xref: /freebsd/sys/kern/vfs_syscalls.c (revision 7fdf597e96a02165cfe22ff357b857d5fa15ed8a)
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 <sys/cdefs.h>
38 #include "opt_capsicum.h"
39 #include "opt_ktrace.h"
40 
41 #include <sys/param.h>
42 #include <sys/systm.h>
43 #ifdef COMPAT_FREEBSD11
44 #include <sys/abi_compat.h>
45 #endif
46 #include <sys/bio.h>
47 #include <sys/buf.h>
48 #include <sys/capsicum.h>
49 #include <sys/disk.h>
50 #include <sys/malloc.h>
51 #include <sys/mount.h>
52 #include <sys/mutex.h>
53 #include <sys/sysproto.h>
54 #include <sys/namei.h>
55 #include <sys/filedesc.h>
56 #include <sys/kernel.h>
57 #include <sys/fcntl.h>
58 #include <sys/file.h>
59 #include <sys/filio.h>
60 #include <sys/limits.h>
61 #include <sys/linker.h>
62 #include <sys/rwlock.h>
63 #include <sys/sdt.h>
64 #include <sys/stat.h>
65 #include <sys/sx.h>
66 #include <sys/unistd.h>
67 #include <sys/vnode.h>
68 #include <sys/priv.h>
69 #include <sys/proc.h>
70 #include <sys/dirent.h>
71 #include <sys/jail.h>
72 #include <sys/syscallsubr.h>
73 #include <sys/sysctl.h>
74 #ifdef KTRACE
75 #include <sys/ktrace.h>
76 #endif
77 
78 #include <machine/stdarg.h>
79 
80 #include <security/audit/audit.h>
81 #include <security/mac/mac_framework.h>
82 
83 #include <vm/vm.h>
84 #include <vm/vm_object.h>
85 #include <vm/vm_page.h>
86 #include <vm/vnode_pager.h>
87 #include <vm/uma.h>
88 
89 #include <fs/devfs/devfs.h>
90 
91 MALLOC_DEFINE(M_FADVISE, "fadvise", "posix_fadvise(2) information");
92 
93 static int kern_chflagsat(struct thread *td, int fd, const char *path,
94     enum uio_seg pathseg, u_long flags, int atflag);
95 static int setfflags(struct thread *td, struct vnode *, u_long);
96 static int getutimes(const struct timeval *, enum uio_seg, struct timespec *);
97 static int getutimens(const struct timespec *, enum uio_seg,
98     struct timespec *, int *);
99 static int setutimes(struct thread *td, struct vnode *,
100     const struct timespec *, int, int);
101 static int vn_access(struct vnode *vp, int user_flags, struct ucred *cred,
102     struct thread *td);
103 static int kern_fhlinkat(struct thread *td, int fd, const char *path,
104     enum uio_seg pathseg, fhandle_t *fhp);
105 static int kern_readlink_vp(struct vnode *vp, char *buf, enum uio_seg bufseg,
106     size_t count, struct thread *td);
107 static int kern_linkat_vp(struct thread *td, struct vnode *vp, int fd,
108     const char *path, enum uio_seg segflag);
109 
110 uint64_t
111 at2cnpflags(u_int at_flags, u_int mask)
112 {
113 	uint64_t res;
114 
115 	MPASS((at_flags & (AT_SYMLINK_FOLLOW | AT_SYMLINK_NOFOLLOW)) !=
116 	    (AT_SYMLINK_FOLLOW | AT_SYMLINK_NOFOLLOW));
117 
118 	res = 0;
119 	at_flags &= mask;
120 	if ((at_flags & AT_RESOLVE_BENEATH) != 0)
121 		res |= RBENEATH;
122 	if ((at_flags & AT_SYMLINK_FOLLOW) != 0)
123 		res |= FOLLOW;
124 	/* NOFOLLOW is pseudo flag */
125 	if ((mask & AT_SYMLINK_NOFOLLOW) != 0) {
126 		res |= (at_flags & AT_SYMLINK_NOFOLLOW) != 0 ? NOFOLLOW :
127 		    FOLLOW;
128 	}
129 	if ((mask & AT_EMPTY_PATH) != 0 && (at_flags & AT_EMPTY_PATH) != 0)
130 		res |= EMPTYPATH;
131 	return (res);
132 }
133 
134 int
135 kern_sync(struct thread *td)
136 {
137 	struct mount *mp, *nmp;
138 	int save;
139 
140 	mtx_lock(&mountlist_mtx);
141 	for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
142 		if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK)) {
143 			nmp = TAILQ_NEXT(mp, mnt_list);
144 			continue;
145 		}
146 		if ((mp->mnt_flag & MNT_RDONLY) == 0 &&
147 		    vn_start_write(NULL, &mp, V_NOWAIT) == 0) {
148 			save = curthread_pflags_set(TDP_SYNCIO);
149 			vfs_periodic(mp, MNT_NOWAIT);
150 			VFS_SYNC(mp, MNT_NOWAIT);
151 			curthread_pflags_restore(save);
152 			vn_finished_write(mp);
153 		}
154 		mtx_lock(&mountlist_mtx);
155 		nmp = TAILQ_NEXT(mp, mnt_list);
156 		vfs_unbusy(mp);
157 	}
158 	mtx_unlock(&mountlist_mtx);
159 	return (0);
160 }
161 
162 /*
163  * Sync each mounted filesystem.
164  */
165 #ifndef _SYS_SYSPROTO_H_
166 struct sync_args {
167 	int     dummy;
168 };
169 #endif
170 /* ARGSUSED */
171 int
172 sys_sync(struct thread *td, struct sync_args *uap)
173 {
174 
175 	return (kern_sync(td));
176 }
177 
178 /*
179  * Change filesystem quotas.
180  */
181 #ifndef _SYS_SYSPROTO_H_
182 struct quotactl_args {
183 	char *path;
184 	int cmd;
185 	int uid;
186 	caddr_t arg;
187 };
188 #endif
189 int
190 sys_quotactl(struct thread *td, struct quotactl_args *uap)
191 {
192 	struct mount *mp;
193 	struct nameidata nd;
194 	int error;
195 	bool mp_busy;
196 
197 	AUDIT_ARG_CMD(uap->cmd);
198 	AUDIT_ARG_UID(uap->uid);
199 	if (!prison_allow(td->td_ucred, PR_ALLOW_QUOTAS))
200 		return (EPERM);
201 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
202 	    uap->path);
203 	if ((error = namei(&nd)) != 0)
204 		return (error);
205 	NDFREE_PNBUF(&nd);
206 	mp = nd.ni_vp->v_mount;
207 	vfs_ref(mp);
208 	vput(nd.ni_vp);
209 	error = vfs_busy(mp, 0);
210 	if (error != 0) {
211 		vfs_rel(mp);
212 		return (error);
213 	}
214 	mp_busy = true;
215 	error = VFS_QUOTACTL(mp, uap->cmd, uap->uid, uap->arg, &mp_busy);
216 
217 	/*
218 	 * Since quota on/off operations typically need to open quota
219 	 * files, the implementation may need to unbusy the mount point
220 	 * before calling into namei.  Otherwise, unmount might be
221 	 * started between two vfs_busy() invocations (first is ours,
222 	 * second is from mount point cross-walk code in lookup()),
223 	 * causing deadlock.
224 	 *
225 	 * Avoid unbusying mp if the implementation indicates it has
226 	 * already done so.
227 	 */
228 	if (mp_busy)
229 		vfs_unbusy(mp);
230 	vfs_rel(mp);
231 	return (error);
232 }
233 
234 /*
235  * Used by statfs conversion routines to scale the block size up if
236  * necessary so that all of the block counts are <= 'max_size'.  Note
237  * that 'max_size' should be a bitmask, i.e. 2^n - 1 for some non-zero
238  * value of 'n'.
239  */
240 void
241 statfs_scale_blocks(struct statfs *sf, long max_size)
242 {
243 	uint64_t count;
244 	int shift;
245 
246 	KASSERT(powerof2(max_size + 1), ("%s: invalid max_size", __func__));
247 
248 	/*
249 	 * Attempt to scale the block counts to give a more accurate
250 	 * overview to userland of the ratio of free space to used
251 	 * space.  To do this, find the largest block count and compute
252 	 * a divisor that lets it fit into a signed integer <= max_size.
253 	 */
254 	if (sf->f_bavail < 0)
255 		count = -sf->f_bavail;
256 	else
257 		count = sf->f_bavail;
258 	count = MAX(sf->f_blocks, MAX(sf->f_bfree, count));
259 	if (count <= max_size)
260 		return;
261 
262 	count >>= flsl(max_size);
263 	shift = 0;
264 	while (count > 0) {
265 		shift++;
266 		count >>=1;
267 	}
268 
269 	sf->f_bsize <<= shift;
270 	sf->f_blocks >>= shift;
271 	sf->f_bfree >>= shift;
272 	sf->f_bavail >>= shift;
273 }
274 
275 static int
276 kern_do_statfs(struct thread *td, struct mount *mp, struct statfs *buf)
277 {
278 	int error;
279 
280 	if (mp == NULL)
281 		return (EBADF);
282 	error = vfs_busy(mp, 0);
283 	vfs_rel(mp);
284 	if (error != 0)
285 		return (error);
286 #ifdef MAC
287 	error = mac_mount_check_stat(td->td_ucred, mp);
288 	if (error != 0)
289 		goto out;
290 #endif
291 	error = VFS_STATFS(mp, buf);
292 	if (error != 0)
293 		goto out;
294 	if (priv_check_cred_vfs_generation(td->td_ucred)) {
295 		buf->f_fsid.val[0] = buf->f_fsid.val[1] = 0;
296 		prison_enforce_statfs(td->td_ucred, mp, buf);
297 	}
298 out:
299 	vfs_unbusy(mp);
300 	return (error);
301 }
302 
303 /*
304  * Get filesystem statistics.
305  */
306 #ifndef _SYS_SYSPROTO_H_
307 struct statfs_args {
308 	char *path;
309 	struct statfs *buf;
310 };
311 #endif
312 int
313 sys_statfs(struct thread *td, struct statfs_args *uap)
314 {
315 	struct statfs *sfp;
316 	int error;
317 
318 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
319 	error = kern_statfs(td, uap->path, UIO_USERSPACE, sfp);
320 	if (error == 0)
321 		error = copyout(sfp, uap->buf, sizeof(struct statfs));
322 	free(sfp, M_STATFS);
323 	return (error);
324 }
325 
326 int
327 kern_statfs(struct thread *td, const char *path, enum uio_seg pathseg,
328     struct statfs *buf)
329 {
330 	struct mount *mp;
331 	struct nameidata nd;
332 	int error;
333 
334 	NDINIT(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path);
335 	error = namei(&nd);
336 	if (error != 0)
337 		return (error);
338 	NDFREE_PNBUF(&nd);
339 	mp = vfs_ref_from_vp(nd.ni_vp);
340 	vrele(nd.ni_vp);
341 	return (kern_do_statfs(td, mp, buf));
342 }
343 
344 /*
345  * Get filesystem statistics.
346  */
347 #ifndef _SYS_SYSPROTO_H_
348 struct fstatfs_args {
349 	int fd;
350 	struct statfs *buf;
351 };
352 #endif
353 int
354 sys_fstatfs(struct thread *td, struct fstatfs_args *uap)
355 {
356 	struct statfs *sfp;
357 	int error;
358 
359 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
360 	error = kern_fstatfs(td, uap->fd, sfp);
361 	if (error == 0)
362 		error = copyout(sfp, uap->buf, sizeof(struct statfs));
363 	free(sfp, M_STATFS);
364 	return (error);
365 }
366 
367 int
368 kern_fstatfs(struct thread *td, int fd, struct statfs *buf)
369 {
370 	struct file *fp;
371 	struct mount *mp;
372 	struct vnode *vp;
373 	int error;
374 
375 	AUDIT_ARG_FD(fd);
376 	error = getvnode_path(td, fd, &cap_fstatfs_rights, &fp);
377 	if (error != 0)
378 		return (error);
379 	vp = fp->f_vnode;
380 #ifdef AUDIT
381 	if (AUDITING_TD(td)) {
382 		vn_lock(vp, LK_SHARED | LK_RETRY);
383 		AUDIT_ARG_VNODE1(vp);
384 		VOP_UNLOCK(vp);
385 	}
386 #endif
387 	mp = vfs_ref_from_vp(vp);
388 	fdrop(fp, td);
389 	return (kern_do_statfs(td, mp, buf));
390 }
391 
392 /*
393  * Get statistics on all filesystems.
394  */
395 #ifndef _SYS_SYSPROTO_H_
396 struct getfsstat_args {
397 	struct statfs *buf;
398 	long bufsize;
399 	int mode;
400 };
401 #endif
402 int
403 sys_getfsstat(struct thread *td, struct getfsstat_args *uap)
404 {
405 	size_t count;
406 	int error;
407 
408 	if (uap->bufsize < 0 || uap->bufsize > SIZE_MAX)
409 		return (EINVAL);
410 	error = kern_getfsstat(td, &uap->buf, uap->bufsize, &count,
411 	    UIO_USERSPACE, uap->mode);
412 	if (error == 0)
413 		td->td_retval[0] = count;
414 	return (error);
415 }
416 
417 /*
418  * If (bufsize > 0 && bufseg == UIO_SYSSPACE)
419  *	The caller is responsible for freeing memory which will be allocated
420  *	in '*buf'.
421  */
422 int
423 kern_getfsstat(struct thread *td, struct statfs **buf, size_t bufsize,
424     size_t *countp, enum uio_seg bufseg, int mode)
425 {
426 	struct mount *mp, *nmp;
427 	struct statfs *sfsp, *sp, *sptmp, *tofree;
428 	size_t count, maxcount;
429 	int error;
430 
431 	switch (mode) {
432 	case MNT_WAIT:
433 	case MNT_NOWAIT:
434 		break;
435 	default:
436 		if (bufseg == UIO_SYSSPACE)
437 			*buf = NULL;
438 		return (EINVAL);
439 	}
440 restart:
441 	maxcount = bufsize / sizeof(struct statfs);
442 	if (bufsize == 0) {
443 		sfsp = NULL;
444 		tofree = NULL;
445 	} else if (bufseg == UIO_USERSPACE) {
446 		sfsp = *buf;
447 		tofree = NULL;
448 	} else /* if (bufseg == UIO_SYSSPACE) */ {
449 		count = 0;
450 		mtx_lock(&mountlist_mtx);
451 		TAILQ_FOREACH(mp, &mountlist, mnt_list) {
452 			count++;
453 		}
454 		mtx_unlock(&mountlist_mtx);
455 		if (maxcount > count)
456 			maxcount = count;
457 		tofree = sfsp = *buf = malloc(maxcount * sizeof(struct statfs),
458 		    M_STATFS, M_WAITOK);
459 	}
460 
461 	count = 0;
462 
463 	/*
464 	 * If there is no target buffer they only want the count.
465 	 *
466 	 * This could be TAILQ_FOREACH but it is open-coded to match the original
467 	 * code below.
468 	 */
469 	if (sfsp == NULL) {
470 		mtx_lock(&mountlist_mtx);
471 		for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
472 			if (prison_canseemount(td->td_ucred, mp) != 0) {
473 				nmp = TAILQ_NEXT(mp, mnt_list);
474 				continue;
475 			}
476 #ifdef MAC
477 			if (mac_mount_check_stat(td->td_ucred, mp) != 0) {
478 				nmp = TAILQ_NEXT(mp, mnt_list);
479 				continue;
480 			}
481 #endif
482 			count++;
483 			nmp = TAILQ_NEXT(mp, mnt_list);
484 		}
485 		mtx_unlock(&mountlist_mtx);
486 		*countp = count;
487 		return (0);
488 	}
489 
490 	/*
491 	 * They want the entire thing.
492 	 *
493 	 * Short-circuit the corner case of no room for anything, avoids
494 	 * relocking below.
495 	 */
496 	if (maxcount < 1) {
497 		goto out;
498 	}
499 
500 	mtx_lock(&mountlist_mtx);
501 	for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
502 		if (prison_canseemount(td->td_ucred, mp) != 0) {
503 			nmp = TAILQ_NEXT(mp, mnt_list);
504 			continue;
505 		}
506 #ifdef MAC
507 		if (mac_mount_check_stat(td->td_ucred, mp) != 0) {
508 			nmp = TAILQ_NEXT(mp, mnt_list);
509 			continue;
510 		}
511 #endif
512 		if (mode == MNT_WAIT) {
513 			if (vfs_busy(mp, MBF_MNTLSTLOCK) != 0) {
514 				/*
515 				 * If vfs_busy() failed, and MBF_NOWAIT
516 				 * wasn't passed, then the mp is gone.
517 				 * Furthermore, because of MBF_MNTLSTLOCK,
518 				 * the mountlist_mtx was dropped.  We have
519 				 * no other choice than to start over.
520 				 */
521 				mtx_unlock(&mountlist_mtx);
522 				free(tofree, M_STATFS);
523 				goto restart;
524 			}
525 		} else {
526 			if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK) != 0) {
527 				nmp = TAILQ_NEXT(mp, mnt_list);
528 				continue;
529 			}
530 		}
531 		sp = &mp->mnt_stat;
532 		/*
533 		 * If MNT_NOWAIT is specified, do not refresh
534 		 * the fsstat cache.
535 		 */
536 		if (mode != MNT_NOWAIT) {
537 			error = VFS_STATFS(mp, sp);
538 			if (error != 0) {
539 				mtx_lock(&mountlist_mtx);
540 				nmp = TAILQ_NEXT(mp, mnt_list);
541 				vfs_unbusy(mp);
542 				continue;
543 			}
544 		}
545 		if (priv_check_cred_vfs_generation(td->td_ucred)) {
546 			sptmp = malloc(sizeof(struct statfs), M_STATFS,
547 			    M_WAITOK);
548 			*sptmp = *sp;
549 			sptmp->f_fsid.val[0] = sptmp->f_fsid.val[1] = 0;
550 			prison_enforce_statfs(td->td_ucred, mp, sptmp);
551 			sp = sptmp;
552 		} else
553 			sptmp = NULL;
554 		if (bufseg == UIO_SYSSPACE) {
555 			bcopy(sp, sfsp, sizeof(*sp));
556 			free(sptmp, M_STATFS);
557 		} else /* if (bufseg == UIO_USERSPACE) */ {
558 			error = copyout(sp, sfsp, sizeof(*sp));
559 			free(sptmp, M_STATFS);
560 			if (error != 0) {
561 				vfs_unbusy(mp);
562 				return (error);
563 			}
564 		}
565 		sfsp++;
566 		count++;
567 
568 		if (count == maxcount) {
569 			vfs_unbusy(mp);
570 			goto out;
571 		}
572 
573 		mtx_lock(&mountlist_mtx);
574 		nmp = TAILQ_NEXT(mp, mnt_list);
575 		vfs_unbusy(mp);
576 	}
577 	mtx_unlock(&mountlist_mtx);
578 out:
579 	*countp = count;
580 	return (0);
581 }
582 
583 #ifdef COMPAT_FREEBSD4
584 /*
585  * Get old format filesystem statistics.
586  */
587 static void freebsd4_cvtstatfs(struct statfs *, struct ostatfs *);
588 
589 #ifndef _SYS_SYSPROTO_H_
590 struct freebsd4_statfs_args {
591 	char *path;
592 	struct ostatfs *buf;
593 };
594 #endif
595 int
596 freebsd4_statfs(struct thread *td, struct freebsd4_statfs_args *uap)
597 {
598 	struct ostatfs osb;
599 	struct statfs *sfp;
600 	int error;
601 
602 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
603 	error = kern_statfs(td, uap->path, UIO_USERSPACE, sfp);
604 	if (error == 0) {
605 		freebsd4_cvtstatfs(sfp, &osb);
606 		error = copyout(&osb, uap->buf, sizeof(osb));
607 	}
608 	free(sfp, M_STATFS);
609 	return (error);
610 }
611 
612 /*
613  * Get filesystem statistics.
614  */
615 #ifndef _SYS_SYSPROTO_H_
616 struct freebsd4_fstatfs_args {
617 	int fd;
618 	struct ostatfs *buf;
619 };
620 #endif
621 int
622 freebsd4_fstatfs(struct thread *td, struct freebsd4_fstatfs_args *uap)
623 {
624 	struct ostatfs osb;
625 	struct statfs *sfp;
626 	int error;
627 
628 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
629 	error = kern_fstatfs(td, uap->fd, sfp);
630 	if (error == 0) {
631 		freebsd4_cvtstatfs(sfp, &osb);
632 		error = copyout(&osb, uap->buf, sizeof(osb));
633 	}
634 	free(sfp, M_STATFS);
635 	return (error);
636 }
637 
638 /*
639  * Get statistics on all filesystems.
640  */
641 #ifndef _SYS_SYSPROTO_H_
642 struct freebsd4_getfsstat_args {
643 	struct ostatfs *buf;
644 	long bufsize;
645 	int mode;
646 };
647 #endif
648 int
649 freebsd4_getfsstat(struct thread *td, struct freebsd4_getfsstat_args *uap)
650 {
651 	struct statfs *buf, *sp;
652 	struct ostatfs osb;
653 	size_t count, size;
654 	int error;
655 
656 	if (uap->bufsize < 0)
657 		return (EINVAL);
658 	count = uap->bufsize / sizeof(struct ostatfs);
659 	if (count > SIZE_MAX / sizeof(struct statfs))
660 		return (EINVAL);
661 	size = count * sizeof(struct statfs);
662 	error = kern_getfsstat(td, &buf, size, &count, UIO_SYSSPACE,
663 	    uap->mode);
664 	if (error == 0)
665 		td->td_retval[0] = count;
666 	if (size != 0) {
667 		sp = buf;
668 		while (count != 0 && error == 0) {
669 			freebsd4_cvtstatfs(sp, &osb);
670 			error = copyout(&osb, uap->buf, sizeof(osb));
671 			sp++;
672 			uap->buf++;
673 			count--;
674 		}
675 		free(buf, M_STATFS);
676 	}
677 	return (error);
678 }
679 
680 /*
681  * Implement fstatfs() for (NFS) file handles.
