xref: /freebsd/sys/kern/vfs_syscalls.c (revision 1603881667360c015f6685131f2f25474fa67a72)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (c) 1989, 1993
5  *	The Regents of the University of California.  All rights reserved.
6  * (c) UNIX System Laboratories, Inc.
7  * All or some portions of this file are derived from material licensed
8  * to the University of California by American Telephone and Telegraph
9  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
10  * the permission of UNIX System Laboratories, Inc.
11  *
12  * Redistribution and use in source and binary forms, with or without
13  * modification, are permitted provided that the following conditions
14  * are met:
15  * 1. Redistributions of source code must retain the above copyright
16  *    notice, this list of conditions and the following disclaimer.
17  * 2. Redistributions in binary form must reproduce the above copyright
18  *    notice, this list of conditions and the following disclaimer in the
19  *    documentation and/or other materials provided with the distribution.
20  * 3. Neither the name of the University nor the names of its contributors
21  *    may be used to endorse or promote products derived from this software
22  *    without specific prior written permission.
23  *
24  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
25  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
28  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
29  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
30  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
31  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
32  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
33  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
34  * SUCH DAMAGE.
35  *
36  *	@(#)vfs_syscalls.c	8.13 (Berkeley) 4/15/94
37  */
38 
39 #include <sys/cdefs.h>
40 __FBSDID("$FreeBSD$");
41 
42 #include "opt_capsicum.h"
43 #include "opt_ktrace.h"
44 
45 #include <sys/param.h>
46 #include <sys/systm.h>
47 #include <sys/bio.h>
48 #include <sys/buf.h>
49 #include <sys/capsicum.h>
50 #include <sys/disk.h>
51 #include <sys/sysent.h>
52 #include <sys/malloc.h>
53 #include <sys/mount.h>
54 #include <sys/mutex.h>
55 #include <sys/sysproto.h>
56 #include <sys/namei.h>
57 #include <sys/filedesc.h>
58 #include <sys/kernel.h>
59 #include <sys/fcntl.h>
60 #include <sys/file.h>
61 #include <sys/filio.h>
62 #include <sys/limits.h>
63 #include <sys/linker.h>
64 #include <sys/rwlock.h>
65 #include <sys/sdt.h>
66 #include <sys/stat.h>
67 #include <sys/sx.h>
68 #include <sys/unistd.h>
69 #include <sys/vnode.h>
70 #include <sys/priv.h>
71 #include <sys/proc.h>
72 #include <sys/dirent.h>
73 #include <sys/jail.h>
74 #include <sys/syscallsubr.h>
75 #include <sys/sysctl.h>
76 #ifdef KTRACE
77 #include <sys/ktrace.h>
78 #endif
79 
80 #include <machine/stdarg.h>
81 
82 #include <security/audit/audit.h>
83 #include <security/mac/mac_framework.h>
84 
85 #include <vm/vm.h>
86 #include <vm/vm_object.h>
87 #include <vm/vm_page.h>
88 #include <vm/uma.h>
89 
90 #include <fs/devfs/devfs.h>
91 
92 MALLOC_DEFINE(M_FADVISE, "fadvise", "posix_fadvise(2) information");
93 
94 static int kern_chflagsat(struct thread *td, int fd, const char *path,
95     enum uio_seg pathseg, u_long flags, int atflag);
96 static int setfflags(struct thread *td, struct vnode *, u_long);
97 static int getutimes(const struct timeval *, enum uio_seg, struct timespec *);
98 static int getutimens(const struct timespec *, enum uio_seg,
99     struct timespec *, int *);
100 static int setutimes(struct thread *td, struct vnode *,
101     const struct timespec *, int, int);
102 static int vn_access(struct vnode *vp, int user_flags, struct ucred *cred,
103     struct thread *td);
104 static int kern_fhlinkat(struct thread *td, int fd, const char *path,
105     enum uio_seg pathseg, fhandle_t *fhp);
106 static int kern_readlink_vp(struct vnode *vp, char *buf, enum uio_seg bufseg,
107     size_t count, struct thread *td);
108 static int kern_linkat_vp(struct thread *td, struct vnode *vp, int fd,
109     const char *path, enum uio_seg segflag);
110 
111 static uint64_t
112 at2cnpflags(u_int at_flags, u_int mask)
113 {
114 	u_int64_t res;
115 
116 	MPASS((at_flags & (AT_SYMLINK_FOLLOW | AT_SYMLINK_NOFOLLOW)) !=
117 	    (AT_SYMLINK_FOLLOW | AT_SYMLINK_NOFOLLOW));
118 
119 	res = 0;
120 	at_flags &= mask;
121 	if ((at_flags & AT_RESOLVE_BENEATH) != 0)
122 		res |= RBENEATH;
123 	if ((at_flags & AT_SYMLINK_FOLLOW) != 0)
124 		res |= FOLLOW;
125 	/* NOFOLLOW is pseudo flag */
126 	if ((mask & AT_SYMLINK_NOFOLLOW) != 0) {
127 		res |= (at_flags & AT_SYMLINK_NOFOLLOW) != 0 ? NOFOLLOW :
128 		    FOLLOW;
129 	}
130 	if ((mask & AT_EMPTY_PATH) != 0 && (at_flags & AT_EMPTY_PATH) != 0)
131 		res |= EMPTYPATH;
132 	return (res);
133 }
134 
135 int
136 kern_sync(struct thread *td)
137 {
138 	struct mount *mp, *nmp;
139 	int save;
140 
141 	mtx_lock(&mountlist_mtx);
142 	for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
143 		if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK)) {
144 			nmp = TAILQ_NEXT(mp, mnt_list);
145 			continue;
146 		}
147 		if ((mp->mnt_flag & MNT_RDONLY) == 0 &&
148 		    vn_start_write(NULL, &mp, V_NOWAIT) == 0) {
149 			save = curthread_pflags_set(TDP_SYNCIO);
150 			vfs_periodic(mp, MNT_NOWAIT);
151 			VFS_SYNC(mp, MNT_NOWAIT);
152 			curthread_pflags_restore(save);
153 			vn_finished_write(mp);
154 		}
155 		mtx_lock(&mountlist_mtx);
156 		nmp = TAILQ_NEXT(mp, mnt_list);
157 		vfs_unbusy(mp);
158 	}
159 	mtx_unlock(&mountlist_mtx);
160 	return (0);
161 }
162 
163 /*
164  * Sync each mounted filesystem.
165  */
166 #ifndef _SYS_SYSPROTO_H_
167 struct sync_args {
168 	int     dummy;
169 };
170 #endif
171 /* ARGSUSED */
172 int
173 sys_sync(struct thread *td, struct sync_args *uap)
174 {
175 
176 	return (kern_sync(td));
177 }
178 
179 /*
180  * Change filesystem quotas.
181  */
182 #ifndef _SYS_SYSPROTO_H_
183 struct quotactl_args {
184 	char *path;
185 	int cmd;
186 	int uid;
187 	caddr_t arg;
188 };
189 #endif
190 int
191 sys_quotactl(struct thread *td, struct quotactl_args *uap)
192 {
193 	struct mount *mp;
194 	struct nameidata nd;
195 	int error;
196 	bool mp_busy;
197 
198 	AUDIT_ARG_CMD(uap->cmd);
199 	AUDIT_ARG_UID(uap->uid);
200 	if (!prison_allow(td->td_ucred, PR_ALLOW_QUOTAS))
201 		return (EPERM);
202 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
203 	    uap->path, td);
204 	if ((error = namei(&nd)) != 0)
205 		return (error);
206 	NDFREE(&nd, NDF_ONLY_PNBUF);
207 	mp = nd.ni_vp->v_mount;
208 	vfs_ref(mp);
209 	vput(nd.ni_vp);
210 	error = vfs_busy(mp, 0);
211 	if (error != 0) {
212 		vfs_rel(mp);
213 		return (error);
214 	}
215 	mp_busy = true;
216 	error = VFS_QUOTACTL(mp, uap->cmd, uap->uid, uap->arg, &mp_busy);
217 
218 	/*
219 	 * Since quota on/off operations typically need to open quota
220 	 * files, the implementation may need to unbusy the mount point
221 	 * before calling into namei.  Otherwise, unmount might be
222 	 * started between two vfs_busy() invocations (first is ours,
223 	 * second is from mount point cross-walk code in lookup()),
224 	 * causing deadlock.
225 	 *
226 	 * Avoid unbusying mp if the implementation indicates it has
227 	 * already done so.
228 	 */
229 	if (mp_busy)
230 		vfs_unbusy(mp);
231 	vfs_rel(mp);
232 	return (error);
233 }
234 
235 /*
236  * Used by statfs conversion routines to scale the block size up if
237  * necessary so that all of the block counts are <= 'max_size'.  Note
238  * that 'max_size' should be a bitmask, i.e. 2^n - 1 for some non-zero
239  * value of 'n'.
240  */
241 void
242 statfs_scale_blocks(struct statfs *sf, long max_size)
243 {
244 	uint64_t count;
245 	int shift;
246 
247 	KASSERT(powerof2(max_size + 1), ("%s: invalid max_size", __func__));
248 
249 	/*
250 	 * Attempt to scale the block counts to give a more accurate
251 	 * overview to userland of the ratio of free space to used
252 	 * space.  To do this, find the largest block count and compute
253 	 * a divisor that lets it fit into a signed integer <= max_size.
254 	 */
255 	if (sf->f_bavail < 0)
256 		count = -sf->f_bavail;
257 	else
258 		count = sf->f_bavail;
259 	count = MAX(sf->f_blocks, MAX(sf->f_bfree, count));
260 	if (count <= max_size)
261 		return;
262 
263 	count >>= flsl(max_size);
264 	shift = 0;
265 	while (count > 0) {
266 		shift++;
267 		count >>=1;
268 	}
269 
270 	sf->f_bsize <<= shift;
271 	sf->f_blocks >>= shift;
272 	sf->f_bfree >>= shift;
273 	sf->f_bavail >>= shift;
274 }
275 
276 static int
277 kern_do_statfs(struct thread *td, struct mount *mp, struct statfs *buf)
278 {
279 	int error;
280 
281 	if (mp == NULL)
282 		return (EBADF);
283 	error = vfs_busy(mp, 0);
284 	vfs_rel(mp);
285 	if (error != 0)
286 		return (error);
287 #ifdef MAC
288 	error = mac_mount_check_stat(td->td_ucred, mp);
289 	if (error != 0)
290 		goto out;
291 #endif
292 	error = VFS_STATFS(mp, buf);
293 	if (error != 0)
294 		goto out;
295 	if (priv_check_cred_vfs_generation(td->td_ucred)) {
296 		buf->f_fsid.val[0] = buf->f_fsid.val[1] = 0;
297 		prison_enforce_statfs(td->td_ucred, mp, buf);
298 	}
299 out:
300 	vfs_unbusy(mp);
301 	return (error);
302 }
303 
304 /*
305  * Get filesystem statistics.
306  */
307 #ifndef _SYS_SYSPROTO_H_
308 struct statfs_args {
309 	char *path;
310 	struct statfs *buf;
311 };
312 #endif
313 int
314 sys_statfs(struct thread *td, struct statfs_args *uap)
315 {
316 	struct statfs *sfp;
317 	int error;
318 
319 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
320 	error = kern_statfs(td, uap->path, UIO_USERSPACE, sfp);
321 	if (error == 0)
322 		error = copyout(sfp, uap->buf, sizeof(struct statfs));
323 	free(sfp, M_STATFS);
324 	return (error);
325 }
326 
327 int
328 kern_statfs(struct thread *td, const char *path, enum uio_seg pathseg,
329     struct statfs *buf)
330 {
331 	struct mount *mp;
332 	struct nameidata nd;
333 	int error;
334 
335 	NDINIT(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, td);
336 	error = namei(&nd);
337 	if (error != 0)
338 		return (error);
339 	mp = vfs_ref_from_vp(nd.ni_vp);
340 	NDFREE_NOTHING(&nd);
341 	vrele(nd.ni_vp);
342 	return (kern_do_statfs(td, mp, buf));
343 }
344 
345 /*
346  * Get filesystem statistics.
347  */
348 #ifndef _SYS_SYSPROTO_H_
349 struct fstatfs_args {
350 	int fd;
351 	struct statfs *buf;
352 };
353 #endif
354 int
355 sys_fstatfs(struct thread *td, struct fstatfs_args *uap)
356 {
357 	struct statfs *sfp;
358 	int error;
359 
360 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
361 	error = kern_fstatfs(td, uap->fd, sfp);
362 	if (error == 0)
363 		error = copyout(sfp, uap->buf, sizeof(struct statfs));
364 	free(sfp, M_STATFS);
365 	return (error);
366 }
367 
368 int
369 kern_fstatfs(struct thread *td, int fd, struct statfs *buf)
370 {
371 	struct file *fp;
372 	struct mount *mp;
373 	struct vnode *vp;
374 	int error;
375 
376 	AUDIT_ARG_FD(fd);
377 	error = getvnode_path(td, fd, &cap_fstatfs_rights, &fp);
378 	if (error != 0)
379 		return (error);
380 	vp = fp->f_vnode;
381 #ifdef AUDIT
382 	if (AUDITING_TD(td)) {
383 		vn_lock(vp, LK_SHARED | LK_RETRY);
384 		AUDIT_ARG_VNODE1(vp);
385 		VOP_UNLOCK(vp);
386 	}
387 #endif
388 	mp = vfs_ref_from_vp(vp);
389 	fdrop(fp, td);
390 	return (kern_do_statfs(td, mp, buf));
391 }
392 
393 /*
394  * Get statistics on all filesystems.
395  */
396 #ifndef _SYS_SYSPROTO_H_
397 struct getfsstat_args {
398 	struct statfs *buf;
399 	long bufsize;
400 	int mode;
401 };
402 #endif
403 int
404 sys_getfsstat(struct thread *td, struct getfsstat_args *uap)
405 {
406 	size_t count;
407 	int error;
408 
409 	if (uap->bufsize < 0 || uap->bufsize > SIZE_MAX)
410 		return (EINVAL);
411 	error = kern_getfsstat(td, &uap->buf, uap->bufsize, &count,
412 	    UIO_USERSPACE, uap->mode);
413 	if (error == 0)
414 		td->td_retval[0] = count;
415 	return (error);
416 }
417 
418 /*
419  * If (bufsize > 0 && bufseg == UIO_SYSSPACE)
420  *	The caller is responsible for freeing memory which will be allocated
421  *	in '*buf'.
422  */
423 int
424 kern_getfsstat(struct thread *td, struct statfs **buf, size_t bufsize,
425     size_t *countp, enum uio_seg bufseg, int mode)
426 {
427 	struct mount *mp, *nmp;
428 	struct statfs *sfsp, *sp, *sptmp, *tofree;
429 	size_t count, maxcount;
430 	int error;
431 
432 	switch (mode) {
433 	case MNT_WAIT:
434 	case MNT_NOWAIT:
435 		break;
436 	default:
437 		if (bufseg == UIO_SYSSPACE)
438 			*buf = NULL;
439 		return (EINVAL);
440 	}
441 restart:
442 	maxcount = bufsize / sizeof(struct statfs);
443 	if (bufsize == 0) {
444 		sfsp = NULL;
445 		tofree = NULL;
446 	} else if (bufseg == UIO_USERSPACE) {
447 		sfsp = *buf;
448 		tofree = NULL;
449 	} else /* if (bufseg == UIO_SYSSPACE) */ {
450 		count = 0;
451 		mtx_lock(&mountlist_mtx);
452 		TAILQ_FOREACH(mp, &mountlist, mnt_list) {
453 			count++;
454 		}
455 		mtx_unlock(&mountlist_mtx);
456 		if (maxcount > count)
457 			maxcount = count;
458 		tofree = sfsp = *buf = malloc(maxcount * sizeof(struct statfs),
459 		    M_STATFS, M_WAITOK);
460 	}
461 
462 	count = 0;
463 
464 	/*
465 	 * If there is no target buffer they only want the count.
466 	 *
467 	 * This could be TAILQ_FOREACH but it is open-coded to match the original
468 	 * code below.
469 	 */
470 	if (sfsp == NULL) {
471 		mtx_lock(&mountlist_mtx);
472 		for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
473 			if (prison_canseemount(td->td_ucred, mp) != 0) {
474 				nmp = TAILQ_NEXT(mp, mnt_list);
475 				continue;
476 			}
477 #ifdef MAC
478 			if (mac_mount_check_stat(td->td_ucred, mp) != 0) {
479 				nmp = TAILQ_NEXT(mp, mnt_list);
480 				continue;
481 			}
482 #endif
483 			count++;
484 			nmp = TAILQ_NEXT(mp, mnt_list);
485 		}
486 		mtx_unlock(&mountlist_mtx);
487 		*countp = count;
488 		return (0);
489 	}
490 
491 	/*
492 	 * They want the entire thing.
493 	 *
494 	 * Short-circuit the corner case of no room for anything, avoids
495 	 * relocking below.
496 	 */
497 	if (maxcount < 1) {
498 		goto out;
499 	}
500 
501 	mtx_lock(&mountlist_mtx);
502 	for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
503 		if (prison_canseemount(td->td_ucred, mp) != 0) {
504 			nmp = TAILQ_NEXT(mp, mnt_list);
505 			continue;
506 		}
507 #ifdef MAC
508 		if (mac_mount_check_stat(td->td_ucred, mp) != 0) {
509 			nmp = TAILQ_NEXT(mp, mnt_list);
510 			continue;
511 		}
512 #endif
513 		if (mode == MNT_WAIT) {
514 			if (vfs_busy(mp, MBF_MNTLSTLOCK) != 0) {
515 				/*
516 				 * If vfs_busy() failed, and MBF_NOWAIT
517 				 * wasn't passed, then the mp is gone.
518 				 * Furthermore, because of MBF_MNTLSTLOCK,
519 				 * the mountlist_mtx was dropped.  We have
520 				 * no other choice than to start over.
521 				 */
522 				mtx_unlock(&mountlist_mtx);
523 				free(tofree, M_STATFS);
524 				goto restart;
525 			}
526 		} else {
527 			if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK) != 0) {
528 				nmp = TAILQ_NEXT(mp, mnt_list);
529 				continue;
530 			}
531 		}
532 		sp = &mp->mnt_stat;
533 		/*
534 		 * If MNT_NOWAIT is specified, do not refresh
535 		 * the fsstat cache.
536 		 */
537 		if (mode != MNT_NOWAIT) {
538 			error = VFS_STATFS(mp, sp);
539 			if (error != 0) {
540 				mtx_lock(&mountlist_mtx);
541 				nmp = TAILQ_NEXT(mp, mnt_list);
542 				vfs_unbusy(mp);
543 				continue;
544 			}
545 		}
546 		if (priv_check_cred_vfs_generation(td->td_ucred)) {
547 			sptmp = malloc(sizeof(struct statfs), M_STATFS,
548 			    M_WAITOK);
549 			*sptmp = *sp;
550 			sptmp->f_fsid.val[0] = sptmp->f_fsid.val[1] = 0;
551 			prison_enforce_statfs(td->td_ucred, mp, sptmp);
552 			sp = sptmp;
553 		} else
554 			sptmp = NULL;
555 		if (bufseg == UIO_SYSSPACE) {
556 			bcopy(sp, sfsp, sizeof(*sp));
557 			free(sptmp, M_STATFS);
558 		} else /* if (bufseg == UIO_USERSPACE) */ {
559 			error = copyout(sp, sfsp, sizeof(*sp));
560 			free(sptmp, M_STATFS);
561 			if (error != 0) {
562 				vfs_unbusy(mp);
563 				return (error);
564 			}
565 		}
566 		sfsp++;
567 		count++;
568 
569 		if (count == maxcount) {
570 			vfs_unbusy(mp);
571 			goto out;
572 		}
573 
574 		mtx_lock(&mountlist_mtx);
575 		nmp = TAILQ_NEXT(mp, mnt_list);
576 		vfs_unbusy(mp);
577 	}
578 	mtx_unlock(&mountlist_mtx);
579 out:
580 	*countp = count;
581 	return (0);
582 }
583 
584 #ifdef COMPAT_FREEBSD4
585 /*
586  * Get old format filesystem statistics.
587  */
588 static void freebsd4_cvtstatfs(struct statfs *, struct ostatfs *);
589 
590 #ifndef _SYS_SYSPROTO_H_
591 struct freebsd4_statfs_args {
592 	char *path;
593 	struct ostatfs *buf;
594 };
595 #endif
596 int
597 freebsd4_statfs(struct thread *td, struct freebsd4_statfs_args *uap)
598 {
599 	struct ostatfs osb;
600 	struct statfs *sfp;
601 	int error;
602 
603 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
604 	error = kern_statfs(td, uap->path, UIO_USERSPACE, sfp);
605 	if (error == 0) {
606 		freebsd4_cvtstatfs(sfp, &osb);
607 		error = copyout(&osb, uap->buf, sizeof(osb));
608 	}
609 	free(sfp, M_STATFS);
610 	return (error);
611 }
612 
613 /*
614  * Get filesystem statistics.
615  */
616 #ifndef _SYS_SYSPROTO_H_
617 struct freebsd4_fstatfs_args {
618 	int fd;
619 	struct ostatfs *buf;
620 };
621 #endif
622 int
623 freebsd4_fstatfs(struct thread *td, struct freebsd4_fstatfs_args *uap)
624 {
625 	struct ostatfs osb;
626 	struct statfs *sfp;
627 	int error;
628 
629 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
630 	error = kern_fstatfs(td, uap->fd, sfp);
631 	if (error == 0) {
632 		freebsd4_cvtstatfs(sfp, &osb);
633 		error = copyout(&osb, uap->buf, sizeof(osb));
634 	}
635 	free(sfp, M_STATFS);
636 	return (error);
637 }
638 
639 /*
640  * Get statistics on all filesystems.
