xref: /freebsd/sys/kern/vfs_syscalls.c (revision 0440b3d2cbf83afb55209a9938b31a48912f222c)
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 
37 #include "opt_capsicum.h"
38 #include "opt_ktrace.h"
39 
40 #include <sys/systm.h>
41 #ifdef COMPAT_FREEBSD11
42 #include <sys/abi_compat.h>
43 #endif
44 #include <sys/bio.h>
45 #include <sys/buf.h>
46 #include <sys/capsicum.h>
47 #include <sys/disk.h>
48 #include <sys/dirent.h>
49 #include <sys/fcntl.h>
50 #include <sys/file.h>
51 #include <sys/filedesc.h>
52 #include <sys/filio.h>
53 #include <sys/jail.h>
54 #include <sys/kernel.h>
55 #ifdef KTRACE
56 #include <sys/ktrace.h>
57 #endif
58 #include <sys/limits.h>
59 #include <sys/linker.h>
60 #include <sys/malloc.h>
61 #include <sys/mount.h>
62 #include <sys/mutex.h>
63 #include <sys/namei.h>
64 #include <sys/priv.h>
65 #include <sys/proc.h>
66 #include <sys/rwlock.h>
67 #include <sys/sdt.h>
68 #include <sys/stat.h>
69 #include <sys/sx.h>
70 #include <sys/syscallsubr.h>
71 #include <sys/sysctl.h>
72 #include <sys/sysproto.h>
73 #include <sys/unistd.h>
74 #include <sys/vnode.h>
75 
76 #include <machine/stdarg.h>
77 
78 #include <security/audit/audit.h>
79 #include <security/mac/mac_framework.h>
80 
81 #include <vm/vm.h>
82 #include <vm/vm_object.h>
83 #include <vm/vm_page.h>
84 #include <vm/vnode_pager.h>
85 #include <vm/uma.h>
86 
87 #include <fs/devfs/devfs.h>
88 
89 MALLOC_DEFINE(M_FADVISE, "fadvise", "posix_fadvise(2) information");
90 
91 static int kern_chflagsat(struct thread *td, int fd, const char *path,
92     enum uio_seg pathseg, u_long flags, int atflag);
93 static int setfflags(struct thread *td, struct vnode *, u_long);
94 static int getutimes(const struct timeval *, enum uio_seg, struct timespec *);
95 static int getutimens(const struct timespec *, enum uio_seg,
96     struct timespec *, int *);
97 static int setutimes(struct thread *td, struct vnode *,
98     const struct timespec *, int, int);
99 static int vn_access(struct vnode *vp, int user_flags, struct ucred *cred,
100     struct thread *td);
101 static int kern_fhlinkat(struct thread *td, int fd, const char *path,
102     enum uio_seg pathseg, fhandle_t *fhp);
103 static int kern_readlink_vp(struct vnode *vp, char *buf, enum uio_seg bufseg,
104     size_t count, struct thread *td);
105 static int kern_linkat_vp(struct thread *td, struct vnode *vp, int fd,
106     const char *path, enum uio_seg segflag);
107 
108 uint64_t
109 at2cnpflags(u_int at_flags, u_int mask)
110 {
111 	uint64_t res;
112 
113 	MPASS((at_flags & (AT_SYMLINK_FOLLOW | AT_SYMLINK_NOFOLLOW)) !=
114 	    (AT_SYMLINK_FOLLOW | AT_SYMLINK_NOFOLLOW));
115 
116 	res = 0;
117 	at_flags &= mask;
118 	if ((at_flags & AT_RESOLVE_BENEATH) != 0)
119 		res |= RBENEATH;
120 	if ((at_flags & AT_SYMLINK_FOLLOW) != 0)
121 		res |= FOLLOW;
122 	/* NOFOLLOW is pseudo flag */
123 	if ((mask & AT_SYMLINK_NOFOLLOW) != 0) {
124 		res |= (at_flags & AT_SYMLINK_NOFOLLOW) != 0 ? NOFOLLOW :
125 		    FOLLOW;
126 	}
127 	if ((mask & AT_EMPTY_PATH) != 0 && (at_flags & AT_EMPTY_PATH) != 0)
128 		res |= EMPTYPATH;
129 	return (res);
130 }
131 
132 int
133 kern_sync(struct thread *td)
134 {
135 	struct mount *mp, *nmp;
136 	int save;
137 
138 	mtx_lock(&mountlist_mtx);
139 	for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
140 		if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK)) {
141 			nmp = TAILQ_NEXT(mp, mnt_list);
142 			continue;
143 		}
144 		if ((mp->mnt_flag & MNT_RDONLY) == 0 &&
145 		    vn_start_write(NULL, &mp, V_NOWAIT) == 0) {
146 			save = curthread_pflags_set(TDP_SYNCIO);
147 			vfs_periodic(mp, MNT_NOWAIT);
148 			VFS_SYNC(mp, MNT_NOWAIT);
149 			curthread_pflags_restore(save);
150 			vn_finished_write(mp);
151 		}
152 		mtx_lock(&mountlist_mtx);
153 		nmp = TAILQ_NEXT(mp, mnt_list);
154 		vfs_unbusy(mp);
155 	}
156 	mtx_unlock(&mountlist_mtx);
157 	return (0);
158 }
159 
160 /*
161  * Sync each mounted filesystem.
162  */
163 #ifndef _SYS_SYSPROTO_H_
164 struct sync_args {
165 	int     dummy;
166 };
167 #endif
168 /* ARGSUSED */
169 int
170 sys_sync(struct thread *td, struct sync_args *uap)
171 {
172 
173 	return (kern_sync(td));
174 }
175 
176 /*
177  * Change filesystem quotas.
178  */
179 #ifndef _SYS_SYSPROTO_H_
180 struct quotactl_args {
181 	char *path;
182 	int cmd;
183 	int uid;
184 	caddr_t arg;
185 };
186 #endif
187 int
188 sys_quotactl(struct thread *td, struct quotactl_args *uap)
189 {
190 	struct mount *mp;
191 	struct nameidata nd;
192 	int error;
193 	bool mp_busy;
194 
195 	AUDIT_ARG_CMD(uap->cmd);
196 	AUDIT_ARG_UID(uap->uid);
197 	if (!prison_allow(td->td_ucred, PR_ALLOW_QUOTAS))
198 		return (EPERM);
199 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
200 	    uap->path);
201 	if ((error = namei(&nd)) != 0)
202 		return (error);
203 	NDFREE_PNBUF(&nd);
204 	mp = nd.ni_vp->v_mount;
205 	vfs_ref(mp);
206 	vput(nd.ni_vp);
207 	error = vfs_busy(mp, 0);
208 	if (error != 0) {
209 		vfs_rel(mp);
210 		return (error);
211 	}
212 	mp_busy = true;
213 	error = VFS_QUOTACTL(mp, uap->cmd, uap->uid, uap->arg, &mp_busy);
214 
215 	/*
216 	 * Since quota on/off operations typically need to open quota
217 	 * files, the implementation may need to unbusy the mount point
218 	 * before calling into namei.  Otherwise, unmount might be
219 	 * started between two vfs_busy() invocations (first is ours,
220 	 * second is from mount point cross-walk code in lookup()),
221 	 * causing deadlock.
222 	 *
223 	 * Avoid unbusying mp if the implementation indicates it has
224 	 * already done so.
225 	 */
226 	if (mp_busy)
227 		vfs_unbusy(mp);
228 	vfs_rel(mp);
229 	return (error);
230 }
231 
232 /*
233  * Used by statfs conversion routines to scale the block size up if
234  * necessary so that all of the block counts are <= 'max_size'.  Note
235  * that 'max_size' should be a bitmask, i.e. 2^n - 1 for some non-zero
236  * value of 'n'.
237  */
238 void
239 statfs_scale_blocks(struct statfs *sf, long max_size)
240 {
241 	uint64_t count;
242 	int shift;
243 
244 	KASSERT(powerof2(max_size + 1), ("%s: invalid max_size", __func__));
245 
246 	/*
247 	 * Attempt to scale the block counts to give a more accurate
248 	 * overview to userland of the ratio of free space to used
249 	 * space.  To do this, find the largest block count and compute
250 	 * a divisor that lets it fit into a signed integer <= max_size.
251 	 */
252 	if (sf->f_bavail < 0)
253 		count = -sf->f_bavail;
254 	else
255 		count = sf->f_bavail;
256 	count = MAX(sf->f_blocks, MAX(sf->f_bfree, count));
257 	if (count <= max_size)
258 		return;
259 
260 	count >>= flsl(max_size);
261 	shift = 0;
262 	while (count > 0) {
263 		shift++;
264 		count >>=1;
265 	}
266 
267 	sf->f_bsize <<= shift;
268 	sf->f_blocks >>= shift;
269 	sf->f_bfree >>= shift;
270 	sf->f_bavail >>= shift;
271 }
272 
273 static int
274 kern_do_statfs(struct thread *td, struct mount *mp, struct statfs *buf)
275 {
276 	int error;
277 
278 	if (mp == NULL)
279 		return (EBADF);
280 	error = vfs_busy(mp, 0);
281 	vfs_rel(mp);
282 	if (error != 0)
283 		return (error);
284 #ifdef MAC
285 	error = mac_mount_check_stat(td->td_ucred, mp);
286 	if (error != 0)
287 		goto out;
288 #endif
289 	error = VFS_STATFS(mp, buf);
290 	if (error != 0)
291 		goto out;
292 	if (priv_check_cred_vfs_generation(td->td_ucred)) {
293 		buf->f_fsid.val[0] = buf->f_fsid.val[1] = 0;
294 		prison_enforce_statfs(td->td_ucred, mp, buf);
295 	}
296 out:
297 	vfs_unbusy(mp);
298 	return (error);
299 }
300 
301 /*
302  * Get filesystem statistics.
303  */
304 #ifndef _SYS_SYSPROTO_H_
305 struct statfs_args {
306 	char *path;
307 	struct statfs *buf;
308 };
309 #endif
310 int
311 sys_statfs(struct thread *td, struct statfs_args *uap)
312 {
313 	struct statfs *sfp;
314 	int error;
315 
316 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
317 	error = kern_statfs(td, uap->path, UIO_USERSPACE, sfp);
318 	if (error == 0)
319 		error = copyout(sfp, uap->buf, sizeof(struct statfs));
320 	free(sfp, M_STATFS);
321 	return (error);
322 }
323 
324 int
325 kern_statfs(struct thread *td, const char *path, enum uio_seg pathseg,
326     struct statfs *buf)
327 {
328 	struct mount *mp;
329 	struct nameidata nd;
330 	int error;
331 
332 	NDINIT(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path);
333 	error = namei(&nd);
334 	if (error != 0)
335 		return (error);
336 	NDFREE_PNBUF(&nd);
337 	mp = vfs_ref_from_vp(nd.ni_vp);
338 	vrele(nd.ni_vp);
339 	return (kern_do_statfs(td, mp, buf));
340 }
341 
342 /*
343  * Get filesystem statistics.
344  */
345 #ifndef _SYS_SYSPROTO_H_
346 struct fstatfs_args {
347 	int fd;
348 	struct statfs *buf;
349 };
350 #endif
351 int
352 sys_fstatfs(struct thread *td, struct fstatfs_args *uap)
353 {
354 	struct statfs *sfp;
355 	int error;
356 
357 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
358 	error = kern_fstatfs(td, uap->fd, sfp);
359 	if (error == 0)
360 		error = copyout(sfp, uap->buf, sizeof(struct statfs));
361 	free(sfp, M_STATFS);
362 	return (error);
363 }
364 
365 int
366 kern_fstatfs(struct thread *td, int fd, struct statfs *buf)
367 {
368 	struct file *fp;
369 	struct mount *mp;
370 	struct vnode *vp;
371 	int error;
372 
373 	AUDIT_ARG_FD(fd);
374 	error = getvnode_path(td, fd, &cap_fstatfs_rights, &fp);
375 	if (error != 0)
376 		return (error);
377 	vp = fp->f_vnode;
378 #ifdef AUDIT
379 	if (AUDITING_TD(td)) {
380 		vn_lock(vp, LK_SHARED | LK_RETRY);
381 		AUDIT_ARG_VNODE1(vp);
382 		VOP_UNLOCK(vp);
383 	}
384 #endif
385 	mp = vfs_ref_from_vp(vp);
386 	fdrop(fp, td);
387 	return (kern_do_statfs(td, mp, buf));
388 }
389 
390 /*
391  * Get statistics on all filesystems.
392  */
393 #ifndef _SYS_SYSPROTO_H_
394 struct getfsstat_args {
395 	struct statfs *buf;
396 	long bufsize;
397 	int mode;
398 };
399 #endif
400 int
401 sys_getfsstat(struct thread *td, struct getfsstat_args *uap)
402 {
403 	size_t count;
404 	int error;
405 
406 	if (uap->bufsize < 0 || uap->bufsize > SIZE_MAX)
407 		return (EINVAL);
408 	error = kern_getfsstat(td, &uap->buf, uap->bufsize, &count,
409 	    UIO_USERSPACE, uap->mode);
410 	if (error == 0)
411 		td->td_retval[0] = count;
412 	return (error);
413 }
414 
415 /*
416  * If (bufsize > 0 && bufseg == UIO_SYSSPACE)
417  *	The caller is responsible for freeing memory which will be allocated
418  *	in '*buf'.
419  */
420 int
421 kern_getfsstat(struct thread *td, struct statfs **buf, size_t bufsize,
422     size_t *countp, enum uio_seg bufseg, int mode)
423 {
424 	struct mount *mp, *nmp;
425 	struct statfs *sfsp, *sp, *sptmp, *tofree;
426 	size_t count, maxcount;
427 	int error;
428 
429 	switch (mode) {
430 	case MNT_WAIT:
431 	case MNT_NOWAIT:
432 		break;
433 	default:
434 		if (bufseg == UIO_SYSSPACE)
435 			*buf = NULL;
436 		return (EINVAL);
437 	}
438 restart:
439 	maxcount = bufsize / sizeof(struct statfs);
440 	if (bufsize == 0) {
441 		sfsp = NULL;
442 		tofree = NULL;
443 	} else if (bufseg == UIO_USERSPACE) {
444 		sfsp = *buf;
445 		tofree = NULL;
446 	} else /* if (bufseg == UIO_SYSSPACE) */ {
447 		count = 0;
448 		mtx_lock(&mountlist_mtx);
449 		TAILQ_FOREACH(mp, &mountlist, mnt_list) {
450 			count++;
451 		}
452 		mtx_unlock(&mountlist_mtx);
453 		if (maxcount > count)
454 			maxcount = count;
455 		tofree = sfsp = *buf = malloc(maxcount * sizeof(struct statfs),
456 		    M_STATFS, M_WAITOK);
457 	}
458 
459 	count = 0;
460 
461 	/*
462 	 * If there is no target buffer they only want the count.
463 	 *
464 	 * This could be TAILQ_FOREACH but it is open-coded to match the original
465 	 * code below.
466 	 */
467 	if (sfsp == NULL) {
468 		mtx_lock(&mountlist_mtx);
469 		for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
470 			if (prison_canseemount(td->td_ucred, mp) != 0) {
471 				nmp = TAILQ_NEXT(mp, mnt_list);
472 				continue;
473 			}
474 #ifdef MAC
475 			if (mac_mount_check_stat(td->td_ucred, mp) != 0) {
476 				nmp = TAILQ_NEXT(mp, mnt_list);
477 				continue;
478 			}
479 #endif
480 			count++;
481 			nmp = TAILQ_NEXT(mp, mnt_list);
482 		}
483 		mtx_unlock(&mountlist_mtx);
484 		*countp = count;
485 		return (0);
486 	}
487 
488 	/*
489 	 * They want the entire thing.
490 	 *
491 	 * Short-circuit the corner case of no room for anything, avoids
492 	 * relocking below.
493 	 */
494 	if (maxcount < 1) {
495 		goto out;
496 	}
497 
498 	mtx_lock(&mountlist_mtx);
499 	for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
500 		if (prison_canseemount(td->td_ucred, mp) != 0) {
501 			nmp = TAILQ_NEXT(mp, mnt_list);
502 			continue;
503 		}
504 #ifdef MAC
505 		if (mac_mount_check_stat(td->td_ucred, mp) != 0) {
506 			nmp = TAILQ_NEXT(mp, mnt_list);
507 			continue;
508 		}
509 #endif
510 		if (mode == MNT_WAIT) {
511 			if (vfs_busy(mp, MBF_MNTLSTLOCK) != 0) {
512 				/*
513 				 * If vfs_busy() failed, and MBF_NOWAIT
514 				 * wasn't passed, then the mp is gone.
515 				 * Furthermore, because of MBF_MNTLSTLOCK,
516 				 * the mountlist_mtx was dropped.  We have
517 				 * no other choice than to start over.
518 				 */
519 				mtx_unlock(&mountlist_mtx);
520 				free(tofree, M_STATFS);
521 				goto restart;
522 			}
523 		} else {
524 			if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK) != 0) {
525 				nmp = TAILQ_NEXT(mp, mnt_list);
526 				continue;
527 			}
528 		}
529 		sp = &mp->mnt_stat;
530 		/*
531 		 * If MNT_NOWAIT is specified, do not refresh
532 		 * the fsstat cache.
533 		 */
534 		if (mode != MNT_NOWAIT) {
535 			error = VFS_STATFS(mp, sp);
536 			if (error != 0) {
537 				mtx_lock(&mountlist_mtx);
538 				nmp = TAILQ_NEXT(mp, mnt_list);
539 				vfs_unbusy(mp);
540 				continue;
541 			}
542 		}
543 		if (priv_check_cred_vfs_generation(td->td_ucred)) {
544 			sptmp = malloc(sizeof(struct statfs), M_STATFS,
545 			    M_WAITOK);
546 			*sptmp = *sp;
547 			sptmp->f_fsid.val[0] = sptmp->f_fsid.val[1] = 0;
548 			prison_enforce_statfs(td->td_ucred, mp, sptmp);
549 			sp = sptmp;
550 		} else
551 			sptmp = NULL;
552 		if (bufseg == UIO_SYSSPACE) {
553 			bcopy(sp, sfsp, sizeof(*sp));
554 			free(sptmp, M_STATFS);
555 		} else /* if (bufseg == UIO_USERSPACE) */ {
556 			error = copyout(sp, sfsp, sizeof(*sp));
557 			free(sptmp, M_STATFS);
558 			if (error != 0) {
559 				vfs_unbusy(mp);
560 				return (error);
561 			}
562 		}
563 		sfsp++;
564 		count++;
565 
566 		if (count == maxcount) {
567 			vfs_unbusy(mp);
568 			goto out;
569 		}
570 
571 		mtx_lock(&mountlist_mtx);
572 		nmp = TAILQ_NEXT(mp, mnt_list);
573 		vfs_unbusy(mp);
574 	}
575 	mtx_unlock(&mountlist_mtx);
576 out:
577 	*countp = count;
578 	return (0);
579 }
580 
581 #ifdef COMPAT_FREEBSD4
582 /*
583  * Get old format filesystem statistics.
584  */
585 static void freebsd4_cvtstatfs(struct statfs *, struct ostatfs *);
586 
587 #ifndef _SYS_SYSPROTO_H_
588 struct freebsd4_statfs_args {
589 	char *path;
590 	struct ostatfs *buf;
591 };
592 #endif
593 int
594 freebsd4_statfs(struct thread *td, struct freebsd4_statfs_args *uap)
595 {
596 	struct ostatfs osb;
597 	struct statfs *sfp;
598 	int error;
599 
600 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
601 	error = kern_statfs(td, uap->path, UIO_USERSPACE, sfp);
602 	if (error == 0) {
603 		freebsd4_cvtstatfs(sfp, &osb);
604 		error = copyout(&osb, uap->buf, sizeof(osb));
605 	}
606 	free(sfp, M_STATFS);
607 	return (error);
608 }
609 
610 /*
611  * Get filesystem statistics.
612  */
613 #ifndef _SYS_SYSPROTO_H_
614 struct freebsd4_fstatfs_args {
615 	int fd;
616 	struct ostatfs *buf;
617 };
618 #endif
619 int
620 freebsd4_fstatfs(struct thread *td, struct freebsd4_fstatfs_args *uap)
621 {
622 	struct ostatfs osb;
623 	struct statfs *sfp;
624 	int error;
625 
626 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
627 	error = kern_fstatfs(td, uap->fd, sfp);
628 	if (error == 0) {
629 		freebsd4_cvtstatfs(sfp, &osb);
630 		error = copyout(&osb, uap->buf, sizeof(osb));
631 	}
632 	free(sfp, M_STATFS);
633 	return (error);
634 }
635 
636 /*
637  * Get statistics on all filesystems.
638  */
639 #ifndef _SYS_SYSPROTO_H_
640 struct freebsd4_getfsstat_args {
641 	struct ostatfs *buf;
642 	long bufsize;
643 	int mode;
644 };
645 #endif
646 int
647 freebsd4_getfsstat(struct thread *td, struct freebsd4_getfsstat_args *uap)
648 {
649 	struct statfs *buf, *sp;
650 	struct ostatfs osb;
651 	size_t count, size;
652 	int error;
653 
654 	if (uap->bufsize < 0)
655 		return (EINVAL);
656 	count = uap->bufsize / sizeof(struct ostatfs);
657 	if (count > SIZE_MAX / sizeof(struct statfs))
658 		return (EINVAL);
659 	size = count * sizeof(struct statfs);
660 	error = kern_getfsstat(td, &buf, size, &count, UIO_SYSSPACE,
661 	    uap->mode);
662 	if (error == 0)
663 		td->td_retval[0] = count;
664 	if (size != 0) {
665 		sp = buf;
666 		while (count != 0 && error == 0) {
667 			freebsd4_cvtstatfs(sp, &osb);
668 			error = copyout(&osb, uap->buf, sizeof(osb));
669 			sp++;
670 			uap->buf++;
671 			count--;
672 		}
673 		free(buf, M_STATFS);
674 	}
675 	return (error);
676 }
677 
678 /*
679  * Implement fstatfs() for (NFS) file handles.
680  */
681 #ifndef _SYS_SYSPROTO_H_
682 struct freebsd4_fhstatfs_args {
683 	struct fhandle *u_fhp;
684 	struct ostatfs *buf;
685 };
686 #endif
687 int
688 freebsd4_fhstatfs(struct thread *td, struct freebsd4_fhstatfs_args *uap)
689 {
690 	struct ostatfs osb;
691 	struct statfs *sfp;
692 	fhandle_t fh;
693 	int error;
694 
695 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
696 	if (error != 0)
697 		return (error);
698 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
699 	error = kern_fhstatfs(td, fh, sfp);
700 	if (error == 0) {
701 		freebsd4_cvtstatfs(sfp, &osb);
702 		error = copyout(&osb, uap->buf, sizeof(osb));
703 	}
704 	free(sfp, M_STATFS);
705 	return (error);
706 }
707 
708 /*
709  * Convert a new format statfs structure to an old format statfs structure.
