xref: /freebsd/sys/kern/vfs_syscalls.c (revision 410556f1f10fd35b350102725fd8504c3cb0afc8)
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 <fs/devfs/devfs.h>
91 
92 #include <ufs/ufs/quota.h>
93 
94 MALLOC_DEFINE(M_FADVISE, "fadvise", "posix_fadvise(2) information");
95 
96 static int kern_chflagsat(struct thread *td, int fd, const char *path,
97     enum uio_seg pathseg, u_long flags, int atflag);
98 static int setfflags(struct thread *td, struct vnode *, u_long);
99 static int getutimes(const struct timeval *, enum uio_seg, struct timespec *);
100 static int getutimens(const struct timespec *, enum uio_seg,
101     struct timespec *, int *);
102 static int setutimes(struct thread *td, struct vnode *,
103     const struct timespec *, int, int);
104 static int vn_access(struct vnode *vp, int user_flags, struct ucred *cred,
105     struct thread *td);
106 static int kern_fhlinkat(struct thread *td, int fd, const char *path,
107     enum uio_seg pathseg, fhandle_t *fhp);
108 static int kern_readlink_vp(struct vnode *vp, char *buf, enum uio_seg bufseg,
109     size_t count, struct thread *td);
110 static int kern_linkat_vp(struct thread *td, struct vnode *vp, int fd,
111     const char *path, enum uio_seg segflag);
112 
113 static uint64_t
114 at2cnpflags(u_int at_flags, u_int mask)
115 {
116 	u_int64_t res;
117 
118 	MPASS((at_flags & (AT_SYMLINK_FOLLOW | AT_SYMLINK_NOFOLLOW)) !=
119 	    (AT_SYMLINK_FOLLOW | AT_SYMLINK_NOFOLLOW));
120 
121 	res = 0;
122 	at_flags &= mask;
123 	if ((at_flags & AT_RESOLVE_BENEATH) != 0)
124 		res |= RBENEATH;
125 	if ((at_flags & AT_SYMLINK_FOLLOW) != 0)
126 		res |= FOLLOW;
127 	/* NOFOLLOW is pseudo flag */
128 	if ((mask & AT_SYMLINK_NOFOLLOW) != 0) {
129 		res |= (at_flags & AT_SYMLINK_NOFOLLOW) != 0 ? NOFOLLOW :
130 		    FOLLOW;
131 	}
132 	return (res);
133 }
134 
135 int
136 kern_sync(struct thread *td)
137 {
138 	struct mount *mp, *nmp;
139 	int save;
140 
141 	mtx_lock(&mountlist_mtx);
142 	for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
143 		if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK)) {
144 			nmp = TAILQ_NEXT(mp, mnt_list);
145 			continue;
146 		}
147 		if ((mp->mnt_flag & MNT_RDONLY) == 0 &&
148 		    vn_start_write(NULL, &mp, V_NOWAIT) == 0) {
149 			save = curthread_pflags_set(TDP_SYNCIO);
150 			vfs_periodic(mp, MNT_NOWAIT);
151 			VFS_SYNC(mp, MNT_NOWAIT);
152 			curthread_pflags_restore(save);
153 			vn_finished_write(mp);
154 		}
155 		mtx_lock(&mountlist_mtx);
156 		nmp = TAILQ_NEXT(mp, mnt_list);
157 		vfs_unbusy(mp);
158 	}
159 	mtx_unlock(&mountlist_mtx);
160 	return (0);
161 }
162 
163 /*
164  * Sync each mounted filesystem.
165  */
166 #ifndef _SYS_SYSPROTO_H_
167 struct sync_args {
168 	int     dummy;
169 };
170 #endif
171 /* ARGSUSED */
172 int
173 sys_sync(struct thread *td, struct sync_args *uap)
174 {
175 
176 	return (kern_sync(td));
177 }
178 
179 /*
180  * Change filesystem quotas.
181  */
182 #ifndef _SYS_SYSPROTO_H_
183 struct quotactl_args {
184 	char *path;
185 	int cmd;
186 	int uid;
187 	caddr_t arg;
188 };
189 #endif
190 int
191 sys_quotactl(struct thread *td, struct quotactl_args *uap)
192 {
193 	struct mount *mp;
194 	struct nameidata nd;
195 	int error;
196 
197 	AUDIT_ARG_CMD(uap->cmd);
198 	AUDIT_ARG_UID(uap->uid);
199 	if (!prison_allow(td->td_ucred, PR_ALLOW_QUOTAS))
200 		return (EPERM);
201 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
202 	    uap->path, td);
203 	if ((error = namei(&nd)) != 0)
204 		return (error);
205 	NDFREE(&nd, NDF_ONLY_PNBUF);
206 	mp = nd.ni_vp->v_mount;
207 	vfs_ref(mp);
208 	vput(nd.ni_vp);
209 	error = vfs_busy(mp, 0);
210 	if (error != 0) {
211 		vfs_rel(mp);
212 		return (error);
213 	}
214 	error = VFS_QUOTACTL(mp, uap->cmd, uap->uid, uap->arg);
215 
216 	/*
217 	 * Since quota on operation typically needs to open quota
218 	 * file, the Q_QUOTAON handler needs to unbusy the mount point
219 	 * before calling into namei.  Otherwise, unmount might be
220 	 * started between two vfs_busy() invocations (first is our,
221 	 * second is from mount point cross-walk code in lookup()),
222 	 * causing deadlock.
223 	 *
224 	 * Require that Q_QUOTAON handles the vfs_busy() reference on
225 	 * its own, always returning with ubusied mount point.
226 	 */
227 	if ((uap->cmd >> SUBCMDSHIFT) != Q_QUOTAON &&
228 	    (uap->cmd >> SUBCMDSHIFT) != Q_QUOTAOFF)
229 		vfs_unbusy(mp);
230 	vfs_rel(mp);
231 	return (error);
232 }
233 
234 /*
235  * Used by statfs conversion routines to scale the block size up if
236  * necessary so that all of the block counts are <= 'max_size'.  Note
237  * that 'max_size' should be a bitmask, i.e. 2^n - 1 for some non-zero
238  * value of 'n'.
239  */
240 void
241 statfs_scale_blocks(struct statfs *sf, long max_size)
242 {
243 	uint64_t count;
244 	int shift;
245 
246 	KASSERT(powerof2(max_size + 1), ("%s: invalid max_size", __func__));
247 
248 	/*
249 	 * Attempt to scale the block counts to give a more accurate
250 	 * overview to userland of the ratio of free space to used
251 	 * space.  To do this, find the largest block count and compute
252 	 * a divisor that lets it fit into a signed integer <= max_size.
253 	 */
254 	if (sf->f_bavail < 0)
255 		count = -sf->f_bavail;
256 	else
257 		count = sf->f_bavail;
258 	count = MAX(sf->f_blocks, MAX(sf->f_bfree, count));
259 	if (count <= max_size)
260 		return;
261 
262 	count >>= flsl(max_size);
263 	shift = 0;
264 	while (count > 0) {
265 		shift++;
266 		count >>=1;
267 	}
268 
269 	sf->f_bsize <<= shift;
270 	sf->f_blocks >>= shift;
271 	sf->f_bfree >>= shift;
272 	sf->f_bavail >>= shift;
273 }
274 
275 static int
276 kern_do_statfs(struct thread *td, struct mount *mp, struct statfs *buf)
277 {
278 	int error;
279 
280 	if (mp == NULL)
281 		return (EBADF);
282 	error = vfs_busy(mp, 0);
283 	vfs_rel(mp);
284 	if (error != 0)
285 		return (error);
286 #ifdef MAC
287 	error = mac_mount_check_stat(td->td_ucred, mp);
288 	if (error != 0)
289 		goto out;
290 #endif
291 	error = VFS_STATFS(mp, buf);
292 	if (error != 0)
293 		goto out;
294 	if (priv_check_cred_vfs_generation(td->td_ucred)) {
295 		buf->f_fsid.val[0] = buf->f_fsid.val[1] = 0;
296 		prison_enforce_statfs(td->td_ucred, mp, buf);
297 	}
298 out:
299 	vfs_unbusy(mp);
300 	return (error);
301 }
302 
303 /*
304  * Get filesystem statistics.
305  */
306 #ifndef _SYS_SYSPROTO_H_
307 struct statfs_args {
308 	char *path;
309 	struct statfs *buf;
310 };
311 #endif
312 int
313 sys_statfs(struct thread *td, struct statfs_args *uap)
314 {
315 	struct statfs *sfp;
316 	int error;
317 
318 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
319 	error = kern_statfs(td, uap->path, UIO_USERSPACE, sfp);
320 	if (error == 0)
321 		error = copyout(sfp, uap->buf, sizeof(struct statfs));
322 	free(sfp, M_STATFS);
323 	return (error);
324 }
325 
326 int
327 kern_statfs(struct thread *td, const char *path, enum uio_seg pathseg,
328     struct statfs *buf)
329 {
330 	struct mount *mp;
331 	struct nameidata nd;
332 	int error;
333 
334 	NDINIT(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, td);
335 	error = namei(&nd);
336 	if (error != 0)
337 		return (error);
338 	mp = vfs_ref_from_vp(nd.ni_vp);
339 	NDFREE_NOTHING(&nd);
340 	vrele(nd.ni_vp);
341 	return (kern_do_statfs(td, mp, buf));
342 }
343 
344 /*
345  * Get filesystem statistics.
346  */
347 #ifndef _SYS_SYSPROTO_H_
348 struct fstatfs_args {
349 	int fd;
350 	struct statfs *buf;
351 };
352 #endif
353 int
354 sys_fstatfs(struct thread *td, struct fstatfs_args *uap)
355 {
356 	struct statfs *sfp;
357 	int error;
358 
359 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
360 	error = kern_fstatfs(td, uap->fd, sfp);
361 	if (error == 0)
362 		error = copyout(sfp, uap->buf, sizeof(struct statfs));
363 	free(sfp, M_STATFS);
364 	return (error);
365 }
366 
367 int
368 kern_fstatfs(struct thread *td, int fd, struct statfs *buf)
369 {
370 	struct file *fp;
371 	struct mount *mp;
372 	struct vnode *vp;
373 	int error;
374 
375 	AUDIT_ARG_FD(fd);
376 	error = getvnode(td, fd, &cap_fstatfs_rights, &fp);
377 	if (error != 0)
378 		return (error);
379 	vp = fp->f_vnode;
380 #ifdef AUDIT
381 	if (AUDITING_TD(td)) {
382 		vn_lock(vp, LK_SHARED | LK_RETRY);
383 		AUDIT_ARG_VNODE1(vp);
384 		VOP_UNLOCK(vp);
385 	}
386 #endif
387 	mp = vfs_ref_from_vp(vp);
388 	fdrop(fp, td);
389 	return (kern_do_statfs(td, mp, buf));
390 }
391 
392 /*
393  * Get statistics on all filesystems.
