, UIO_USERSPACE, uap->fhp));
4396 }
4397 
4398 static int
4399 kern_fhlinkat(struct thread *td, int fd, const char *path,
4400     enum uio_seg pathseg, fhandle_t *fhp)
4401 {
4402 	fhandle_t fh;
4403 	struct mount *mp;
4404 	struct vnode *vp;
4405 	int error;
4406 
4407 	error = priv_check(td, PRIV_VFS_GETFH);
4408 	if (error != 0)
4409 		return (error);
4410 	error = copyin(fhp, &fh, sizeof(fh));
4411 	if (error != 0)
4412 		return (error);
4413 	do {
4414 		bwillwrite();
4415 		if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4416 			return (ESTALE);
4417 		error = VFS_FHTOVP(mp, &fh.fh_fid, LK_SHARED, &vp);
4418 		vfs_unbusy(mp);
4419 		if (error != 0)
4420 			return (error);
4421 		VOP_UNLOCK(vp);
4422 		error = kern_linkat_vp(td, vp, fd, path, pathseg);
4423 	} while (error == EAGAIN || error == ERELOOKUP);
4424 	return (error);
4425 }
4426 
4427 #ifndef _SYS_SYSPROTO_H_
4428 struct fhreadlink_args {
4429 	fhandle_t *fhp;
4430 	char *buf;
4431 	size_t bufsize;
4432 };
4433 #endif
4434 int
4435 sys_fhreadlink(struct thread *td, struct fhreadlink_args *uap)
4436 {
4437 	fhandle_t fh;
4438 	struct mount *mp;
4439 	struct vnode *vp;
4440 	int error;
4441 
4442 	error = priv_check(td, PRIV_VFS_GETFH);
4443 	if (error != 0)
4444 		return (error);
4445 	if (uap->bufsize > IOSIZE_MAX)
4446 		return (EINVAL);
4447 	error = copyin(uap->fhp, &fh, sizeof(fh));
4448 	if (error != 0)
4449 		return (error);
4450 	if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4451 		return (ESTALE);
4452 	error = VFS_FHTOVP(mp, &fh.fh_fid, LK_SHARED, &vp);
4453 	vfs_unbusy(mp);
4454 	if (error != 0)
4455 		return (error);
4456 	error = kern_readlink_vp(vp, uap->buf, UIO_USERSPACE, uap->bufsize, td);
4457 	vput(vp);
4458 	return (error);
4459 }
4460 
4461 /*
4462  * syscall for the rpc.lockd to use to translate a NFS file handle into an
4463  * open descriptor.
4464  *
4465  * warning: do not remove the priv_check() call or this becomes one giant
4466  * security hole.
4467  */
4468 #ifndef _SYS_SYSPROTO_H_
4469 struct fhopen_args {
4470 	const struct fhandle *u_fhp;
4471 	int flags;
4472 };
4473 #endif
4474 int
4475 sys_fhopen(struct thread *td, struct fhopen_args *uap)
4476 {
4477 	return (kern_fhopen(td, uap->u_fhp, uap->flags));
4478 }
4479 
4480 int
4481 kern_fhopen(struct thread *td, const struct fhandle *u_fhp, int flags)
4482 {
4483 	struct mount *mp;
4484 	struct vnode *vp;
4485 	struct fhandle fhp;
4486 	struct file *fp;
4487 	int fmode, error;
4488 	int indx;
4489 
4490 	error = priv_check(td, PRIV_VFS_FHOPEN);
4491 	if (error != 0)
4492 		return (error);
4493 	indx = -1;
4494 	fmode = FFLAGS(flags);
4495 	/* why not allow a non-read/write open for our lockd? */
4496 	if (((fmode & (FREAD | FWRITE)) == 0) || (fmode & O_CREAT))
4497 		return (EINVAL);
4498 	error = copyin(u_fhp, &fhp, sizeof(fhp));
4499 	if (error != 0)
4500 		return(error);
4501 	/* find the mount point */
4502 	mp = vfs_busyfs(&fhp.fh_fsid);
4503 	if (mp == NULL)
4504 		return (ESTALE);
4505 	/* now give me my vnode, it gets returned to me locked */
4506 	error = VFS_FHTOVP(mp, &fhp.fh_fid, LK_EXCLUSIVE, &vp);
4507 	vfs_unbusy(mp);
4508 	if (error != 0)
4509 		return (error);
4510 
4511 	error = falloc_noinstall(td, &fp);
4512 	if (error != 0) {
4513 		vput(vp);
4514 		return (error);
4515 	}
4516 	/*
4517 	 * An extra reference on `fp' has been held for us by
4518 	 * falloc_noinstall().
