xref: /freebsd/sys/kern/vfs_syscalls.c (revision d0ba1baed3f6e4936a0c1b89c25f6c59168ef6de)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (c) 1989, 1993
5  *	The Regents of the University of California.  All rights reserved.
6  * (c) UNIX System Laboratories, Inc.
7  * All or some portions of this file are derived from material licensed
8  * to the University of California by American Telephone and Telegraph
9  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
10  * the permission of UNIX System Laboratories, Inc.
11  *
12  * Redistribution and use in source and binary forms, with or without
13  * modification, are permitted provided that the following conditions
14  * are met:
15  * 1. Redistributions of source code must retain the above copyright
16  *    notice, this list of conditions and the following disclaimer.
17  * 2. Redistributions in binary form must reproduce the above copyright
18  *    notice, this list of conditions and the following disclaimer in the
19  *    documentation and/or other materials provided with the distribution.
20  * 3. Neither the name of the University nor the names of its contributors
21  *    may be used to endorse or promote products derived from this software
22  *    without specific prior written permission.
23  *
24  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
25  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
28  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
29  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
30  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
31  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
32  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
33  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
34  * SUCH DAMAGE.
35  *
36  *	@(#)vfs_syscalls.c	8.13 (Berkeley) 4/15/94
37  */
38 
39 #include <sys/cdefs.h>
40 __FBSDID("$FreeBSD$");
41 
42 #include "opt_capsicum.h"
43 #include "opt_ktrace.h"
44 
45 #include <sys/param.h>
46 #include <sys/systm.h>
47 #include <sys/bio.h>
48 #include <sys/buf.h>
49 #include <sys/capsicum.h>
50 #include <sys/disk.h>
51 #include <sys/sysent.h>
52 #include <sys/malloc.h>
53 #include <sys/mount.h>
54 #include <sys/mutex.h>
55 #include <sys/sysproto.h>
56 #include <sys/namei.h>
57 #include <sys/filedesc.h>
58 #include <sys/kernel.h>
59 #include <sys/fcntl.h>
60 #include <sys/file.h>
61 #include <sys/filio.h>
62 #include <sys/limits.h>
63 #include <sys/linker.h>
64 #include <sys/rwlock.h>
65 #include <sys/sdt.h>
66 #include <sys/stat.h>
67 #include <sys/sx.h>
68 #include <sys/unistd.h>
69 #include <sys/vnode.h>
70 #include <sys/priv.h>
71 #include <sys/proc.h>
72 #include <sys/dirent.h>
73 #include <sys/jail.h>
74 #include <sys/syscallsubr.h>
75 #include <sys/sysctl.h>
76 #ifdef KTRACE
77 #include <sys/ktrace.h>
78 #endif
79 
80 #include <machine/stdarg.h>
81 
82 #include <security/audit/audit.h>
83 #include <security/mac/mac_framework.h>
84 
85 #include <vm/vm.h>
86 #include <vm/vm_object.h>
87 #include <vm/vm_page.h>
88 #include <vm/uma.h>
89 
90 #include <ufs/ufs/quota.h>
91 
92 MALLOC_DEFINE(M_FADVISE, "fadvise", "posix_fadvise(2) information");
93 
94 SDT_PROVIDER_DEFINE(vfs);
95 SDT_PROBE_DEFINE2(vfs, , stat, mode, "char *", "int");
96 SDT_PROBE_DEFINE2(vfs, , stat, reg, "char *", "int");
97 
98 static int kern_chflagsat(struct thread *td, int fd, const char *path,
99     enum uio_seg pathseg, u_long flags, int atflag);
100 static int setfflags(struct thread *td, struct vnode *, u_long);
101 static int getutimes(const struct timeval *, enum uio_seg, struct timespec *);
102 static int getutimens(const struct timespec *, enum uio_seg,
103     struct timespec *, int *);
104 static int setutimes(struct thread *td, struct vnode *,
105     const struct timespec *, int, int);
106 static int vn_access(struct vnode *vp, int user_flags, struct ucred *cred,
107     struct thread *td);
108 
109 /*
110  * Sync each mounted filesystem.
111  */
112 #ifndef _SYS_SYSPROTO_H_
113 struct sync_args {
114 	int     dummy;
115 };
116 #endif
117 /* ARGSUSED */
118 int
119 sys_sync(struct thread *td, struct sync_args *uap)
120 {
121 	struct mount *mp, *nmp;
122 	int save;
123 
124 	mtx_lock(&mountlist_mtx);
125 	for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
126 		if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK)) {
127 			nmp = TAILQ_NEXT(mp, mnt_list);
128 			continue;
129 		}
130 		if ((mp->mnt_flag & MNT_RDONLY) == 0 &&
131 		    vn_start_write(NULL, &mp, V_NOWAIT) == 0) {
132 			save = curthread_pflags_set(TDP_SYNCIO);
133 			vfs_msync(mp, MNT_NOWAIT);
134 			VFS_SYNC(mp, MNT_NOWAIT);
135 			curthread_pflags_restore(save);
136 			vn_finished_write(mp);
137 		}
138 		mtx_lock(&mountlist_mtx);
139 		nmp = TAILQ_NEXT(mp, mnt_list);
140 		vfs_unbusy(mp);
141 	}
142 	mtx_unlock(&mountlist_mtx);
143 	return (0);
144 }
145 
146 /*
147  * Change filesystem quotas.
148  */
149 #ifndef _SYS_SYSPROTO_H_
150 struct quotactl_args {
151 	char *path;
152 	int cmd;
153 	int uid;
154 	caddr_t arg;
155 };
156 #endif
157 int
158 sys_quotactl(struct thread *td, struct quotactl_args *uap)
159 {
160 	struct mount *mp;
161 	struct nameidata nd;
162 	int error;
163 
164 	AUDIT_ARG_CMD(uap->cmd);
165 	AUDIT_ARG_UID(uap->uid);
166 	if (!prison_allow(td->td_ucred, PR_ALLOW_QUOTAS))
167 		return (EPERM);
168 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
169 	    uap->path, td);
170 	if ((error = namei(&nd)) != 0)
171 		return (error);
172 	NDFREE(&nd, NDF_ONLY_PNBUF);
173 	mp = nd.ni_vp->v_mount;
174 	vfs_ref(mp);
175 	vput(nd.ni_vp);
176 	error = vfs_busy(mp, 0);
177 	vfs_rel(mp);
178 	if (error != 0)
179 		return (error);
180 	error = VFS_QUOTACTL(mp, uap->cmd, uap->uid, uap->arg);
181 
182 	/*
183 	 * Since quota on operation typically needs to open quota
184 	 * file, the Q_QUOTAON handler needs to unbusy the mount point
185 	 * before calling into namei.  Otherwise, unmount might be
186 	 * started between two vfs_busy() invocations (first is our,
187 	 * second is from mount point cross-walk code in lookup()),
188 	 * causing deadlock.
189 	 *
190 	 * Require that Q_QUOTAON handles the vfs_busy() reference on
191 	 * its own, always returning with ubusied mount point.
192 	 */
193 	if ((uap->cmd >> SUBCMDSHIFT) != Q_QUOTAON)
194 		vfs_unbusy(mp);
195 	return (error);
196 }
197 
198 /*
199  * Used by statfs conversion routines to scale the block size up if
200  * necessary so that all of the block counts are <= 'max_size'.  Note
201  * that 'max_size' should be a bitmask, i.e. 2^n - 1 for some non-zero
202  * value of 'n'.
203  */
204 void
205 statfs_scale_blocks(struct statfs *sf, long max_size)
206 {
207 	uint64_t count;
208 	int shift;
209 
210 	KASSERT(powerof2(max_size + 1), ("%s: invalid max_size", __func__));
211 
212 	/*
213 	 * Attempt to scale the block counts to give a more accurate
214 	 * overview to userland of the ratio of free space to used
215 	 * space.  To do this, find the largest block count and compute
216 	 * a divisor that lets it fit into a signed integer <= max_size.
217 	 */
218 	if (sf->f_bavail < 0)
219 		count = -sf->f_bavail;
220 	else
221 		count = sf->f_bavail;
222 	count = MAX(sf->f_blocks, MAX(sf->f_bfree, count));
223 	if (count <= max_size)
224 		return;
225 
226 	count >>= flsl(max_size);
227 	shift = 0;
228 	while (count > 0) {
229 		shift++;
230 		count >>=1;
231 	}
232 
233 	sf->f_bsize <<= shift;
234 	sf->f_blocks >>= shift;
235 	sf->f_bfree >>= shift;
236 	sf->f_bavail >>= shift;
237 }
238 
239 static int
240 kern_do_statfs(struct thread *td, struct mount *mp, struct statfs *buf)
241 {
242 	struct statfs *sp;
243 	int error;
244 
245 	if (mp == NULL)
246 		return (EBADF);
247 	error = vfs_busy(mp, 0);
248 	vfs_rel(mp);
249 	if (error != 0)
250 		return (error);
251 #ifdef MAC
252 	error = mac_mount_check_stat(td->td_ucred, mp);
253 	if (error != 0)
254 		goto out;
255 #endif
256 	/*
257 	 * Set these in case the underlying filesystem fails to do so.
258 	 */
259 	sp = &mp->mnt_stat;
260 	sp->f_version = STATFS_VERSION;
261 	sp->f_namemax = NAME_MAX;
262 	sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
263 	error = VFS_STATFS(mp, sp);
264 	if (error != 0)
265 		goto out;
266 	*buf = *sp;
267 	if (priv_check(td, PRIV_VFS_GENERATION)) {
268 		buf->f_fsid.val[0] = buf->f_fsid.val[1] = 0;
269 		prison_enforce_statfs(td->td_ucred, mp, buf);
270 	}
271 out:
272 	vfs_unbusy(mp);
273 	return (error);
274 }
275 
276 /*
277  * Get filesystem statistics.
278  */
279 #ifndef _SYS_SYSPROTO_H_
280 struct statfs_args {
281 	char *path;
282 	struct statfs *buf;
283 };
284 #endif
285 int
286 sys_statfs(struct thread *td, struct statfs_args *uap)
287 {
288 	struct statfs *sfp;
289 	int error;
290 
291 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
292 	error = kern_statfs(td, uap->path, UIO_USERSPACE, sfp);
293 	if (error == 0)
294 		error = copyout(sfp, uap->buf, sizeof(struct statfs));
295 	free(sfp, M_STATFS);
296 	return (error);
297 }
298 
299 int
300 kern_statfs(struct thread *td, char *path, enum uio_seg pathseg,
301     struct statfs *buf)
302 {
303 	struct mount *mp;
304 	struct nameidata nd;
305 	int error;
306 
307 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
308 	    pathseg, path, td);
309 	error = namei(&nd);
310 	if (error != 0)
311 		return (error);
312 	mp = nd.ni_vp->v_mount;
313 	vfs_ref(mp);
314 	NDFREE(&nd, NDF_ONLY_PNBUF);
315 	vput(nd.ni_vp);
316 	return (kern_do_statfs(td, mp, buf));
317 }
318 
319 /*
320  * Get filesystem statistics.
321  */
322 #ifndef _SYS_SYSPROTO_H_
323 struct fstatfs_args {
324 	int fd;
325 	struct statfs *buf;
326 };
327 #endif
328 int
329 sys_fstatfs(struct thread *td, struct fstatfs_args *uap)
330 {
331 	struct statfs *sfp;
332 	int error;
333 
334 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
335 	error = kern_fstatfs(td, uap->fd, sfp);
336 	if (error == 0)
337 		error = copyout(sfp, uap->buf, sizeof(struct statfs));
338 	free(sfp, M_STATFS);
339 	return (error);
340 }
341 
342 int
343 kern_fstatfs(struct thread *td, int fd, struct statfs *buf)
344 {
345 	struct file *fp;
346 	struct mount *mp;
347 	struct vnode *vp;
348 	cap_rights_t rights;
349 	int error;
350 
351 	AUDIT_ARG_FD(fd);
352 	error = getvnode(td, fd, cap_rights_init(&rights, CAP_FSTATFS), &fp);
353 	if (error != 0)
354 		return (error);
355 	vp = fp->f_vnode;
356 	vn_lock(vp, LK_SHARED | LK_RETRY);
357 #ifdef AUDIT
358 	AUDIT_ARG_VNODE1(vp);
359 #endif
360 	mp = vp->v_mount;
361 	if (mp != NULL)
362 		vfs_ref(mp);
363 	VOP_UNLOCK(vp, 0);
364 	fdrop(fp, td);
365 	return (kern_do_statfs(td, mp, buf));
366 }
367 
368 /*
369  * Get statistics on all filesystems.
370  */
371 #ifndef _SYS_SYSPROTO_H_
372 struct getfsstat_args {
373 	struct statfs *buf;
374 	long bufsize;
375 	int mode;
376 };
377 #endif
378 int
379 sys_getfsstat(struct thread *td, struct getfsstat_args *uap)
380 {
381 	size_t count;
382 	int error;
383 
384 	if (uap->bufsize < 0 || uap->bufsize > SIZE_MAX)
385 		return (EINVAL);
386 	error = kern_getfsstat(td, &uap->buf, uap->bufsize, &count,
387 	    UIO_USERSPACE, uap->mode);
388 	if (error == 0)
389 		td->td_retval[0] = count;
390 	return (error);
391 }
392 
393 /*
394  * If (bufsize > 0 && bufseg == UIO_SYSSPACE)
395  *	The caller is responsible for freeing memory which will be allocated
396  *	in '*buf'.
397  */
398 int
399 kern_getfsstat(struct thread *td, struct statfs **buf, size_t bufsize,
400     size_t *countp, enum uio_seg bufseg, int mode)
401 {
402 	struct mount *mp, *nmp;
403 	struct statfs *sfsp, *sp, *sptmp, *tofree;
404 	size_t count, maxcount;
405 	int error;
406 
407 	switch (mode) {
408 	case MNT_WAIT:
409 	case MNT_NOWAIT:
410 		break;
411 	default:
412 		if (bufseg == UIO_SYSSPACE)
413 			*buf = NULL;
414 		return (EINVAL);
415 	}
416 restart:
417 	maxcount = bufsize / sizeof(struct statfs);
418 	if (bufsize == 0) {
419 		sfsp = NULL;
420 		tofree = NULL;
421 	} else if (bufseg == UIO_USERSPACE) {
422 		sfsp = *buf;
423 		tofree = NULL;
424 	} else /* if (bufseg == UIO_SYSSPACE) */ {
425 		count = 0;
426 		mtx_lock(&mountlist_mtx);
427 		TAILQ_FOREACH(mp, &mountlist, mnt_list) {
428 			count++;
429 		}
430 		mtx_unlock(&mountlist_mtx);
431 		if (maxcount > count)
432 			maxcount = count;
433 		tofree = sfsp = *buf = malloc(maxcount * sizeof(struct statfs),
434 		    M_STATFS, M_WAITOK);
435 	}
436 	count = 0;
437 	mtx_lock(&mountlist_mtx);
438 	for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
439 		if (prison_canseemount(td->td_ucred, mp) != 0) {
440 			nmp = TAILQ_NEXT(mp, mnt_list);
441 			continue;
442 		}
443 #ifdef MAC
444 		if (mac_mount_check_stat(td->td_ucred, mp) != 0) {
445 			nmp = TAILQ_NEXT(mp, mnt_list);
446 			continue;
447 		}
448 #endif
449 		if (mode == MNT_WAIT) {
450 			if (vfs_busy(mp, MBF_MNTLSTLOCK) != 0) {
451 				/*
452 				 * If vfs_busy() failed, and MBF_NOWAIT
453 				 * wasn't passed, then the mp is gone.
454 				 * Furthermore, because of MBF_MNTLSTLOCK,
455 				 * the mountlist_mtx was dropped.  We have
456 				 * no other choice than to start over.
457 				 */
458 				mtx_unlock(&mountlist_mtx);
459 				free(tofree, M_STATFS);
460 				goto restart;
461 			}
462 		} else {
463 			if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK) != 0) {
464 				nmp = TAILQ_NEXT(mp, mnt_list);
465 				continue;
466 			}
467 		}
468 		if (sfsp != NULL && count < maxcount) {
469 			sp = &mp->mnt_stat;
470 			/*
471 			 * Set these in case the underlying filesystem
472 			 * fails to do so.
473 			 */
474 			sp->f_version = STATFS_VERSION;
475 			sp->f_namemax = NAME_MAX;
476 			sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
477 			/*
478 			 * If MNT_NOWAIT is specified, do not refresh
479 			 * the fsstat cache.
480 			 */
481 			if (mode != MNT_NOWAIT) {
482 				error = VFS_STATFS(mp, sp);
483 				if (error != 0) {
484 					mtx_lock(&mountlist_mtx);
485 					nmp = TAILQ_NEXT(mp, mnt_list);
486 					vfs_unbusy(mp);
487 					continue;
488 				}
489 			}
490 			if (priv_check(td, PRIV_VFS_GENERATION)) {
491 				sptmp = malloc(sizeof(struct statfs), M_STATFS,
492 				    M_WAITOK);
493 				*sptmp = *sp;
494 				sptmp->f_fsid.val[0] = sptmp->f_fsid.val[1] = 0;
495 				prison_enforce_statfs(td->td_ucred, mp, sptmp);
496 				sp = sptmp;
497 			} else
498 				sptmp = NULL;
499 			if (bufseg == UIO_SYSSPACE) {
500 				bcopy(sp, sfsp, sizeof(*sp));
501 				free(sptmp, M_STATFS);
502 			} else /* if (bufseg == UIO_USERSPACE) */ {
503 				error = copyout(sp, sfsp, sizeof(*sp));
504 				free(sptmp, M_STATFS);
505 				if (error != 0) {
506 					vfs_unbusy(mp);
507 					return (error);
508 				}
509 			}
510 			sfsp++;
511 		}
512 		count++;
513 		mtx_lock(&mountlist_mtx);
514 		nmp = TAILQ_NEXT(mp, mnt_list);
515 		vfs_unbusy(mp);
516 	}
517 	mtx_unlock(&mountlist_mtx);
518 	if (sfsp != NULL && count > maxcount)
519 		*countp = maxcount;
520 	else
521 		*countp = count;
522 	return (0);
523 }
524 
525 #ifdef COMPAT_FREEBSD4
526 /*
527  * Get old format filesystem statistics.
528  */
529 static void freebsd4_cvtstatfs(struct statfs *, struct ostatfs *);
530 
531 #ifndef _SYS_SYSPROTO_H_
532 struct freebsd4_statfs_args {
533 	char *path;
534 	struct ostatfs *buf;
535 };
536 #endif
537 int
538 freebsd4_statfs(struct thread *td, struct freebsd4_statfs_args *uap)
539 {
540 	struct ostatfs osb;
541 	struct statfs *sfp;
542 	int error;
543 
544 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
545 	error = kern_statfs(td, uap->path, UIO_USERSPACE, sfp);
546 	if (error == 0) {
547 		freebsd4_cvtstatfs(sfp, &osb);
548 		error = copyout(&osb, uap->buf, sizeof(osb));
549 	}
550 	free(sfp, M_STATFS);
551 	return (error);
552 }
553 
554 /*
555  * Get filesystem statistics.
556  */
557 #ifndef _SYS_SYSPROTO_H_
558 struct freebsd4_fstatfs_args {
559 	int fd;
560 	struct ostatfs *buf;
561 };
562 #endif
563 int
564 freebsd4_fstatfs(struct thread *td, struct freebsd4_fstatfs_args *uap)
565 {
566 	struct ostatfs osb;
567 	struct statfs *sfp;
568 	int error;
569 
570 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
571 	error = kern_fstatfs(td, uap->fd, sfp);
572 	if (error == 0) {
573 		freebsd4_cvtstatfs(sfp, &osb);
574 		error = copyout(&osb, uap->buf, sizeof(osb));
575 	}
576 	free(sfp, M_STATFS);
577 	return (error);
578 }
579 
580 /*
581  * Get statistics on all filesystems.