710  */
711 static void
712 freebsd4_cvtstatfs(struct statfs *nsp, struct ostatfs *osp)
713 {
714 
715 	statfs_scale_blocks(nsp, LONG_MAX);
716 	bzero(osp, sizeof(*osp));
717 	osp->f_bsize = nsp->f_bsize;
718 	osp->f_iosize = MIN(nsp->f_iosize, LONG_MAX);
719 	osp->f_blocks = nsp->f_blocks;
720 	osp->f_bfree = nsp->f_bfree;
721 	osp->f_bavail = nsp->f_bavail;
722 	osp->f_files = MIN(nsp->f_files, LONG_MAX);
723 	osp->f_ffree = MIN(nsp->f_ffree, LONG_MAX);
724 	osp->f_owner = nsp->f_owner;
725 	osp->f_type = nsp->f_type;
726 	osp->f_flags = nsp->f_flags;
727 	osp->f_syncwrites = MIN(nsp->f_syncwrites, LONG_MAX);
728 	osp->f_asyncwrites = MIN(nsp->f_asyncwrites, LONG_MAX);
729 	osp->f_syncreads = MIN(nsp->f_syncreads, LONG_MAX);
730 	osp->f_asyncreads = MIN(nsp->f_asyncreads, LONG_MAX);
731 	strlcpy(osp->f_fstypename, nsp->f_fstypename,
732 	    MIN(MFSNAMELEN, OMFSNAMELEN));
733 	strlcpy(osp->f_mntonname, nsp->f_mntonname,
734 	    MIN(MNAMELEN, OMNAMELEN));
735 	strlcpy(osp->f_mntfromname, nsp->f_mntfromname,
736 	    MIN(MNAMELEN, OMNAMELEN));
737 	osp->f_fsid = nsp->f_fsid;
738 }
739 #endif /* COMPAT_FREEBSD4 */
740 
741 #if defined(COMPAT_FREEBSD11)
742 /*
743  * Get old format filesystem statistics.
744  */
745 static void freebsd11_cvtstatfs(struct statfs *, struct freebsd11_statfs *);
746 
747 int
748 freebsd11_statfs(struct thread *td, struct freebsd11_statfs_args *uap)
749 {
750 	struct freebsd11_statfs osb;
751 	struct statfs *sfp;
752 	int error;
753 
754 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
755 	error = kern_statfs(td, uap->path, UIO_USERSPACE, sfp);
756 	if (error == 0) {
757 		freebsd11_cvtstatfs(sfp, &osb);
758 		error = copyout(&osb, uap->buf, sizeof(osb));
759 	}
760 	free(sfp, M_STATFS);
761 	return (error);
762 }
763 
764 /*
765  * Get filesystem statistics.
766  */
767 int
768 freebsd11_fstatfs(struct thread *td, struct freebsd11_fstatfs_args *uap)
769 {
770 	struct freebsd11_statfs osb;
771 	struct statfs *sfp;
772 	int error;
773 
774 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
775 	error = kern_fstatfs(td, uap->fd, sfp);
776 	if (error == 0) {
777 		freebsd11_cvtstatfs(sfp, &osb);
778 		error = copyout(&osb, uap->buf, sizeof(osb));
779 	}
780 	free(sfp, M_STATFS);
781 	return (error);
782 }
783 
784 /*
785  * Get statistics on all filesystems.
786  */
787 int
788 freebsd11_getfsstat(struct thread *td, struct freebsd11_getfsstat_args *uap)
789 {
790 	return (kern_freebsd11_getfsstat(td, uap->buf, uap->bufsize, uap->mode));
791 }
792 
793 int
794 kern_freebsd11_getfsstat(struct thread *td, struct freebsd11_statfs * ubuf,
795     long bufsize, int mode)
796 {
797 	struct freebsd11_statfs osb;
798 	struct statfs *buf, *sp;
799 	size_t count, size;
800 	int error;
801 
802 	if (bufsize < 0)
803 		return (EINVAL);
804 
805 	count = bufsize / sizeof(struct ostatfs);
806 	size = count * sizeof(struct statfs);
807 	error = kern_getfsstat(td, &buf, size, &count, UIO_SYSSPACE, mode);
808 	if (error == 0)
809 		td->td_retval[0] = count;
810 	if (size > 0) {
811 		sp = buf;
812 		while (count > 0 && error == 0) {
813 			freebsd11_cvtstatfs(sp, &osb);
814 			error = copyout(&osb, ubuf, sizeof(osb));
815 			sp++;
816 			ubuf++;
817 			count--;
818 		}
819 		free(buf, M_STATFS);
820 	}
821 	return (error);
822 }
823 
824 /*
825  * Implement fstatfs() for (NFS) file handles.
826  */
827 int
828 freebsd11_fhstatfs(struct thread *td, struct freebsd11_fhstatfs_args *uap)
829 {
830 	struct freebsd11_statfs osb;
831 	struct statfs *sfp;
832 	fhandle_t fh;
833 	int error;
834 
835 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
836 	if (error)
837 		return (error);
838 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
839 	error = kern_fhstatfs(td, fh, sfp);
840 	if (error == 0) {
841 		freebsd11_cvtstatfs(sfp, &osb);
842 		error = copyout(&osb, uap->buf, sizeof(osb));
843 	}
844 	free(sfp, M_STATFS);
845 	return (error);
846 }
847 
848 /*
849  * Convert a new format statfs structure to an old format statfs structure.
850  */
851 static void
852 freebsd11_cvtstatfs(struct statfs *nsp, struct freebsd11_statfs *osp)
853 {
854 
855 	bzero(osp, sizeof(*osp));
856 	osp->f_version = FREEBSD11_STATFS_VERSION;
857 	osp->f_type = nsp->f_type;
858 	osp->f_flags = nsp->f_flags;
859 	osp->f_bsize = nsp->f_bsize;
860 	osp->f_iosize = nsp->f_iosize;
861 	osp->f_blocks = nsp->f_blocks;
862 	osp->f_bfree = nsp->f_bfree;
863 	osp->f_bavail = nsp->f_bavail;
864 	osp->f_files = nsp->f_files;
865 	osp->f_ffree = nsp->f_ffree;
866 	osp->f_syncwrites = nsp->f_syncwrites;
867 	osp->f_asyncwrites = nsp->f_asyncwrites;
868 	osp->f_syncreads = nsp->f_syncreads;
869 	osp->f_asyncreads = nsp->f_asyncreads;
870 	osp->f_namemax = nsp->f_namemax;
871 	osp->f_owner = nsp->f_owner;
872 	osp->f_fsid = nsp->f_fsid;
873 	strlcpy(osp->f_fstypename, nsp->f_fstypename,
874 	    MIN(MFSNAMELEN, sizeof(osp->f_fstypename)));
875 	strlcpy(osp->f_mntonname, nsp->f_mntonname,
876 	    MIN(MNAMELEN, sizeof(osp->f_mntonname)));
877 	strlcpy(osp->f_mntfromname, nsp->f_mntfromname,
878 	    MIN(MNAMELEN, sizeof(osp->f_mntfromname)));
879 }
880 #endif /* COMPAT_FREEBSD11 */
881 
882 /*
883  * Change current working directory to a given file descriptor.
884  */
885 #ifndef _SYS_SYSPROTO_H_
886 struct fchdir_args {
887 	int	fd;
888 };
889 #endif
890 int
891 sys_fchdir(struct thread *td, struct fchdir_args *uap)
892 {
893 	struct vnode *vp, *tdp;
894 	struct mount *mp;
895 	struct file *fp;
896 	int error;
897 
898 	AUDIT_ARG_FD(uap->fd);
899 	error = getvnode_path(td, uap->fd, &cap_fchdir_rights,
900 	    &fp);
901 	if (error != 0)
902 		return (error);
903 	vp = fp->f_vnode;
904 	vrefact(vp);
905 	fdrop(fp, td);
906 	vn_lock(vp, LK_SHARED | LK_RETRY);
907 	AUDIT_ARG_VNODE1(vp);
908 	error = change_dir(vp, td);
909 	while (!error && (mp = vp->v_mountedhere) != NULL) {
910 		if (vfs_busy(mp, 0))
911 			continue;
912 		error = VFS_ROOT(mp, LK_SHARED, &tdp);
913 		vfs_unbusy(mp);
914 		if (error != 0)
915 			break;
916 		vput(vp);
917 		vp = tdp;
918 	}
919 	if (error != 0) {
920 		vput(vp);
921 		return (error);
922 	}
923 	VOP_UNLOCK(vp);
924 	pwd_chdir(td, vp);
925 	return (0);
926 }
927 
928 /*
929  * Change current working directory (``.'').
930  */
931 #ifndef _SYS_SYSPROTO_H_
932 struct chdir_args {
933 	char	*path;
934 };
935 #endif
936 int
937 sys_chdir(struct thread *td, struct chdir_args *uap)
938 {
939 
940 	return (kern_chdir(td, uap->path, UIO_USERSPACE));
941 }
942 
943 int
944 kern_chdir(struct thread *td, const char *path, enum uio_seg pathseg)
945 {
946 	struct nameidata nd;
947 	int error;
948 
949 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
950 	    pathseg, path);
951 	if ((error = namei(&nd)) != 0)
952 		return (error);
953 	if ((error = change_dir(nd.ni_vp, td)) != 0) {
954 		vput(nd.ni_vp);
955 		NDFREE_PNBUF(&nd);
956 		return (error);
957 	}
958 	VOP_UNLOCK(nd.ni_vp);
959 	NDFREE_PNBUF(&nd);
960 	pwd_chdir(td, nd.ni_vp);
961 	return (0);
962 }
963 
964 static int unprivileged_chroot = 0;
965 SYSCTL_INT(_security_bsd, OID_AUTO, unprivileged_chroot, CTLFLAG_RW,
966     &unprivileged_chroot, 0,
967     "Unprivileged processes can use chroot(2)");
968 
969 /*
970  * Takes locked vnode, unlocks it before returning.
971  */
972 static int
973 kern_chroot(struct thread *td, struct vnode *vp)
974 {
975 	struct proc *p;
976 	int error;
977 
978 	error = priv_check(td, PRIV_VFS_CHROOT);
979 	if (error != 0) {
980 		p = td->td_proc;
981 		PROC_LOCK(p);
982 		if (unprivileged_chroot == 0 ||
983 		    (p->p_flag2 & P2_NO_NEW_PRIVS) == 0) {
984 			PROC_UNLOCK(p);
985 			goto e_vunlock;
986 		}
987 		PROC_UNLOCK(p);
988 	}
989 
990 	error = change_dir(vp, td);
991 	if (error != 0)
992 		goto e_vunlock;
993 #ifdef MAC
994 	error = mac_vnode_check_chroot(td->td_ucred, vp);
995 	if (error != 0)
996 		goto e_vunlock;
997 #endif
998 	VOP_UNLOCK(vp);
999 	error = pwd_chroot(td, vp);
1000 	vrele(vp);
1001 	return (error);
1002 e_vunlock:
1003 	vput(vp);
1004 	return (error);
1005 }
1006 
1007 /*
1008  * Change notion of root (``/'') directory.
1009  */
1010 #ifndef _SYS_SYSPROTO_H_
1011 struct chroot_args {
1012 	char	*path;
1013 };
1014 #endif
1015 int
1016 sys_chroot(struct thread *td, struct chroot_args *uap)
1017 {
1018 	struct nameidata nd;
1019 	int error;
1020 
1021 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
1022 	    UIO_USERSPACE, uap->path);
1023 	error = namei(&nd);
1024 	if (error != 0)
1025 		return (error);
1026 	NDFREE_PNBUF(&nd);
1027 	error = kern_chroot(td, nd.ni_vp);
1028 	return (error);
1029 }
1030 
1031 /*
1032  * Change notion of root directory to a given file descriptor.
1033  */
1034 #ifndef _SYS_SYSPROTO_H_
1035 struct fchroot_args {
1036 	int	fd;
1037 };
1038 #endif
1039 int
1040 sys_fchroot(struct thread *td, struct fchroot_args *uap)
1041 {
1042 	struct vnode *vp;
1043 	struct file *fp;
1044 	int error;
1045 
1046 	error = getvnode_path(td, uap->fd, &cap_fchroot_rights, &fp);
1047 	if (error != 0)
1048 		return (error);
1049 	vp = fp->f_vnode;
1050 	vrefact(vp);
1051 	fdrop(fp, td);
1052 	vn_lock(vp, LK_SHARED | LK_RETRY);
1053 	error = kern_chroot(td, vp);
1054 	return (error);
1055 }
1056 
1057 /*
1058  * Common routine for chroot and chdir.  Callers must provide a locked vnode
1059  * instance.
1060  */
1061 int
1062 change_dir(struct vnode *vp, struct thread *td)
1063 {
1064 #ifdef MAC
1065 	int error;
1066 #endif
1067 
1068 	ASSERT_VOP_LOCKED(vp, "change_dir(): vp not locked");
1069 	if (vp->v_type != VDIR)
1070 		return (ENOTDIR);
1071 #ifdef MAC
1072 	error = mac_vnode_check_chdir(td->td_ucred, vp);
1073 	if (error != 0)
1074 		return (error);
1075 #endif
1076 	return (VOP_ACCESS(vp, VEXEC, td->td_ucred, td));
1077 }
1078 
1079 static __inline void
1080 flags_to_rights(int flags, cap_rights_t *rightsp)
1081 {
1082 	if (flags & O_EXEC) {
1083 		cap_rights_set_one(rightsp, CAP_FEXECVE);
1084 		if (flags & O_PATH)
1085 			return;
1086 	} else {
1087 		switch ((flags & O_ACCMODE)) {
1088 		case O_RDONLY:
1089 			cap_rights_set_one(rightsp, CAP_READ);
1090 			break;
1091 		case O_RDWR:
1092 			cap_rights_set_one(rightsp, CAP_READ);
1093 			/* FALLTHROUGH */
1094 		case O_WRONLY:
1095 			cap_rights_set_one(rightsp, CAP_WRITE);
1096 			if (!(flags & (O_APPEND | O_TRUNC)))
1097 				cap_rights_set_one(rightsp, CAP_SEEK);
1098 			break;
1099 		}
1100 	}
1101 
1102 	if (flags & O_CREAT)
1103 		cap_rights_set_one(rightsp, CAP_CREATE);
1104 
1105 	if (flags & O_TRUNC)
1106 		cap_rights_set_one(rightsp, CAP_FTRUNCATE);
1107 
1108 	if (flags & (O_SYNC | O_FSYNC))
1109 		cap_rights_set_one(rightsp, CAP_FSYNC);
1110 
1111 	if (flags & (O_EXLOCK | O_SHLOCK))
1112 		cap_rights_set_one(rightsp, CAP_FLOCK);
1113 }
1114 
1115 /*
1116  * Check permissions, allocate an open file structure, and call the device
1117  * open routine if any.
1118  */
1119 #ifndef _SYS_SYSPROTO_H_
1120 struct open_args {
1121 	char	*path;
1122 	int	flags;
1123 	int	mode;
1124 };
1125 #endif
1126 int
1127 sys_open(struct thread *td, struct open_args *uap)
1128 {
1129 
1130 	return (kern_openat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1131 	    uap->flags, uap->mode));
1132 }
1133 
1134 #ifndef _SYS_SYSPROTO_H_
1135 struct openat_args {
1136 	int	fd;
1137 	char	*path;
1138 	int	flag;
1139 	int	mode;
1140 };
1141 #endif
1142 int
1143 sys_openat(struct thread *td, struct openat_args *uap)
1144 {
1145 
1146 	AUDIT_ARG_FD(uap->fd);
1147 	return (kern_openat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
1148 	    uap->mode));
1149 }
1150 
1151 /*
1152  * Validate open(2) flags and convert access mode flags (O_RDONLY etc.) to their
1153  * in-kernel representations (FREAD etc.).
1154  */
1155 static int
1156 openflags(int *flagsp)
1157 {
1158 	int flags;
1159 
1160 	/*
1161 	 * Only one of the O_EXEC, O_RDONLY, O_WRONLY and O_RDWR flags
1162 	 * may be specified.  On the other hand, for O_PATH any mode
1163 	 * except O_EXEC is ignored.
1164 	 */
1165 	flags = *flagsp;
1166 	if ((flags & O_PATH) != 0) {
1167 		flags &= ~O_ACCMODE;
1168 	} else if ((flags & O_EXEC) != 0) {
1169 		if ((flags & O_ACCMODE) != 0)
1170 			return (EINVAL);
1171 	} else if ((flags & O_ACCMODE) == O_ACCMODE) {
1172 		return (EINVAL);
1173 	} else {
1174 		flags = FFLAGS(flags);
1175 	}
1176 	*flagsp = flags;
1177 	return (0);
1178 }
1179 
1180 static void
1181 finit_open(struct file *fp, struct vnode *vp, int flags)
1182 {
1183 	/*
1184 	 * Store the vnode, for any f_type. Typically, the vnode use count is
1185 	 * decremented by a direct call to vnops.fo_close() for files that
1186 	 * switched type.
1187 	 */
1188 	fp->f_vnode = vp;
1189 
1190 	/*
1191 	 * If the file wasn't claimed by devfs or fifofs, bind it to the normal
1192 	 * vnode operations here.
1193 	 */
1194 	if (fp->f_ops == &badfileops) {
1195 		KASSERT(vp->v_type != VFIFO || (flags & O_PATH) != 0,
1196 		    ("Unexpected fifo fp %p vp %p", fp, vp));
1197 		if ((flags & O_PATH) != 0) {
1198 			finit(fp, (flags & FMASK) | (fp->f_flag & FKQALLOWED),
1199 			    DTYPE_VNODE, NULL, &path_fileops);
1200 		} else {
1201 			finit_vnode(fp, flags, NULL, &vnops);
1202 		}
1203 	}
1204 }
1205 
1206 /*
1207  * If fpp != NULL, opened file is not installed into the file
1208  * descriptor table, instead it is returned in *fpp.  This is
1209  * incompatible with fdopen(), in which case we return EINVAL.
1210  */
1211 static int
1212 openatfp(struct thread *td, int dirfd, const char *path,
1213     enum uio_seg pathseg, int flags, int mode, struct file **fpp)
1214 {
1215 	struct proc *p;
1216 	struct filedesc *fdp;
1217 	struct pwddesc *pdp;
1218 	struct file *fp;
1219 	struct vnode *vp;
1220 	struct filecaps *fcaps;
1221 	struct nameidata nd;
1222 	cap_rights_t rights;
1223 	int cmode, error, indx;
1224 
1225 	indx = -1;
1226 	p = td->td_proc;
1227 	fdp = p->p_fd;
1228 	pdp = p->p_pd;
1229 
1230 	AUDIT_ARG_FFLAGS(flags);
1231 	AUDIT_ARG_MODE(mode);
1232 	cap_rights_init_one(&rights, CAP_LOOKUP);
1233 	flags_to_rights(flags, &rights);
1234 
1235 	error = openflags(&flags);
1236 	if (error != 0)
1237 		return (error);
1238 
1239 	/*
1240 	 * Allocate a file structure. The descriptor to reference it
1241 	 * is allocated and used by finstall_refed() below.
1242 	 */
1243 	error = falloc_noinstall(td, &fp);
1244 	if (error != 0)
1245 		return (error);
1246 	/* Set the flags early so the finit in devfs can pick them up. */
1247 	fp->f_flag = flags & FMASK;
1248 	cmode = ((mode & ~pdp->pd_cmask) & ALLPERMS) & ~S_ISTXT;
1249 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1 | WANTIOCTLCAPS,
1250 	    pathseg, path, dirfd, &rights);
1251 	td->td_dupfd = -1;		/* XXX check for fdopen */
1252 	error = vn_open_cred(&nd, &flags, cmode, VN_OPEN_WANTIOCTLCAPS,
1253 	    td->td_ucred, fp);
1254 	if (error != 0) {
1255 		/*
1256 		 * If the vn_open replaced the method vector, something
1257 		 * wonderous happened deep below and we just pass it up
1258 		 * pretending we know what we do.
1259 		 */
1260 		if (error == ENXIO && fp->f_ops != &badfileops) {
1261 			MPASS((flags & O_PATH) == 0);
1262 			goto success;
1263 		}
1264 
1265 		/*
1266 		 * Handle special fdopen() case. bleh.
1267 		 *
1268 		 * Don't do this for relative (capability) lookups; we don't
1269 		 * understand exactly what would happen, and we don't think
1270 		 * that it ever should.
1271 		 */
1272 		if ((nd.ni_resflags & NIRES_STRICTREL) == 0 &&
1273 		    (error == ENODEV || error == ENXIO) &&
1274 		    td->td_dupfd >= 0) {
1275 			MPASS(fpp == NULL);
1276 			error = dupfdopen(td, fdp, td->td_dupfd, flags, error,
1277 			    &indx);
1278 			if (error == 0)
1279 				goto success;
1280 		}
1281 
1282 		goto bad;
1283 	}
1284 	td->td_dupfd = 0;
1285 	NDFREE_PNBUF(&nd);
1286 	vp = nd.ni_vp;
1287 
1288 	finit_open(fp, vp, flags);
1289 	VOP_UNLOCK(vp);
1290 	if (flags & O_TRUNC) {
1291 		error = fo_truncate(fp, 0, td->td_ucred, td);
1292 		if (error != 0)
1293 			goto bad;
1294 	}
1295 success:
1296 	if (fpp != NULL) {
1297 		MPASS(error == 0);
1298 		NDFREE_IOCTLCAPS(&nd);
1299 		*fpp = fp;
1300 		return (0);
1301 	}
1302 
1303 	/*
1304 	 * If we haven't already installed the FD (for dupfdopen), do so now.