394  */
395 #ifndef _SYS_SYSPROTO_H_
396 struct getfsstat_args {
397 	struct statfs *buf;
398 	long bufsize;
399 	int mode;
400 };
401 #endif
402 int
403 sys_getfsstat(struct thread *td, struct getfsstat_args *uap)
404 {
405 	size_t count;
406 	int error;
407 
408 	if (uap->bufsize < 0 || uap->bufsize > SIZE_MAX)
409 		return (EINVAL);
410 	error = kern_getfsstat(td, &uap->buf, uap->bufsize, &count,
411 	    UIO_USERSPACE, uap->mode);
412 	if (error == 0)
413 		td->td_retval[0] = count;
414 	return (error);
415 }
416 
417 /*
418  * If (bufsize > 0 && bufseg == UIO_SYSSPACE)
419  *	The caller is responsible for freeing memory which will be allocated
420  *	in '*buf'.
421  */
422 int
423 kern_getfsstat(struct thread *td, struct statfs **buf, size_t bufsize,
424     size_t *countp, enum uio_seg bufseg, int mode)
425 {
426 	struct mount *mp, *nmp;
427 	struct statfs *sfsp, *sp, *sptmp, *tofree;
428 	size_t count, maxcount;
429 	int error;
430 
431 	switch (mode) {
432 	case MNT_WAIT:
433 	case MNT_NOWAIT:
434 		break;
435 	default:
436 		if (bufseg == UIO_SYSSPACE)
437 			*buf = NULL;
438 		return (EINVAL);
439 	}
440 restart:
441 	maxcount = bufsize / sizeof(struct statfs);
442 	if (bufsize == 0) {
443 		sfsp = NULL;
444 		tofree = NULL;
445 	} else if (bufseg == UIO_USERSPACE) {
446 		sfsp = *buf;
447 		tofree = NULL;
448 	} else /* if (bufseg == UIO_SYSSPACE) */ {
449 		count = 0;
450 		mtx_lock(&mountlist_mtx);
451 		TAILQ_FOREACH(mp, &mountlist, mnt_list) {
452 			count++;
453 		}
454 		mtx_unlock(&mountlist_mtx);
455 		if (maxcount > count)
456 			maxcount = count;
457 		tofree = sfsp = *buf = malloc(maxcount * sizeof(struct statfs),
458 		    M_STATFS, M_WAITOK);
459 	}
460 
461 	count = 0;
462 
463 	/*
464 	 * If there is no target buffer they only want the count.
465 	 *
466 	 * This could be TAILQ_FOREACH but it is open-coded to match the original
467 	 * code below.
468 	 */
469 	if (sfsp == NULL) {
470 		mtx_lock(&mountlist_mtx);
471 		for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
472 			if (prison_canseemount(td->td_ucred, mp) != 0) {
473 				nmp = TAILQ_NEXT(mp, mnt_list);
474 				continue;
475 			}
476 #ifdef MAC
477 			if (mac_mount_check_stat(td->td_ucred, mp) != 0) {
478 				nmp = TAILQ_NEXT(mp, mnt_list);
479 				continue;
480 			}
481 #endif
482 			count++;
483 			nmp = TAILQ_NEXT(mp, mnt_list);
484 		}
485 		mtx_unlock(&mountlist_mtx);
486 		*countp = count;
487 		return (0);
488 	}
489 
490 	/*
491 	 * They want the entire thing.
492 	 *
493 	 * Short-circuit the corner case of no room for anything, avoids
494 	 * relocking below.
495 	 */
496 	if (maxcount < 1) {
497 		goto out;
498 	}
499 
500 	mtx_lock(&mountlist_mtx);
501 	for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
502 		if (prison_canseemount(td->td_ucred, mp) != 0) {
503 			nmp = TAILQ_NEXT(mp, mnt_list);
504 			continue;
505 		}
506 #ifdef MAC
507 		if (mac_mount_check_stat(td->td_ucred, mp) != 0) {
508 			nmp = TAILQ_NEXT(mp, mnt_list);
509 			continue;
510 		}
511 #endif
512 		if (mode == MNT_WAIT) {
513 			if (vfs_busy(mp, MBF_MNTLSTLOCK) != 0) {
514 				/*
515 				 * If vfs_busy() failed, and MBF_NOWAIT
516 				 * wasn't passed, then the mp is gone.
517 				 * Furthermore, because of MBF_MNTLSTLOCK,
518 				 * the mountlist_mtx was dropped.  We have
519 				 * no other choice than to start over.
520 				 */
521 				mtx_unlock(&mountlist_mtx);
522 				free(tofree, M_STATFS);
523 				goto restart;
524 			}
525 		} else {
526 			if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK) != 0) {
527 				nmp = TAILQ_NEXT(mp, mnt_list);
528 				continue;
529 			}
530 		}
531 		sp = &mp->mnt_stat;
532 		/*
533 		 * If MNT_NOWAIT is specified, do not refresh
534 		 * the fsstat cache.
535 		 */
536 		if (mode != MNT_NOWAIT) {
537 			error = VFS_STATFS(mp, sp);
538 			if (error != 0) {
539 				mtx_lock(&mountlist_mtx);
540 				nmp = TAILQ_NEXT(mp, mnt_list);
541 				vfs_unbusy(mp);
542 				continue;
543 			}
544 		}
545 		if (priv_check_cred_vfs_generation(td->td_ucred)) {
546 			sptmp = malloc(sizeof(struct statfs), M_STATFS,
547 			    M_WAITOK);
548 			*sptmp = *sp;
549 			sptmp->f_fsid.val[0] = sptmp->f_fsid.val[1] = 0;
550 			prison_enforce_statfs(td->td_ucred, mp, sptmp);
551 			sp = sptmp;
552 		} else
553 			sptmp = NULL;
554 		if (bufseg == UIO_SYSSPACE) {
555 			bcopy(sp, sfsp, sizeof(*sp));
556 			free(sptmp, M_STATFS);
557 		} else /* if (bufseg == UIO_USERSPACE) */ {
558 			error = copyout(sp, sfsp, sizeof(*sp));
559 			free(sptmp, M_STATFS);
560 			if (error != 0) {
561 				vfs_unbusy(mp);
562 				return (error);
563 			}
564 		}
565 		sfsp++;
566 		count++;
567 
568 		if (count == maxcount) {
569 			vfs_unbusy(mp);
570 			goto out;
571 		}
572 
573 		mtx_lock(&mountlist_mtx);
574 		nmp = TAILQ_NEXT(mp, mnt_list);
575 		vfs_unbusy(mp);
576 	}
577 	mtx_unlock(&mountlist_mtx);
578 out:
579 	*countp = count;
580 	return (0);
581 }
582 
583 #ifdef COMPAT_FREEBSD4
584 /*
585  * Get old format filesystem statistics.
586  */
587 static void freebsd4_cvtstatfs(struct statfs *, struct ostatfs *);
588 
589 #ifndef _SYS_SYSPROTO_H_
590 struct freebsd4_statfs_args {
591 	char *path;
592 	struct ostatfs *buf;
593 };
594 #endif
595 int
596 freebsd4_statfs(struct thread *td, struct freebsd4_statfs_args *uap)
597 {
598 	struct ostatfs osb;
599 	struct statfs *sfp;
600 	int error;
601 
602 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
603 	error = kern_statfs(td, uap->path, UIO_USERSPACE, sfp);
604 	if (error == 0) {
605 		freebsd4_cvtstatfs(sfp, &osb);
606 		error = copyout(&osb, uap->buf, sizeof(osb));
607 	}
608 	free(sfp, M_STATFS);
609 	return (error);
610 }
611 
612 /*
613  * Get filesystem statistics.
614  */
615 #ifndef _SYS_SYSPROTO_H_
616 struct freebsd4_fstatfs_args {
617 	int fd;
618 	struct ostatfs *buf;
619 };
620 #endif
621 int
622 freebsd4_fstatfs(struct thread *td, struct freebsd4_fstatfs_args *uap)
623 {
624 	struct ostatfs osb;
625 	struct statfs *sfp;
626 	int error;
627 
628 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
629 	error = kern_fstatfs(td, uap->fd, sfp);
630 	if (error == 0) {
631 		freebsd4_cvtstatfs(sfp, &osb);
632 		error = copyout(&osb, uap->buf, sizeof(osb));
633 	}
634 	free(sfp, M_STATFS);
635 	return (error);
636 }
637 
638 /*
639  * Get statistics on all filesystems.
640  */
641 #ifndef _SYS_SYSPROTO_H_
642 struct freebsd4_getfsstat_args {
643 	struct ostatfs *buf;
644 	long bufsize;
645 	int mode;
646 };
647 #endif
648 int
649 freebsd4_getfsstat(struct thread *td, struct freebsd4_getfsstat_args *uap)
650 {
651 	struct statfs *buf, *sp;
652 	struct ostatfs osb;
653 	size_t count, size;
654 	int error;
655 
656 	if (uap->bufsize < 0)
657 		return (EINVAL);
658 	count = uap->bufsize / sizeof(struct ostatfs);
659 	if (count > SIZE_MAX / sizeof(struct statfs))
660 		return (EINVAL);
661 	size = count * sizeof(struct statfs);
662 	error = kern_getfsstat(td, &buf, size, &count, UIO_SYSSPACE,
663 	    uap->mode);
664 	if (error == 0)
665 		td->td_retval[0] = count;
666 	if (size != 0) {
667 		sp = buf;
668 		while (count != 0 && error == 0) {
669 			freebsd4_cvtstatfs(sp, &osb);
670 			error = copyout(&osb, uap->buf, sizeof(osb));
671 			sp++;
672 			uap->buf++;
673 			count--;
674 		}
675 		free(buf, M_STATFS);
676 	}
677 	return (error);
678 }
679 
680 /*
681  * Implement fstatfs() for (NFS) file handles.
682  */
683 #ifndef _SYS_SYSPROTO_H_
684 struct freebsd4_fhstatfs_args {
685 	struct fhandle *u_fhp;
686 	struct ostatfs *buf;
687 };
688 #endif
689 int
690 freebsd4_fhstatfs(struct thread *td, struct freebsd4_fhstatfs_args *uap)
691 {
692 	struct ostatfs osb;
693 	struct statfs *sfp;
694 	fhandle_t fh;
695 	int error;
696 
697 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
698 	if (error != 0)
699 		return (error);
700 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
701 	error = kern_fhstatfs(td, fh, sfp);
702 	if (error == 0) {
703 		freebsd4_cvtstatfs(sfp, &osb);
704 		error = copyout(&osb, uap->buf, sizeof(osb));
705 	}
706 	free(sfp, M_STATFS);
707 	return (error);
708 }
709 
710 /*
711  * Convert a new format statfs structure to an old format statfs structure.