682  */
683 #ifndef _SYS_SYSPROTO_H_
684 struct freebsd4_fhstatfs_args {
685 	struct fhandle *u_fhp;
686 	struct ostatfs *buf;
687 };
688 #endif
689 int
690 freebsd4_fhstatfs(struct thread *td, struct freebsd4_fhstatfs_args *uap)
691 {
692 	struct ostatfs osb;
693 	struct statfs *sfp;
694 	fhandle_t fh;
695 	int error;
696 
697 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
698 	if (error != 0)
699 		return (error);
700 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
701 	error = kern_fhstatfs(td, fh, sfp);
702 	if (error == 0) {
703 		freebsd4_cvtstatfs(sfp, &osb);
704 		error = copyout(&osb, uap->buf, sizeof(osb));
705 	}
706 	free(sfp, M_STATFS);
707 	return (error);
708 }
709 
710 /*
711  * Convert a new format statfs structure to an old format statfs structure.
712  */
713 static void
714 freebsd4_cvtstatfs(struct statfs *nsp, struct ostatfs *osp)
715 {
716 
717 	statfs_scale_blocks(nsp, LONG_MAX);
718 	bzero(osp, sizeof(*osp));
719 	osp->f_bsize = nsp->f_bsize;
720 	osp->f_iosize = MIN(nsp->f_iosize, LONG_MAX);
721 	osp->f_blocks = nsp->f_blocks;
722 	osp->f_bfree = nsp->f_bfree;
723 	osp->f_bavail = nsp->f_bavail;
724 	osp->f_files = MIN(nsp->f_files, LONG_MAX);
725 	osp->f_ffree = MIN(nsp->f_ffree, LONG_MAX);
726 	osp->f_owner = nsp->f_owner;
727 	osp->f_type = nsp->f_type;
728 	osp->f_flags = nsp->f_flags;
729 	osp->f_syncwrites = MIN(nsp->f_syncwrites, LONG_MAX);
730 	osp->f_asyncwrites = MIN(nsp->f_asyncwrites, LONG_MAX);
731 	osp->f_syncreads = MIN(nsp->f_syncreads, LONG_MAX);
732 	osp->f_asyncreads = MIN(nsp->f_asyncreads, LONG_MAX);
733 	strlcpy(osp->f_fstypename, nsp->f_fstypename,
734 	    MIN(MFSNAMELEN, OMFSNAMELEN));
735 	strlcpy(osp->f_mntonname, nsp->f_mntonname,
736 	    MIN(MNAMELEN, OMNAMELEN));
737 	strlcpy(osp->f_mntfromname, nsp->f_mntfromname,
738 	    MIN(MNAMELEN, OMNAMELEN));
739 	osp->f_fsid = nsp->f_fsid;
740 }
741 #endif /* COMPAT_FREEBSD4 */
742 
743 #if defined(COMPAT_FREEBSD11)
744 /*
745  * Get old format filesystem statistics.
746  */
747 static void freebsd11_cvtstatfs(struct statfs *, struct freebsd11_statfs *);
748 
749 int
750 freebsd11_statfs(struct thread *td, struct freebsd11_statfs_args *uap)
751 {
752 	struct freebsd11_statfs osb;
753 	struct statfs *sfp;
754 	int error;
755 
756 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
757 	error = kern_statfs(td, uap->path, UIO_USERSPACE, sfp);
758 	if (error == 0) {
759 		freebsd11_cvtstatfs(sfp, &osb);
760 		error = copyout(&osb, uap->buf, sizeof(osb));
761 	}
762 	free(sfp, M_STATFS);
763 	return (error);
764 }
765 
766 /*
767  * Get filesystem statistics.
768  */
769 int
770 freebsd11_fstatfs(struct thread *td, struct freebsd11_fstatfs_args *uap)
771 {
772 	struct freebsd11_statfs osb;
773 	struct statfs *sfp;
774 	int error;
775 
776 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
777 	error = kern_fstatfs(td, uap->fd, sfp);
778 	if (error == 0) {
779 		freebsd11_cvtstatfs(sfp, &osb);
780 		error = copyout(&osb, uap->buf, sizeof(osb));
781 	}
782 	free(sfp, M_STATFS);
783 	return (error);
784 }
785 
786 /*
787  * Get statistics on all filesystems.
788  */
789 int
790 freebsd11_getfsstat(struct thread *td, struct freebsd11_getfsstat_args *uap)
791 {
792 	return (kern_freebsd11_getfsstat(td, uap->buf, uap->bufsize, uap->mode));
793 }
794 
795 int
796 kern_freebsd11_getfsstat(struct thread *td, struct freebsd11_statfs * ubuf,
797     long bufsize, int mode)
798 {
799 	struct freebsd11_statfs osb;
800 	struct statfs *buf, *sp;
801 	size_t count, size;
802 	int error;
803 
804 	if (bufsize < 0)
805 		return (EINVAL);
806 
807 	count = bufsize / sizeof(struct ostatfs);
808 	size = count * sizeof(struct statfs);
809 	error = kern_getfsstat(td, &buf, size, &count, UIO_SYSSPACE, mode);
810 	if (error == 0)
811 		td->td_retval[0] = count;
812 	if (size > 0) {
813 		sp = buf;
814 		while (count > 0 && error == 0) {
815 			freebsd11_cvtstatfs(sp, &osb);
816 			error = copyout(&osb, ubuf, sizeof(osb));
817 			sp++;
818 			ubuf++;
819 			count--;
820 		}
821 		free(buf, M_STATFS);
822 	}
823 	return (error);
824 }
825 
826 /*
827  * Implement fstatfs() for (NFS) file handles.
828  */
829 int
830 freebsd11_fhstatfs(struct thread *td, struct freebsd11_fhstatfs_args *uap)
831 {
832 	struct freebsd11_statfs osb;
833 	struct statfs *sfp;
834 	fhandle_t fh;
835 	int error;
836 
837 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
838 	if (error)
839 		return (error);
840 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
841 	error = kern_fhstatfs(td, fh, sfp);
842 	if (error == 0) {
843 		freebsd11_cvtstatfs(sfp, &osb);
844 		error = copyout(&osb, uap->buf, sizeof(osb));
845 	}
846 	free(sfp, M_STATFS);
847 	return (error);
848 }
849 
850 /*
851  * Convert a new format statfs structure to an old format statfs structure.
852  */
853 static void
854 freebsd11_cvtstatfs(struct statfs *nsp, struct freebsd11_statfs *osp)
855 {
856 
857 	bzero(osp, sizeof(*osp));
858 	osp->f_version = FREEBSD11_STATFS_VERSION;
859 	osp->f_type = nsp->f_type;
860 	osp->f_flags = nsp->f_flags;
861 	osp->f_bsize = nsp->f_bsize;
862 	osp->f_iosize = nsp->f_iosize;
863 	osp->f_blocks = nsp->f_blocks;
864 	osp->f_bfree = nsp->f_bfree;
865 	osp->f_bavail = nsp->f_bavail;
866 	osp->f_files = nsp->f_files;
867 	osp->f_ffree = nsp->f_ffree;
868 	osp->f_syncwrites = nsp->f_syncwrites;
869 	osp->f_asyncwrites = nsp->f_asyncwrites;
870 	osp->f_syncreads = nsp->f_syncreads;
871 	osp->f_asyncreads = nsp->f_asyncreads;
872 	osp->f_namemax = nsp->f_namemax;
873 	osp->f_owner = nsp->f_owner;
874 	osp->f_fsid = nsp->f_fsid;
875 	strlcpy(osp->f_fstypename, nsp->f_fstypename,
876 	    MIN(MFSNAMELEN, sizeof(osp->f_fstypename)));
877 	strlcpy(osp->f_mntonname, nsp->f_mntonname,
878 	    MIN(MNAMELEN, sizeof(osp->f_mntonname)));
879 	strlcpy(osp->f_mntfromname, nsp->f_mntfromname,
880 	    MIN(MNAMELEN, sizeof(osp->f_mntfromname)));
881 }
882 #endif /* COMPAT_FREEBSD11 */
883 
884 /*
885  * Change current working directory to a given file descriptor.
886  */
887 #ifndef _SYS_SYSPROTO_H_
888 struct fchdir_args {
889 	int	fd;
890 };
891 #endif
892 int
893 sys_fchdir(struct thread *td, struct fchdir_args *uap)
894 {
895 	struct vnode *vp, *tdp;
896 	struct mount *mp;
897 	struct file *fp;
898 	int error;
899 
900 	AUDIT_ARG_FD(uap->fd);
901 	error = getvnode_path(td, uap->fd, &cap_fchdir_rights,
902 	    &fp);
903 	if (error != 0)
904 		return (error);
905 	vp = fp->f_vnode;
906 	vrefact(vp);
907 	fdrop(fp, td);
908 	vn_lock(vp, LK_SHARED | LK_RETRY);
909 	AUDIT_ARG_VNODE1(vp);
910 	error = change_dir(vp, td);
911 	while (!error && (mp = vp->v_mountedhere) != NULL) {
912 		if (vfs_busy(mp, 0))
913 			continue;
914 		error = VFS_ROOT(mp, LK_SHARED, &tdp);
915 		vfs_unbusy(mp);
916 		if (error != 0)
917 			break;
918 		vput(vp);
919 		vp = tdp;
920 	}
921 	if (error != 0) {
922 		vput(vp);
923 		return (error);
924 	}
925 	VOP_UNLOCK(vp);
926 	pwd_chdir(td, vp);
927 	return (0);
928 }
929 
930 /*
931  * Change current working directory (``.'').
932  */
933 #ifndef _SYS_SYSPROTO_H_
934 struct chdir_args {
935 	char	*path;
936 };
937 #endif
938 int
939 sys_chdir(struct thread *td, struct chdir_args *uap)
940 {
941 
942 	return (kern_chdir(td, uap->path, UIO_USERSPACE));
943 }
944 
945 int
946 kern_chdir(struct thread *td, const char *path, enum uio_seg pathseg)
947 {
948 	struct nameidata nd;
949 	int error;
950 
951 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
952 	    pathseg, path);
953 	if ((error = namei(&nd)) != 0)
954 		return (error);
955 	if ((error = change_dir(nd.ni_vp, td)) != 0) {
956 		vput(nd.ni_vp);
957 		NDFREE_PNBUF(&nd);
958 		return (error);
959 	}
960 	VOP_UNLOCK(nd.ni_vp);
961 	NDFREE_PNBUF(&nd);
962 	pwd_chdir(td, nd.ni_vp);
963 	return (0);
964 }
965 
966 static int unprivileged_chroot = 0;
967 SYSCTL_INT(_security_bsd, OID_AUTO, unprivileged_chroot, CTLFLAG_RW,
968     &unprivileged_chroot, 0,
969     "Unprivileged processes can use chroot(2)");
970 
971 /*
972  * Takes locked vnode, unlocks it before returning.
973  */
974 static int
975 kern_chroot(struct thread *td, struct vnode *vp)
976 {
977 	struct proc *p;
978 	int error;
979 
980 	error = priv_check(td, PRIV_VFS_CHROOT);
981 	if (error != 0) {
982 		p = td->td_proc;
983 		PROC_LOCK(p);
984 		if (unprivileged_chroot == 0 ||
985 		    (p->p_flag2 & P2_NO_NEW_PRIVS) == 0) {
986 			PROC_UNLOCK(p);
987 			goto e_vunlock;
988 		}
989 		PROC_UNLOCK(p);
990 	}
991 
992 	error = change_dir(vp, td);
993 	if (error != 0)
994 		goto e_vunlock;
995 #ifdef MAC
996 	error = mac_vnode_check_chroot(td->td_ucred, vp);
997 	if (error != 0)
998 		goto e_vunlock;
999 #endif
1000 	VOP_UNLOCK(vp);
1001 	error = pwd_chroot(td, vp);
1002 	vrele(vp);
1003 	return (error);
1004 e_vunlock:
1005 	vput(vp);
1006 	return (error);
1007 }
1008 
1009 /*
1010  * Change notion of root (``/'') directory.
1011  */
1012 #ifndef _SYS_SYSPROTO_H_
1013 struct chroot_args {
1014 	char	*path;
1015 };
1016 #endif
1017 int
1018 sys_chroot(struct thread *td, struct chroot_args *uap)
1019 {
1020 	struct nameidata nd;
1021 	int error;
1022 
1023 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
1024 	    UIO_USERSPACE, uap->path);
1025 	error = namei(&nd);
1026 	if (error != 0)
1027 		return (error);
1028 	NDFREE_PNBUF(&nd);
1029 	error = kern_chroot(td, nd.ni_vp);
1030 	return (error);
1031 }
1032 
1033 /*
1034  * Change notion of root directory to a given file descriptor.
1035  */
1036 #ifndef _SYS_SYSPROTO_H_
1037 struct fchroot_args {
1038 	int	fd;
1039 };
1040 #endif
1041 int
1042 sys_fchroot(struct thread *td, struct fchroot_args *uap)
1043 {
1044 	struct vnode *vp;
1045 	struct file *fp;
1046 	int error;
1047 
1048 	error = getvnode_path(td, uap->fd, &cap_fchroot_rights, &fp);
1049 	if (error != 0)
1050 		return (error);
1051 	vp = fp->f_vnode;
1052 	vrefact(vp);
1053 	fdrop(fp, td);
1054 	vn_lock(vp, LK_SHARED | LK_RETRY);
1055 	error = kern_chroot(td, vp);
1056 	return (error);
1057 }
1058 
1059 /*
1060  * Common routine for chroot and chdir.  Callers must provide a locked vnode
1061  * instance.
1062  */
1063 int
1064 change_dir(struct vnode *vp, struct thread *td)
1065 {
1066 #ifdef MAC
1067 	int error;
1068 #endif
1069 
1070 	ASSERT_VOP_LOCKED(vp, "change_dir(): vp not locked");
1071 	if (vp->v_type != VDIR)
1072 		return (ENOTDIR);
1073 #ifdef MAC
1074 	error = mac_vnode_check_chdir(td->td_ucred, vp);
1075 	if (error != 0)
1076 		return (error);
1077 #endif
1078 	return (VOP_ACCESS(vp, VEXEC, td->td_ucred, td));
1079 }
1080 
1081 static __inline void
1082 flags_to_rights(int flags, cap_rights_t *rightsp)
1083 {
1084 	if (flags & O_EXEC) {
1085 		cap_rights_set_one(rightsp, CAP_FEXECVE);
1086 		if (flags & O_PATH)
1087 			return;
1088 	} else {
1089 		switch ((flags & O_ACCMODE)) {
1090 		case O_RDONLY:
1091 			cap_rights_set_one(rightsp, CAP_READ);
1092 			break;
1093 		case O_RDWR:
1094 			cap_rights_set_one(rightsp, CAP_READ);
1095 			/* FALLTHROUGH */
1096 		case O_WRONLY:
1097 			cap_rights_set_one(rightsp, CAP_WRITE);
1098 			if (!(flags & (O_APPEND | O_TRUNC)))
1099 				cap_rights_set_one(rightsp, CAP_SEEK);
1100 			break;
1101 		}
1102 	}
1103 
1104 	if (flags & O_CREAT)
1105 		cap_rights_set_one(rightsp, CAP_CREATE);
1106 
1107 	if (flags & O_TRUNC)
1108 		cap_rights_set_one(rightsp, CAP_FTRUNCATE);
1109 
1110 	if (flags & (O_SYNC | O_FSYNC))
1111 		cap_rights_set_one(rightsp, CAP_FSYNC);
1112 
1113 	if (flags & (O_EXLOCK | O_SHLOCK))
1114 		cap_rights_set_one(rightsp, CAP_FLOCK);
1115 }
1116 
1117 /*
1118  * Check permissions, allocate an open file structure, and call the device
1119  * open routine if any.
1120  */
1121 #ifndef _SYS_SYSPROTO_H_
1122 struct open_args {
1123 	char	*path;
1124 	int	flags;
1125 	int	mode;
1126 };
1127 #endif
1128 int
1129 sys_open(struct thread *td, struct open_args *uap)
1130 {
1131 
1132 	return (kern_openat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1133 	    uap->flags, uap->mode));
1134 }
1135 
1136 #ifndef _SYS_SYSPROTO_H_
1137 struct openat_args {
1138 	int	fd;
1139 	char	*path;
1140 	int	flag;
1141 	int	mode;
1142 };
1143 #endif
1144 int
1145 sys_openat(struct thread *td, struct openat_args *uap)
1146 {
1147 
1148 	AUDIT_ARG_FD(uap->fd);
1149 	return (kern_openat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
1150 	    uap->mode));
1151 }
1152 
1153 /*
1154  * If fpp != NULL, opened file is not installed into the file
1155  * descriptor table, instead it is returned in *fpp.  This is
1156  * incompatible with fdopen(), in which case we return EINVAL.
1157  */
1158 static int
1159 openatfp(struct thread *td, int dirfd, const char *path,
1160     enum uio_seg pathseg, int flags, int mode, struct file **fpp)
1161 {
1162 	struct proc *p;
1163 	struct filedesc *fdp;
1164 	struct pwddesc *pdp;
1165 	struct file *fp;
1166 	struct vnode *vp;
1167 	struct filecaps *fcaps;
1168 	struct nameidata nd;
1169 	cap_rights_t rights;
1170 	int cmode, error, indx;
1171 
1172 	indx = -1;
1173 	p = td->td_proc;
1174 	fdp = p->p_fd;
1175 	pdp = p->p_pd;
1176 
1177 	AUDIT_ARG_FFLAGS(flags);
1178 	AUDIT_ARG_MODE(mode);
1179 	cap_rights_init_one(&rights, CAP_LOOKUP);
1180 	flags_to_rights(flags, &rights);
1181 
1182 	/*
1183 	 * Only one of the O_EXEC, O_RDONLY, O_WRONLY and O_RDWR flags
1184 	 * may be specified.  On the other hand, for O_PATH any mode
1185 	 * except O_EXEC is ignored.
1186 	 */
1187 	if ((flags & O_PATH) != 0) {
1188 		flags &= ~(O_CREAT | O_ACCMODE);
1189 	} else if ((flags & O_EXEC) != 0) {
1190 		if (flags & O_ACCMODE)
1191 			return (EINVAL);
1192 	} else if ((flags & O_ACCMODE) == O_ACCMODE) {
1193 		return (EINVAL);
1194 	} else {
1195 		flags = FFLAGS(flags);
1196 	}
1197 
1198 	/*
1199 	 * Allocate a file structure. The descriptor to reference it
1200 	 * is allocated and used by finstall_refed() below.
1201 	 */
1202 	error = falloc_noinstall(td, &fp);
1203 	if (error != 0)
1204 		return (error);
1205 	/* Set the flags early so the finit in devfs can pick them up. */
1206 	fp->f_flag = flags & FMASK;
1207 	cmode = ((mode & ~pdp->pd_cmask) & ALLPERMS) & ~S_ISTXT;
1208 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1 | WANTIOCTLCAPS,
1209 	    pathseg, path, dirfd, &rights);
1210 	td->td_dupfd = -1;		/* XXX check for fdopen */
1211 	error = vn_open_cred(&nd, &flags, cmode, VN_OPEN_WANTIOCTLCAPS,
1212 	    td->td_ucred, fp);
1213 	if (error != 0) {
1214 		/*
1215 		 * If the vn_open replaced the method vector, something
1216 		 * wonderous happened deep below and we just pass it up
1217 		 * pretending we know what we do.
1218 		 */
1219 		if (error == ENXIO && fp->f_ops != &badfileops) {
1220 			MPASS((flags & O_PATH) == 0);
1221 			goto success;
1222 		}
1223 
1224 		/*
1225 		 * Handle special fdopen() case. bleh.
1226 		 *
1227 		 * Don't do this for relative (capability) lookups; we don't
1228 		 * understand exactly what would happen, and we don't think
1229 		 * that it ever should.
1230 		 */
1231 		if ((nd.ni_resflags & NIRES_STRICTREL) == 0 &&
1232 		    (error == ENODEV || error == ENXIO) &&
1233 		    td->td_dupfd >= 0) {
1234 			MPASS(fpp == NULL);
1235 			error = dupfdopen(td, fdp, td->td_dupfd, flags, error,
1236 			    &indx);
1237 			if (error == 0)
1238 				goto success;
1239 		}
1240 
1241 		goto bad;
1242 	}
1243 	td->td_dupfd = 0;
1244 	NDFREE_PNBUF(&nd);
1245 	vp = nd.ni_vp;
1246 
1247 	/*
1248 	 * Store the vnode, for any f_type. Typically, the vnode use
1249 	 * count is decremented by direct call to vn_closefile() for
1250 	 * files that switched type in the cdevsw fdopen() method.
1251 	 */
1252 	fp->f_vnode = vp;
1253 
1254 	/*
1255 	 * If the file wasn't claimed by devfs bind it to the normal
1256 	 * vnode operations here.
1257 	 */
1258 	if (fp->f_ops == &badfileops) {
1259 		KASSERT(vp->v_type != VFIFO || (flags & O_PATH) != 0,
1260 		    ("Unexpected fifo fp %p vp %p", fp, vp));
1261 		if ((flags & O_PATH) != 0) {
1262 			finit(fp, (flags & FMASK) | (fp->f_flag & FKQALLOWED),
1263 			    DTYPE_VNODE, NULL, &path_fileops);
1264 		} else {
1265 			finit_vnode(fp, flags, NULL, &vnops);
1266 		}
1267 	}
1268 
1269 	VOP_UNLOCK(vp);
1270 	if (flags & O_TRUNC) {
1271 		error = fo_truncate(fp, 0, td->td_ucred, td);
1272 		if (error != 0)
1273 			goto bad;
1274 	}
1275 success:
1276 	if (fpp != NULL) {
1277 		MPASS(error == 0);
1278 		NDFREE_IOCTLCAPS(&nd);
1279 		*fpp = fp;
1280 		return (0);
1281 	}
1282 
1283 	/*
1284 	 * If we haven't already installed the FD (for dupfdopen), do so now.