1305 	 */
1306 	if (indx == -1) {
1307 #ifdef CAPABILITIES
1308 		if ((nd.ni_resflags & NIRES_STRICTREL) != 0)
1309 			fcaps = &nd.ni_filecaps;
1310 		else
1311 #endif
1312 			fcaps = NULL;
1313 		error = finstall_refed(td, fp, &indx, flags, fcaps);
1314 		/* On success finstall_refed() consumes fcaps. */
1315 		if (error != 0) {
1316 			goto bad;
1317 		}
1318 	} else {
1319 		NDFREE_IOCTLCAPS(&nd);
1320 		falloc_abort(td, fp);
1321 	}
1322 
1323 	td->td_retval[0] = indx;
1324 	return (0);
1325 bad:
1326 	KASSERT(indx == -1, ("indx=%d, should be -1", indx));
1327 	NDFREE_IOCTLCAPS(&nd);
1328 	falloc_abort(td, fp);
1329 	return (error);
1330 }
1331 
1332 int
1333 kern_openat(struct thread *td, int dirfd, const char *path,
1334     enum uio_seg pathseg, int flags, int mode)
1335 {
1336 	return (openatfp(td, dirfd, path, pathseg, flags, mode, NULL));
1337 }
1338 
1339 int
1340 kern_openatfp(struct thread *td, int dirfd, const char *path,
1341     enum uio_seg pathseg, int flags, int mode, struct file **fpp)
1342 {
1343 	int error, old_dupfd;
1344 
1345 	old_dupfd = td->td_dupfd;
1346 	td->td_dupfd = -1;
1347 	error = openatfp(td, dirfd, path, pathseg, flags, mode, fpp);
1348 	td->td_dupfd = old_dupfd;
1349 	return (error);
1350 }
1351 
1352 #ifdef COMPAT_43
1353 /*
1354  * Create a file.
1355  */
1356 #ifndef _SYS_SYSPROTO_H_
1357 struct ocreat_args {
1358 	char	*path;
1359 	int	mode;
1360 };
1361 #endif
1362 int
1363 ocreat(struct thread *td, struct ocreat_args *uap)
1364 {
1365 
1366 	return (kern_openat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1367 	    O_WRONLY | O_CREAT | O_TRUNC, uap->mode));
1368 }
1369 #endif /* COMPAT_43 */
1370 
1371 /*
1372  * Create a special file.
1373  */
1374 #ifndef _SYS_SYSPROTO_H_
1375 struct mknodat_args {
1376 	int	fd;
1377 	char	*path;
1378 	mode_t	mode;
1379 	dev_t	dev;
1380 };
1381 #endif
1382 int
1383 sys_mknodat(struct thread *td, struct mknodat_args *uap)
1384 {
1385 
1386 	return (kern_mknodat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode,
1387 	    uap->dev));
1388 }
1389 
1390 #if defined(COMPAT_FREEBSD11)
1391 int
1392 freebsd11_mknod(struct thread *td,
1393     struct freebsd11_mknod_args *uap)
1394 {
1395 
1396 	return (kern_mknodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1397 	    uap->mode, uap->dev));
1398 }
1399 
1400 int
1401 freebsd11_mknodat(struct thread *td,
1402     struct freebsd11_mknodat_args *uap)
1403 {
1404 
1405 	return (kern_mknodat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode,
1406 	    uap->dev));
1407 }
1408 #endif /* COMPAT_FREEBSD11 */
1409 
1410 int
1411 kern_mknodat(struct thread *td, int fd, const char *path, enum uio_seg pathseg,
1412     int mode, dev_t dev)
1413 {
1414 	struct vnode *vp;
1415 	struct mount *mp;
1416 	struct vattr vattr;
1417 	struct nameidata nd;
1418 	int error, whiteout = 0;
1419 
1420 	AUDIT_ARG_MODE(mode);
1421 	AUDIT_ARG_DEV(dev);
1422 	switch (mode & S_IFMT) {
1423 	case S_IFCHR:
1424 	case S_IFBLK:
1425 		error = priv_check(td, PRIV_VFS_MKNOD_DEV);
1426 		if (error == 0 && dev == VNOVAL)
1427 			error = EINVAL;
1428 		break;
1429 	case S_IFWHT:
1430 		error = priv_check(td, PRIV_VFS_MKNOD_WHT);
1431 		break;
1432 	case S_IFIFO:
1433 		if (dev == 0)
1434 			return (kern_mkfifoat(td, fd, path, pathseg, mode));
1435 		/* FALLTHROUGH */
1436 	default:
1437 		error = EINVAL;
1438 		break;
1439 	}
1440 	if (error != 0)
1441 		return (error);
1442 	NDPREINIT(&nd);
1443 restart:
1444 	bwillwrite();
1445 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | AUDITVNODE1 | NOCACHE,
1446 	    pathseg, path, fd, &cap_mknodat_rights);
1447 	if ((error = namei(&nd)) != 0)
1448 		return (error);
1449 	vp = nd.ni_vp;
1450 	if (vp != NULL) {
1451 		NDFREE_PNBUF(&nd);
1452 		if (vp == nd.ni_dvp)
1453 			vrele(nd.ni_dvp);
1454 		else
1455 			vput(nd.ni_dvp);
1456 		vrele(vp);
1457 		return (EEXIST);
1458 	} else if ((vn_irflag_read(nd.ni_dvp) & VIRF_NAMEDDIR) != 0) {
1459 		NDFREE_PNBUF(&nd);
1460 		vput(nd.ni_dvp);
1461 		return (EINVAL);
1462 	} else {
1463 		VATTR_NULL(&vattr);
1464 		vattr.va_mode = (mode & ALLPERMS) &
1465 		    ~td->td_proc->p_pd->pd_cmask;
1466 		vattr.va_rdev = dev;
1467 		whiteout = 0;
1468 
1469 		switch (mode & S_IFMT) {
1470 		case S_IFCHR:
1471 			vattr.va_type = VCHR;
1472 			break;
1473 		case S_IFBLK:
1474 			vattr.va_type = VBLK;
1475 			break;
1476 		case S_IFWHT:
1477 			whiteout = 1;
1478 			break;
1479 		default:
1480 			panic("kern_mknod: invalid mode");
1481 		}
1482 	}
1483 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1484 		NDFREE_PNBUF(&nd);
1485 		vput(nd.ni_dvp);
1486 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | V_PCATCH)) != 0)
1487 			return (error);
1488 		goto restart;
1489 	}
1490 #ifdef MAC
1491 	if (error == 0 && !whiteout)
1492 		error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp,
1493 		    &nd.ni_cnd, &vattr);
1494 #endif
1495 	if (error == 0) {
1496 		if (whiteout)
1497 			error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE);
1498 		else {
1499 			error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp,
1500 						&nd.ni_cnd, &vattr);
1501 		}
1502 	}
1503 	VOP_VPUT_PAIR(nd.ni_dvp, error == 0 && !whiteout ? &nd.ni_vp : NULL,
1504 	    true);
1505 	vn_finished_write(mp);
1506 	NDFREE_PNBUF(&nd);
1507 	if (error == ERELOOKUP)
1508 		goto restart;
1509 	return (error);
1510 }
1511 
1512 /*
1513  * Create a named pipe.
1514  */
1515 #ifndef _SYS_SYSPROTO_H_
1516 struct mkfifo_args {
1517 	char	*path;
1518 	int	mode;
1519 };
1520 #endif
1521 int
1522 sys_mkfifo(struct thread *td, struct mkfifo_args *uap)
1523 {
1524 
1525 	return (kern_mkfifoat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1526 	    uap->mode));
1527 }
1528 
1529 #ifndef _SYS_SYSPROTO_H_
1530 struct mkfifoat_args {
1531 	int	fd;
1532 	char	*path;
1533 	mode_t	mode;
1534 };
1535 #endif
1536 int
1537 sys_mkfifoat(struct thread *td, struct mkfifoat_args *uap)
1538 {
1539 
1540 	return (kern_mkfifoat(td, uap->fd, uap->path, UIO_USERSPACE,
1541 	    uap->mode));
1542 }
1543 
1544 int
1545 kern_mkfifoat(struct thread *td, int fd, const char *path,
1546     enum uio_seg pathseg, int mode)
1547 {
1548 	struct mount *mp;
1549 	struct vattr vattr;
1550 	struct nameidata nd;
1551 	int error;
1552 
1553 	AUDIT_ARG_MODE(mode);
1554 	NDPREINIT(&nd);
1555 restart:
1556 	bwillwrite();
1557 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | AUDITVNODE1 | NOCACHE,
1558 	    pathseg, path, fd, &cap_mkfifoat_rights);
1559 	if ((error = namei(&nd)) != 0)
1560 		return (error);
1561 	if (nd.ni_vp != NULL) {
1562 		NDFREE_PNBUF(&nd);
1563 		if (nd.ni_vp == nd.ni_dvp)
1564 			vrele(nd.ni_dvp);
1565 		else
1566 			vput(nd.ni_dvp);
1567 		vrele(nd.ni_vp);
1568 		return (EEXIST);
1569 	}
1570 	if ((vn_irflag_read(nd.ni_dvp) & VIRF_NAMEDDIR) != 0) {
1571 		NDFREE_PNBUF(&nd);
1572 		vput(nd.ni_dvp);
1573 		return (EINVAL);
1574 	}
1575 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1576 		NDFREE_PNBUF(&nd);
1577 		vput(nd.ni_dvp);
1578 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | V_PCATCH)) != 0)
1579 			return (error);
1580 		goto restart;
1581 	}
1582 	VATTR_NULL(&vattr);
1583 	vattr.va_type = VFIFO;
1584 	vattr.va_mode = (mode & ALLPERMS) & ~td->td_proc->p_pd->pd_cmask;
1585 #ifdef MAC
1586 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1587 	    &vattr);
1588 	if (error != 0)
1589 		goto out;
1590 #endif
1591 	error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
1592 #ifdef MAC
1593 out:
1594 #endif
1595 	VOP_VPUT_PAIR(nd.ni_dvp, error == 0 ? &nd.ni_vp : NULL, true);
1596 	vn_finished_write(mp);
1597 	NDFREE_PNBUF(&nd);
1598 	if (error == ERELOOKUP)
1599 		goto restart;
1600 	return (error);
1601 }
1602 
1603 /*
1604  * Make a hard file link.
1605  */
1606 #ifndef _SYS_SYSPROTO_H_
1607 struct link_args {
1608 	char	*path;
1609 	char	*link;
1610 };
1611 #endif
1612 int
1613 sys_link(struct thread *td, struct link_args *uap)
1614 {
1615 
1616 	return (kern_linkat(td, AT_FDCWD, AT_FDCWD, uap->path, uap->link,
1617 	    UIO_USERSPACE, AT_SYMLINK_FOLLOW));
1618 }
1619 
1620 #ifndef _SYS_SYSPROTO_H_
1621 struct linkat_args {
1622 	int	fd1;
1623 	char	*path1;
1624 	int	fd2;
1625 	char	*path2;
1626 	int	flag;
1627 };
1628 #endif
1629 int
1630 sys_linkat(struct thread *td, struct linkat_args *uap)
1631 {
1632 
1633 	return (kern_linkat(td, uap->fd1, uap->fd2, uap->path1, uap->path2,
1634 	    UIO_USERSPACE, uap->flag));
1635 }
1636 
1637 int hardlink_check_uid = 0;
1638 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_uid, CTLFLAG_RW,
1639     &hardlink_check_uid, 0,
1640     "Unprivileged processes cannot create hard links to files owned by other "
1641     "users");
1642 static int hardlink_check_gid = 0;
1643 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_gid, CTLFLAG_RW,
1644     &hardlink_check_gid, 0,
1645     "Unprivileged processes cannot create hard links to files owned by other "
1646     "groups");
1647 
1648 static int
1649 can_hardlink(struct vnode *vp, struct ucred *cred)
1650 {
1651 	struct vattr va;
1652 	int error;
1653 
1654 	if (!hardlink_check_uid && !hardlink_check_gid)
1655 		return (0);
1656 
1657 	error = VOP_GETATTR(vp, &va, cred);
1658 	if (error != 0)
1659 		return (error);
1660 
1661 	if (hardlink_check_uid && cred->cr_uid != va.va_uid) {
1662 		error = priv_check_cred(cred, PRIV_VFS_LINK);
1663 		if (error != 0)
1664 			return (error);
1665 	}
1666 
1667 	if (hardlink_check_gid && !groupmember(va.va_gid, cred)) {
1668 		error = priv_check_cred(cred, PRIV_VFS_LINK);
1669 		if (error != 0)
1670 			return (error);
1671 	}
1672 
1673 	return (0);
1674 }
1675 
1676 int
1677 kern_linkat(struct thread *td, int fd1, int fd2, const char *path1,
1678     const char *path2, enum uio_seg segflag, int flag)
1679 {
1680 	struct nameidata nd;
1681 	int error;
1682 
1683 	if ((flag & ~(AT_SYMLINK_FOLLOW | AT_RESOLVE_BENEATH |
1684 	    AT_EMPTY_PATH)) != 0)
1685 		return (EINVAL);
1686 
1687 	NDPREINIT(&nd);
1688 	do {
1689 		bwillwrite();
1690 		NDINIT_ATRIGHTS(&nd, LOOKUP, AUDITVNODE1 | at2cnpflags(flag,
1691 		    AT_SYMLINK_FOLLOW | AT_RESOLVE_BENEATH | AT_EMPTY_PATH),
1692 		    segflag, path1, fd1, &cap_linkat_source_rights);
1693 		if ((error = namei(&nd)) != 0)
1694 			return (error);
1695 		NDFREE_PNBUF(&nd);
1696 		if ((nd.ni_resflags & NIRES_EMPTYPATH) != 0) {
1697 			error = priv_check(td, PRIV_VFS_FHOPEN);
1698 			if (error != 0) {
1699 				vrele(nd.ni_vp);
1700 				return (error);
1701 			}
1702 		}
1703 		error = kern_linkat_vp(td, nd.ni_vp, fd2, path2, segflag);
1704 	} while (error ==  EAGAIN || error == ERELOOKUP);
1705 	return (error);
1706 }
1707 
1708 static int
1709 kern_linkat_vp(struct thread *td, struct vnode *vp, int fd, const char *path,
1710     enum uio_seg segflag)
1711 {
1712 	struct nameidata nd;
1713 	struct mount *mp;
1714 	int error;
1715 
1716 	if (vp->v_type == VDIR) {
1717 		vrele(vp);
1718 		return (EPERM);		/* POSIX */
1719 	}
1720 	if ((vn_irflag_read(vp) & (VIRF_NAMEDDIR | VIRF_NAMEDATTR)) != 0) {
1721 		vrele(vp);
1722 		return (EINVAL);
1723 	}
1724 	NDINIT_ATRIGHTS(&nd, CREATE,
1725 	    LOCKPARENT | AUDITVNODE2 | NOCACHE, segflag, path, fd,
1726 	    &cap_linkat_target_rights);
1727 	if ((error = namei(&nd)) == 0) {
1728 		if (nd.ni_vp != NULL) {
1729 			NDFREE_PNBUF(&nd);
1730 			if (nd.ni_dvp == nd.ni_vp)
1731 				vrele(nd.ni_dvp);
1732 			else
1733 				vput(nd.ni_dvp);
1734 			vrele(nd.ni_vp);
1735 			vrele(vp);
1736 			return (EEXIST);
1737 		} else if (nd.ni_dvp->v_mount != vp->v_mount) {
1738 			/*
1739 			 * Cross-device link.  No need to recheck
1740 			 * vp->v_type, since it cannot change, except
1741 			 * to VBAD.
1742 			 */
1743 			NDFREE_PNBUF(&nd);
1744 			vput(nd.ni_dvp);
1745 			vrele(vp);
1746 			return (EXDEV);
1747 		} else if (vn_lock(vp, LK_EXCLUSIVE) == 0) {
1748 			error = can_hardlink(vp, td->td_ucred);
1749 #ifdef MAC
1750 			if (error == 0)
1751 				error = mac_vnode_check_link(td->td_ucred,
1752 				    nd.ni_dvp, vp, &nd.ni_cnd);
1753 #endif
1754 			if (error != 0) {
1755 				vput(vp);
1756 				vput(nd.ni_dvp);
1757 				NDFREE_PNBUF(&nd);
1758 				return (error);
1759 			}
1760 			error = vn_start_write(vp, &mp, V_NOWAIT);
1761 			if (error != 0) {
1762 				vput(vp);
1763 				vput(nd.ni_dvp);
1764 				NDFREE_PNBUF(&nd);
1765 				error = vn_start_write(NULL, &mp,
1766 				    V_XSLEEP | V_PCATCH);
1767 				if (error != 0)
1768 					return (error);
1769 				return (EAGAIN);
1770 			}
1771 			error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
1772 			VOP_VPUT_PAIR(nd.ni_dvp, &vp, true);
1773 			vn_finished_write(mp);
1774 			NDFREE_PNBUF(&nd);
1775 			vp = NULL;
1776 		} else {
1777 			vput(nd.ni_dvp);
1778 			NDFREE_PNBUF(&nd);
1779 			vrele(vp);
1780 			return (EAGAIN);
1781 		}
1782 	}
1783 	if (vp != NULL)
1784 		vrele(vp);
1785 	return (error);
1786 }
1787 
1788 /*
1789  * Make a symbolic link.
1790  */
1791 #ifndef _SYS_SYSPROTO_H_
1792 struct symlink_args {
1793 	char	*path;
1794 	char	*link;
1795 };
1796 #endif
1797 int
1798 sys_symlink(struct thread *td, struct symlink_args *uap)
1799 {
1800 
1801 	return (kern_symlinkat(td, uap->path, AT_FDCWD, uap->link,
1802 	    UIO_USERSPACE));
1803 }
1804 
1805 #ifndef _SYS_SYSPROTO_H_
1806 struct symlinkat_args {
1807 	char	*path;
1808 	int	fd;
1809 	char	*path2;
1810 };
1811 #endif
1812 int
1813 sys_symlinkat(struct thread *td, struct symlinkat_args *uap)
1814 {
1815 
1816 	return (kern_symlinkat(td, uap->path1, uap->fd, uap->path2,
1817 	    UIO_USERSPACE));
1818 }
1819 
1820 int
1821 kern_symlinkat(struct thread *td, const char *path1, int fd, const char *path2,
1822     enum uio_seg segflg)
1823 {
1824 	struct mount *mp;
1825 	struct vattr vattr;
1826 	const char *syspath;
1827 	char *tmppath;
1828 	struct nameidata nd;
1829 	int error;
1830 
1831 	if (segflg == UIO_SYSSPACE) {
1832 		syspath = path1;
1833 	} else {
1834 		tmppath = uma_zalloc(namei_zone, M_WAITOK);
1835 		if ((error = copyinstr(path1, tmppath, MAXPATHLEN, NULL)) != 0)
1836 			goto out;
1837 		syspath = tmppath;
1838 	}
1839 	AUDIT_ARG_TEXT(syspath);
1840 	NDPREINIT(&nd);
1841 restart:
1842 	bwillwrite();
1843 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | AUDITVNODE1 | NOCACHE, segflg,
1844 	    path2, fd, &cap_symlinkat_rights);
1845 	if ((error = namei(&nd)) != 0)
1846 		goto out;
1847 	if (nd.ni_vp) {
1848 		NDFREE_PNBUF(&nd);
1849 		if (nd.ni_vp == nd.ni_dvp)
1850 			vrele(nd.ni_dvp);
1851 		else
1852 			vput(nd.ni_dvp);
1853 		vrele(nd.ni_vp);
1854 		nd.ni_vp = NULL;
1855 		error = EEXIST;
1856 		goto out;
1857 	}
1858 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1859 		NDFREE_PNBUF(&nd);
1860 		vput(nd.ni_dvp);
1861 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | V_PCATCH)) != 0)
1862 			goto out;
1863 		goto restart;
1864 	}
1865 	if ((vn_irflag_read(nd.ni_dvp) & VIRF_NAMEDDIR) != 0) {
1866 		error = EINVAL;
1867 		goto out;
1868 	}
1869 	VATTR_NULL(&vattr);
1870 	vattr.va_mode = ACCESSPERMS &~ td->td_proc->p_pd->pd_cmask;
1871 #ifdef MAC
1872 	vattr.va_type = VLNK;
1873 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1874 	    &vattr);
1875 	if (error != 0)
1876 		goto out2;
1877 #endif
1878 	error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, syspath);
1879 #ifdef MAC
1880 out2:
1881 #endif
1882 	VOP_VPUT_PAIR(nd.ni_dvp, error == 0 ? &nd.ni_vp : NULL, true);
1883 	vn_finished_write(mp);
1884 	NDFREE_PNBUF(&nd);
1885 	if (error == ERELOOKUP)
1886 		goto restart;
1887 out:
1888 	if (segflg != UIO_SYSSPACE)
1889 		uma_zfree(namei_zone, tmppath);
1890 	return (error);
1891 }
1892 
1893 /*
1894  * Delete a whiteout from the filesystem.
1895  */
1896 #ifndef _SYS_SYSPROTO_H_
1897 struct undelete_args {
1898 	char *path;
1899 };
1900 #endif
1901 int
1902 sys_undelete(struct thread *td, struct undelete_args *uap)
1903 {
1904 	struct mount *mp;
1905 	struct nameidata nd;
1906 	int error;
1907 
1908 	NDPREINIT(&nd);
1909 restart:
1910 	bwillwrite();
1911 	NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | AUDITVNODE1,
1912 	    UIO_USERSPACE, uap->path);
1913 	error = namei(&nd);
1914 	if (error != 0)
1915 		return (error);
1916 
1917 	if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) {
1918 		NDFREE_PNBUF(&nd);
1919 		if (nd.ni_vp == nd.ni_dvp)
1920 			vrele(nd.ni_dvp);
1921 		else
1922 			vput(nd.ni_dvp);
1923 		if (nd.ni_vp)
1924 			vrele(nd.ni_vp);
1925 		return (EEXIST);
1926 	}
1927 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1928 		NDFREE_PNBUF(&nd);
1929 		vput(nd.ni_dvp);
1930 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | V_PCATCH)) != 0)
1931 			return (error);
1932 		goto restart;
1933 	}
1934 	error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE);
1935 	NDFREE_PNBUF(&nd);
1936 	vput(nd.ni_dvp);
1937 	vn_finished_write(mp);
1938 	if (error == ERELOOKUP)
1939 		goto restart;
1940 	return (error);
1941 }
1942 
1943 /*
1944  * Delete a name from the filesystem.
