xref: /freebsd/sys/kern/vfs_syscalls.c (revision 52d973f52c07b94909a6487be373c269988dc151)
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 #ifdef COMPAT_FREEBSD11
48 #include <sys/abi_compat.h>
49 #endif
50 #include <sys/bio.h>
51 #include <sys/buf.h>
52 #include <sys/capsicum.h>
53 #include <sys/disk.h>
54 #include <sys/sysent.h>
55 #include <sys/malloc.h>
56 #include <sys/mount.h>
57 #include <sys/mutex.h>
58 #include <sys/sysproto.h>
59 #include <sys/namei.h>
60 #include <sys/filedesc.h>
61 #include <sys/kernel.h>
62 #include <sys/fcntl.h>
63 #include <sys/file.h>
64 #include <sys/filio.h>
65 #include <sys/limits.h>
66 #include <sys/linker.h>
67 #include <sys/rwlock.h>
68 #include <sys/sdt.h>
69 #include <sys/stat.h>
70 #include <sys/sx.h>
71 #include <sys/unistd.h>
72 #include <sys/vnode.h>
73 #include <sys/priv.h>
74 #include <sys/proc.h>
75 #include <sys/dirent.h>
76 #include <sys/jail.h>
77 #include <sys/syscallsubr.h>
78 #include <sys/sysctl.h>
79 #ifdef KTRACE
80 #include <sys/ktrace.h>
81 #endif
82 
83 #include <machine/stdarg.h>
84 
85 #include <security/audit/audit.h>
86 #include <security/mac/mac_framework.h>
87 
88 #include <vm/vm.h>
89 #include <vm/vm_object.h>
90 #include <vm/vm_page.h>
91 #include <vm/uma.h>
92 
93 #include <fs/devfs/devfs.h>
94 
95 MALLOC_DEFINE(M_FADVISE, "fadvise", "posix_fadvise(2) information");
96 
97 static int kern_chflagsat(struct thread *td, int fd, const char *path,
98     enum uio_seg pathseg, u_long flags, int atflag);
99 static int setfflags(struct thread *td, struct vnode *, u_long);
100 static int getutimes(const struct timeval *, enum uio_seg, struct timespec *);
101 static int getutimens(const struct timespec *, enum uio_seg,
102     struct timespec *, int *);
103 static int setutimes(struct thread *td, struct vnode *,
104     const struct timespec *, int, int);
105 static int vn_access(struct vnode *vp, int user_flags, struct ucred *cred,
106     struct thread *td);
107 static int kern_fhlinkat(struct thread *td, int fd, const char *path,
108     enum uio_seg pathseg, fhandle_t *fhp);
109 static int kern_readlink_vp(struct vnode *vp, char *buf, enum uio_seg bufseg,
110     size_t count, struct thread *td);
111 static int kern_linkat_vp(struct thread *td, struct vnode *vp, int fd,
112     const char *path, enum uio_seg segflag);
113 
114 static uint64_t
115 at2cnpflags(u_int at_flags, u_int mask)
116 {
117 	u_int64_t res;
118 
119 	MPASS((at_flags & (AT_SYMLINK_FOLLOW | AT_SYMLINK_NOFOLLOW)) !=
120 	    (AT_SYMLINK_FOLLOW | AT_SYMLINK_NOFOLLOW));
121 
122 	res = 0;
123 	at_flags &= mask;
124 	if ((at_flags & AT_RESOLVE_BENEATH) != 0)
125 		res |= RBENEATH;
126 	if ((at_flags & AT_SYMLINK_FOLLOW) != 0)
127 		res |= FOLLOW;
128 	/* NOFOLLOW is pseudo flag */
129 	if ((mask & AT_SYMLINK_NOFOLLOW) != 0) {
130 		res |= (at_flags & AT_SYMLINK_NOFOLLOW) != 0 ? NOFOLLOW :
131 		    FOLLOW;
132 	}
133 	if ((mask & AT_EMPTY_PATH) != 0 && (at_flags & AT_EMPTY_PATH) != 0)
134 		res |= EMPTYPATH;
135 	return (res);
136 }
137 
138 int
139 kern_sync(struct thread *td)
140 {
141 	struct mount *mp, *nmp;
142 	int save;
143 
144 	mtx_lock(&mountlist_mtx);
145 	for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
146 		if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK)) {
147 			nmp = TAILQ_NEXT(mp, mnt_list);
148 			continue;
149 		}
150 		if ((mp->mnt_flag & MNT_RDONLY) == 0 &&
151 		    vn_start_write(NULL, &mp, V_NOWAIT) == 0) {
152 			save = curthread_pflags_set(TDP_SYNCIO);
153 			vfs_periodic(mp, MNT_NOWAIT);
154 			VFS_SYNC(mp, MNT_NOWAIT);
155 			curthread_pflags_restore(save);
156 			vn_finished_write(mp);
157 		}
158 		mtx_lock(&mountlist_mtx);
159 		nmp = TAILQ_NEXT(mp, mnt_list);
160 		vfs_unbusy(mp);
161 	}
162 	mtx_unlock(&mountlist_mtx);
163 	return (0);
164 }
165 
166 /*
167  * Sync each mounted filesystem.
168  */
169 #ifndef _SYS_SYSPROTO_H_
170 struct sync_args {
171 	int     dummy;
172 };
173 #endif
174 /* ARGSUSED */
175 int
176 sys_sync(struct thread *td, struct sync_args *uap)
177 {
178 
179 	return (kern_sync(td));
180 }
181 
182 /*
183  * Change filesystem quotas.
184  */
185 #ifndef _SYS_SYSPROTO_H_
186 struct quotactl_args {
187 	char *path;
188 	int cmd;
189 	int uid;
190 	caddr_t arg;
191 };
192 #endif
193 int
194 sys_quotactl(struct thread *td, struct quotactl_args *uap)
195 {
196 	struct mount *mp;
197 	struct nameidata nd;
198 	int error;
199 	bool mp_busy;
200 
201 	AUDIT_ARG_CMD(uap->cmd);
202 	AUDIT_ARG_UID(uap->uid);
203 	if (!prison_allow(td->td_ucred, PR_ALLOW_QUOTAS))
204 		return (EPERM);
205 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
206 	    uap->path, td);
207 	if ((error = namei(&nd)) != 0)
208 		return (error);
209 	NDFREE(&nd, NDF_ONLY_PNBUF);
210 	mp = nd.ni_vp->v_mount;
211 	vfs_ref(mp);
212 	vput(nd.ni_vp);
213 	error = vfs_busy(mp, 0);
214 	if (error != 0) {
215 		vfs_rel(mp);
216 		return (error);
217 	}
218 	mp_busy = true;
219 	error = VFS_QUOTACTL(mp, uap->cmd, uap->uid, uap->arg, &mp_busy);
220 
221 	/*
222 	 * Since quota on/off operations typically need to open quota
223 	 * files, the implementation may need to unbusy the mount point
224 	 * before calling into namei.  Otherwise, unmount might be
225 	 * started between two vfs_busy() invocations (first is ours,
226 	 * second is from mount point cross-walk code in lookup()),
227 	 * causing deadlock.
228 	 *
229 	 * Avoid unbusying mp if the implementation indicates it has
230 	 * already done so.
231 	 */
232 	if (mp_busy)
233 		vfs_unbusy(mp);
234 	vfs_rel(mp);
235 	return (error);
236 }
237 
238 /*
239  * Used by statfs conversion routines to scale the block size up if
240  * necessary so that all of the block counts are <= 'max_size'.  Note
241  * that 'max_size' should be a bitmask, i.e. 2^n - 1 for some non-zero
242  * value of 'n'.
243  */
244 void
245 statfs_scale_blocks(struct statfs *sf, long max_size)
246 {
247 	uint64_t count;
248 	int shift;
249 
250 	KASSERT(powerof2(max_size + 1), ("%s: invalid max_size", __func__));
251 
252 	/*
253 	 * Attempt to scale the block counts to give a more accurate
254 	 * overview to userland of the ratio of free space to used
255 	 * space.  To do this, find the largest block count and compute
256 	 * a divisor that lets it fit into a signed integer <= max_size.
257 	 */
258 	if (sf->f_bavail < 0)
259 		count = -sf->f_bavail;
260 	else
261 		count = sf->f_bavail;
262 	count = MAX(sf->f_blocks, MAX(sf->f_bfree, count));
263 	if (count <= max_size)
264 		return;
265 
266 	count >>= flsl(max_size);
267 	shift = 0;
268 	while (count > 0) {
269 		shift++;
270 		count >>=1;
271 	}
272 
273 	sf->f_bsize <<= shift;
274 	sf->f_blocks >>= shift;
275 	sf->f_bfree >>= shift;
276 	sf->f_bavail >>= shift;
277 }
278 
279 static int
280 kern_do_statfs(struct thread *td, struct mount *mp, struct statfs *buf)
281 {
282 	int error;
283 
284 	if (mp == NULL)
285 		return (EBADF);
286 	error = vfs_busy(mp, 0);
287 	vfs_rel(mp);
288 	if (error != 0)
289 		return (error);
290 #ifdef MAC
291 	error = mac_mount_check_stat(td->td_ucred, mp);
292 	if (error != 0)
293 		goto out;
294 #endif
295 	error = VFS_STATFS(mp, buf);
296 	if (error != 0)
297 		goto out;
298 	if (priv_check_cred_vfs_generation(td->td_ucred)) {
299 		buf->f_fsid.val[0] = buf->f_fsid.val[1] = 0;
300 		prison_enforce_statfs(td->td_ucred, mp, buf);
301 	}
302 out:
303 	vfs_unbusy(mp);
304 	return (error);
305 }
306 
307 /*
308  * Get filesystem statistics.
309  */
310 #ifndef _SYS_SYSPROTO_H_
311 struct statfs_args {
312 	char *path;
313 	struct statfs *buf;
314 };
315 #endif
316 int
317 sys_statfs(struct thread *td, struct statfs_args *uap)
318 {
319 	struct statfs *sfp;
320 	int error;
321 
322 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
323 	error = kern_statfs(td, uap->path, UIO_USERSPACE, sfp);
324 	if (error == 0)
325 		error = copyout(sfp, uap->buf, sizeof(struct statfs));
326 	free(sfp, M_STATFS);
327 	return (error);
328 }
329 
330 int
331 kern_statfs(struct thread *td, const char *path, enum uio_seg pathseg,
332     struct statfs *buf)
333 {
334 	struct mount *mp;
335 	struct nameidata nd;
336 	int error;
337 
338 	NDINIT(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, td);
339 	error = namei(&nd);
340 	if (error != 0)
341 		return (error);
342 	mp = vfs_ref_from_vp(nd.ni_vp);
343 	NDFREE_NOTHING(&nd);
344 	vrele(nd.ni_vp);
345 	return (kern_do_statfs(td, mp, buf));
346 }
347 
348 /*
349  * Get filesystem statistics.
350  */
351 #ifndef _SYS_SYSPROTO_H_
352 struct fstatfs_args {
353 	int fd;
354 	struct statfs *buf;
355 };
356 #endif
357 int
358 sys_fstatfs(struct thread *td, struct fstatfs_args *uap)
359 {
360 	struct statfs *sfp;
361 	int error;
362 
363 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
364 	error = kern_fstatfs(td, uap->fd, sfp);
365 	if (error == 0)
366 		error = copyout(sfp, uap->buf, sizeof(struct statfs));
367 	free(sfp, M_STATFS);
368 	return (error);
369 }
370 
371 int
372 kern_fstatfs(struct thread *td, int fd, struct statfs *buf)
373 {
374 	struct file *fp;
375 	struct mount *mp;
376 	struct vnode *vp;
377 	int error;
378 
379 	AUDIT_ARG_FD(fd);
380 	error = getvnode_path(td, fd, &cap_fstatfs_rights, &fp);
381 	if (error != 0)
382 		return (error);
383 	vp = fp->f_vnode;
384 #ifdef AUDIT
385 	if (AUDITING_TD(td)) {
386 		vn_lock(vp, LK_SHARED | LK_RETRY);
387 		AUDIT_ARG_VNODE1(vp);
388 		VOP_UNLOCK(vp);
389 	}
390 #endif
391 	mp = vfs_ref_from_vp(vp);
392 	fdrop(fp, td);
393 	return (kern_do_statfs(td, mp, buf));
394 }
395 
396 /*
397  * Get statistics on all filesystems.
398  */
399 #ifndef _SYS_SYSPROTO_H_
400 struct getfsstat_args {
401 	struct statfs *buf;
402 	long bufsize;
403 	int mode;
404 };
405 #endif
406 int
407 sys_getfsstat(struct thread *td, struct getfsstat_args *uap)
408 {
409 	size_t count;
410 	int error;
411 
412 	if (uap->bufsize < 0 || uap->bufsize > SIZE_MAX)
413 		return (EINVAL);
414 	error = kern_getfsstat(td, &uap->buf, uap->bufsize, &count,
415 	    UIO_USERSPACE, uap->mode);
416 	if (error == 0)
417 		td->td_retval[0] = count;
418 	return (error);
419 }
420 
421 /*
422  * If (bufsize > 0 && bufseg == UIO_SYSSPACE)
423  *	The caller is responsible for freeing memory which will be allocated
424  *	in '*buf'.
425  */
426 int
427 kern_getfsstat(struct thread *td, struct statfs **buf, size_t bufsize,
428     size_t *countp, enum uio_seg bufseg, int mode)
429 {
430 	struct mount *mp, *nmp;
431 	struct statfs *sfsp, *sp, *sptmp, *tofree;
432 	size_t count, maxcount;
433 	int error;
434 
435 	switch (mode) {
436 	case MNT_WAIT:
437 	case MNT_NOWAIT:
438 		break;
439 	default:
440 		if (bufseg == UIO_SYSSPACE)
441 			*buf = NULL;
442 		return (EINVAL);
443 	}
444 restart:
445 	maxcount = bufsize / sizeof(struct statfs);
446 	if (bufsize == 0) {
447 		sfsp = NULL;
448 		tofree = NULL;
449 	} else if (bufseg == UIO_USERSPACE) {
450 		sfsp = *buf;
451 		tofree = NULL;
452 	} else /* if (bufseg == UIO_SYSSPACE) */ {
453 		count = 0;
454 		mtx_lock(&mountlist_mtx);
455 		TAILQ_FOREACH(mp, &mountlist, mnt_list) {
456 			count++;
457 		}
458 		mtx_unlock(&mountlist_mtx);
459 		if (maxcount > count)
460 			maxcount = count;
461 		tofree = sfsp = *buf = malloc(maxcount * sizeof(struct statfs),
462 		    M_STATFS, M_WAITOK);
463 	}
464 
465 	count = 0;
466 
467 	/*
468 	 * If there is no target buffer they only want the count.
469 	 *
470 	 * This could be TAILQ_FOREACH but it is open-coded to match the original
471 	 * code below.
472 	 */
473 	if (sfsp == NULL) {
474 		mtx_lock(&mountlist_mtx);
475 		for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
476 			if (prison_canseemount(td->td_ucred, mp) != 0) {
477 				nmp = TAILQ_NEXT(mp, mnt_list);
478 				continue;
479 			}
480 #ifdef MAC
481 			if (mac_mount_check_stat(td->td_ucred, mp) != 0) {
482 				nmp = TAILQ_NEXT(mp, mnt_list);
483 				continue;
484 			}
485 #endif
486 			count++;
487 			nmp = TAILQ_NEXT(mp, mnt_list);
488 		}
489 		mtx_unlock(&mountlist_mtx);
490 		*countp = count;
491 		return (0);
492 	}
493 
494 	/*
495 	 * They want the entire thing.
496 	 *
497 	 * Short-circuit the corner case of no room for anything, avoids
498 	 * relocking below.
499 	 */
500 	if (maxcount < 1) {
501 		goto out;
502 	}
503 
504 	mtx_lock(&mountlist_mtx);
505 	for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
506 		if (prison_canseemount(td->td_ucred, mp) != 0) {
507 			nmp = TAILQ_NEXT(mp, mnt_list);
508 			continue;
509 		}
510 #ifdef MAC
511 		if (mac_mount_check_stat(td->td_ucred, mp) != 0) {
512 			nmp = TAILQ_NEXT(mp, mnt_list);
513 			continue;
514 		}
515 #endif
516 		if (mode == MNT_WAIT) {
517 			if (vfs_busy(mp, MBF_MNTLSTLOCK) != 0) {
518 				/*
519 				 * If vfs_busy() failed, and MBF_NOWAIT
520 				 * wasn't passed, then the mp is gone.
521 				 * Furthermore, because of MBF_MNTLSTLOCK,
522 				 * the mountlist_mtx was dropped.  We have
523 				 * no other choice than to start over.
524 				 */
525 				mtx_unlock(&mountlist_mtx);
526 				free(tofree, M_STATFS);
527 				goto restart;
528 			}
529 		} else {
530 			if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK) != 0) {
531 				nmp = TAILQ_NEXT(mp, mnt_list);
532 				continue;
533 			}
534 		}
535 		sp = &mp->mnt_stat;
536 		/*
537 		 * If MNT_NOWAIT is specified, do not refresh
538 		 * the fsstat cache.
539 		 */
540 		if (mode != MNT_NOWAIT) {
541 			error = VFS_STATFS(mp, sp);
542 			if (error != 0) {
543 				mtx_lock(&mountlist_mtx);
544 				nmp = TAILQ_NEXT(mp, mnt_list);
545 				vfs_unbusy(mp);
546 				continue;
547 			}
548 		}
549 		if (priv_check_cred_vfs_generation(td->td_ucred)) {
550 			sptmp = malloc(sizeof(struct statfs), M_STATFS,
551 			    M_WAITOK);
552 			*sptmp = *sp;
553 			sptmp->f_fsid.val[0] = sptmp->f_fsid.val[1] = 0;
554 			prison_enforce_statfs(td->td_ucred, mp, sptmp);
555 			sp = sptmp;
556 		} else
557 			sptmp = NULL;
558 		if (bufseg == UIO_SYSSPACE) {
559 			bcopy(sp, sfsp, sizeof(*sp));
560 			free(sptmp, M_STATFS);
561 		} else /* if (bufseg == UIO_USERSPACE) */ {
562 			error = copyout(sp, sfsp, sizeof(*sp));
563 			free(sptmp, M_STATFS);
564 			if (error != 0) {
565 				vfs_unbusy(mp);
566 				return (error);
567 			}
568 		}
569 		sfsp++;
570 		count++;
571 
572 		if (count == maxcount) {
573 			vfs_unbusy(mp);
574 			goto out;
575 		}
576 
577 		mtx_lock(&mountlist_mtx);
578 		nmp = TAILQ_NEXT(mp, mnt_list);
579 		vfs_unbusy(mp);
580 	}
581 	mtx_unlock(&mountlist_mtx);
582 out:
583 	*countp = count;
584 	return (0);
585 }
586 
587 #ifdef COMPAT_FREEBSD4
588 /*
589  * Get old format filesystem statistics.
590  */
591 static void freebsd4_cvtstatfs(struct statfs *, struct ostatfs *);
592 
593 #ifndef _SYS_SYSPROTO_H_
594 struct freebsd4_statfs_args {
595 	char *path;
596 	struct ostatfs *buf;
597 };
598 #endif
599 int
600 freebsd4_statfs(struct thread *td, struct freebsd4_statfs_args *uap)
601 {
602 	struct ostatfs osb;
603 	struct statfs *sfp;
604 	int error;
605 
606 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
607 	error = kern_statfs(td, uap->path, UIO_USERSPACE, sfp);
608 	if (error == 0) {
609 		freebsd4_cvtstatfs(sfp, &osb);
610 		error = copyout(&osb, uap->buf, sizeof(osb));
611 	}
612 	free(sfp, M_STATFS);
613 	return (error);
614 }
615 
616 /*
617  * Get filesystem statistics.
618  */
619 #ifndef _SYS_SYSPROTO_H_
620 struct freebsd4_fstatfs_args {
621 	int fd;
622 	struct ostatfs *buf;
623 };
624 #endif
625 int
626 freebsd4_fstatfs(struct thread *td, struct freebsd4_fstatfs_args *uap)
627 {
628 	struct ostatfs osb;
629 	struct statfs *sfp;
630 	int error;
631 
632 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
633 	error = kern_fstatfs(td, uap->fd, sfp);
634 	if (error == 0) {
635 		freebsd4_cvtstatfs(sfp, &osb);
636 		error = copyout(&osb, uap->buf, sizeof(osb));
637 	}
638 	free(sfp, M_STATFS);
639 	return (error);
640 }
641 
642 /*
643  * Get statistics on all filesystems.
644  */
645 #ifndef _SYS_SYSPROTO_H_
646 struct freebsd4_getfsstat_args {
647 	struct ostatfs *buf;
648 	long bufsize;
649 	int mode;
650 };
651 #endif
652 int
653 freebsd4_getfsstat(struct thread *td, struct freebsd4_getfsstat_args *uap)
654 {
655 	struct statfs *buf, *sp;
656 	struct ostatfs osb;
657 	size_t count, size;
658 	int error;
659 
660 	if (uap->bufsize < 0)
661 		return (EINVAL);
662 	count = uap->bufsize / sizeof(struct ostatfs);
663 	if (count > SIZE_MAX / sizeof(struct statfs))
664 		return (EINVAL);
665 	size = count * sizeof(struct statfs);
666 	error = kern_getfsstat(td, &buf, size, &count, UIO_SYSSPACE,
667 	    uap->mode);
668 	if (error == 0)
669 		td->td_retval[0] = count;
670 	if (size != 0) {
671 		sp = buf;
672 		while (count != 0 && error == 0) {
673 			freebsd4_cvtstatfs(sp, &osb);
674 			error = copyout(&osb, uap->buf, sizeof(osb));
675 			sp++;
676 			uap->buf++;
677 			count--;
678 		}
679 		free(buf, M_STATFS);
680 	}
681 	return (error);
682 }
683 
684 /*
685  * Implement fstatfs() for (NFS) file handles.
