xref: /freebsd/sys/kern/vfs_syscalls.c (revision 2e0d67c3ed7ec760b2d54f8258304b2e6359418b)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (c) 1989, 1993
5  *	The Regents of the University of California.  All rights reserved.
6  * (c) UNIX System Laboratories, Inc.
7  * All or some portions of this file are derived from material licensed
8  * to the University of California by American Telephone and Telegraph
9  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
10  * the permission of UNIX System Laboratories, Inc.
11  *
12  * Redistribution and use in source and binary forms, with or without
13  * modification, are permitted provided that the following conditions
14  * are met:
15  * 1. Redistributions of source code must retain the above copyright
16  *    notice, this list of conditions and the following disclaimer.
17  * 2. Redistributions in binary form must reproduce the above copyright
18  *    notice, this list of conditions and the following disclaimer in the
19  *    documentation and/or other materials provided with the distribution.
20  * 3. Neither the name of the University nor the names of its contributors
21  *    may be used to endorse or promote products derived from this software
22  *    without specific prior written permission.
23  *
24  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
25  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
28  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
29  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
30  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
31  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
32  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
33  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
34  * SUCH DAMAGE.
35  *
36  *	@(#)vfs_syscalls.c	8.13 (Berkeley) 4/15/94
37  */
38 
39 #include <sys/cdefs.h>
40 __FBSDID("$FreeBSD$");
41 
42 #include "opt_capsicum.h"
43 #include "opt_ktrace.h"
44 
45 #include <sys/param.h>
46 #include <sys/systm.h>
47 #include <sys/bio.h>
48 #include <sys/buf.h>
49 #include <sys/capsicum.h>
50 #include <sys/disk.h>
51 #include <sys/sysent.h>
52 #include <sys/malloc.h>
53 #include <sys/mount.h>
54 #include <sys/mutex.h>
55 #include <sys/sysproto.h>
56 #include <sys/namei.h>
57 #include <sys/filedesc.h>
58 #include <sys/kernel.h>
59 #include <sys/fcntl.h>
60 #include <sys/file.h>
61 #include <sys/filio.h>
62 #include <sys/limits.h>
63 #include <sys/linker.h>
64 #include <sys/rwlock.h>
65 #include <sys/sdt.h>
66 #include <sys/stat.h>
67 #include <sys/sx.h>
68 #include <sys/unistd.h>
69 #include <sys/vnode.h>
70 #include <sys/priv.h>
71 #include <sys/proc.h>
72 #include <sys/dirent.h>
73 #include <sys/jail.h>
74 #include <sys/syscallsubr.h>
75 #include <sys/sysctl.h>
76 #ifdef KTRACE
77 #include <sys/ktrace.h>
78 #endif
79 
80 #include <machine/stdarg.h>
81 
82 #include <security/audit/audit.h>
83 #include <security/mac/mac_framework.h>
84 
85 #include <vm/vm.h>
86 #include <vm/vm_object.h>
87 #include <vm/vm_page.h>
88 #include <vm/uma.h>
89 
90 #include <fs/devfs/devfs.h>
91 
92 MALLOC_DEFINE(M_FADVISE, "fadvise", "posix_fadvise(2) information");
93 
94 static int kern_chflagsat(struct thread *td, int fd, const char *path,
95     enum uio_seg pathseg, u_long flags, int atflag);
96 static int setfflags(struct thread *td, struct vnode *, u_long);
97 static int getutimes(const struct timeval *, enum uio_seg, struct timespec *);
98 static int getutimens(const struct timespec *, enum uio_seg,
99     struct timespec *, int *);
100 static int setutimes(struct thread *td, struct vnode *,
101     const struct timespec *, int, int);
102 static int vn_access(struct vnode *vp, int user_flags, struct ucred *cred,
103     struct thread *td);
104 static int kern_fhlinkat(struct thread *td, int fd, const char *path,
105     enum uio_seg pathseg, fhandle_t *fhp);
106 static int kern_readlink_vp(struct vnode *vp, char *buf, enum uio_seg bufseg,
107     size_t count, struct thread *td);
108 static int kern_linkat_vp(struct thread *td, struct vnode *vp, int fd,
109     const char *path, enum uio_seg segflag);
110 
111 static uint64_t
112 at2cnpflags(u_int at_flags, u_int mask)
113 {
114 	u_int64_t res;
115 
116 	MPASS((at_flags & (AT_SYMLINK_FOLLOW | AT_SYMLINK_NOFOLLOW)) !=
117 	    (AT_SYMLINK_FOLLOW | AT_SYMLINK_NOFOLLOW));
118 
119 	res = 0;
120 	at_flags &= mask;
121 	if ((at_flags & AT_RESOLVE_BENEATH) != 0)
122 		res |= RBENEATH;
123 	if ((at_flags & AT_SYMLINK_FOLLOW) != 0)
124 		res |= FOLLOW;
125 	/* NOFOLLOW is pseudo flag */
126 	if ((mask & AT_SYMLINK_NOFOLLOW) != 0) {
127 		res |= (at_flags & AT_SYMLINK_NOFOLLOW) != 0 ? NOFOLLOW :
128 		    FOLLOW;
129 	}
130 	if ((mask & AT_EMPTY_PATH) != 0 && (at_flags & AT_EMPTY_PATH) != 0)
131 		res |= EMPTYPATH;
132 	return (res);
133 }
134 
135 int
136 kern_sync(struct thread *td)
137 {
138 	struct mount *mp, *nmp;
139 	int save;
140 
141 	mtx_lock(&mountlist_mtx);
142 	for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
143 		if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK)) {
144 			nmp = TAILQ_NEXT(mp, mnt_list);
145 			continue;
146 		}
147 		if ((mp->mnt_flag & MNT_RDONLY) == 0 &&
148 		    vn_start_write(NULL, &mp, V_NOWAIT) == 0) {
149 			save = curthread_pflags_set(TDP_SYNCIO);
150 			vfs_periodic(mp, MNT_NOWAIT);
151 			VFS_SYNC(mp, MNT_NOWAIT);
152 			curthread_pflags_restore(save);
153 			vn_finished_write(mp);
154 		}
155 		mtx_lock(&mountlist_mtx);
156 		nmp = TAILQ_NEXT(mp, mnt_list);
157 		vfs_unbusy(mp);
158 	}
159 	mtx_unlock(&mountlist_mtx);
160 	return (0);
161 }
162 
163 /*
164  * Sync each mounted filesystem.
165  */
166 #ifndef _SYS_SYSPROTO_H_
167 struct sync_args {
168 	int     dummy;
169 };
170 #endif
171 /* ARGSUSED */
172 int
173 sys_sync(struct thread *td, struct sync_args *uap)
174 {
175 
176 	return (kern_sync(td));
177 }
178 
179 /*
180  * Change filesystem quotas.
181  */
182 #ifndef _SYS_SYSPROTO_H_
183 struct quotactl_args {
184 	char *path;
185 	int cmd;
186 	int uid;
187 	caddr_t arg;
188 };
189 #endif
190 int
191 sys_quotactl(struct thread *td, struct quotactl_args *uap)
192 {
193 	struct mount *mp;
194 	struct nameidata nd;
195 	int error;
196 	bool mp_busy;
197 
198 	AUDIT_ARG_CMD(uap->cmd);
199 	AUDIT_ARG_UID(uap->uid);
200 	if (!prison_allow(td->td_ucred, PR_ALLOW_QUOTAS))
201 		return (EPERM);
202 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
203 	    uap->path, td);
204 	if ((error = namei(&nd)) != 0)
205 		return (error);
206 	NDFREE(&nd, NDF_ONLY_PNBUF);
207 	mp = nd.ni_vp->v_mount;
208 	vfs_ref(mp);
209 	vput(nd.ni_vp);
210 	error = vfs_busy(mp, 0);
211 	if (error != 0) {
212 		vfs_rel(mp);
213 		return (error);
214 	}
215 	mp_busy = true;
216 	error = VFS_QUOTACTL(mp, uap->cmd, uap->uid, uap->arg, &mp_busy);
217 
218 	/*
219 	 * Since quota on/off operations typically need to open quota
220 	 * files, the implementation may need to unbusy the mount point
221 	 * before calling into namei.  Otherwise, unmount might be
222 	 * started between two vfs_busy() invocations (first is ours,
223 	 * second is from mount point cross-walk code in lookup()),
224 	 * causing deadlock.
225 	 *
226 	 * Avoid unbusying mp if the implementation indicates it has
227 	 * already done so.
228 	 */
229 	if (mp_busy)
230 		vfs_unbusy(mp);
231 	vfs_rel(mp);
232 	return (error);
233 }
234 
235 /*
236  * Used by statfs conversion routines to scale the block size up if
237  * necessary so that all of the block counts are <= 'max_size'.  Note
238  * that 'max_size' should be a bitmask, i.e. 2^n - 1 for some non-zero
239  * value of 'n'.
240  */
241 void
242 statfs_scale_blocks(struct statfs *sf, long max_size)
243 {
244 	uint64_t count;
245 	int shift;
246 
247 	KASSERT(powerof2(max_size + 1), ("%s: invalid max_size", __func__));
248 
249 	/*
250 	 * Attempt to scale the block counts to give a more accurate
251 	 * overview to userland of the ratio of free space to used
252 	 * space.  To do this, find the largest block count and compute
253 	 * a divisor that lets it fit into a signed integer <= max_size.
254 	 */
255 	if (sf->f_bavail < 0)
256 		count = -sf->f_bavail;
257 	else
258 		count = sf->f_bavail;
259 	count = MAX(sf->f_blocks, MAX(sf->f_bfree, count));
260 	if (count <= max_size)
261 		return;
262 
263 	count >>= flsl(max_size);
264 	shift = 0;
265 	while (count > 0) {
266 		shift++;
267 		count >>=1;
268 	}
269 
270 	sf->f_bsize <<= shift;
271 	sf->f_blocks >>= shift;
272 	sf->f_bfree >>= shift;
273 	sf->f_bavail >>= shift;
274 }
275 
276 static int
277 kern_do_statfs(struct thread *td, struct mount *mp, struct statfs *buf)
278 {
279 	int error;
280 
281 	if (mp == NULL)
282 		return (EBADF);
283 	error = vfs_busy(mp, 0);
284 	vfs_rel(mp);
285 	if (error != 0)
286 		return (error);
287 #ifdef MAC
288 	error = mac_mount_check_stat(td->td_ucred, mp);
289 	if (error != 0)
290 		goto out;
291 #endif
292 	error = VFS_STATFS(mp, buf);
293 	if (error != 0)
294 		goto out;
295 	if (priv_check_cred_vfs_generation(td->td_ucred)) {
296 		buf->f_fsid.val[0] = buf->f_fsid.val[1] = 0;
297 		prison_enforce_statfs(td->td_ucred, mp, buf);
298 	}
299 out:
300 	vfs_unbusy(mp);
301 	return (error);
302 }
303 
304 /*
305  * Get filesystem statistics.
306  */
307 #ifndef _SYS_SYSPROTO_H_
308 struct statfs_args {
309 	char *path;
310 	struct statfs *buf;
311 };
312 #endif
313 int
314 sys_statfs(struct thread *td, struct statfs_args *uap)
315 {
316 	struct statfs *sfp;
317 	int error;
318 
319 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
320 	error = kern_statfs(td, uap->path, UIO_USERSPACE, sfp);
321 	if (error == 0)
322 		error = copyout(sfp, uap->buf, sizeof(struct statfs));
323 	free(sfp, M_STATFS);
324 	return (error);
325 }
326 
327 int
328 kern_statfs(struct thread *td, const char *path, enum uio_seg pathseg,
329     struct statfs *buf)
330 {
331 	struct mount *mp;
332 	struct nameidata nd;
333 	int error;
334 
335 	NDINIT(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, td);
336 	error = namei(&nd);
337 	if (error != 0)
338 		return (error);
339 	mp = vfs_ref_from_vp(nd.ni_vp);
340 	NDFREE_NOTHING(&nd);
341 	vrele(nd.ni_vp);
342 	return (kern_do_statfs(td, mp, buf));
343 }
344 
345 /*
346  * Get filesystem statistics.
347  */
348 #ifndef _SYS_SYSPROTO_H_
349 struct fstatfs_args {
350 	int fd;
351 	struct statfs *buf;
352 };
353 #endif
354 int
355 sys_fstatfs(struct thread *td, struct fstatfs_args *uap)
356 {
357 	struct statfs *sfp;
358 	int error;
359 
360 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
361 	error = kern_fstatfs(td, uap->fd, sfp);
362 	if (error == 0)
363 		error = copyout(sfp, uap->buf, sizeof(struct statfs));
364 	free(sfp, M_STATFS);
365 	return (error);
366 }
367 
368 int
369 kern_fstatfs(struct thread *td, int fd, struct statfs *buf)
370 {
371 	struct file *fp;
372 	struct mount *mp;
373 	struct vnode *vp;
374 	int error;
375 
376 	AUDIT_ARG_FD(fd);
377 	error = getvnode_path(td, fd, &cap_fstatfs_rights, &fp);
378 	if (error != 0)
379 		return (error);
380 	vp = fp->f_vnode;
381 #ifdef AUDIT
382 	if (AUDITING_TD(td)) {
383 		vn_lock(vp, LK_SHARED | LK_RETRY);
384 		AUDIT_ARG_VNODE1(vp);
385 		VOP_UNLOCK(vp);
386 	}
387 #endif
388 	mp = vfs_ref_from_vp(vp);
389 	fdrop(fp, td);
390 	return (kern_do_statfs(td, mp, buf));
391 }
392 
393 /*
394  * Get statistics on all filesystems.
395  */
396 #ifndef _SYS_SYSPROTO_H_
397 struct getfsstat_args {
398 	struct statfs *buf;
399 	long bufsize;
400 	int mode;
401 };
402 #endif
403 int
404 sys_getfsstat(struct thread *td, struct getfsstat_args *uap)
405 {
406 	size_t count;
407 	int error;
408 
409 	if (uap->bufsize < 0 || uap->bufsize > SIZE_MAX)
410 		return (EINVAL);
411 	error = kern_getfsstat(td, &uap->buf, uap->bufsize, &count,
412 	    UIO_USERSPACE, uap->mode);
413 	if (error == 0)
414 		td->td_retval[0] = count;
415 	return (error);
416 }
417 
418 /*
419  * If (bufsize > 0 && bufseg == UIO_SYSSPACE)
420  *	The caller is responsible for freeing memory which will be allocated
421  *	in '*buf'.
422  */
423 int
424 kern_getfsstat(struct thread *td, struct statfs **buf, size_t bufsize,
425     size_t *countp, enum uio_seg bufseg, int mode)
426 {
427 	struct mount *mp, *nmp;
428 	struct statfs *sfsp, *sp, *sptmp, *tofree;
429 	size_t count, maxcount;
430 	int error;
431 
432 	switch (mode) {
433 	case MNT_WAIT:
434 	case MNT_NOWAIT:
435 		break;
436 	default:
437 		if (bufseg == UIO_SYSSPACE)
438 			*buf = NULL;
439 		return (EINVAL);
440 	}
441 restart:
442 	maxcount = bufsize / sizeof(struct statfs);
443 	if (bufsize == 0) {
444 		sfsp = NULL;
445 		tofree = NULL;
446 	} else if (bufseg == UIO_USERSPACE) {
447 		sfsp = *buf;
448 		tofree = NULL;
449 	} else /* if (bufseg == UIO_SYSSPACE) */ {
450 		count = 0;
451 		mtx_lock(&mountlist_mtx);
452 		TAILQ_FOREACH(mp, &mountlist, mnt_list) {
453 			count++;
454 		}
455 		mtx_unlock(&mountlist_mtx);
456 		if (maxcount > count)
457 			maxcount = count;
458 		tofree = sfsp = *buf = malloc(maxcount * sizeof(struct statfs),
459 		    M_STATFS, M_WAITOK);
460 	}
461 
462 	count = 0;
463 
464 	/*
465 	 * If there is no target buffer they only want the count.
466 	 *
467 	 * This could be TAILQ_FOREACH but it is open-coded to match the original
468 	 * code below.
469 	 */
470 	if (sfsp == NULL) {
471 		mtx_lock(&mountlist_mtx);
472 		for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
473 			if (prison_canseemount(td->td_ucred, mp) != 0) {
474 				nmp = TAILQ_NEXT(mp, mnt_list);
475 				continue;
476 			}
477 #ifdef MAC
478 			if (mac_mount_check_stat(td->td_ucred, mp) != 0) {
479 				nmp = TAILQ_NEXT(mp, mnt_list);
480 				continue;
481 			}
482 #endif
483 			count++;
484 			nmp = TAILQ_NEXT(mp, mnt_list);
485 		}
486 		mtx_unlock(&mountlist_mtx);
487 		*countp = count;
488 		return (0);
489 	}
490 
491 	/*
492 	 * They want the entire thing.
493 	 *
494 	 * Short-circuit the corner case of no room for anything, avoids
495 	 * relocking below.
496 	 */
497 	if (maxcount < 1) {
498 		goto out;
499 	}
500 
501 	mtx_lock(&mountlist_mtx);
502 	for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
503 		if (prison_canseemount(td->td_ucred, mp) != 0) {
504 			nmp = TAILQ_NEXT(mp, mnt_list);
505 			continue;
506 		}
507 #ifdef MAC
508 		if (mac_mount_check_stat(td->td_ucred, mp) != 0) {
509 			nmp = TAILQ_NEXT(mp, mnt_list);
510 			continue;
511 		}
512 #endif
513 		if (mode == MNT_WAIT) {
514 			if (vfs_busy(mp, MBF_MNTLSTLOCK) != 0) {
515 				/*
516 				 * If vfs_busy() failed, and MBF_NOWAIT
517 				 * wasn't passed, then the mp is gone.
518 				 * Furthermore, because of MBF_MNTLSTLOCK,
519 				 * the mountlist_mtx was dropped.  We have
520 				 * no other choice than to start over.
521 				 */
522 				mtx_unlock(&mountlist_mtx);
523 				free(tofree, M_STATFS);
524 				goto restart;
525 			}
526 		} else {
527 			if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK) != 0) {
528 				nmp = TAILQ_NEXT(mp, mnt_list);
529 				continue;
530 			}
531 		}
532 		sp = &mp->mnt_stat;
533 		/*
534 		 * If MNT_NOWAIT is specified, do not refresh
535 		 * the fsstat cache.
536 		 */
537 		if (mode != MNT_NOWAIT) {
538 			error = VFS_STATFS(mp, sp);
539 			if (error != 0) {
540 				mtx_lock(&mountlist_mtx);
541 				nmp = TAILQ_NEXT(mp, mnt_list);
542 				vfs_unbusy(mp);
543 				continue;
544 			}
545 		}
546 		if (priv_check_cred_vfs_generation(td->td_ucred)) {
547 			sptmp = malloc(sizeof(struct statfs), M_STATFS,
548 			    M_WAITOK);
549 			*sptmp = *sp;
550 			sptmp->f_fsid.val[0] = sptmp->f_fsid.val[1] = 0;
551 			prison_enforce_statfs(td->td_ucred, mp, sptmp);
552 			sp = sptmp;
553 		} else
554 			sptmp = NULL;
555 		if (bufseg == UIO_SYSSPACE) {
556 			bcopy(sp, sfsp, sizeof(*sp));
557 			free(sptmp, M_STATFS);
558 		} else /* if (bufseg == UIO_USERSPACE) */ {
559 			error = copyout(sp, sfsp, sizeof(*sp));
560 			free(sptmp, M_STATFS);
561 			if (error != 0) {
562 				vfs_unbusy(mp);
563 				return (error);
564 			}
565 		}
566 		sfsp++;
567 		count++;
568 
569 		if (count == maxcount) {
570 			vfs_unbusy(mp);
571 			goto out;
572 		}
573 
574 		mtx_lock(&mountlist_mtx);
575 		nmp = TAILQ_NEXT(mp, mnt_list);
576 		vfs_unbusy(mp);
577 	}
578 	mtx_unlock(&mountlist_mtx);
579 out:
580 	*countp = count;
581 	return (0);
582 }
583 
584 #ifdef COMPAT_FREEBSD4
585 /*
586  * Get old format filesystem statistics.
587  */
588 static void freebsd4_cvtstatfs(struct statfs *, struct ostatfs *);
589 
590 #ifndef _SYS_SYSPROTO_H_
591 struct freebsd4_statfs_args {
592 	char *path;
593 	struct ostatfs *buf;
594 };
595 #endif
596 int
597 freebsd4_statfs(struct thread *td, struct freebsd4_statfs_args *uap)
598 {
599 	struct ostatfs osb;
600 	struct statfs *sfp;
601 	int error;
602 
603 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
604 	error = kern_statfs(td, uap->path, UIO_USERSPACE, sfp);
605 	if (error == 0) {
606 		freebsd4_cvtstatfs(sfp, &osb);
607 		error = copyout(&osb, uap->buf, sizeof(osb));
608 	}
609 	free(sfp, M_STATFS);
610 	return (error);
611 }
612 
613 /*
614  * Get filesystem statistics.
615  */
616 #ifndef _SYS_SYSPROTO_H_
617 struct freebsd4_fstatfs_args {
618 	int fd;
619 	struct ostatfs *buf;
620 };
621 #endif
622 int
623 freebsd4_fstatfs(struct thread *td, struct freebsd4_fstatfs_args *uap)
624 {
625 	struct ostatfs osb;
626 	struct statfs *sfp;
627 	int error;
628 
629 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
630 	error = kern_fstatfs(td, uap->fd, sfp);
631 	if (error == 0) {
632 		freebsd4_cvtstatfs(sfp, &osb);
633 		error = copyout(&osb, uap->buf, sizeof(osb));
634 	}
635 	free(sfp, M_STATFS);
636 	return (error);
637 }
638 
639 /*
640  * Get statistics on all filesystems.