1945  */
1946 #ifndef _SYS_SYSPROTO_H_
1947 struct unlink_args {
1948 	char	*path;
1949 };
1950 #endif
1951 int
1952 sys_unlink(struct thread *td, struct unlink_args *uap)
1953 {
1954 
1955 	return (kern_funlinkat(td, AT_FDCWD, uap->path, FD_NONE, UIO_USERSPACE,
1956 	    0, 0));
1957 }
1958 
1959 static int
1960 kern_funlinkat_ex(struct thread *td, int dfd, const char *path, int fd,
1961     int flag, enum uio_seg pathseg, ino_t oldinum)
1962 {
1963 
1964 	if ((flag & ~(AT_REMOVEDIR | AT_RESOLVE_BENEATH)) != 0)
1965 		return (EINVAL);
1966 
1967 	if ((flag & AT_REMOVEDIR) != 0)
1968 		return (kern_frmdirat(td, dfd, path, fd, UIO_USERSPACE, 0));
1969 
1970 	return (kern_funlinkat(td, dfd, path, fd, UIO_USERSPACE, 0, 0));
1971 }
1972 
1973 #ifndef _SYS_SYSPROTO_H_
1974 struct unlinkat_args {
1975 	int	fd;
1976 	char	*path;
1977 	int	flag;
1978 };
1979 #endif
1980 int
1981 sys_unlinkat(struct thread *td, struct unlinkat_args *uap)
1982 {
1983 
1984 	return (kern_funlinkat_ex(td, uap->fd, uap->path, FD_NONE, uap->flag,
1985 	    UIO_USERSPACE, 0));
1986 }
1987 
1988 #ifndef _SYS_SYSPROTO_H_
1989 struct funlinkat_args {
1990 	int		dfd;
1991 	const char	*path;
1992 	int		fd;
1993 	int		flag;
1994 };
1995 #endif
1996 int
1997 sys_funlinkat(struct thread *td, struct funlinkat_args *uap)
1998 {
1999 
2000 	return (kern_funlinkat_ex(td, uap->dfd, uap->path, uap->fd, uap->flag,
2001 	    UIO_USERSPACE, 0));
2002 }
2003 
2004 int
2005 kern_funlinkat(struct thread *td, int dfd, const char *path, int fd,
2006     enum uio_seg pathseg, int flag, ino_t oldinum)
2007 {
2008 	struct mount *mp;
2009 	struct file *fp;
2010 	struct vnode *vp;
2011 	struct nameidata nd;
2012 	struct stat sb;
2013 	int error;
2014 
2015 	fp = NULL;
2016 	if (fd != FD_NONE) {
2017 		error = getvnode_path(td, fd, &cap_no_rights, &fp);
2018 		if (error != 0)
2019 			return (error);
2020 	}
2021 
2022 	NDPREINIT(&nd);
2023 restart:
2024 	bwillwrite();
2025 	NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1 |
2026 	    at2cnpflags(flag, AT_RESOLVE_BENEATH),
2027 	    pathseg, path, dfd, &cap_unlinkat_rights);
2028 	if ((error = namei(&nd)) != 0) {
2029 		if (error == EINVAL)
2030 			error = EPERM;
2031 		goto fdout;
2032 	}
2033 	vp = nd.ni_vp;
2034 	if (vp->v_type == VDIR && oldinum == 0) {
2035 		error = EPERM;		/* POSIX */
2036 	} else if (oldinum != 0 &&
2037 	    ((error = VOP_STAT(vp, &sb, td->td_ucred, NOCRED)) == 0) &&
2038 	    sb.st_ino != oldinum) {
2039 		error = EIDRM;	/* Identifier removed */
2040 	} else if (fp != NULL && fp->f_vnode != vp) {
2041 		if (VN_IS_DOOMED(fp->f_vnode))
2042 			error = EBADF;
2043 		else
2044 			error = EDEADLK;
2045 	} else {
2046 		/*
2047 		 * The root of a mounted filesystem cannot be deleted.
2048 		 *
2049 		 * XXX: can this only be a VDIR case?
2050 		 */
2051 		if (vp->v_vflag & VV_ROOT)
2052 			error = EBUSY;
2053 	}
2054 	if (error == 0) {
2055 		if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
2056 			NDFREE_PNBUF(&nd);
2057 			vput(nd.ni_dvp);
2058 			if (vp == nd.ni_dvp)
2059 				vrele(vp);
2060 			else
2061 				vput(vp);
2062 			if ((error = vn_start_write(NULL, &mp,
2063 			    V_XSLEEP | V_PCATCH)) != 0) {
2064 				goto fdout;
2065 			}
2066 			goto restart;
2067 		}
2068 #ifdef MAC
2069 		error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
2070 		    &nd.ni_cnd);
2071 		if (error != 0)
2072 			goto out;
2073 #endif
2074 		error = VOP_REMOVE(nd.ni_dvp, vp, &nd.ni_cnd);
2075 #ifdef MAC
2076 out:
2077 #endif
2078 		vn_finished_write(mp);
2079 	}
2080 	NDFREE_PNBUF(&nd);
2081 	vput(nd.ni_dvp);
2082 	if (vp == nd.ni_dvp)
2083 		vrele(vp);
2084 	else
2085 		vput(vp);
2086 	if (error == ERELOOKUP)
2087 		goto restart;
2088 fdout:
2089 	if (fp != NULL)
2090 		fdrop(fp, td);
2091 	return (error);
2092 }
2093 
2094 /*
2095  * Reposition read/write file offset.
2096  */
2097 #ifndef _SYS_SYSPROTO_H_
2098 struct lseek_args {
2099 	int	fd;
2100 	int	pad;
2101 	off_t	offset;
2102 	int	whence;
2103 };
2104 #endif
2105 int
2106 sys_lseek(struct thread *td, struct lseek_args *uap)
2107 {
2108 
2109 	return (kern_lseek(td, uap->fd, uap->offset, uap->whence));
2110 }
2111 
2112 int
2113 kern_lseek(struct thread *td, int fd, off_t offset, int whence)
2114 {
2115 	struct file *fp;
2116 	int error;
2117 
2118 	AUDIT_ARG_FD(fd);
2119 	error = fget(td, fd, &cap_seek_rights, &fp);
2120 	if (error != 0)
2121 		return (error);
2122 	error = (fp->f_ops->fo_flags & DFLAG_SEEKABLE) != 0 ?
2123 	    fo_seek(fp, offset, whence, td) : ESPIPE;
2124 	fdrop(fp, td);
2125 	return (error);
2126 }
2127 
2128 #if defined(COMPAT_43)
2129 /*
2130  * Reposition read/write file offset.
2131  */
2132 #ifndef _SYS_SYSPROTO_H_
2133 struct olseek_args {
2134 	int	fd;
2135 	long	offset;
2136 	int	whence;
2137 };
2138 #endif
2139 int
2140 olseek(struct thread *td, struct olseek_args *uap)
2141 {
2142 
2143 	return (kern_lseek(td, uap->fd, uap->offset, uap->whence));
2144 }
2145 #endif /* COMPAT_43 */
2146 
2147 #if defined(COMPAT_FREEBSD6)
2148 /* Version with the 'pad' argument */
2149 int
2150 freebsd6_lseek(struct thread *td, struct freebsd6_lseek_args *uap)
2151 {
2152 
2153 	return (kern_lseek(td, uap->fd, uap->offset, uap->whence));
2154 }
2155 #endif
2156 
2157 /*
2158  * Check access permissions using passed credentials.
2159  */
2160 static int
2161 vn_access(struct vnode *vp, int user_flags, struct ucred *cred,
2162      struct thread *td)
2163 {
2164 	accmode_t accmode;
2165 	int error;
2166 
2167 	/* Flags == 0 means only check for existence. */
2168 	if (user_flags == 0)
2169 		return (0);
2170 
2171 	accmode = 0;
2172 	if (user_flags & R_OK)
2173 		accmode |= VREAD;
2174 	if (user_flags & W_OK)
2175 		accmode |= VWRITE;
2176 	if (user_flags & X_OK)
2177 		accmode |= VEXEC;
2178 #ifdef MAC
2179 	error = mac_vnode_check_access(cred, vp, accmode);
2180 	if (error != 0)
2181 		return (error);
2182 #endif
2183 	if ((accmode & VWRITE) == 0 || (error = vn_writechk(vp)) == 0)
2184 		error = VOP_ACCESS(vp, accmode, cred, td);
2185 	return (error);
2186 }
2187 
2188 /*
2189  * Check access permissions using "real" credentials.
2190  */
2191 #ifndef _SYS_SYSPROTO_H_
2192 struct access_args {
2193 	char	*path;
2194 	int	amode;
2195 };
2196 #endif
2197 int
2198 sys_access(struct thread *td, struct access_args *uap)
2199 {
2200 
2201 	return (kern_accessat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2202 	    0, uap->amode));
2203 }
2204 
2205 #ifndef _SYS_SYSPROTO_H_
2206 struct faccessat_args {
2207 	int	dirfd;
2208 	char	*path;
2209 	int	amode;
2210 	int	flag;
2211 }
2212 #endif
2213 int
2214 sys_faccessat(struct thread *td, struct faccessat_args *uap)
2215 {
2216 
2217 	return (kern_accessat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
2218 	    uap->amode));
2219 }
2220 
2221 int
2222 kern_accessat(struct thread *td, int fd, const char *path,
2223     enum uio_seg pathseg, int flag, int amode)
2224 {
2225 	struct ucred *cred, *usecred;
2226 	struct vnode *vp;
2227 	struct nameidata nd;
2228 	int error;
2229 
2230 	if ((flag & ~(AT_EACCESS | AT_RESOLVE_BENEATH | AT_EMPTY_PATH |
2231 	    AT_SYMLINK_NOFOLLOW)) != 0)
2232 		return (EINVAL);
2233 	if (amode != F_OK && (amode & ~(R_OK | W_OK | X_OK)) != 0)
2234 		return (EINVAL);
2235 
2236 	/*
2237 	 * Create and modify a temporary credential instead of one that
2238 	 * is potentially shared (if we need one).
2239 	 */
2240 	cred = td->td_ucred;
2241 	if ((flag & AT_EACCESS) == 0 &&
2242 	    ((cred->cr_uid != cred->cr_ruid ||
2243 	    cred->cr_rgid != cred->cr_groups[0]))) {
2244 		usecred = crdup(cred);
2245 		usecred->cr_uid = cred->cr_ruid;
2246 		usecred->cr_groups[0] = cred->cr_rgid;
2247 		td->td_ucred = usecred;
2248 	} else
2249 		usecred = cred;
2250 	AUDIT_ARG_VALUE(amode);
2251 	NDINIT_ATRIGHTS(&nd, LOOKUP, LOCKSHARED | LOCKLEAF |
2252 	    AUDITVNODE1 | at2cnpflags(flag, AT_RESOLVE_BENEATH | AT_SYMLINK_NOFOLLOW |
2253 	    AT_EMPTY_PATH), pathseg, path, fd, &cap_fstat_rights);
2254 	if ((error = namei(&nd)) != 0)
2255 		goto out;
2256 	vp = nd.ni_vp;
2257 
2258 	error = vn_access(vp, amode, usecred, td);
2259 	NDFREE_PNBUF(&nd);
2260 	vput(vp);
2261 out:
2262 	if (usecred != cred) {
2263 		td->td_ucred = cred;
2264 		crfree(usecred);
2265 	}
2266 	return (error);
2267 }
2268 
2269 /*
2270  * Check access permissions using "effective" credentials.
2271  */
2272 #ifndef _SYS_SYSPROTO_H_
2273 struct eaccess_args {
2274 	char	*path;
2275 	int	amode;
2276 };
2277 #endif
2278 int
2279 sys_eaccess(struct thread *td, struct eaccess_args *uap)
2280 {
2281 
2282 	return (kern_accessat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2283 	    AT_EACCESS, uap->amode));
2284 }
2285 
2286 #if defined(COMPAT_43)
2287 /*
2288  * Get file status; this version follows links.
2289  */
2290 #ifndef _SYS_SYSPROTO_H_
2291 struct ostat_args {
2292 	char	*path;
2293 	struct ostat *ub;
2294 };
2295 #endif
2296 int
2297 ostat(struct thread *td, struct ostat_args *uap)
2298 {
2299 	struct stat sb;
2300 	struct ostat osb;
2301 	int error;
2302 
2303 	error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE, &sb);
2304 	if (error != 0)
2305 		return (error);
2306 	cvtstat(&sb, &osb);
2307 	return (copyout(&osb, uap->ub, sizeof (osb)));
2308 }
2309 
2310 /*
2311  * Get file status; this version does not follow links.
2312  */
2313 #ifndef _SYS_SYSPROTO_H_
2314 struct olstat_args {
2315 	char	*path;
2316 	struct ostat *ub;
2317 };
2318 #endif
2319 int
2320 olstat(struct thread *td, struct olstat_args *uap)
2321 {
2322 	struct stat sb;
2323 	struct ostat osb;
2324 	int error;
2325 
2326 	error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2327 	    UIO_USERSPACE, &sb);
2328 	if (error != 0)
2329 		return (error);
2330 	cvtstat(&sb, &osb);
2331 	return (copyout(&osb, uap->ub, sizeof (osb)));
2332 }
2333 
2334 /*
2335  * Convert from an old to a new stat structure.
2336  * XXX: many values are blindly truncated.
2337  */
2338 void
2339 cvtstat(struct stat *st, struct ostat *ost)
2340 {
2341 
2342 	bzero(ost, sizeof(*ost));
2343 	ost->st_dev = st->st_dev;
2344 	ost->st_ino = st->st_ino;
2345 	ost->st_mode = st->st_mode;
2346 	ost->st_nlink = st->st_nlink;
2347 	ost->st_uid = st->st_uid;
2348 	ost->st_gid = st->st_gid;
2349 	ost->st_rdev = st->st_rdev;
2350 	ost->st_size = MIN(st->st_size, INT32_MAX);
2351 	ost->st_atim = st->st_atim;
2352 	ost->st_mtim = st->st_mtim;
2353 	ost->st_ctim = st->st_ctim;
2354 	ost->st_blksize = st->st_blksize;
2355 	ost->st_blocks = st->st_blocks;
2356 	ost->st_flags = st->st_flags;
2357 	ost->st_gen = st->st_gen;
2358 }
2359 #endif /* COMPAT_43 */
2360 
2361 #if defined(COMPAT_43) || defined(COMPAT_FREEBSD11)
2362 int ino64_trunc_error;
2363 SYSCTL_INT(_vfs, OID_AUTO, ino64_trunc_error, CTLFLAG_RW,
2364     &ino64_trunc_error, 0,
2365     "Error on truncation of device, file or inode number, or link count");
2366 
2367 int
2368 freebsd11_cvtstat(struct stat *st, struct freebsd11_stat *ost)
2369 {
2370 
2371 	ost->st_dev = st->st_dev;
2372 	if (ost->st_dev != st->st_dev) {
2373 		switch (ino64_trunc_error) {
2374 		default:
2375 			/*
2376 			 * Since dev_t is almost raw, don't clamp to the
2377 			 * maximum for case 2, but ignore the error.
2378 			 */
2379 			break;
2380 		case 1:
2381 			return (EOVERFLOW);
2382 		}
2383 	}
2384 	ost->st_ino = st->st_ino;
2385 	if (ost->st_ino != st->st_ino) {
2386 		switch (ino64_trunc_error) {
2387 		default:
2388 		case 0:
2389 			break;
2390 		case 1:
2391 			return (EOVERFLOW);
2392 		case 2:
2393 			ost->st_ino = UINT32_MAX;
2394 			break;
2395 		}
2396 	}
2397 	ost->st_mode = st->st_mode;
2398 	ost->st_nlink = st->st_nlink;
2399 	if (ost->st_nlink != st->st_nlink) {
2400 		switch (ino64_trunc_error) {
2401 		default:
2402 		case 0:
2403 			break;
2404 		case 1:
2405 			return (EOVERFLOW);
2406 		case 2:
2407 			ost->st_nlink = UINT16_MAX;
2408 			break;
2409 		}
2410 	}
2411 	ost->st_uid = st->st_uid;
2412 	ost->st_gid = st->st_gid;
2413 	ost->st_rdev = st->st_rdev;
2414 	if (ost->st_rdev != st->st_rdev) {
2415 		switch (ino64_trunc_error) {
2416 		default:
2417 			break;
2418 		case 1:
2419 			return (EOVERFLOW);
2420 		}
2421 	}
2422 	ost->st_atim = st->st_atim;
2423 	ost->st_mtim = st->st_mtim;
2424 	ost->st_ctim = st->st_ctim;
2425 	ost->st_size = st->st_size;
2426 	ost->st_blocks = st->st_blocks;
2427 	ost->st_blksize = st->st_blksize;
2428 	ost->st_flags = st->st_flags;
2429 	ost->st_gen = st->st_gen;
2430 	ost->st_lspare = 0;
2431 	ost->st_birthtim = st->st_birthtim;
2432 	bzero((char *)&ost->st_birthtim + sizeof(ost->st_birthtim),
2433 	    sizeof(*ost) - offsetof(struct freebsd11_stat,
2434 	    st_birthtim) - sizeof(ost->st_birthtim));
2435 	return (0);
2436 }
2437 
2438 int
2439 freebsd11_stat(struct thread *td, struct freebsd11_stat_args* uap)
2440 {
2441 	struct stat sb;
2442 	struct freebsd11_stat osb;
2443 	int error;
2444 
2445 	error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE, &sb);
2446 	if (error != 0)
2447 		return (error);
2448 	error = freebsd11_cvtstat(&sb, &osb);
2449 	if (error == 0)
2450 		error = copyout(&osb, uap->ub, sizeof(osb));
2451 	return (error);
2452 }
2453 
2454 int
2455 freebsd11_lstat(struct thread *td, struct freebsd11_lstat_args* uap)
2456 {
2457 	struct stat sb;
2458 	struct freebsd11_stat osb;
2459 	int error;
2460 
2461 	error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2462 	    UIO_USERSPACE, &sb);
2463 	if (error != 0)
2464 		return (error);
2465 	error = freebsd11_cvtstat(&sb, &osb);
2466 	if (error == 0)
2467 		error = copyout(&osb, uap->ub, sizeof(osb));
2468 	return (error);
2469 }
2470 
2471 int
2472 freebsd11_fhstat(struct thread *td, struct freebsd11_fhstat_args* uap)
2473 {
2474 	struct fhandle fh;
2475 	struct stat sb;
2476 	struct freebsd11_stat osb;
2477 	int error;
2478 
2479 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
2480 	if (error != 0)
2481 		return (error);
2482 	error = kern_fhstat(td, fh, &sb);
2483 	if (error != 0)
2484 		return (error);
2485 	error = freebsd11_cvtstat(&sb, &osb);
2486 	if (error == 0)
2487 		error = copyout(&osb, uap->sb, sizeof(osb));
2488 	return (error);
2489 }
2490 
2491 int
2492 freebsd11_fstatat(struct thread *td, struct freebsd11_fstatat_args* uap)
2493 {
2494 	struct stat sb;
2495 	struct freebsd11_stat osb;
2496 	int error;
2497 
2498 	error = kern_statat(td, uap->flag, uap->fd, uap->path,
2499 	    UIO_USERSPACE, &sb);
2500 	if (error != 0)
2501 		return (error);
2502 	error = freebsd11_cvtstat(&sb, &osb);
2503 	if (error == 0)
2504 		error = copyout(&osb, uap->buf, sizeof(osb));
2505 	return (error);
2506 }
2507 #endif	/* COMPAT_FREEBSD11 */
2508 
2509 /*
2510  * Get file status
2511  */
2512 #ifndef _SYS_SYSPROTO_H_
2513 struct fstatat_args {
2514 	int	fd;
2515 	char	*path;
2516 	struct stat	*buf;
2517 	int	flag;
2518 }
2519 #endif
2520 int
2521 sys_fstatat(struct thread *td, struct fstatat_args *uap)
2522 {
2523 	struct stat sb;
2524 	int error;
2525 
2526 	error = kern_statat(td, uap->flag, uap->fd, uap->path,
2527 	    UIO_USERSPACE, &sb);
2528 	if (error == 0)
2529 		error = copyout(&sb, uap->buf, sizeof (sb));
2530 	return (error);
2531 }
2532 
2533 int
2534 kern_statat(struct thread *td, int flag, int fd, const char *path,
2535     enum uio_seg pathseg, struct stat *sbp)
2536 {
2537 	struct nameidata nd;
2538 	int error;
2539 
2540 	if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH |
2541 	    AT_EMPTY_PATH)) != 0)
2542 		return (EINVAL);
2543 
2544 	NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(flag, AT_RESOLVE_BENEATH |
2545 	    AT_SYMLINK_NOFOLLOW | AT_EMPTY_PATH) | LOCKSHARED | LOCKLEAF |
2546 	    AUDITVNODE1, pathseg, path, fd, &cap_fstat_rights);
2547 
2548 	if ((error = namei(&nd)) != 0) {
2549 		if (error == ENOTDIR &&
2550 		    (nd.ni_resflags & NIRES_EMPTYPATH) != 0)
2551 			error = kern_fstat(td, fd, sbp);
2552 		return (error);
2553 	}
2554 	error = VOP_STAT(nd.ni_vp, sbp, td->td_ucred, NOCRED);
2555 	NDFREE_PNBUF(&nd);
2556 	vput(nd.ni_vp);
2557 #ifdef __STAT_TIME_T_EXT
2558 	sbp->st_atim_ext = 0;
2559 	sbp->st_mtim_ext = 0;
2560 	sbp->st_ctim_ext = 0;
2561 	sbp->st_btim_ext = 0;
2562 #endif
2563 #ifdef KTRACE
2564 	if (KTRPOINT(td, KTR_STRUCT))
2565 		ktrstat_error(sbp, error);
2566 #endif
2567 	return (error);
2568 }
2569 
2570 #if defined(COMPAT_FREEBSD11)
2571 /*
2572  * Implementation of the NetBSD [l]stat() functions.
2573  */
2574 int
2575 freebsd11_cvtnstat(struct stat *sb, struct nstat *nsb)
2576 {
2577 	struct freebsd11_stat sb11;
2578 	int error;
2579 
2580 	error = freebsd11_cvtstat(sb, &sb11);
2581 	if (error != 0)
2582 		return (error);
2583 
2584 	bzero(nsb, sizeof(*nsb));
2585 	CP(sb11, *nsb, st_dev);
2586 	CP(sb11, *nsb, st_ino);
2587 	CP(sb11, *nsb, st_mode);
2588 	CP(sb11, *nsb, st_nlink);
2589 	CP(sb11, *nsb, st_uid);
2590 	CP(sb11, *nsb, st_gid);
2591 	CP(sb11, *nsb, st_rdev);
2592 	CP(sb11, *nsb, st_atim);
2593 	CP(sb11, *nsb, st_mtim);
2594 	CP(sb11, *nsb, st_ctim);
2595 	CP(sb11, *nsb, st_size);
2596 	CP(sb11, *nsb, st_blocks);
2597 	CP(sb11, *nsb, st_blksize);
2598 	CP(sb11, *nsb, st_flags);
2599 	CP(sb11, *nsb, st_gen);
2600 	CP(sb11, *nsb, st_birthtim);
2601 	return (0);
2602 }
2603 
2604 #ifndef _SYS_SYSPROTO_H_
2605 struct freebsd11_nstat_args {
2606 	char	*path;
2607 	struct nstat *ub;
2608 };
2609 #endif
2610 int
2611 freebsd11_nstat(struct thread *td, struct freebsd11_nstat_args *uap)
2612 {
2613 	struct stat sb;
2614 	struct nstat nsb;
2615 	int error;
2616 
2617 	error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE, &sb);
2618 	if (error != 0)
2619 		return (error);
2620 	error = freebsd11_cvtnstat(&sb, &nsb);
2621 	if (error == 0)
2622 		error = copyout(&nsb, uap->ub, sizeof (nsb));
2623 	return (error);
2624 }
2625 
2626 /*
2627  * NetBSD lstat.  Get file status; this version does not follow links.