641  */
642 #ifndef _SYS_SYSPROTO_H_
643 struct freebsd4_getfsstat_args {
644 	struct ostatfs *buf;
645 	long bufsize;
646 	int mode;
647 };
648 #endif
649 int
650 freebsd4_getfsstat(struct thread *td, struct freebsd4_getfsstat_args *uap)
651 {
652 	struct statfs *buf, *sp;
653 	struct ostatfs osb;
654 	size_t count, size;
655 	int error;
656 
657 	if (uap->bufsize < 0)
658 		return (EINVAL);
659 	count = uap->bufsize / sizeof(struct ostatfs);
660 	if (count > SIZE_MAX / sizeof(struct statfs))
661 		return (EINVAL);
662 	size = count * sizeof(struct statfs);
663 	error = kern_getfsstat(td, &buf, size, &count, UIO_SYSSPACE,
664 	    uap->mode);
665 	if (error == 0)
666 		td->td_retval[0] = count;
667 	if (size != 0) {
668 		sp = buf;
669 		while (count != 0 && error == 0) {
670 			freebsd4_cvtstatfs(sp, &osb);
671 			error = copyout(&osb, uap->buf, sizeof(osb));
672 			sp++;
673 			uap->buf++;
674 			count--;
675 		}
676 		free(buf, M_STATFS);
677 	}
678 	return (error);
679 }
680 
681 /*
682  * Implement fstatfs() for (NFS) file handles.
683  */
684 #ifndef _SYS_SYSPROTO_H_
685 struct freebsd4_fhstatfs_args {
686 	struct fhandle *u_fhp;
687 	struct ostatfs *buf;
688 };
689 #endif
690 int
691 freebsd4_fhstatfs(struct thread *td, struct freebsd4_fhstatfs_args *uap)
692 {
693 	struct ostatfs osb;
694 	struct statfs *sfp;
695 	fhandle_t fh;
696 	int error;
697 
698 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
699 	if (error != 0)
700 		return (error);
701 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
702 	error = kern_fhstatfs(td, fh, sfp);
703 	if (error == 0) {
704 		freebsd4_cvtstatfs(sfp, &osb);
705 		error = copyout(&osb, uap->buf, sizeof(osb));
706 	}
707 	free(sfp, M_STATFS);
708 	return (error);
709 }
710 
711 /*
712  * Convert a new format statfs structure to an old format statfs structure.
713  */
714 static void
715 freebsd4_cvtstatfs(struct statfs *nsp, struct ostatfs *osp)
716 {
717 
718 	statfs_scale_blocks(nsp, LONG_MAX);
719 	bzero(osp, sizeof(*osp));
720 	osp->f_bsize = nsp->f_bsize;
721 	osp->f_iosize = MIN(nsp->f_iosize, LONG_MAX);
722 	osp->f_blocks = nsp->f_blocks;
723 	osp->f_bfree = nsp->f_bfree;
724 	osp->f_bavail = nsp->f_bavail;
725 	osp->f_files = MIN(nsp->f_files, LONG_MAX);
726 	osp->f_ffree = MIN(nsp->f_ffree, LONG_MAX);
727 	osp->f_owner = nsp->f_owner;
728 	osp->f_type = nsp->f_type;
729 	osp->f_flags = nsp->f_flags;
730 	osp->f_syncwrites = MIN(nsp->f_syncwrites, LONG_MAX);
731 	osp->f_asyncwrites = MIN(nsp->f_asyncwrites, LONG_MAX);
732 	osp->f_syncreads = MIN(nsp->f_syncreads, LONG_MAX);
733 	osp->f_asyncreads = MIN(nsp->f_asyncreads, LONG_MAX);
734 	strlcpy(osp->f_fstypename, nsp->f_fstypename,
735 	    MIN(MFSNAMELEN, OMFSNAMELEN));
736 	strlcpy(osp->f_mntonname, nsp->f_mntonname,
737 	    MIN(MNAMELEN, OMNAMELEN));
738 	strlcpy(osp->f_mntfromname, nsp->f_mntfromname,
739 	    MIN(MNAMELEN, OMNAMELEN));
740 	osp->f_fsid = nsp->f_fsid;
741 }
742 #endif /* COMPAT_FREEBSD4 */
743 
744 #if defined(COMPAT_FREEBSD11)
745 /*
746  * Get old format filesystem statistics.
747  */
748 static void freebsd11_cvtstatfs(struct statfs *, struct freebsd11_statfs *);
749 
750 int
751 freebsd11_statfs(struct thread *td, struct freebsd11_statfs_args *uap)
752 {
753 	struct freebsd11_statfs osb;
754 	struct statfs *sfp;
755 	int error;
756 
757 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
758 	error = kern_statfs(td, uap->path, UIO_USERSPACE, sfp);
759 	if (error == 0) {
760 		freebsd11_cvtstatfs(sfp, &osb);
761 		error = copyout(&osb, uap->buf, sizeof(osb));
762 	}
763 	free(sfp, M_STATFS);
764 	return (error);
765 }
766 
767 /*
768  * Get filesystem statistics.
769  */
770 int
771 freebsd11_fstatfs(struct thread *td, struct freebsd11_fstatfs_args *uap)
772 {
773 	struct freebsd11_statfs osb;
774 	struct statfs *sfp;
775 	int error;
776 
777 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
778 	error = kern_fstatfs(td, uap->fd, sfp);
779 	if (error == 0) {
780 		freebsd11_cvtstatfs(sfp, &osb);
781 		error = copyout(&osb, uap->buf, sizeof(osb));
782 	}
783 	free(sfp, M_STATFS);
784 	return (error);
785 }
786 
787 /*
788  * Get statistics on all filesystems.
789  */
790 int
791 freebsd11_getfsstat(struct thread *td, struct freebsd11_getfsstat_args *uap)
792 {
793 	struct freebsd11_statfs osb;
794 	struct statfs *buf, *sp;
795 	size_t count, size;
796 	int error;
797 
798 	count = uap->bufsize / sizeof(struct ostatfs);
799 	size = count * sizeof(struct statfs);
800 	error = kern_getfsstat(td, &buf, size, &count, UIO_SYSSPACE,
801 	    uap->mode);
802 	if (error == 0)
803 		td->td_retval[0] = count;
804 	if (size > 0) {
805 		sp = buf;
806 		while (count > 0 && error == 0) {
807 			freebsd11_cvtstatfs(sp, &osb);
808 			error = copyout(&osb, uap->buf, sizeof(osb));
809 			sp++;
810 			uap->buf++;
811 			count--;
812 		}
813 		free(buf, M_STATFS);
814 	}
815 	return (error);
816 }
817 
818 /*
819  * Implement fstatfs() for (NFS) file handles.
820  */
821 int
822 freebsd11_fhstatfs(struct thread *td, struct freebsd11_fhstatfs_args *uap)
823 {
824 	struct freebsd11_statfs osb;
825 	struct statfs *sfp;
826 	fhandle_t fh;
827 	int error;
828 
829 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
830 	if (error)
831 		return (error);
832 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
833 	error = kern_fhstatfs(td, fh, sfp);
834 	if (error == 0) {
835 		freebsd11_cvtstatfs(sfp, &osb);
836 		error = copyout(&osb, uap->buf, sizeof(osb));
837 	}
838 	free(sfp, M_STATFS);
839 	return (error);
840 }
841 
842 /*
843  * Convert a new format statfs structure to an old format statfs structure.
844  */
845 static void
846 freebsd11_cvtstatfs(struct statfs *nsp, struct freebsd11_statfs *osp)
847 {
848 
849 	bzero(osp, sizeof(*osp));
850 	osp->f_version = FREEBSD11_STATFS_VERSION;
851 	osp->f_type = nsp->f_type;
852 	osp->f_flags = nsp->f_flags;
853 	osp->f_bsize = nsp->f_bsize;
854 	osp->f_iosize = nsp->f_iosize;
855 	osp->f_blocks = nsp->f_blocks;
856 	osp->f_bfree = nsp->f_bfree;
857 	osp->f_bavail = nsp->f_bavail;
858 	osp->f_files = nsp->f_files;
859 	osp->f_ffree = nsp->f_ffree;
860 	osp->f_syncwrites = nsp->f_syncwrites;
861 	osp->f_asyncwrites = nsp->f_asyncwrites;
862 	osp->f_syncreads = nsp->f_syncreads;
863 	osp->f_asyncreads = nsp->f_asyncreads;
864 	osp->f_namemax = nsp->f_namemax;
865 	osp->f_owner = nsp->f_owner;
866 	osp->f_fsid = nsp->f_fsid;
867 	strlcpy(osp->f_fstypename, nsp->f_fstypename,
868 	    MIN(MFSNAMELEN, sizeof(osp->f_fstypename)));
869 	strlcpy(osp->f_mntonname, nsp->f_mntonname,
870 	    MIN(MNAMELEN, sizeof(osp->f_mntonname)));
871 	strlcpy(osp->f_mntfromname, nsp->f_mntfromname,
872 	    MIN(MNAMELEN, sizeof(osp->f_mntfromname)));
873 }
874 #endif /* COMPAT_FREEBSD11 */
875 
876 /*
877  * Change current working directory to a given file descriptor.
878  */
879 #ifndef _SYS_SYSPROTO_H_
880 struct fchdir_args {
881 	int	fd;
882 };
883 #endif
884 int
885 sys_fchdir(struct thread *td, struct fchdir_args *uap)
886 {
887 	struct vnode *vp, *tdp;
888 	struct mount *mp;
889 	struct file *fp;
890 	int error;
891 
892 	AUDIT_ARG_FD(uap->fd);
893 	error = getvnode_path(td, uap->fd, &cap_fchdir_rights,
894 	    &fp);
895 	if (error != 0)
896 		return (error);
897 	vp = fp->f_vnode;
898 	vref(vp);
899 	fdrop(fp, td);
900 	vn_lock(vp, LK_SHARED | LK_RETRY);
901 	AUDIT_ARG_VNODE1(vp);
902 	error = change_dir(vp, td);
903 	while (!error && (mp = vp->v_mountedhere) != NULL) {
904 		if (vfs_busy(mp, 0))
905 			continue;
906 		error = VFS_ROOT(mp, LK_SHARED, &tdp);
907 		vfs_unbusy(mp);
908 		if (error != 0)
909 			break;
910 		vput(vp);
911 		vp = tdp;
912 	}
913 	if (error != 0) {
914 		vput(vp);
915 		return (error);
916 	}
917 	VOP_UNLOCK(vp);
918 	pwd_chdir(td, vp);
919 	return (0);
920 }
921 
922 /*
923  * Change current working directory (``.'').
924  */
925 #ifndef _SYS_SYSPROTO_H_
926 struct chdir_args {
927 	char	*path;
928 };
929 #endif
930 int
931 sys_chdir(struct thread *td, struct chdir_args *uap)
932 {
933 
934 	return (kern_chdir(td, uap->path, UIO_USERSPACE));
935 }
936 
937 int
938 kern_chdir(struct thread *td, const char *path, enum uio_seg pathseg)
939 {
940 	struct nameidata nd;
941 	int error;
942 
943 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
944 	    pathseg, path, td);
945 	if ((error = namei(&nd)) != 0)
946 		return (error);
947 	if ((error = change_dir(nd.ni_vp, td)) != 0) {
948 		vput(nd.ni_vp);
949 		NDFREE_NOTHING(&nd);
950 		return (error);
951 	}
952 	VOP_UNLOCK(nd.ni_vp);
953 	NDFREE_NOTHING(&nd);
954 	pwd_chdir(td, nd.ni_vp);
955 	return (0);
956 }
957 
958 /*
959  * Change notion of root (``/'') directory.
960  */
961 #ifndef _SYS_SYSPROTO_H_
962 struct chroot_args {
963 	char	*path;
964 };
965 #endif
966 int
967 sys_chroot(struct thread *td, struct chroot_args *uap)
968 {
969 	struct nameidata nd;
970 	int error;
971 
972 	error = priv_check(td, PRIV_VFS_CHROOT);
973 	if (error != 0)
974 		return (error);
975 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
976 	    UIO_USERSPACE, uap->path, td);
977 	error = namei(&nd);
978 	if (error != 0)
979 		goto error;
980 	error = change_dir(nd.ni_vp, td);
981 	if (error != 0)
982 		goto e_vunlock;
983 #ifdef MAC
984 	error = mac_vnode_check_chroot(td->td_ucred, nd.ni_vp);
985 	if (error != 0)
986 		goto e_vunlock;
987 #endif
988 	VOP_UNLOCK(nd.ni_vp);
989 	error = pwd_chroot(td, nd.ni_vp);
990 	vrele(nd.ni_vp);
991 	NDFREE_NOTHING(&nd);
992 	return (error);
993 e_vunlock:
994 	vput(nd.ni_vp);
995 error:
996 	NDFREE_NOTHING(&nd);
997 	return (error);
998 }
999 
1000 /*
1001  * Common routine for chroot and chdir.  Callers must provide a locked vnode
1002  * instance.
1003  */
1004 int
1005 change_dir(struct vnode *vp, struct thread *td)
1006 {
1007 #ifdef MAC
1008 	int error;
1009 #endif
1010 
1011 	ASSERT_VOP_LOCKED(vp, "change_dir(): vp not locked");
1012 	if (vp->v_type != VDIR)
1013 		return (ENOTDIR);
1014 #ifdef MAC
1015 	error = mac_vnode_check_chdir(td->td_ucred, vp);
1016 	if (error != 0)
1017 		return (error);
1018 #endif
1019 	return (VOP_ACCESS(vp, VEXEC, td->td_ucred, td));
1020 }
1021 
1022 static __inline void
1023 flags_to_rights(int flags, cap_rights_t *rightsp)
1024 {
1025 	if (flags & O_EXEC) {
1026 		cap_rights_set_one(rightsp, CAP_FEXECVE);
1027 		if (flags & O_PATH)
1028 			return;
1029 	} else {
1030 		switch ((flags & O_ACCMODE)) {
1031 		case O_RDONLY:
1032 			cap_rights_set_one(rightsp, CAP_READ);
1033 			break;
1034 		case O_RDWR:
1035 			cap_rights_set_one(rightsp, CAP_READ);
1036 			/* FALLTHROUGH */
1037 		case O_WRONLY:
1038 			cap_rights_set_one(rightsp, CAP_WRITE);
1039 			if (!(flags & (O_APPEND | O_TRUNC)))
1040 				cap_rights_set_one(rightsp, CAP_SEEK);
1041 			break;
1042 		}
1043 	}
1044 
1045 	if (flags & O_CREAT)
1046 		cap_rights_set_one(rightsp, CAP_CREATE);
1047 
1048 	if (flags & O_TRUNC)
1049 		cap_rights_set_one(rightsp, CAP_FTRUNCATE);
1050 
1051 	if (flags & (O_SYNC | O_FSYNC))
1052 		cap_rights_set_one(rightsp, CAP_FSYNC);
1053 
1054 	if (flags & (O_EXLOCK | O_SHLOCK))
1055 		cap_rights_set_one(rightsp, CAP_FLOCK);
1056 }
1057 
1058 /*
1059  * Check permissions, allocate an open file structure, and call the device
1060  * open routine if any.
1061  */
1062 #ifndef _SYS_SYSPROTO_H_
1063 struct open_args {
1064 	char	*path;
1065 	int	flags;
1066 	int	mode;
1067 };
1068 #endif
1069 int
1070 sys_open(struct thread *td, struct open_args *uap)
1071 {
1072 
1073 	return (kern_openat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1074 	    uap->flags, uap->mode));
1075 }
1076 
1077 #ifndef _SYS_SYSPROTO_H_
1078 struct openat_args {
1079 	int	fd;
1080 	char	*path;
1081 	int	flag;
1082 	int	mode;
1083 };
1084 #endif
1085 int
1086 sys_openat(struct thread *td, struct openat_args *uap)
1087 {
1088 
1089 	AUDIT_ARG_FD(uap->fd);
1090 	return (kern_openat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
1091 	    uap->mode));
1092 }
1093 
1094 int
1095 kern_openat(struct thread *td, int fd, const char *path, enum uio_seg pathseg,
1096     int flags, int mode)
1097 {
1098 	struct proc *p = td->td_proc;
1099 	struct filedesc *fdp;
1100 	struct pwddesc *pdp;
1101 	struct file *fp;
1102 	struct vnode *vp;
1103 	struct nameidata nd;
1104 	cap_rights_t rights;
1105 	int cmode, error, indx;
1106 
1107 	indx = -1;
1108 	fdp = p->p_fd;
1109 	pdp = p->p_pd;
1110 
1111 	AUDIT_ARG_FFLAGS(flags);
1112 	AUDIT_ARG_MODE(mode);
1113 	cap_rights_init_one(&rights, CAP_LOOKUP);
1114 	flags_to_rights(flags, &rights);
1115 
1116 	/*
1117 	 * Only one of the O_EXEC, O_RDONLY, O_WRONLY and O_RDWR flags
1118 	 * may be specified.  On the other hand, for O_PATH any mode
1119 	 * except O_EXEC is ignored.
1120 	 */
1121 	if ((flags & O_PATH) != 0) {
1122 		flags &= ~(O_CREAT | O_ACCMODE);
1123 	} else if ((flags & O_EXEC) != 0) {
1124 		if (flags & O_ACCMODE)
1125 			return (EINVAL);
1126 	} else if ((flags & O_ACCMODE) == O_ACCMODE) {
1127 		return (EINVAL);
1128 	} else {
1129 		flags = FFLAGS(flags);
1130 	}
1131 
1132 	/*
1133 	 * Allocate a file structure. The descriptor to reference it
1134 	 * is allocated and used by finstall_refed() below.
1135 	 */
1136 	error = falloc_noinstall(td, &fp);
1137 	if (error != 0)
1138 		return (error);
1139 	/* Set the flags early so the finit in devfs can pick them up. */
1140 	fp->f_flag = flags & FMASK;
1141 	cmode = ((mode & ~pdp->pd_cmask) & ALLPERMS) & ~S_ISTXT;
1142 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, fd,
1143 	    &rights, td);
1144 	td->td_dupfd = -1;		/* XXX check for fdopen */
1145 	error = vn_open(&nd, &flags, cmode, fp);
1146 	if (error != 0) {
1147 		/*
1148 		 * If the vn_open replaced the method vector, something
1149 		 * wonderous happened deep below and we just pass it up
1150 		 * pretending we know what we do.
1151 		 */
1152 		if (error == ENXIO && fp->f_ops != &badfileops) {
1153 			MPASS((flags & O_PATH) == 0);
1154 			goto success;
1155 		}
1156 
1157 		/*
1158 		 * Handle special fdopen() case. bleh.
1159 		 *
1160 		 * Don't do this for relative (capability) lookups; we don't
1161 		 * understand exactly what would happen, and we don't think
1162 		 * that it ever should.
1163 		 */
1164 		if ((nd.ni_resflags & NIRES_STRICTREL) == 0 &&
1165 		    (error == ENODEV || error == ENXIO) &&
1166 		    td->td_dupfd >= 0) {
1167 			error = dupfdopen(td, fdp, td->td_dupfd, flags, error,
1168 			    &indx);
1169 			if (error == 0)
1170 				goto success;
1171 		}
1172 
1173 		goto bad;
1174 	}
1175 	td->td_dupfd = 0;
1176 	NDFREE(&nd, NDF_ONLY_PNBUF);
1177 	vp = nd.ni_vp;
1178 
1179 	/*
1180 	 * Store the vnode, for any f_type. Typically, the vnode use
1181 	 * count is decremented by direct call to vn_closefile() for
1182 	 * files that switched type in the cdevsw fdopen() method.
1183 	 */
1184 	fp->f_vnode = vp;
1185 
1186 	/*
1187 	 * If the file wasn't claimed by devfs bind it to the normal
1188 	 * vnode operations here.
1189 	 */
1190 	if (fp->f_ops == &badfileops) {
1191 		KASSERT(vp->v_type != VFIFO || (flags & O_PATH) != 0,
1192 		    ("Unexpected fifo fp %p vp %p", fp, vp));
1193 		if ((flags & O_PATH) != 0) {
1194 			finit(fp, (flags & FMASK) | (fp->f_flag & FKQALLOWED),
1195 			    DTYPE_VNODE, NULL, &path_fileops);
1196 			vhold(vp);
1197 			vunref(vp);
1198 		} else {
1199 			finit_vnode(fp, flags, NULL, &vnops);
1200 		}
1201 	}
1202 
1203 	VOP_UNLOCK(vp);
1204 	if (flags & O_TRUNC) {
1205 		error = fo_truncate(fp, 0, td->td_ucred, td);
1206 		if (error != 0)
1207 			goto bad;
1208 	}
1209 success:
1210 	/*
1211 	 * If we haven't already installed the FD (for dupfdopen), do so now.
1212 	 */
1213 	if (indx == -1) {
1214 		struct filecaps *fcaps;
1215 
1216 #ifdef CAPABILITIES
1217 		if ((nd.ni_resflags & NIRES_STRICTREL) != 0)
1218 			fcaps = &nd.ni_filecaps;
1219 		else
1220 #endif
1221 			fcaps = NULL;
1222 		error = finstall_refed(td, fp, &indx, flags, fcaps);
1223 		/* On success finstall_refed() consumes fcaps. */
1224 		if (error != 0) {
1225 			filecaps_free(&nd.ni_filecaps);
1226 			goto bad;
1227 		}
1228 	} else {
1229 		filecaps_free(&nd.ni_filecaps);
1230 		falloc_abort(td, fp);
1231 	}
1232 
1233 	td->td_retval[0] = indx;
1234 	return (0);
1235 bad:
1236 	KASSERT(indx == -1, ("indx=%d, should be -1", indx));
1237 	falloc_abort(td, fp);
1238 	return (error);
1239 }
1240 
1241 #ifdef COMPAT_43
1242 /*
1243  * Create a file.