582  */
583 #ifndef _SYS_SYSPROTO_H_
584 struct freebsd4_getfsstat_args {
585 	struct ostatfs *buf;
586 	long bufsize;
587 	int mode;
588 };
589 #endif
590 int
591 freebsd4_getfsstat(struct thread *td, struct freebsd4_getfsstat_args *uap)
592 {
593 	struct statfs *buf, *sp;
594 	struct ostatfs osb;
595 	size_t count, size;
596 	int error;
597 
598 	if (uap->bufsize < 0)
599 		return (EINVAL);
600 	count = uap->bufsize / sizeof(struct ostatfs);
601 	if (count > SIZE_MAX / sizeof(struct statfs))
602 		return (EINVAL);
603 	size = count * sizeof(struct statfs);
604 	error = kern_getfsstat(td, &buf, size, &count, UIO_SYSSPACE,
605 	    uap->mode);
606 	if (error == 0)
607 		td->td_retval[0] = count;
608 	if (size != 0) {
609 		sp = buf;
610 		while (count != 0 && error == 0) {
611 			freebsd4_cvtstatfs(sp, &osb);
612 			error = copyout(&osb, uap->buf, sizeof(osb));
613 			sp++;
614 			uap->buf++;
615 			count--;
616 		}
617 		free(buf, M_STATFS);
618 	}
619 	return (error);
620 }
621 
622 /*
623  * Implement fstatfs() for (NFS) file handles.
624  */
625 #ifndef _SYS_SYSPROTO_H_
626 struct freebsd4_fhstatfs_args {
627 	struct fhandle *u_fhp;
628 	struct ostatfs *buf;
629 };
630 #endif
631 int
632 freebsd4_fhstatfs(struct thread *td, struct freebsd4_fhstatfs_args *uap)
633 {
634 	struct ostatfs osb;
635 	struct statfs *sfp;
636 	fhandle_t fh;
637 	int error;
638 
639 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
640 	if (error != 0)
641 		return (error);
642 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
643 	error = kern_fhstatfs(td, fh, sfp);
644 	if (error == 0) {
645 		freebsd4_cvtstatfs(sfp, &osb);
646 		error = copyout(&osb, uap->buf, sizeof(osb));
647 	}
648 	free(sfp, M_STATFS);
649 	return (error);
650 }
651 
652 /*
653  * Convert a new format statfs structure to an old format statfs structure.
654  */
655 static void
656 freebsd4_cvtstatfs(struct statfs *nsp, struct ostatfs *osp)
657 {
658 
659 	statfs_scale_blocks(nsp, LONG_MAX);
660 	bzero(osp, sizeof(*osp));
661 	osp->f_bsize = nsp->f_bsize;
662 	osp->f_iosize = MIN(nsp->f_iosize, LONG_MAX);
663 	osp->f_blocks = nsp->f_blocks;
664 	osp->f_bfree = nsp->f_bfree;
665 	osp->f_bavail = nsp->f_bavail;
666 	osp->f_files = MIN(nsp->f_files, LONG_MAX);
667 	osp->f_ffree = MIN(nsp->f_ffree, LONG_MAX);
668 	osp->f_owner = nsp->f_owner;
669 	osp->f_type = nsp->f_type;
670 	osp->f_flags = nsp->f_flags;
671 	osp->f_syncwrites = MIN(nsp->f_syncwrites, LONG_MAX);
672 	osp->f_asyncwrites = MIN(nsp->f_asyncwrites, LONG_MAX);
673 	osp->f_syncreads = MIN(nsp->f_syncreads, LONG_MAX);
674 	osp->f_asyncreads = MIN(nsp->f_asyncreads, LONG_MAX);
675 	strlcpy(osp->f_fstypename, nsp->f_fstypename,
676 	    MIN(MFSNAMELEN, OMFSNAMELEN));
677 	strlcpy(osp->f_mntonname, nsp->f_mntonname,
678 	    MIN(MNAMELEN, OMNAMELEN));
679 	strlcpy(osp->f_mntfromname, nsp->f_mntfromname,
680 	    MIN(MNAMELEN, OMNAMELEN));
681 	osp->f_fsid = nsp->f_fsid;
682 }
683 #endif /* COMPAT_FREEBSD4 */
684 
685 #if defined(COMPAT_FREEBSD11)
686 /*
687  * Get old format filesystem statistics.
688  */
689 static void freebsd11_cvtstatfs(struct statfs *, struct freebsd11_statfs *);
690 
691 int
692 freebsd11_statfs(struct thread *td, struct freebsd11_statfs_args *uap)
693 {
694 	struct freebsd11_statfs osb;
695 	struct statfs *sfp;
696 	int error;
697 
698 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
699 	error = kern_statfs(td, uap->path, UIO_USERSPACE, sfp);
700 	if (error == 0) {
701 		freebsd11_cvtstatfs(sfp, &osb);
702 		error = copyout(&osb, uap->buf, sizeof(osb));
703 	}
704 	free(sfp, M_STATFS);
705 	return (error);
706 }
707 
708 /*
709  * Get filesystem statistics.
710  */
711 int
712 freebsd11_fstatfs(struct thread *td, struct freebsd11_fstatfs_args *uap)
713 {
714 	struct freebsd11_statfs osb;
715 	struct statfs *sfp;
716 	int error;
717 
718 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
719 	error = kern_fstatfs(td, uap->fd, sfp);
720 	if (error == 0) {
721 		freebsd11_cvtstatfs(sfp, &osb);
722 		error = copyout(&osb, uap->buf, sizeof(osb));
723 	}
724 	free(sfp, M_STATFS);
725 	return (error);
726 }
727 
728 /*
729  * Get statistics on all filesystems.
730  */
731 int
732 freebsd11_getfsstat(struct thread *td, struct freebsd11_getfsstat_args *uap)
733 {
734 	struct freebsd11_statfs osb;
735 	struct statfs *buf, *sp;
736 	size_t count, size;
737 	int error;
738 
739 	count = uap->bufsize / sizeof(struct ostatfs);
740 	size = count * sizeof(struct statfs);
741 	error = kern_getfsstat(td, &buf, size, &count, UIO_SYSSPACE,
742 	    uap->mode);
743 	if (error == 0)
744 		td->td_retval[0] = count;
745 	if (size > 0) {
746 		sp = buf;
747 		while (count > 0 && error == 0) {
748 			freebsd11_cvtstatfs(sp, &osb);
749 			error = copyout(&osb, uap->buf, sizeof(osb));
750 			sp++;
751 			uap->buf++;
752 			count--;
753 		}
754 		free(buf, M_STATFS);
755 	}
756 	return (error);
757 }
758 
759 /*
760  * Implement fstatfs() for (NFS) file handles.
761  */
762 int
763 freebsd11_fhstatfs(struct thread *td, struct freebsd11_fhstatfs_args *uap)
764 {
765 	struct freebsd11_statfs osb;
766 	struct statfs *sfp;
767 	fhandle_t fh;
768 	int error;
769 
770 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
771 	if (error)
772 		return (error);
773 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
774 	error = kern_fhstatfs(td, fh, sfp);
775 	if (error == 0) {
776 		freebsd11_cvtstatfs(sfp, &osb);
777 		error = copyout(&osb, uap->buf, sizeof(osb));
778 	}
779 	free(sfp, M_STATFS);
780 	return (error);
781 }
782 
783 /*
784  * Convert a new format statfs structure to an old format statfs structure.
785  */
786 static void
787 freebsd11_cvtstatfs(struct statfs *nsp, struct freebsd11_statfs *osp)
788 {
789 
790 	bzero(osp, sizeof(*osp));
791 	osp->f_version = FREEBSD11_STATFS_VERSION;
792 	osp->f_type = nsp->f_type;
793 	osp->f_flags = nsp->f_flags;
794 	osp->f_bsize = nsp->f_bsize;
795 	osp->f_iosize = nsp->f_iosize;
796 	osp->f_blocks = nsp->f_blocks;
797 	osp->f_bfree = nsp->f_bfree;
798 	osp->f_bavail = nsp->f_bavail;
799 	osp->f_files = nsp->f_files;
800 	osp->f_ffree = nsp->f_ffree;
801 	osp->f_syncwrites = nsp->f_syncwrites;
802 	osp->f_asyncwrites = nsp->f_asyncwrites;
803 	osp->f_syncreads = nsp->f_syncreads;
804 	osp->f_asyncreads = nsp->f_asyncreads;
805 	osp->f_namemax = nsp->f_namemax;
806 	osp->f_owner = nsp->f_owner;
807 	osp->f_fsid = nsp->f_fsid;
808 	strlcpy(osp->f_fstypename, nsp->f_fstypename,
809 	    MIN(MFSNAMELEN, sizeof(osp->f_fstypename)));
810 	strlcpy(osp->f_mntonname, nsp->f_mntonname,
811 	    MIN(MNAMELEN, sizeof(osp->f_mntonname)));
812 	strlcpy(osp->f_mntfromname, nsp->f_mntfromname,
813 	    MIN(MNAMELEN, sizeof(osp->f_mntfromname)));
814 }
815 #endif /* COMPAT_FREEBSD11 */
816 
817 /*
818  * Change current working directory to a given file descriptor.
819  */
820 #ifndef _SYS_SYSPROTO_H_
821 struct fchdir_args {
822 	int	fd;
823 };
824 #endif
825 int
826 sys_fchdir(struct thread *td, struct fchdir_args *uap)
827 {
828 	struct vnode *vp, *tdp;
829 	struct mount *mp;
830 	struct file *fp;
831 	int error;
832 
833 	AUDIT_ARG_FD(uap->fd);
834 	error = getvnode(td, uap->fd, &cap_fchdir_rights,
835 	    &fp);
836 	if (error != 0)
837 		return (error);
838 	vp = fp->f_vnode;
839 	vrefact(vp);
840 	fdrop(fp, td);
841 	vn_lock(vp, LK_SHARED | LK_RETRY);
842 	AUDIT_ARG_VNODE1(vp);
843 	error = change_dir(vp, td);
844 	while (!error && (mp = vp->v_mountedhere) != NULL) {
845 		if (vfs_busy(mp, 0))
846 			continue;
847 		error = VFS_ROOT(mp, LK_SHARED, &tdp);
848 		vfs_unbusy(mp);
849 		if (error != 0)
850 			break;
851 		vput(vp);
852 		vp = tdp;
853 	}
854 	if (error != 0) {
855 		vput(vp);
856 		return (error);
857 	}
858 	VOP_UNLOCK(vp, 0);
859 	pwd_chdir(td, vp);
860 	return (0);
861 }
862 
863 /*
864  * Change current working directory (``.'').
865  */
866 #ifndef _SYS_SYSPROTO_H_
867 struct chdir_args {
868 	char	*path;
869 };
870 #endif
871 int
872 sys_chdir(struct thread *td, struct chdir_args *uap)
873 {
874 
875 	return (kern_chdir(td, uap->path, UIO_USERSPACE));
876 }
877 
878 int
879 kern_chdir(struct thread *td, char *path, enum uio_seg pathseg)
880 {
881 	struct nameidata nd;
882 	int error;
883 
884 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
885 	    pathseg, path, td);
886 	if ((error = namei(&nd)) != 0)
887 		return (error);
888 	if ((error = change_dir(nd.ni_vp, td)) != 0) {
889 		vput(nd.ni_vp);
890 		NDFREE(&nd, NDF_ONLY_PNBUF);
891 		return (error);
892 	}
893 	VOP_UNLOCK(nd.ni_vp, 0);
894 	NDFREE(&nd, NDF_ONLY_PNBUF);
895 	pwd_chdir(td, nd.ni_vp);
896 	return (0);
897 }
898 
899 /*
900  * Change notion of root (``/'') directory.
901  */
902 #ifndef _SYS_SYSPROTO_H_
903 struct chroot_args {
904 	char	*path;
905 };
906 #endif
907 int
908 sys_chroot(struct thread *td, struct chroot_args *uap)
909 {
910 	struct nameidata nd;
911 	int error;
912 
913 	error = priv_check(td, PRIV_VFS_CHROOT);
914 	if (error != 0)
915 		return (error);
916 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
917 	    UIO_USERSPACE, uap->path, td);
918 	error = namei(&nd);
919 	if (error != 0)
920 		goto error;
921 	error = change_dir(nd.ni_vp, td);
922 	if (error != 0)
923 		goto e_vunlock;
924 #ifdef MAC
925 	error = mac_vnode_check_chroot(td->td_ucred, nd.ni_vp);
926 	if (error != 0)
927 		goto e_vunlock;
928 #endif
929 	VOP_UNLOCK(nd.ni_vp, 0);
930 	error = pwd_chroot(td, nd.ni_vp);
931 	vrele(nd.ni_vp);
932 	NDFREE(&nd, NDF_ONLY_PNBUF);
933 	return (error);
934 e_vunlock:
935 	vput(nd.ni_vp);
936 error:
937 	NDFREE(&nd, NDF_ONLY_PNBUF);
938 	return (error);
939 }
940 
941 /*
942  * Common routine for chroot and chdir.  Callers must provide a locked vnode
943  * instance.
944  */
945 int
946 change_dir(struct vnode *vp, struct thread *td)
947 {
948 #ifdef MAC
949 	int error;
950 #endif
951 
952 	ASSERT_VOP_LOCKED(vp, "change_dir(): vp not locked");
953 	if (vp->v_type != VDIR)
954 		return (ENOTDIR);
955 #ifdef MAC
956 	error = mac_vnode_check_chdir(td->td_ucred, vp);
957 	if (error != 0)
958 		return (error);
959 #endif
960 	return (VOP_ACCESS(vp, VEXEC, td->td_ucred, td));
961 }
962 
963 static __inline void
964 flags_to_rights(int flags, cap_rights_t *rightsp)
965 {
966 
967 	if (flags & O_EXEC) {
968 		cap_rights_set(rightsp, CAP_FEXECVE);
969 	} else {
970 		switch ((flags & O_ACCMODE)) {
971 		case O_RDONLY:
972 			cap_rights_set(rightsp, CAP_READ);
973 			break;
974 		case O_RDWR:
975 			cap_rights_set(rightsp, CAP_READ);
976 			/* FALLTHROUGH */
977 		case O_WRONLY:
978 			cap_rights_set(rightsp, CAP_WRITE);
979 			if (!(flags & (O_APPEND | O_TRUNC)))
980 				cap_rights_set(rightsp, CAP_SEEK);
981 			break;
982 		}
983 	}
984 
985 	if (flags & O_CREAT)
986 		cap_rights_set(rightsp, CAP_CREATE);
987 
988 	if (flags & O_TRUNC)
989 		cap_rights_set(rightsp, CAP_FTRUNCATE);
990 
991 	if (flags & (O_SYNC | O_FSYNC))
992 		cap_rights_set(rightsp, CAP_FSYNC);
993 
994 	if (flags & (O_EXLOCK | O_SHLOCK))
995 		cap_rights_set(rightsp, CAP_FLOCK);
996 }
997 
998 /*
999  * Check permissions, allocate an open file structure, and call the device
1000  * open routine if any.
1001  */
1002 #ifndef _SYS_SYSPROTO_H_
1003 struct open_args {
1004 	char	*path;
1005 	int	flags;
1006 	int	mode;
1007 };
1008 #endif
1009 int
1010 sys_open(struct thread *td, struct open_args *uap)
1011 {
1012 
1013 	return (kern_openat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1014 	    uap->flags, uap->mode));
1015 }
1016 
1017 #ifndef _SYS_SYSPROTO_H_
1018 struct openat_args {
1019 	int	fd;
1020 	char	*path;
1021 	int	flag;
1022 	int	mode;
1023 };
1024 #endif
1025 int
1026 sys_openat(struct thread *td, struct openat_args *uap)
1027 {
1028 
1029 	AUDIT_ARG_FD(uap->fd);
1030 	return (kern_openat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
1031 	    uap->mode));
1032 }
1033 
1034 int
1035 kern_openat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1036     int flags, int mode)
1037 {
1038 	struct proc *p = td->td_proc;
1039 	struct filedesc *fdp = p->p_fd;
1040 	struct file *fp;
1041 	struct vnode *vp;
1042 	struct nameidata nd;
1043 	cap_rights_t rights;
1044 	int cmode, error, indx;
1045 
1046 	indx = -1;
1047 
1048 	AUDIT_ARG_FFLAGS(flags);
1049 	AUDIT_ARG_MODE(mode);
1050 	cap_rights_init(&rights, CAP_LOOKUP);
1051 	flags_to_rights(flags, &rights);
1052 	/*
1053 	 * Only one of the O_EXEC, O_RDONLY, O_WRONLY and O_RDWR flags
1054 	 * may be specified.
1055 	 */
1056 	if (flags & O_EXEC) {
1057 		if (flags & O_ACCMODE)
1058 			return (EINVAL);
1059 	} else if ((flags & O_ACCMODE) == O_ACCMODE) {
1060 		return (EINVAL);
1061 	} else {
1062 		flags = FFLAGS(flags);
1063 	}
1064 
1065 	/*
1066 	 * Allocate a file structure. The descriptor to reference it
1067 	 * is allocated and set by finstall() below.
1068 	 */
1069 	error = falloc_noinstall(td, &fp);
1070 	if (error != 0)
1071 		return (error);
1072 	/*
1073 	 * An extra reference on `fp' has been held for us by
1074 	 * falloc_noinstall().
1075 	 */
1076 	/* Set the flags early so the finit in devfs can pick them up. */
1077 	fp->f_flag = flags & FMASK;
1078 	cmode = ((mode & ~fdp->fd_cmask) & ALLPERMS) & ~S_ISTXT;
1079 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, fd,
1080 	    &rights, td);
1081 	td->td_dupfd = -1;		/* XXX check for fdopen */
1082 	error = vn_open(&nd, &flags, cmode, fp);
1083 	if (error != 0) {
1084 		/*
1085 		 * If the vn_open replaced the method vector, something
1086 		 * wonderous happened deep below and we just pass it up
1087 		 * pretending we know what we do.
1088 		 */
1089 		if (error == ENXIO && fp->f_ops != &badfileops)
1090 			goto success;
1091 
1092 		/*
1093 		 * Handle special fdopen() case. bleh.
1094 		 *
1095 		 * Don't do this for relative (capability) lookups; we don't
1096 		 * understand exactly what would happen, and we don't think
1097 		 * that it ever should.
1098 		 */
1099 		if ((nd.ni_lcf & NI_LCF_STRICTRELATIVE) == 0 &&
1100 		    (error == ENODEV || error == ENXIO) &&
1101 		    td->td_dupfd >= 0) {
1102 			error = dupfdopen(td, fdp, td->td_dupfd, flags, error,
1103 			    &indx);
1104 			if (error == 0)
1105 				goto success;
1106 		}
1107 
1108 		goto bad;
1109 	}
1110 	td->td_dupfd = 0;
1111 	NDFREE(&nd, NDF_ONLY_PNBUF);
1112 	vp = nd.ni_vp;
1113 
1114 	/*
1115 	 * Store the vnode, for any f_type. Typically, the vnode use
1116 	 * count is decremented by direct call to vn_closefile() for
1117 	 * files that switched type in the cdevsw fdopen() method.
1118 	 */
1119 	fp->f_vnode = vp;
1120 	/*
1121 	 * If the file wasn't claimed by devfs bind it to the normal
1122 	 * vnode operations here.
1123 	 */
1124 	if (fp->f_ops == &badfileops) {
1125 		KASSERT(vp->v_type != VFIFO, ("Unexpected fifo."));
1126 		fp->f_seqcount = 1;
1127 		finit(fp, (flags & FMASK) | (fp->f_flag & FHASLOCK),
1128 		    DTYPE_VNODE, vp, &vnops);
1129 	}
1130 
1131 	VOP_UNLOCK(vp, 0);
1132 	if (flags & O_TRUNC) {
1133 		error = fo_truncate(fp, 0, td->td_ucred, td);
1134 		if (error != 0)
1135 			goto bad;
1136 	}
1137 success:
1138 	/*
1139 	 * If we haven't already installed the FD (for dupfdopen), do so now.