641  */
642 #ifndef _SYS_SYSPROTO_H_
643 struct freebsd4_getfsstat_args {
644 	struct ostatfs *buf;
645 	long bufsize;
646 	int mode;
647 };
648 #endif
649 int
650 freebsd4_getfsstat(struct thread *td, struct freebsd4_getfsstat_args *uap)
651 {
652 	struct statfs *buf, *sp;
653 	struct ostatfs osb;
654 	size_t count, size;
655 	int error;
656 
657 	if (uap->bufsize < 0)
658 		return (EINVAL);
659 	count = uap->bufsize / sizeof(struct ostatfs);
660 	if (count > SIZE_MAX / sizeof(struct statfs))
661 		return (EINVAL);
662 	size = count * sizeof(struct statfs);
663 	error = kern_getfsstat(td, &buf, size, &count, UIO_SYSSPACE,
664 	    uap->mode);
665 	if (error == 0)
666 		td->td_retval[0] = count;
667 	if (size != 0) {
668 		sp = buf;
669 		while (count != 0 && error == 0) {
670 			freebsd4_cvtstatfs(sp, &osb);
671 			error = copyout(&osb, uap->buf, sizeof(osb));
672 			sp++;
673 			uap->buf++;
674 			count--;
675 		}
676 		free(buf, M_STATFS);
677 	}
678 	return (error);
679 }
680 
681 /*
682  * Implement fstatfs() for (NFS) file handles.
683  */
684 #ifndef _SYS_SYSPROTO_H_
685 struct freebsd4_fhstatfs_args {
686 	struct fhandle *u_fhp;
687 	struct ostatfs *buf;
688 };
689 #endif
690 int
691 freebsd4_fhstatfs(struct thread *td, struct freebsd4_fhstatfs_args *uap)
692 {
693 	struct ostatfs osb;
694 	struct statfs *sfp;
695 	fhandle_t fh;
696 	int error;
697 
698 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
699 	if (error != 0)
700 		return (error);
701 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
702 	error = kern_fhstatfs(td, fh, sfp);
703 	if (error == 0) {
704 		freebsd4_cvtstatfs(sfp, &osb);
705 		error = copyout(&osb, uap->buf, sizeof(osb));
706 	}
707 	free(sfp, M_STATFS);
708 	return (error);
709 }
710 
711 /*
712  * Convert a new format statfs structure to an old format statfs structure.
713  */
714 static void
715 freebsd4_cvtstatfs(struct statfs *nsp, struct ostatfs *osp)
716 {
717 
718 	statfs_scale_blocks(nsp, LONG_MAX);
719 	bzero(osp, sizeof(*osp));
720 	osp->f_bsize = nsp->f_bsize;
721 	osp->f_iosize = MIN(nsp->f_iosize, LONG_MAX);
722 	osp->f_blocks = nsp->f_blocks;
723 	osp->f_bfree = nsp->f_bfree;
724 	osp->f_bavail = nsp->f_bavail;
725 	osp->f_files = MIN(nsp->f_files, LONG_MAX);
726 	osp->f_ffree = MIN(nsp->f_ffree, LONG_MAX);
727 	osp->f_owner = nsp->f_owner;
728 	osp->f_type = nsp->f_type;
729 	osp->f_flags = nsp->f_flags;
730 	osp->f_syncwrites = MIN(nsp->f_syncwrites, LONG_MAX);
731 	osp->f_asyncwrites = MIN(nsp->f_asyncwrites, LONG_MAX);
732 	osp->f_syncreads = MIN(nsp->f_syncreads, LONG_MAX);
733 	osp->f_asyncreads = MIN(nsp->f_asyncreads, LONG_MAX);
734 	strlcpy(osp->f_fstypename, nsp->f_fstypename,
735 	    MIN(MFSNAMELEN, OMFSNAMELEN));
736 	strlcpy(osp->f_mntonname, nsp->f_mntonname,
737 	    MIN(MNAMELEN, OMNAMELEN));
738 	strlcpy(osp->f_mntfromname, nsp->f_mntfromname,
739 	    MIN(MNAMELEN, OMNAMELEN));
740 	osp->f_fsid = nsp->f_fsid;
741 }
742 #endif /* COMPAT_FREEBSD4 */
743 
744 #if defined(COMPAT_FREEBSD11)
745 /*
746  * Get old format filesystem statistics.
747  */
748 static void freebsd11_cvtstatfs(struct statfs *, struct freebsd11_statfs *);
749 
750 int
751 freebsd11_statfs(struct thread *td, struct freebsd11_statfs_args *uap)
752 {
753 	struct freebsd11_statfs osb;
754 	struct statfs *sfp;
755 	int error;
756 
757 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
758 	error = kern_statfs(td, uap->path, UIO_USERSPACE, sfp);
759 	if (error == 0) {
760 		freebsd11_cvtstatfs(sfp, &osb);
761 		error = copyout(&osb, uap->buf, sizeof(osb));
762 	}
763 	free(sfp, M_STATFS);
764 	return (error);
765 }
766 
767 /*
768  * Get filesystem statistics.
769  */
770 int
771 freebsd11_fstatfs(struct thread *td, struct freebsd11_fstatfs_args *uap)
772 {
773 	struct freebsd11_statfs osb;
774 	struct statfs *sfp;
775 	int error;
776 
777 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
778 	error = kern_fstatfs(td, uap->fd, sfp);
779 	if (error == 0) {
780 		freebsd11_cvtstatfs(sfp, &osb);
781 		error = copyout(&osb, uap->buf, sizeof(osb));
782 	}
783 	free(sfp, M_STATFS);
784 	return (error);
785 }
786 
787 /*
788  * Get statistics on all filesystems.
789  */
790 int
791 freebsd11_getfsstat(struct thread *td, struct freebsd11_getfsstat_args *uap)
792 {
793 	struct freebsd11_statfs osb;
794 	struct statfs *buf, *sp;
795 	size_t count, size;
796 	int error;
797 
798 	count = uap->bufsize / sizeof(struct ostatfs);
799 	size = count * sizeof(struct statfs);
800 	error = kern_getfsstat(td, &buf, size, &count, UIO_SYSSPACE,
801 	    uap->mode);
802 	if (error == 0)
803 		td->td_retval[0] = count;
804 	if (size > 0) {
805 		sp = buf;
806 		while (count > 0 && error == 0) {
807 			freebsd11_cvtstatfs(sp, &osb);
808 			error = copyout(&osb, uap->buf, sizeof(osb));
809 			sp++;
810 			uap->buf++;
811 			count--;
812 		}
813 		free(buf, M_STATFS);
814 	}
815 	return (error);
816 }
817 
818 /*
819  * Implement fstatfs() for (NFS) file handles.
820  */
821 int
822 freebsd11_fhstatfs(struct thread *td, struct freebsd11_fhstatfs_args *uap)
823 {
824 	struct freebsd11_statfs osb;
825 	struct statfs *sfp;
826 	fhandle_t fh;
827 	int error;
828 
829 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
830 	if (error)
831 		return (error);
832 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
833 	error = kern_fhstatfs(td, fh, sfp);
834 	if (error == 0) {
835 		freebsd11_cvtstatfs(sfp, &osb);
836 		error = copyout(&osb, uap->buf, sizeof(osb));
837 	}
838 	free(sfp, M_STATFS);
839 	return (error);
840 }
841 
842 /*
843  * Convert a new format statfs structure to an old format statfs structure.
844  */
845 static void
846 freebsd11_cvtstatfs(struct statfs *nsp, struct freebsd11_statfs *osp)
847 {
848 
849 	bzero(osp, sizeof(*osp));
850 	osp->f_version = FREEBSD11_STATFS_VERSION;
851 	osp->f_type = nsp->f_type;
852 	osp->f_flags = nsp->f_flags;
853 	osp->f_bsize = nsp->f_bsize;
854 	osp->f_iosize = nsp->f_iosize;
855 	osp->f_blocks = nsp->f_blocks;
856 	osp->f_bfree = nsp->f_bfree;
857 	osp->f_bavail = nsp->f_bavail;
858 	osp->f_files = nsp->f_files;
859 	osp->f_ffree = nsp->f_ffree;
860 	osp->f_syncwrites = nsp->f_syncwrites;
861 	osp->f_asyncwrites = nsp->f_asyncwrites;
862 	osp->f_syncreads = nsp->f_syncreads;
863 	osp->f_asyncreads = nsp->f_asyncreads;
864 	osp->f_namemax = nsp->f_namemax;
865 	osp->f_owner = nsp->f_owner;
866 	osp->f_fsid = nsp->f_fsid;
867 	strlcpy(osp->f_fstypename, nsp->f_fstypename,
868 	    MIN(MFSNAMELEN, sizeof(osp->f_fstypename)));
869 	strlcpy(osp->f_mntonname, nsp->f_mntonname,
870 	    MIN(MNAMELEN, sizeof(osp->f_mntonname)));
871 	strlcpy(osp->f_mntfromname, nsp->f_mntfromname,
872 	    MIN(MNAMELEN, sizeof(osp->f_mntfromname)));
873 }
874 #endif /* COMPAT_FREEBSD11 */
875 
876 /*
877  * Change current working directory to a given file descriptor.
878  */
879 #ifndef _SYS_SYSPROTO_H_
880 struct fchdir_args {
881 	int	fd;
882 };
883 #endif
884 int
885 sys_fchdir(struct thread *td, struct fchdir_args *uap)
886 {
887 	struct vnode *vp, *tdp;
888 	struct mount *mp;
889 	struct file *fp;
890 	int error;
891 
892 	AUDIT_ARG_FD(uap->fd);
893 	error = getvnode_path(td, uap->fd, &cap_fchdir_rights,
894 	    &fp);
895 	if (error != 0)
896 		return (error);
897 	vp = fp->f_vnode;
898 	vref(vp);
899 	fdrop(fp, td);
900 	vn_lock(vp, LK_SHARED | LK_RETRY);
901 	AUDIT_ARG_VNODE1(vp);
902 	error = change_dir(vp, td);
903 	while (!error && (mp = vp->v_mountedhere) != NULL) {
904 		if (vfs_busy(mp, 0))
905 			continue;
906 		error = VFS_ROOT(mp, LK_SHARED, &tdp);
907 		vfs_unbusy(mp);
908 		if (error != 0)
909 			break;
910 		vput(vp);
911 		vp = tdp;
912 	}
913 	if (error != 0) {
914 		vput(vp);
915 		return (error);
916 	}
917 	VOP_UNLOCK(vp);
918 	pwd_chdir(td, vp);
919 	return (0);
920 }
921 
922 /*
923  * Change current working directory (``.'').
924  */
925 #ifndef _SYS_SYSPROTO_H_
926 struct chdir_args {
927 	char	*path;
928 };
929 #endif
930 int
931 sys_chdir(struct thread *td, struct chdir_args *uap)
932 {
933 
934 	return (kern_chdir(td, uap->path, UIO_USERSPACE));
935 }
936 
937 int
938 kern_chdir(struct thread *td, const char *path, enum uio_seg pathseg)
939 {
940 	struct nameidata nd;
941 	int error;
942 
943 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
944 	    pathseg, path, td);
945 	if ((error = namei(&nd)) != 0)
946 		return (error);
947 	if ((error = change_dir(nd.ni_vp, td)) != 0) {
948 		vput(nd.ni_vp);
949 		NDFREE_NOTHING(&nd);
950 		return (error);
951 	}
952 	VOP_UNLOCK(nd.ni_vp);
953 	NDFREE_NOTHING(&nd);
954 	pwd_chdir(td, nd.ni_vp);
955 	return (0);
956 }
957 
958 static int unprivileged_chroot = 0;
959 SYSCTL_INT(_security_bsd, OID_AUTO, unprivileged_chroot, CTLFLAG_RW,
960     &unprivileged_chroot, 0,
961     "Unprivileged processes can use chroot(2)");
962 /*
963  * Change notion of root (``/'') directory.
964  */
965 #ifndef _SYS_SYSPROTO_H_
966 struct chroot_args {
967 	char	*path;
968 };
969 #endif
970 int
971 sys_chroot(struct thread *td, struct chroot_args *uap)
972 {
973 	struct nameidata nd;
974 	struct proc *p;
975 	int error;
976 
977 	error = priv_check(td, PRIV_VFS_CHROOT);
978 	if (error != 0) {
979 		p = td->td_proc;
980 		PROC_LOCK(p);
981 		if (unprivileged_chroot == 0 ||
982 		    (p->p_flag2 & P2_NO_NEW_PRIVS) == 0) {
983 			PROC_UNLOCK(p);
984 			return (error);
985 		}
986 		PROC_UNLOCK(p);
987 	}
988 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
989 	    UIO_USERSPACE, uap->path, td);
990 	error = namei(&nd);
991 	if (error != 0)
992 		goto error;
993 	error = change_dir(nd.ni_vp, td);
994 	if (error != 0)
995 		goto e_vunlock;
996 #ifdef MAC
997 	error = mac_vnode_check_chroot(td->td_ucred, nd.ni_vp);
998 	if (error != 0)
999 		goto e_vunlock;
1000 #endif
1001 	VOP_UNLOCK(nd.ni_vp);
1002 	error = pwd_chroot(td, nd.ni_vp);
1003 	vrele(nd.ni_vp);
1004 	NDFREE_NOTHING(&nd);
1005 	return (error);
1006 e_vunlock:
1007 	vput(nd.ni_vp);
1008 error:
1009 	NDFREE_NOTHING(&nd);
1010 	return (error);
1011 }
1012 
1013 /*
1014  * Common routine for chroot and chdir.  Callers must provide a locked vnode
1015  * instance.
1016  */
1017 int
1018 change_dir(struct vnode *vp, struct thread *td)
1019 {
1020 #ifdef MAC
1021 	int error;
1022 #endif
1023 
1024 	ASSERT_VOP_LOCKED(vp, "change_dir(): vp not locked");
1025 	if (vp->v_type != VDIR)
1026 		return (ENOTDIR);
1027 #ifdef MAC
1028 	error = mac_vnode_check_chdir(td->td_ucred, vp);
1029 	if (error != 0)
1030 		return (error);
1031 #endif
1032 	return (VOP_ACCESS(vp, VEXEC, td->td_ucred, td));
1033 }
1034 
1035 static __inline void
1036 flags_to_rights(int flags, cap_rights_t *rightsp)
1037 {
1038 	if (flags & O_EXEC) {
1039 		cap_rights_set_one(rightsp, CAP_FEXECVE);
1040 		if (flags & O_PATH)
1041 			return;
1042 	} else {
1043 		switch ((flags & O_ACCMODE)) {
1044 		case O_RDONLY:
1045 			cap_rights_set_one(rightsp, CAP_READ);
1046 			break;
1047 		case O_RDWR:
1048 			cap_rights_set_one(rightsp, CAP_READ);
1049 			/* FALLTHROUGH */
1050 		case O_WRONLY:
1051 			cap_rights_set_one(rightsp, CAP_WRITE);
1052 			if (!(flags & (O_APPEND | O_TRUNC)))
1053 				cap_rights_set_one(rightsp, CAP_SEEK);
1054 			break;
1055 		}
1056 	}
1057 
1058 	if (flags & O_CREAT)
1059 		cap_rights_set_one(rightsp, CAP_CREATE);
1060 
1061 	if (flags & O_TRUNC)
1062 		cap_rights_set_one(rightsp, CAP_FTRUNCATE);
1063 
1064 	if (flags & (O_SYNC | O_FSYNC))
1065 		cap_rights_set_one(rightsp, CAP_FSYNC);
1066 
1067 	if (flags & (O_EXLOCK | O_SHLOCK))
1068 		cap_rights_set_one(rightsp, CAP_FLOCK);
1069 }
1070 
1071 /*
1072  * Check permissions, allocate an open file structure, and call the device
1073  * open routine if any.
1074  */
1075 #ifndef _SYS_SYSPROTO_H_
1076 struct open_args {
1077 	char	*path;
1078 	int	flags;
1079 	int	mode;
1080 };
1081 #endif
1082 int
1083 sys_open(struct thread *td, struct open_args *uap)
1084 {
1085 
1086 	return (kern_openat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1087 	    uap->flags, uap->mode));
1088 }
1089 
1090 #ifndef _SYS_SYSPROTO_H_
1091 struct openat_args {
1092 	int	fd;
1093 	char	*path;
1094 	int	flag;
1095 	int	mode;
1096 };
1097 #endif
1098 int
1099 sys_openat(struct thread *td, struct openat_args *uap)
1100 {
1101 
1102 	AUDIT_ARG_FD(uap->fd);
1103 	return (kern_openat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
1104 	    uap->mode));
1105 }
1106 
1107 int
1108 kern_openat(struct thread *td, int fd, const char *path, enum uio_seg pathseg,
1109     int flags, int mode)
1110 {
1111 	struct proc *p = td->td_proc;
1112 	struct filedesc *fdp;
1113 	struct pwddesc *pdp;
1114 	struct file *fp;
1115 	struct vnode *vp;
1116 	struct nameidata nd;
1117 	cap_rights_t rights;
1118 	int cmode, error, indx;
1119 
1120 	indx = -1;
1121 	fdp = p->p_fd;
1122 	pdp = p->p_pd;
1123 
1124 	AUDIT_ARG_FFLAGS(flags);
1125 	AUDIT_ARG_MODE(mode);
1126 	cap_rights_init_one(&rights, CAP_LOOKUP);
1127 	flags_to_rights(flags, &rights);
1128 
1129 	/*
1130 	 * Only one of the O_EXEC, O_RDONLY, O_WRONLY and O_RDWR flags
1131 	 * may be specified.  On the other hand, for O_PATH any mode
1132 	 * except O_EXEC is ignored.
1133 	 */
1134 	if ((flags & O_PATH) != 0) {
1135 		flags &= ~(O_CREAT | O_ACCMODE);
1136 	} else if ((flags & O_EXEC) != 0) {
1137 		if (flags & O_ACCMODE)
1138 			return (EINVAL);
1139 	} else if ((flags & O_ACCMODE) == O_ACCMODE) {
1140 		return (EINVAL);
1141 	} else {
1142 		flags = FFLAGS(flags);
1143 	}
1144 
1145 	/*
1146 	 * Allocate a file structure. The descriptor to reference it
1147 	 * is allocated and used by finstall_refed() below.
1148 	 */
1149 	error = falloc_noinstall(td, &fp);
1150 	if (error != 0)
1151 		return (error);
1152 	/* Set the flags early so the finit in devfs can pick them up. */
1153 	fp->f_flag = flags & FMASK;
1154 	cmode = ((mode & ~pdp->pd_cmask) & ALLPERMS) & ~S_ISTXT;
1155 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, fd,
1156 	    &rights, td);
1157 	td->td_dupfd = -1;		/* XXX check for fdopen */
1158 	error = vn_open(&nd, &flags, cmode, fp);
1159 	if (error != 0) {
1160 		/*
1161 		 * If the vn_open replaced the method vector, something
1162 		 * wonderous happened deep below and we just pass it up
1163 		 * pretending we know what we do.
1164 		 */
1165 		if (error == ENXIO && fp->f_ops != &badfileops) {
1166 			MPASS((flags & O_PATH) == 0);
1167 			goto success;
1168 		}
1169 
1170 		/*
1171 		 * Handle special fdopen() case. bleh.
1172 		 *
1173 		 * Don't do this for relative (capability) lookups; we don't
1174 		 * understand exactly what would happen, and we don't think
1175 		 * that it ever should.
1176 		 */
1177 		if ((nd.ni_resflags & NIRES_STRICTREL) == 0 &&
1178 		    (error == ENODEV || error == ENXIO) &&
1179 		    td->td_dupfd >= 0) {
1180 			error = dupfdopen(td, fdp, td->td_dupfd, flags, error,
1181 			    &indx);
1182 			if (error == 0)
1183 				goto success;
1184 		}
1185 
1186 		goto bad;
1187 	}
1188 	td->td_dupfd = 0;
1189 	NDFREE(&nd, NDF_ONLY_PNBUF);
1190 	vp = nd.ni_vp;
1191 
1192 	/*
1193 	 * Store the vnode, for any f_type. Typically, the vnode use
1194 	 * count is decremented by direct call to vn_closefile() for
1195 	 * files that switched type in the cdevsw fdopen() method.
1196 	 */
1197 	fp->f_vnode = vp;
1198 
1199 	/*
1200 	 * If the file wasn't claimed by devfs bind it to the normal
1201 	 * vnode operations here.
1202 	 */
1203 	if (fp->f_ops == &badfileops) {
1204 		KASSERT(vp->v_type != VFIFO || (flags & O_PATH) != 0,
1205 		    ("Unexpected fifo fp %p vp %p", fp, vp));
1206 		if ((flags & O_PATH) != 0) {
1207 			finit(fp, (flags & FMASK) | (fp->f_flag & FKQALLOWED),
1208 			    DTYPE_VNODE, NULL, &path_fileops);
1209 			vhold(vp);
1210 			vunref(vp);
1211 		} else {
1212 			finit_vnode(fp, flags, NULL, &vnops);
1213 		}
1214 	}
1215 
1216 	VOP_UNLOCK(vp);
1217 	if (flags & O_TRUNC) {
1218 		error = fo_truncate(fp, 0, td->td_ucred, td);
1219 		if (error != 0)
1220 			goto bad;
1221 	}
1222 success:
1223 	/*
1224 	 * If we haven't already installed the FD (for dupfdopen), do so now.
1225 	 */
1226 	if (indx == -1) {
1227 		struct filecaps *fcaps;
1228 
1229 #ifdef CAPABILITIES
1230 		if ((nd.ni_resflags & NIRES_STRICTREL) != 0)
1231 			fcaps = &nd.ni_filecaps;
1232 		else
1233 #endif
1234 			fcaps = NULL;
1235 		error = finstall_refed(td, fp, &indx, flags, fcaps);
1236 		/* On success finstall_refed() consumes fcaps. */
1237 		if (error != 0) {
1238 			filecaps_free(&nd.ni_filecaps);
1239 			goto bad;
1240 		}
1241 	} else {
1242 		filecaps_free(&nd.ni_filecaps);
1243 		falloc_abort(td, fp);
1244 	}
1245 
1246 	td->td_retval[0] = indx;
1247 	return (0);
1248 bad:
1249 	KASSERT(indx == -1, ("indx=%d, should be -1", indx));
1250 	falloc_abort(td, fp);
1251 	return (error);
1252 }
1253 
1254 #ifdef COMPAT_43
1255 /*
1256  * Create a file.