1244  */
1245 #ifndef _SYS_SYSPROTO_H_
1246 struct ocreat_args {
1247 	char	*path;
1248 	int	mode;
1249 };
1250 #endif
1251 int
1252 ocreat(struct thread *td, struct ocreat_args *uap)
1253 {
1254 
1255 	return (kern_openat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1256 	    O_WRONLY | O_CREAT | O_TRUNC, uap->mode));
1257 }
1258 #endif /* COMPAT_43 */
1259 
1260 /*
1261  * Create a special file.
1262  */
1263 #ifndef _SYS_SYSPROTO_H_
1264 struct mknodat_args {
1265 	int	fd;
1266 	char	*path;
1267 	mode_t	mode;
1268 	dev_t	dev;
1269 };
1270 #endif
1271 int
1272 sys_mknodat(struct thread *td, struct mknodat_args *uap)
1273 {
1274 
1275 	return (kern_mknodat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode,
1276 	    uap->dev));
1277 }
1278 
1279 #if defined(COMPAT_FREEBSD11)
1280 int
1281 freebsd11_mknod(struct thread *td,
1282     struct freebsd11_mknod_args *uap)
1283 {
1284 
1285 	return (kern_mknodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1286 	    uap->mode, uap->dev));
1287 }
1288 
1289 int
1290 freebsd11_mknodat(struct thread *td,
1291     struct freebsd11_mknodat_args *uap)
1292 {
1293 
1294 	return (kern_mknodat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode,
1295 	    uap->dev));
1296 }
1297 #endif /* COMPAT_FREEBSD11 */
1298 
1299 int
1300 kern_mknodat(struct thread *td, int fd, const char *path, enum uio_seg pathseg,
1301     int mode, dev_t dev)
1302 {
1303 	struct vnode *vp;
1304 	struct mount *mp;
1305 	struct vattr vattr;
1306 	struct nameidata nd;
1307 	int error, whiteout = 0;
1308 
1309 	AUDIT_ARG_MODE(mode);
1310 	AUDIT_ARG_DEV(dev);
1311 	switch (mode & S_IFMT) {
1312 	case S_IFCHR:
1313 	case S_IFBLK:
1314 		error = priv_check(td, PRIV_VFS_MKNOD_DEV);
1315 		if (error == 0 && dev == VNOVAL)
1316 			error = EINVAL;
1317 		break;
1318 	case S_IFWHT:
1319 		error = priv_check(td, PRIV_VFS_MKNOD_WHT);
1320 		break;
1321 	case S_IFIFO:
1322 		if (dev == 0)
1323 			return (kern_mkfifoat(td, fd, path, pathseg, mode));
1324 		/* FALLTHROUGH */
1325 	default:
1326 		error = EINVAL;
1327 		break;
1328 	}
1329 	if (error != 0)
1330 		return (error);
1331 restart:
1332 	bwillwrite();
1333 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
1334 	    NOCACHE, pathseg, path, fd, &cap_mknodat_rights,
1335 	    td);
1336 	if ((error = namei(&nd)) != 0)
1337 		return (error);
1338 	vp = nd.ni_vp;
1339 	if (vp != NULL) {
1340 		NDFREE(&nd, NDF_ONLY_PNBUF);
1341 		if (vp == nd.ni_dvp)
1342 			vrele(nd.ni_dvp);
1343 		else
1344 			vput(nd.ni_dvp);
1345 		vrele(vp);
1346 		return (EEXIST);
1347 	} else {
1348 		VATTR_NULL(&vattr);
1349 		vattr.va_mode = (mode & ALLPERMS) &
1350 		    ~td->td_proc->p_pd->pd_cmask;
1351 		vattr.va_rdev = dev;
1352 		whiteout = 0;
1353 
1354 		switch (mode & S_IFMT) {
1355 		case S_IFCHR:
1356 			vattr.va_type = VCHR;
1357 			break;
1358 		case S_IFBLK:
1359 			vattr.va_type = VBLK;
1360 			break;
1361 		case S_IFWHT:
1362 			whiteout = 1;
1363 			break;
1364 		default:
1365 			panic("kern_mknod: invalid mode");
1366 		}
1367 	}
1368 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1369 		NDFREE(&nd, NDF_ONLY_PNBUF);
1370 		vput(nd.ni_dvp);
1371 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1372 			return (error);
1373 		goto restart;
1374 	}
1375 #ifdef MAC
1376 	if (error == 0 && !whiteout)
1377 		error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp,
1378 		    &nd.ni_cnd, &vattr);
1379 #endif
1380 	if (error == 0) {
1381 		if (whiteout)
1382 			error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE);
1383 		else {
1384 			error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp,
1385 						&nd.ni_cnd, &vattr);
1386 		}
1387 	}
1388 	VOP_VPUT_PAIR(nd.ni_dvp, error == 0 && !whiteout ? &nd.ni_vp : NULL,
1389 	    true);
1390 	vn_finished_write(mp);
1391 	NDFREE(&nd, NDF_ONLY_PNBUF);
1392 	if (error == ERELOOKUP)
1393 		goto restart;
1394 	return (error);
1395 }
1396 
1397 /*
1398  * Create a named pipe.
1399  */
1400 #ifndef _SYS_SYSPROTO_H_
1401 struct mkfifo_args {
1402 	char	*path;
1403 	int	mode;
1404 };
1405 #endif
1406 int
1407 sys_mkfifo(struct thread *td, struct mkfifo_args *uap)
1408 {
1409 
1410 	return (kern_mkfifoat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1411 	    uap->mode));
1412 }
1413 
1414 #ifndef _SYS_SYSPROTO_H_
1415 struct mkfifoat_args {
1416 	int	fd;
1417 	char	*path;
1418 	mode_t	mode;
1419 };
1420 #endif
1421 int
1422 sys_mkfifoat(struct thread *td, struct mkfifoat_args *uap)
1423 {
1424 
1425 	return (kern_mkfifoat(td, uap->fd, uap->path, UIO_USERSPACE,
1426 	    uap->mode));
1427 }
1428 
1429 int
1430 kern_mkfifoat(struct thread *td, int fd, const char *path,
1431     enum uio_seg pathseg, int mode)
1432 {
1433 	struct mount *mp;
1434 	struct vattr vattr;
1435 	struct nameidata nd;
1436 	int error;
1437 
1438 	AUDIT_ARG_MODE(mode);
1439 restart:
1440 	bwillwrite();
1441 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
1442 	    NOCACHE, pathseg, path, fd, &cap_mkfifoat_rights,
1443 	    td);
1444 	if ((error = namei(&nd)) != 0)
1445 		return (error);
1446 	if (nd.ni_vp != NULL) {
1447 		NDFREE(&nd, NDF_ONLY_PNBUF);
1448 		if (nd.ni_vp == nd.ni_dvp)
1449 			vrele(nd.ni_dvp);
1450 		else
1451 			vput(nd.ni_dvp);
1452 		vrele(nd.ni_vp);
1453 		return (EEXIST);
1454 	}
1455 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1456 		NDFREE(&nd, NDF_ONLY_PNBUF);
1457 		vput(nd.ni_dvp);
1458 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1459 			return (error);
1460 		goto restart;
1461 	}
1462 	VATTR_NULL(&vattr);
1463 	vattr.va_type = VFIFO;
1464 	vattr.va_mode = (mode & ALLPERMS) & ~td->td_proc->p_pd->pd_cmask;
1465 #ifdef MAC
1466 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1467 	    &vattr);
1468 	if (error != 0)
1469 		goto out;
1470 #endif
1471 	error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
1472 #ifdef MAC
1473 out:
1474 #endif
1475 	VOP_VPUT_PAIR(nd.ni_dvp, error == 0 ? &nd.ni_vp : NULL, true);
1476 	vn_finished_write(mp);
1477 	NDFREE(&nd, NDF_ONLY_PNBUF);
1478 	if (error == ERELOOKUP)
1479 		goto restart;
1480 	return (error);
1481 }
1482 
1483 /*
1484  * Make a hard file link.
1485  */
1486 #ifndef _SYS_SYSPROTO_H_
1487 struct link_args {
1488 	char	*path;
1489 	char	*link;
1490 };
1491 #endif
1492 int
1493 sys_link(struct thread *td, struct link_args *uap)
1494 {
1495 
1496 	return (kern_linkat(td, AT_FDCWD, AT_FDCWD, uap->path, uap->link,
1497 	    UIO_USERSPACE, AT_SYMLINK_FOLLOW));
1498 }
1499 
1500 #ifndef _SYS_SYSPROTO_H_
1501 struct linkat_args {
1502 	int	fd1;
1503 	char	*path1;
1504 	int	fd2;
1505 	char	*path2;
1506 	int	flag;
1507 };
1508 #endif
1509 int
1510 sys_linkat(struct thread *td, struct linkat_args *uap)
1511 {
1512 	int flag;
1513 
1514 	flag = uap->flag;
1515 	if ((flag & ~(AT_SYMLINK_FOLLOW | AT_RESOLVE_BENEATH |
1516 	    AT_EMPTY_PATH)) != 0)
1517 		return (EINVAL);
1518 
1519 	return (kern_linkat(td, uap->fd1, uap->fd2, uap->path1, uap->path2,
1520 	    UIO_USERSPACE, flag));
1521 }
1522 
1523 int hardlink_check_uid = 0;
1524 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_uid, CTLFLAG_RW,
1525     &hardlink_check_uid, 0,
1526     "Unprivileged processes cannot create hard links to files owned by other "
1527     "users");
1528 static int hardlink_check_gid = 0;
1529 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_gid, CTLFLAG_RW,
1530     &hardlink_check_gid, 0,
1531     "Unprivileged processes cannot create hard links to files owned by other "
1532     "groups");
1533 
1534 static int
1535 can_hardlink(struct vnode *vp, struct ucred *cred)
1536 {
1537 	struct vattr va;
1538 	int error;
1539 
1540 	if (!hardlink_check_uid && !hardlink_check_gid)
1541 		return (0);
1542 
1543 	error = VOP_GETATTR(vp, &va, cred);
1544 	if (error != 0)
1545 		return (error);
1546 
1547 	if (hardlink_check_uid && cred->cr_uid != va.va_uid) {
1548 		error = priv_check_cred(cred, PRIV_VFS_LINK);
1549 		if (error != 0)
1550 			return (error);
1551 	}
1552 
1553 	if (hardlink_check_gid && !groupmember(va.va_gid, cred)) {
1554 		error = priv_check_cred(cred, PRIV_VFS_LINK);
1555 		if (error != 0)
1556 			return (error);
1557 	}
1558 
1559 	return (0);
1560 }
1561 
1562 int
1563 kern_linkat(struct thread *td, int fd1, int fd2, const char *path1,
1564     const char *path2, enum uio_seg segflag, int flag)
1565 {
1566 	struct nameidata nd;
1567 	int error;
1568 
1569 	do {
1570 		bwillwrite();
1571 		NDINIT_ATRIGHTS(&nd, LOOKUP, AUDITVNODE1 | at2cnpflags(flag,
1572 		    AT_SYMLINK_FOLLOW | AT_RESOLVE_BENEATH | AT_EMPTY_PATH),
1573 		    segflag, path1, fd1, &cap_linkat_source_rights, td);
1574 		if ((error = namei(&nd)) != 0)
1575 			return (error);
1576 		NDFREE(&nd, NDF_ONLY_PNBUF);
1577 		if ((nd.ni_resflags & NIRES_EMPTYPATH) != 0) {
1578 			error = priv_check(td, PRIV_VFS_FHOPEN);
1579 			if (error != 0) {
1580 				vrele(nd.ni_vp);
1581 				return (error);
1582 			}
1583 		}
1584 		error = kern_linkat_vp(td, nd.ni_vp, fd2, path2, segflag);
1585 	} while (error ==  EAGAIN || error == ERELOOKUP);
1586 	return (error);
1587 }
1588 
1589 static int
1590 kern_linkat_vp(struct thread *td, struct vnode *vp, int fd, const char *path,
1591     enum uio_seg segflag)
1592 {
1593 	struct nameidata nd;
1594 	struct mount *mp;
1595 	int error;
1596 
1597 	if (vp->v_type == VDIR) {
1598 		vrele(vp);
1599 		return (EPERM);		/* POSIX */
1600 	}
1601 	NDINIT_ATRIGHTS(&nd, CREATE,
1602 	    LOCKPARENT | SAVENAME | AUDITVNODE2 | NOCACHE, segflag, path, fd,
1603 	    &cap_linkat_target_rights, td);
1604 	if ((error = namei(&nd)) == 0) {
1605 		if (nd.ni_vp != NULL) {
1606 			NDFREE(&nd, NDF_ONLY_PNBUF);
1607 			if (nd.ni_dvp == nd.ni_vp)
1608 				vrele(nd.ni_dvp);
1609 			else
1610 				vput(nd.ni_dvp);
1611 			vrele(nd.ni_vp);
1612 			vrele(vp);
1613 			return (EEXIST);
1614 		} else if (nd.ni_dvp->v_mount != vp->v_mount) {
1615 			/*
1616 			 * Cross-device link.  No need to recheck
1617 			 * vp->v_type, since it cannot change, except
1618 			 * to VBAD.
1619 			 */
1620 			NDFREE(&nd, NDF_ONLY_PNBUF);
1621 			vput(nd.ni_dvp);
1622 			vrele(vp);
1623 			return (EXDEV);
1624 		} else if ((error = vn_lock(vp, LK_EXCLUSIVE)) == 0) {
1625 			error = can_hardlink(vp, td->td_ucred);
1626 #ifdef MAC
1627 			if (error == 0)
1628 				error = mac_vnode_check_link(td->td_ucred,
1629 				    nd.ni_dvp, vp, &nd.ni_cnd);
1630 #endif
1631 			if (error != 0) {
1632 				vput(vp);
1633 				vput(nd.ni_dvp);
1634 				NDFREE(&nd, NDF_ONLY_PNBUF);
1635 				return (error);
1636 			}
1637 			error = vn_start_write(vp, &mp, V_NOWAIT);
1638 			if (error != 0) {
1639 				vput(vp);
1640 				vput(nd.ni_dvp);
1641 				NDFREE(&nd, NDF_ONLY_PNBUF);
1642 				error = vn_start_write(NULL, &mp,
1643 				    V_XSLEEP | PCATCH);
1644 				if (error != 0)
1645 					return (error);
1646 				return (EAGAIN);
1647 			}
1648 			error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
1649 			VOP_VPUT_PAIR(nd.ni_dvp, &vp, true);
1650 			vn_finished_write(mp);
1651 			NDFREE(&nd, NDF_ONLY_PNBUF);
1652 			vp = NULL;
1653 		} else {
1654 			vput(nd.ni_dvp);
1655 			NDFREE(&nd, NDF_ONLY_PNBUF);
1656 			vrele(vp);
1657 			return (EAGAIN);
1658 		}
1659 	}
1660 	if (vp != NULL)
1661 		vrele(vp);
1662 	return (error);
1663 }
1664 
1665 /*
1666  * Make a symbolic link.
1667  */
1668 #ifndef _SYS_SYSPROTO_H_
1669 struct symlink_args {
1670 	char	*path;
1671 	char	*link;
1672 };
1673 #endif
1674 int
1675 sys_symlink(struct thread *td, struct symlink_args *uap)
1676 {
1677 
1678 	return (kern_symlinkat(td, uap->path, AT_FDCWD, uap->link,
1679 	    UIO_USERSPACE));
1680 }
1681 
1682 #ifndef _SYS_SYSPROTO_H_
1683 struct symlinkat_args {
1684 	char	*path;
1685 	int	fd;
1686 	char	*path2;
1687 };
1688 #endif
1689 int
1690 sys_symlinkat(struct thread *td, struct symlinkat_args *uap)
1691 {
1692 
1693 	return (kern_symlinkat(td, uap->path1, uap->fd, uap->path2,
1694 	    UIO_USERSPACE));
1695 }
1696 
1697 int
1698 kern_symlinkat(struct thread *td, const char *path1, int fd, const char *path2,
1699     enum uio_seg segflg)
1700 {
1701 	struct mount *mp;
1702 	struct vattr vattr;
1703 	const char *syspath;
1704 	char *tmppath;
1705 	struct nameidata nd;
1706 	int error;
1707 
1708 	if (segflg == UIO_SYSSPACE) {
1709 		syspath = path1;
1710 	} else {
1711 		tmppath = uma_zalloc(namei_zone, M_WAITOK);
1712 		if ((error = copyinstr(path1, tmppath, MAXPATHLEN, NULL)) != 0)
1713 			goto out;
1714 		syspath = tmppath;
1715 	}
1716 	AUDIT_ARG_TEXT(syspath);
1717 restart:
1718 	bwillwrite();
1719 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
1720 	    NOCACHE, segflg, path2, fd, &cap_symlinkat_rights,
1721 	    td);
1722 	if ((error = namei(&nd)) != 0)
1723 		goto out;
1724 	if (nd.ni_vp) {
1725 		NDFREE(&nd, NDF_ONLY_PNBUF);
1726 		if (nd.ni_vp == nd.ni_dvp)
1727 			vrele(nd.ni_dvp);
1728 		else
1729 			vput(nd.ni_dvp);
1730 		vrele(nd.ni_vp);
1731 		nd.ni_vp = NULL;
1732 		error = EEXIST;
1733 		goto out;
1734 	}
1735 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1736 		NDFREE(&nd, NDF_ONLY_PNBUF);
1737 		vput(nd.ni_dvp);
1738 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1739 			goto out;
1740 		goto restart;
1741 	}
1742 	VATTR_NULL(&vattr);
1743 	vattr.va_mode = ACCESSPERMS &~ td->td_proc->p_pd->pd_cmask;
1744 #ifdef MAC
1745 	vattr.va_type = VLNK;
1746 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1747 	    &vattr);
1748 	if (error != 0)
1749 		goto out2;
1750 #endif
1751 	error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, syspath);
1752 #ifdef MAC
1753 out2:
1754 #endif
1755 	VOP_VPUT_PAIR(nd.ni_dvp, error == 0 ? &nd.ni_vp : NULL, true);
1756 	vn_finished_write(mp);
1757 	NDFREE(&nd, NDF_ONLY_PNBUF);
1758 	if (error == ERELOOKUP)
1759 		goto restart;
1760 out:
1761 	if (segflg != UIO_SYSSPACE)
1762 		uma_zfree(namei_zone, tmppath);
1763 	return (error);
1764 }
1765 
1766 /*
1767  * Delete a whiteout from the filesystem.
1768  */
1769 #ifndef _SYS_SYSPROTO_H_
1770 struct undelete_args {
1771 	char *path;
1772 };
1773 #endif
1774 int
1775 sys_undelete(struct thread *td, struct undelete_args *uap)
1776 {
1777 	struct mount *mp;
1778 	struct nameidata nd;
1779 	int error;
1780 
1781 restart:
1782 	bwillwrite();
1783 	NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | AUDITVNODE1,
1784 	    UIO_USERSPACE, uap->path, td);
1785 	error = namei(&nd);
1786 	if (error != 0)
1787 		return (error);
1788 
1789 	if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) {
1790 		NDFREE(&nd, NDF_ONLY_PNBUF);
1791 		if (nd.ni_vp == nd.ni_dvp)
1792 			vrele(nd.ni_dvp);
1793 		else
1794 			vput(nd.ni_dvp);
1795 		if (nd.ni_vp)
1796 			vrele(nd.ni_vp);
1797 		return (EEXIST);
1798 	}
1799 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1800 		NDFREE(&nd, NDF_ONLY_PNBUF);
1801 		vput(nd.ni_dvp);
1802 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1803 			return (error);
1804 		goto restart;
1805 	}
1806 	error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE);
1807 	NDFREE(&nd, NDF_ONLY_PNBUF);
1808 	vput(nd.ni_dvp);
1809 	vn_finished_write(mp);
1810 	if (error == ERELOOKUP)
1811 		goto restart;
1812 	return (error);
1813 }
1814 
1815 /*
1816  * Delete a name from the filesystem.
1817  */
1818 #ifndef _SYS_SYSPROTO_H_
1819 struct unlink_args {
1820 	char	*path;
1821 };
1822 #endif
1823 int
1824 sys_unlink(struct thread *td, struct unlink_args *uap)
1825 {
1826 
1827 	return (kern_funlinkat(td, AT_FDCWD, uap->path, FD_NONE, UIO_USERSPACE,
1828 	    0, 0));
1829 }
1830 
1831 static int
1832 kern_funlinkat_ex(struct thread *td, int dfd, const char *path, int fd,
1833     int flag, enum uio_seg pathseg, ino_t oldinum)
1834 {
1835 
1836 	if ((flag & ~(AT_REMOVEDIR | AT_RESOLVE_BENEATH)) != 0)
1837 		return (EINVAL);
1838 
1839 	if ((flag & AT_REMOVEDIR) != 0)
1840 		return (kern_frmdirat(td, dfd, path, fd, UIO_USERSPACE, 0));
1841 
1842 	return (kern_funlinkat(td, dfd, path, fd, UIO_USERSPACE, 0, 0));
1843 }
1844 
1845 #ifndef _SYS_SYSPROTO_H_
1846 struct unlinkat_args {
1847 	int	fd;
1848 	char	*path;
1849 	int	flag;
1850 };
1851 #endif
1852 int
1853 sys_unlinkat(struct thread *td, struct unlinkat_args *uap)
1854 {
1855 
1856 	return (kern_funlinkat_ex(td, uap->fd, uap->path, FD_NONE, uap->flag,
1857 	    UIO_USERSPACE, 0));
1858 }
1859 
1860 #ifndef _SYS_SYSPROTO_H_
1861 struct funlinkat_args {
1862 	int		dfd;
1863 	const char	*path;
1864 	int		fd;
1865 	int		flag;
1866 };
1867 #endif
1868 int
1869 sys_funlinkat(struct thread *td, struct funlinkat_args *uap)
1870 {
1871 
1872 	return (kern_funlinkat_ex(td, uap->dfd, uap->path, uap->fd, uap->flag,
1873 	    UIO_USERSPACE, 0));
1874 }
1875 
1876 int
1877 kern_funlinkat(struct thread *td, int dfd, const char *path, int fd,
1878     enum uio_seg pathseg, int flag, ino_t oldinum)
1879 {
1880 	struct mount *mp;
1881 	struct file *fp;
1882 	struct vnode *vp;
1883 	struct nameidata nd;
1884 	struct stat sb;
1885 	int error;
1886 
1887 	fp = NULL;
1888 	if (fd != FD_NONE) {
1889 		error = getvnode_path(td, fd, &cap_no_rights, &fp);
1890 		if (error != 0)
1891 			return (error);
1892 	}
1893 
1894 restart:
1895 	bwillwrite();
1896 	NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1 |
1897 	    at2cnpflags(flag, AT_RESOLVE_BENEATH),
1898 	    pathseg, path, dfd, &cap_unlinkat_rights, td);
1899 	if ((error = namei(&nd)) != 0) {
1900 		if (error == EINVAL)
1901 			error = EPERM;
1902 		goto fdout;
1903 	}
1904 	vp = nd.ni_vp;
1905 	if (vp->v_type == VDIR && oldinum == 0) {
1906 		error = EPERM;		/* POSIX */
1907 	} else if (oldinum != 0 &&
1908 	    ((error = VOP_STAT(vp, &sb, td->td_ucred, NOCRED, td)) == 0) &&
1909 	    sb.st_ino != oldinum) {
1910 		error = EIDRM;	/* Identifier removed */
1911 	} else if (fp != NULL && fp->f_vnode != vp) {
1912 		if (VN_IS_DOOMED(fp->f_vnode))
1913 			error = EBADF;
1914 		else
1915 			error = EDEADLK;
1916 	} else {
1917 		/*
1918 		 * The root of a mounted filesystem cannot be deleted.