686  */
687 #ifndef _SYS_SYSPROTO_H_
688 struct freebsd4_fhstatfs_args {
689 	struct fhandle *u_fhp;
690 	struct ostatfs *buf;
691 };
692 #endif
693 int
694 freebsd4_fhstatfs(struct thread *td, struct freebsd4_fhstatfs_args *uap)
695 {
696 	struct ostatfs osb;
697 	struct statfs *sfp;
698 	fhandle_t fh;
699 	int error;
700 
701 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
702 	if (error != 0)
703 		return (error);
704 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
705 	error = kern_fhstatfs(td, fh, sfp);
706 	if (error == 0) {
707 		freebsd4_cvtstatfs(sfp, &osb);
708 		error = copyout(&osb, uap->buf, sizeof(osb));
709 	}
710 	free(sfp, M_STATFS);
711 	return (error);
712 }
713 
714 /*
715  * Convert a new format statfs structure to an old format statfs structure.
716  */
717 static void
718 freebsd4_cvtstatfs(struct statfs *nsp, struct ostatfs *osp)
719 {
720 
721 	statfs_scale_blocks(nsp, LONG_MAX);
722 	bzero(osp, sizeof(*osp));
723 	osp->f_bsize = nsp->f_bsize;
724 	osp->f_iosize = MIN(nsp->f_iosize, LONG_MAX);
725 	osp->f_blocks = nsp->f_blocks;
726 	osp->f_bfree = nsp->f_bfree;
727 	osp->f_bavail = nsp->f_bavail;
728 	osp->f_files = MIN(nsp->f_files, LONG_MAX);
729 	osp->f_ffree = MIN(nsp->f_ffree, LONG_MAX);
730 	osp->f_owner = nsp->f_owner;
731 	osp->f_type = nsp->f_type;
732 	osp->f_flags = nsp->f_flags;
733 	osp->f_syncwrites = MIN(nsp->f_syncwrites, LONG_MAX);
734 	osp->f_asyncwrites = MIN(nsp->f_asyncwrites, LONG_MAX);
735 	osp->f_syncreads = MIN(nsp->f_syncreads, LONG_MAX);
736 	osp->f_asyncreads = MIN(nsp->f_asyncreads, LONG_MAX);
737 	strlcpy(osp->f_fstypename, nsp->f_fstypename,
738 	    MIN(MFSNAMELEN, OMFSNAMELEN));
739 	strlcpy(osp->f_mntonname, nsp->f_mntonname,
740 	    MIN(MNAMELEN, OMNAMELEN));
741 	strlcpy(osp->f_mntfromname, nsp->f_mntfromname,
742 	    MIN(MNAMELEN, OMNAMELEN));
743 	osp->f_fsid = nsp->f_fsid;
744 }
745 #endif /* COMPAT_FREEBSD4 */
746 
747 #if defined(COMPAT_FREEBSD11)
748 /*
749  * Get old format filesystem statistics.
750  */
751 static void freebsd11_cvtstatfs(struct statfs *, struct freebsd11_statfs *);
752 
753 int
754 freebsd11_statfs(struct thread *td, struct freebsd11_statfs_args *uap)
755 {
756 	struct freebsd11_statfs osb;
757 	struct statfs *sfp;
758 	int error;
759 
760 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
761 	error = kern_statfs(td, uap->path, UIO_USERSPACE, sfp);
762 	if (error == 0) {
763 		freebsd11_cvtstatfs(sfp, &osb);
764 		error = copyout(&osb, uap->buf, sizeof(osb));
765 	}
766 	free(sfp, M_STATFS);
767 	return (error);
768 }
769 
770 /*
771  * Get filesystem statistics.
772  */
773 int
774 freebsd11_fstatfs(struct thread *td, struct freebsd11_fstatfs_args *uap)
775 {
776 	struct freebsd11_statfs osb;
777 	struct statfs *sfp;
778 	int error;
779 
780 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
781 	error = kern_fstatfs(td, uap->fd, sfp);
782 	if (error == 0) {
783 		freebsd11_cvtstatfs(sfp, &osb);
784 		error = copyout(&osb, uap->buf, sizeof(osb));
785 	}
786 	free(sfp, M_STATFS);
787 	return (error);
788 }
789 
790 /*
791  * Get statistics on all filesystems.
792  */
793 int
794 freebsd11_getfsstat(struct thread *td, struct freebsd11_getfsstat_args *uap)
795 {
796 	return (kern_freebsd11_getfsstat(td, uap->buf, uap->bufsize, uap->mode));
797 }
798 
799 int
800 kern_freebsd11_getfsstat(struct thread *td, struct freebsd11_statfs * ubuf,
801     long bufsize, int mode)
802 {
803 	struct freebsd11_statfs osb;
804 	struct statfs *buf, *sp;
805 	size_t count, size;
806 	int error;
807 
808 	if (bufsize < 0)
809 		return (EINVAL);
810 
811 	count = bufsize / sizeof(struct ostatfs);
812 	size = count * sizeof(struct statfs);
813 	error = kern_getfsstat(td, &buf, size, &count, UIO_SYSSPACE, mode);
814 	if (error == 0)
815 		td->td_retval[0] = count;
816 	if (size > 0) {
817 		sp = buf;
818 		while (count > 0 && error == 0) {
819 			freebsd11_cvtstatfs(sp, &osb);
820 			error = copyout(&osb, ubuf, sizeof(osb));
821 			sp++;
822 			ubuf++;
823 			count--;
824 		}
825 		free(buf, M_STATFS);
826 	}
827 	return (error);
828 }
829 
830 /*
831  * Implement fstatfs() for (NFS) file handles.
832  */
833 int
834 freebsd11_fhstatfs(struct thread *td, struct freebsd11_fhstatfs_args *uap)
835 {
836 	struct freebsd11_statfs osb;
837 	struct statfs *sfp;
838 	fhandle_t fh;
839 	int error;
840 
841 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
842 	if (error)
843 		return (error);
844 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
845 	error = kern_fhstatfs(td, fh, sfp);
846 	if (error == 0) {
847 		freebsd11_cvtstatfs(sfp, &osb);
848 		error = copyout(&osb, uap->buf, sizeof(osb));
849 	}
850 	free(sfp, M_STATFS);
851 	return (error);
852 }
853 
854 /*
855  * Convert a new format statfs structure to an old format statfs structure.
856  */
857 static void
858 freebsd11_cvtstatfs(struct statfs *nsp, struct freebsd11_statfs *osp)
859 {
860 
861 	bzero(osp, sizeof(*osp));
862 	osp->f_version = FREEBSD11_STATFS_VERSION;
863 	osp->f_type = nsp->f_type;
864 	osp->f_flags = nsp->f_flags;
865 	osp->f_bsize = nsp->f_bsize;
866 	osp->f_iosize = nsp->f_iosize;
867 	osp->f_blocks = nsp->f_blocks;
868 	osp->f_bfree = nsp->f_bfree;
869 	osp->f_bavail = nsp->f_bavail;
870 	osp->f_files = nsp->f_files;
871 	osp->f_ffree = nsp->f_ffree;
872 	osp->f_syncwrites = nsp->f_syncwrites;
873 	osp->f_asyncwrites = nsp->f_asyncwrites;
874 	osp->f_syncreads = nsp->f_syncreads;
875 	osp->f_asyncreads = nsp->f_asyncreads;
876 	osp->f_namemax = nsp->f_namemax;
877 	osp->f_owner = nsp->f_owner;
878 	osp->f_fsid = nsp->f_fsid;
879 	strlcpy(osp->f_fstypename, nsp->f_fstypename,
880 	    MIN(MFSNAMELEN, sizeof(osp->f_fstypename)));
881 	strlcpy(osp->f_mntonname, nsp->f_mntonname,
882 	    MIN(MNAMELEN, sizeof(osp->f_mntonname)));
883 	strlcpy(osp->f_mntfromname, nsp->f_mntfromname,
884 	    MIN(MNAMELEN, sizeof(osp->f_mntfromname)));
885 }
886 #endif /* COMPAT_FREEBSD11 */
887 
888 /*
889  * Change current working directory to a given file descriptor.
890  */
891 #ifndef _SYS_SYSPROTO_H_
892 struct fchdir_args {
893 	int	fd;
894 };
895 #endif
896 int
897 sys_fchdir(struct thread *td, struct fchdir_args *uap)
898 {
899 	struct vnode *vp, *tdp;
900 	struct mount *mp;
901 	struct file *fp;
902 	int error;
903 
904 	AUDIT_ARG_FD(uap->fd);
905 	error = getvnode_path(td, uap->fd, &cap_fchdir_rights,
906 	    &fp);
907 	if (error != 0)
908 		return (error);
909 	vp = fp->f_vnode;
910 	vref(vp);
911 	fdrop(fp, td);
912 	vn_lock(vp, LK_SHARED | LK_RETRY);
913 	AUDIT_ARG_VNODE1(vp);
914 	error = change_dir(vp, td);
915 	while (!error && (mp = vp->v_mountedhere) != NULL) {
916 		if (vfs_busy(mp, 0))
917 			continue;
918 		error = VFS_ROOT(mp, LK_SHARED, &tdp);
919 		vfs_unbusy(mp);
920 		if (error != 0)
921 			break;
922 		vput(vp);
923 		vp = tdp;
924 	}
925 	if (error != 0) {
926 		vput(vp);
927 		return (error);
928 	}
929 	VOP_UNLOCK(vp);
930 	pwd_chdir(td, vp);
931 	return (0);
932 }
933 
934 /*
935  * Change current working directory (``.'').
936  */
937 #ifndef _SYS_SYSPROTO_H_
938 struct chdir_args {
939 	char	*path;
940 };
941 #endif
942 int
943 sys_chdir(struct thread *td, struct chdir_args *uap)
944 {
945 
946 	return (kern_chdir(td, uap->path, UIO_USERSPACE));
947 }
948 
949 int
950 kern_chdir(struct thread *td, const char *path, enum uio_seg pathseg)
951 {
952 	struct nameidata nd;
953 	int error;
954 
955 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
956 	    pathseg, path, td);
957 	if ((error = namei(&nd)) != 0)
958 		return (error);
959 	if ((error = change_dir(nd.ni_vp, td)) != 0) {
960 		vput(nd.ni_vp);
961 		NDFREE_NOTHING(&nd);
962 		return (error);
963 	}
964 	VOP_UNLOCK(nd.ni_vp);
965 	NDFREE_NOTHING(&nd);
966 	pwd_chdir(td, nd.ni_vp);
967 	return (0);
968 }
969 
970 static int unprivileged_chroot = 0;
971 SYSCTL_INT(_security_bsd, OID_AUTO, unprivileged_chroot, CTLFLAG_RW,
972     &unprivileged_chroot, 0,
973     "Unprivileged processes can use chroot(2)");
974 /*
975  * Change notion of root (``/'') directory.
976  */
977 #ifndef _SYS_SYSPROTO_H_
978 struct chroot_args {
979 	char	*path;
980 };
981 #endif
982 int
983 sys_chroot(struct thread *td, struct chroot_args *uap)
984 {
985 	struct nameidata nd;
986 	struct proc *p;
987 	int error;
988 
989 	error = priv_check(td, PRIV_VFS_CHROOT);
990 	if (error != 0) {
991 		p = td->td_proc;
992 		PROC_LOCK(p);
993 		if (unprivileged_chroot == 0 ||
994 		    (p->p_flag2 & P2_NO_NEW_PRIVS) == 0) {
995 			PROC_UNLOCK(p);
996 			return (error);
997 		}
998 		PROC_UNLOCK(p);
999 	}
1000 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
1001 	    UIO_USERSPACE, uap->path, td);
1002 	error = namei(&nd);
1003 	if (error != 0)
1004 		goto error;
1005 	error = change_dir(nd.ni_vp, td);
1006 	if (error != 0)
1007 		goto e_vunlock;
1008 #ifdef MAC
1009 	error = mac_vnode_check_chroot(td->td_ucred, nd.ni_vp);
1010 	if (error != 0)
1011 		goto e_vunlock;
1012 #endif
1013 	VOP_UNLOCK(nd.ni_vp);
1014 	error = pwd_chroot(td, nd.ni_vp);
1015 	vrele(nd.ni_vp);
1016 	NDFREE_NOTHING(&nd);
1017 	return (error);
1018 e_vunlock:
1019 	vput(nd.ni_vp);
1020 error:
1021 	NDFREE_NOTHING(&nd);
1022 	return (error);
1023 }
1024 
1025 /*
1026  * Common routine for chroot and chdir.  Callers must provide a locked vnode
1027  * instance.
1028  */
1029 int
1030 change_dir(struct vnode *vp, struct thread *td)
1031 {
1032 #ifdef MAC
1033 	int error;
1034 #endif
1035 
1036 	ASSERT_VOP_LOCKED(vp, "change_dir(): vp not locked");
1037 	if (vp->v_type != VDIR)
1038 		return (ENOTDIR);
1039 #ifdef MAC
1040 	error = mac_vnode_check_chdir(td->td_ucred, vp);
1041 	if (error != 0)
1042 		return (error);
1043 #endif
1044 	return (VOP_ACCESS(vp, VEXEC, td->td_ucred, td));
1045 }
1046 
1047 static __inline void
1048 flags_to_rights(int flags, cap_rights_t *rightsp)
1049 {
1050 	if (flags & O_EXEC) {
1051 		cap_rights_set_one(rightsp, CAP_FEXECVE);
1052 		if (flags & O_PATH)
1053 			return;
1054 	} else {
1055 		switch ((flags & O_ACCMODE)) {
1056 		case O_RDONLY:
1057 			cap_rights_set_one(rightsp, CAP_READ);
1058 			break;
1059 		case O_RDWR:
1060 			cap_rights_set_one(rightsp, CAP_READ);
1061 			/* FALLTHROUGH */
1062 		case O_WRONLY:
1063 			cap_rights_set_one(rightsp, CAP_WRITE);
1064 			if (!(flags & (O_APPEND | O_TRUNC)))
1065 				cap_rights_set_one(rightsp, CAP_SEEK);
1066 			break;
1067 		}
1068 	}
1069 
1070 	if (flags & O_CREAT)
1071 		cap_rights_set_one(rightsp, CAP_CREATE);
1072 
1073 	if (flags & O_TRUNC)
1074 		cap_rights_set_one(rightsp, CAP_FTRUNCATE);
1075 
1076 	if (flags & (O_SYNC | O_FSYNC))
1077 		cap_rights_set_one(rightsp, CAP_FSYNC);
1078 
1079 	if (flags & (O_EXLOCK | O_SHLOCK))
1080 		cap_rights_set_one(rightsp, CAP_FLOCK);
1081 }
1082 
1083 /*
1084  * Check permissions, allocate an open file structure, and call the device
1085  * open routine if any.
1086  */
1087 #ifndef _SYS_SYSPROTO_H_
1088 struct open_args {
1089 	char	*path;
1090 	int	flags;
1091 	int	mode;
1092 };
1093 #endif
1094 int
1095 sys_open(struct thread *td, struct open_args *uap)
1096 {
1097 
1098 	return (kern_openat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1099 	    uap->flags, uap->mode));
1100 }
1101 
1102 #ifndef _SYS_SYSPROTO_H_
1103 struct openat_args {
1104 	int	fd;
1105 	char	*path;
1106 	int	flag;
1107 	int	mode;
1108 };
1109 #endif
1110 int
1111 sys_openat(struct thread *td, struct openat_args *uap)
1112 {
1113 
1114 	AUDIT_ARG_FD(uap->fd);
1115 	return (kern_openat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
1116 	    uap->mode));
1117 }
1118 
1119 int
1120 kern_openat(struct thread *td, int fd, const char *path, enum uio_seg pathseg,
1121     int flags, int mode)
1122 {
1123 	struct proc *p = td->td_proc;
1124 	struct filedesc *fdp;
1125 	struct pwddesc *pdp;
1126 	struct file *fp;
1127 	struct vnode *vp;
1128 	struct nameidata nd;
1129 	cap_rights_t rights;
1130 	int cmode, error, indx;
1131 
1132 	indx = -1;
1133 	fdp = p->p_fd;
1134 	pdp = p->p_pd;
1135 
1136 	AUDIT_ARG_FFLAGS(flags);
1137 	AUDIT_ARG_MODE(mode);
1138 	cap_rights_init_one(&rights, CAP_LOOKUP);
1139 	flags_to_rights(flags, &rights);
1140 
1141 	/*
1142 	 * Only one of the O_EXEC, O_RDONLY, O_WRONLY and O_RDWR flags
1143 	 * may be specified.  On the other hand, for O_PATH any mode
1144 	 * except O_EXEC is ignored.
1145 	 */
1146 	if ((flags & O_PATH) != 0) {
1147 		flags &= ~(O_CREAT | O_ACCMODE);
1148 	} else if ((flags & O_EXEC) != 0) {
1149 		if (flags & O_ACCMODE)
1150 			return (EINVAL);
1151 	} else if ((flags & O_ACCMODE) == O_ACCMODE) {
1152 		return (EINVAL);
1153 	} else {
1154 		flags = FFLAGS(flags);
1155 	}
1156 
1157 	/*
1158 	 * Allocate a file structure. The descriptor to reference it
1159 	 * is allocated and used by finstall_refed() below.
1160 	 */
1161 	error = falloc_noinstall(td, &fp);
1162 	if (error != 0)
1163 		return (error);
1164 	/* Set the flags early so the finit in devfs can pick them up. */
1165 	fp->f_flag = flags & FMASK;
1166 	cmode = ((mode & ~pdp->pd_cmask) & ALLPERMS) & ~S_ISTXT;
1167 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, fd,
1168 	    &rights, td);
1169 	td->td_dupfd = -1;		/* XXX check for fdopen */
1170 	error = vn_open(&nd, &flags, cmode, fp);
1171 	if (error != 0) {
1172 		/*
1173 		 * If the vn_open replaced the method vector, something
1174 		 * wonderous happened deep below and we just pass it up
1175 		 * pretending we know what we do.
1176 		 */
1177 		if (error == ENXIO && fp->f_ops != &badfileops) {
1178 			MPASS((flags & O_PATH) == 0);
1179 			goto success;
1180 		}
1181 
1182 		/*
1183 		 * Handle special fdopen() case. bleh.
1184 		 *
1185 		 * Don't do this for relative (capability) lookups; we don't
1186 		 * understand exactly what would happen, and we don't think
1187 		 * that it ever should.
1188 		 */
1189 		if ((nd.ni_resflags & NIRES_STRICTREL) == 0 &&
1190 		    (error == ENODEV || error == ENXIO) &&
1191 		    td->td_dupfd >= 0) {
1192 			error = dupfdopen(td, fdp, td->td_dupfd, flags, error,
1193 			    &indx);
1194 			if (error == 0)
1195 				goto success;
1196 		}
1197 
1198 		goto bad;
1199 	}
1200 	td->td_dupfd = 0;
1201 	NDFREE(&nd, NDF_ONLY_PNBUF);
1202 	vp = nd.ni_vp;
1203 
1204 	/*
1205 	 * Store the vnode, for any f_type. Typically, the vnode use
1206 	 * count is decremented by direct call to vn_closefile() for
1207 	 * files that switched type in the cdevsw fdopen() method.
1208 	 */
1209 	fp->f_vnode = vp;
1210 
1211 	/*
1212 	 * If the file wasn't claimed by devfs bind it to the normal
1213 	 * vnode operations here.
1214 	 */
1215 	if (fp->f_ops == &badfileops) {
1216 		KASSERT(vp->v_type != VFIFO || (flags & O_PATH) != 0,
1217 		    ("Unexpected fifo fp %p vp %p", fp, vp));
1218 		if ((flags & O_PATH) != 0) {
1219 			finit(fp, (flags & FMASK) | (fp->f_flag & FKQALLOWED),
1220 			    DTYPE_VNODE, NULL, &path_fileops);
1221 			vhold(vp);
1222 			vunref(vp);
1223 		} else {
1224 			finit_vnode(fp, flags, NULL, &vnops);
1225 		}
1226 	}
1227 
1228 	VOP_UNLOCK(vp);
1229 	if (flags & O_TRUNC) {
1230 		error = fo_truncate(fp, 0, td->td_ucred, td);
1231 		if (error != 0)
1232 			goto bad;
1233 	}
1234 success:
1235 	/*
1236 	 * If we haven't already installed the FD (for dupfdopen), do so now.
1237 	 */
1238 	if (indx == -1) {
1239 		struct filecaps *fcaps;
1240 
1241 #ifdef CAPABILITIES
1242 		if ((nd.ni_resflags & NIRES_STRICTREL) != 0)
1243 			fcaps = &nd.ni_filecaps;
1244 		else
1245 #endif
1246 			fcaps = NULL;
1247 		error = finstall_refed(td, fp, &indx, flags, fcaps);
1248 		/* On success finstall_refed() consumes fcaps. */
1249 		if (error != 0) {
1250 			filecaps_free(&nd.ni_filecaps);
1251 			goto bad;
1252 		}
1253 	} else {
1254 		filecaps_free(&nd.ni_filecaps);
1255 		falloc_abort(td, fp);
1256 	}
1257 
1258 	td->td_retval[0] = indx;
1259 	return (0);
1260 bad:
1261 	KASSERT(indx == -1, ("indx=%d, should be -1", indx));
1262 	falloc_abort(td, fp);
1263 	return (error);
1264 }
1265 
1266 #ifdef COMPAT_43
1267 /*
1268  * Create a file.
1269  */
1270 #ifndef _SYS_SYSPROTO_H_
1271 struct ocreat_args {
1272 	char	*path;
1273 	int	mode;
1274 };
1275 #endif
1276 int
1277 ocreat(struct thread *td, struct ocreat_args *uap)
1278 {
1279 
1280 	return (kern_openat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1281 	    O_WRONLY | O_CREAT | O_TRUNC, uap->mode));
1282 }
1283 #endif /* COMPAT_43 */
1284 
1285 /*
1286  * Create a special file.