2628  */
2629 #ifndef _SYS_SYSPROTO_H_
2630 struct freebsd11_nlstat_args {
2631 	char	*path;
2632 	struct nstat *ub;
2633 };
2634 #endif
2635 int
2636 freebsd11_nlstat(struct thread *td, struct freebsd11_nlstat_args *uap)
2637 {
2638 	struct stat sb;
2639 	struct nstat nsb;
2640 	int error;
2641 
2642 	error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2643 	    UIO_USERSPACE, &sb);
2644 	if (error != 0)
2645 		return (error);
2646 	error = freebsd11_cvtnstat(&sb, &nsb);
2647 	if (error == 0)
2648 		error = copyout(&nsb, uap->ub, sizeof (nsb));
2649 	return (error);
2650 }
2651 #endif /* COMPAT_FREEBSD11 */
2652 
2653 /*
2654  * Get configurable pathname variables.
2655  */
2656 #ifndef _SYS_SYSPROTO_H_
2657 struct pathconf_args {
2658 	char	*path;
2659 	int	name;
2660 };
2661 #endif
2662 int
2663 sys_pathconf(struct thread *td, struct pathconf_args *uap)
2664 {
2665 	long value;
2666 	int error;
2667 
2668 	error = kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name, FOLLOW,
2669 	    &value);
2670 	if (error == 0)
2671 		td->td_retval[0] = value;
2672 	return (error);
2673 }
2674 
2675 #ifndef _SYS_SYSPROTO_H_
2676 struct lpathconf_args {
2677 	char	*path;
2678 	int	name;
2679 };
2680 #endif
2681 int
2682 sys_lpathconf(struct thread *td, struct lpathconf_args *uap)
2683 {
2684 	long value;
2685 	int error;
2686 
2687 	error = kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name,
2688 	    NOFOLLOW, &value);
2689 	if (error == 0)
2690 		td->td_retval[0] = value;
2691 	return (error);
2692 }
2693 
2694 int
2695 kern_pathconf(struct thread *td, const char *path, enum uio_seg pathseg,
2696     int name, u_long flags, long *valuep)
2697 {
2698 	struct nameidata nd;
2699 	int error;
2700 
2701 	NDINIT(&nd, LOOKUP, LOCKSHARED | LOCKLEAF | AUDITVNODE1 | flags,
2702 	       pathseg, path);
2703 	if ((error = namei(&nd)) != 0)
2704 		return (error);
2705 	NDFREE_PNBUF(&nd);
2706 
2707 	error = VOP_PATHCONF(nd.ni_vp, name, valuep);
2708 	vput(nd.ni_vp);
2709 	return (error);
2710 }
2711 
2712 /*
2713  * Return target name of a symbolic link.
2714  */
2715 #ifndef _SYS_SYSPROTO_H_
2716 struct readlink_args {
2717 	char	*path;
2718 	char	*buf;
2719 	size_t	count;
2720 };
2721 #endif
2722 int
2723 sys_readlink(struct thread *td, struct readlink_args *uap)
2724 {
2725 
2726 	return (kern_readlinkat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2727 	    uap->buf, UIO_USERSPACE, uap->count));
2728 }
2729 #ifndef _SYS_SYSPROTO_H_
2730 struct readlinkat_args {
2731 	int	fd;
2732 	char	*path;
2733 	char	*buf;
2734 	size_t	bufsize;
2735 };
2736 #endif
2737 int
2738 sys_readlinkat(struct thread *td, struct readlinkat_args *uap)
2739 {
2740 
2741 	return (kern_readlinkat(td, uap->fd, uap->path, UIO_USERSPACE,
2742 	    uap->buf, UIO_USERSPACE, uap->bufsize));
2743 }
2744 
2745 int
2746 kern_readlinkat(struct thread *td, int fd, const char *path,
2747     enum uio_seg pathseg, char *buf, enum uio_seg bufseg, size_t count)
2748 {
2749 	struct vnode *vp;
2750 	struct nameidata nd;
2751 	int error;
2752 
2753 	if (count > IOSIZE_MAX)
2754 		return (EINVAL);
2755 
2756 	NDINIT_AT(&nd, LOOKUP, NOFOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1 |
2757 	    EMPTYPATH, pathseg, path, fd);
2758 
2759 	if ((error = namei(&nd)) != 0)
2760 		return (error);
2761 	NDFREE_PNBUF(&nd);
2762 	vp = nd.ni_vp;
2763 
2764 	error = kern_readlink_vp(vp, buf, bufseg, count, td);
2765 	vput(vp);
2766 
2767 	return (error);
2768 }
2769 
2770 /*
2771  * Helper function to readlink from a vnode
2772  */
2773 static int
2774 kern_readlink_vp(struct vnode *vp, char *buf, enum uio_seg bufseg, size_t count,
2775     struct thread *td)
2776 {
2777 	struct iovec aiov;
2778 	struct uio auio;
2779 	int error;
2780 
2781 	ASSERT_VOP_LOCKED(vp, "kern_readlink_vp(): vp not locked");
2782 #ifdef MAC
2783 	error = mac_vnode_check_readlink(td->td_ucred, vp);
2784 	if (error != 0)
2785 		return (error);
2786 #endif
2787 	if (vp->v_type != VLNK && (vp->v_vflag & VV_READLINK) == 0)
2788 		return (EINVAL);
2789 
2790 	aiov.iov_base = buf;
2791 	aiov.iov_len = count;
2792 	auio.uio_iov = &aiov;
2793 	auio.uio_iovcnt = 1;
2794 	auio.uio_offset = 0;
2795 	auio.uio_rw = UIO_READ;
2796 	auio.uio_segflg = bufseg;
2797 	auio.uio_td = td;
2798 	auio.uio_resid = count;
2799 	error = VOP_READLINK(vp, &auio, td->td_ucred);
2800 	td->td_retval[0] = count - auio.uio_resid;
2801 	return (error);
2802 }
2803 
2804 /*
2805  * Common implementation code for chflags() and fchflags().
2806  */
2807 static int
2808 setfflags(struct thread *td, struct vnode *vp, u_long flags)
2809 {
2810 	struct mount *mp;
2811 	struct vattr vattr;
2812 	int error;
2813 
2814 	/* We can't support the value matching VNOVAL. */
2815 	if (flags == VNOVAL)
2816 		return (EOPNOTSUPP);
2817 
2818 	/*
2819 	 * Prevent non-root users from setting flags on devices.  When
2820 	 * a device is reused, users can retain ownership of the device
2821 	 * if they are allowed to set flags and programs assume that
2822 	 * chown can't fail when done as root.
2823 	 */
2824 	if (vp->v_type == VCHR || vp->v_type == VBLK) {
2825 		error = priv_check(td, PRIV_VFS_CHFLAGS_DEV);
2826 		if (error != 0)
2827 			return (error);
2828 	}
2829 
2830 	if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0)
2831 		return (error);
2832 	VATTR_NULL(&vattr);
2833 	vattr.va_flags = flags;
2834 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2835 #ifdef MAC
2836 	error = mac_vnode_check_setflags(td->td_ucred, vp, vattr.va_flags);
2837 	if (error == 0)
2838 #endif
2839 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
2840 	VOP_UNLOCK(vp);
2841 	vn_finished_write(mp);
2842 	return (error);
2843 }
2844 
2845 /*
2846  * Change flags of a file given a path name.
2847  */
2848 #ifndef _SYS_SYSPROTO_H_
2849 struct chflags_args {
2850 	const char *path;
2851 	u_long	flags;
2852 };
2853 #endif
2854 int
2855 sys_chflags(struct thread *td, struct chflags_args *uap)
2856 {
2857 
2858 	return (kern_chflagsat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2859 	    uap->flags, 0));
2860 }
2861 
2862 #ifndef _SYS_SYSPROTO_H_
2863 struct chflagsat_args {
2864 	int	fd;
2865 	const char *path;
2866 	u_long	flags;
2867 	int	atflag;
2868 }
2869 #endif
2870 int
2871 sys_chflagsat(struct thread *td, struct chflagsat_args *uap)
2872 {
2873 
2874 	return (kern_chflagsat(td, uap->fd, uap->path, UIO_USERSPACE,
2875 	    uap->flags, uap->atflag));
2876 }
2877 
2878 /*
2879  * Same as chflags() but doesn't follow symlinks.
2880  */
2881 #ifndef _SYS_SYSPROTO_H_
2882 struct lchflags_args {
2883 	const char *path;
2884 	u_long flags;
2885 };
2886 #endif
2887 int
2888 sys_lchflags(struct thread *td, struct lchflags_args *uap)
2889 {
2890 
2891 	return (kern_chflagsat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2892 	    uap->flags, AT_SYMLINK_NOFOLLOW));
2893 }
2894 
2895 static int
2896 kern_chflagsat(struct thread *td, int fd, const char *path,
2897     enum uio_seg pathseg, u_long flags, int atflag)
2898 {
2899 	struct nameidata nd;
2900 	int error;
2901 
2902 	if ((atflag & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH |
2903 	    AT_EMPTY_PATH)) != 0)
2904 		return (EINVAL);
2905 
2906 	AUDIT_ARG_FFLAGS(flags);
2907 	NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(atflag, AT_SYMLINK_NOFOLLOW |
2908 	    AT_RESOLVE_BENEATH | AT_EMPTY_PATH) | AUDITVNODE1, pathseg, path,
2909 	    fd, &cap_fchflags_rights);
2910 	if ((error = namei(&nd)) != 0)
2911 		return (error);
2912 	NDFREE_PNBUF(&nd);
2913 	error = setfflags(td, nd.ni_vp, flags);
2914 	vrele(nd.ni_vp);
2915 	return (error);
2916 }
2917 
2918 /*
2919  * Change flags of a file given a file descriptor.
2920  */
2921 #ifndef _SYS_SYSPROTO_H_
2922 struct fchflags_args {
2923 	int	fd;
2924 	u_long	flags;
2925 };
2926 #endif
2927 int
2928 sys_fchflags(struct thread *td, struct fchflags_args *uap)
2929 {
2930 	struct file *fp;
2931 	int error;
2932 
2933 	AUDIT_ARG_FD(uap->fd);
2934 	AUDIT_ARG_FFLAGS(uap->flags);
2935 	error = getvnode(td, uap->fd, &cap_fchflags_rights,
2936 	    &fp);
2937 	if (error != 0)
2938 		return (error);
2939 #ifdef AUDIT
2940 	if (AUDITING_TD(td)) {
2941 		vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
2942 		AUDIT_ARG_VNODE1(fp->f_vnode);
2943 		VOP_UNLOCK(fp->f_vnode);
2944 	}
2945 #endif
2946 	error = setfflags(td, fp->f_vnode, uap->flags);
2947 	fdrop(fp, td);
2948 	return (error);
2949 }
2950 
2951 /*
2952  * Common implementation code for chmod(), lchmod() and fchmod().
2953  */
2954 int
2955 setfmode(struct thread *td, struct ucred *cred, struct vnode *vp, int mode)
2956 {
2957 	struct mount *mp;
2958 	struct vattr vattr;
2959 	int error;
2960 
2961 	if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0)
2962 		return (error);
2963 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2964 	VATTR_NULL(&vattr);
2965 	vattr.va_mode = mode & ALLPERMS;
2966 #ifdef MAC
2967 	error = mac_vnode_check_setmode(cred, vp, vattr.va_mode);
2968 	if (error == 0)
2969 #endif
2970 		error = VOP_SETATTR(vp, &vattr, cred);
2971 	VOP_UNLOCK(vp);
2972 	vn_finished_write(mp);
2973 	return (error);
2974 }
2975 
2976 /*
2977  * Change mode of a file given path name.
2978  */
2979 #ifndef _SYS_SYSPROTO_H_
2980 struct chmod_args {
2981 	char	*path;
2982 	int	mode;
2983 };
2984 #endif
2985 int
2986 sys_chmod(struct thread *td, struct chmod_args *uap)
2987 {
2988 
2989 	return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2990 	    uap->mode, 0));
2991 }
2992 
2993 #ifndef _SYS_SYSPROTO_H_
2994 struct fchmodat_args {
2995 	int	dirfd;
2996 	char	*path;
2997 	mode_t	mode;
2998 	int	flag;
2999 }
3000 #endif
3001 int
3002 sys_fchmodat(struct thread *td, struct fchmodat_args *uap)
3003 {
3004 
3005 	return (kern_fchmodat(td, uap->fd, uap->path, UIO_USERSPACE,
3006 	    uap->mode, uap->flag));
3007 }
3008 
3009 /*
3010  * Change mode of a file given path name (don't follow links.)
3011  */
3012 #ifndef _SYS_SYSPROTO_H_
3013 struct lchmod_args {
3014 	char	*path;
3015 	int	mode;
3016 };
3017 #endif
3018 int
3019 sys_lchmod(struct thread *td, struct lchmod_args *uap)
3020 {
3021 
3022 	return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
3023 	    uap->mode, AT_SYMLINK_NOFOLLOW));
3024 }
3025 
3026 int
3027 kern_fchmodat(struct thread *td, int fd, const char *path,
3028     enum uio_seg pathseg, mode_t mode, int flag)
3029 {
3030 	struct nameidata nd;
3031 	int error;
3032 
3033 	if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH |
3034 	    AT_EMPTY_PATH)) != 0)
3035 		return (EINVAL);
3036 
3037 	AUDIT_ARG_MODE(mode);
3038 	NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(flag, AT_SYMLINK_NOFOLLOW |
3039 	    AT_RESOLVE_BENEATH | AT_EMPTY_PATH) | AUDITVNODE1, pathseg, path,
3040 	    fd, &cap_fchmod_rights);
3041 	if ((error = namei(&nd)) != 0)
3042 		return (error);
3043 	NDFREE_PNBUF(&nd);
3044 	error = setfmode(td, td->td_ucred, nd.ni_vp, mode);
3045 	vrele(nd.ni_vp);
3046 	return (error);
3047 }
3048 
3049 /*
3050  * Change mode of a file given a file descriptor.
3051  */
3052 #ifndef _SYS_SYSPROTO_H_
3053 struct fchmod_args {
3054 	int	fd;
3055 	int	mode;
3056 };
3057 #endif
3058 int
3059 sys_fchmod(struct thread *td, struct fchmod_args *uap)
3060 {
3061 	struct file *fp;
3062 	int error;
3063 
3064 	AUDIT_ARG_FD(uap->fd);
3065 	AUDIT_ARG_MODE(uap->mode);
3066 
3067 	error = fget(td, uap->fd, &cap_fchmod_rights, &fp);
3068 	if (error != 0)
3069 		return (error);
3070 	error = fo_chmod(fp, uap->mode, td->td_ucred, td);
3071 	fdrop(fp, td);
3072 	return (error);
3073 }
3074 
3075 /*
3076  * Common implementation for chown(), lchown(), and fchown()
3077  */
3078 int
3079 setfown(struct thread *td, struct ucred *cred, struct vnode *vp, uid_t uid,
3080     gid_t gid)
3081 {
3082 	struct mount *mp;
3083 	struct vattr vattr;
3084 	int error;
3085 
3086 	if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0)
3087 		return (error);
3088 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3089 	VATTR_NULL(&vattr);
3090 	vattr.va_uid = uid;
3091 	vattr.va_gid = gid;
3092 #ifdef MAC
3093 	error = mac_vnode_check_setowner(cred, vp, vattr.va_uid,
3094 	    vattr.va_gid);
3095 	if (error == 0)
3096 #endif
3097 		error = VOP_SETATTR(vp, &vattr, cred);
3098 	VOP_UNLOCK(vp);
3099 	vn_finished_write(mp);
3100 	return (error);
3101 }
3102 
3103 /*
3104  * Set ownership given a path name.
3105  */
3106 #ifndef _SYS_SYSPROTO_H_
3107 struct chown_args {
3108 	char	*path;
3109 	int	uid;
3110 	int	gid;
3111 };
3112 #endif
3113 int
3114 sys_chown(struct thread *td, struct chown_args *uap)
3115 {
3116 
3117 	return (kern_fchownat(td, AT_FDCWD, uap->path, UIO_USERSPACE, uap->uid,
3118 	    uap->gid, 0));
3119 }
3120 
3121 #ifndef _SYS_SYSPROTO_H_
3122 struct fchownat_args {
3123 	int fd;
3124 	const char * path;
3125 	uid_t uid;
3126 	gid_t gid;
3127 	int flag;
3128 };
3129 #endif
3130 int
3131 sys_fchownat(struct thread *td, struct fchownat_args *uap)
3132 {
3133 
3134 	return (kern_fchownat(td, uap->fd, uap->path, UIO_USERSPACE, uap->uid,
3135 	    uap->gid, uap->flag));
3136 }
3137 
3138 int
3139 kern_fchownat(struct thread *td, int fd, const char *path,
3140     enum uio_seg pathseg, int uid, int gid, int flag)
3141 {
3142 	struct nameidata nd;
3143 	int error;
3144 
3145 	if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH |
3146 	    AT_EMPTY_PATH)) != 0)
3147 		return (EINVAL);
3148 
3149 	AUDIT_ARG_OWNER(uid, gid);
3150 	NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(flag, AT_SYMLINK_NOFOLLOW |
3151 	    AT_RESOLVE_BENEATH | AT_EMPTY_PATH) | AUDITVNODE1, pathseg, path,
3152 	    fd, &cap_fchown_rights);
3153 
3154 	if ((error = namei(&nd)) != 0)
3155 		return (error);
3156 	NDFREE_PNBUF(&nd);
3157 	error = setfown(td, td->td_ucred, nd.ni_vp, uid, gid);
3158 	vrele(nd.ni_vp);
3159 	return (error);
3160 }
3161 
3162 /*
3163  * Set ownership given a path name, do not cross symlinks.
3164  */
3165 #ifndef _SYS_SYSPROTO_H_
3166 struct lchown_args {
3167 	char	*path;
3168 	int	uid;
3169 	int	gid;
3170 };
3171 #endif
3172 int
3173 sys_lchown(struct thread *td, struct lchown_args *uap)
3174 {
3175 
3176 	return (kern_fchownat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
3177 	    uap->uid, uap->gid, AT_SYMLINK_NOFOLLOW));
3178 }
3179 
3180 /*
3181  * Set ownership given a file descriptor.
3182  */
3183 #ifndef _SYS_SYSPROTO_H_
3184 struct fchown_args {
3185 	int	fd;
3186 	int	uid;
3187 	int	gid;
3188 };
3189 #endif
3190 int
3191 sys_fchown(struct thread *td, struct fchown_args *uap)
3192 {
3193 	struct file *fp;
3194 	int error;
3195 
3196 	AUDIT_ARG_FD(uap->fd);
3197 	AUDIT_ARG_OWNER(uap->uid, uap->gid);
3198 	error = fget(td, uap->fd, &cap_fchown_rights, &fp);
3199 	if (error != 0)
3200 		return (error);
3201 	error = fo_chown(fp, uap->uid, uap->gid, td->td_ucred, td);
3202 	fdrop(fp, td);
3203 	return (error);
3204 }
3205 
3206 /*
3207  * Common implementation code for utimes(), lutimes(), and futimes().
3208  */
3209 static int
3210 getutimes(const struct timeval *usrtvp, enum uio_seg tvpseg,
3211     struct timespec *tsp)
3212 {
3213 	struct timeval tv[2];
3214 	const struct timeval *tvp;
3215 	int error;
3216 
3217 	if (usrtvp == NULL) {
3218 		vfs_timestamp(&tsp[0]);
3219 		tsp[1] = tsp[0];
3220 	} else {
3221 		if (tvpseg == UIO_SYSSPACE) {
3222 			tvp = usrtvp;
3223 		} else {
3224 			if ((error = copyin(usrtvp, tv, sizeof(tv))) != 0)
3225 				return (error);
3226 			tvp = tv;
3227 		}
3228 
3229 		if (tvp[0].tv_usec < 0 || tvp[0].tv_usec >= 1000000 ||
3230 		    tvp[1].tv_usec < 0 || tvp[1].tv_usec >= 1000000)
3231 			return (EINVAL);
3232 		TIMEVAL_TO_TIMESPEC(&tvp[0], &tsp[0]);
3233 		TIMEVAL_TO_TIMESPEC(&tvp[1], &tsp[1]);
3234 	}
3235 	return (0);
3236 }
3237 
3238 /*
3239  * Common implementation code for futimens(), utimensat().
3240  */
3241 #define	UTIMENS_NULL	0x1
3242 #define	UTIMENS_EXIT	0x2
3243 static int
3244 getutimens(const struct timespec *usrtsp, enum uio_seg tspseg,
3245     struct timespec *tsp, int *retflags)
3246 {
3247 	struct timespec tsnow;
3248 	int error;
3249 
3250 	vfs_timestamp(&tsnow);
3251 	*retflags = 0;
3252 	if (usrtsp == NULL) {
3253 		tsp[0] = tsnow;
3254 		tsp[1] = tsnow;
3255 		*retflags |= UTIMENS_NULL;
3256 		return (0);
3257 	}
3258 	if (tspseg == UIO_SYSSPACE) {
3259 		tsp[0] = usrtsp[0];
3260 		tsp[1] = usrtsp[1];
3261 	} else if ((error = copyin(usrtsp, tsp, sizeof(*tsp) * 2)) != 0)
3262 		return (error);
3263 	if (tsp[0].tv_nsec == UTIME_OMIT && tsp[1].tv_nsec == UTIME_OMIT)
3264 		*retflags |= UTIMENS_EXIT;
3265 	if (tsp[0].tv_nsec == UTIME_NOW && tsp[1].tv_nsec == UTIME_NOW)
3266 		*retflags |= UTIMENS_NULL;
3267 	if (tsp[0].tv_nsec == UTIME_OMIT)
3268 		tsp[0].tv_sec = VNOVAL;
3269 	else if (tsp[0].tv_nsec == UTIME_NOW)
3270 		tsp[0] = tsnow;
3271 	else if (tsp[0].tv_nsec < 0 || tsp[0].tv_nsec >= 1000000000L)
3272 		return (EINVAL);
3273 	if (tsp[1].tv_nsec == UTIME_OMIT)
3274 		tsp[1].tv_sec = VNOVAL;
3275 	else if (tsp[1].tv_nsec == UTIME_NOW)
3276 		tsp[1] = tsnow;
3277 	else if (tsp[1].tv_nsec < 0 || tsp[1].tv_nsec >= 1000000000L)
3278 		return (EINVAL);
3279 
3280 	return (0);
3281 }
3282 
3283 /*
3284  * Common implementation code for utimes(), lutimes(), futimes(), futimens(),
3285  * and utimensat().