1919 		 *
1920 		 * XXX: can this only be a VDIR case?
1921 		 */
1922 		if (vp->v_vflag & VV_ROOT)
1923 			error = EBUSY;
1924 	}
1925 	if (error == 0) {
1926 		if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1927 			NDFREE(&nd, NDF_ONLY_PNBUF);
1928 			vput(nd.ni_dvp);
1929 			if (vp == nd.ni_dvp)
1930 				vrele(vp);
1931 			else
1932 				vput(vp);
1933 			if ((error = vn_start_write(NULL, &mp,
1934 			    V_XSLEEP | PCATCH)) != 0) {
1935 				goto fdout;
1936 			}
1937 			goto restart;
1938 		}
1939 #ifdef MAC
1940 		error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
1941 		    &nd.ni_cnd);
1942 		if (error != 0)
1943 			goto out;
1944 #endif
1945 		vfs_notify_upper(vp, VFS_NOTIFY_UPPER_UNLINK);
1946 		error = VOP_REMOVE(nd.ni_dvp, vp, &nd.ni_cnd);
1947 #ifdef MAC
1948 out:
1949 #endif
1950 		vn_finished_write(mp);
1951 	}
1952 	NDFREE(&nd, NDF_ONLY_PNBUF);
1953 	vput(nd.ni_dvp);
1954 	if (vp == nd.ni_dvp)
1955 		vrele(vp);
1956 	else
1957 		vput(vp);
1958 	if (error == ERELOOKUP)
1959 		goto restart;
1960 fdout:
1961 	if (fp != NULL)
1962 		fdrop(fp, td);
1963 	return (error);
1964 }
1965 
1966 /*
1967  * Reposition read/write file offset.
1968  */
1969 #ifndef _SYS_SYSPROTO_H_
1970 struct lseek_args {
1971 	int	fd;
1972 	int	pad;
1973 	off_t	offset;
1974 	int	whence;
1975 };
1976 #endif
1977 int
1978 sys_lseek(struct thread *td, struct lseek_args *uap)
1979 {
1980 
1981 	return (kern_lseek(td, uap->fd, uap->offset, uap->whence));
1982 }
1983 
1984 int
1985 kern_lseek(struct thread *td, int fd, off_t offset, int whence)
1986 {
1987 	struct file *fp;
1988 	int error;
1989 
1990 	AUDIT_ARG_FD(fd);
1991 	error = fget(td, fd, &cap_seek_rights, &fp);
1992 	if (error != 0)
1993 		return (error);
1994 	error = (fp->f_ops->fo_flags & DFLAG_SEEKABLE) != 0 ?
1995 	    fo_seek(fp, offset, whence, td) : ESPIPE;
1996 	fdrop(fp, td);
1997 	return (error);
1998 }
1999 
2000 #if defined(COMPAT_43)
2001 /*
2002  * Reposition read/write file offset.
2003  */
2004 #ifndef _SYS_SYSPROTO_H_
2005 struct olseek_args {
2006 	int	fd;
2007 	long	offset;
2008 	int	whence;
2009 };
2010 #endif
2011 int
2012 olseek(struct thread *td, struct olseek_args *uap)
2013 {
2014 
2015 	return (kern_lseek(td, uap->fd, uap->offset, uap->whence));
2016 }
2017 #endif /* COMPAT_43 */
2018 
2019 #if defined(COMPAT_FREEBSD6)
2020 /* Version with the 'pad' argument */
2021 int
2022 freebsd6_lseek(struct thread *td, struct freebsd6_lseek_args *uap)
2023 {
2024 
2025 	return (kern_lseek(td, uap->fd, uap->offset, uap->whence));
2026 }
2027 #endif
2028 
2029 /*
2030  * Check access permissions using passed credentials.
2031  */
2032 static int
2033 vn_access(struct vnode *vp, int user_flags, struct ucred *cred,
2034      struct thread *td)
2035 {
2036 	accmode_t accmode;
2037 	int error;
2038 
2039 	/* Flags == 0 means only check for existence. */
2040 	if (user_flags == 0)
2041 		return (0);
2042 
2043 	accmode = 0;
2044 	if (user_flags & R_OK)
2045 		accmode |= VREAD;
2046 	if (user_flags & W_OK)
2047 		accmode |= VWRITE;
2048 	if (user_flags & X_OK)
2049 		accmode |= VEXEC;
2050 #ifdef MAC
2051 	error = mac_vnode_check_access(cred, vp, accmode);
2052 	if (error != 0)
2053 		return (error);
2054 #endif
2055 	if ((accmode & VWRITE) == 0 || (error = vn_writechk(vp)) == 0)
2056 		error = VOP_ACCESS(vp, accmode, cred, td);
2057 	return (error);
2058 }
2059 
2060 /*
2061  * Check access permissions using "real" credentials.
2062  */
2063 #ifndef _SYS_SYSPROTO_H_
2064 struct access_args {
2065 	char	*path;
2066 	int	amode;
2067 };
2068 #endif
2069 int
2070 sys_access(struct thread *td, struct access_args *uap)
2071 {
2072 
2073 	return (kern_accessat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2074 	    0, uap->amode));
2075 }
2076 
2077 #ifndef _SYS_SYSPROTO_H_
2078 struct faccessat_args {
2079 	int	dirfd;
2080 	char	*path;
2081 	int	amode;
2082 	int	flag;
2083 }
2084 #endif
2085 int
2086 sys_faccessat(struct thread *td, struct faccessat_args *uap)
2087 {
2088 
2089 	return (kern_accessat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
2090 	    uap->amode));
2091 }
2092 
2093 int
2094 kern_accessat(struct thread *td, int fd, const char *path,
2095     enum uio_seg pathseg, int flag, int amode)
2096 {
2097 	struct ucred *cred, *usecred;
2098 	struct vnode *vp;
2099 	struct nameidata nd;
2100 	int error;
2101 
2102 	if ((flag & ~(AT_EACCESS | AT_RESOLVE_BENEATH | AT_EMPTY_PATH)) != 0)
2103 		return (EINVAL);
2104 	if (amode != F_OK && (amode & ~(R_OK | W_OK | X_OK)) != 0)
2105 		return (EINVAL);
2106 
2107 	/*
2108 	 * Create and modify a temporary credential instead of one that
2109 	 * is potentially shared (if we need one).
2110 	 */
2111 	cred = td->td_ucred;
2112 	if ((flag & AT_EACCESS) == 0 &&
2113 	    ((cred->cr_uid != cred->cr_ruid ||
2114 	    cred->cr_rgid != cred->cr_groups[0]))) {
2115 		usecred = crdup(cred);
2116 		usecred->cr_uid = cred->cr_ruid;
2117 		usecred->cr_groups[0] = cred->cr_rgid;
2118 		td->td_ucred = usecred;
2119 	} else
2120 		usecred = cred;
2121 	AUDIT_ARG_VALUE(amode);
2122 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF |
2123 	    AUDITVNODE1 | at2cnpflags(flag, AT_RESOLVE_BENEATH |
2124 	    AT_EMPTY_PATH), pathseg, path, fd, &cap_fstat_rights, td);
2125 	if ((error = namei(&nd)) != 0)
2126 		goto out;
2127 	vp = nd.ni_vp;
2128 
2129 	error = vn_access(vp, amode, usecred, td);
2130 	NDFREE_NOTHING(&nd);
2131 	vput(vp);
2132 out:
2133 	if (usecred != cred) {
2134 		td->td_ucred = cred;
2135 		crfree(usecred);
2136 	}
2137 	return (error);
2138 }
2139 
2140 /*
2141  * Check access permissions using "effective" credentials.
2142  */
2143 #ifndef _SYS_SYSPROTO_H_
2144 struct eaccess_args {
2145 	char	*path;
2146 	int	amode;
2147 };
2148 #endif
2149 int
2150 sys_eaccess(struct thread *td, struct eaccess_args *uap)
2151 {
2152 
2153 	return (kern_accessat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2154 	    AT_EACCESS, uap->amode));
2155 }
2156 
2157 #if defined(COMPAT_43)
2158 /*
2159  * Get file status; this version follows links.
2160  */
2161 #ifndef _SYS_SYSPROTO_H_
2162 struct ostat_args {
2163 	char	*path;
2164 	struct ostat *ub;
2165 };
2166 #endif
2167 int
2168 ostat(struct thread *td, struct ostat_args *uap)
2169 {
2170 	struct stat sb;
2171 	struct ostat osb;
2172 	int error;
2173 
2174 	error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE,
2175 	    &sb, NULL);
2176 	if (error != 0)
2177 		return (error);
2178 	cvtstat(&sb, &osb);
2179 	return (copyout(&osb, uap->ub, sizeof (osb)));
2180 }
2181 
2182 /*
2183  * Get file status; this version does not follow links.
2184  */
2185 #ifndef _SYS_SYSPROTO_H_
2186 struct olstat_args {
2187 	char	*path;
2188 	struct ostat *ub;
2189 };
2190 #endif
2191 int
2192 olstat(struct thread *td, struct olstat_args *uap)
2193 {
2194 	struct stat sb;
2195 	struct ostat osb;
2196 	int error;
2197 
2198 	error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2199 	    UIO_USERSPACE, &sb, NULL);
2200 	if (error != 0)
2201 		return (error);
2202 	cvtstat(&sb, &osb);
2203 	return (copyout(&osb, uap->ub, sizeof (osb)));
2204 }
2205 
2206 /*
2207  * Convert from an old to a new stat structure.
2208  * XXX: many values are blindly truncated.
2209  */
2210 void
2211 cvtstat(struct stat *st, struct ostat *ost)
2212 {
2213 
2214 	bzero(ost, sizeof(*ost));
2215 	ost->st_dev = st->st_dev;
2216 	ost->st_ino = st->st_ino;
2217 	ost->st_mode = st->st_mode;
2218 	ost->st_nlink = st->st_nlink;
2219 	ost->st_uid = st->st_uid;
2220 	ost->st_gid = st->st_gid;
2221 	ost->st_rdev = st->st_rdev;
2222 	ost->st_size = MIN(st->st_size, INT32_MAX);
2223 	ost->st_atim = st->st_atim;
2224 	ost->st_mtim = st->st_mtim;
2225 	ost->st_ctim = st->st_ctim;
2226 	ost->st_blksize = st->st_blksize;
2227 	ost->st_blocks = st->st_blocks;
2228 	ost->st_flags = st->st_flags;
2229 	ost->st_gen = st->st_gen;
2230 }
2231 #endif /* COMPAT_43 */
2232 
2233 #if defined(COMPAT_43) || defined(COMPAT_FREEBSD11)
2234 int ino64_trunc_error;
2235 SYSCTL_INT(_vfs, OID_AUTO, ino64_trunc_error, CTLFLAG_RW,
2236     &ino64_trunc_error, 0,
2237     "Error on truncation of device, file or inode number, or link count");
2238 
2239 int
2240 freebsd11_cvtstat(struct stat *st, struct freebsd11_stat *ost)
2241 {
2242 
2243 	ost->st_dev = st->st_dev;
2244 	if (ost->st_dev != st->st_dev) {
2245 		switch (ino64_trunc_error) {
2246 		default:
2247 			/*
2248 			 * Since dev_t is almost raw, don't clamp to the
2249 			 * maximum for case 2, but ignore the error.
2250 			 */
2251 			break;
2252 		case 1:
2253 			return (EOVERFLOW);
2254 		}
2255 	}
2256 	ost->st_ino = st->st_ino;
2257 	if (ost->st_ino != st->st_ino) {
2258 		switch (ino64_trunc_error) {
2259 		default:
2260 		case 0:
2261 			break;
2262 		case 1:
2263 			return (EOVERFLOW);
2264 		case 2:
2265 			ost->st_ino = UINT32_MAX;
2266 			break;
2267 		}
2268 	}
2269 	ost->st_mode = st->st_mode;
2270 	ost->st_nlink = st->st_nlink;
2271 	if (ost->st_nlink != st->st_nlink) {
2272 		switch (ino64_trunc_error) {
2273 		default:
2274 		case 0:
2275 			break;
2276 		case 1:
2277 			return (EOVERFLOW);
2278 		case 2:
2279 			ost->st_nlink = UINT16_MAX;
2280 			break;
2281 		}
2282 	}
2283 	ost->st_uid = st->st_uid;
2284 	ost->st_gid = st->st_gid;
2285 	ost->st_rdev = st->st_rdev;
2286 	if (ost->st_rdev != st->st_rdev) {
2287 		switch (ino64_trunc_error) {
2288 		default:
2289 			break;
2290 		case 1:
2291 			return (EOVERFLOW);
2292 		}
2293 	}
2294 	ost->st_atim = st->st_atim;
2295 	ost->st_mtim = st->st_mtim;
2296 	ost->st_ctim = st->st_ctim;
2297 	ost->st_size = st->st_size;
2298 	ost->st_blocks = st->st_blocks;
2299 	ost->st_blksize = st->st_blksize;
2300 	ost->st_flags = st->st_flags;
2301 	ost->st_gen = st->st_gen;
2302 	ost->st_lspare = 0;
2303 	ost->st_birthtim = st->st_birthtim;
2304 	bzero((char *)&ost->st_birthtim + sizeof(ost->st_birthtim),
2305 	    sizeof(*ost) - offsetof(struct freebsd11_stat,
2306 	    st_birthtim) - sizeof(ost->st_birthtim));
2307 	return (0);
2308 }
2309 
2310 int
2311 freebsd11_stat(struct thread *td, struct freebsd11_stat_args* uap)
2312 {
2313 	struct stat sb;
2314 	struct freebsd11_stat osb;
2315 	int error;
2316 
2317 	error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE,
2318 	    &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_lstat(struct thread *td, struct freebsd11_lstat_args* uap)
2329 {
2330 	struct stat sb;
2331 	struct freebsd11_stat osb;
2332 	int error;
2333 
2334 	error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2335 	    UIO_USERSPACE, &sb, NULL);
2336 	if (error != 0)
2337 		return (error);
2338 	error = freebsd11_cvtstat(&sb, &osb);
2339 	if (error == 0)
2340 		error = copyout(&osb, uap->ub, sizeof(osb));
2341 	return (error);
2342 }
2343 
2344 int
2345 freebsd11_fhstat(struct thread *td, struct freebsd11_fhstat_args* uap)
2346 {
2347 	struct fhandle fh;
2348 	struct stat sb;
2349 	struct freebsd11_stat osb;
2350 	int error;
2351 
2352 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
2353 	if (error != 0)
2354 		return (error);
2355 	error = kern_fhstat(td, fh, &sb);
2356 	if (error != 0)
2357 		return (error);
2358 	error = freebsd11_cvtstat(&sb, &osb);
2359 	if (error == 0)
2360 		error = copyout(&osb, uap->sb, sizeof(osb));
2361 	return (error);
2362 }
2363 
2364 int
2365 freebsd11_fstatat(struct thread *td, struct freebsd11_fstatat_args* uap)
2366 {
2367 	struct stat sb;
2368 	struct freebsd11_stat osb;
2369 	int error;
2370 
2371 	error = kern_statat(td, uap->flag, uap->fd, uap->path,
2372 	    UIO_USERSPACE, &sb, NULL);
2373 	if (error != 0)
2374 		return (error);
2375 	error = freebsd11_cvtstat(&sb, &osb);
2376 	if (error == 0)
2377 		error = copyout(&osb, uap->buf, sizeof(osb));
2378 	return (error);
2379 }
2380 #endif	/* COMPAT_FREEBSD11 */
2381 
2382 /*
2383  * Get file status
2384  */
2385 #ifndef _SYS_SYSPROTO_H_
2386 struct fstatat_args {
2387 	int	fd;
2388 	char	*path;
2389 	struct stat	*buf;
2390 	int	flag;
2391 }
2392 #endif
2393 int
2394 sys_fstatat(struct thread *td, struct fstatat_args *uap)
2395 {
2396 	struct stat sb;
2397 	int error;
2398 
2399 	error = kern_statat(td, uap->flag, uap->fd, uap->path,
2400 	    UIO_USERSPACE, &sb, NULL);
2401 	if (error == 0)
2402 		error = copyout(&sb, uap->buf, sizeof (sb));
2403 	return (error);
2404 }
2405 
2406 int
2407 kern_statat(struct thread *td, int flag, int fd, const char *path,
2408     enum uio_seg pathseg, struct stat *sbp,
2409     void (*hook)(struct vnode *vp, struct stat *sbp))
2410 {
2411 	struct nameidata nd;
2412 	int error;
2413 
2414 	if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH |
2415 	    AT_EMPTY_PATH)) != 0)
2416 		return (EINVAL);
2417 
2418 	NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(flag, AT_RESOLVE_BENEATH |
2419 	    AT_SYMLINK_NOFOLLOW | AT_EMPTY_PATH) | LOCKSHARED | LOCKLEAF |
2420 	    AUDITVNODE1, pathseg, path, fd, &cap_fstat_rights, td);
2421 
2422 	if ((error = namei(&nd)) != 0)
2423 		return (error);
2424 	error = VOP_STAT(nd.ni_vp, sbp, td->td_ucred, NOCRED, td);
2425 	if (error == 0) {
2426 		if (__predict_false(hook != NULL))
2427 			hook(nd.ni_vp, sbp);
2428 	}
2429 	NDFREE_NOTHING(&nd);
2430 	vput(nd.ni_vp);
2431 #ifdef __STAT_TIME_T_EXT
2432 	sbp->st_atim_ext = 0;
2433 	sbp->st_mtim_ext = 0;
2434 	sbp->st_ctim_ext = 0;
2435 	sbp->st_btim_ext = 0;
2436 #endif
2437 #ifdef KTRACE
2438 	if (KTRPOINT(td, KTR_STRUCT))
2439 		ktrstat_error(sbp, error);
2440 #endif
2441 	return (error);
2442 }
2443 
2444 #if defined(COMPAT_FREEBSD11)
2445 /*
2446  * Implementation of the NetBSD [l]stat() functions.
2447  */
2448 void
2449 freebsd11_cvtnstat(struct stat *sb, struct nstat *nsb)
2450 {
2451 
2452 	bzero(nsb, sizeof(*nsb));
2453 	nsb->st_dev = sb->st_dev;
2454 	nsb->st_ino = sb->st_ino;
2455 	nsb->st_mode = sb->st_mode;
2456 	nsb->st_nlink = sb->st_nlink;
2457 	nsb->st_uid = sb->st_uid;
2458 	nsb->st_gid = sb->st_gid;
2459 	nsb->st_rdev = sb->st_rdev;
2460 	nsb->st_atim = sb->st_atim;
2461 	nsb->st_mtim = sb->st_mtim;
2462 	nsb->st_ctim = sb->st_ctim;
2463 	nsb->st_size = sb->st_size;
2464 	nsb->st_blocks = sb->st_blocks;
2465 	nsb->st_blksize = sb->st_blksize;
2466 	nsb->st_flags = sb->st_flags;
2467 	nsb->st_gen = sb->st_gen;
2468 	nsb->st_birthtim = sb->st_birthtim;
2469 }
2470 
2471 #ifndef _SYS_SYSPROTO_H_
2472 struct freebsd11_nstat_args {
2473 	char	*path;
2474 	struct nstat *ub;
2475 };
2476 #endif
2477 int
2478 freebsd11_nstat(struct thread *td, struct freebsd11_nstat_args *uap)
2479 {
2480 	struct stat sb;
2481 	struct nstat nsb;
2482 	int error;
2483 
2484 	error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE,
2485 	    &sb, NULL);
2486 	if (error != 0)
2487 		return (error);
2488 	freebsd11_cvtnstat(&sb, &nsb);
2489 	return (copyout(&nsb, uap->ub, sizeof (nsb)));
2490 }
2491 
2492 /*
2493  * NetBSD lstat.  Get file status; this version does not follow links.
2494  */
2495 #ifndef _SYS_SYSPROTO_H_
2496 struct freebsd11_nlstat_args {
2497 	char	*path;
2498 	struct nstat *ub;
2499 };
2500 #endif
2501 int
2502 freebsd11_nlstat(struct thread *td, struct freebsd11_nlstat_args *uap)
2503 {
2504 	struct stat sb;
2505 	struct nstat nsb;
2506 	int error;
2507 
2508 	error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2509 	    UIO_USERSPACE, &sb, NULL);
2510 	if (error != 0)
2511 		return (error);
2512 	freebsd11_cvtnstat(&sb, &nsb);
2513 	return (copyout(&nsb, uap->ub, sizeof (nsb)));
2514 }
2515 #endif /* COMPAT_FREEBSD11 */
2516 
2517 /*
2518  * Get configurable pathname variables.