1287  */
1288 #ifndef _SYS_SYSPROTO_H_
1289 struct mknodat_args {
1290 	int	fd;
1291 	char	*path;
1292 	mode_t	mode;
1293 	dev_t	dev;
1294 };
1295 #endif
1296 int
1297 sys_mknodat(struct thread *td, struct mknodat_args *uap)
1298 {
1299 
1300 	return (kern_mknodat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode,
1301 	    uap->dev));
1302 }
1303 
1304 #if defined(COMPAT_FREEBSD11)
1305 int
1306 freebsd11_mknod(struct thread *td,
1307     struct freebsd11_mknod_args *uap)
1308 {
1309 
1310 	return (kern_mknodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1311 	    uap->mode, uap->dev));
1312 }
1313 
1314 int
1315 freebsd11_mknodat(struct thread *td,
1316     struct freebsd11_mknodat_args *uap)
1317 {
1318 
1319 	return (kern_mknodat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode,
1320 	    uap->dev));
1321 }
1322 #endif /* COMPAT_FREEBSD11 */
1323 
1324 int
1325 kern_mknodat(struct thread *td, int fd, const char *path, enum uio_seg pathseg,
1326     int mode, dev_t dev)
1327 {
1328 	struct vnode *vp;
1329 	struct mount *mp;
1330 	struct vattr vattr;
1331 	struct nameidata nd;
1332 	int error, whiteout = 0;
1333 
1334 	AUDIT_ARG_MODE(mode);
1335 	AUDIT_ARG_DEV(dev);
1336 	switch (mode & S_IFMT) {
1337 	case S_IFCHR:
1338 	case S_IFBLK:
1339 		error = priv_check(td, PRIV_VFS_MKNOD_DEV);
1340 		if (error == 0 && dev == VNOVAL)
1341 			error = EINVAL;
1342 		break;
1343 	case S_IFWHT:
1344 		error = priv_check(td, PRIV_VFS_MKNOD_WHT);
1345 		break;
1346 	case S_IFIFO:
1347 		if (dev == 0)
1348 			return (kern_mkfifoat(td, fd, path, pathseg, mode));
1349 		/* FALLTHROUGH */
1350 	default:
1351 		error = EINVAL;
1352 		break;
1353 	}
1354 	if (error != 0)
1355 		return (error);
1356 	NDPREINIT(&nd);
1357 restart:
1358 	bwillwrite();
1359 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
1360 	    NOCACHE, pathseg, path, fd, &cap_mknodat_rights,
1361 	    td);
1362 	if ((error = namei(&nd)) != 0)
1363 		return (error);
1364 	vp = nd.ni_vp;
1365 	if (vp != NULL) {
1366 		NDFREE(&nd, NDF_ONLY_PNBUF);
1367 		if (vp == nd.ni_dvp)
1368 			vrele(nd.ni_dvp);
1369 		else
1370 			vput(nd.ni_dvp);
1371 		vrele(vp);
1372 		return (EEXIST);
1373 	} else {
1374 		VATTR_NULL(&vattr);
1375 		vattr.va_mode = (mode & ALLPERMS) &
1376 		    ~td->td_proc->p_pd->pd_cmask;
1377 		vattr.va_rdev = dev;
1378 		whiteout = 0;
1379 
1380 		switch (mode & S_IFMT) {
1381 		case S_IFCHR:
1382 			vattr.va_type = VCHR;
1383 			break;
1384 		case S_IFBLK:
1385 			vattr.va_type = VBLK;
1386 			break;
1387 		case S_IFWHT:
1388 			whiteout = 1;
1389 			break;
1390 		default:
1391 			panic("kern_mknod: invalid mode");
1392 		}
1393 	}
1394 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1395 		NDFREE(&nd, NDF_ONLY_PNBUF);
1396 		vput(nd.ni_dvp);
1397 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1398 			return (error);
1399 		goto restart;
1400 	}
1401 #ifdef MAC
1402 	if (error == 0 && !whiteout)
1403 		error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp,
1404 		    &nd.ni_cnd, &vattr);
1405 #endif
1406 	if (error == 0) {
1407 		if (whiteout)
1408 			error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE);
1409 		else {
1410 			error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp,
1411 						&nd.ni_cnd, &vattr);
1412 		}
1413 	}
1414 	VOP_VPUT_PAIR(nd.ni_dvp, error == 0 && !whiteout ? &nd.ni_vp : NULL,
1415 	    true);
1416 	vn_finished_write(mp);
1417 	NDFREE(&nd, NDF_ONLY_PNBUF);
1418 	if (error == ERELOOKUP)
1419 		goto restart;
1420 	return (error);
1421 }
1422 
1423 /*
1424  * Create a named pipe.
1425  */
1426 #ifndef _SYS_SYSPROTO_H_
1427 struct mkfifo_args {
1428 	char	*path;
1429 	int	mode;
1430 };
1431 #endif
1432 int
1433 sys_mkfifo(struct thread *td, struct mkfifo_args *uap)
1434 {
1435 
1436 	return (kern_mkfifoat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1437 	    uap->mode));
1438 }
1439 
1440 #ifndef _SYS_SYSPROTO_H_
1441 struct mkfifoat_args {
1442 	int	fd;
1443 	char	*path;
1444 	mode_t	mode;
1445 };
1446 #endif
1447 int
1448 sys_mkfifoat(struct thread *td, struct mkfifoat_args *uap)
1449 {
1450 
1451 	return (kern_mkfifoat(td, uap->fd, uap->path, UIO_USERSPACE,
1452 	    uap->mode));
1453 }
1454 
1455 int
1456 kern_mkfifoat(struct thread *td, int fd, const char *path,
1457     enum uio_seg pathseg, int mode)
1458 {
1459 	struct mount *mp;
1460 	struct vattr vattr;
1461 	struct nameidata nd;
1462 	int error;
1463 
1464 	AUDIT_ARG_MODE(mode);
1465 	NDPREINIT(&nd);
1466 restart:
1467 	bwillwrite();
1468 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
1469 	    NOCACHE, pathseg, path, fd, &cap_mkfifoat_rights,
1470 	    td);
1471 	if ((error = namei(&nd)) != 0)
1472 		return (error);
1473 	if (nd.ni_vp != NULL) {
1474 		NDFREE(&nd, NDF_ONLY_PNBUF);
1475 		if (nd.ni_vp == nd.ni_dvp)
1476 			vrele(nd.ni_dvp);
1477 		else
1478 			vput(nd.ni_dvp);
1479 		vrele(nd.ni_vp);
1480 		return (EEXIST);
1481 	}
1482 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1483 		NDFREE(&nd, NDF_ONLY_PNBUF);
1484 		vput(nd.ni_dvp);
1485 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1486 			return (error);
1487 		goto restart;
1488 	}
1489 	VATTR_NULL(&vattr);
1490 	vattr.va_type = VFIFO;
1491 	vattr.va_mode = (mode & ALLPERMS) & ~td->td_proc->p_pd->pd_cmask;
1492 #ifdef MAC
1493 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1494 	    &vattr);
1495 	if (error != 0)
1496 		goto out;
1497 #endif
1498 	error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
1499 #ifdef MAC
1500 out:
1501 #endif
1502 	VOP_VPUT_PAIR(nd.ni_dvp, error == 0 ? &nd.ni_vp : NULL, true);
1503 	vn_finished_write(mp);
1504 	NDFREE(&nd, NDF_ONLY_PNBUF);
1505 	if (error == ERELOOKUP)
1506 		goto restart;
1507 	return (error);
1508 }
1509 
1510 /*
1511  * Make a hard file link.
1512  */
1513 #ifndef _SYS_SYSPROTO_H_
1514 struct link_args {
1515 	char	*path;
1516 	char	*link;
1517 };
1518 #endif
1519 int
1520 sys_link(struct thread *td, struct link_args *uap)
1521 {
1522 
1523 	return (kern_linkat(td, AT_FDCWD, AT_FDCWD, uap->path, uap->link,
1524 	    UIO_USERSPACE, AT_SYMLINK_FOLLOW));
1525 }
1526 
1527 #ifndef _SYS_SYSPROTO_H_
1528 struct linkat_args {
1529 	int	fd1;
1530 	char	*path1;
1531 	int	fd2;
1532 	char	*path2;
1533 	int	flag;
1534 };
1535 #endif
1536 int
1537 sys_linkat(struct thread *td, struct linkat_args *uap)
1538 {
1539 
1540 	return (kern_linkat(td, uap->fd1, uap->fd2, uap->path1, uap->path2,
1541 	    UIO_USERSPACE, uap->flag));
1542 }
1543 
1544 int hardlink_check_uid = 0;
1545 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_uid, CTLFLAG_RW,
1546     &hardlink_check_uid, 0,
1547     "Unprivileged processes cannot create hard links to files owned by other "
1548     "users");
1549 static int hardlink_check_gid = 0;
1550 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_gid, CTLFLAG_RW,
1551     &hardlink_check_gid, 0,
1552     "Unprivileged processes cannot create hard links to files owned by other "
1553     "groups");
1554 
1555 static int
1556 can_hardlink(struct vnode *vp, struct ucred *cred)
1557 {
1558 	struct vattr va;
1559 	int error;
1560 
1561 	if (!hardlink_check_uid && !hardlink_check_gid)
1562 		return (0);
1563 
1564 	error = VOP_GETATTR(vp, &va, cred);
1565 	if (error != 0)
1566 		return (error);
1567 
1568 	if (hardlink_check_uid && cred->cr_uid != va.va_uid) {
1569 		error = priv_check_cred(cred, PRIV_VFS_LINK);
1570 		if (error != 0)
1571 			return (error);
1572 	}
1573 
1574 	if (hardlink_check_gid && !groupmember(va.va_gid, cred)) {
1575 		error = priv_check_cred(cred, PRIV_VFS_LINK);
1576 		if (error != 0)
1577 			return (error);
1578 	}
1579 
1580 	return (0);
1581 }
1582 
1583 int
1584 kern_linkat(struct thread *td, int fd1, int fd2, const char *path1,
1585     const char *path2, enum uio_seg segflag, int flag)
1586 {
1587 	struct nameidata nd;
1588 	int error;
1589 
1590 	if ((flag & ~(AT_SYMLINK_FOLLOW | AT_RESOLVE_BENEATH |
1591 	    AT_EMPTY_PATH)) != 0)
1592 		return (EINVAL);
1593 
1594 	NDPREINIT(&nd);
1595 	do {
1596 		bwillwrite();
1597 		NDINIT_ATRIGHTS(&nd, LOOKUP, AUDITVNODE1 | at2cnpflags(flag,
1598 		    AT_SYMLINK_FOLLOW | AT_RESOLVE_BENEATH | AT_EMPTY_PATH),
1599 		    segflag, path1, fd1, &cap_linkat_source_rights, td);
1600 		if ((error = namei(&nd)) != 0)
1601 			return (error);
1602 		NDFREE(&nd, NDF_ONLY_PNBUF);
1603 		if ((nd.ni_resflags & NIRES_EMPTYPATH) != 0) {
1604 			error = priv_check(td, PRIV_VFS_FHOPEN);
1605 			if (error != 0) {
1606 				vrele(nd.ni_vp);
1607 				return (error);
1608 			}
1609 		}
1610 		error = kern_linkat_vp(td, nd.ni_vp, fd2, path2, segflag);
1611 	} while (error ==  EAGAIN || error == ERELOOKUP);
1612 	return (error);
1613 }
1614 
1615 static int
1616 kern_linkat_vp(struct thread *td, struct vnode *vp, int fd, const char *path,
1617     enum uio_seg segflag)
1618 {
1619 	struct nameidata nd;
1620 	struct mount *mp;
1621 	int error;
1622 
1623 	if (vp->v_type == VDIR) {
1624 		vrele(vp);
1625 		return (EPERM);		/* POSIX */
1626 	}
1627 	NDINIT_ATRIGHTS(&nd, CREATE,
1628 	    LOCKPARENT | SAVENAME | AUDITVNODE2 | NOCACHE, segflag, path, fd,
1629 	    &cap_linkat_target_rights, td);
1630 	if ((error = namei(&nd)) == 0) {
1631 		if (nd.ni_vp != NULL) {
1632 			NDFREE(&nd, NDF_ONLY_PNBUF);
1633 			if (nd.ni_dvp == nd.ni_vp)
1634 				vrele(nd.ni_dvp);
1635 			else
1636 				vput(nd.ni_dvp);
1637 			vrele(nd.ni_vp);
1638 			vrele(vp);
1639 			return (EEXIST);
1640 		} else if (nd.ni_dvp->v_mount != vp->v_mount) {
1641 			/*
1642 			 * Cross-device link.  No need to recheck
1643 			 * vp->v_type, since it cannot change, except
1644 			 * to VBAD.
1645 			 */
1646 			NDFREE(&nd, NDF_ONLY_PNBUF);
1647 			vput(nd.ni_dvp);
1648 			vrele(vp);
1649 			return (EXDEV);
1650 		} else if ((error = vn_lock(vp, LK_EXCLUSIVE)) == 0) {
1651 			error = can_hardlink(vp, td->td_ucred);
1652 #ifdef MAC
1653 			if (error == 0)
1654 				error = mac_vnode_check_link(td->td_ucred,
1655 				    nd.ni_dvp, vp, &nd.ni_cnd);
1656 #endif
1657 			if (error != 0) {
1658 				vput(vp);
1659 				vput(nd.ni_dvp);
1660 				NDFREE(&nd, NDF_ONLY_PNBUF);
1661 				return (error);
1662 			}
1663 			error = vn_start_write(vp, &mp, V_NOWAIT);
1664 			if (error != 0) {
1665 				vput(vp);
1666 				vput(nd.ni_dvp);
1667 				NDFREE(&nd, NDF_ONLY_PNBUF);
1668 				error = vn_start_write(NULL, &mp,
1669 				    V_XSLEEP | PCATCH);
1670 				if (error != 0)
1671 					return (error);
1672 				return (EAGAIN);
1673 			}
1674 			error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
1675 			VOP_VPUT_PAIR(nd.ni_dvp, &vp, true);
1676 			vn_finished_write(mp);
1677 			NDFREE(&nd, NDF_ONLY_PNBUF);
1678 			vp = NULL;
1679 		} else {
1680 			vput(nd.ni_dvp);
1681 			NDFREE(&nd, NDF_ONLY_PNBUF);
1682 			vrele(vp);
1683 			return (EAGAIN);
1684 		}
1685 	}
1686 	if (vp != NULL)
1687 		vrele(vp);
1688 	return (error);
1689 }
1690 
1691 /*
1692  * Make a symbolic link.
1693  */
1694 #ifndef _SYS_SYSPROTO_H_
1695 struct symlink_args {
1696 	char	*path;
1697 	char	*link;
1698 };
1699 #endif
1700 int
1701 sys_symlink(struct thread *td, struct symlink_args *uap)
1702 {
1703 
1704 	return (kern_symlinkat(td, uap->path, AT_FDCWD, uap->link,
1705 	    UIO_USERSPACE));
1706 }
1707 
1708 #ifndef _SYS_SYSPROTO_H_
1709 struct symlinkat_args {
1710 	char	*path;
1711 	int	fd;
1712 	char	*path2;
1713 };
1714 #endif
1715 int
1716 sys_symlinkat(struct thread *td, struct symlinkat_args *uap)
1717 {
1718 
1719 	return (kern_symlinkat(td, uap->path1, uap->fd, uap->path2,
1720 	    UIO_USERSPACE));
1721 }
1722 
1723 int
1724 kern_symlinkat(struct thread *td, const char *path1, int fd, const char *path2,
1725     enum uio_seg segflg)
1726 {
1727 	struct mount *mp;
1728 	struct vattr vattr;
1729 	const char *syspath;
1730 	char *tmppath;
1731 	struct nameidata nd;
1732 	int error;
1733 
1734 	if (segflg == UIO_SYSSPACE) {
1735 		syspath = path1;
1736 	} else {
1737 		tmppath = uma_zalloc(namei_zone, M_WAITOK);
1738 		if ((error = copyinstr(path1, tmppath, MAXPATHLEN, NULL)) != 0)
1739 			goto out;
1740 		syspath = tmppath;
1741 	}
1742 	AUDIT_ARG_TEXT(syspath);
1743 	NDPREINIT(&nd);
1744 restart:
1745 	bwillwrite();
1746 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
1747 	    NOCACHE, segflg, path2, fd, &cap_symlinkat_rights,
1748 	    td);
1749 	if ((error = namei(&nd)) != 0)
1750 		goto out;
1751 	if (nd.ni_vp) {
1752 		NDFREE(&nd, NDF_ONLY_PNBUF);
1753 		if (nd.ni_vp == nd.ni_dvp)
1754 			vrele(nd.ni_dvp);
1755 		else
1756 			vput(nd.ni_dvp);
1757 		vrele(nd.ni_vp);
1758 		nd.ni_vp = NULL;
1759 		error = EEXIST;
1760 		goto out;
1761 	}
1762 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1763 		NDFREE(&nd, NDF_ONLY_PNBUF);
1764 		vput(nd.ni_dvp);
1765 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1766 			goto out;
1767 		goto restart;
1768 	}
1769 	VATTR_NULL(&vattr);
1770 	vattr.va_mode = ACCESSPERMS &~ td->td_proc->p_pd->pd_cmask;
1771 #ifdef MAC
1772 	vattr.va_type = VLNK;
1773 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1774 	    &vattr);
1775 	if (error != 0)
1776 		goto out2;
1777 #endif
1778 	error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, syspath);
1779 #ifdef MAC
1780 out2:
1781 #endif
1782 	VOP_VPUT_PAIR(nd.ni_dvp, error == 0 ? &nd.ni_vp : NULL, true);
1783 	vn_finished_write(mp);
1784 	NDFREE(&nd, NDF_ONLY_PNBUF);
1785 	if (error == ERELOOKUP)
1786 		goto restart;
1787 out:
1788 	if (segflg != UIO_SYSSPACE)
1789 		uma_zfree(namei_zone, tmppath);
1790 	return (error);
1791 }
1792 
1793 /*
1794  * Delete a whiteout from the filesystem.
1795  */
1796 #ifndef _SYS_SYSPROTO_H_
1797 struct undelete_args {
1798 	char *path;
1799 };
1800 #endif
1801 int
1802 sys_undelete(struct thread *td, struct undelete_args *uap)
1803 {
1804 	struct mount *mp;
1805 	struct nameidata nd;
1806 	int error;
1807 
1808 	NDPREINIT(&nd);
1809 restart:
1810 	bwillwrite();
1811 	NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | AUDITVNODE1,
1812 	    UIO_USERSPACE, uap->path, td);
1813 	error = namei(&nd);
1814 	if (error != 0)
1815 		return (error);
1816 
1817 	if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) {
1818 		NDFREE(&nd, NDF_ONLY_PNBUF);
1819 		if (nd.ni_vp == nd.ni_dvp)
1820 			vrele(nd.ni_dvp);
1821 		else
1822 			vput(nd.ni_dvp);
1823 		if (nd.ni_vp)
1824 			vrele(nd.ni_vp);
1825 		return (EEXIST);
1826 	}
1827 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1828 		NDFREE(&nd, NDF_ONLY_PNBUF);
1829 		vput(nd.ni_dvp);
1830 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1831 			return (error);
1832 		goto restart;
1833 	}
1834 	error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE);
1835 	NDFREE(&nd, NDF_ONLY_PNBUF);
1836 	vput(nd.ni_dvp);
1837 	vn_finished_write(mp);
1838 	if (error == ERELOOKUP)
1839 		goto restart;
1840 	return (error);
1841 }
1842 
1843 /*
1844  * Delete a name from the filesystem.
1845  */
1846 #ifndef _SYS_SYSPROTO_H_
1847 struct unlink_args {
1848 	char	*path;
1849 };
1850 #endif
1851 int
1852 sys_unlink(struct thread *td, struct unlink_args *uap)
1853 {
1854 
1855 	return (kern_funlinkat(td, AT_FDCWD, uap->path, FD_NONE, UIO_USERSPACE,
1856 	    0, 0));
1857 }
1858 
1859 static int
1860 kern_funlinkat_ex(struct thread *td, int dfd, const char *path, int fd,
1861     int flag, enum uio_seg pathseg, ino_t oldinum)
1862 {
1863 
1864 	if ((flag & ~(AT_REMOVEDIR | AT_RESOLVE_BENEATH)) != 0)
1865 		return (EINVAL);
1866 
1867 	if ((flag & AT_REMOVEDIR) != 0)
1868 		return (kern_frmdirat(td, dfd, path, fd, UIO_USERSPACE, 0));
1869 
1870 	return (kern_funlinkat(td, dfd, path, fd, UIO_USERSPACE, 0, 0));
1871 }
1872 
1873 #ifndef _SYS_SYSPROTO_H_
1874 struct unlinkat_args {
1875 	int	fd;
1876 	char	*path;
1877 	int	flag;
1878 };
1879 #endif
1880 int
1881 sys_unlinkat(struct thread *td, struct unlinkat_args *uap)
1882 {
1883 
1884 	return (kern_funlinkat_ex(td, uap->fd, uap->path, FD_NONE, uap->flag,
1885 	    UIO_USERSPACE, 0));
1886 }
1887 
1888 #ifndef _SYS_SYSPROTO_H_
1889 struct funlinkat_args {
1890 	int		dfd;
1891 	const char	*path;
1892 	int		fd;
1893 	int		flag;
1894 };
1895 #endif
1896 int
1897 sys_funlinkat(struct thread *td, struct funlinkat_args *uap)
1898 {
1899 
1900 	return (kern_funlinkat_ex(td, uap->dfd, uap->path, uap->fd, uap->flag,
1901 	    UIO_USERSPACE, 0));
1902 }
1903 
1904 int
1905 kern_funlinkat(struct thread *td, int dfd, const char *path, int fd,
1906     enum uio_seg pathseg, int flag, ino_t oldinum)
1907 {
1908 	struct mount *mp;
1909 	struct file *fp;
1910 	struct vnode *vp;
1911 	struct nameidata nd;
1912 	struct stat sb;
1913 	int error;
1914 
1915 	fp = NULL;
1916 	if (fd != FD_NONE) {
1917 		error = getvnode_path(td, fd, &cap_no_rights, &fp);
1918 		if (error != 0)
1919 			return (error);
1920 	}
1921 
1922 	NDPREINIT(&nd);
1923 restart:
1924 	bwillwrite();
1925 	NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1 |
1926 	    at2cnpflags(flag, AT_RESOLVE_BENEATH),
1927 	    pathseg, path, dfd, &cap_unlinkat_rights, td);
1928 	if ((error = namei(&nd)) != 0) {
1929 		if (error == EINVAL)
1930 			error = EPERM;
1931 		goto fdout;
1932 	}
1933 	vp = nd.ni_vp;
1934 	if (vp->v_type == VDIR && oldinum == 0) {
1935 		error = EPERM;		/* POSIX */
1936 	} else if (oldinum != 0 &&
1937 	    ((error = VOP_STAT(vp, &sb, td->td_ucred, NOCRED)) == 0) &&
1938 	    sb.st_ino != oldinum) {
1939 		error = EIDRM;	/* Identifier removed */
1940 	} else if (fp != NULL && fp->f_vnode != vp) {
1941 		if (VN_IS_DOOMED(fp->f_vnode))
1942 			error = EBADF;
1943 		else
1944 			error = EDEADLK;
1945 	} else {
1946 		/*
1947 		 * The root of a mounted filesystem cannot be deleted.