4519 	 */
4520 
4521 #ifdef INVARIANTS
4522 	td->td_dupfd = -1;
4523 #endif
4524 	error = vn_open_vnode(vp, fmode, td->td_ucred, td, fp);
4525 	if (error != 0) {
4526 		KASSERT(fp->f_ops == &badfileops,
4527 		    ("VOP_OPEN in fhopen() set f_ops"));
4528 		KASSERT(td->td_dupfd < 0,
4529 		    ("fhopen() encountered fdopen()"));
4530 
4531 		vput(vp);
4532 		goto bad;
4533 	}
4534 #ifdef INVARIANTS
4535 	td->td_dupfd = 0;
4536 #endif
4537 	fp->f_vnode = vp;
4538 	finit_vnode(fp, fmode, NULL, &vnops);
4539 	VOP_UNLOCK(vp);
4540 	if ((fmode & O_TRUNC) != 0) {
4541 		error = fo_truncate(fp, 0, td->td_ucred, td);
4542 		if (error != 0)
4543 			goto bad;
4544 	}
4545 
4546 	error = finstall(td, fp, &indx, fmode, NULL);
4547 bad:
4548 	fdrop(fp, td);
4549 	td->td_retval[0] = indx;
4550 	return (error);
4551 }
4552 
4553 /*
4554  * Stat an (NFS) file handle.
4555  */
4556 #ifndef _SYS_SYSPROTO_H_
4557 struct fhstat_args {
4558 	struct fhandle *u_fhp;
4559 	struct stat *sb;
4560 };
4561 #endif
4562 int
4563 sys_fhstat(struct thread *td, struct fhstat_args *uap)
4564 {
4565 	struct stat sb;
4566 	struct fhandle fh;
4567 	int error;
4568 
4569 	error = copyin(uap->u_fhp, &fh, sizeof(fh));
4570 	if (error != 0)
4571 		return (error);
4572 	error = kern_fhstat(td, fh, &sb);
4573 	if (error == 0)
4574 		error = copyout(&sb, uap->sb, sizeof(sb));
4575 	return (error);
4576 }
4577 
4578 int
4579 kern_fhstat(struct thread *td, struct fhandle fh, struct stat *sb)
4580 {
4581 	struct mount *mp;
4582 	struct vnode *vp;
4583 	int error;
4584 
4585 	error = priv_check(td, PRIV_VFS_FHSTAT);
4586 	if (error != 0)
4587 		return (error);
4588 	if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4589 		return (ESTALE);
4590 	error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4591 	vfs_unbusy(mp);
4592 	if (error != 0)
4593 		return (error);
4594 	error = VOP_STAT(vp, sb, td->td_ucred, NOCRED, td);
4595 	vput(vp);
4596 	return (error);
4597 }
4598 
4599 /*
4600  * Implement fstatfs() for (NFS) file handles.