2519  */
2520 #ifndef _SYS_SYSPROTO_H_
2521 struct pathconf_args {
2522 	char	*path;
2523 	int	name;
2524 };
2525 #endif
2526 int
2527 sys_pathconf(struct thread *td, struct pathconf_args *uap)
2528 {
2529 	long value;
2530 	int error;
2531 
2532 	error = kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name, FOLLOW,
2533 	    &value);
2534 	if (error == 0)
2535 		td->td_retval[0] = value;
2536 	return (error);
2537 }
2538 
2539 #ifndef _SYS_SYSPROTO_H_
2540 struct lpathconf_args {
2541 	char	*path;
2542 	int	name;
2543 };
2544 #endif
2545 int
2546 sys_lpathconf(struct thread *td, struct lpathconf_args *uap)
2547 {
2548 	long value;
2549 	int error;
2550 
2551 	error = kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name,
2552 	    NOFOLLOW, &value);
2553 	if (error == 0)
2554 		td->td_retval[0] = value;
2555 	return (error);
2556 }
2557 
2558 int
2559 kern_pathconf(struct thread *td, const char *path, enum uio_seg pathseg,
2560     int name, u_long flags, long *valuep)
2561 {
2562 	struct nameidata nd;
2563 	int error;
2564 
2565 	NDINIT(&nd, LOOKUP, LOCKSHARED | LOCKLEAF | AUDITVNODE1 | flags,
2566 	    pathseg, path, td);
2567 	if ((error = namei(&nd)) != 0)
2568 		return (error);
2569 	NDFREE_NOTHING(&nd);
2570 
2571 	error = VOP_PATHCONF(nd.ni_vp, name, valuep);
2572 	vput(nd.ni_vp);
2573 	return (error);
2574 }
2575 
2576 /*
2577  * Return target name of a symbolic link.
2578  */
2579 #ifndef _SYS_SYSPROTO_H_
2580 struct readlink_args {
2581 	char	*path;
2582 	char	*buf;
2583 	size_t	count;
2584 };
2585 #endif
2586 int
2587 sys_readlink(struct thread *td, struct readlink_args *uap)
2588 {
2589 
2590 	return (kern_readlinkat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2591 	    uap->buf, UIO_USERSPACE, uap->count));
2592 }
2593 #ifndef _SYS_SYSPROTO_H_
2594 struct readlinkat_args {
2595 	int	fd;
2596 	char	*path;
2597 	char	*buf;
2598 	size_t	bufsize;
2599 };
2600 #endif
2601 int
2602 sys_readlinkat(struct thread *td, struct readlinkat_args *uap)
2603 {
2604 
2605 	return (kern_readlinkat(td, uap->fd, uap->path, UIO_USERSPACE,
2606 	    uap->buf, UIO_USERSPACE, uap->bufsize));
2607 }
2608 
2609 int
2610 kern_readlinkat(struct thread *td, int fd, const char *path,
2611     enum uio_seg pathseg, char *buf, enum uio_seg bufseg, size_t count)
2612 {
2613 	struct vnode *vp;
2614 	struct nameidata nd;
2615 	int error;
2616 
2617 	if (count > IOSIZE_MAX)
2618 		return (EINVAL);
2619 
2620 	NDINIT_AT(&nd, LOOKUP, NOFOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
2621 	    pathseg, path, fd, td);
2622 
2623 	if ((error = namei(&nd)) != 0)
2624 		return (error);
2625 	NDFREE_NOTHING(&nd);
2626 	vp = nd.ni_vp;
2627 
2628 	error = kern_readlink_vp(vp, buf, bufseg, count, td);
2629 	vput(vp);
2630 
2631 	return (error);
2632 }
2633 
2634 /*
2635  * Helper function to readlink from a vnode
2636  */
2637 static int
2638 kern_readlink_vp(struct vnode *vp, char *buf, enum uio_seg bufseg, size_t count,
2639     struct thread *td)
2640 {
2641 	struct iovec aiov;
2642 	struct uio auio;
2643 	int error;
2644 
2645 	ASSERT_VOP_LOCKED(vp, "kern_readlink_vp(): vp not locked");
2646 #ifdef MAC
2647 	error = mac_vnode_check_readlink(td->td_ucred, vp);
2648 	if (error != 0)
2649 		return (error);
2650 #endif
2651 	if (vp->v_type != VLNK && (vp->v_vflag & VV_READLINK) == 0)
2652 		return (EINVAL);
2653 
2654 	aiov.iov_base = buf;
2655 	aiov.iov_len = count;
2656 	auio.uio_iov = &aiov;
2657 	auio.uio_iovcnt = 1;
2658 	auio.uio_offset = 0;
2659 	auio.uio_rw = UIO_READ;
2660 	auio.uio_segflg = bufseg;
2661 	auio.uio_td = td;
2662 	auio.uio_resid = count;
2663 	error = VOP_READLINK(vp, &auio, td->td_ucred);
2664 	td->td_retval[0] = count - auio.uio_resid;
2665 	return (error);
2666 }
2667 
2668 /*
2669  * Common implementation code for chflags() and fchflags().
2670  */
2671 static int
2672 setfflags(struct thread *td, struct vnode *vp, u_long flags)
2673 {
2674 	struct mount *mp;
2675 	struct vattr vattr;
2676 	int error;
2677 
2678 	/* We can't support the value matching VNOVAL. */
2679 	if (flags == VNOVAL)
2680 		return (EOPNOTSUPP);
2681 
2682 	/*
2683 	 * Prevent non-root users from setting flags on devices.  When
2684 	 * a device is reused, users can retain ownership of the device
2685 	 * if they are allowed to set flags and programs assume that
2686 	 * chown can't fail when done as root.
2687 	 */
2688 	if (vp->v_type == VCHR || vp->v_type == VBLK) {
2689 		error = priv_check(td, PRIV_VFS_CHFLAGS_DEV);
2690 		if (error != 0)
2691 			return (error);
2692 	}
2693 
2694 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2695 		return (error);
2696 	VATTR_NULL(&vattr);
2697 	vattr.va_flags = flags;
2698 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2699 #ifdef MAC
2700 	error = mac_vnode_check_setflags(td->td_ucred, vp, vattr.va_flags);
2701 	if (error == 0)
2702 #endif
2703 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
2704 	VOP_UNLOCK(vp);
2705 	vn_finished_write(mp);
2706 	return (error);
2707 }
2708 
2709 /*
2710  * Change flags of a file given a path name.
2711  */
2712 #ifndef _SYS_SYSPROTO_H_
2713 struct chflags_args {
2714 	const char *path;
2715 	u_long	flags;
2716 };
2717 #endif
2718 int
2719 sys_chflags(struct thread *td, struct chflags_args *uap)
2720 {
2721 
2722 	return (kern_chflagsat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2723 	    uap->flags, 0));
2724 }
2725 
2726 #ifndef _SYS_SYSPROTO_H_
2727 struct chflagsat_args {
2728 	int	fd;
2729 	const char *path;
2730 	u_long	flags;
2731 	int	atflag;
2732 }
2733 #endif
2734 int
2735 sys_chflagsat(struct thread *td, struct chflagsat_args *uap)
2736 {
2737 
2738 	if ((uap->atflag & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH |
2739 	    AT_EMPTY_PATH)) != 0)
2740 		return (EINVAL);
2741 
2742 	return (kern_chflagsat(td, uap->fd, uap->path, UIO_USERSPACE,
2743 	    uap->flags, uap->atflag));
2744 }
2745 
2746 /*
2747  * Same as chflags() but doesn't follow symlinks.
2748  */
2749 #ifndef _SYS_SYSPROTO_H_
2750 struct lchflags_args {
2751 	const char *path;
2752 	u_long flags;
2753 };
2754 #endif
2755 int
2756 sys_lchflags(struct thread *td, struct lchflags_args *uap)
2757 {
2758 
2759 	return (kern_chflagsat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2760 	    uap->flags, AT_SYMLINK_NOFOLLOW));
2761 }
2762 
2763 static int
2764 kern_chflagsat(struct thread *td, int fd, const char *path,
2765     enum uio_seg pathseg, u_long flags, int atflag)
2766 {
2767 	struct nameidata nd;
2768 	int error;
2769 
2770 	AUDIT_ARG_FFLAGS(flags);
2771 	NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(atflag, AT_SYMLINK_NOFOLLOW |
2772 	    AT_RESOLVE_BENEATH | AT_EMPTY_PATH) | AUDITVNODE1, pathseg, path,
2773 	    fd, &cap_fchflags_rights, td);
2774 	if ((error = namei(&nd)) != 0)
2775 		return (error);
2776 	NDFREE_NOTHING(&nd);
2777 	error = setfflags(td, nd.ni_vp, flags);
2778 	vrele(nd.ni_vp);
2779 	return (error);
2780 }
2781 
2782 /*
2783  * Change flags of a file given a file descriptor.
2784  */
2785 #ifndef _SYS_SYSPROTO_H_
2786 struct fchflags_args {
2787 	int	fd;
2788 	u_long	flags;
2789 };
2790 #endif
2791 int
2792 sys_fchflags(struct thread *td, struct fchflags_args *uap)
2793 {
2794 	struct file *fp;
2795 	int error;
2796 
2797 	AUDIT_ARG_FD(uap->fd);
2798 	AUDIT_ARG_FFLAGS(uap->flags);
2799 	error = getvnode(td, uap->fd, &cap_fchflags_rights,
2800 	    &fp);
2801 	if (error != 0)
2802 		return (error);
2803 #ifdef AUDIT
2804 	if (AUDITING_TD(td)) {
2805 		vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
2806 		AUDIT_ARG_VNODE1(fp->f_vnode);
2807 		VOP_UNLOCK(fp->f_vnode);
2808 	}
2809 #endif
2810 	error = setfflags(td, fp->f_vnode, uap->flags);
2811 	fdrop(fp, td);
2812 	return (error);
2813 }
2814 
2815 /*
2816  * Common implementation code for chmod(), lchmod() and fchmod().
2817  */
2818 int
2819 setfmode(struct thread *td, struct ucred *cred, struct vnode *vp, int mode)
2820 {
2821 	struct mount *mp;
2822 	struct vattr vattr;
2823 	int error;
2824 
2825 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2826 		return (error);
2827 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2828 	VATTR_NULL(&vattr);
2829 	vattr.va_mode = mode & ALLPERMS;
2830 #ifdef MAC
2831 	error = mac_vnode_check_setmode(cred, vp, vattr.va_mode);
2832 	if (error == 0)
2833 #endif
2834 		error = VOP_SETATTR(vp, &vattr, cred);
2835 	VOP_UNLOCK(vp);
2836 	vn_finished_write(mp);
2837 	return (error);
2838 }
2839 
2840 /*
2841  * Change mode of a file given path name.
2842  */
2843 #ifndef _SYS_SYSPROTO_H_
2844 struct chmod_args {
2845 	char	*path;
2846 	int	mode;
2847 };
2848 #endif
2849 int
2850 sys_chmod(struct thread *td, struct chmod_args *uap)
2851 {
2852 
2853 	return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2854 	    uap->mode, 0));
2855 }
2856 
2857 #ifndef _SYS_SYSPROTO_H_
2858 struct fchmodat_args {
2859 	int	dirfd;
2860 	char	*path;
2861 	mode_t	mode;
2862 	int	flag;
2863 }
2864 #endif
2865 int
2866 sys_fchmodat(struct thread *td, struct fchmodat_args *uap)
2867 {
2868 
2869 	if ((uap->flag & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH |
2870 	    AT_EMPTY_PATH)) != 0)
2871 		return (EINVAL);
2872 
2873 	return (kern_fchmodat(td, uap->fd, uap->path, UIO_USERSPACE,
2874 	    uap->mode, uap->flag));
2875 }
2876 
2877 /*
2878  * Change mode of a file given path name (don't follow links.)
2879  */
2880 #ifndef _SYS_SYSPROTO_H_
2881 struct lchmod_args {
2882 	char	*path;
2883 	int	mode;
2884 };
2885 #endif
2886 int
2887 sys_lchmod(struct thread *td, struct lchmod_args *uap)
2888 {
2889 
2890 	return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2891 	    uap->mode, AT_SYMLINK_NOFOLLOW));
2892 }
2893 
2894 int
2895 kern_fchmodat(struct thread *td, int fd, const char *path,
2896     enum uio_seg pathseg, mode_t mode, int flag)
2897 {
2898 	struct nameidata nd;
2899 	int error;
2900 
2901 	AUDIT_ARG_MODE(mode);
2902 	NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(flag, AT_SYMLINK_NOFOLLOW |
2903 	    AT_RESOLVE_BENEATH | AT_EMPTY_PATH) | AUDITVNODE1, pathseg, path,
2904 	    fd, &cap_fchmod_rights, td);
2905 	if ((error = namei(&nd)) != 0)
2906 		return (error);
2907 	NDFREE_NOTHING(&nd);
2908 	error = setfmode(td, td->td_ucred, nd.ni_vp, mode);
2909 	vrele(nd.ni_vp);
2910 	return (error);
2911 }
2912 
2913 /*
2914  * Change mode of a file given a file descriptor.
2915  */
2916 #ifndef _SYS_SYSPROTO_H_
2917 struct fchmod_args {
2918 	int	fd;
2919 	int	mode;
2920 };
2921 #endif
2922 int
2923 sys_fchmod(struct thread *td, struct fchmod_args *uap)
2924 {
2925 	struct file *fp;
2926 	int error;
2927 
2928 	AUDIT_ARG_FD(uap->fd);
2929 	AUDIT_ARG_MODE(uap->mode);
2930 
2931 	error = fget(td, uap->fd, &cap_fchmod_rights, &fp);
2932 	if (error != 0)
2933 		return (error);
2934 	error = fo_chmod(fp, uap->mode, td->td_ucred, td);
2935 	fdrop(fp, td);
2936 	return (error);
2937 }
2938 
2939 /*
2940  * Common implementation for chown(), lchown(), and fchown()
2941  */
2942 int
2943 setfown(struct thread *td, struct ucred *cred, struct vnode *vp, uid_t uid,
2944     gid_t gid)
2945 {
2946 	struct mount *mp;
2947 	struct vattr vattr;
2948 	int error;
2949 
2950 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2951 		return (error);
2952 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2953 	VATTR_NULL(&vattr);
2954 	vattr.va_uid = uid;
2955 	vattr.va_gid = gid;
2956 #ifdef MAC
2957 	error = mac_vnode_check_setowner(cred, vp, vattr.va_uid,
2958 	    vattr.va_gid);
2959 	if (error == 0)
2960 #endif
2961 		error = VOP_SETATTR(vp, &vattr, cred);
2962 	VOP_UNLOCK(vp);
2963 	vn_finished_write(mp);
2964 	return (error);
2965 }
2966 
2967 /*
2968  * Set ownership given a path name.
2969  */
2970 #ifndef _SYS_SYSPROTO_H_
2971 struct chown_args {
2972 	char	*path;
2973 	int	uid;
2974 	int	gid;
2975 };
2976 #endif
2977 int
2978 sys_chown(struct thread *td, struct chown_args *uap)
2979 {
2980 
2981 	return (kern_fchownat(td, AT_FDCWD, uap->path, UIO_USERSPACE, uap->uid,
2982 	    uap->gid, 0));
2983 }
2984 
2985 #ifndef _SYS_SYSPROTO_H_
2986 struct fchownat_args {
2987 	int fd;
2988 	const char * path;
2989 	uid_t uid;
2990 	gid_t gid;
2991 	int flag;
2992 };
2993 #endif
2994 int
2995 sys_fchownat(struct thread *td, struct fchownat_args *uap)
2996 {
2997 
2998 	if ((uap->flag & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH |
2999 	    AT_EMPTY_PATH)) != 0)
3000 		return (EINVAL);
3001 
3002 	return (kern_fchownat(td, uap->fd, uap->path, UIO_USERSPACE, uap->uid,
3003 	    uap->gid, uap->flag));
3004 }
3005 
3006 int
3007 kern_fchownat(struct thread *td, int fd, const char *path,
3008     enum uio_seg pathseg, int uid, int gid, int flag)
3009 {
3010 	struct nameidata nd;
3011 	int error;
3012 
3013 	AUDIT_ARG_OWNER(uid, gid);
3014 	NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(flag, AT_SYMLINK_NOFOLLOW |
3015 	    AT_RESOLVE_BENEATH | AT_EMPTY_PATH) | AUDITVNODE1, pathseg, path,
3016 	    fd, &cap_fchown_rights, td);
3017 
3018 	if ((error = namei(&nd)) != 0)
3019 		return (error);
3020 	NDFREE_NOTHING(&nd);
3021 	error = setfown(td, td->td_ucred, nd.ni_vp, uid, gid);
3022 	vrele(nd.ni_vp);
3023 	return (error);
3024 }
3025 
3026 /*
3027  * Set ownership given a path name, do not cross symlinks.
3028  */
3029 #ifndef _SYS_SYSPROTO_H_
3030 struct lchown_args {
3031 	char	*path;
3032 	int	uid;
3033 	int	gid;
3034 };
3035 #endif
3036 int
3037 sys_lchown(struct thread *td, struct lchown_args *uap)
3038 {
3039 
3040 	return (kern_fchownat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
3041 	    uap->uid, uap->gid, AT_SYMLINK_NOFOLLOW));
3042 }
3043 
3044 /*
3045  * Set ownership given a file descriptor.
3046  */
3047 #ifndef _SYS_SYSPROTO_H_
3048 struct fchown_args {
3049 	int	fd;
3050 	int	uid;
3051 	int	gid;
3052 };
3053 #endif
3054 int
3055 sys_fchown(struct thread *td, struct fchown_args *uap)
3056 {
3057 	struct file *fp;
3058 	int error;
3059 
3060 	AUDIT_ARG_FD(uap->fd);
3061 	AUDIT_ARG_OWNER(uap->uid, uap->gid);
3062 	error = fget(td, uap->fd, &cap_fchown_rights, &fp);
3063 	if (error != 0)
3064 		return (error);
3065 	error = fo_chown(fp, uap->uid, uap->gid, td->td_ucred, td);
3066 	fdrop(fp, td);
3067 	return (error);
3068 }
3069 
3070 /*
3071  * Common implementation code for utimes(), lutimes(), and futimes().
3072  */
3073 static int
3074 getutimes(const struct timeval *usrtvp, enum uio_seg tvpseg,
3075     struct timespec *tsp)
3076 {
3077 	struct timeval tv[2];
3078 	const struct timeval *tvp;
3079 	int error;
3080 
3081 	if (usrtvp == NULL) {
3082 		vfs_timestamp(&tsp[0]);
3083 		tsp[1] = tsp[0];
3084 	} else {
3085 		if (tvpseg == UIO_SYSSPACE) {
3086 			tvp = usrtvp;
3087 		} else {
3088 			if ((error = copyin(usrtvp, tv, sizeof(tv))) != 0)
3089 				return (error);
3090 			tvp = tv;
3091 		}
3092 
3093 		if (tvp[0].tv_usec < 0 || tvp[0].tv_usec >= 1000000 ||
3094 		    tvp[1].tv_usec < 0 || tvp[1].tv_usec >= 1000000)
3095 			return (EINVAL);
3096 		TIMEVAL_TO_TIMESPEC(&tvp[0], &tsp[0]);
3097 		TIMEVAL_TO_TIMESPEC(&tvp[1], &tsp[1]);
3098 	}
3099 	return (0);
3100 }
3101 
3102 /*
3103  * Common implementation code for futimens(), utimensat().
3104  */
3105 #define	UTIMENS_NULL	0x1
3106 #define	UTIMENS_EXIT	0x2
3107 static int
3108 getutimens(const struct timespec *usrtsp, enum uio_seg tspseg,
3109     struct timespec *tsp, int *retflags)
3110 {
3111 	struct timespec tsnow;
3112 	int error;
3113 
3114 	vfs_timestamp(&tsnow);
3115 	*retflags = 0;
3116 	if (usrtsp == NULL) {
3117 		tsp[0] = tsnow;
3118 		tsp[1] = tsnow;
3119 		*retflags |= UTIMENS_NULL;
3120 		return (0);
3121 	}
3122 	if (tspseg == UIO_SYSSPACE) {
3123 		tsp[0] = usrtsp[0];
3124 		tsp[1] = usrtsp[1];
3125 	} else if ((error = copyin(usrtsp, tsp, sizeof(*tsp) * 2)) != 0)
3126 		return (error);
3127 	if (tsp[0].tv_nsec == UTIME_OMIT && tsp[1].tv_nsec == UTIME_OMIT)
3128 		*retflags |= UTIMENS_EXIT;
3129 	if (tsp[0].tv_nsec == UTIME_NOW && tsp[1].tv_nsec == UTIME_NOW)
3130 		*retflags |= UTIMENS_NULL;
3131 	if (tsp[0].tv_nsec == UTIME_OMIT)
3132 		tsp[0].tv_sec = VNOVAL;
3133 	else if (tsp[0].tv_nsec == UTIME_NOW)
3134 		tsp[0] = tsnow;
3135 	else if (tsp[0].tv_nsec < 0 || tsp[0].tv_nsec >= 1000000000L)
3136 		return (EINVAL);
3137 	if (tsp[1].tv_nsec == UTIME_OMIT)
3138 		tsp[1].tv_sec = VNOVAL;
3139 	else if (tsp[1].tv_nsec == UTIME_NOW)
3140 		tsp[1] = tsnow;
3141 	else if (tsp[1].tv_nsec < 0 || tsp[1].tv_nsec >= 1000000000L)
3142 		return (EINVAL);
3143 
3144 	return (0);
3145 }
3146 
3147 /*
3148  * Common implementation code for utimes(), lutimes(), futimes(), futimens(),
3149  * and utimensat().