1285 	 */
1286 	if (indx == -1) {
1287 #ifdef CAPABILITIES
1288 		if ((nd.ni_resflags & NIRES_STRICTREL) != 0)
1289 			fcaps = &nd.ni_filecaps;
1290 		else
1291 #endif
1292 			fcaps = NULL;
1293 		error = finstall_refed(td, fp, &indx, flags, fcaps);
1294 		/* On success finstall_refed() consumes fcaps. */
1295 		if (error != 0) {
1296 			goto bad;
1297 		}
1298 	} else {
1299 		NDFREE_IOCTLCAPS(&nd);
1300 		falloc_abort(td, fp);
1301 	}
1302 
1303 	td->td_retval[0] = indx;
1304 	return (0);
1305 bad:
1306 	KASSERT(indx == -1, ("indx=%d, should be -1", indx));
1307 	NDFREE_IOCTLCAPS(&nd);
1308 	falloc_abort(td, fp);
1309 	return (error);
1310 }
1311 
1312 int
1313 kern_openat(struct thread *td, int dirfd, const char *path,
1314     enum uio_seg pathseg, int flags, int mode)
1315 {
1316 	return (openatfp(td, dirfd, path, pathseg, flags, mode, NULL));
1317 }
1318 
1319 int
1320 kern_openatfp(struct thread *td, int dirfd, const char *path,
1321     enum uio_seg pathseg, int flags, int mode, struct file **fpp)
1322 {
1323 	int error, old_dupfd;
1324 
1325 	old_dupfd = td->td_dupfd;
1326 	td->td_dupfd = -1;
1327 	error = openatfp(td, dirfd, path, pathseg, flags, mode, fpp);
1328 	td->td_dupfd = old_dupfd;
1329 	return (error);
1330 }
1331 
1332 #ifdef COMPAT_43
1333 /*
1334  * Create a file.
1335  */
1336 #ifndef _SYS_SYSPROTO_H_
1337 struct ocreat_args {
1338 	char	*path;
1339 	int	mode;
1340 };
1341 #endif
1342 int
1343 ocreat(struct thread *td, struct ocreat_args *uap)
1344 {
1345 
1346 	return (kern_openat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1347 	    O_WRONLY | O_CREAT | O_TRUNC, uap->mode));
1348 }
1349 #endif /* COMPAT_43 */
1350 
1351 /*
1352  * Create a special file.
1353  */
1354 #ifndef _SYS_SYSPROTO_H_
1355 struct mknodat_args {
1356 	int	fd;
1357 	char	*path;
1358 	mode_t	mode;
1359 	dev_t	dev;
1360 };
1361 #endif
1362 int
1363 sys_mknodat(struct thread *td, struct mknodat_args *uap)
1364 {
1365 
1366 	return (kern_mknodat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode,
1367 	    uap->dev));
1368 }
1369 
1370 #if defined(COMPAT_FREEBSD11)
1371 int
1372 freebsd11_mknod(struct thread *td,
1373     struct freebsd11_mknod_args *uap)
1374 {
1375 
1376 	return (kern_mknodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1377 	    uap->mode, uap->dev));
1378 }
1379 
1380 int
1381 freebsd11_mknodat(struct thread *td,
1382     struct freebsd11_mknodat_args *uap)
1383 {
1384 
1385 	return (kern_mknodat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode,
1386 	    uap->dev));
1387 }
1388 #endif /* COMPAT_FREEBSD11 */
1389 
1390 int
1391 kern_mknodat(struct thread *td, int fd, const char *path, enum uio_seg pathseg,
1392     int mode, dev_t dev)
1393 {
1394 	struct vnode *vp;
1395 	struct mount *mp;
1396 	struct vattr vattr;
1397 	struct nameidata nd;
1398 	int error, whiteout = 0;
1399 
1400 	AUDIT_ARG_MODE(mode);
1401 	AUDIT_ARG_DEV(dev);
1402 	switch (mode & S_IFMT) {
1403 	case S_IFCHR:
1404 	case S_IFBLK:
1405 		error = priv_check(td, PRIV_VFS_MKNOD_DEV);
1406 		if (error == 0 && dev == VNOVAL)
1407 			error = EINVAL;
1408 		break;
1409 	case S_IFWHT:
1410 		error = priv_check(td, PRIV_VFS_MKNOD_WHT);
1411 		break;
1412 	case S_IFIFO:
1413 		if (dev == 0)
1414 			return (kern_mkfifoat(td, fd, path, pathseg, mode));
1415 		/* FALLTHROUGH */
1416 	default:
1417 		error = EINVAL;
1418 		break;
1419 	}
1420 	if (error != 0)
1421 		return (error);
1422 	NDPREINIT(&nd);
1423 restart:
1424 	bwillwrite();
1425 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | AUDITVNODE1 | NOCACHE,
1426 	    pathseg, path, fd, &cap_mknodat_rights);
1427 	if ((error = namei(&nd)) != 0)
1428 		return (error);
1429 	vp = nd.ni_vp;
1430 	if (vp != NULL) {
1431 		NDFREE_PNBUF(&nd);
1432 		if (vp == nd.ni_dvp)
1433 			vrele(nd.ni_dvp);
1434 		else
1435 			vput(nd.ni_dvp);
1436 		vrele(vp);
1437 		return (EEXIST);
1438 	} else {
1439 		VATTR_NULL(&vattr);
1440 		vattr.va_mode = (mode & ALLPERMS) &
1441 		    ~td->td_proc->p_pd->pd_cmask;
1442 		vattr.va_rdev = dev;
1443 		whiteout = 0;
1444 
1445 		switch (mode & S_IFMT) {
1446 		case S_IFCHR:
1447 			vattr.va_type = VCHR;
1448 			break;
1449 		case S_IFBLK:
1450 			vattr.va_type = VBLK;
1451 			break;
1452 		case S_IFWHT:
1453 			whiteout = 1;
1454 			break;
1455 		default:
1456 			panic("kern_mknod: invalid mode");
1457 		}
1458 	}
1459 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1460 		NDFREE_PNBUF(&nd);
1461 		vput(nd.ni_dvp);
1462 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | V_PCATCH)) != 0)
1463 			return (error);
1464 		goto restart;
1465 	}
1466 #ifdef MAC
1467 	if (error == 0 && !whiteout)
1468 		error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp,
1469 		    &nd.ni_cnd, &vattr);
1470 #endif
1471 	if (error == 0) {
1472 		if (whiteout)
1473 			error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE);
1474 		else {
1475 			error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp,
1476 						&nd.ni_cnd, &vattr);
1477 		}
1478 	}
1479 	VOP_VPUT_PAIR(nd.ni_dvp, error == 0 && !whiteout ? &nd.ni_vp : NULL,
1480 	    true);
1481 	vn_finished_write(mp);
1482 	NDFREE_PNBUF(&nd);
1483 	if (error == ERELOOKUP)
1484 		goto restart;
1485 	return (error);
1486 }
1487 
1488 /*
1489  * Create a named pipe.
1490  */
1491 #ifndef _SYS_SYSPROTO_H_
1492 struct mkfifo_args {
1493 	char	*path;
1494 	int	mode;
1495 };
1496 #endif
1497 int
1498 sys_mkfifo(struct thread *td, struct mkfifo_args *uap)
1499 {
1500 
1501 	return (kern_mkfifoat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1502 	    uap->mode));
1503 }
1504 
1505 #ifndef _SYS_SYSPROTO_H_
1506 struct mkfifoat_args {
1507 	int	fd;
1508 	char	*path;
1509 	mode_t	mode;
1510 };
1511 #endif
1512 int
1513 sys_mkfifoat(struct thread *td, struct mkfifoat_args *uap)
1514 {
1515 
1516 	return (kern_mkfifoat(td, uap->fd, uap->path, UIO_USERSPACE,
1517 	    uap->mode));
1518 }
1519 
1520 int
1521 kern_mkfifoat(struct thread *td, int fd, const char *path,
1522     enum uio_seg pathseg, int mode)
1523 {
1524 	struct mount *mp;
1525 	struct vattr vattr;
1526 	struct nameidata nd;
1527 	int error;
1528 
1529 	AUDIT_ARG_MODE(mode);
1530 	NDPREINIT(&nd);
1531 restart:
1532 	bwillwrite();
1533 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | AUDITVNODE1 | NOCACHE,
1534 	    pathseg, path, fd, &cap_mkfifoat_rights);
1535 	if ((error = namei(&nd)) != 0)
1536 		return (error);
1537 	if (nd.ni_vp != NULL) {
1538 		NDFREE_PNBUF(&nd);
1539 		if (nd.ni_vp == nd.ni_dvp)
1540 			vrele(nd.ni_dvp);
1541 		else
1542 			vput(nd.ni_dvp);
1543 		vrele(nd.ni_vp);
1544 		return (EEXIST);
1545 	}
1546 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1547 		NDFREE_PNBUF(&nd);
1548 		vput(nd.ni_dvp);
1549 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | V_PCATCH)) != 0)
1550 			return (error);
1551 		goto restart;
1552 	}
1553 	VATTR_NULL(&vattr);
1554 	vattr.va_type = VFIFO;
1555 	vattr.va_mode = (mode & ALLPERMS) & ~td->td_proc->p_pd->pd_cmask;
1556 #ifdef MAC
1557 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1558 	    &vattr);
1559 	if (error != 0)
1560 		goto out;
1561 #endif
1562 	error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
1563 #ifdef MAC
1564 out:
1565 #endif
1566 	VOP_VPUT_PAIR(nd.ni_dvp, error == 0 ? &nd.ni_vp : NULL, true);
1567 	vn_finished_write(mp);
1568 	NDFREE_PNBUF(&nd);
1569 	if (error == ERELOOKUP)
1570 		goto restart;
1571 	return (error);
1572 }
1573 
1574 /*
1575  * Make a hard file link.
1576  */
1577 #ifndef _SYS_SYSPROTO_H_
1578 struct link_args {
1579 	char	*path;
1580 	char	*link;
1581 };
1582 #endif
1583 int
1584 sys_link(struct thread *td, struct link_args *uap)
1585 {
1586 
1587 	return (kern_linkat(td, AT_FDCWD, AT_FDCWD, uap->path, uap->link,
1588 	    UIO_USERSPACE, AT_SYMLINK_FOLLOW));
1589 }
1590 
1591 #ifndef _SYS_SYSPROTO_H_
1592 struct linkat_args {
1593 	int	fd1;
1594 	char	*path1;
1595 	int	fd2;
1596 	char	*path2;
1597 	int	flag;
1598 };
1599 #endif
1600 int
1601 sys_linkat(struct thread *td, struct linkat_args *uap)
1602 {
1603 
1604 	return (kern_linkat(td, uap->fd1, uap->fd2, uap->path1, uap->path2,
1605 	    UIO_USERSPACE, uap->flag));
1606 }
1607 
1608 int hardlink_check_uid = 0;
1609 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_uid, CTLFLAG_RW,
1610     &hardlink_check_uid, 0,
1611     "Unprivileged processes cannot create hard links to files owned by other "
1612     "users");
1613 static int hardlink_check_gid = 0;
1614 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_gid, CTLFLAG_RW,
1615     &hardlink_check_gid, 0,
1616     "Unprivileged processes cannot create hard links to files owned by other "
1617     "groups");
1618 
1619 static int
1620 can_hardlink(struct vnode *vp, struct ucred *cred)
1621 {
1622 	struct vattr va;
1623 	int error;
1624 
1625 	if (!hardlink_check_uid && !hardlink_check_gid)
1626 		return (0);
1627 
1628 	error = VOP_GETATTR(vp, &va, cred);
1629 	if (error != 0)
1630 		return (error);
1631 
1632 	if (hardlink_check_uid && cred->cr_uid != va.va_uid) {
1633 		error = priv_check_cred(cred, PRIV_VFS_LINK);
1634 		if (error != 0)
1635 			return (error);
1636 	}
1637 
1638 	if (hardlink_check_gid && !groupmember(va.va_gid, cred)) {
1639 		error = priv_check_cred(cred, PRIV_VFS_LINK);
1640 		if (error != 0)
1641 			return (error);
1642 	}
1643 
1644 	return (0);
1645 }
1646 
1647 int
1648 kern_linkat(struct thread *td, int fd1, int fd2, const char *path1,
1649     const char *path2, enum uio_seg segflag, int flag)
1650 {
1651 	struct nameidata nd;
1652 	int error;
1653 
1654 	if ((flag & ~(AT_SYMLINK_FOLLOW | AT_RESOLVE_BENEATH |
1655 	    AT_EMPTY_PATH)) != 0)
1656 		return (EINVAL);
1657 
1658 	NDPREINIT(&nd);
1659 	do {
1660 		bwillwrite();
1661 		NDINIT_ATRIGHTS(&nd, LOOKUP, AUDITVNODE1 | at2cnpflags(flag,
1662 		    AT_SYMLINK_FOLLOW | AT_RESOLVE_BENEATH | AT_EMPTY_PATH),
1663 		    segflag, path1, fd1, &cap_linkat_source_rights);
1664 		if ((error = namei(&nd)) != 0)
1665 			return (error);
1666 		NDFREE_PNBUF(&nd);
1667 		if ((nd.ni_resflags & NIRES_EMPTYPATH) != 0) {
1668 			error = priv_check(td, PRIV_VFS_FHOPEN);
1669 			if (error != 0) {
1670 				vrele(nd.ni_vp);
1671 				return (error);
1672 			}
1673 		}
1674 		error = kern_linkat_vp(td, nd.ni_vp, fd2, path2, segflag);
1675 	} while (error ==  EAGAIN || error == ERELOOKUP);
1676 	return (error);
1677 }
1678 
1679 static int
1680 kern_linkat_vp(struct thread *td, struct vnode *vp, int fd, const char *path,
1681     enum uio_seg segflag)
1682 {
1683 	struct nameidata nd;
1684 	struct mount *mp;
1685 	int error;
1686 
1687 	if (vp->v_type == VDIR) {
1688 		vrele(vp);
1689 		return (EPERM);		/* POSIX */
1690 	}
1691 	NDINIT_ATRIGHTS(&nd, CREATE,
1692 	    LOCKPARENT | AUDITVNODE2 | NOCACHE, segflag, path, fd,
1693 	    &cap_linkat_target_rights);
1694 	if ((error = namei(&nd)) == 0) {
1695 		if (nd.ni_vp != NULL) {
1696 			NDFREE_PNBUF(&nd);
1697 			if (nd.ni_dvp == nd.ni_vp)
1698 				vrele(nd.ni_dvp);
1699 			else
1700 				vput(nd.ni_dvp);
1701 			vrele(nd.ni_vp);
1702 			vrele(vp);
1703 			return (EEXIST);
1704 		} else if (nd.ni_dvp->v_mount != vp->v_mount) {
1705 			/*
1706 			 * Cross-device link.  No need to recheck
1707 			 * vp->v_type, since it cannot change, except
1708 			 * to VBAD.
1709 			 */
1710 			NDFREE_PNBUF(&nd);
1711 			vput(nd.ni_dvp);
1712 			vrele(vp);
1713 			return (EXDEV);
1714 		} else if (vn_lock(vp, LK_EXCLUSIVE) == 0) {
1715 			error = can_hardlink(vp, td->td_ucred);
1716 #ifdef MAC
1717 			if (error == 0)
1718 				error = mac_vnode_check_link(td->td_ucred,
1719 				    nd.ni_dvp, vp, &nd.ni_cnd);
1720 #endif
1721 			if (error != 0) {
1722 				vput(vp);
1723 				vput(nd.ni_dvp);
1724 				NDFREE_PNBUF(&nd);
1725 				return (error);
1726 			}
1727 			error = vn_start_write(vp, &mp, V_NOWAIT);
1728 			if (error != 0) {
1729 				vput(vp);
1730 				vput(nd.ni_dvp);
1731 				NDFREE_PNBUF(&nd);
1732 				error = vn_start_write(NULL, &mp,
1733 				    V_XSLEEP | V_PCATCH);
1734 				if (error != 0)
1735 					return (error);
1736 				return (EAGAIN);
1737 			}
1738 			error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
1739 			VOP_VPUT_PAIR(nd.ni_dvp, &vp, true);
1740 			vn_finished_write(mp);
1741 			NDFREE_PNBUF(&nd);
1742 			vp = NULL;
1743 		} else {
1744 			vput(nd.ni_dvp);
1745 			NDFREE_PNBUF(&nd);
1746 			vrele(vp);
1747 			return (EAGAIN);
1748 		}
1749 	}
1750 	if (vp != NULL)
1751 		vrele(vp);
1752 	return (error);
1753 }
1754 
1755 /*
1756  * Make a symbolic link.
1757  */
1758 #ifndef _SYS_SYSPROTO_H_
1759 struct symlink_args {
1760 	char	*path;
1761 	char	*link;
1762 };
1763 #endif
1764 int
1765 sys_symlink(struct thread *td, struct symlink_args *uap)
1766 {
1767 
1768 	return (kern_symlinkat(td, uap->path, AT_FDCWD, uap->link,
1769 	    UIO_USERSPACE));
1770 }
1771 
1772 #ifndef _SYS_SYSPROTO_H_
1773 struct symlinkat_args {
1774 	char	*path;
1775 	int	fd;
1776 	char	*path2;
1777 };
1778 #endif
1779 int
1780 sys_symlinkat(struct thread *td, struct symlinkat_args *uap)
1781 {
1782 
1783 	return (kern_symlinkat(td, uap->path1, uap->fd, uap->path2,
1784 	    UIO_USERSPACE));
1785 }
1786 
1787 int
1788 kern_symlinkat(struct thread *td, const char *path1, int fd, const char *path2,
1789     enum uio_seg segflg)
1790 {
1791 	struct mount *mp;
1792 	struct vattr vattr;
1793 	const char *syspath;
1794 	char *tmppath;
1795 	struct nameidata nd;
1796 	int error;
1797 
1798 	if (segflg == UIO_SYSSPACE) {
1799 		syspath = path1;
1800 	} else {
1801 		tmppath = uma_zalloc(namei_zone, M_WAITOK);
1802 		if ((error = copyinstr(path1, tmppath, MAXPATHLEN, NULL)) != 0)
1803 			goto out;
1804 		syspath = tmppath;
1805 	}
1806 	AUDIT_ARG_TEXT(syspath);
1807 	NDPREINIT(&nd);
1808 restart:
1809 	bwillwrite();
1810 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | AUDITVNODE1 | NOCACHE, segflg,
1811 	    path2, fd, &cap_symlinkat_rights);
1812 	if ((error = namei(&nd)) != 0)
1813 		goto out;
1814 	if (nd.ni_vp) {
1815 		NDFREE_PNBUF(&nd);
1816 		if (nd.ni_vp == nd.ni_dvp)
1817 			vrele(nd.ni_dvp);
1818 		else
1819 			vput(nd.ni_dvp);
1820 		vrele(nd.ni_vp);
1821 		nd.ni_vp = NULL;
1822 		error = EEXIST;
1823 		goto out;
1824 	}
1825 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1826 		NDFREE_PNBUF(&nd);
1827 		vput(nd.ni_dvp);
1828 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | V_PCATCH)) != 0)
1829 			goto out;
1830 		goto restart;
1831 	}
1832 	VATTR_NULL(&vattr);
1833 	vattr.va_mode = ACCESSPERMS &~ td->td_proc->p_pd->pd_cmask;
1834 #ifdef MAC
1835 	vattr.va_type = VLNK;
1836 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1837 	    &vattr);
1838 	if (error != 0)
1839 		goto out2;
1840 #endif
1841 	error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, syspath);
1842 #ifdef MAC
1843 out2:
1844 #endif
1845 	VOP_VPUT_PAIR(nd.ni_dvp, error == 0 ? &nd.ni_vp : NULL, true);
1846 	vn_finished_write(mp);
1847 	NDFREE_PNBUF(&nd);
1848 	if (error == ERELOOKUP)
1849 		goto restart;
1850 out:
1851 	if (segflg != UIO_SYSSPACE)
1852 		uma_zfree(namei_zone, tmppath);
1853 	return (error);
1854 }
1855 
1856 /*
1857  * Delete a whiteout from the filesystem.
1858  */
1859 #ifndef _SYS_SYSPROTO_H_
1860 struct undelete_args {
1861 	char *path;
1862 };
1863 #endif
1864 int
1865 sys_undelete(struct thread *td, struct undelete_args *uap)
1866 {
1867 	struct mount *mp;
1868 	struct nameidata nd;
1869 	int error;
1870 
1871 	NDPREINIT(&nd);
1872 restart:
1873 	bwillwrite();
1874 	NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | AUDITVNODE1,
1875 	    UIO_USERSPACE, uap->path);
1876 	error = namei(&nd);
1877 	if (error != 0)
1878 		return (error);
1879 
1880 	if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) {
1881 		NDFREE_PNBUF(&nd);
1882 		if (nd.ni_vp == nd.ni_dvp)
1883 			vrele(nd.ni_dvp);
1884 		else
1885 			vput(nd.ni_dvp);
1886 		if (nd.ni_vp)
1887 			vrele(nd.ni_vp);
1888 		return (EEXIST);
1889 	}
1890 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1891 		NDFREE_PNBUF(&nd);
1892 		vput(nd.ni_dvp);
1893 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | V_PCATCH)) != 0)
1894 			return (error);
1895 		goto restart;
1896 	}
1897 	error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE);
1898 	NDFREE_PNBUF(&nd);
1899 	vput(nd.ni_dvp);
1900 	vn_finished_write(mp);
1901 	if (error == ERELOOKUP)
1902 		goto restart;
1903 	return (error);
1904 }
1905 
1906 /*
1907  * Delete a name from the filesystem.