1257  */
1258 #ifndef _SYS_SYSPROTO_H_
1259 struct ocreat_args {
1260 	char	*path;
1261 	int	mode;
1262 };
1263 #endif
1264 int
1265 ocreat(struct thread *td, struct ocreat_args *uap)
1266 {
1267 
1268 	return (kern_openat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1269 	    O_WRONLY | O_CREAT | O_TRUNC, uap->mode));
1270 }
1271 #endif /* COMPAT_43 */
1272 
1273 /*
1274  * Create a special file.
1275  */
1276 #ifndef _SYS_SYSPROTO_H_
1277 struct mknodat_args {
1278 	int	fd;
1279 	char	*path;
1280 	mode_t	mode;
1281 	dev_t	dev;
1282 };
1283 #endif
1284 int
1285 sys_mknodat(struct thread *td, struct mknodat_args *uap)
1286 {
1287 
1288 	return (kern_mknodat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode,
1289 	    uap->dev));
1290 }
1291 
1292 #if defined(COMPAT_FREEBSD11)
1293 int
1294 freebsd11_mknod(struct thread *td,
1295     struct freebsd11_mknod_args *uap)
1296 {
1297 
1298 	return (kern_mknodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1299 	    uap->mode, uap->dev));
1300 }
1301 
1302 int
1303 freebsd11_mknodat(struct thread *td,
1304     struct freebsd11_mknodat_args *uap)
1305 {
1306 
1307 	return (kern_mknodat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode,
1308 	    uap->dev));
1309 }
1310 #endif /* COMPAT_FREEBSD11 */
1311 
1312 int
1313 kern_mknodat(struct thread *td, int fd, const char *path, enum uio_seg pathseg,
1314     int mode, dev_t dev)
1315 {
1316 	struct vnode *vp;
1317 	struct mount *mp;
1318 	struct vattr vattr;
1319 	struct nameidata nd;
1320 	int error, whiteout = 0;
1321 
1322 	AUDIT_ARG_MODE(mode);
1323 	AUDIT_ARG_DEV(dev);
1324 	switch (mode & S_IFMT) {
1325 	case S_IFCHR:
1326 	case S_IFBLK:
1327 		error = priv_check(td, PRIV_VFS_MKNOD_DEV);
1328 		if (error == 0 && dev == VNOVAL)
1329 			error = EINVAL;
1330 		break;
1331 	case S_IFWHT:
1332 		error = priv_check(td, PRIV_VFS_MKNOD_WHT);
1333 		break;
1334 	case S_IFIFO:
1335 		if (dev == 0)
1336 			return (kern_mkfifoat(td, fd, path, pathseg, mode));
1337 		/* FALLTHROUGH */
1338 	default:
1339 		error = EINVAL;
1340 		break;
1341 	}
1342 	if (error != 0)
1343 		return (error);
1344 	NDPREINIT(&nd);
1345 restart:
1346 	bwillwrite();
1347 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
1348 	    NOCACHE, pathseg, path, fd, &cap_mknodat_rights,
1349 	    td);
1350 	if ((error = namei(&nd)) != 0)
1351 		return (error);
1352 	vp = nd.ni_vp;
1353 	if (vp != NULL) {
1354 		NDFREE(&nd, NDF_ONLY_PNBUF);
1355 		if (vp == nd.ni_dvp)
1356 			vrele(nd.ni_dvp);
1357 		else
1358 			vput(nd.ni_dvp);
1359 		vrele(vp);
1360 		return (EEXIST);
1361 	} else {
1362 		VATTR_NULL(&vattr);
1363 		vattr.va_mode = (mode & ALLPERMS) &
1364 		    ~td->td_proc->p_pd->pd_cmask;
1365 		vattr.va_rdev = dev;
1366 		whiteout = 0;
1367 
1368 		switch (mode & S_IFMT) {
1369 		case S_IFCHR:
1370 			vattr.va_type = VCHR;
1371 			break;
1372 		case S_IFBLK:
1373 			vattr.va_type = VBLK;
1374 			break;
1375 		case S_IFWHT:
1376 			whiteout = 1;
1377 			break;
1378 		default:
1379 			panic("kern_mknod: invalid mode");
1380 		}
1381 	}
1382 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1383 		NDFREE(&nd, NDF_ONLY_PNBUF);
1384 		vput(nd.ni_dvp);
1385 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1386 			return (error);
1387 		goto restart;
1388 	}
1389 #ifdef MAC
1390 	if (error == 0 && !whiteout)
1391 		error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp,
1392 		    &nd.ni_cnd, &vattr);
1393 #endif
1394 	if (error == 0) {
1395 		if (whiteout)
1396 			error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE);
1397 		else {
1398 			error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp,
1399 						&nd.ni_cnd, &vattr);
1400 		}
1401 	}
1402 	VOP_VPUT_PAIR(nd.ni_dvp, error == 0 && !whiteout ? &nd.ni_vp : NULL,
1403 	    true);
1404 	vn_finished_write(mp);
1405 	NDFREE(&nd, NDF_ONLY_PNBUF);
1406 	if (error == ERELOOKUP)
1407 		goto restart;
1408 	return (error);
1409 }
1410 
1411 /*
1412  * Create a named pipe.
1413  */
1414 #ifndef _SYS_SYSPROTO_H_
1415 struct mkfifo_args {
1416 	char	*path;
1417 	int	mode;
1418 };
1419 #endif
1420 int
1421 sys_mkfifo(struct thread *td, struct mkfifo_args *uap)
1422 {
1423 
1424 	return (kern_mkfifoat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1425 	    uap->mode));
1426 }
1427 
1428 #ifndef _SYS_SYSPROTO_H_
1429 struct mkfifoat_args {
1430 	int	fd;
1431 	char	*path;
1432 	mode_t	mode;
1433 };
1434 #endif
1435 int
1436 sys_mkfifoat(struct thread *td, struct mkfifoat_args *uap)
1437 {
1438 
1439 	return (kern_mkfifoat(td, uap->fd, uap->path, UIO_USERSPACE,
1440 	    uap->mode));
1441 }
1442 
1443 int
1444 kern_mkfifoat(struct thread *td, int fd, const char *path,
1445     enum uio_seg pathseg, int mode)
1446 {
1447 	struct mount *mp;
1448 	struct vattr vattr;
1449 	struct nameidata nd;
1450 	int error;
1451 
1452 	AUDIT_ARG_MODE(mode);
1453 	NDPREINIT(&nd);
1454 restart:
1455 	bwillwrite();
1456 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
1457 	    NOCACHE, pathseg, path, fd, &cap_mkfifoat_rights,
1458 	    td);
1459 	if ((error = namei(&nd)) != 0)
1460 		return (error);
1461 	if (nd.ni_vp != NULL) {
1462 		NDFREE(&nd, NDF_ONLY_PNBUF);
1463 		if (nd.ni_vp == nd.ni_dvp)
1464 			vrele(nd.ni_dvp);
1465 		else
1466 			vput(nd.ni_dvp);
1467 		vrele(nd.ni_vp);
1468 		return (EEXIST);
1469 	}
1470 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1471 		NDFREE(&nd, NDF_ONLY_PNBUF);
1472 		vput(nd.ni_dvp);
1473 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1474 			return (error);
1475 		goto restart;
1476 	}
1477 	VATTR_NULL(&vattr);
1478 	vattr.va_type = VFIFO;
1479 	vattr.va_mode = (mode & ALLPERMS) & ~td->td_proc->p_pd->pd_cmask;
1480 #ifdef MAC
1481 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1482 	    &vattr);
1483 	if (error != 0)
1484 		goto out;
1485 #endif
1486 	error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
1487 #ifdef MAC
1488 out:
1489 #endif
1490 	VOP_VPUT_PAIR(nd.ni_dvp, error == 0 ? &nd.ni_vp : NULL, true);
1491 	vn_finished_write(mp);
1492 	NDFREE(&nd, NDF_ONLY_PNBUF);
1493 	if (error == ERELOOKUP)
1494 		goto restart;
1495 	return (error);
1496 }
1497 
1498 /*
1499  * Make a hard file link.
1500  */
1501 #ifndef _SYS_SYSPROTO_H_
1502 struct link_args {
1503 	char	*path;
1504 	char	*link;
1505 };
1506 #endif
1507 int
1508 sys_link(struct thread *td, struct link_args *uap)
1509 {
1510 
1511 	return (kern_linkat(td, AT_FDCWD, AT_FDCWD, uap->path, uap->link,
1512 	    UIO_USERSPACE, AT_SYMLINK_FOLLOW));
1513 }
1514 
1515 #ifndef _SYS_SYSPROTO_H_
1516 struct linkat_args {
1517 	int	fd1;
1518 	char	*path1;
1519 	int	fd2;
1520 	char	*path2;
1521 	int	flag;
1522 };
1523 #endif
1524 int
1525 sys_linkat(struct thread *td, struct linkat_args *uap)
1526 {
1527 
1528 	return (kern_linkat(td, uap->fd1, uap->fd2, uap->path1, uap->path2,
1529 	    UIO_USERSPACE, uap->flag));
1530 }
1531 
1532 int hardlink_check_uid = 0;
1533 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_uid, CTLFLAG_RW,
1534     &hardlink_check_uid, 0,
1535     "Unprivileged processes cannot create hard links to files owned by other "
1536     "users");
1537 static int hardlink_check_gid = 0;
1538 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_gid, CTLFLAG_RW,
1539     &hardlink_check_gid, 0,
1540     "Unprivileged processes cannot create hard links to files owned by other "
1541     "groups");
1542 
1543 static int
1544 can_hardlink(struct vnode *vp, struct ucred *cred)
1545 {
1546 	struct vattr va;
1547 	int error;
1548 
1549 	if (!hardlink_check_uid && !hardlink_check_gid)
1550 		return (0);
1551 
1552 	error = VOP_GETATTR(vp, &va, cred);
1553 	if (error != 0)
1554 		return (error);
1555 
1556 	if (hardlink_check_uid && cred->cr_uid != va.va_uid) {
1557 		error = priv_check_cred(cred, PRIV_VFS_LINK);
1558 		if (error != 0)
1559 			return (error);
1560 	}
1561 
1562 	if (hardlink_check_gid && !groupmember(va.va_gid, cred)) {
1563 		error = priv_check_cred(cred, PRIV_VFS_LINK);
1564 		if (error != 0)
1565 			return (error);
1566 	}
1567 
1568 	return (0);
1569 }
1570 
1571 int
1572 kern_linkat(struct thread *td, int fd1, int fd2, const char *path1,
1573     const char *path2, enum uio_seg segflag, int flag)
1574 {
1575 	struct nameidata nd;
1576 	int error;
1577 
1578 	if ((flag & ~(AT_SYMLINK_FOLLOW | AT_RESOLVE_BENEATH |
1579 	    AT_EMPTY_PATH)) != 0)
1580 		return (EINVAL);
1581 
1582 	NDPREINIT(&nd);
1583 	do {
1584 		bwillwrite();
1585 		NDINIT_ATRIGHTS(&nd, LOOKUP, AUDITVNODE1 | at2cnpflags(flag,
1586 		    AT_SYMLINK_FOLLOW | AT_RESOLVE_BENEATH | AT_EMPTY_PATH),
1587 		    segflag, path1, fd1, &cap_linkat_source_rights, td);
1588 		if ((error = namei(&nd)) != 0)
1589 			return (error);
1590 		NDFREE(&nd, NDF_ONLY_PNBUF);
1591 		if ((nd.ni_resflags & NIRES_EMPTYPATH) != 0) {
1592 			error = priv_check(td, PRIV_VFS_FHOPEN);
1593 			if (error != 0) {
1594 				vrele(nd.ni_vp);
1595 				return (error);
1596 			}
1597 		}
1598 		error = kern_linkat_vp(td, nd.ni_vp, fd2, path2, segflag);
1599 	} while (error ==  EAGAIN || error == ERELOOKUP);
1600 	return (error);
1601 }
1602 
1603 static int
1604 kern_linkat_vp(struct thread *td, struct vnode *vp, int fd, const char *path,
1605     enum uio_seg segflag)
1606 {
1607 	struct nameidata nd;
1608 	struct mount *mp;
1609 	int error;
1610 
1611 	if (vp->v_type == VDIR) {
1612 		vrele(vp);
1613 		return (EPERM);		/* POSIX */
1614 	}
1615 	NDINIT_ATRIGHTS(&nd, CREATE,
1616 	    LOCKPARENT | SAVENAME | AUDITVNODE2 | NOCACHE, segflag, path, fd,
1617 	    &cap_linkat_target_rights, td);
1618 	if ((error = namei(&nd)) == 0) {
1619 		if (nd.ni_vp != NULL) {
1620 			NDFREE(&nd, NDF_ONLY_PNBUF);
1621 			if (nd.ni_dvp == nd.ni_vp)
1622 				vrele(nd.ni_dvp);
1623 			else
1624 				vput(nd.ni_dvp);
1625 			vrele(nd.ni_vp);
1626 			vrele(vp);
1627 			return (EEXIST);
1628 		} else if (nd.ni_dvp->v_mount != vp->v_mount) {
1629 			/*
1630 			 * Cross-device link.  No need to recheck
1631 			 * vp->v_type, since it cannot change, except
1632 			 * to VBAD.
1633 			 */
1634 			NDFREE(&nd, NDF_ONLY_PNBUF);
1635 			vput(nd.ni_dvp);
1636 			vrele(vp);
1637 			return (EXDEV);
1638 		} else if ((error = vn_lock(vp, LK_EXCLUSIVE)) == 0) {
1639 			error = can_hardlink(vp, td->td_ucred);
1640 #ifdef MAC
1641 			if (error == 0)
1642 				error = mac_vnode_check_link(td->td_ucred,
1643 				    nd.ni_dvp, vp, &nd.ni_cnd);
1644 #endif
1645 			if (error != 0) {
1646 				vput(vp);
1647 				vput(nd.ni_dvp);
1648 				NDFREE(&nd, NDF_ONLY_PNBUF);
1649 				return (error);
1650 			}
1651 			error = vn_start_write(vp, &mp, V_NOWAIT);
1652 			if (error != 0) {
1653 				vput(vp);
1654 				vput(nd.ni_dvp);
1655 				NDFREE(&nd, NDF_ONLY_PNBUF);
1656 				error = vn_start_write(NULL, &mp,
1657 				    V_XSLEEP | PCATCH);
1658 				if (error != 0)
1659 					return (error);
1660 				return (EAGAIN);
1661 			}
1662 			error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
1663 			VOP_VPUT_PAIR(nd.ni_dvp, &vp, true);
1664 			vn_finished_write(mp);
1665 			NDFREE(&nd, NDF_ONLY_PNBUF);
1666 			vp = NULL;
1667 		} else {
1668 			vput(nd.ni_dvp);
1669 			NDFREE(&nd, NDF_ONLY_PNBUF);
1670 			vrele(vp);
1671 			return (EAGAIN);
1672 		}
1673 	}
1674 	if (vp != NULL)
1675 		vrele(vp);
1676 	return (error);
1677 }
1678 
1679 /*
1680  * Make a symbolic link.
1681  */
1682 #ifndef _SYS_SYSPROTO_H_
1683 struct symlink_args {
1684 	char	*path;
1685 	char	*link;
1686 };
1687 #endif
1688 int
1689 sys_symlink(struct thread *td, struct symlink_args *uap)
1690 {
1691 
1692 	return (kern_symlinkat(td, uap->path, AT_FDCWD, uap->link,
1693 	    UIO_USERSPACE));
1694 }
1695 
1696 #ifndef _SYS_SYSPROTO_H_
1697 struct symlinkat_args {
1698 	char	*path;
1699 	int	fd;
1700 	char	*path2;
1701 };
1702 #endif
1703 int
1704 sys_symlinkat(struct thread *td, struct symlinkat_args *uap)
1705 {
1706 
1707 	return (kern_symlinkat(td, uap->path1, uap->fd, uap->path2,
1708 	    UIO_USERSPACE));
1709 }
1710 
1711 int
1712 kern_symlinkat(struct thread *td, const char *path1, int fd, const char *path2,
1713     enum uio_seg segflg)
1714 {
1715 	struct mount *mp;
1716 	struct vattr vattr;
1717 	const char *syspath;
1718 	char *tmppath;
1719 	struct nameidata nd;
1720 	int error;
1721 
1722 	if (segflg == UIO_SYSSPACE) {
1723 		syspath = path1;
1724 	} else {
1725 		tmppath = uma_zalloc(namei_zone, M_WAITOK);
1726 		if ((error = copyinstr(path1, tmppath, MAXPATHLEN, NULL)) != 0)
1727 			goto out;
1728 		syspath = tmppath;
1729 	}
1730 	AUDIT_ARG_TEXT(syspath);
1731 	NDPREINIT(&nd);
1732 restart:
1733 	bwillwrite();
1734 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
1735 	    NOCACHE, segflg, path2, fd, &cap_symlinkat_rights,
1736 	    td);
1737 	if ((error = namei(&nd)) != 0)
1738 		goto out;
1739 	if (nd.ni_vp) {
1740 		NDFREE(&nd, NDF_ONLY_PNBUF);
1741 		if (nd.ni_vp == nd.ni_dvp)
1742 			vrele(nd.ni_dvp);
1743 		else
1744 			vput(nd.ni_dvp);
1745 		vrele(nd.ni_vp);
1746 		nd.ni_vp = NULL;
1747 		error = EEXIST;
1748 		goto out;
1749 	}
1750 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1751 		NDFREE(&nd, NDF_ONLY_PNBUF);
1752 		vput(nd.ni_dvp);
1753 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1754 			goto out;
1755 		goto restart;
1756 	}
1757 	VATTR_NULL(&vattr);
1758 	vattr.va_mode = ACCESSPERMS &~ td->td_proc->p_pd->pd_cmask;
1759 #ifdef MAC
1760 	vattr.va_type = VLNK;
1761 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1762 	    &vattr);
1763 	if (error != 0)
1764 		goto out2;
1765 #endif
1766 	error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, syspath);
1767 #ifdef MAC
1768 out2:
1769 #endif
1770 	VOP_VPUT_PAIR(nd.ni_dvp, error == 0 ? &nd.ni_vp : NULL, true);
1771 	vn_finished_write(mp);
1772 	NDFREE(&nd, NDF_ONLY_PNBUF);
1773 	if (error == ERELOOKUP)
1774 		goto restart;
1775 out:
1776 	if (segflg != UIO_SYSSPACE)
1777 		uma_zfree(namei_zone, tmppath);
1778 	return (error);
1779 }
1780 
1781 /*
1782  * Delete a whiteout from the filesystem.
1783  */
1784 #ifndef _SYS_SYSPROTO_H_
1785 struct undelete_args {
1786 	char *path;
1787 };
1788 #endif
1789 int
1790 sys_undelete(struct thread *td, struct undelete_args *uap)
1791 {
1792 	struct mount *mp;
1793 	struct nameidata nd;
1794 	int error;
1795 
1796 	NDPREINIT(&nd);
1797 restart:
1798 	bwillwrite();
1799 	NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | AUDITVNODE1,
1800 	    UIO_USERSPACE, uap->path, td);
1801 	error = namei(&nd);
1802 	if (error != 0)
1803 		return (error);
1804 
1805 	if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) {
1806 		NDFREE(&nd, NDF_ONLY_PNBUF);
1807 		if (nd.ni_vp == nd.ni_dvp)
1808 			vrele(nd.ni_dvp);
1809 		else
1810 			vput(nd.ni_dvp);
1811 		if (nd.ni_vp)
1812 			vrele(nd.ni_vp);
1813 		return (EEXIST);
1814 	}
1815 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1816 		NDFREE(&nd, NDF_ONLY_PNBUF);
1817 		vput(nd.ni_dvp);
1818 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1819 			return (error);
1820 		goto restart;
1821 	}
1822 	error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE);
1823 	NDFREE(&nd, NDF_ONLY_PNBUF);
1824 	vput(nd.ni_dvp);
1825 	vn_finished_write(mp);
1826 	if (error == ERELOOKUP)
1827 		goto restart;
1828 	return (error);
1829 }
1830 
1831 /*
1832  * Delete a name from the filesystem.