712  */
713 static void
714 freebsd4_cvtstatfs(struct statfs *nsp, struct ostatfs *osp)
715 {
716 
717 	statfs_scale_blocks(nsp, LONG_MAX);
718 	bzero(osp, sizeof(*osp));
719 	osp->f_bsize = nsp->f_bsize;
720 	osp->f_iosize = MIN(nsp->f_iosize, LONG_MAX);
721 	osp->f_blocks = nsp->f_blocks;
722 	osp->f_bfree = nsp->f_bfree;
723 	osp->f_bavail = nsp->f_bavail;
724 	osp->f_files = MIN(nsp->f_files, LONG_MAX);
725 	osp->f_ffree = MIN(nsp->f_ffree, LONG_MAX);
726 	osp->f_owner = nsp->f_owner;
727 	osp->f_type = nsp->f_type;
728 	osp->f_flags = nsp->f_flags;
729 	osp->f_syncwrites = MIN(nsp->f_syncwrites, LONG_MAX);
730 	osp->f_asyncwrites = MIN(nsp->f_asyncwrites, LONG_MAX);
731 	osp->f_syncreads = MIN(nsp->f_syncreads, LONG_MAX);
732 	osp->f_asyncreads = MIN(nsp->f_asyncreads, LONG_MAX);
733 	strlcpy(osp->f_fstypename, nsp->f_fstypename,
734 	    MIN(MFSNAMELEN, OMFSNAMELEN));
735 	strlcpy(osp->f_mntonname, nsp->f_mntonname,
736 	    MIN(MNAMELEN, OMNAMELEN));
737 	strlcpy(osp->f_mntfromname, nsp->f_mntfromname,
738 	    MIN(MNAMELEN, OMNAMELEN));
739 	osp->f_fsid = nsp->f_fsid;
740 }
741 #endif /* COMPAT_FREEBSD4 */
742 
743 #if defined(COMPAT_FREEBSD11)
744 /*
745  * Get old format filesystem statistics.
746  */
747 static void freebsd11_cvtstatfs(struct statfs *, struct freebsd11_statfs *);
748 
749 int
750 freebsd11_statfs(struct thread *td, struct freebsd11_statfs_args *uap)
751 {
752 	struct freebsd11_statfs osb;
753 	struct statfs *sfp;
754 	int error;
755 
756 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
757 	error = kern_statfs(td, uap->path, UIO_USERSPACE, sfp);
758 	if (error == 0) {
759 		freebsd11_cvtstatfs(sfp, &osb);
760 		error = copyout(&osb, uap->buf, sizeof(osb));
761 	}
762 	free(sfp, M_STATFS);
763 	return (error);
764 }
765 
766 /*
767  * Get filesystem statistics.
768  */
769 int
770 freebsd11_fstatfs(struct thread *td, struct freebsd11_fstatfs_args *uap)
771 {
772 	struct freebsd11_statfs osb;
773 	struct statfs *sfp;
774 	int error;
775 
776 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
777 	error = kern_fstatfs(td, uap->fd, sfp);
778 	if (error == 0) {
779 		freebsd11_cvtstatfs(sfp, &osb);
780 		error = copyout(&osb, uap->buf, sizeof(osb));
781 	}
782 	free(sfp, M_STATFS);
783 	return (error);
784 }
785 
786 /*
787  * Get statistics on all filesystems.
788  */
789 int
790 freebsd11_getfsstat(struct thread *td, struct freebsd11_getfsstat_args *uap)
791 {
792 	struct freebsd11_statfs osb;
793 	struct statfs *buf, *sp;
794 	size_t count, size;
795 	int error;
796 
797 	count = uap->bufsize / sizeof(struct ostatfs);
798 	size = count * sizeof(struct statfs);
799 	error = kern_getfsstat(td, &buf, size, &count, UIO_SYSSPACE,
800 	    uap->mode);
801 	if (error == 0)
802 		td->td_retval[0] = count;
803 	if (size > 0) {
804 		sp = buf;
805 		while (count > 0 && error == 0) {
806 			freebsd11_cvtstatfs(sp, &osb);
807 			error = copyout(&osb, uap->buf, sizeof(osb));
808 			sp++;
809 			uap->buf++;
810 			count--;
811 		}
812 		free(buf, M_STATFS);
813 	}
814 	return (error);
815 }
816 
817 /*
818  * Implement fstatfs() for (NFS) file handles.
819  */
820 int
821 freebsd11_fhstatfs(struct thread *td, struct freebsd11_fhstatfs_args *uap)
822 {
823 	struct freebsd11_statfs osb;
824 	struct statfs *sfp;
825 	fhandle_t fh;
826 	int error;
827 
828 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
829 	if (error)
830 		return (error);
831 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
832 	error = kern_fhstatfs(td, fh, sfp);
833 	if (error == 0) {
834 		freebsd11_cvtstatfs(sfp, &osb);
835 		error = copyout(&osb, uap->buf, sizeof(osb));
836 	}
837 	free(sfp, M_STATFS);
838 	return (error);
839 }
840 
841 /*
842  * Convert a new format statfs structure to an old format statfs structure.
843  */
844 static void
845 freebsd11_cvtstatfs(struct statfs *nsp, struct freebsd11_statfs *osp)
846 {
847 
848 	bzero(osp, sizeof(*osp));
849 	osp->f_version = FREEBSD11_STATFS_VERSION;
850 	osp->f_type = nsp->f_type;
851 	osp->f_flags = nsp->f_flags;
852 	osp->f_bsize = nsp->f_bsize;
853 	osp->f_iosize = nsp->f_iosize;
854 	osp->f_blocks = nsp->f_blocks;
855 	osp->f_bfree = nsp->f_bfree;
856 	osp->f_bavail = nsp->f_bavail;
857 	osp->f_files = nsp->f_files;
858 	osp->f_ffree = nsp->f_ffree;
859 	osp->f_syncwrites = nsp->f_syncwrites;
860 	osp->f_asyncwrites = nsp->f_asyncwrites;
861 	osp->f_syncreads = nsp->f_syncreads;
862 	osp->f_asyncreads = nsp->f_asyncreads;
863 	osp->f_namemax = nsp->f_namemax;
864 	osp->f_owner = nsp->f_owner;
865 	osp->f_fsid = nsp->f_fsid;
866 	strlcpy(osp->f_fstypename, nsp->f_fstypename,
867 	    MIN(MFSNAMELEN, sizeof(osp->f_fstypename)));
868 	strlcpy(osp->f_mntonname, nsp->f_mntonname,
869 	    MIN(MNAMELEN, sizeof(osp->f_mntonname)));
870 	strlcpy(osp->f_mntfromname, nsp->f_mntfromname,
871 	    MIN(MNAMELEN, sizeof(osp->f_mntfromname)));
872 }
873 #endif /* COMPAT_FREEBSD11 */
874 
875 /*
876  * Change current working directory to a given file descriptor.
877  */
878 #ifndef _SYS_SYSPROTO_H_
879 struct fchdir_args {
880 	int	fd;
881 };
882 #endif
883 int
884 sys_fchdir(struct thread *td, struct fchdir_args *uap)
885 {
886 	struct vnode *vp, *tdp;
887 	struct mount *mp;
888 	struct file *fp;
889 	int error;
890 
891 	AUDIT_ARG_FD(uap->fd);
892 	error = getvnode(td, uap->fd, &cap_fchdir_rights,
893 	    &fp);
894 	if (error != 0)
895 		return (error);
896 	vp = fp->f_vnode;
897 	vrefact(vp);
898 	fdrop(fp, td);
899 	vn_lock(vp, LK_SHARED | LK_RETRY);
900 	AUDIT_ARG_VNODE1(vp);
901 	error = change_dir(vp, td);
902 	while (!error && (mp = vp->v_mountedhere) != NULL) {
903 		if (vfs_busy(mp, 0))
904 			continue;
905 		error = VFS_ROOT(mp, LK_SHARED, &tdp);
906 		vfs_unbusy(mp);
907 		if (error != 0)
908 			break;
909 		vput(vp);
910 		vp = tdp;
911 	}
912 	if (error != 0) {
913 		vput(vp);
914 		return (error);
915 	}
916 	VOP_UNLOCK(vp);
917 	pwd_chdir(td, vp);
918 	return (0);
919 }
920 
921 /*
922  * Change current working directory (``.'').
923  */
924 #ifndef _SYS_SYSPROTO_H_
925 struct chdir_args {
926 	char	*path;
927 };
928 #endif
929 int
930 sys_chdir(struct thread *td, struct chdir_args *uap)
931 {
932 
933 	return (kern_chdir(td, uap->path, UIO_USERSPACE));
934 }
935 
936 int
937 kern_chdir(struct thread *td, const char *path, enum uio_seg pathseg)
938 {
939 	struct nameidata nd;
940 	int error;
941 
942 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
943 	    pathseg, path, td);
944 	if ((error = namei(&nd)) != 0)
945 		return (error);
946 	if ((error = change_dir(nd.ni_vp, td)) != 0) {
947 		vput(nd.ni_vp);
948 		NDFREE_NOTHING(&nd);
949 		return (error);
950 	}
951 	VOP_UNLOCK(nd.ni_vp);
952 	NDFREE_NOTHING(&nd);
953 	pwd_chdir(td, nd.ni_vp);
954 	return (0);
955 }
956 
957 /*
958  * Change notion of root (``/'') directory.
959  */
960 #ifndef _SYS_SYSPROTO_H_
961 struct chroot_args {
962 	char	*path;
963 };
964 #endif
965 int
966 sys_chroot(struct thread *td, struct chroot_args *uap)
967 {
968 	struct nameidata nd;
969 	int error;
970 
971 	error = priv_check(td, PRIV_VFS_CHROOT);
972 	if (error != 0)
973 		return (error);
974 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
975 	    UIO_USERSPACE, uap->path, td);
976 	error = namei(&nd);
977 	if (error != 0)
978 		goto error;
979 	error = change_dir(nd.ni_vp, td);
980 	if (error != 0)
981 		goto e_vunlock;
982 #ifdef MAC
983 	error = mac_vnode_check_chroot(td->td_ucred, nd.ni_vp);
984 	if (error != 0)
985 		goto e_vunlock;
986 #endif
987 	VOP_UNLOCK(nd.ni_vp);
988 	error = pwd_chroot(td, nd.ni_vp);
989 	vrele(nd.ni_vp);
990 	NDFREE_NOTHING(&nd);
991 	return (error);
992 e_vunlock:
993 	vput(nd.ni_vp);
994 error:
995 	NDFREE_NOTHING(&nd);
996 	return (error);
997 }
998 
999 /*
1000  * Common routine for chroot and chdir.  Callers must provide a locked vnode
1001  * instance.