1908  */
1909 #ifndef _SYS_SYSPROTO_H_
1910 struct unlink_args {
1911 	char	*path;
1912 };
1913 #endif
1914 int
1915 sys_unlink(struct thread *td, struct unlink_args *uap)
1916 {
1917 
1918 	return (kern_funlinkat(td, AT_FDCWD, uap->path, FD_NONE, UIO_USERSPACE,
1919 	    0, 0));
1920 }
1921 
1922 static int
1923 kern_funlinkat_ex(struct thread *td, int dfd, const char *path, int fd,
1924     int flag, enum uio_seg pathseg, ino_t oldinum)
1925 {
1926 
1927 	if ((flag & ~(AT_REMOVEDIR | AT_RESOLVE_BENEATH)) != 0)
1928 		return (EINVAL);
1929 
1930 	if ((flag & AT_REMOVEDIR) != 0)
1931 		return (kern_frmdirat(td, dfd, path, fd, UIO_USERSPACE, 0));
1932 
1933 	return (kern_funlinkat(td, dfd, path, fd, UIO_USERSPACE, 0, 0));
1934 }
1935 
1936 #ifndef _SYS_SYSPROTO_H_
1937 struct unlinkat_args {
1938 	int	fd;
1939 	char	*path;
1940 	int	flag;
1941 };
1942 #endif
1943 int
1944 sys_unlinkat(struct thread *td, struct unlinkat_args *uap)
1945 {
1946 
1947 	return (kern_funlinkat_ex(td, uap->fd, uap->path, FD_NONE, uap->flag,
1948 	    UIO_USERSPACE, 0));
1949 }
1950 
1951 #ifndef _SYS_SYSPROTO_H_
1952 struct funlinkat_args {
1953 	int		dfd;
1954 	const char	*path;
1955 	int		fd;
1956 	int		flag;
1957 };
1958 #endif
1959 int
1960 sys_funlinkat(struct thread *td, struct funlinkat_args *uap)
1961 {
1962 
1963 	return (kern_funlinkat_ex(td, uap->dfd, uap->path, uap->fd, uap->flag,
1964 	    UIO_USERSPACE, 0));
1965 }
1966 
1967 int
1968 kern_funlinkat(struct thread *td, int dfd, const char *path, int fd,
1969     enum uio_seg pathseg, int flag, ino_t oldinum)
1970 {
1971 	struct mount *mp;
1972 	struct file *fp;
1973 	struct vnode *vp;
1974 	struct nameidata nd;
1975 	struct stat sb;
1976 	int error;
1977 
1978 	fp = NULL;
1979 	if (fd != FD_NONE) {
1980 		error = getvnode_path(td, fd, &cap_no_rights, &fp);
1981 		if (error != 0)
1982 			return (error);
1983 	}
1984 
1985 	NDPREINIT(&nd);
1986 restart:
1987 	bwillwrite();
1988 	NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1 |
1989 	    at2cnpflags(flag, AT_RESOLVE_BENEATH),
1990 	    pathseg, path, dfd, &cap_unlinkat_rights);
1991 	if ((error = namei(&nd)) != 0) {
1992 		if (error == EINVAL)
1993 			error = EPERM;
1994 		goto fdout;
1995 	}
1996 	vp = nd.ni_vp;
1997 	if (vp->v_type == VDIR && oldinum == 0) {
1998 		error = EPERM;		/* POSIX */
1999 	} else if (oldinum != 0 &&
2000 	    ((error = VOP_STAT(vp, &sb, td->td_ucred, NOCRED)) == 0) &&
2001 	    sb.st_ino != oldinum) {
2002 		error = EIDRM;	/* Identifier removed */
2003 	} else if (fp != NULL && fp->f_vnode != vp) {
2004 		if (VN_IS_DOOMED(fp->f_vnode))
2005 			error = EBADF;
2006 		else
2007 			error = EDEADLK;
2008 	} else {
2009 		/*
2010 		 * The root of a mounted filesystem cannot be deleted.
2011 		 *
2012 		 * XXX: can this only be a VDIR case?
2013 		 */
2014 		if (vp->v_vflag & VV_ROOT)
2015 			error = EBUSY;
2016 	}
2017 	if (error == 0) {
2018 		if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
2019 			NDFREE_PNBUF(&nd);
2020 			vput(nd.ni_dvp);
2021 			if (vp == nd.ni_dvp)
2022 				vrele(vp);
2023 			else
2024 				vput(vp);
2025 			if ((error = vn_start_write(NULL, &mp,
2026 			    V_XSLEEP | V_PCATCH)) != 0) {
2027 				goto fdout;
2028 			}
2029 			goto restart;
2030 		}
2031 #ifdef MAC
2032 		error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
2033 		    &nd.ni_cnd);
2034 		if (error != 0)
2035 			goto out;
2036 #endif
2037 		vfs_notify_upper(vp, VFS_NOTIFY_UPPER_UNLINK);
2038 		error = VOP_REMOVE(nd.ni_dvp, vp, &nd.ni_cnd);
2039 #ifdef MAC
2040 out:
2041 #endif
2042 		vn_finished_write(mp);
2043 	}
2044 	NDFREE_PNBUF(&nd);
2045 	vput(nd.ni_dvp);
2046 	if (vp == nd.ni_dvp)
2047 		vrele(vp);
2048 	else
2049 		vput(vp);
2050 	if (error == ERELOOKUP)
2051 		goto restart;
2052 fdout:
2053 	if (fp != NULL)
2054 		fdrop(fp, td);
2055 	return (error);
2056 }
2057 
2058 /*
2059  * Reposition read/write file offset.
2060  */
2061 #ifndef _SYS_SYSPROTO_H_
2062 struct lseek_args {
2063 	int	fd;
2064 	int	pad;
2065 	off_t	offset;
2066 	int	whence;
2067 };
2068 #endif
2069 int
2070 sys_lseek(struct thread *td, struct lseek_args *uap)
2071 {
2072 
2073 	return (kern_lseek(td, uap->fd, uap->offset, uap->whence));
2074 }
2075 
2076 int
2077 kern_lseek(struct thread *td, int fd, off_t offset, int whence)
2078 {
2079 	struct file *fp;
2080 	int error;
2081 
2082 	AUDIT_ARG_FD(fd);
2083 	error = fget(td, fd, &cap_seek_rights, &fp);
2084 	if (error != 0)
2085 		return (error);
2086 	error = (fp->f_ops->fo_flags & DFLAG_SEEKABLE) != 0 ?
2087 	    fo_seek(fp, offset, whence, td) : ESPIPE;
2088 	fdrop(fp, td);
2089 	return (error);
2090 }
2091 
2092 #if defined(COMPAT_43)
2093 /*
2094  * Reposition read/write file offset.
2095  */
2096 #ifndef _SYS_SYSPROTO_H_
2097 struct olseek_args {
2098 	int	fd;
2099 	long	offset;
2100 	int	whence;
2101 };
2102 #endif
2103 int
2104 olseek(struct thread *td, struct olseek_args *uap)
2105 {
2106 
2107 	return (kern_lseek(td, uap->fd, uap->offset, uap->whence));
2108 }
2109 #endif /* COMPAT_43 */
2110 
2111 #if defined(COMPAT_FREEBSD6)
2112 /* Version with the 'pad' argument */
2113 int
2114 freebsd6_lseek(struct thread *td, struct freebsd6_lseek_args *uap)
2115 {
2116 
2117 	return (kern_lseek(td, uap->fd, uap->offset, uap->whence));
2118 }
2119 #endif
2120 
2121 /*
2122  * Check access permissions using passed credentials.
2123  */
2124 static int
2125 vn_access(struct vnode *vp, int user_flags, struct ucred *cred,
2126      struct thread *td)
2127 {
2128 	accmode_t accmode;
2129 	int error;
2130 
2131 	/* Flags == 0 means only check for existence. */
2132 	if (user_flags == 0)
2133 		return (0);
2134 
2135 	accmode = 0;
2136 	if (user_flags & R_OK)
2137 		accmode |= VREAD;
2138 	if (user_flags & W_OK)
2139 		accmode |= VWRITE;
2140 	if (user_flags & X_OK)
2141 		accmode |= VEXEC;
2142 #ifdef MAC
2143 	error = mac_vnode_check_access(cred, vp, accmode);
2144 	if (error != 0)
2145 		return (error);
2146 #endif
2147 	if ((accmode & VWRITE) == 0 || (error = vn_writechk(vp)) == 0)
2148 		error = VOP_ACCESS(vp, accmode, cred, td);
2149 	return (error);
2150 }
2151 
2152 /*
2153  * Check access permissions using "real" credentials.
2154  */
2155 #ifndef _SYS_SYSPROTO_H_
2156 struct access_args {
2157 	char	*path;
2158 	int	amode;
2159 };
2160 #endif
2161 int
2162 sys_access(struct thread *td, struct access_args *uap)
2163 {
2164 
2165 	return (kern_accessat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2166 	    0, uap->amode));
2167 }
2168 
2169 #ifndef _SYS_SYSPROTO_H_
2170 struct faccessat_args {
2171 	int	dirfd;
2172 	char	*path;
2173 	int	amode;
2174 	int	flag;
2175 }
2176 #endif
2177 int
2178 sys_faccessat(struct thread *td, struct faccessat_args *uap)
2179 {
2180 
2181 	return (kern_accessat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
2182 	    uap->amode));
2183 }
2184 
2185 int
2186 kern_accessat(struct thread *td, int fd, const char *path,
2187     enum uio_seg pathseg, int flag, int amode)
2188 {
2189 	struct ucred *cred, *usecred;
2190 	struct vnode *vp;
2191 	struct nameidata nd;
2192 	int error;
2193 
2194 	if ((flag & ~(AT_EACCESS | AT_RESOLVE_BENEATH | AT_EMPTY_PATH |
2195 	    AT_SYMLINK_NOFOLLOW)) != 0)
2196 		return (EINVAL);
2197 	if (amode != F_OK && (amode & ~(R_OK | W_OK | X_OK)) != 0)
2198 		return (EINVAL);
2199 
2200 	/*
2201 	 * Create and modify a temporary credential instead of one that
2202 	 * is potentially shared (if we need one).
2203 	 */
2204 	cred = td->td_ucred;
2205 	if ((flag & AT_EACCESS) == 0 &&
2206 	    ((cred->cr_uid != cred->cr_ruid ||
2207 	    cred->cr_rgid != cred->cr_groups[0]))) {
2208 		usecred = crdup(cred);
2209 		usecred->cr_uid = cred->cr_ruid;
2210 		usecred->cr_groups[0] = cred->cr_rgid;
2211 		td->td_ucred = usecred;
2212 	} else
2213 		usecred = cred;
2214 	AUDIT_ARG_VALUE(amode);
2215 	NDINIT_ATRIGHTS(&nd, LOOKUP, LOCKSHARED | LOCKLEAF |
2216 	    AUDITVNODE1 | at2cnpflags(flag, AT_RESOLVE_BENEATH | AT_SYMLINK_NOFOLLOW |
2217 	    AT_EMPTY_PATH), pathseg, path, fd, &cap_fstat_rights);
2218 	if ((error = namei(&nd)) != 0)
2219 		goto out;
2220 	vp = nd.ni_vp;
2221 
2222 	error = vn_access(vp, amode, usecred, td);
2223 	NDFREE_PNBUF(&nd);
2224 	vput(vp);
2225 out:
2226 	if (usecred != cred) {
2227 		td->td_ucred = cred;
2228 		crfree(usecred);
2229 	}
2230 	return (error);
2231 }
2232 
2233 /*
2234  * Check access permissions using "effective" credentials.
2235  */
2236 #ifndef _SYS_SYSPROTO_H_
2237 struct eaccess_args {
2238 	char	*path;
2239 	int	amode;
2240 };
2241 #endif
2242 int
2243 sys_eaccess(struct thread *td, struct eaccess_args *uap)
2244 {
2245 
2246 	return (kern_accessat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2247 	    AT_EACCESS, uap->amode));
2248 }
2249 
2250 #if defined(COMPAT_43)
2251 /*
2252  * Get file status; this version follows links.
2253  */
2254 #ifndef _SYS_SYSPROTO_H_
2255 struct ostat_args {
2256 	char	*path;
2257 	struct ostat *ub;
2258 };
2259 #endif
2260 int
2261 ostat(struct thread *td, struct ostat_args *uap)
2262 {
2263 	struct stat sb;
2264 	struct ostat osb;
2265 	int error;
2266 
2267 	error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE, &sb);
2268 	if (error != 0)
2269 		return (error);
2270 	cvtstat(&sb, &osb);
2271 	return (copyout(&osb, uap->ub, sizeof (osb)));
2272 }
2273 
2274 /*
2275  * Get file status; this version does not follow links.
2276  */
2277 #ifndef _SYS_SYSPROTO_H_
2278 struct olstat_args {
2279 	char	*path;
2280 	struct ostat *ub;
2281 };
2282 #endif
2283 int
2284 olstat(struct thread *td, struct olstat_args *uap)
2285 {
2286 	struct stat sb;
2287 	struct ostat osb;
2288 	int error;
2289 
2290 	error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2291 	    UIO_USERSPACE, &sb);
2292 	if (error != 0)
2293 		return (error);
2294 	cvtstat(&sb, &osb);
2295 	return (copyout(&osb, uap->ub, sizeof (osb)));
2296 }
2297 
2298 /*
2299  * Convert from an old to a new stat structure.
2300  * XXX: many values are blindly truncated.
2301  */
2302 void
2303 cvtstat(struct stat *st, struct ostat *ost)
2304 {
2305 
2306 	bzero(ost, sizeof(*ost));
2307 	ost->st_dev = st->st_dev;
2308 	ost->st_ino = st->st_ino;
2309 	ost->st_mode = st->st_mode;
2310 	ost->st_nlink = st->st_nlink;
2311 	ost->st_uid = st->st_uid;
2312 	ost->st_gid = st->st_gid;
2313 	ost->st_rdev = st->st_rdev;
2314 	ost->st_size = MIN(st->st_size, INT32_MAX);
2315 	ost->st_atim = st->st_atim;
2316 	ost->st_mtim = st->st_mtim;
2317 	ost->st_ctim = st->st_ctim;
2318 	ost->st_blksize = st->st_blksize;
2319 	ost->st_blocks = st->st_blocks;
2320 	ost->st_flags = st->st_flags;
2321 	ost->st_gen = st->st_gen;
2322 }
2323 #endif /* COMPAT_43 */
2324 
2325 #if defined(COMPAT_43) || defined(COMPAT_FREEBSD11)
2326 int ino64_trunc_error;
2327 SYSCTL_INT(_vfs, OID_AUTO, ino64_trunc_error, CTLFLAG_RW,
2328     &ino64_trunc_error, 0,
2329     "Error on truncation of device, file or inode number, or link count");
2330 
2331 int
2332 freebsd11_cvtstat(struct stat *st, struct freebsd11_stat *ost)
2333 {
2334 
2335 	ost->st_dev = st->st_dev;
2336 	if (ost->st_dev != st->st_dev) {
2337 		switch (ino64_trunc_error) {
2338 		default:
2339 			/*
2340 			 * Since dev_t is almost raw, don't clamp to the
2341 			 * maximum for case 2, but ignore the error.
2342 			 */
2343 			break;
2344 		case 1:
2345 			return (EOVERFLOW);
2346 		}
2347 	}
2348 	ost->st_ino = st->st_ino;
2349 	if (ost->st_ino != st->st_ino) {
2350 		switch (ino64_trunc_error) {
2351 		default:
2352 		case 0:
2353 			break;
2354 		case 1:
2355 			return (EOVERFLOW);
2356 		case 2:
2357 			ost->st_ino = UINT32_MAX;
2358 			break;
2359 		}
2360 	}
2361 	ost->st_mode = st->st_mode;
2362 	ost->st_nlink = st->st_nlink;
2363 	if (ost->st_nlink != st->st_nlink) {
2364 		switch (ino64_trunc_error) {
2365 		default:
2366 		case 0:
2367 			break;
2368 		case 1:
2369 			return (EOVERFLOW);
2370 		case 2:
2371 			ost->st_nlink = UINT16_MAX;
2372 			break;
2373 		}
2374 	}
2375 	ost->st_uid = st->st_uid;
2376 	ost->st_gid = st->st_gid;
2377 	ost->st_rdev = st->st_rdev;
2378 	if (ost->st_rdev != st->st_rdev) {
2379 		switch (ino64_trunc_error) {
2380 		default:
2381 			break;
2382 		case 1:
2383 			return (EOVERFLOW);
2384 		}
2385 	}
2386 	ost->st_atim = st->st_atim;
2387 	ost->st_mtim = st->st_mtim;
2388 	ost->st_ctim = st->st_ctim;
2389 	ost->st_size = st->st_size;
2390 	ost->st_blocks = st->st_blocks;
2391 	ost->st_blksize = st->st_blksize;
2392 	ost->st_flags = st->st_flags;
2393 	ost->st_gen = st->st_gen;
2394 	ost->st_lspare = 0;
2395 	ost->st_birthtim = st->st_birthtim;
2396 	bzero((char *)&ost->st_birthtim + sizeof(ost->st_birthtim),
2397 	    sizeof(*ost) - offsetof(struct freebsd11_stat,
2398 	    st_birthtim) - sizeof(ost->st_birthtim));
2399 	return (0);
2400 }
2401 
2402 int
2403 freebsd11_stat(struct thread *td, struct freebsd11_stat_args* uap)
2404 {
2405 	struct stat sb;
2406 	struct freebsd11_stat osb;
2407 	int error;
2408 
2409 	error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE, &sb);
2410 	if (error != 0)
2411 		return (error);
2412 	error = freebsd11_cvtstat(&sb, &osb);
2413 	if (error == 0)
2414 		error = copyout(&osb, uap->ub, sizeof(osb));
2415 	return (error);
2416 }
2417 
2418 int
2419 freebsd11_lstat(struct thread *td, struct freebsd11_lstat_args* uap)
2420 {
2421 	struct stat sb;
2422 	struct freebsd11_stat osb;
2423 	int error;
2424 
2425 	error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2426 	    UIO_USERSPACE, &sb);
2427 	if (error != 0)
2428 		return (error);
2429 	error = freebsd11_cvtstat(&sb, &osb);
2430 	if (error == 0)
2431 		error = copyout(&osb, uap->ub, sizeof(osb));
2432 	return (error);
2433 }
2434 
2435 int
2436 freebsd11_fhstat(struct thread *td, struct freebsd11_fhstat_args* uap)
2437 {
2438 	struct fhandle fh;
2439 	struct stat sb;
2440 	struct freebsd11_stat osb;
2441 	int error;
2442 
2443 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
2444 	if (error != 0)
2445 		return (error);
2446 	error = kern_fhstat(td, fh, &sb);
2447 	if (error != 0)
2448 		return (error);
2449 	error = freebsd11_cvtstat(&sb, &osb);
2450 	if (error == 0)
2451 		error = copyout(&osb, uap->sb, sizeof(osb));
2452 	return (error);
2453 }
2454 
2455 int
2456 freebsd11_fstatat(struct thread *td, struct freebsd11_fstatat_args* uap)
2457 {
2458 	struct stat sb;
2459 	struct freebsd11_stat osb;
2460 	int error;
2461 
2462 	error = kern_statat(td, uap->flag, uap->fd, uap->path,
2463 	    UIO_USERSPACE, &sb);
2464 	if (error != 0)
2465 		return (error);
2466 	error = freebsd11_cvtstat(&sb, &osb);
2467 	if (error == 0)
2468 		error = copyout(&osb, uap->buf, sizeof(osb));
2469 	return (error);
2470 }
2471 #endif	/* COMPAT_FREEBSD11 */
2472 
2473 /*
2474  * Get file status
2475  */
2476 #ifndef _SYS_SYSPROTO_H_
2477 struct fstatat_args {
2478 	int	fd;
2479 	char	*path;
2480 	struct stat	*buf;
2481 	int	flag;
2482 }
2483 #endif
2484 int
2485 sys_fstatat(struct thread *td, struct fstatat_args *uap)
2486 {
2487 	struct stat sb;
2488 	int error;
2489 
2490 	error = kern_statat(td, uap->flag, uap->fd, uap->path,
2491 	    UIO_USERSPACE, &sb);
2492 	if (error == 0)
2493 		error = copyout(&sb, uap->buf, sizeof (sb));
2494 	return (error);
2495 }
2496 
2497 int
2498 kern_statat(struct thread *td, int flag, int fd, const char *path,
2499     enum uio_seg pathseg, struct stat *sbp)
2500 {
2501 	struct nameidata nd;
2502 	int error;
2503 
2504 	if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH |
2505 	    AT_EMPTY_PATH)) != 0)
2506 		return (EINVAL);
2507 
2508 	NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(flag, AT_RESOLVE_BENEATH |
2509 	    AT_SYMLINK_NOFOLLOW | AT_EMPTY_PATH) | LOCKSHARED | LOCKLEAF |
2510 	    AUDITVNODE1, pathseg, path, fd, &cap_fstat_rights);
2511 
2512 	if ((error = namei(&nd)) != 0) {
2513 		if (error == ENOTDIR &&
2514 		    (nd.ni_resflags & NIRES_EMPTYPATH) != 0)
2515 			error = kern_fstat(td, fd, sbp);
2516 		return (error);
2517 	}
2518 	error = VOP_STAT(nd.ni_vp, sbp, td->td_ucred, NOCRED);
2519 	NDFREE_PNBUF(&nd);
2520 	vput(nd.ni_vp);
2521 #ifdef __STAT_TIME_T_EXT
2522 	sbp->st_atim_ext = 0;
2523 	sbp->st_mtim_ext = 0;
2524 	sbp->st_ctim_ext = 0;
2525 	sbp->st_btim_ext = 0;
2526 #endif
2527 #ifdef KTRACE
2528 	if (KTRPOINT(td, KTR_STRUCT))
2529 		ktrstat_error(sbp, error);
2530 #endif
2531 	return (error);
2532 }
2533 
2534 #if defined(COMPAT_FREEBSD11)
2535 /*
2536  * Implementation of the NetBSD [l]stat() functions.