1833  */
1834 #ifndef _SYS_SYSPROTO_H_
1835 struct unlink_args {
1836 	char	*path;
1837 };
1838 #endif
1839 int
1840 sys_unlink(struct thread *td, struct unlink_args *uap)
1841 {
1842 
1843 	return (kern_funlinkat(td, AT_FDCWD, uap->path, FD_NONE, UIO_USERSPACE,
1844 	    0, 0));
1845 }
1846 
1847 static int
1848 kern_funlinkat_ex(struct thread *td, int dfd, const char *path, int fd,
1849     int flag, enum uio_seg pathseg, ino_t oldinum)
1850 {
1851 
1852 	if ((flag & ~(AT_REMOVEDIR | AT_RESOLVE_BENEATH)) != 0)
1853 		return (EINVAL);
1854 
1855 	if ((flag & AT_REMOVEDIR) != 0)
1856 		return (kern_frmdirat(td, dfd, path, fd, UIO_USERSPACE, 0));
1857 
1858 	return (kern_funlinkat(td, dfd, path, fd, UIO_USERSPACE, 0, 0));
1859 }
1860 
1861 #ifndef _SYS_SYSPROTO_H_
1862 struct unlinkat_args {
1863 	int	fd;
1864 	char	*path;
1865 	int	flag;
1866 };
1867 #endif
1868 int
1869 sys_unlinkat(struct thread *td, struct unlinkat_args *uap)
1870 {
1871 
1872 	return (kern_funlinkat_ex(td, uap->fd, uap->path, FD_NONE, uap->flag,
1873 	    UIO_USERSPACE, 0));
1874 }
1875 
1876 #ifndef _SYS_SYSPROTO_H_
1877 struct funlinkat_args {
1878 	int		dfd;
1879 	const char	*path;
1880 	int		fd;
1881 	int		flag;
1882 };
1883 #endif
1884 int
1885 sys_funlinkat(struct thread *td, struct funlinkat_args *uap)
1886 {
1887 
1888 	return (kern_funlinkat_ex(td, uap->dfd, uap->path, uap->fd, uap->flag,
1889 	    UIO_USERSPACE, 0));
1890 }
1891 
1892 int
1893 kern_funlinkat(struct thread *td, int dfd, const char *path, int fd,
1894     enum uio_seg pathseg, int flag, ino_t oldinum)
1895 {
1896 	struct mount *mp;
1897 	struct file *fp;
1898 	struct vnode *vp;
1899 	struct nameidata nd;
1900 	struct stat sb;
1901 	int error;
1902 
1903 	fp = NULL;
1904 	if (fd != FD_NONE) {
1905 		error = getvnode_path(td, fd, &cap_no_rights, &fp);
1906 		if (error != 0)
1907 			return (error);
1908 	}
1909 
1910 	NDPREINIT(&nd);
1911 restart:
1912 	bwillwrite();
1913 	NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1 |
1914 	    at2cnpflags(flag, AT_RESOLVE_BENEATH),
1915 	    pathseg, path, dfd, &cap_unlinkat_rights, td);
1916 	if ((error = namei(&nd)) != 0) {
1917 		if (error == EINVAL)
1918 			error = EPERM;
1919 		goto fdout;
1920 	}
1921 	vp = nd.ni_vp;
1922 	if (vp->v_type == VDIR && oldinum == 0) {
1923 		error = EPERM;		/* POSIX */
1924 	} else if (oldinum != 0 &&
1925 	    ((error = VOP_STAT(vp, &sb, td->td_ucred, NOCRED)) == 0) &&
1926 	    sb.st_ino != oldinum) {
1927 		error = EIDRM;	/* Identifier removed */
1928 	} else if (fp != NULL && fp->f_vnode != vp) {
1929 		if (VN_IS_DOOMED(fp->f_vnode))
1930 			error = EBADF;
1931 		else
1932 			error = EDEADLK;
1933 	} else {
1934 		/*
1935 		 * The root of a mounted filesystem cannot be deleted.
1936 		 *
1937 		 * XXX: can this only be a VDIR case?
1938 		 */
1939 		if (vp->v_vflag & VV_ROOT)
1940 			error = EBUSY;
1941 	}
1942 	if (error == 0) {
1943 		if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1944 			NDFREE(&nd, NDF_ONLY_PNBUF);
1945 			vput(nd.ni_dvp);
1946 			if (vp == nd.ni_dvp)
1947 				vrele(vp);
1948 			else
1949 				vput(vp);
1950 			if ((error = vn_start_write(NULL, &mp,
1951 			    V_XSLEEP | PCATCH)) != 0) {
1952 				goto fdout;
1953 			}
1954 			goto restart;
1955 		}
1956 #ifdef MAC
1957 		error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
1958 		    &nd.ni_cnd);
1959 		if (error != 0)
1960 			goto out;
1961 #endif
1962 		vfs_notify_upper(vp, VFS_NOTIFY_UPPER_UNLINK);
1963 		error = VOP_REMOVE(nd.ni_dvp, vp, &nd.ni_cnd);
1964 #ifdef MAC
1965 out:
1966 #endif
1967 		vn_finished_write(mp);
1968 	}
1969 	NDFREE(&nd, NDF_ONLY_PNBUF);
1970 	vput(nd.ni_dvp);
1971 	if (vp == nd.ni_dvp)
1972 		vrele(vp);
1973 	else
1974 		vput(vp);
1975 	if (error == ERELOOKUP)
1976 		goto restart;
1977 fdout:
1978 	if (fp != NULL)
1979 		fdrop(fp, td);
1980 	return (error);
1981 }
1982 
1983 /*
1984  * Reposition read/write file offset.
1985  */
1986 #ifndef _SYS_SYSPROTO_H_
1987 struct lseek_args {
1988 	int	fd;
1989 	int	pad;
1990 	off_t	offset;
1991 	int	whence;
1992 };
1993 #endif
1994 int
1995 sys_lseek(struct thread *td, struct lseek_args *uap)
1996 {
1997 
1998 	return (kern_lseek(td, uap->fd, uap->offset, uap->whence));
1999 }
2000 
2001 int
2002 kern_lseek(struct thread *td, int fd, off_t offset, int whence)
2003 {
2004 	struct file *fp;
2005 	int error;
2006 
2007 	AUDIT_ARG_FD(fd);
2008 	error = fget(td, fd, &cap_seek_rights, &fp);
2009 	if (error != 0)
2010 		return (error);
2011 	error = (fp->f_ops->fo_flags & DFLAG_SEEKABLE) != 0 ?
2012 	    fo_seek(fp, offset, whence, td) : ESPIPE;
2013 	fdrop(fp, td);
2014 	return (error);
2015 }
2016 
2017 #if defined(COMPAT_43)
2018 /*
2019  * Reposition read/write file offset.
2020  */
2021 #ifndef _SYS_SYSPROTO_H_
2022 struct olseek_args {
2023 	int	fd;
2024 	long	offset;
2025 	int	whence;
2026 };
2027 #endif
2028 int
2029 olseek(struct thread *td, struct olseek_args *uap)
2030 {
2031 
2032 	return (kern_lseek(td, uap->fd, uap->offset, uap->whence));
2033 }
2034 #endif /* COMPAT_43 */
2035 
2036 #if defined(COMPAT_FREEBSD6)
2037 /* Version with the 'pad' argument */
2038 int
2039 freebsd6_lseek(struct thread *td, struct freebsd6_lseek_args *uap)
2040 {
2041 
2042 	return (kern_lseek(td, uap->fd, uap->offset, uap->whence));
2043 }
2044 #endif
2045 
2046 /*
2047  * Check access permissions using passed credentials.
2048  */
2049 static int
2050 vn_access(struct vnode *vp, int user_flags, struct ucred *cred,
2051      struct thread *td)
2052 {
2053 	accmode_t accmode;
2054 	int error;
2055 
2056 	/* Flags == 0 means only check for existence. */
2057 	if (user_flags == 0)
2058 		return (0);
2059 
2060 	accmode = 0;
2061 	if (user_flags & R_OK)
2062 		accmode |= VREAD;
2063 	if (user_flags & W_OK)
2064 		accmode |= VWRITE;
2065 	if (user_flags & X_OK)
2066 		accmode |= VEXEC;
2067 #ifdef MAC
2068 	error = mac_vnode_check_access(cred, vp, accmode);
2069 	if (error != 0)
2070 		return (error);
2071 #endif
2072 	if ((accmode & VWRITE) == 0 || (error = vn_writechk(vp)) == 0)
2073 		error = VOP_ACCESS(vp, accmode, cred, td);
2074 	return (error);
2075 }
2076 
2077 /*
2078  * Check access permissions using "real" credentials.
2079  */
2080 #ifndef _SYS_SYSPROTO_H_
2081 struct access_args {
2082 	char	*path;
2083 	int	amode;
2084 };
2085 #endif
2086 int
2087 sys_access(struct thread *td, struct access_args *uap)
2088 {
2089 
2090 	return (kern_accessat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2091 	    0, uap->amode));
2092 }
2093 
2094 #ifndef _SYS_SYSPROTO_H_
2095 struct faccessat_args {
2096 	int	dirfd;
2097 	char	*path;
2098 	int	amode;
2099 	int	flag;
2100 }
2101 #endif
2102 int
2103 sys_faccessat(struct thread *td, struct faccessat_args *uap)
2104 {
2105 
2106 	return (kern_accessat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
2107 	    uap->amode));
2108 }
2109 
2110 int
2111 kern_accessat(struct thread *td, int fd, const char *path,
2112     enum uio_seg pathseg, int flag, int amode)
2113 {
2114 	struct ucred *cred, *usecred;
2115 	struct vnode *vp;
2116 	struct nameidata nd;
2117 	int error;
2118 
2119 	if ((flag & ~(AT_EACCESS | AT_RESOLVE_BENEATH | AT_EMPTY_PATH)) != 0)
2120 		return (EINVAL);
2121 	if (amode != F_OK && (amode & ~(R_OK | W_OK | X_OK)) != 0)
2122 		return (EINVAL);
2123 
2124 	/*
2125 	 * Create and modify a temporary credential instead of one that
2126 	 * is potentially shared (if we need one).
2127 	 */
2128 	cred = td->td_ucred;
2129 	if ((flag & AT_EACCESS) == 0 &&
2130 	    ((cred->cr_uid != cred->cr_ruid ||
2131 	    cred->cr_rgid != cred->cr_groups[0]))) {
2132 		usecred = crdup(cred);
2133 		usecred->cr_uid = cred->cr_ruid;
2134 		usecred->cr_groups[0] = cred->cr_rgid;
2135 		td->td_ucred = usecred;
2136 	} else
2137 		usecred = cred;
2138 	AUDIT_ARG_VALUE(amode);
2139 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF |
2140 	    AUDITVNODE1 | at2cnpflags(flag, AT_RESOLVE_BENEATH |
2141 	    AT_EMPTY_PATH), pathseg, path, fd, &cap_fstat_rights, td);
2142 	if ((error = namei(&nd)) != 0)
2143 		goto out;
2144 	vp = nd.ni_vp;
2145 
2146 	error = vn_access(vp, amode, usecred, td);
2147 	NDFREE_NOTHING(&nd);
2148 	vput(vp);
2149 out:
2150 	if (usecred != cred) {
2151 		td->td_ucred = cred;
2152 		crfree(usecred);
2153 	}
2154 	return (error);
2155 }
2156 
2157 /*
2158  * Check access permissions using "effective" credentials.
2159  */
2160 #ifndef _SYS_SYSPROTO_H_
2161 struct eaccess_args {
2162 	char	*path;
2163 	int	amode;
2164 };
2165 #endif
2166 int
2167 sys_eaccess(struct thread *td, struct eaccess_args *uap)
2168 {
2169 
2170 	return (kern_accessat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2171 	    AT_EACCESS, uap->amode));
2172 }
2173 
2174 #if defined(COMPAT_43)
2175 /*
2176  * Get file status; this version follows links.
2177  */
2178 #ifndef _SYS_SYSPROTO_H_
2179 struct ostat_args {
2180 	char	*path;
2181 	struct ostat *ub;
2182 };
2183 #endif
2184 int
2185 ostat(struct thread *td, struct ostat_args *uap)
2186 {
2187 	struct stat sb;
2188 	struct ostat osb;
2189 	int error;
2190 
2191 	error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE,
2192 	    &sb, NULL);
2193 	if (error != 0)
2194 		return (error);
2195 	cvtstat(&sb, &osb);
2196 	return (copyout(&osb, uap->ub, sizeof (osb)));
2197 }
2198 
2199 /*
2200  * Get file status; this version does not follow links.
2201  */
2202 #ifndef _SYS_SYSPROTO_H_
2203 struct olstat_args {
2204 	char	*path;
2205 	struct ostat *ub;
2206 };
2207 #endif
2208 int
2209 olstat(struct thread *td, struct olstat_args *uap)
2210 {
2211 	struct stat sb;
2212 	struct ostat osb;
2213 	int error;
2214 
2215 	error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2216 	    UIO_USERSPACE, &sb, NULL);
2217 	if (error != 0)
2218 		return (error);
2219 	cvtstat(&sb, &osb);
2220 	return (copyout(&osb, uap->ub, sizeof (osb)));
2221 }
2222 
2223 /*
2224  * Convert from an old to a new stat structure.
2225  * XXX: many values are blindly truncated.
2226  */
2227 void
2228 cvtstat(struct stat *st, struct ostat *ost)
2229 {
2230 
2231 	bzero(ost, sizeof(*ost));
2232 	ost->st_dev = st->st_dev;
2233 	ost->st_ino = st->st_ino;
2234 	ost->st_mode = st->st_mode;
2235 	ost->st_nlink = st->st_nlink;
2236 	ost->st_uid = st->st_uid;
2237 	ost->st_gid = st->st_gid;
2238 	ost->st_rdev = st->st_rdev;
2239 	ost->st_size = MIN(st->st_size, INT32_MAX);
2240 	ost->st_atim = st->st_atim;
2241 	ost->st_mtim = st->st_mtim;
2242 	ost->st_ctim = st->st_ctim;
2243 	ost->st_blksize = st->st_blksize;
2244 	ost->st_blocks = st->st_blocks;
2245 	ost->st_flags = st->st_flags;
2246 	ost->st_gen = st->st_gen;
2247 }
2248 #endif /* COMPAT_43 */
2249 
2250 #if defined(COMPAT_43) || defined(COMPAT_FREEBSD11)
2251 int ino64_trunc_error;
2252 SYSCTL_INT(_vfs, OID_AUTO, ino64_trunc_error, CTLFLAG_RW,
2253     &ino64_trunc_error, 0,
2254     "Error on truncation of device, file or inode number, or link count");
2255 
2256 int
2257 freebsd11_cvtstat(struct stat *st, struct freebsd11_stat *ost)
2258 {
2259 
2260 	ost->st_dev = st->st_dev;
2261 	if (ost->st_dev != st->st_dev) {
2262 		switch (ino64_trunc_error) {
2263 		default:
2264 			/*
2265 			 * Since dev_t is almost raw, don't clamp to the
2266 			 * maximum for case 2, but ignore the error.
2267 			 */
2268 			break;
2269 		case 1:
2270 			return (EOVERFLOW);
2271 		}
2272 	}
2273 	ost->st_ino = st->st_ino;
2274 	if (ost->st_ino != st->st_ino) {
2275 		switch (ino64_trunc_error) {
2276 		default:
2277 		case 0:
2278 			break;
2279 		case 1:
2280 			return (EOVERFLOW);
2281 		case 2:
2282 			ost->st_ino = UINT32_MAX;
2283 			break;
2284 		}
2285 	}
2286 	ost->st_mode = st->st_mode;
2287 	ost->st_nlink = st->st_nlink;
2288 	if (ost->st_nlink != st->st_nlink) {
2289 		switch (ino64_trunc_error) {
2290 		default:
2291 		case 0:
2292 			break;
2293 		case 1:
2294 			return (EOVERFLOW);
2295 		case 2:
2296 			ost->st_nlink = UINT16_MAX;
2297 			break;
2298 		}
2299 	}
2300 	ost->st_uid = st->st_uid;
2301 	ost->st_gid = st->st_gid;
2302 	ost->st_rdev = st->st_rdev;
2303 	if (ost->st_rdev != st->st_rdev) {
2304 		switch (ino64_trunc_error) {
2305 		default:
2306 			break;
2307 		case 1:
2308 			return (EOVERFLOW);
2309 		}
2310 	}
2311 	ost->st_atim = st->st_atim;
2312 	ost->st_mtim = st->st_mtim;
2313 	ost->st_ctim = st->st_ctim;
2314 	ost->st_size = st->st_size;
2315 	ost->st_blocks = st->st_blocks;
2316 	ost->st_blksize = st->st_blksize;
2317 	ost->st_flags = st->st_flags;
2318 	ost->st_gen = st->st_gen;
2319 	ost->st_lspare = 0;
2320 	ost->st_birthtim = st->st_birthtim;
2321 	bzero((char *)&ost->st_birthtim + sizeof(ost->st_birthtim),
2322 	    sizeof(*ost) - offsetof(struct freebsd11_stat,
2323 	    st_birthtim) - sizeof(ost->st_birthtim));
2324 	return (0);
2325 }
2326 
2327 int
2328 freebsd11_stat(struct thread *td, struct freebsd11_stat_args* uap)
2329 {
2330 	struct stat sb;
2331 	struct freebsd11_stat osb;
2332 	int error;
2333 
2334 	error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE,
2335 	    &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_lstat(struct thread *td, struct freebsd11_lstat_args* uap)
2346 {
2347 	struct stat sb;
2348 	struct freebsd11_stat osb;
2349 	int error;
2350 
2351 	error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2352 	    UIO_USERSPACE, &sb, NULL);
2353 	if (error != 0)
2354 		return (error);
2355 	error = freebsd11_cvtstat(&sb, &osb);
2356 	if (error == 0)
2357 		error = copyout(&osb, uap->ub, sizeof(osb));
2358 	return (error);
2359 }
2360 
2361 int
2362 freebsd11_fhstat(struct thread *td, struct freebsd11_fhstat_args* uap)
2363 {
2364 	struct fhandle fh;
2365 	struct stat sb;
2366 	struct freebsd11_stat osb;
2367 	int error;
2368 
2369 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
2370 	if (error != 0)
2371 		return (error);
2372 	error = kern_fhstat(td, fh, &sb);
2373 	if (error != 0)
2374 		return (error);
2375 	error = freebsd11_cvtstat(&sb, &osb);
2376 	if (error == 0)
2377 		error = copyout(&osb, uap->sb, sizeof(osb));
2378 	return (error);
2379 }
2380 
2381 int
2382 freebsd11_fstatat(struct thread *td, struct freebsd11_fstatat_args* uap)
2383 {
2384 	struct stat sb;
2385 	struct freebsd11_stat osb;
2386 	int error;
2387 
2388 	error = kern_statat(td, uap->flag, uap->fd, uap->path,
2389 	    UIO_USERSPACE, &sb, NULL);
2390 	if (error != 0)
2391 		return (error);
2392 	error = freebsd11_cvtstat(&sb, &osb);
2393 	if (error == 0)
2394 		error = copyout(&osb, uap->buf, sizeof(osb));
2395 	return (error);
2396 }
2397 #endif	/* COMPAT_FREEBSD11 */
2398 
2399 /*
2400  * Get file status
2401  */
2402 #ifndef _SYS_SYSPROTO_H_
2403 struct fstatat_args {
2404 	int	fd;
2405 	char	*path;
2406 	struct stat	*buf;
2407 	int	flag;
2408 }
2409 #endif
2410 int
2411 sys_fstatat(struct thread *td, struct fstatat_args *uap)
2412 {
2413 	struct stat sb;
2414 	int error;
2415 
2416 	error = kern_statat(td, uap->flag, uap->fd, uap->path,
2417 	    UIO_USERSPACE, &sb, NULL);
2418 	if (error == 0)
2419 		error = copyout(&sb, uap->buf, sizeof (sb));
2420 	return (error);
2421 }
2422 
2423 int
2424 kern_statat(struct thread *td, int flag, int fd, const char *path,
2425     enum uio_seg pathseg, struct stat *sbp,
2426     void (*hook)(struct vnode *vp, struct stat *sbp))
2427 {
2428 	struct nameidata nd;
2429 	int error;
2430 
2431 	if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH |
2432 	    AT_EMPTY_PATH)) != 0)
2433 		return (EINVAL);
2434 
2435 	NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(flag, AT_RESOLVE_BENEATH |
2436 	    AT_SYMLINK_NOFOLLOW | AT_EMPTY_PATH) | LOCKSHARED | LOCKLEAF |
2437 	    AUDITVNODE1, pathseg, path, fd, &cap_fstat_rights, td);
2438 
2439 	if ((error = namei(&nd)) != 0) {
2440 		if (error == ENOTDIR &&
2441 		    (nd.ni_resflags & NIRES_EMPTYPATH) != 0)
2442 			error = kern_fstat(td, fd, sbp);
2443 		return (error);
2444 	}
2445 	error = VOP_STAT(nd.ni_vp, sbp, td->td_ucred, NOCRED);
2446 	if (error == 0) {
2447 		if (__predict_false(hook != NULL))
2448 			hook(nd.ni_vp, sbp);
2449 	}
2450 	NDFREE_NOTHING(&nd);
2451 	vput(nd.ni_vp);
2452 #ifdef __STAT_TIME_T_EXT
2453 	sbp->st_atim_ext = 0;
2454 	sbp->st_mtim_ext = 0;
2455 	sbp->st_ctim_ext = 0;
2456 	sbp->st_btim_ext = 0;
2457 #endif
2458 #ifdef KTRACE
2459 	if (KTRPOINT(td, KTR_STRUCT))
2460 		ktrstat_error(sbp, error);
2461 #endif
2462 	return (error);
2463 }
2464 
2465 #if defined(COMPAT_FREEBSD11)
2466 /*
2467  * Implementation of the NetBSD [l]stat() functions.
2468  */
2469 void
2470 freebsd11_cvtnstat(struct stat *sb, struct nstat *nsb)
2471 {
2472 
2473 	bzero(nsb, sizeof(*nsb));
2474 	nsb->st_dev = sb->st_dev;
2475 	nsb->st_ino = sb->st_ino;
2476 	nsb->st_mode = sb->st_mode;
2477 	nsb->st_nlink = sb->st_nlink;
2478 	nsb->st_uid = sb->st_uid;
2479 	nsb->st_gid = sb->st_gid;
2480 	nsb->st_rdev = sb->st_rdev;
2481 	nsb->st_atim = sb->st_atim;
2482 	nsb->st_mtim = sb->st_mtim;
2483 	nsb->st_ctim = sb->st_ctim;
2484 	nsb->st_size = sb->st_size;
2485 	nsb->st_blocks = sb->st_blocks;
2486 	nsb->st_blksize = sb->st_blksize;
2487 	nsb->st_flags = sb->st_flags;
2488 	nsb->st_gen = sb->st_gen;
2489 	nsb->st_birthtim = sb->st_birthtim;
2490 }
2491 
2492 #ifndef _SYS_SYSPROTO_H_
2493 struct freebsd11_nstat_args {
2494 	char	*path;
2495 	struct nstat *ub;
2496 };
2497 #endif
2498 int
2499 freebsd11_nstat(struct thread *td, struct freebsd11_nstat_args *uap)
2500 {
2501 	struct stat sb;
2502 	struct nstat nsb;
2503 	int error;
2504 
2505 	error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE,
2506 	    &sb, NULL);
2507 	if (error != 0)
2508 		return (error);
2509 	freebsd11_cvtnstat(&sb, &nsb);
2510 	return (copyout(&nsb, uap->ub, sizeof (nsb)));
2511 }
2512 
2513 /*
2514  * NetBSD lstat.  Get file status; this version does not follow links.