3286  */
3287 static int
3288 setutimes(struct thread *td, struct vnode *vp, const struct timespec *ts,
3289     int numtimes, int nullflag)
3290 {
3291 	struct mount *mp;
3292 	struct vattr vattr;
3293 	int error;
3294 	bool setbirthtime;
3295 
3296 	setbirthtime = false;
3297 	vattr.va_birthtime.tv_sec = VNOVAL;
3298 	vattr.va_birthtime.tv_nsec = 0;
3299 
3300 	if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0)
3301 		return (error);
3302 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3303 	if (numtimes < 3 && VOP_GETATTR(vp, &vattr, td->td_ucred) == 0 &&
3304 	    timespeccmp(&ts[1], &vattr.va_birthtime, < ))
3305 		setbirthtime = true;
3306 	VATTR_NULL(&vattr);
3307 	vattr.va_atime = ts[0];
3308 	vattr.va_mtime = ts[1];
3309 	if (setbirthtime)
3310 		vattr.va_birthtime = ts[1];
3311 	if (numtimes > 2)
3312 		vattr.va_birthtime = ts[2];
3313 	if (nullflag)
3314 		vattr.va_vaflags |= VA_UTIMES_NULL;
3315 #ifdef MAC
3316 	error = mac_vnode_check_setutimes(td->td_ucred, vp, vattr.va_atime,
3317 	    vattr.va_mtime);
3318 #endif
3319 	if (error == 0)
3320 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3321 	VOP_UNLOCK(vp);
3322 	vn_finished_write(mp);
3323 	return (error);
3324 }
3325 
3326 /*
3327  * Set the access and modification times of a file.
3328  */
3329 #ifndef _SYS_SYSPROTO_H_
3330 struct utimes_args {
3331 	char	*path;
3332 	struct	timeval *tptr;
3333 };
3334 #endif
3335 int
3336 sys_utimes(struct thread *td, struct utimes_args *uap)
3337 {
3338 
3339 	return (kern_utimesat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
3340 	    uap->tptr, UIO_USERSPACE));
3341 }
3342 
3343 #ifndef _SYS_SYSPROTO_H_
3344 struct futimesat_args {
3345 	int fd;
3346 	const char * path;
3347 	const struct timeval * times;
3348 };
3349 #endif
3350 int
3351 sys_futimesat(struct thread *td, struct futimesat_args *uap)
3352 {
3353 
3354 	return (kern_utimesat(td, uap->fd, uap->path, UIO_USERSPACE,
3355 	    uap->times, UIO_USERSPACE));
3356 }
3357 
3358 int
3359 kern_utimesat(struct thread *td, int fd, const char *path,
3360     enum uio_seg pathseg, const struct timeval *tptr, enum uio_seg tptrseg)
3361 {
3362 	struct nameidata nd;
3363 	struct timespec ts[2];
3364 	int error;
3365 
3366 	if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3367 		return (error);
3368 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, fd,
3369 	    &cap_futimes_rights);
3370 
3371 	if ((error = namei(&nd)) != 0)
3372 		return (error);
3373 	NDFREE_PNBUF(&nd);
3374 	error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3375 	vrele(nd.ni_vp);
3376 	return (error);
3377 }
3378 
3379 /*
3380  * Set the access and modification times of a file.
3381  */
3382 #ifndef _SYS_SYSPROTO_H_
3383 struct lutimes_args {
3384 	char	*path;
3385 	struct	timeval *tptr;
3386 };
3387 #endif
3388 int
3389 sys_lutimes(struct thread *td, struct lutimes_args *uap)
3390 {
3391 
3392 	return (kern_lutimes(td, uap->path, UIO_USERSPACE, uap->tptr,
3393 	    UIO_USERSPACE));
3394 }
3395 
3396 int
3397 kern_lutimes(struct thread *td, const char *path, enum uio_seg pathseg,
3398     const struct timeval *tptr, enum uio_seg tptrseg)
3399 {
3400 	struct timespec ts[2];
3401 	struct nameidata nd;
3402 	int error;
3403 
3404 	if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3405 		return (error);
3406 	NDINIT(&nd, LOOKUP, NOFOLLOW | AUDITVNODE1, pathseg, path);
3407 	if ((error = namei(&nd)) != 0)
3408 		return (error);
3409 	NDFREE_PNBUF(&nd);
3410 	error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3411 	vrele(nd.ni_vp);
3412 	return (error);
3413 }
3414 
3415 /*
3416  * Set the access and modification times of a file.
3417  */
3418 #ifndef _SYS_SYSPROTO_H_
3419 struct futimes_args {
3420 	int	fd;
3421 	struct	timeval *tptr;
3422 };
3423 #endif
3424 int
3425 sys_futimes(struct thread *td, struct futimes_args *uap)
3426 {
3427 
3428 	return (kern_futimes(td, uap->fd, uap->tptr, UIO_USERSPACE));
3429 }
3430 
3431 int
3432 kern_futimes(struct thread *td, int fd, const struct timeval *tptr,
3433     enum uio_seg tptrseg)
3434 {
3435 	struct timespec ts[2];
3436 	struct file *fp;
3437 	int error;
3438 
3439 	AUDIT_ARG_FD(fd);
3440 	error = getutimes(tptr, tptrseg, ts);
3441 	if (error != 0)
3442 		return (error);
3443 	error = getvnode(td, fd, &cap_futimes_rights, &fp);
3444 	if (error != 0)
3445 		return (error);
3446 #ifdef AUDIT
3447 	if (AUDITING_TD(td)) {
3448 		vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
3449 		AUDIT_ARG_VNODE1(fp->f_vnode);
3450 		VOP_UNLOCK(fp->f_vnode);
3451 	}
3452 #endif
3453 	error = setutimes(td, fp->f_vnode, ts, 2, tptr == NULL);
3454 	fdrop(fp, td);
3455 	return (error);
3456 }
3457 
3458 int
3459 sys_futimens(struct thread *td, struct futimens_args *uap)
3460 {
3461 
3462 	return (kern_futimens(td, uap->fd, uap->times, UIO_USERSPACE));
3463 }
3464 
3465 int
3466 kern_futimens(struct thread *td, int fd, const struct timespec *tptr,
3467     enum uio_seg tptrseg)
3468 {
3469 	struct timespec ts[2];
3470 	struct file *fp;
3471 	int error, flags;
3472 
3473 	AUDIT_ARG_FD(fd);
3474 	error = getutimens(tptr, tptrseg, ts, &flags);
3475 	if (error != 0)
3476 		return (error);
3477 	if (flags & UTIMENS_EXIT)
3478 		return (0);
3479 	error = getvnode(td, fd, &cap_futimes_rights, &fp);
3480 	if (error != 0)
3481 		return (error);
3482 #ifdef AUDIT
3483 	if (AUDITING_TD(td)) {
3484 		vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
3485 		AUDIT_ARG_VNODE1(fp->f_vnode);
3486 		VOP_UNLOCK(fp->f_vnode);
3487 	}
3488 #endif
3489 	error = setutimes(td, fp->f_vnode, ts, 2, flags & UTIMENS_NULL);
3490 	fdrop(fp, td);
3491 	return (error);
3492 }
3493 
3494 int
3495 sys_utimensat(struct thread *td, struct utimensat_args *uap)
3496 {
3497 
3498 	return (kern_utimensat(td, uap->fd, uap->path, UIO_USERSPACE,
3499 	    uap->times, UIO_USERSPACE, uap->flag));
3500 }
3501 
3502 int
3503 kern_utimensat(struct thread *td, int fd, const char *path,
3504     enum uio_seg pathseg, const struct timespec *tptr, enum uio_seg tptrseg,
3505     int flag)
3506 {
3507 	struct nameidata nd;
3508 	struct timespec ts[2];
3509 	int error, flags;
3510 
3511 	if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH |
3512 	    AT_EMPTY_PATH)) != 0)
3513 		return (EINVAL);
3514 
3515 	if ((error = getutimens(tptr, tptrseg, ts, &flags)) != 0)
3516 		return (error);
3517 	NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(flag, AT_SYMLINK_NOFOLLOW |
3518 	    AT_RESOLVE_BENEATH | AT_EMPTY_PATH) | AUDITVNODE1,
3519 	    pathseg, path, fd, &cap_futimes_rights);
3520 	if ((error = namei(&nd)) != 0)
3521 		return (error);
3522 	/*
3523 	 * We are allowed to call namei() regardless of 2xUTIME_OMIT.
3524 	 * POSIX states:
3525 	 * "If both tv_nsec fields are UTIME_OMIT... EACCESS may be detected."
3526 	 * "Search permission is denied by a component of the path prefix."
3527 	 */
3528 	NDFREE_PNBUF(&nd);
3529 	if ((flags & UTIMENS_EXIT) == 0)
3530 		error = setutimes(td, nd.ni_vp, ts, 2, flags & UTIMENS_NULL);
3531 	vrele(nd.ni_vp);
3532 	return (error);
3533 }
3534 
3535 /*
3536  * Truncate a file given its path name.
3537  */
3538 #ifndef _SYS_SYSPROTO_H_
3539 struct truncate_args {
3540 	char	*path;
3541 	int	pad;
3542 	off_t	length;
3543 };
3544 #endif
3545 int
3546 sys_truncate(struct thread *td, struct truncate_args *uap)
3547 {
3548 
3549 	return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3550 }
3551 
3552 int
3553 kern_truncate(struct thread *td, const char *path, enum uio_seg pathseg,
3554     off_t length)
3555 {
3556 	struct mount *mp;
3557 	struct vnode *vp;
3558 	void *rl_cookie;
3559 	struct nameidata nd;
3560 	int error;
3561 
3562 	if (length < 0)
3563 		return (EINVAL);
3564 	NDPREINIT(&nd);
3565 retry:
3566 	NDINIT(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path);
3567 	if ((error = namei(&nd)) != 0)
3568 		return (error);
3569 	vp = nd.ni_vp;
3570 	NDFREE_PNBUF(&nd);
3571 	rl_cookie = vn_rangelock_wlock(vp, 0, OFF_MAX);
3572 	if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) {
3573 		vn_rangelock_unlock(vp, rl_cookie);
3574 		vrele(vp);
3575 		return (error);
3576 	}
3577 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3578 	if (vp->v_type == VDIR) {
3579 		error = EISDIR;
3580 		goto out;
3581 	}
3582 #ifdef MAC
3583 	error = mac_vnode_check_write(td->td_ucred, NOCRED, vp);
3584 	if (error != 0)
3585 		goto out;
3586 #endif
3587 	error = VOP_ACCESS(vp, VWRITE, td->td_ucred, td);
3588 	if (error != 0)
3589 		goto out;
3590 
3591 	error = vn_truncate_locked(vp, length, false, td->td_ucred);
3592 out:
3593 	VOP_UNLOCK(vp);
3594 	vn_finished_write(mp);
3595 	vn_rangelock_unlock(vp, rl_cookie);
3596 	vrele(vp);
3597 	if (error == ERELOOKUP)
3598 		goto retry;
3599 	return (error);
3600 }
3601 
3602 #if defined(COMPAT_43)
3603 /*
3604  * Truncate a file given its path name.
3605  */
3606 #ifndef _SYS_SYSPROTO_H_
3607 struct otruncate_args {
3608 	char	*path;
3609 	long	length;
3610 };
3611 #endif
3612 int
3613 otruncate(struct thread *td, struct otruncate_args *uap)
3614 {
3615 
3616 	return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3617 }
3618 #endif /* COMPAT_43 */
3619 
3620 #if defined(COMPAT_FREEBSD6)
3621 /* Versions with the pad argument */
3622 int
3623 freebsd6_truncate(struct thread *td, struct freebsd6_truncate_args *uap)
3624 {
3625 
3626 	return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3627 }
3628 
3629 int
3630 freebsd6_ftruncate(struct thread *td, struct freebsd6_ftruncate_args *uap)
3631 {
3632 
3633 	return (kern_ftruncate(td, uap->fd, uap->length));
3634 }
3635 #endif
3636 
3637 int
3638 kern_fsync(struct thread *td, int fd, bool fullsync)
3639 {
3640 	struct vnode *vp;
3641 	struct mount *mp;
3642 	struct file *fp;
3643 	int error;
3644 
3645 	AUDIT_ARG_FD(fd);
3646 	error = getvnode(td, fd, &cap_fsync_rights, &fp);
3647 	if (error != 0)
3648 		return (error);
3649 	vp = fp->f_vnode;
3650 #if 0
3651 	if (!fullsync)
3652 		/* XXXKIB: compete outstanding aio writes */;
3653 #endif
3654 retry:
3655 	error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH);
3656 	if (error != 0)
3657 		goto drop;
3658 	vn_lock(vp, vn_lktype_write(mp, vp) | LK_RETRY);
3659 	AUDIT_ARG_VNODE1(vp);
3660 	vnode_pager_clean_async(vp);
3661 	error = fullsync ? VOP_FSYNC(vp, MNT_WAIT, td) : VOP_FDATASYNC(vp, td);
3662 	VOP_UNLOCK(vp);
3663 	vn_finished_write(mp);
3664 	if (error == ERELOOKUP)
3665 		goto retry;
3666 drop:
3667 	fdrop(fp, td);
3668 	return (error);
3669 }
3670 
3671 /*
3672  * Sync an open file.
3673  */
3674 #ifndef _SYS_SYSPROTO_H_
3675 struct fsync_args {
3676 	int	fd;
3677 };
3678 #endif
3679 int
3680 sys_fsync(struct thread *td, struct fsync_args *uap)
3681 {
3682 
3683 	return (kern_fsync(td, uap->fd, true));
3684 }
3685 
3686 int
3687 sys_fdatasync(struct thread *td, struct fdatasync_args *uap)
3688 {
3689 
3690 	return (kern_fsync(td, uap->fd, false));
3691 }
3692 
3693 /*
3694  * Rename files.  Source and destination must either both be directories, or
3695  * both not be directories.  If target is a directory, it must be empty.
3696  */
3697 #ifndef _SYS_SYSPROTO_H_
3698 struct rename_args {
3699 	char	*from;
3700 	char	*to;
3701 };
3702 #endif
3703 int
3704 sys_rename(struct thread *td, struct rename_args *uap)
3705 {
3706 
3707 	return (kern_renameat(td, AT_FDCWD, uap->from, AT_FDCWD,
3708 	    uap->to, UIO_USERSPACE));
3709 }
3710 
3711 #ifndef _SYS_SYSPROTO_H_
3712 struct renameat_args {
3713 	int	oldfd;
3714 	char	*old;
3715 	int	newfd;
3716 	char	*new;
3717 };
3718 #endif
3719 int
3720 sys_renameat(struct thread *td, struct renameat_args *uap)
3721 {
3722 
3723 	return (kern_renameat(td, uap->oldfd, uap->old, uap->newfd, uap->new,
3724 	    UIO_USERSPACE));
3725 }
3726 
3727 #ifdef MAC
3728 static int
3729 kern_renameat_mac(struct thread *td, int oldfd, const char *old, int newfd,
3730     const char *new, enum uio_seg pathseg, struct nameidata *fromnd)
3731 {
3732 	int error;
3733 
3734 	NDINIT_ATRIGHTS(fromnd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1,
3735 	    pathseg, old, oldfd, &cap_renameat_source_rights);
3736 	if ((error = namei(fromnd)) != 0)
3737 		return (error);
3738 	error = mac_vnode_check_rename_from(td->td_ucred, fromnd->ni_dvp,
3739 	    fromnd->ni_vp, &fromnd->ni_cnd);
3740 	VOP_UNLOCK(fromnd->ni_dvp);
3741 	if (fromnd->ni_dvp != fromnd->ni_vp)
3742 		VOP_UNLOCK(fromnd->ni_vp);
3743 	if (error != 0) {
3744 		NDFREE_PNBUF(fromnd);
3745 		vrele(fromnd->ni_dvp);
3746 		vrele(fromnd->ni_vp);
3747 	}
3748 	return (error);
3749 }
3750 #endif
3751 
3752 int
3753 kern_renameat(struct thread *td, int oldfd, const char *old, int newfd,
3754     const char *new, enum uio_seg pathseg)
3755 {
3756 	struct mount *mp = NULL;
3757 	struct vnode *tvp, *fvp, *tdvp;
3758 	struct nameidata fromnd, tond;
3759 	uint64_t tondflags;
3760 	int error;
3761 	short irflag;
3762 
3763 again:
3764 	bwillwrite();
3765 #ifdef MAC
3766 	if (mac_vnode_check_rename_from_enabled()) {
3767 		error = kern_renameat_mac(td, oldfd, old, newfd, new, pathseg,
3768 		    &fromnd);
3769 		if (error != 0)
3770 			return (error);
3771 	} else {
3772 #endif
3773 	NDINIT_ATRIGHTS(&fromnd, DELETE, WANTPARENT | AUDITVNODE1,
3774 	    pathseg, old, oldfd, &cap_renameat_source_rights);
3775 	if ((error = namei(&fromnd)) != 0)
3776 		return (error);
3777 #ifdef MAC
3778 	}
3779 #endif
3780 	fvp = fromnd.ni_vp;
3781 	tondflags = LOCKPARENT | LOCKLEAF | NOCACHE | AUDITVNODE2;
3782 	if (fromnd.ni_vp->v_type == VDIR)
3783 		tondflags |= WILLBEDIR;
3784 	NDINIT_ATRIGHTS(&tond, RENAME, tondflags, pathseg, new, newfd,
3785 	    &cap_renameat_target_rights);
3786 	if ((error = namei(&tond)) != 0) {
3787 		/* Translate error code for rename("dir1", "dir2/."). */
3788 		if (error == EISDIR && fvp->v_type == VDIR)
3789 			error = EINVAL;
3790 		NDFREE_PNBUF(&fromnd);
3791 		vrele(fromnd.ni_dvp);
3792 		vrele(fvp);
3793 		goto out1;
3794 	}
3795 	tdvp = tond.ni_dvp;
3796 	tvp = tond.ni_vp;
3797 	error = vn_start_write(fvp, &mp, V_NOWAIT);
3798 	if (error != 0) {
3799 		NDFREE_PNBUF(&fromnd);
3800 		NDFREE_PNBUF(&tond);
3801 		if (tvp != NULL)
3802 			vput(tvp);
3803 		if (tdvp == tvp)
3804 			vrele(tdvp);
3805 		else
3806 			vput(tdvp);
3807 		vrele(fromnd.ni_dvp);
3808 		vrele(fvp);
3809 		error = vn_start_write(NULL, &mp, V_XSLEEP | V_PCATCH);
3810 		if (error != 0)
3811 			return (error);
3812 		goto again;
3813 	}
3814 	irflag = vn_irflag_read(fvp);
3815 	if (((irflag & VIRF_NAMEDATTR) != 0 && tdvp != fromnd.ni_dvp) ||
3816 	    (irflag & VIRF_NAMEDDIR) != 0) {
3817 		error = EINVAL;
3818 		goto out;
3819 	}
3820 	if (tvp != NULL) {
3821 		if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
3822 			error = ENOTDIR;
3823 			goto out;
3824 		} else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
3825 			error = EISDIR;
3826 			goto out;
3827 		}
3828 #ifdef CAPABILITIES
3829 		if (newfd != AT_FDCWD && (tond.ni_resflags & NIRES_ABS) == 0) {
3830 			/*
3831 			 * If the target already exists we require CAP_UNLINKAT
3832 			 * from 'newfd', when newfd was used for the lookup.
3833 			 */
3834 			error = cap_check(&tond.ni_filecaps.fc_rights,
3835 			    &cap_unlinkat_rights);
3836 			if (error != 0)
3837 				goto out;
3838 		}
3839 #endif
3840 	}
3841 	if (fvp == tdvp) {
3842 		error = EINVAL;
3843 		goto out;
3844 	}
3845 	/*
3846 	 * If the source is the same as the destination (that is, if they
3847 	 * are links to the same vnode), then there is nothing to do.
3848 	 */
3849 	if (fvp == tvp)
3850 		error = ERESTART;
3851 #ifdef MAC
3852 	else
3853 		error = mac_vnode_check_rename_to(td->td_ucred, tdvp,
3854 		    tond.ni_vp, fromnd.ni_dvp == tdvp, &tond.ni_cnd);
3855 #endif
3856 out:
3857 	if (error == 0) {
3858 		error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
3859 		    tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
3860 		NDFREE_PNBUF(&fromnd);
3861 		NDFREE_PNBUF(&tond);
3862 	} else {
3863 		NDFREE_PNBUF(&fromnd);
3864 		NDFREE_PNBUF(&tond);
3865 		if (tvp != NULL)
3866 			vput(tvp);
3867 		if (tdvp == tvp)
3868 			vrele(tdvp);
3869 		else
3870 			vput(tdvp);
3871 		vrele(fromnd.ni_dvp);
3872 		vrele(fvp);
3873 	}
3874 	vn_finished_write(mp);
3875 out1:
3876 	if (error == ERESTART)
3877 		return (0);
3878 	if (error == ERELOOKUP)
3879 		goto again;
3880 	return (error);
3881 }
3882 
3883 /*
3884  * Make a directory file.
3885  */
3886 #ifndef _SYS_SYSPROTO_H_
3887 struct mkdir_args {
3888 	char	*path;
3889 	int	mode;
3890 };
3891 #endif
3892 int
3893 sys_mkdir(struct thread *td, struct mkdir_args *uap)
3894 {
3895 
3896 	return (kern_mkdirat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
3897 	    uap->mode));
3898 }
3899 
3900 #ifndef _SYS_SYSPROTO_H_
3901 struct mkdirat_args {
3902 	int	fd;
3903 	char	*path;
3904 	mode_t	mode;
3905 };
3906 #endif
3907 int
3908 sys_mkdirat(struct thread *td, struct mkdirat_args *uap)
3909 {
3910 
3911 	return (kern_mkdirat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode));
3912 }
3913 
3914 int
3915 kern_mkdirat(struct thread *td, int fd, const char *path, enum uio_seg segflg,
3916     int mode)
3917 {
3918 	struct mount *mp;
3919 	struct vattr vattr;
3920 	struct nameidata nd;
3921 	int error;
3922 
3923 	AUDIT_ARG_MODE(mode);
3924 	NDPREINIT(&nd);
3925 restart:
3926 	bwillwrite();
3927 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | AUDITVNODE1 |
3928 	    NC_NOMAKEENTRY | NC_KEEPPOSENTRY | FAILIFEXISTS | WILLBEDIR,
3929 	    segflg, path, fd, &cap_mkdirat_rights);
3930 	if ((error = namei(&nd)) != 0)
3931 		return (error);
3932 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3933 		NDFREE_PNBUF(&nd);
3934 		vput(nd.ni_dvp);
3935 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | V_PCATCH)) != 0)
3936 			return (error);
3937 		goto restart;
3938 	}
3939 	if ((vn_irflag_read(nd.ni_dvp) & VIRF_NAMEDDIR) != 0) {
3940 		error = EINVAL;
3941 		goto out;
3942 	}
3943 	VATTR_NULL(&vattr);
3944 	vattr.va_type = VDIR;
3945 	vattr.va_mode = (mode & ACCESSPERMS) &~ td->td_proc->p_pd->pd_cmask;
3946 #ifdef MAC
3947 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
3948 	    &vattr);
3949 	if (error != 0)
3950 		goto out;
3951 #endif
3952 	error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
3953 out:
3954 	NDFREE_PNBUF(&nd);
3955 	VOP_VPUT_PAIR(nd.ni_dvp, error == 0 ? &nd.ni_vp : NULL, true);
3956 	vn_finished_write(mp);
3957 	if (error == ERELOOKUP)
3958 		goto restart;
3959 	return (error);
3960 }
3961 
3962 /*
3963  * Remove a directory file.