2537  */
2538 int
2539 freebsd11_cvtnstat(struct stat *sb, struct nstat *nsb)
2540 {
2541 	struct freebsd11_stat sb11;
2542 	int error;
2543 
2544 	error = freebsd11_cvtstat(sb, &sb11);
2545 	if (error != 0)
2546 		return (error);
2547 
2548 	bzero(nsb, sizeof(*nsb));
2549 	CP(sb11, *nsb, st_dev);
2550 	CP(sb11, *nsb, st_ino);
2551 	CP(sb11, *nsb, st_mode);
2552 	CP(sb11, *nsb, st_nlink);
2553 	CP(sb11, *nsb, st_uid);
2554 	CP(sb11, *nsb, st_gid);
2555 	CP(sb11, *nsb, st_rdev);
2556 	CP(sb11, *nsb, st_atim);
2557 	CP(sb11, *nsb, st_mtim);
2558 	CP(sb11, *nsb, st_ctim);
2559 	CP(sb11, *nsb, st_size);
2560 	CP(sb11, *nsb, st_blocks);
2561 	CP(sb11, *nsb, st_blksize);
2562 	CP(sb11, *nsb, st_flags);
2563 	CP(sb11, *nsb, st_gen);
2564 	CP(sb11, *nsb, st_birthtim);
2565 	return (0);
2566 }
2567 
2568 #ifndef _SYS_SYSPROTO_H_
2569 struct freebsd11_nstat_args {
2570 	char	*path;
2571 	struct nstat *ub;
2572 };
2573 #endif
2574 int
2575 freebsd11_nstat(struct thread *td, struct freebsd11_nstat_args *uap)
2576 {
2577 	struct stat sb;
2578 	struct nstat nsb;
2579 	int error;
2580 
2581 	error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE, &sb);
2582 	if (error != 0)
2583 		return (error);
2584 	error = freebsd11_cvtnstat(&sb, &nsb);
2585 	if (error == 0)
2586 		error = copyout(&nsb, uap->ub, sizeof (nsb));
2587 	return (error);
2588 }
2589 
2590 /*
2591  * NetBSD lstat.  Get file status; this version does not follow links.
2592  */
2593 #ifndef _SYS_SYSPROTO_H_
2594 struct freebsd11_nlstat_args {
2595 	char	*path;
2596 	struct nstat *ub;
2597 };
2598 #endif
2599 int
2600 freebsd11_nlstat(struct thread *td, struct freebsd11_nlstat_args *uap)
2601 {
2602 	struct stat sb;
2603 	struct nstat nsb;
2604 	int error;
2605 
2606 	error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2607 	    UIO_USERSPACE, &sb);
2608 	if (error != 0)
2609 		return (error);
2610 	error = freebsd11_cvtnstat(&sb, &nsb);
2611 	if (error == 0)
2612 		error = copyout(&nsb, uap->ub, sizeof (nsb));
2613 	return (error);
2614 }
2615 #endif /* COMPAT_FREEBSD11 */
2616 
2617 /*
2618  * Get configurable pathname variables.
2619  */
2620 #ifndef _SYS_SYSPROTO_H_
2621 struct pathconf_args {
2622 	char	*path;
2623 	int	name;
2624 };
2625 #endif
2626 int
2627 sys_pathconf(struct thread *td, struct pathconf_args *uap)
2628 {
2629 	long value;
2630 	int error;
2631 
2632 	error = kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name, FOLLOW,
2633 	    &value);
2634 	if (error == 0)
2635 		td->td_retval[0] = value;
2636 	return (error);
2637 }
2638 
2639 #ifndef _SYS_SYSPROTO_H_
2640 struct lpathconf_args {
2641 	char	*path;
2642 	int	name;
2643 };
2644 #endif
2645 int
2646 sys_lpathconf(struct thread *td, struct lpathconf_args *uap)
2647 {
2648 	long value;
2649 	int error;
2650 
2651 	error = kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name,
2652 	    NOFOLLOW, &value);
2653 	if (error == 0)
2654 		td->td_retval[0] = value;
2655 	return (error);
2656 }
2657 
2658 int
2659 kern_pathconf(struct thread *td, const char *path, enum uio_seg pathseg,
2660     int name, u_long flags, long *valuep)
2661 {
2662 	struct nameidata nd;
2663 	int error;
2664 
2665 	NDINIT(&nd, LOOKUP, LOCKSHARED | LOCKLEAF | AUDITVNODE1 | flags,
2666 	       pathseg, path);
2667 	if ((error = namei(&nd)) != 0)
2668 		return (error);
2669 	NDFREE_PNBUF(&nd);
2670 
2671 	error = VOP_PATHCONF(nd.ni_vp, name, valuep);
2672 	vput(nd.ni_vp);
2673 	return (error);
2674 }
2675 
2676 /*
2677  * Return target name of a symbolic link.
2678  */
2679 #ifndef _SYS_SYSPROTO_H_
2680 struct readlink_args {
2681 	char	*path;
2682 	char	*buf;
2683 	size_t	count;
2684 };
2685 #endif
2686 int
2687 sys_readlink(struct thread *td, struct readlink_args *uap)
2688 {
2689 
2690 	return (kern_readlinkat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2691 	    uap->buf, UIO_USERSPACE, uap->count));
2692 }
2693 #ifndef _SYS_SYSPROTO_H_
2694 struct readlinkat_args {
2695 	int	fd;
2696 	char	*path;
2697 	char	*buf;
2698 	size_t	bufsize;
2699 };
2700 #endif
2701 int
2702 sys_readlinkat(struct thread *td, struct readlinkat_args *uap)
2703 {
2704 
2705 	return (kern_readlinkat(td, uap->fd, uap->path, UIO_USERSPACE,
2706 	    uap->buf, UIO_USERSPACE, uap->bufsize));
2707 }
2708 
2709 int
2710 kern_readlinkat(struct thread *td, int fd, const char *path,
2711     enum uio_seg pathseg, char *buf, enum uio_seg bufseg, size_t count)
2712 {
2713 	struct vnode *vp;
2714 	struct nameidata nd;
2715 	int error;
2716 
2717 	if (count > IOSIZE_MAX)
2718 		return (EINVAL);
2719 
2720 	NDINIT_AT(&nd, LOOKUP, NOFOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1 |
2721 	    EMPTYPATH, pathseg, path, fd);
2722 
2723 	if ((error = namei(&nd)) != 0)
2724 		return (error);
2725 	NDFREE_PNBUF(&nd);
2726 	vp = nd.ni_vp;
2727 
2728 	error = kern_readlink_vp(vp, buf, bufseg, count, td);
2729 	vput(vp);
2730 
2731 	return (error);
2732 }
2733 
2734 /*
2735  * Helper function to readlink from a vnode
2736  */
2737 static int
2738 kern_readlink_vp(struct vnode *vp, char *buf, enum uio_seg bufseg, size_t count,
2739     struct thread *td)
2740 {
2741 	struct iovec aiov;
2742 	struct uio auio;
2743 	int error;
2744 
2745 	ASSERT_VOP_LOCKED(vp, "kern_readlink_vp(): vp not locked");
2746 #ifdef MAC
2747 	error = mac_vnode_check_readlink(td->td_ucred, vp);
2748 	if (error != 0)
2749 		return (error);
2750 #endif
2751 	if (vp->v_type != VLNK && (vp->v_vflag & VV_READLINK) == 0)
2752 		return (EINVAL);
2753 
2754 	aiov.iov_base = buf;
2755 	aiov.iov_len = count;
2756 	auio.uio_iov = &aiov;
2757 	auio.uio_iovcnt = 1;
2758 	auio.uio_offset = 0;
2759 	auio.uio_rw = UIO_READ;
2760 	auio.uio_segflg = bufseg;
2761 	auio.uio_td = td;
2762 	auio.uio_resid = count;
2763 	error = VOP_READLINK(vp, &auio, td->td_ucred);
2764 	td->td_retval[0] = count - auio.uio_resid;
2765 	return (error);
2766 }
2767 
2768 /*
2769  * Common implementation code for chflags() and fchflags().
2770  */
2771 static int
2772 setfflags(struct thread *td, struct vnode *vp, u_long flags)
2773 {
2774 	struct mount *mp;
2775 	struct vattr vattr;
2776 	int error;
2777 
2778 	/* We can't support the value matching VNOVAL. */
2779 	if (flags == VNOVAL)
2780 		return (EOPNOTSUPP);
2781 
2782 	/*
2783 	 * Prevent non-root users from setting flags on devices.  When
2784 	 * a device is reused, users can retain ownership of the device
2785 	 * if they are allowed to set flags and programs assume that
2786 	 * chown can't fail when done as root.
2787 	 */
2788 	if (vp->v_type == VCHR || vp->v_type == VBLK) {
2789 		error = priv_check(td, PRIV_VFS_CHFLAGS_DEV);
2790 		if (error != 0)
2791 			return (error);
2792 	}
2793 
2794 	if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0)
2795 		return (error);
2796 	VATTR_NULL(&vattr);
2797 	vattr.va_flags = flags;
2798 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2799 #ifdef MAC
2800 	error = mac_vnode_check_setflags(td->td_ucred, vp, vattr.va_flags);
2801 	if (error == 0)
2802 #endif
2803 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
2804 	VOP_UNLOCK(vp);
2805 	vn_finished_write(mp);
2806 	return (error);
2807 }
2808 
2809 /*
2810  * Change flags of a file given a path name.
2811  */
2812 #ifndef _SYS_SYSPROTO_H_
2813 struct chflags_args {
2814 	const char *path;
2815 	u_long	flags;
2816 };
2817 #endif
2818 int
2819 sys_chflags(struct thread *td, struct chflags_args *uap)
2820 {
2821 
2822 	return (kern_chflagsat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2823 	    uap->flags, 0));
2824 }
2825 
2826 #ifndef _SYS_SYSPROTO_H_
2827 struct chflagsat_args {
2828 	int	fd;
2829 	const char *path;
2830 	u_long	flags;
2831 	int	atflag;
2832 }
2833 #endif
2834 int
2835 sys_chflagsat(struct thread *td, struct chflagsat_args *uap)
2836 {
2837 
2838 	return (kern_chflagsat(td, uap->fd, uap->path, UIO_USERSPACE,
2839 	    uap->flags, uap->atflag));
2840 }
2841 
2842 /*
2843  * Same as chflags() but doesn't follow symlinks.
2844  */
2845 #ifndef _SYS_SYSPROTO_H_
2846 struct lchflags_args {
2847 	const char *path;
2848 	u_long flags;
2849 };
2850 #endif
2851 int
2852 sys_lchflags(struct thread *td, struct lchflags_args *uap)
2853 {
2854 
2855 	return (kern_chflagsat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2856 	    uap->flags, AT_SYMLINK_NOFOLLOW));
2857 }
2858 
2859 static int
2860 kern_chflagsat(struct thread *td, int fd, const char *path,
2861     enum uio_seg pathseg, u_long flags, int atflag)
2862 {
2863 	struct nameidata nd;
2864 	int error;
2865 
2866 	if ((atflag & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH |
2867 	    AT_EMPTY_PATH)) != 0)
2868 		return (EINVAL);
2869 
2870 	AUDIT_ARG_FFLAGS(flags);
2871 	NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(atflag, AT_SYMLINK_NOFOLLOW |
2872 	    AT_RESOLVE_BENEATH | AT_EMPTY_PATH) | AUDITVNODE1, pathseg, path,
2873 	    fd, &cap_fchflags_rights);
2874 	if ((error = namei(&nd)) != 0)
2875 		return (error);
2876 	NDFREE_PNBUF(&nd);
2877 	error = setfflags(td, nd.ni_vp, flags);
2878 	vrele(nd.ni_vp);
2879 	return (error);
2880 }
2881 
2882 /*
2883  * Change flags of a file given a file descriptor.
2884  */
2885 #ifndef _SYS_SYSPROTO_H_
2886 struct fchflags_args {
2887 	int	fd;
2888 	u_long	flags;
2889 };
2890 #endif
2891 int
2892 sys_fchflags(struct thread *td, struct fchflags_args *uap)
2893 {
2894 	struct file *fp;
2895 	int error;
2896 
2897 	AUDIT_ARG_FD(uap->fd);
2898 	AUDIT_ARG_FFLAGS(uap->flags);
2899 	error = getvnode(td, uap->fd, &cap_fchflags_rights,
2900 	    &fp);
2901 	if (error != 0)
2902 		return (error);
2903 #ifdef AUDIT
2904 	if (AUDITING_TD(td)) {
2905 		vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
2906 		AUDIT_ARG_VNODE1(fp->f_vnode);
2907 		VOP_UNLOCK(fp->f_vnode);
2908 	}
2909 #endif
2910 	error = setfflags(td, fp->f_vnode, uap->flags);
2911 	fdrop(fp, td);
2912 	return (error);
2913 }
2914 
2915 /*
2916  * Common implementation code for chmod(), lchmod() and fchmod().
2917  */
2918 int
2919 setfmode(struct thread *td, struct ucred *cred, struct vnode *vp, int mode)
2920 {
2921 	struct mount *mp;
2922 	struct vattr vattr;
2923 	int error;
2924 
2925 	if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0)
2926 		return (error);
2927 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2928 	VATTR_NULL(&vattr);
2929 	vattr.va_mode = mode & ALLPERMS;
2930 #ifdef MAC
2931 	error = mac_vnode_check_setmode(cred, vp, vattr.va_mode);
2932 	if (error == 0)
2933 #endif
2934 		error = VOP_SETATTR(vp, &vattr, cred);
2935 	VOP_UNLOCK(vp);
2936 	vn_finished_write(mp);
2937 	return (error);
2938 }
2939 
2940 /*
2941  * Change mode of a file given path name.
2942  */
2943 #ifndef _SYS_SYSPROTO_H_
2944 struct chmod_args {
2945 	char	*path;
2946 	int	mode;
2947 };
2948 #endif
2949 int
2950 sys_chmod(struct thread *td, struct chmod_args *uap)
2951 {
2952 
2953 	return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2954 	    uap->mode, 0));
2955 }
2956 
2957 #ifndef _SYS_SYSPROTO_H_
2958 struct fchmodat_args {
2959 	int	dirfd;
2960 	char	*path;
2961 	mode_t	mode;
2962 	int	flag;
2963 }
2964 #endif
2965 int
2966 sys_fchmodat(struct thread *td, struct fchmodat_args *uap)
2967 {
2968 
2969 	return (kern_fchmodat(td, uap->fd, uap->path, UIO_USERSPACE,
2970 	    uap->mode, uap->flag));
2971 }
2972 
2973 /*
2974  * Change mode of a file given path name (don't follow links.)
2975  */
2976 #ifndef _SYS_SYSPROTO_H_
2977 struct lchmod_args {
2978 	char	*path;
2979 	int	mode;
2980 };
2981 #endif
2982 int
2983 sys_lchmod(struct thread *td, struct lchmod_args *uap)
2984 {
2985 
2986 	return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2987 	    uap->mode, AT_SYMLINK_NOFOLLOW));
2988 }
2989 
2990 int
2991 kern_fchmodat(struct thread *td, int fd, const char *path,
2992     enum uio_seg pathseg, mode_t mode, int flag)
2993 {
2994 	struct nameidata nd;
2995 	int error;
2996 
2997 	if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH |
2998 	    AT_EMPTY_PATH)) != 0)
2999 		return (EINVAL);
3000 
3001 	AUDIT_ARG_MODE(mode);
3002 	NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(flag, AT_SYMLINK_NOFOLLOW |
3003 	    AT_RESOLVE_BENEATH | AT_EMPTY_PATH) | AUDITVNODE1, pathseg, path,
3004 	    fd, &cap_fchmod_rights);
3005 	if ((error = namei(&nd)) != 0)
3006 		return (error);
3007 	NDFREE_PNBUF(&nd);
3008 	error = setfmode(td, td->td_ucred, nd.ni_vp, mode);
3009 	vrele(nd.ni_vp);
3010 	return (error);
3011 }
3012 
3013 /*
3014  * Change mode of a file given a file descriptor.
3015  */
3016 #ifndef _SYS_SYSPROTO_H_
3017 struct fchmod_args {
3018 	int	fd;
3019 	int	mode;
3020 };
3021 #endif
3022 int
3023 sys_fchmod(struct thread *td, struct fchmod_args *uap)
3024 {
3025 	struct file *fp;
3026 	int error;
3027 
3028 	AUDIT_ARG_FD(uap->fd);
3029 	AUDIT_ARG_MODE(uap->mode);
3030 
3031 	error = fget(td, uap->fd, &cap_fchmod_rights, &fp);
3032 	if (error != 0)
3033 		return (error);
3034 	error = fo_chmod(fp, uap->mode, td->td_ucred, td);
3035 	fdrop(fp, td);
3036 	return (error);
3037 }
3038 
3039 /*
3040  * Common implementation for chown(), lchown(), and fchown()
3041  */
3042 int
3043 setfown(struct thread *td, struct ucred *cred, struct vnode *vp, uid_t uid,
3044     gid_t gid)
3045 {
3046 	struct mount *mp;
3047 	struct vattr vattr;
3048 	int error;
3049 
3050 	if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0)
3051 		return (error);
3052 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3053 	VATTR_NULL(&vattr);
3054 	vattr.va_uid = uid;
3055 	vattr.va_gid = gid;
3056 #ifdef MAC
3057 	error = mac_vnode_check_setowner(cred, vp, vattr.va_uid,
3058 	    vattr.va_gid);
3059 	if (error == 0)
3060 #endif
3061 		error = VOP_SETATTR(vp, &vattr, cred);
3062 	VOP_UNLOCK(vp);
3063 	vn_finished_write(mp);
3064 	return (error);
3065 }
3066 
3067 /*
3068  * Set ownership given a path name.
3069  */
3070 #ifndef _SYS_SYSPROTO_H_
3071 struct chown_args {
3072 	char	*path;
3073 	int	uid;
3074 	int	gid;
3075 };
3076 #endif
3077 int
3078 sys_chown(struct thread *td, struct chown_args *uap)
3079 {
3080 
3081 	return (kern_fchownat(td, AT_FDCWD, uap->path, UIO_USERSPACE, uap->uid,
3082 	    uap->gid, 0));
3083 }
3084 
3085 #ifndef _SYS_SYSPROTO_H_
3086 struct fchownat_args {
3087 	int fd;
3088 	const char * path;
3089 	uid_t uid;
3090 	gid_t gid;
3091 	int flag;
3092 };
3093 #endif
3094 int
3095 sys_fchownat(struct thread *td, struct fchownat_args *uap)
3096 {
3097 
3098 	return (kern_fchownat(td, uap->fd, uap->path, UIO_USERSPACE, uap->uid,
3099 	    uap->gid, uap->flag));
3100 }
3101 
3102 int
3103 kern_fchownat(struct thread *td, int fd, const char *path,
3104     enum uio_seg pathseg, int uid, int gid, int flag)
3105 {
3106 	struct nameidata nd;
3107 	int error;
3108 
3109 	if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH |
3110 	    AT_EMPTY_PATH)) != 0)
3111 		return (EINVAL);
3112 
3113 	AUDIT_ARG_OWNER(uid, gid);
3114 	NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(flag, AT_SYMLINK_NOFOLLOW |
3115 	    AT_RESOLVE_BENEATH | AT_EMPTY_PATH) | AUDITVNODE1, pathseg, path,
3116 	    fd, &cap_fchown_rights);
3117 
3118 	if ((error = namei(&nd)) != 0)
3119 		return (error);
3120 	NDFREE_PNBUF(&nd);
3121 	error = setfown(td, td->td_ucred, nd.ni_vp, uid, gid);
3122 	vrele(nd.ni_vp);
3123 	return (error);
3124 }
3125 
3126 /*
3127  * Set ownership given a path name, do not cross symlinks.
3128  */
3129 #ifndef _SYS_SYSPROTO_H_
3130 struct lchown_args {
3131 	char	*path;
3132 	int	uid;
3133 	int	gid;
3134 };
3135 #endif
3136 int
3137 sys_lchown(struct thread *td, struct lchown_args *uap)
3138 {
3139 
3140 	return (kern_fchownat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
3141 	    uap->uid, uap->gid, AT_SYMLINK_NOFOLLOW));
3142 }
3143 
3144 /*
3145  * Set ownership given a file descriptor.
3146  */
3147 #ifndef _SYS_SYSPROTO_H_
3148 struct fchown_args {
3149 	int	fd;
3150 	int	uid;
3151 	int	gid;
3152 };
3153 #endif
3154 int
3155 sys_fchown(struct thread *td, struct fchown_args *uap)
3156 {
3157 	struct file *fp;
3158 	int error;
3159 
3160 	AUDIT_ARG_FD(uap->fd);
3161 	AUDIT_ARG_OWNER(uap->uid, uap->gid);
3162 	error = fget(td, uap->fd, &cap_fchown_rights, &fp);
3163 	if (error != 0)
3164 		return (error);
3165 	error = fo_chown(fp, uap->uid, uap->gid, td->td_ucred, td);
3166 	fdrop(fp, td);
3167 	return (error);
3168 }
3169 
3170 /*
3171  * Common implementation code for utimes(), lutimes(), and futimes().
3172  */
3173 static int
3174 getutimes(const struct timeval *usrtvp, enum uio_seg tvpseg,
3175     struct timespec *tsp)
3176 {
3177 	struct timeval tv[2];
3178 	const struct timeval *tvp;
3179 	int error;
3180 
3181 	if (usrtvp == NULL) {
3182 		vfs_timestamp(&tsp[0]);
3183 		tsp[1] = tsp[0];
3184 	} else {
3185 		if (tvpseg == UIO_SYSSPACE) {
3186 			tvp = usrtvp;
3187 		} else {
3188 			if ((error = copyin(usrtvp, tv, sizeof(tv))) != 0)
3189 				return (error);
3190 			tvp = tv;
3191 		}
3192 
3193 		if (tvp[0].tv_usec < 0 || tvp[0].tv_usec >= 1000000 ||
3194 		    tvp[1].tv_usec < 0 || tvp[1].tv_usec >= 1000000)
3195 			return (EINVAL);
3196 		TIMEVAL_TO_TIMESPEC(&tvp[0], &tsp[0]);
3197 		TIMEVAL_TO_TIMESPEC(&tvp[1], &tsp[1]);
3198 	}
3199 	return (0);
3200 }
3201 
3202 /*
3203  * Common implementation code for futimens(), utimensat().