2515  */
2516 #ifndef _SYS_SYSPROTO_H_
2517 struct freebsd11_nlstat_args {
2518 	char	*path;
2519 	struct nstat *ub;
2520 };
2521 #endif
2522 int
2523 freebsd11_nlstat(struct thread *td, struct freebsd11_nlstat_args *uap)
2524 {
2525 	struct stat sb;
2526 	struct nstat nsb;
2527 	int error;
2528 
2529 	error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2530 	    UIO_USERSPACE, &sb, NULL);
2531 	if (error != 0)
2532 		return (error);
2533 	freebsd11_cvtnstat(&sb, &nsb);
2534 	return (copyout(&nsb, uap->ub, sizeof (nsb)));
2535 }
2536 #endif /* COMPAT_FREEBSD11 */
2537 
2538 /*
2539  * Get configurable pathname variables.
2540  */
2541 #ifndef _SYS_SYSPROTO_H_
2542 struct pathconf_args {
2543 	char	*path;
2544 	int	name;
2545 };
2546 #endif
2547 int
2548 sys_pathconf(struct thread *td, struct pathconf_args *uap)
2549 {
2550 	long value;
2551 	int error;
2552 
2553 	error = kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name, FOLLOW,
2554 	    &value);
2555 	if (error == 0)
2556 		td->td_retval[0] = value;
2557 	return (error);
2558 }
2559 
2560 #ifndef _SYS_SYSPROTO_H_
2561 struct lpathconf_args {
2562 	char	*path;
2563 	int	name;
2564 };
2565 #endif
2566 int
2567 sys_lpathconf(struct thread *td, struct lpathconf_args *uap)
2568 {
2569 	long value;
2570 	int error;
2571 
2572 	error = kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name,
2573 	    NOFOLLOW, &value);
2574 	if (error == 0)
2575 		td->td_retval[0] = value;
2576 	return (error);
2577 }
2578 
2579 int
2580 kern_pathconf(struct thread *td, const char *path, enum uio_seg pathseg,
2581     int name, u_long flags, long *valuep)
2582 {
2583 	struct nameidata nd;
2584 	int error;
2585 
2586 	NDINIT(&nd, LOOKUP, LOCKSHARED | LOCKLEAF | AUDITVNODE1 | flags,
2587 	    pathseg, path, td);
2588 	if ((error = namei(&nd)) != 0)
2589 		return (error);
2590 	NDFREE_NOTHING(&nd);
2591 
2592 	error = VOP_PATHCONF(nd.ni_vp, name, valuep);
2593 	vput(nd.ni_vp);
2594 	return (error);
2595 }
2596 
2597 /*
2598  * Return target name of a symbolic link.
2599  */
2600 #ifndef _SYS_SYSPROTO_H_
2601 struct readlink_args {
2602 	char	*path;
2603 	char	*buf;
2604 	size_t	count;
2605 };
2606 #endif
2607 int
2608 sys_readlink(struct thread *td, struct readlink_args *uap)
2609 {
2610 
2611 	return (kern_readlinkat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2612 	    uap->buf, UIO_USERSPACE, uap->count));
2613 }
2614 #ifndef _SYS_SYSPROTO_H_
2615 struct readlinkat_args {
2616 	int	fd;
2617 	char	*path;
2618 	char	*buf;
2619 	size_t	bufsize;
2620 };
2621 #endif
2622 int
2623 sys_readlinkat(struct thread *td, struct readlinkat_args *uap)
2624 {
2625 
2626 	return (kern_readlinkat(td, uap->fd, uap->path, UIO_USERSPACE,
2627 	    uap->buf, UIO_USERSPACE, uap->bufsize));
2628 }
2629 
2630 int
2631 kern_readlinkat(struct thread *td, int fd, const char *path,
2632     enum uio_seg pathseg, char *buf, enum uio_seg bufseg, size_t count)
2633 {
2634 	struct vnode *vp;
2635 	struct nameidata nd;
2636 	int error;
2637 
2638 	if (count > IOSIZE_MAX)
2639 		return (EINVAL);
2640 
2641 	NDINIT_AT(&nd, LOOKUP, NOFOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1 |
2642 	    EMPTYPATH, pathseg, path, fd, td);
2643 
2644 	if ((error = namei(&nd)) != 0)
2645 		return (error);
2646 	NDFREE_NOTHING(&nd);
2647 	vp = nd.ni_vp;
2648 
2649 	error = kern_readlink_vp(vp, buf, bufseg, count, td);
2650 	vput(vp);
2651 
2652 	return (error);
2653 }
2654 
2655 /*
2656  * Helper function to readlink from a vnode
2657  */
2658 static int
2659 kern_readlink_vp(struct vnode *vp, char *buf, enum uio_seg bufseg, size_t count,
2660     struct thread *td)
2661 {
2662 	struct iovec aiov;
2663 	struct uio auio;
2664 	int error;
2665 
2666 	ASSERT_VOP_LOCKED(vp, "kern_readlink_vp(): vp not locked");
2667 #ifdef MAC
2668 	error = mac_vnode_check_readlink(td->td_ucred, vp);
2669 	if (error != 0)
2670 		return (error);
2671 #endif
2672 	if (vp->v_type != VLNK && (vp->v_vflag & VV_READLINK) == 0)
2673 		return (EINVAL);
2674 
2675 	aiov.iov_base = buf;
2676 	aiov.iov_len = count;
2677 	auio.uio_iov = &aiov;
2678 	auio.uio_iovcnt = 1;
2679 	auio.uio_offset = 0;
2680 	auio.uio_rw = UIO_READ;
2681 	auio.uio_segflg = bufseg;
2682 	auio.uio_td = td;
2683 	auio.uio_resid = count;
2684 	error = VOP_READLINK(vp, &auio, td->td_ucred);
2685 	td->td_retval[0] = count - auio.uio_resid;
2686 	return (error);
2687 }
2688 
2689 /*
2690  * Common implementation code for chflags() and fchflags().
2691  */
2692 static int
2693 setfflags(struct thread *td, struct vnode *vp, u_long flags)
2694 {
2695 	struct mount *mp;
2696 	struct vattr vattr;
2697 	int error;
2698 
2699 	/* We can't support the value matching VNOVAL. */
2700 	if (flags == VNOVAL)
2701 		return (EOPNOTSUPP);
2702 
2703 	/*
2704 	 * Prevent non-root users from setting flags on devices.  When
2705 	 * a device is reused, users can retain ownership of the device
2706 	 * if they are allowed to set flags and programs assume that
2707 	 * chown can't fail when done as root.
2708 	 */
2709 	if (vp->v_type == VCHR || vp->v_type == VBLK) {
2710 		error = priv_check(td, PRIV_VFS_CHFLAGS_DEV);
2711 		if (error != 0)
2712 			return (error);
2713 	}
2714 
2715 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2716 		return (error);
2717 	VATTR_NULL(&vattr);
2718 	vattr.va_flags = flags;
2719 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2720 #ifdef MAC
2721 	error = mac_vnode_check_setflags(td->td_ucred, vp, vattr.va_flags);
2722 	if (error == 0)
2723 #endif
2724 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
2725 	VOP_UNLOCK(vp);
2726 	vn_finished_write(mp);
2727 	return (error);
2728 }
2729 
2730 /*
2731  * Change flags of a file given a path name.
2732  */
2733 #ifndef _SYS_SYSPROTO_H_
2734 struct chflags_args {
2735 	const char *path;
2736 	u_long	flags;
2737 };
2738 #endif
2739 int
2740 sys_chflags(struct thread *td, struct chflags_args *uap)
2741 {
2742 
2743 	return (kern_chflagsat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2744 	    uap->flags, 0));
2745 }
2746 
2747 #ifndef _SYS_SYSPROTO_H_
2748 struct chflagsat_args {
2749 	int	fd;
2750 	const char *path;
2751 	u_long	flags;
2752 	int	atflag;
2753 }
2754 #endif
2755 int
2756 sys_chflagsat(struct thread *td, struct chflagsat_args *uap)
2757 {
2758 
2759 	return (kern_chflagsat(td, uap->fd, uap->path, UIO_USERSPACE,
2760 	    uap->flags, uap->atflag));
2761 }
2762 
2763 /*
2764  * Same as chflags() but doesn't follow symlinks.
2765  */
2766 #ifndef _SYS_SYSPROTO_H_
2767 struct lchflags_args {
2768 	const char *path;
2769 	u_long flags;
2770 };
2771 #endif
2772 int
2773 sys_lchflags(struct thread *td, struct lchflags_args *uap)
2774 {
2775 
2776 	return (kern_chflagsat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2777 	    uap->flags, AT_SYMLINK_NOFOLLOW));
2778 }
2779 
2780 static int
2781 kern_chflagsat(struct thread *td, int fd, const char *path,
2782     enum uio_seg pathseg, u_long flags, int atflag)
2783 {
2784 	struct nameidata nd;
2785 	int error;
2786 
2787 	if ((atflag & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH |
2788 	    AT_EMPTY_PATH)) != 0)
2789 		return (EINVAL);
2790 
2791 	AUDIT_ARG_FFLAGS(flags);
2792 	NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(atflag, AT_SYMLINK_NOFOLLOW |
2793 	    AT_RESOLVE_BENEATH | AT_EMPTY_PATH) | AUDITVNODE1, pathseg, path,
2794 	    fd, &cap_fchflags_rights, td);
2795 	if ((error = namei(&nd)) != 0)
2796 		return (error);
2797 	NDFREE_NOTHING(&nd);
2798 	error = setfflags(td, nd.ni_vp, flags);
2799 	vrele(nd.ni_vp);
2800 	return (error);
2801 }
2802 
2803 /*
2804  * Change flags of a file given a file descriptor.
2805  */
2806 #ifndef _SYS_SYSPROTO_H_
2807 struct fchflags_args {
2808 	int	fd;
2809 	u_long	flags;
2810 };
2811 #endif
2812 int
2813 sys_fchflags(struct thread *td, struct fchflags_args *uap)
2814 {
2815 	struct file *fp;
2816 	int error;
2817 
2818 	AUDIT_ARG_FD(uap->fd);
2819 	AUDIT_ARG_FFLAGS(uap->flags);
2820 	error = getvnode(td, uap->fd, &cap_fchflags_rights,
2821 	    &fp);
2822 	if (error != 0)
2823 		return (error);
2824 #ifdef AUDIT
2825 	if (AUDITING_TD(td)) {
2826 		vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
2827 		AUDIT_ARG_VNODE1(fp->f_vnode);
2828 		VOP_UNLOCK(fp->f_vnode);
2829 	}
2830 #endif
2831 	error = setfflags(td, fp->f_vnode, uap->flags);
2832 	fdrop(fp, td);
2833 	return (error);
2834 }
2835 
2836 /*
2837  * Common implementation code for chmod(), lchmod() and fchmod().
2838  */
2839 int
2840 setfmode(struct thread *td, struct ucred *cred, struct vnode *vp, int mode)
2841 {
2842 	struct mount *mp;
2843 	struct vattr vattr;
2844 	int error;
2845 
2846 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2847 		return (error);
2848 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2849 	VATTR_NULL(&vattr);
2850 	vattr.va_mode = mode & ALLPERMS;
2851 #ifdef MAC
2852 	error = mac_vnode_check_setmode(cred, vp, vattr.va_mode);
2853 	if (error == 0)
2854 #endif
2855 		error = VOP_SETATTR(vp, &vattr, cred);
2856 	VOP_UNLOCK(vp);
2857 	vn_finished_write(mp);
2858 	return (error);
2859 }
2860 
2861 /*
2862  * Change mode of a file given path name.
2863  */
2864 #ifndef _SYS_SYSPROTO_H_
2865 struct chmod_args {
2866 	char	*path;
2867 	int	mode;
2868 };
2869 #endif
2870 int
2871 sys_chmod(struct thread *td, struct chmod_args *uap)
2872 {
2873 
2874 	return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2875 	    uap->mode, 0));
2876 }
2877 
2878 #ifndef _SYS_SYSPROTO_H_
2879 struct fchmodat_args {
2880 	int	dirfd;
2881 	char	*path;
2882 	mode_t	mode;
2883 	int	flag;
2884 }
2885 #endif
2886 int
2887 sys_fchmodat(struct thread *td, struct fchmodat_args *uap)
2888 {
2889 
2890 	return (kern_fchmodat(td, uap->fd, uap->path, UIO_USERSPACE,
2891 	    uap->mode, uap->flag));
2892 }
2893 
2894 /*
2895  * Change mode of a file given path name (don't follow links.)
2896  */
2897 #ifndef _SYS_SYSPROTO_H_
2898 struct lchmod_args {
2899 	char	*path;
2900 	int	mode;
2901 };
2902 #endif
2903 int
2904 sys_lchmod(struct thread *td, struct lchmod_args *uap)
2905 {
2906 
2907 	return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2908 	    uap->mode, AT_SYMLINK_NOFOLLOW));
2909 }
2910 
2911 int
2912 kern_fchmodat(struct thread *td, int fd, const char *path,
2913     enum uio_seg pathseg, mode_t mode, int flag)
2914 {
2915 	struct nameidata nd;
2916 	int error;
2917 
2918 	if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH |
2919 	    AT_EMPTY_PATH)) != 0)
2920 		return (EINVAL);
2921 
2922 	AUDIT_ARG_MODE(mode);
2923 	NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(flag, AT_SYMLINK_NOFOLLOW |
2924 	    AT_RESOLVE_BENEATH | AT_EMPTY_PATH) | AUDITVNODE1, pathseg, path,
2925 	    fd, &cap_fchmod_rights, td);
2926 	if ((error = namei(&nd)) != 0)
2927 		return (error);
2928 	NDFREE_NOTHING(&nd);
2929 	error = setfmode(td, td->td_ucred, nd.ni_vp, mode);
2930 	vrele(nd.ni_vp);
2931 	return (error);
2932 }
2933 
2934 /*
2935  * Change mode of a file given a file descriptor.
2936  */
2937 #ifndef _SYS_SYSPROTO_H_
2938 struct fchmod_args {
2939 	int	fd;
2940 	int	mode;
2941 };
2942 #endif
2943 int
2944 sys_fchmod(struct thread *td, struct fchmod_args *uap)
2945 {
2946 	struct file *fp;
2947 	int error;
2948 
2949 	AUDIT_ARG_FD(uap->fd);
2950 	AUDIT_ARG_MODE(uap->mode);
2951 
2952 	error = fget(td, uap->fd, &cap_fchmod_rights, &fp);
2953 	if (error != 0)
2954 		return (error);
2955 	error = fo_chmod(fp, uap->mode, td->td_ucred, td);
2956 	fdrop(fp, td);
2957 	return (error);
2958 }
2959 
2960 /*
2961  * Common implementation for chown(), lchown(), and fchown()
2962  */
2963 int
2964 setfown(struct thread *td, struct ucred *cred, struct vnode *vp, uid_t uid,
2965     gid_t gid)
2966 {
2967 	struct mount *mp;
2968 	struct vattr vattr;
2969 	int error;
2970 
2971 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2972 		return (error);
2973 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2974 	VATTR_NULL(&vattr);
2975 	vattr.va_uid = uid;
2976 	vattr.va_gid = gid;
2977 #ifdef MAC
2978 	error = mac_vnode_check_setowner(cred, vp, vattr.va_uid,
2979 	    vattr.va_gid);
2980 	if (error == 0)
2981 #endif
2982 		error = VOP_SETATTR(vp, &vattr, cred);
2983 	VOP_UNLOCK(vp);
2984 	vn_finished_write(mp);
2985 	return (error);
2986 }
2987 
2988 /*
2989  * Set ownership given a path name.
2990  */
2991 #ifndef _SYS_SYSPROTO_H_
2992 struct chown_args {
2993 	char	*path;
2994 	int	uid;
2995 	int	gid;
2996 };
2997 #endif
2998 int
2999 sys_chown(struct thread *td, struct chown_args *uap)
3000 {
3001 
3002 	return (kern_fchownat(td, AT_FDCWD, uap->path, UIO_USERSPACE, uap->uid,
3003 	    uap->gid, 0));
3004 }
3005 
3006 #ifndef _SYS_SYSPROTO_H_
3007 struct fchownat_args {
3008 	int fd;
3009 	const char * path;
3010 	uid_t uid;
3011 	gid_t gid;
3012 	int flag;
3013 };
3014 #endif
3015 int
3016 sys_fchownat(struct thread *td, struct fchownat_args *uap)
3017 {
3018 
3019 	return (kern_fchownat(td, uap->fd, uap->path, UIO_USERSPACE, uap->uid,
3020 	    uap->gid, uap->flag));
3021 }
3022 
3023 int
3024 kern_fchownat(struct thread *td, int fd, const char *path,
3025     enum uio_seg pathseg, int uid, int gid, int flag)
3026 {
3027 	struct nameidata nd;
3028 	int error;
3029 
3030 	if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH |
3031 	    AT_EMPTY_PATH)) != 0)
3032 		return (EINVAL);
3033 
3034 	AUDIT_ARG_OWNER(uid, gid);
3035 	NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(flag, AT_SYMLINK_NOFOLLOW |
3036 	    AT_RESOLVE_BENEATH | AT_EMPTY_PATH) | AUDITVNODE1, pathseg, path,
3037 	    fd, &cap_fchown_rights, td);
3038 
3039 	if ((error = namei(&nd)) != 0)
3040 		return (error);
3041 	NDFREE_NOTHING(&nd);
3042 	error = setfown(td, td->td_ucred, nd.ni_vp, uid, gid);
3043 	vrele(nd.ni_vp);
3044 	return (error);
3045 }
3046 
3047 /*
3048  * Set ownership given a path name, do not cross symlinks.
3049  */
3050 #ifndef _SYS_SYSPROTO_H_
3051 struct lchown_args {
3052 	char	*path;
3053 	int	uid;
3054 	int	gid;
3055 };
3056 #endif
3057 int
3058 sys_lchown(struct thread *td, struct lchown_args *uap)
3059 {
3060 
3061 	return (kern_fchownat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
3062 	    uap->uid, uap->gid, AT_SYMLINK_NOFOLLOW));
3063 }
3064 
3065 /*
3066  * Set ownership given a file descriptor.
3067  */
3068 #ifndef _SYS_SYSPROTO_H_
3069 struct fchown_args {
3070 	int	fd;
3071 	int	uid;
3072 	int	gid;
3073 };
3074 #endif
3075 int
3076 sys_fchown(struct thread *td, struct fchown_args *uap)
3077 {
3078 	struct file *fp;
3079 	int error;
3080 
3081 	AUDIT_ARG_FD(uap->fd);
3082 	AUDIT_ARG_OWNER(uap->uid, uap->gid);
3083 	error = fget(td, uap->fd, &cap_fchown_rights, &fp);
3084 	if (error != 0)
3085 		return (error);
3086 	error = fo_chown(fp, uap->uid, uap->gid, td->td_ucred, td);
3087 	fdrop(fp, td);
3088 	return (error);
3089 }
3090 
3091 /*
3092  * Common implementation code for utimes(), lutimes(), and futimes().
3093  */
3094 static int
3095 getutimes(const struct timeval *usrtvp, enum uio_seg tvpseg,
3096     struct timespec *tsp)
3097 {
3098 	struct timeval tv[2];
3099 	const struct timeval *tvp;
3100 	int error;
3101 
3102 	if (usrtvp == NULL) {
3103 		vfs_timestamp(&tsp[0]);
3104 		tsp[1] = tsp[0];
3105 	} else {
3106 		if (tvpseg == UIO_SYSSPACE) {
3107 			tvp = usrtvp;
3108 		} else {
3109 			if ((error = copyin(usrtvp, tv, sizeof(tv))) != 0)
3110 				return (error);
3111 			tvp = tv;
3112 		}
3113 
3114 		if (tvp[0].tv_usec < 0 || tvp[0].tv_usec >= 1000000 ||
3115 		    tvp[1].tv_usec < 0 || tvp[1].tv_usec >= 1000000)
3116 			return (EINVAL);
3117 		TIMEVAL_TO_TIMESPEC(&tvp[0], &tsp[0]);
3118 		TIMEVAL_TO_TIMESPEC(&tvp[1], &tsp[1]);
3119 	}
3120 	return (0);
3121 }
3122 
3123 /*
3124  * Common implementation code for futimens(), utimensat().
3125  */
3126 #define	UTIMENS_NULL	0x1
3127 #define	UTIMENS_EXIT	0x2
3128 static int
3129 getutimens(const struct timespec *usrtsp, enum uio_seg tspseg,
3130     struct timespec *tsp, int *retflags)
3131 {
3132 	struct timespec tsnow;
3133 	int error;
3134 
3135 	vfs_timestamp(&tsnow);
3136 	*retflags = 0;
3137 	if (usrtsp == NULL) {
3138 		tsp[0] = tsnow;
3139 		tsp[1] = tsnow;
3140 		*retflags |= UTIMENS_NULL;
3141 		return (0);
3142 	}
3143 	if (tspseg == UIO_SYSSPACE) {
3144 		tsp[0] = usrtsp[0];
3145 		tsp[1] = usrtsp[1];
3146 	} else if ((error = copyin(usrtsp, tsp, sizeof(*tsp) * 2)) != 0)
3147 		return (error);
3148 	if (tsp[0].tv_nsec == UTIME_OMIT && tsp[1].tv_nsec == UTIME_OMIT)
3149 		*retflags |= UTIMENS_EXIT;
3150 	if (tsp[0].tv_nsec == UTIME_NOW && tsp[1].tv_nsec == UTIME_NOW)
3151 		*retflags |= UTIMENS_NULL;
3152 	if (tsp[0].tv_nsec == UTIME_OMIT)
3153 		tsp[0].tv_sec = VNOVAL;
3154 	else if (tsp[0].tv_nsec == UTIME_NOW)
3155 		tsp[0] = tsnow;
3156 	else if (tsp[0].tv_nsec < 0 || tsp[0].tv_nsec >= 1000000000L)
3157 		return (EINVAL);
3158 	if (tsp[1].tv_nsec == UTIME_OMIT)
3159 		tsp[1].tv_sec = VNOVAL;
3160 	else if (tsp[1].tv_nsec == UTIME_NOW)
3161 		tsp[1] = tsnow;
3162 	else if (tsp[1].tv_nsec < 0 || tsp[1].tv_nsec >= 1000000000L)
3163 		return (EINVAL);
3164 
3165 	return (0);
3166 }
3167 
3168 /*
3169  * Common implementation code for utimes(), lutimes(), futimes(), futimens(),
3170  * and utimensat().