1140 	 */
1141 	if (indx == -1) {
1142 		struct filecaps *fcaps;
1143 
1144 #ifdef CAPABILITIES
1145 		if ((nd.ni_lcf & NI_LCF_STRICTRELATIVE) != 0)
1146 			fcaps = &nd.ni_filecaps;
1147 		else
1148 #endif
1149 			fcaps = NULL;
1150 		error = finstall(td, fp, &indx, flags, fcaps);
1151 		/* On success finstall() consumes fcaps. */
1152 		if (error != 0) {
1153 			filecaps_free(&nd.ni_filecaps);
1154 			goto bad;
1155 		}
1156 	} else {
1157 		filecaps_free(&nd.ni_filecaps);
1158 	}
1159 
1160 	/*
1161 	 * Release our private reference, leaving the one associated with
1162 	 * the descriptor table intact.
1163 	 */
1164 	fdrop(fp, td);
1165 	td->td_retval[0] = indx;
1166 	return (0);
1167 bad:
1168 	KASSERT(indx == -1, ("indx=%d, should be -1", indx));
1169 	fdrop(fp, td);
1170 	return (error);
1171 }
1172 
1173 #ifdef COMPAT_43
1174 /*
1175  * Create a file.
1176  */
1177 #ifndef _SYS_SYSPROTO_H_
1178 struct ocreat_args {
1179 	char	*path;
1180 	int	mode;
1181 };
1182 #endif
1183 int
1184 ocreat(struct thread *td, struct ocreat_args *uap)
1185 {
1186 
1187 	return (kern_openat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1188 	    O_WRONLY | O_CREAT | O_TRUNC, uap->mode));
1189 }
1190 #endif /* COMPAT_43 */
1191 
1192 /*
1193  * Create a special file.
1194  */
1195 #ifndef _SYS_SYSPROTO_H_
1196 struct mknodat_args {
1197 	int	fd;
1198 	char	*path;
1199 	mode_t	mode;
1200 	dev_t	dev;
1201 };
1202 #endif
1203 int
1204 sys_mknodat(struct thread *td, struct mknodat_args *uap)
1205 {
1206 
1207 	return (kern_mknodat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode,
1208 	    uap->dev));
1209 }
1210 
1211 #if defined(COMPAT_FREEBSD11)
1212 int
1213 freebsd11_mknod(struct thread *td,
1214     struct freebsd11_mknod_args *uap)
1215 {
1216 
1217 	return (kern_mknodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1218 	    uap->mode, uap->dev));
1219 }
1220 
1221 int
1222 freebsd11_mknodat(struct thread *td,
1223     struct freebsd11_mknodat_args *uap)
1224 {
1225 
1226 	return (kern_mknodat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode,
1227 	    uap->dev));
1228 }
1229 #endif /* COMPAT_FREEBSD11 */
1230 
1231 int
1232 kern_mknodat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1233     int mode, dev_t dev)
1234 {
1235 	struct vnode *vp;
1236 	struct mount *mp;
1237 	struct vattr vattr;
1238 	struct nameidata nd;
1239 	cap_rights_t rights;
1240 	int error, whiteout = 0;
1241 
1242 	AUDIT_ARG_MODE(mode);
1243 	AUDIT_ARG_DEV(dev);
1244 	switch (mode & S_IFMT) {
1245 	case S_IFCHR:
1246 	case S_IFBLK:
1247 		error = priv_check(td, PRIV_VFS_MKNOD_DEV);
1248 		if (error == 0 && dev == VNOVAL)
1249 			error = EINVAL;
1250 		break;
1251 	case S_IFWHT:
1252 		error = priv_check(td, PRIV_VFS_MKNOD_WHT);
1253 		break;
1254 	case S_IFIFO:
1255 		if (dev == 0)
1256 			return (kern_mkfifoat(td, fd, path, pathseg, mode));
1257 		/* FALLTHROUGH */
1258 	default:
1259 		error = EINVAL;
1260 		break;
1261 	}
1262 	if (error != 0)
1263 		return (error);
1264 restart:
1265 	bwillwrite();
1266 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
1267 	    NOCACHE, pathseg, path, fd, cap_rights_init(&rights, CAP_MKNODAT),
1268 	    td);
1269 	if ((error = namei(&nd)) != 0)
1270 		return (error);
1271 	vp = nd.ni_vp;
1272 	if (vp != NULL) {
1273 		NDFREE(&nd, NDF_ONLY_PNBUF);
1274 		if (vp == nd.ni_dvp)
1275 			vrele(nd.ni_dvp);
1276 		else
1277 			vput(nd.ni_dvp);
1278 		vrele(vp);
1279 		return (EEXIST);
1280 	} else {
1281 		VATTR_NULL(&vattr);
1282 		vattr.va_mode = (mode & ALLPERMS) &
1283 		    ~td->td_proc->p_fd->fd_cmask;
1284 		vattr.va_rdev = dev;
1285 		whiteout = 0;
1286 
1287 		switch (mode & S_IFMT) {
1288 		case S_IFCHR:
1289 			vattr.va_type = VCHR;
1290 			break;
1291 		case S_IFBLK:
1292 			vattr.va_type = VBLK;
1293 			break;
1294 		case S_IFWHT:
1295 			whiteout = 1;
1296 			break;
1297 		default:
1298 			panic("kern_mknod: invalid mode");
1299 		}
1300 	}
1301 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1302 		NDFREE(&nd, NDF_ONLY_PNBUF);
1303 		vput(nd.ni_dvp);
1304 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1305 			return (error);
1306 		goto restart;
1307 	}
1308 #ifdef MAC
1309 	if (error == 0 && !whiteout)
1310 		error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp,
1311 		    &nd.ni_cnd, &vattr);
1312 #endif
1313 	if (error == 0) {
1314 		if (whiteout)
1315 			error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE);
1316 		else {
1317 			error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp,
1318 						&nd.ni_cnd, &vattr);
1319 			if (error == 0)
1320 				vput(nd.ni_vp);
1321 		}
1322 	}
1323 	NDFREE(&nd, NDF_ONLY_PNBUF);
1324 	vput(nd.ni_dvp);
1325 	vn_finished_write(mp);
1326 	return (error);
1327 }
1328 
1329 /*
1330  * Create a named pipe.
1331  */
1332 #ifndef _SYS_SYSPROTO_H_
1333 struct mkfifo_args {
1334 	char	*path;
1335 	int	mode;
1336 };
1337 #endif
1338 int
1339 sys_mkfifo(struct thread *td, struct mkfifo_args *uap)
1340 {
1341 
1342 	return (kern_mkfifoat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1343 	    uap->mode));
1344 }
1345 
1346 #ifndef _SYS_SYSPROTO_H_
1347 struct mkfifoat_args {
1348 	int	fd;
1349 	char	*path;
1350 	mode_t	mode;
1351 };
1352 #endif
1353 int
1354 sys_mkfifoat(struct thread *td, struct mkfifoat_args *uap)
1355 {
1356 
1357 	return (kern_mkfifoat(td, uap->fd, uap->path, UIO_USERSPACE,
1358 	    uap->mode));
1359 }
1360 
1361 int
1362 kern_mkfifoat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1363     int mode)
1364 {
1365 	struct mount *mp;
1366 	struct vattr vattr;
1367 	struct nameidata nd;
1368 	cap_rights_t rights;
1369 	int error;
1370 
1371 	AUDIT_ARG_MODE(mode);
1372 restart:
1373 	bwillwrite();
1374 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
1375 	    NOCACHE, pathseg, path, fd, cap_rights_init(&rights, CAP_MKFIFOAT),
1376 	    td);
1377 	if ((error = namei(&nd)) != 0)
1378 		return (error);
1379 	if (nd.ni_vp != NULL) {
1380 		NDFREE(&nd, NDF_ONLY_PNBUF);
1381 		if (nd.ni_vp == nd.ni_dvp)
1382 			vrele(nd.ni_dvp);
1383 		else
1384 			vput(nd.ni_dvp);
1385 		vrele(nd.ni_vp);
1386 		return (EEXIST);
1387 	}
1388 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1389 		NDFREE(&nd, NDF_ONLY_PNBUF);
1390 		vput(nd.ni_dvp);
1391 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1392 			return (error);
1393 		goto restart;
1394 	}
1395 	VATTR_NULL(&vattr);
1396 	vattr.va_type = VFIFO;
1397 	vattr.va_mode = (mode & ALLPERMS) & ~td->td_proc->p_fd->fd_cmask;
1398 #ifdef MAC
1399 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1400 	    &vattr);
1401 	if (error != 0)
1402 		goto out;
1403 #endif
1404 	error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
1405 	if (error == 0)
1406 		vput(nd.ni_vp);
1407 #ifdef MAC
1408 out:
1409 #endif
1410 	vput(nd.ni_dvp);
1411 	vn_finished_write(mp);
1412 	NDFREE(&nd, NDF_ONLY_PNBUF);
1413 	return (error);
1414 }
1415 
1416 /*
1417  * Make a hard file link.
1418  */
1419 #ifndef _SYS_SYSPROTO_H_
1420 struct link_args {
1421 	char	*path;
1422 	char	*link;
1423 };
1424 #endif
1425 int
1426 sys_link(struct thread *td, struct link_args *uap)
1427 {
1428 
1429 	return (kern_linkat(td, AT_FDCWD, AT_FDCWD, uap->path, uap->link,
1430 	    UIO_USERSPACE, FOLLOW));
1431 }
1432 
1433 #ifndef _SYS_SYSPROTO_H_
1434 struct linkat_args {
1435 	int	fd1;
1436 	char	*path1;
1437 	int	fd2;
1438 	char	*path2;
1439 	int	flag;
1440 };
1441 #endif
1442 int
1443 sys_linkat(struct thread *td, struct linkat_args *uap)
1444 {
1445 	int flag;
1446 
1447 	flag = uap->flag;
1448 	if (flag & ~AT_SYMLINK_FOLLOW)
1449 		return (EINVAL);
1450 
1451 	return (kern_linkat(td, uap->fd1, uap->fd2, uap->path1, uap->path2,
1452 	    UIO_USERSPACE, (flag & AT_SYMLINK_FOLLOW) ? FOLLOW : NOFOLLOW));
1453 }
1454 
1455 int hardlink_check_uid = 0;
1456 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_uid, CTLFLAG_RW,
1457     &hardlink_check_uid, 0,
1458     "Unprivileged processes cannot create hard links to files owned by other "
1459     "users");
1460 static int hardlink_check_gid = 0;
1461 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_gid, CTLFLAG_RW,
1462     &hardlink_check_gid, 0,
1463     "Unprivileged processes cannot create hard links to files owned by other "
1464     "groups");
1465 
1466 static int
1467 can_hardlink(struct vnode *vp, struct ucred *cred)
1468 {
1469 	struct vattr va;
1470 	int error;
1471 
1472 	if (!hardlink_check_uid && !hardlink_check_gid)
1473 		return (0);
1474 
1475 	error = VOP_GETATTR(vp, &va, cred);
1476 	if (error != 0)
1477 		return (error);
1478 
1479 	if (hardlink_check_uid && cred->cr_uid != va.va_uid) {
1480 		error = priv_check_cred(cred, PRIV_VFS_LINK, 0);
1481 		if (error != 0)
1482 			return (error);
1483 	}
1484 
1485 	if (hardlink_check_gid && !groupmember(va.va_gid, cred)) {
1486 		error = priv_check_cred(cred, PRIV_VFS_LINK, 0);
1487 		if (error != 0)
1488 			return (error);
1489 	}
1490 
1491 	return (0);
1492 }
1493 
1494 int
1495 kern_linkat(struct thread *td, int fd1, int fd2, char *path1, char *path2,
1496     enum uio_seg segflg, int follow)
1497 {
1498 	struct vnode *vp;
1499 	struct mount *mp;
1500 	struct nameidata nd;
1501 	cap_rights_t rights;
1502 	int error;
1503 
1504 again:
1505 	bwillwrite();
1506 	NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, segflg, path1, fd1,
1507 	    cap_rights_init(&rights, CAP_LINKAT_SOURCE), td);
1508 
1509 	if ((error = namei(&nd)) != 0)
1510 		return (error);
1511 	NDFREE(&nd, NDF_ONLY_PNBUF);
1512 	vp = nd.ni_vp;
1513 	if (vp->v_type == VDIR) {
1514 		vrele(vp);
1515 		return (EPERM);		/* POSIX */
1516 	}
1517 	NDINIT_ATRIGHTS(&nd, CREATE,
1518 	    LOCKPARENT | SAVENAME | AUDITVNODE2 | NOCACHE, segflg, path2, fd2,
1519 	    cap_rights_init(&rights, CAP_LINKAT_TARGET), td);
1520 	if ((error = namei(&nd)) == 0) {
1521 		if (nd.ni_vp != NULL) {
1522 			NDFREE(&nd, NDF_ONLY_PNBUF);
1523 			if (nd.ni_dvp == nd.ni_vp)
1524 				vrele(nd.ni_dvp);
1525 			else
1526 				vput(nd.ni_dvp);
1527 			vrele(nd.ni_vp);
1528 			vrele(vp);
1529 			return (EEXIST);
1530 		} else if (nd.ni_dvp->v_mount != vp->v_mount) {
1531 			/*
1532 			 * Cross-device link.  No need to recheck
1533 			 * vp->v_type, since it cannot change, except
1534 			 * to VBAD.
1535 			 */
1536 			NDFREE(&nd, NDF_ONLY_PNBUF);
1537 			vput(nd.ni_dvp);
1538 			vrele(vp);
1539 			return (EXDEV);
1540 		} else if ((error = vn_lock(vp, LK_EXCLUSIVE)) == 0) {
1541 			error = can_hardlink(vp, td->td_ucred);
1542 #ifdef MAC
1543 			if (error == 0)
1544 				error = mac_vnode_check_link(td->td_ucred,
1545 				    nd.ni_dvp, vp, &nd.ni_cnd);
1546 #endif
1547 			if (error != 0) {
1548 				vput(vp);
1549 				vput(nd.ni_dvp);
1550 				NDFREE(&nd, NDF_ONLY_PNBUF);
1551 				return (error);
1552 			}
1553 			error = vn_start_write(vp, &mp, V_NOWAIT);
1554 			if (error != 0) {
1555 				vput(vp);
1556 				vput(nd.ni_dvp);
1557 				NDFREE(&nd, NDF_ONLY_PNBUF);
1558 				error = vn_start_write(NULL, &mp,
1559 				    V_XSLEEP | PCATCH);
1560 				if (error != 0)
1561 					return (error);
1562 				goto again;
1563 			}
1564 			error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
1565 			VOP_UNLOCK(vp, 0);
1566 			vput(nd.ni_dvp);
1567 			vn_finished_write(mp);
1568 			NDFREE(&nd, NDF_ONLY_PNBUF);
1569 		} else {
1570 			vput(nd.ni_dvp);
1571 			NDFREE(&nd, NDF_ONLY_PNBUF);
1572 			vrele(vp);
1573 			goto again;
1574 		}
1575 	}
1576 	vrele(vp);
1577 	return (error);
1578 }
1579 
1580 /*
1581  * Make a symbolic link.
1582  */
1583 #ifndef _SYS_SYSPROTO_H_
1584 struct symlink_args {
1585 	char	*path;
1586 	char	*link;
1587 };
1588 #endif
1589 int
1590 sys_symlink(struct thread *td, struct symlink_args *uap)
1591 {
1592 
1593 	return (kern_symlinkat(td, uap->path, AT_FDCWD, uap->link,
1594 	    UIO_USERSPACE));
1595 }
1596 
1597 #ifndef _SYS_SYSPROTO_H_
1598 struct symlinkat_args {
1599 	char	*path;
1600 	int	fd;
1601 	char	*path2;
1602 };
1603 #endif
1604 int
1605 sys_symlinkat(struct thread *td, struct symlinkat_args *uap)
1606 {
1607 
1608 	return (kern_symlinkat(td, uap->path1, uap->fd, uap->path2,
1609 	    UIO_USERSPACE));
1610 }
1611 
1612 int
1613 kern_symlinkat(struct thread *td, char *path1, int fd, char *path2,
1614     enum uio_seg segflg)
1615 {
1616 	struct mount *mp;
1617 	struct vattr vattr;
1618 	char *syspath;
1619 	struct nameidata nd;
1620 	int error;
1621 	cap_rights_t rights;
1622 
1623 	if (segflg == UIO_SYSSPACE) {
1624 		syspath = path1;
1625 	} else {
1626 		syspath = uma_zalloc(namei_zone, M_WAITOK);
1627 		if ((error = copyinstr(path1, syspath, MAXPATHLEN, NULL)) != 0)
1628 			goto out;
1629 	}
1630 	AUDIT_ARG_TEXT(syspath);
1631 restart:
1632 	bwillwrite();
1633 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
1634 	    NOCACHE, segflg, path2, fd, cap_rights_init(&rights, CAP_SYMLINKAT),
1635 	    td);
1636 	if ((error = namei(&nd)) != 0)
1637 		goto out;
1638 	if (nd.ni_vp) {
1639 		NDFREE(&nd, NDF_ONLY_PNBUF);
1640 		if (nd.ni_vp == nd.ni_dvp)
1641 			vrele(nd.ni_dvp);
1642 		else
1643 			vput(nd.ni_dvp);
1644 		vrele(nd.ni_vp);
1645 		error = EEXIST;
1646 		goto out;
1647 	}
1648 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1649 		NDFREE(&nd, NDF_ONLY_PNBUF);
1650 		vput(nd.ni_dvp);
1651 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1652 			goto out;
1653 		goto restart;
1654 	}
1655 	VATTR_NULL(&vattr);
1656 	vattr.va_mode = ACCESSPERMS &~ td->td_proc->p_fd->fd_cmask;
1657 #ifdef MAC
1658 	vattr.va_type = VLNK;
1659 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1660 	    &vattr);
1661 	if (error != 0)
1662 		goto out2;
1663 #endif
1664 	error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, syspath);
1665 	if (error == 0)
1666 		vput(nd.ni_vp);
1667 #ifdef MAC
1668 out2:
1669 #endif
1670 	NDFREE(&nd, NDF_ONLY_PNBUF);
1671 	vput(nd.ni_dvp);
1672 	vn_finished_write(mp);
1673 out:
1674 	if (segflg != UIO_SYSSPACE)
1675 		uma_zfree(namei_zone, syspath);
1676 	return (error);
1677 }
1678 
1679 /*
1680  * Delete a whiteout from the filesystem.
1681  */
1682 #ifndef _SYS_SYSPROTO_H_
1683 struct undelete_args {
1684 	char *path;
1685 };
1686 #endif
1687 int
1688 sys_undelete(struct thread *td, struct undelete_args *uap)
1689 {
1690 	struct mount *mp;
1691 	struct nameidata nd;
1692 	int error;
1693 
1694 restart:
1695 	bwillwrite();
1696 	NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | AUDITVNODE1,
1697 	    UIO_USERSPACE, uap->path, td);
1698 	error = namei(&nd);
1699 	if (error != 0)
1700 		return (error);
1701 
1702 	if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) {
1703 		NDFREE(&nd, NDF_ONLY_PNBUF);
1704 		if (nd.ni_vp == nd.ni_dvp)
1705 			vrele(nd.ni_dvp);
1706 		else
1707 			vput(nd.ni_dvp);
1708 		if (nd.ni_vp)
1709 			vrele(nd.ni_vp);
1710 		return (EEXIST);
1711 	}
1712 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1713 		NDFREE(&nd, NDF_ONLY_PNBUF);
1714 		vput(nd.ni_dvp);
1715 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1716 			return (error);
1717 		goto restart;
1718 	}
1719 	error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE);
1720 	NDFREE(&nd, NDF_ONLY_PNBUF);
1721 	vput(nd.ni_dvp);
1722 	vn_finished_write(mp);
1723 	return (error);
1724 }
1725 
1726 /*
1727  * Delete a name from the filesystem.