3150  */
3151 static int
3152 setutimes(struct thread *td, struct vnode *vp, const struct timespec *ts,
3153     int numtimes, int nullflag)
3154 {
3155 	struct mount *mp;
3156 	struct vattr vattr;
3157 	int error, setbirthtime;
3158 
3159 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
3160 		return (error);
3161 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3162 	setbirthtime = 0;
3163 	if (numtimes < 3 && !VOP_GETATTR(vp, &vattr, td->td_ucred) &&
3164 	    timespeccmp(&ts[1], &vattr.va_birthtime, < ))
3165 		setbirthtime = 1;
3166 	VATTR_NULL(&vattr);
3167 	vattr.va_atime = ts[0];
3168 	vattr.va_mtime = ts[1];
3169 	if (setbirthtime)
3170 		vattr.va_birthtime = ts[1];
3171 	if (numtimes > 2)
3172 		vattr.va_birthtime = ts[2];
3173 	if (nullflag)
3174 		vattr.va_vaflags |= VA_UTIMES_NULL;
3175 #ifdef MAC
3176 	error = mac_vnode_check_setutimes(td->td_ucred, vp, vattr.va_atime,
3177 	    vattr.va_mtime);
3178 #endif
3179 	if (error == 0)
3180 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3181 	VOP_UNLOCK(vp);
3182 	vn_finished_write(mp);
3183 	return (error);
3184 }
3185 
3186 /*
3187  * Set the access and modification times of a file.
3188  */
3189 #ifndef _SYS_SYSPROTO_H_
3190 struct utimes_args {
3191 	char	*path;
3192 	struct	timeval *tptr;
3193 };
3194 #endif
3195 int
3196 sys_utimes(struct thread *td, struct utimes_args *uap)
3197 {
3198 
3199 	return (kern_utimesat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
3200 	    uap->tptr, UIO_USERSPACE));
3201 }
3202 
3203 #ifndef _SYS_SYSPROTO_H_
3204 struct futimesat_args {
3205 	int fd;
3206 	const char * path;
3207 	const struct timeval * times;
3208 };
3209 #endif
3210 int
3211 sys_futimesat(struct thread *td, struct futimesat_args *uap)
3212 {
3213 
3214 	return (kern_utimesat(td, uap->fd, uap->path, UIO_USERSPACE,
3215 	    uap->times, UIO_USERSPACE));
3216 }
3217 
3218 int
3219 kern_utimesat(struct thread *td, int fd, const char *path,
3220     enum uio_seg pathseg, struct timeval *tptr, enum uio_seg tptrseg)
3221 {
3222 	struct nameidata nd;
3223 	struct timespec ts[2];
3224 	int error;
3225 
3226 	if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3227 		return (error);
3228 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, fd,
3229 	    &cap_futimes_rights, td);
3230 
3231 	if ((error = namei(&nd)) != 0)
3232 		return (error);
3233 	NDFREE_NOTHING(&nd);
3234 	error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3235 	vrele(nd.ni_vp);
3236 	return (error);
3237 }
3238 
3239 /*
3240  * Set the access and modification times of a file.
3241  */
3242 #ifndef _SYS_SYSPROTO_H_
3243 struct lutimes_args {
3244 	char	*path;
3245 	struct	timeval *tptr;
3246 };
3247 #endif
3248 int
3249 sys_lutimes(struct thread *td, struct lutimes_args *uap)
3250 {
3251 
3252 	return (kern_lutimes(td, uap->path, UIO_USERSPACE, uap->tptr,
3253 	    UIO_USERSPACE));
3254 }
3255 
3256 int
3257 kern_lutimes(struct thread *td, const char *path, enum uio_seg pathseg,
3258     struct timeval *tptr, enum uio_seg tptrseg)
3259 {
3260 	struct timespec ts[2];
3261 	struct nameidata nd;
3262 	int error;
3263 
3264 	if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3265 		return (error);
3266 	NDINIT(&nd, LOOKUP, NOFOLLOW | AUDITVNODE1, pathseg, path, td);
3267 	if ((error = namei(&nd)) != 0)
3268 		return (error);
3269 	NDFREE_NOTHING(&nd);
3270 	error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3271 	vrele(nd.ni_vp);
3272 	return (error);
3273 }
3274 
3275 /*
3276  * Set the access and modification times of a file.
3277  */
3278 #ifndef _SYS_SYSPROTO_H_
3279 struct futimes_args {
3280 	int	fd;
3281 	struct	timeval *tptr;
3282 };
3283 #endif
3284 int
3285 sys_futimes(struct thread *td, struct futimes_args *uap)
3286 {
3287 
3288 	return (kern_futimes(td, uap->fd, uap->tptr, UIO_USERSPACE));
3289 }
3290 
3291 int
3292 kern_futimes(struct thread *td, int fd, struct timeval *tptr,
3293     enum uio_seg tptrseg)
3294 {
3295 	struct timespec ts[2];
3296 	struct file *fp;
3297 	int error;
3298 
3299 	AUDIT_ARG_FD(fd);
3300 	error = getutimes(tptr, tptrseg, ts);
3301 	if (error != 0)
3302 		return (error);
3303 	error = getvnode(td, fd, &cap_futimes_rights, &fp);
3304 	if (error != 0)
3305 		return (error);
3306 #ifdef AUDIT
3307 	if (AUDITING_TD(td)) {
3308 		vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
3309 		AUDIT_ARG_VNODE1(fp->f_vnode);
3310 		VOP_UNLOCK(fp->f_vnode);
3311 	}
3312 #endif
3313 	error = setutimes(td, fp->f_vnode, ts, 2, tptr == NULL);
3314 	fdrop(fp, td);
3315 	return (error);
3316 }
3317 
3318 int
3319 sys_futimens(struct thread *td, struct futimens_args *uap)
3320 {
3321 
3322 	return (kern_futimens(td, uap->fd, uap->times, UIO_USERSPACE));
3323 }
3324 
3325 int
3326 kern_futimens(struct thread *td, int fd, struct timespec *tptr,
3327     enum uio_seg tptrseg)
3328 {
3329 	struct timespec ts[2];
3330 	struct file *fp;
3331 	int error, flags;
3332 
3333 	AUDIT_ARG_FD(fd);
3334 	error = getutimens(tptr, tptrseg, ts, &flags);
3335 	if (error != 0)
3336 		return (error);
3337 	if (flags & UTIMENS_EXIT)
3338 		return (0);
3339 	error = getvnode(td, fd, &cap_futimes_rights, &fp);
3340 	if (error != 0)
3341 		return (error);
3342 #ifdef AUDIT
3343 	if (AUDITING_TD(td)) {
3344 		vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
3345 		AUDIT_ARG_VNODE1(fp->f_vnode);
3346 		VOP_UNLOCK(fp->f_vnode);
3347 	}
3348 #endif
3349 	error = setutimes(td, fp->f_vnode, ts, 2, flags & UTIMENS_NULL);
3350 	fdrop(fp, td);
3351 	return (error);
3352 }
3353 
3354 int
3355 sys_utimensat(struct thread *td, struct utimensat_args *uap)
3356 {
3357 
3358 	return (kern_utimensat(td, uap->fd, uap->path, UIO_USERSPACE,
3359 	    uap->times, UIO_USERSPACE, uap->flag));
3360 }
3361 
3362 int
3363 kern_utimensat(struct thread *td, int fd, const char *path,
3364     enum uio_seg pathseg, struct timespec *tptr, enum uio_seg tptrseg,
3365     int flag)
3366 {
3367 	struct nameidata nd;
3368 	struct timespec ts[2];
3369 	int error, flags;
3370 
3371 	if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH |
3372 	    AT_EMPTY_PATH)) != 0)
3373 		return (EINVAL);
3374 
3375 	if ((error = getutimens(tptr, tptrseg, ts, &flags)) != 0)
3376 		return (error);
3377 	NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(flag, AT_SYMLINK_NOFOLLOW |
3378 	    AT_RESOLVE_BENEATH | AT_EMPTY_PATH) | AUDITVNODE1,
3379 	    pathseg, path, fd, &cap_futimes_rights, td);
3380 	if ((error = namei(&nd)) != 0)
3381 		return (error);
3382 	/*
3383 	 * We are allowed to call namei() regardless of 2xUTIME_OMIT.
3384 	 * POSIX states:
3385 	 * "If both tv_nsec fields are UTIME_OMIT... EACCESS may be detected."
3386 	 * "Search permission is denied by a component of the path prefix."
3387 	 */
3388 	NDFREE_NOTHING(&nd);
3389 	if ((flags & UTIMENS_EXIT) == 0)
3390 		error = setutimes(td, nd.ni_vp, ts, 2, flags & UTIMENS_NULL);
3391 	vrele(nd.ni_vp);
3392 	return (error);
3393 }
3394 
3395 /*
3396  * Truncate a file given its path name.
3397  */
3398 #ifndef _SYS_SYSPROTO_H_
3399 struct truncate_args {
3400 	char	*path;
3401 	int	pad;
3402 	off_t	length;
3403 };
3404 #endif
3405 int
3406 sys_truncate(struct thread *td, struct truncate_args *uap)
3407 {
3408 
3409 	return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3410 }
3411 
3412 int
3413 kern_truncate(struct thread *td, const char *path, enum uio_seg pathseg,
3414     off_t length)
3415 {
3416 	struct mount *mp;
3417 	struct vnode *vp;
3418 	void *rl_cookie;
3419 	struct vattr vattr;
3420 	struct nameidata nd;
3421 	int error;
3422 
3423 	if (length < 0)
3424 		return (EINVAL);
3425 retry:
3426 	NDINIT(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, td);
3427 	if ((error = namei(&nd)) != 0)
3428 		return (error);
3429 	vp = nd.ni_vp;
3430 	rl_cookie = vn_rangelock_wlock(vp, 0, OFF_MAX);
3431 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) {
3432 		vn_rangelock_unlock(vp, rl_cookie);
3433 		vrele(vp);
3434 		return (error);
3435 	}
3436 	NDFREE(&nd, NDF_ONLY_PNBUF);
3437 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3438 	if (vp->v_type == VDIR)
3439 		error = EISDIR;
3440 #ifdef MAC
3441 	else if ((error = mac_vnode_check_write(td->td_ucred, NOCRED, vp))) {
3442 	}
3443 #endif
3444 	else if ((error = vn_writechk(vp)) == 0 &&
3445 	    (error = VOP_ACCESS(vp, VWRITE, td->td_ucred, td)) == 0) {
3446 		VATTR_NULL(&vattr);
3447 		vattr.va_size = length;
3448 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3449 	}
3450 	VOP_UNLOCK(vp);
3451 	vn_finished_write(mp);
3452 	vn_rangelock_unlock(vp, rl_cookie);
3453 	vrele(vp);
3454 	if (error == ERELOOKUP)
3455 		goto retry;
3456 	return (error);
3457 }
3458 
3459 #if defined(COMPAT_43)
3460 /*
3461  * Truncate a file given its path name.
3462  */
3463 #ifndef _SYS_SYSPROTO_H_
3464 struct otruncate_args {
3465 	char	*path;
3466 	long	length;
3467 };
3468 #endif
3469 int
3470 otruncate(struct thread *td, struct otruncate_args *uap)
3471 {
3472 
3473 	return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3474 }
3475 #endif /* COMPAT_43 */
3476 
3477 #if defined(COMPAT_FREEBSD6)
3478 /* Versions with the pad argument */
3479 int
3480 freebsd6_truncate(struct thread *td, struct freebsd6_truncate_args *uap)
3481 {
3482 
3483 	return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3484 }
3485 
3486 int
3487 freebsd6_ftruncate(struct thread *td, struct freebsd6_ftruncate_args *uap)
3488 {
3489 
3490 	return (kern_ftruncate(td, uap->fd, uap->length));
3491 }
3492 #endif
3493 
3494 int
3495 kern_fsync(struct thread *td, int fd, bool fullsync)
3496 {
3497 	struct vnode *vp;
3498 	struct mount *mp;
3499 	struct file *fp;
3500 	int error, lock_flags;
3501 
3502 	AUDIT_ARG_FD(fd);
3503 	error = getvnode(td, fd, &cap_fsync_rights, &fp);
3504 	if (error != 0)
3505 		return (error);
3506 	vp = fp->f_vnode;
3507 #if 0
3508 	if (!fullsync)
3509 		/* XXXKIB: compete outstanding aio writes */;
3510 #endif
3511 retry:
3512 	error = vn_start_write(vp, &mp, V_WAIT | PCATCH);
3513 	if (error != 0)
3514 		goto drop;
3515 	if (MNT_SHARED_WRITES(mp) ||
3516 	    ((mp == NULL) && MNT_SHARED_WRITES(vp->v_mount))) {
3517 		lock_flags = LK_SHARED;
3518 	} else {
3519 		lock_flags = LK_EXCLUSIVE;
3520 	}
3521 	vn_lock(vp, lock_flags | LK_RETRY);
3522 	AUDIT_ARG_VNODE1(vp);
3523 	if (vp->v_object != NULL) {
3524 		VM_OBJECT_WLOCK(vp->v_object);
3525 		vm_object_page_clean(vp->v_object, 0, 0, 0);
3526 		VM_OBJECT_WUNLOCK(vp->v_object);
3527 	}
3528 	error = fullsync ? VOP_FSYNC(vp, MNT_WAIT, td) : VOP_FDATASYNC(vp, td);
3529 	VOP_UNLOCK(vp);
3530 	vn_finished_write(mp);
3531 	if (error == ERELOOKUP)
3532 		goto retry;
3533 drop:
3534 	fdrop(fp, td);
3535 	return (error);
3536 }
3537 
3538 /*
3539  * Sync an open file.
3540  */
3541 #ifndef _SYS_SYSPROTO_H_
3542 struct fsync_args {
3543 	int	fd;
3544 };
3545 #endif
3546 int
3547 sys_fsync(struct thread *td, struct fsync_args *uap)
3548 {
3549 
3550 	return (kern_fsync(td, uap->fd, true));
3551 }
3552 
3553 int
3554 sys_fdatasync(struct thread *td, struct fdatasync_args *uap)
3555 {
3556 
3557 	return (kern_fsync(td, uap->fd, false));
3558 }
3559 
3560 /*
3561  * Rename files.  Source and destination must either both be directories, or
3562  * both not be directories.  If target is a directory, it must be empty.
3563  */
3564 #ifndef _SYS_SYSPROTO_H_
3565 struct rename_args {
3566 	char	*from;
3567 	char	*to;
3568 };
3569 #endif
3570 int
3571 sys_rename(struct thread *td, struct rename_args *uap)
3572 {
3573 
3574 	return (kern_renameat(td, AT_FDCWD, uap->from, AT_FDCWD,
3575 	    uap->to, UIO_USERSPACE));
3576 }
3577 
3578 #ifndef _SYS_SYSPROTO_H_
3579 struct renameat_args {
3580 	int	oldfd;
3581 	char	*old;
3582 	int	newfd;
3583 	char	*new;
3584 };
3585 #endif
3586 int
3587 sys_renameat(struct thread *td, struct renameat_args *uap)
3588 {
3589 
3590 	return (kern_renameat(td, uap->oldfd, uap->old, uap->newfd, uap->new,
3591 	    UIO_USERSPACE));
3592 }
3593 
3594 #ifdef MAC
3595 static int
3596 kern_renameat_mac(struct thread *td, int oldfd, const char *old, int newfd,
3597     const char *new, enum uio_seg pathseg, struct nameidata *fromnd)
3598 {
3599 	int error;
3600 
3601 	NDINIT_ATRIGHTS(fromnd, DELETE, LOCKPARENT | LOCKLEAF | SAVESTART |
3602 	    AUDITVNODE1, pathseg, old, oldfd, &cap_renameat_source_rights, td);
3603 	if ((error = namei(fromnd)) != 0)
3604 		return (error);
3605 	error = mac_vnode_check_rename_from(td->td_ucred, fromnd->ni_dvp,
3606 	    fromnd->ni_vp, &fromnd->ni_cnd);
3607 	VOP_UNLOCK(fromnd->ni_dvp);
3608 	if (fromnd->ni_dvp != fromnd->ni_vp)
3609 		VOP_UNLOCK(fromnd->ni_vp);
3610 	if (error != 0) {
3611 		NDFREE(fromnd, NDF_ONLY_PNBUF);
3612 		vrele(fromnd->ni_dvp);
3613 		vrele(fromnd->ni_vp);
3614 		if (fromnd->ni_startdir)
3615 			vrele(fromnd->ni_startdir);
3616 	}
3617 	return (error);
3618 }
3619 #endif
3620 
3621 int
3622 kern_renameat(struct thread *td, int oldfd, const char *old, int newfd,
3623     const char *new, enum uio_seg pathseg)
3624 {
3625 	struct mount *mp = NULL;
3626 	struct vnode *tvp, *fvp, *tdvp;
3627 	struct nameidata fromnd, tond;
3628 	u_int64_t tondflags;
3629 	int error;
3630 
3631 again:
3632 	bwillwrite();
3633 #ifdef MAC
3634 	if (mac_vnode_check_rename_from_enabled()) {
3635 		error = kern_renameat_mac(td, oldfd, old, newfd, new, pathseg,
3636 		    &fromnd);
3637 		if (error != 0)
3638 			return (error);
3639 	} else {
3640 #endif
3641 	NDINIT_ATRIGHTS(&fromnd, DELETE, WANTPARENT | SAVESTART | AUDITVNODE1,
3642 	    pathseg, old, oldfd, &cap_renameat_source_rights, td);
3643 	if ((error = namei(&fromnd)) != 0)
3644 		return (error);
3645 #ifdef MAC
3646 	}
3647 #endif
3648 	fvp = fromnd.ni_vp;
3649 	tondflags = LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART | AUDITVNODE2;
3650 	if (fromnd.ni_vp->v_type == VDIR)
3651 		tondflags |= WILLBEDIR;
3652 	NDINIT_ATRIGHTS(&tond, RENAME, tondflags, pathseg, new, newfd,
3653 	    &cap_renameat_target_rights, td);
3654 	if ((error = namei(&tond)) != 0) {
3655 		/* Translate error code for rename("dir1", "dir2/."). */
3656 		if (error == EISDIR && fvp->v_type == VDIR)
3657 			error = EINVAL;
3658 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3659 		vrele(fromnd.ni_dvp);
3660 		vrele(fvp);
3661 		goto out1;
3662 	}
3663 	tdvp = tond.ni_dvp;
3664 	tvp = tond.ni_vp;
3665 	error = vn_start_write(fvp, &mp, V_NOWAIT);
3666 	if (error != 0) {
3667 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3668 		NDFREE(&tond, NDF_ONLY_PNBUF);
3669 		if (tvp != NULL)
3670 			vput(tvp);
3671 		if (tdvp == tvp)
3672 			vrele(tdvp);
3673 		else
3674 			vput(tdvp);
3675 		vrele(fromnd.ni_dvp);
3676 		vrele(fvp);
3677 		vrele(tond.ni_startdir);
3678 		if (fromnd.ni_startdir != NULL)
3679 			vrele(fromnd.ni_startdir);
3680 		error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH);
3681 		if (error != 0)
3682 			return (error);
3683 		goto again;
3684 	}
3685 	if (tvp != NULL) {
3686 		if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
3687 			error = ENOTDIR;
3688 			goto out;
3689 		} else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
3690 			error = EISDIR;
3691 			goto out;
3692 		}
3693 #ifdef CAPABILITIES
3694 		if (newfd != AT_FDCWD && (tond.ni_resflags & NIRES_ABS) == 0) {
3695 			/*
3696 			 * If the target already exists we require CAP_UNLINKAT
3697 			 * from 'newfd', when newfd was used for the lookup.
3698 			 */
3699 			error = cap_check(&tond.ni_filecaps.fc_rights,
3700 			    &cap_unlinkat_rights);
3701 			if (error != 0)
3702 				goto out;
3703 		}
3704 #endif
3705 	}
3706 	if (fvp == tdvp) {
3707 		error = EINVAL;
3708 		goto out;
3709 	}
3710 	/*
3711 	 * If the source is the same as the destination (that is, if they
3712 	 * are links to the same vnode), then there is nothing to do.
3713 	 */
3714 	if (fvp == tvp)
3715 		error = ERESTART;
3716 #ifdef MAC
3717 	else
3718 		error = mac_vnode_check_rename_to(td->td_ucred, tdvp,
3719 		    tond.ni_vp, fromnd.ni_dvp == tdvp, &tond.ni_cnd);
3720 #endif
3721 out:
3722 	if (error == 0) {
3723 		error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
3724 		    tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
3725 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3726 		NDFREE(&tond, NDF_ONLY_PNBUF);
3727 	} else {
3728 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3729 		NDFREE(&tond, NDF_ONLY_PNBUF);
3730 		if (tvp != NULL)
3731 			vput(tvp);
3732 		if (tdvp == tvp)
3733 			vrele(tdvp);
3734 		else
3735 			vput(tdvp);
3736 		vrele(fromnd.ni_dvp);
3737 		vrele(fvp);
3738 	}
3739 	vrele(tond.ni_startdir);
3740 	vn_finished_write(mp);
3741 out1:
3742 	if (fromnd.ni_startdir)
3743 		vrele(fromnd.ni_startdir);
3744 	if (error == ERESTART)
3745 		return (0);
3746 	if (error == ERELOOKUP)
3747 		goto again;
3748 	return (error);
3749 }
3750 
3751 /*
3752  * Make a directory file.