1002  */
1003 int
1004 change_dir(struct vnode *vp, struct thread *td)
1005 {
1006 #ifdef MAC
1007 	int error;
1008 #endif
1009 
1010 	ASSERT_VOP_LOCKED(vp, "change_dir(): vp not locked");
1011 	if (vp->v_type != VDIR)
1012 		return (ENOTDIR);
1013 #ifdef MAC
1014 	error = mac_vnode_check_chdir(td->td_ucred, vp);
1015 	if (error != 0)
1016 		return (error);
1017 #endif
1018 	return (VOP_ACCESS(vp, VEXEC, td->td_ucred, td));
1019 }
1020 
1021 static __inline void
1022 flags_to_rights(int flags, cap_rights_t *rightsp)
1023 {
1024 
1025 	if (flags & O_EXEC) {
1026 		cap_rights_set_one(rightsp, CAP_FEXECVE);
1027 	} else {
1028 		switch ((flags & O_ACCMODE)) {
1029 		case O_RDONLY:
1030 			cap_rights_set_one(rightsp, CAP_READ);
1031 			break;
1032 		case O_RDWR:
1033 			cap_rights_set_one(rightsp, CAP_READ);
1034 			/* FALLTHROUGH */
1035 		case O_WRONLY:
1036 			cap_rights_set_one(rightsp, CAP_WRITE);
1037 			if (!(flags & (O_APPEND | O_TRUNC)))
1038 				cap_rights_set_one(rightsp, CAP_SEEK);
1039 			break;
1040 		}
1041 	}
1042 
1043 	if (flags & O_CREAT)
1044 		cap_rights_set_one(rightsp, CAP_CREATE);
1045 
1046 	if (flags & O_TRUNC)
1047 		cap_rights_set_one(rightsp, CAP_FTRUNCATE);
1048 
1049 	if (flags & (O_SYNC | O_FSYNC))
1050 		cap_rights_set_one(rightsp, CAP_FSYNC);
1051 
1052 	if (flags & (O_EXLOCK | O_SHLOCK))
1053 		cap_rights_set_one(rightsp, CAP_FLOCK);
1054 }
1055 
1056 /*
1057  * Check permissions, allocate an open file structure, and call the device
1058  * open routine if any.
1059  */
1060 #ifndef _SYS_SYSPROTO_H_
1061 struct open_args {
1062 	char	*path;
1063 	int	flags;
1064 	int	mode;
1065 };
1066 #endif
1067 int
1068 sys_open(struct thread *td, struct open_args *uap)
1069 {
1070 
1071 	return (kern_openat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1072 	    uap->flags, uap->mode));
1073 }
1074 
1075 #ifndef _SYS_SYSPROTO_H_
1076 struct openat_args {
1077 	int	fd;
1078 	char	*path;
1079 	int	flag;
1080 	int	mode;
1081 };
1082 #endif
1083 int
1084 sys_openat(struct thread *td, struct openat_args *uap)
1085 {
1086 
1087 	AUDIT_ARG_FD(uap->fd);
1088 	return (kern_openat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
1089 	    uap->mode));
1090 }
1091 
1092 int
1093 kern_openat(struct thread *td, int fd, const char *path, enum uio_seg pathseg,
1094     int flags, int mode)
1095 {
1096 	struct proc *p = td->td_proc;
1097 	struct filedesc *fdp;
1098 	struct pwddesc *pdp;
1099 	struct file *fp;
1100 	struct vnode *vp;
1101 	struct nameidata nd;
1102 	cap_rights_t rights;
1103 	int cmode, error, indx;
1104 
1105 	indx = -1;
1106 	fdp = p->p_fd;
1107 	pdp = p->p_pd;
1108 
1109 	AUDIT_ARG_FFLAGS(flags);
1110 	AUDIT_ARG_MODE(mode);
1111 	cap_rights_init_one(&rights, CAP_LOOKUP);
1112 	flags_to_rights(flags, &rights);
1113 	/*
1114 	 * Only one of the O_EXEC, O_RDONLY, O_WRONLY and O_RDWR flags
1115 	 * may be specified.
1116 	 */
1117 	if (flags & O_EXEC) {
1118 		if (flags & O_ACCMODE)
1119 			return (EINVAL);
1120 	} else if ((flags & O_ACCMODE) == O_ACCMODE) {
1121 		return (EINVAL);
1122 	} else {
1123 		flags = FFLAGS(flags);
1124 	}
1125 
1126 	/*
1127 	 * Allocate a file structure. The descriptor to reference it
1128 	 * is allocated and used by finstall_refed() below.
1129 	 */
1130 	error = falloc_noinstall(td, &fp);
1131 	if (error != 0)
1132 		return (error);
1133 	/* Set the flags early so the finit in devfs can pick them up. */
1134 	fp->f_flag = flags & FMASK;
1135 	cmode = ((mode & ~pdp->pd_cmask) & ALLPERMS) & ~S_ISTXT;
1136 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, fd,
1137 	    &rights, td);
1138 	td->td_dupfd = -1;		/* XXX check for fdopen */
1139 	error = vn_open(&nd, &flags, cmode, fp);
1140 	if (error != 0) {
1141 		/*
1142 		 * If the vn_open replaced the method vector, something
1143 		 * wonderous happened deep below and we just pass it up
1144 		 * pretending we know what we do.
1145 		 */
1146 		if (error == ENXIO && fp->f_ops != &badfileops)
1147 			goto success;
1148 
1149 		/*
1150 		 * Handle special fdopen() case. bleh.
1151 		 *
1152 		 * Don't do this for relative (capability) lookups; we don't
1153 		 * understand exactly what would happen, and we don't think
1154 		 * that it ever should.
1155 		 */
1156 		if ((nd.ni_resflags & NIRES_STRICTREL) == 0 &&
1157 		    (error == ENODEV || error == ENXIO) &&
1158 		    td->td_dupfd >= 0) {
1159 			error = dupfdopen(td, fdp, td->td_dupfd, flags, error,
1160 			    &indx);
1161 			if (error == 0)
1162 				goto success;
1163 		}
1164 
1165 		goto bad;
1166 	}
1167 	td->td_dupfd = 0;
1168 	NDFREE(&nd, NDF_ONLY_PNBUF);
1169 	vp = nd.ni_vp;
1170 
1171 	/*
1172 	 * Store the vnode, for any f_type. Typically, the vnode use
1173 	 * count is decremented by direct call to vn_closefile() for
1174 	 * files that switched type in the cdevsw fdopen() method.
1175 	 */
1176 	fp->f_vnode = vp;
1177 	/*
1178 	 * If the file wasn't claimed by devfs bind it to the normal
1179 	 * vnode operations here.
1180 	 */
1181 	if (fp->f_ops == &badfileops) {
1182 		KASSERT(vp->v_type != VFIFO,
1183 		    ("Unexpected fifo fp %p vp %p", fp, vp));
1184 		finit_vnode(fp, flags, NULL, &vnops);
1185 	}
1186 
1187 	VOP_UNLOCK(vp);
1188 	if (flags & O_TRUNC) {
1189 		error = fo_truncate(fp, 0, td->td_ucred, td);
1190 		if (error != 0)
1191 			goto bad;
1192 	}
1193 success:
1194 	/*
1195 	 * If we haven't already installed the FD (for dupfdopen), do so now.
1196 	 */
1197 	if (indx == -1) {
1198 		struct filecaps *fcaps;
1199 
1200 #ifdef CAPABILITIES
1201 		if ((nd.ni_resflags & NIRES_STRICTREL) != 0)
1202 			fcaps = &nd.ni_filecaps;
1203 		else
1204 #endif
1205 			fcaps = NULL;
1206 		error = finstall_refed(td, fp, &indx, flags, fcaps);
1207 		/* On success finstall_refed() consumes fcaps. */
1208 		if (error != 0) {
1209 			filecaps_free(&nd.ni_filecaps);
1210 			goto bad;
1211 		}
1212 	} else {
1213 		filecaps_free(&nd.ni_filecaps);
1214 		falloc_abort(td, fp);
1215 	}
1216 
1217 	td->td_retval[0] = indx;
1218 	return (0);
1219 bad:
1220 	KASSERT(indx == -1, ("indx=%d, should be -1", indx));
1221 	falloc_abort(td, fp);
1222 	return (error);
1223 }
1224 
1225 #ifdef COMPAT_43
1226 /*
1227  * Create a file.
1228  */
1229 #ifndef _SYS_SYSPROTO_H_
1230 struct ocreat_args {
1231 	char	*path;
1232 	int	mode;
1233 };
1234 #endif
1235 int
1236 ocreat(struct thread *td, struct ocreat_args *uap)
1237 {
1238 
1239 	return (kern_openat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1240 	    O_WRONLY | O_CREAT | O_TRUNC, uap->mode));
1241 }
1242 #endif /* COMPAT_43 */
1243 
1244 /*
1245  * Create a special file.