3171  */
3172 static int
3173 setutimes(struct thread *td, struct vnode *vp, const struct timespec *ts,
3174     int numtimes, int nullflag)
3175 {
3176 	struct mount *mp;
3177 	struct vattr vattr;
3178 	int error;
3179 	bool setbirthtime;
3180 
3181 	setbirthtime = false;
3182 	vattr.va_birthtime.tv_sec = VNOVAL;
3183 	vattr.va_birthtime.tv_nsec = 0;
3184 
3185 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
3186 		return (error);
3187 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3188 	if (numtimes < 3 && VOP_GETATTR(vp, &vattr, td->td_ucred) == 0 &&
3189 	    timespeccmp(&ts[1], &vattr.va_birthtime, < ))
3190 		setbirthtime = true;
3191 	VATTR_NULL(&vattr);
3192 	vattr.va_atime = ts[0];
3193 	vattr.va_mtime = ts[1];
3194 	if (setbirthtime)
3195 		vattr.va_birthtime = ts[1];
3196 	if (numtimes > 2)
3197 		vattr.va_birthtime = ts[2];
3198 	if (nullflag)
3199 		vattr.va_vaflags |= VA_UTIMES_NULL;
3200 #ifdef MAC
3201 	error = mac_vnode_check_setutimes(td->td_ucred, vp, vattr.va_atime,
3202 	    vattr.va_mtime);
3203 #endif
3204 	if (error == 0)
3205 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3206 	VOP_UNLOCK(vp);
3207 	vn_finished_write(mp);
3208 	return (error);
3209 }
3210 
3211 /*
3212  * Set the access and modification times of a file.
3213  */
3214 #ifndef _SYS_SYSPROTO_H_
3215 struct utimes_args {
3216 	char	*path;
3217 	struct	timeval *tptr;
3218 };
3219 #endif
3220 int
3221 sys_utimes(struct thread *td, struct utimes_args *uap)
3222 {
3223 
3224 	return (kern_utimesat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
3225 	    uap->tptr, UIO_USERSPACE));
3226 }
3227 
3228 #ifndef _SYS_SYSPROTO_H_
3229 struct futimesat_args {
3230 	int fd;
3231 	const char * path;
3232 	const struct timeval * times;
3233 };
3234 #endif
3235 int
3236 sys_futimesat(struct thread *td, struct futimesat_args *uap)
3237 {
3238 
3239 	return (kern_utimesat(td, uap->fd, uap->path, UIO_USERSPACE,
3240 	    uap->times, UIO_USERSPACE));
3241 }
3242 
3243 int
3244 kern_utimesat(struct thread *td, int fd, const char *path,
3245     enum uio_seg pathseg, struct timeval *tptr, enum uio_seg tptrseg)
3246 {
3247 	struct nameidata nd;
3248 	struct timespec ts[2];
3249 	int error;
3250 
3251 	if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3252 		return (error);
3253 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, fd,
3254 	    &cap_futimes_rights, td);
3255 
3256 	if ((error = namei(&nd)) != 0)
3257 		return (error);
3258 	NDFREE_NOTHING(&nd);
3259 	error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3260 	vrele(nd.ni_vp);
3261 	return (error);
3262 }
3263 
3264 /*
3265  * Set the access and modification times of a file.
3266  */
3267 #ifndef _SYS_SYSPROTO_H_
3268 struct lutimes_args {
3269 	char	*path;
3270 	struct	timeval *tptr;
3271 };
3272 #endif
3273 int
3274 sys_lutimes(struct thread *td, struct lutimes_args *uap)
3275 {
3276 
3277 	return (kern_lutimes(td, uap->path, UIO_USERSPACE, uap->tptr,
3278 	    UIO_USERSPACE));
3279 }
3280 
3281 int
3282 kern_lutimes(struct thread *td, const char *path, enum uio_seg pathseg,
3283     struct timeval *tptr, enum uio_seg tptrseg)
3284 {
3285 	struct timespec ts[2];
3286 	struct nameidata nd;
3287 	int error;
3288 
3289 	if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3290 		return (error);
3291 	NDINIT(&nd, LOOKUP, NOFOLLOW | AUDITVNODE1, pathseg, path, td);
3292 	if ((error = namei(&nd)) != 0)
3293 		return (error);
3294 	NDFREE_NOTHING(&nd);
3295 	error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3296 	vrele(nd.ni_vp);
3297 	return (error);
3298 }
3299 
3300 /*
3301  * Set the access and modification times of a file.
3302  */
3303 #ifndef _SYS_SYSPROTO_H_
3304 struct futimes_args {
3305 	int	fd;
3306 	struct	timeval *tptr;
3307 };
3308 #endif
3309 int
3310 sys_futimes(struct thread *td, struct futimes_args *uap)
3311 {
3312 
3313 	return (kern_futimes(td, uap->fd, uap->tptr, UIO_USERSPACE));
3314 }
3315 
3316 int
3317 kern_futimes(struct thread *td, int fd, struct timeval *tptr,
3318     enum uio_seg tptrseg)
3319 {
3320 	struct timespec ts[2];
3321 	struct file *fp;
3322 	int error;
3323 
3324 	AUDIT_ARG_FD(fd);
3325 	error = getutimes(tptr, tptrseg, ts);
3326 	if (error != 0)
3327 		return (error);
3328 	error = getvnode(td, fd, &cap_futimes_rights, &fp);
3329 	if (error != 0)
3330 		return (error);
3331 #ifdef AUDIT
3332 	if (AUDITING_TD(td)) {
3333 		vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
3334 		AUDIT_ARG_VNODE1(fp->f_vnode);
3335 		VOP_UNLOCK(fp->f_vnode);
3336 	}
3337 #endif
3338 	error = setutimes(td, fp->f_vnode, ts, 2, tptr == NULL);
3339 	fdrop(fp, td);
3340 	return (error);
3341 }
3342 
3343 int
3344 sys_futimens(struct thread *td, struct futimens_args *uap)
3345 {
3346 
3347 	return (kern_futimens(td, uap->fd, uap->times, UIO_USERSPACE));
3348 }
3349 
3350 int
3351 kern_futimens(struct thread *td, int fd, struct timespec *tptr,
3352     enum uio_seg tptrseg)
3353 {
3354 	struct timespec ts[2];
3355 	struct file *fp;
3356 	int error, flags;
3357 
3358 	AUDIT_ARG_FD(fd);
3359 	error = getutimens(tptr, tptrseg, ts, &flags);
3360 	if (error != 0)
3361 		return (error);
3362 	if (flags & UTIMENS_EXIT)
3363 		return (0);
3364 	error = getvnode(td, fd, &cap_futimes_rights, &fp);
3365 	if (error != 0)
3366 		return (error);
3367 #ifdef AUDIT
3368 	if (AUDITING_TD(td)) {
3369 		vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
3370 		AUDIT_ARG_VNODE1(fp->f_vnode);
3371 		VOP_UNLOCK(fp->f_vnode);
3372 	}
3373 #endif
3374 	error = setutimes(td, fp->f_vnode, ts, 2, flags & UTIMENS_NULL);
3375 	fdrop(fp, td);
3376 	return (error);
3377 }
3378 
3379 int
3380 sys_utimensat(struct thread *td, struct utimensat_args *uap)
3381 {
3382 
3383 	return (kern_utimensat(td, uap->fd, uap->path, UIO_USERSPACE,
3384 	    uap->times, UIO_USERSPACE, uap->flag));
3385 }
3386 
3387 int
3388 kern_utimensat(struct thread *td, int fd, const char *path,
3389     enum uio_seg pathseg, struct timespec *tptr, enum uio_seg tptrseg,
3390     int flag)
3391 {
3392 	struct nameidata nd;
3393 	struct timespec ts[2];
3394 	int error, flags;
3395 
3396 	if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH |
3397 	    AT_EMPTY_PATH)) != 0)
3398 		return (EINVAL);
3399 
3400 	if ((error = getutimens(tptr, tptrseg, ts, &flags)) != 0)
3401 		return (error);
3402 	NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(flag, AT_SYMLINK_NOFOLLOW |
3403 	    AT_RESOLVE_BENEATH | AT_EMPTY_PATH) | AUDITVNODE1,
3404 	    pathseg, path, fd, &cap_futimes_rights, td);
3405 	if ((error = namei(&nd)) != 0)
3406 		return (error);
3407 	/*
3408 	 * We are allowed to call namei() regardless of 2xUTIME_OMIT.
3409 	 * POSIX states:
3410 	 * "If both tv_nsec fields are UTIME_OMIT... EACCESS may be detected."
3411 	 * "Search permission is denied by a component of the path prefix."
3412 	 */
3413 	NDFREE_NOTHING(&nd);
3414 	if ((flags & UTIMENS_EXIT) == 0)
3415 		error = setutimes(td, nd.ni_vp, ts, 2, flags & UTIMENS_NULL);
3416 	vrele(nd.ni_vp);
3417 	return (error);
3418 }
3419 
3420 /*
3421  * Truncate a file given its path name.
3422  */
3423 #ifndef _SYS_SYSPROTO_H_
3424 struct truncate_args {
3425 	char	*path;
3426 	int	pad;
3427 	off_t	length;
3428 };
3429 #endif
3430 int
3431 sys_truncate(struct thread *td, struct truncate_args *uap)
3432 {
3433 
3434 	return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3435 }
3436 
3437 int
3438 kern_truncate(struct thread *td, const char *path, enum uio_seg pathseg,
3439     off_t length)
3440 {
3441 	struct mount *mp;
3442 	struct vnode *vp;
3443 	void *rl_cookie;
3444 	struct vattr vattr;
3445 	struct nameidata nd;
3446 	int error;
3447 
3448 	if (length < 0)
3449 		return (EINVAL);
3450 	NDPREINIT(&nd);
3451 retry:
3452 	NDINIT(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, td);
3453 	if ((error = namei(&nd)) != 0)
3454 		return (error);
3455 	vp = nd.ni_vp;
3456 	rl_cookie = vn_rangelock_wlock(vp, 0, OFF_MAX);
3457 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) {
3458 		vn_rangelock_unlock(vp, rl_cookie);
3459 		vrele(vp);
3460 		return (error);
3461 	}
3462 	NDFREE(&nd, NDF_ONLY_PNBUF);
3463 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3464 	if (vp->v_type == VDIR)
3465 		error = EISDIR;
3466 #ifdef MAC
3467 	else if ((error = mac_vnode_check_write(td->td_ucred, NOCRED, vp))) {
3468 	}
3469 #endif
3470 	else if ((error = vn_writechk(vp)) == 0 &&
3471 	    (error = VOP_ACCESS(vp, VWRITE, td->td_ucred, td)) == 0) {
3472 		VATTR_NULL(&vattr);
3473 		vattr.va_size = length;
3474 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3475 	}
3476 	VOP_UNLOCK(vp);
3477 	vn_finished_write(mp);
3478 	vn_rangelock_unlock(vp, rl_cookie);
3479 	vrele(vp);
3480 	if (error == ERELOOKUP)
3481 		goto retry;
3482 	return (error);
3483 }
3484 
3485 #if defined(COMPAT_43)
3486 /*
3487  * Truncate a file given its path name.
3488  */
3489 #ifndef _SYS_SYSPROTO_H_
3490 struct otruncate_args {
3491 	char	*path;
3492 	long	length;
3493 };
3494 #endif
3495 int
3496 otruncate(struct thread *td, struct otruncate_args *uap)
3497 {
3498 
3499 	return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3500 }
3501 #endif /* COMPAT_43 */
3502 
3503 #if defined(COMPAT_FREEBSD6)
3504 /* Versions with the pad argument */
3505 int
3506 freebsd6_truncate(struct thread *td, struct freebsd6_truncate_args *uap)
3507 {
3508 
3509 	return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3510 }
3511 
3512 int
3513 freebsd6_ftruncate(struct thread *td, struct freebsd6_ftruncate_args *uap)
3514 {
3515 
3516 	return (kern_ftruncate(td, uap->fd, uap->length));
3517 }
3518 #endif
3519 
3520 int
3521 kern_fsync(struct thread *td, int fd, bool fullsync)
3522 {
3523 	struct vnode *vp;
3524 	struct mount *mp;
3525 	struct file *fp;
3526 	int error;
3527 
3528 	AUDIT_ARG_FD(fd);
3529 	error = getvnode(td, fd, &cap_fsync_rights, &fp);
3530 	if (error != 0)
3531 		return (error);
3532 	vp = fp->f_vnode;
3533 #if 0
3534 	if (!fullsync)
3535 		/* XXXKIB: compete outstanding aio writes */;
3536 #endif
3537 retry:
3538 	error = vn_start_write(vp, &mp, V_WAIT | PCATCH);
3539 	if (error != 0)
3540 		goto drop;
3541 	vn_lock(vp, vn_lktype_write(mp, vp) | LK_RETRY);
3542 	AUDIT_ARG_VNODE1(vp);
3543 	if (vp->v_object != NULL) {
3544 		VM_OBJECT_WLOCK(vp->v_object);
3545 		vm_object_page_clean(vp->v_object, 0, 0, 0);
3546 		VM_OBJECT_WUNLOCK(vp->v_object);
3547 	}
3548 	error = fullsync ? VOP_FSYNC(vp, MNT_WAIT, td) : VOP_FDATASYNC(vp, td);
3549 	VOP_UNLOCK(vp);
3550 	vn_finished_write(mp);
3551 	if (error == ERELOOKUP)
3552 		goto retry;
3553 drop:
3554 	fdrop(fp, td);
3555 	return (error);
3556 }
3557 
3558 /*
3559  * Sync an open file.
3560  */
3561 #ifndef _SYS_SYSPROTO_H_
3562 struct fsync_args {
3563 	int	fd;
3564 };
3565 #endif
3566 int
3567 sys_fsync(struct thread *td, struct fsync_args *uap)
3568 {
3569 
3570 	return (kern_fsync(td, uap->fd, true));
3571 }
3572 
3573 int
3574 sys_fdatasync(struct thread *td, struct fdatasync_args *uap)
3575 {
3576 
3577 	return (kern_fsync(td, uap->fd, false));
3578 }
3579 
3580 /*
3581  * Rename files.  Source and destination must either both be directories, or
3582  * both not be directories.  If target is a directory, it must be empty.
3583  */
3584 #ifndef _SYS_SYSPROTO_H_
3585 struct rename_args {
3586 	char	*from;
3587 	char	*to;
3588 };
3589 #endif
3590 int
3591 sys_rename(struct thread *td, struct rename_args *uap)
3592 {
3593 
3594 	return (kern_renameat(td, AT_FDCWD, uap->from, AT_FDCWD,
3595 	    uap->to, UIO_USERSPACE));
3596 }
3597 
3598 #ifndef _SYS_SYSPROTO_H_
3599 struct renameat_args {
3600 	int	oldfd;
3601 	char	*old;
3602 	int	newfd;
3603 	char	*new;
3604 };
3605 #endif
3606 int
3607 sys_renameat(struct thread *td, struct renameat_args *uap)
3608 {
3609 
3610 	return (kern_renameat(td, uap->oldfd, uap->old, uap->newfd, uap->new,
3611 	    UIO_USERSPACE));
3612 }
3613 
3614 #ifdef MAC
3615 static int
3616 kern_renameat_mac(struct thread *td, int oldfd, const char *old, int newfd,
3617     const char *new, enum uio_seg pathseg, struct nameidata *fromnd)
3618 {
3619 	int error;
3620 
3621 	NDINIT_ATRIGHTS(fromnd, DELETE, LOCKPARENT | LOCKLEAF | SAVESTART |
3622 	    AUDITVNODE1, pathseg, old, oldfd, &cap_renameat_source_rights, td);
3623 	if ((error = namei(fromnd)) != 0)
3624 		return (error);
3625 	error = mac_vnode_check_rename_from(td->td_ucred, fromnd->ni_dvp,
3626 	    fromnd->ni_vp, &fromnd->ni_cnd);
3627 	VOP_UNLOCK(fromnd->ni_dvp);
3628 	if (fromnd->ni_dvp != fromnd->ni_vp)
3629 		VOP_UNLOCK(fromnd->ni_vp);
3630 	if (error != 0) {
3631 		NDFREE(fromnd, NDF_ONLY_PNBUF);
3632 		vrele(fromnd->ni_dvp);
3633 		vrele(fromnd->ni_vp);
3634 		if (fromnd->ni_startdir)
3635 			vrele(fromnd->ni_startdir);
3636 	}
3637 	return (error);
3638 }
3639 #endif
3640 
3641 int
3642 kern_renameat(struct thread *td, int oldfd, const char *old, int newfd,
3643     const char *new, enum uio_seg pathseg)
3644 {
3645 	struct mount *mp = NULL;
3646 	struct vnode *tvp, *fvp, *tdvp;
3647 	struct nameidata fromnd, tond;
3648 	u_int64_t tondflags;
3649 	int error;
3650 
3651 again:
3652 	bwillwrite();
3653 #ifdef MAC
3654 	if (mac_vnode_check_rename_from_enabled()) {
3655 		error = kern_renameat_mac(td, oldfd, old, newfd, new, pathseg,
3656 		    &fromnd);
3657 		if (error != 0)
3658 			return (error);
3659 	} else {
3660 #endif
3661 	NDINIT_ATRIGHTS(&fromnd, DELETE, WANTPARENT | SAVESTART | AUDITVNODE1,
3662 	    pathseg, old, oldfd, &cap_renameat_source_rights, td);
3663 	if ((error = namei(&fromnd)) != 0)
3664 		return (error);
3665 #ifdef MAC
3666 	}
3667 #endif
3668 	fvp = fromnd.ni_vp;
3669 	tondflags = LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART | AUDITVNODE2;
3670 	if (fromnd.ni_vp->v_type == VDIR)
3671 		tondflags |= WILLBEDIR;
3672 	NDINIT_ATRIGHTS(&tond, RENAME, tondflags, pathseg, new, newfd,
3673 	    &cap_renameat_target_rights, td);
3674 	if ((error = namei(&tond)) != 0) {
3675 		/* Translate error code for rename("dir1", "dir2/."). */
3676 		if (error == EISDIR && fvp->v_type == VDIR)
3677 			error = EINVAL;
3678 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3679 		vrele(fromnd.ni_dvp);
3680 		vrele(fvp);
3681 		goto out1;
3682 	}
3683 	tdvp = tond.ni_dvp;
3684 	tvp = tond.ni_vp;
3685 	error = vn_start_write(fvp, &mp, V_NOWAIT);
3686 	if (error != 0) {
3687 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3688 		NDFREE(&tond, NDF_ONLY_PNBUF);
3689 		if (tvp != NULL)
3690 			vput(tvp);
3691 		if (tdvp == tvp)
3692 			vrele(tdvp);
3693 		else
3694 			vput(tdvp);
3695 		vrele(fromnd.ni_dvp);
3696 		vrele(fvp);
3697 		vrele(tond.ni_startdir);
3698 		if (fromnd.ni_startdir != NULL)
3699 			vrele(fromnd.ni_startdir);
3700 		error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH);
3701 		if (error != 0)
3702 			return (error);
3703 		goto again;
3704 	}
3705 	if (tvp != NULL) {
3706 		if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
3707 			error = ENOTDIR;
3708 			goto out;
3709 		} else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
3710 			error = EISDIR;
3711 			goto out;
3712 		}
3713 #ifdef CAPABILITIES
3714 		if (newfd != AT_FDCWD && (tond.ni_resflags & NIRES_ABS) == 0) {
3715 			/*
3716 			 * If the target already exists we require CAP_UNLINKAT
3717 			 * from 'newfd', when newfd was used for the lookup.
3718 			 */
3719 			error = cap_check(&tond.ni_filecaps.fc_rights,
3720 			    &cap_unlinkat_rights);
3721 			if (error != 0)
3722 				goto out;
3723 		}
3724 #endif
3725 	}
3726 	if (fvp == tdvp) {
3727 		error = EINVAL;
3728 		goto out;
3729 	}
3730 	/*
3731 	 * If the source is the same as the destination (that is, if they
3732 	 * are links to the same vnode), then there is nothing to do.
3733 	 */
3734 	if (fvp == tvp)
3735 		error = ERESTART;
3736 #ifdef MAC
3737 	else
3738 		error = mac_vnode_check_rename_to(td->td_ucred, tdvp,
3739 		    tond.ni_vp, fromnd.ni_dvp == tdvp, &tond.ni_cnd);
3740 #endif
3741 out:
3742 	if (error == 0) {
3743 		error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
3744 		    tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
3745 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3746 		NDFREE(&tond, NDF_ONLY_PNBUF);
3747 	} else {
3748 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3749 		NDFREE(&tond, NDF_ONLY_PNBUF);
3750 		if (tvp != NULL)
3751 			vput(tvp);
3752 		if (tdvp == tvp)
3753 			vrele(tdvp);
3754 		else
3755 			vput(tdvp);
3756 		vrele(fromnd.ni_dvp);
3757 		vrele(fvp);
3758 	}
3759 	vrele(tond.ni_startdir);
3760 	vn_finished_write(mp);
3761 out1:
3762 	if (fromnd.ni_startdir)
3763 		vrele(fromnd.ni_startdir);
3764 	if (error == ERESTART)
3765 		return (0);
3766 	if (error == ERELOOKUP)
3767 		goto again;
3768 	return (error);
3769 }
3770 
3771 /*
3772  * Make a directory file.