3204  */
3205 #define	UTIMENS_NULL	0x1
3206 #define	UTIMENS_EXIT	0x2
3207 static int
3208 getutimens(const struct timespec *usrtsp, enum uio_seg tspseg,
3209     struct timespec *tsp, int *retflags)
3210 {
3211 	struct timespec tsnow;
3212 	int error;
3213 
3214 	vfs_timestamp(&tsnow);
3215 	*retflags = 0;
3216 	if (usrtsp == NULL) {
3217 		tsp[0] = tsnow;
3218 		tsp[1] = tsnow;
3219 		*retflags |= UTIMENS_NULL;
3220 		return (0);
3221 	}
3222 	if (tspseg == UIO_SYSSPACE) {
3223 		tsp[0] = usrtsp[0];
3224 		tsp[1] = usrtsp[1];
3225 	} else if ((error = copyin(usrtsp, tsp, sizeof(*tsp) * 2)) != 0)
3226 		return (error);
3227 	if (tsp[0].tv_nsec == UTIME_OMIT && tsp[1].tv_nsec == UTIME_OMIT)
3228 		*retflags |= UTIMENS_EXIT;
3229 	if (tsp[0].tv_nsec == UTIME_NOW && tsp[1].tv_nsec == UTIME_NOW)
3230 		*retflags |= UTIMENS_NULL;
3231 	if (tsp[0].tv_nsec == UTIME_OMIT)
3232 		tsp[0].tv_sec = VNOVAL;
3233 	else if (tsp[0].tv_nsec == UTIME_NOW)
3234 		tsp[0] = tsnow;
3235 	else if (tsp[0].tv_nsec < 0 || tsp[0].tv_nsec >= 1000000000L)
3236 		return (EINVAL);
3237 	if (tsp[1].tv_nsec == UTIME_OMIT)
3238 		tsp[1].tv_sec = VNOVAL;
3239 	else if (tsp[1].tv_nsec == UTIME_NOW)
3240 		tsp[1] = tsnow;
3241 	else if (tsp[1].tv_nsec < 0 || tsp[1].tv_nsec >= 1000000000L)
3242 		return (EINVAL);
3243 
3244 	return (0);
3245 }
3246 
3247 /*
3248  * Common implementation code for utimes(), lutimes(), futimes(), futimens(),
3249  * and utimensat().
3250  */
3251 static int
3252 setutimes(struct thread *td, struct vnode *vp, const struct timespec *ts,
3253     int numtimes, int nullflag)
3254 {
3255 	struct mount *mp;
3256 	struct vattr vattr;
3257 	int error;
3258 	bool setbirthtime;
3259 
3260 	setbirthtime = false;
3261 	vattr.va_birthtime.tv_sec = VNOVAL;
3262 	vattr.va_birthtime.tv_nsec = 0;
3263 
3264 	if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0)
3265 		return (error);
3266 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3267 	if (numtimes < 3 && VOP_GETATTR(vp, &vattr, td->td_ucred) == 0 &&
3268 	    timespeccmp(&ts[1], &vattr.va_birthtime, < ))
3269 		setbirthtime = true;
3270 	VATTR_NULL(&vattr);
3271 	vattr.va_atime = ts[0];
3272 	vattr.va_mtime = ts[1];
3273 	if (setbirthtime)
3274 		vattr.va_birthtime = ts[1];
3275 	if (numtimes > 2)
3276 		vattr.va_birthtime = ts[2];
3277 	if (nullflag)
3278 		vattr.va_vaflags |= VA_UTIMES_NULL;
3279 #ifdef MAC
3280 	error = mac_vnode_check_setutimes(td->td_ucred, vp, vattr.va_atime,
3281 	    vattr.va_mtime);
3282 #endif
3283 	if (error == 0)
3284 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3285 	VOP_UNLOCK(vp);
3286 	vn_finished_write(mp);
3287 	return (error);
3288 }
3289 
3290 /*
3291  * Set the access and modification times of a file.
3292  */
3293 #ifndef _SYS_SYSPROTO_H_
3294 struct utimes_args {
3295 	char	*path;
3296 	struct	timeval *tptr;
3297 };
3298 #endif
3299 int
3300 sys_utimes(struct thread *td, struct utimes_args *uap)
3301 {
3302 
3303 	return (kern_utimesat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
3304 	    uap->tptr, UIO_USERSPACE));
3305 }
3306 
3307 #ifndef _SYS_SYSPROTO_H_
3308 struct futimesat_args {
3309 	int fd;
3310 	const char * path;
3311 	const struct timeval * times;
3312 };
3313 #endif
3314 int
3315 sys_futimesat(struct thread *td, struct futimesat_args *uap)
3316 {
3317 
3318 	return (kern_utimesat(td, uap->fd, uap->path, UIO_USERSPACE,
3319 	    uap->times, UIO_USERSPACE));
3320 }
3321 
3322 int
3323 kern_utimesat(struct thread *td, int fd, const char *path,
3324     enum uio_seg pathseg, const struct timeval *tptr, enum uio_seg tptrseg)
3325 {
3326 	struct nameidata nd;
3327 	struct timespec ts[2];
3328 	int error;
3329 
3330 	if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3331 		return (error);
3332 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, fd,
3333 	    &cap_futimes_rights);
3334 
3335 	if ((error = namei(&nd)) != 0)
3336 		return (error);
3337 	NDFREE_PNBUF(&nd);
3338 	error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3339 	vrele(nd.ni_vp);
3340 	return (error);
3341 }
3342 
3343 /*
3344  * Set the access and modification times of a file.
3345  */
3346 #ifndef _SYS_SYSPROTO_H_
3347 struct lutimes_args {
3348 	char	*path;
3349 	struct	timeval *tptr;
3350 };
3351 #endif
3352 int
3353 sys_lutimes(struct thread *td, struct lutimes_args *uap)
3354 {
3355 
3356 	return (kern_lutimes(td, uap->path, UIO_USERSPACE, uap->tptr,
3357 	    UIO_USERSPACE));
3358 }
3359 
3360 int
3361 kern_lutimes(struct thread *td, const char *path, enum uio_seg pathseg,
3362     const struct timeval *tptr, enum uio_seg tptrseg)
3363 {
3364 	struct timespec ts[2];
3365 	struct nameidata nd;
3366 	int error;
3367 
3368 	if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3369 		return (error);
3370 	NDINIT(&nd, LOOKUP, NOFOLLOW | AUDITVNODE1, pathseg, path);
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 futimes_args {
3384 	int	fd;
3385 	struct	timeval *tptr;
3386 };
3387 #endif
3388 int
3389 sys_futimes(struct thread *td, struct futimes_args *uap)
3390 {
3391 
3392 	return (kern_futimes(td, uap->fd, uap->tptr, UIO_USERSPACE));
3393 }
3394 
3395 int
3396 kern_futimes(struct thread *td, int fd, const struct timeval *tptr,
3397     enum uio_seg tptrseg)
3398 {
3399 	struct timespec ts[2];
3400 	struct file *fp;
3401 	int error;
3402 
3403 	AUDIT_ARG_FD(fd);
3404 	error = getutimes(tptr, tptrseg, ts);
3405 	if (error != 0)
3406 		return (error);
3407 	error = getvnode(td, fd, &cap_futimes_rights, &fp);
3408 	if (error != 0)
3409 		return (error);
3410 #ifdef AUDIT
3411 	if (AUDITING_TD(td)) {
3412 		vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
3413 		AUDIT_ARG_VNODE1(fp->f_vnode);
3414 		VOP_UNLOCK(fp->f_vnode);
3415 	}
3416 #endif
3417 	error = setutimes(td, fp->f_vnode, ts, 2, tptr == NULL);
3418 	fdrop(fp, td);
3419 	return (error);
3420 }
3421 
3422 int
3423 sys_futimens(struct thread *td, struct futimens_args *uap)
3424 {
3425 
3426 	return (kern_futimens(td, uap->fd, uap->times, UIO_USERSPACE));
3427 }
3428 
3429 int
3430 kern_futimens(struct thread *td, int fd, const struct timespec *tptr,
3431     enum uio_seg tptrseg)
3432 {
3433 	struct timespec ts[2];
3434 	struct file *fp;
3435 	int error, flags;
3436 
3437 	AUDIT_ARG_FD(fd);
3438 	error = getutimens(tptr, tptrseg, ts, &flags);
3439 	if (error != 0)
3440 		return (error);
3441 	if (flags & UTIMENS_EXIT)
3442 		return (0);
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, flags & UTIMENS_NULL);
3454 	fdrop(fp, td);
3455 	return (error);
3456 }
3457 
3458 int
3459 sys_utimensat(struct thread *td, struct utimensat_args *uap)
3460 {
3461 
3462 	return (kern_utimensat(td, uap->fd, uap->path, UIO_USERSPACE,
3463 	    uap->times, UIO_USERSPACE, uap->flag));
3464 }
3465 
3466 int
3467 kern_utimensat(struct thread *td, int fd, const char *path,
3468     enum uio_seg pathseg, const struct timespec *tptr, enum uio_seg tptrseg,
3469     int flag)
3470 {
3471 	struct nameidata nd;
3472 	struct timespec ts[2];
3473 	int error, flags;
3474 
3475 	if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH |
3476 	    AT_EMPTY_PATH)) != 0)
3477 		return (EINVAL);
3478 
3479 	if ((error = getutimens(tptr, tptrseg, ts, &flags)) != 0)
3480 		return (error);
3481 	NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(flag, AT_SYMLINK_NOFOLLOW |
3482 	    AT_RESOLVE_BENEATH | AT_EMPTY_PATH) | AUDITVNODE1,
3483 	    pathseg, path, fd, &cap_futimes_rights);
3484 	if ((error = namei(&nd)) != 0)
3485 		return (error);
3486 	/*
3487 	 * We are allowed to call namei() regardless of 2xUTIME_OMIT.
3488 	 * POSIX states:
3489 	 * "If both tv_nsec fields are UTIME_OMIT... EACCESS may be detected."
3490 	 * "Search permission is denied by a component of the path prefix."
3491 	 */
3492 	NDFREE_PNBUF(&nd);
3493 	if ((flags & UTIMENS_EXIT) == 0)
3494 		error = setutimes(td, nd.ni_vp, ts, 2, flags & UTIMENS_NULL);
3495 	vrele(nd.ni_vp);
3496 	return (error);
3497 }
3498 
3499 /*
3500  * Truncate a file given its path name.
3501  */
3502 #ifndef _SYS_SYSPROTO_H_
3503 struct truncate_args {
3504 	char	*path;
3505 	int	pad;
3506 	off_t	length;
3507 };
3508 #endif
3509 int
3510 sys_truncate(struct thread *td, struct truncate_args *uap)
3511 {
3512 
3513 	return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3514 }
3515 
3516 int
3517 kern_truncate(struct thread *td, const char *path, enum uio_seg pathseg,
3518     off_t length)
3519 {
3520 	struct mount *mp;
3521 	struct vnode *vp;
3522 	void *rl_cookie;
3523 	struct nameidata nd;
3524 	int error;
3525 
3526 	if (length < 0)
3527 		return (EINVAL);
3528 	NDPREINIT(&nd);
3529 retry:
3530 	NDINIT(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path);
3531 	if ((error = namei(&nd)) != 0)
3532 		return (error);
3533 	vp = nd.ni_vp;
3534 	NDFREE_PNBUF(&nd);
3535 	rl_cookie = vn_rangelock_wlock(vp, 0, OFF_MAX);
3536 	if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) {
3537 		vn_rangelock_unlock(vp, rl_cookie);
3538 		vrele(vp);
3539 		return (error);
3540 	}
3541 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3542 	if (vp->v_type == VDIR) {
3543 		error = EISDIR;
3544 		goto out;
3545 	}
3546 #ifdef MAC
3547 	error = mac_vnode_check_write(td->td_ucred, NOCRED, vp);
3548 	if (error != 0)
3549 		goto out;
3550 #endif
3551 	error = VOP_ACCESS(vp, VWRITE, td->td_ucred, td);
3552 	if (error != 0)
3553 		goto out;
3554 
3555 	error = vn_truncate_locked(vp, length, false, td->td_ucred);
3556 out:
3557 	VOP_UNLOCK(vp);
3558 	vn_finished_write(mp);
3559 	vn_rangelock_unlock(vp, rl_cookie);
3560 	vrele(vp);
3561 	if (error == ERELOOKUP)
3562 		goto retry;
3563 	return (error);
3564 }
3565 
3566 #if defined(COMPAT_43)
3567 /*
3568  * Truncate a file given its path name.
3569  */
3570 #ifndef _SYS_SYSPROTO_H_
3571 struct otruncate_args {
3572 	char	*path;
3573 	long	length;
3574 };
3575 #endif
3576 int
3577 otruncate(struct thread *td, struct otruncate_args *uap)
3578 {
3579 
3580 	return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3581 }
3582 #endif /* COMPAT_43 */
3583 
3584 #if defined(COMPAT_FREEBSD6)
3585 /* Versions with the pad argument */
3586 int
3587 freebsd6_truncate(struct thread *td, struct freebsd6_truncate_args *uap)
3588 {
3589 
3590 	return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3591 }
3592 
3593 int
3594 freebsd6_ftruncate(struct thread *td, struct freebsd6_ftruncate_args *uap)
3595 {
3596 
3597 	return (kern_ftruncate(td, uap->fd, uap->length));
3598 }
3599 #endif
3600 
3601 int
3602 kern_fsync(struct thread *td, int fd, bool fullsync)
3603 {
3604 	struct vnode *vp;
3605 	struct mount *mp;
3606 	struct file *fp;
3607 	int error;
3608 
3609 	AUDIT_ARG_FD(fd);
3610 	error = getvnode(td, fd, &cap_fsync_rights, &fp);
3611 	if (error != 0)
3612 		return (error);
3613 	vp = fp->f_vnode;
3614 #if 0
3615 	if (!fullsync)
3616 		/* XXXKIB: compete outstanding aio writes */;
3617 #endif
3618 retry:
3619 	error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH);
3620 	if (error != 0)
3621 		goto drop;
3622 	vn_lock(vp, vn_lktype_write(mp, vp) | LK_RETRY);
3623 	AUDIT_ARG_VNODE1(vp);
3624 	vnode_pager_clean_async(vp);
3625 	error = fullsync ? VOP_FSYNC(vp, MNT_WAIT, td) : VOP_FDATASYNC(vp, td);
3626 	VOP_UNLOCK(vp);
3627 	vn_finished_write(mp);
3628 	if (error == ERELOOKUP)
3629 		goto retry;
3630 drop:
3631 	fdrop(fp, td);
3632 	return (error);
3633 }
3634 
3635 /*
3636  * Sync an open file.
3637  */
3638 #ifndef _SYS_SYSPROTO_H_
3639 struct fsync_args {
3640 	int	fd;
3641 };
3642 #endif
3643 int
3644 sys_fsync(struct thread *td, struct fsync_args *uap)
3645 {
3646 
3647 	return (kern_fsync(td, uap->fd, true));
3648 }
3649 
3650 int
3651 sys_fdatasync(struct thread *td, struct fdatasync_args *uap)
3652 {
3653 
3654 	return (kern_fsync(td, uap->fd, false));
3655 }
3656 
3657 /*
3658  * Rename files.  Source and destination must either both be directories, or
3659  * both not be directories.  If target is a directory, it must be empty.
3660  */
3661 #ifndef _SYS_SYSPROTO_H_
3662 struct rename_args {
3663 	char	*from;
3664 	char	*to;
3665 };
3666 #endif
3667 int
3668 sys_rename(struct thread *td, struct rename_args *uap)
3669 {
3670 
3671 	return (kern_renameat(td, AT_FDCWD, uap->from, AT_FDCWD,
3672 	    uap->to, UIO_USERSPACE));
3673 }
3674 
3675 #ifndef _SYS_SYSPROTO_H_
3676 struct renameat_args {
3677 	int	oldfd;
3678 	char	*old;
3679 	int	newfd;
3680 	char	*new;
3681 };
3682 #endif
3683 int
3684 sys_renameat(struct thread *td, struct renameat_args *uap)
3685 {
3686 
3687 	return (kern_renameat(td, uap->oldfd, uap->old, uap->newfd, uap->new,
3688 	    UIO_USERSPACE));
3689 }
3690 
3691 #ifdef MAC
3692 static int
3693 kern_renameat_mac(struct thread *td, int oldfd, const char *old, int newfd,
3694     const char *new, enum uio_seg pathseg, struct nameidata *fromnd)
3695 {
3696 	int error;
3697 
3698 	NDINIT_ATRIGHTS(fromnd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1,
3699 	    pathseg, old, oldfd, &cap_renameat_source_rights);
3700 	if ((error = namei(fromnd)) != 0)
3701 		return (error);
3702 	error = mac_vnode_check_rename_from(td->td_ucred, fromnd->ni_dvp,
3703 	    fromnd->ni_vp, &fromnd->ni_cnd);
3704 	VOP_UNLOCK(fromnd->ni_dvp);
3705 	if (fromnd->ni_dvp != fromnd->ni_vp)
3706 		VOP_UNLOCK(fromnd->ni_vp);
3707 	if (error != 0) {
3708 		NDFREE_PNBUF(fromnd);
3709 		vrele(fromnd->ni_dvp);
3710 		vrele(fromnd->ni_vp);
3711 	}
3712 	return (error);
3713 }
3714 #endif
3715 
3716 int
3717 kern_renameat(struct thread *td, int oldfd, const char *old, int newfd,
3718     const char *new, enum uio_seg pathseg)
3719 {
3720 	struct mount *mp = NULL;
3721 	struct vnode *tvp, *fvp, *tdvp;
3722 	struct nameidata fromnd, tond;
3723 	uint64_t tondflags;
3724 	int error;
3725 
3726 again:
3727 	bwillwrite();
3728 #ifdef MAC
3729 	if (mac_vnode_check_rename_from_enabled()) {
3730 		error = kern_renameat_mac(td, oldfd, old, newfd, new, pathseg,
3731 		    &fromnd);
3732 		if (error != 0)
3733 			return (error);
3734 	} else {
3735 #endif
3736 	NDINIT_ATRIGHTS(&fromnd, DELETE, WANTPARENT | AUDITVNODE1,
3737 	    pathseg, old, oldfd, &cap_renameat_source_rights);
3738 	if ((error = namei(&fromnd)) != 0)
3739 		return (error);
3740 #ifdef MAC
3741 	}
3742 #endif
3743 	fvp = fromnd.ni_vp;
3744 	tondflags = LOCKPARENT | LOCKLEAF | NOCACHE | AUDITVNODE2;
3745 	if (fromnd.ni_vp->v_type == VDIR)
3746 		tondflags |= WILLBEDIR;
3747 	NDINIT_ATRIGHTS(&tond, RENAME, tondflags, pathseg, new, newfd,
3748 	    &cap_renameat_target_rights);
3749 	if ((error = namei(&tond)) != 0) {
3750 		/* Translate error code for rename("dir1", "dir2/."). */
3751 		if (error == EISDIR && fvp->v_type == VDIR)
3752 			error = EINVAL;
3753 		NDFREE_PNBUF(&fromnd);
3754 		vrele(fromnd.ni_dvp);
3755 		vrele(fvp);
3756 		goto out1;
3757 	}
3758 	tdvp = tond.ni_dvp;
3759 	tvp = tond.ni_vp;
3760 	error = vn_start_write(fvp, &mp, V_NOWAIT);
3761 	if (error != 0) {
3762 		NDFREE_PNBUF(&fromnd);
3763 		NDFREE_PNBUF(&tond);
3764 		if (tvp != NULL)
3765 			vput(tvp);
3766 		if (tdvp == tvp)
3767 			vrele(tdvp);
3768 		else
3769 			vput(tdvp);
3770 		vrele(fromnd.ni_dvp);
3771 		vrele(fvp);
3772 		error = vn_start_write(NULL, &mp, V_XSLEEP | V_PCATCH);
3773 		if (error != 0)
3774 			return (error);
3775 		goto again;
3776 	}
3777 	if (tvp != NULL) {
3778 		if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
3779 			error = ENOTDIR;
3780 			goto out;
3781 		} else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
3782 			error = EISDIR;
3783 			goto out;
3784 		}
3785 #ifdef CAPABILITIES
3786 		if (newfd != AT_FDCWD && (tond.ni_resflags & NIRES_ABS) == 0) {
3787 			/*
3788 			 * If the target already exists we require CAP_UNLINKAT
3789 			 * from 'newfd', when newfd was used for the lookup.
3790 			 */
3791 			error = cap_check(&tond.ni_filecaps.fc_rights,
3792 			    &cap_unlinkat_rights);
3793 			if (error != 0)
3794 				goto out;
3795 		}
3796 #endif
3797 	}
3798 	if (fvp == tdvp) {
3799 		error = EINVAL;
3800 		goto out;
3801 	}
3802 	/*
3803 	 * If the source is the same as the destination (that is, if they
3804 	 * are links to the same vnode), then there is nothing to do.
3805 	 */
3806 	if (fvp == tvp)
3807 		error = ERESTART;
3808 #ifdef MAC
3809 	else
3810 		error = mac_vnode_check_rename_to(td->td_ucred, tdvp,
3811 		    tond.ni_vp, fromnd.ni_dvp == tdvp, &tond.ni_cnd);
3812 #endif
3813 out:
3814 	if (error == 0) {
3815 		error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
3816 		    tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
3817 		NDFREE_PNBUF(&fromnd);
3818 		NDFREE_PNBUF(&tond);
3819 	} else {
3820 		NDFREE_PNBUF(&fromnd);
3821 		NDFREE_PNBUF(&tond);
3822 		if (tvp != NULL)
3823 			vput(tvp);
3824 		if (tdvp == tvp)
3825 			vrele(tdvp);
3826 		else
3827 			vput(tdvp);
3828 		vrele(fromnd.ni_dvp);
3829 		vrele(fvp);
3830 	}
3831 	vn_finished_write(mp);
3832 out1:
3833 	if (error == ERESTART)
3834 		return (0);
3835 	if (error == ERELOOKUP)
3836 		goto again;
3837 	return (error);
3838 }
3839 
3840 /*
3841  * Make a directory file.