4601  */
4602 #ifndef _SYS_SYSPROTO_H_
4603 struct fhstatfs_args {
4604 	struct fhandle *u_fhp;
4605 	struct statfs *buf;
4606 };
4607 #endif
4608 int
4609 sys_fhstatfs(struct thread *td, struct fhstatfs_args *uap)
4610 {
4611 	struct statfs *sfp;
4612 	fhandle_t fh;
4613 	int error;
4614 
4615 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
4616 	if (error != 0)
4617 		return (error);
4618 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
4619 	error = kern_fhstatfs(td, fh, sfp);
4620 	if (error == 0)
4621 		error = copyout(sfp, uap->buf, sizeof(*sfp));
4622 	free(sfp, M_STATFS);
4623 	return (error);
4624 }
4625 
4626 int
4627 kern_fhstatfs(struct thread *td, fhandle_t fh, struct statfs *buf)
4628 {
4629 	struct mount *mp;
4630 	struct vnode *vp;
4631 	int error;
4632 
4633 	error = priv_check(td, PRIV_VFS_FHSTATFS);
4634 	if (error != 0)
4635 		return (error);
4636 	if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4637 		return (ESTALE);
4638 	error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4639 	if (error != 0) {
4640 		vfs_unbusy(mp);
4641 		return (error);
4642 	}
4643 	vput(vp);
4644 	error = prison_canseemount(td->td_ucred, mp);
4645 	if (error != 0)
4646 		goto out;
4647 #ifdef MAC
4648 	error = mac_mount_check_stat(td->td_ucred, mp);
4649 	if (error != 0)
4650 		goto out;
4651 #endif
4652 	error = VFS_STATFS(mp, buf);
4653 out:
4654 	vfs_unbusy(mp);
4655 	return (error);
4656 }
4657 
4658 /*
4659  * Unlike madvise(2), we do not make a best effort to remember every
4660  * possible caching hint.  Instead, we remember the last setting with
4661  * the exception that we will allow POSIX_FADV_NORMAL to adjust the
4662  * region of any current setting.
4663  */
4664 int
4665 kern_posix_fadvise(struct thread *td, int fd, off_t offset, off_t len,
4666     int advice)
4667 {
4668 	struct fadvise_info *fa, *new;
4669 	struct file *fp;
4670 	struct vnode *vp;
4671 	off_t end;
4672 	int error;
4673 
4674 	if (offset < 0 || len < 0 || offset > OFF_MAX - len)
4675 		return (EINVAL);
4676 	AUDIT_ARG_VALUE(advice);
4677 	switch (advice) {
4678 	case POSIX_FADV_SEQUENTIAL:
4679 	case POSIX_FADV_RANDOM:
4680 	case POSIX_FADV_NOREUSE:
4681 		new = malloc(sizeof(*fa), M_FADVISE, M_WAITOK);
4682 		break;
4683 	case POSIX_FADV_NORMAL:
4684 	case POSIX_FADV_WILLNEED:
4685 	case POSIX_FADV_DONTNEED:
4686 		new = NULL;
4687 		break;
4688 	default:
4689 		return (EINVAL);
4690 	}
4691 	/* XXX: CAP_POSIX_FADVISE? */
4692 	AUDIT_ARG_FD(fd);
4693 	error = fget(td, fd, &cap_no_rights, &fp);
4694 	if (error != 0)
4695 		goto out;
4696 	AUDIT_ARG_FILE(td->td_proc, fp);
4697 	if ((fp->f_ops->fo_flags & DFLAG_SEEKABLE) == 0) {
4698 		error = ESPIPE;
4699 		goto out;
4700 	}
4701 	if (fp->f_type != DTYPE_VNODE) {
4702 		error = ENODEV;
4703 		goto out;
4704 	}
4705 	vp = fp->f_vnode;
4706 	if (vp->v_type != VREG) {
4707 		error = ENODEV;
4708 		goto out;
4709 	}
4710 	if (len == 0)
4711 		end = OFF_MAX;
4712 	else
4713 		end = offset + len - 1;
4714 	switch (advice) {
4715 	case POSIX_FADV_SEQUENTIAL:
4716 	case POSIX_FADV_RANDOM:
4717 	case POSIX_FADV_NOREUSE:
4718 		/*
4719 		 * Try to merge any existing non-standard region with
4720 		 * this new region if possible, otherwise create a new
4721 		 * non-standard region for this request.
4722 		 */
4723 		mtx_pool_lock(mtxpool_sleep, fp);
4724 		fa = fp->f_advice;
4725 		if (fa != NULL && fa->fa_advice == advice &&
4726 		    ((fa->fa_start <= end && fa->fa_end >= offset) ||
4727 		    (end != OFF_MAX && fa->fa_start == end + 1) ||
4728 		    (fa->fa_end != OFF_MAX && fa->fa_end + 1 == offset))) {
4729 			if (offset < fa->fa_start)
4730 				fa->fa_start = offset;
4731 			if (end > fa->fa_end)
4732 				fa->fa_end = end;
4733 		} else {
4734 			new->fa_advice = advice;
4735 			new->fa_start = offset;
4736 			new->fa_end = end;
4737 			fp->f_advice = new;
4738 			new = fa;
4739 		}
4740 		mtx_pool_unlock(mtxpool_sleep, fp);
4741 		break;
4742 	case POSIX_FADV_NORMAL:
4743 		/*
4744 		 * If a the "normal" region overlaps with an existing
4745 		 * non-standard region, trim or remove the
4746 		 * non-standard region.
