xref: /freebsd/sys/kern/vfs_syscalls.c (revision dd4f32ae62426a10a84b4322756d82c06c202c4e)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (c) 1989, 1993
5  *	The Regents of the University of California.  All rights reserved.
6  * (c) UNIX System Laboratories, Inc.
7  * All or some portions of this file are derived from material licensed
8  * to the University of California by American Telephone and Telegraph
9  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
10  * the permission of UNIX System Laboratories, Inc.
11  *
12  * Redistribution and use in source and binary forms, with or without
13  * modification, are permitted provided that the following conditions
14  * are met:
15  * 1. Redistributions of source code must retain the above copyright
16  *    notice, this list of conditions and the following disclaimer.
17  * 2. Redistributions in binary form must reproduce the above copyright
18  *    notice, this list of conditions and the following disclaimer in the
19  *    documentation and/or other materials provided with the distribution.
20  * 3. Neither the name of the University nor the names of its contributors
21  *    may be used to endorse or promote products derived from this software
22  *    without specific prior written permission.
23  *
24  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
25  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
28  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
29  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
30  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
31  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
32  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
33  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
34  * SUCH DAMAGE.
35  *
36  *	@(#)vfs_syscalls.c	8.13 (Berkeley) 4/15/94
37  */
38 
39 #include <sys/cdefs.h>
40 __FBSDID("$FreeBSD$");
41 
42 #include "opt_capsicum.h"
43 #include "opt_ktrace.h"
44 
45 #include <sys/param.h>
46 #include <sys/systm.h>
47 #ifdef COMPAT_FREEBSD11
48 #include <sys/abi_compat.h>
49 #endif
50 #include <sys/bio.h>
51 #include <sys/buf.h>
52 #include <sys/capsicum.h>
53 #include <sys/disk.h>
54 #include <sys/malloc.h>
55 #include <sys/mount.h>
56 #include <sys/mutex.h>
57 #include <sys/sysproto.h>
58 #include <sys/namei.h>
59 #include <sys/filedesc.h>
60 #include <sys/kernel.h>
61 #include <sys/fcntl.h>
62 #include <sys/file.h>
63 #include <sys/filio.h>
64 #include <sys/limits.h>
65 #include <sys/linker.h>
66 #include <sys/rwlock.h>
67 #include <sys/sdt.h>
68 #include <sys/stat.h>
69 #include <sys/sx.h>
70 #include <sys/unistd.h>
71 #include <sys/vnode.h>
72 #include <sys/priv.h>
73 #include <sys/proc.h>
74 #include <sys/dirent.h>
75 #include <sys/jail.h>
76 #include <sys/syscallsubr.h>
77 #include <sys/sysctl.h>
78 #ifdef KTRACE
79 #include <sys/ktrace.h>
80 #endif
81 
82 #include <machine/stdarg.h>
83 
84 #include <security/audit/audit.h>
85 #include <security/mac/mac_framework.h>
86 
87 #include <vm/vm.h>
88 #include <vm/vm_object.h>
89 #include <vm/vm_page.h>
90 #include <vm/uma.h>
91 
92 #include <fs/devfs/devfs.h>
93 
94 MALLOC_DEFINE(M_FADVISE, "fadvise", "posix_fadvise(2) information");
95 
96 static int kern_chflagsat(struct thread *td, int fd, const char *path,
97     enum uio_seg pathseg, u_long flags, int atflag);
98 static int setfflags(struct thread *td, struct vnode *, u_long);
99 static int getutimes(const struct timeval *, enum uio_seg, struct timespec *);
100 static int getutimens(const struct timespec *, enum uio_seg,
101     struct timespec *, int *);
102 static int setutimes(struct thread *td, struct vnode *,
103     const struct timespec *, int, int);
104 static int vn_access(struct vnode *vp, int user_flags, struct ucred *cred,
105     struct thread *td);
106 static int kern_fhlinkat(struct thread *td, int fd, const char *path,
107     enum uio_seg pathseg, fhandle_t *fhp);
108 static int kern_readlink_vp(struct vnode *vp, char *buf, enum uio_seg bufseg,
109     size_t count, struct thread *td);
110 static int kern_linkat_vp(struct thread *td, struct vnode *vp, int fd,
111     const char *path, enum uio_seg segflag);
112 
113 uint64_t
114 at2cnpflags(u_int at_flags, u_int mask)
115 {
116 	uint64_t res;
117 
118 	MPASS((at_flags & (AT_SYMLINK_FOLLOW | AT_SYMLINK_NOFOLLOW)) !=
119 	    (AT_SYMLINK_FOLLOW | AT_SYMLINK_NOFOLLOW));
120 
121 	res = 0;
122 	at_flags &= mask;
123 	if ((at_flags & AT_RESOLVE_BENEATH) != 0)
124 		res |= RBENEATH;
125 	if ((at_flags & AT_SYMLINK_FOLLOW) != 0)
126 		res |= FOLLOW;
127 	/* NOFOLLOW is pseudo flag */
128 	if ((mask & AT_SYMLINK_NOFOLLOW) != 0) {
129 		res |= (at_flags & AT_SYMLINK_NOFOLLOW) != 0 ? NOFOLLOW :
130 		    FOLLOW;
131 	}
132 	if ((mask & AT_EMPTY_PATH) != 0 && (at_flags & AT_EMPTY_PATH) != 0)
133 		res |= EMPTYPATH;
134 	return (res);
135 }
136 
137 int
138 kern_sync(struct thread *td)
139 {
140 	struct mount *mp, *nmp;
141 	int save;
142 
143 	mtx_lock(&mountlist_mtx);
144 	for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
145 		if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK)) {
146 			nmp = TAILQ_NEXT(mp, mnt_list);
147 			continue;
148 		}
149 		if ((mp->mnt_flag & MNT_RDONLY) == 0 &&
150 		    vn_start_write(NULL, &mp, V_NOWAIT) == 0) {
151 			save = curthread_pflags_set(TDP_SYNCIO);
152 			vfs_periodic(mp, MNT_NOWAIT);
153 			VFS_SYNC(mp, MNT_NOWAIT);
154 			curthread_pflags_restore(save);
155 			vn_finished_write(mp);
156 		}
157 		mtx_lock(&mountlist_mtx);
158 		nmp = TAILQ_NEXT(mp, mnt_list);
159 		vfs_unbusy(mp);
160 	}
161 	mtx_unlock(&mountlist_mtx);
162 	return (0);
163 }
164 
165 /*
166  * Sync each mounted filesystem.
167  */
168 #ifndef _SYS_SYSPROTO_H_
169 struct sync_args {
170 	int     dummy;
171 };
172 #endif
173 /* ARGSUSED */
174 int
175 sys_sync(struct thread *td, struct sync_args *uap)
176 {
177 
178 	return (kern_sync(td));
179 }
180 
181 /*
182  * Change filesystem quotas.
183  */
184 #ifndef _SYS_SYSPROTO_H_
185 struct quotactl_args {
186 	char *path;
187 	int cmd;
188 	int uid;
189 	caddr_t arg;
190 };
191 #endif
192 int
193 sys_quotactl(struct thread *td, struct quotactl_args *uap)
194 {
195 	struct mount *mp;
196 	struct nameidata nd;
197 	int error;
198 	bool mp_busy;
199 
200 	AUDIT_ARG_CMD(uap->cmd);
201 	AUDIT_ARG_UID(uap->uid);
202 	if (!prison_allow(td->td_ucred, PR_ALLOW_QUOTAS))
203 		return (EPERM);
204 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
205 	    uap->path);
206 	if ((error = namei(&nd)) != 0)
207 		return (error);
208 	NDFREE_PNBUF(&nd);
209 	mp = nd.ni_vp->v_mount;
210 	vfs_ref(mp);
211 	vput(nd.ni_vp);
212 	error = vfs_busy(mp, 0);
213 	if (error != 0) {
214 		vfs_rel(mp);
215 		return (error);
216 	}
217 	mp_busy = true;
218 	error = VFS_QUOTACTL(mp, uap->cmd, uap->uid, uap->arg, &mp_busy);
219 
220 	/*
221 	 * Since quota on/off operations typically need to open quota
222 	 * files, the implementation may need to unbusy the mount point
223 	 * before calling into namei.  Otherwise, unmount might be
224 	 * started between two vfs_busy() invocations (first is ours,
225 	 * second is from mount point cross-walk code in lookup()),
226 	 * causing deadlock.
227 	 *
228 	 * Avoid unbusying mp if the implementation indicates it has
229 	 * already done so.
230 	 */
231 	if (mp_busy)
232 		vfs_unbusy(mp);
233 	vfs_rel(mp);
234 	return (error);
235 }
236 
237 /*
238  * Used by statfs conversion routines to scale the block size up if
239  * necessary so that all of the block counts are <= 'max_size'.  Note
240  * that 'max_size' should be a bitmask, i.e. 2^n - 1 for some non-zero
241  * value of 'n'.
242  */
243 void
244 statfs_scale_blocks(struct statfs *sf, long max_size)
245 {
246 	uint64_t count;
247 	int shift;
248 
249 	KASSERT(powerof2(max_size + 1), ("%s: invalid max_size", __func__));
250 
251 	/*
252 	 * Attempt to scale the block counts to give a more accurate
253 	 * overview to userland of the ratio of free space to used
254 	 * space.  To do this, find the largest block count and compute
255 	 * a divisor that lets it fit into a signed integer <= max_size.
256 	 */
257 	if (sf->f_bavail < 0)
258 		count = -sf->f_bavail;
259 	else
260 		count = sf->f_bavail;
261 	count = MAX(sf->f_blocks, MAX(sf->f_bfree, count));
262 	if (count <= max_size)
263 		return;
264 
265 	count >>= flsl(max_size);
266 	shift = 0;
267 	while (count > 0) {
268 		shift++;
269 		count >>=1;
270 	}
271 
272 	sf->f_bsize <<= shift;
273 	sf->f_blocks >>= shift;
274 	sf->f_bfree >>= shift;
275 	sf->f_bavail >>= shift;
276 }
277 
278 static int
279 kern_do_statfs(struct thread *td, struct mount *mp, struct statfs *buf)
280 {
281 	int error;
282 
283 	if (mp == NULL)
284 		return (EBADF);
285 	error = vfs_busy(mp, 0);
286 	vfs_rel(mp);
287 	if (error != 0)
288 		return (error);
289 #ifdef MAC
290 	error = mac_mount_check_stat(td->td_ucred, mp);
291 	if (error != 0)
292 		goto out;
293 #endif
294 	error = VFS_STATFS(mp, buf);
295 	if (error != 0)
296 		goto out;
297 	if (priv_check_cred_vfs_generation(td->td_ucred)) {
298 		buf->f_fsid.val[0] = buf->f_fsid.val[1] = 0;
299 		prison_enforce_statfs(td->td_ucred, mp, buf);
300 	}
301 out:
302 	vfs_unbusy(mp);
303 	return (error);
304 }
305 
306 /*
307  * Get filesystem statistics.
308  */
309 #ifndef _SYS_SYSPROTO_H_
310 struct statfs_args {
311 	char *path;
312 	struct statfs *buf;
313 };
314 #endif
315 int
316 sys_statfs(struct thread *td, struct statfs_args *uap)
317 {
318 	struct statfs *sfp;
319 	int error;
320 
321 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
322 	error = kern_statfs(td, uap->path, UIO_USERSPACE, sfp);
323 	if (error == 0)
324 		error = copyout(sfp, uap->buf, sizeof(struct statfs));
325 	free(sfp, M_STATFS);
326 	return (error);
327 }
328 
329 int
330 kern_statfs(struct thread *td, const char *path, enum uio_seg pathseg,
331     struct statfs *buf)
332 {
333 	struct mount *mp;
334 	struct nameidata nd;
335 	int error;
336 
337 	NDINIT(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path);
338 	error = namei(&nd);
339 	if (error != 0)
340 		return (error);
341 	NDFREE_PNBUF(&nd);
342 	mp = vfs_ref_from_vp(nd.ni_vp);
343 	vrele(nd.ni_vp);
344 	return (kern_do_statfs(td, mp, buf));
345 }
346 
347 /*
348  * Get filesystem statistics.
349  */
350 #ifndef _SYS_SYSPROTO_H_
351 struct fstatfs_args {
352 	int fd;
353 	struct statfs *buf;
354 };
355 #endif
356 int
357 sys_fstatfs(struct thread *td, struct fstatfs_args *uap)
358 {
359 	struct statfs *sfp;
360 	int error;
361 
362 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
363 	error = kern_fstatfs(td, uap->fd, sfp);
364 	if (error == 0)
365 		error = copyout(sfp, uap->buf, sizeof(struct statfs));
366 	free(sfp, M_STATFS);
367 	return (error);
368 }
369 
370 int
371 kern_fstatfs(struct thread *td, int fd, struct statfs *buf)
372 {
373 	struct file *fp;
374 	struct mount *mp;
375 	struct vnode *vp;
376 	int error;
377 
378 	AUDIT_ARG_FD(fd);
379 	error = getvnode_path(td, fd, &cap_fstatfs_rights, &fp);
380 	if (error != 0)
381 		return (error);
382 	vp = fp->f_vnode;
383 #ifdef AUDIT
384 	if (AUDITING_TD(td)) {
385 		vn_lock(vp, LK_SHARED | LK_RETRY);
386 		AUDIT_ARG_VNODE1(vp);
387 		VOP_UNLOCK(vp);
388 	}
389 #endif
390 	mp = vfs_ref_from_vp(vp);
391 	fdrop(fp, td);
392 	return (kern_do_statfs(td, mp, buf));
393 }
394 
395 /*
396  * Get statistics on all filesystems.
397  */
398 #ifndef _SYS_SYSPROTO_H_
399 struct getfsstat_args {
400 	struct statfs *buf;
401 	long bufsize;
402 	int mode;
403 };
404 #endif
405 int
406 sys_getfsstat(struct thread *td, struct getfsstat_args *uap)
407 {
408 	size_t count;
409 	int error;
410 
411 	if (uap->bufsize < 0 || uap->bufsize > SIZE_MAX)
412 		return (EINVAL);
413 	error = kern_getfsstat(td, &uap->buf, uap->bufsize, &count,
414 	    UIO_USERSPACE, uap->mode);
415 	if (error == 0)
416 		td->td_retval[0] = count;
417 	return (error);
418 }
419 
420 /*
421  * If (bufsize > 0 && bufseg == UIO_SYSSPACE)
422  *	The caller is responsible for freeing memory which will be allocated
423  *	in '*buf'.
424  */
425 int
426 kern_getfsstat(struct thread *td, struct statfs **buf, size_t bufsize,
427     size_t *countp, enum uio_seg bufseg, int mode)
428 {
429 	struct mount *mp, *nmp;
430 	struct statfs *sfsp, *sp, *sptmp, *tofree;
431 	size_t count, maxcount;
432 	int error;
433 
434 	switch (mode) {
435 	case MNT_WAIT:
436 	case MNT_NOWAIT:
437 		break;
438 	default:
439 		if (bufseg == UIO_SYSSPACE)
440 			*buf = NULL;
441 		return (EINVAL);
442 	}
443 restart:
444 	maxcount = bufsize / sizeof(struct statfs);
445 	if (bufsize == 0) {
446 		sfsp = NULL;
447 		tofree = NULL;
448 	} else if (bufseg == UIO_USERSPACE) {
449 		sfsp = *buf;
450 		tofree = NULL;
451 	} else /* if (bufseg == UIO_SYSSPACE) */ {
452 		count = 0;
453 		mtx_lock(&mountlist_mtx);
454 		TAILQ_FOREACH(mp, &mountlist, mnt_list) {
455 			count++;
456 		}
457 		mtx_unlock(&mountlist_mtx);
458 		if (maxcount > count)
459 			maxcount = count;
460 		tofree = sfsp = *buf = malloc(maxcount * sizeof(struct statfs),
461 		    M_STATFS, M_WAITOK);
462 	}
463 
464 	count = 0;
465 
466 	/*
467 	 * If there is no target buffer they only want the count.
468 	 *
469 	 * This could be TAILQ_FOREACH but it is open-coded to match the original
470 	 * code below.
471 	 */
472 	if (sfsp == NULL) {
473 		mtx_lock(&mountlist_mtx);
474 		for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
475 			if (prison_canseemount(td->td_ucred, mp) != 0) {
476 				nmp = TAILQ_NEXT(mp, mnt_list);
477 				continue;
478 			}
479 #ifdef MAC
480 			if (mac_mount_check_stat(td->td_ucred, mp) != 0) {
481 				nmp = TAILQ_NEXT(mp, mnt_list);
482 				continue;
483 			}
484 #endif
485 			count++;
486 			nmp = TAILQ_NEXT(mp, mnt_list);
487 		}
488 		mtx_unlock(&mountlist_mtx);
489 		*countp = count;
490 		return (0);
491 	}
492 
493 	/*
494 	 * They want the entire thing.
495 	 *
496 	 * Short-circuit the corner case of no room for anything, avoids
497 	 * relocking below.
498 	 */
499 	if (maxcount < 1) {
500 		goto out;
501 	}
502 
503 	mtx_lock(&mountlist_mtx);
504 	for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
505 		if (prison_canseemount(td->td_ucred, mp) != 0) {
506 			nmp = TAILQ_NEXT(mp, mnt_list);
507 			continue;
508 		}
509 #ifdef MAC
510 		if (mac_mount_check_stat(td->td_ucred, mp) != 0) {
511 			nmp = TAILQ_NEXT(mp, mnt_list);
512 			continue;
513 		}
514 #endif
515 		if (mode == MNT_WAIT) {
516 			if (vfs_busy(mp, MBF_MNTLSTLOCK) != 0) {
517 				/*
518 				 * If vfs_busy() failed, and MBF_NOWAIT
519 				 * wasn't passed, then the mp is gone.
520 				 * Furthermore, because of MBF_MNTLSTLOCK,
521 				 * the mountlist_mtx was dropped.  We have
522 				 * no other choice than to start over.
523 				 */
524 				mtx_unlock(&mountlist_mtx);
525 				free(tofree, M_STATFS);
526 				goto restart;
527 			}
528 		} else {
529 			if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK) != 0) {
530 				nmp = TAILQ_NEXT(mp, mnt_list);
531 				continue;
532 			}
533 		}
534 		sp = &mp->mnt_stat;
535 		/*
536 		 * If MNT_NOWAIT is specified, do not refresh
537 		 * the fsstat cache.
538 		 */
539 		if (mode != MNT_NOWAIT) {
540 			error = VFS_STATFS(mp, sp);
541 			if (error != 0) {
542 				mtx_lock(&mountlist_mtx);
543 				nmp = TAILQ_NEXT(mp, mnt_list);
544 				vfs_unbusy(mp);
545 				continue;
546 			}
547 		}
548 		if (priv_check_cred_vfs_generation(td->td_ucred)) {
549 			sptmp = malloc(sizeof(struct statfs), M_STATFS,
550 			    M_WAITOK);
551 			*sptmp = *sp;
552 			sptmp->f_fsid.val[0] = sptmp->f_fsid.val[1] = 0;
553 			prison_enforce_statfs(td->td_ucred, mp, sptmp);
554 			sp = sptmp;
555 		} else
556 			sptmp = NULL;
557 		if (bufseg == UIO_SYSSPACE) {
558 			bcopy(sp, sfsp, sizeof(*sp));
559 			free(sptmp, M_STATFS);
560 		} else /* if (bufseg == UIO_USERSPACE) */ {
561 			error = copyout(sp, sfsp, sizeof(*sp));
562 			free(sptmp, M_STATFS);
563 			if (error != 0) {
564 				vfs_unbusy(mp);
565 				return (error);
566 			}
567 		}
568 		sfsp++;
569 		count++;
570 
571 		if (count == maxcount) {
572 			vfs_unbusy(mp);
573 			goto out;
574 		}
575 
576 		mtx_lock(&mountlist_mtx);
577 		nmp = TAILQ_NEXT(mp, mnt_list);
578 		vfs_unbusy(mp);
579 	}
580 	mtx_unlock(&mountlist_mtx);
581 out:
582 	*countp = count;
583 	return (0);
584 }
585 
586 #ifdef COMPAT_FREEBSD4
587 /*
588  * Get old format filesystem statistics.
589  */
590 static void freebsd4_cvtstatfs(struct statfs *, struct ostatfs *);
591 
592 #ifndef _SYS_SYSPROTO_H_
593 struct freebsd4_statfs_args {
594 	char *path;
595 	struct ostatfs *buf;
596 };
597 #endif
598 int
599 freebsd4_statfs(struct thread *td, struct freebsd4_statfs_args *uap)
600 {
601 	struct ostatfs osb;
602 	struct statfs *sfp;
603 	int error;
604 
605 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
606 	error = kern_statfs(td, uap->path, UIO_USERSPACE, sfp);
607 	if (error == 0) {
608 		freebsd4_cvtstatfs(sfp, &osb);
609 		error = copyout(&osb, uap->buf, sizeof(osb));
610 	}
611 	free(sfp, M_STATFS);
612 	return (error);
613 }
614 
615 /*
616  * Get filesystem statistics.
617  */
618 #ifndef _SYS_SYSPROTO_H_
619 struct freebsd4_fstatfs_args {
620 	int fd;
621 	struct ostatfs *buf;
622 };
623 #endif
624 int
625 freebsd4_fstatfs(struct thread *td, struct freebsd4_fstatfs_args *uap)
626 {
627 	struct ostatfs osb;
628 	struct statfs *sfp;
629 	int error;
630 
631 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
632 	error = kern_fstatfs(td, uap->fd, sfp);
633 	if (error == 0) {
634 		freebsd4_cvtstatfs(sfp, &osb);
635 		error = copyout(&osb, uap->buf, sizeof(osb));
636 	}
637 	free(sfp, M_STATFS);
638 	return (error);
639 }
640 
641 /*
642  * Get statistics on all filesystems.