3773  */
3774 #ifndef _SYS_SYSPROTO_H_
3775 struct mkdir_args {
3776 	char	*path;
3777 	int	mode;
3778 };
3779 #endif
3780 int
3781 sys_mkdir(struct thread *td, struct mkdir_args *uap)
3782 {
3783 
3784 	return (kern_mkdirat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
3785 	    uap->mode));
3786 }
3787 
3788 #ifndef _SYS_SYSPROTO_H_
3789 struct mkdirat_args {
3790 	int	fd;
3791 	char	*path;
3792 	mode_t	mode;
3793 };
3794 #endif
3795 int
3796 sys_mkdirat(struct thread *td, struct mkdirat_args *uap)
3797 {
3798 
3799 	return (kern_mkdirat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode));
3800 }
3801 
3802 int
3803 kern_mkdirat(struct thread *td, int fd, const char *path, enum uio_seg segflg,
3804     int mode)
3805 {
3806 	struct mount *mp;
3807 	struct vattr vattr;
3808 	struct nameidata nd;
3809 	int error;
3810 
3811 	AUDIT_ARG_MODE(mode);
3812 	NDPREINIT(&nd);
3813 restart:
3814 	bwillwrite();
3815 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
3816 	    NC_NOMAKEENTRY | NC_KEEPPOSENTRY | FAILIFEXISTS | WILLBEDIR,
3817 	    segflg, path, fd, &cap_mkdirat_rights, td);
3818 	if ((error = namei(&nd)) != 0)
3819 		return (error);
3820 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3821 		NDFREE(&nd, NDF_ONLY_PNBUF);
3822 		vput(nd.ni_dvp);
3823 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3824 			return (error);
3825 		goto restart;
3826 	}
3827 	VATTR_NULL(&vattr);
3828 	vattr.va_type = VDIR;
3829 	vattr.va_mode = (mode & ACCESSPERMS) &~ td->td_proc->p_pd->pd_cmask;
3830 #ifdef MAC
3831 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
3832 	    &vattr);
3833 	if (error != 0)
3834 		goto out;
3835 #endif
3836 	error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
3837 #ifdef MAC
3838 out:
3839 #endif
3840 	NDFREE(&nd, NDF_ONLY_PNBUF);
3841 	VOP_VPUT_PAIR(nd.ni_dvp, error == 0 ? &nd.ni_vp : NULL, true);
3842 	vn_finished_write(mp);
3843 	if (error == ERELOOKUP)
3844 		goto restart;
3845 	return (error);
3846 }
3847 
3848 /*
3849  * Remove a directory file.
3850  */
3851 #ifndef _SYS_SYSPROTO_H_
3852 struct rmdir_args {
3853 	char	*path;
3854 };
3855 #endif
3856 int
3857 sys_rmdir(struct thread *td, struct rmdir_args *uap)
3858 {
3859 
3860 	return (kern_frmdirat(td, AT_FDCWD, uap->path, FD_NONE, UIO_USERSPACE,
3861 	    0));
3862 }
3863 
3864 int
3865 kern_frmdirat(struct thread *td, int dfd, const char *path, int fd,
3866     enum uio_seg pathseg, int flag)
3867 {
3868 	struct mount *mp;
3869 	struct vnode *vp;
3870 	struct file *fp;
3871 	struct nameidata nd;
3872 	cap_rights_t rights;
3873 	int error;
3874 
3875 	fp = NULL;
3876 	if (fd != FD_NONE) {
3877 		error = getvnode(td, fd, cap_rights_init_one(&rights,
3878 		    CAP_LOOKUP), &fp);
3879 		if (error != 0)
3880 			return (error);
3881 	}
3882 
3883 	NDPREINIT(&nd);
3884 restart:
3885 	bwillwrite();
3886 	NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1 |
3887 	    at2cnpflags(flag, AT_RESOLVE_BENEATH),
3888 	    pathseg, path, dfd, &cap_unlinkat_rights, td);
3889 	if ((error = namei(&nd)) != 0)
3890 		goto fdout;
3891 	vp = nd.ni_vp;
3892 	if (vp->v_type != VDIR) {
3893 		error = ENOTDIR;
3894 		goto out;
3895 	}
3896 	/*
3897 	 * No rmdir "." please.
3898 	 */
3899 	if (nd.ni_dvp == vp) {
3900 		error = EINVAL;
3901 		goto out;
3902 	}
3903 	/*
3904 	 * The root of a mounted filesystem cannot be deleted.
3905 	 */
3906 	if (vp->v_vflag & VV_ROOT) {
3907 		error = EBUSY;
3908 		goto out;
3909 	}
3910 
3911 	if (fp != NULL && fp->f_vnode != vp) {
3912 		if (VN_IS_DOOMED(fp->f_vnode))
3913 			error = EBADF;
3914 		else
3915 			error = EDEADLK;
3916 		goto out;
3917 	}
3918 
3919 #ifdef MAC
3920 	error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
3921 	    &nd.ni_cnd);
3922 	if (error != 0)
3923 		goto out;
3924 #endif
3925 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3926 		NDFREE(&nd, NDF_ONLY_PNBUF);
3927 		vput(vp);
3928 		if (nd.ni_dvp == vp)
3929 			vrele(nd.ni_dvp);
3930 		else
3931 			vput(nd.ni_dvp);
3932 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3933 			goto fdout;
3934 		goto restart;
3935 	}
3936 	vfs_notify_upper(vp, VFS_NOTIFY_UPPER_UNLINK);
3937 	error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
3938 	vn_finished_write(mp);
3939 out:
3940 	NDFREE(&nd, NDF_ONLY_PNBUF);
3941 	vput(vp);
3942 	if (nd.ni_dvp == vp)
3943 		vrele(nd.ni_dvp);
3944 	else
3945 		vput(nd.ni_dvp);
3946 	if (error == ERELOOKUP)
3947 		goto restart;
3948 fdout:
3949 	if (fp != NULL)
3950 		fdrop(fp, td);
3951 	return (error);
3952 }
3953 
3954 #if defined(COMPAT_43) || defined(COMPAT_FREEBSD11)
3955 int
3956 freebsd11_kern_getdirentries(struct thread *td, int fd, char *ubuf, u_int count,
3957     long *basep, void (*func)(struct freebsd11_dirent *))
3958 {
3959 	struct freebsd11_dirent dstdp;
3960 	struct dirent *dp, *edp;
3961 	char *dirbuf;
3962 	off_t base;
3963 	ssize_t resid, ucount;
3964 	int error;
3965 
3966 	/* XXX arbitrary sanity limit on `count'. */
3967 	count = min(count, 64 * 1024);
3968 
3969 	dirbuf = malloc(count, M_TEMP, M_WAITOK);
3970 
3971 	error = kern_getdirentries(td, fd, dirbuf, count, &base, &resid,
3972 	    UIO_SYSSPACE);
3973 	if (error != 0)
3974 		goto done;
3975 	if (basep != NULL)
3976 		*basep = base;
3977 
3978 	ucount = 0;
3979 	for (dp = (struct dirent *)dirbuf,
3980 	    edp = (struct dirent *)&dirbuf[count - resid];
3981 	    ucount < count && dp < edp; ) {
3982 		if (dp->d_reclen == 0)
3983 			break;
3984 		MPASS(dp->d_reclen >= _GENERIC_DIRLEN(0));
3985 		if (dp->d_namlen >= sizeof(dstdp.d_name))
3986 			continue;
3987 		dstdp.d_type = dp->d_type;
3988 		dstdp.d_namlen = dp->d_namlen;
3989 		dstdp.d_fileno = dp->d_fileno;		/* truncate */
3990 		if (dstdp.d_fileno != dp->d_fileno) {
3991 			switch (ino64_trunc_error) {
3992 			default:
3993 			case 0:
3994 				break;
3995 			case 1:
3996 				error = EOVERFLOW;
3997 				goto done;
3998 			case 2:
3999 				dstdp.d_fileno = UINT32_MAX;
4000 				break;
4001 			}
4002 		}
4003 		dstdp.d_reclen = sizeof(dstdp) - sizeof(dstdp.d_name) +
4004 		    ((dp->d_namlen + 1 + 3) &~ 3);
4005 		bcopy(dp->d_name, dstdp.d_name, dstdp.d_namlen);
4006 		bzero(dstdp.d_name + dstdp.d_namlen,
4007 		    dstdp.d_reclen - offsetof(struct freebsd11_dirent, d_name) -
4008 		    dstdp.d_namlen);
4009 		MPASS(dstdp.d_reclen <= dp->d_reclen);
4010 		MPASS(ucount + dstdp.d_reclen <= count);
4011 		if (func != NULL)
4012 			func(&dstdp);
4013 		error = copyout(&dstdp, ubuf + ucount, dstdp.d_reclen);
4014 		if (error != 0)
4015 			break;
4016 		dp = (struct dirent *)((char *)dp + dp->d_reclen);
4017 		ucount += dstdp.d_reclen;
4018 	}
4019 
4020 done:
4021 	free(dirbuf, M_TEMP);
4022 	if (error == 0)
4023 		td->td_retval[0] = ucount;
4024 	return (error);
4025 }
4026 #endif /* COMPAT */
4027 
4028 #ifdef COMPAT_43
4029 static void
4030 ogetdirentries_cvt(struct freebsd11_dirent *dp)
4031 {
4032 #if (BYTE_ORDER == LITTLE_ENDIAN)
4033 	/*
4034 	 * The expected low byte of dp->d_namlen is our dp->d_type.
4035 	 * The high MBZ byte of dp->d_namlen is our dp->d_namlen.
4036 	 */
4037 	dp->d_type = dp->d_namlen;
4038 	dp->d_namlen = 0;
4039 #else
4040 	/*
4041 	 * The dp->d_type is the high byte of the expected dp->d_namlen,
4042 	 * so must be zero'ed.
4043 	 */
4044 	dp->d_type = 0;
4045 #endif
4046 }
4047 
4048 /*
4049  * Read a block of directory entries in a filesystem independent format.
4050  */
4051 #ifndef _SYS_SYSPROTO_H_
4052 struct ogetdirentries_args {
4053 	int	fd;
4054 	char	*buf;
4055 	u_int	count;
4056 	long	*basep;
4057 };
4058 #endif
4059 int
4060 ogetdirentries(struct thread *td, struct ogetdirentries_args *uap)
4061 {
4062 	long loff;
4063 	int error;
4064 
4065 	error = kern_ogetdirentries(td, uap, &loff);
4066 	if (error == 0)
4067 		error = copyout(&loff, uap->basep, sizeof(long));
4068 	return (error);
4069 }
4070 
4071 int
4072 kern_ogetdirentries(struct thread *td, struct ogetdirentries_args *uap,
4073     long *ploff)
4074 {
4075 	long base;
4076 	int error;
4077 
4078 	/* XXX arbitrary sanity limit on `count'. */
4079 	if (uap->count > 64 * 1024)
4080 		return (EINVAL);
4081 
4082 	error = freebsd11_kern_getdirentries(td, uap->fd, uap->buf, uap->count,
4083 	    &base, ogetdirentries_cvt);
4084 
4085 	if (error == 0 && uap->basep != NULL)
4086 		error = copyout(&base, uap->basep, sizeof(long));
4087 
4088 	return (error);
4089 }
4090 #endif /* COMPAT_43 */
4091 
4092 #if defined(COMPAT_FREEBSD11)
4093 #ifndef _SYS_SYSPROTO_H_
4094 struct freebsd11_getdirentries_args {
4095 	int	fd;
4096 	char	*buf;
4097 	u_int	count;
4098 	long	*basep;
4099 };
4100 #endif
4101 int
4102 freebsd11_getdirentries(struct thread *td,
4103     struct freebsd11_getdirentries_args *uap)
4104 {
4105 	long base;
4106 	int error;
4107 
4108 	error = freebsd11_kern_getdirentries(td, uap->fd, uap->buf, uap->count,
4109 	    &base, NULL);
4110 
4111 	if (error == 0 && uap->basep != NULL)
4112 		error = copyout(&base, uap->basep, sizeof(long));
4113 	return (error);
4114 }
4115 
4116 int
4117 freebsd11_getdents(struct thread *td, struct freebsd11_getdents_args *uap)
4118 {
4119 	struct freebsd11_getdirentries_args ap;
4120 
4121 	ap.fd = uap->fd;
4122 	ap.buf = uap->buf;
4123 	ap.count = uap->count;
4124 	ap.basep = NULL;
4125 	return (freebsd11_getdirentries(td, &ap));
4126 }
4127 #endif /* COMPAT_FREEBSD11 */
4128 
4129 /*
4130  * Read a block of directory entries in a filesystem independent format.
4131  */
4132 int
4133 sys_getdirentries(struct thread *td, struct getdirentries_args *uap)
4134 {
4135 	off_t base;
4136 	int error;
4137 
4138 	error = kern_getdirentries(td, uap->fd, uap->buf, uap->count, &base,
4139 	    NULL, UIO_USERSPACE);
4140 	if (error != 0)
4141 		return (error);
4142 	if (uap->basep != NULL)
4143 		error = copyout(&base, uap->basep, sizeof(off_t));
4144 	return (error);
4145 }
4146 
4147 int
4148 kern_getdirentries(struct thread *td, int fd, char *buf, size_t count,
4149     off_t *basep, ssize_t *residp, enum uio_seg bufseg)
4150 {
4151 	struct vnode *vp;
4152 	struct file *fp;
4153 	struct uio auio;
4154 	struct iovec aiov;
4155 	off_t loff;
4156 	int error, eofflag;
4157 	off_t foffset;
4158 
4159 	AUDIT_ARG_FD(fd);
4160 	if (count > IOSIZE_MAX)
4161 		return (EINVAL);
4162 	auio.uio_resid = count;
4163 	error = getvnode(td, fd, &cap_read_rights, &fp);
4164 	if (error != 0)
4165 		return (error);
4166 	if ((fp->f_flag & FREAD) == 0) {
4167 		fdrop(fp, td);
4168 		return (EBADF);
4169 	}
4170 	vp = fp->f_vnode;
4171 	foffset = foffset_lock(fp, 0);
4172 unionread:
4173 	if (vp->v_type != VDIR) {
4174 		error = EINVAL;
4175 		goto fail;
4176 	}
4177 	aiov.iov_base = buf;
4178 	aiov.iov_len = count;
4179 	auio.uio_iov = &aiov;
4180 	auio.uio_iovcnt = 1;
4181 	auio.uio_rw = UIO_READ;
4182 	auio.uio_segflg = bufseg;
4183 	auio.uio_td = td;
4184 	vn_lock(vp, LK_SHARED | LK_RETRY);
4185 	AUDIT_ARG_VNODE1(vp);
4186 	loff = auio.uio_offset = foffset;
4187 #ifdef MAC
4188 	error = mac_vnode_check_readdir(td->td_ucred, vp);
4189 	if (error == 0)
4190 #endif
4191 		error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, NULL,
4192 		    NULL);
4193 	foffset = auio.uio_offset;
4194 	if (error != 0) {
4195 		VOP_UNLOCK(vp);
4196 		goto fail;
4197 	}
4198 	if (count == auio.uio_resid &&
4199 	    (vp->v_vflag & VV_ROOT) &&
4200 	    (vp->v_mount->mnt_flag & MNT_UNION)) {
4201 		struct vnode *tvp = vp;
4202 
4203 		vp = vp->v_mount->mnt_vnodecovered;
4204 		VREF(vp);
4205 		fp->f_vnode = vp;
4206 		foffset = 0;
4207 		vput(tvp);
4208 		goto unionread;
4209 	}
4210 	VOP_UNLOCK(vp);
4211 	*basep = loff;
4212 	if (residp != NULL)
4213 		*residp = auio.uio_resid;
4214 	td->td_retval[0] = count - auio.uio_resid;
4215 fail:
4216 	foffset_unlock(fp, foffset, 0);
4217 	fdrop(fp, td);
4218 	return (error);
4219 }
4220 
4221 /*
4222  * Set the mode mask for creation of filesystem nodes.
4223  */
4224 #ifndef _SYS_SYSPROTO_H_
4225 struct umask_args {
4226 	int	newmask;
4227 };
4228 #endif
4229 int
4230 sys_umask(struct thread *td, struct umask_args *uap)
4231 {
4232 	struct pwddesc *pdp;
4233 
4234 	pdp = td->td_proc->p_pd;
4235 	PWDDESC_XLOCK(pdp);
4236 	td->td_retval[0] = pdp->pd_cmask;
4237 	pdp->pd_cmask = uap->newmask & ALLPERMS;
4238 	PWDDESC_XUNLOCK(pdp);
4239 	return (0);
4240 }
4241 
4242 /*
4243  * Void all references to file by ripping underlying filesystem away from
4244  * vnode.
4245  */
4246 #ifndef _SYS_SYSPROTO_H_
4247 struct revoke_args {
4248 	char	*path;
4249 };
4250 #endif
4251 int
4252 sys_revoke(struct thread *td, struct revoke_args *uap)
4253 {
4254 	struct vnode *vp;
4255 	struct vattr vattr;
4256 	struct nameidata nd;
4257 	int error;
4258 
4259 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
4260 	    uap->path, td);
4261 	if ((error = namei(&nd)) != 0)
4262 		return (error);
4263 	vp = nd.ni_vp;
4264 	NDFREE_NOTHING(&nd);
4265 	if (vp->v_type != VCHR || vp->v_rdev == NULL) {
4266 		error = EINVAL;
4267 		goto out;
4268 	}
4269 #ifdef MAC
4270 	error = mac_vnode_check_revoke(td->td_ucred, vp);
4271 	if (error != 0)
4272 		goto out;
4273 #endif
4274 	error = VOP_GETATTR(vp, &vattr, td->td_ucred);
4275 	if (error != 0)
4276 		goto out;
4277 	if (td->td_ucred->cr_uid != vattr.va_uid) {
4278 		error = priv_check(td, PRIV_VFS_ADMIN);
4279 		if (error != 0)
4280 			goto out;
4281 	}
4282 	if (devfs_usecount(vp) > 0)
4283 		VOP_REVOKE(vp, REVOKEALL);
4284 out:
4285 	vput(vp);
4286 	return (error);
4287 }
4288 
4289 /*
4290  * This variant of getvnode() allows O_PATH files.  Caller should
4291  * ensure that returned file and vnode are only used for compatible
4292  * semantics.
4293  */
4294 int
4295 getvnode_path(struct thread *td, int fd, cap_rights_t *rightsp,
4296     struct file **fpp)
4297 {
4298 	struct file *fp;
4299 	int error;
4300 
4301 	error = fget_unlocked(td->td_proc->p_fd, fd, rightsp, &fp);
4302 	if (error != 0)
4303 		return (error);
4304 
4305 	/*
4306 	 * The file could be not of the vnode type, or it may be not
4307 	 * yet fully initialized, in which case the f_vnode pointer
4308 	 * may be set, but f_ops is still badfileops.  E.g.,
4309 	 * devfs_open() transiently create such situation to
4310 	 * facilitate csw d_fdopen().
4311 	 *
4312 	 * Dupfdopen() handling in kern_openat() installs the
4313 	 * half-baked file into the process descriptor table, allowing
4314 	 * other thread to dereference it. Guard against the race by
4315 	 * checking f_ops.
4316 	 */
4317 	if (__predict_false(fp->f_vnode == NULL || fp->f_ops == &badfileops)) {
4318 		fdrop(fp, td);
4319 		return (EINVAL);
4320 	}
4321 
4322 	*fpp = fp;
4323 	return (0);
4324 }
4325 
4326 /*
4327  * Convert a user file descriptor to a kernel file entry and check
4328  * that, if it is a capability, the correct rights are present.
4329  * A reference on the file entry is held upon returning.
4330  */
4331 int
4332 getvnode(struct thread *td, int fd, cap_rights_t *rightsp, struct file **fpp)
4333 {
4334 	int error;
4335 
4336 	error = getvnode_path(td, fd, rightsp, fpp);
4337 	if (__predict_false(error != 0))
4338 		return (error);
4339 
4340 	/*
4341 	 * Filter out O_PATH file descriptors, most getvnode() callers
4342 	 * do not call fo_ methods.
4343 	 */
4344 	if (__predict_false((*fpp)->f_ops == &path_fileops)) {
4345 		fdrop(*fpp, td);
4346 		error = EBADF;
4347 	}
4348 
4349 	return (error);
4350 }
4351 
4352 /*
4353  * Get an (NFS) file handle.
4354  */
4355 #ifndef _SYS_SYSPROTO_H_
4356 struct lgetfh_args {
4357 	char *fname;
4358 	fhandle_t *fhp;
4359 };
4360 #endif
4361 int
4362 sys_lgetfh(struct thread *td, struct lgetfh_args *uap)
4363 {
4364 
4365 	return (kern_getfhat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->fname,
4366 	    UIO_USERSPACE, uap->fhp, UIO_USERSPACE));
4367 }
4368 
4369 #ifndef _SYS_SYSPROTO_H_
4370 struct getfh_args {
4371 	char *fname;
4372 	fhandle_t *fhp;
4373 };
4374 #endif
4375 int
4376 sys_getfh(struct thread *td, struct getfh_args *uap)
4377 {
4378 
4379 	return (kern_getfhat(td, 0, AT_FDCWD, uap->fname, UIO_USERSPACE,
4380 	    uap->fhp, UIO_USERSPACE));
4381 }
4382 
4383 /*
4384  * syscall for the rpc.lockd to use to translate an open descriptor into
4385  * a NFS file handle.
4386  *
4387  * warning: do not remove the priv_check() call or this becomes one giant
4388  * security hole.