1948 		 *
1949 		 * XXX: can this only be a VDIR case?
1950 		 */
1951 		if (vp->v_vflag & VV_ROOT)
1952 			error = EBUSY;
1953 	}
1954 	if (error == 0) {
1955 		if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1956 			NDFREE(&nd, NDF_ONLY_PNBUF);
1957 			vput(nd.ni_dvp);
1958 			if (vp == nd.ni_dvp)
1959 				vrele(vp);
1960 			else
1961 				vput(vp);
1962 			if ((error = vn_start_write(NULL, &mp,
1963 			    V_XSLEEP | PCATCH)) != 0) {
1964 				goto fdout;
1965 			}
1966 			goto restart;
1967 		}
1968 #ifdef MAC
1969 		error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
1970 		    &nd.ni_cnd);
1971 		if (error != 0)
1972 			goto out;
1973 #endif
1974 		vfs_notify_upper(vp, VFS_NOTIFY_UPPER_UNLINK);
1975 		error = VOP_REMOVE(nd.ni_dvp, vp, &nd.ni_cnd);
1976 #ifdef MAC
1977 out:
1978 #endif
1979 		vn_finished_write(mp);
1980 	}
1981 	NDFREE(&nd, NDF_ONLY_PNBUF);
1982 	vput(nd.ni_dvp);
1983 	if (vp == nd.ni_dvp)
1984 		vrele(vp);
1985 	else
1986 		vput(vp);
1987 	if (error == ERELOOKUP)
1988 		goto restart;
1989 fdout:
1990 	if (fp != NULL)
1991 		fdrop(fp, td);
1992 	return (error);
1993 }
1994 
1995 /*
1996  * Reposition read/write file offset.
1997  */
1998 #ifndef _SYS_SYSPROTO_H_
1999 struct lseek_args {
2000 	int	fd;
2001 	int	pad;
2002 	off_t	offset;
2003 	int	whence;
2004 };
2005 #endif
2006 int
2007 sys_lseek(struct thread *td, struct lseek_args *uap)
2008 {
2009 
2010 	return (kern_lseek(td, uap->fd, uap->offset, uap->whence));
2011 }
2012 
2013 int
2014 kern_lseek(struct thread *td, int fd, off_t offset, int whence)
2015 {
2016 	struct file *fp;
2017 	int error;
2018 
2019 	AUDIT_ARG_FD(fd);
2020 	error = fget(td, fd, &cap_seek_rights, &fp);
2021 	if (error != 0)
2022 		return (error);
2023 	error = (fp->f_ops->fo_flags & DFLAG_SEEKABLE) != 0 ?
2024 	    fo_seek(fp, offset, whence, td) : ESPIPE;
2025 	fdrop(fp, td);
2026 	return (error);
2027 }
2028 
2029 #if defined(COMPAT_43)
2030 /*
2031  * Reposition read/write file offset.
2032  */
2033 #ifndef _SYS_SYSPROTO_H_
2034 struct olseek_args {
2035 	int	fd;
2036 	long	offset;
2037 	int	whence;
2038 };
2039 #endif
2040 int
2041 olseek(struct thread *td, struct olseek_args *uap)
2042 {
2043 
2044 	return (kern_lseek(td, uap->fd, uap->offset, uap->whence));
2045 }
2046 #endif /* COMPAT_43 */
2047 
2048 #if defined(COMPAT_FREEBSD6)
2049 /* Version with the 'pad' argument */
2050 int
2051 freebsd6_lseek(struct thread *td, struct freebsd6_lseek_args *uap)
2052 {
2053 
2054 	return (kern_lseek(td, uap->fd, uap->offset, uap->whence));
2055 }
2056 #endif
2057 
2058 /*
2059  * Check access permissions using passed credentials.
2060  */
2061 static int
2062 vn_access(struct vnode *vp, int user_flags, struct ucred *cred,
2063      struct thread *td)
2064 {
2065 	accmode_t accmode;
2066 	int error;
2067 
2068 	/* Flags == 0 means only check for existence. */
2069 	if (user_flags == 0)
2070 		return (0);
2071 
2072 	accmode = 0;
2073 	if (user_flags & R_OK)
2074 		accmode |= VREAD;
2075 	if (user_flags & W_OK)
2076 		accmode |= VWRITE;
2077 	if (user_flags & X_OK)
2078 		accmode |= VEXEC;
2079 #ifdef MAC
2080 	error = mac_vnode_check_access(cred, vp, accmode);
2081 	if (error != 0)
2082 		return (error);
2083 #endif
2084 	if ((accmode & VWRITE) == 0 || (error = vn_writechk(vp)) == 0)
2085 		error = VOP_ACCESS(vp, accmode, cred, td);
2086 	return (error);
2087 }
2088 
2089 /*
2090  * Check access permissions using "real" credentials.
2091  */
2092 #ifndef _SYS_SYSPROTO_H_
2093 struct access_args {
2094 	char	*path;
2095 	int	amode;
2096 };
2097 #endif
2098 int
2099 sys_access(struct thread *td, struct access_args *uap)
2100 {
2101 
2102 	return (kern_accessat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2103 	    0, uap->amode));
2104 }
2105 
2106 #ifndef _SYS_SYSPROTO_H_
2107 struct faccessat_args {
2108 	int	dirfd;
2109 	char	*path;
2110 	int	amode;
2111 	int	flag;
2112 }
2113 #endif
2114 int
2115 sys_faccessat(struct thread *td, struct faccessat_args *uap)
2116 {
2117 
2118 	return (kern_accessat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
2119 	    uap->amode));
2120 }
2121 
2122 int
2123 kern_accessat(struct thread *td, int fd, const char *path,
2124     enum uio_seg pathseg, int flag, int amode)
2125 {
2126 	struct ucred *cred, *usecred;
2127 	struct vnode *vp;
2128 	struct nameidata nd;
2129 	int error;
2130 
2131 	if ((flag & ~(AT_EACCESS | AT_RESOLVE_BENEATH | AT_EMPTY_PATH)) != 0)
2132 		return (EINVAL);
2133 	if (amode != F_OK && (amode & ~(R_OK | W_OK | X_OK)) != 0)
2134 		return (EINVAL);
2135 
2136 	/*
2137 	 * Create and modify a temporary credential instead of one that
2138 	 * is potentially shared (if we need one).
2139 	 */
2140 	cred = td->td_ucred;
2141 	if ((flag & AT_EACCESS) == 0 &&
2142 	    ((cred->cr_uid != cred->cr_ruid ||
2143 	    cred->cr_rgid != cred->cr_groups[0]))) {
2144 		usecred = crdup(cred);
2145 		usecred->cr_uid = cred->cr_ruid;
2146 		usecred->cr_groups[0] = cred->cr_rgid;
2147 		td->td_ucred = usecred;
2148 	} else
2149 		usecred = cred;
2150 	AUDIT_ARG_VALUE(amode);
2151 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF |
2152 	    AUDITVNODE1 | at2cnpflags(flag, AT_RESOLVE_BENEATH |
2153 	    AT_EMPTY_PATH), pathseg, path, fd, &cap_fstat_rights, td);
2154 	if ((error = namei(&nd)) != 0)
2155 		goto out;
2156 	vp = nd.ni_vp;
2157 
2158 	error = vn_access(vp, amode, usecred, td);
2159 	NDFREE_NOTHING(&nd);
2160 	vput(vp);
2161 out:
2162 	if (usecred != cred) {
2163 		td->td_ucred = cred;
2164 		crfree(usecred);
2165 	}
2166 	return (error);
2167 }
2168 
2169 /*
2170  * Check access permissions using "effective" credentials.
2171  */
2172 #ifndef _SYS_SYSPROTO_H_
2173 struct eaccess_args {
2174 	char	*path;
2175 	int	amode;
2176 };
2177 #endif
2178 int
2179 sys_eaccess(struct thread *td, struct eaccess_args *uap)
2180 {
2181 
2182 	return (kern_accessat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2183 	    AT_EACCESS, uap->amode));
2184 }
2185 
2186 #if defined(COMPAT_43)
2187 /*
2188  * Get file status; this version follows links.
2189  */
2190 #ifndef _SYS_SYSPROTO_H_
2191 struct ostat_args {
2192 	char	*path;
2193 	struct ostat *ub;
2194 };
2195 #endif
2196 int
2197 ostat(struct thread *td, struct ostat_args *uap)
2198 {
2199 	struct stat sb;
2200 	struct ostat osb;
2201 	int error;
2202 
2203 	error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE,
2204 	    &sb, NULL);
2205 	if (error != 0)
2206 		return (error);
2207 	cvtstat(&sb, &osb);
2208 	return (copyout(&osb, uap->ub, sizeof (osb)));
2209 }
2210 
2211 /*
2212  * Get file status; this version does not follow links.
2213  */
2214 #ifndef _SYS_SYSPROTO_H_
2215 struct olstat_args {
2216 	char	*path;
2217 	struct ostat *ub;
2218 };
2219 #endif
2220 int
2221 olstat(struct thread *td, struct olstat_args *uap)
2222 {
2223 	struct stat sb;
2224 	struct ostat osb;
2225 	int error;
2226 
2227 	error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2228 	    UIO_USERSPACE, &sb, NULL);
2229 	if (error != 0)
2230 		return (error);
2231 	cvtstat(&sb, &osb);
2232 	return (copyout(&osb, uap->ub, sizeof (osb)));
2233 }
2234 
2235 /*
2236  * Convert from an old to a new stat structure.
2237  * XXX: many values are blindly truncated.
2238  */
2239 void
2240 cvtstat(struct stat *st, struct ostat *ost)
2241 {
2242 
2243 	bzero(ost, sizeof(*ost));
2244 	ost->st_dev = st->st_dev;
2245 	ost->st_ino = st->st_ino;
2246 	ost->st_mode = st->st_mode;
2247 	ost->st_nlink = st->st_nlink;
2248 	ost->st_uid = st->st_uid;
2249 	ost->st_gid = st->st_gid;
2250 	ost->st_rdev = st->st_rdev;
2251 	ost->st_size = MIN(st->st_size, INT32_MAX);
2252 	ost->st_atim = st->st_atim;
2253 	ost->st_mtim = st->st_mtim;
2254 	ost->st_ctim = st->st_ctim;
2255 	ost->st_blksize = st->st_blksize;
2256 	ost->st_blocks = st->st_blocks;
2257 	ost->st_flags = st->st_flags;
2258 	ost->st_gen = st->st_gen;
2259 }
2260 #endif /* COMPAT_43 */
2261 
2262 #if defined(COMPAT_43) || defined(COMPAT_FREEBSD11)
2263 int ino64_trunc_error;
2264 SYSCTL_INT(_vfs, OID_AUTO, ino64_trunc_error, CTLFLAG_RW,
2265     &ino64_trunc_error, 0,
2266     "Error on truncation of device, file or inode number, or link count");
2267 
2268 int
2269 freebsd11_cvtstat(struct stat *st, struct freebsd11_stat *ost)
2270 {
2271 
2272 	ost->st_dev = st->st_dev;
2273 	if (ost->st_dev != st->st_dev) {
2274 		switch (ino64_trunc_error) {
2275 		default:
2276 			/*
2277 			 * Since dev_t is almost raw, don't clamp to the
2278 			 * maximum for case 2, but ignore the error.
2279 			 */
2280 			break;
2281 		case 1:
2282 			return (EOVERFLOW);
2283 		}
2284 	}
2285 	ost->st_ino = st->st_ino;
2286 	if (ost->st_ino != st->st_ino) {
2287 		switch (ino64_trunc_error) {
2288 		default:
2289 		case 0:
2290 			break;
2291 		case 1:
2292 			return (EOVERFLOW);
2293 		case 2:
2294 			ost->st_ino = UINT32_MAX;
2295 			break;
2296 		}
2297 	}
2298 	ost->st_mode = st->st_mode;
2299 	ost->st_nlink = st->st_nlink;
2300 	if (ost->st_nlink != st->st_nlink) {
2301 		switch (ino64_trunc_error) {
2302 		default:
2303 		case 0:
2304 			break;
2305 		case 1:
2306 			return (EOVERFLOW);
2307 		case 2:
2308 			ost->st_nlink = UINT16_MAX;
2309 			break;
2310 		}
2311 	}
2312 	ost->st_uid = st->st_uid;
2313 	ost->st_gid = st->st_gid;
2314 	ost->st_rdev = st->st_rdev;
2315 	if (ost->st_rdev != st->st_rdev) {
2316 		switch (ino64_trunc_error) {
2317 		default:
2318 			break;
2319 		case 1:
2320 			return (EOVERFLOW);
2321 		}
2322 	}
2323 	ost->st_atim = st->st_atim;
2324 	ost->st_mtim = st->st_mtim;
2325 	ost->st_ctim = st->st_ctim;
2326 	ost->st_size = st->st_size;
2327 	ost->st_blocks = st->st_blocks;
2328 	ost->st_blksize = st->st_blksize;
2329 	ost->st_flags = st->st_flags;
2330 	ost->st_gen = st->st_gen;
2331 	ost->st_lspare = 0;
2332 	ost->st_birthtim = st->st_birthtim;
2333 	bzero((char *)&ost->st_birthtim + sizeof(ost->st_birthtim),
2334 	    sizeof(*ost) - offsetof(struct freebsd11_stat,
2335 	    st_birthtim) - sizeof(ost->st_birthtim));
2336 	return (0);
2337 }
2338 
2339 int
2340 freebsd11_stat(struct thread *td, struct freebsd11_stat_args* uap)
2341 {
2342 	struct stat sb;
2343 	struct freebsd11_stat osb;
2344 	int error;
2345 
2346 	error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE,
2347 	    &sb, NULL);
2348 	if (error != 0)
2349 		return (error);
2350 	error = freebsd11_cvtstat(&sb, &osb);
2351 	if (error == 0)
2352 		error = copyout(&osb, uap->ub, sizeof(osb));
2353 	return (error);
2354 }
2355 
2356 int
2357 freebsd11_lstat(struct thread *td, struct freebsd11_lstat_args* uap)
2358 {
2359 	struct stat sb;
2360 	struct freebsd11_stat osb;
2361 	int error;
2362 
2363 	error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2364 	    UIO_USERSPACE, &sb, NULL);
2365 	if (error != 0)
2366 		return (error);
2367 	error = freebsd11_cvtstat(&sb, &osb);
2368 	if (error == 0)
2369 		error = copyout(&osb, uap->ub, sizeof(osb));
2370 	return (error);
2371 }
2372 
2373 int
2374 freebsd11_fhstat(struct thread *td, struct freebsd11_fhstat_args* uap)
2375 {
2376 	struct fhandle fh;
2377 	struct stat sb;
2378 	struct freebsd11_stat osb;
2379 	int error;
2380 
2381 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
2382 	if (error != 0)
2383 		return (error);
2384 	error = kern_fhstat(td, fh, &sb);
2385 	if (error != 0)
2386 		return (error);
2387 	error = freebsd11_cvtstat(&sb, &osb);
2388 	if (error == 0)
2389 		error = copyout(&osb, uap->sb, sizeof(osb));
2390 	return (error);
2391 }
2392 
2393 int
2394 freebsd11_fstatat(struct thread *td, struct freebsd11_fstatat_args* uap)
2395 {
2396 	struct stat sb;
2397 	struct freebsd11_stat osb;
2398 	int error;
2399 
2400 	error = kern_statat(td, uap->flag, uap->fd, uap->path,
2401 	    UIO_USERSPACE, &sb, NULL);
2402 	if (error != 0)
2403 		return (error);
2404 	error = freebsd11_cvtstat(&sb, &osb);
2405 	if (error == 0)
2406 		error = copyout(&osb, uap->buf, sizeof(osb));
2407 	return (error);
2408 }
2409 #endif	/* COMPAT_FREEBSD11 */
2410 
2411 /*
2412  * Get file status
2413  */
2414 #ifndef _SYS_SYSPROTO_H_
2415 struct fstatat_args {
2416 	int	fd;
2417 	char	*path;
2418 	struct stat	*buf;
2419 	int	flag;
2420 }
2421 #endif
2422 int
2423 sys_fstatat(struct thread *td, struct fstatat_args *uap)
2424 {
2425 	struct stat sb;
2426 	int error;
2427 
2428 	error = kern_statat(td, uap->flag, uap->fd, uap->path,
2429 	    UIO_USERSPACE, &sb, NULL);
2430 	if (error == 0)
2431 		error = copyout(&sb, uap->buf, sizeof (sb));
2432 	return (error);
2433 }
2434 
2435 int
2436 kern_statat(struct thread *td, int flag, int fd, const char *path,
2437     enum uio_seg pathseg, struct stat *sbp,
2438     void (*hook)(struct vnode *vp, struct stat *sbp))
2439 {
2440 	struct nameidata nd;
2441 	int error;
2442 
2443 	if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH |
2444 	    AT_EMPTY_PATH)) != 0)
2445 		return (EINVAL);
2446 
2447 	NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(flag, AT_RESOLVE_BENEATH |
2448 	    AT_SYMLINK_NOFOLLOW | AT_EMPTY_PATH) | LOCKSHARED | LOCKLEAF |
2449 	    AUDITVNODE1, pathseg, path, fd, &cap_fstat_rights, td);
2450 
2451 	if ((error = namei(&nd)) != 0) {
2452 		if (error == ENOTDIR &&
2453 		    (nd.ni_resflags & NIRES_EMPTYPATH) != 0)
2454 			error = kern_fstat(td, fd, sbp);
2455 		return (error);
2456 	}
2457 	error = VOP_STAT(nd.ni_vp, sbp, td->td_ucred, NOCRED);
2458 	if (error == 0) {
2459 		if (__predict_false(hook != NULL))
2460 			hook(nd.ni_vp, sbp);
2461 	}
2462 	NDFREE_NOTHING(&nd);
2463 	vput(nd.ni_vp);
2464 #ifdef __STAT_TIME_T_EXT
2465 	sbp->st_atim_ext = 0;
2466 	sbp->st_mtim_ext = 0;
2467 	sbp->st_ctim_ext = 0;
2468 	sbp->st_btim_ext = 0;
2469 #endif
2470 #ifdef KTRACE
2471 	if (KTRPOINT(td, KTR_STRUCT))
2472 		ktrstat_error(sbp, error);
2473 #endif
2474 	return (error);
2475 }
2476 
2477 #if defined(COMPAT_FREEBSD11)
2478 /*
2479  * Implementation of the NetBSD [l]stat() functions.
2480  */
2481 int
2482 freebsd11_cvtnstat(struct stat *sb, struct nstat *nsb)
2483 {
2484 	struct freebsd11_stat sb11;
2485 	int error;
2486 
2487 	error = freebsd11_cvtstat(sb, &sb11);
2488 	if (error != 0)
2489 		return (error);
2490 
2491 	bzero(nsb, sizeof(*nsb));
2492 	CP(sb11, *nsb, st_dev);
2493 	CP(sb11, *nsb, st_ino);
2494 	CP(sb11, *nsb, st_mode);
2495 	CP(sb11, *nsb, st_nlink);
2496 	CP(sb11, *nsb, st_uid);
2497 	CP(sb11, *nsb, st_gid);
2498 	CP(sb11, *nsb, st_rdev);
2499 	CP(sb11, *nsb, st_atim);
2500 	CP(sb11, *nsb, st_mtim);
2501 	CP(sb11, *nsb, st_ctim);
2502 	CP(sb11, *nsb, st_size);
2503 	CP(sb11, *nsb, st_blocks);
2504 	CP(sb11, *nsb, st_blksize);
2505 	CP(sb11, *nsb, st_flags);
2506 	CP(sb11, *nsb, st_gen);
2507 	CP(sb11, *nsb, st_birthtim);
2508 	return (0);
2509 }
2510 
2511 #ifndef _SYS_SYSPROTO_H_
2512 struct freebsd11_nstat_args {
2513 	char	*path;
2514 	struct nstat *ub;
2515 };
2516 #endif
2517 int
2518 freebsd11_nstat(struct thread *td, struct freebsd11_nstat_args *uap)
2519 {
2520 	struct stat sb;
2521 	struct nstat nsb;
2522 	int error;
2523 
2524 	error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE,
2525 	    &sb, NULL);
2526 	if (error != 0)
2527 		return (error);
2528 	error = freebsd11_cvtnstat(&sb, &nsb);
2529 	if (error == 0)
2530 		error = copyout(&nsb, uap->ub, sizeof (nsb));
2531 	return (error);
2532 }
2533 
2534 /*
2535  * NetBSD lstat.  Get file status; this version does not follow links.