643  */
644 #ifndef _SYS_SYSPROTO_H_
645 struct freebsd4_getfsstat_args {
646 	struct ostatfs *buf;
647 	long bufsize;
648 	int mode;
649 };
650 #endif
651 int
652 freebsd4_getfsstat(struct thread *td, struct freebsd4_getfsstat_args *uap)
653 {
654 	struct statfs *buf, *sp;
655 	struct ostatfs osb;
656 	size_t count, size;
657 	int error;
658 
659 	if (uap->bufsize < 0)
660 		return (EINVAL);
661 	count = uap->bufsize / sizeof(struct ostatfs);
662 	if (count > SIZE_MAX / sizeof(struct statfs))
663 		return (EINVAL);
664 	size = count * sizeof(struct statfs);
665 	error = kern_getfsstat(td, &buf, size, &count, UIO_SYSSPACE,
666 	    uap->mode);
667 	if (error == 0)
668 		td->td_retval[0] = count;
669 	if (size != 0) {
670 		sp = buf;
671 		while (count != 0 && error == 0) {
672 			freebsd4_cvtstatfs(sp, &osb);
673 			error = copyout(&osb, uap->buf, sizeof(osb));
674 			sp++;
675 			uap->buf++;
676 			count--;
677 		}
678 		free(buf, M_STATFS);
679 	}
680 	return (error);
681 }
682 
683 /*
684  * Implement fstatfs() for (NFS) file handles.
685  */
686 #ifndef _SYS_SYSPROTO_H_
687 struct freebsd4_fhstatfs_args {
688 	struct fhandle *u_fhp;
689 	struct ostatfs *buf;
690 };
691 #endif
692 int
693 freebsd4_fhstatfs(struct thread *td, struct freebsd4_fhstatfs_args *uap)
694 {
695 	struct ostatfs osb;
696 	struct statfs *sfp;
697 	fhandle_t fh;
698 	int error;
699 
700 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
701 	if (error != 0)
702 		return (error);
703 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
704 	error = kern_fhstatfs(td, fh, sfp);
705 	if (error == 0) {
706 		freebsd4_cvtstatfs(sfp, &osb);
707 		error = copyout(&osb, uap->buf, sizeof(osb));
708 	}
709 	free(sfp, M_STATFS);
710 	return (error);
711 }
712 
713 /*
714  * Convert a new format statfs structure to an old format statfs structure.
715  */
716 static void
717 freebsd4_cvtstatfs(struct statfs *nsp, struct ostatfs *osp)
718 {
719 
720 	statfs_scale_blocks(nsp, LONG_MAX);
721 	bzero(osp, sizeof(*osp));
722 	osp->f_bsize = nsp->f_bsize;
723 	osp->f_iosize = MIN(nsp->f_iosize, LONG_MAX);
724 	osp->f_blocks = nsp->f_blocks;
725 	osp->f_bfree = nsp->f_bfree;
726 	osp->f_bavail = nsp->f_bavail;
727 	osp->f_files = MIN(nsp->f_files, LONG_MAX);
728 	osp->f_ffree = MIN(nsp->f_ffree, LONG_MAX);
729 	osp->f_owner = nsp->f_owner;
730 	osp->f_type = nsp->f_type;
731 	osp->f_flags = nsp->f_flags;
732 	osp->f_syncwrites = MIN(nsp->f_syncwrites, LONG_MAX);
733 	osp->f_asyncwrites = MIN(nsp->f_asyncwrites, LONG_MAX);
734 	osp->f_syncreads = MIN(nsp->f_syncreads, LONG_MAX);
735 	osp->f_asyncreads = MIN(nsp->f_asyncreads, LONG_MAX);
736 	strlcpy(osp->f_fstypename, nsp->f_fstypename,
737 	    MIN(MFSNAMELEN, OMFSNAMELEN));
738 	strlcpy(osp->f_mntonname, nsp->f_mntonname,
739 	    MIN(MNAMELEN, OMNAMELEN));
740 	strlcpy(osp->f_mntfromname, nsp->f_mntfromname,
741 	    MIN(MNAMELEN, OMNAMELEN));
742 	osp->f_fsid = nsp->f_fsid;
743 }
744 #endif /* COMPAT_FREEBSD4 */
745 
746 #if defined(COMPAT_FREEBSD11)
747 /*
748  * Get old format filesystem statistics.
749  */
750 static void freebsd11_cvtstatfs(struct statfs *, struct freebsd11_statfs *);
751 
752 int
753 freebsd11_statfs(struct thread *td, struct freebsd11_statfs_args *uap)
754 {
755 	struct freebsd11_statfs osb;
756 	struct statfs *sfp;
757 	int error;
758 
759 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
760 	error = kern_statfs(td, uap->path, UIO_USERSPACE, sfp);
761 	if (error == 0) {
762 		freebsd11_cvtstatfs(sfp, &osb);
763 		error = copyout(&osb, uap->buf, sizeof(osb));
764 	}
765 	free(sfp, M_STATFS);
766 	return (error);
767 }
768 
769 /*
770  * Get filesystem statistics.
771  */
772 int
773 freebsd11_fstatfs(struct thread *td, struct freebsd11_fstatfs_args *uap)
774 {
775 	struct freebsd11_statfs osb;
776 	struct statfs *sfp;
777 	int error;
778 
779 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
780 	error = kern_fstatfs(td, uap->fd, sfp);
781 	if (error == 0) {
782 		freebsd11_cvtstatfs(sfp, &osb);
783 		error = copyout(&osb, uap->buf, sizeof(osb));
784 	}
785 	free(sfp, M_STATFS);
786 	return (error);
787 }
788 
789 /*
790  * Get statistics on all filesystems.
791  */
792 int
793 freebsd11_getfsstat(struct thread *td, struct freebsd11_getfsstat_args *uap)
794 {
795 	return (kern_freebsd11_getfsstat(td, uap->buf, uap->bufsize, uap->mode));
796 }
797 
798 int
799 kern_freebsd11_getfsstat(struct thread *td, struct freebsd11_statfs * ubuf,
800     long bufsize, int mode)
801 {
802 	struct freebsd11_statfs osb;
803 	struct statfs *buf, *sp;
804 	size_t count, size;
805 	int error;
806 
807 	if (bufsize < 0)
808 		return (EINVAL);
809 
810 	count = bufsize / sizeof(struct ostatfs);
811 	size = count * sizeof(struct statfs);
812 	error = kern_getfsstat(td, &buf, size, &count, UIO_SYSSPACE, mode);
813 	if (error == 0)
814 		td->td_retval[0] = count;
815 	if (size > 0) {
816 		sp = buf;
817 		while (count > 0 && error == 0) {
818 			freebsd11_cvtstatfs(sp, &osb);
819 			error = copyout(&osb, ubuf, sizeof(osb));
820 			sp++;
821 			ubuf++;
822 			count--;
823 		}
824 		free(buf, M_STATFS);
825 	}
826 	return (error);
827 }
828 
829 /*
830  * Implement fstatfs() for (NFS) file handles.
831  */
832 int
833 freebsd11_fhstatfs(struct thread *td, struct freebsd11_fhstatfs_args *uap)
834 {
835 	struct freebsd11_statfs osb;
836 	struct statfs *sfp;
837 	fhandle_t fh;
838 	int error;
839 
840 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
841 	if (error)
842 		return (error);
843 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
844 	error = kern_fhstatfs(td, fh, sfp);
845 	if (error == 0) {
846 		freebsd11_cvtstatfs(sfp, &osb);
847 		error = copyout(&osb, uap->buf, sizeof(osb));
848 	}
849 	free(sfp, M_STATFS);
850 	return (error);
851 }
852 
853 /*
854  * Convert a new format statfs structure to an old format statfs structure.
855  */
856 static void
857 freebsd11_cvtstatfs(struct statfs *nsp, struct freebsd11_statfs *osp)
858 {
859 
860 	bzero(osp, sizeof(*osp));
861 	osp->f_version = FREEBSD11_STATFS_VERSION;
862 	osp->f_type = nsp->f_type;
863 	osp->f_flags = nsp->f_flags;
864 	osp->f_bsize = nsp->f_bsize;
865 	osp->f_iosize = nsp->f_iosize;
866 	osp->f_blocks = nsp->f_blocks;
867 	osp->f_bfree = nsp->f_bfree;
868 	osp->f_bavail = nsp->f_bavail;
869 	osp->f_files = nsp->f_files;
870 	osp->f_ffree = nsp->f_ffree;
871 	osp->f_syncwrites = nsp->f_syncwrites;
872 	osp->f_asyncwrites = nsp->f_asyncwrites;
873 	osp->f_syncreads = nsp->f_syncreads;
874 	osp->f_asyncreads = nsp->f_asyncreads;
875 	osp->f_namemax = nsp->f_namemax;
876 	osp->f_owner = nsp->f_owner;
877 	osp->f_fsid = nsp->f_fsid;
878 	strlcpy(osp->f_fstypename, nsp->f_fstypename,
879 	    MIN(MFSNAMELEN, sizeof(osp->f_fstypename)));
880 	strlcpy(osp->f_mntonname, nsp->f_mntonname,
881 	    MIN(MNAMELEN, sizeof(osp->f_mntonname)));
882 	strlcpy(osp->f_mntfromname, nsp->f_mntfromname,
883 	    MIN(MNAMELEN, sizeof(osp->f_mntfromname)));
884 }
885 #endif /* COMPAT_FREEBSD11 */
886 
887 /*
888  * Change current working directory to a given file descriptor.
889  */
890 #ifndef _SYS_SYSPROTO_H_
891 struct fchdir_args {
892 	int	fd;
893 };
894 #endif
895 int
896 sys_fchdir(struct thread *td, struct fchdir_args *uap)
897 {
898 	struct vnode *vp, *tdp;
899 	struct mount *mp;
900 	struct file *fp;
901 	int error;
902 
903 	AUDIT_ARG_FD(uap->fd);
904 	error = getvnode_path(td, uap->fd, &cap_fchdir_rights,
905 	    &fp);
906 	if (error != 0)
907 		return (error);
908 	vp = fp->f_vnode;
909 	vrefact(vp);
910 	fdrop(fp, td);
911 	vn_lock(vp, LK_SHARED | LK_RETRY);
912 	AUDIT_ARG_VNODE1(vp);
913 	error = change_dir(vp, td);
914 	while (!error && (mp = vp->v_mountedhere) != NULL) {
915 		if (vfs_busy(mp, 0))
916 			continue;
917 		error = VFS_ROOT(mp, LK_SHARED, &tdp);
918 		vfs_unbusy(mp);
919 		if (error != 0)
920 			break;
921 		vput(vp);
922 		vp = tdp;
923 	}
924 	if (error != 0) {
925 		vput(vp);
926 		return (error);
927 	}
928 	VOP_UNLOCK(vp);
929 	pwd_chdir(td, vp);
930 	return (0);
931 }
932 
933 /*
934  * Change current working directory (``.'').
935  */
936 #ifndef _SYS_SYSPROTO_H_
937 struct chdir_args {
938 	char	*path;
939 };
940 #endif
941 int
942 sys_chdir(struct thread *td, struct chdir_args *uap)
943 {
944 
945 	return (kern_chdir(td, uap->path, UIO_USERSPACE));
946 }
947 
948 int
949 kern_chdir(struct thread *td, const char *path, enum uio_seg pathseg)
950 {
951 	struct nameidata nd;
952 	int error;
953 
954 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
955 	    pathseg, path);
956 	if ((error = namei(&nd)) != 0)
957 		return (error);
958 	if ((error = change_dir(nd.ni_vp, td)) != 0) {
959 		vput(nd.ni_vp);
960 		NDFREE_PNBUF(&nd);
961 		return (error);
962 	}
963 	VOP_UNLOCK(nd.ni_vp);
964 	NDFREE_PNBUF(&nd);
965 	pwd_chdir(td, nd.ni_vp);
966 	return (0);
967 }
968 
969 static int unprivileged_chroot = 0;
970 SYSCTL_INT(_security_bsd, OID_AUTO, unprivileged_chroot, CTLFLAG_RW,
971     &unprivileged_chroot, 0,
972     "Unprivileged processes can use chroot(2)");
973 /*
974  * Change notion of root (``/'') directory.
975  */
976 #ifndef _SYS_SYSPROTO_H_
977 struct chroot_args {
978 	char	*path;
979 };
980 #endif
981 int
982 sys_chroot(struct thread *td, struct chroot_args *uap)
983 {
984 	struct nameidata nd;
985 	struct proc *p;
986 	int error;
987 
988 	error = priv_check(td, PRIV_VFS_CHROOT);
989 	if (error != 0) {
990 		p = td->td_proc;
991 		PROC_LOCK(p);
992 		if (unprivileged_chroot == 0 ||
993 		    (p->p_flag2 & P2_NO_NEW_PRIVS) == 0) {
994 			PROC_UNLOCK(p);
995 			return (error);
996 		}
997 		PROC_UNLOCK(p);
998 	}
999 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
1000 	    UIO_USERSPACE, uap->path);
1001 	error = namei(&nd);
1002 	if (error != 0)
1003 		return (error);
1004 	NDFREE_PNBUF(&nd);
1005 	error = change_dir(nd.ni_vp, td);
1006 	if (error != 0)
1007 		goto e_vunlock;
1008 #ifdef MAC
1009 	error = mac_vnode_check_chroot(td->td_ucred, nd.ni_vp);
1010 	if (error != 0)
1011 		goto e_vunlock;
1012 #endif
1013 	VOP_UNLOCK(nd.ni_vp);
1014 	error = pwd_chroot(td, nd.ni_vp);
1015 	vrele(nd.ni_vp);
1016 	return (error);
1017 e_vunlock:
1018 	vput(nd.ni_vp);
1019 	return (error);
1020 }
1021 
1022 /*
1023  * Common routine for chroot and chdir.  Callers must provide a locked vnode
1024  * instance.
1025  */
1026 int
1027 change_dir(struct vnode *vp, struct thread *td)
1028 {
1029 #ifdef MAC
1030 	int error;
1031 #endif
1032 
1033 	ASSERT_VOP_LOCKED(vp, "change_dir(): vp not locked");
1034 	if (vp->v_type != VDIR)
1035 		return (ENOTDIR);
1036 #ifdef MAC
1037 	error = mac_vnode_check_chdir(td->td_ucred, vp);
1038 	if (error != 0)
1039 		return (error);
1040 #endif
1041 	return (VOP_ACCESS(vp, VEXEC, td->td_ucred, td));
1042 }
1043 
1044 static __inline void
1045 flags_to_rights(int flags, cap_rights_t *rightsp)
1046 {
1047 	if (flags & O_EXEC) {
1048 		cap_rights_set_one(rightsp, CAP_FEXECVE);
1049 		if (flags & O_PATH)
1050 			return;
1051 	} else {
1052 		switch ((flags & O_ACCMODE)) {
1053 		case O_RDONLY:
1054 			cap_rights_set_one(rightsp, CAP_READ);
1055 			break;
1056 		case O_RDWR:
1057 			cap_rights_set_one(rightsp, CAP_READ);
1058 			/* FALLTHROUGH */
1059 		case O_WRONLY:
1060 			cap_rights_set_one(rightsp, CAP_WRITE);
1061 			if (!(flags & (O_APPEND | O_TRUNC)))
1062 				cap_rights_set_one(rightsp, CAP_SEEK);
1063 			break;
1064 		}
1065 	}
1066 
1067 	if (flags & O_CREAT)
1068 		cap_rights_set_one(rightsp, CAP_CREATE);
1069 
1070 	if (flags & O_TRUNC)
1071 		cap_rights_set_one(rightsp, CAP_FTRUNCATE);
1072 
1073 	if (flags & (O_SYNC | O_FSYNC))
1074 		cap_rights_set_one(rightsp, CAP_FSYNC);
1075 
1076 	if (flags & (O_EXLOCK | O_SHLOCK))
1077 		cap_rights_set_one(rightsp, CAP_FLOCK);
1078 }
1079 
1080 /*
1081  * Check permissions, allocate an open file structure, and call the device
1082  * open routine if any.
1083  */
1084 #ifndef _SYS_SYSPROTO_H_
1085 struct open_args {
1086 	char	*path;
1087 	int	flags;
1088 	int	mode;
1089 };
1090 #endif
1091 int
1092 sys_open(struct thread *td, struct open_args *uap)
1093 {
1094 
1095 	return (kern_openat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1096 	    uap->flags, uap->mode));
1097 }
1098 
1099 #ifndef _SYS_SYSPROTO_H_
1100 struct openat_args {
1101 	int	fd;
1102 	char	*path;
1103 	int	flag;
1104 	int	mode;
1105 };
1106 #endif
1107 int
1108 sys_openat(struct thread *td, struct openat_args *uap)
1109 {
1110 
1111 	AUDIT_ARG_FD(uap->fd);
1112 	return (kern_openat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
1113 	    uap->mode));
1114 }
1115 
1116 int
1117 kern_openat(struct thread *td, int fd, const char *path, enum uio_seg pathseg,
1118     int flags, int mode)
1119 {
1120 	struct proc *p = td->td_proc;
1121 	struct filedesc *fdp;
1122 	struct pwddesc *pdp;
1123 	struct file *fp;
1124 	struct vnode *vp;
1125 	struct nameidata nd;
1126 	cap_rights_t rights;
1127 	int cmode, error, indx;
1128 
1129 	indx = -1;
1130 	fdp = p->p_fd;
1131 	pdp = p->p_pd;
1132 
1133 	AUDIT_ARG_FFLAGS(flags);
1134 	AUDIT_ARG_MODE(mode);
1135 	cap_rights_init_one(&rights, CAP_LOOKUP);
1136 	flags_to_rights(flags, &rights);
1137 
1138 	/*
1139 	 * Only one of the O_EXEC, O_RDONLY, O_WRONLY and O_RDWR flags
1140 	 * may be specified.  On the other hand, for O_PATH any mode
1141 	 * except O_EXEC is ignored.
1142 	 */
1143 	if ((flags & O_PATH) != 0) {
1144 		flags &= ~(O_CREAT | O_ACCMODE);
1145 	} else if ((flags & O_EXEC) != 0) {
1146 		if (flags & O_ACCMODE)
1147 			return (EINVAL);
1148 	} else if ((flags & O_ACCMODE) == O_ACCMODE) {
1149 		return (EINVAL);
1150 	} else {
1151 		flags = FFLAGS(flags);
1152 	}
1153 
1154 	/*
1155 	 * Allocate a file structure. The descriptor to reference it
1156 	 * is allocated and used by finstall_refed() below.
1157 	 */
1158 	error = falloc_noinstall(td, &fp);
1159 	if (error != 0)
1160 		return (error);
1161 	/* Set the flags early so the finit in devfs can pick them up. */
1162 	fp->f_flag = flags & FMASK;
1163 	cmode = ((mode & ~pdp->pd_cmask) & ALLPERMS) & ~S_ISTXT;
1164 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1 | WANTIOCTLCAPS,
1165 	    pathseg, path, fd, &rights);
1166 	td->td_dupfd = -1;		/* XXX check for fdopen */
1167 	error = vn_open_cred(&nd, &flags, cmode, VN_OPEN_WANTIOCTLCAPS,
1168 	    td->td_ucred, fp);
1169 	if (error != 0) {
1170 		/*
1171 		 * If the vn_open replaced the method vector, something
1172 		 * wonderous happened deep below and we just pass it up
1173 		 * pretending we know what we do.
1174 		 */
1175 		if (error == ENXIO && fp->f_ops != &badfileops) {
1176 			MPASS((flags & O_PATH) == 0);
1177 			goto success;
1178 		}
1179 
1180 		/*
1181 		 * Handle special fdopen() case. bleh.
1182 		 *
1183 		 * Don't do this for relative (capability) lookups; we don't
1184 		 * understand exactly what would happen, and we don't think
1185 		 * that it ever should.
1186 		 */
1187 		if ((nd.ni_resflags & NIRES_STRICTREL) == 0 &&
1188 		    (error == ENODEV || error == ENXIO) &&
1189 		    td->td_dupfd >= 0) {
1190 			error = dupfdopen(td, fdp, td->td_dupfd, flags, error,
1191 			    &indx);
1192 			if (error == 0)
1193 				goto success;
1194 		}
1195 
1196 		goto bad;
1197 	}
1198 	td->td_dupfd = 0;
1199 	NDFREE_PNBUF(&nd);
1200 	vp = nd.ni_vp;
1201 
1202 	/*
1203 	 * Store the vnode, for any f_type. Typically, the vnode use
1204 	 * count is decremented by direct call to vn_closefile() for
1205 	 * files that switched type in the cdevsw fdopen() method.
1206 	 */
1207 	fp->f_vnode = vp;
1208 
1209 	/*
1210 	 * If the file wasn't claimed by devfs bind it to the normal
1211 	 * vnode operations here.
1212 	 */
1213 	if (fp->f_ops == &badfileops) {
1214 		KASSERT(vp->v_type != VFIFO || (flags & O_PATH) != 0,
1215 		    ("Unexpected fifo fp %p vp %p", fp, vp));
1216 		if ((flags & O_PATH) != 0) {
1217 			finit(fp, (flags & FMASK) | (fp->f_flag & FKQALLOWED),
1218 			    DTYPE_VNODE, NULL, &path_fileops);
1219 		} else {
1220 			finit_vnode(fp, flags, NULL, &vnops);
1221 		}
1222 	}
1223 
1224 	VOP_UNLOCK(vp);
1225 	if (flags & O_TRUNC) {
1226 		error = fo_truncate(fp, 0, td->td_ucred, td);
1227 		if (error != 0)
1228 			goto bad;
1229 	}
1230 success:
1231 	/*
1232 	 * If we haven't already installed the FD (for dupfdopen), do so now.
1233 	 */
1234 	if (indx == -1) {
1235 		struct filecaps *fcaps;
1236 
1237 #ifdef CAPABILITIES
1238 		if ((nd.ni_resflags & NIRES_STRICTREL) != 0)
1239 			fcaps = &nd.ni_filecaps;
1240 		else
1241 #endif
1242 			fcaps = NULL;
1243 		error = finstall_refed(td, fp, &indx, flags, fcaps);
1244 		/* On success finstall_refed() consumes fcaps. */
1245 		if (error != 0) {
1246 			goto bad;
1247 		}
1248 	} else {
1249 		NDFREE_IOCTLCAPS(&nd);
1250 		falloc_abort(td, fp);
1251 	}
1252 
1253 	td->td_retval[0] = indx;
1254 	return (0);
1255 bad:
1256 	KASSERT(indx == -1, ("indx=%d, should be -1", indx));
1257 	NDFREE_IOCTLCAPS(&nd);
1258 	falloc_abort(td, fp);
1259 	return (error);
1260 }
1261 
1262 #ifdef COMPAT_43
1263 /*
1264  * Create a file.