3964  */
3965 #ifndef _SYS_SYSPROTO_H_
3966 struct rmdir_args {
3967 	char	*path;
3968 };
3969 #endif
3970 int
3971 sys_rmdir(struct thread *td, struct rmdir_args *uap)
3972 {
3973 
3974 	return (kern_frmdirat(td, AT_FDCWD, uap->path, FD_NONE, UIO_USERSPACE,
3975 	    0));
3976 }
3977 
3978 int
3979 kern_frmdirat(struct thread *td, int dfd, const char *path, int fd,
3980     enum uio_seg pathseg, int flag)
3981 {
3982 	struct mount *mp;
3983 	struct vnode *vp;
3984 	struct file *fp;
3985 	struct nameidata nd;
3986 	cap_rights_t rights;
3987 	int error;
3988 
3989 	fp = NULL;
3990 	if (fd != FD_NONE) {
3991 		error = getvnode(td, fd, cap_rights_init_one(&rights,
3992 		    CAP_LOOKUP), &fp);
3993 		if (error != 0)
3994 			return (error);
3995 	}
3996 
3997 	NDPREINIT(&nd);
3998 restart:
3999 	bwillwrite();
4000 	NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1 |
4001 	    at2cnpflags(flag, AT_RESOLVE_BENEATH),
4002 	    pathseg, path, dfd, &cap_unlinkat_rights);
4003 	if ((error = namei(&nd)) != 0)
4004 		goto fdout;
4005 	vp = nd.ni_vp;
4006 	if (vp->v_type != VDIR) {
4007 		error = ENOTDIR;
4008 		goto out;
4009 	}
4010 	/*
4011 	 * No rmdir "." please.
4012 	 */
4013 	if (nd.ni_dvp == vp) {
4014 		error = EINVAL;
4015 		goto out;
4016 	}
4017 	/*
4018 	 * The root of a mounted filesystem cannot be deleted.
4019 	 */
4020 	if (vp->v_vflag & VV_ROOT) {
4021 		error = EBUSY;
4022 		goto out;
4023 	}
4024 
4025 	if (fp != NULL && fp->f_vnode != vp) {
4026 		if (VN_IS_DOOMED(fp->f_vnode))
4027 			error = EBADF;
4028 		else
4029 			error = EDEADLK;
4030 		goto out;
4031 	}
4032 
4033 #ifdef MAC
4034 	error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
4035 	    &nd.ni_cnd);
4036 	if (error != 0)
4037 		goto out;
4038 #endif
4039 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
4040 		NDFREE_PNBUF(&nd);
4041 		vput(vp);
4042 		if (nd.ni_dvp == vp)
4043 			vrele(nd.ni_dvp);
4044 		else
4045 			vput(nd.ni_dvp);
4046 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | V_PCATCH)) != 0)
4047 			goto fdout;
4048 		goto restart;
4049 	}
4050 	error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
4051 	vn_finished_write(mp);
4052 out:
4053 	NDFREE_PNBUF(&nd);
4054 	vput(vp);
4055 	if (nd.ni_dvp == vp)
4056 		vrele(nd.ni_dvp);
4057 	else
4058 		vput(nd.ni_dvp);
4059 	if (error == ERELOOKUP)
4060 		goto restart;
4061 fdout:
4062 	if (fp != NULL)
4063 		fdrop(fp, td);
4064 	return (error);
4065 }
4066 
4067 #if defined(COMPAT_43) || defined(COMPAT_FREEBSD11)
4068 int
4069 freebsd11_kern_getdirentries(struct thread *td, int fd, char *ubuf, u_int count,
4070     long *basep, void (*func)(struct freebsd11_dirent *))
4071 {
4072 	struct freebsd11_dirent dstdp;
4073 	struct dirent *dp, *edp;
4074 	char *dirbuf;
4075 	off_t base;
4076 	ssize_t resid, ucount;
4077 	int error;
4078 
4079 	/* XXX arbitrary sanity limit on `count'. */
4080 	count = min(count, 64 * 1024);
4081 
4082 	dirbuf = malloc(count, M_TEMP, M_WAITOK);
4083 
4084 	error = kern_getdirentries(td, fd, dirbuf, count, &base, &resid,
4085 	    UIO_SYSSPACE);
4086 	if (error != 0)
4087 		goto done;
4088 	if (basep != NULL)
4089 		*basep = base;
4090 
4091 	ucount = 0;
4092 	for (dp = (struct dirent *)dirbuf,
4093 	    edp = (struct dirent *)&dirbuf[count - resid];
4094 	    ucount < count && dp < edp; ) {
4095 		if (dp->d_reclen == 0)
4096 			break;
4097 		MPASS(dp->d_reclen >= _GENERIC_DIRLEN(0));
4098 		if (dp->d_namlen >= sizeof(dstdp.d_name))
4099 			continue;
4100 		dstdp.d_type = dp->d_type;
4101 		dstdp.d_namlen = dp->d_namlen;
4102 		dstdp.d_fileno = dp->d_fileno;		/* truncate */
4103 		if (dstdp.d_fileno != dp->d_fileno) {
4104 			switch (ino64_trunc_error) {
4105 			default:
4106 			case 0:
4107 				break;
4108 			case 1:
4109 				error = EOVERFLOW;
4110 				goto done;
4111 			case 2:
4112 				dstdp.d_fileno = UINT32_MAX;
4113 				break;
4114 			}
4115 		}
4116 		dstdp.d_reclen = sizeof(dstdp) - sizeof(dstdp.d_name) +
4117 		    ((dp->d_namlen + 1 + 3) &~ 3);
4118 		bcopy(dp->d_name, dstdp.d_name, dstdp.d_namlen);
4119 		bzero(dstdp.d_name + dstdp.d_namlen,
4120 		    dstdp.d_reclen - offsetof(struct freebsd11_dirent, d_name) -
4121 		    dstdp.d_namlen);
4122 		MPASS(dstdp.d_reclen <= dp->d_reclen);
4123 		MPASS(ucount + dstdp.d_reclen <= count);
4124 		if (func != NULL)
4125 			func(&dstdp);
4126 		error = copyout(&dstdp, ubuf + ucount, dstdp.d_reclen);
4127 		if (error != 0)
4128 			break;
4129 		dp = (struct dirent *)((char *)dp + dp->d_reclen);
4130 		ucount += dstdp.d_reclen;
4131 	}
4132 
4133 done:
4134 	free(dirbuf, M_TEMP);
4135 	if (error == 0)
4136 		td->td_retval[0] = ucount;
4137 	return (error);
4138 }
4139 #endif /* COMPAT */
4140 
4141 #ifdef COMPAT_43
4142 static void
4143 ogetdirentries_cvt(struct freebsd11_dirent *dp)
4144 {
4145 #if (BYTE_ORDER == LITTLE_ENDIAN)
4146 	/*
4147 	 * The expected low byte of dp->d_namlen is our dp->d_type.
4148 	 * The high MBZ byte of dp->d_namlen is our dp->d_namlen.
4149 	 */
4150 	dp->d_type = dp->d_namlen;
4151 	dp->d_namlen = 0;
4152 #else
4153 	/*
4154 	 * The dp->d_type is the high byte of the expected dp->d_namlen,
4155 	 * so must be zero'ed.
4156 	 */
4157 	dp->d_type = 0;
4158 #endif
4159 }
4160 
4161 /*
4162  * Read a block of directory entries in a filesystem independent format.
4163  */
4164 #ifndef _SYS_SYSPROTO_H_
4165 struct ogetdirentries_args {
4166 	int	fd;
4167 	char	*buf;
4168 	u_int	count;
4169 	long	*basep;
4170 };
4171 #endif
4172 int
4173 ogetdirentries(struct thread *td, struct ogetdirentries_args *uap)
4174 {
4175 	long loff;
4176 	int error;
4177 
4178 	error = kern_ogetdirentries(td, uap, &loff);
4179 	if (error == 0)
4180 		error = copyout(&loff, uap->basep, sizeof(long));
4181 	return (error);
4182 }
4183 
4184 int
4185 kern_ogetdirentries(struct thread *td, struct ogetdirentries_args *uap,
4186     long *ploff)
4187 {
4188 	long base;
4189 	int error;
4190 
4191 	/* XXX arbitrary sanity limit on `count'. */
4192 	if (uap->count > 64 * 1024)
4193 		return (EINVAL);
4194 
4195 	error = freebsd11_kern_getdirentries(td, uap->fd, uap->buf, uap->count,
4196 	    &base, ogetdirentries_cvt);
4197 
4198 	if (error == 0 && uap->basep != NULL)
4199 		error = copyout(&base, uap->basep, sizeof(long));
4200 
4201 	return (error);
4202 }
4203 #endif /* COMPAT_43 */
4204 
4205 #if defined(COMPAT_FREEBSD11)
4206 #ifndef _SYS_SYSPROTO_H_
4207 struct freebsd11_getdirentries_args {
4208 	int	fd;
4209 	char	*buf;
4210 	u_int	count;
4211 	long	*basep;
4212 };
4213 #endif
4214 int
4215 freebsd11_getdirentries(struct thread *td,
4216     struct freebsd11_getdirentries_args *uap)
4217 {
4218 	long base;
4219 	int error;
4220 
4221 	error = freebsd11_kern_getdirentries(td, uap->fd, uap->buf, uap->count,
4222 	    &base, NULL);
4223 
4224 	if (error == 0 && uap->basep != NULL)
4225 		error = copyout(&base, uap->basep, sizeof(long));
4226 	return (error);
4227 }
4228 
4229 int
4230 freebsd11_getdents(struct thread *td, struct freebsd11_getdents_args *uap)
4231 {
4232 	struct freebsd11_getdirentries_args ap;
4233 
4234 	ap.fd = uap->fd;
4235 	ap.buf = uap->buf;
4236 	ap.count = uap->count;
4237 	ap.basep = NULL;
4238 	return (freebsd11_getdirentries(td, &ap));
4239 }
4240 #endif /* COMPAT_FREEBSD11 */
4241 
4242 /*
4243  * Read a block of directory entries in a filesystem independent format.
4244  */
4245 int
4246 sys_getdirentries(struct thread *td, struct getdirentries_args *uap)
4247 {
4248 	off_t base;
4249 	int error;
4250 
4251 	error = kern_getdirentries(td, uap->fd, uap->buf, uap->count, &base,
4252 	    NULL, UIO_USERSPACE);
4253 	if (error != 0)
4254 		return (error);
4255 	if (uap->basep != NULL)
4256 		error = copyout(&base, uap->basep, sizeof(off_t));
4257 	return (error);
4258 }
4259 
4260 int
4261 kern_getdirentries(struct thread *td, int fd, char *buf, size_t count,
4262     off_t *basep, ssize_t *residp, enum uio_seg bufseg)
4263 {
4264 	struct vnode *vp;
4265 	struct file *fp;
4266 	struct uio auio;
4267 	struct iovec aiov;
4268 	off_t loff;
4269 	int error, eofflag;
4270 	off_t foffset;
4271 
4272 	AUDIT_ARG_FD(fd);
4273 	if (count > IOSIZE_MAX)
4274 		return (EINVAL);
4275 	auio.uio_resid = count;
4276 	error = getvnode(td, fd, &cap_read_rights, &fp);
4277 	if (error != 0)
4278 		return (error);
4279 	if ((fp->f_flag & FREAD) == 0) {
4280 		fdrop(fp, td);
4281 		return (EBADF);
4282 	}
4283 	vp = fp->f_vnode;
4284 	foffset = foffset_lock(fp, 0);
4285 unionread:
4286 	if (vp->v_type != VDIR) {
4287 		error = EINVAL;
4288 		goto fail;
4289 	}
4290 	if (__predict_false((vp->v_vflag & VV_UNLINKED) != 0)) {
4291 		error = ENOENT;
4292 		goto fail;
4293 	}
4294 	aiov.iov_base = buf;
4295 	aiov.iov_len = count;
4296 	auio.uio_iov = &aiov;
4297 	auio.uio_iovcnt = 1;
4298 	auio.uio_rw = UIO_READ;
4299 	auio.uio_segflg = bufseg;
4300 	auio.uio_td = td;
4301 	vn_lock(vp, LK_SHARED | LK_RETRY);
4302 	AUDIT_ARG_VNODE1(vp);
4303 	loff = auio.uio_offset = foffset;
4304 #ifdef MAC
4305 	error = mac_vnode_check_readdir(td->td_ucred, vp);
4306 	if (error == 0)
4307 #endif
4308 		error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, NULL,
4309 		    NULL);
4310 	foffset = auio.uio_offset;
4311 	if (error != 0) {
4312 		VOP_UNLOCK(vp);
4313 		goto fail;
4314 	}
4315 	if (count == auio.uio_resid &&
4316 	    (vp->v_vflag & VV_ROOT) &&
4317 	    (vp->v_mount->mnt_flag & MNT_UNION)) {
4318 		struct vnode *tvp = vp;
4319 
4320 		vp = vp->v_mount->mnt_vnodecovered;
4321 		VREF(vp);
4322 		fp->f_vnode = vp;
4323 		foffset = 0;
4324 		vput(tvp);
4325 		goto unionread;
4326 	}
4327 	VOP_UNLOCK(vp);
4328 	*basep = loff;
4329 	if (residp != NULL)
4330 		*residp = auio.uio_resid;
4331 	td->td_retval[0] = count - auio.uio_resid;
4332 fail:
4333 	foffset_unlock(fp, foffset, 0);
4334 	fdrop(fp, td);
4335 	return (error);
4336 }
4337 
4338 /*
4339  * Set the mode mask for creation of filesystem nodes.
4340  */
4341 #ifndef _SYS_SYSPROTO_H_
4342 struct umask_args {
4343 	int	newmask;
4344 };
4345 #endif
4346 int
4347 sys_umask(struct thread *td, struct umask_args *uap)
4348 {
4349 	struct pwddesc *pdp;
4350 
4351 	pdp = td->td_proc->p_pd;
4352 	PWDDESC_XLOCK(pdp);
4353 	td->td_retval[0] = pdp->pd_cmask;
4354 	pdp->pd_cmask = uap->newmask & ALLPERMS;
4355 	PWDDESC_XUNLOCK(pdp);
4356 	return (0);
4357 }
4358 
4359 /*
4360  * Void all references to file by ripping underlying filesystem away from
4361  * vnode.
4362  */
4363 #ifndef _SYS_SYSPROTO_H_
4364 struct revoke_args {
4365 	char	*path;
4366 };
4367 #endif
4368 int
4369 sys_revoke(struct thread *td, struct revoke_args *uap)
4370 {
4371 	struct vnode *vp;
4372 	struct vattr vattr;
4373 	struct nameidata nd;
4374 	int error;
4375 
4376 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
4377 	    uap->path);
4378 	if ((error = namei(&nd)) != 0)
4379 		return (error);
4380 	vp = nd.ni_vp;
4381 	NDFREE_PNBUF(&nd);
4382 	if (vp->v_type != VCHR || vp->v_rdev == NULL) {
4383 		error = EINVAL;
4384 		goto out;
4385 	}
4386 #ifdef MAC
4387 	error = mac_vnode_check_revoke(td->td_ucred, vp);
4388 	if (error != 0)
4389 		goto out;
4390 #endif
4391 	error = VOP_GETATTR(vp, &vattr, td->td_ucred);
4392 	if (error != 0)
4393 		goto out;
4394 	if (td->td_ucred->cr_uid != vattr.va_uid) {
4395 		error = priv_check(td, PRIV_VFS_ADMIN);
4396 		if (error != 0)
4397 			goto out;
4398 	}
4399 	if (devfs_usecount(vp) > 0)
4400 		VOP_REVOKE(vp, REVOKEALL);
4401 out:
4402 	vput(vp);
4403 	return (error);
4404 }
4405 
4406 /*
4407  * This variant of getvnode() allows O_PATH files.  Caller should
4408  * ensure that returned file and vnode are only used for compatible
4409  * semantics.
4410  */
4411 int
4412 getvnode_path(struct thread *td, int fd, const cap_rights_t *rightsp,
4413     struct file **fpp)
4414 {
4415 	struct file *fp;
4416 	int error;
4417 
4418 	error = fget_unlocked(td, fd, rightsp, &fp);
4419 	if (error != 0)
4420 		return (error);
4421 
4422 	/*
4423 	 * The file could be not of the vnode type, or it may be not
4424 	 * yet fully initialized, in which case the f_vnode pointer
4425 	 * may be set, but f_ops is still badfileops.  E.g.,
4426 	 * devfs_open() transiently create such situation to
4427 	 * facilitate csw d_fdopen().
4428 	 *
4429 	 * Dupfdopen() handling in kern_openat() installs the
4430 	 * half-baked file into the process descriptor table, allowing
4431 	 * other thread to dereference it. Guard against the race by
4432 	 * checking f_ops.
4433 	 */
4434 	if (__predict_false(fp->f_vnode == NULL || fp->f_ops == &badfileops)) {
4435 		fdrop(fp, td);
4436 		*fpp = NULL;
4437 		return (EINVAL);
4438 	}
4439 
4440 	*fpp = fp;
4441 	return (0);
4442 }
4443 
4444 /*
4445  * Convert a user file descriptor to a kernel file entry and check
4446  * that, if it is a capability, the correct rights are present.
4447  * A reference on the file entry is held upon returning.
4448  */
4449 int
4450 getvnode(struct thread *td, int fd, const cap_rights_t *rightsp,
4451     struct file **fpp)
4452 {
4453 	int error;
4454 
4455 	error = getvnode_path(td, fd, rightsp, fpp);
4456 	if (__predict_false(error != 0))
4457 		return (error);
4458 
4459 	/*
4460 	 * Filter out O_PATH file descriptors, most getvnode() callers
4461 	 * do not call fo_ methods.
4462 	 */
4463 	if (__predict_false((*fpp)->f_ops == &path_fileops)) {
4464 		fdrop(*fpp, td);
4465 		*fpp = NULL;
4466 		error = EBADF;
4467 	}
4468 
4469 	return (error);
4470 }
4471 
4472 /*
4473  * Get an (NFS) file handle.
4474  */
4475 #ifndef _SYS_SYSPROTO_H_
4476 struct lgetfh_args {
4477 	char *fname;
4478 	fhandle_t *fhp;
4479 };
4480 #endif
4481 int
4482 sys_lgetfh(struct thread *td, struct lgetfh_args *uap)
4483 {
4484 
4485 	return (kern_getfhat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->fname,
4486 	    UIO_USERSPACE, uap->fhp, UIO_USERSPACE));
4487 }
4488 
4489 #ifndef _SYS_SYSPROTO_H_
4490 struct getfh_args {
4491 	char *fname;
4492 	fhandle_t *fhp;
4493 };
4494 #endif
4495 int
4496 sys_getfh(struct thread *td, struct getfh_args *uap)
4497 {
4498 
4499 	return (kern_getfhat(td, 0, AT_FDCWD, uap->fname, UIO_USERSPACE,
4500 	    uap->fhp, UIO_USERSPACE));
4501 }
4502 
4503 /*
4504  * syscall for the rpc.lockd to use to translate an open descriptor into
4505  * a NFS file handle.
4506  *
4507  * warning: do not remove the priv_check() call or this becomes one giant
4508  * security hole.