2536  */
2537 #ifndef _SYS_SYSPROTO_H_
2538 struct freebsd11_nlstat_args {
2539 	char	*path;
2540 	struct nstat *ub;
2541 };
2542 #endif
2543 int
2544 freebsd11_nlstat(struct thread *td, struct freebsd11_nlstat_args *uap)
2545 {
2546 	struct stat sb;
2547 	struct nstat nsb;
2548 	int error;
2549 
2550 	error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2551 	    UIO_USERSPACE, &sb, NULL);
2552 	if (error != 0)
2553 		return (error);
2554 	error = freebsd11_cvtnstat(&sb, &nsb);
2555 	if (error == 0)
2556 		error = copyout(&nsb, uap->ub, sizeof (nsb));
2557 	return (error);
2558 }
2559 #endif /* COMPAT_FREEBSD11 */
2560 
2561 /*
2562  * Get configurable pathname variables.
2563  */
2564 #ifndef _SYS_SYSPROTO_H_
2565 struct pathconf_args {
2566 	char	*path;
2567 	int	name;
2568 };
2569 #endif
2570 int
2571 sys_pathconf(struct thread *td, struct pathconf_args *uap)
2572 {
2573 	long value;
2574 	int error;
2575 
2576 	error = kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name, FOLLOW,
2577 	    &value);
2578 	if (error == 0)
2579 		td->td_retval[0] = value;
2580 	return (error);
2581 }
2582 
2583 #ifndef _SYS_SYSPROTO_H_
2584 struct lpathconf_args {
2585 	char	*path;
2586 	int	name;
2587 };
2588 #endif
2589 int
2590 sys_lpathconf(struct thread *td, struct lpathconf_args *uap)
2591 {
2592 	long value;
2593 	int error;
2594 
2595 	error = kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name,
2596 	    NOFOLLOW, &value);
2597 	if (error == 0)
2598 		td->td_retval[0] = value;
2599 	return (error);
2600 }
2601 
2602 int
2603 kern_pathconf(struct thread *td, const char *path, enum uio_seg pathseg,
2604     int name, u_long flags, long *valuep)
2605 {
2606 	struct nameidata nd;
2607 	int error;
2608 
2609 	NDINIT(&nd, LOOKUP, LOCKSHARED | LOCKLEAF | AUDITVNODE1 | flags,
2610 	    pathseg, path, td);
2611 	if ((error = namei(&nd)) != 0)
2612 		return (error);
2613 	NDFREE_NOTHING(&nd);
2614 
2615 	error = VOP_PATHCONF(nd.ni_vp, name, valuep);
2616 	vput(nd.ni_vp);
2617 	return (error);
2618 }
2619 
2620 /*
2621  * Return target name of a symbolic link.
2622  */
2623 #ifndef _SYS_SYSPROTO_H_
2624 struct readlink_args {
2625 	char	*path;
2626 	char	*buf;
2627 	size_t	count;
2628 };
2629 #endif
2630 int
2631 sys_readlink(struct thread *td, struct readlink_args *uap)
2632 {
2633 
2634 	return (kern_readlinkat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2635 	    uap->buf, UIO_USERSPACE, uap->count));
2636 }
2637 #ifndef _SYS_SYSPROTO_H_
2638 struct readlinkat_args {
2639 	int	fd;
2640 	char	*path;
2641 	char	*buf;
2642 	size_t	bufsize;
2643 };
2644 #endif
2645 int
2646 sys_readlinkat(struct thread *td, struct readlinkat_args *uap)
2647 {
2648 
2649 	return (kern_readlinkat(td, uap->fd, uap->path, UIO_USERSPACE,
2650 	    uap->buf, UIO_USERSPACE, uap->bufsize));
2651 }
2652 
2653 int
2654 kern_readlinkat(struct thread *td, int fd, const char *path,
2655     enum uio_seg pathseg, char *buf, enum uio_seg bufseg, size_t count)
2656 {
2657 	struct vnode *vp;
2658 	struct nameidata nd;
2659 	int error;
2660 
2661 	if (count > IOSIZE_MAX)
2662 		return (EINVAL);
2663 
2664 	NDINIT_AT(&nd, LOOKUP, NOFOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1 |
2665 	    EMPTYPATH, pathseg, path, fd, td);
2666 
2667 	if ((error = namei(&nd)) != 0)
2668 		return (error);
2669 	NDFREE_NOTHING(&nd);
2670 	vp = nd.ni_vp;
2671 
2672 	error = kern_readlink_vp(vp, buf, bufseg, count, td);
2673 	vput(vp);
2674 
2675 	return (error);
2676 }
2677 
2678 /*
2679  * Helper function to readlink from a vnode
2680  */
2681 static int
2682 kern_readlink_vp(struct vnode *vp, char *buf, enum uio_seg bufseg, size_t count,
2683     struct thread *td)
2684 {
2685 	struct iovec aiov;
2686 	struct uio auio;
2687 	int error;
2688 
2689 	ASSERT_VOP_LOCKED(vp, "kern_readlink_vp(): vp not locked");
2690 #ifdef MAC
2691 	error = mac_vnode_check_readlink(td->td_ucred, vp);
2692 	if (error != 0)
2693 		return (error);
2694 #endif
2695 	if (vp->v_type != VLNK && (vp->v_vflag & VV_READLINK) == 0)
2696 		return (EINVAL);
2697 
2698 	aiov.iov_base = buf;
2699 	aiov.iov_len = count;
2700 	auio.uio_iov = &aiov;
2701 	auio.uio_iovcnt = 1;
2702 	auio.uio_offset = 0;
2703 	auio.uio_rw = UIO_READ;
2704 	auio.uio_segflg = bufseg;
2705 	auio.uio_td = td;
2706 	auio.uio_resid = count;
2707 	error = VOP_READLINK(vp, &auio, td->td_ucred);
2708 	td->td_retval[0] = count - auio.uio_resid;
2709 	return (error);
2710 }
2711 
2712 /*
2713  * Common implementation code for chflags() and fchflags().
2714  */
2715 static int
2716 setfflags(struct thread *td, struct vnode *vp, u_long flags)
2717 {
2718 	struct mount *mp;
2719 	struct vattr vattr;
2720 	int error;
2721 
2722 	/* We can't support the value matching VNOVAL. */
2723 	if (flags == VNOVAL)
2724 		return (EOPNOTSUPP);
2725 
2726 	/*
2727 	 * Prevent non-root users from setting flags on devices.  When
2728 	 * a device is reused, users can retain ownership of the device
2729 	 * if they are allowed to set flags and programs assume that
2730 	 * chown can't fail when done as root.
2731 	 */
2732 	if (vp->v_type == VCHR || vp->v_type == VBLK) {
2733 		error = priv_check(td, PRIV_VFS_CHFLAGS_DEV);
2734 		if (error != 0)
2735 			return (error);
2736 	}
2737 
2738 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2739 		return (error);
2740 	VATTR_NULL(&vattr);
2741 	vattr.va_flags = flags;
2742 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2743 #ifdef MAC
2744 	error = mac_vnode_check_setflags(td->td_ucred, vp, vattr.va_flags);
2745 	if (error == 0)
2746 #endif
2747 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
2748 	VOP_UNLOCK(vp);
2749 	vn_finished_write(mp);
2750 	return (error);
2751 }
2752 
2753 /*
2754  * Change flags of a file given a path name.
2755  */
2756 #ifndef _SYS_SYSPROTO_H_
2757 struct chflags_args {
2758 	const char *path;
2759 	u_long	flags;
2760 };
2761 #endif
2762 int
2763 sys_chflags(struct thread *td, struct chflags_args *uap)
2764 {
2765 
2766 	return (kern_chflagsat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2767 	    uap->flags, 0));
2768 }
2769 
2770 #ifndef _SYS_SYSPROTO_H_
2771 struct chflagsat_args {
2772 	int	fd;
2773 	const char *path;
2774 	u_long	flags;
2775 	int	atflag;
2776 }
2777 #endif
2778 int
2779 sys_chflagsat(struct thread *td, struct chflagsat_args *uap)
2780 {
2781 
2782 	return (kern_chflagsat(td, uap->fd, uap->path, UIO_USERSPACE,
2783 	    uap->flags, uap->atflag));
2784 }
2785 
2786 /*
2787  * Same as chflags() but doesn't follow symlinks.
2788  */
2789 #ifndef _SYS_SYSPROTO_H_
2790 struct lchflags_args {
2791 	const char *path;
2792 	u_long flags;
2793 };
2794 #endif
2795 int
2796 sys_lchflags(struct thread *td, struct lchflags_args *uap)
2797 {
2798 
2799 	return (kern_chflagsat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2800 	    uap->flags, AT_SYMLINK_NOFOLLOW));
2801 }
2802 
2803 static int
2804 kern_chflagsat(struct thread *td, int fd, const char *path,
2805     enum uio_seg pathseg, u_long flags, int atflag)
2806 {
2807 	struct nameidata nd;
2808 	int error;
2809 
2810 	if ((atflag & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH |
2811 	    AT_EMPTY_PATH)) != 0)
2812 		return (EINVAL);
2813 
2814 	AUDIT_ARG_FFLAGS(flags);
2815 	NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(atflag, AT_SYMLINK_NOFOLLOW |
2816 	    AT_RESOLVE_BENEATH | AT_EMPTY_PATH) | AUDITVNODE1, pathseg, path,
2817 	    fd, &cap_fchflags_rights, td);
2818 	if ((error = namei(&nd)) != 0)
2819 		return (error);
2820 	NDFREE_NOTHING(&nd);
2821 	error = setfflags(td, nd.ni_vp, flags);
2822 	vrele(nd.ni_vp);
2823 	return (error);
2824 }
2825 
2826 /*
2827  * Change flags of a file given a file descriptor.
2828  */
2829 #ifndef _SYS_SYSPROTO_H_
2830 struct fchflags_args {
2831 	int	fd;
2832 	u_long	flags;
2833 };
2834 #endif
2835 int
2836 sys_fchflags(struct thread *td, struct fchflags_args *uap)
2837 {
2838 	struct file *fp;
2839 	int error;
2840 
2841 	AUDIT_ARG_FD(uap->fd);
2842 	AUDIT_ARG_FFLAGS(uap->flags);
2843 	error = getvnode(td, uap->fd, &cap_fchflags_rights,
2844 	    &fp);
2845 	if (error != 0)
2846 		return (error);
2847 #ifdef AUDIT
2848 	if (AUDITING_TD(td)) {
2849 		vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
2850 		AUDIT_ARG_VNODE1(fp->f_vnode);
2851 		VOP_UNLOCK(fp->f_vnode);
2852 	}
2853 #endif
2854 	error = setfflags(td, fp->f_vnode, uap->flags);
2855 	fdrop(fp, td);
2856 	return (error);
2857 }
2858 
2859 /*
2860  * Common implementation code for chmod(), lchmod() and fchmod().
2861  */
2862 int
2863 setfmode(struct thread *td, struct ucred *cred, struct vnode *vp, int mode)
2864 {
2865 	struct mount *mp;
2866 	struct vattr vattr;
2867 	int error;
2868 
2869 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2870 		return (error);
2871 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2872 	VATTR_NULL(&vattr);
2873 	vattr.va_mode = mode & ALLPERMS;
2874 #ifdef MAC
2875 	error = mac_vnode_check_setmode(cred, vp, vattr.va_mode);
2876 	if (error == 0)
2877 #endif
2878 		error = VOP_SETATTR(vp, &vattr, cred);
2879 	VOP_UNLOCK(vp);
2880 	vn_finished_write(mp);
2881 	return (error);
2882 }
2883 
2884 /*
2885  * Change mode of a file given path name.
2886  */
2887 #ifndef _SYS_SYSPROTO_H_
2888 struct chmod_args {
2889 	char	*path;
2890 	int	mode;
2891 };
2892 #endif
2893 int
2894 sys_chmod(struct thread *td, struct chmod_args *uap)
2895 {
2896 
2897 	return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2898 	    uap->mode, 0));
2899 }
2900 
2901 #ifndef _SYS_SYSPROTO_H_
2902 struct fchmodat_args {
2903 	int	dirfd;
2904 	char	*path;
2905 	mode_t	mode;
2906 	int	flag;
2907 }
2908 #endif
2909 int
2910 sys_fchmodat(struct thread *td, struct fchmodat_args *uap)
2911 {
2912 
2913 	return (kern_fchmodat(td, uap->fd, uap->path, UIO_USERSPACE,
2914 	    uap->mode, uap->flag));
2915 }
2916 
2917 /*
2918  * Change mode of a file given path name (don't follow links.)
2919  */
2920 #ifndef _SYS_SYSPROTO_H_
2921 struct lchmod_args {
2922 	char	*path;
2923 	int	mode;
2924 };
2925 #endif
2926 int
2927 sys_lchmod(struct thread *td, struct lchmod_args *uap)
2928 {
2929 
2930 	return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2931 	    uap->mode, AT_SYMLINK_NOFOLLOW));
2932 }
2933 
2934 int
2935 kern_fchmodat(struct thread *td, int fd, const char *path,
2936     enum uio_seg pathseg, mode_t mode, int flag)
2937 {
2938 	struct nameidata nd;
2939 	int error;
2940 
2941 	if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH |
2942 	    AT_EMPTY_PATH)) != 0)
2943 		return (EINVAL);
2944 
2945 	AUDIT_ARG_MODE(mode);
2946 	NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(flag, AT_SYMLINK_NOFOLLOW |
2947 	    AT_RESOLVE_BENEATH | AT_EMPTY_PATH) | AUDITVNODE1, pathseg, path,
2948 	    fd, &cap_fchmod_rights, td);
2949 	if ((error = namei(&nd)) != 0)
2950 		return (error);
2951 	NDFREE_NOTHING(&nd);
2952 	error = setfmode(td, td->td_ucred, nd.ni_vp, mode);
2953 	vrele(nd.ni_vp);
2954 	return (error);
2955 }
2956 
2957 /*
2958  * Change mode of a file given a file descriptor.
2959  */
2960 #ifndef _SYS_SYSPROTO_H_
2961 struct fchmod_args {
2962 	int	fd;
2963 	int	mode;
2964 };
2965 #endif
2966 int
2967 sys_fchmod(struct thread *td, struct fchmod_args *uap)
2968 {
2969 	struct file *fp;
2970 	int error;
2971 
2972 	AUDIT_ARG_FD(uap->fd);
2973 	AUDIT_ARG_MODE(uap->mode);
2974 
2975 	error = fget(td, uap->fd, &cap_fchmod_rights, &fp);
2976 	if (error != 0)
2977 		return (error);
2978 	error = fo_chmod(fp, uap->mode, td->td_ucred, td);
2979 	fdrop(fp, td);
2980 	return (error);
2981 }
2982 
2983 /*
2984  * Common implementation for chown(), lchown(), and fchown()
2985  */
2986 int
2987 setfown(struct thread *td, struct ucred *cred, struct vnode *vp, uid_t uid,
2988     gid_t gid)
2989 {
2990 	struct mount *mp;
2991 	struct vattr vattr;
2992 	int error;
2993 
2994 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2995 		return (error);
2996 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2997 	VATTR_NULL(&vattr);
2998 	vattr.va_uid = uid;
2999 	vattr.va_gid = gid;
3000 #ifdef MAC
3001 	error = mac_vnode_check_setowner(cred, vp, vattr.va_uid,
3002 	    vattr.va_gid);
3003 	if (error == 0)
3004 #endif
3005 		error = VOP_SETATTR(vp, &vattr, cred);
3006 	VOP_UNLOCK(vp);
3007 	vn_finished_write(mp);
3008 	return (error);
3009 }
3010 
3011 /*
3012  * Set ownership given a path name.
3013  */
3014 #ifndef _SYS_SYSPROTO_H_
3015 struct chown_args {
3016 	char	*path;
3017 	int	uid;
3018 	int	gid;
3019 };
3020 #endif
3021 int
3022 sys_chown(struct thread *td, struct chown_args *uap)
3023 {
3024 
3025 	return (kern_fchownat(td, AT_FDCWD, uap->path, UIO_USERSPACE, uap->uid,
3026 	    uap->gid, 0));
3027 }
3028 
3029 #ifndef _SYS_SYSPROTO_H_
3030 struct fchownat_args {
3031 	int fd;
3032 	const char * path;
3033 	uid_t uid;
3034 	gid_t gid;
3035 	int flag;
3036 };
3037 #endif
3038 int
3039 sys_fchownat(struct thread *td, struct fchownat_args *uap)
3040 {
3041 
3042 	return (kern_fchownat(td, uap->fd, uap->path, UIO_USERSPACE, uap->uid,
3043 	    uap->gid, uap->flag));
3044 }
3045 
3046 int
3047 kern_fchownat(struct thread *td, int fd, const char *path,
3048     enum uio_seg pathseg, int uid, int gid, int flag)
3049 {
3050 	struct nameidata nd;
3051 	int error;
3052 
3053 	if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH |
3054 	    AT_EMPTY_PATH)) != 0)
3055 		return (EINVAL);
3056 
3057 	AUDIT_ARG_OWNER(uid, gid);
3058 	NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(flag, AT_SYMLINK_NOFOLLOW |
3059 	    AT_RESOLVE_BENEATH | AT_EMPTY_PATH) | AUDITVNODE1, pathseg, path,
3060 	    fd, &cap_fchown_rights, td);
3061 
3062 	if ((error = namei(&nd)) != 0)
3063 		return (error);
3064 	NDFREE_NOTHING(&nd);
3065 	error = setfown(td, td->td_ucred, nd.ni_vp, uid, gid);
3066 	vrele(nd.ni_vp);
3067 	return (error);
3068 }
3069 
3070 /*
3071  * Set ownership given a path name, do not cross symlinks.
3072  */
3073 #ifndef _SYS_SYSPROTO_H_
3074 struct lchown_args {
3075 	char	*path;
3076 	int	uid;
3077 	int	gid;
3078 };
3079 #endif
3080 int
3081 sys_lchown(struct thread *td, struct lchown_args *uap)
3082 {
3083 
3084 	return (kern_fchownat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
3085 	    uap->uid, uap->gid, AT_SYMLINK_NOFOLLOW));
3086 }
3087 
3088 /*
3089  * Set ownership given a file descriptor.
3090  */
3091 #ifndef _SYS_SYSPROTO_H_
3092 struct fchown_args {
3093 	int	fd;
3094 	int	uid;
3095 	int	gid;
3096 };
3097 #endif
3098 int
3099 sys_fchown(struct thread *td, struct fchown_args *uap)
3100 {
3101 	struct file *fp;
3102 	int error;
3103 
3104 	AUDIT_ARG_FD(uap->fd);
3105 	AUDIT_ARG_OWNER(uap->uid, uap->gid);
3106 	error = fget(td, uap->fd, &cap_fchown_rights, &fp);
3107 	if (error != 0)
3108 		return (error);
3109 	error = fo_chown(fp, uap->uid, uap->gid, td->td_ucred, td);
3110 	fdrop(fp, td);
3111 	return (error);
3112 }
3113 
3114 /*
3115  * Common implementation code for utimes(), lutimes(), and futimes().
3116  */
3117 static int
3118 getutimes(const struct timeval *usrtvp, enum uio_seg tvpseg,
3119     struct timespec *tsp)
3120 {
3121 	struct timeval tv[2];
3122 	const struct timeval *tvp;
3123 	int error;
3124 
3125 	if (usrtvp == NULL) {
3126 		vfs_timestamp(&tsp[0]);
3127 		tsp[1] = tsp[0];
3128 	} else {
3129 		if (tvpseg == UIO_SYSSPACE) {
3130 			tvp = usrtvp;
3131 		} else {
3132 			if ((error = copyin(usrtvp, tv, sizeof(tv))) != 0)
3133 				return (error);
3134 			tvp = tv;
3135 		}
3136 
3137 		if (tvp[0].tv_usec < 0 || tvp[0].tv_usec >= 1000000 ||
3138 		    tvp[1].tv_usec < 0 || tvp[1].tv_usec >= 1000000)
3139 			return (EINVAL);
3140 		TIMEVAL_TO_TIMESPEC(&tvp[0], &tsp[0]);
3141 		TIMEVAL_TO_TIMESPEC(&tvp[1], &tsp[1]);
3142 	}
3143 	return (0);
3144 }
3145 
3146 /*
3147  * Common implementation code for futimens(), utimensat().
3148  */
3149 #define	UTIMENS_NULL	0x1
3150 #define	UTIMENS_EXIT	0x2
3151 static int
3152 getutimens(const struct timespec *usrtsp, enum uio_seg tspseg,
3153     struct timespec *tsp, int *retflags)
3154 {
3155 	struct timespec tsnow;
3156 	int error;
3157 
3158 	vfs_timestamp(&tsnow);
3159 	*retflags = 0;
3160 	if (usrtsp == NULL) {
3161 		tsp[0] = tsnow;
3162 		tsp[1] = tsnow;
3163 		*retflags |= UTIMENS_NULL;
3164 		return (0);
3165 	}
3166 	if (tspseg == UIO_SYSSPACE) {
3167 		tsp[0] = usrtsp[0];
3168 		tsp[1] = usrtsp[1];
3169 	} else if ((error = copyin(usrtsp, tsp, sizeof(*tsp) * 2)) != 0)
3170 		return (error);
3171 	if (tsp[0].tv_nsec == UTIME_OMIT && tsp[1].tv_nsec == UTIME_OMIT)
3172 		*retflags |= UTIMENS_EXIT;
3173 	if (tsp[0].tv_nsec == UTIME_NOW && tsp[1].tv_nsec == UTIME_NOW)
3174 		*retflags |= UTIMENS_NULL;
3175 	if (tsp[0].tv_nsec == UTIME_OMIT)
3176 		tsp[0].tv_sec = VNOVAL;
3177 	else if (tsp[0].tv_nsec == UTIME_NOW)
3178 		tsp[0] = tsnow;
3179 	else if (tsp[0].tv_nsec < 0 || tsp[0].tv_nsec >= 1000000000L)
3180 		return (EINVAL);
3181 	if (tsp[1].tv_nsec == UTIME_OMIT)
3182 		tsp[1].tv_sec = VNOVAL;
3183 	else if (tsp[1].tv_nsec == UTIME_NOW)
3184 		tsp[1] = tsnow;
3185 	else if (tsp[1].tv_nsec < 0 || tsp[1].tv_nsec >= 1000000000L)
3186 		return (EINVAL);
3187 
3188 	return (0);
3189 }
3190 
3191 /*
3192  * Common implementation code for utimes(), lutimes(), futimes(), futimens(),
3193  * and utimensat().