1265  */
1266 #ifndef _SYS_SYSPROTO_H_
1267 struct ocreat_args {
1268 	char	*path;
1269 	int	mode;
1270 };
1271 #endif
1272 int
1273 ocreat(struct thread *td, struct ocreat_args *uap)
1274 {
1275 
1276 	return (kern_openat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1277 	    O_WRONLY | O_CREAT | O_TRUNC, uap->mode));
1278 }
1279 #endif /* COMPAT_43 */
1280 
1281 /*
1282  * Create a special file.
1283  */
1284 #ifndef _SYS_SYSPROTO_H_
1285 struct mknodat_args {
1286 	int	fd;
1287 	char	*path;
1288 	mode_t	mode;
1289 	dev_t	dev;
1290 };
1291 #endif
1292 int
1293 sys_mknodat(struct thread *td, struct mknodat_args *uap)
1294 {
1295 
1296 	return (kern_mknodat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode,
1297 	    uap->dev));
1298 }
1299 
1300 #if defined(COMPAT_FREEBSD11)
1301 int
1302 freebsd11_mknod(struct thread *td,
1303     struct freebsd11_mknod_args *uap)
1304 {
1305 
1306 	return (kern_mknodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1307 	    uap->mode, uap->dev));
1308 }
1309 
1310 int
1311 freebsd11_mknodat(struct thread *td,
1312     struct freebsd11_mknodat_args *uap)
1313 {
1314 
1315 	return (kern_mknodat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode,
1316 	    uap->dev));
1317 }
1318 #endif /* COMPAT_FREEBSD11 */
1319 
1320 int
1321 kern_mknodat(struct thread *td, int fd, const char *path, enum uio_seg pathseg,
1322     int mode, dev_t dev)
1323 {
1324 	struct vnode *vp;
1325 	struct mount *mp;
1326 	struct vattr vattr;
1327 	struct nameidata nd;
1328 	int error, whiteout = 0;
1329 
1330 	AUDIT_ARG_MODE(mode);
1331 	AUDIT_ARG_DEV(dev);
1332 	switch (mode & S_IFMT) {
1333 	case S_IFCHR:
1334 	case S_IFBLK:
1335 		error = priv_check(td, PRIV_VFS_MKNOD_DEV);
1336 		if (error == 0 && dev == VNOVAL)
1337 			error = EINVAL;
1338 		break;
1339 	case S_IFWHT:
1340 		error = priv_check(td, PRIV_VFS_MKNOD_WHT);
1341 		break;
1342 	case S_IFIFO:
1343 		if (dev == 0)
1344 			return (kern_mkfifoat(td, fd, path, pathseg, mode));
1345 		/* FALLTHROUGH */
1346 	default:
1347 		error = EINVAL;
1348 		break;
1349 	}
1350 	if (error != 0)
1351 		return (error);
1352 	NDPREINIT(&nd);
1353 restart:
1354 	bwillwrite();
1355 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | AUDITVNODE1 | NOCACHE,
1356 	    pathseg, path, fd, &cap_mknodat_rights);
1357 	if ((error = namei(&nd)) != 0)
1358 		return (error);
1359 	vp = nd.ni_vp;
1360 	if (vp != NULL) {
1361 		NDFREE_PNBUF(&nd);
1362 		if (vp == nd.ni_dvp)
1363 			vrele(nd.ni_dvp);
1364 		else
1365 			vput(nd.ni_dvp);
1366 		vrele(vp);
1367 		return (EEXIST);
1368 	} else {
1369 		VATTR_NULL(&vattr);
1370 		vattr.va_mode = (mode & ALLPERMS) &
1371 		    ~td->td_proc->p_pd->pd_cmask;
1372 		vattr.va_rdev = dev;
1373 		whiteout = 0;
1374 
1375 		switch (mode & S_IFMT) {
1376 		case S_IFCHR:
1377 			vattr.va_type = VCHR;
1378 			break;
1379 		case S_IFBLK:
1380 			vattr.va_type = VBLK;
1381 			break;
1382 		case S_IFWHT:
1383 			whiteout = 1;
1384 			break;
1385 		default:
1386 			panic("kern_mknod: invalid mode");
1387 		}
1388 	}
1389 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1390 		NDFREE_PNBUF(&nd);
1391 		vput(nd.ni_dvp);
1392 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | V_PCATCH)) != 0)
1393 			return (error);
1394 		goto restart;
1395 	}
1396 #ifdef MAC
1397 	if (error == 0 && !whiteout)
1398 		error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp,
1399 		    &nd.ni_cnd, &vattr);
1400 #endif
1401 	if (error == 0) {
1402 		if (whiteout)
1403 			error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE);
1404 		else {
1405 			error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp,
1406 						&nd.ni_cnd, &vattr);
1407 		}
1408 	}
1409 	VOP_VPUT_PAIR(nd.ni_dvp, error == 0 && !whiteout ? &nd.ni_vp : NULL,
1410 	    true);
1411 	vn_finished_write(mp);
1412 	NDFREE_PNBUF(&nd);
1413 	if (error == ERELOOKUP)
1414 		goto restart;
1415 	return (error);
1416 }
1417 
1418 /*
1419  * Create a named pipe.
1420  */
1421 #ifndef _SYS_SYSPROTO_H_
1422 struct mkfifo_args {
1423 	char	*path;
1424 	int	mode;
1425 };
1426 #endif
1427 int
1428 sys_mkfifo(struct thread *td, struct mkfifo_args *uap)
1429 {
1430 
1431 	return (kern_mkfifoat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1432 	    uap->mode));
1433 }
1434 
1435 #ifndef _SYS_SYSPROTO_H_
1436 struct mkfifoat_args {
1437 	int	fd;
1438 	char	*path;
1439 	mode_t	mode;
1440 };
1441 #endif
1442 int
1443 sys_mkfifoat(struct thread *td, struct mkfifoat_args *uap)
1444 {
1445 
1446 	return (kern_mkfifoat(td, uap->fd, uap->path, UIO_USERSPACE,
1447 	    uap->mode));
1448 }
1449 
1450 int
1451 kern_mkfifoat(struct thread *td, int fd, const char *path,
1452     enum uio_seg pathseg, int mode)
1453 {
1454 	struct mount *mp;
1455 	struct vattr vattr;
1456 	struct nameidata nd;
1457 	int error;
1458 
1459 	AUDIT_ARG_MODE(mode);
1460 	NDPREINIT(&nd);
1461 restart:
1462 	bwillwrite();
1463 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | AUDITVNODE1 | NOCACHE,
1464 	    pathseg, path, fd, &cap_mkfifoat_rights);
1465 	if ((error = namei(&nd)) != 0)
1466 		return (error);
1467 	if (nd.ni_vp != NULL) {
1468 		NDFREE_PNBUF(&nd);
1469 		if (nd.ni_vp == nd.ni_dvp)
1470 			vrele(nd.ni_dvp);
1471 		else
1472 			vput(nd.ni_dvp);
1473 		vrele(nd.ni_vp);
1474 		return (EEXIST);
1475 	}
1476 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1477 		NDFREE_PNBUF(&nd);
1478 		vput(nd.ni_dvp);
1479 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | V_PCATCH)) != 0)
1480 			return (error);
1481 		goto restart;
1482 	}
1483 	VATTR_NULL(&vattr);
1484 	vattr.va_type = VFIFO;
1485 	vattr.va_mode = (mode & ALLPERMS) & ~td->td_proc->p_pd->pd_cmask;
1486 #ifdef MAC
1487 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1488 	    &vattr);
1489 	if (error != 0)
1490 		goto out;
1491 #endif
1492 	error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
1493 #ifdef MAC
1494 out:
1495 #endif
1496 	VOP_VPUT_PAIR(nd.ni_dvp, error == 0 ? &nd.ni_vp : NULL, true);
1497 	vn_finished_write(mp);
1498 	NDFREE_PNBUF(&nd);
1499 	if (error == ERELOOKUP)
1500 		goto restart;
1501 	return (error);
1502 }
1503 
1504 /*
1505  * Make a hard file link.
1506  */
1507 #ifndef _SYS_SYSPROTO_H_
1508 struct link_args {
1509 	char	*path;
1510 	char	*link;
1511 };
1512 #endif
1513 int
1514 sys_link(struct thread *td, struct link_args *uap)
1515 {
1516 
1517 	return (kern_linkat(td, AT_FDCWD, AT_FDCWD, uap->path, uap->link,
1518 	    UIO_USERSPACE, AT_SYMLINK_FOLLOW));
1519 }
1520 
1521 #ifndef _SYS_SYSPROTO_H_
1522 struct linkat_args {
1523 	int	fd1;
1524 	char	*path1;
1525 	int	fd2;
1526 	char	*path2;
1527 	int	flag;
1528 };
1529 #endif
1530 int
1531 sys_linkat(struct thread *td, struct linkat_args *uap)
1532 {
1533 
1534 	return (kern_linkat(td, uap->fd1, uap->fd2, uap->path1, uap->path2,
1535 	    UIO_USERSPACE, uap->flag));
1536 }
1537 
1538 int hardlink_check_uid = 0;
1539 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_uid, CTLFLAG_RW,
1540     &hardlink_check_uid, 0,
1541     "Unprivileged processes cannot create hard links to files owned by other "
1542     "users");
1543 static int hardlink_check_gid = 0;
1544 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_gid, CTLFLAG_RW,
1545     &hardlink_check_gid, 0,
1546     "Unprivileged processes cannot create hard links to files owned by other "
1547     "groups");
1548 
1549 static int
1550 can_hardlink(struct vnode *vp, struct ucred *cred)
1551 {
1552 	struct vattr va;
1553 	int error;
1554 
1555 	if (!hardlink_check_uid && !hardlink_check_gid)
1556 		return (0);
1557 
1558 	error = VOP_GETATTR(vp, &va, cred);
1559 	if (error != 0)
1560 		return (error);
1561 
1562 	if (hardlink_check_uid && cred->cr_uid != va.va_uid) {
1563 		error = priv_check_cred(cred, PRIV_VFS_LINK);
1564 		if (error != 0)
1565 			return (error);
1566 	}
1567 
1568 	if (hardlink_check_gid && !groupmember(va.va_gid, cred)) {
1569 		error = priv_check_cred(cred, PRIV_VFS_LINK);
1570 		if (error != 0)
1571 			return (error);
1572 	}
1573 
1574 	return (0);
1575 }
1576 
1577 int
1578 kern_linkat(struct thread *td, int fd1, int fd2, const char *path1,
1579     const char *path2, enum uio_seg segflag, int flag)
1580 {
1581 	struct nameidata nd;
1582 	int error;
1583 
1584 	if ((flag & ~(AT_SYMLINK_FOLLOW | AT_RESOLVE_BENEATH |
1585 	    AT_EMPTY_PATH)) != 0)
1586 		return (EINVAL);
1587 
1588 	NDPREINIT(&nd);
1589 	do {
1590 		bwillwrite();
1591 		NDINIT_ATRIGHTS(&nd, LOOKUP, AUDITVNODE1 | at2cnpflags(flag,
1592 		    AT_SYMLINK_FOLLOW | AT_RESOLVE_BENEATH | AT_EMPTY_PATH),
1593 		    segflag, path1, fd1, &cap_linkat_source_rights);
1594 		if ((error = namei(&nd)) != 0)
1595 			return (error);
1596 		NDFREE_PNBUF(&nd);
1597 		if ((nd.ni_resflags & NIRES_EMPTYPATH) != 0) {
1598 			error = priv_check(td, PRIV_VFS_FHOPEN);
1599 			if (error != 0) {
1600 				vrele(nd.ni_vp);
1601 				return (error);
1602 			}
1603 		}
1604 		error = kern_linkat_vp(td, nd.ni_vp, fd2, path2, segflag);
1605 	} while (error ==  EAGAIN || error == ERELOOKUP);
1606 	return (error);
1607 }
1608 
1609 static int
1610 kern_linkat_vp(struct thread *td, struct vnode *vp, int fd, const char *path,
1611     enum uio_seg segflag)
1612 {
1613 	struct nameidata nd;
1614 	struct mount *mp;
1615 	int error;
1616 
1617 	if (vp->v_type == VDIR) {
1618 		vrele(vp);
1619 		return (EPERM);		/* POSIX */
1620 	}
1621 	NDINIT_ATRIGHTS(&nd, CREATE,
1622 	    LOCKPARENT | AUDITVNODE2 | NOCACHE, segflag, path, fd,
1623 	    &cap_linkat_target_rights);
1624 	if ((error = namei(&nd)) == 0) {
1625 		if (nd.ni_vp != NULL) {
1626 			NDFREE_PNBUF(&nd);
1627 			if (nd.ni_dvp == nd.ni_vp)
1628 				vrele(nd.ni_dvp);
1629 			else
1630 				vput(nd.ni_dvp);
1631 			vrele(nd.ni_vp);
1632 			vrele(vp);
1633 			return (EEXIST);
1634 		} else if (nd.ni_dvp->v_mount != vp->v_mount) {
1635 			/*
1636 			 * Cross-device link.  No need to recheck
1637 			 * vp->v_type, since it cannot change, except
1638 			 * to VBAD.
1639 			 */
1640 			NDFREE_PNBUF(&nd);
1641 			vput(nd.ni_dvp);
1642 			vrele(vp);
1643 			return (EXDEV);
1644 		} else if (vn_lock(vp, LK_EXCLUSIVE) == 0) {
1645 			error = can_hardlink(vp, td->td_ucred);
1646 #ifdef MAC
1647 			if (error == 0)
1648 				error = mac_vnode_check_link(td->td_ucred,
1649 				    nd.ni_dvp, vp, &nd.ni_cnd);
1650 #endif
1651 			if (error != 0) {
1652 				vput(vp);
1653 				vput(nd.ni_dvp);
1654 				NDFREE_PNBUF(&nd);
1655 				return (error);
1656 			}
1657 			error = vn_start_write(vp, &mp, V_NOWAIT);
1658 			if (error != 0) {
1659 				vput(vp);
1660 				vput(nd.ni_dvp);
1661 				NDFREE_PNBUF(&nd);
1662 				error = vn_start_write(NULL, &mp,
1663 				    V_XSLEEP | V_PCATCH);
1664 				if (error != 0)
1665 					return (error);
1666 				return (EAGAIN);
1667 			}
1668 			error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
1669 			VOP_VPUT_PAIR(nd.ni_dvp, &vp, true);
1670 			vn_finished_write(mp);
1671 			NDFREE_PNBUF(&nd);
1672 			vp = NULL;
1673 		} else {
1674 			vput(nd.ni_dvp);
1675 			NDFREE_PNBUF(&nd);
1676 			vrele(vp);
1677 			return (EAGAIN);
1678 		}
1679 	}
1680 	if (vp != NULL)
1681 		vrele(vp);
1682 	return (error);
1683 }
1684 
1685 /*
1686  * Make a symbolic link.
1687  */
1688 #ifndef _SYS_SYSPROTO_H_
1689 struct symlink_args {
1690 	char	*path;
1691 	char	*link;
1692 };
1693 #endif
1694 int
1695 sys_symlink(struct thread *td, struct symlink_args *uap)
1696 {
1697 
1698 	return (kern_symlinkat(td, uap->path, AT_FDCWD, uap->link,
1699 	    UIO_USERSPACE));
1700 }
1701 
1702 #ifndef _SYS_SYSPROTO_H_
1703 struct symlinkat_args {
1704 	char	*path;
1705 	int	fd;
1706 	char	*path2;
1707 };
1708 #endif
1709 int
1710 sys_symlinkat(struct thread *td, struct symlinkat_args *uap)
1711 {
1712 
1713 	return (kern_symlinkat(td, uap->path1, uap->fd, uap->path2,
1714 	    UIO_USERSPACE));
1715 }
1716 
1717 int
1718 kern_symlinkat(struct thread *td, const char *path1, int fd, const char *path2,
1719     enum uio_seg segflg)
1720 {
1721 	struct mount *mp;
1722 	struct vattr vattr;
1723 	const char *syspath;
1724 	char *tmppath;
1725 	struct nameidata nd;
1726 	int error;
1727 
1728 	if (segflg == UIO_SYSSPACE) {
1729 		syspath = path1;
1730 	} else {
1731 		tmppath = uma_zalloc(namei_zone, M_WAITOK);
1732 		if ((error = copyinstr(path1, tmppath, MAXPATHLEN, NULL)) != 0)
1733 			goto out;
1734 		syspath = tmppath;
1735 	}
1736 	AUDIT_ARG_TEXT(syspath);
1737 	NDPREINIT(&nd);
1738 restart:
1739 	bwillwrite();
1740 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | AUDITVNODE1 | NOCACHE, segflg,
1741 	    path2, fd, &cap_symlinkat_rights);
1742 	if ((error = namei(&nd)) != 0)
1743 		goto out;
1744 	if (nd.ni_vp) {
1745 		NDFREE_PNBUF(&nd);
1746 		if (nd.ni_vp == nd.ni_dvp)
1747 			vrele(nd.ni_dvp);
1748 		else
1749 			vput(nd.ni_dvp);
1750 		vrele(nd.ni_vp);
1751 		nd.ni_vp = NULL;
1752 		error = EEXIST;
1753 		goto out;
1754 	}
1755 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1756 		NDFREE_PNBUF(&nd);
1757 		vput(nd.ni_dvp);
1758 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | V_PCATCH)) != 0)
1759 			goto out;
1760 		goto restart;
1761 	}
1762 	VATTR_NULL(&vattr);
1763 	vattr.va_mode = ACCESSPERMS &~ td->td_proc->p_pd->pd_cmask;
1764 #ifdef MAC
1765 	vattr.va_type = VLNK;
1766 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1767 	    &vattr);
1768 	if (error != 0)
1769 		goto out2;
1770 #endif
1771 	error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, syspath);
1772 #ifdef MAC
1773 out2:
1774 #endif
1775 	VOP_VPUT_PAIR(nd.ni_dvp, error == 0 ? &nd.ni_vp : NULL, true);
1776 	vn_finished_write(mp);
1777 	NDFREE_PNBUF(&nd);
1778 	if (error == ERELOOKUP)
1779 		goto restart;
1780 out:
1781 	if (segflg != UIO_SYSSPACE)
1782 		uma_zfree(namei_zone, tmppath);
1783 	return (error);
1784 }
1785 
1786 /*
1787  * Delete a whiteout from the filesystem.
1788  */
1789 #ifndef _SYS_SYSPROTO_H_
1790 struct undelete_args {
1791 	char *path;
1792 };
1793 #endif
1794 int
1795 sys_undelete(struct thread *td, struct undelete_args *uap)
1796 {
1797 	struct mount *mp;
1798 	struct nameidata nd;
1799 	int error;
1800 
1801 	NDPREINIT(&nd);
1802 restart:
1803 	bwillwrite();
1804 	NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | AUDITVNODE1,
1805 	    UIO_USERSPACE, uap->path);
1806 	error = namei(&nd);
1807 	if (error != 0)
1808 		return (error);
1809 
1810 	if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) {
1811 		NDFREE_PNBUF(&nd);
1812 		if (nd.ni_vp == nd.ni_dvp)
1813 			vrele(nd.ni_dvp);
1814 		else
1815 			vput(nd.ni_dvp);
1816 		if (nd.ni_vp)
1817 			vrele(nd.ni_vp);
1818 		return (EEXIST);
1819 	}
1820 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1821 		NDFREE_PNBUF(&nd);
1822 		vput(nd.ni_dvp);
1823 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | V_PCATCH)) != 0)
1824 			return (error);
1825 		goto restart;
1826 	}
1827 	error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE);
1828 	NDFREE_PNBUF(&nd);
1829 	vput(nd.ni_dvp);
1830 	vn_finished_write(mp);
1831 	if (error == ERELOOKUP)
1832 		goto restart;
1833 	return (error);
1834 }
1835 
1836 /*
1837  * Delete a name from the filesystem.