1246  */
1247 #ifndef _SYS_SYSPROTO_H_
1248 struct mknodat_args {
1249 	int	fd;
1250 	char	*path;
1251 	mode_t	mode;
1252 	dev_t	dev;
1253 };
1254 #endif
1255 int
1256 sys_mknodat(struct thread *td, struct mknodat_args *uap)
1257 {
1258 
1259 	return (kern_mknodat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode,
1260 	    uap->dev));
1261 }
1262 
1263 #if defined(COMPAT_FREEBSD11)
1264 int
1265 freebsd11_mknod(struct thread *td,
1266     struct freebsd11_mknod_args *uap)
1267 {
1268 
1269 	return (kern_mknodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1270 	    uap->mode, uap->dev));
1271 }
1272 
1273 int
1274 freebsd11_mknodat(struct thread *td,
1275     struct freebsd11_mknodat_args *uap)
1276 {
1277 
1278 	return (kern_mknodat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode,
1279 	    uap->dev));
1280 }
1281 #endif /* COMPAT_FREEBSD11 */
1282 
1283 int
1284 kern_mknodat(struct thread *td, int fd, const char *path, enum uio_seg pathseg,
1285     int mode, dev_t dev)
1286 {
1287 	struct vnode *vp;
1288 	struct mount *mp;
1289 	struct vattr vattr;
1290 	struct nameidata nd;
1291 	int error, whiteout = 0;
1292 
1293 	AUDIT_ARG_MODE(mode);
1294 	AUDIT_ARG_DEV(dev);
1295 	switch (mode & S_IFMT) {
1296 	case S_IFCHR:
1297 	case S_IFBLK:
1298 		error = priv_check(td, PRIV_VFS_MKNOD_DEV);
1299 		if (error == 0 && dev == VNOVAL)
1300 			error = EINVAL;
1301 		break;
1302 	case S_IFWHT:
1303 		error = priv_check(td, PRIV_VFS_MKNOD_WHT);
1304 		break;
1305 	case S_IFIFO:
1306 		if (dev == 0)
1307 			return (kern_mkfifoat(td, fd, path, pathseg, mode));
1308 		/* FALLTHROUGH */
1309 	default:
1310 		error = EINVAL;
1311 		break;
1312 	}
1313 	if (error != 0)
1314 		return (error);
1315 restart:
1316 	bwillwrite();
1317 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
1318 	    NOCACHE, pathseg, path, fd, &cap_mknodat_rights,
1319 	    td);
1320 	if ((error = namei(&nd)) != 0)
1321 		return (error);
1322 	vp = nd.ni_vp;
1323 	if (vp != NULL) {
1324 		NDFREE(&nd, NDF_ONLY_PNBUF);
1325 		if (vp == nd.ni_dvp)
1326 			vrele(nd.ni_dvp);
1327 		else
1328 			vput(nd.ni_dvp);
1329 		vrele(vp);
1330 		return (EEXIST);
1331 	} else {
1332 		VATTR_NULL(&vattr);
1333 		vattr.va_mode = (mode & ALLPERMS) &
1334 		    ~td->td_proc->p_pd->pd_cmask;
1335 		vattr.va_rdev = dev;
1336 		whiteout = 0;
1337 
1338 		switch (mode & S_IFMT) {
1339 		case S_IFCHR:
1340 			vattr.va_type = VCHR;
1341 			break;
1342 		case S_IFBLK:
1343 			vattr.va_type = VBLK;
1344 			break;
1345 		case S_IFWHT:
1346 			whiteout = 1;
1347 			break;
1348 		default:
1349 			panic("kern_mknod: invalid mode");
1350 		}
1351 	}
1352 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1353 		NDFREE(&nd, NDF_ONLY_PNBUF);
1354 		vput(nd.ni_dvp);
1355 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1356 			return (error);
1357 		goto restart;
1358 	}
1359 #ifdef MAC
1360 	if (error == 0 && !whiteout)
1361 		error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp,
1362 		    &nd.ni_cnd, &vattr);
1363 #endif
1364 	if (error == 0) {
1365 		if (whiteout)
1366 			error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE);
1367 		else {
1368 			error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp,
1369 						&nd.ni_cnd, &vattr);
1370 		}
1371 	}
1372 	VOP_VPUT_PAIR(nd.ni_dvp, error == 0 && !whiteout ? &nd.ni_vp : NULL,
1373 	    true);
1374 	vn_finished_write(mp);
1375 	NDFREE(&nd, NDF_ONLY_PNBUF);
1376 	if (error == ERELOOKUP)
1377 		goto restart;
1378 	return (error);
1379 }
1380 
1381 /*
1382  * Create a named pipe.
1383  */
1384 #ifndef _SYS_SYSPROTO_H_
1385 struct mkfifo_args {
1386 	char	*path;
1387 	int	mode;
1388 };
1389 #endif
1390 int
1391 sys_mkfifo(struct thread *td, struct mkfifo_args *uap)
1392 {
1393 
1394 	return (kern_mkfifoat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1395 	    uap->mode));
1396 }
1397 
1398 #ifndef _SYS_SYSPROTO_H_
1399 struct mkfifoat_args {
1400 	int	fd;
1401 	char	*path;
1402 	mode_t	mode;
1403 };
1404 #endif
1405 int
1406 sys_mkfifoat(struct thread *td, struct mkfifoat_args *uap)
1407 {
1408 
1409 	return (kern_mkfifoat(td, uap->fd, uap->path, UIO_USERSPACE,
1410 	    uap->mode));
1411 }
1412 
1413 int
1414 kern_mkfifoat(struct thread *td, int fd, const char *path,
1415     enum uio_seg pathseg, int mode)
1416 {
1417 	struct mount *mp;
1418 	struct vattr vattr;
1419 	struct nameidata nd;
1420 	int error;
1421 
1422 	AUDIT_ARG_MODE(mode);
1423 restart:
1424 	bwillwrite();
1425 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
1426 	    NOCACHE, pathseg, path, fd, &cap_mkfifoat_rights,
1427 	    td);
1428 	if ((error = namei(&nd)) != 0)
1429 		return (error);
1430 	if (nd.ni_vp != NULL) {
1431 		NDFREE(&nd, NDF_ONLY_PNBUF);
1432 		if (nd.ni_vp == nd.ni_dvp)
1433 			vrele(nd.ni_dvp);
1434 		else
1435 			vput(nd.ni_dvp);
1436 		vrele(nd.ni_vp);
1437 		return (EEXIST);
1438 	}
1439 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1440 		NDFREE(&nd, NDF_ONLY_PNBUF);
1441 		vput(nd.ni_dvp);
1442 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1443 			return (error);
1444 		goto restart;
1445 	}
1446 	VATTR_NULL(&vattr);
1447 	vattr.va_type = VFIFO;
1448 	vattr.va_mode = (mode & ALLPERMS) & ~td->td_proc->p_pd->pd_cmask;
1449 #ifdef MAC
1450 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1451 	    &vattr);
1452 	if (error != 0)
1453 		goto out;
1454 #endif
1455 	error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
1456 #ifdef MAC
1457 out:
1458 #endif
1459 	VOP_VPUT_PAIR(nd.ni_dvp, error == 0 ? &nd.ni_vp : NULL, true);
1460 	vn_finished_write(mp);
1461 	NDFREE(&nd, NDF_ONLY_PNBUF);
1462 	if (error == ERELOOKUP)
1463 		goto restart;
1464 	return (error);
1465 }
1466 
1467 /*
1468  * Make a hard file link.
1469  */
1470 #ifndef _SYS_SYSPROTO_H_
1471 struct link_args {
1472 	char	*path;
1473 	char	*link;
1474 };
1475 #endif
1476 int
1477 sys_link(struct thread *td, struct link_args *uap)
1478 {
1479 
1480 	return (kern_linkat(td, AT_FDCWD, AT_FDCWD, uap->path, uap->link,
1481 	    UIO_USERSPACE, FOLLOW));
1482 }
1483 
1484 #ifndef _SYS_SYSPROTO_H_
1485 struct linkat_args {
1486 	int	fd1;
1487 	char	*path1;
1488 	int	fd2;
1489 	char	*path2;
1490 	int	flag;
1491 };
1492 #endif
1493 int
1494 sys_linkat(struct thread *td, struct linkat_args *uap)
1495 {
1496 	int flag;
1497 
1498 	flag = uap->flag;
1499 	if ((flag & ~(AT_SYMLINK_FOLLOW | AT_RESOLVE_BENEATH)) != 0)
1500 		return (EINVAL);
1501 
1502 	return (kern_linkat(td, uap->fd1, uap->fd2, uap->path1, uap->path2,
1503 	    UIO_USERSPACE, at2cnpflags(flag, AT_SYMLINK_FOLLOW |
1504 	    AT_RESOLVE_BENEATH)));
1505 }
1506 
1507 int hardlink_check_uid = 0;
1508 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_uid, CTLFLAG_RW,
1509     &hardlink_check_uid, 0,
1510     "Unprivileged processes cannot create hard links to files owned by other "
1511     "users");
1512 static int hardlink_check_gid = 0;
1513 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_gid, CTLFLAG_RW,
1514     &hardlink_check_gid, 0,
1515     "Unprivileged processes cannot create hard links to files owned by other "
1516     "groups");
1517 
1518 static int
1519 can_hardlink(struct vnode *vp, struct ucred *cred)
1520 {
1521 	struct vattr va;
1522 	int error;
1523 
1524 	if (!hardlink_check_uid && !hardlink_check_gid)
1525 		return (0);
1526 
1527 	error = VOP_GETATTR(vp, &va, cred);
1528 	if (error != 0)
1529 		return (error);
1530 
1531 	if (hardlink_check_uid && cred->cr_uid != va.va_uid) {
1532 		error = priv_check_cred(cred, PRIV_VFS_LINK);
1533 		if (error != 0)
1534 			return (error);
1535 	}
1536 
1537 	if (hardlink_check_gid && !groupmember(va.va_gid, cred)) {
1538 		error = priv_check_cred(cred, PRIV_VFS_LINK);
1539 		if (error != 0)
1540 			return (error);
1541 	}
1542 
1543 	return (0);
1544 }
1545 
1546 int
1547 kern_linkat(struct thread *td, int fd1, int fd2, const char *path1,
1548     const char *path2, enum uio_seg segflag, int follow)
1549 {
1550 	struct nameidata nd;
1551 	int error;
1552 
1553 	do {
1554 		bwillwrite();
1555 		NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, segflag,
1556 		    path1, fd1, &cap_linkat_source_rights, td);
1557 		if ((error = namei(&nd)) != 0)
1558 			return (error);
1559 		NDFREE(&nd, NDF_ONLY_PNBUF);
1560 		error = kern_linkat_vp(td, nd.ni_vp, fd2, path2, segflag);
1561 	} while (error ==  EAGAIN || error == ERELOOKUP);
1562 	return (error);
1563 }
1564 
1565 static int
1566 kern_linkat_vp(struct thread *td, struct vnode *vp, int fd, const char *path,
1567     enum uio_seg segflag)
1568 {
1569 	struct nameidata nd;
1570 	struct mount *mp;
1571 	int error;
1572 
1573 	if (vp->v_type == VDIR) {
1574 		vrele(vp);
1575 		return (EPERM);		/* POSIX */
1576 	}
1577 	NDINIT_ATRIGHTS(&nd, CREATE,
1578 	    LOCKPARENT | SAVENAME | AUDITVNODE2 | NOCACHE, segflag, path, fd,
1579 	    &cap_linkat_target_rights, td);
1580 	if ((error = namei(&nd)) == 0) {
1581 		if (nd.ni_vp != NULL) {
1582 			NDFREE(&nd, NDF_ONLY_PNBUF);
1583 			if (nd.ni_dvp == nd.ni_vp)
1584 				vrele(nd.ni_dvp);
1585 			else
1586 				vput(nd.ni_dvp);
1587 			vrele(nd.ni_vp);
1588 			vrele(vp);
1589 			return (EEXIST);
1590 		} else if (nd.ni_dvp->v_mount != vp->v_mount) {
1591 			/*
1592 			 * Cross-device link.  No need to recheck
1593 			 * vp->v_type, since it cannot change, except
1594 			 * to VBAD.