3194  */
3195 static int
3196 setutimes(struct thread *td, struct vnode *vp, const struct timespec *ts,
3197     int numtimes, int nullflag)
3198 {
3199 	struct mount *mp;
3200 	struct vattr vattr;
3201 	int error;
3202 	bool setbirthtime;
3203 
3204 	setbirthtime = false;
3205 	vattr.va_birthtime.tv_sec = VNOVAL;
3206 	vattr.va_birthtime.tv_nsec = 0;
3207 
3208 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
3209 		return (error);
3210 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3211 	if (numtimes < 3 && VOP_GETATTR(vp, &vattr, td->td_ucred) == 0 &&
3212 	    timespeccmp(&ts[1], &vattr.va_birthtime, < ))
3213 		setbirthtime = true;
3214 	VATTR_NULL(&vattr);
3215 	vattr.va_atime = ts[0];
3216 	vattr.va_mtime = ts[1];
3217 	if (setbirthtime)
3218 		vattr.va_birthtime = ts[1];
3219 	if (numtimes > 2)
3220 		vattr.va_birthtime = ts[2];
3221 	if (nullflag)
3222 		vattr.va_vaflags |= VA_UTIMES_NULL;
3223 #ifdef MAC
3224 	error = mac_vnode_check_setutimes(td->td_ucred, vp, vattr.va_atime,
3225 	    vattr.va_mtime);
3226 #endif
3227 	if (error == 0)
3228 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3229 	VOP_UNLOCK(vp);
3230 	vn_finished_write(mp);
3231 	return (error);
3232 }
3233 
3234 /*
3235  * Set the access and modification times of a file.
3236  */
3237 #ifndef _SYS_SYSPROTO_H_
3238 struct utimes_args {
3239 	char	*path;
3240 	struct	timeval *tptr;
3241 };
3242 #endif
3243 int
3244 sys_utimes(struct thread *td, struct utimes_args *uap)
3245 {
3246 
3247 	return (kern_utimesat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
3248 	    uap->tptr, UIO_USERSPACE));
3249 }
3250 
3251 #ifndef _SYS_SYSPROTO_H_
3252 struct futimesat_args {
3253 	int fd;
3254 	const char * path;
3255 	const struct timeval * times;
3256 };
3257 #endif
3258 int
3259 sys_futimesat(struct thread *td, struct futimesat_args *uap)
3260 {
3261 
3262 	return (kern_utimesat(td, uap->fd, uap->path, UIO_USERSPACE,
3263 	    uap->times, UIO_USERSPACE));
3264 }
3265 
3266 int
3267 kern_utimesat(struct thread *td, int fd, const char *path,
3268     enum uio_seg pathseg, const struct timeval *tptr, enum uio_seg tptrseg)
3269 {
3270 	struct nameidata nd;
3271 	struct timespec ts[2];
3272 	int error;
3273 
3274 	if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3275 		return (error);
3276 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, fd,
3277 	    &cap_futimes_rights, td);
3278 
3279 	if ((error = namei(&nd)) != 0)
3280 		return (error);
3281 	NDFREE_NOTHING(&nd);
3282 	error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3283 	vrele(nd.ni_vp);
3284 	return (error);
3285 }
3286 
3287 /*
3288  * Set the access and modification times of a file.
3289  */
3290 #ifndef _SYS_SYSPROTO_H_
3291 struct lutimes_args {
3292 	char	*path;
3293 	struct	timeval *tptr;
3294 };
3295 #endif
3296 int
3297 sys_lutimes(struct thread *td, struct lutimes_args *uap)
3298 {
3299 
3300 	return (kern_lutimes(td, uap->path, UIO_USERSPACE, uap->tptr,
3301 	    UIO_USERSPACE));
3302 }
3303 
3304 int
3305 kern_lutimes(struct thread *td, const char *path, enum uio_seg pathseg,
3306     const struct timeval *tptr, enum uio_seg tptrseg)
3307 {
3308 	struct timespec ts[2];
3309 	struct nameidata nd;
3310 	int error;
3311 
3312 	if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3313 		return (error);
3314 	NDINIT(&nd, LOOKUP, NOFOLLOW | AUDITVNODE1, pathseg, path, td);
3315 	if ((error = namei(&nd)) != 0)
3316 		return (error);
3317 	NDFREE_NOTHING(&nd);
3318 	error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3319 	vrele(nd.ni_vp);
3320 	return (error);
3321 }
3322 
3323 /*
3324  * Set the access and modification times of a file.
3325  */
3326 #ifndef _SYS_SYSPROTO_H_
3327 struct futimes_args {
3328 	int	fd;
3329 	struct	timeval *tptr;
3330 };
3331 #endif
3332 int
3333 sys_futimes(struct thread *td, struct futimes_args *uap)
3334 {
3335 
3336 	return (kern_futimes(td, uap->fd, uap->tptr, UIO_USERSPACE));
3337 }
3338 
3339 int
3340 kern_futimes(struct thread *td, int fd, const struct timeval *tptr,
3341     enum uio_seg tptrseg)
3342 {
3343 	struct timespec ts[2];
3344 	struct file *fp;
3345 	int error;
3346 
3347 	AUDIT_ARG_FD(fd);
3348 	error = getutimes(tptr, tptrseg, ts);
3349 	if (error != 0)
3350 		return (error);
3351 	error = getvnode(td, fd, &cap_futimes_rights, &fp);
3352 	if (error != 0)
3353 		return (error);
3354 #ifdef AUDIT
3355 	if (AUDITING_TD(td)) {
3356 		vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
3357 		AUDIT_ARG_VNODE1(fp->f_vnode);
3358 		VOP_UNLOCK(fp->f_vnode);
3359 	}
3360 #endif
3361 	error = setutimes(td, fp->f_vnode, ts, 2, tptr == NULL);
3362 	fdrop(fp, td);
3363 	return (error);
3364 }
3365 
3366 int
3367 sys_futimens(struct thread *td, struct futimens_args *uap)
3368 {
3369 
3370 	return (kern_futimens(td, uap->fd, uap->times, UIO_USERSPACE));
3371 }
3372 
3373 int
3374 kern_futimens(struct thread *td, int fd, const struct timespec *tptr,
3375     enum uio_seg tptrseg)
3376 {
3377 	struct timespec ts[2];
3378 	struct file *fp;
3379 	int error, flags;
3380 
3381 	AUDIT_ARG_FD(fd);
3382 	error = getutimens(tptr, tptrseg, ts, &flags);
3383 	if (error != 0)
3384 		return (error);
3385 	if (flags & UTIMENS_EXIT)
3386 		return (0);
3387 	error = getvnode(td, fd, &cap_futimes_rights, &fp);
3388 	if (error != 0)
3389 		return (error);
3390 #ifdef AUDIT
3391 	if (AUDITING_TD(td)) {
3392 		vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
3393 		AUDIT_ARG_VNODE1(fp->f_vnode);
3394 		VOP_UNLOCK(fp->f_vnode);
3395 	}
3396 #endif
3397 	error = setutimes(td, fp->f_vnode, ts, 2, flags & UTIMENS_NULL);
3398 	fdrop(fp, td);
3399 	return (error);
3400 }
3401 
3402 int
3403 sys_utimensat(struct thread *td, struct utimensat_args *uap)
3404 {
3405 
3406 	return (kern_utimensat(td, uap->fd, uap->path, UIO_USERSPACE,
3407 	    uap->times, UIO_USERSPACE, uap->flag));
3408 }
3409 
3410 int
3411 kern_utimensat(struct thread *td, int fd, const char *path,
3412     enum uio_seg pathseg, const struct timespec *tptr, enum uio_seg tptrseg,
3413     int flag)
3414 {
3415 	struct nameidata nd;
3416 	struct timespec ts[2];
3417 	int error, flags;
3418 
3419 	if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH |
3420 	    AT_EMPTY_PATH)) != 0)
3421 		return (EINVAL);
3422 
3423 	if ((error = getutimens(tptr, tptrseg, ts, &flags)) != 0)
3424 		return (error);
3425 	NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(flag, AT_SYMLINK_NOFOLLOW |
3426 	    AT_RESOLVE_BENEATH | AT_EMPTY_PATH) | AUDITVNODE1,
3427 	    pathseg, path, fd, &cap_futimes_rights, td);
3428 	if ((error = namei(&nd)) != 0)
3429 		return (error);
3430 	/*
3431 	 * We are allowed to call namei() regardless of 2xUTIME_OMIT.
3432 	 * POSIX states:
3433 	 * "If both tv_nsec fields are UTIME_OMIT... EACCESS may be detected."
3434 	 * "Search permission is denied by a component of the path prefix."
3435 	 */
3436 	NDFREE_NOTHING(&nd);
3437 	if ((flags & UTIMENS_EXIT) == 0)
3438 		error = setutimes(td, nd.ni_vp, ts, 2, flags & UTIMENS_NULL);
3439 	vrele(nd.ni_vp);
3440 	return (error);
3441 }
3442 
3443 /*
3444  * Truncate a file given its path name.
3445  */
3446 #ifndef _SYS_SYSPROTO_H_
3447 struct truncate_args {
3448 	char	*path;
3449 	int	pad;
3450 	off_t	length;
3451 };
3452 #endif
3453 int
3454 sys_truncate(struct thread *td, struct truncate_args *uap)
3455 {
3456 
3457 	return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3458 }
3459 
3460 int
3461 kern_truncate(struct thread *td, const char *path, enum uio_seg pathseg,
3462     off_t length)
3463 {
3464 	struct mount *mp;
3465 	struct vnode *vp;
3466 	void *rl_cookie;
3467 	struct vattr vattr;
3468 	struct nameidata nd;
3469 	int error;
3470 
3471 	if (length < 0)
3472 		return (EINVAL);
3473 	NDPREINIT(&nd);
3474 retry:
3475 	NDINIT(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, td);
3476 	if ((error = namei(&nd)) != 0)
3477 		return (error);
3478 	vp = nd.ni_vp;
3479 	rl_cookie = vn_rangelock_wlock(vp, 0, OFF_MAX);
3480 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) {
3481 		vn_rangelock_unlock(vp, rl_cookie);
3482 		vrele(vp);
3483 		return (error);
3484 	}
3485 	NDFREE(&nd, NDF_ONLY_PNBUF);
3486 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3487 	if (vp->v_type == VDIR)
3488 		error = EISDIR;
3489 #ifdef MAC
3490 	else if ((error = mac_vnode_check_write(td->td_ucred, NOCRED, vp))) {
3491 	}
3492 #endif
3493 	else if ((error = vn_writechk(vp)) == 0 &&
3494 	    (error = VOP_ACCESS(vp, VWRITE, td->td_ucred, td)) == 0) {
3495 		VATTR_NULL(&vattr);
3496 		vattr.va_size = length;
3497 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3498 	}
3499 	VOP_UNLOCK(vp);
3500 	vn_finished_write(mp);
3501 	vn_rangelock_unlock(vp, rl_cookie);
3502 	vrele(vp);
3503 	if (error == ERELOOKUP)
3504 		goto retry;
3505 	return (error);
3506 }
3507 
3508 #if defined(COMPAT_43)
3509 /*
3510  * Truncate a file given its path name.
3511  */
3512 #ifndef _SYS_SYSPROTO_H_
3513 struct otruncate_args {
3514 	char	*path;
3515 	long	length;
3516 };
3517 #endif
3518 int
3519 otruncate(struct thread *td, struct otruncate_args *uap)
3520 {
3521 
3522 	return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3523 }
3524 #endif /* COMPAT_43 */
3525 
3526 #if defined(COMPAT_FREEBSD6)
3527 /* Versions with the pad argument */
3528 int
3529 freebsd6_truncate(struct thread *td, struct freebsd6_truncate_args *uap)
3530 {
3531 
3532 	return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3533 }
3534 
3535 int
3536 freebsd6_ftruncate(struct thread *td, struct freebsd6_ftruncate_args *uap)
3537 {
3538 
3539 	return (kern_ftruncate(td, uap->fd, uap->length));
3540 }
3541 #endif
3542 
3543 int
3544 kern_fsync(struct thread *td, int fd, bool fullsync)
3545 {
3546 	struct vnode *vp;
3547 	struct mount *mp;
3548 	struct file *fp;
3549 	int error;
3550 
3551 	AUDIT_ARG_FD(fd);
3552 	error = getvnode(td, fd, &cap_fsync_rights, &fp);
3553 	if (error != 0)
3554 		return (error);
3555 	vp = fp->f_vnode;
3556 #if 0
3557 	if (!fullsync)
3558 		/* XXXKIB: compete outstanding aio writes */;
3559 #endif
3560 retry:
3561 	error = vn_start_write(vp, &mp, V_WAIT | PCATCH);
3562 	if (error != 0)
3563 		goto drop;
3564 	vn_lock(vp, vn_lktype_write(mp, vp) | LK_RETRY);
3565 	AUDIT_ARG_VNODE1(vp);
3566 	if (vp->v_object != NULL) {
3567 		VM_OBJECT_WLOCK(vp->v_object);
3568 		vm_object_page_clean(vp->v_object, 0, 0, 0);
3569 		VM_OBJECT_WUNLOCK(vp->v_object);
3570 	}
3571 	error = fullsync ? VOP_FSYNC(vp, MNT_WAIT, td) : VOP_FDATASYNC(vp, td);
3572 	VOP_UNLOCK(vp);
3573 	vn_finished_write(mp);
3574 	if (error == ERELOOKUP)
3575 		goto retry;
3576 drop:
3577 	fdrop(fp, td);
3578 	return (error);
3579 }
3580 
3581 /*
3582  * Sync an open file.
3583  */
3584 #ifndef _SYS_SYSPROTO_H_
3585 struct fsync_args {
3586 	int	fd;
3587 };
3588 #endif
3589 int
3590 sys_fsync(struct thread *td, struct fsync_args *uap)
3591 {
3592 
3593 	return (kern_fsync(td, uap->fd, true));
3594 }
3595 
3596 int
3597 sys_fdatasync(struct thread *td, struct fdatasync_args *uap)
3598 {
3599 
3600 	return (kern_fsync(td, uap->fd, false));
3601 }
3602 
3603 /*
3604  * Rename files.  Source and destination must either both be directories, or
3605  * both not be directories.  If target is a directory, it must be empty.
3606  */
3607 #ifndef _SYS_SYSPROTO_H_
3608 struct rename_args {
3609 	char	*from;
3610 	char	*to;
3611 };
3612 #endif
3613 int
3614 sys_rename(struct thread *td, struct rename_args *uap)
3615 {
3616 
3617 	return (kern_renameat(td, AT_FDCWD, uap->from, AT_FDCWD,
3618 	    uap->to, UIO_USERSPACE));
3619 }
3620 
3621 #ifndef _SYS_SYSPROTO_H_
3622 struct renameat_args {
3623 	int	oldfd;
3624 	char	*old;
3625 	int	newfd;
3626 	char	*new;
3627 };
3628 #endif
3629 int
3630 sys_renameat(struct thread *td, struct renameat_args *uap)
3631 {
3632 
3633 	return (kern_renameat(td, uap->oldfd, uap->old, uap->newfd, uap->new,
3634 	    UIO_USERSPACE));
3635 }
3636 
3637 #ifdef MAC
3638 static int
3639 kern_renameat_mac(struct thread *td, int oldfd, const char *old, int newfd,
3640     const char *new, enum uio_seg pathseg, struct nameidata *fromnd)
3641 {
3642 	int error;
3643 
3644 	NDINIT_ATRIGHTS(fromnd, DELETE, LOCKPARENT | LOCKLEAF | SAVESTART |
3645 	    AUDITVNODE1, pathseg, old, oldfd, &cap_renameat_source_rights, td);
3646 	if ((error = namei(fromnd)) != 0)
3647 		return (error);
3648 	error = mac_vnode_check_rename_from(td->td_ucred, fromnd->ni_dvp,
3649 	    fromnd->ni_vp, &fromnd->ni_cnd);
3650 	VOP_UNLOCK(fromnd->ni_dvp);
3651 	if (fromnd->ni_dvp != fromnd->ni_vp)
3652 		VOP_UNLOCK(fromnd->ni_vp);
3653 	if (error != 0) {
3654 		NDFREE(fromnd, NDF_ONLY_PNBUF);
3655 		vrele(fromnd->ni_dvp);
3656 		vrele(fromnd->ni_vp);
3657 		if (fromnd->ni_startdir)
3658 			vrele(fromnd->ni_startdir);
3659 	}
3660 	return (error);
3661 }
3662 #endif
3663 
3664 int
3665 kern_renameat(struct thread *td, int oldfd, const char *old, int newfd,
3666     const char *new, enum uio_seg pathseg)
3667 {
3668 	struct mount *mp = NULL;
3669 	struct vnode *tvp, *fvp, *tdvp;
3670 	struct nameidata fromnd, tond;
3671 	u_int64_t tondflags;
3672 	int error;
3673 
3674 again:
3675 	bwillwrite();
3676 #ifdef MAC
3677 	if (mac_vnode_check_rename_from_enabled()) {
3678 		error = kern_renameat_mac(td, oldfd, old, newfd, new, pathseg,
3679 		    &fromnd);
3680 		if (error != 0)
3681 			return (error);
3682 	} else {
3683 #endif
3684 	NDINIT_ATRIGHTS(&fromnd, DELETE, WANTPARENT | SAVESTART | AUDITVNODE1,
3685 	    pathseg, old, oldfd, &cap_renameat_source_rights, td);
3686 	if ((error = namei(&fromnd)) != 0)
3687 		return (error);
3688 #ifdef MAC
3689 	}
3690 #endif
3691 	fvp = fromnd.ni_vp;
3692 	tondflags = LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART | AUDITVNODE2;
3693 	if (fromnd.ni_vp->v_type == VDIR)
3694 		tondflags |= WILLBEDIR;
3695 	NDINIT_ATRIGHTS(&tond, RENAME, tondflags, pathseg, new, newfd,
3696 	    &cap_renameat_target_rights, td);
3697 	if ((error = namei(&tond)) != 0) {
3698 		/* Translate error code for rename("dir1", "dir2/."). */
3699 		if (error == EISDIR && fvp->v_type == VDIR)
3700 			error = EINVAL;
3701 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3702 		vrele(fromnd.ni_dvp);
3703 		vrele(fvp);
3704 		goto out1;
3705 	}
3706 	tdvp = tond.ni_dvp;
3707 	tvp = tond.ni_vp;
3708 	error = vn_start_write(fvp, &mp, V_NOWAIT);
3709 	if (error != 0) {
3710 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3711 		NDFREE(&tond, NDF_ONLY_PNBUF);
3712 		if (tvp != NULL)
3713 			vput(tvp);
3714 		if (tdvp == tvp)
3715 			vrele(tdvp);
3716 		else
3717 			vput(tdvp);
3718 		vrele(fromnd.ni_dvp);
3719 		vrele(fvp);
3720 		vrele(tond.ni_startdir);
3721 		if (fromnd.ni_startdir != NULL)
3722 			vrele(fromnd.ni_startdir);
3723 		error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH);
3724 		if (error != 0)
3725 			return (error);
3726 		goto again;
3727 	}
3728 	if (tvp != NULL) {
3729 		if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
3730 			error = ENOTDIR;
3731 			goto out;
3732 		} else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
3733 			error = EISDIR;
3734 			goto out;
3735 		}
3736 #ifdef CAPABILITIES
3737 		if (newfd != AT_FDCWD && (tond.ni_resflags & NIRES_ABS) == 0) {
3738 			/*
3739 			 * If the target already exists we require CAP_UNLINKAT
3740 			 * from 'newfd', when newfd was used for the lookup.
3741 			 */
3742 			error = cap_check(&tond.ni_filecaps.fc_rights,
3743 			    &cap_unlinkat_rights);
3744 			if (error != 0)
3745 				goto out;
3746 		}
3747 #endif
3748 	}
3749 	if (fvp == tdvp) {
3750 		error = EINVAL;
3751 		goto out;
3752 	}
3753 	/*
3754 	 * If the source is the same as the destination (that is, if they
3755 	 * are links to the same vnode), then there is nothing to do.
3756 	 */
3757 	if (fvp == tvp)
3758 		error = ERESTART;
3759 #ifdef MAC
3760 	else
3761 		error = mac_vnode_check_rename_to(td->td_ucred, tdvp,
3762 		    tond.ni_vp, fromnd.ni_dvp == tdvp, &tond.ni_cnd);
3763 #endif
3764 out:
3765 	if (error == 0) {
3766 		error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
3767 		    tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
3768 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3769 		NDFREE(&tond, NDF_ONLY_PNBUF);
3770 	} else {
3771 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3772 		NDFREE(&tond, NDF_ONLY_PNBUF);
3773 		if (tvp != NULL)
3774 			vput(tvp);
3775 		if (tdvp == tvp)
3776 			vrele(tdvp);
3777 		else
3778 			vput(tdvp);
3779 		vrele(fromnd.ni_dvp);
3780 		vrele(fvp);
3781 	}
3782 	vrele(tond.ni_startdir);
3783 	vn_finished_write(mp);
3784 out1:
3785 	if (fromnd.ni_startdir)
3786 		vrele(fromnd.ni_startdir);
3787 	if (error == ERESTART)
3788 		return (0);
3789 	if (error == ERELOOKUP)
3790 		goto again;
3791 	return (error);
3792 }
3793 
3794 /*
3795  * Make a directory file.