1728  */
1729 #ifndef _SYS_SYSPROTO_H_
1730 struct unlink_args {
1731 	char	*path;
1732 };
1733 #endif
1734 int
1735 sys_unlink(struct thread *td, struct unlink_args *uap)
1736 {
1737 
1738 	return (kern_unlinkat(td, AT_FDCWD, uap->path, UIO_USERSPACE, 0));
1739 }
1740 
1741 #ifndef _SYS_SYSPROTO_H_
1742 struct unlinkat_args {
1743 	int	fd;
1744 	char	*path;
1745 	int	flag;
1746 };
1747 #endif
1748 int
1749 sys_unlinkat(struct thread *td, struct unlinkat_args *uap)
1750 {
1751 	int flag = uap->flag;
1752 	int fd = uap->fd;
1753 	char *path = uap->path;
1754 
1755 	if (flag & ~AT_REMOVEDIR)
1756 		return (EINVAL);
1757 
1758 	if (flag & AT_REMOVEDIR)
1759 		return (kern_rmdirat(td, fd, path, UIO_USERSPACE));
1760 	else
1761 		return (kern_unlinkat(td, fd, path, UIO_USERSPACE, 0));
1762 }
1763 
1764 int
1765 kern_unlinkat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1766     ino_t oldinum)
1767 {
1768 	struct mount *mp;
1769 	struct vnode *vp;
1770 	struct nameidata nd;
1771 	struct stat sb;
1772 	cap_rights_t rights;
1773 	int error;
1774 
1775 restart:
1776 	bwillwrite();
1777 	NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1,
1778 	    pathseg, path, fd, cap_rights_init(&rights, CAP_UNLINKAT), td);
1779 	if ((error = namei(&nd)) != 0)
1780 		return (error == EINVAL ? EPERM : error);
1781 	vp = nd.ni_vp;
1782 	if (vp->v_type == VDIR && oldinum == 0) {
1783 		error = EPERM;		/* POSIX */
1784 	} else if (oldinum != 0 &&
1785 		  ((error = vn_stat(vp, &sb, td->td_ucred, NOCRED, td)) == 0) &&
1786 		  sb.st_ino != oldinum) {
1787 			error = EIDRM;	/* Identifier removed */
1788 	} else {
1789 		/*
1790 		 * The root of a mounted filesystem cannot be deleted.
1791 		 *
1792 		 * XXX: can this only be a VDIR case?
1793 		 */
1794 		if (vp->v_vflag & VV_ROOT)
1795 			error = EBUSY;
1796 	}
1797 	if (error == 0) {
1798 		if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1799 			NDFREE(&nd, NDF_ONLY_PNBUF);
1800 			vput(nd.ni_dvp);
1801 			if (vp == nd.ni_dvp)
1802 				vrele(vp);
1803 			else
1804 				vput(vp);
1805 			if ((error = vn_start_write(NULL, &mp,
1806 			    V_XSLEEP | PCATCH)) != 0)
1807 				return (error);
1808 			goto restart;
1809 		}
1810 #ifdef MAC
1811 		error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
1812 		    &nd.ni_cnd);
1813 		if (error != 0)
1814 			goto out;
1815 #endif
1816 		vfs_notify_upper(vp, VFS_NOTIFY_UPPER_UNLINK);
1817 		error = VOP_REMOVE(nd.ni_dvp, vp, &nd.ni_cnd);
1818 #ifdef MAC
1819 out:
1820 #endif
1821 		vn_finished_write(mp);
1822 	}
1823 	NDFREE(&nd, NDF_ONLY_PNBUF);
1824 	vput(nd.ni_dvp);
1825 	if (vp == nd.ni_dvp)
1826 		vrele(vp);
1827 	else
1828 		vput(vp);
1829 	return (error);
1830 }
1831 
1832 /*
1833  * Reposition read/write file offset.
1834  */
1835 #ifndef _SYS_SYSPROTO_H_
1836 struct lseek_args {
1837 	int	fd;
1838 	int	pad;
1839 	off_t	offset;
1840 	int	whence;
1841 };
1842 #endif
1843 int
1844 sys_lseek(struct thread *td, struct lseek_args *uap)
1845 {
1846 
1847 	return (kern_lseek(td, uap->fd, uap->offset, uap->whence));
1848 }
1849 
1850 int
1851 kern_lseek(struct thread *td, int fd, off_t offset, int whence)
1852 {
1853 	struct file *fp;
1854 	cap_rights_t rights;
1855 	int error;
1856 
1857 	AUDIT_ARG_FD(fd);
1858 	error = fget(td, fd, cap_rights_init(&rights, CAP_SEEK), &fp);
1859 	if (error != 0)
1860 		return (error);
1861 	error = (fp->f_ops->fo_flags & DFLAG_SEEKABLE) != 0 ?
1862 	    fo_seek(fp, offset, whence, td) : ESPIPE;
1863 	fdrop(fp, td);
1864 	return (error);
1865 }
1866 
1867 #if defined(COMPAT_43)
1868 /*
1869  * Reposition read/write file offset.
1870  */
1871 #ifndef _SYS_SYSPROTO_H_
1872 struct olseek_args {
1873 	int	fd;
1874 	long	offset;
1875 	int	whence;
1876 };
1877 #endif
1878 int
1879 olseek(struct thread *td, struct olseek_args *uap)
1880 {
1881 
1882 	return (kern_lseek(td, uap->fd, uap->offset, uap->whence));
1883 }
1884 #endif /* COMPAT_43 */
1885 
1886 #if defined(COMPAT_FREEBSD6)
1887 /* Version with the 'pad' argument */
1888 int
1889 freebsd6_lseek(struct thread *td, struct freebsd6_lseek_args *uap)
1890 {
1891 
1892 	return (kern_lseek(td, uap->fd, uap->offset, uap->whence));
1893 }
1894 #endif
1895 
1896 /*
1897  * Check access permissions using passed credentials.
1898  */
1899 static int
1900 vn_access(struct vnode *vp, int user_flags, struct ucred *cred,
1901      struct thread *td)
1902 {
1903 	accmode_t accmode;
1904 	int error;
1905 
1906 	/* Flags == 0 means only check for existence. */
1907 	if (user_flags == 0)
1908 		return (0);
1909 
1910 	accmode = 0;
1911 	if (user_flags & R_OK)
1912 		accmode |= VREAD;
1913 	if (user_flags & W_OK)
1914 		accmode |= VWRITE;
1915 	if (user_flags & X_OK)
1916 		accmode |= VEXEC;
1917 #ifdef MAC
1918 	error = mac_vnode_check_access(cred, vp, accmode);
1919 	if (error != 0)
1920 		return (error);
1921 #endif
1922 	if ((accmode & VWRITE) == 0 || (error = vn_writechk(vp)) == 0)
1923 		error = VOP_ACCESS(vp, accmode, cred, td);
1924 	return (error);
1925 }
1926 
1927 /*
1928  * Check access permissions using "real" credentials.
1929  */
1930 #ifndef _SYS_SYSPROTO_H_
1931 struct access_args {
1932 	char	*path;
1933 	int	amode;
1934 };
1935 #endif
1936 int
1937 sys_access(struct thread *td, struct access_args *uap)
1938 {
1939 
1940 	return (kern_accessat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1941 	    0, uap->amode));
1942 }
1943 
1944 #ifndef _SYS_SYSPROTO_H_
1945 struct faccessat_args {
1946 	int	dirfd;
1947 	char	*path;
1948 	int	amode;
1949 	int	flag;
1950 }
1951 #endif
1952 int
1953 sys_faccessat(struct thread *td, struct faccessat_args *uap)
1954 {
1955 
1956 	return (kern_accessat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
1957 	    uap->amode));
1958 }
1959 
1960 int
1961 kern_accessat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1962     int flag, int amode)
1963 {
1964 	struct ucred *cred, *usecred;
1965 	struct vnode *vp;
1966 	struct nameidata nd;
1967 	cap_rights_t rights;
1968 	int error;
1969 
1970 	if (flag & ~AT_EACCESS)
1971 		return (EINVAL);
1972 	if (amode != F_OK && (amode & ~(R_OK | W_OK | X_OK)) != 0)
1973 		return (EINVAL);
1974 
1975 	/*
1976 	 * Create and modify a temporary credential instead of one that
1977 	 * is potentially shared (if we need one).
1978 	 */
1979 	cred = td->td_ucred;
1980 	if ((flag & AT_EACCESS) == 0 &&
1981 	    ((cred->cr_uid != cred->cr_ruid ||
1982 	    cred->cr_rgid != cred->cr_groups[0]))) {
1983 		usecred = crdup(cred);
1984 		usecred->cr_uid = cred->cr_ruid;
1985 		usecred->cr_groups[0] = cred->cr_rgid;
1986 		td->td_ucred = usecred;
1987 	} else
1988 		usecred = cred;
1989 	AUDIT_ARG_VALUE(amode);
1990 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF |
1991 	    AUDITVNODE1, pathseg, path, fd, cap_rights_init(&rights, CAP_FSTAT),
1992 	    td);
1993 	if ((error = namei(&nd)) != 0)
1994 		goto out;
1995 	vp = nd.ni_vp;
1996 
1997 	error = vn_access(vp, amode, usecred, td);
1998 	NDFREE(&nd, NDF_ONLY_PNBUF);
1999 	vput(vp);
2000 out:
2001 	if (usecred != cred) {
2002 		td->td_ucred = cred;
2003 		crfree(usecred);
2004 	}
2005 	return (error);
2006 }
2007 
2008 /*
2009  * Check access permissions using "effective" credentials.
2010  */
2011 #ifndef _SYS_SYSPROTO_H_
2012 struct eaccess_args {
2013 	char	*path;
2014 	int	amode;
2015 };
2016 #endif
2017 int
2018 sys_eaccess(struct thread *td, struct eaccess_args *uap)
2019 {
2020 
2021 	return (kern_accessat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2022 	    AT_EACCESS, uap->amode));
2023 }
2024 
2025 #if defined(COMPAT_43)
2026 /*
2027  * Get file status; this version follows links.
2028  */
2029 #ifndef _SYS_SYSPROTO_H_
2030 struct ostat_args {
2031 	char	*path;
2032 	struct ostat *ub;
2033 };
2034 #endif
2035 int
2036 ostat(struct thread *td, struct ostat_args *uap)
2037 {
2038 	struct stat sb;
2039 	struct ostat osb;
2040 	int error;
2041 
2042 	error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE,
2043 	    &sb, NULL);
2044 	if (error != 0)
2045 		return (error);
2046 	cvtstat(&sb, &osb);
2047 	return (copyout(&osb, uap->ub, sizeof (osb)));
2048 }
2049 
2050 /*
2051  * Get file status; this version does not follow links.
2052  */
2053 #ifndef _SYS_SYSPROTO_H_
2054 struct olstat_args {
2055 	char	*path;
2056 	struct ostat *ub;
2057 };
2058 #endif
2059 int
2060 olstat(struct thread *td, struct olstat_args *uap)
2061 {
2062 	struct stat sb;
2063 	struct ostat osb;
2064 	int error;
2065 
2066 	error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2067 	    UIO_USERSPACE, &sb, NULL);
2068 	if (error != 0)
2069 		return (error);
2070 	cvtstat(&sb, &osb);
2071 	return (copyout(&osb, uap->ub, sizeof (osb)));
2072 }
2073 
2074 /*
2075  * Convert from an old to a new stat structure.
2076  */
2077 void
2078 cvtstat(struct stat *st, struct ostat *ost)
2079 {
2080 
2081 	bzero(ost, sizeof(*ost));
2082 	ost->st_dev = st->st_dev;
2083 	ost->st_ino = st->st_ino;
2084 	ost->st_mode = st->st_mode;
2085 	ost->st_nlink = st->st_nlink;
2086 	ost->st_uid = st->st_uid;
2087 	ost->st_gid = st->st_gid;
2088 	ost->st_rdev = st->st_rdev;
2089 	if (st->st_size < (quad_t)1 << 32)
2090 		ost->st_size = st->st_size;
2091 	else
2092 		ost->st_size = -2;
2093 	ost->st_atim = st->st_atim;
2094 	ost->st_mtim = st->st_mtim;
2095 	ost->st_ctim = st->st_ctim;
2096 	ost->st_blksize = st->st_blksize;
2097 	ost->st_blocks = st->st_blocks;
2098 	ost->st_flags = st->st_flags;
2099 	ost->st_gen = st->st_gen;
2100 }
2101 #endif /* COMPAT_43 */
2102 
2103 #if defined(COMPAT_43) || defined(COMPAT_FREEBSD11)
2104 int ino64_trunc_error;
2105 SYSCTL_INT(_vfs, OID_AUTO, ino64_trunc_error, CTLFLAG_RW,
2106     &ino64_trunc_error, 0,
2107     "Error on truncation of inode number, device id or link count");
2108 int
2109 freebsd11_cvtstat(struct stat *st, struct freebsd11_stat *ost)
2110 {
2111 
2112 	ost->st_dev = st->st_dev;
2113 	ost->st_ino = st->st_ino;
2114 	if (ost->st_ino != st->st_ino) {
2115 		switch (ino64_trunc_error) {
2116 		default:
2117 		case 0:
2118 			break;
2119 		case 1:
2120 			return (EOVERFLOW);
2121 		case 2:
2122 			ost->st_ino = UINT32_MAX;
2123 			break;
2124 		}
2125 	}
2126 	ost->st_mode = st->st_mode;
2127 	ost->st_nlink = st->st_nlink;
2128 	if (ost->st_nlink != st->st_nlink) {
2129 		switch (ino64_trunc_error) {
2130 		default:
2131 		case 0:
2132 			break;
2133 		case 1:
2134 			return (EOVERFLOW);
2135 		case 2:
2136 			ost->st_nlink = UINT16_MAX;
2137 			break;
2138 		}
2139 	}
2140 	ost->st_uid = st->st_uid;
2141 	ost->st_gid = st->st_gid;
2142 	ost->st_rdev = st->st_rdev;
2143 	ost->st_atim = st->st_atim;
2144 	ost->st_mtim = st->st_mtim;
2145 	ost->st_ctim = st->st_ctim;
2146 	ost->st_size = st->st_size;
2147 	ost->st_blocks = st->st_blocks;
2148 	ost->st_blksize = st->st_blksize;
2149 	ost->st_flags = st->st_flags;
2150 	ost->st_gen = st->st_gen;
2151 	ost->st_lspare = 0;
2152 	ost->st_birthtim = st->st_birthtim;
2153 	bzero((char *)&ost->st_birthtim + sizeof(ost->st_birthtim),
2154 	    sizeof(*ost) - offsetof(struct freebsd11_stat,
2155 	    st_birthtim) - sizeof(ost->st_birthtim));
2156 	return (0);
2157 }
2158 
2159 int
2160 freebsd11_stat(struct thread *td, struct freebsd11_stat_args* uap)
2161 {
2162 	struct stat sb;
2163 	struct freebsd11_stat osb;
2164 	int error;
2165 
2166 	error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE,
2167 	    &sb, NULL);
2168 	if (error != 0)
2169 		return (error);
2170 	error = freebsd11_cvtstat(&sb, &osb);
2171 	if (error == 0)
2172 		error = copyout(&osb, uap->ub, sizeof(osb));
2173 	return (error);
2174 }
2175 
2176 int
2177 freebsd11_lstat(struct thread *td, struct freebsd11_lstat_args* uap)
2178 {
2179 	struct stat sb;
2180 	struct freebsd11_stat osb;
2181 	int error;
2182 
2183 	error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2184 	    UIO_USERSPACE, &sb, NULL);
2185 	if (error != 0)
2186 		return (error);
2187 	error = freebsd11_cvtstat(&sb, &osb);
2188 	if (error == 0)
2189 		error = copyout(&osb, uap->ub, sizeof(osb));
2190 	return (error);
2191 }
2192 
2193 int
2194 freebsd11_fhstat(struct thread *td, struct freebsd11_fhstat_args* uap)
2195 {
2196 	struct fhandle fh;
2197 	struct stat sb;
2198 	struct freebsd11_stat osb;
2199 	int error;
2200 
2201 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
2202 	if (error != 0)
2203 		return (error);
2204 	error = kern_fhstat(td, fh, &sb);
2205 	if (error != 0)
2206 		return (error);
2207 	error = freebsd11_cvtstat(&sb, &osb);
2208 	if (error == 0)
2209 		error = copyout(&osb, uap->sb, sizeof(osb));
2210 	return (error);
2211 }
2212 
2213 int
2214 freebsd11_fstatat(struct thread *td, struct freebsd11_fstatat_args* uap)
2215 {
2216 	struct stat sb;
2217 	struct freebsd11_stat osb;
2218 	int error;
2219 
2220 	error = kern_statat(td, uap->flag, uap->fd, uap->path,
2221 	    UIO_USERSPACE, &sb, NULL);
2222 	if (error != 0)
2223 		return (error);
2224 	error = freebsd11_cvtstat(&sb, &osb);
2225 	if (error == 0)
2226 		error = copyout(&osb, uap->buf, sizeof(osb));
2227 	return (error);
2228 }
2229 #endif	/* COMPAT_FREEBSD11 */
2230 
2231 /*
2232  * Get file status
2233  */
2234 #ifndef _SYS_SYSPROTO_H_
2235 struct fstatat_args {
2236 	int	fd;
2237 	char	*path;
2238 	struct stat	*buf;
2239 	int	flag;
2240 }
2241 #endif
2242 int
2243 sys_fstatat(struct thread *td, struct fstatat_args *uap)
2244 {
2245 	struct stat sb;
2246 	int error;
2247 
2248 	error = kern_statat(td, uap->flag, uap->fd, uap->path,
2249 	    UIO_USERSPACE, &sb, NULL);
2250 	if (error == 0)
2251 		error = copyout(&sb, uap->buf, sizeof (sb));
2252 	return (error);
2253 }
2254 
2255 int
2256 kern_statat(struct thread *td, int flag, int fd, char *path,
2257     enum uio_seg pathseg, struct stat *sbp,
2258     void (*hook)(struct vnode *vp, struct stat *sbp))
2259 {
2260 	struct nameidata nd;
2261 	struct stat sb;
2262 	int error;
2263 
2264 	if (flag & ~AT_SYMLINK_NOFOLLOW)
2265 		return (EINVAL);
2266 
2267 	NDINIT_ATRIGHTS(&nd, LOOKUP, ((flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW :
2268 	    FOLLOW) | LOCKSHARED | LOCKLEAF | AUDITVNODE1, pathseg, path, fd,
2269 	    &cap_fstat_rights, td);
2270 
2271 	if ((error = namei(&nd)) != 0)
2272 		return (error);
2273 	error = vn_stat(nd.ni_vp, &sb, td->td_ucred, NOCRED, td);
2274 	if (error == 0) {
2275 		SDT_PROBE2(vfs, , stat, mode, path, sb.st_mode);
2276 		if (S_ISREG(sb.st_mode))
2277 			SDT_PROBE2(vfs, , stat, reg, path, pathseg);
2278 		if (__predict_false(hook != NULL))
2279 			hook(nd.ni_vp, &sb);
2280 	}
2281 	NDFREE(&nd, NDF_ONLY_PNBUF);
2282 	vput(nd.ni_vp);
2283 	if (error != 0)
2284 		return (error);
2285 #ifdef __STAT_TIME_T_EXT
2286 	sb.st_atim_ext = 0;
2287 	sb.st_mtim_ext = 0;
2288 	sb.st_ctim_ext = 0;
2289 	sb.st_btim_ext = 0;
2290 #endif
2291 	*sbp = sb;
2292 #ifdef KTRACE
2293 	if (KTRPOINT(td, KTR_STRUCT))
2294 		ktrstat(&sb);
2295 #endif
2296 	return (0);
2297 }
2298 
2299 #if defined(COMPAT_FREEBSD11)
2300 /*
2301  * Implementation of the NetBSD [l]stat() functions.
2302  */
2303 void
2304 freebsd11_cvtnstat(struct stat *sb, struct nstat *nsb)
2305 {
2306 
2307 	bzero(nsb, sizeof(*nsb));
2308 	nsb->st_dev = sb->st_dev;
2309 	nsb->st_ino = sb->st_ino;
2310 	nsb->st_mode = sb->st_mode;
2311 	nsb->st_nlink = sb->st_nlink;
2312 	nsb->st_uid = sb->st_uid;
2313 	nsb->st_gid = sb->st_gid;
2314 	nsb->st_rdev = sb->st_rdev;
2315 	nsb->st_atim = sb->st_atim;
2316 	nsb->st_mtim = sb->st_mtim;
2317 	nsb->st_ctim = sb->st_ctim;
2318 	nsb->st_size = sb->st_size;
2319 	nsb->st_blocks = sb->st_blocks;
2320 	nsb->st_blksize = sb->st_blksize;
2321 	nsb->st_flags = sb->st_flags;
2322 	nsb->st_gen = sb->st_gen;
2323 	nsb->st_birthtim = sb->st_birthtim;
2324 }
2325 
2326 #ifndef _SYS_SYSPROTO_H_
2327 struct freebsd11_nstat_args {
2328 	char	*path;
2329 	struct nstat *ub;
2330 };
2331 #endif
2332 int
2333 freebsd11_nstat(struct thread *td, struct freebsd11_nstat_args *uap)
2334 {
2335 	struct stat sb;
2336 	struct nstat nsb;
2337 	int error;
2338 
2339 	error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE,
2340 	    &sb, NULL);
2341 	if (error != 0)
2342 		return (error);
2343 	freebsd11_cvtnstat(&sb, &nsb);
2344 	return (copyout(&nsb, uap->ub, sizeof (nsb)));
2345 }
2346 
2347 /*
2348  * NetBSD lstat.  Get file status; this version does not follow links.