3753  */
3754 #ifndef _SYS_SYSPROTO_H_
3755 struct mkdir_args {
3756 	char	*path;
3757 	int	mode;
3758 };
3759 #endif
3760 int
3761 sys_mkdir(struct thread *td, struct mkdir_args *uap)
3762 {
3763 
3764 	return (kern_mkdirat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
3765 	    uap->mode));
3766 }
3767 
3768 #ifndef _SYS_SYSPROTO_H_
3769 struct mkdirat_args {
3770 	int	fd;
3771 	char	*path;
3772 	mode_t	mode;
3773 };
3774 #endif
3775 int
3776 sys_mkdirat(struct thread *td, struct mkdirat_args *uap)
3777 {
3778 
3779 	return (kern_mkdirat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode));
3780 }
3781 
3782 int
3783 kern_mkdirat(struct thread *td, int fd, const char *path, enum uio_seg segflg,
3784     int mode)
3785 {
3786 	struct mount *mp;
3787 	struct vattr vattr;
3788 	struct nameidata nd;
3789 	int error;
3790 
3791 	AUDIT_ARG_MODE(mode);
3792 restart:
3793 	bwillwrite();
3794 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
3795 	    NC_NOMAKEENTRY | NC_KEEPPOSENTRY | FAILIFEXISTS | WILLBEDIR,
3796 	    segflg, path, fd, &cap_mkdirat_rights, td);
3797 	if ((error = namei(&nd)) != 0)
3798 		return (error);
3799 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3800 		NDFREE(&nd, NDF_ONLY_PNBUF);
3801 		vput(nd.ni_dvp);
3802 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3803 			return (error);
3804 		goto restart;
3805 	}
3806 	VATTR_NULL(&vattr);
3807 	vattr.va_type = VDIR;
3808 	vattr.va_mode = (mode & ACCESSPERMS) &~ td->td_proc->p_pd->pd_cmask;
3809 #ifdef MAC
3810 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
3811 	    &vattr);
3812 	if (error != 0)
3813 		goto out;
3814 #endif
3815 	error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
3816 #ifdef MAC
3817 out:
3818 #endif
3819 	NDFREE(&nd, NDF_ONLY_PNBUF);
3820 	VOP_VPUT_PAIR(nd.ni_dvp, error == 0 ? &nd.ni_vp : NULL, true);
3821 	vn_finished_write(mp);
3822 	if (error == ERELOOKUP)
3823 		goto restart;
3824 	return (error);
3825 }
3826 
3827 /*
3828  * Remove a directory file.
3829  */
3830 #ifndef _SYS_SYSPROTO_H_
3831 struct rmdir_args {
3832 	char	*path;
3833 };
3834 #endif
3835 int
3836 sys_rmdir(struct thread *td, struct rmdir_args *uap)
3837 {
3838 
3839 	return (kern_frmdirat(td, AT_FDCWD, uap->path, FD_NONE, UIO_USERSPACE,
3840 	    0));
3841 }
3842 
3843 int
3844 kern_frmdirat(struct thread *td, int dfd, const char *path, int fd,
3845     enum uio_seg pathseg, int flag)
3846 {
3847 	struct mount *mp;
3848 	struct vnode *vp;
3849 	struct file *fp;
3850 	struct nameidata nd;
3851 	cap_rights_t rights;
3852 	int error;
3853 
3854 	fp = NULL;
3855 	if (fd != FD_NONE) {
3856 		error = getvnode(td, fd, cap_rights_init_one(&rights,
3857 		    CAP_LOOKUP), &fp);
3858 		if (error != 0)
3859 			return (error);
3860 	}
3861 
3862 restart:
3863 	bwillwrite();
3864 	NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1 |
3865 	    at2cnpflags(flag, AT_RESOLVE_BENEATH),
3866 	    pathseg, path, dfd, &cap_unlinkat_rights, td);
3867 	if ((error = namei(&nd)) != 0)
3868 		goto fdout;
3869 	vp = nd.ni_vp;
3870 	if (vp->v_type != VDIR) {
3871 		error = ENOTDIR;
3872 		goto out;
3873 	}
3874 	/*
3875 	 * No rmdir "." please.
3876 	 */
3877 	if (nd.ni_dvp == vp) {
3878 		error = EINVAL;
3879 		goto out;
3880 	}
3881 	/*
3882 	 * The root of a mounted filesystem cannot be deleted.
3883 	 */
3884 	if (vp->v_vflag & VV_ROOT) {
3885 		error = EBUSY;
3886 		goto out;
3887 	}
3888 
3889 	if (fp != NULL && fp->f_vnode != vp) {
3890 		if (VN_IS_DOOMED(fp->f_vnode))
3891 			error = EBADF;
3892 		else
3893 			error = EDEADLK;
3894 		goto out;
3895 	}
3896 
3897 #ifdef MAC
3898 	error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
3899 	    &nd.ni_cnd);
3900 	if (error != 0)
3901 		goto out;
3902 #endif
3903 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3904 		NDFREE(&nd, NDF_ONLY_PNBUF);
3905 		vput(vp);
3906 		if (nd.ni_dvp == vp)
3907 			vrele(nd.ni_dvp);
3908 		else
3909 			vput(nd.ni_dvp);
3910 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3911 			goto fdout;
3912 		goto restart;
3913 	}
3914 	vfs_notify_upper(vp, VFS_NOTIFY_UPPER_UNLINK);
3915 	error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
3916 	vn_finished_write(mp);
3917 out:
3918 	NDFREE(&nd, NDF_ONLY_PNBUF);
3919 	vput(vp);
3920 	if (nd.ni_dvp == vp)
3921 		vrele(nd.ni_dvp);
3922 	else
3923 		vput(nd.ni_dvp);
3924 	if (error == ERELOOKUP)
3925 		goto restart;
3926 fdout:
3927 	if (fp != NULL)
3928 		fdrop(fp, td);
3929 	return (error);
3930 }
3931 
3932 #if defined(COMPAT_43) || defined(COMPAT_FREEBSD11)
3933 int
3934 freebsd11_kern_getdirentries(struct thread *td, int fd, char *ubuf, u_int count,
3935     long *basep, void (*func)(struct freebsd11_dirent *))
3936 {
3937 	struct freebsd11_dirent dstdp;
3938 	struct dirent *dp, *edp;
3939 	char *dirbuf;
3940 	off_t base;
3941 	ssize_t resid, ucount;
3942 	int error;
3943 
3944 	/* XXX arbitrary sanity limit on `count'. */
3945 	count = min(count, 64 * 1024);
3946 
3947 	dirbuf = malloc(count, M_TEMP, M_WAITOK);
3948 
3949 	error = kern_getdirentries(td, fd, dirbuf, count, &base, &resid,
3950 	    UIO_SYSSPACE);
3951 	if (error != 0)
3952 		goto done;
3953 	if (basep != NULL)
3954 		*basep = base;
3955 
3956 	ucount = 0;
3957 	for (dp = (struct dirent *)dirbuf,
3958 	    edp = (struct dirent *)&dirbuf[count - resid];
3959 	    ucount < count && dp < edp; ) {
3960 		if (dp->d_reclen == 0)
3961 			break;
3962 		MPASS(dp->d_reclen >= _GENERIC_DIRLEN(0));
3963 		if (dp->d_namlen >= sizeof(dstdp.d_name))
3964 			continue;
3965 		dstdp.d_type = dp->d_type;
3966 		dstdp.d_namlen = dp->d_namlen;
3967 		dstdp.d_fileno = dp->d_fileno;		/* truncate */
3968 		if (dstdp.d_fileno != dp->d_fileno) {
3969 			switch (ino64_trunc_error) {
3970 			default:
3971 			case 0:
3972 				break;
3973 			case 1:
3974 				error = EOVERFLOW;
3975 				goto done;
3976 			case 2:
3977 				dstdp.d_fileno = UINT32_MAX;
3978 				break;
3979 			}
3980 		}
3981 		dstdp.d_reclen = sizeof(dstdp) - sizeof(dstdp.d_name) +
3982 		    ((dp->d_namlen + 1 + 3) &~ 3);
3983 		bcopy(dp->d_name, dstdp.d_name, dstdp.d_namlen);
3984 		bzero(dstdp.d_name + dstdp.d_namlen,
3985 		    dstdp.d_reclen - offsetof(struct freebsd11_dirent, d_name) -
3986 		    dstdp.d_namlen);
3987 		MPASS(dstdp.d_reclen <= dp->d_reclen);
3988 		MPASS(ucount + dstdp.d_reclen <= count);
3989 		if (func != NULL)
3990 			func(&dstdp);
3991 		error = copyout(&dstdp, ubuf + ucount, dstdp.d_reclen);
3992 		if (error != 0)
3993 			break;
3994 		dp = (struct dirent *)((char *)dp + dp->d_reclen);
3995 		ucount += dstdp.d_reclen;
3996 	}
3997 
3998 done:
3999 	free(dirbuf, M_TEMP);
4000 	if (error == 0)
4001 		td->td_retval[0] = ucount;
4002 	return (error);
4003 }
4004 #endif /* COMPAT */
4005 
4006 #ifdef COMPAT_43
4007 static void
4008 ogetdirentries_cvt(struct freebsd11_dirent *dp)
4009 {
4010 #if (BYTE_ORDER == LITTLE_ENDIAN)
4011 	/*
4012 	 * The expected low byte of dp->d_namlen is our dp->d_type.
4013 	 * The high MBZ byte of dp->d_namlen is our dp->d_namlen.
4014 	 */
4015 	dp->d_type = dp->d_namlen;
4016 	dp->d_namlen = 0;
4017 #else
4018 	/*
4019 	 * The dp->d_type is the high byte of the expected dp->d_namlen,
4020 	 * so must be zero'ed.
4021 	 */
4022 	dp->d_type = 0;
4023 #endif
4024 }
4025 
4026 /*
4027  * Read a block of directory entries in a filesystem independent format.
4028  */
4029 #ifndef _SYS_SYSPROTO_H_
4030 struct ogetdirentries_args {
4031 	int	fd;
4032 	char	*buf;
4033 	u_int	count;
4034 	long	*basep;
4035 };
4036 #endif
4037 int
4038 ogetdirentries(struct thread *td, struct ogetdirentries_args *uap)
4039 {
4040 	long loff;
4041 	int error;
4042 
4043 	error = kern_ogetdirentries(td, uap, &loff);
4044 	if (error == 0)
4045 		error = copyout(&loff, uap->basep, sizeof(long));
4046 	return (error);
4047 }
4048 
4049 int
4050 kern_ogetdirentries(struct thread *td, struct ogetdirentries_args *uap,
4051     long *ploff)
4052 {
4053 	long base;
4054 	int error;
4055 
4056 	/* XXX arbitrary sanity limit on `count'. */
4057 	if (uap->count > 64 * 1024)
4058 		return (EINVAL);
4059 
4060 	error = freebsd11_kern_getdirentries(td, uap->fd, uap->buf, uap->count,
4061 	    &base, ogetdirentries_cvt);
4062 
4063 	if (error == 0 && uap->basep != NULL)
4064 		error = copyout(&base, uap->basep, sizeof(long));
4065 
4066 	return (error);
4067 }
4068 #endif /* COMPAT_43 */
4069 
4070 #if defined(COMPAT_FREEBSD11)
4071 #ifndef _SYS_SYSPROTO_H_
4072 struct freebsd11_getdirentries_args {
4073 	int	fd;
4074 	char	*buf;
4075 	u_int	count;
4076 	long	*basep;
4077 };
4078 #endif
4079 int
4080 freebsd11_getdirentries(struct thread *td,
4081     struct freebsd11_getdirentries_args *uap)
4082 {
4083 	long base;
4084 	int error;
4085 
4086 	error = freebsd11_kern_getdirentries(td, uap->fd, uap->buf, uap->count,
4087 	    &base, NULL);
4088 
4089 	if (error == 0 && uap->basep != NULL)
4090 		error = copyout(&base, uap->basep, sizeof(long));
4091 	return (error);
4092 }
4093 
4094 int
4095 freebsd11_getdents(struct thread *td, struct freebsd11_getdents_args *uap)
4096 {
4097 	struct freebsd11_getdirentries_args ap;
4098 
4099 	ap.fd = uap->fd;
4100 	ap.buf = uap->buf;
4101 	ap.count = uap->count;
4102 	ap.basep = NULL;
4103 	return (freebsd11_getdirentries(td, &ap));
4104 }
4105 #endif /* COMPAT_FREEBSD11 */
4106 
4107 /*
4108  * Read a block of directory entries in a filesystem independent format.
4109  */
4110 int
4111 sys_getdirentries(struct thread *td, struct getdirentries_args *uap)
4112 {
4113 	off_t base;
4114 	int error;
4115 
4116 	error = kern_getdirentries(td, uap->fd, uap->buf, uap->count, &base,
4117 	    NULL, UIO_USERSPACE);
4118 	if (error != 0)
4119 		return (error);
4120 	if (uap->basep != NULL)
4121 		error = copyout(&base, uap->basep, sizeof(off_t));
4122 	return (error);
4123 }
4124 
4125 int
4126 kern_getdirentries(struct thread *td, int fd, char *buf, size_t count,
4127     off_t *basep, ssize_t *residp, enum uio_seg bufseg)
4128 {
4129 	struct vnode *vp;
4130 	struct file *fp;
4131 	struct uio auio;
4132 	struct iovec aiov;
4133 	off_t loff;
4134 	int error, eofflag;
4135 	off_t foffset;
4136 
4137 	AUDIT_ARG_FD(fd);
4138 	if (count > IOSIZE_MAX)
4139 		return (EINVAL);
4140 	auio.uio_resid = count;
4141 	error = getvnode(td, fd, &cap_read_rights, &fp);
4142 	if (error != 0)
4143 		return (error);
4144 	if ((fp->f_flag & FREAD) == 0) {
4145 		fdrop(fp, td);
4146 		return (EBADF);
4147 	}
4148 	vp = fp->f_vnode;
4149 	foffset = foffset_lock(fp, 0);
4150 unionread:
4151 	if (vp->v_type != VDIR) {
4152 		error = EINVAL;
4153 		goto fail;
4154 	}
4155 	aiov.iov_base = buf;
4156 	aiov.iov_len = count;
4157 	auio.uio_iov = &aiov;
4158 	auio.uio_iovcnt = 1;
4159 	auio.uio_rw = UIO_READ;
4160 	auio.uio_segflg = bufseg;
4161 	auio.uio_td = td;
4162 	vn_lock(vp, LK_SHARED | LK_RETRY);
4163 	AUDIT_ARG_VNODE1(vp);
4164 	loff = auio.uio_offset = foffset;
4165 #ifdef MAC
4166 	error = mac_vnode_check_readdir(td->td_ucred, vp);
4167 	if (error == 0)
4168 #endif
4169 		error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, NULL,
4170 		    NULL);
4171 	foffset = auio.uio_offset;
4172 	if (error != 0) {
4173 		VOP_UNLOCK(vp);
4174 		goto fail;
4175 	}
4176 	if (count == auio.uio_resid &&
4177 	    (vp->v_vflag & VV_ROOT) &&
4178 	    (vp->v_mount->mnt_flag & MNT_UNION)) {
4179 		struct vnode *tvp = vp;
4180 
4181 		vp = vp->v_mount->mnt_vnodecovered;
4182 		VREF(vp);
4183 		fp->f_vnode = vp;
4184 		foffset = 0;
4185 		vput(tvp);
4186 		goto unionread;
4187 	}
4188 	VOP_UNLOCK(vp);
4189 	*basep = loff;
4190 	if (residp != NULL)
4191 		*residp = auio.uio_resid;
4192 	td->td_retval[0] = count - auio.uio_resid;
4193 fail:
4194 	foffset_unlock(fp, foffset, 0);
4195 	fdrop(fp, td);
4196 	return (error);
4197 }
4198 
4199 /*
4200  * Set the mode mask for creation of filesystem nodes.
4201  */
4202 #ifndef _SYS_SYSPROTO_H_
4203 struct umask_args {
4204 	int	newmask;
4205 };
4206 #endif
4207 int
4208 sys_umask(struct thread *td, struct umask_args *uap)
4209 {
4210 	struct pwddesc *pdp;
4211 
4212 	pdp = td->td_proc->p_pd;
4213 	PWDDESC_XLOCK(pdp);
4214 	td->td_retval[0] = pdp->pd_cmask;
4215 	pdp->pd_cmask = uap->newmask & ALLPERMS;
4216 	PWDDESC_XUNLOCK(pdp);
4217 	return (0);
4218 }
4219 
4220 /*
4221  * Void all references to file by ripping underlying filesystem away from
4222  * vnode.
4223  */
4224 #ifndef _SYS_SYSPROTO_H_
4225 struct revoke_args {
4226 	char	*path;
4227 };
4228 #endif
4229 int
4230 sys_revoke(struct thread *td, struct revoke_args *uap)
4231 {
4232 	struct vnode *vp;
4233 	struct vattr vattr;
4234 	struct nameidata nd;
4235 	int error;
4236 
4237 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
4238 	    uap->path, td);
4239 	if ((error = namei(&nd)) != 0)
4240 		return (error);
4241 	vp = nd.ni_vp;
4242 	NDFREE_NOTHING(&nd);
4243 	if (vp->v_type != VCHR || vp->v_rdev == NULL) {
4244 		error = EINVAL;
4245 		goto out;
4246 	}
4247 #ifdef MAC
4248 	error = mac_vnode_check_revoke(td->td_ucred, vp);
4249 	if (error != 0)
4250 		goto out;
4251 #endif
4252 	error = VOP_GETATTR(vp, &vattr, td->td_ucred);
4253 	if (error != 0)
4254 		goto out;
4255 	if (td->td_ucred->cr_uid != vattr.va_uid) {
4256 		error = priv_check(td, PRIV_VFS_ADMIN);
4257 		if (error != 0)
4258 			goto out;
4259 	}
4260 	if (devfs_usecount(vp) > 0)
4261 		VOP_REVOKE(vp, REVOKEALL);
4262 out:
4263 	vput(vp);
4264 	return (error);
4265 }
4266 
4267 /*
4268  * This variant of getvnode() allows O_PATH files.  Caller should
4269  * ensure that returned file and vnode are only used for compatible
4270  * semantics.
4271  */
4272 int
4273 getvnode_path(struct thread *td, int fd, cap_rights_t *rightsp,
4274     struct file **fpp)
4275 {
4276 	struct file *fp;
4277 	int error;
4278 
4279 	error = fget_unlocked(td->td_proc->p_fd, fd, rightsp, &fp);
4280 	if (error != 0)
4281 		return (error);
4282 
4283 	/*
4284 	 * The file could be not of the vnode type, or it may be not
4285 	 * yet fully initialized, in which case the f_vnode pointer
4286 	 * may be set, but f_ops is still badfileops.  E.g.,
4287 	 * devfs_open() transiently create such situation to
4288 	 * facilitate csw d_fdopen().
4289 	 *
4290 	 * Dupfdopen() handling in kern_openat() installs the
4291 	 * half-baked file into the process descriptor table, allowing
4292 	 * other thread to dereference it. Guard against the race by
4293 	 * checking f_ops.
4294 	 */
4295 	if (fp->f_vnode == NULL || fp->f_ops == &badfileops) {
4296 		fdrop(fp, td);
4297 		return (EINVAL);
4298 	}
4299 
4300 	*fpp = fp;
4301 	return (0);
4302 }
4303 
4304 /*
4305  * Convert a user file descriptor to a kernel file entry and check
4306  * that, if it is a capability, the correct rights are present.
4307  * A reference on the file entry is held upon returning.
4308  */
4309 int
4310 getvnode(struct thread *td, int fd, cap_rights_t *rightsp, struct file **fpp)
4311 {
4312 	int error;
4313 
4314 	error = getvnode_path(td, fd, rightsp, fpp);
4315 
4316 	/*
4317 	 * Filter out O_PATH file descriptors, most getvnode() callers
4318 	 * do not call fo_ methods.
4319 	 */
4320 	if (error == 0 && (*fpp)->f_ops == &path_fileops) {
4321 		fdrop(*fpp, td);
4322 		error = EBADF;
4323 	}
4324 
4325 	return (error);
4326 }
4327 
4328 /*
4329  * Get an (NFS) file handle.
4330  */
4331 #ifndef _SYS_SYSPROTO_H_
4332 struct lgetfh_args {
4333 	char *fname;
4334 	fhandle_t *fhp;
4335 };
4336 #endif
4337 int
4338 sys_lgetfh(struct thread *td, struct lgetfh_args *uap)
4339 {
4340 
4341 	return (kern_getfhat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->fname,
4342 	    UIO_USERSPACE, uap->fhp, UIO_USERSPACE));
4343 }
4344 
4345 #ifndef _SYS_SYSPROTO_H_
4346 struct getfh_args {
4347 	char *fname;
4348 	fhandle_t *fhp;
4349 };
4350 #endif
4351 int
4352 sys_getfh(struct thread *td, struct getfh_args *uap)
4353 {
4354 
4355 	return (kern_getfhat(td, 0, AT_FDCWD, uap->fname, UIO_USERSPACE,
4356 	    uap->fhp, UIO_USERSPACE));
4357 }
4358 
4359 /*
4360  * syscall for the rpc.lockd to use to translate an open descriptor into
4361  * a NFS file handle.
4362  *
4363  * warning: do not remove the priv_check() call or this becomes one giant
4364  * security hole.