3796  */
3797 #ifndef _SYS_SYSPROTO_H_
3798 struct mkdir_args {
3799 	char	*path;
3800 	int	mode;
3801 };
3802 #endif
3803 int
3804 sys_mkdir(struct thread *td, struct mkdir_args *uap)
3805 {
3806 
3807 	return (kern_mkdirat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
3808 	    uap->mode));
3809 }
3810 
3811 #ifndef _SYS_SYSPROTO_H_
3812 struct mkdirat_args {
3813 	int	fd;
3814 	char	*path;
3815 	mode_t	mode;
3816 };
3817 #endif
3818 int
3819 sys_mkdirat(struct thread *td, struct mkdirat_args *uap)
3820 {
3821 
3822 	return (kern_mkdirat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode));
3823 }
3824 
3825 int
3826 kern_mkdirat(struct thread *td, int fd, const char *path, enum uio_seg segflg,
3827     int mode)
3828 {
3829 	struct mount *mp;
3830 	struct vattr vattr;
3831 	struct nameidata nd;
3832 	int error;
3833 
3834 	AUDIT_ARG_MODE(mode);
3835 	NDPREINIT(&nd);
3836 restart:
3837 	bwillwrite();
3838 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
3839 	    NC_NOMAKEENTRY | NC_KEEPPOSENTRY | FAILIFEXISTS | WILLBEDIR,
3840 	    segflg, path, fd, &cap_mkdirat_rights, td);
3841 	if ((error = namei(&nd)) != 0)
3842 		return (error);
3843 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3844 		NDFREE(&nd, NDF_ONLY_PNBUF);
3845 		vput(nd.ni_dvp);
3846 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3847 			return (error);
3848 		goto restart;
3849 	}
3850 	VATTR_NULL(&vattr);
3851 	vattr.va_type = VDIR;
3852 	vattr.va_mode = (mode & ACCESSPERMS) &~ td->td_proc->p_pd->pd_cmask;
3853 #ifdef MAC
3854 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
3855 	    &vattr);
3856 	if (error != 0)
3857 		goto out;
3858 #endif
3859 	error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
3860 #ifdef MAC
3861 out:
3862 #endif
3863 	NDFREE(&nd, NDF_ONLY_PNBUF);
3864 	VOP_VPUT_PAIR(nd.ni_dvp, error == 0 ? &nd.ni_vp : NULL, true);
3865 	vn_finished_write(mp);
3866 	if (error == ERELOOKUP)
3867 		goto restart;
3868 	return (error);
3869 }
3870 
3871 /*
3872  * Remove a directory file.
3873  */
3874 #ifndef _SYS_SYSPROTO_H_
3875 struct rmdir_args {
3876 	char	*path;
3877 };
3878 #endif
3879 int
3880 sys_rmdir(struct thread *td, struct rmdir_args *uap)
3881 {
3882 
3883 	return (kern_frmdirat(td, AT_FDCWD, uap->path, FD_NONE, UIO_USERSPACE,
3884 	    0));
3885 }
3886 
3887 int
3888 kern_frmdirat(struct thread *td, int dfd, const char *path, int fd,
3889     enum uio_seg pathseg, int flag)
3890 {
3891 	struct mount *mp;
3892 	struct vnode *vp;
3893 	struct file *fp;
3894 	struct nameidata nd;
3895 	cap_rights_t rights;
3896 	int error;
3897 
3898 	fp = NULL;
3899 	if (fd != FD_NONE) {
3900 		error = getvnode(td, fd, cap_rights_init_one(&rights,
3901 		    CAP_LOOKUP), &fp);
3902 		if (error != 0)
3903 			return (error);
3904 	}
3905 
3906 	NDPREINIT(&nd);
3907 restart:
3908 	bwillwrite();
3909 	NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1 |
3910 	    at2cnpflags(flag, AT_RESOLVE_BENEATH),
3911 	    pathseg, path, dfd, &cap_unlinkat_rights, td);
3912 	if ((error = namei(&nd)) != 0)
3913 		goto fdout;
3914 	vp = nd.ni_vp;
3915 	if (vp->v_type != VDIR) {
3916 		error = ENOTDIR;
3917 		goto out;
3918 	}
3919 	/*
3920 	 * No rmdir "." please.
3921 	 */
3922 	if (nd.ni_dvp == vp) {
3923 		error = EINVAL;
3924 		goto out;
3925 	}
3926 	/*
3927 	 * The root of a mounted filesystem cannot be deleted.
3928 	 */
3929 	if (vp->v_vflag & VV_ROOT) {
3930 		error = EBUSY;
3931 		goto out;
3932 	}
3933 
3934 	if (fp != NULL && fp->f_vnode != vp) {
3935 		if (VN_IS_DOOMED(fp->f_vnode))
3936 			error = EBADF;
3937 		else
3938 			error = EDEADLK;
3939 		goto out;
3940 	}
3941 
3942 #ifdef MAC
3943 	error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
3944 	    &nd.ni_cnd);
3945 	if (error != 0)
3946 		goto out;
3947 #endif
3948 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3949 		NDFREE(&nd, NDF_ONLY_PNBUF);
3950 		vput(vp);
3951 		if (nd.ni_dvp == vp)
3952 			vrele(nd.ni_dvp);
3953 		else
3954 			vput(nd.ni_dvp);
3955 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3956 			goto fdout;
3957 		goto restart;
3958 	}
3959 	vfs_notify_upper(vp, VFS_NOTIFY_UPPER_UNLINK);
3960 	error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
3961 	vn_finished_write(mp);
3962 out:
3963 	NDFREE(&nd, NDF_ONLY_PNBUF);
3964 	vput(vp);
3965 	if (nd.ni_dvp == vp)
3966 		vrele(nd.ni_dvp);
3967 	else
3968 		vput(nd.ni_dvp);
3969 	if (error == ERELOOKUP)
3970 		goto restart;
3971 fdout:
3972 	if (fp != NULL)
3973 		fdrop(fp, td);
3974 	return (error);
3975 }
3976 
3977 #if defined(COMPAT_43) || defined(COMPAT_FREEBSD11)
3978 int
3979 freebsd11_kern_getdirentries(struct thread *td, int fd, char *ubuf, u_int count,
3980     long *basep, void (*func)(struct freebsd11_dirent *))
3981 {
3982 	struct freebsd11_dirent dstdp;
3983 	struct dirent *dp, *edp;
3984 	char *dirbuf;
3985 	off_t base;
3986 	ssize_t resid, ucount;
3987 	int error;
3988 
3989 	/* XXX arbitrary sanity limit on `count'. */
3990 	count = min(count, 64 * 1024);
3991 
3992 	dirbuf = malloc(count, M_TEMP, M_WAITOK);
3993 
3994 	error = kern_getdirentries(td, fd, dirbuf, count, &base, &resid,
3995 	    UIO_SYSSPACE);
3996 	if (error != 0)
3997 		goto done;
3998 	if (basep != NULL)
3999 		*basep = base;
4000 
4001 	ucount = 0;
4002 	for (dp = (struct dirent *)dirbuf,
4003 	    edp = (struct dirent *)&dirbuf[count - resid];
4004 	    ucount < count && dp < edp; ) {
4005 		if (dp->d_reclen == 0)
4006 			break;
4007 		MPASS(dp->d_reclen >= _GENERIC_DIRLEN(0));
4008 		if (dp->d_namlen >= sizeof(dstdp.d_name))
4009 			continue;
4010 		dstdp.d_type = dp->d_type;
4011 		dstdp.d_namlen = dp->d_namlen;
4012 		dstdp.d_fileno = dp->d_fileno;		/* truncate */
4013 		if (dstdp.d_fileno != dp->d_fileno) {
4014 			switch (ino64_trunc_error) {
4015 			default:
4016 			case 0:
4017 				break;
4018 			case 1:
4019 				error = EOVERFLOW;
4020 				goto done;
4021 			case 2:
4022 				dstdp.d_fileno = UINT32_MAX;
4023 				break;
4024 			}
4025 		}
4026 		dstdp.d_reclen = sizeof(dstdp) - sizeof(dstdp.d_name) +
4027 		    ((dp->d_namlen + 1 + 3) &~ 3);
4028 		bcopy(dp->d_name, dstdp.d_name, dstdp.d_namlen);
4029 		bzero(dstdp.d_name + dstdp.d_namlen,
4030 		    dstdp.d_reclen - offsetof(struct freebsd11_dirent, d_name) -
4031 		    dstdp.d_namlen);
4032 		MPASS(dstdp.d_reclen <= dp->d_reclen);
4033 		MPASS(ucount + dstdp.d_reclen <= count);
4034 		if (func != NULL)
4035 			func(&dstdp);
4036 		error = copyout(&dstdp, ubuf + ucount, dstdp.d_reclen);
4037 		if (error != 0)
4038 			break;
4039 		dp = (struct dirent *)((char *)dp + dp->d_reclen);
4040 		ucount += dstdp.d_reclen;
4041 	}
4042 
4043 done:
4044 	free(dirbuf, M_TEMP);
4045 	if (error == 0)
4046 		td->td_retval[0] = ucount;
4047 	return (error);
4048 }
4049 #endif /* COMPAT */
4050 
4051 #ifdef COMPAT_43
4052 static void
4053 ogetdirentries_cvt(struct freebsd11_dirent *dp)
4054 {
4055 #if (BYTE_ORDER == LITTLE_ENDIAN)
4056 	/*
4057 	 * The expected low byte of dp->d_namlen is our dp->d_type.
4058 	 * The high MBZ byte of dp->d_namlen is our dp->d_namlen.
4059 	 */
4060 	dp->d_type = dp->d_namlen;
4061 	dp->d_namlen = 0;
4062 #else
4063 	/*
4064 	 * The dp->d_type is the high byte of the expected dp->d_namlen,
4065 	 * so must be zero'ed.
4066 	 */
4067 	dp->d_type = 0;
4068 #endif
4069 }
4070 
4071 /*
4072  * Read a block of directory entries in a filesystem independent format.
4073  */
4074 #ifndef _SYS_SYSPROTO_H_
4075 struct ogetdirentries_args {
4076 	int	fd;
4077 	char	*buf;
4078 	u_int	count;
4079 	long	*basep;
4080 };
4081 #endif
4082 int
4083 ogetdirentries(struct thread *td, struct ogetdirentries_args *uap)
4084 {
4085 	long loff;
4086 	int error;
4087 
4088 	error = kern_ogetdirentries(td, uap, &loff);
4089 	if (error == 0)
4090 		error = copyout(&loff, uap->basep, sizeof(long));
4091 	return (error);
4092 }
4093 
4094 int
4095 kern_ogetdirentries(struct thread *td, struct ogetdirentries_args *uap,
4096     long *ploff)
4097 {
4098 	long base;
4099 	int error;
4100 
4101 	/* XXX arbitrary sanity limit on `count'. */
4102 	if (uap->count > 64 * 1024)
4103 		return (EINVAL);
4104 
4105 	error = freebsd11_kern_getdirentries(td, uap->fd, uap->buf, uap->count,
4106 	    &base, ogetdirentries_cvt);
4107 
4108 	if (error == 0 && uap->basep != NULL)
4109 		error = copyout(&base, uap->basep, sizeof(long));
4110 
4111 	return (error);
4112 }
4113 #endif /* COMPAT_43 */
4114 
4115 #if defined(COMPAT_FREEBSD11)
4116 #ifndef _SYS_SYSPROTO_H_
4117 struct freebsd11_getdirentries_args {
4118 	int	fd;
4119 	char	*buf;
4120 	u_int	count;
4121 	long	*basep;
4122 };
4123 #endif
4124 int
4125 freebsd11_getdirentries(struct thread *td,
4126     struct freebsd11_getdirentries_args *uap)
4127 {
4128 	long base;
4129 	int error;
4130 
4131 	error = freebsd11_kern_getdirentries(td, uap->fd, uap->buf, uap->count,
4132 	    &base, NULL);
4133 
4134 	if (error == 0 && uap->basep != NULL)
4135 		error = copyout(&base, uap->basep, sizeof(long));
4136 	return (error);
4137 }
4138 
4139 int
4140 freebsd11_getdents(struct thread *td, struct freebsd11_getdents_args *uap)
4141 {
4142 	struct freebsd11_getdirentries_args ap;
4143 
4144 	ap.fd = uap->fd;
4145 	ap.buf = uap->buf;
4146 	ap.count = uap->count;
4147 	ap.basep = NULL;
4148 	return (freebsd11_getdirentries(td, &ap));
4149 }
4150 #endif /* COMPAT_FREEBSD11 */
4151 
4152 /*
4153  * Read a block of directory entries in a filesystem independent format.
4154  */
4155 int
4156 sys_getdirentries(struct thread *td, struct getdirentries_args *uap)
4157 {
4158 	off_t base;
4159 	int error;
4160 
4161 	error = kern_getdirentries(td, uap->fd, uap->buf, uap->count, &base,
4162 	    NULL, UIO_USERSPACE);
4163 	if (error != 0)
4164 		return (error);
4165 	if (uap->basep != NULL)
4166 		error = copyout(&base, uap->basep, sizeof(off_t));
4167 	return (error);
4168 }
4169 
4170 int
4171 kern_getdirentries(struct thread *td, int fd, char *buf, size_t count,
4172     off_t *basep, ssize_t *residp, enum uio_seg bufseg)
4173 {
4174 	struct vnode *vp;
4175 	struct file *fp;
4176 	struct uio auio;
4177 	struct iovec aiov;
4178 	off_t loff;
4179 	int error, eofflag;
4180 	off_t foffset;
4181 
4182 	AUDIT_ARG_FD(fd);
4183 	if (count > IOSIZE_MAX)
4184 		return (EINVAL);
4185 	auio.uio_resid = count;
4186 	error = getvnode(td, fd, &cap_read_rights, &fp);
4187 	if (error != 0)
4188 		return (error);
4189 	if ((fp->f_flag & FREAD) == 0) {
4190 		fdrop(fp, td);
4191 		return (EBADF);
4192 	}
4193 	vp = fp->f_vnode;
4194 	foffset = foffset_lock(fp, 0);
4195 unionread:
4196 	if (vp->v_type != VDIR) {
4197 		error = EINVAL;
4198 		goto fail;
4199 	}
4200 	aiov.iov_base = buf;
4201 	aiov.iov_len = count;
4202 	auio.uio_iov = &aiov;
4203 	auio.uio_iovcnt = 1;
4204 	auio.uio_rw = UIO_READ;
4205 	auio.uio_segflg = bufseg;
4206 	auio.uio_td = td;
4207 	vn_lock(vp, LK_SHARED | LK_RETRY);
4208 	AUDIT_ARG_VNODE1(vp);
4209 	loff = auio.uio_offset = foffset;
4210 #ifdef MAC
4211 	error = mac_vnode_check_readdir(td->td_ucred, vp);
4212 	if (error == 0)
4213 #endif
4214 		error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, NULL,
4215 		    NULL);
4216 	foffset = auio.uio_offset;
4217 	if (error != 0) {
4218 		VOP_UNLOCK(vp);
4219 		goto fail;
4220 	}
4221 	if (count == auio.uio_resid &&
4222 	    (vp->v_vflag & VV_ROOT) &&
4223 	    (vp->v_mount->mnt_flag & MNT_UNION)) {
4224 		struct vnode *tvp = vp;
4225 
4226 		vp = vp->v_mount->mnt_vnodecovered;
4227 		VREF(vp);
4228 		fp->f_vnode = vp;
4229 		foffset = 0;
4230 		vput(tvp);
4231 		goto unionread;
4232 	}
4233 	VOP_UNLOCK(vp);
4234 	*basep = loff;
4235 	if (residp != NULL)
4236 		*residp = auio.uio_resid;
4237 	td->td_retval[0] = count - auio.uio_resid;
4238 fail:
4239 	foffset_unlock(fp, foffset, 0);
4240 	fdrop(fp, td);
4241 	return (error);
4242 }
4243 
4244 /*
4245  * Set the mode mask for creation of filesystem nodes.
4246  */
4247 #ifndef _SYS_SYSPROTO_H_
4248 struct umask_args {
4249 	int	newmask;
4250 };
4251 #endif
4252 int
4253 sys_umask(struct thread *td, struct umask_args *uap)
4254 {
4255 	struct pwddesc *pdp;
4256 
4257 	pdp = td->td_proc->p_pd;
4258 	PWDDESC_XLOCK(pdp);
4259 	td->td_retval[0] = pdp->pd_cmask;
4260 	pdp->pd_cmask = uap->newmask & ALLPERMS;
4261 	PWDDESC_XUNLOCK(pdp);
4262 	return (0);
4263 }
4264 
4265 /*
4266  * Void all references to file by ripping underlying filesystem away from
4267  * vnode.
4268  */
4269 #ifndef _SYS_SYSPROTO_H_
4270 struct revoke_args {
4271 	char	*path;
4272 };
4273 #endif
4274 int
4275 sys_revoke(struct thread *td, struct revoke_args *uap)
4276 {
4277 	struct vnode *vp;
4278 	struct vattr vattr;
4279 	struct nameidata nd;
4280 	int error;
4281 
4282 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
4283 	    uap->path, td);
4284 	if ((error = namei(&nd)) != 0)
4285 		return (error);
4286 	vp = nd.ni_vp;
4287 	NDFREE_NOTHING(&nd);
4288 	if (vp->v_type != VCHR || vp->v_rdev == NULL) {
4289 		error = EINVAL;
4290 		goto out;
4291 	}
4292 #ifdef MAC
4293 	error = mac_vnode_check_revoke(td->td_ucred, vp);
4294 	if (error != 0)
4295 		goto out;
4296 #endif
4297 	error = VOP_GETATTR(vp, &vattr, td->td_ucred);
4298 	if (error != 0)
4299 		goto out;
4300 	if (td->td_ucred->cr_uid != vattr.va_uid) {
4301 		error = priv_check(td, PRIV_VFS_ADMIN);
4302 		if (error != 0)
4303 			goto out;
4304 	}
4305 	if (devfs_usecount(vp) > 0)
4306 		VOP_REVOKE(vp, REVOKEALL);
4307 out:
4308 	vput(vp);
4309 	return (error);
4310 }
4311 
4312 /*
4313  * This variant of getvnode() allows O_PATH files.  Caller should
4314  * ensure that returned file and vnode are only used for compatible
4315  * semantics.
4316  */
4317 int
4318 getvnode_path(struct thread *td, int fd, cap_rights_t *rightsp,
4319     struct file **fpp)
4320 {
4321 	struct file *fp;
4322 	int error;
4323 
4324 	error = fget_unlocked(td->td_proc->p_fd, fd, rightsp, &fp);
4325 	if (error != 0)
4326 		return (error);
4327 
4328 	/*
4329 	 * The file could be not of the vnode type, or it may be not
4330 	 * yet fully initialized, in which case the f_vnode pointer
4331 	 * may be set, but f_ops is still badfileops.  E.g.,
4332 	 * devfs_open() transiently create such situation to
4333 	 * facilitate csw d_fdopen().
4334 	 *
4335 	 * Dupfdopen() handling in kern_openat() installs the
4336 	 * half-baked file into the process descriptor table, allowing
4337 	 * other thread to dereference it. Guard against the race by
4338 	 * checking f_ops.
4339 	 */
4340 	if (__predict_false(fp->f_vnode == NULL || fp->f_ops == &badfileops)) {
4341 		fdrop(fp, td);
4342 		return (EINVAL);
4343 	}
4344 
4345 	*fpp = fp;
4346 	return (0);
4347 }
4348 
4349 /*
4350  * Convert a user file descriptor to a kernel file entry and check
4351  * that, if it is a capability, the correct rights are present.
4352  * A reference on the file entry is held upon returning.
4353  */
4354 int
4355 getvnode(struct thread *td, int fd, cap_rights_t *rightsp, struct file **fpp)
4356 {
4357 	int error;
4358 
4359 	error = getvnode_path(td, fd, rightsp, fpp);
4360 	if (__predict_false(error != 0))
4361 		return (error);
4362 
4363 	/*
4364 	 * Filter out O_PATH file descriptors, most getvnode() callers
4365 	 * do not call fo_ methods.
4366 	 */
4367 	if (__predict_false((*fpp)->f_ops == &path_fileops)) {
4368 		fdrop(*fpp, td);
4369 		error = EBADF;
4370 	}
4371 
4372 	return (error);
4373 }
4374 
4375 /*
4376  * Get an (NFS) file handle.
4377  */
4378 #ifndef _SYS_SYSPROTO_H_
4379 struct lgetfh_args {
4380 	char *fname;
4381 	fhandle_t *fhp;
4382 };
4383 #endif
4384 int
4385 sys_lgetfh(struct thread *td, struct lgetfh_args *uap)
4386 {
4387 
4388 	return (kern_getfhat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->fname,
4389 	    UIO_USERSPACE, uap->fhp, UIO_USERSPACE));
4390 }
4391 
4392 #ifndef _SYS_SYSPROTO_H_
4393 struct getfh_args {
4394 	char *fname;
4395 	fhandle_t *fhp;
4396 };
4397 #endif
4398 int
4399 sys_getfh(struct thread *td, struct getfh_args *uap)
4400 {
4401 
4402 	return (kern_getfhat(td, 0, AT_FDCWD, uap->fname, UIO_USERSPACE,
4403 	    uap->fhp, UIO_USERSPACE));
4404 }
4405 
4406 /*
4407  * syscall for the rpc.lockd to use to translate an open descriptor into
4408  * a NFS file handle.
4409  *
4410  * warning: do not remove the priv_check() call or this becomes one giant
4411  * security hole.