2349  */
2350 #ifndef _SYS_SYSPROTO_H_
2351 struct freebsd11_nlstat_args {
2352 	char	*path;
2353 	struct nstat *ub;
2354 };
2355 #endif
2356 int
2357 freebsd11_nlstat(struct thread *td, struct freebsd11_nlstat_args *uap)
2358 {
2359 	struct stat sb;
2360 	struct nstat nsb;
2361 	int error;
2362 
2363 	error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2364 	    UIO_USERSPACE, &sb, NULL);
2365 	if (error != 0)
2366 		return (error);
2367 	freebsd11_cvtnstat(&sb, &nsb);
2368 	return (copyout(&nsb, uap->ub, sizeof (nsb)));
2369 }
2370 #endif /* COMPAT_FREEBSD11 */
2371 
2372 /*
2373  * Get configurable pathname variables.
2374  */
2375 #ifndef _SYS_SYSPROTO_H_
2376 struct pathconf_args {
2377 	char	*path;
2378 	int	name;
2379 };
2380 #endif
2381 int
2382 sys_pathconf(struct thread *td, struct pathconf_args *uap)
2383 {
2384 	long value;
2385 	int error;
2386 
2387 	error = kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name, FOLLOW,
2388 	    &value);
2389 	if (error == 0)
2390 		td->td_retval[0] = value;
2391 	return (error);
2392 }
2393 
2394 #ifndef _SYS_SYSPROTO_H_
2395 struct lpathconf_args {
2396 	char	*path;
2397 	int	name;
2398 };
2399 #endif
2400 int
2401 sys_lpathconf(struct thread *td, struct lpathconf_args *uap)
2402 {
2403 	long value;
2404 	int error;
2405 
2406 	error = kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name,
2407 	    NOFOLLOW, &value);
2408 	if (error == 0)
2409 		td->td_retval[0] = value;
2410 	return (error);
2411 }
2412 
2413 int
2414 kern_pathconf(struct thread *td, char *path, enum uio_seg pathseg, int name,
2415     u_long flags, long *valuep)
2416 {
2417 	struct nameidata nd;
2418 	int error;
2419 
2420 	NDINIT(&nd, LOOKUP, LOCKSHARED | LOCKLEAF | AUDITVNODE1 | flags,
2421 	    pathseg, path, td);
2422 	if ((error = namei(&nd)) != 0)
2423 		return (error);
2424 	NDFREE(&nd, NDF_ONLY_PNBUF);
2425 
2426 	error = VOP_PATHCONF(nd.ni_vp, name, valuep);
2427 	vput(nd.ni_vp);
2428 	return (error);
2429 }
2430 
2431 /*
2432  * Return target name of a symbolic link.
2433  */
2434 #ifndef _SYS_SYSPROTO_H_
2435 struct readlink_args {
2436 	char	*path;
2437 	char	*buf;
2438 	size_t	count;
2439 };
2440 #endif
2441 int
2442 sys_readlink(struct thread *td, struct readlink_args *uap)
2443 {
2444 
2445 	return (kern_readlinkat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2446 	    uap->buf, UIO_USERSPACE, uap->count));
2447 }
2448 #ifndef _SYS_SYSPROTO_H_
2449 struct readlinkat_args {
2450 	int	fd;
2451 	char	*path;
2452 	char	*buf;
2453 	size_t	bufsize;
2454 };
2455 #endif
2456 int
2457 sys_readlinkat(struct thread *td, struct readlinkat_args *uap)
2458 {
2459 
2460 	return (kern_readlinkat(td, uap->fd, uap->path, UIO_USERSPACE,
2461 	    uap->buf, UIO_USERSPACE, uap->bufsize));
2462 }
2463 
2464 int
2465 kern_readlinkat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2466     char *buf, enum uio_seg bufseg, size_t count)
2467 {
2468 	struct vnode *vp;
2469 	struct iovec aiov;
2470 	struct uio auio;
2471 	struct nameidata nd;
2472 	int error;
2473 
2474 	if (count > IOSIZE_MAX)
2475 		return (EINVAL);
2476 
2477 	NDINIT_AT(&nd, LOOKUP, NOFOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
2478 	    pathseg, path, fd, td);
2479 
2480 	if ((error = namei(&nd)) != 0)
2481 		return (error);
2482 	NDFREE(&nd, NDF_ONLY_PNBUF);
2483 	vp = nd.ni_vp;
2484 #ifdef MAC
2485 	error = mac_vnode_check_readlink(td->td_ucred, vp);
2486 	if (error != 0) {
2487 		vput(vp);
2488 		return (error);
2489 	}
2490 #endif
2491 	if (vp->v_type != VLNK && (vp->v_vflag & VV_READLINK) == 0)
2492 		error = EINVAL;
2493 	else {
2494 		aiov.iov_base = buf;
2495 		aiov.iov_len = count;
2496 		auio.uio_iov = &aiov;
2497 		auio.uio_iovcnt = 1;
2498 		auio.uio_offset = 0;
2499 		auio.uio_rw = UIO_READ;
2500 		auio.uio_segflg = bufseg;
2501 		auio.uio_td = td;
2502 		auio.uio_resid = count;
2503 		error = VOP_READLINK(vp, &auio, td->td_ucred);
2504 		td->td_retval[0] = count - auio.uio_resid;
2505 	}
2506 	vput(vp);
2507 	return (error);
2508 }
2509 
2510 /*
2511  * Common implementation code for chflags() and fchflags().
2512  */
2513 static int
2514 setfflags(struct thread *td, struct vnode *vp, u_long flags)
2515 {
2516 	struct mount *mp;
2517 	struct vattr vattr;
2518 	int error;
2519 
2520 	/* We can't support the value matching VNOVAL. */
2521 	if (flags == VNOVAL)
2522 		return (EOPNOTSUPP);
2523 
2524 	/*
2525 	 * Prevent non-root users from setting flags on devices.  When
2526 	 * a device is reused, users can retain ownership of the device
2527 	 * if they are allowed to set flags and programs assume that
2528 	 * chown can't fail when done as root.
2529 	 */
2530 	if (vp->v_type == VCHR || vp->v_type == VBLK) {
2531 		error = priv_check(td, PRIV_VFS_CHFLAGS_DEV);
2532 		if (error != 0)
2533 			return (error);
2534 	}
2535 
2536 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2537 		return (error);
2538 	VATTR_NULL(&vattr);
2539 	vattr.va_flags = flags;
2540 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2541 #ifdef MAC
2542 	error = mac_vnode_check_setflags(td->td_ucred, vp, vattr.va_flags);
2543 	if (error == 0)
2544 #endif
2545 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
2546 	VOP_UNLOCK(vp, 0);
2547 	vn_finished_write(mp);
2548 	return (error);
2549 }
2550 
2551 /*
2552  * Change flags of a file given a path name.
2553  */
2554 #ifndef _SYS_SYSPROTO_H_
2555 struct chflags_args {
2556 	const char *path;
2557 	u_long	flags;
2558 };
2559 #endif
2560 int
2561 sys_chflags(struct thread *td, struct chflags_args *uap)
2562 {
2563 
2564 	return (kern_chflagsat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2565 	    uap->flags, 0));
2566 }
2567 
2568 #ifndef _SYS_SYSPROTO_H_
2569 struct chflagsat_args {
2570 	int	fd;
2571 	const char *path;
2572 	u_long	flags;
2573 	int	atflag;
2574 }
2575 #endif
2576 int
2577 sys_chflagsat(struct thread *td, struct chflagsat_args *uap)
2578 {
2579 	int fd = uap->fd;
2580 	const char *path = uap->path;
2581 	u_long flags = uap->flags;
2582 	int atflag = uap->atflag;
2583 
2584 	if (atflag & ~AT_SYMLINK_NOFOLLOW)
2585 		return (EINVAL);
2586 
2587 	return (kern_chflagsat(td, fd, path, UIO_USERSPACE, flags, atflag));
2588 }
2589 
2590 /*
2591  * Same as chflags() but doesn't follow symlinks.
2592  */
2593 #ifndef _SYS_SYSPROTO_H_
2594 struct lchflags_args {
2595 	const char *path;
2596 	u_long flags;
2597 };
2598 #endif
2599 int
2600 sys_lchflags(struct thread *td, struct lchflags_args *uap)
2601 {
2602 
2603 	return (kern_chflagsat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2604 	    uap->flags, AT_SYMLINK_NOFOLLOW));
2605 }
2606 
2607 static int
2608 kern_chflagsat(struct thread *td, int fd, const char *path,
2609     enum uio_seg pathseg, u_long flags, int atflag)
2610 {
2611 	struct nameidata nd;
2612 	cap_rights_t rights;
2613 	int error, follow;
2614 
2615 	AUDIT_ARG_FFLAGS(flags);
2616 	follow = (atflag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
2617 	NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, pathseg, path, fd,
2618 	    cap_rights_init(&rights, CAP_FCHFLAGS), td);
2619 	if ((error = namei(&nd)) != 0)
2620 		return (error);
2621 	NDFREE(&nd, NDF_ONLY_PNBUF);
2622 	error = setfflags(td, nd.ni_vp, flags);
2623 	vrele(nd.ni_vp);
2624 	return (error);
2625 }
2626 
2627 /*
2628  * Change flags of a file given a file descriptor.
2629  */
2630 #ifndef _SYS_SYSPROTO_H_
2631 struct fchflags_args {
2632 	int	fd;
2633 	u_long	flags;
2634 };
2635 #endif
2636 int
2637 sys_fchflags(struct thread *td, struct fchflags_args *uap)
2638 {
2639 	struct file *fp;
2640 	cap_rights_t rights;
2641 	int error;
2642 
2643 	AUDIT_ARG_FD(uap->fd);
2644 	AUDIT_ARG_FFLAGS(uap->flags);
2645 	error = getvnode(td, uap->fd, cap_rights_init(&rights, CAP_FCHFLAGS),
2646 	    &fp);
2647 	if (error != 0)
2648 		return (error);
2649 #ifdef AUDIT
2650 	vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
2651 	AUDIT_ARG_VNODE1(fp->f_vnode);
2652 	VOP_UNLOCK(fp->f_vnode, 0);
2653 #endif
2654 	error = setfflags(td, fp->f_vnode, uap->flags);
2655 	fdrop(fp, td);
2656 	return (error);
2657 }
2658 
2659 /*
2660  * Common implementation code for chmod(), lchmod() and fchmod().
2661  */
2662 int
2663 setfmode(struct thread *td, struct ucred *cred, struct vnode *vp, int mode)
2664 {
2665 	struct mount *mp;
2666 	struct vattr vattr;
2667 	int error;
2668 
2669 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2670 		return (error);
2671 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2672 	VATTR_NULL(&vattr);
2673 	vattr.va_mode = mode & ALLPERMS;
2674 #ifdef MAC
2675 	error = mac_vnode_check_setmode(cred, vp, vattr.va_mode);
2676 	if (error == 0)
2677 #endif
2678 		error = VOP_SETATTR(vp, &vattr, cred);
2679 	VOP_UNLOCK(vp, 0);
2680 	vn_finished_write(mp);
2681 	return (error);
2682 }
2683 
2684 /*
2685  * Change mode of a file given path name.
2686  */
2687 #ifndef _SYS_SYSPROTO_H_
2688 struct chmod_args {
2689 	char	*path;
2690 	int	mode;
2691 };
2692 #endif
2693 int
2694 sys_chmod(struct thread *td, struct chmod_args *uap)
2695 {
2696 
2697 	return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2698 	    uap->mode, 0));
2699 }
2700 
2701 #ifndef _SYS_SYSPROTO_H_
2702 struct fchmodat_args {
2703 	int	dirfd;
2704 	char	*path;
2705 	mode_t	mode;
2706 	int	flag;
2707 }
2708 #endif
2709 int
2710 sys_fchmodat(struct thread *td, struct fchmodat_args *uap)
2711 {
2712 	int flag = uap->flag;
2713 	int fd = uap->fd;
2714 	char *path = uap->path;
2715 	mode_t mode = uap->mode;
2716 
2717 	if (flag & ~AT_SYMLINK_NOFOLLOW)
2718 		return (EINVAL);
2719 
2720 	return (kern_fchmodat(td, fd, path, UIO_USERSPACE, mode, flag));
2721 }
2722 
2723 /*
2724  * Change mode of a file given path name (don't follow links.)
2725  */
2726 #ifndef _SYS_SYSPROTO_H_
2727 struct lchmod_args {
2728 	char	*path;
2729 	int	mode;
2730 };
2731 #endif
2732 int
2733 sys_lchmod(struct thread *td, struct lchmod_args *uap)
2734 {
2735 
2736 	return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2737 	    uap->mode, AT_SYMLINK_NOFOLLOW));
2738 }
2739 
2740 int
2741 kern_fchmodat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2742     mode_t mode, int flag)
2743 {
2744 	struct nameidata nd;
2745 	cap_rights_t rights;
2746 	int error, follow;
2747 
2748 	AUDIT_ARG_MODE(mode);
2749 	follow = (flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
2750 	NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, pathseg, path, fd,
2751 	    cap_rights_init(&rights, CAP_FCHMOD), td);
2752 	if ((error = namei(&nd)) != 0)
2753 		return (error);
2754 	NDFREE(&nd, NDF_ONLY_PNBUF);
2755 	error = setfmode(td, td->td_ucred, nd.ni_vp, mode);
2756 	vrele(nd.ni_vp);
2757 	return (error);
2758 }
2759 
2760 /*
2761  * Change mode of a file given a file descriptor.
2762  */
2763 #ifndef _SYS_SYSPROTO_H_
2764 struct fchmod_args {
2765 	int	fd;
2766 	int	mode;
2767 };
2768 #endif
2769 int
2770 sys_fchmod(struct thread *td, struct fchmod_args *uap)
2771 {
2772 	struct file *fp;
2773 	cap_rights_t rights;
2774 	int error;
2775 
2776 	AUDIT_ARG_FD(uap->fd);
2777 	AUDIT_ARG_MODE(uap->mode);
2778 
2779 	error = fget(td, uap->fd, cap_rights_init(&rights, CAP_FCHMOD), &fp);
2780 	if (error != 0)
2781 		return (error);
2782 	error = fo_chmod(fp, uap->mode, td->td_ucred, td);
2783 	fdrop(fp, td);
2784 	return (error);
2785 }
2786 
2787 /*
2788  * Common implementation for chown(), lchown(), and fchown()
2789  */
2790 int
2791 setfown(struct thread *td, struct ucred *cred, struct vnode *vp, uid_t uid,
2792     gid_t gid)
2793 {
2794 	struct mount *mp;
2795 	struct vattr vattr;
2796 	int error;
2797 
2798 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2799 		return (error);
2800 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2801 	VATTR_NULL(&vattr);
2802 	vattr.va_uid = uid;
2803 	vattr.va_gid = gid;
2804 #ifdef MAC
2805 	error = mac_vnode_check_setowner(cred, vp, vattr.va_uid,
2806 	    vattr.va_gid);
2807 	if (error == 0)
2808 #endif
2809 		error = VOP_SETATTR(vp, &vattr, cred);
2810 	VOP_UNLOCK(vp, 0);
2811 	vn_finished_write(mp);
2812 	return (error);
2813 }
2814 
2815 /*
2816  * Set ownership given a path name.
2817  */
2818 #ifndef _SYS_SYSPROTO_H_
2819 struct chown_args {
2820 	char	*path;
2821 	int	uid;
2822 	int	gid;
2823 };
2824 #endif
2825 int
2826 sys_chown(struct thread *td, struct chown_args *uap)
2827 {
2828 
2829 	return (kern_fchownat(td, AT_FDCWD, uap->path, UIO_USERSPACE, uap->uid,
2830 	    uap->gid, 0));
2831 }
2832 
2833 #ifndef _SYS_SYSPROTO_H_
2834 struct fchownat_args {
2835 	int fd;
2836 	const char * path;
2837 	uid_t uid;
2838 	gid_t gid;
2839 	int flag;
2840 };
2841 #endif
2842 int
2843 sys_fchownat(struct thread *td, struct fchownat_args *uap)
2844 {
2845 	int flag;
2846 
2847 	flag = uap->flag;
2848 	if (flag & ~AT_SYMLINK_NOFOLLOW)
2849 		return (EINVAL);
2850 
2851 	return (kern_fchownat(td, uap->fd, uap->path, UIO_USERSPACE, uap->uid,
2852 	    uap->gid, uap->flag));
2853 }
2854 
2855 int
2856 kern_fchownat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2857     int uid, int gid, int flag)
2858 {
2859 	struct nameidata nd;
2860 	cap_rights_t rights;
2861 	int error, follow;
2862 
2863 	AUDIT_ARG_OWNER(uid, gid);
2864 	follow = (flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
2865 	NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, pathseg, path, fd,
2866 	    cap_rights_init(&rights, CAP_FCHOWN), td);
2867 
2868 	if ((error = namei(&nd)) != 0)
2869 		return (error);
2870 	NDFREE(&nd, NDF_ONLY_PNBUF);
2871 	error = setfown(td, td->td_ucred, nd.ni_vp, uid, gid);
2872 	vrele(nd.ni_vp);
2873 	return (error);
2874 }
2875 
2876 /*
2877  * Set ownership given a path name, do not cross symlinks.
2878  */
2879 #ifndef _SYS_SYSPROTO_H_
2880 struct lchown_args {
2881 	char	*path;
2882 	int	uid;
2883 	int	gid;
2884 };
2885 #endif
2886 int
2887 sys_lchown(struct thread *td, struct lchown_args *uap)
2888 {
2889 
2890 	return (kern_fchownat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2891 	    uap->uid, uap->gid, AT_SYMLINK_NOFOLLOW));
2892 }
2893 
2894 /*
2895  * Set ownership given a file descriptor.
2896  */
2897 #ifndef _SYS_SYSPROTO_H_
2898 struct fchown_args {
2899 	int	fd;
2900 	int	uid;
2901 	int	gid;
2902 };
2903 #endif
2904 int
2905 sys_fchown(struct thread *td, struct fchown_args *uap)
2906 {
2907 	struct file *fp;
2908 	cap_rights_t rights;
2909 	int error;
2910 
2911 	AUDIT_ARG_FD(uap->fd);
2912 	AUDIT_ARG_OWNER(uap->uid, uap->gid);
2913 	error = fget(td, uap->fd, cap_rights_init(&rights, CAP_FCHOWN), &fp);
2914 	if (error != 0)
2915 		return (error);
2916 	error = fo_chown(fp, uap->uid, uap->gid, td->td_ucred, td);
2917 	fdrop(fp, td);
2918 	return (error);
2919 }
2920 
2921 /*
2922  * Common implementation code for utimes(), lutimes(), and futimes().