1838  */
1839 #ifndef _SYS_SYSPROTO_H_
1840 struct unlink_args {
1841 	char	*path;
1842 };
1843 #endif
1844 int
1845 sys_unlink(struct thread *td, struct unlink_args *uap)
1846 {
1847 
1848 	return (kern_funlinkat(td, AT_FDCWD, uap->path, FD_NONE, UIO_USERSPACE,
1849 	    0, 0));
1850 }
1851 
1852 static int
1853 kern_funlinkat_ex(struct thread *td, int dfd, const char *path, int fd,
1854     int flag, enum uio_seg pathseg, ino_t oldinum)
1855 {
1856 
1857 	if ((flag & ~(AT_REMOVEDIR | AT_RESOLVE_BENEATH)) != 0)
1858 		return (EINVAL);
1859 
1860 	if ((flag & AT_REMOVEDIR) != 0)
1861 		return (kern_frmdirat(td, dfd, path, fd, UIO_USERSPACE, 0));
1862 
1863 	return (kern_funlinkat(td, dfd, path, fd, UIO_USERSPACE, 0, 0));
1864 }
1865 
1866 #ifndef _SYS_SYSPROTO_H_
1867 struct unlinkat_args {
1868 	int	fd;
1869 	char	*path;
1870 	int	flag;
1871 };
1872 #endif
1873 int
1874 sys_unlinkat(struct thread *td, struct unlinkat_args *uap)
1875 {
1876 
1877 	return (kern_funlinkat_ex(td, uap->fd, uap->path, FD_NONE, uap->flag,
1878 	    UIO_USERSPACE, 0));
1879 }
1880 
1881 #ifndef _SYS_SYSPROTO_H_
1882 struct funlinkat_args {
1883 	int		dfd;
1884 	const char	*path;
1885 	int		fd;
1886 	int		flag;
1887 };
1888 #endif
1889 int
1890 sys_funlinkat(struct thread *td, struct funlinkat_args *uap)
1891 {
1892 
1893 	return (kern_funlinkat_ex(td, uap->dfd, uap->path, uap->fd, uap->flag,
1894 	    UIO_USERSPACE, 0));
1895 }
1896 
1897 int
1898 kern_funlinkat(struct thread *td, int dfd, const char *path, int fd,
1899     enum uio_seg pathseg, int flag, ino_t oldinum)
1900 {
1901 	struct mount *mp;
1902 	struct file *fp;
1903 	struct vnode *vp;
1904 	struct nameidata nd;
1905 	struct stat sb;
1906 	int error;
1907 
1908 	fp = NULL;
1909 	if (fd != FD_NONE) {
1910 		error = getvnode_path(td, fd, &cap_no_rights, &fp);
1911 		if (error != 0)
1912 			return (error);
1913 	}
1914 
1915 	NDPREINIT(&nd);
1916 restart:
1917 	bwillwrite();
1918 	NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1 |
1919 	    at2cnpflags(flag, AT_RESOLVE_BENEATH),
1920 	    pathseg, path, dfd, &cap_unlinkat_rights);
1921 	if ((error = namei(&nd)) != 0) {
1922 		if (error == EINVAL)
1923 			error = EPERM;
1924 		goto fdout;
1925 	}
1926 	vp = nd.ni_vp;
1927 	if (vp->v_type == VDIR && oldinum == 0) {
1928 		error = EPERM;		/* POSIX */
1929 	} else if (oldinum != 0 &&
1930 	    ((error = VOP_STAT(vp, &sb, td->td_ucred, NOCRED)) == 0) &&
1931 	    sb.st_ino != oldinum) {
1932 		error = EIDRM;	/* Identifier removed */
1933 	} else if (fp != NULL && fp->f_vnode != vp) {
1934 		if (VN_IS_DOOMED(fp->f_vnode))
1935 			error = EBADF;
1936 		else
1937 			error = EDEADLK;
1938 	} else {
1939 		/*
1940 		 * The root of a mounted filesystem cannot be deleted.
1941 		 *
1942 		 * XXX: can this only be a VDIR case?
1943 		 */
1944 		if (vp->v_vflag & VV_ROOT)
1945 			error = EBUSY;
1946 	}
1947 	if (error == 0) {
1948 		if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1949 			NDFREE_PNBUF(&nd);
1950 			vput(nd.ni_dvp);
1951 			if (vp == nd.ni_dvp)
1952 				vrele(vp);
1953 			else
1954 				vput(vp);
1955 			if ((error = vn_start_write(NULL, &mp,
1956 			    V_XSLEEP | V_PCATCH)) != 0) {
1957 				goto fdout;
1958 			}
1959 			goto restart;
1960 		}
1961 #ifdef MAC
1962 		error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
1963 		    &nd.ni_cnd);
1964 		if (error != 0)
1965 			goto out;
1966 #endif
1967 		vfs_notify_upper(vp, VFS_NOTIFY_UPPER_UNLINK);
1968 		error = VOP_REMOVE(nd.ni_dvp, vp, &nd.ni_cnd);
1969 #ifdef MAC
1970 out:
1971 #endif
1972 		vn_finished_write(mp);
1973 	}
1974 	NDFREE_PNBUF(&nd);
1975 	vput(nd.ni_dvp);
1976 	if (vp == nd.ni_dvp)
1977 		vrele(vp);
1978 	else
1979 		vput(vp);
1980 	if (error == ERELOOKUP)
1981 		goto restart;
1982 fdout:
1983 	if (fp != NULL)
1984 		fdrop(fp, td);
1985 	return (error);
1986 }
1987 
1988 /*
1989  * Reposition read/write file offset.
1990  */
1991 #ifndef _SYS_SYSPROTO_H_
1992 struct lseek_args {
1993 	int	fd;
1994 	int	pad;
1995 	off_t	offset;
1996 	int	whence;
1997 };
1998 #endif
1999 int
2000 sys_lseek(struct thread *td, struct lseek_args *uap)
2001 {
2002 
2003 	return (kern_lseek(td, uap->fd, uap->offset, uap->whence));
2004 }
2005 
2006 int
2007 kern_lseek(struct thread *td, int fd, off_t offset, int whence)
2008 {
2009 	struct file *fp;
2010 	int error;
2011 
2012 	AUDIT_ARG_FD(fd);
2013 	error = fget(td, fd, &cap_seek_rights, &fp);
2014 	if (error != 0)
2015 		return (error);
2016 	error = (fp->f_ops->fo_flags & DFLAG_SEEKABLE) != 0 ?
2017 	    fo_seek(fp, offset, whence, td) : ESPIPE;
2018 	fdrop(fp, td);
2019 	return (error);
2020 }
2021 
2022 #if defined(COMPAT_43)
2023 /*
2024  * Reposition read/write file offset.
2025  */
2026 #ifndef _SYS_SYSPROTO_H_
2027 struct olseek_args {
2028 	int	fd;
2029 	long	offset;
2030 	int	whence;
2031 };
2032 #endif
2033 int
2034 olseek(struct thread *td, struct olseek_args *uap)
2035 {
2036 
2037 	return (kern_lseek(td, uap->fd, uap->offset, uap->whence));
2038 }
2039 #endif /* COMPAT_43 */
2040 
2041 #if defined(COMPAT_FREEBSD6)
2042 /* Version with the 'pad' argument */
2043 int
2044 freebsd6_lseek(struct thread *td, struct freebsd6_lseek_args *uap)
2045 {
2046 
2047 	return (kern_lseek(td, uap->fd, uap->offset, uap->whence));
2048 }
2049 #endif
2050 
2051 /*
2052  * Check access permissions using passed credentials.
2053  */
2054 static int
2055 vn_access(struct vnode *vp, int user_flags, struct ucred *cred,
2056      struct thread *td)
2057 {
2058 	accmode_t accmode;
2059 	int error;
2060 
2061 	/* Flags == 0 means only check for existence. */
2062 	if (user_flags == 0)
2063 		return (0);
2064 
2065 	accmode = 0;
2066 	if (user_flags & R_OK)
2067 		accmode |= VREAD;
2068 	if (user_flags & W_OK)
2069 		accmode |= VWRITE;
2070 	if (user_flags & X_OK)
2071 		accmode |= VEXEC;
2072 #ifdef MAC
2073 	error = mac_vnode_check_access(cred, vp, accmode);
2074 	if (error != 0)
2075 		return (error);
2076 #endif
2077 	if ((accmode & VWRITE) == 0 || (error = vn_writechk(vp)) == 0)
2078 		error = VOP_ACCESS(vp, accmode, cred, td);
2079 	return (error);
2080 }
2081 
2082 /*
2083  * Check access permissions using "real" credentials.
2084  */
2085 #ifndef _SYS_SYSPROTO_H_
2086 struct access_args {
2087 	char	*path;
2088 	int	amode;
2089 };
2090 #endif
2091 int
2092 sys_access(struct thread *td, struct access_args *uap)
2093 {
2094 
2095 	return (kern_accessat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2096 	    0, uap->amode));
2097 }
2098 
2099 #ifndef _SYS_SYSPROTO_H_
2100 struct faccessat_args {
2101 	int	dirfd;
2102 	char	*path;
2103 	int	amode;
2104 	int	flag;
2105 }
2106 #endif
2107 int
2108 sys_faccessat(struct thread *td, struct faccessat_args *uap)
2109 {
2110 
2111 	return (kern_accessat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
2112 	    uap->amode));
2113 }
2114 
2115 int
2116 kern_accessat(struct thread *td, int fd, const char *path,
2117     enum uio_seg pathseg, int flag, int amode)
2118 {
2119 	struct ucred *cred, *usecred;
2120 	struct vnode *vp;
2121 	struct nameidata nd;
2122 	int error;
2123 
2124 	if ((flag & ~(AT_EACCESS | AT_RESOLVE_BENEATH | AT_EMPTY_PATH)) != 0)
2125 		return (EINVAL);
2126 	if (amode != F_OK && (amode & ~(R_OK | W_OK | X_OK)) != 0)
2127 		return (EINVAL);
2128 
2129 	/*
2130 	 * Create and modify a temporary credential instead of one that
2131 	 * is potentially shared (if we need one).
2132 	 */
2133 	cred = td->td_ucred;
2134 	if ((flag & AT_EACCESS) == 0 &&
2135 	    ((cred->cr_uid != cred->cr_ruid ||
2136 	    cred->cr_rgid != cred->cr_groups[0]))) {
2137 		usecred = crdup(cred);
2138 		usecred->cr_uid = cred->cr_ruid;
2139 		usecred->cr_groups[0] = cred->cr_rgid;
2140 		td->td_ucred = usecred;
2141 	} else
2142 		usecred = cred;
2143 	AUDIT_ARG_VALUE(amode);
2144 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF |
2145 	    AUDITVNODE1 | at2cnpflags(flag, AT_RESOLVE_BENEATH |
2146 	    AT_EMPTY_PATH), pathseg, path, fd, &cap_fstat_rights);
2147 	if ((error = namei(&nd)) != 0)
2148 		goto out;
2149 	vp = nd.ni_vp;
2150 
2151 	error = vn_access(vp, amode, usecred, td);
2152 	NDFREE_PNBUF(&nd);
2153 	vput(vp);
2154 out:
2155 	if (usecred != cred) {
2156 		td->td_ucred = cred;
2157 		crfree(usecred);
2158 	}
2159 	return (error);
2160 }
2161 
2162 /*
2163  * Check access permissions using "effective" credentials.
2164  */
2165 #ifndef _SYS_SYSPROTO_H_
2166 struct eaccess_args {
2167 	char	*path;
2168 	int	amode;
2169 };
2170 #endif
2171 int
2172 sys_eaccess(struct thread *td, struct eaccess_args *uap)
2173 {
2174 
2175 	return (kern_accessat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2176 	    AT_EACCESS, uap->amode));
2177 }
2178 
2179 #if defined(COMPAT_43)
2180 /*
2181  * Get file status; this version follows links.
2182  */
2183 #ifndef _SYS_SYSPROTO_H_
2184 struct ostat_args {
2185 	char	*path;
2186 	struct ostat *ub;
2187 };
2188 #endif
2189 int
2190 ostat(struct thread *td, struct ostat_args *uap)
2191 {
2192 	struct stat sb;
2193 	struct ostat osb;
2194 	int error;
2195 
2196 	error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE, &sb);
2197 	if (error != 0)
2198 		return (error);
2199 	cvtstat(&sb, &osb);
2200 	return (copyout(&osb, uap->ub, sizeof (osb)));
2201 }
2202 
2203 /*
2204  * Get file status; this version does not follow links.
2205  */
2206 #ifndef _SYS_SYSPROTO_H_
2207 struct olstat_args {
2208 	char	*path;
2209 	struct ostat *ub;
2210 };
2211 #endif
2212 int
2213 olstat(struct thread *td, struct olstat_args *uap)
2214 {
2215 	struct stat sb;
2216 	struct ostat osb;
2217 	int error;
2218 
2219 	error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2220 	    UIO_USERSPACE, &sb);
2221 	if (error != 0)
2222 		return (error);
2223 	cvtstat(&sb, &osb);
2224 	return (copyout(&osb, uap->ub, sizeof (osb)));
2225 }
2226 
2227 /*
2228  * Convert from an old to a new stat structure.
2229  * XXX: many values are blindly truncated.
2230  */
2231 void
2232 cvtstat(struct stat *st, struct ostat *ost)
2233 {
2234 
2235 	bzero(ost, sizeof(*ost));
2236 	ost->st_dev = st->st_dev;
2237 	ost->st_ino = st->st_ino;
2238 	ost->st_mode = st->st_mode;
2239 	ost->st_nlink = st->st_nlink;
2240 	ost->st_uid = st->st_uid;
2241 	ost->st_gid = st->st_gid;
2242 	ost->st_rdev = st->st_rdev;
2243 	ost->st_size = MIN(st->st_size, INT32_MAX);
2244 	ost->st_atim = st->st_atim;
2245 	ost->st_mtim = st->st_mtim;
2246 	ost->st_ctim = st->st_ctim;
2247 	ost->st_blksize = st->st_blksize;
2248 	ost->st_blocks = st->st_blocks;
2249 	ost->st_flags = st->st_flags;
2250 	ost->st_gen = st->st_gen;
2251 }
2252 #endif /* COMPAT_43 */
2253 
2254 #if defined(COMPAT_43) || defined(COMPAT_FREEBSD11)
2255 int ino64_trunc_error;
2256 SYSCTL_INT(_vfs, OID_AUTO, ino64_trunc_error, CTLFLAG_RW,
2257     &ino64_trunc_error, 0,
2258     "Error on truncation of device, file or inode number, or link count");
2259 
2260 int
2261 freebsd11_cvtstat(struct stat *st, struct freebsd11_stat *ost)
2262 {
2263 
2264 	ost->st_dev = st->st_dev;
2265 	if (ost->st_dev != st->st_dev) {
2266 		switch (ino64_trunc_error) {
2267 		default:
2268 			/*
2269 			 * Since dev_t is almost raw, don't clamp to the
2270 			 * maximum for case 2, but ignore the error.
2271 			 */
2272 			break;
2273 		case 1:
2274 			return (EOVERFLOW);
2275 		}
2276 	}
2277 	ost->st_ino = st->st_ino;
2278 	if (ost->st_ino != st->st_ino) {
2279 		switch (ino64_trunc_error) {
2280 		default:
2281 		case 0:
2282 			break;
2283 		case 1:
2284 			return (EOVERFLOW);
2285 		case 2:
2286 			ost->st_ino = UINT32_MAX;
2287 			break;
2288 		}
2289 	}
2290 	ost->st_mode = st->st_mode;
2291 	ost->st_nlink = st->st_nlink;
2292 	if (ost->st_nlink != st->st_nlink) {
2293 		switch (ino64_trunc_error) {
2294 		default:
2295 		case 0:
2296 			break;
2297 		case 1:
2298 			return (EOVERFLOW);
2299 		case 2:
2300 			ost->st_nlink = UINT16_MAX;
2301 			break;
2302 		}
2303 	}
2304 	ost->st_uid = st->st_uid;
2305 	ost->st_gid = st->st_gid;
2306 	ost->st_rdev = st->st_rdev;
2307 	if (ost->st_rdev != st->st_rdev) {
2308 		switch (ino64_trunc_error) {
2309 		default:
2310 			break;
2311 		case 1:
2312 			return (EOVERFLOW);
2313 		}
2314 	}
2315 	ost->st_atim = st->st_atim;
2316 	ost->st_mtim = st->st_mtim;
2317 	ost->st_ctim = st->st_ctim;
2318 	ost->st_size = st->st_size;
2319 	ost->st_blocks = st->st_blocks;
2320 	ost->st_blksize = st->st_blksize;
2321 	ost->st_flags = st->st_flags;
2322 	ost->st_gen = st->st_gen;
2323 	ost->st_lspare = 0;
2324 	ost->st_birthtim = st->st_birthtim;
2325 	bzero((char *)&ost->st_birthtim + sizeof(ost->st_birthtim),
2326 	    sizeof(*ost) - offsetof(struct freebsd11_stat,
2327 	    st_birthtim) - sizeof(ost->st_birthtim));
2328 	return (0);
2329 }
2330 
2331 int
2332 freebsd11_stat(struct thread *td, struct freebsd11_stat_args* uap)
2333 {
2334 	struct stat sb;
2335 	struct freebsd11_stat osb;
2336 	int error;
2337 
2338 	error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE, &sb);
2339 	if (error != 0)
2340 		return (error);
2341 	error = freebsd11_cvtstat(&sb, &osb);
2342 	if (error == 0)
2343 		error = copyout(&osb, uap->ub, sizeof(osb));
2344 	return (error);
2345 }
2346 
2347 int
2348 freebsd11_lstat(struct thread *td, struct freebsd11_lstat_args* uap)
2349 {
2350 	struct stat sb;
2351 	struct freebsd11_stat osb;
2352 	int error;
2353 
2354 	error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2355 	    UIO_USERSPACE, &sb);
2356 	if (error != 0)
2357 		return (error);
2358 	error = freebsd11_cvtstat(&sb, &osb);
2359 	if (error == 0)
2360 		error = copyout(&osb, uap->ub, sizeof(osb));
2361 	return (error);
2362 }
2363 
2364 int
2365 freebsd11_fhstat(struct thread *td, struct freebsd11_fhstat_args* uap)
2366 {
2367 	struct fhandle fh;
2368 	struct stat sb;
2369 	struct freebsd11_stat osb;
2370 	int error;
2371 
2372 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
2373 	if (error != 0)
2374 		return (error);
2375 	error = kern_fhstat(td, fh, &sb);
2376 	if (error != 0)
2377 		return (error);
2378 	error = freebsd11_cvtstat(&sb, &osb);
2379 	if (error == 0)
2380 		error = copyout(&osb, uap->sb, sizeof(osb));
2381 	return (error);
2382 }
2383 
2384 int
2385 freebsd11_fstatat(struct thread *td, struct freebsd11_fstatat_args* uap)
2386 {
2387 	struct stat sb;
2388 	struct freebsd11_stat osb;
2389 	int error;
2390 
2391 	error = kern_statat(td, uap->flag, uap->fd, uap->path,
2392 	    UIO_USERSPACE, &sb);
2393 	if (error != 0)
2394 		return (error);
2395 	error = freebsd11_cvtstat(&sb, &osb);
2396 	if (error == 0)
2397 		error = copyout(&osb, uap->buf, sizeof(osb));
2398 	return (error);
2399 }
2400 #endif	/* COMPAT_FREEBSD11 */
2401 
2402 /*
2403  * Get file status
2404  */
2405 #ifndef _SYS_SYSPROTO_H_
2406 struct fstatat_args {
2407 	int	fd;
2408 	char	*path;
2409 	struct stat	*buf;
2410 	int	flag;
2411 }
2412 #endif
2413 int
2414 sys_fstatat(struct thread *td, struct fstatat_args *uap)
2415 {
2416 	struct stat sb;
2417 	int error;
2418 
2419 	error = kern_statat(td, uap->flag, uap->fd, uap->path,
2420 	    UIO_USERSPACE, &sb);
2421 	if (error == 0)
2422 		error = copyout(&sb, uap->buf, sizeof (sb));
2423 	return (error);
2424 }
2425 
2426 int
2427 kern_statat(struct thread *td, int flag, int fd, const char *path,
2428     enum uio_seg pathseg, struct stat *sbp)
2429 {
2430 	struct nameidata nd;
2431 	int error;
2432 
2433 	if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH |
2434 	    AT_EMPTY_PATH)) != 0)
2435 		return (EINVAL);
2436 
2437 	NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(flag, AT_RESOLVE_BENEATH |
2438 	    AT_SYMLINK_NOFOLLOW | AT_EMPTY_PATH) | LOCKSHARED | LOCKLEAF |
2439 	    AUDITVNODE1, pathseg, path, fd, &cap_fstat_rights);
2440 
2441 	if ((error = namei(&nd)) != 0) {
2442 		if (error == ENOTDIR &&
2443 		    (nd.ni_resflags & NIRES_EMPTYPATH) != 0)
2444 			error = kern_fstat(td, fd, sbp);
2445 		return (error);
2446 	}
2447 	error = VOP_STAT(nd.ni_vp, sbp, td->td_ucred, NOCRED);
2448 	NDFREE_PNBUF(&nd);
2449 	vput(nd.ni_vp);
2450 #ifdef __STAT_TIME_T_EXT
2451 	sbp->st_atim_ext = 0;
2452 	sbp->st_mtim_ext = 0;
2453 	sbp->st_ctim_ext = 0;
2454 	sbp->st_btim_ext = 0;
2455 #endif
2456 #ifdef KTRACE
2457 	if (KTRPOINT(td, KTR_STRUCT))
2458 		ktrstat_error(sbp, error);
2459 #endif
2460 	return (error);
2461 }
2462 
2463 #if defined(COMPAT_FREEBSD11)
2464 /*
2465  * Implementation of the NetBSD [l]stat() functions.
2466  */
2467 int
2468 freebsd11_cvtnstat(struct stat *sb, struct nstat *nsb)
2469 {
2470 	struct freebsd11_stat sb11;
2471 	int error;
2472 
2473 	error = freebsd11_cvtstat(sb, &sb11);
2474 	if (error != 0)
2475 		return (error);
2476 
2477 	bzero(nsb, sizeof(*nsb));
2478 	CP(sb11, *nsb, st_dev);
2479 	CP(sb11, *nsb, st_ino);
2480 	CP(sb11, *nsb, st_mode);
2481 	CP(sb11, *nsb, st_nlink);
2482 	CP(sb11, *nsb, st_uid);
2483 	CP(sb11, *nsb, st_gid);
2484 	CP(sb11, *nsb, st_rdev);
2485 	CP(sb11, *nsb, st_atim);
2486 	CP(sb11, *nsb, st_mtim);
2487 	CP(sb11, *nsb, st_ctim);
2488 	CP(sb11, *nsb, st_size);
2489 	CP(sb11, *nsb, st_blocks);
2490 	CP(sb11, *nsb, st_blksize);
2491 	CP(sb11, *nsb, st_flags);
2492 	CP(sb11, *nsb, st_gen);
2493 	CP(sb11, *nsb, st_birthtim);
2494 	return (0);
2495 }
2496 
2497 #ifndef _SYS_SYSPROTO_H_
2498 struct freebsd11_nstat_args {
2499 	char	*path;
2500 	struct nstat *ub;
2501 };
2502 #endif
2503 int
2504 freebsd11_nstat(struct thread *td, struct freebsd11_nstat_args *uap)
2505 {
2506 	struct stat sb;
2507 	struct nstat nsb;
2508 	int error;
2509 
2510 	error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE, &sb);
2511 	if (error != 0)
2512 		return (error);
2513 	error = freebsd11_cvtnstat(&sb, &nsb);
2514 	if (error == 0)
2515 		error = copyout(&nsb, uap->ub, sizeof (nsb));
2516 	return (error);
2517 }
2518 
2519 /*
2520  * NetBSD lstat.  Get file status; this version does not follow links.