4412  */
4413 #ifndef _SYS_SYSPROTO_H_
4414 struct getfhat_args {
4415 	int fd;
4416 	char *path;
4417 	fhandle_t *fhp;
4418 	int flags;
4419 };
4420 #endif
4421 int
4422 sys_getfhat(struct thread *td, struct getfhat_args *uap)
4423 {
4424 
4425 	return (kern_getfhat(td, uap->flags, uap->fd, uap->path, UIO_USERSPACE,
4426 	    uap->fhp, UIO_USERSPACE));
4427 }
4428 
4429 int
4430 kern_getfhat(struct thread *td, int flags, int fd, const char *path,
4431     enum uio_seg pathseg, fhandle_t *fhp, enum uio_seg fhseg)
4432 {
4433 	struct nameidata nd;
4434 	fhandle_t fh;
4435 	struct vnode *vp;
4436 	int error;
4437 
4438 	if ((flags & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH)) != 0)
4439 		return (EINVAL);
4440 	error = priv_check(td, PRIV_VFS_GETFH);
4441 	if (error != 0)
4442 		return (error);
4443 	NDINIT_AT(&nd, LOOKUP, at2cnpflags(flags, AT_SYMLINK_NOFOLLOW |
4444 	    AT_RESOLVE_BENEATH) | LOCKLEAF | AUDITVNODE1, pathseg, path,
4445 	    fd, td);
4446 	error = namei(&nd);
4447 	if (error != 0)
4448 		return (error);
4449 	NDFREE_NOTHING(&nd);
4450 	vp = nd.ni_vp;
4451 	bzero(&fh, sizeof(fh));
4452 	fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
4453 	error = VOP_VPTOFH(vp, &fh.fh_fid);
4454 	vput(vp);
4455 	if (error == 0) {
4456 		if (fhseg == UIO_USERSPACE)
4457 			error = copyout(&fh, fhp, sizeof (fh));
4458 		else
4459 			memcpy(fhp, &fh, sizeof(fh));
4460 	}
4461 	return (error);
4462 }
4463 
4464 #ifndef _SYS_SYSPROTO_H_
4465 struct fhlink_args {
4466 	fhandle_t *fhp;
4467 	const char *to;
4468 };
4469 #endif
4470 int
4471 sys_fhlink(struct thread *td, struct fhlink_args *uap)
4472 {
4473 
4474 	return (kern_fhlinkat(td, AT_FDCWD, uap->to, UIO_USERSPACE, uap->fhp));
4475 }
4476 
4477 #ifndef _SYS_SYSPROTO_H_
4478 struct fhlinkat_args {
4479 	fhandle_t *fhp;
4480 	int tofd;
4481 	const char *to;
4482 };
4483 #endif
4484 int
4485 sys_fhlinkat(struct thread *td, struct fhlinkat_args *uap)
4486 {
4487 
4488 	return (kern_fhlinkat(td, uap->tofd, uap->to, UIO_USERSPACE, uap->fhp));
4489 }
4490 
4491 static int
4492 kern_fhlinkat(struct thread *td, int fd, const char *path,
4493     enum uio_seg pathseg, fhandle_t *fhp)
4494 {
4495 	fhandle_t fh;
4496 	struct mount *mp;
4497 	struct vnode *vp;
4498 	int error;
4499 
4500 	error = priv_check(td, PRIV_VFS_GETFH);
4501 	if (error != 0)
4502 		return (error);
4503 	error = copyin(fhp, &fh, sizeof(fh));
4504 	if (error != 0)
4505 		return (error);
4506 	do {
4507 		bwillwrite();
4508 		if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4509 			return (ESTALE);
4510 		error = VFS_FHTOVP(mp, &fh.fh_fid, LK_SHARED, &vp);
4511 		vfs_unbusy(mp);
4512 		if (error != 0)
4513 			return (error);
4514 		VOP_UNLOCK(vp);
4515 		error = kern_linkat_vp(td, vp, fd, path, pathseg);
4516 	} while (error == EAGAIN || error == ERELOOKUP);
4517 	return (error);
4518 }
4519 
4520 #ifndef _SYS_SYSPROTO_H_
4521 struct fhreadlink_args {
4522 	fhandle_t *fhp;
4523 	char *buf;
4524 	size_t bufsize;
4525 };
4526 #endif
4527 int
4528 sys_fhreadlink(struct thread *td, struct fhreadlink_args *uap)
4529 {
4530 	fhandle_t fh;
4531 	struct mount *mp;
4532 	struct vnode *vp;
4533 	int error;
4534 
4535 	error = priv_check(td, PRIV_VFS_GETFH);
4536 	if (error != 0)
4537 		return (error);
4538 	if (uap->bufsize > IOSIZE_MAX)
4539 		return (EINVAL);
4540 	error = copyin(uap->fhp, &fh, sizeof(fh));
4541 	if (error != 0)
4542 		return (error);
4543 	if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4544 		return (ESTALE);
4545 	error = VFS_FHTOVP(mp, &fh.fh_fid, LK_SHARED, &vp);
4546 	vfs_unbusy(mp);
4547 	if (error != 0)
4548 		return (error);
4549 	error = kern_readlink_vp(vp, uap->buf, UIO_USERSPACE, uap->bufsize, td);
4550 	vput(vp);
4551 	return (error);
4552 }
4553 
4554 /*
4555  * syscall for the rpc.lockd to use to translate a NFS file handle into an
4556  * open descriptor.
4557  *
4558  * warning: do not remove the priv_check() call or this becomes one giant
4559  * security hole.
4560  */
4561 #ifndef _SYS_SYSPROTO_H_
4562 struct fhopen_args {
4563 	const struct fhandle *u_fhp;
4564 	int flags;
4565 };
4566 #endif
4567 int
4568 sys_fhopen(struct thread *td, struct fhopen_args *uap)
4569 {
4570 	return (kern_fhopen(td, uap->u_fhp, uap->flags));
4571 }
4572 
4573 int
4574 kern_fhopen(struct thread *td, const struct fhandle *u_fhp, int flags)
4575 {
4576 	struct mount *mp;
4577 	struct vnode *vp;
4578 	struct fhandle fhp;
4579 	struct file *fp;
4580 	int fmode, error;
4581 	int indx;
4582 
4583 	error = priv_check(td, PRIV_VFS_FHOPEN);
4584 	if (error != 0)
4585 		return (error);
4586 	indx = -1;
4587 	fmode = FFLAGS(flags);
4588 	/* why not allow a non-read/write open for our lockd? */
4589 	if (((fmode & (FREAD | FWRITE)) == 0) || (fmode & O_CREAT))
4590 		return (EINVAL);
4591 	error = copyin(u_fhp, &fhp, sizeof(fhp));
4592 	if (error != 0)
4593 		return(error);
4594 	/* find the mount point */
4595 	mp = vfs_busyfs(&fhp.fh_fsid);
4596 	if (mp == NULL)
4597 		return (ESTALE);
4598 	/* now give me my vnode, it gets returned to me locked */
4599 	error = VFS_FHTOVP(mp, &fhp.fh_fid, LK_EXCLUSIVE, &vp);
4600 	vfs_unbusy(mp);
4601 	if (error != 0)
4602 		return (error);
4603 
4604 	error = falloc_noinstall(td, &fp);
4605 	if (error != 0) {
4606 		vput(vp);
4607 		return (error);
4608 	}
4609 	/*
4610 	 * An extra reference on `fp' has been held for us by
4611 	 * falloc_noinstall().
4612 	 */
4613 
4614 #ifdef INVARIANTS
4615 	td->td_dupfd = -1;
4616 #endif
4617 	error = vn_open_vnode(vp, fmode, td->td_ucred, td, fp);
4618 	if (error != 0) {
4619 		KASSERT(fp->f_ops == &badfileops,
4620 		    ("VOP_OPEN in fhopen() set f_ops"));
4621 		KASSERT(td->td_dupfd < 0,
4622 		    ("fhopen() encountered fdopen()"));
4623 
4624 		vput(vp);
4625 		goto bad;
4626 	}
4627 #ifdef INVARIANTS
4628 	td->td_dupfd = 0;
4629 #endif
4630 	fp->f_vnode = vp;
4631 	finit_vnode(fp, fmode, NULL, &vnops);
4632 	VOP_UNLOCK(vp);
4633 	if ((fmode & O_TRUNC) != 0) {
4634 		error = fo_truncate(fp, 0, td->td_ucred, td);
4635 		if (error != 0)
4636 			goto bad;
4637 	}
4638 
4639 	error = finstall(td, fp, &indx, fmode, NULL);
4640 bad:
4641 	fdrop(fp, td);
4642 	td->td_retval[0] = indx;
4643 	return (error);
4644 }
4645 
4646 /*
4647  * Stat an (NFS) file handle.
4648  */
4649 #ifndef _SYS_SYSPROTO_H_
4650 struct fhstat_args {
4651 	struct fhandle *u_fhp;
4652 	struct stat *sb;
4653 };
4654 #endif
4655 int
4656 sys_fhstat(struct thread *td, struct fhstat_args *uap)
4657 {
4658 	struct stat sb;
4659 	struct fhandle fh;
4660 	int error;
4661 
4662 	error = copyin(uap->u_fhp, &fh, sizeof(fh));
4663 	if (error != 0)
4664 		return (error);
4665 	error = kern_fhstat(td, fh, &sb);
4666 	if (error == 0)
4667 		error = copyout(&sb, uap->sb, sizeof(sb));
4668 	return (error);
4669 }
4670 
4671 int
4672 kern_fhstat(struct thread *td, struct fhandle fh, struct stat *sb)
4673 {
4674 	struct mount *mp;
4675 	struct vnode *vp;
4676 	int error;
4677 
4678 	error = priv_check(td, PRIV_VFS_FHSTAT);
4679 	if (error != 0)
4680 		return (error);
4681 	if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4682 		return (ESTALE);
4683 	error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4684 	vfs_unbusy(mp);
4685 	if (error != 0)
4686 		return (error);
4687 	error = VOP_STAT(vp, sb, td->td_ucred, NOCRED);
4688 	vput(vp);
4689 	return (error);
4690 }
4691 
4692 /*
4693  * Implement fstatfs() for (NFS) file handles.
4694  */
4695 #ifndef _SYS_SYSPROTO_H_
4696 struct fhstatfs_args {
4697 	struct fhandle *u_fhp;
4698 	struct statfs *buf;
4699 };
4700 #endif
4701 int
4702 sys_fhstatfs(struct thread *td, struct fhstatfs_args *uap)
4703 {
4704 	struct statfs *sfp;
4705 	fhandle_t fh;
4706 	int error;
4707 
4708 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
4709 	if (error != 0)
4710 		return (error);
4711 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
4712 	error = kern_fhstatfs(td, fh, sfp);
4713 	if (error == 0)
4714 		error = copyout(sfp, uap->buf, sizeof(*sfp));
4715 	free(sfp, M_STATFS);
4716 	return (error);
4717 }
4718 
4719 int
4720 kern_fhstatfs(struct thread *td, fhandle_t fh, struct statfs *buf)
4721 {
4722 	struct mount *mp;
4723 	struct vnode *vp;
4724 	int error;
4725 
4726 	error = priv_check(td, PRIV_VFS_FHSTATFS);
4727 	if (error != 0)
4728 		return (error);
4729 	if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4730 		return (ESTALE);
4731 	error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4732 	if (error != 0) {
4733 		vfs_unbusy(mp);
4734 		return (error);
4735 	}
4736 	vput(vp);
4737 	error = prison_canseemount(td->td_ucred, mp);
4738 	if (error != 0)
4739 		goto out;
4740 #ifdef MAC
4741 	error = mac_mount_check_stat(td->td_ucred, mp);
4742 	if (error != 0)
4743 		goto out;
4744 #endif
4745 	error = VFS_STATFS(mp, buf);
4746 out:
4747 	vfs_unbusy(mp);
4748 	return (error);
4749 }
4750 
4751 /*
4752  * Unlike madvise(2), we do not make a best effort to remember every
4753  * possible caching hint.  Instead, we remember the last setting with
4754  * the exception that we will allow POSIX_FADV_NORMAL to adjust the
4755  * region of any current setting.
4756  */
4757 int
4758 kern_posix_fadvise(struct thread *td, int fd, off_t offset, off_t len,
4759     int advice)
4760 {
4761 	struct fadvise_info *fa, *new;
4762 	struct file *fp;
4763 	struct vnode *vp;
4764 	off_t end;
4765 	int error;
4766 
4767 	if (offset < 0 || len < 0 || offset > OFF_MAX - len)
4768 		return (EINVAL);
4769 	AUDIT_ARG_VALUE(advice);
4770 	switch (advice) {
4771 	case POSIX_FADV_SEQUENTIAL:
4772 	case POSIX_FADV_RANDOM:
4773 	case POSIX_FADV_NOREUSE:
4774 		new = malloc(sizeof(*fa), M_FADVISE, M_WAITOK);
4775 		break;
4776 	case POSIX_FADV_NORMAL:
4777 	case POSIX_FADV_WILLNEED:
4778 	case POSIX_FADV_DONTNEED:
4779 		new = NULL;
4780 		break;
4781 	default:
4782 		return (EINVAL);
4783 	}
4784 	/* XXX: CAP_POSIX_FADVISE? */
4785 	AUDIT_ARG_FD(fd);
4786 	error = fget(td, fd, &cap_no_rights, &fp);
4787 	if (error != 0)
4788 		goto out;
4789 	AUDIT_ARG_FILE(td->td_proc, fp);
4790 	if ((fp->f_ops->fo_flags & DFLAG_SEEKABLE) == 0) {
4791 		error = ESPIPE;
4792 		goto out;
4793 	}
4794 	if (fp->f_type != DTYPE_VNODE) {
4795 		error = ENODEV;
4796 		goto out;
4797 	}
4798 	vp = fp->f_vnode;
4799 	if (vp->v_type != VREG) {
4800 		error = ENODEV;
4801 		goto out;
4802 	}
4803 	if (len == 0)
4804 		end = OFF_MAX;
4805 	else
4806 		end = offset + len - 1;
4807 	switch (advice) {
4808 	case POSIX_FADV_SEQUENTIAL:
4809 	case POSIX_FADV_RANDOM:
4810 	case POSIX_FADV_NOREUSE:
4811 		/*
4812 		 * Try to merge any existing non-standard region with
4813 		 * this new region if possible, otherwise create a new
4814 		 * non-standard region for this request.
4815 		 */
4816 		mtx_pool_lock(mtxpool_sleep, fp);
4817 		fa = fp->f_advice;
4818 		if (fa != NULL && fa->fa_advice == advice &&
4819 		    ((fa->fa_start <= end && fa->fa_end >= offset) ||
4820 		    (end != OFF_MAX && fa->fa_start == end + 1) ||
4821 		    (fa->fa_end != OFF_MAX && fa->fa_end + 1 == offset))) {
4822 			if (offset < fa->fa_start)
4823 				fa->fa_start = offset;
4824 			if (end > fa->fa_end)
4825 				fa->fa_end = end;
4826 		} else {
4827 			new->fa_advice = advice;
4828 			new->fa_start = offset;
4829 			new->fa_end = end;
4830 			fp->f_advice = new;
4831 			new = fa;
4832 		}
4833 		mtx_pool_unlock(mtxpool_sleep, fp);
4834 		break;
4835 	case POSIX_FADV_NORMAL:
4836 		/*
4837 		 * If a the "normal" region overlaps with an existing
4838 		 * non-standard region, trim or remove the
4839 		 * non-standard region.
4840 		 */
4841 		mtx_pool_lock(mtxpool_sleep, fp);
4842 		fa = fp->f_advice;
4843 		if (fa != NULL) {
4844 			if (offset <= fa->fa_start && end >= fa->fa_end) {
4845 				new = fa;
4846 				fp->f_advice = NULL;
4847 			} else if (offset <= fa->fa_start &&
4848 			    end >= fa->fa_start)
4849 				fa->fa_start = end + 1;
4850 			else if (offset <= fa->fa_end && end >= fa->fa_end)
4851 				fa->fa_end = offset - 1;
4852 			else if (offset >= fa->fa_start && end <= fa->fa_end) {
4853 				/*
4854 				 * If the "normal" region is a middle
4855 				 * portion of the existing
4856 				 * non-standard region, just remove
4857 				 * the whole thing rather than picking
4858 				 * one side or the other to
4859 				 * preserve.
4860 				 */
4861 				new = fa;
4862 				fp->f_advice = NULL;
4863 			}
4864 		}
4865 		mtx_pool_unlock(mtxpool_sleep, fp);
4866 		break;
4867 	case POSIX_FADV_WILLNEED:
4868 	case POSIX_FADV_DONTNEED:
4869 		error = VOP_ADVISE(vp, offset, end, advice);
4870 		break;
4871 	}
4872 out:
4873 	if (fp != NULL)
4874 		fdrop(fp, td);
4875 	free(new, M_FADVISE);
4876 	return (error);
4877 }
4878 
4879 int
4880 sys_posix_fadvise(struct thread *td, struct posix_fadvise_args *uap)
4881 {
4882 	int error;
4883 
4884 	error = kern_posix_fadvise(td, uap->fd, uap->offset, uap->len,
4885 	    uap->advice);
4886 	return (kern_posix_error(td, error));
4887 }
4888 
4889 int
4890 kern_copy_file_range(struct thread *td, int infd, off_t *inoffp, int outfd,
4891     off_t *outoffp, size_t len, unsigned int flags)
4892 {
4893 	struct file *infp, *outfp;
4894 	struct vnode *invp, *outvp;
4895 	int error;
4896 	size_t retlen;
4897 	void *rl_rcookie, *rl_wcookie;
4898 	off_t savinoff, savoutoff;
4899 
4900 	infp = outfp = NULL;
4901 	rl_rcookie = rl_wcookie = NULL;
4902 	savinoff = -1;
4903 	error = 0;
4904 	retlen = 0;
4905 
4906 	if (flags != 0) {
4907 		error = EINVAL;
4908 		goto out;
4909 	}
4910 	if (len > SSIZE_MAX)
4911 		/*
4912 		 * Although the len argument is size_t, the return argument
4913 		 * is ssize_t (which is signed).  Therefore a size that won't
4914 		 * fit in ssize_t can't be returned.
4915 		 */
4916 		len = SSIZE_MAX;
4917 
4918 	/* Get the file structures for the file descriptors. */
4919 	error = fget_read(td, infd, &cap_read_rights, &infp);
4920 	if (error != 0)
4921 		goto out;
4922 	if (infp->f_ops == &badfileops) {
4923 		error = EBADF;
4924 		goto out;
4925 	}
4926 	if (infp->f_vnode == NULL) {
4927 		error = EINVAL;
4928 		goto out;
4929 	}
4930 	error = fget_write(td, outfd, &cap_write_rights, &outfp);
4931 	if (error != 0)
4932 		goto out;
4933 	if (outfp->f_ops == &badfileops) {
4934 		error = EBADF;
4935 		goto out;
4936 	}
4937 	if (outfp->f_vnode == NULL) {
4938 		error = EINVAL;
4939 		goto out;
4940 	}
4941 
4942 	/* Set the offset pointers to the correct place. */
4943 	if (inoffp == NULL)
4944 		inoffp = &infp->f_offset;
4945 	if (outoffp == NULL)
4946 		outoffp = &outfp->f_offset;
4947 	savinoff = *inoffp;
4948 	savoutoff = *outoffp;
4949 
4950 	invp = infp->f_vnode;
4951 	outvp = outfp->f_vnode;
4952 	/* Sanity check the f_flag bits. */
4953 	if ((outfp->f_flag & (FWRITE | FAPPEND)) != FWRITE ||
4954 	    (infp->f_flag & FREAD) == 0) {
4955 		error = EBADF;
4956 		goto out;
4957 	}
4958 
4959 	/* If len == 0, just return 0. */
4960 	if (len == 0)
4961 		goto out;
4962 
4963 	/*
4964 	 * If infp and outfp refer to the same file, the byte ranges cannot
4965 	 * overlap.
4966 	 */
4967 	if (invp == outvp && ((savinoff <= savoutoff && savinoff + len >
4968 	    savoutoff) || (savinoff > savoutoff && savoutoff + len >
4969 	    savinoff))) {
4970 		error = EINVAL;
4971 		goto out;
4972 	}
4973 
4974 	/* Range lock the byte ranges for both invp and outvp. */
4975 	for (;;) {
4976 		rl_wcookie = vn_rangelock_wlock(outvp, *outoffp, *outoffp +
4977 		    len);
4978 		rl_rcookie = vn_rangelock_tryrlock(invp, *inoffp, *inoffp +
4979 		    len);
4980 		if (rl_rcookie != NULL)
4981 			break;
4982 		vn_rangelock_unlock(outvp, rl_wcookie);
4983 		rl_rcookie = vn_rangelock_rlock(invp, *inoffp, *inoffp + len);
4984 		vn_rangelock_unlock(invp, rl_rcookie);
4985 	}
4986 
4987 	retlen = len;
4988 	error = vn_copy_file_range(invp, inoffp, outvp, outoffp, &retlen,
4989 	    flags, infp->f_cred, outfp->f_cred, td);
4990 out:
4991 	if (rl_rcookie != NULL)
4992 		vn_rangelock_unlock(invp, rl_rcookie);
4993 	if (rl_wcookie != NULL)
4994 		vn_rangelock_unlock(outvp, rl_wcookie);
4995 	if (savinoff != -1 && (error == EINTR || error == ERESTART)) {
4996 		*inoffp = savinoff;
4997 		*outoffp = savoutoff;
4998 	}
4999 	if (outfp != NULL)
5000 		fdrop(outfp, td);
5001 	if (infp != NULL)
5002 		fdrop(infp, td);
5003 	td->td_retval[0] = retlen;
5004 	return (error);
5005 }
5006 
5007 int
5008 sys_copy_file_range(struct thread *td, struct copy_file_range_args *uap)
5009 {
5010 	off_t inoff, outoff, *inoffp, *outoffp;
5011 	int error;
5012 
5013 	inoffp = outoffp = NULL;
5014 	if (uap->inoffp != NULL) {
5015 		error = copyin(uap->inoffp, &inoff, sizeof(off_t));
5016 		if (error != 0)
5017 			return (error);
5018 		inoffp = &inoff;
5019 	}
5020 	if (uap->outoffp != NULL) {
5021 		error = copyin(uap->outoffp, &outoff, sizeof(off_t));
5022 		if (error != 0)
5023 			return (error);
5024 		outoffp = &outoff;
5025 	}
5026 	error = kern_copy_file_range(td, uap->infd, inoffp, uap->outfd,
5027 	    outoffp, uap->len, uap->flags);
5028 	if (error == 0 && uap->inoffp != NULL)
5029 		error = copyout(inoffp, uap->inoffp, sizeof(off_t));
5030 	if (error == 0 && uap->outoffp != NULL)
5031 		error = copyout(outoffp, uap->outoffp, sizeof(off_t));
5032 	return (error);
5033 }
5034