2923  */
2924 static int
2925 getutimes(const struct timeval *usrtvp, enum uio_seg tvpseg,
2926     struct timespec *tsp)
2927 {
2928 	struct timeval tv[2];
2929 	const struct timeval *tvp;
2930 	int error;
2931 
2932 	if (usrtvp == NULL) {
2933 		vfs_timestamp(&tsp[0]);
2934 		tsp[1] = tsp[0];
2935 	} else {
2936 		if (tvpseg == UIO_SYSSPACE) {
2937 			tvp = usrtvp;
2938 		} else {
2939 			if ((error = copyin(usrtvp, tv, sizeof(tv))) != 0)
2940 				return (error);
2941 			tvp = tv;
2942 		}
2943 
2944 		if (tvp[0].tv_usec < 0 || tvp[0].tv_usec >= 1000000 ||
2945 		    tvp[1].tv_usec < 0 || tvp[1].tv_usec >= 1000000)
2946 			return (EINVAL);
2947 		TIMEVAL_TO_TIMESPEC(&tvp[0], &tsp[0]);
2948 		TIMEVAL_TO_TIMESPEC(&tvp[1], &tsp[1]);
2949 	}
2950 	return (0);
2951 }
2952 
2953 /*
2954  * Common implementation code for futimens(), utimensat().
2955  */
2956 #define	UTIMENS_NULL	0x1
2957 #define	UTIMENS_EXIT	0x2
2958 static int
2959 getutimens(const struct timespec *usrtsp, enum uio_seg tspseg,
2960     struct timespec *tsp, int *retflags)
2961 {
2962 	struct timespec tsnow;
2963 	int error;
2964 
2965 	vfs_timestamp(&tsnow);
2966 	*retflags = 0;
2967 	if (usrtsp == NULL) {
2968 		tsp[0] = tsnow;
2969 		tsp[1] = tsnow;
2970 		*retflags |= UTIMENS_NULL;
2971 		return (0);
2972 	}
2973 	if (tspseg == UIO_SYSSPACE) {
2974 		tsp[0] = usrtsp[0];
2975 		tsp[1] = usrtsp[1];
2976 	} else if ((error = copyin(usrtsp, tsp, sizeof(*tsp) * 2)) != 0)
2977 		return (error);
2978 	if (tsp[0].tv_nsec == UTIME_OMIT && tsp[1].tv_nsec == UTIME_OMIT)
2979 		*retflags |= UTIMENS_EXIT;
2980 	if (tsp[0].tv_nsec == UTIME_NOW && tsp[1].tv_nsec == UTIME_NOW)
2981 		*retflags |= UTIMENS_NULL;
2982 	if (tsp[0].tv_nsec == UTIME_OMIT)
2983 		tsp[0].tv_sec = VNOVAL;
2984 	else if (tsp[0].tv_nsec == UTIME_NOW)
2985 		tsp[0] = tsnow;
2986 	else if (tsp[0].tv_nsec < 0 || tsp[0].tv_nsec >= 1000000000L)
2987 		return (EINVAL);
2988 	if (tsp[1].tv_nsec == UTIME_OMIT)
2989 		tsp[1].tv_sec = VNOVAL;
2990 	else if (tsp[1].tv_nsec == UTIME_NOW)
2991 		tsp[1] = tsnow;
2992 	else if (tsp[1].tv_nsec < 0 || tsp[1].tv_nsec >= 1000000000L)
2993 		return (EINVAL);
2994 
2995 	return (0);
2996 }
2997 
2998 /*
2999  * Common implementation code for utimes(), lutimes(), futimes(), futimens(),
3000  * and utimensat().
3001  */
3002 static int
3003 setutimes(struct thread *td, struct vnode *vp, const struct timespec *ts,
3004     int numtimes, int nullflag)
3005 {
3006 	struct mount *mp;
3007 	struct vattr vattr;
3008 	int error, setbirthtime;
3009 
3010 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
3011 		return (error);
3012 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3013 	setbirthtime = 0;
3014 	if (numtimes < 3 && !VOP_GETATTR(vp, &vattr, td->td_ucred) &&
3015 	    timespeccmp(&ts[1], &vattr.va_birthtime, < ))
3016 		setbirthtime = 1;
3017 	VATTR_NULL(&vattr);
3018 	vattr.va_atime = ts[0];
3019 	vattr.va_mtime = ts[1];
3020 	if (setbirthtime)
3021 		vattr.va_birthtime = ts[1];
3022 	if (numtimes > 2)
3023 		vattr.va_birthtime = ts[2];
3024 	if (nullflag)
3025 		vattr.va_vaflags |= VA_UTIMES_NULL;
3026 #ifdef MAC
3027 	error = mac_vnode_check_setutimes(td->td_ucred, vp, vattr.va_atime,
3028 	    vattr.va_mtime);
3029 #endif
3030 	if (error == 0)
3031 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3032 	VOP_UNLOCK(vp, 0);
3033 	vn_finished_write(mp);
3034 	return (error);
3035 }
3036 
3037 /*
3038  * Set the access and modification times of a file.
3039  */
3040 #ifndef _SYS_SYSPROTO_H_
3041 struct utimes_args {
3042 	char	*path;
3043 	struct	timeval *tptr;
3044 };
3045 #endif
3046 int
3047 sys_utimes(struct thread *td, struct utimes_args *uap)
3048 {
3049 
3050 	return (kern_utimesat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
3051 	    uap->tptr, UIO_USERSPACE));
3052 }
3053 
3054 #ifndef _SYS_SYSPROTO_H_
3055 struct futimesat_args {
3056 	int fd;
3057 	const char * path;
3058 	const struct timeval * times;
3059 };
3060 #endif
3061 int
3062 sys_futimesat(struct thread *td, struct futimesat_args *uap)
3063 {
3064 
3065 	return (kern_utimesat(td, uap->fd, uap->path, UIO_USERSPACE,
3066 	    uap->times, UIO_USERSPACE));
3067 }
3068 
3069 int
3070 kern_utimesat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
3071     struct timeval *tptr, enum uio_seg tptrseg)
3072 {
3073 	struct nameidata nd;
3074 	struct timespec ts[2];
3075 	cap_rights_t rights;
3076 	int error;
3077 
3078 	if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3079 		return (error);
3080 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, fd,
3081 	    cap_rights_init(&rights, CAP_FUTIMES), td);
3082 
3083 	if ((error = namei(&nd)) != 0)
3084 		return (error);
3085 	NDFREE(&nd, NDF_ONLY_PNBUF);
3086 	error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3087 	vrele(nd.ni_vp);
3088 	return (error);
3089 }
3090 
3091 /*
3092  * Set the access and modification times of a file.
3093  */
3094 #ifndef _SYS_SYSPROTO_H_
3095 struct lutimes_args {
3096 	char	*path;
3097 	struct	timeval *tptr;
3098 };
3099 #endif
3100 int
3101 sys_lutimes(struct thread *td, struct lutimes_args *uap)
3102 {
3103 
3104 	return (kern_lutimes(td, uap->path, UIO_USERSPACE, uap->tptr,
3105 	    UIO_USERSPACE));
3106 }
3107 
3108 int
3109 kern_lutimes(struct thread *td, char *path, enum uio_seg pathseg,
3110     struct timeval *tptr, enum uio_seg tptrseg)
3111 {
3112 	struct timespec ts[2];
3113 	struct nameidata nd;
3114 	int error;
3115 
3116 	if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3117 		return (error);
3118 	NDINIT(&nd, LOOKUP, NOFOLLOW | AUDITVNODE1, pathseg, path, td);
3119 	if ((error = namei(&nd)) != 0)
3120 		return (error);
3121 	NDFREE(&nd, NDF_ONLY_PNBUF);
3122 	error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3123 	vrele(nd.ni_vp);
3124 	return (error);
3125 }
3126 
3127 /*
3128  * Set the access and modification times of a file.
3129  */
3130 #ifndef _SYS_SYSPROTO_H_
3131 struct futimes_args {
3132 	int	fd;
3133 	struct	timeval *tptr;
3134 };
3135 #endif
3136 int
3137 sys_futimes(struct thread *td, struct futimes_args *uap)
3138 {
3139 
3140 	return (kern_futimes(td, uap->fd, uap->tptr, UIO_USERSPACE));
3141 }
3142 
3143 int
3144 kern_futimes(struct thread *td, int fd, struct timeval *tptr,
3145     enum uio_seg tptrseg)
3146 {
3147 	struct timespec ts[2];
3148 	struct file *fp;
3149 	cap_rights_t rights;
3150 	int error;
3151 
3152 	AUDIT_ARG_FD(fd);
3153 	error = getutimes(tptr, tptrseg, ts);
3154 	if (error != 0)
3155 		return (error);
3156 	error = getvnode(td, fd, cap_rights_init(&rights, CAP_FUTIMES), &fp);
3157 	if (error != 0)
3158 		return (error);
3159 #ifdef AUDIT
3160 	vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
3161 	AUDIT_ARG_VNODE1(fp->f_vnode);
3162 	VOP_UNLOCK(fp->f_vnode, 0);
3163 #endif
3164 	error = setutimes(td, fp->f_vnode, ts, 2, tptr == NULL);
3165 	fdrop(fp, td);
3166 	return (error);
3167 }
3168 
3169 int
3170 sys_futimens(struct thread *td, struct futimens_args *uap)
3171 {
3172 
3173 	return (kern_futimens(td, uap->fd, uap->times, UIO_USERSPACE));
3174 }
3175 
3176 int
3177 kern_futimens(struct thread *td, int fd, struct timespec *tptr,
3178     enum uio_seg tptrseg)
3179 {
3180 	struct timespec ts[2];
3181 	struct file *fp;
3182 	cap_rights_t rights;
3183 	int error, flags;
3184 
3185 	AUDIT_ARG_FD(fd);
3186 	error = getutimens(tptr, tptrseg, ts, &flags);
3187 	if (error != 0)
3188 		return (error);
3189 	if (flags & UTIMENS_EXIT)
3190 		return (0);
3191 	error = getvnode(td, fd, cap_rights_init(&rights, CAP_FUTIMES), &fp);
3192 	if (error != 0)
3193 		return (error);
3194 #ifdef AUDIT
3195 	vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
3196 	AUDIT_ARG_VNODE1(fp->f_vnode);
3197 	VOP_UNLOCK(fp->f_vnode, 0);
3198 #endif
3199 	error = setutimes(td, fp->f_vnode, ts, 2, flags & UTIMENS_NULL);
3200 	fdrop(fp, td);
3201 	return (error);
3202 }
3203 
3204 int
3205 sys_utimensat(struct thread *td, struct utimensat_args *uap)
3206 {
3207 
3208 	return (kern_utimensat(td, uap->fd, uap->path, UIO_USERSPACE,
3209 	    uap->times, UIO_USERSPACE, uap->flag));
3210 }
3211 
3212 int
3213 kern_utimensat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
3214     struct timespec *tptr, enum uio_seg tptrseg, int flag)
3215 {
3216 	struct nameidata nd;
3217 	struct timespec ts[2];
3218 	cap_rights_t rights;
3219 	int error, flags;
3220 
3221 	if (flag & ~AT_SYMLINK_NOFOLLOW)
3222 		return (EINVAL);
3223 
3224 	if ((error = getutimens(tptr, tptrseg, ts, &flags)) != 0)
3225 		return (error);
3226 	NDINIT_ATRIGHTS(&nd, LOOKUP, ((flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW :
3227 	    FOLLOW) | AUDITVNODE1, pathseg, path, fd,
3228 	    cap_rights_init(&rights, CAP_FUTIMES), td);
3229 	if ((error = namei(&nd)) != 0)
3230 		return (error);
3231 	/*
3232 	 * We are allowed to call namei() regardless of 2xUTIME_OMIT.
3233 	 * POSIX states:
3234 	 * "If both tv_nsec fields are UTIME_OMIT... EACCESS may be detected."
3235 	 * "Search permission is denied by a component of the path prefix."
3236 	 */
3237 	NDFREE(&nd, NDF_ONLY_PNBUF);
3238 	if ((flags & UTIMENS_EXIT) == 0)
3239 		error = setutimes(td, nd.ni_vp, ts, 2, flags & UTIMENS_NULL);
3240 	vrele(nd.ni_vp);
3241 	return (error);
3242 }
3243 
3244 /*
3245  * Truncate a file given its path name.
3246  */
3247 #ifndef _SYS_SYSPROTO_H_
3248 struct truncate_args {
3249 	char	*path;
3250 	int	pad;
3251 	off_t	length;
3252 };
3253 #endif
3254 int
3255 sys_truncate(struct thread *td, struct truncate_args *uap)
3256 {
3257 
3258 	return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3259 }
3260 
3261 int
3262 kern_truncate(struct thread *td, char *path, enum uio_seg pathseg, off_t length)
3263 {
3264 	struct mount *mp;
3265 	struct vnode *vp;
3266 	void *rl_cookie;
3267 	struct vattr vattr;
3268 	struct nameidata nd;
3269 	int error;
3270 
3271 	if (length < 0)
3272 		return(EINVAL);
3273 	NDINIT(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, td);
3274 	if ((error = namei(&nd)) != 0)
3275 		return (error);
3276 	vp = nd.ni_vp;
3277 	rl_cookie = vn_rangelock_wlock(vp, 0, OFF_MAX);
3278 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) {
3279 		vn_rangelock_unlock(vp, rl_cookie);
3280 		vrele(vp);
3281 		return (error);
3282 	}
3283 	NDFREE(&nd, NDF_ONLY_PNBUF);
3284 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3285 	if (vp->v_type == VDIR)
3286 		error = EISDIR;
3287 #ifdef MAC
3288 	else if ((error = mac_vnode_check_write(td->td_ucred, NOCRED, vp))) {
3289 	}
3290 #endif
3291 	else if ((error = vn_writechk(vp)) == 0 &&
3292 	    (error = VOP_ACCESS(vp, VWRITE, td->td_ucred, td)) == 0) {
3293 		VATTR_NULL(&vattr);
3294 		vattr.va_size = length;
3295 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3296 	}
3297 	VOP_UNLOCK(vp, 0);
3298 	vn_finished_write(mp);
3299 	vn_rangelock_unlock(vp, rl_cookie);
3300 	vrele(vp);
3301 	return (error);
3302 }
3303 
3304 #if defined(COMPAT_43)
3305 /*
3306  * Truncate a file given its path name.
3307  */
3308 #ifndef _SYS_SYSPROTO_H_
3309 struct otruncate_args {
3310 	char	*path;
3311 	long	length;
3312 };
3313 #endif
3314 int
3315 otruncate(struct thread *td, struct otruncate_args *uap)
3316 {
3317 
3318 	return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3319 }
3320 #endif /* COMPAT_43 */
3321 
3322 #if defined(COMPAT_FREEBSD6)
3323 /* Versions with the pad argument */
3324 int
3325 freebsd6_truncate(struct thread *td, struct freebsd6_truncate_args *uap)
3326 {
3327 
3328 	return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3329 }
3330 
3331 int
3332 freebsd6_ftruncate(struct thread *td, struct freebsd6_ftruncate_args *uap)
3333 {
3334 
3335 	return (kern_ftruncate(td, uap->fd, uap->length));
3336 }
3337 #endif
3338 
3339 int
3340 kern_fsync(struct thread *td, int fd, bool fullsync)
3341 {
3342 	struct vnode *vp;
3343 	struct mount *mp;
3344 	struct file *fp;
3345 	cap_rights_t rights;
3346 	int error, lock_flags;
3347 
3348 	AUDIT_ARG_FD(fd);
3349 	error = getvnode(td, fd, cap_rights_init(&rights, CAP_FSYNC), &fp);
3350 	if (error != 0)
3351 		return (error);
3352 	vp = fp->f_vnode;
3353 #if 0
3354 	if (!fullsync)
3355 		/* XXXKIB: compete outstanding aio writes */;
3356 #endif
3357 	error = vn_start_write(vp, &mp, V_WAIT | PCATCH);
3358 	if (error != 0)
3359 		goto drop;
3360 	if (MNT_SHARED_WRITES(mp) ||
3361 	    ((mp == NULL) && MNT_SHARED_WRITES(vp->v_mount))) {
3362 		lock_flags = LK_SHARED;
3363 	} else {
3364 		lock_flags = LK_EXCLUSIVE;
3365 	}
3366 	vn_lock(vp, lock_flags | LK_RETRY);
3367 	AUDIT_ARG_VNODE1(vp);
3368 	if (vp->v_object != NULL) {
3369 		VM_OBJECT_WLOCK(vp->v_object);
3370 		vm_object_page_clean(vp->v_object, 0, 0, 0);
3371 		VM_OBJECT_WUNLOCK(vp->v_object);
3372 	}
3373 	error = fullsync ? VOP_FSYNC(vp, MNT_WAIT, td) : VOP_FDATASYNC(vp, td);
3374 	VOP_UNLOCK(vp, 0);
3375 	vn_finished_write(mp);
3376 drop:
3377 	fdrop(fp, td);
3378 	return (error);
3379 }
3380 
3381 /*
3382  * Sync an open file.
3383  */
3384 #ifndef _SYS_SYSPROTO_H_
3385 struct fsync_args {
3386 	int	fd;
3387 };
3388 #endif
3389 int
3390 sys_fsync(struct thread *td, struct fsync_args *uap)
3391 {
3392 
3393 	return (kern_fsync(td, uap->fd, true));
3394 }
3395 
3396 int
3397 sys_fdatasync(struct thread *td, struct fdatasync_args *uap)
3398 {
3399 
3400 	return (kern_fsync(td, uap->fd, false));
3401 }
3402 
3403 /*
3404  * Rename files.  Source and destination must either both be directories, or
3405  * both not be directories.  If target is a directory, it must be empty.
3406  */
3407 #ifndef _SYS_SYSPROTO_H_
3408 struct rename_args {
3409 	char	*from;
3410 	char	*to;
3411 };
3412 #endif
3413 int
3414 sys_rename(struct thread *td, struct rename_args *uap)
3415 {
3416 
3417 	return (kern_renameat(td, AT_FDCWD, uap->from, AT_FDCWD,
3418 	    uap->to, UIO_USERSPACE));
3419 }
3420 
3421 #ifndef _SYS_SYSPROTO_H_
3422 struct renameat_args {
3423 	int	oldfd;
3424 	char	*old;
3425 	int	newfd;
3426 	char	*new;
3427 };
3428 #endif
3429 int
3430 sys_renameat(struct thread *td, struct renameat_args *uap)
3431 {
3432 
3433 	return (kern_renameat(td, uap->oldfd, uap->old, uap->newfd, uap->new,
3434 	    UIO_USERSPACE));
3435 }
3436 
3437 int
3438 kern_renameat(struct thread *td, int oldfd, char *old, int newfd, char *new,
3439     enum uio_seg pathseg)
3440 {
3441 	struct mount *mp = NULL;
3442 	struct vnode *tvp, *fvp, *tdvp;
3443 	struct nameidata fromnd, tond;
3444 	cap_rights_t rights;
3445 	int error;
3446 
3447 again:
3448 	bwillwrite();
3449 #ifdef MAC
3450 	NDINIT_ATRIGHTS(&fromnd, DELETE, LOCKPARENT | LOCKLEAF | SAVESTART |
3451 	    AUDITVNODE1, pathseg, old, oldfd,
3452 	    cap_rights_init(&rights, CAP_RENAMEAT_SOURCE), td);
3453 #else
3454 	NDINIT_ATRIGHTS(&fromnd, DELETE, WANTPARENT | SAVESTART | AUDITVNODE1,
3455 	    pathseg, old, oldfd,
3456 	    cap_rights_init(&rights, CAP_RENAMEAT_SOURCE), td);
3457 #endif
3458 
3459 	if ((error = namei(&fromnd)) != 0)
3460 		return (error);
3461 #ifdef MAC
3462 	error = mac_vnode_check_rename_from(td->td_ucred, fromnd.ni_dvp,
3463 	    fromnd.ni_vp, &fromnd.ni_cnd);
3464 	VOP_UNLOCK(fromnd.ni_dvp, 0);
3465 	if (fromnd.ni_dvp != fromnd.ni_vp)
3466 		VOP_UNLOCK(fromnd.ni_vp, 0);
3467 #endif
3468 	fvp = fromnd.ni_vp;
3469 	NDINIT_ATRIGHTS(&tond, RENAME, LOCKPARENT | LOCKLEAF | NOCACHE |
3470 	    SAVESTART | AUDITVNODE2, pathseg, new, newfd,
3471 	    cap_rights_init(&rights, CAP_RENAMEAT_TARGET), td);
3472 	if (fromnd.ni_vp->v_type == VDIR)
3473 		tond.ni_cnd.cn_flags |= WILLBEDIR;
3474 	if ((error = namei(&tond)) != 0) {
3475 		/* Translate error code for rename("dir1", "dir2/."). */
3476 		if (error == EISDIR && fvp->v_type == VDIR)
3477 			error = EINVAL;
3478 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3479 		vrele(fromnd.ni_dvp);
3480 		vrele(fvp);
3481 		goto out1;
3482 	}
3483 	tdvp = tond.ni_dvp;
3484 	tvp = tond.ni_vp;
3485 	error = vn_start_write(fvp, &mp, V_NOWAIT);
3486 	if (error != 0) {
3487 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3488 		NDFREE(&tond, NDF_ONLY_PNBUF);
3489 		if (tvp != NULL)
3490 			vput(tvp);
3491 		if (tdvp == tvp)
3492 			vrele(tdvp);
3493 		else
3494 			vput(tdvp);
3495 		vrele(fromnd.ni_dvp);
3496 		vrele(fvp);
3497 		vrele(tond.ni_startdir);
3498 		if (fromnd.ni_startdir != NULL)
3499 			vrele(fromnd.ni_startdir);
3500 		error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH);
3501 		if (error != 0)
3502 			return (error);
3503 		goto again;
3504 	}
3505 	if (tvp != NULL) {
3506 		if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
3507 			error = ENOTDIR;
3508 			goto out;
3509 		} else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
3510 			error = EISDIR;
3511 			goto out;
3512 		}
3513 #ifdef CAPABILITIES
3514 		if (newfd != AT_FDCWD) {
3515 			/*
3516 			 * If the target already exists we require CAP_UNLINKAT
3517 			 * from 'newfd'.