1595 			 */
1596 			NDFREE(&nd, NDF_ONLY_PNBUF);
1597 			vput(nd.ni_dvp);
1598 			vrele(vp);
1599 			return (EXDEV);
1600 		} else if ((error = vn_lock(vp, LK_EXCLUSIVE)) == 0) {
1601 			error = can_hardlink(vp, td->td_ucred);
1602 #ifdef MAC
1603 			if (error == 0)
1604 				error = mac_vnode_check_link(td->td_ucred,
1605 				    nd.ni_dvp, vp, &nd.ni_cnd);
1606 #endif
1607 			if (error != 0) {
1608 				vput(vp);
1609 				vput(nd.ni_dvp);
1610 				NDFREE(&nd, NDF_ONLY_PNBUF);
1611 				return (error);
1612 			}
1613 			error = vn_start_write(vp, &mp, V_NOWAIT);
1614 			if (error != 0) {
1615 				vput(vp);
1616 				vput(nd.ni_dvp);
1617 				NDFREE(&nd, NDF_ONLY_PNBUF);
1618 				error = vn_start_write(NULL, &mp,
1619 				    V_XSLEEP | PCATCH);
1620 				if (error != 0)
1621 					return (error);
1622 				return (EAGAIN);
1623 			}
1624 			error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
1625 			VOP_VPUT_PAIR(nd.ni_dvp, &vp, true);
1626 			vn_finished_write(mp);
1627 			NDFREE(&nd, NDF_ONLY_PNBUF);
1628 			vp = NULL;
1629 		} else {
1630 			vput(nd.ni_dvp);
1631 			NDFREE(&nd, NDF_ONLY_PNBUF);
1632 			vrele(vp);
1633 			return (EAGAIN);
1634 		}
1635 	}
1636 	if (vp != NULL)
1637 		vrele(vp);
1638 	return (error);
1639 }
1640 
1641 /*
1642  * Make a symbolic link.
1643  */
1644 #ifndef _SYS_SYSPROTO_H_
1645 struct symlink_args {
1646 	char	*path;
1647 	char	*link;
1648 };
1649 #endif
1650 int
1651 sys_symlink(struct thread *td, struct symlink_args *uap)
1652 {
1653 
1654 	return (kern_symlinkat(td, uap->path, AT_FDCWD, uap->link,
1655 	    UIO_USERSPACE));
1656 }
1657 
1658 #ifndef _SYS_SYSPROTO_H_
1659 struct symlinkat_args {
1660 	char	*path;
1661 	int	fd;
1662 	char	*path2;
1663 };
1664 #endif
1665 int
1666 sys_symlinkat(struct thread *td, struct symlinkat_args *uap)
1667 {
1668 
1669 	return (kern_symlinkat(td, uap->path1, uap->fd, uap->path2,
1670 	    UIO_USERSPACE));
1671 }
1672 
1673 int
1674 kern_symlinkat(struct thread *td, const char *path1, int fd, const char *path2,
1675     enum uio_seg segflg)
1676 {
1677 	struct mount *mp;
1678 	struct vattr vattr;
1679 	const char *syspath;
1680 	char *tmppath;
1681 	struct nameidata nd;
1682 	int error;
1683 
1684 	if (segflg == UIO_SYSSPACE) {
1685 		syspath = path1;
1686 	} else {
1687 		tmppath = uma_zalloc(namei_zone, M_WAITOK);
1688 		if ((error = copyinstr(path1, tmppath, MAXPATHLEN, NULL)) != 0)
1689 			goto out;
1690 		syspath = tmppath;
1691 	}
1692 	AUDIT_ARG_TEXT(syspath);
1693 restart:
1694 	bwillwrite();
1695 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
1696 	    NOCACHE, segflg, path2, fd, &cap_symlinkat_rights,
1697 	    td);
1698 	if ((error = namei(&nd)) != 0)
1699 		goto out;
1700 	if (nd.ni_vp) {
1701 		NDFREE(&nd, NDF_ONLY_PNBUF);
1702 		if (nd.ni_vp == nd.ni_dvp)
1703 			vrele(nd.ni_dvp);
1704 		else
1705 			vput(nd.ni_dvp);
1706 		vrele(nd.ni_vp);
1707 		nd.ni_vp = NULL;
1708 		error = EEXIST;
1709 		goto out;
1710 	}
1711 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1712 		NDFREE(&nd, NDF_ONLY_PNBUF);
1713 		vput(nd.ni_dvp);
1714 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1715 			goto out;
1716 		goto restart;
1717 	}
1718 	VATTR_NULL(&vattr);
1719 	vattr.va_mode = ACCESSPERMS &~ td->td_proc->p_pd->pd_cmask;
1720 #ifdef MAC
1721 	vattr.va_type = VLNK;
1722 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1723 	    &vattr);
1724 	if (error != 0)
1725 		goto out2;
1726 #endif
1727 	error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, syspath);
1728 #ifdef MAC
1729 out2:
1730 #endif
1731 	VOP_VPUT_PAIR(nd.ni_dvp, error == 0 ? &nd.ni_vp : NULL, true);
1732 	vn_finished_write(mp);
1733 	NDFREE(&nd, NDF_ONLY_PNBUF);
1734 	if (error == ERELOOKUP)
1735 		goto restart;
1736 out:
1737 	if (segflg != UIO_SYSSPACE)
1738 		uma_zfree(namei_zone, tmppath);
1739 	return (error);
1740 }
1741 
1742 /*
1743  * Delete a whiteout from the filesystem.
1744  */
1745 #ifndef _SYS_SYSPROTO_H_
1746 struct undelete_args {
1747 	char *path;
1748 };
1749 #endif
1750 int
1751 sys_undelete(struct thread *td, struct undelete_args *uap)
1752 {
1753 	struct mount *mp;
1754 	struct nameidata nd;
1755 	int error;
1756 
1757 restart:
1758 	bwillwrite();
1759 	NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | AUDITVNODE1,
1760 	    UIO_USERSPACE, uap->path, td);
1761 	error = namei(&nd);
1762 	if (error != 0)
1763 		return (error);
1764 
1765 	if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) {
1766 		NDFREE(&nd, NDF_ONLY_PNBUF);
1767 		if (nd.ni_vp == nd.ni_dvp)
1768 			vrele(nd.ni_dvp);
1769 		else
1770 			vput(nd.ni_dvp);
1771 		if (nd.ni_vp)
1772 			vrele(nd.ni_vp);
1773 		return (EEXIST);
1774 	}
1775 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1776 		NDFREE(&nd, NDF_ONLY_PNBUF);
1777 		vput(nd.ni_dvp);
1778 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1779 			return (error);
1780 		goto restart;
1781 	}
1782 	error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE);
1783 	NDFREE(&nd, NDF_ONLY_PNBUF);
1784 	vput(nd.ni_dvp);
1785 	vn_finished_write(mp);
1786 	if (error == ERELOOKUP)
1787 		goto restart;
1788 	return (error);
1789 }
1790 
1791 /*
1792  * Delete a name from the filesystem.
1793  */
1794 #ifndef _SYS_SYSPROTO_H_
1795 struct unlink_args {
1796 	char	*path;
1797 };
1798 #endif
1799 int
1800 sys_unlink(struct thread *td, struct unlink_args *uap)
1801 {
1802 
1803 	return (kern_funlinkat(td, AT_FDCWD, uap->path, FD_NONE, UIO_USERSPACE,
1804 	    0, 0));
1805 }
1806 
1807 static int
1808 kern_funlinkat_ex(struct thread *td, int dfd, const char *path, int fd,
1809     int flag, enum uio_seg pathseg, ino_t oldinum)
1810 {
1811 
1812 	if ((flag & ~(AT_REMOVEDIR | AT_RESOLVE_BENEATH)) != 0)
1813 		return (EINVAL);
1814 
1815 	if ((flag & AT_REMOVEDIR) != 0)
1816 		return (kern_frmdirat(td, dfd, path, fd, UIO_USERSPACE, 0));
1817 
1818 	return (kern_funlinkat(td, dfd, path, fd, UIO_USERSPACE, 0, 0));
1819 }
1820 
1821 #ifndef _SYS_SYSPROTO_H_
1822 struct unlinkat_args {
1823 	int	fd;
1824 	char	*path;
1825 	int	flag;
1826 };
1827 #endif
1828 int
1829 sys_unlinkat(struct thread *td, struct unlinkat_args *uap)
1830 {
1831 
1832 	return (kern_funlinkat_ex(td, uap->fd, uap->path, FD_NONE, uap->flag,
1833 	    UIO_USERSPACE, 0));
1834 }
1835 
1836 #ifndef _SYS_SYSPROTO_H_
1837 struct funlinkat_args {
1838 	int		dfd;
1839 	const char	*path;
1840 	int		fd;
1841 	int		flag;
1842 };
1843 #endif
1844 int
1845 sys_funlinkat(struct thread *td, struct funlinkat_args *uap)
1846 {
1847 
1848 	return (kern_funlinkat_ex(td, uap->dfd, uap->path, uap->fd, uap->flag,
1849 	    UIO_USERSPACE, 0));
1850 }
1851 
1852 int
1853 kern_funlinkat(struct thread *td, int dfd, const char *path, int fd,
1854     enum uio_seg pathseg, int flag, ino_t oldinum)
1855 {
1856 	struct mount *mp;
1857 	struct file *fp;
1858 	struct vnode *vp;
1859 	struct nameidata nd;
1860 	struct stat sb;
1861 	int error;
1862 
1863 	fp = NULL;
1864 	if (fd != FD_NONE) {
1865 		error = getvnode(td, fd, &cap_no_rights, &fp);
1866 		if (error != 0)
1867 			return (error);
1868 	}
1869 
1870 restart:
1871 	bwillwrite();
1872 	NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1 |
1873 	    at2cnpflags(flag, AT_RESOLVE_BENEATH),
1874 	    pathseg, path, dfd, &cap_unlinkat_rights, td);
1875 	if ((error = namei(&nd)) != 0) {
1876 		if (error == EINVAL)
1877 			error = EPERM;
1878 		goto fdout;
1879 	}
1880 	vp = nd.ni_vp;
1881 	if (vp->v_type == VDIR && oldinum == 0) {
1882 		error = EPERM;		/* POSIX */
1883 	} else if (oldinum != 0 &&
1884 		  ((error = VOP_STAT(vp, &sb, td->td_ucred, NOCRED, td)) == 0) &&
1885 		  sb.st_ino != oldinum) {
1886 		error = EIDRM;	/* Identifier removed */
1887 	} else if (fp != NULL && fp->f_vnode != vp) {
1888 		if (VN_IS_DOOMED(fp->f_vnode))
1889 			error = EBADF;
1890 		else
1891 			error = EDEADLK;
1892 	} else {
1893 		/*
1894 		 * The root of a mounted filesystem cannot be deleted.