4389  */
4390 #ifndef _SYS_SYSPROTO_H_
4391 struct getfhat_args {
4392 	int fd;
4393 	char *path;
4394 	fhandle_t *fhp;
4395 	int flags;
4396 };
4397 #endif
4398 int
4399 sys_getfhat(struct thread *td, struct getfhat_args *uap)
4400 {
4401 
4402 	return (kern_getfhat(td, uap->flags, uap->fd, uap->path, UIO_USERSPACE,
4403 	    uap->fhp, UIO_USERSPACE));
4404 }
4405 
4406 int
4407 kern_getfhat(struct thread *td, int flags, int fd, const char *path,
4408     enum uio_seg pathseg, fhandle_t *fhp, enum uio_seg fhseg)
4409 {
4410 	struct nameidata nd;
4411 	fhandle_t fh;
4412 	struct vnode *vp;
4413 	int error;
4414 
4415 	if ((flags & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH)) != 0)
4416 		return (EINVAL);
4417 	error = priv_check(td, PRIV_VFS_GETFH);
4418 	if (error != 0)
4419 		return (error);
4420 	NDINIT_AT(&nd, LOOKUP, at2cnpflags(flags, AT_SYMLINK_NOFOLLOW |
4421 	    AT_RESOLVE_BENEATH) | LOCKLEAF | AUDITVNODE1, pathseg, path,
4422 	    fd, td);
4423 	error = namei(&nd);
4424 	if (error != 0)
4425 		return (error);
4426 	NDFREE_NOTHING(&nd);
4427 	vp = nd.ni_vp;
4428 	bzero(&fh, sizeof(fh));
4429 	fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
4430 	error = VOP_VPTOFH(vp, &fh.fh_fid);
4431 	vput(vp);
4432 	if (error == 0) {
4433 		if (fhseg == UIO_USERSPACE)
4434 			error = copyout(&fh, fhp, sizeof (fh));
4435 		else
4436 			memcpy(fhp, &fh, sizeof(fh));
4437 	}
4438 	return (error);
4439 }
4440 
4441 #ifndef _SYS_SYSPROTO_H_
4442 struct fhlink_args {
4443 	fhandle_t *fhp;
4444 	const char *to;
4445 };
4446 #endif
4447 int
4448 sys_fhlink(struct thread *td, struct fhlink_args *uap)
4449 {
4450 
4451 	return (kern_fhlinkat(td, AT_FDCWD, uap->to, UIO_USERSPACE, uap->fhp));
4452 }
4453 
4454 #ifndef _SYS_SYSPROTO_H_
4455 struct fhlinkat_args {
4456 	fhandle_t *fhp;
4457 	int tofd;
4458 	const char *to;
4459 };
4460 #endif
4461 int
4462 sys_fhlinkat(struct thread *td, struct fhlinkat_args *uap)
4463 {
4464 
4465 	return (kern_fhlinkat(td, uap->tofd, uap->to, UIO_USERSPACE, uap->fhp));
4466 }
4467 
4468 static int
4469 kern_fhlinkat(struct thread *td, int fd, const char *path,
4470     enum uio_seg pathseg, fhandle_t *fhp)
4471 {
4472 	fhandle_t fh;
4473 	struct mount *mp;
4474 	struct vnode *vp;
4475 	int error;
4476 
4477 	error = priv_check(td, PRIV_VFS_GETFH);
4478 	if (error != 0)
4479 		return (error);
4480 	error = copyin(fhp, &fh, sizeof(fh));
4481 	if (error != 0)
4482 		return (error);
4483 	do {
4484 		bwillwrite();
4485 		if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4486 			return (ESTALE);
4487 		error = VFS_FHTOVP(mp, &fh.fh_fid, LK_SHARED, &vp);
4488 		vfs_unbusy(mp);
4489 		if (error != 0)
4490 			return (error);
4491 		VOP_UNLOCK(vp);
4492 		error = kern_linkat_vp(td, vp, fd, path, pathseg);
4493 	} while (error == EAGAIN || error == ERELOOKUP);
4494 	return (error);
4495 }
4496 
4497 #ifndef _SYS_SYSPROTO_H_
4498 struct fhreadlink_args {
4499 	fhandle_t *fhp;
4500 	char *buf;
4501 	size_t bufsize;
4502 };
4503 #endif
4504 int
4505 sys_fhreadlink(struct thread *td, struct fhreadlink_args *uap)
4506 {
4507 	fhandle_t fh;
4508 	struct mount *mp;
4509 	struct vnode *vp;
4510 	int error;
4511 
4512 	error = priv_check(td, PRIV_VFS_GETFH);
4513 	if (error != 0)
4514 		return (error);
4515 	if (uap->bufsize > IOSIZE_MAX)
4516 		return (EINVAL);
4517 	error = copyin(uap->fhp, &fh, sizeof(fh));
4518 	if (error != 0)
4519 		return (error);
4520 	if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4521 		return (ESTALE);
4522 	error = VFS_FHTOVP(mp, &fh.fh_fid, LK_SHARED, &vp);
4523 	vfs_unbusy(mp);
4524 	if (error != 0)
4525 		return (error);
4526 	error = kern_readlink_vp(vp, uap->buf, UIO_USERSPACE, uap->bufsize, td);
4527 	vput(vp);
4528 	return (error);
4529 }
4530 
4531 /*
4532  * syscall for the rpc.lockd to use to translate a NFS file handle into an
4533  * open descriptor.
4534  *
4535  * warning: do not remove the priv_check() call or this becomes one giant
4536  * security hole.
4537  */
4538 #ifndef _SYS_SYSPROTO_H_
4539 struct fhopen_args {
4540 	const struct fhandle *u_fhp;
4541 	int flags;
4542 };
4543 #endif
4544 int
4545 sys_fhopen(struct thread *td, struct fhopen_args *uap)
4546 {
4547 	return (kern_fhopen(td, uap->u_fhp, uap->flags));
4548 }
4549 
4550 int
4551 kern_fhopen(struct thread *td, const struct fhandle *u_fhp, int flags)
4552 {
4553 	struct mount *mp;
4554 	struct vnode *vp;
4555 	struct fhandle fhp;
4556 	struct file *fp;
4557 	int fmode, error;
4558 	int indx;
4559 
4560 	error = priv_check(td, PRIV_VFS_FHOPEN);
4561 	if (error != 0)
4562 		return (error);
4563 	indx = -1;
4564 	fmode = FFLAGS(flags);
4565 	/* why not allow a non-read/write open for our lockd? */
4566 	if (((fmode & (FREAD | FWRITE)) == 0) || (fmode & O_CREAT))
4567 		return (EINVAL);
4568 	error = copyin(u_fhp, &fhp, sizeof(fhp));
4569 	if (error != 0)
4570 		return(error);
4571 	/* find the mount point */
4572 	mp = vfs_busyfs(&fhp.fh_fsid);
4573 	if (mp == NULL)
4574 		return (ESTALE);
4575 	/* now give me my vnode, it gets returned to me locked */
4576 	error = VFS_FHTOVP(mp, &fhp.fh_fid, LK_EXCLUSIVE, &vp);
4577 	vfs_unbusy(mp);
4578 	if (error != 0)
4579 		return (error);
4580 
4581 	error = falloc_noinstall(td, &fp);
4582 	if (error != 0) {
4583 		vput(vp);
4584 		return (error);
4585 	}
4586 	/*
4587 	 * An extra reference on `fp' has been held for us by
4588 	 * falloc_noinstall().
4589 	 */
4590 
4591 #ifdef INVARIANTS
4592 	td->td_dupfd = -1;
4593 #endif
4594 	error = vn_open_vnode(vp, fmode, td->td_ucred, td, fp);
4595 	if (error != 0) {
4596 		KASSERT(fp->f_ops == &badfileops,
4597 		    ("VOP_OPEN in fhopen() set f_ops"));
4598 		KASSERT(td->td_dupfd < 0,
4599 		    ("fhopen() encountered fdopen()"));
4600 
4601 		vput(vp);
4602 		goto bad;
4603 	}
4604 #ifdef INVARIANTS
4605 	td->td_dupfd = 0;
4606 #endif
4607 	fp->f_vnode = vp;
4608 	finit_vnode(fp, fmode, NULL, &vnops);
4609 	VOP_UNLOCK(vp);
4610 	if ((fmode & O_TRUNC) != 0) {
4611 		error = fo_truncate(fp, 0, td->td_ucred, td);
4612 		if (error != 0)
4613 			goto bad;
4614 	}
4615 
4616 	error = finstall(td, fp, &indx, fmode, NULL);
4617 bad:
4618 	fdrop(fp, td);
4619 	td->td_retval[0] = indx;
4620 	return (error);
4621 }
4622 
4623 /*
4624  * Stat an (NFS) file handle.
4625  */
4626 #ifndef _SYS_SYSPROTO_H_
4627 struct fhstat_args {
4628 	struct fhandle *u_fhp;
4629 	struct stat *sb;
4630 };
4631 #endif
4632 int
4633 sys_fhstat(struct thread *td, struct fhstat_args *uap)
4634 {
4635 	struct stat sb;
4636 	struct fhandle fh;
4637 	int error;
4638 
4639 	error = copyin(uap->u_fhp, &fh, sizeof(fh));
4640 	if (error != 0)
4641 		return (error);
4642 	error = kern_fhstat(td, fh, &sb);
4643 	if (error == 0)
4644 		error = copyout(&sb, uap->sb, sizeof(sb));
4645 	return (error);
4646 }
4647 
4648 int
4649 kern_fhstat(struct thread *td, struct fhandle fh, struct stat *sb)
4650 {
4651 	struct mount *mp;
4652 	struct vnode *vp;
4653 	int error;
4654 
4655 	error = priv_check(td, PRIV_VFS_FHSTAT);
4656 	if (error != 0)
4657 		return (error);
4658 	if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4659 		return (ESTALE);
4660 	error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4661 	vfs_unbusy(mp);
4662 	if (error != 0)
4663 		return (error);
4664 	error = VOP_STAT(vp, sb, td->td_ucred, NOCRED);
4665 	vput(vp);
4666 	return (error);
4667 }
4668 
4669 /*
4670  * Implement fstatfs() for (NFS) file handles.
4671  */
4672 #ifndef _SYS_SYSPROTO_H_
4673 struct fhstatfs_args {
4674 	struct fhandle *u_fhp;
4675 	struct statfs *buf;
4676 };
4677 #endif
4678 int
4679 sys_fhstatfs(struct thread *td, struct fhstatfs_args *uap)
4680 {
4681 	struct statfs *sfp;
4682 	fhandle_t fh;
4683 	int error;
4684 
4685 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
4686 	if (error != 0)
4687 		return (error);
4688 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
4689 	error = kern_fhstatfs(td, fh, sfp);
4690 	if (error == 0)
4691 		error = copyout(sfp, uap->buf, sizeof(*sfp));
4692 	free(sfp, M_STATFS);
4693 	return (error);
4694 }
4695 
4696 int
4697 kern_fhstatfs(struct thread *td, fhandle_t fh, struct statfs *buf)
4698 {
4699 	struct mount *mp;
4700 	struct vnode *vp;
4701 	int error;
4702 
4703 	error = priv_check(td, PRIV_VFS_FHSTATFS);
4704 	if (error != 0)
4705 		return (error);
4706 	if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4707 		return (ESTALE);
4708 	error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4709 	if (error != 0) {
4710 		vfs_unbusy(mp);
4711 		return (error);
4712 	}
4713 	vput(vp);
4714 	error = prison_canseemount(td->td_ucred, mp);
4715 	if (error != 0)
4716 		goto out;
4717 #ifdef MAC
4718 	error = mac_mount_check_stat(td->td_ucred, mp);
4719 	if (error != 0)
4720 		goto out;
4721 #endif
4722 	error = VFS_STATFS(mp, buf);
4723 out:
4724 	vfs_unbusy(mp);
4725 	return (error);
4726 }
4727 
4728 /*
4729  * Unlike madvise(2), we do not make a best effort to remember every
4730  * possible caching hint.  Instead, we remember the last setting with
4731  * the exception that we will allow POSIX_FADV_NORMAL to adjust the
4732  * region of any current setting.
4733  */
4734 int
4735 kern_posix_fadvise(struct thread *td, int fd, off_t offset, off_t len,
4736     int advice)
4737 {
4738 	struct fadvise_info *fa, *new;
4739 	struct file *fp;
4740 	struct vnode *vp;
4741 	off_t end;
4742 	int error;
4743 
4744 	if (offset < 0 || len < 0 || offset > OFF_MAX - len)
4745 		return (EINVAL);
4746 	AUDIT_ARG_VALUE(advice);
4747 	switch (advice) {
4748 	case POSIX_FADV_SEQUENTIAL:
4749 	case POSIX_FADV_RANDOM:
4750 	case POSIX_FADV_NOREUSE:
4751 		new = malloc(sizeof(*fa), M_FADVISE, M_WAITOK);
4752 		break;
4753 	case POSIX_FADV_NORMAL:
4754 	case POSIX_FADV_WILLNEED:
4755 	case POSIX_FADV_DONTNEED:
4756 		new = NULL;
4757 		break;
4758 	default:
4759 		return (EINVAL);
4760 	}
4761 	/* XXX: CAP_POSIX_FADVISE? */
4762 	AUDIT_ARG_FD(fd);
4763 	error = fget(td, fd, &cap_no_rights, &fp);
4764 	if (error != 0)
4765 		goto out;
4766 	AUDIT_ARG_FILE(td->td_proc, fp);
4767 	if ((fp->f_ops->fo_flags & DFLAG_SEEKABLE) == 0) {
4768 		error = ESPIPE;
4769 		goto out;
4770 	}
4771 	if (fp->f_type != DTYPE_VNODE) {
4772 		error = ENODEV;
4773 		goto out;
4774 	}
4775 	vp = fp->f_vnode;
4776 	if (vp->v_type != VREG) {
4777 		error = ENODEV;
4778 		goto out;
4779 	}
4780 	if (len == 0)
4781 		end = OFF_MAX;
4782 	else
4783 		end = offset + len - 1;
4784 	switch (advice) {
4785 	case POSIX_FADV_SEQUENTIAL:
4786 	case POSIX_FADV_RANDOM:
4787 	case POSIX_FADV_NOREUSE:
4788 		/*
4789 		 * Try to merge any existing non-standard region with
4790 		 * this new region if possible, otherwise create a new
4791 		 * non-standard region for this request.
4792 		 */
4793 		mtx_pool_lock(mtxpool_sleep, fp);
4794 		fa = fp->f_advice;
4795 		if (fa != NULL && fa->fa_advice == advice &&
4796 		    ((fa->fa_start <= end && fa->fa_end >= offset) ||
4797 		    (end != OFF_MAX && fa->fa_start == end + 1) ||
4798 		    (fa->fa_end != OFF_MAX && fa->fa_end + 1 == offset))) {
4799 			if (offset < fa->fa_start)
4800 				fa->fa_start = offset;
4801 			if (end > fa->fa_end)
4802 				fa->fa_end = end;
4803 		} else {
4804 			new->fa_advice = advice;
4805 			new->fa_start = offset;
4806 			new->fa_end = end;
4807 			fp->f_advice = new;
4808 			new = fa;
4809 		}
4810 		mtx_pool_unlock(mtxpool_sleep, fp);
4811 		break;
4812 	case POSIX_FADV_NORMAL:
4813 		/*
4814 		 * If a the "normal" region overlaps with an existing
4815 		 * non-standard region, trim or remove the
4816 		 * non-standard region.
4817 		 */
4818 		mtx_pool_lock(mtxpool_sleep, fp);
4819 		fa = fp->f_advice;
4820 		if (fa != NULL) {
4821 			if (offset <= fa->fa_start && end >= fa->fa_end) {
4822 				new = fa;
4823 				fp->f_advice = NULL;
4824 			} else if (offset <= fa->fa_start &&
4825 			    end >= fa->fa_start)
4826 				fa->fa_start = end + 1;
4827 			else if (offset <= fa->fa_end && end >= fa->fa_end)
4828 				fa->fa_end = offset - 1;
4829 			else if (offset >= fa->fa_start && end <= fa->fa_end) {
4830 				/*
4831 				 * If the "normal" region is a middle
4832 				 * portion of the existing
4833 				 * non-standard region, just remove
4834 				 * the whole thing rather than picking
4835 				 * one side or the other to
4836 				 * preserve.
4837 				 */
4838 				new = fa;
4839 				fp->f_advice = NULL;
4840 			}
4841 		}
4842 		mtx_pool_unlock(mtxpool_sleep, fp);
4843 		break;
4844 	case POSIX_FADV_WILLNEED:
4845 	case POSIX_FADV_DONTNEED:
4846 		error = VOP_ADVISE(vp, offset, end, advice);
4847 		break;
4848 	}
4849 out:
4850 	if (fp != NULL)
4851 		fdrop(fp, td);
4852 	free(new, M_FADVISE);
4853 	return (error);
4854 }
4855 
4856 int
4857 sys_posix_fadvise(struct thread *td, struct posix_fadvise_args *uap)
4858 {
4859 	int error;
4860 
4861 	error = kern_posix_fadvise(td, uap->fd, uap->offset, uap->len,
4862 	    uap->advice);
4863 	return (kern_posix_error(td, error));
4864 }
4865 
4866 int
4867 kern_copy_file_range(struct thread *td, int infd, off_t *inoffp, int outfd,
4868     off_t *outoffp, size_t len, unsigned int flags)
4869 {
4870 	struct file *infp, *outfp;
4871 	struct vnode *invp, *outvp;
4872 	int error;
4873 	size_t retlen;
4874 	void *rl_rcookie, *rl_wcookie;
4875 	off_t savinoff, savoutoff;
4876 
4877 	infp = outfp = NULL;
4878 	rl_rcookie = rl_wcookie = NULL;
4879 	savinoff = -1;
4880 	error = 0;
4881 	retlen = 0;
4882 
4883 	if (flags != 0) {
4884 		error = EINVAL;
4885 		goto out;
4886 	}
4887 	if (len > SSIZE_MAX)
4888 		/*
4889 		 * Although the len argument is size_t, the return argument
4890 		 * is ssize_t (which is signed).  Therefore a size that won't
4891 		 * fit in ssize_t can't be returned.
4892 		 */
4893 		len = SSIZE_MAX;
4894 
4895 	/* Get the file structures for the file descriptors. */
4896 	error = fget_read(td, infd, &cap_read_rights, &infp);
4897 	if (error != 0)
4898 		goto out;
4899 	if (infp->f_ops == &badfileops) {
4900 		error = EBADF;
4901 		goto out;
4902 	}
4903 	if (infp->f_vnode == NULL) {
4904 		error = EINVAL;
4905 		goto out;
4906 	}
4907 	error = fget_write(td, outfd, &cap_write_rights, &outfp);
4908 	if (error != 0)
4909 		goto out;
4910 	if (outfp->f_ops == &badfileops) {
4911 		error = EBADF;
4912 		goto out;
4913 	}
4914 	if (outfp->f_vnode == NULL) {
4915 		error = EINVAL;
4916 		goto out;
4917 	}
4918 
4919 	/* Set the offset pointers to the correct place. */
4920 	if (inoffp == NULL)
4921 		inoffp = &infp->f_offset;
4922 	if (outoffp == NULL)
4923 		outoffp = &outfp->f_offset;
4924 	savinoff = *inoffp;
4925 	savoutoff = *outoffp;
4926 
4927 	invp = infp->f_vnode;
4928 	outvp = outfp->f_vnode;
4929 	/* Sanity check the f_flag bits. */
4930 	if ((outfp->f_flag & (FWRITE | FAPPEND)) != FWRITE ||
4931 	    (infp->f_flag & FREAD) == 0) {
4932 		error = EBADF;
4933 		goto out;
4934 	}
4935 
4936 	/* If len == 0, just return 0. */
4937 	if (len == 0)
4938 		goto out;
4939 
4940 	/*
4941 	 * If infp and outfp refer to the same file, the byte ranges cannot
4942 	 * overlap.
4943 	 */
4944 	if (invp == outvp && ((savinoff <= savoutoff && savinoff + len >
4945 	    savoutoff) || (savinoff > savoutoff && savoutoff + len >
4946 	    savinoff))) {
4947 		error = EINVAL;
4948 		goto out;
4949 	}
4950 
4951 	/* Range lock the byte ranges for both invp and outvp. */
4952 	for (;;) {
4953 		rl_wcookie = vn_rangelock_wlock(outvp, *outoffp, *outoffp +
4954 		    len);
4955 		rl_rcookie = vn_rangelock_tryrlock(invp, *inoffp, *inoffp +
4956 		    len);
4957 		if (rl_rcookie != NULL)
4958 			break;
4959 		vn_rangelock_unlock(outvp, rl_wcookie);
4960 		rl_rcookie = vn_rangelock_rlock(invp, *inoffp, *inoffp + len);
4961 		vn_rangelock_unlock(invp, rl_rcookie);
4962 	}
4963 
4964 	retlen = len;
4965 	error = vn_copy_file_range(invp, inoffp, outvp, outoffp, &retlen,
4966 	    flags, infp->f_cred, outfp->f_cred, td);
4967 out:
4968 	if (rl_rcookie != NULL)
4969 		vn_rangelock_unlock(invp, rl_rcookie);
4970 	if (rl_wcookie != NULL)
4971 		vn_rangelock_unlock(outvp, rl_wcookie);
4972 	if (savinoff != -1 && (error == EINTR || error == ERESTART)) {
4973 		*inoffp = savinoff;
4974 		*outoffp = savoutoff;
4975 	}
4976 	if (outfp != NULL)
4977 		fdrop(outfp, td);
4978 	if (infp != NULL)
4979 		fdrop(infp, td);
4980 	td->td_retval[0] = retlen;
4981 	return (error);
4982 }
4983 
4984 int
4985 sys_copy_file_range(struct thread *td, struct copy_file_range_args *uap)
4986 {
4987 	off_t inoff, outoff, *inoffp, *outoffp;
4988 	int error;
4989 
4990 	inoffp = outoffp = NULL;
4991 	if (uap->inoffp != NULL) {
4992 		error = copyin(uap->inoffp, &inoff, sizeof(off_t));
4993 		if (error != 0)
4994 			return (error);
4995 		inoffp = &inoff;
4996 	}
4997 	if (uap->outoffp != NULL) {
4998 		error = copyin(uap->outoffp, &outoff, sizeof(off_t));
4999 		if (error != 0)
5000 			return (error);
5001 		outoffp = &outoff;
5002 	}
5003 	error = kern_copy_file_range(td, uap->infd, inoffp, uap->outfd,
5004 	    outoffp, uap->len, uap->flags);
5005 	if (error == 0 && uap->inoffp != NULL)
5006 		error = copyout(inoffp, uap->inoffp, sizeof(off_t));
5007 	if (error == 0 && uap->outoffp != NULL)
5008 		error = copyout(outoffp, uap->outoffp, sizeof(off_t));
5009 	return (error);
5010 }
5011