3518 			 */
3519 			error = cap_check(&tond.ni_filecaps.fc_rights,
3520 			    cap_rights_init(&rights, CAP_UNLINKAT));
3521 			if (error != 0)
3522 				goto out;
3523 		}
3524 #endif
3525 	}
3526 	if (fvp == tdvp) {
3527 		error = EINVAL;
3528 		goto out;
3529 	}
3530 	/*
3531 	 * If the source is the same as the destination (that is, if they
3532 	 * are links to the same vnode), then there is nothing to do.
3533 	 */
3534 	if (fvp == tvp)
3535 		error = -1;
3536 #ifdef MAC
3537 	else
3538 		error = mac_vnode_check_rename_to(td->td_ucred, tdvp,
3539 		    tond.ni_vp, fromnd.ni_dvp == tdvp, &tond.ni_cnd);
3540 #endif
3541 out:
3542 	if (error == 0) {
3543 		error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
3544 		    tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
3545 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3546 		NDFREE(&tond, NDF_ONLY_PNBUF);
3547 	} else {
3548 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3549 		NDFREE(&tond, NDF_ONLY_PNBUF);
3550 		if (tvp != NULL)
3551 			vput(tvp);
3552 		if (tdvp == tvp)
3553 			vrele(tdvp);
3554 		else
3555 			vput(tdvp);
3556 		vrele(fromnd.ni_dvp);
3557 		vrele(fvp);
3558 	}
3559 	vrele(tond.ni_startdir);
3560 	vn_finished_write(mp);
3561 out1:
3562 	if (fromnd.ni_startdir)
3563 		vrele(fromnd.ni_startdir);
3564 	if (error == -1)
3565 		return (0);
3566 	return (error);
3567 }
3568 
3569 /*
3570  * Make a directory file.
3571  */
3572 #ifndef _SYS_SYSPROTO_H_
3573 struct mkdir_args {
3574 	char	*path;
3575 	int	mode;
3576 };
3577 #endif
3578 int
3579 sys_mkdir(struct thread *td, struct mkdir_args *uap)
3580 {
3581 
3582 	return (kern_mkdirat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
3583 	    uap->mode));
3584 }
3585 
3586 #ifndef _SYS_SYSPROTO_H_
3587 struct mkdirat_args {
3588 	int	fd;
3589 	char	*path;
3590 	mode_t	mode;
3591 };
3592 #endif
3593 int
3594 sys_mkdirat(struct thread *td, struct mkdirat_args *uap)
3595 {
3596 
3597 	return (kern_mkdirat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode));
3598 }
3599 
3600 int
3601 kern_mkdirat(struct thread *td, int fd, char *path, enum uio_seg segflg,
3602     int mode)
3603 {
3604 	struct mount *mp;
3605 	struct vnode *vp;
3606 	struct vattr vattr;
3607 	struct nameidata nd;
3608 	cap_rights_t rights;
3609 	int error;
3610 
3611 	AUDIT_ARG_MODE(mode);
3612 restart:
3613 	bwillwrite();
3614 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
3615 	    NOCACHE, segflg, path, fd, cap_rights_init(&rights, CAP_MKDIRAT),
3616 	    td);
3617 	nd.ni_cnd.cn_flags |= WILLBEDIR;
3618 	if ((error = namei(&nd)) != 0)
3619 		return (error);
3620 	vp = nd.ni_vp;
3621 	if (vp != NULL) {
3622 		NDFREE(&nd, NDF_ONLY_PNBUF);
3623 		/*
3624 		 * XXX namei called with LOCKPARENT but not LOCKLEAF has
3625 		 * the strange behaviour of leaving the vnode unlocked
3626 		 * if the target is the same vnode as the parent.
3627 		 */
3628 		if (vp == nd.ni_dvp)
3629 			vrele(nd.ni_dvp);
3630 		else
3631 			vput(nd.ni_dvp);
3632 		vrele(vp);
3633 		return (EEXIST);
3634 	}
3635 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3636 		NDFREE(&nd, NDF_ONLY_PNBUF);
3637 		vput(nd.ni_dvp);
3638 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3639 			return (error);
3640 		goto restart;
3641 	}
3642 	VATTR_NULL(&vattr);
3643 	vattr.va_type = VDIR;
3644 	vattr.va_mode = (mode & ACCESSPERMS) &~ td->td_proc->p_fd->fd_cmask;
3645 #ifdef MAC
3646 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
3647 	    &vattr);
3648 	if (error != 0)
3649 		goto out;
3650 #endif
3651 	error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
3652 #ifdef MAC
3653 out:
3654 #endif
3655 	NDFREE(&nd, NDF_ONLY_PNBUF);
3656 	vput(nd.ni_dvp);
3657 	if (error == 0)
3658 		vput(nd.ni_vp);
3659 	vn_finished_write(mp);
3660 	return (error);
3661 }
3662 
3663 /*
3664  * Remove a directory file.
3665  */
3666 #ifndef _SYS_SYSPROTO_H_
3667 struct rmdir_args {
3668 	char	*path;
3669 };
3670 #endif
3671 int
3672 sys_rmdir(struct thread *td, struct rmdir_args *uap)
3673 {
3674 
3675 	return (kern_rmdirat(td, AT_FDCWD, uap->path, UIO_USERSPACE));
3676 }
3677 
3678 int
3679 kern_rmdirat(struct thread *td, int fd, char *path, enum uio_seg pathseg)
3680 {
3681 	struct mount *mp;
3682 	struct vnode *vp;
3683 	struct nameidata nd;
3684 	cap_rights_t rights;
3685 	int error;
3686 
3687 restart:
3688 	bwillwrite();
3689 	NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1,
3690 	    pathseg, path, fd, cap_rights_init(&rights, CAP_UNLINKAT), td);
3691 	if ((error = namei(&nd)) != 0)
3692 		return (error);
3693 	vp = nd.ni_vp;
3694 	if (vp->v_type != VDIR) {
3695 		error = ENOTDIR;
3696 		goto out;
3697 	}
3698 	/*
3699 	 * No rmdir "." please.
3700 	 */
3701 	if (nd.ni_dvp == vp) {
3702 		error = EINVAL;
3703 		goto out;
3704 	}
3705 	/*
3706 	 * The root of a mounted filesystem cannot be deleted.
3707 	 */
3708 	if (vp->v_vflag & VV_ROOT) {
3709 		error = EBUSY;
3710 		goto out;
3711 	}
3712 #ifdef MAC
3713 	error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
3714 	    &nd.ni_cnd);
3715 	if (error != 0)
3716 		goto out;
3717 #endif
3718 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3719 		NDFREE(&nd, NDF_ONLY_PNBUF);
3720 		vput(vp);
3721 		if (nd.ni_dvp == vp)
3722 			vrele(nd.ni_dvp);
3723 		else
3724 			vput(nd.ni_dvp);
3725 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3726 			return (error);
3727 		goto restart;
3728 	}
3729 	vfs_notify_upper(vp, VFS_NOTIFY_UPPER_UNLINK);
3730 	error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
3731 	vn_finished_write(mp);
3732 out:
3733 	NDFREE(&nd, NDF_ONLY_PNBUF);
3734 	vput(vp);
3735 	if (nd.ni_dvp == vp)
3736 		vrele(nd.ni_dvp);
3737 	else
3738 		vput(nd.ni_dvp);
3739 	return (error);
3740 }
3741 
3742 #if defined(COMPAT_43) || defined(COMPAT_FREEBSD11)
3743 int
3744 freebsd11_kern_getdirentries(struct thread *td, int fd, char *ubuf, u_int count,
3745     long *basep, void (*func)(struct freebsd11_dirent *))
3746 {
3747 	struct freebsd11_dirent dstdp;
3748 	struct dirent *dp, *edp;
3749 	char *dirbuf;
3750 	off_t base;
3751 	ssize_t resid, ucount;
3752 	int error;
3753 
3754 	/* XXX arbitrary sanity limit on `count'. */
3755 	count = min(count, 64 * 1024);
3756 
3757 	dirbuf = malloc(count, M_TEMP, M_WAITOK);
3758 
3759 	error = kern_getdirentries(td, fd, dirbuf, count, &base, &resid,
3760 	    UIO_SYSSPACE);
3761 	if (error != 0)
3762 		goto done;
3763 	if (basep != NULL)
3764 		*basep = base;
3765 
3766 	ucount = 0;
3767 	for (dp = (struct dirent *)dirbuf,
3768 	    edp = (struct dirent *)&dirbuf[count - resid];
3769 	    ucount < count && dp < edp; ) {
3770 		if (dp->d_reclen == 0)
3771 			break;
3772 		MPASS(dp->d_reclen >= _GENERIC_DIRLEN(0));
3773 		if (dp->d_namlen >= sizeof(dstdp.d_name))
3774 			continue;
3775 		dstdp.d_type = dp->d_type;
3776 		dstdp.d_namlen = dp->d_namlen;
3777 		dstdp.d_fileno = dp->d_fileno;		/* truncate */
3778 		if (dstdp.d_fileno != dp->d_fileno) {
3779 			switch (ino64_trunc_error) {
3780 			default:
3781 			case 0:
3782 				break;
3783 			case 1:
3784 				error = EOVERFLOW;
3785 				goto done;
3786 			case 2:
3787 				dstdp.d_fileno = UINT32_MAX;
3788 				break;
3789 			}
3790 		}
3791 		dstdp.d_reclen = sizeof(dstdp) - sizeof(dstdp.d_name) +
3792 		    ((dp->d_namlen + 1 + 3) &~ 3);
3793 		bcopy(dp->d_name, dstdp.d_name, dstdp.d_namlen);
3794 		bzero(dstdp.d_name + dstdp.d_namlen,
3795 		    dstdp.d_reclen - offsetof(struct freebsd11_dirent, d_name) -
3796 		    dstdp.d_namlen);
3797 		MPASS(dstdp.d_reclen <= dp->d_reclen);
3798 		MPASS(ucount + dstdp.d_reclen <= count);
3799 		if (func != NULL)
3800 			func(&dstdp);
3801 		error = copyout(&dstdp, ubuf + ucount, dstdp.d_reclen);
3802 		if (error != 0)
3803 			break;
3804 		dp = (struct dirent *)((char *)dp + dp->d_reclen);
3805 		ucount += dstdp.d_reclen;
3806 	}
3807 
3808 done:
3809 	free(dirbuf, M_TEMP);
3810 	if (error == 0)
3811 		td->td_retval[0] = ucount;
3812 	return (error);
3813 }
3814 #endif /* COMPAT */
3815 
3816 #ifdef COMPAT_43
3817 static void
3818 ogetdirentries_cvt(struct freebsd11_dirent *dp)
3819 {
3820 #if (BYTE_ORDER == LITTLE_ENDIAN)
3821 	/*
3822 	 * The expected low byte of dp->d_namlen is our dp->d_type.
3823 	 * The high MBZ byte of dp->d_namlen is our dp->d_namlen.
3824 	 */
3825 	dp->d_type = dp->d_namlen;
3826 	dp->d_namlen = 0;
3827 #else
3828 	/*
3829 	 * The dp->d_type is the high byte of the expected dp->d_namlen,
3830 	 * so must be zero'ed.
3831 	 */
3832 	dp->d_type = 0;
3833 #endif
3834 }
3835 
3836 /*
3837  * Read a block of directory entries in a filesystem independent format.
3838  */
3839 #ifndef _SYS_SYSPROTO_H_
3840 struct ogetdirentries_args {
3841 	int	fd;
3842 	char	*buf;
3843 	u_int	count;
3844 	long	*basep;
3845 };
3846 #endif
3847 int
3848 ogetdirentries(struct thread *td, struct ogetdirentries_args *uap)
3849 {
3850 	long loff;
3851 	int error;
3852 
3853 	error = kern_ogetdirentries(td, uap, &loff);
3854 	if (error == 0)
3855 		error = copyout(&loff, uap->basep, sizeof(long));
3856 	return (error);
3857 }
3858 
3859 int
3860 kern_ogetdirentries(struct thread *td, struct ogetdirentries_args *uap,
3861     long *ploff)
3862 {
3863 	long base;
3864 	int error;
3865 
3866 	/* XXX arbitrary sanity limit on `count'. */
3867 	if (uap->count > 64 * 1024)
3868 		return (EINVAL);
3869 
3870 	error = freebsd11_kern_getdirentries(td, uap->fd, uap->buf, uap->count,
3871 	    &base, ogetdirentries_cvt);
3872 
3873 	if (error == 0 && uap->basep != NULL)
3874 		error = copyout(&base, uap->basep, sizeof(long));
3875 
3876 	return (error);
3877 }
3878 #endif /* COMPAT_43 */
3879 
3880 #if defined(COMPAT_FREEBSD11)
3881 #ifndef _SYS_SYSPROTO_H_
3882 struct freebsd11_getdirentries_args {
3883 	int	fd;
3884 	char	*buf;
3885 	u_int	count;
3886 	long	*basep;
3887 };
3888 #endif
3889 int
3890 freebsd11_getdirentries(struct thread *td,
3891     struct freebsd11_getdirentries_args *uap)
3892 {
3893 	long base;
3894 	int error;
3895 
3896 	error = freebsd11_kern_getdirentries(td, uap->fd, uap->buf, uap->count,
3897 	    &base, NULL);
3898 
3899 	if (error == 0 && uap->basep != NULL)
3900 		error = copyout(&base, uap->basep, sizeof(long));
3901 	return (error);
3902 }
3903 
3904 int
3905 freebsd11_getdents(struct thread *td, struct freebsd11_getdents_args *uap)
3906 {
3907 	struct freebsd11_getdirentries_args ap;
3908 
3909 	ap.fd = uap->fd;
3910 	ap.buf = uap->buf;
3911 	ap.count = uap->count;
3912 	ap.basep = NULL;
3913 	return (freebsd11_getdirentries(td, &ap));
3914 }
3915 #endif /* COMPAT_FREEBSD11 */
3916 
3917 /*
3918  * Read a block of directory entries in a filesystem independent format.
3919  */
3920 int
3921 sys_getdirentries(struct thread *td, struct getdirentries_args *uap)
3922 {
3923 	off_t base;
3924 	int error;
3925 
3926 	error = kern_getdirentries(td, uap->fd, uap->buf, uap->count, &base,
3927 	    NULL, UIO_USERSPACE);
3928 	if (error != 0)
3929 		return (error);
3930 	if (uap->basep != NULL)
3931 		error = copyout(&base, uap->basep, sizeof(off_t));
3932 	return (error);
3933 }
3934 
3935 int
3936 kern_getdirentries(struct thread *td, int fd, char *buf, size_t count,
3937     off_t *basep, ssize_t *residp, enum uio_seg bufseg)
3938 {
3939 	struct vnode *vp;
3940 	struct file *fp;
3941 	struct uio auio;
3942 	struct iovec aiov;
3943 	off_t loff;
3944 	int error, eofflag;
3945 	off_t foffset;
3946 
3947 	AUDIT_ARG_FD(fd);
3948 	if (count > IOSIZE_MAX)
3949 		return (EINVAL);
3950 	auio.uio_resid = count;
3951 	error = getvnode(td, fd, &cap_read_rights, &fp);
3952 	if (error != 0)
3953 		return (error);
3954 	if ((fp->f_flag & FREAD) == 0) {
3955 		fdrop(fp, td);
3956 		return (EBADF);
3957 	}
3958 	vp = fp->f_vnode;
3959 	foffset = foffset_lock(fp, 0);
3960 unionread:
3961 	if (vp->v_type != VDIR) {
3962 		error = EINVAL;
3963 		goto fail;
3964 	}
3965 	aiov.iov_base = buf;
3966 	aiov.iov_len = count;
3967 	auio.uio_iov = &aiov;
3968 	auio.uio_iovcnt = 1;
3969 	auio.uio_rw = UIO_READ;
3970 	auio.uio_segflg = bufseg;
3971 	auio.uio_td = td;
3972 	vn_lock(vp, LK_SHARED | LK_RETRY);
3973 	AUDIT_ARG_VNODE1(vp);
3974 	loff = auio.uio_offset = foffset;
3975 #ifdef MAC
3976 	error = mac_vnode_check_readdir(td->td_ucred, vp);
3977 	if (error == 0)
3978 #endif
3979 		error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, NULL,
3980 		    NULL);
3981 	foffset = auio.uio_offset;
3982 	if (error != 0) {
3983 		VOP_UNLOCK(vp, 0);
3984 		goto fail;
3985 	}
3986 	if (count == auio.uio_resid &&
3987 	    (vp->v_vflag & VV_ROOT) &&
3988 	    (vp->v_mount->mnt_flag & MNT_UNION)) {
3989 		struct vnode *tvp = vp;
3990 
3991 		vp = vp->v_mount->mnt_vnodecovered;
3992 		VREF(vp);
3993 		fp->f_vnode = vp;
3994 		fp->f_data = vp;
3995 		foffset = 0;
3996 		vput(tvp);
3997 		goto unionread;
3998 	}
3999 	VOP_UNLOCK(vp, 0);
4000 	*basep = loff;
4001 	if (residp != NULL)
4002 		*residp = auio.uio_resid;
4003 	td->td_retval[0] = count - auio.uio_resid;
4004 fail:
4005 	foffset_unlock(fp, foffset, 0);
4006 	fdrop(fp, td);
4007 	return (error);
4008 }
4009 
4010 /*
4011  * Set the mode mask for creation of filesystem nodes.
4012  */
4013 #ifndef _SYS_SYSPROTO_H_
4014 struct umask_args {
4015 	int	newmask;
4016 };
4017 #endif
4018 int
4019 sys_umask(struct thread *td, struct umask_args *uap)
4020 {
4021 	struct filedesc *fdp;
4022 
4023 	fdp = td->td_proc->p_fd;
4024 	FILEDESC_XLOCK(fdp);
4025 	td->td_retval[0] = fdp->fd_cmask;
4026 	fdp->fd_cmask = uap->newmask & ALLPERMS;
4027 	FILEDESC_XUNLOCK(fdp);
4028 	return (0);
4029 }
4030 
4031 /*
4032  * Void all references to file by ripping underlying filesystem away from
4033  * vnode.