4365  */
4366 #ifndef _SYS_SYSPROTO_H_
4367 struct getfhat_args {
4368 	int fd;
4369 	char *path;
4370 	fhandle_t *fhp;
4371 	int flags;
4372 };
4373 #endif
4374 int
4375 sys_getfhat(struct thread *td, struct getfhat_args *uap)
4376 {
4377 
4378 	if ((uap->flags & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH)) != 0)
4379 		return (EINVAL);
4380 	return (kern_getfhat(td, uap->flags, uap->fd, uap->path, UIO_USERSPACE,
4381 	    uap->fhp, UIO_USERSPACE));
4382 }
4383 
4384 int
4385 kern_getfhat(struct thread *td, int flags, int fd, const char *path,
4386     enum uio_seg pathseg, fhandle_t *fhp, enum uio_seg fhseg)
4387 {
4388 	struct nameidata nd;
4389 	fhandle_t fh;
4390 	struct vnode *vp;
4391 	int error;
4392 
4393 	error = priv_check(td, PRIV_VFS_GETFH);
4394 	if (error != 0)
4395 		return (error);
4396 	NDINIT_AT(&nd, LOOKUP, at2cnpflags(flags, AT_SYMLINK_NOFOLLOW |
4397 	    AT_RESOLVE_BENEATH) | LOCKLEAF | AUDITVNODE1, pathseg, path,
4398 	    fd, td);
4399 	error = namei(&nd);
4400 	if (error != 0)
4401 		return (error);
4402 	NDFREE_NOTHING(&nd);
4403 	vp = nd.ni_vp;
4404 	bzero(&fh, sizeof(fh));
4405 	fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
4406 	error = VOP_VPTOFH(vp, &fh.fh_fid);
4407 	vput(vp);
4408 	if (error == 0) {
4409 		if (fhseg == UIO_USERSPACE)
4410 			error = copyout(&fh, fhp, sizeof (fh));
4411 		else
4412 			memcpy(fhp, &fh, sizeof(fh));
4413 	}
4414 	return (error);
4415 }
4416 
4417 #ifndef _SYS_SYSPROTO_H_
4418 struct fhlink_args {
4419 	fhandle_t *fhp;
4420 	const char *to;
4421 };
4422 #endif
4423 int
4424 sys_fhlink(struct thread *td, struct fhlink_args *uap)
4425 {
4426 
4427 	return (kern_fhlinkat(td, AT_FDCWD, uap->to, UIO_USERSPACE, uap->fhp));
4428 }
4429 
4430 #ifndef _SYS_SYSPROTO_H_
4431 struct fhlinkat_args {
4432 	fhandle_t *fhp;
4433 	int tofd;
4434 	const char *to;
4435 };
4436 #endif
4437 int
4438 sys_fhlinkat(struct thread *td, struct fhlinkat_args *uap)
4439 {
4440 
4441 	return (kern_fhlinkat(td, uap->tofd, uap->to, UIO_USERSPACE, uap->fhp));
4442 }
4443 
4444 static int
4445 kern_fhlinkat(struct thread *td, int fd, const char *path,
4446     enum uio_seg pathseg, fhandle_t *fhp)
4447 {
4448 	fhandle_t fh;
4449 	struct mount *mp;
4450 	struct vnode *vp;
4451 	int error;
4452 
4453 	error = priv_check(td, PRIV_VFS_GETFH);
4454 	if (error != 0)
4455 		return (error);
4456 	error = copyin(fhp, &fh, sizeof(fh));
4457 	if (error != 0)
4458 		return (error);
4459 	do {
4460 		bwillwrite();
4461 		if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4462 			return (ESTALE);
4463 		error = VFS_FHTOVP(mp, &fh.fh_fid, LK_SHARED, &vp);
4464 		vfs_unbusy(mp);
4465 		if (error != 0)
4466 			return (error);
4467 		VOP_UNLOCK(vp);
4468 		error = kern_linkat_vp(td, vp, fd, path, pathseg);
4469 	} while (error == EAGAIN || error == ERELOOKUP);
4470 	return (error);
4471 }
4472 
4473 #ifndef _SYS_SYSPROTO_H_
4474 struct fhreadlink_args {
4475 	fhandle_t *fhp;
4476 	char *buf;
4477 	size_t bufsize;
4478 };
4479 #endif
4480 int
4481 sys_fhreadlink(struct thread *td, struct fhreadlink_args *uap)
4482 {
4483 	fhandle_t fh;
4484 	struct mount *mp;
4485 	struct vnode *vp;
4486 	int error;
4487 
4488 	error = priv_check(td, PRIV_VFS_GETFH);
4489 	if (error != 0)
4490 		return (error);
4491 	if (uap->bufsize > IOSIZE_MAX)
4492 		return (EINVAL);
4493 	error = copyin(uap->fhp, &fh, sizeof(fh));
4494 	if (error != 0)
4495 		return (error);
4496 	if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4497 		return (ESTALE);
4498 	error = VFS_FHTOVP(mp, &fh.fh_fid, LK_SHARED, &vp);
4499 	vfs_unbusy(mp);
4500 	if (error != 0)
4501 		return (error);
4502 	error = kern_readlink_vp(vp, uap->buf, UIO_USERSPACE, uap->bufsize, td);
4503 	vput(vp);
4504 	return (error);
4505 }
4506 
4507 /*
4508  * syscall for the rpc.lockd to use to translate a NFS file handle into an
4509  * open descriptor.
4510  *
4511  * warning: do not remove the priv_check() call or this becomes one giant
4512  * security hole.
4513  */
4514 #ifndef _SYS_SYSPROTO_H_
4515 struct fhopen_args {
4516 	const struct fhandle *u_fhp;
4517 	int flags;
4518 };
4519 #endif
4520 int
4521 sys_fhopen(struct thread *td, struct fhopen_args *uap)
4522 {
4523 	return (kern_fhopen(td, uap->u_fhp, uap->flags));
4524 }
4525 
4526 int
4527 kern_fhopen(struct thread *td, const struct fhandle *u_fhp, int flags)
4528 {
4529 	struct mount *mp;
4530 	struct vnode *vp;
4531 	struct fhandle fhp;
4532 	struct file *fp;
4533 	int fmode, error;
4534 	int indx;
4535 
4536 	error = priv_check(td, PRIV_VFS_FHOPEN);
4537 	if (error != 0)
4538 		return (error);
4539 	indx = -1;
4540 	fmode = FFLAGS(flags);
4541 	/* why not allow a non-read/write open for our lockd? */
4542 	if (((fmode & (FREAD | FWRITE)) == 0) || (fmode & O_CREAT))
4543 		return (EINVAL);
4544 	error = copyin(u_fhp, &fhp, sizeof(fhp));
4545 	if (error != 0)
4546 		return(error);
4547 	/* find the mount point */
4548 	mp = vfs_busyfs(&fhp.fh_fsid);
4549 	if (mp == NULL)
4550 		return (ESTALE);
4551 	/* now give me my vnode, it gets returned to me locked */
4552 	error = VFS_FHTOVP(mp, &fhp.fh_fid, LK_EXCLUSIVE, &vp);
4553 	vfs_unbusy(mp);
4554 	if (error != 0)
4555 		return (error);
4556 
4557 	error = falloc_noinstall(td, &fp);
4558 	if (error != 0) {
4559 		vput(vp);
4560 		return (error);
4561 	}
4562 	/*
4563 	 * An extra reference on `fp' has been held for us by
4564 	 * falloc_noinstall().
4565 	 */
4566 
4567 #ifdef INVARIANTS
4568 	td->td_dupfd = -1;
4569 #endif
4570 	error = vn_open_vnode(vp, fmode, td->td_ucred, td, fp);
4571 	if (error != 0) {
4572 		KASSERT(fp->f_ops == &badfileops,
4573 		    ("VOP_OPEN in fhopen() set f_ops"));
4574 		KASSERT(td->td_dupfd < 0,
4575 		    ("fhopen() encountered fdopen()"));
4576 
4577 		vput(vp);
4578 		goto bad;
4579 	}
4580 #ifdef INVARIANTS
4581 	td->td_dupfd = 0;
4582 #endif
4583 	fp->f_vnode = vp;
4584 	finit_vnode(fp, fmode, NULL, &vnops);
4585 	VOP_UNLOCK(vp);
4586 	if ((fmode & O_TRUNC) != 0) {
4587 		error = fo_truncate(fp, 0, td->td_ucred, td);
4588 		if (error != 0)
4589 			goto bad;
4590 	}
4591 
4592 	error = finstall(td, fp, &indx, fmode, NULL);
4593 bad:
4594 	fdrop(fp, td);
4595 	td->td_retval[0] = indx;
4596 	return (error);
4597 }
4598 
4599 /*
4600  * Stat an (NFS) file handle.
4601  */
4602 #ifndef _SYS_SYSPROTO_H_
4603 struct fhstat_args {
4604 	struct fhandle *u_fhp;
4605 	struct stat *sb;
4606 };
4607 #endif
4608 int
4609 sys_fhstat(struct thread *td, struct fhstat_args *uap)
4610 {
4611 	struct stat sb;
4612 	struct fhandle fh;
4613 	int error;
4614 
4615 	error = copyin(uap->u_fhp, &fh, sizeof(fh));
4616 	if (error != 0)
4617 		return (error);
4618 	error = kern_fhstat(td, fh, &sb);
4619 	if (error == 0)
4620 		error = copyout(&sb, uap->sb, sizeof(sb));
4621 	return (error);
4622 }
4623 
4624 int
4625 kern_fhstat(struct thread *td, struct fhandle fh, struct stat *sb)
4626 {
4627 	struct mount *mp;
4628 	struct vnode *vp;
4629 	int error;
4630 
4631 	error = priv_check(td, PRIV_VFS_FHSTAT);
4632 	if (error != 0)
4633 		return (error);
4634 	if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4635 		return (ESTALE);
4636 	error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4637 	vfs_unbusy(mp);
4638 	if (error != 0)
4639 		return (error);
4640 	error = VOP_STAT(vp, sb, td->td_ucred, NOCRED, td);
4641 	vput(vp);
4642 	return (error);
4643 }
4644 
4645 /*
4646  * Implement fstatfs() for (NFS) file handles.
4647  */
4648 #ifndef _SYS_SYSPROTO_H_
4649 struct fhstatfs_args {
4650 	struct fhandle *u_fhp;
4651 	struct statfs *buf;
4652 };
4653 #endif
4654 int
4655 sys_fhstatfs(struct thread *td, struct fhstatfs_args *uap)
4656 {
4657 	struct statfs *sfp;
4658 	fhandle_t fh;
4659 	int error;
4660 
4661 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
4662 	if (error != 0)
4663 		return (error);
4664 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
4665 	error = kern_fhstatfs(td, fh, sfp);
4666 	if (error == 0)
4667 		error = copyout(sfp, uap->buf, sizeof(*sfp));
4668 	free(sfp, M_STATFS);
4669 	return (error);
4670 }
4671 
4672 int
4673 kern_fhstatfs(struct thread *td, fhandle_t fh, struct statfs *buf)
4674 {
4675 	struct mount *mp;
4676 	struct vnode *vp;
4677 	int error;
4678 
4679 	error = priv_check(td, PRIV_VFS_FHSTATFS);
4680 	if (error != 0)
4681 		return (error);
4682 	if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4683 		return (ESTALE);
4684 	error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4685 	if (error != 0) {
4686 		vfs_unbusy(mp);
4687 		return (error);
4688 	}
4689 	vput(vp);
4690 	error = prison_canseemount(td->td_ucred, mp);
4691 	if (error != 0)
4692 		goto out;
4693 #ifdef MAC
4694 	error = mac_mount_check_stat(td->td_ucred, mp);
4695 	if (error != 0)
4696 		goto out;
4697 #endif
4698 	error = VFS_STATFS(mp, buf);
4699 out:
4700 	vfs_unbusy(mp);
4701 	return (error);
4702 }
4703 
4704 /*
4705  * Unlike madvise(2), we do not make a best effort to remember every
4706  * possible caching hint.  Instead, we remember the last setting with
4707  * the exception that we will allow POSIX_FADV_NORMAL to adjust the
4708  * region of any current setting.
4709  */
4710 int
4711 kern_posix_fadvise(struct thread *td, int fd, off_t offset, off_t len,
4712     int advice)
4713 {
4714 	struct fadvise_info *fa, *new;
4715 	struct file *fp;
4716 	struct vnode *vp;
4717 	off_t end;
4718 	int error;
4719 
4720 	if (offset < 0 || len < 0 || offset > OFF_MAX - len)
4721 		return (EINVAL);
4722 	AUDIT_ARG_VALUE(advice);
4723 	switch (advice) {
4724 	case POSIX_FADV_SEQUENTIAL:
4725 	case POSIX_FADV_RANDOM:
4726 	case POSIX_FADV_NOREUSE:
4727 		new = malloc(sizeof(*fa), M_FADVISE, M_WAITOK);
4728 		break;
4729 	case POSIX_FADV_NORMAL:
4730 	case POSIX_FADV_WILLNEED:
4731 	case POSIX_FADV_DONTNEED:
4732 		new = NULL;
4733 		break;
4734 	default:
4735 		return (EINVAL);
4736 	}
4737 	/* XXX: CAP_POSIX_FADVISE? */
4738 	AUDIT_ARG_FD(fd);
4739 	error = fget(td, fd, &cap_no_rights, &fp);
4740 	if (error != 0)
4741 		goto out;
4742 	AUDIT_ARG_FILE(td->td_proc, fp);
4743 	if ((fp->f_ops->fo_flags & DFLAG_SEEKABLE) == 0) {
4744 		error = ESPIPE;
4745 		goto out;
4746 	}
4747 	if (fp->f_type != DTYPE_VNODE) {
4748 		error = ENODEV;
4749 		goto out;
4750 	}
4751 	vp = fp->f_vnode;
4752 	if (vp->v_type != VREG) {
4753 		error = ENODEV;
4754 		goto out;
4755 	}
4756 	if (len == 0)
4757 		end = OFF_MAX;
4758 	else
4759 		end = offset + len - 1;
4760 	switch (advice) {
4761 	case POSIX_FADV_SEQUENTIAL:
4762 	case POSIX_FADV_RANDOM:
4763 	case POSIX_FADV_NOREUSE:
4764 		/*
4765 		 * Try to merge any existing non-standard region with
4766 		 * this new region if possible, otherwise create a new
4767 		 * non-standard region for this request.
4768 		 */
4769 		mtx_pool_lock(mtxpool_sleep, fp);
4770 		fa = fp->f_advice;
4771 		if (fa != NULL && fa->fa_advice == advice &&
4772 		    ((fa->fa_start <= end && fa->fa_end >= offset) ||
4773 		    (end != OFF_MAX && fa->fa_start == end + 1) ||
4774 		    (fa->fa_end != OFF_MAX && fa->fa_end + 1 == offset))) {
4775 			if (offset < fa->fa_start)
4776 				fa->fa_start = offset;
4777 			if (end > fa->fa_end)
4778 				fa->fa_end = end;
4779 		} else {
4780 			new->fa_advice = advice;
4781 			new->fa_start = offset;
4782 			new->fa_end = end;
4783 			fp->f_advice = new;
4784 			new = fa;
4785 		}
4786 		mtx_pool_unlock(mtxpool_sleep, fp);
4787 		break;
4788 	case POSIX_FADV_NORMAL:
4789 		/*
4790 		 * If a the "normal" region overlaps with an existing
4791 		 * non-standard region, trim or remove the
4792 		 * non-standard region.
4793 		 */
4794 		mtx_pool_lock(mtxpool_sleep, fp);
4795 		fa = fp->f_advice;
4796 		if (fa != NULL) {
4797 			if (offset <= fa->fa_start && end >= fa->fa_end) {
4798 				new = fa;
4799 				fp->f_advice = NULL;
4800 			} else if (offset <= fa->fa_start &&
4801 			    end >= fa->fa_start)
4802 				fa->fa_start = end + 1;
4803 			else if (offset <= fa->fa_end && end >= fa->fa_end)
4804 				fa->fa_end = offset - 1;
4805 			else if (offset >= fa->fa_start && end <= fa->fa_end) {
4806 				/*
4807 				 * If the "normal" region is a middle
4808 				 * portion of the existing
4809 				 * non-standard region, just remove
4810 				 * the whole thing rather than picking
4811 				 * one side or the other to
4812 				 * preserve.
4813 				 */
4814 				new = fa;
4815 				fp->f_advice = NULL;
4816 			}
4817 		}
4818 		mtx_pool_unlock(mtxpool_sleep, fp);
4819 		break;
4820 	case POSIX_FADV_WILLNEED:
4821 	case POSIX_FADV_DONTNEED:
4822 		error = VOP_ADVISE(vp, offset, end, advice);
4823 		break;
4824 	}
4825 out:
4826 	if (fp != NULL)
4827 		fdrop(fp, td);
4828 	free(new, M_FADVISE);
4829 	return (error);
4830 }
4831 
4832 int
4833 sys_posix_fadvise(struct thread *td, struct posix_fadvise_args *uap)
4834 {
4835 	int error;
4836 
4837 	error = kern_posix_fadvise(td, uap->fd, uap->offset, uap->len,
4838 	    uap->advice);
4839 	return (kern_posix_error(td, error));
4840 }
4841 
4842 int
4843 kern_copy_file_range(struct thread *td, int infd, off_t *inoffp, int outfd,
4844     off_t *outoffp, size_t len, unsigned int flags)
4845 {
4846 	struct file *infp, *outfp;
4847 	struct vnode *invp, *outvp;
4848 	int error;
4849 	size_t retlen;
4850 	void *rl_rcookie, *rl_wcookie;
4851 	off_t savinoff, savoutoff;
4852 
4853 	infp = outfp = NULL;
4854 	rl_rcookie = rl_wcookie = NULL;
4855 	savinoff = -1;
4856 	error = 0;
4857 	retlen = 0;
4858 
4859 	if (flags != 0) {
4860 		error = EINVAL;
4861 		goto out;
4862 	}
4863 	if (len > SSIZE_MAX)
4864 		/*
4865 		 * Although the len argument is size_t, the return argument
4866 		 * is ssize_t (which is signed).  Therefore a size that won't
4867 		 * fit in ssize_t can't be returned.
4868 		 */
4869 		len = SSIZE_MAX;
4870 
4871 	/* Get the file structures for the file descriptors. */
4872 	error = fget_read(td, infd, &cap_read_rights, &infp);
4873 	if (error != 0)
4874 		goto out;
4875 	if (infp->f_ops == &badfileops) {
4876 		error = EBADF;
4877 		goto out;
4878 	}
4879 	if (infp->f_vnode == NULL) {
4880 		error = EINVAL;
4881 		goto out;
4882 	}
4883 	error = fget_write(td, outfd, &cap_write_rights, &outfp);
4884 	if (error != 0)
4885 		goto out;
4886 	if (outfp->f_ops == &badfileops) {
4887 		error = EBADF;
4888 		goto out;
4889 	}
4890 	if (outfp->f_vnode == NULL) {
4891 		error = EINVAL;
4892 		goto out;
4893 	}
4894 
4895 	/* Set the offset pointers to the correct place. */
4896 	if (inoffp == NULL)
4897 		inoffp = &infp->f_offset;
4898 	if (outoffp == NULL)
4899 		outoffp = &outfp->f_offset;
4900 	savinoff = *inoffp;
4901 	savoutoff = *outoffp;
4902 
4903 	invp = infp->f_vnode;
4904 	outvp = outfp->f_vnode;
4905 	/* Sanity check the f_flag bits. */
4906 	if ((outfp->f_flag & (FWRITE | FAPPEND)) != FWRITE ||
4907 	    (infp->f_flag & FREAD) == 0) {
4908 		error = EBADF;
4909 		goto out;
4910 	}
4911 
4912 	/* If len == 0, just return 0. */
4913 	if (len == 0)
4914 		goto out;
4915 
4916 	/*
4917 	 * If infp and outfp refer to the same file, the byte ranges cannot
4918 	 * overlap.
4919 	 */
4920 	if (invp == outvp && ((savinoff <= savoutoff && savinoff + len >
4921 	    savoutoff) || (savinoff > savoutoff && savoutoff + len >
4922 	    savinoff))) {
4923 		error = EINVAL;
4924 		goto out;
4925 	}
4926 
4927 	/* Range lock the byte ranges for both invp and outvp. */
4928 	for (;;) {
4929 		rl_wcookie = vn_rangelock_wlock(outvp, *outoffp, *outoffp +
4930 		    len);
4931 		rl_rcookie = vn_rangelock_tryrlock(invp, *inoffp, *inoffp +
4932 		    len);
4933 		if (rl_rcookie != NULL)
4934 			break;
4935 		vn_rangelock_unlock(outvp, rl_wcookie);
4936 		rl_rcookie = vn_rangelock_rlock(invp, *inoffp, *inoffp + len);
4937 		vn_rangelock_unlock(invp, rl_rcookie);
4938 	}
4939 
4940 	retlen = len;
4941 	error = vn_copy_file_range(invp, inoffp, outvp, outoffp, &retlen,
4942 	    flags, infp->f_cred, outfp->f_cred, td);
4943 out:
4944 	if (rl_rcookie != NULL)
4945 		vn_rangelock_unlock(invp, rl_rcookie);
4946 	if (rl_wcookie != NULL)
4947 		vn_rangelock_unlock(outvp, rl_wcookie);
4948 	if (savinoff != -1 && (error == EINTR || error == ERESTART)) {
4949 		*inoffp = savinoff;
4950 		*outoffp = savoutoff;
4951 	}
4952 	if (outfp != NULL)
4953 		fdrop(outfp, td);
4954 	if (infp != NULL)
4955 		fdrop(infp, td);
4956 	td->td_retval[0] = retlen;
4957 	return (error);
4958 }
4959 
4960 int
4961 sys_copy_file_range(struct thread *td, struct copy_file_range_args *uap)
4962 {
4963 	off_t inoff, outoff, *inoffp, *outoffp;
4964 	int error;
4965 
4966 	inoffp = outoffp = NULL;
4967 	if (uap->inoffp != NULL) {
4968 		error = copyin(uap->inoffp, &inoff, sizeof(off_t));
4969 		if (error != 0)
4970 			return (error);
4971 		inoffp = &inoff;
4972 	}
4973 	if (uap->outoffp != NULL) {
4974 		error = copyin(uap->outoffp, &outoff, sizeof(off_t));
4975 		if (error != 0)
4976 			return (error);
4977 		outoffp = &outoff;
4978 	}
4979 	error = kern_copy_file_range(td, uap->infd, inoffp, uap->outfd,
4980 	    outoffp, uap->len, uap->flags);
4981 	if (error == 0 && uap->inoffp != NULL)
4982 		error = copyout(inoffp, uap->inoffp, sizeof(off_t));
4983 	if (error == 0 && uap->outoffp != NULL)
4984 		error = copyout(outoffp, uap->outoffp, sizeof(off_t));
4985 	return (error);
4986 }
4987