1895 		 *
1896 		 * XXX: can this only be a VDIR case?
1897 		 */
1898 		if (vp->v_vflag & VV_ROOT)
1899 			error = EBUSY;
1900 	}
1901 	if (error == 0) {
1902 		if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1903 			NDFREE(&nd, NDF_ONLY_PNBUF);
1904 			vput(nd.ni_dvp);
1905 			if (vp == nd.ni_dvp)
1906 				vrele(vp);
1907 			else
1908 				vput(vp);
1909 			if ((error = vn_start_write(NULL, &mp,
1910 			    V_XSLEEP | PCATCH)) != 0) {
1911 				goto fdout;
1912 			}
1913 			goto restart;
1914 		}
1915 #ifdef MAC
1916 		error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
1917 		    &nd.ni_cnd);
1918 		if (error != 0)
1919 			goto out;
1920 #endif
1921 		vfs_notify_upper(vp, VFS_NOTIFY_UPPER_UNLINK);
1922 		error = VOP_REMOVE(nd.ni_dvp, vp, &nd.ni_cnd);
1923 #ifdef MAC
1924 out:
1925 #endif
1926 		vn_finished_write(mp);
1927 	}
1928 	NDFREE(&nd, NDF_ONLY_PNBUF);
1929 	vput(nd.ni_dvp);
1930 	if (vp == nd.ni_dvp)
1931 		vrele(vp);
1932 	else
1933 		vput(vp);
1934 	if (error == ERELOOKUP)
1935 		goto restart;
1936 fdout:
1937 	if (fp != NULL)
1938 		fdrop(fp, td);
1939 	return (error);
1940 }
1941 
1942 /*
1943  * Reposition read/write file offset.
1944  */
1945 #ifndef _SYS_SYSPROTO_H_
1946 struct lseek_args {
1947 	int	fd;
1948 	int	pad;
1949 	off_t	offset;
1950 	int	whence;
1951 };
1952 #endif
1953 int
1954 sys_lseek(struct thread *td, struct lseek_args *uap)
1955 {
1956 
1957 	return (kern_lseek(td, uap->fd, uap->offset, uap->whence));
1958 }
1959 
1960 int
1961 kern_lseek(struct thread *td, int fd, off_t offset, int whence)
1962 {
1963 	struct file *fp;
1964 	int error;
1965 
1966 	AUDIT_ARG_FD(fd);
1967 	error = fget(td, fd, &cap_seek_rights, &fp);
1968 	if (error != 0)
1969 		return (error);
1970 	error = (fp->f_ops->fo_flags & DFLAG_SEEKABLE) != 0 ?
1971 	    fo_seek(fp, offset, whence, td) : ESPIPE;
1972 	fdrop(fp, td);
1973 	return (error);
1974 }
1975 
1976 #if defined(COMPAT_43)
1977 /*
1978  * Reposition read/write file offset.
1979  */
1980 #ifndef _SYS_SYSPROTO_H_
1981 struct olseek_args {
1982 	int	fd;
1983 	long	offset;
1984 	int	whence;
1985 };
1986 #endif
1987 int
1988 olseek(struct thread *td, struct olseek_args *uap)
1989 {
1990 
1991 	return (kern_lseek(td, uap->fd, uap->offset, uap->whence));
1992 }
1993 #endif /* COMPAT_43 */
1994 
1995 #if defined(COMPAT_FREEBSD6)
1996 /* Version with the 'pad' argument */
1997 int
1998 freebsd6_lseek(struct thread *td, struct freebsd6_lseek_args *uap)
1999 {
2000 
2001 	return (kern_lseek(td, uap->fd, uap->offset, uap->whence));
2002 }
2003 #endif
2004 
2005 /*
2006  * Check access permissions using passed credentials.
2007  */
2008 static int
2009 vn_access(struct vnode *vp, int user_flags, struct ucred *cred,
2010      struct thread *td)
2011 {
2012 	accmode_t accmode;
2013 	int error;
2014 
2015 	/* Flags == 0 means only check for existence. */
2016 	if (user_flags == 0)
2017 		return (0);
2018 
2019 	accmode = 0;
2020 	if (user_flags & R_OK)
2021 		accmode |= VREAD;
2022 	if (user_flags & W_OK)
2023 		accmode |= VWRITE;
2024 	if (user_flags & X_OK)
2025 		accmode |= VEXEC;
2026 #ifdef MAC
2027 	error = mac_vnode_check_access(cred, vp, accmode);
2028 	if (error != 0)
2029 		return (error);
2030 #endif
2031 	if ((accmode & VWRITE) == 0 || (error = vn_writechk(vp)) == 0)
2032 		error = VOP_ACCESS(vp, accmode, cred, td);
2033 	return (error);
2034 }
2035 
2036 /*
2037  * Check access permissions using "real" credentials.
2038  */
2039 #ifndef _SYS_SYSPROTO_H_
2040 struct access_args {
2041 	char	*path;
2042 	int	amode;
2043 };
2044 #endif
2045 int
2046 sys_access(struct thread *td, struct access_args *uap)
2047 {
2048 
2049 	return (kern_accessat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2050 	    0, uap->amode));
2051 }
2052 
2053 #ifndef _SYS_SYSPROTO_H_
2054 struct faccessat_args {
2055 	int	dirfd;
2056 	char	*path;
2057 	int	amode;
2058 	int	flag;
2059 }
2060 #endif
2061 int
2062 sys_faccessat(struct thread *td, struct faccessat_args *uap)
2063 {
2064 
2065 	return (kern_accessat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
2066 	    uap->amode));
2067 }
2068 
2069 int
2070 kern_accessat(struct thread *td, int fd, const char *path,
2071     enum uio_seg pathseg, int flag, int amode)
2072 {
2073 	struct ucred *cred, *usecred;
2074 	struct vnode *vp;
2075 	struct nameidata nd;
2076 	int error;
2077 
2078 	if ((flag & ~(AT_EACCESS | AT_RESOLVE_BENEATH)) != 0)
2079 		return (EINVAL);
2080 	if (amode != F_OK && (amode & ~(R_OK | W_OK | X_OK)) != 0)
2081 		return (EINVAL);
2082 
2083 	/*
2084 	 * Create and modify a temporary credential instead of one that
2085 	 * is potentially shared (if we need one).
2086 	 */
2087 	cred = td->td_ucred;
2088 	if ((flag & AT_EACCESS) == 0 &&
2089 	    ((cred->cr_uid != cred->cr_ruid ||
2090 	    cred->cr_rgid != cred->cr_groups[0]))) {
2091 		usecred = crdup(cred);
2092 		usecred->cr_uid = cred->cr_ruid;
2093 		usecred->cr_groups[0] = cred->cr_rgid;
2094 		td->td_ucred = usecred;
2095 	} else
2096 		usecred = cred;
2097 	AUDIT_ARG_VALUE(amode);
2098 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF |
2099 	    AUDITVNODE1 | at2cnpflags(flag, AT_RESOLVE_BENEATH),
2100 	    pathseg, path, fd, &cap_fstat_rights, td);
2101 	if ((error = namei(&nd)) != 0)
2102 		goto out;
2103 	vp = nd.ni_vp;
2104 
2105 	error = vn_access(vp, amode, usecred, td);
2106 	NDFREE_NOTHING(&nd);
2107 	vput(vp);
2108 out:
2109 	if (usecred != cred) {
2110 		td->td_ucred = cred;
2111 		crfree(usecred);
2112 	}
2113 	return (error);
2114 }
2115 
2116 /*
2117  * Check access permissions using "effective" credentials.
2118  */
2119 #ifndef _SYS_SYSPROTO_H_
2120 struct eaccess_args {
2121 	char	*path;
2122 	int	amode;
2123 };
2124 #endif
2125 int
2126 sys_eaccess(struct thread *td, struct eaccess_args *uap)
2127 {
2128 
2129 	return (kern_accessat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2130 	    AT_EACCESS, uap->amode));
2131 }
2132 
2133 #if defined(COMPAT_43)
2134 /*
2135  * Get file status; this version follows links.
2136  */
2137 #ifndef _SYS_SYSPROTO_H_
2138 struct ostat_args {
2139 	char	*path;
2140 	struct ostat *ub;
2141 };
2142 #endif
2143 int
2144 ostat(struct thread *td, struct ostat_args *uap)
2145 {
2146 	struct stat sb;
2147 	struct ostat osb;
2148 	int error;
2149 
2150 	error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE,
2151 	    &sb, NULL);
2152 	if (error != 0)
2153 		return (error);
2154 	cvtstat(&sb, &osb);
2155 	return (copyout(&osb, uap->ub, sizeof (osb)));
2156 }
2157 
2158 /*
2159  * Get file status; this version does not follow links.
2160  */
2161 #ifndef _SYS_SYSPROTO_H_
2162 struct olstat_args {
2163 	char	*path;
2164 	struct ostat *ub;
2165 };
2166 #endif
2167 int
2168 olstat(struct thread *td, struct olstat_args *uap)
2169 {
2170 	struct stat sb;
2171 	struct ostat osb;
2172 	int error;
2173 
2174 	error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2175 	    UIO_USERSPACE, &sb, NULL);
2176 	if (error != 0)
2177 		return (error);
2178 	cvtstat(&sb, &osb);
2179 	return (copyout(&osb, uap->ub, sizeof (osb)));
2180 }
2181 
2182 /*
2183  * Convert from an old to a new stat structure.
2184  * XXX: many values are blindly truncated.
2185  */
2186 void
2187 cvtstat(struct stat *st, struct ostat *ost)
2188 {
2189 
2190 	bzero(ost, sizeof(*ost));
2191 	ost->st_dev = st->st_dev;
2192 	ost->st_ino = st->st_ino;
2193 	ost->st_mode = st->st_mode;
2194 	ost->st_nlink = st->st_nlink;
2195 	ost->st_uid = st->st_uid;
2196 	ost->st_gid = st->st_gid;
2197 	ost->st_rdev = st->st_rdev;
2198 	ost->st_size = MIN(st->st_size, INT32_MAX);
2199 	ost->st_atim = st->st_atim;
2200 	ost->st_mtim = st->st_mtim;
2201 	ost->st_ctim = st->st_ctim;
2202 	ost->st_blksize = st->st_blksize;
2203 	ost->st_blocks = st->st_blocks;
2204 	ost->st_flags = st->st_flags;
2205 	ost->st_gen = st->st_gen;
2206 }
2207 #endif /* COMPAT_43 */
2208 
2209 #if defined(COMPAT_43) || defined(COMPAT_FREEBSD11)
2210 int ino64_trunc_error;
2211 SYSCTL_INT(_vfs, OID_AUTO, ino64_trunc_error, CTLFLAG_RW,
2212     &ino64_trunc_error, 0,
2213     "Error on truncation of device, file or inode number, or link count");
2214 
2215 int
2216 freebsd11_cvtstat(struct stat *st, struct freebsd11_stat *ost)
2217 {
2218 
2219 	ost->st_dev = st->st_dev;
2220 	if (ost->st_dev != st->st_dev) {
2221 		switch (ino64_trunc_error) {
2222 		default:
2223 			/*
2224 			 * Since dev_t is almost raw, don't clamp to the
2225 			 * maximum for case 2, but ignore the error.