4034  */
4035 #ifndef _SYS_SYSPROTO_H_
4036 struct revoke_args {
4037 	char	*path;
4038 };
4039 #endif
4040 int
4041 sys_revoke(struct thread *td, struct revoke_args *uap)
4042 {
4043 	struct vnode *vp;
4044 	struct vattr vattr;
4045 	struct nameidata nd;
4046 	int error;
4047 
4048 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
4049 	    uap->path, td);
4050 	if ((error = namei(&nd)) != 0)
4051 		return (error);
4052 	vp = nd.ni_vp;
4053 	NDFREE(&nd, NDF_ONLY_PNBUF);
4054 	if (vp->v_type != VCHR || vp->v_rdev == NULL) {
4055 		error = EINVAL;
4056 		goto out;
4057 	}
4058 #ifdef MAC
4059 	error = mac_vnode_check_revoke(td->td_ucred, vp);
4060 	if (error != 0)
4061 		goto out;
4062 #endif
4063 	error = VOP_GETATTR(vp, &vattr, td->td_ucred);
4064 	if (error != 0)
4065 		goto out;
4066 	if (td->td_ucred->cr_uid != vattr.va_uid) {
4067 		error = priv_check(td, PRIV_VFS_ADMIN);
4068 		if (error != 0)
4069 			goto out;
4070 	}
4071 	if (vcount(vp) > 1)
4072 		VOP_REVOKE(vp, REVOKEALL);
4073 out:
4074 	vput(vp);
4075 	return (error);
4076 }
4077 
4078 /*
4079  * Convert a user file descriptor to a kernel file entry and check that, if it
4080  * is a capability, the correct rights are present. A reference on the file
4081  * entry is held upon returning.
4082  */
4083 int
4084 getvnode(struct thread *td, int fd, cap_rights_t *rightsp, struct file **fpp)
4085 {
4086 	struct file *fp;
4087 	int error;
4088 
4089 	error = fget_unlocked(td->td_proc->p_fd, fd, rightsp, &fp, NULL);
4090 	if (error != 0)
4091 		return (error);
4092 
4093 	/*
4094 	 * The file could be not of the vnode type, or it may be not
4095 	 * yet fully initialized, in which case the f_vnode pointer
4096 	 * may be set, but f_ops is still badfileops.  E.g.,
4097 	 * devfs_open() transiently create such situation to
4098 	 * facilitate csw d_fdopen().
4099 	 *
4100 	 * Dupfdopen() handling in kern_openat() installs the
4101 	 * half-baked file into the process descriptor table, allowing
4102 	 * other thread to dereference it. Guard against the race by
4103 	 * checking f_ops.
4104 	 */
4105 	if (fp->f_vnode == NULL || fp->f_ops == &badfileops) {
4106 		fdrop(fp, td);
4107 		return (EINVAL);
4108 	}
4109 	*fpp = fp;
4110 	return (0);
4111 }
4112 
4113 
4114 /*
4115  * Get an (NFS) file handle.
4116  */
4117 #ifndef _SYS_SYSPROTO_H_
4118 struct lgetfh_args {
4119 	char	*fname;
4120 	fhandle_t *fhp;
4121 };
4122 #endif
4123 int
4124 sys_lgetfh(struct thread *td, struct lgetfh_args *uap)
4125 {
4126 	struct nameidata nd;
4127 	fhandle_t fh;
4128 	struct vnode *vp;
4129 	int error;
4130 
4131 	error = priv_check(td, PRIV_VFS_GETFH);
4132 	if (error != 0)
4133 		return (error);
4134 	NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
4135 	    uap->fname, td);
4136 	error = namei(&nd);
4137 	if (error != 0)
4138 		return (error);
4139 	NDFREE(&nd, NDF_ONLY_PNBUF);
4140 	vp = nd.ni_vp;
4141 	bzero(&fh, sizeof(fh));
4142 	fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
4143 	error = VOP_VPTOFH(vp, &fh.fh_fid);
4144 	vput(vp);
4145 	if (error == 0)
4146 		error = copyout(&fh, uap->fhp, sizeof (fh));
4147 	return (error);
4148 }
4149 
4150 #ifndef _SYS_SYSPROTO_H_
4151 struct getfh_args {
4152 	char	*fname;
4153 	fhandle_t *fhp;
4154 };
4155 #endif
4156 int
4157 sys_getfh(struct thread *td, struct getfh_args *uap)
4158 {
4159 	struct nameidata nd;
4160 	fhandle_t fh;
4161 	struct vnode *vp;
4162 	int error;
4163 
4164 	error = priv_check(td, PRIV_VFS_GETFH);
4165 	if (error != 0)
4166 		return (error);
4167 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
4168 	    uap->fname, td);
4169 	error = namei(&nd);
4170 	if (error != 0)
4171 		return (error);
4172 	NDFREE(&nd, NDF_ONLY_PNBUF);
4173 	vp = nd.ni_vp;
4174 	bzero(&fh, sizeof(fh));
4175 	fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
4176 	error = VOP_VPTOFH(vp, &fh.fh_fid);
4177 	vput(vp);
4178 	if (error == 0)
4179 		error = copyout(&fh, uap->fhp, sizeof (fh));
4180 	return (error);
4181 }
4182 
4183 /*
4184  * syscall for the rpc.lockd to use to translate a NFS file handle into an
4185  * open descriptor.
4186  *
4187  * warning: do not remove the priv_check() call or this becomes one giant
4188  * security hole.
4189  */
4190 #ifndef _SYS_SYSPROTO_H_
4191 struct fhopen_args {
4192 	const struct fhandle *u_fhp;
4193 	int flags;
4194 };
4195 #endif
4196 int
4197 sys_fhopen(struct thread *td, struct fhopen_args *uap)
4198 {
4199 	struct mount *mp;
4200 	struct vnode *vp;
4201 	struct fhandle fhp;
4202 	struct file *fp;
4203 	int fmode, error;
4204 	int indx;
4205 
4206 	error = priv_check(td, PRIV_VFS_FHOPEN);
4207 	if (error != 0)
4208 		return (error);
4209 	indx = -1;
4210 	fmode = FFLAGS(uap->flags);
4211 	/* why not allow a non-read/write open for our lockd? */
4212 	if (((fmode & (FREAD | FWRITE)) == 0) || (fmode & O_CREAT))
4213 		return (EINVAL);
4214 	error = copyin(uap->u_fhp, &fhp, sizeof(fhp));
4215 	if (error != 0)
4216 		return(error);
4217 	/* find the mount point */
4218 	mp = vfs_busyfs(&fhp.fh_fsid);
4219 	if (mp == NULL)
4220 		return (ESTALE);
4221 	/* now give me my vnode, it gets returned to me locked */
4222 	error = VFS_FHTOVP(mp, &fhp.fh_fid, LK_EXCLUSIVE, &vp);
4223 	vfs_unbusy(mp);
4224 	if (error != 0)
4225 		return (error);
4226 
4227 	error = falloc_noinstall(td, &fp);
4228 	if (error != 0) {
4229 		vput(vp);
4230 		return (error);
4231 	}
4232 	/*
4233 	 * An extra reference on `fp' has been held for us by
4234 	 * falloc_noinstall().
4235 	 */
4236 
4237 #ifdef INVARIANTS
4238 	td->td_dupfd = -1;
4239 #endif
4240 	error = vn_open_vnode(vp, fmode, td->td_ucred, td, fp);
4241 	if (error != 0) {
4242 		KASSERT(fp->f_ops == &badfileops,
4243 		    ("VOP_OPEN in fhopen() set f_ops"));
4244 		KASSERT(td->td_dupfd < 0,
4245 		    ("fhopen() encountered fdopen()"));
4246 
4247 		vput(vp);
4248 		goto bad;
4249 	}
4250 #ifdef INVARIANTS
4251 	td->td_dupfd = 0;
4252 #endif
4253 	fp->f_vnode = vp;
4254 	fp->f_seqcount = 1;
4255 	finit(fp, (fmode & FMASK) | (fp->f_flag & FHASLOCK), DTYPE_VNODE, vp,
4256 	    &vnops);
4257 	VOP_UNLOCK(vp, 0);
4258 	if ((fmode & O_TRUNC) != 0) {
4259 		error = fo_truncate(fp, 0, td->td_ucred, td);
4260 		if (error != 0)
4261 			goto bad;
4262 	}
4263 
4264 	error = finstall(td, fp, &indx, fmode, NULL);
4265 bad:
4266 	fdrop(fp, td);
4267 	td->td_retval[0] = indx;
4268 	return (error);
4269 }
4270 
4271 /*
4272  * Stat an (NFS) file handle.
4273  */
4274 #ifndef _SYS_SYSPROTO_H_
4275 struct fhstat_args {
4276 	struct fhandle *u_fhp;
4277 	struct stat *sb;
4278 };
4279 #endif
4280 int
4281 sys_fhstat(struct thread *td, struct fhstat_args *uap)
4282 {
4283 	struct stat sb;
4284 	struct fhandle fh;
4285 	int error;
4286 
4287 	error = copyin(uap->u_fhp, &fh, sizeof(fh));
4288 	if (error != 0)
4289 		return (error);
4290 	error = kern_fhstat(td, fh, &sb);
4291 	if (error == 0)
4292 		error = copyout(&sb, uap->sb, sizeof(sb));
4293 	return (error);
4294 }
4295 
4296 int
4297 kern_fhstat(struct thread *td, struct fhandle fh, struct stat *sb)
4298 {
4299 	struct mount *mp;
4300 	struct vnode *vp;
4301 	int error;
4302 
4303 	error = priv_check(td, PRIV_VFS_FHSTAT);
4304 	if (error != 0)
4305 		return (error);
4306 	if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4307 		return (ESTALE);
4308 	error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4309 	vfs_unbusy(mp);
4310 	if (error != 0)
4311 		return (error);
4312 	error = vn_stat(vp, sb, td->td_ucred, NOCRED, td);
4313 	vput(vp);
4314 	return (error);
4315 }
4316 
4317 /*
4318  * Implement fstatfs() for (NFS) file handles.
4319  */
4320 #ifndef _SYS_SYSPROTO_H_
4321 struct fhstatfs_args {
4322 	struct fhandle *u_fhp;
4323 	struct statfs *buf;
4324 };
4325 #endif
4326 int
4327 sys_fhstatfs(struct thread *td, struct fhstatfs_args *uap)
4328 {
4329 	struct statfs *sfp;
4330 	fhandle_t fh;
4331 	int error;
4332 
4333 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
4334 	if (error != 0)
4335 		return (error);
4336 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
4337 	error = kern_fhstatfs(td, fh, sfp);
4338 	if (error == 0)
4339 		error = copyout(sfp, uap->buf, sizeof(*sfp));
4340 	free(sfp, M_STATFS);
4341 	return (error);
4342 }
4343 
4344 int
4345 kern_fhstatfs(struct thread *td, fhandle_t fh, struct statfs *buf)
4346 {
4347 	struct statfs *sp;
4348 	struct mount *mp;
4349 	struct vnode *vp;
4350 	int error;
4351 
4352 	error = priv_check(td, PRIV_VFS_FHSTATFS);
4353 	if (error != 0)
4354 		return (error);
4355 	if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4356 		return (ESTALE);
4357 	error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4358 	if (error != 0) {
4359 		vfs_unbusy(mp);
4360 		return (error);
4361 	}
4362 	vput(vp);
4363 	error = prison_canseemount(td->td_ucred, mp);
4364 	if (error != 0)
4365 		goto out;
4366 #ifdef MAC
4367 	error = mac_mount_check_stat(td->td_ucred, mp);
4368 	if (error != 0)
4369 		goto out;
4370 #endif
4371 	/*
4372 	 * Set these in case the underlying filesystem fails to do so.
4373 	 */
4374 	sp = &mp->mnt_stat;
4375 	sp->f_version = STATFS_VERSION;
4376 	sp->f_namemax = NAME_MAX;
4377 	sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
4378 	error = VFS_STATFS(mp, sp);
4379 	if (error == 0)
4380 		*buf = *sp;
4381 out:
4382 	vfs_unbusy(mp);
4383 	return (error);
4384 }
4385 
4386 int
4387 kern_posix_fallocate(struct thread *td, int fd, off_t offset, off_t len)
4388 {
4389 	struct file *fp;
4390 	struct mount *mp;
4391 	struct vnode *vp;
4392 	off_t olen, ooffset;
4393 	int error;
4394 #ifdef AUDIT
4395 	int audited_vnode1 = 0;
4396 #endif
4397 
4398 	AUDIT_ARG_FD(fd);
4399 	if (offset < 0 || len <= 0)
4400 		return (EINVAL);
4401 	/* Check for wrap. */
4402 	if (offset > OFF_MAX - len)
4403 		return (EFBIG);
4404 	AUDIT_ARG_FD(fd);
4405 	error = fget(td, fd, &cap_pwrite_rights, &fp);
4406 	if (error != 0)
4407 		return (error);
4408 	AUDIT_ARG_FILE(td->td_proc, fp);
4409 	if ((fp->f_ops->fo_flags & DFLAG_SEEKABLE) == 0) {
4410 		error = ESPIPE;
4411 		goto out;
4412 	}
4413 	if ((fp->f_flag & FWRITE) == 0) {
4414 		error = EBADF;
4415 		goto out;
4416 	}
4417 	if (fp->f_type != DTYPE_VNODE) {
4418 		error = ENODEV;
4419 		goto out;
4420 	}
4421 	vp = fp->f_vnode;
4422 	if (vp->v_type != VREG) {
4423 		error = ENODEV;
4424 		goto out;
4425 	}
4426 
4427 	/* Allocating blocks may take a long time, so iterate. */
4428 	for (;;) {
4429 		olen = len;
4430 		ooffset = offset;
4431 
4432 		bwillwrite();
4433 		mp = NULL;
4434 		error = vn_start_write(vp, &mp, V_WAIT | PCATCH);
4435 		if (error != 0)
4436 			break;
4437 		error = vn_lock(vp, LK_EXCLUSIVE);
4438 		if (error != 0) {
4439 			vn_finished_write(mp);
4440 			break;
4441 		}
4442 #ifdef AUDIT
4443 		if (!audited_vnode1) {
4444 			AUDIT_ARG_VNODE1(vp);
4445 			audited_vnode1 = 1;
4446 		}
4447 #endif
4448 #ifdef MAC
4449 		error = mac_vnode_check_write(td->td_ucred, fp->f_cred, vp);
4450 		if (error == 0)
4451 #endif
4452 			error = VOP_ALLOCATE(vp, &offset, &len);
4453 		VOP_UNLOCK(vp, 0);
4454 		vn_finished_write(mp);
4455 
4456 		if (olen + ooffset != offset + len) {
4457 			panic("offset + len changed from %jx/%jx to %jx/%jx",
4458 			    ooffset, olen, offset, len);
4459 		}
4460 		if (error != 0 || len == 0)
4461 			break;
4462 		KASSERT(olen > len, ("Iteration did not make progress?"));
4463 		maybe_yield();
4464 	}
4465  out:
4466 	fdrop(fp, td);
4467 	return (error);
4468 }
4469 
4470 int
4471 sys_posix_fallocate(struct thread *td, struct posix_fallocate_args *uap)
4472 {
4473 	int error;
4474 
4475 	error = kern_posix_fallocate(td, uap->fd, uap->offset, uap->len);
4476 	return (kern_posix_error(td, error));
4477 }
4478 
4479 /*
4480  * Unlike madvise(2), we do not make a best effort to remember every
4481  * possible caching hint.  Instead, we remember the last setting with
4482  * the exception that we will allow POSIX_FADV_NORMAL to adjust the
4483  * region of any current setting.
4484  */
4485 int
4486 kern_posix_fadvise(struct thread *td, int fd, off_t offset, off_t len,
4487     int advice)
4488 {
4489 	struct fadvise_info *fa, *new;
4490 	struct file *fp;
4491 	struct vnode *vp;
4492 	off_t end;
4493 	int error;
4494 
4495 	if (offset < 0 || len < 0 || offset > OFF_MAX - len)
4496 		return (EINVAL);
4497 	AUDIT_ARG_VALUE(advice);
4498 	switch (advice) {
4499 	case POSIX_FADV_SEQUENTIAL:
4500 	case POSIX_FADV_RANDOM:
4501 	case POSIX_FADV_NOREUSE:
4502 		new = malloc(sizeof(*fa), M_FADVISE, M_WAITOK);
4503 		break;
4504 	case POSIX_FADV_NORMAL:
4505 	case POSIX_FADV_WILLNEED:
4506 	case POSIX_FADV_DONTNEED:
4507 		new = NULL;
4508 		break;
4509 	default:
4510 		return (EINVAL);
4511 	}
4512 	/* XXX: CAP_POSIX_FADVISE? */
4513 	AUDIT_ARG_FD(fd);
4514 	error = fget(td, fd, &cap_no_rights, &fp);
4515 	if (error != 0)
4516 		goto out;
4517 	AUDIT_ARG_FILE(td->td_proc, fp);
4518 	if ((fp->f_ops->fo_flags & DFLAG_SEEKABLE) == 0) {
4519 		error = ESPIPE;
4520 		goto out;
4521 	}
4522 	if (fp->f_type != DTYPE_VNODE) {
4523 		error = ENODEV;
4524 		goto out;
4525 	}
4526 	vp = fp->f_vnode;
4527 	if (vp->v_type != VREG) {
4528 		error = ENODEV;
4529 		goto out;
4530 	}
4531 	if (len == 0)
4532 		end = OFF_MAX;
4533 	else
4534 		end = offset + len - 1;
4535 	switch (advice) {
4536 	case POSIX_FADV_SEQUENTIAL:
4537 	case POSIX_FADV_RANDOM:
4538 	case POSIX_FADV_NOREUSE:
4539 		/*
4540 		 * Try to merge any existing non-standard region with
4541 		 * this new region if possible, otherwise create a new
4542 		 * non-standard region for this request.
4543 		 */
4544 		mtx_pool_lock(mtxpool_sleep, fp);
4545 		fa = fp->f_advice;
4546 		if (fa != NULL && fa->fa_advice == advice &&
4547 		    ((fa->fa_start <= end && fa->fa_end >= offset) ||
4548 		    (end != OFF_MAX && fa->fa_start == end + 1) ||
4549 		    (fa->fa_end != OFF_MAX && fa->fa_end + 1 == offset))) {
4550 			if (offset < fa->fa_start)
4551 				fa->fa_start = offset;
4552 			if (end > fa->fa_end)
4553 				fa->fa_end = end;
4554 		} else {
4555 			new->fa_advice = advice;
4556 			new->fa_start = offset;
4557 			new->fa_end = end;
4558 			fp->f_advice = new;
4559 			new = fa;
4560 		}
4561 		mtx_pool_unlock(mtxpool_sleep, fp);
4562 		break;
4563 	case POSIX_FADV_NORMAL:
4564 		/*
4565 		 * If a the "normal" region overlaps with an existing
4566 		 * non-standard region, trim or remove the
4567 		 * non-standard region.
4568 		 */
4569 		mtx_pool_lock(mtxpool_sleep, fp);
4570 		fa = fp->f_advice;
4571 		if (fa != NULL) {
4572 			if (offset <= fa->fa_start && end >= fa->fa_end) {
4573 				new = fa;
4574 				fp->f_advice = NULL;
4575 			} else if (offset <= fa->fa_start &&
4576 			    end >= fa->fa_start)
4577 				fa->fa_start = end + 1;
4578 			else if (offset <= fa->fa_end && end >= fa->fa_end)
4579 				fa->fa_end = offset - 1;
4580 			else if (offset >= fa->fa_start && end <= fa->fa_end) {
4581 				/*
4582 				 * If the "normal" region is a middle
4583 				 * portion of the existing
4584 				 * non-standard region, just remove
4585 				 * the whole thing rather than picking
4586 				 * one side or the other to
4587 				 * preserve.
4588 				 */
4589 				new = fa;
4590 				fp->f_advice = NULL;
4591 			}
4592 		}
4593 		mtx_pool_unlock(mtxpool_sleep, fp);
4594 		break;
4595 	case POSIX_FADV_WILLNEED:
4596 	case POSIX_FADV_DONTNEED:
4597 		error = VOP_ADVISE(vp, offset, end, advice);
4598 		break;
4599 	}
4600 out:
4601 	if (fp != NULL)
4602 		fdrop(fp, td);
4603 	free(new, M_FADVISE);
4604 	return (error);
4605 }
4606 
4607 int
4608 sys_posix_fadvise(struct thread *td, struct posix_fadvise_args *uap)
4609 {
4610 	int error;
4611 
4612 	error = kern_posix_fadvise(td, uap->fd, uap->offset, uap->len,
4613 	    uap->advice);
4614 	return (kern_posix_error(td, error));
4615 }
4616