2521  */
2522 #ifndef _SYS_SYSPROTO_H_
2523 struct freebsd11_nlstat_args {
2524 	char	*path;
2525 	struct nstat *ub;
2526 };
2527 #endif
2528 int
2529 freebsd11_nlstat(struct thread *td, struct freebsd11_nlstat_args *uap)
2530 {
2531 	struct stat sb;
2532 	struct nstat nsb;
2533 	int error;
2534 
2535 	error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2536 	    UIO_USERSPACE, &sb);
2537 	if (error != 0)
2538 		return (error);
2539 	error = freebsd11_cvtnstat(&sb, &nsb);
2540 	if (error == 0)
2541 		error = copyout(&nsb, uap->ub, sizeof (nsb));
2542 	return (error);
2543 }
2544 #endif /* COMPAT_FREEBSD11 */
2545 
2546 /*
2547  * Get configurable pathname variables.
2548  */
2549 #ifndef _SYS_SYSPROTO_H_
2550 struct pathconf_args {
2551 	char	*path;
2552 	int	name;
2553 };
2554 #endif
2555 int
2556 sys_pathconf(struct thread *td, struct pathconf_args *uap)
2557 {
2558 	long value;
2559 	int error;
2560 
2561 	error = kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name, FOLLOW,
2562 	    &value);
2563 	if (error == 0)
2564 		td->td_retval[0] = value;
2565 	return (error);
2566 }
2567 
2568 #ifndef _SYS_SYSPROTO_H_
2569 struct lpathconf_args {
2570 	char	*path;
2571 	int	name;
2572 };
2573 #endif
2574 int
2575 sys_lpathconf(struct thread *td, struct lpathconf_args *uap)
2576 {
2577 	long value;
2578 	int error;
2579 
2580 	error = kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name,
2581 	    NOFOLLOW, &value);
2582 	if (error == 0)
2583 		td->td_retval[0] = value;
2584 	return (error);
2585 }
2586 
2587 int
2588 kern_pathconf(struct thread *td, const char *path, enum uio_seg pathseg,
2589     int name, u_long flags, long *valuep)
2590 {
2591 	struct nameidata nd;
2592 	int error;
2593 
2594 	NDINIT(&nd, LOOKUP, LOCKSHARED | LOCKLEAF | AUDITVNODE1 | flags,
2595 	       pathseg, path);
2596 	if ((error = namei(&nd)) != 0)
2597 		return (error);
2598 	NDFREE_PNBUF(&nd);
2599 
2600 	error = VOP_PATHCONF(nd.ni_vp, name, valuep);
2601 	vput(nd.ni_vp);
2602 	return (error);
2603 }
2604 
2605 /*
2606  * Return target name of a symbolic link.
2607  */
2608 #ifndef _SYS_SYSPROTO_H_
2609 struct readlink_args {
2610 	char	*path;
2611 	char	*buf;
2612 	size_t	count;
2613 };
2614 #endif
2615 int
2616 sys_readlink(struct thread *td, struct readlink_args *uap)
2617 {
2618 
2619 	return (kern_readlinkat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2620 	    uap->buf, UIO_USERSPACE, uap->count));
2621 }
2622 #ifndef _SYS_SYSPROTO_H_
2623 struct readlinkat_args {
2624 	int	fd;
2625 	char	*path;
2626 	char	*buf;
2627 	size_t	bufsize;
2628 };
2629 #endif
2630 int
2631 sys_readlinkat(struct thread *td, struct readlinkat_args *uap)
2632 {
2633 
2634 	return (kern_readlinkat(td, uap->fd, uap->path, UIO_USERSPACE,
2635 	    uap->buf, UIO_USERSPACE, uap->bufsize));
2636 }
2637 
2638 int
2639 kern_readlinkat(struct thread *td, int fd, const char *path,
2640     enum uio_seg pathseg, char *buf, enum uio_seg bufseg, size_t count)
2641 {
2642 	struct vnode *vp;
2643 	struct nameidata nd;
2644 	int error;
2645 
2646 	if (count > IOSIZE_MAX)
2647 		return (EINVAL);
2648 
2649 	NDINIT_AT(&nd, LOOKUP, NOFOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1 |
2650 	    EMPTYPATH, pathseg, path, fd);
2651 
2652 	if ((error = namei(&nd)) != 0)
2653 		return (error);
2654 	NDFREE_PNBUF(&nd);
2655 	vp = nd.ni_vp;
2656 
2657 	error = kern_readlink_vp(vp, buf, bufseg, count, td);
2658 	vput(vp);
2659 
2660 	return (error);
2661 }
2662 
2663 /*
2664  * Helper function to readlink from a vnode
2665  */
2666 static int
2667 kern_readlink_vp(struct vnode *vp, char *buf, enum uio_seg bufseg, size_t count,
2668     struct thread *td)
2669 {
2670 	struct iovec aiov;
2671 	struct uio auio;
2672 	int error;
2673 
2674 	ASSERT_VOP_LOCKED(vp, "kern_readlink_vp(): vp not locked");
2675 #ifdef MAC
2676 	error = mac_vnode_check_readlink(td->td_ucred, vp);
2677 	if (error != 0)
2678 		return (error);
2679 #endif
2680 	if (vp->v_type != VLNK && (vp->v_vflag & VV_READLINK) == 0)
2681 		return (EINVAL);
2682 
2683 	aiov.iov_base = buf;
2684 	aiov.iov_len = count;
2685 	auio.uio_iov = &aiov;
2686 	auio.uio_iovcnt = 1;
2687 	auio.uio_offset = 0;
2688 	auio.uio_rw = UIO_READ;
2689 	auio.uio_segflg = bufseg;
2690 	auio.uio_td = td;
2691 	auio.uio_resid = count;
2692 	error = VOP_READLINK(vp, &auio, td->td_ucred);
2693 	td->td_retval[0] = count - auio.uio_resid;
2694 	return (error);
2695 }
2696 
2697 /*
2698  * Common implementation code for chflags() and fchflags().
2699  */
2700 static int
2701 setfflags(struct thread *td, struct vnode *vp, u_long flags)
2702 {
2703 	struct mount *mp;
2704 	struct vattr vattr;
2705 	int error;
2706 
2707 	/* We can't support the value matching VNOVAL. */
2708 	if (flags == VNOVAL)
2709 		return (EOPNOTSUPP);
2710 
2711 	/*
2712 	 * Prevent non-root users from setting flags on devices.  When
2713 	 * a device is reused, users can retain ownership of the device
2714 	 * if they are allowed to set flags and programs assume that
2715 	 * chown can't fail when done as root.
2716 	 */
2717 	if (vp->v_type == VCHR || vp->v_type == VBLK) {
2718 		error = priv_check(td, PRIV_VFS_CHFLAGS_DEV);
2719 		if (error != 0)
2720 			return (error);
2721 	}
2722 
2723 	if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0)
2724 		return (error);
2725 	VATTR_NULL(&vattr);
2726 	vattr.va_flags = flags;
2727 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2728 #ifdef MAC
2729 	error = mac_vnode_check_setflags(td->td_ucred, vp, vattr.va_flags);
2730 	if (error == 0)
2731 #endif
2732 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
2733 	VOP_UNLOCK(vp);
2734 	vn_finished_write(mp);
2735 	return (error);
2736 }
2737 
2738 /*
2739  * Change flags of a file given a path name.
2740  */
2741 #ifndef _SYS_SYSPROTO_H_
2742 struct chflags_args {
2743 	const char *path;
2744 	u_long	flags;
2745 };
2746 #endif
2747 int
2748 sys_chflags(struct thread *td, struct chflags_args *uap)
2749 {
2750 
2751 	return (kern_chflagsat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2752 	    uap->flags, 0));
2753 }
2754 
2755 #ifndef _SYS_SYSPROTO_H_
2756 struct chflagsat_args {
2757 	int	fd;
2758 	const char *path;
2759 	u_long	flags;
2760 	int	atflag;
2761 }
2762 #endif
2763 int
2764 sys_chflagsat(struct thread *td, struct chflagsat_args *uap)
2765 {
2766 
2767 	return (kern_chflagsat(td, uap->fd, uap->path, UIO_USERSPACE,
2768 	    uap->flags, uap->atflag));
2769 }
2770 
2771 /*
2772  * Same as chflags() but doesn't follow symlinks.
2773  */
2774 #ifndef _SYS_SYSPROTO_H_
2775 struct lchflags_args {
2776 	const char *path;
2777 	u_long flags;
2778 };
2779 #endif
2780 int
2781 sys_lchflags(struct thread *td, struct lchflags_args *uap)
2782 {
2783 
2784 	return (kern_chflagsat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2785 	    uap->flags, AT_SYMLINK_NOFOLLOW));
2786 }
2787 
2788 static int
2789 kern_chflagsat(struct thread *td, int fd, const char *path,
2790     enum uio_seg pathseg, u_long flags, int atflag)
2791 {
2792 	struct nameidata nd;
2793 	int error;
2794 
2795 	if ((atflag & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH |
2796 	    AT_EMPTY_PATH)) != 0)
2797 		return (EINVAL);
2798 
2799 	AUDIT_ARG_FFLAGS(flags);
2800 	NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(atflag, AT_SYMLINK_NOFOLLOW |
2801 	    AT_RESOLVE_BENEATH | AT_EMPTY_PATH) | AUDITVNODE1, pathseg, path,
2802 	    fd, &cap_fchflags_rights);
2803 	if ((error = namei(&nd)) != 0)
2804 		return (error);
2805 	NDFREE_PNBUF(&nd);
2806 	error = setfflags(td, nd.ni_vp, flags);
2807 	vrele(nd.ni_vp);
2808 	return (error);
2809 }
2810 
2811 /*
2812  * Change flags of a file given a file descriptor.
2813  */
2814 #ifndef _SYS_SYSPROTO_H_
2815 struct fchflags_args {
2816 	int	fd;
2817 	u_long	flags;
2818 };
2819 #endif
2820 int
2821 sys_fchflags(struct thread *td, struct fchflags_args *uap)
2822 {
2823 	struct file *fp;
2824 	int error;
2825 
2826 	AUDIT_ARG_FD(uap->fd);
2827 	AUDIT_ARG_FFLAGS(uap->flags);
2828 	error = getvnode(td, uap->fd, &cap_fchflags_rights,
2829 	    &fp);
2830 	if (error != 0)
2831 		return (error);
2832 #ifdef AUDIT
2833 	if (AUDITING_TD(td)) {
2834 		vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
2835 		AUDIT_ARG_VNODE1(fp->f_vnode);
2836 		VOP_UNLOCK(fp->f_vnode);
2837 	}
2838 #endif
2839 	error = setfflags(td, fp->f_vnode, uap->flags);
2840 	fdrop(fp, td);
2841 	return (error);
2842 }
2843 
2844 /*
2845  * Common implementation code for chmod(), lchmod() and fchmod().
2846  */
2847 int
2848 setfmode(struct thread *td, struct ucred *cred, struct vnode *vp, int mode)
2849 {
2850 	struct mount *mp;
2851 	struct vattr vattr;
2852 	int error;
2853 
2854 	if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0)
2855 		return (error);
2856 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2857 	VATTR_NULL(&vattr);
2858 	vattr.va_mode = mode & ALLPERMS;
2859 #ifdef MAC
2860 	error = mac_vnode_check_setmode(cred, vp, vattr.va_mode);
2861 	if (error == 0)
2862 #endif
2863 		error = VOP_SETATTR(vp, &vattr, cred);
2864 	VOP_UNLOCK(vp);
2865 	vn_finished_write(mp);
2866 	return (error);
2867 }
2868 
2869 /*
2870  * Change mode of a file given path name.
2871  */
2872 #ifndef _SYS_SYSPROTO_H_
2873 struct chmod_args {
2874 	char	*path;
2875 	int	mode;
2876 };
2877 #endif
2878 int
2879 sys_chmod(struct thread *td, struct chmod_args *uap)
2880 {
2881 
2882 	return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2883 	    uap->mode, 0));
2884 }
2885 
2886 #ifndef _SYS_SYSPROTO_H_
2887 struct fchmodat_args {
2888 	int	dirfd;
2889 	char	*path;
2890 	mode_t	mode;
2891 	int	flag;
2892 }
2893 #endif
2894 int
2895 sys_fchmodat(struct thread *td, struct fchmodat_args *uap)
2896 {
2897 
2898 	return (kern_fchmodat(td, uap->fd, uap->path, UIO_USERSPACE,
2899 	    uap->mode, uap->flag));
2900 }
2901 
2902 /*
2903  * Change mode of a file given path name (don't follow links.)
2904  */
2905 #ifndef _SYS_SYSPROTO_H_
2906 struct lchmod_args {
2907 	char	*path;
2908 	int	mode;
2909 };
2910 #endif
2911 int
2912 sys_lchmod(struct thread *td, struct lchmod_args *uap)
2913 {
2914 
2915 	return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2916 	    uap->mode, AT_SYMLINK_NOFOLLOW));
2917 }
2918 
2919 int
2920 kern_fchmodat(struct thread *td, int fd, const char *path,
2921     enum uio_seg pathseg, mode_t mode, int flag)
2922 {
2923 	struct nameidata nd;
2924 	int error;
2925 
2926 	if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH |
2927 	    AT_EMPTY_PATH)) != 0)
2928 		return (EINVAL);
2929 
2930 	AUDIT_ARG_MODE(mode);
2931 	NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(flag, AT_SYMLINK_NOFOLLOW |
2932 	    AT_RESOLVE_BENEATH | AT_EMPTY_PATH) | AUDITVNODE1, pathseg, path,
2933 	    fd, &cap_fchmod_rights);
2934 	if ((error = namei(&nd)) != 0)
2935 		return (error);
2936 	NDFREE_PNBUF(&nd);
2937 	error = setfmode(td, td->td_ucred, nd.ni_vp, mode);
2938 	vrele(nd.ni_vp);
2939 	return (error);
2940 }
2941 
2942 /*
2943  * Change mode of a file given a file descriptor.
2944  */
2945 #ifndef _SYS_SYSPROTO_H_
2946 struct fchmod_args {
2947 	int	fd;
2948 	int	mode;
2949 };
2950 #endif
2951 int
2952 sys_fchmod(struct thread *td, struct fchmod_args *uap)
2953 {
2954 	struct file *fp;
2955 	int error;
2956 
2957 	AUDIT_ARG_FD(uap->fd);
2958 	AUDIT_ARG_MODE(uap->mode);
2959 
2960 	error = fget(td, uap->fd, &cap_fchmod_rights, &fp);
2961 	if (error != 0)
2962 		return (error);
2963 	error = fo_chmod(fp, uap->mode, td->td_ucred, td);
2964 	fdrop(fp, td);
2965 	return (error);
2966 }
2967 
2968 /*
2969  * Common implementation for chown(), lchown(), and fchown()
2970  */
2971 int
2972 setfown(struct thread *td, struct ucred *cred, struct vnode *vp, uid_t uid,
2973     gid_t gid)
2974 {
2975 	struct mount *mp;
2976 	struct vattr vattr;
2977 	int error;
2978 
2979 	if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0)
2980 		return (error);
2981 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2982 	VATTR_NULL(&vattr);
2983 	vattr.va_uid = uid;
2984 	vattr.va_gid = gid;
2985 #ifdef MAC
2986 	error = mac_vnode_check_setowner(cred, vp, vattr.va_uid,
2987 	    vattr.va_gid);
2988 	if (error == 0)
2989 #endif
2990 		error = VOP_SETATTR(vp, &vattr, cred);
2991 	VOP_UNLOCK(vp);
2992 	vn_finished_write(mp);
2993 	return (error);
2994 }
2995 
2996 /*
2997  * Set ownership given a path name.
2998  */
2999 #ifndef _SYS_SYSPROTO_H_
3000 struct chown_args {
3001 	char	*path;
3002 	int	uid;
3003 	int	gid;
3004 };
3005 #endif
3006 int
3007 sys_chown(struct thread *td, struct chown_args *uap)
3008 {
3009 
3010 	return (kern_fchownat(td, AT_FDCWD, uap->path, UIO_USERSPACE, uap->uid,
3011 	    uap->gid, 0));
3012 }
3013 
3014 #ifndef _SYS_SYSPROTO_H_
3015 struct fchownat_args {
3016 	int fd;
3017 	const char * path;
3018 	uid_t uid;
3019 	gid_t gid;
3020 	int flag;
3021 };
3022 #endif
3023 int
3024 sys_fchownat(struct thread *td, struct fchownat_args *uap)
3025 {
3026 
3027 	return (kern_fchownat(td, uap->fd, uap->path, UIO_USERSPACE, uap->uid,
3028 	    uap->gid, uap->flag));
3029 }
3030 
3031 int
3032 kern_fchownat(struct thread *td, int fd, const char *path,
3033     enum uio_seg pathseg, int uid, int gid, int flag)
3034 {
3035 	struct nameidata nd;
3036 	int error;
3037 
3038 	if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH |
3039 	    AT_EMPTY_PATH)) != 0)
3040 		return (EINVAL);
3041 
3042 	AUDIT_ARG_OWNER(uid, gid);
3043 	NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(flag, AT_SYMLINK_NOFOLLOW |
3044 	    AT_RESOLVE_BENEATH | AT_EMPTY_PATH) | AUDITVNODE1, pathseg, path,
3045 	    fd, &cap_fchown_rights);
3046 
3047 	if ((error = namei(&nd)) != 0)
3048 		return (error);
3049 	NDFREE_PNBUF(&nd);
3050 	error = setfown(td, td->td_ucred, nd.ni_vp, uid, gid);
3051 	vrele(nd.ni_vp);
3052 	return (error);
3053 }
3054 
3055 /*
3056  * Set ownership given a path name, do not cross symlinks.
3057  */
3058 #ifndef _SYS_SYSPROTO_H_
3059 struct lchown_args {
3060 	char	*path;
3061 	int	uid;
3062 	int	gid;
3063 };
3064 #endif
3065 int
3066 sys_lchown(struct thread *td, struct lchown_args *uap)
3067 {
3068 
3069 	return (kern_fchownat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
3070 	    uap->uid, uap->gid, AT_SYMLINK_NOFOLLOW));
3071 }
3072 
3073 /*
3074  * Set ownership given a file descriptor.
3075  */
3076 #ifndef _SYS_SYSPROTO_H_
3077 struct fchown_args {
3078 	int	fd;
3079 	int	uid;
3080 	int	gid;
3081 };
3082 #endif
3083 int
3084 sys_fchown(struct thread *td, struct fchown_args *uap)
3085 {
3086 	struct file *fp;
3087 	int error;
3088 
3089 	AUDIT_ARG_FD(uap->fd);
3090 	AUDIT_ARG_OWNER(uap->uid, uap->gid);
3091 	error = fget(td, uap->fd, &cap_fchown_rights, &fp);
3092 	if (error != 0)
3093 		return (error);
3094 	error = fo_chown(fp, uap->uid, uap->gid, td->td_ucred, td);
3095 	fdrop(fp, td);
3096 	return (error);
3097 }
3098 
3099 /*
3100  * Common implementation code for utimes(), lutimes(), and futimes().
3101  */
3102 static int
3103 getutimes(const struct timeval *usrtvp, enum uio_seg tvpseg,
3104     struct timespec *tsp)
3105 {
3106 	struct timeval tv[2];
3107 	const struct timeval *tvp;
3108 	int error;
3109 
3110 	if (usrtvp == NULL) {
3111 		vfs_timestamp(&tsp[0]);
3112 		tsp[1] = tsp[0];
3113 	} else {
3114 		if (tvpseg == UIO_SYSSPACE) {
3115 			tvp = usrtvp;
3116 		} else {
3117 			if ((error = copyin(usrtvp, tv, sizeof(tv))) != 0)
3118 				return (error);
3119 			tvp = tv;
3120 		}
3121 
3122 		if (tvp[0].tv_usec < 0 || tvp[0].tv_usec >= 1000000 ||
3123 		    tvp[1].tv_usec < 0 || tvp[1].tv_usec >= 1000000)
3124 			return (EINVAL);
3125 		TIMEVAL_TO_TIMESPEC(&tvp[0], &tsp[0]);
3126 		TIMEVAL_TO_TIMESPEC(&tvp[1], &tsp[1]);
3127 	}
3128 	return (0);
3129 }
3130 
3131 /*
3132  * Common implementation code for futimens(), utimensat().
3133  */
3134 #define	UTIMENS_NULL	0x1
3135 #define	UTIMENS_EXIT	0x2
3136 static int
3137 getutimens(const struct timespec *usrtsp, enum uio_seg tspseg,
3138     struct timespec *tsp, int *retflags)
3139 {
3140 	struct timespec tsnow;
3141 	int error;
3142 
3143 	vfs_timestamp(&tsnow);
3144 	*retflags = 0;
3145 	if (usrtsp == NULL) {
3146 		tsp[0] = tsnow;
3147 		tsp[1] = tsnow;
3148 		*retflags |= UTIMENS_NULL;
3149 		return (0);
3150 	}
3151 	if (tspseg == UIO_SYSSPACE) {
3152 		tsp[0] = usrtsp[0];
3153 		tsp[1] = usrtsp[1];
3154 	} else if ((error = copyin(usrtsp, tsp, sizeof(*tsp) * 2)) != 0)
3155 		return (error);
3156 	if (tsp[0].tv_nsec == UTIME_OMIT && tsp[1].tv_nsec == UTIME_OMIT)
3157 		*retflags |= UTIMENS_EXIT;
3158 	if (tsp[0].tv_nsec == UTIME_NOW && tsp[1].tv_nsec == UTIME_NOW)
3159 		*retflags |= UTIMENS_NULL;
3160 	if (tsp[0].tv_nsec == UTIME_OMIT)
3161 		tsp[0].tv_sec = VNOVAL;
3162 	else if (tsp[0].tv_nsec == UTIME_NOW)
3163 		tsp[0] = tsnow;
3164 	else if (tsp[0].tv_nsec < 0 || tsp[0].tv_nsec >= 1000000000L)
3165 		return (EINVAL);
3166 	if (tsp[1].tv_nsec == UTIME_OMIT)
3167 		tsp[1].tv_sec = VNOVAL;
3168 	else if (tsp[1].tv_nsec == UTIME_NOW)
3169 		tsp[1] = tsnow;
3170 	else if (tsp[1].tv_nsec < 0 || tsp[1].tv_nsec >= 1000000000L)
3171 		return (EINVAL);
3172 
3173 	return (0);
3174 }
3175 
3176 /*
3177  * Common implementation code for utimes(), lutimes(), futimes(), futimens(),
3178  * and utimensat().