3842  */
3843 #ifndef _SYS_SYSPROTO_H_
3844 struct mkdir_args {
3845 	char	*path;
3846 	int	mode;
3847 };
3848 #endif
3849 int
3850 sys_mkdir(struct thread *td, struct mkdir_args *uap)
3851 {
3852 
3853 	return (kern_mkdirat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
3854 	    uap->mode));
3855 }
3856 
3857 #ifndef _SYS_SYSPROTO_H_
3858 struct mkdirat_args {
3859 	int	fd;
3860 	char	*path;
3861 	mode_t	mode;
3862 };
3863 #endif
3864 int
3865 sys_mkdirat(struct thread *td, struct mkdirat_args *uap)
3866 {
3867 
3868 	return (kern_mkdirat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode));
3869 }
3870 
3871 int
3872 kern_mkdirat(struct thread *td, int fd, const char *path, enum uio_seg segflg,
3873     int mode)
3874 {
3875 	struct mount *mp;
3876 	struct vattr vattr;
3877 	struct nameidata nd;
3878 	int error;
3879 
3880 	AUDIT_ARG_MODE(mode);
3881 	NDPREINIT(&nd);
3882 restart:
3883 	bwillwrite();
3884 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | AUDITVNODE1 |
3885 	    NC_NOMAKEENTRY | NC_KEEPPOSENTRY | FAILIFEXISTS | WILLBEDIR,
3886 	    segflg, path, fd, &cap_mkdirat_rights);
3887 	if ((error = namei(&nd)) != 0)
3888 		return (error);
3889 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3890 		NDFREE_PNBUF(&nd);
3891 		vput(nd.ni_dvp);
3892 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | V_PCATCH)) != 0)
3893 			return (error);
3894 		goto restart;
3895 	}
3896 	VATTR_NULL(&vattr);
3897 	vattr.va_type = VDIR;
3898 	vattr.va_mode = (mode & ACCESSPERMS) &~ td->td_proc->p_pd->pd_cmask;
3899 #ifdef MAC
3900 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
3901 	    &vattr);
3902 	if (error != 0)
3903 		goto out;
3904 #endif
3905 	error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
3906 #ifdef MAC
3907 out:
3908 #endif
3909 	NDFREE_PNBUF(&nd);
3910 	VOP_VPUT_PAIR(nd.ni_dvp, error == 0 ? &nd.ni_vp : NULL, true);
3911 	vn_finished_write(mp);
3912 	if (error == ERELOOKUP)
3913 		goto restart;
3914 	return (error);
3915 }
3916 
3917 /*
3918  * Remove a directory file.
3919  */
3920 #ifndef _SYS_SYSPROTO_H_
3921 struct rmdir_args {
3922 	char	*path;
3923 };
3924 #endif
3925 int
3926 sys_rmdir(struct thread *td, struct rmdir_args *uap)
3927 {
3928 
3929 	return (kern_frmdirat(td, AT_FDCWD, uap->path, FD_NONE, UIO_USERSPACE,
3930 	    0));
3931 }
3932 
3933 int
3934 kern_frmdirat(struct thread *td, int dfd, const char *path, int fd,
3935     enum uio_seg pathseg, int flag)
3936 {
3937 	struct mount *mp;
3938 	struct vnode *vp;
3939 	struct file *fp;
3940 	struct nameidata nd;
3941 	cap_rights_t rights;
3942 	int error;
3943 
3944 	fp = NULL;
3945 	if (fd != FD_NONE) {
3946 		error = getvnode(td, fd, cap_rights_init_one(&rights,
3947 		    CAP_LOOKUP), &fp);
3948 		if (error != 0)
3949 			return (error);
3950 	}
3951 
3952 	NDPREINIT(&nd);
3953 restart:
3954 	bwillwrite();
3955 	NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1 |
3956 	    at2cnpflags(flag, AT_RESOLVE_BENEATH),
3957 	    pathseg, path, dfd, &cap_unlinkat_rights);
3958 	if ((error = namei(&nd)) != 0)
3959 		goto fdout;
3960 	vp = nd.ni_vp;
3961 	if (vp->v_type != VDIR) {
3962 		error = ENOTDIR;
3963 		goto out;
3964 	}
3965 	/*
3966 	 * No rmdir "." please.
3967 	 */
3968 	if (nd.ni_dvp == vp) {
3969 		error = EINVAL;
3970 		goto out;
3971 	}
3972 	/*
3973 	 * The root of a mounted filesystem cannot be deleted.
3974 	 */
3975 	if (vp->v_vflag & VV_ROOT) {
3976 		error = EBUSY;
3977 		goto out;
3978 	}
3979 
3980 	if (fp != NULL && fp->f_vnode != vp) {
3981 		if (VN_IS_DOOMED(fp->f_vnode))
3982 			error = EBADF;
3983 		else
3984 			error = EDEADLK;
3985 		goto out;
3986 	}
3987 
3988 #ifdef MAC
3989 	error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
3990 	    &nd.ni_cnd);
3991 	if (error != 0)
3992 		goto out;
3993 #endif
3994 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3995 		NDFREE_PNBUF(&nd);
3996 		vput(vp);
3997 		if (nd.ni_dvp == vp)
3998 			vrele(nd.ni_dvp);
3999 		else
4000 			vput(nd.ni_dvp);
4001 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | V_PCATCH)) != 0)
4002 			goto fdout;
4003 		goto restart;
4004 	}
4005 	vfs_notify_upper(vp, VFS_NOTIFY_UPPER_UNLINK);
4006 	error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
4007 	vn_finished_write(mp);
4008 out:
4009 	NDFREE_PNBUF(&nd);
4010 	vput(vp);
4011 	if (nd.ni_dvp == vp)
4012 		vrele(nd.ni_dvp);
4013 	else
4014 		vput(nd.ni_dvp);
4015 	if (error == ERELOOKUP)
4016 		goto restart;
4017 fdout:
4018 	if (fp != NULL)
4019 		fdrop(fp, td);
4020 	return (error);
4021 }
4022 
4023 #if defined(COMPAT_43) || defined(COMPAT_FREEBSD11)
4024 int
4025 freebsd11_kern_getdirentries(struct thread *td, int fd, char *ubuf, u_int count,
4026     long *basep, void (*func)(struct freebsd11_dirent *))
4027 {
4028 	struct freebsd11_dirent dstdp;
4029 	struct dirent *dp, *edp;
4030 	char *dirbuf;
4031 	off_t base;
4032 	ssize_t resid, ucount;
4033 	int error;
4034 
4035 	/* XXX arbitrary sanity limit on `count'. */
4036 	count = min(count, 64 * 1024);
4037 
4038 	dirbuf = malloc(count, M_TEMP, M_WAITOK);
4039 
4040 	error = kern_getdirentries(td, fd, dirbuf, count, &base, &resid,
4041 	    UIO_SYSSPACE);
4042 	if (error != 0)
4043 		goto done;
4044 	if (basep != NULL)
4045 		*basep = base;
4046 
4047 	ucount = 0;
4048 	for (dp = (struct dirent *)dirbuf,
4049 	    edp = (struct dirent *)&dirbuf[count - resid];
4050 	    ucount < count && dp < edp; ) {
4051 		if (dp->d_reclen == 0)
4052 			break;
4053 		MPASS(dp->d_reclen >= _GENERIC_DIRLEN(0));
4054 		if (dp->d_namlen >= sizeof(dstdp.d_name))
4055 			continue;
4056 		dstdp.d_type = dp->d_type;
4057 		dstdp.d_namlen = dp->d_namlen;
4058 		dstdp.d_fileno = dp->d_fileno;		/* truncate */
4059 		if (dstdp.d_fileno != dp->d_fileno) {
4060 			switch (ino64_trunc_error) {
4061 			default:
4062 			case 0:
4063 				break;
4064 			case 1:
4065 				error = EOVERFLOW;
4066 				goto done;
4067 			case 2:
4068 				dstdp.d_fileno = UINT32_MAX;
4069 				break;
4070 			}
4071 		}
4072 		dstdp.d_reclen = sizeof(dstdp) - sizeof(dstdp.d_name) +
4073 		    ((dp->d_namlen + 1 + 3) &~ 3);
4074 		bcopy(dp->d_name, dstdp.d_name, dstdp.d_namlen);
4075 		bzero(dstdp.d_name + dstdp.d_namlen,
4076 		    dstdp.d_reclen - offsetof(struct freebsd11_dirent, d_name) -
4077 		    dstdp.d_namlen);
4078 		MPASS(dstdp.d_reclen <= dp->d_reclen);
4079 		MPASS(ucount + dstdp.d_reclen <= count);
4080 		if (func != NULL)
4081 			func(&dstdp);
4082 		error = copyout(&dstdp, ubuf + ucount, dstdp.d_reclen);
4083 		if (error != 0)
4084 			break;
4085 		dp = (struct dirent *)((char *)dp + dp->d_reclen);
4086 		ucount += dstdp.d_reclen;
4087 	}
4088 
4089 done:
4090 	free(dirbuf, M_TEMP);
4091 	if (error == 0)
4092 		td->td_retval[0] = ucount;
4093 	return (error);
4094 }
4095 #endif /* COMPAT */
4096 
4097 #ifdef COMPAT_43
4098 static void
4099 ogetdirentries_cvt(struct freebsd11_dirent *dp)
4100 {
4101 #if (BYTE_ORDER == LITTLE_ENDIAN)
4102 	/*
4103 	 * The expected low byte of dp->d_namlen is our dp->d_type.
4104 	 * The high MBZ byte of dp->d_namlen is our dp->d_namlen.
4105 	 */
4106 	dp->d_type = dp->d_namlen;
4107 	dp->d_namlen = 0;
4108 #else
4109 	/*
4110 	 * The dp->d_type is the high byte of the expected dp->d_namlen,
4111 	 * so must be zero'ed.
4112 	 */
4113 	dp->d_type = 0;
4114 #endif
4115 }
4116 
4117 /*
4118  * Read a block of directory entries in a filesystem independent format.
4119  */
4120 #ifndef _SYS_SYSPROTO_H_
4121 struct ogetdirentries_args {
4122 	int	fd;
4123 	char	*buf;
4124 	u_int	count;
4125 	long	*basep;
4126 };
4127 #endif
4128 int
4129 ogetdirentries(struct thread *td, struct ogetdirentries_args *uap)
4130 {
4131 	long loff;
4132 	int error;
4133 
4134 	error = kern_ogetdirentries(td, uap, &loff);
4135 	if (error == 0)
4136 		error = copyout(&loff, uap->basep, sizeof(long));
4137 	return (error);
4138 }
4139 
4140 int
4141 kern_ogetdirentries(struct thread *td, struct ogetdirentries_args *uap,
4142     long *ploff)
4143 {
4144 	long base;
4145 	int error;
4146 
4147 	/* XXX arbitrary sanity limit on `count'. */
4148 	if (uap->count > 64 * 1024)
4149 		return (EINVAL);
4150 
4151 	error = freebsd11_kern_getdirentries(td, uap->fd, uap->buf, uap->count,
4152 	    &base, ogetdirentries_cvt);
4153 
4154 	if (error == 0 && uap->basep != NULL)
4155 		error = copyout(&base, uap->basep, sizeof(long));
4156 
4157 	return (error);
4158 }
4159 #endif /* COMPAT_43 */
4160 
4161 #if defined(COMPAT_FREEBSD11)
4162 #ifndef _SYS_SYSPROTO_H_
4163 struct freebsd11_getdirentries_args {
4164 	int	fd;
4165 	char	*buf;
4166 	u_int	count;
4167 	long	*basep;
4168 };
4169 #endif
4170 int
4171 freebsd11_getdirentries(struct thread *td,
4172     struct freebsd11_getdirentries_args *uap)
4173 {
4174 	long base;
4175 	int error;
4176 
4177 	error = freebsd11_kern_getdirentries(td, uap->fd, uap->buf, uap->count,
4178 	    &base, NULL);
4179 
4180 	if (error == 0 && uap->basep != NULL)
4181 		error = copyout(&base, uap->basep, sizeof(long));
4182 	return (error);
4183 }
4184 
4185 int
4186 freebsd11_getdents(struct thread *td, struct freebsd11_getdents_args *uap)
4187 {
4188 	struct freebsd11_getdirentries_args ap;
4189 
4190 	ap.fd = uap->fd;
4191 	ap.buf = uap->buf;
4192 	ap.count = uap->count;
4193 	ap.basep = NULL;
4194 	return (freebsd11_getdirentries(td, &ap));
4195 }
4196 #endif /* COMPAT_FREEBSD11 */
4197 
4198 /*
4199  * Read a block of directory entries in a filesystem independent format.
4200  */
4201 int
4202 sys_getdirentries(struct thread *td, struct getdirentries_args *uap)
4203 {
4204 	off_t base;
4205 	int error;
4206 
4207 	error = kern_getdirentries(td, uap->fd, uap->buf, uap->count, &base,
4208 	    NULL, UIO_USERSPACE);
4209 	if (error != 0)
4210 		return (error);
4211 	if (uap->basep != NULL)
4212 		error = copyout(&base, uap->basep, sizeof(off_t));
4213 	return (error);
4214 }
4215 
4216 int
4217 kern_getdirentries(struct thread *td, int fd, char *buf, size_t count,
4218     off_t *basep, ssize_t *residp, enum uio_seg bufseg)
4219 {
4220 	struct vnode *vp;
4221 	struct file *fp;
4222 	struct uio auio;
4223 	struct iovec aiov;
4224 	off_t loff;
4225 	int error, eofflag;
4226 	off_t foffset;
4227 
4228 	AUDIT_ARG_FD(fd);
4229 	if (count > IOSIZE_MAX)
4230 		return (EINVAL);
4231 	auio.uio_resid = count;
4232 	error = getvnode(td, fd, &cap_read_rights, &fp);
4233 	if (error != 0)
4234 		return (error);
4235 	if ((fp->f_flag & FREAD) == 0) {
4236 		fdrop(fp, td);
4237 		return (EBADF);
4238 	}
4239 	vp = fp->f_vnode;
4240 	foffset = foffset_lock(fp, 0);
4241 unionread:
4242 	if (vp->v_type != VDIR) {
4243 		error = EINVAL;
4244 		goto fail;
4245 	}
4246 	if (__predict_false((vp->v_vflag & VV_UNLINKED) != 0)) {
4247 		error = ENOENT;
4248 		goto fail;
4249 	}
4250 	aiov.iov_base = buf;
4251 	aiov.iov_len = count;
4252 	auio.uio_iov = &aiov;
4253 	auio.uio_iovcnt = 1;
4254 	auio.uio_rw = UIO_READ;
4255 	auio.uio_segflg = bufseg;
4256 	auio.uio_td = td;
4257 	vn_lock(vp, LK_SHARED | LK_RETRY);
4258 	AUDIT_ARG_VNODE1(vp);
4259 	loff = auio.uio_offset = foffset;
4260 #ifdef MAC
4261 	error = mac_vnode_check_readdir(td->td_ucred, vp);
4262 	if (error == 0)
4263 #endif
4264 		error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, NULL,
4265 		    NULL);
4266 	foffset = auio.uio_offset;
4267 	if (error != 0) {
4268 		VOP_UNLOCK(vp);
4269 		goto fail;
4270 	}
4271 	if (count == auio.uio_resid &&
4272 	    (vp->v_vflag & VV_ROOT) &&
4273 	    (vp->v_mount->mnt_flag & MNT_UNION)) {
4274 		struct vnode *tvp = vp;
4275 
4276 		vp = vp->v_mount->mnt_vnodecovered;
4277 		VREF(vp);
4278 		fp->f_vnode = vp;
4279 		foffset = 0;
4280 		vput(tvp);
4281 		goto unionread;
4282 	}
4283 	VOP_UNLOCK(vp);
4284 	*basep = loff;
4285 	if (residp != NULL)
4286 		*residp = auio.uio_resid;
4287 	td->td_retval[0] = count - auio.uio_resid;
4288 fail:
4289 	foffset_unlock(fp, foffset, 0);
4290 	fdrop(fp, td);
4291 	return (error);
4292 }
4293 
4294 /*
4295  * Set the mode mask for creation of filesystem nodes.
4296  */
4297 #ifndef _SYS_SYSPROTO_H_
4298 struct umask_args {
4299 	int	newmask;
4300 };
4301 #endif
4302 int
4303 sys_umask(struct thread *td, struct umask_args *uap)
4304 {
4305 	struct pwddesc *pdp;
4306 
4307 	pdp = td->td_proc->p_pd;
4308 	PWDDESC_XLOCK(pdp);
4309 	td->td_retval[0] = pdp->pd_cmask;
4310 	pdp->pd_cmask = uap->newmask & ALLPERMS;
4311 	PWDDESC_XUNLOCK(pdp);
4312 	return (0);
4313 }
4314 
4315 /*
4316  * Void all references to file by ripping underlying filesystem away from
4317  * vnode.
4318  */
4319 #ifndef _SYS_SYSPROTO_H_
4320 struct revoke_args {
4321 	char	*path;
4322 };
4323 #endif
4324 int
4325 sys_revoke(struct thread *td, struct revoke_args *uap)
4326 {
4327 	struct vnode *vp;
4328 	struct vattr vattr;
4329 	struct nameidata nd;
4330 	int error;
4331 
4332 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
4333 	    uap->path);
4334 	if ((error = namei(&nd)) != 0)
4335 		return (error);
4336 	vp = nd.ni_vp;
4337 	NDFREE_PNBUF(&nd);
4338 	if (vp->v_type != VCHR || vp->v_rdev == NULL) {
4339 		error = EINVAL;
4340 		goto out;
4341 	}
4342 #ifdef MAC
4343 	error = mac_vnode_check_revoke(td->td_ucred, vp);
4344 	if (error != 0)
4345 		goto out;
4346 #endif
4347 	error = VOP_GETATTR(vp, &vattr, td->td_ucred);
4348 	if (error != 0)
4349 		goto out;
4350 	if (td->td_ucred->cr_uid != vattr.va_uid) {
4351 		error = priv_check(td, PRIV_VFS_ADMIN);
4352 		if (error != 0)
4353 			goto out;
4354 	}
4355 	if (devfs_usecount(vp) > 0)
4356 		VOP_REVOKE(vp, REVOKEALL);
4357 out:
4358 	vput(vp);
4359 	return (error);
4360 }
4361 
4362 /*
4363  * This variant of getvnode() allows O_PATH files.  Caller should
4364  * ensure that returned file and vnode are only used for compatible
4365  * semantics.
4366  */
4367 int
4368 getvnode_path(struct thread *td, int fd, cap_rights_t *rightsp,
4369     struct file **fpp)
4370 {
4371 	struct file *fp;
4372 	int error;
4373 
4374 	error = fget_unlocked(td, fd, rightsp, &fp);
4375 	if (error != 0)
4376 		return (error);
4377 
4378 	/*
4379 	 * The file could be not of the vnode type, or it may be not
4380 	 * yet fully initialized, in which case the f_vnode pointer
4381 	 * may be set, but f_ops is still badfileops.  E.g.,
4382 	 * devfs_open() transiently create such situation to
4383 	 * facilitate csw d_fdopen().
4384 	 *
4385 	 * Dupfdopen() handling in kern_openat() installs the
4386 	 * half-baked file into the process descriptor table, allowing
4387 	 * other thread to dereference it. Guard against the race by
4388 	 * checking f_ops.
4389 	 */
4390 	if (__predict_false(fp->f_vnode == NULL || fp->f_ops == &badfileops)) {
4391 		fdrop(fp, td);
4392 		*fpp = NULL;
4393 		return (EINVAL);
4394 	}
4395 
4396 	*fpp = fp;
4397 	return (0);
4398 }
4399 
4400 /*
4401  * Convert a user file descriptor to a kernel file entry and check
4402  * that, if it is a capability, the correct rights are present.
4403  * A reference on the file entry is held upon returning.
4404  */
4405 int
4406 getvnode(struct thread *td, int fd, cap_rights_t *rightsp, struct file **fpp)
4407 {
4408 	int error;
4409 
4410 	error = getvnode_path(td, fd, rightsp, fpp);
4411 	if (__predict_false(error != 0))
4412 		return (error);
4413 
4414 	/*
4415 	 * Filter out O_PATH file descriptors, most getvnode() callers
4416 	 * do not call fo_ methods.
4417 	 */
4418 	if (__predict_false((*fpp)->f_ops == &path_fileops)) {
4419 		fdrop(*fpp, td);
4420 		*fpp = NULL;
4421 		error = EBADF;
4422 	}
4423 
4424 	return (error);
4425 }
4426 
4427 /*
4428  * Get an (NFS) file handle.
4429  */
4430 #ifndef _SYS_SYSPROTO_H_
4431 struct lgetfh_args {
4432 	char *fname;
4433 	fhandle_t *fhp;
4434 };
4435 #endif
4436 int
4437 sys_lgetfh(struct thread *td, struct lgetfh_args *uap)
4438 {
4439 
4440 	return (kern_getfhat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->fname,
4441 	    UIO_USERSPACE, uap->fhp, UIO_USERSPACE));
4442 }
4443 
4444 #ifndef _SYS_SYSPROTO_H_
4445 struct getfh_args {
4446 	char *fname;
4447 	fhandle_t *fhp;
4448 };
4449 #endif
4450 int
4451 sys_getfh(struct thread *td, struct getfh_args *uap)
4452 {
4453 
4454 	return (kern_getfhat(td, 0, AT_FDCWD, uap->fname, UIO_USERSPACE,
4455 	    uap->fhp, UIO_USERSPACE));
4456 }
4457 
4458 /*
4459  * syscall for the rpc.lockd to use to translate an open descriptor into
4460  * a NFS file handle.
4461  *
4462  * warning: do not remove the priv_check() call or this becomes one giant
4463  * security hole.