4747 		 */
4748 		mtx_pool_lock(mtxpool_sleep, fp);
4749 		fa = fp->f_advice;
4750 		if (fa != NULL) {
4751 			if (offset <= fa->fa_start && end >= fa->fa_end) {
4752 				new = fa;
4753 				fp->f_advice = NULL;
4754 			} else if (offset <= fa->fa_start &&
4755 			    end >= fa->fa_start)
4756 				fa->fa_start = end + 1;
4757 			else if (offset <= fa->fa_end && end >= fa->fa_end)
4758 				fa->fa_end = offset - 1;
4759 			else if (offset >= fa->fa_start && end <= fa->fa_end) {
4760 				/*
4761 				 * If the "normal" region is a middle
4762 				 * portion of the existing
4763 				 * non-standard region, just remove
4764 				 * the whole thing rather than picking
4765 				 * one side or the other to
4766 				 * preserve.
4767 				 */
4768 				new = fa;
4769 				fp->f_advice = NULL;
4770 			}
4771 		}
4772 		mtx_pool_unlock(mtxpool_sleep, fp);
4773 		break;
4774 	case POSIX_FADV_WILLNEED:
4775 	case POSIX_FADV_DONTNEED:
4776 		error = VOP_ADVISE(vp, offset, end, advice);
4777 		break;
4778 	}
4779 out:
4780 	if (fp != NULL)
4781 		fdrop(fp, td);
4782 	free(new, M_FADVISE);
4783 	return (error);
4784 }
4785 
4786 int
4787 sys_posix_fadvise(struct thread *td, struct posix_fadvise_args *uap)
4788 {
4789 	int error;
4790 
4791 	error = kern_posix_fadvise(td, uap->fd, uap->offset, uap->len,
4792 	    uap->advice);
4793 	return (kern_posix_error(td, error));
4794 }
4795 
4796 int
4797 kern_copy_file_range(struct thread *td, int infd, off_t *inoffp, int outfd,
4798     off_t *outoffp, size_t len, unsigned int flags)
4799 {
4800 	struct file *infp, *outfp;
4801 	struct vnode *invp, *outvp;
4802 	int error;
4803 	size_t retlen;
4804 	void *rl_rcookie, *rl_wcookie;
4805 	off_t savinoff, savoutoff;
4806 
4807 	infp = outfp = NULL;
4808 	rl_rcookie = rl_wcookie = NULL;
4809 	savinoff = -1;
4810 	error = 0;
4811 	retlen = 0;
4812 
4813 	if (flags != 0) {
4814 		error = EINVAL;
4815 		goto out;
4816 	}
4817 	if (len > SSIZE_MAX)
4818 		/*
4819 		 * Although the len argument is size_t, the return argument
4820 		 * is ssize_t (which is signed).  Therefore a size that won't
4821 		 * fit in ssize_t can't be returned.