4509  */
4510 #ifndef _SYS_SYSPROTO_H_
4511 struct getfhat_args {
4512 	int fd;
4513 	char *path;
4514 	fhandle_t *fhp;
4515 	int flags;
4516 };
4517 #endif
4518 int
4519 sys_getfhat(struct thread *td, struct getfhat_args *uap)
4520 {
4521 
4522 	return (kern_getfhat(td, uap->flags, uap->fd, uap->path, UIO_USERSPACE,
4523 	    uap->fhp, UIO_USERSPACE));
4524 }
4525 
4526 int
4527 kern_getfhat(struct thread *td, int flags, int fd, const char *path,
4528     enum uio_seg pathseg, fhandle_t *fhp, enum uio_seg fhseg)
4529 {
4530 	struct nameidata nd;
4531 	fhandle_t fh;
4532 	struct vnode *vp;
4533 	int error;
4534 
4535 	if ((flags & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH)) != 0)
4536 		return (EINVAL);
4537 	error = priv_check(td, PRIV_VFS_GETFH);
4538 	if (error != 0)
4539 		return (error);
4540 	NDINIT_AT(&nd, LOOKUP, at2cnpflags(flags, AT_SYMLINK_NOFOLLOW |
4541 	    AT_RESOLVE_BENEATH) | LOCKLEAF | AUDITVNODE1, pathseg, path,
4542 	    fd);
4543 	error = namei(&nd);
4544 	if (error != 0)
4545 		return (error);
4546 	NDFREE_PNBUF(&nd);
4547 	vp = nd.ni_vp;
4548 	bzero(&fh, sizeof(fh));
4549 	fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
4550 	error = VOP_VPTOFH(vp, &fh.fh_fid);
4551 	vput(vp);
4552 	if (error == 0) {
4553 		if (fhseg == UIO_USERSPACE)
4554 			error = copyout(&fh, fhp, sizeof (fh));
4555 		else
4556 			memcpy(fhp, &fh, sizeof(fh));
4557 	}
4558 	return (error);
4559 }
4560 
4561 #ifndef _SYS_SYSPROTO_H_
4562 struct fhlink_args {
4563 	fhandle_t *fhp;
4564 	const char *to;
4565 };
4566 #endif
4567 int
4568 sys_fhlink(struct thread *td, struct fhlink_args *uap)
4569 {
4570 
4571 	return (kern_fhlinkat(td, AT_FDCWD, uap->to, UIO_USERSPACE, uap->fhp));
4572 }
4573 
4574 #ifndef _SYS_SYSPROTO_H_
4575 struct fhlinkat_args {
4576 	fhandle_t *fhp;
4577 	int tofd;
4578 	const char *to;
4579 };
4580 #endif
4581 int
4582 sys_fhlinkat(struct thread *td, struct fhlinkat_args *uap)
4583 {
4584 
4585 	return (kern_fhlinkat(td, uap->tofd, uap->to, UIO_USERSPACE, uap->fhp));
4586 }
4587 
4588 static int
4589 kern_fhlinkat(struct thread *td, int fd, const char *path,
4590     enum uio_seg pathseg, fhandle_t *fhp)
4591 {
4592 	fhandle_t fh;
4593 	struct mount *mp;
4594 	struct vnode *vp;
4595 	int error;
4596 
4597 	error = priv_check(td, PRIV_VFS_GETFH);
4598 	if (error != 0)
4599 		return (error);
4600 	error = copyin(fhp, &fh, sizeof(fh));
4601 	if (error != 0)
4602 		return (error);
4603 	do {
4604 		bwillwrite();
4605 		if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4606 			return (ESTALE);
4607 		error = VFS_FHTOVP(mp, &fh.fh_fid, LK_SHARED, &vp);
4608 		vfs_unbusy(mp);
4609 		if (error != 0)
4610 			return (error);
4611 		VOP_UNLOCK(vp);
4612 		error = kern_linkat_vp(td, vp, fd, path, pathseg);
4613 	} while (error == EAGAIN || error == ERELOOKUP);
4614 	return (error);
4615 }
4616 
4617 #ifndef _SYS_SYSPROTO_H_
4618 struct fhreadlink_args {
4619 	fhandle_t *fhp;
4620 	char *buf;
4621 	size_t bufsize;
4622 };
4623 #endif
4624 int
4625 sys_fhreadlink(struct thread *td, struct fhreadlink_args *uap)
4626 {
4627 	fhandle_t fh;
4628 	struct mount *mp;
4629 	struct vnode *vp;
4630 	int error;
4631 
4632 	error = priv_check(td, PRIV_VFS_GETFH);
4633 	if (error != 0)
4634 		return (error);
4635 	if (uap->bufsize > IOSIZE_MAX)
4636 		return (EINVAL);
4637 	error = copyin(uap->fhp, &fh, sizeof(fh));
4638 	if (error != 0)
4639 		return (error);
4640 	if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4641 		return (ESTALE);
4642 	error = VFS_FHTOVP(mp, &fh.fh_fid, LK_SHARED, &vp);
4643 	vfs_unbusy(mp);
4644 	if (error != 0)
4645 		return (error);
4646 	error = kern_readlink_vp(vp, uap->buf, UIO_USERSPACE, uap->bufsize, td);
4647 	vput(vp);
4648 	return (error);
4649 }
4650 
4651 /*
4652  * syscall for the rpc.lockd to use to translate a NFS file handle into an
4653  * open descriptor.
4654  *
4655  * warning: do not remove the priv_check() call or this becomes one giant
4656  * security hole.
4657  */
4658 #ifndef _SYS_SYSPROTO_H_
4659 struct fhopen_args {
4660 	const struct fhandle *u_fhp;
4661 	int flags;
4662 };
4663 #endif
4664 int
4665 sys_fhopen(struct thread *td, struct fhopen_args *uap)
4666 {
4667 	return (kern_fhopen(td, uap->u_fhp, uap->flags));
4668 }
4669 
4670 int
4671 kern_fhopen(struct thread *td, const struct fhandle *u_fhp, int flags)
4672 {
4673 	struct mount *mp;
4674 	struct vnode *vp;
4675 	struct fhandle fhp;
4676 	struct file *fp;
4677 	int error, indx;
4678 	bool named_attr;
4679 
4680 	error = priv_check(td, PRIV_VFS_FHOPEN);
4681 	if (error != 0)
4682 		return (error);
4683 
4684 	indx = -1;
4685 	if ((flags & O_CREAT) != 0)
4686 		return (EINVAL);
4687 	error = openflags(&flags);
4688 	if (error != 0)
4689 		return (error);
4690 	error = copyin(u_fhp, &fhp, sizeof(fhp));
4691 	if (error != 0)
4692 		return (error);
4693 	/* find the mount point */
4694 	mp = vfs_busyfs(&fhp.fh_fsid);
4695 	if (mp == NULL)
4696 		return (ESTALE);
4697 	/* now give me my vnode, it gets returned to me locked */
4698 	error = VFS_FHTOVP(mp, &fhp.fh_fid, LK_EXCLUSIVE, &vp);
4699 	vfs_unbusy(mp);
4700 	if (error != 0)
4701 		return (error);
4702 
4703 	/*
4704 	 * Check to see if the file handle refers to a named attribute
4705 	 * directory or attribute.  If it does, the O_NAMEDATTR flag
4706 	 * must have been specified.
4707 	 */
4708 	named_attr = (vn_irflag_read(vp) &
4709 	    (VIRF_NAMEDDIR | VIRF_NAMEDATTR)) != 0;
4710 	if ((named_attr && (flags & O_NAMEDATTR) == 0) ||
4711 	    (!named_attr && (flags & O_NAMEDATTR) != 0)) {
4712 		vput(vp);
4713 		return (ENOATTR);
4714 	}
4715 
4716 	error = falloc_noinstall(td, &fp);
4717 	if (error != 0) {
4718 		vput(vp);
4719 		return (error);
4720 	}
4721 	/* Set the flags early so the finit in devfs can pick them up. */
4722 	fp->f_flag = flags & FMASK;
4723 
4724 #ifdef INVARIANTS
4725 	td->td_dupfd = -1;
4726 #endif
4727 	error = vn_open_vnode(vp, flags, td->td_ucred, td, fp);
4728 	if (error != 0) {
4729 		KASSERT(fp->f_ops == &badfileops,
4730 		    ("VOP_OPEN in fhopen() set f_ops"));
4731 		KASSERT(td->td_dupfd < 0,
4732 		    ("fhopen() encountered fdopen()"));
4733 
4734 		vput(vp);
4735 		goto bad;
4736 	}
4737 #ifdef INVARIANTS
4738 	td->td_dupfd = 0;
4739 #endif
4740 	finit_open(fp, vp, flags);
4741 	VOP_UNLOCK(vp);
4742 	if ((flags & O_TRUNC) != 0) {
4743 		error = fo_truncate(fp, 0, td->td_ucred, td);
4744 		if (error != 0)
4745 			goto bad;
4746 	}
4747 
4748 	error = finstall(td, fp, &indx, flags, NULL);
4749 bad:
4750 	fdrop(fp, td);
4751 	td->td_retval[0] = indx;
4752 	return (error);
4753 }
4754 
4755 /*
4756  * Stat an (NFS) file handle.
4757  */
4758 #ifndef _SYS_SYSPROTO_H_
4759 struct fhstat_args {
4760 	struct fhandle *u_fhp;
4761 	struct stat *sb;
4762 };
4763 #endif
4764 int
4765 sys_fhstat(struct thread *td, struct fhstat_args *uap)
4766 {
4767 	struct stat sb;
4768 	struct fhandle fh;
4769 	int error;
4770 
4771 	error = copyin(uap->u_fhp, &fh, sizeof(fh));
4772 	if (error != 0)
4773 		return (error);
4774 	error = kern_fhstat(td, fh, &sb);
4775 	if (error == 0)
4776 		error = copyout(&sb, uap->sb, sizeof(sb));
4777 	return (error);
4778 }
4779 
4780 int
4781 kern_fhstat(struct thread *td, struct fhandle fh, struct stat *sb)
4782 {
4783 	struct mount *mp;
4784 	struct vnode *vp;
4785 	int error;
4786 
4787 	error = priv_check(td, PRIV_VFS_FHSTAT);
4788 	if (error != 0)
4789 		return (error);
4790 	if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4791 		return (ESTALE);
4792 	error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4793 	vfs_unbusy(mp);
4794 	if (error != 0)
4795 		return (error);
4796 	error = VOP_STAT(vp, sb, td->td_ucred, NOCRED);
4797 	vput(vp);
4798 	return (error);
4799 }
4800 
4801 /*
4802  * Implement fstatfs() for (NFS) file handles.
4803  */
4804 #ifndef _SYS_SYSPROTO_H_
4805 struct fhstatfs_args {
4806 	struct fhandle *u_fhp;
4807 	struct statfs *buf;
4808 };
4809 #endif
4810 int
4811 sys_fhstatfs(struct thread *td, struct fhstatfs_args *uap)
4812 {
4813 	struct statfs *sfp;
4814 	fhandle_t fh;
4815 	int error;
4816 
4817 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
4818 	if (error != 0)
4819 		return (error);
4820 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
4821 	error = kern_fhstatfs(td, fh, sfp);
4822 	if (error == 0)
4823 		error = copyout(sfp, uap->buf, sizeof(*sfp));
4824 	free(sfp, M_STATFS);
4825 	return (error);
4826 }
4827 
4828 int
4829 kern_fhstatfs(struct thread *td, fhandle_t fh, struct statfs *buf)
4830 {
4831 	struct mount *mp;
4832 	struct vnode *vp;
4833 	int error;
4834 
4835 	error = priv_check(td, PRIV_VFS_FHSTATFS);
4836 	if (error != 0)
4837 		return (error);
4838 	if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4839 		return (ESTALE);
4840 	error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4841 	if (error != 0) {
4842 		vfs_unbusy(mp);
4843 		return (error);
4844 	}
4845 	vput(vp);
4846 	error = prison_canseemount(td->td_ucred, mp);
4847 	if (error != 0)
4848 		goto out;
4849 #ifdef MAC
4850 	error = mac_mount_check_stat(td->td_ucred, mp);
4851 	if (error != 0)
4852 		goto out;
4853 #endif
4854 	error = VFS_STATFS(mp, buf);
4855 out:
4856 	vfs_unbusy(mp);
4857 	return (error);
4858 }
4859 
4860 /*
4861  * Unlike madvise(2), we do not make a best effort to remember every
4862  * possible caching hint.  Instead, we remember the last setting with
4863  * the exception that we will allow POSIX_FADV_NORMAL to adjust the
4864  * region of any current setting.
4865  */
4866 int
4867 kern_posix_fadvise(struct thread *td, int fd, off_t offset, off_t len,
4868     int advice)
4869 {
4870 	struct fadvise_info *fa, *new;
4871 	struct file *fp;
4872 	struct vnode *vp;
4873 	off_t end;
4874 	int error;
4875 
4876 	if (offset < 0 || len < 0 || offset > OFF_MAX - len)
4877 		return (EINVAL);
4878 	AUDIT_ARG_VALUE(advice);
4879 	switch (advice) {
4880 	case POSIX_FADV_SEQUENTIAL:
4881 	case POSIX_FADV_RANDOM:
4882 	case POSIX_FADV_NOREUSE:
4883 		new = malloc(sizeof(*fa), M_FADVISE, M_WAITOK);
4884 		break;
4885 	case POSIX_FADV_NORMAL:
4886 	case POSIX_FADV_WILLNEED:
4887 	case POSIX_FADV_DONTNEED:
4888 		new = NULL;
4889 		break;
4890 	default:
4891 		return (EINVAL);
4892 	}
4893 	/* XXX: CAP_POSIX_FADVISE? */
4894 	AUDIT_ARG_FD(fd);
4895 	error = fget(td, fd, &cap_no_rights, &fp);
4896 	if (error != 0)
4897 		goto out;
4898 	AUDIT_ARG_FILE(td->td_proc, fp);
4899 	if ((fp->f_ops->fo_flags & DFLAG_SEEKABLE) == 0) {
4900 		error = ESPIPE;
4901 		goto out;
4902 	}
4903 	if (fp->f_type != DTYPE_VNODE) {
4904 		error = ENODEV;
4905 		goto out;
4906 	}
4907 	vp = fp->f_vnode;
4908 	if (vp->v_type != VREG) {
4909 		error = ENODEV;
4910 		goto out;
4911 	}
4912 	if (len == 0)
4913 		end = OFF_MAX;
4914 	else
4915 		end = offset + len - 1;
4916 	switch (advice) {
4917 	case POSIX_FADV_SEQUENTIAL:
4918 	case POSIX_FADV_RANDOM:
4919 	case POSIX_FADV_NOREUSE:
4920 		/*
4921 		 * Try to merge any existing non-standard region with
4922 		 * this new region if possible, otherwise create a new
4923 		 * non-standard region for this request.
4924 		 */
4925 		mtx_pool_lock(mtxpool_sleep, fp);
4926 		fa = fp->f_advice;
4927 		if (fa != NULL && fa->fa_advice == advice &&
4928 		    ((fa->fa_start <= end && fa->fa_end >= offset) ||
4929 		    (end != OFF_MAX && fa->fa_start == end + 1) ||
4930 		    (fa->fa_end != OFF_MAX && fa->fa_end + 1 == offset))) {
4931 			if (offset < fa->fa_start)
4932 				fa->fa_start = offset;
4933 			if (end > fa->fa_end)
4934 				fa->fa_end = end;
4935 		} else {
4936 			new->fa_advice = advice;
4937 			new->fa_start = offset;
4938 			new->fa_end = end;
4939 			fp->f_advice = new;
4940 			new = fa;
4941 		}
4942 		mtx_pool_unlock(mtxpool_sleep, fp);
4943 		break;
4944 	case POSIX_FADV_NORMAL:
4945 		/*
4946 		 * If a the "normal" region overlaps with an existing
4947 		 * non-standard region, trim or remove the
4948 		 * non-standard region.
4949 		 */
4950 		mtx_pool_lock(mtxpool_sleep, fp);
4951 		fa = fp->f_advice;
4952 		if (fa != NULL) {
4953 			if (offset <= fa->fa_start && end >= fa->fa_end) {
4954 				new = fa;
4955 				fp->f_advice = NULL;
4956 			} else if (offset <= fa->fa_start &&
4957 			    end >= fa->fa_start)
4958 				fa->fa_start = end + 1;
4959 			else if (offset <= fa->fa_end && end >= fa->fa_end)
4960 				fa->fa_end = offset - 1;
4961 			else if (offset >= fa->fa_start && end <= fa->fa_end) {
4962 				/*
4963 				 * If the "normal" region is a middle
4964 				 * portion of the existing
4965 				 * non-standard region, just remove
4966 				 * the whole thing rather than picking
4967 				 * one side or the other to
4968 				 * preserve.
4969 				 */
4970 				new = fa;
4971 				fp->f_advice = NULL;
4972 			}
4973 		}
4974 		mtx_pool_unlock(mtxpool_sleep, fp);
4975 		break;
4976 	case POSIX_FADV_WILLNEED:
4977 	case POSIX_FADV_DONTNEED:
4978 		error = VOP_ADVISE(vp, offset, end, advice);
4979 		break;
4980 	}
4981 out:
4982 	if (fp != NULL)
4983 		fdrop(fp, td);
4984 	free(new, M_FADVISE);
4985 	return (error);
4986 }
4987 
4988 int
4989 sys_posix_fadvise(struct thread *td, struct posix_fadvise_args *uap)
4990 {
4991 	int error;
4992 
4993 	error = kern_posix_fadvise(td, uap->fd, uap->offset, uap->len,
4994 	    uap->advice);
4995 	return (kern_posix_error(td, error));
4996 }
4997 
4998 int
4999 kern_copy_file_range(struct thread *td, int infd, off_t *inoffp, int outfd,
5000     off_t *outoffp, size_t len, unsigned int flags)
5001 {
5002 	struct file *infp, *infp1, *outfp, *outfp1;
5003 	struct vnode *invp, *outvp;
5004 	int error;
5005 	size_t retlen;
5006 	void *rl_rcookie, *rl_wcookie;
5007 	off_t inoff, outoff, savinoff, savoutoff;
5008 	bool foffsets_locked;
5009 
5010 	infp = outfp = NULL;
5011 	rl_rcookie = rl_wcookie = NULL;
5012 	foffsets_locked = false;
5013 	error = 0;
5014 	retlen = 0;
5015 
5016 	if (flags != 0) {
5017 		error = EINVAL;
5018 		goto out;
5019 	}
5020 	if (len > SSIZE_MAX)
5021 		/*
5022 		 * Although the len argument is size_t, the return argument
5023 		 * is ssize_t (which is signed).  Therefore a size that won't
5024 		 * fit in ssize_t can't be returned.
5025 		 */
5026 		len = SSIZE_MAX;
5027 
5028 	/* Get the file structures for the file descriptors. */
5029 	error = fget_read(td, infd,
5030 	    inoffp != NULL ? &cap_pread_rights : &cap_read_rights, &infp);
5031 	if (error != 0)
5032 		goto out;
5033 	if (infp->f_ops == &badfileops) {
5034 		error = EBADF;
5035 		goto out;
5036 	}
5037 	if (infp->f_vnode == NULL) {
5038 		error = EINVAL;
5039 		goto out;
5040 	}
5041 	error = fget_write(td, outfd,
5042 	    outoffp != NULL ? &cap_pwrite_rights : &cap_write_rights, &outfp);
5043 	if (error != 0)
5044 		goto out;
5045 	if (outfp->f_ops == &badfileops) {
5046 		error = EBADF;
5047 		goto out;
5048 	}
5049 	if (outfp->f_vnode == NULL) {
5050 		error = EINVAL;
5051 		goto out;
5052 	}
5053 
5054 	/*
5055 	 * Figure out which file offsets we're reading from and writing to.
5056 	 * If the offsets come from the file descriptions, we need to lock them,
5057 	 * and locking both offsets requires a loop to avoid deadlocks.
5058 	 */
5059 	infp1 = outfp1 = NULL;
5060 	if (inoffp != NULL)
5061 		inoff = *inoffp;
5062 	else
5063 		infp1 = infp;
5064 	if (outoffp != NULL)
5065 		outoff = *outoffp;
5066 	else
5067 		outfp1 = outfp;
5068 	if (infp1 != NULL || outfp1 != NULL) {
5069 		if (infp1 == outfp1) {
5070 			/*
5071 			 * Overlapping ranges are not allowed.  A more thorough
5072 			 * check appears below, but we must not lock the same
5073 			 * offset twice.
5074 			 */
5075 			error = EINVAL;
5076 			goto out;
5077 		}
5078 		foffset_lock_pair(infp1, &inoff, outfp1, &outoff, 0);
5079 		foffsets_locked = true;
5080 	}
5081 	savinoff = inoff;
5082 	savoutoff = outoff;
5083 
5084 	invp = infp->f_vnode;
5085 	outvp = outfp->f_vnode;
5086 	/* Sanity check the f_flag bits. */
5087 	if ((outfp->f_flag & (FWRITE | FAPPEND)) != FWRITE ||
5088 	    (infp->f_flag & FREAD) == 0) {
5089 		error = EBADF;
5090 		goto out;
5091 	}
5092 
5093 	/* If len == 0, just return 0. */
5094 	if (len == 0)
5095 		goto out;
5096 
5097 	/*
5098 	 * Make sure that the ranges we check and lock below are valid.  Note
5099 	 * that len is clamped to SSIZE_MAX above.
5100 	 */
5101 	if (inoff < 0 || outoff < 0) {
5102 		error = EINVAL;
5103 		goto out;
5104 	}
5105 
5106 	/*
5107 	 * If infp and outfp refer to the same file, the byte ranges cannot
5108 	 * overlap.
5109 	 */
5110 	if (invp == outvp) {
5111 		if ((inoff <= outoff && inoff + len > outoff) ||
5112 		    (inoff > outoff && outoff + len > inoff)) {
5113 			error = EINVAL;
5114 			goto out;
5115 		}
5116 		rangelock_may_recurse(&invp->v_rl);
5117 	}
5118 
5119 	/* Range lock the byte ranges for both invp and outvp. */
5120 	for (;;) {
5121 		rl_wcookie = vn_rangelock_wlock(outvp, outoff, outoff + len);
5122 		rl_rcookie = vn_rangelock_tryrlock(invp, inoff, inoff + len);
5123 		if (rl_rcookie != NULL)
5124 			break;
5125 		vn_rangelock_unlock(outvp, rl_wcookie);
5126 		rl_rcookie = vn_rangelock_rlock(invp, inoff, inoff + len);
5127 		vn_rangelock_unlock(invp, rl_rcookie);
5128 	}
5129 
5130 	retlen = len;
5131 	error = vn_copy_file_range(invp, &inoff, outvp, &outoff, &retlen,
5132 	    flags, infp->f_cred, outfp->f_cred, td);
5133 out:
5134 	if (rl_rcookie != NULL)
5135 		vn_rangelock_unlock(invp, rl_rcookie);
5136 	if (rl_wcookie != NULL)
5137 		vn_rangelock_unlock(outvp, rl_wcookie);
5138 	if (foffsets_locked) {
5139 		if (error == EINTR || error == ERESTART) {
5140 			inoff = savinoff;
5141 			outoff = savoutoff;
5142 		}
5143 		if (inoffp == NULL)
5144 			foffset_unlock(infp, inoff, 0);
5145 		else
5146 			*inoffp = inoff;
5147 		if (outoffp == NULL)
5148 			foffset_unlock(outfp, outoff, 0);
5149 		else
5150 			*outoffp = outoff;
5151 	}
5152 	if (outfp != NULL)
5153 		fdrop(outfp, td);
5154 	if (infp != NULL)
5155 		fdrop(infp, td);
5156 	td->td_retval[0] = retlen;
5157 	return (error);
5158 }
5159 
5160 int
5161 sys_copy_file_range(struct thread *td, struct copy_file_range_args *uap)
5162 {
5163 	off_t inoff, outoff, *inoffp, *outoffp;
5164 	int error;
5165 
5166 	inoffp = outoffp = NULL;
5167 	if (uap->inoffp != NULL) {
5168 		error = copyin(uap->inoffp, &inoff, sizeof(off_t));
5169 		if (error != 0)
5170 			return (error);
5171 		inoffp = &inoff;
5172 	}
5173 	if (uap->outoffp != NULL) {
5174 		error = copyin(uap->outoffp, &outoff, sizeof(off_t));
5175 		if (error != 0)
5176 			return (error);
5177 		outoffp = &outoff;
5178 	}
5179 	error = kern_copy_file_range(td, uap->infd, inoffp, uap->outfd,
5180 	    outoffp, uap->len, uap->flags);
5181 	if (error == 0 && uap->inoffp != NULL)
5182 		error = copyout(inoffp, uap->inoffp, sizeof(off_t));
5183 	if (error == 0 && uap->outoffp != NULL)
5184 		error = copyout(outoffp, uap->outoffp, sizeof(off_t));
5185 	return (error);
5186 }
5187