3179  */
3180 static int
3181 setutimes(struct thread *td, struct vnode *vp, const struct timespec *ts,
3182     int numtimes, int nullflag)
3183 {
3184 	struct mount *mp;
3185 	struct vattr vattr;
3186 	int error;
3187 	bool setbirthtime;
3188 
3189 	setbirthtime = false;
3190 	vattr.va_birthtime.tv_sec = VNOVAL;
3191 	vattr.va_birthtime.tv_nsec = 0;
3192 
3193 	if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0)
3194 		return (error);
3195 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3196 	if (numtimes < 3 && VOP_GETATTR(vp, &vattr, td->td_ucred) == 0 &&
3197 	    timespeccmp(&ts[1], &vattr.va_birthtime, < ))
3198 		setbirthtime = true;
3199 	VATTR_NULL(&vattr);
3200 	vattr.va_atime = ts[0];
3201 	vattr.va_mtime = ts[1];
3202 	if (setbirthtime)
3203 		vattr.va_birthtime = ts[1];
3204 	if (numtimes > 2)
3205 		vattr.va_birthtime = ts[2];
3206 	if (nullflag)
3207 		vattr.va_vaflags |= VA_UTIMES_NULL;
3208 #ifdef MAC
3209 	error = mac_vnode_check_setutimes(td->td_ucred, vp, vattr.va_atime,
3210 	    vattr.va_mtime);
3211 #endif
3212 	if (error == 0)
3213 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3214 	VOP_UNLOCK(vp);
3215 	vn_finished_write(mp);
3216 	return (error);
3217 }
3218 
3219 /*
3220  * Set the access and modification times of a file.
3221  */
3222 #ifndef _SYS_SYSPROTO_H_
3223 struct utimes_args {
3224 	char	*path;
3225 	struct	timeval *tptr;
3226 };
3227 #endif
3228 int
3229 sys_utimes(struct thread *td, struct utimes_args *uap)
3230 {
3231 
3232 	return (kern_utimesat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
3233 	    uap->tptr, UIO_USERSPACE));
3234 }
3235 
3236 #ifndef _SYS_SYSPROTO_H_
3237 struct futimesat_args {
3238 	int fd;
3239 	const char * path;
3240 	const struct timeval * times;
3241 };
3242 #endif
3243 int
3244 sys_futimesat(struct thread *td, struct futimesat_args *uap)
3245 {
3246 
3247 	return (kern_utimesat(td, uap->fd, uap->path, UIO_USERSPACE,
3248 	    uap->times, UIO_USERSPACE));
3249 }
3250 
3251 int
3252 kern_utimesat(struct thread *td, int fd, const char *path,
3253     enum uio_seg pathseg, const struct timeval *tptr, enum uio_seg tptrseg)
3254 {
3255 	struct nameidata nd;
3256 	struct timespec ts[2];
3257 	int error;
3258 
3259 	if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3260 		return (error);
3261 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, fd,
3262 	    &cap_futimes_rights);
3263 
3264 	if ((error = namei(&nd)) != 0)
3265 		return (error);
3266 	NDFREE_PNBUF(&nd);
3267 	error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3268 	vrele(nd.ni_vp);
3269 	return (error);
3270 }
3271 
3272 /*
3273  * Set the access and modification times of a file.
3274  */
3275 #ifndef _SYS_SYSPROTO_H_
3276 struct lutimes_args {
3277 	char	*path;
3278 	struct	timeval *tptr;
3279 };
3280 #endif
3281 int
3282 sys_lutimes(struct thread *td, struct lutimes_args *uap)
3283 {
3284 
3285 	return (kern_lutimes(td, uap->path, UIO_USERSPACE, uap->tptr,
3286 	    UIO_USERSPACE));
3287 }
3288 
3289 int
3290 kern_lutimes(struct thread *td, const char *path, enum uio_seg pathseg,
3291     const struct timeval *tptr, enum uio_seg tptrseg)
3292 {
3293 	struct timespec ts[2];
3294 	struct nameidata nd;
3295 	int error;
3296 
3297 	if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3298 		return (error);
3299 	NDINIT(&nd, LOOKUP, NOFOLLOW | AUDITVNODE1, pathseg, path);
3300 	if ((error = namei(&nd)) != 0)
3301 		return (error);
3302 	NDFREE_PNBUF(&nd);
3303 	error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3304 	vrele(nd.ni_vp);
3305 	return (error);
3306 }
3307 
3308 /*
3309  * Set the access and modification times of a file.
3310  */
3311 #ifndef _SYS_SYSPROTO_H_
3312 struct futimes_args {
3313 	int	fd;
3314 	struct	timeval *tptr;
3315 };
3316 #endif
3317 int
3318 sys_futimes(struct thread *td, struct futimes_args *uap)
3319 {
3320 
3321 	return (kern_futimes(td, uap->fd, uap->tptr, UIO_USERSPACE));
3322 }
3323 
3324 int
3325 kern_futimes(struct thread *td, int fd, const struct timeval *tptr,
3326     enum uio_seg tptrseg)
3327 {
3328 	struct timespec ts[2];
3329 	struct file *fp;
3330 	int error;
3331 
3332 	AUDIT_ARG_FD(fd);
3333 	error = getutimes(tptr, tptrseg, ts);
3334 	if (error != 0)
3335 		return (error);
3336 	error = getvnode(td, fd, &cap_futimes_rights, &fp);
3337 	if (error != 0)
3338 		return (error);
3339 #ifdef AUDIT
3340 	if (AUDITING_TD(td)) {
3341 		vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
3342 		AUDIT_ARG_VNODE1(fp->f_vnode);
3343 		VOP_UNLOCK(fp->f_vnode);
3344 	}
3345 #endif
3346 	error = setutimes(td, fp->f_vnode, ts, 2, tptr == NULL);
3347 	fdrop(fp, td);
3348 	return (error);
3349 }
3350 
3351 int
3352 sys_futimens(struct thread *td, struct futimens_args *uap)
3353 {
3354 
3355 	return (kern_futimens(td, uap->fd, uap->times, UIO_USERSPACE));
3356 }
3357 
3358 int
3359 kern_futimens(struct thread *td, int fd, const struct timespec *tptr,
3360     enum uio_seg tptrseg)
3361 {
3362 	struct timespec ts[2];
3363 	struct file *fp;
3364 	int error, flags;
3365 
3366 	AUDIT_ARG_FD(fd);
3367 	error = getutimens(tptr, tptrseg, ts, &flags);
3368 	if (error != 0)
3369 		return (error);
3370 	if (flags & UTIMENS_EXIT)
3371 		return (0);
3372 	error = getvnode(td, fd, &cap_futimes_rights, &fp);
3373 	if (error != 0)
3374 		return (error);
3375 #ifdef AUDIT
3376 	if (AUDITING_TD(td)) {
3377 		vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
3378 		AUDIT_ARG_VNODE1(fp->f_vnode);
3379 		VOP_UNLOCK(fp->f_vnode);
3380 	}
3381 #endif
3382 	error = setutimes(td, fp->f_vnode, ts, 2, flags & UTIMENS_NULL);
3383 	fdrop(fp, td);
3384 	return (error);
3385 }
3386 
3387 int
3388 sys_utimensat(struct thread *td, struct utimensat_args *uap)
3389 {
3390 
3391 	return (kern_utimensat(td, uap->fd, uap->path, UIO_USERSPACE,
3392 	    uap->times, UIO_USERSPACE, uap->flag));
3393 }
3394 
3395 int
3396 kern_utimensat(struct thread *td, int fd, const char *path,
3397     enum uio_seg pathseg, const struct timespec *tptr, enum uio_seg tptrseg,
3398     int flag)
3399 {
3400 	struct nameidata nd;
3401 	struct timespec ts[2];
3402 	int error, flags;
3403 
3404 	if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH |
3405 	    AT_EMPTY_PATH)) != 0)
3406 		return (EINVAL);
3407 
3408 	if ((error = getutimens(tptr, tptrseg, ts, &flags)) != 0)
3409 		return (error);
3410 	NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(flag, AT_SYMLINK_NOFOLLOW |
3411 	    AT_RESOLVE_BENEATH | AT_EMPTY_PATH) | AUDITVNODE1,
3412 	    pathseg, path, fd, &cap_futimes_rights);
3413 	if ((error = namei(&nd)) != 0)
3414 		return (error);
3415 	/*
3416 	 * We are allowed to call namei() regardless of 2xUTIME_OMIT.
3417 	 * POSIX states:
3418 	 * "If both tv_nsec fields are UTIME_OMIT... EACCESS may be detected."
3419 	 * "Search permission is denied by a component of the path prefix."
3420 	 */
3421 	NDFREE_PNBUF(&nd);
3422 	if ((flags & UTIMENS_EXIT) == 0)
3423 		error = setutimes(td, nd.ni_vp, ts, 2, flags & UTIMENS_NULL);
3424 	vrele(nd.ni_vp);
3425 	return (error);
3426 }
3427 
3428 /*
3429  * Truncate a file given its path name.
3430  */
3431 #ifndef _SYS_SYSPROTO_H_
3432 struct truncate_args {
3433 	char	*path;
3434 	int	pad;
3435 	off_t	length;
3436 };
3437 #endif
3438 int
3439 sys_truncate(struct thread *td, struct truncate_args *uap)
3440 {
3441 
3442 	return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3443 }
3444 
3445 int
3446 kern_truncate(struct thread *td, const char *path, enum uio_seg pathseg,
3447     off_t length)
3448 {
3449 	struct mount *mp;
3450 	struct vnode *vp;
3451 	void *rl_cookie;
3452 	struct nameidata nd;
3453 	int error;
3454 
3455 	if (length < 0)
3456 		return (EINVAL);
3457 	NDPREINIT(&nd);
3458 retry:
3459 	NDINIT(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path);
3460 	if ((error = namei(&nd)) != 0)
3461 		return (error);
3462 	vp = nd.ni_vp;
3463 	NDFREE_PNBUF(&nd);
3464 	rl_cookie = vn_rangelock_wlock(vp, 0, OFF_MAX);
3465 	if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) {
3466 		vn_rangelock_unlock(vp, rl_cookie);
3467 		vrele(vp);
3468 		return (error);
3469 	}
3470 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3471 	if (vp->v_type == VDIR) {
3472 		error = EISDIR;
3473 		goto out;
3474 	}
3475 #ifdef MAC
3476 	error = mac_vnode_check_write(td->td_ucred, NOCRED, vp);
3477 	if (error != 0)
3478 		goto out;
3479 #endif
3480 	error = VOP_ACCESS(vp, VWRITE, td->td_ucred, td);
3481 	if (error != 0)
3482 		goto out;
3483 
3484 	error = vn_truncate_locked(vp, length, false, td->td_ucred);
3485 out:
3486 	VOP_UNLOCK(vp);
3487 	vn_finished_write(mp);
3488 	vn_rangelock_unlock(vp, rl_cookie);
3489 	vrele(vp);
3490 	if (error == ERELOOKUP)
3491 		goto retry;
3492 	return (error);
3493 }
3494 
3495 #if defined(COMPAT_43)
3496 /*
3497  * Truncate a file given its path name.
3498  */
3499 #ifndef _SYS_SYSPROTO_H_
3500 struct otruncate_args {
3501 	char	*path;
3502 	long	length;
3503 };
3504 #endif
3505 int
3506 otruncate(struct thread *td, struct otruncate_args *uap)
3507 {
3508 
3509 	return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3510 }
3511 #endif /* COMPAT_43 */
3512 
3513 #if defined(COMPAT_FREEBSD6)
3514 /* Versions with the pad argument */
3515 int
3516 freebsd6_truncate(struct thread *td, struct freebsd6_truncate_args *uap)
3517 {
3518 
3519 	return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3520 }
3521 
3522 int
3523 freebsd6_ftruncate(struct thread *td, struct freebsd6_ftruncate_args *uap)
3524 {
3525 
3526 	return (kern_ftruncate(td, uap->fd, uap->length));
3527 }
3528 #endif
3529 
3530 int
3531 kern_fsync(struct thread *td, int fd, bool fullsync)
3532 {
3533 	struct vnode *vp;
3534 	struct mount *mp;
3535 	struct file *fp;
3536 	int error;
3537 
3538 	AUDIT_ARG_FD(fd);
3539 	error = getvnode(td, fd, &cap_fsync_rights, &fp);
3540 	if (error != 0)
3541 		return (error);
3542 	vp = fp->f_vnode;
3543 #if 0
3544 	if (!fullsync)
3545 		/* XXXKIB: compete outstanding aio writes */;
3546 #endif
3547 retry:
3548 	error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH);
3549 	if (error != 0)
3550 		goto drop;
3551 	vn_lock(vp, vn_lktype_write(mp, vp) | LK_RETRY);
3552 	AUDIT_ARG_VNODE1(vp);
3553 	if (vp->v_object != NULL) {
3554 		VM_OBJECT_WLOCK(vp->v_object);
3555 		vm_object_page_clean(vp->v_object, 0, 0, 0);
3556 		VM_OBJECT_WUNLOCK(vp->v_object);
3557 	}
3558 	error = fullsync ? VOP_FSYNC(vp, MNT_WAIT, td) : VOP_FDATASYNC(vp, td);
3559 	VOP_UNLOCK(vp);
3560 	vn_finished_write(mp);
3561 	if (error == ERELOOKUP)
3562 		goto retry;
3563 drop:
3564 	fdrop(fp, td);
3565 	return (error);
3566 }
3567 
3568 /*
3569  * Sync an open file.
3570  */
3571 #ifndef _SYS_SYSPROTO_H_
3572 struct fsync_args {
3573 	int	fd;
3574 };
3575 #endif
3576 int
3577 sys_fsync(struct thread *td, struct fsync_args *uap)
3578 {
3579 
3580 	return (kern_fsync(td, uap->fd, true));
3581 }
3582 
3583 int
3584 sys_fdatasync(struct thread *td, struct fdatasync_args *uap)
3585 {
3586 
3587 	return (kern_fsync(td, uap->fd, false));
3588 }
3589 
3590 /*
3591  * Rename files.  Source and destination must either both be directories, or
3592  * both not be directories.  If target is a directory, it must be empty.
3593  */
3594 #ifndef _SYS_SYSPROTO_H_
3595 struct rename_args {
3596 	char	*from;
3597 	char	*to;
3598 };
3599 #endif
3600 int
3601 sys_rename(struct thread *td, struct rename_args *uap)
3602 {
3603 
3604 	return (kern_renameat(td, AT_FDCWD, uap->from, AT_FDCWD,
3605 	    uap->to, UIO_USERSPACE));
3606 }
3607 
3608 #ifndef _SYS_SYSPROTO_H_
3609 struct renameat_args {
3610 	int	oldfd;
3611 	char	*old;
3612 	int	newfd;
3613 	char	*new;
3614 };
3615 #endif
3616 int
3617 sys_renameat(struct thread *td, struct renameat_args *uap)
3618 {
3619 
3620 	return (kern_renameat(td, uap->oldfd, uap->old, uap->newfd, uap->new,
3621 	    UIO_USERSPACE));
3622 }
3623 
3624 #ifdef MAC
3625 static int
3626 kern_renameat_mac(struct thread *td, int oldfd, const char *old, int newfd,
3627     const char *new, enum uio_seg pathseg, struct nameidata *fromnd)
3628 {
3629 	int error;
3630 
3631 	NDINIT_ATRIGHTS(fromnd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1,
3632 	    pathseg, old, oldfd, &cap_renameat_source_rights);
3633 	if ((error = namei(fromnd)) != 0)
3634 		return (error);
3635 	error = mac_vnode_check_rename_from(td->td_ucred, fromnd->ni_dvp,
3636 	    fromnd->ni_vp, &fromnd->ni_cnd);
3637 	VOP_UNLOCK(fromnd->ni_dvp);
3638 	if (fromnd->ni_dvp != fromnd->ni_vp)
3639 		VOP_UNLOCK(fromnd->ni_vp);
3640 	if (error != 0) {
3641 		NDFREE_PNBUF(fromnd);
3642 		vrele(fromnd->ni_dvp);
3643 		vrele(fromnd->ni_vp);
3644 	}
3645 	return (error);
3646 }
3647 #endif
3648 
3649 int
3650 kern_renameat(struct thread *td, int oldfd, const char *old, int newfd,
3651     const char *new, enum uio_seg pathseg)
3652 {
3653 	struct mount *mp = NULL;
3654 	struct vnode *tvp, *fvp, *tdvp;
3655 	struct nameidata fromnd, tond;
3656 	uint64_t tondflags;
3657 	int error;
3658 
3659 again:
3660 	bwillwrite();
3661 #ifdef MAC
3662 	if (mac_vnode_check_rename_from_enabled()) {
3663 		error = kern_renameat_mac(td, oldfd, old, newfd, new, pathseg,
3664 		    &fromnd);
3665 		if (error != 0)
3666 			return (error);
3667 	} else {
3668 #endif
3669 	NDINIT_ATRIGHTS(&fromnd, DELETE, WANTPARENT | AUDITVNODE1,
3670 	    pathseg, old, oldfd, &cap_renameat_source_rights);
3671 	if ((error = namei(&fromnd)) != 0)
3672 		return (error);
3673 #ifdef MAC
3674 	}
3675 #endif
3676 	fvp = fromnd.ni_vp;
3677 	tondflags = LOCKPARENT | LOCKLEAF | NOCACHE | AUDITVNODE2;
3678 	if (fromnd.ni_vp->v_type == VDIR)
3679 		tondflags |= WILLBEDIR;
3680 	NDINIT_ATRIGHTS(&tond, RENAME, tondflags, pathseg, new, newfd,
3681 	    &cap_renameat_target_rights);
3682 	if ((error = namei(&tond)) != 0) {
3683 		/* Translate error code for rename("dir1", "dir2/."). */
3684 		if (error == EISDIR && fvp->v_type == VDIR)
3685 			error = EINVAL;
3686 		NDFREE_PNBUF(&fromnd);
3687 		vrele(fromnd.ni_dvp);
3688 		vrele(fvp);
3689 		goto out1;
3690 	}
3691 	tdvp = tond.ni_dvp;
3692 	tvp = tond.ni_vp;
3693 	error = vn_start_write(fvp, &mp, V_NOWAIT);
3694 	if (error != 0) {
3695 		NDFREE_PNBUF(&fromnd);
3696 		NDFREE_PNBUF(&tond);
3697 		if (tvp != NULL)
3698 			vput(tvp);
3699 		if (tdvp == tvp)
3700 			vrele(tdvp);
3701 		else
3702 			vput(tdvp);
3703 		vrele(fromnd.ni_dvp);
3704 		vrele(fvp);
3705 		error = vn_start_write(NULL, &mp, V_XSLEEP | V_PCATCH);
3706 		if (error != 0)
3707 			return (error);
3708 		goto again;
3709 	}
3710 	if (tvp != NULL) {
3711 		if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
3712 			error = ENOTDIR;
3713 			goto out;
3714 		} else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
3715 			error = EISDIR;
3716 			goto out;
3717 		}
3718 #ifdef CAPABILITIES
3719 		if (newfd != AT_FDCWD && (tond.ni_resflags & NIRES_ABS) == 0) {
3720 			/*
3721 			 * If the target already exists we require CAP_UNLINKAT
3722 			 * from 'newfd', when newfd was used for the lookup.
3723 			 */
3724 			error = cap_check(&tond.ni_filecaps.fc_rights,
3725 			    &cap_unlinkat_rights);
3726 			if (error != 0)
3727 				goto out;
3728 		}
3729 #endif
3730 	}
3731 	if (fvp == tdvp) {
3732 		error = EINVAL;
3733 		goto out;
3734 	}
3735 	/*
3736 	 * If the source is the same as the destination (that is, if they
3737 	 * are links to the same vnode), then there is nothing to do.
3738 	 */
3739 	if (fvp == tvp)
3740 		error = ERESTART;
3741 #ifdef MAC
3742 	else
3743 		error = mac_vnode_check_rename_to(td->td_ucred, tdvp,
3744 		    tond.ni_vp, fromnd.ni_dvp == tdvp, &tond.ni_cnd);
3745 #endif
3746 out:
3747 	if (error == 0) {
3748 		error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
3749 		    tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
3750 		NDFREE_PNBUF(&fromnd);
3751 		NDFREE_PNBUF(&tond);
3752 	} else {
3753 		NDFREE_PNBUF(&fromnd);
3754 		NDFREE_PNBUF(&tond);
3755 		if (tvp != NULL)
3756 			vput(tvp);
3757 		if (tdvp == tvp)
3758 			vrele(tdvp);
3759 		else
3760 			vput(tdvp);
3761 		vrele(fromnd.ni_dvp);
3762 		vrele(fvp);
3763 	}
3764 	vn_finished_write(mp);
3765 out1:
3766 	if (error == ERESTART)
3767 		return (0);
3768 	if (error == ERELOOKUP)
3769 		goto again;
3770 	return (error);
3771 }
3772 
3773 /*
3774  * Make a directory file.