4464  */
4465 #ifndef _SYS_SYSPROTO_H_
4466 struct getfhat_args {
4467 	int fd;
4468 	char *path;
4469 	fhandle_t *fhp;
4470 	int flags;
4471 };
4472 #endif
4473 int
4474 sys_getfhat(struct thread *td, struct getfhat_args *uap)
4475 {
4476 
4477 	return (kern_getfhat(td, uap->flags, uap->fd, uap->path, UIO_USERSPACE,
4478 	    uap->fhp, UIO_USERSPACE));
4479 }
4480 
4481 int
4482 kern_getfhat(struct thread *td, int flags, int fd, const char *path,
4483     enum uio_seg pathseg, fhandle_t *fhp, enum uio_seg fhseg)
4484 {
4485 	struct nameidata nd;
4486 	fhandle_t fh;
4487 	struct vnode *vp;
4488 	int error;
4489 
4490 	if ((flags & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH)) != 0)
4491 		return (EINVAL);
4492 	error = priv_check(td, PRIV_VFS_GETFH);
4493 	if (error != 0)
4494 		return (error);
4495 	NDINIT_AT(&nd, LOOKUP, at2cnpflags(flags, AT_SYMLINK_NOFOLLOW |
4496 	    AT_RESOLVE_BENEATH) | LOCKLEAF | AUDITVNODE1, pathseg, path,
4497 	    fd);
4498 	error = namei(&nd);
4499 	if (error != 0)
4500 		return (error);
4501 	NDFREE_PNBUF(&nd);
4502 	vp = nd.ni_vp;
4503 	bzero(&fh, sizeof(fh));
4504 	fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
4505 	error = VOP_VPTOFH(vp, &fh.fh_fid);
4506 	vput(vp);
4507 	if (error == 0) {
4508 		if (fhseg == UIO_USERSPACE)
4509 			error = copyout(&fh, fhp, sizeof (fh));
4510 		else
4511 			memcpy(fhp, &fh, sizeof(fh));
4512 	}
4513 	return (error);
4514 }
4515 
4516 #ifndef _SYS_SYSPROTO_H_
4517 struct fhlink_args {
4518 	fhandle_t *fhp;
4519 	const char *to;
4520 };
4521 #endif
4522 int
4523 sys_fhlink(struct thread *td, struct fhlink_args *uap)
4524 {
4525 
4526 	return (kern_fhlinkat(td, AT_FDCWD, uap->to, UIO_USERSPACE, uap->fhp));
4527 }
4528 
4529 #ifndef _SYS_SYSPROTO_H_
4530 struct fhlinkat_args {
4531 	fhandle_t *fhp;
4532 	int tofd;
4533 	const char *to;
4534 };
4535 #endif
4536 int
4537 sys_fhlinkat(struct thread *td, struct fhlinkat_args *uap)
4538 {
4539 
4540 	return (kern_fhlinkat(td, uap->tofd, uap->to, UIO_USERSPACE, uap->fhp));
4541 }
4542 
4543 static int
4544 kern_fhlinkat(struct thread *td, int fd, const char *path,
4545     enum uio_seg pathseg, fhandle_t *fhp)
4546 {
4547 	fhandle_t fh;
4548 	struct mount *mp;
4549 	struct vnode *vp;
4550 	int error;
4551 
4552 	error = priv_check(td, PRIV_VFS_GETFH);
4553 	if (error != 0)
4554 		return (error);
4555 	error = copyin(fhp, &fh, sizeof(fh));
4556 	if (error != 0)
4557 		return (error);
4558 	do {
4559 		bwillwrite();
4560 		if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4561 			return (ESTALE);
4562 		error = VFS_FHTOVP(mp, &fh.fh_fid, LK_SHARED, &vp);
4563 		vfs_unbusy(mp);
4564 		if (error != 0)
4565 			return (error);
4566 		VOP_UNLOCK(vp);
4567 		error = kern_linkat_vp(td, vp, fd, path, pathseg);
4568 	} while (error == EAGAIN || error == ERELOOKUP);
4569 	return (error);
4570 }
4571 
4572 #ifndef _SYS_SYSPROTO_H_
4573 struct fhreadlink_args {
4574 	fhandle_t *fhp;
4575 	char *buf;
4576 	size_t bufsize;
4577 };
4578 #endif
4579 int
4580 sys_fhreadlink(struct thread *td, struct fhreadlink_args *uap)
4581 {
4582 	fhandle_t fh;
4583 	struct mount *mp;
4584 	struct vnode *vp;
4585 	int error;
4586 
4587 	error = priv_check(td, PRIV_VFS_GETFH);
4588 	if (error != 0)
4589 		return (error);
4590 	if (uap->bufsize > IOSIZE_MAX)
4591 		return (EINVAL);
4592 	error = copyin(uap->fhp, &fh, sizeof(fh));
4593 	if (error != 0)
4594 		return (error);
4595 	if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4596 		return (ESTALE);
4597 	error = VFS_FHTOVP(mp, &fh.fh_fid, LK_SHARED, &vp);
4598 	vfs_unbusy(mp);
4599 	if (error != 0)
4600 		return (error);
4601 	error = kern_readlink_vp(vp, uap->buf, UIO_USERSPACE, uap->bufsize, td);
4602 	vput(vp);
4603 	return (error);
4604 }
4605 
4606 /*
4607  * syscall for the rpc.lockd to use to translate a NFS file handle into an
4608  * open descriptor.
4609  *
4610  * warning: do not remove the priv_check() call or this becomes one giant
4611  * security hole.
4612  */
4613 #ifndef _SYS_SYSPROTO_H_
4614 struct fhopen_args {
4615 	const struct fhandle *u_fhp;
4616 	int flags;
4617 };
4618 #endif
4619 int
4620 sys_fhopen(struct thread *td, struct fhopen_args *uap)
4621 {
4622 	return (kern_fhopen(td, uap->u_fhp, uap->flags));
4623 }
4624 
4625 int
4626 kern_fhopen(struct thread *td, const struct fhandle *u_fhp, int flags)
4627 {
4628 	struct mount *mp;
4629 	struct vnode *vp;
4630 	struct fhandle fhp;
4631 	struct file *fp;
4632 	int fmode, error;
4633 	int indx;
4634 
4635 	error = priv_check(td, PRIV_VFS_FHOPEN);
4636 	if (error != 0)
4637 		return (error);
4638 	indx = -1;
4639 	fmode = FFLAGS(flags);
4640 	/* why not allow a non-read/write open for our lockd? */
4641 	if (((fmode & (FREAD | FWRITE)) == 0) || (fmode & O_CREAT))
4642 		return (EINVAL);
4643 	error = copyin(u_fhp, &fhp, sizeof(fhp));
4644 	if (error != 0)
4645 		return(error);
4646 	/* find the mount point */
4647 	mp = vfs_busyfs(&fhp.fh_fsid);
4648 	if (mp == NULL)
4649 		return (ESTALE);
4650 	/* now give me my vnode, it gets returned to me locked */
4651 	error = VFS_FHTOVP(mp, &fhp.fh_fid, LK_EXCLUSIVE, &vp);
4652 	vfs_unbusy(mp);
4653 	if (error != 0)
4654 		return (error);
4655 
4656 	error = falloc_noinstall(td, &fp);
4657 	if (error != 0) {
4658 		vput(vp);
4659 		return (error);
4660 	}
4661 	/*
4662 	 * An extra reference on `fp' has been held for us by
4663 	 * falloc_noinstall().
4664 	 */
4665 
4666 #ifdef INVARIANTS
4667 	td->td_dupfd = -1;
4668 #endif
4669 	error = vn_open_vnode(vp, fmode, td->td_ucred, td, fp);
4670 	if (error != 0) {
4671 		KASSERT(fp->f_ops == &badfileops,
4672 		    ("VOP_OPEN in fhopen() set f_ops"));
4673 		KASSERT(td->td_dupfd < 0,
4674 		    ("fhopen() encountered fdopen()"));
4675 
4676 		vput(vp);
4677 		goto bad;
4678 	}
4679 #ifdef INVARIANTS
4680 	td->td_dupfd = 0;
4681 #endif
4682 	fp->f_vnode = vp;
4683 	finit_vnode(fp, fmode, NULL, &vnops);
4684 	VOP_UNLOCK(vp);
4685 	if ((fmode & O_TRUNC) != 0) {
4686 		error = fo_truncate(fp, 0, td->td_ucred, td);
4687 		if (error != 0)
4688 			goto bad;
4689 	}
4690 
4691 	error = finstall(td, fp, &indx, fmode, NULL);
4692 bad:
4693 	fdrop(fp, td);
4694 	td->td_retval[0] = indx;
4695 	return (error);
4696 }
4697 
4698 /*
4699  * Stat an (NFS) file handle.
4700  */
4701 #ifndef _SYS_SYSPROTO_H_
4702 struct fhstat_args {
4703 	struct fhandle *u_fhp;
4704 	struct stat *sb;
4705 };
4706 #endif
4707 int
4708 sys_fhstat(struct thread *td, struct fhstat_args *uap)
4709 {
4710 	struct stat sb;
4711 	struct fhandle fh;
4712 	int error;
4713 
4714 	error = copyin(uap->u_fhp, &fh, sizeof(fh));
4715 	if (error != 0)
4716 		return (error);
4717 	error = kern_fhstat(td, fh, &sb);
4718 	if (error == 0)
4719 		error = copyout(&sb, uap->sb, sizeof(sb));
4720 	return (error);
4721 }
4722 
4723 int
4724 kern_fhstat(struct thread *td, struct fhandle fh, struct stat *sb)
4725 {
4726 	struct mount *mp;
4727 	struct vnode *vp;
4728 	int error;
4729 
4730 	error = priv_check(td, PRIV_VFS_FHSTAT);
4731 	if (error != 0)
4732 		return (error);
4733 	if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4734 		return (ESTALE);
4735 	error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4736 	vfs_unbusy(mp);
4737 	if (error != 0)
4738 		return (error);
4739 	error = VOP_STAT(vp, sb, td->td_ucred, NOCRED);
4740 	vput(vp);
4741 	return (error);
4742 }
4743 
4744 /*
4745  * Implement fstatfs() for (NFS) file handles.
4746  */
4747 #ifndef _SYS_SYSPROTO_H_
4748 struct fhstatfs_args {
4749 	struct fhandle *u_fhp;
4750 	struct statfs *buf;
4751 };
4752 #endif
4753 int
4754 sys_fhstatfs(struct thread *td, struct fhstatfs_args *uap)
4755 {
4756 	struct statfs *sfp;
4757 	fhandle_t fh;
4758 	int error;
4759 
4760 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
4761 	if (error != 0)
4762 		return (error);
4763 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
4764 	error = kern_fhstatfs(td, fh, sfp);
4765 	if (error == 0)
4766 		error = copyout(sfp, uap->buf, sizeof(*sfp));
4767 	free(sfp, M_STATFS);
4768 	return (error);
4769 }
4770 
4771 int
4772 kern_fhstatfs(struct thread *td, fhandle_t fh, struct statfs *buf)
4773 {
4774 	struct mount *mp;
4775 	struct vnode *vp;
4776 	int error;
4777 
4778 	error = priv_check(td, PRIV_VFS_FHSTATFS);
4779 	if (error != 0)
4780 		return (error);
4781 	if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4782 		return (ESTALE);
4783 	error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4784 	if (error != 0) {
4785 		vfs_unbusy(mp);
4786 		return (error);
4787 	}
4788 	vput(vp);
4789 	error = prison_canseemount(td->td_ucred, mp);
4790 	if (error != 0)
4791 		goto out;
4792 #ifdef MAC
4793 	error = mac_mount_check_stat(td->td_ucred, mp);
4794 	if (error != 0)
4795 		goto out;
4796 #endif
4797 	error = VFS_STATFS(mp, buf);
4798 out:
4799 	vfs_unbusy(mp);
4800 	return (error);
4801 }
4802 
4803 /*
4804  * Unlike madvise(2), we do not make a best effort to remember every
4805  * possible caching hint.  Instead, we remember the last setting with
4806  * the exception that we will allow POSIX_FADV_NORMAL to adjust the
4807  * region of any current setting.
4808  */
4809 int
4810 kern_posix_fadvise(struct thread *td, int fd, off_t offset, off_t len,
4811     int advice)
4812 {
4813 	struct fadvise_info *fa, *new;
4814 	struct file *fp;
4815 	struct vnode *vp;
4816 	off_t end;
4817 	int error;
4818 
4819 	if (offset < 0 || len < 0 || offset > OFF_MAX - len)
4820 		return (EINVAL);
4821 	AUDIT_ARG_VALUE(advice);
4822 	switch (advice) {
4823 	case POSIX_FADV_SEQUENTIAL:
4824 	case POSIX_FADV_RANDOM:
4825 	case POSIX_FADV_NOREUSE:
4826 		new = malloc(sizeof(*fa), M_FADVISE, M_WAITOK);
4827 		break;
4828 	case POSIX_FADV_NORMAL:
4829 	case POSIX_FADV_WILLNEED:
4830 	case POSIX_FADV_DONTNEED:
4831 		new = NULL;
4832 		break;
4833 	default:
4834 		return (EINVAL);
4835 	}
4836 	/* XXX: CAP_POSIX_FADVISE? */
4837 	AUDIT_ARG_FD(fd);
4838 	error = fget(td, fd, &cap_no_rights, &fp);
4839 	if (error != 0)
4840 		goto out;
4841 	AUDIT_ARG_FILE(td->td_proc, fp);
4842 	if ((fp->f_ops->fo_flags & DFLAG_SEEKABLE) == 0) {
4843 		error = ESPIPE;
4844 		goto out;
4845 	}
4846 	if (fp->f_type != DTYPE_VNODE) {
4847 		error = ENODEV;
4848 		goto out;
4849 	}
4850 	vp = fp->f_vnode;
4851 	if (vp->v_type != VREG) {
4852 		error = ENODEV;
4853 		goto out;
4854 	}
4855 	if (len == 0)
4856 		end = OFF_MAX;
4857 	else
4858 		end = offset + len - 1;
4859 	switch (advice) {
4860 	case POSIX_FADV_SEQUENTIAL:
4861 	case POSIX_FADV_RANDOM:
4862 	case POSIX_FADV_NOREUSE:
4863 		/*
4864 		 * Try to merge any existing non-standard region with
4865 		 * this new region if possible, otherwise create a new
4866 		 * non-standard region for this request.
4867 		 */
4868 		mtx_pool_lock(mtxpool_sleep, fp);
4869 		fa = fp->f_advice;
4870 		if (fa != NULL && fa->fa_advice == advice &&
4871 		    ((fa->fa_start <= end && fa->fa_end >= offset) ||
4872 		    (end != OFF_MAX && fa->fa_start == end + 1) ||
4873 		    (fa->fa_end != OFF_MAX && fa->fa_end + 1 == offset))) {
4874 			if (offset < fa->fa_start)
4875 				fa->fa_start = offset;
4876 			if (end > fa->fa_end)
4877 				fa->fa_end = end;
4878 		} else {
4879 			new->fa_advice = advice;
4880 			new->fa_start = offset;
4881 			new->fa_end = end;
4882 			fp->f_advice = new;
4883 			new = fa;
4884 		}
4885 		mtx_pool_unlock(mtxpool_sleep, fp);
4886 		break;
4887 	case POSIX_FADV_NORMAL:
4888 		/*
4889 		 * If a the "normal" region overlaps with an existing
4890 		 * non-standard region, trim or remove the
4891 		 * non-standard region.
4892 		 */
4893 		mtx_pool_lock(mtxpool_sleep, fp);
4894 		fa = fp->f_advice;
4895 		if (fa != NULL) {
4896 			if (offset <= fa->fa_start && end >= fa->fa_end) {
4897 				new = fa;
4898 				fp->f_advice = NULL;
4899 			} else if (offset <= fa->fa_start &&
4900 			    end >= fa->fa_start)
4901 				fa->fa_start = end + 1;
4902 			else if (offset <= fa->fa_end && end >= fa->fa_end)
4903 				fa->fa_end = offset - 1;
4904 			else if (offset >= fa->fa_start && end <= fa->fa_end) {
4905 				/*
4906 				 * If the "normal" region is a middle
4907 				 * portion of the existing
4908 				 * non-standard region, just remove
4909 				 * the whole thing rather than picking
4910 				 * one side or the other to
4911 				 * preserve.
4912 				 */
4913 				new = fa;
4914 				fp->f_advice = NULL;
4915 			}
4916 		}
4917 		mtx_pool_unlock(mtxpool_sleep, fp);
4918 		break;
4919 	case POSIX_FADV_WILLNEED:
4920 	case POSIX_FADV_DONTNEED:
4921 		error = VOP_ADVISE(vp, offset, end, advice);
4922 		break;
4923 	}
4924 out:
4925 	if (fp != NULL)
4926 		fdrop(fp, td);
4927 	free(new, M_FADVISE);
4928 	return (error);
4929 }
4930 
4931 int
4932 sys_posix_fadvise(struct thread *td, struct posix_fadvise_args *uap)
4933 {
4934 	int error;
4935 
4936 	error = kern_posix_fadvise(td, uap->fd, uap->offset, uap->len,
4937 	    uap->advice);
4938 	return (kern_posix_error(td, error));
4939 }
4940 
4941 int
4942 kern_copy_file_range(struct thread *td, int infd, off_t *inoffp, int outfd,
4943     off_t *outoffp, size_t len, unsigned int flags)
4944 {
4945 	struct file *infp, *outfp;
4946 	struct vnode *invp, *outvp;
4947 	int error;
4948 	size_t retlen;
4949 	void *rl_rcookie, *rl_wcookie;
4950 	off_t savinoff, savoutoff;
4951 
4952 	infp = outfp = NULL;
4953 	rl_rcookie = rl_wcookie = NULL;
4954 	savinoff = -1;
4955 	error = 0;
4956 	retlen = 0;
4957 
4958 	if (flags != 0) {
4959 		error = EINVAL;
4960 		goto out;
4961 	}
4962 	if (len > SSIZE_MAX)
4963 		/*
4964 		 * Although the len argument is size_t, the return argument
4965 		 * is ssize_t (which is signed).  Therefore a size that won't
4966 		 * fit in ssize_t can't be returned.
4967 		 */
4968 		len = SSIZE_MAX;
4969 
4970 	/* Get the file structures for the file descriptors. */
4971 	error = fget_read(td, infd,
4972 	    inoffp != NULL ? &cap_pread_rights : &cap_read_rights, &infp);
4973 	if (error != 0)
4974 		goto out;
4975 	if (infp->f_ops == &badfileops) {
4976 		error = EBADF;
4977 		goto out;
4978 	}
4979 	if (infp->f_vnode == NULL) {
4980 		error = EINVAL;
4981 		goto out;
4982 	}
4983 	error = fget_write(td, outfd,
4984 	    outoffp != NULL ? &cap_pwrite_rights : &cap_write_rights, &outfp);
4985 	if (error != 0)
4986 		goto out;
4987 	if (outfp->f_ops == &badfileops) {
4988 		error = EBADF;
4989 		goto out;
4990 	}
4991 	if (outfp->f_vnode == NULL) {
4992 		error = EINVAL;
4993 		goto out;
4994 	}
4995 
4996 	/* Set the offset pointers to the correct place. */
4997 	if (inoffp == NULL)
4998 		inoffp = &infp->f_offset;
4999 	if (outoffp == NULL)
5000 		outoffp = &outfp->f_offset;
5001 	savinoff = *inoffp;
5002 	savoutoff = *outoffp;
5003 
5004 	invp = infp->f_vnode;
5005 	outvp = outfp->f_vnode;
5006 	/* Sanity check the f_flag bits. */
5007 	if ((outfp->f_flag & (FWRITE | FAPPEND)) != FWRITE ||
5008 	    (infp->f_flag & FREAD) == 0) {
5009 		error = EBADF;
5010 		goto out;
5011 	}
5012 
5013 	/* If len == 0, just return 0. */
5014 	if (len == 0)
5015 		goto out;
5016 
5017 	/*
5018 	 * If infp and outfp refer to the same file, the byte ranges cannot
5019 	 * overlap.
5020 	 */
5021 	if (invp == outvp) {
5022 		if ((savinoff <= savoutoff && savinoff + len > savoutoff) ||
5023 		    (savinoff > savoutoff && savoutoff + len > savinoff)) {
5024 			error = EINVAL;
5025 			goto out;
5026 		}
5027 		rangelock_may_recurse(&invp->v_rl);
5028 	}
5029 
5030 	/* Range lock the byte ranges for both invp and outvp. */
5031 	for (;;) {
5032 		rl_wcookie = vn_rangelock_wlock(outvp, *outoffp, *outoffp +
5033 		    len);
5034 		rl_rcookie = vn_rangelock_tryrlock(invp, *inoffp, *inoffp +
5035 		    len);
5036 		if (rl_rcookie != NULL)
5037 			break;
5038 		vn_rangelock_unlock(outvp, rl_wcookie);
5039 		rl_rcookie = vn_rangelock_rlock(invp, *inoffp, *inoffp + len);
5040 		vn_rangelock_unlock(invp, rl_rcookie);
5041 	}
5042 
5043 	retlen = len;
5044 	error = vn_copy_file_range(invp, inoffp, outvp, outoffp, &retlen,
5045 	    flags, infp->f_cred, outfp->f_cred, td);
5046 out:
5047 	if (rl_rcookie != NULL)
5048 		vn_rangelock_unlock(invp, rl_rcookie);
5049 	if (rl_wcookie != NULL)
5050 		vn_rangelock_unlock(outvp, rl_wcookie);
5051 	if (savinoff != -1 && (error == EINTR || error == ERESTART)) {
5052 		*inoffp = savinoff;
5053 		*outoffp = savoutoff;
5054 	}
5055 	if (outfp != NULL)
5056 		fdrop(outfp, td);
5057 	if (infp != NULL)
5058 		fdrop(infp, td);
5059 	td->td_retval[0] = retlen;
5060 	return (error);
5061 }
5062 
5063 int
5064 sys_copy_file_range(struct thread *td, struct copy_file_range_args *uap)
5065 {
5066 	off_t inoff, outoff, *inoffp, *outoffp;
5067 	int error;
5068 
5069 	inoffp = outoffp = NULL;
5070 	if (uap->inoffp != NULL) {
5071 		error = copyin(uap->inoffp, &inoff, sizeof(off_t));
5072 		if (error != 0)
5073 			return (error);
5074 		inoffp = &inoff;
5075 	}
5076 	if (uap->outoffp != NULL) {
5077 		error = copyin(uap->outoffp, &outoff, sizeof(off_t));
5078 		if (error != 0)
5079 			return (error);
5080 		outoffp = &outoff;
5081 	}
5082 	error = kern_copy_file_range(td, uap->infd, inoffp, uap->outfd,
5083 	    outoffp, uap->len, uap->flags);
5084 	if (error == 0 && uap->inoffp != NULL)
5085 		error = copyout(inoffp, uap->inoffp, sizeof(off_t));
5086 	if (error == 0 && uap->outoffp != NULL)
5087 		error = copyout(outoffp, uap->outoffp, sizeof(off_t));
5088 	return (error);
5089 }
5090