4822 		 */
4823 		len = SSIZE_MAX;
4824 
4825 	/* Get the file structures for the file descriptors. */
4826 	error = fget_read(td, infd, &cap_read_rights, &infp);
4827 	if (error != 0)
4828 		goto out;
4829 	if (infp->f_ops == &badfileops) {
4830 		error = EBADF;
4831 		goto out;
4832 	}
4833 	if (infp->f_vnode == NULL) {
4834 		error = EINVAL;
4835 		goto out;
4836 	}
4837 	error = fget_write(td, outfd, &cap_write_rights, &outfp);
4838 	if (error != 0)
4839 		goto out;
4840 	if (outfp->f_ops == &badfileops) {
4841 		error = EBADF;
4842 		goto out;
4843 	}
4844 	if (outfp->f_vnode == NULL) {
4845 		error = EINVAL;
4846 		goto out;
4847 	}
4848 
4849 	/* Set the offset pointers to the correct place. */
4850 	if (inoffp == NULL)
4851 		inoffp = &infp->f_offset;
4852 	if (outoffp == NULL)
4853 		outoffp = &outfp->f_offset;
4854 	savinoff = *inoffp;
4855 	savoutoff = *outoffp;
4856 
4857 	invp = infp->f_vnode;
4858 	outvp = outfp->f_vnode;
4859 	/* Sanity check the f_flag bits. */
4860 	if ((outfp->f_flag & (FWRITE | FAPPEND)) != FWRITE ||
4861 	    (infp->f_flag & FREAD) == 0) {
4862 		error = EBADF;
4863 		goto out;
4864 	}
4865 
4866 	/* If len == 0, just return 0. */
4867 	if (len == 0)
4868 		goto out;
4869 
4870 	/*
4871 	 * If infp and outfp refer to the same file, the byte ranges cannot
4872 	 * overlap.
4873 	 */
4874 	if (invp == outvp && ((savinoff <= savoutoff && savinoff + len >
4875 	    savoutoff) || (savinoff > savoutoff && savoutoff + len >
4876 	    savinoff))) {
4877 		error = EINVAL;
4878 		goto out;
4879 	}
4880 
4881 	/* Range lock the byte ranges for both invp and outvp. */
4882 	for (;;) {
4883 		rl_wcookie = vn_rangelock_wlock(outvp, *outoffp, *outoffp +
4884 		    len);
4885 		rl_rcookie = vn_rangelock_tryrlock(invp, *inoffp, *inoffp +
4886 		    len);
4887 		if (rl_rcookie != NULL)
4888 			break;
4889 		vn_rangelock_unlock(outvp, rl_wcookie);
4890 		rl_rcookie = vn_rangelock_rlock(invp, *inoffp, *inoffp + len);
4891 		vn_rangelock_unlock(invp, rl_rcookie);
4892 	}
4893 
4894 	retlen = len;
4895 	error = vn_copy_file_range(invp, inoffp, outvp, outoffp, &retlen,
4896 	    flags, infp->f_cred, outfp->f_cred, td);
4897 out:
4898 	if (rl_rcookie != NULL)
4899 		vn_rangelock_unlock(invp, rl_rcookie);
4900 	if (rl_wcookie != NULL)
4901 		vn_rangelock_unlock(outvp, rl_wcookie);
4902 	if (savinoff != -1 && (error == EINTR || error == ERESTART)) {
4903 		*inoffp = savinoff;
4904 		*outoffp = savoutoff;
4905 	}
4906 	if (outfp != NULL)
4907 		fdrop(outfp, td);
4908 	if (infp != NULL)
4909 		fdrop(infp, td);
4910 	td->td_retval[0] = retlen;
4911 	return (error);
4912 }
4913 
4914 int
4915 sys_copy_file_range(struct thread *td, struct copy_file_range_args *uap)
4916 {
4917 	off_t inoff, outoff, *inoffp, *outoffp;
4918 	int error;
4919 
4920 	inoffp = outoffp = NULL;
4921 	if (uap->inoffp != NULL) {
4922 		error = copyin(uap->inoffp, &inoff, sizeof(off_t));
4923 		if (error != 0)
4924 			return (error);
4925 		inoffp = &inoff;
4926 	}
4927 	if (uap->outoffp != NULL) {
4928 		error = copyin(uap->outoffp, &outoff, sizeof(off_t));
4929 		if (error != 0)
4930 			return (error);
4931 		outoffp = &outoff;
4932 	}
4933 	error = kern_copy_file_range(td, uap->infd, inoffp, uap->outfd,
4934 	    outoffp, uap->len, uap->flags);
4935 	if (error == 0 && uap->inoffp != NULL)
4936 		error = copyout(inoffp, uap->inoffp, sizeof(off_t));
4937 	if (error == 0 && uap->outoffp != NULL)
4938 		error = copyout(outoffp, uap->outoffp, sizeof(off_t));
4939 	return (error);
4940 }
4941