3775  */
3776 #ifndef _SYS_SYSPROTO_H_
3777 struct mkdir_args {
3778 	char	*path;
3779 	int	mode;
3780 };
3781 #endif
3782 int
3783 sys_mkdir(struct thread *td, struct mkdir_args *uap)
3784 {
3785 
3786 	return (kern_mkdirat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
3787 	    uap->mode));
3788 }
3789 
3790 #ifndef _SYS_SYSPROTO_H_
3791 struct mkdirat_args {
3792 	int	fd;
3793 	char	*path;
3794 	mode_t	mode;
3795 };
3796 #endif
3797 int
3798 sys_mkdirat(struct thread *td, struct mkdirat_args *uap)
3799 {
3800 
3801 	return (kern_mkdirat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode));
3802 }
3803 
3804 int
3805 kern_mkdirat(struct thread *td, int fd, const char *path, enum uio_seg segflg,
3806     int mode)
3807 {
3808 	struct mount *mp;
3809 	struct vattr vattr;
3810 	struct nameidata nd;
3811 	int error;
3812 
3813 	AUDIT_ARG_MODE(mode);
3814 	NDPREINIT(&nd);
3815 restart:
3816 	bwillwrite();
3817 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | AUDITVNODE1 |
3818 	    NC_NOMAKEENTRY | NC_KEEPPOSENTRY | FAILIFEXISTS | WILLBEDIR,
3819 	    segflg, path, fd, &cap_mkdirat_rights);
3820 	if ((error = namei(&nd)) != 0)
3821 		return (error);
3822 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3823 		NDFREE_PNBUF(&nd);
3824 		vput(nd.ni_dvp);
3825 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | V_PCATCH)) != 0)
3826 			return (error);
3827 		goto restart;
3828 	}
3829 	VATTR_NULL(&vattr);
3830 	vattr.va_type = VDIR;
3831 	vattr.va_mode = (mode & ACCESSPERMS) &~ td->td_proc->p_pd->pd_cmask;
3832 #ifdef MAC
3833 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
3834 	    &vattr);
3835 	if (error != 0)
3836 		goto out;
3837 #endif
3838 	error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
3839 #ifdef MAC
3840 out:
3841 #endif
3842 	NDFREE_PNBUF(&nd);
3843 	VOP_VPUT_PAIR(nd.ni_dvp, error == 0 ? &nd.ni_vp : NULL, true);
3844 	vn_finished_write(mp);
3845 	if (error == ERELOOKUP)
3846 		goto restart;
3847 	return (error);
3848 }
3849 
3850 /*
3851  * Remove a directory file.
3852  */
3853 #ifndef _SYS_SYSPROTO_H_
3854 struct rmdir_args {
3855 	char	*path;
3856 };
3857 #endif
3858 int
3859 sys_rmdir(struct thread *td, struct rmdir_args *uap)
3860 {
3861 
3862 	return (kern_frmdirat(td, AT_FDCWD, uap->path, FD_NONE, UIO_USERSPACE,
3863 	    0));
3864 }
3865 
3866 int
3867 kern_frmdirat(struct thread *td, int dfd, const char *path, int fd,
3868     enum uio_seg pathseg, int flag)
3869 {
3870 	struct mount *mp;
3871 	struct vnode *vp;
3872 	struct file *fp;
3873 	struct nameidata nd;
3874 	cap_rights_t rights;
3875 	int error;
3876 
3877 	fp = NULL;
3878 	if (fd != FD_NONE) {
3879 		error = getvnode(td, fd, cap_rights_init_one(&rights,
3880 		    CAP_LOOKUP), &fp);
3881 		if (error != 0)
3882 			return (error);
3883 	}
3884 
3885 	NDPREINIT(&nd);
3886 restart:
3887 	bwillwrite();
3888 	NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1 |
3889 	    at2cnpflags(flag, AT_RESOLVE_BENEATH),
3890 	    pathseg, path, dfd, &cap_unlinkat_rights);
3891 	if ((error = namei(&nd)) != 0)
3892 		goto fdout;
3893 	vp = nd.ni_vp;
3894 	if (vp->v_type != VDIR) {
3895 		error = ENOTDIR;
3896 		goto out;
3897 	}
3898 	/*
3899 	 * No rmdir "." please.
3900 	 */
3901 	if (nd.ni_dvp == vp) {
3902 		error = EINVAL;
3903 		goto out;
3904 	}
3905 	/*
3906 	 * The root of a mounted filesystem cannot be deleted.
3907 	 */
3908 	if (vp->v_vflag & VV_ROOT) {
3909 		error = EBUSY;
3910 		goto out;
3911 	}
3912 
3913 	if (fp != NULL && fp->f_vnode != vp) {
3914 		if (VN_IS_DOOMED(fp->f_vnode))
3915 			error = EBADF;
3916 		else
3917 			error = EDEADLK;
3918 		goto out;
3919 	}
3920 
3921 #ifdef MAC
3922 	error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
3923 	    &nd.ni_cnd);
3924 	if (error != 0)
3925 		goto out;
3926 #endif
3927 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3928 		NDFREE_PNBUF(&nd);
3929 		vput(vp);
3930 		if (nd.ni_dvp == vp)
3931 			vrele(nd.ni_dvp);
3932 		else
3933 			vput(nd.ni_dvp);
3934 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | V_PCATCH)) != 0)
3935 			goto fdout;
3936 		goto restart;
3937 	}
3938 	vfs_notify_upper(vp, VFS_NOTIFY_UPPER_UNLINK);
3939 	error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
3940 	vn_finished_write(mp);
3941 out:
3942 	NDFREE_PNBUF(&nd);
3943 	vput(vp);
3944 	if (nd.ni_dvp == vp)
3945 		vrele(nd.ni_dvp);
3946 	else
3947 		vput(nd.ni_dvp);
3948 	if (error == ERELOOKUP)
3949 		goto restart;
3950 fdout:
3951 	if (fp != NULL)
3952 		fdrop(fp, td);
3953 	return (error);
3954 }
3955 
3956 #if defined(COMPAT_43) || defined(COMPAT_FREEBSD11)
3957 int
3958 freebsd11_kern_getdirentries(struct thread *td, int fd, char *ubuf, u_int count,
3959     long *basep, void (*func)(struct freebsd11_dirent *))
3960 {
3961 	struct freebsd11_dirent dstdp;
3962 	struct dirent *dp, *edp;
3963 	char *dirbuf;
3964 	off_t base;
3965 	ssize_t resid, ucount;
3966 	int error;
3967 
3968 	/* XXX arbitrary sanity limit on `count'. */
3969 	count = min(count, 64 * 1024);
3970 
3971 	dirbuf = malloc(count, M_TEMP, M_WAITOK);
3972 
3973 	error = kern_getdirentries(td, fd, dirbuf, count, &base, &resid,
3974 	    UIO_SYSSPACE);
3975 	if (error != 0)
3976 		goto done;
3977 	if (basep != NULL)
3978 		*basep = base;
3979 
3980 	ucount = 0;
3981 	for (dp = (struct dirent *)dirbuf,
3982 	    edp = (struct dirent *)&dirbuf[count - resid];
3983 	    ucount < count && dp < edp; ) {
3984 		if (dp->d_reclen == 0)
3985 			break;
3986 		MPASS(dp->d_reclen >= _GENERIC_DIRLEN(0));
3987 		if (dp->d_namlen >= sizeof(dstdp.d_name))
3988 			continue;
3989 		dstdp.d_type = dp->d_type;
3990 		dstdp.d_namlen = dp->d_namlen;
3991 		dstdp.d_fileno = dp->d_fileno;		/* truncate */
3992 		if (dstdp.d_fileno != dp->d_fileno) {
3993 			switch (ino64_trunc_error) {
3994 			default:
3995 			case 0:
3996 				break;
3997 			case 1:
3998 				error = EOVERFLOW;
3999 				goto done;
4000 			case 2:
4001 				dstdp.d_fileno = UINT32_MAX;
4002 				break;
4003 			}
4004 		}
4005 		dstdp.d_reclen = sizeof(dstdp) - sizeof(dstdp.d_name) +
4006 		    ((dp->d_namlen + 1 + 3) &~ 3);
4007 		bcopy(dp->d_name, dstdp.d_name, dstdp.d_namlen);
4008 		bzero(dstdp.d_name + dstdp.d_namlen,
4009 		    dstdp.d_reclen - offsetof(struct freebsd11_dirent, d_name) -
4010 		    dstdp.d_namlen);
4011 		MPASS(dstdp.d_reclen <= dp->d_reclen);
4012 		MPASS(ucount + dstdp.d_reclen <= count);
4013 		if (func != NULL)
4014 			func(&dstdp);
4015 		error = copyout(&dstdp, ubuf + ucount, dstdp.d_reclen);
4016 		if (error != 0)
4017 			break;
4018 		dp = (struct dirent *)((char *)dp + dp->d_reclen);
4019 		ucount += dstdp.d_reclen;
4020 	}
4021 
4022 done:
4023 	free(dirbuf, M_TEMP);
4024 	if (error == 0)
4025 		td->td_retval[0] = ucount;
4026 	return (error);
4027 }
4028 #endif /* COMPAT */
4029 
4030 #ifdef COMPAT_43
4031 static void
4032 ogetdirentries_cvt(struct freebsd11_dirent *dp)
4033 {
4034 #if (BYTE_ORDER == LITTLE_ENDIAN)
4035 	/*
4036 	 * The expected low byte of dp->d_namlen is our dp->d_type.
4037 	 * The high MBZ byte of dp->d_namlen is our dp->d_namlen.
4038 	 */
4039 	dp->d_type = dp->d_namlen;
4040 	dp->d_namlen = 0;
4041 #else
4042 	/*
4043 	 * The dp->d_type is the high byte of the expected dp->d_namlen,
4044 	 * so must be zero'ed.
4045 	 */
4046 	dp->d_type = 0;
4047 #endif
4048 }
4049 
4050 /*
4051  * Read a block of directory entries in a filesystem independent format.
4052  */
4053 #ifndef _SYS_SYSPROTO_H_
4054 struct ogetdirentries_args {
4055 	int	fd;
4056 	char	*buf;
4057 	u_int	count;
4058 	long	*basep;
4059 };
4060 #endif
4061 int
4062 ogetdirentries(struct thread *td, struct ogetdirentries_args *uap)
4063 {
4064 	long loff;
4065 	int error;
4066 
4067 	error = kern_ogetdirentries(td, uap, &loff);
4068 	if (error == 0)
4069 		error = copyout(&loff, uap->basep, sizeof(long));
4070 	return (error);
4071 }
4072 
4073 int
4074 kern_ogetdirentries(struct thread *td, struct ogetdirentries_args *uap,
4075     long *ploff)
4076 {
4077 	long base;
4078 	int error;
4079 
4080 	/* XXX arbitrary sanity limit on `count'. */
4081 	if (uap->count > 64 * 1024)
4082 		return (EINVAL);
4083 
4084 	error = freebsd11_kern_getdirentries(td, uap->fd, uap->buf, uap->count,
4085 	    &base, ogetdirentries_cvt);
4086 
4087 	if (error == 0 && uap->basep != NULL)
4088 		error = copyout(&base, uap->basep, sizeof(long));
4089 
4090 	return (error);
4091 }
4092 #endif /* COMPAT_43 */
4093 
4094 #if defined(COMPAT_FREEBSD11)
4095 #ifndef _SYS_SYSPROTO_H_
4096 struct freebsd11_getdirentries_args {
4097 	int	fd;
4098 	char	*buf;
4099 	u_int	count;
4100 	long	*basep;
4101 };
4102 #endif
4103 int
4104 freebsd11_getdirentries(struct thread *td,
4105     struct freebsd11_getdirentries_args *uap)
4106 {
4107 	long base;
4108 	int error;
4109 
4110 	error = freebsd11_kern_getdirentries(td, uap->fd, uap->buf, uap->count,
4111 	    &base, NULL);
4112 
4113 	if (error == 0 && uap->basep != NULL)
4114 		error = copyout(&base, uap->basep, sizeof(long));
4115 	return (error);
4116 }
4117 
4118 int
4119 freebsd11_getdents(struct thread *td, struct freebsd11_getdents_args *uap)
4120 {
4121 	struct freebsd11_getdirentries_args ap;
4122 
4123 	ap.fd = uap->fd;
4124 	ap.buf = uap->buf;
4125 	ap.count = uap->count;
4126 	ap.basep = NULL;
4127 	return (freebsd11_getdirentries(td, &ap));
4128 }
4129 #endif /* COMPAT_FREEBSD11 */
4130 
4131 /*
4132  * Read a block of directory entries in a filesystem independent format.
4133  */
4134 int
4135 sys_getdirentries(struct thread *td, struct getdirentries_args *uap)
4136 {
4137 	off_t base;
4138 	int error;
4139 
4140 	error = kern_getdirentries(td, uap->fd, uap->buf, uap->count, &base,
4141 	    NULL, UIO_USERSPACE);
4142 	if (error != 0)
4143 		return (error);
4144 	if (uap->basep != NULL)
4145 		error = copyout(&base, uap->basep, sizeof(off_t));
4146 	return (error);
4147 }
4148 
4149 int
4150 kern_getdirentries(struct thread *td, int fd, char *buf, size_t count,
4151     off_t *basep, ssize_t *residp, enum uio_seg bufseg)
4152 {
4153 	struct vnode *vp;
4154 	struct file *fp;
4155 	struct uio auio;
4156 	struct iovec aiov;
4157 	off_t loff;
4158 	int error, eofflag;
4159 	off_t foffset;
4160 
4161 	AUDIT_ARG_FD(fd);
4162 	if (count > IOSIZE_MAX)
4163 		return (EINVAL);
4164 	auio.uio_resid = count;
4165 	error = getvnode(td, fd, &cap_read_rights, &fp);
4166 	if (error != 0)
4167 		return (error);
4168 	if ((fp->f_flag & FREAD) == 0) {
4169 		fdrop(fp, td);
4170 		return (EBADF);
4171 	}
4172 	vp = fp->f_vnode;
4173 	foffset = foffset_lock(fp, 0);
4174 unionread:
4175 	if (vp->v_type != VDIR) {
4176 		error = EINVAL;
4177 		goto fail;
4178 	}
4179 	if (__predict_false((vp->v_vflag & VV_UNLINKED) != 0)) {
4180 		error = ENOENT;
4181 		goto fail;
4182 	}
4183 	aiov.iov_base = buf;
4184 	aiov.iov_len = count;
4185 	auio.uio_iov = &aiov;
4186 	auio.uio_iovcnt = 1;
4187 	auio.uio_rw = UIO_READ;
4188 	auio.uio_segflg = bufseg;
4189 	auio.uio_td = td;
4190 	vn_lock(vp, LK_SHARED | LK_RETRY);
4191 	AUDIT_ARG_VNODE1(vp);
4192 	loff = auio.uio_offset = foffset;
4193 #ifdef MAC
4194 	error = mac_vnode_check_readdir(td->td_ucred, vp);
4195 	if (error == 0)
4196 #endif
4197 		error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, NULL,
4198 		    NULL);
4199 	foffset = auio.uio_offset;
4200 	if (error != 0) {
4201 		VOP_UNLOCK(vp);
4202 		goto fail;
4203 	}
4204 	if (count == auio.uio_resid &&
4205 	    (vp->v_vflag & VV_ROOT) &&
4206 	    (vp->v_mount->mnt_flag & MNT_UNION)) {
4207 		struct vnode *tvp = vp;
4208 
4209 		vp = vp->v_mount->mnt_vnodecovered;
4210 		VREF(vp);
4211 		fp->f_vnode = vp;
4212 		foffset = 0;
4213 		vput(tvp);
4214 		goto unionread;
4215 	}
4216 	VOP_UNLOCK(vp);
4217 	*basep = loff;
4218 	if (residp != NULL)
4219 		*residp = auio.uio_resid;
4220 	td->td_retval[0] = count - auio.uio_resid;
4221 fail:
4222 	foffset_unlock(fp, foffset, 0);
4223 	fdrop(fp, td);
4224 	return (error);
4225 }
4226 
4227 /*
4228  * Set the mode mask for creation of filesystem nodes.
4229  */
4230 #ifndef _SYS_SYSPROTO_H_
4231 struct umask_args {
4232 	int	newmask;
4233 };
4234 #endif
4235 int
4236 sys_umask(struct thread *td, struct umask_args *uap)
4237 {
4238 	struct pwddesc *pdp;
4239 
4240 	pdp = td->td_proc->p_pd;
4241 	PWDDESC_XLOCK(pdp);
4242 	td->td_retval[0] = pdp->pd_cmask;
4243 	pdp->pd_cmask = uap->newmask & ALLPERMS;
4244 	PWDDESC_XUNLOCK(pdp);
4245 	return (0);
4246 }
4247 
4248 /*
4249  * Void all references to file by ripping underlying filesystem away from
4250  * vnode.
4251  */
4252 #ifndef _SYS_SYSPROTO_H_
4253 struct revoke_args {
4254 	char	*path;
4255 };
4256 #endif
4257 int
4258 sys_revoke(struct thread *td, struct revoke_args *uap)
4259 {
4260 	struct vnode *vp;
4261 	struct vattr vattr;
4262 	struct nameidata nd;
4263 	int error;
4264 
4265 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
4266 	    uap->path);
4267 	if ((error = namei(&nd)) != 0)
4268 		return (error);
4269 	vp = nd.ni_vp;
4270 	NDFREE_PNBUF(&nd);
4271 	if (vp->v_type != VCHR || vp->v_rdev == NULL) {
4272 		error = EINVAL;
4273 		goto out;
4274 	}
4275 #ifdef MAC
4276 	error = mac_vnode_check_revoke(td->td_ucred, vp);
4277 	if (error != 0)
4278 		goto out;
4279 #endif
4280 	error = VOP_GETATTR(vp, &vattr, td->td_ucred);
4281 	if (error != 0)
4282 		goto out;
4283 	if (td->td_ucred->cr_uid != vattr.va_uid) {
4284 		error = priv_check(td, PRIV_VFS_ADMIN);
4285 		if (error != 0)
4286 			goto out;
4287 	}
4288 	if (devfs_usecount(vp) > 0)
4289 		VOP_REVOKE(vp, REVOKEALL);
4290 out:
4291 	vput(vp);
4292 	return (error);
4293 }
4294 
4295 /*
4296  * This variant of getvnode() allows O_PATH files.  Caller should
4297  * ensure that returned file and vnode are only used for compatible
4298  * semantics.
4299  */
4300 int
4301 getvnode_path(struct thread *td, int fd, cap_rights_t *rightsp,
4302     struct file **fpp)
4303 {
4304 	struct file *fp;
4305 	int error;
4306 
4307 	error = fget_unlocked(td, fd, rightsp, &fp);
4308 	if (error != 0)
4309 		return (error);
4310 
4311 	/*
4312 	 * The file could be not of the vnode type, or it may be not
4313 	 * yet fully initialized, in which case the f_vnode pointer
4314 	 * may be set, but f_ops is still badfileops.  E.g.,
4315 	 * devfs_open() transiently create such situation to
4316 	 * facilitate csw d_fdopen().
4317 	 *
4318 	 * Dupfdopen() handling in kern_openat() installs the
4319 	 * half-baked file into the process descriptor table, allowing
4320 	 * other thread to dereference it. Guard against the race by
4321 	 * checking f_ops.
4322 	 */
4323 	if (__predict_false(fp->f_vnode == NULL || fp->f_ops == &badfileops)) {
4324 		fdrop(fp, td);
4325 		*fpp = NULL;
4326 		return (EINVAL);
4327 	}
4328 
4329 	*fpp = fp;
4330 	return (0);
4331 }
4332 
4333 /*
4334  * Convert a user file descriptor to a kernel file entry and check
4335  * that, if it is a capability, the correct rights are present.
4336  * A reference on the file entry is held upon returning.
4337  */
4338 int
4339 getvnode(struct thread *td, int fd, cap_rights_t *rightsp, struct file **fpp)
4340 {
4341 	int error;
4342 
4343 	error = getvnode_path(td, fd, rightsp, fpp);
4344 	if (__predict_false(error != 0))
4345 		return (error);
4346 
4347 	/*
4348 	 * Filter out O_PATH file descriptors, most getvnode() callers
4349 	 * do not call fo_ methods.
4350 	 */
4351 	if (__predict_false((*fpp)->f_ops == &path_fileops)) {
4352 		fdrop(*fpp, td);
4353 		*fpp = NULL;
4354 		error = EBADF;
4355 	}
4356 
4357 	return (error);
4358 }
4359 
4360 /*
4361  * Get an (NFS) file handle.
4362  */
4363 #ifndef _SYS_SYSPROTO_H_
4364 struct lgetfh_args {
4365 	char *fname;
4366 	fhandle_t *fhp;
4367 };
4368 #endif
4369 int
4370 sys_lgetfh(struct thread *td, struct lgetfh_args *uap)
4371 {
4372 
4373 	return (kern_getfhat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->fname,
4374 	    UIO_USERSPACE, uap->fhp, UIO_USERSPACE));
4375 }
4376 
4377 #ifndef _SYS_SYSPROTO_H_
4378 struct getfh_args {
4379 	char *fname;
4380 	fhandle_t *fhp;
4381 };
4382 #endif
4383 int
4384 sys_getfh(struct thread *td, struct getfh_args *uap)
4385 {
4386 
4387 	return (kern_getfhat(td, 0, AT_FDCWD, uap->fname, UIO_USERSPACE,
4388 	    uap->fhp, UIO_USERSPACE));
4389 }
4390 
4391 /*
4392  * syscall for the rpc.lockd to use to translate an open descriptor into
4393  * a NFS file handle.
4394  *
4395  * warning: do not remove the priv_check() call or this becomes one giant
4396  * security hole.
4397  */
4398 #ifndef _SYS_SYSPROTO_H_
4399 struct getfhat_args {
4400 	int fd;
4401 	char *path;
4402 	fhandle_t *fhp;
4403 	int flags;
4404 };
4405 #endif
4406 int
4407 sys_getfhat(struct thread *td, struct getfhat_args *uap)
4408 {
4409 
4410 	return (kern_getfhat(td, uap->flags, uap->fd, uap->path, UIO_USERSPACE,
4411 	    uap->fhp, UIO_USERSPACE));
4412 }
4413 
4414 int
4415 kern_getfhat(struct thread *td, int flags, int fd, const char *path,
4416     enum uio_seg pathseg, fhandle_t *fhp, enum uio_seg fhseg)
4417 {
4418 	struct nameidata nd;
4419 	fhandle_t fh;
4420 	struct vnode *vp;
4421 	int error;
4422 
4423 	if ((flags & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH)) != 0)
4424 		return (EINVAL);
4425 	error = priv_check(td, PRIV_VFS_GETFH);
4426 	if (error != 0)
4427 		return (error);
4428 	NDINIT_AT(&nd, LOOKUP, at2cnpflags(flags, AT_SYMLINK_NOFOLLOW |
4429 	    AT_RESOLVE_BENEATH) | LOCKLEAF | AUDITVNODE1, pathseg, path,
4430 	    fd);
4431 	error = namei(&nd);
4432 	if (error != 0)
4433 		return (error);
4434 	NDFREE_PNBUF(&nd);
4435 	vp = nd.ni_vp;
4436 	bzero(&fh, sizeof(fh));
4437 	fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
4438 	error = VOP_VPTOFH(vp, &fh.fh_fid);
4439 	vput(vp);
4440 	if (error == 0) {
4441 		if (fhseg == UIO_USERSPACE)
4442 			error = copyout(&fh, fhp, sizeof (fh));
4443 		else
4444 			memcpy(fhp, &fh, sizeof(fh));
4445 	}
4446 	return (error);
4447 }
4448 
4449 #ifndef _SYS_SYSPROTO_H_
4450 struct fhlink_args {
4451 	fhandle_t *fhp;
4452 	const char *to;
4453 };
4454 #endif
4455 int
4456 sys_fhlink(struct thread *td, struct fhlink_args *uap)
4457 {
4458 
4459 	return (kern_fhlinkat(td, AT_FDCWD, uap->to, UIO_USERSPACE, uap->fhp));
4460 }
4461 
4462 #ifndef _SYS_SYSPROTO_H_
4463 struct fhlinkat_args {
4464 	fhandle_t *fhp;
4465 	int tofd;
4466 	const char *to;
4467 };
4468 #endif
4469 int
4470 sys_fhlinkat(struct thread *td, struct fhlinkat_args *uap)
4471 {
4472 
4473 	return (kern_fhlinkat(td, uap->tofd, uap->to, UIO_USERSPACE, uap->fhp));
4474 }
4475 
4476 static int
4477 kern_fhlinkat(struct thread *td, int fd, const char *path,
4478     enum uio_seg pathseg, fhandle_t *fhp)
4479 {
4480 	fhandle_t fh;
4481 	struct mount *mp;
4482 	struct vnode *vp;
4483 	int error;
4484 
4485 	error = priv_check(td, PRIV_VFS_GETFH);
4486 	if (error != 0)
4487 		return (error);
4488 	error = copyin(fhp, &fh, sizeof(fh));
4489 	if (error != 0)
4490 		return (error);
4491 	do {
4492 		bwillwrite();
4493 		if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4494 			return (ESTALE);
4495 		error = VFS_FHTOVP(mp, &fh.fh_fid, LK_SHARED, &vp);
4496 		vfs_unbusy(mp);
4497 		if (error != 0)
4498 			return (error);
4499 		VOP_UNLOCK(vp);
4500 		error = kern_linkat_vp(td, vp, fd, path, pathseg);
4501 	} while (error == EAGAIN || error == ERELOOKUP);
4502 	return (error);
4503 }
4504 
4505 #ifndef _SYS_SYSPROTO_H_
4506 struct fhreadlink_args {
4507 	fhandle_t *fhp;
4508 	char *buf;
4509 	size_t bufsize;
4510 };
4511 #endif
4512 int
4513 sys_fhreadlink(struct thread *td, struct fhreadlink_args *uap)
4514 {
4515 	fhandle_t fh;
4516 	struct mount *mp;
4517 	struct vnode *vp;
4518 	int error;
4519 
4520 	error = priv_check(td, PRIV_VFS_GETFH);
4521 	if (error != 0)
4522 		return (error);
4523 	if (uap->bufsize > IOSIZE_MAX)
4524 		return (EINVAL);
4525 	error = copyin(uap->fhp, &fh, sizeof(fh));
4526 	if (error != 0)
4527 		return (error);
4528 	if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4529 		return (ESTALE);
4530 	error = VFS_FHTOVP(mp, &fh.fh_fid, LK_SHARED, &vp);
4531 	vfs_unbusy(mp);
4532 	if (error != 0)
4533 		return (error);
4534 	error = kern_readlink_vp(vp, uap->buf, UIO_USERSPACE, uap->bufsize, td);
4535 	vput(vp);
4536 	return (error);
4537 }
4538 
4539 /*
4540  * syscall for the rpc.lockd to use to translate a NFS file handle into an
4541  * open descriptor.
4542  *
4543  * warning: do not remove the priv_check() call or this becomes one giant
4544  * security hole.
4545  */
4546 #ifndef _SYS_SYSPROTO_H_
4547 struct fhopen_args {
4548 	const struct fhandle *u_fhp;
4549 	int flags;
4550 };
4551 #endif
4552 int
4553 sys_fhopen(struct thread *td, struct fhopen_args *uap)
4554 {
4555 	return (kern_fhopen(td, uap->u_fhp, uap->flags));
4556 }
4557 
4558 int
4559 kern_fhopen(struct thread *td, const struct fhandle *u_fhp, int flags)
4560 {
4561 	struct mount *mp;
4562 	struct vnode *vp;
4563 	struct fhandle fhp;
4564 	struct file *fp;
4565 	int fmode, error;
4566 	int indx;
4567 
4568 	error = priv_check(td, PRIV_VFS_FHOPEN);
4569 	if (error != 0)
4570 		return (error);
4571 	indx = -1;
4572 	fmode = FFLAGS(flags);
4573 	/* why not allow a non-read/write open for our lockd? */
4574 	if (((fmode & (FREAD | FWRITE)) == 0) || (fmode & O_CREAT))
4575 		return (EINVAL);
4576 	error = copyin(u_fhp, &fhp, sizeof(fhp));
4577 	if (error != 0)
4578 		return(error);
4579 	/* find the mount point */
4580 	mp = vfs_busyfs(&fhp.fh_fsid);
4581 	if (mp == NULL)
4582 		return (ESTALE);
4583 	/* now give me my vnode, it gets returned to me locked */
4584 	error = VFS_FHTOVP(mp, &fhp.fh_fid, LK_EXCLUSIVE, &vp);
4585 	vfs_unbusy(mp);
4586 	if (error != 0)
4587 		return (error);
4588 
4589 	error = falloc_noinstall(td, &fp);
4590 	if (error != 0) {
4591 		vput(vp);
4592 		return (error);
4593 	}
4594 	/*
4595 	 * An extra reference on `fp' has been held for us by
4596 	 * falloc_noinstall().
4597 	 */
4598 
4599 #ifdef INVARIANTS
4600 	td->td_dupfd = -1;
4601 #endif
4602 	error = vn_open_vnode(vp, fmode, td->td_ucred, td, fp);
4603 	if (error != 0) {
4604 		KASSERT(fp->f_ops == &badfileops,
4605 		    ("VOP_OPEN in fhopen() set f_ops"));
4606 		KASSERT(td->td_dupfd < 0,
4607 		    ("fhopen() encountered fdopen()"));
4608 
4609 		vput(vp);
4610 		goto bad;
4611 	}
4612 #ifdef INVARIANTS
4613 	td->td_dupfd = 0;
4614 #endif
4615 	fp->f_vnode = vp;
4616 	finit_vnode(fp, fmode, NULL, &vnops);
4617 	VOP_UNLOCK(vp);
4618 	if ((fmode & O_TRUNC) != 0) {
4619 		error = fo_truncate(fp, 0, td->td_ucred, td);
4620 		if (error != 0)
4621 			goto bad;
4622 	}
4623 
4624 	error = finstall(td, fp, &indx, fmode, NULL);
4625 bad:
4626 	fdrop(fp, td);
4627 	td->td_retval[0] = indx;
4628 	return (error);
4629 }
4630 
4631 /*
4632  * Stat an (NFS) file handle.
4633  */
4634 #ifndef _SYS_SYSPROTO_H_
4635 struct fhstat_args {
4636 	struct fhandle *u_fhp;
4637 	struct stat *sb;
4638 };
4639 #endif
4640 int
4641 sys_fhstat(struct thread *td, struct fhstat_args *uap)
4642 {
4643 	struct stat sb;
4644 	struct fhandle fh;
4645 	int error;
4646 
4647 	error = copyin(uap->u_fhp, &fh, sizeof(fh));
4648 	if (error != 0)
4649 		return (error);
4650 	error = kern_fhstat(td, fh, &sb);
4651 	if (error == 0)
4652 		error = copyout(&sb, uap->sb, sizeof(sb));
4653 	return (error);
4654 }
4655 
4656 int
4657 kern_fhstat(struct thread *td, struct fhandle fh, struct stat *sb)
4658 {
4659 	struct mount *mp;
4660 	struct vnode *vp;
4661 	int error;
4662 
4663 	error = priv_check(td, PRIV_VFS_FHSTAT);
4664 	if (error != 0)
4665 		return (error);
4666 	if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4667 		return (ESTALE);
4668 	error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4669 	vfs_unbusy(mp);
4670 	if (error != 0)
4671 		return (error);
4672 	error = VOP_STAT(vp, sb, td->td_ucred, NOCRED);
4673 	vput(vp);
4674 	return (error);
4675 }
4676 
4677 /*
4678  * Implement fstatfs() for (NFS) file handles.
4679  */
4680 #ifndef _SYS_SYSPROTO_H_
4681 struct fhstatfs_args {
4682 	struct fhandle *u_fhp;
4683 	struct statfs *buf;
4684 };
4685 #endif
4686 int
4687 sys_fhstatfs(struct thread *td, struct fhstatfs_args *uap)
4688 {
4689 	struct statfs *sfp;
4690 	fhandle_t fh;
4691 	int error;
4692 
4693 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
4694 	if (error != 0)
4695 		return (error);
4696 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
4697 	error = kern_fhstatfs(td, fh, sfp);
4698 	if (error == 0)
4699 		error = copyout(sfp, uap->buf, sizeof(*sfp));
4700 	free(sfp, M_STATFS);
4701 	return (error);
4702 }
4703 
4704 int
4705 kern_fhstatfs(struct thread *td, fhandle_t fh, struct statfs *buf)
4706 {
4707 	struct mount *mp;
4708 	struct vnode *vp;
4709 	int error;
4710 
4711 	error = priv_check(td, PRIV_VFS_FHSTATFS);
4712 	if (error != 0)
4713 		return (error);
4714 	if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4715 		return (ESTALE);
4716 	error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4717 	if (error != 0) {
4718 		vfs_unbusy(mp);
4719 		return (error);
4720 	}
4721 	vput(vp);
4722 	error = prison_canseemount(td->td_ucred, mp);
4723 	if (error != 0)
4724 		goto out;
4725 #ifdef MAC
4726 	error = mac_mount_check_stat(td->td_ucred, mp);
4727 	if (error != 0)
4728 		goto out;
4729 #endif
4730 	error = VFS_STATFS(mp, buf);
4731 out:
4732 	vfs_unbusy(mp);
4733 	return (error);
4734 }
4735 
4736 /*
4737  * Unlike madvise(2), we do not make a best effort to remember every
4738  * possible caching hint.  Instead, we remember the last setting with
4739  * the exception that we will allow POSIX_FADV_NORMAL to adjust the
4740  * region of any current setting.
4741  */
4742 int
4743 kern_posix_fadvise(struct thread *td, int fd, off_t offset, off_t len,
4744     int advice)
4745 {
4746 	struct fadvise_info *fa, *new;
4747 	struct file *fp;
4748 	struct vnode *vp;
4749 	off_t end;
4750 	int error;
4751 
4752 	if (offset < 0 || len < 0 || offset > OFF_MAX - len)
4753 		return (EINVAL);
4754 	AUDIT_ARG_VALUE(advice);
4755 	switch (advice) {
4756 	case POSIX_FADV_SEQUENTIAL:
4757 	case POSIX_FADV_RANDOM:
4758 	case POSIX_FADV_NOREUSE:
4759 		new = malloc(sizeof(*fa), M_FADVISE, M_WAITOK);
4760 		break;
4761 	case POSIX_FADV_NORMAL:
4762 	case POSIX_FADV_WILLNEED:
4763 	case POSIX_FADV_DONTNEED:
4764 		new = NULL;
4765 		break;
4766 	default:
4767 		return (EINVAL);
4768 	}
4769 	/* XXX: CAP_POSIX_FADVISE? */
4770 	AUDIT_ARG_FD(fd);
4771 	error = fget(td, fd, &cap_no_rights, &fp);
4772 	if (error != 0)
4773 		goto out;
4774 	AUDIT_ARG_FILE(td->td_proc, fp);
4775 	if ((fp->f_ops->fo_flags & DFLAG_SEEKABLE) == 0) {
4776 		error = ESPIPE;
4777 		goto out;
4778 	}
4779 	if (fp->f_type != DTYPE_VNODE) {
4780 		error = ENODEV;
4781 		goto out;
4782 	}
4783 	vp = fp->f_vnode;
4784 	if (vp->v_type != VREG) {
4785 		error = ENODEV;
4786 		goto out;
4787 	}
4788 	if (len == 0)
4789 		end = OFF_MAX;
4790 	else
4791 		end = offset + len - 1;
4792 	switch (advice) {
4793 	case POSIX_FADV_SEQUENTIAL:
4794 	case POSIX_FADV_RANDOM:
4795 	case POSIX_FADV_NOREUSE:
4796 		/*
4797 		 * Try to merge any existing non-standard region with
4798 		 * this new region if possible, otherwise create a new
4799 		 * non-standard region for this request.
4800 		 */
4801 		mtx_pool_lock(mtxpool_sleep, fp);
4802 		fa = fp->f_advice;
4803 		if (fa != NULL && fa->fa_advice == advice &&
4804 		    ((fa->fa_start <= end && fa->fa_end >= offset) ||
4805 		    (end != OFF_MAX && fa->fa_start == end + 1) ||
4806 		    (fa->fa_end != OFF_MAX && fa->fa_end + 1 == offset))) {
4807 			if (offset < fa->fa_start)
4808 				fa->fa_start = offset;
4809 			if (end > fa->fa_end)
4810 				fa->fa_end = end;
4811 		} else {
4812 			new->fa_advice = advice;
4813 			new->fa_start = offset;
4814 			new->fa_end = end;
4815 			fp->f_advice = new;
4816 			new = fa;
4817 		}
4818 		mtx_pool_unlock(mtxpool_sleep, fp);
4819 		break;
4820 	case POSIX_FADV_NORMAL:
4821 		/*
4822 		 * If a the "normal" region overlaps with an existing
4823 		 * non-standard region, trim or remove the
4824 		 * non-standard region.
4825 		 */
4826 		mtx_pool_lock(mtxpool_sleep, fp);
4827 		fa = fp->f_advice;
4828 		if (fa != NULL) {
4829 			if (offset <= fa->fa_start && end >= fa->fa_end) {
4830 				new = fa;
4831 				fp->f_advice = NULL;
4832 			} else if (offset <= fa->fa_start &&
4833 			    end >= fa->fa_start)
4834 				fa->fa_start = end + 1;
4835 			else if (offset <= fa->fa_end && end >= fa->fa_end)
4836 				fa->fa_end = offset - 1;
4837 			else if (offset >= fa->fa_start && end <= fa->fa_end) {
4838 				/*
4839 				 * If the "normal" region is a middle
4840 				 * portion of the existing
4841 				 * non-standard region, just remove
4842 				 * the whole thing rather than picking
4843 				 * one side or the other to
4844 				 * preserve.
4845 				 */
4846 				new = fa;
4847 				fp->f_advice = NULL;
4848 			}
4849 		}
4850 		mtx_pool_unlock(mtxpool_sleep, fp);
4851 		break;
4852 	case POSIX_FADV_WILLNEED:
4853 	case POSIX_FADV_DONTNEED:
4854 		error = VOP_ADVISE(vp, offset, end, advice);
4855 		break;
4856 	}
4857 out:
4858 	if (fp != NULL)
4859 		fdrop(fp, td);
4860 	free(new, M_FADVISE);
4861 	return (error);
4862 }
4863 
4864 int
4865 sys_posix_fadvise(struct thread *td, struct posix_fadvise_args *uap)
4866 {
4867 	int error;
4868 
4869 	error = kern_posix_fadvise(td, uap->fd, uap->offset, uap->len,
4870 	    uap->advice);
4871 	return (kern_posix_error(td, error));
4872 }
4873 
4874 int
4875 kern_copy_file_range(struct thread *td, int infd, off_t *inoffp, int outfd,
4876     off_t *outoffp, size_t len, unsigned int flags)
4877 {
4878 	struct file *infp, *outfp;
4879 	struct vnode *invp, *outvp;
4880 	int error;
4881 	size_t retlen;
4882 	void *rl_rcookie, *rl_wcookie;
4883 	off_t savinoff, savoutoff;
4884 
4885 	infp = outfp = NULL;
4886 	rl_rcookie = rl_wcookie = NULL;
4887 	savinoff = -1;
4888 	error = 0;
4889 	retlen = 0;
4890 
4891 	if (flags != 0) {
4892 		error = EINVAL;
4893 		goto out;
4894 	}
4895 	if (len > SSIZE_MAX)
4896 		/*
4897 		 * Although the len argument is size_t, the return argument
4898 		 * is ssize_t (which is signed).  Therefore a size that won't
4899 		 * fit in ssize_t can't be returned.
4900 		 */
4901 		len = SSIZE_MAX;
4902 
4903 	/* Get the file structures for the file descriptors. */
4904 	error = fget_read(td, infd, &cap_read_rights, &infp);
4905 	if (error != 0)
4906 		goto out;
4907 	if (infp->f_ops == &badfileops) {
4908 		error = EBADF;
4909 		goto out;
4910 	}
4911 	if (infp->f_vnode == NULL) {
4912 		error = EINVAL;
4913 		goto out;
4914 	}
4915 	error = fget_write(td, outfd, &cap_write_rights, &outfp);
4916 	if (error != 0)
4917 		goto out;
4918 	if (outfp->f_ops == &badfileops) {
4919 		error = EBADF;
4920 		goto out;
4921 	}
4922 	if (outfp->f_vnode == NULL) {
4923 		error = EINVAL;
4924 		goto out;
4925 	}
4926 
4927 	/* Set the offset pointers to the correct place. */
4928 	if (inoffp == NULL)
4929 		inoffp = &infp->f_offset;
4930 	if (outoffp == NULL)
4931 		outoffp = &outfp->f_offset;
4932 	savinoff = *inoffp;
4933 	savoutoff = *outoffp;
4934 
4935 	invp = infp->f_vnode;
4936 	outvp = outfp->f_vnode;
4937 	/* Sanity check the f_flag bits. */
4938 	if ((outfp->f_flag & (FWRITE | FAPPEND)) != FWRITE ||
4939 	    (infp->f_flag & FREAD) == 0) {
4940 		error = EBADF;
4941 		goto out;
4942 	}
4943 
4944 	/* If len == 0, just return 0. */
4945 	if (len == 0)
4946 		goto out;
4947 
4948 	/*
4949 	 * If infp and outfp refer to the same file, the byte ranges cannot
4950 	 * overlap.
4951 	 */
4952 	if (invp == outvp && ((savinoff <= savoutoff && savinoff + len >
4953 	    savoutoff) || (savinoff > savoutoff && savoutoff + len >
4954 	    savinoff))) {
4955 		error = EINVAL;
4956 		goto out;
4957 	}
4958 
4959 	/* Range lock the byte ranges for both invp and outvp. */
4960 	for (;;) {
4961 		rl_wcookie = vn_rangelock_wlock(outvp, *outoffp, *outoffp +
4962 		    len);
4963 		rl_rcookie = vn_rangelock_tryrlock(invp, *inoffp, *inoffp +
4964 		    len);
4965 		if (rl_rcookie != NULL)
4966 			break;
4967 		vn_rangelock_unlock(outvp, rl_wcookie);
4968 		rl_rcookie = vn_rangelock_rlock(invp, *inoffp, *inoffp + len);
4969 		vn_rangelock_unlock(invp, rl_rcookie);
4970 	}
4971 
4972 	retlen = len;
4973 	error = vn_copy_file_range(invp, inoffp, outvp, outoffp, &retlen,
4974 	    flags, infp->f_cred, outfp->f_cred, td);
4975 out:
4976 	if (rl_rcookie != NULL)
4977 		vn_rangelock_unlock(invp, rl_rcookie);
4978 	if (rl_wcookie != NULL)
4979 		vn_rangelock_unlock(outvp, rl_wcookie);
4980 	if (savinoff != -1 && (error == EINTR || error == ERESTART)) {
4981 		*inoffp = savinoff;
4982 		*outoffp = savoutoff;
4983 	}
4984 	if (outfp != NULL)
4985 		fdrop(outfp, td);
4986 	if (infp != NULL)
4987 		fdrop(infp, td);
4988 	td->td_retval[0] = retlen;
4989 	return (error);
4990 }
4991 
4992 int
4993 sys_copy_file_range(struct thread *td, struct copy_file_range_args *uap)
4994 {
4995 	off_t inoff, outoff, *inoffp, *outoffp;
4996 	int error;
4997 
4998 	inoffp = outoffp = NULL;
4999 	if (uap->inoffp != NULL) {
5000 		error = copyin(uap->inoffp, &inoff, sizeof(off_t));
5001 		if (error != 0)
5002 			return (error);
5003 		inoffp = &inoff;
5004 	}
5005 	if (uap->outoffp != NULL) {
5006 		error = copyin(uap->outoffp, &outoff, sizeof(off_t));
5007 		if (error != 0)
5008 			return (error);
5009 		outoffp = &outoff;
5010 	}
5011 	error = kern_copy_file_range(td, uap->infd, inoffp, uap->outfd,
5012 	    outoffp, uap->len, uap->flags);
5013 	if (error == 0 && uap->inoffp != NULL)
5014 		error = copyout(inoffp, uap->inoffp, sizeof(off_t));
5015 	if (error == 0 && uap->outoffp != NULL)
5016 		error = copyout(outoffp, uap->outoffp, sizeof(off_t));
5017 	return (error);
5018 }
5019