2226 			 */
2227 			break;
2228 		case 1:
2229 			return (EOVERFLOW);
2230 		}
2231 	}
2232 	ost->st_ino = st->st_ino;
2233 	if (ost->st_ino != st->st_ino) {
2234 		switch (ino64_trunc_error) {
2235 		default:
2236 		case 0:
2237 			break;
2238 		case 1:
2239 			return (EOVERFLOW);
2240 		case 2:
2241 			ost->st_ino = UINT32_MAX;
2242 			break;
2243 		}
2244 	}
2245 	ost->st_mode = st->st_mode;
2246 	ost->st_nlink = st->st_nlink;
2247 	if (ost->st_nlink != st->st_nlink) {
2248 		switch (ino64_trunc_error) {
2249 		default:
2250 		case 0:
2251 			break;
2252 		case 1:
2253 			return (EOVERFLOW);
2254 		case 2:
2255 			ost->st_nlink = UINT16_MAX;
2256 			break;
2257 		}
2258 	}
2259 	ost->st_uid = st->st_uid;
2260 	ost->st_gid = st->st_gid;
2261 	ost->st_rdev = st->st_rdev;
2262 	if (ost->st_rdev != st->st_rdev) {
2263 		switch (ino64_trunc_error) {
2264 		default:
2265 			break;
2266 		case 1:
2267 			return (EOVERFLOW);
2268 		}
2269 	}
2270 	ost->st_atim = st->st_atim;
2271 	ost->st_mtim = st->st_mtim;
2272 	ost->st_ctim = st->st_ctim;
2273 	ost->st_size = st->st_size;
2274 	ost->st_blocks = st->st_blocks;
2275 	ost->st_blksize = st->st_blksize;
2276 	ost->st_flags = st->st_flags;
2277 	ost->st_gen = st->st_gen;
2278 	ost->st_lspare = 0;
2279 	ost->st_birthtim = st->st_birthtim;
2280 	bzero((char *)&ost->st_birthtim + sizeof(ost->st_birthtim),
2281 	    sizeof(*ost) - offsetof(struct freebsd11_stat,
2282 	    st_birthtim) - sizeof(ost->st_birthtim));
2283 	return (0);
2284 }
2285 
2286 int
2287 freebsd11_stat(struct thread *td, struct freebsd11_stat_args* uap)
2288 {
2289 	struct stat sb;
2290 	struct freebsd11_stat osb;
2291 	int error;
2292 
2293 	error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE,
2294 	    &sb, NULL);
2295 	if (error != 0)
2296 		return (error);
2297 	error = freebsd11_cvtstat(&sb, &osb);
2298 	if (error == 0)
2299 		error = copyout(&osb, uap->ub, sizeof(osb));
2300 	return (error);
2301 }
2302 
2303 int
2304 freebsd11_lstat(struct thread *td, struct freebsd11_lstat_args* uap)
2305 {
2306 	struct stat sb;
2307 	struct freebsd11_stat osb;
2308 	int error;
2309 
2310 	error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2311 	    UIO_USERSPACE, &sb, NULL);
2312 	if (error != 0)
2313 		return (error);
2314 	error = freebsd11_cvtstat(&sb, &osb);
2315 	if (error == 0)
2316 		error = copyout(&osb, uap->ub, sizeof(osb));
2317 	return (error);
2318 }
2319 
2320 int
2321 freebsd11_fhstat(struct thread *td, struct freebsd11_fhstat_args* uap)
2322 {
2323 	struct fhandle fh;
2324 	struct stat sb;
2325 	struct freebsd11_stat osb;
2326 	int error;
2327 
2328 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
2329 	if (error != 0)
2330 		return (error);
2331 	error = kern_fhstat(td, fh, &sb);
2332 	if (error != 0)
2333 		return (error);
2334 	error = freebsd11_cvtstat(&sb, &osb);
2335 	if (error == 0)
2336 		error = copyout(&osb, uap->sb, sizeof(osb));
2337 	return (error);
2338 }
2339 
2340 int
2341 freebsd11_fstatat(struct thread *td, struct freebsd11_fstatat_args* uap)
2342 {
2343 	struct stat sb;
2344 	struct freebsd11_stat osb;
2345 	int error;
2346 
2347 	error = kern_statat(td, uap->flag, uap->fd, uap->path,
2348 	    UIO_USERSPACE, &sb, NULL);
2349 	if (error != 0)
2350 		return (error);
2351 	error = freebsd11_cvtstat(&sb, &osb);
2352 	if (error == 0)
2353 		error = copyout(&osb, uap->buf, sizeof(osb));
2354 	return (error);
2355 }
2356 #endif	/* COMPAT_FREEBSD11 */
2357 
2358 /*
2359  * Get file status
2360  */
2361 #ifndef _SYS_SYSPROTO_H_
2362 struct fstatat_args {
2363 	int	fd;
2364 	char	*path;
2365 	struct stat	*buf;
2366 	int	flag;
2367 }
2368 #endif
2369 int
2370 sys_fstatat(struct thread *td, struct fstatat_args *uap)
2371 {
2372 	struct stat sb;
2373 	int error;
2374 
2375 	error = kern_statat(td, uap->flag, uap->fd, uap->path,
2376 	    UIO_USERSPACE, &sb, NULL);
2377 	if (error == 0)
2378 		error = copyout(&sb, uap->buf, sizeof (sb));
2379 	return (error);
2380 }
2381 
2382 int
2383 kern_statat(struct thread *td, int flag, int fd, const char *path,
2384     enum uio_seg pathseg, struct stat *sbp,
2385     void (*hook)(struct vnode *vp, struct stat *sbp))
2386 {
2387 	struct nameidata nd;
2388 	int error;
2389 
2390 	if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH)) != 0)
2391 		return (EINVAL);
2392 
2393 	NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(flag, AT_RESOLVE_BENEATH |
2394 	    AT_SYMLINK_NOFOLLOW) | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
2395 	    pathseg, path, fd, &cap_fstat_rights, td);
2396 
2397 	if ((error = namei(&nd)) != 0)
2398 		return (error);
2399 	error = VOP_STAT(nd.ni_vp, sbp, td->td_ucred, NOCRED, td);
2400 	if (error == 0) {
2401 		if (__predict_false(hook != NULL))
2402 			hook(nd.ni_vp, sbp);
2403 	}
2404 	NDFREE_NOTHING(&nd);
2405 	vput(nd.ni_vp);
2406 #ifdef __STAT_TIME_T_EXT
2407 	sbp->st_atim_ext = 0;
2408 	sbp->st_mtim_ext = 0;
2409 	sbp->st_ctim_ext = 0;
2410 	sbp->st_btim_ext = 0;
2411 #endif
2412 #ifdef KTRACE
2413 	if (KTRPOINT(td, KTR_STRUCT))
2414 		ktrstat_error(sbp, error);
2415 #endif
2416 	return (error);
2417 }
2418 
2419 #if defined(COMPAT_FREEBSD11)
2420 /*
2421  * Implementation of the NetBSD [l]stat() functions.
2422  */
2423 void
2424 freebsd11_cvtnstat(struct stat *sb, struct nstat *nsb)
2425 {
2426 
2427 	bzero(nsb, sizeof(*nsb));
2428 	nsb->st_dev = sb->st_dev;
2429 	nsb->st_ino = sb->st_ino;
2430 	nsb->st_mode = sb->st_mode;
2431 	nsb->st_nlink = sb->st_nlink;
2432 	nsb->st_uid = sb->st_uid;
2433 	nsb->st_gid = sb->st_gid;
2434 	nsb->st_rdev = sb->st_rdev;
2435 	nsb->st_atim = sb->st_atim;
2436 	nsb->st_mtim = sb->st_mtim;
2437 	nsb->st_ctim = sb->st_ctim;
2438 	nsb->st_size = sb->st_size;
2439 	nsb->st_blocks = sb->st_blocks;
2440 	nsb->st_blksize = sb->st_blksize;
2441 	nsb->st_flags = sb->st_flags;
2442 	nsb->st_gen = sb->st_gen;
2443 	nsb->st_birthtim = sb->st_birthtim;
2444 }
2445 
2446 #ifndef _SYS_SYSPROTO_H_
2447 struct freebsd11_nstat_args {
2448 	char	*path;
2449 	struct nstat *ub;
2450 };
2451 #endif
2452 int
2453 freebsd11_nstat(struct thread *td, struct freebsd11_nstat_args *uap)
2454 {
2455 	struct stat sb;
2456 	struct nstat nsb;
2457 	int error;
2458 
2459 	error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE,
2460 	    &sb, NULL);
2461 	if (error != 0)
2462 		return (error);
2463 	freebsd11_cvtnstat(&sb, &nsb);
2464 	return (copyout(&nsb, uap->ub, sizeof (nsb)));
2465 }
2466 
2467 /*
2468  * NetBSD lstat.  Get file status; this version does not follow links.
2469  */
2470 #ifndef _SYS_SYSPROTO_H_
2471 struct freebsd11_nlstat_args {
2472 	char	*path;
2473 	struct nstat *ub;
2474 };
2475 #endif
2476 int
2477 freebsd11_nlstat(struct thread *td, struct freebsd11_nlstat_args *uap)
2478 {
2479 	struct stat sb;
2480 	struct nstat nsb;
2481 	int error;
2482 
2483 	error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2484 	    UIO_USERSPACE, &sb, NULL);
2485 	if (error != 0)
2486 		return (error);
2487 	freebsd11_cvtnstat(&sb, &nsb);
2488 	return (copyout(&nsb, uap->ub, sizeof (nsb)));
2489 }
2490 #endif /* COMPAT_FREEBSD11 */
2491 
2492 /*
2493  * Get configurable pathname variables.
2494  */
2495 #ifndef _SYS_SYSPROTO_H_
2496 struct pathconf_args {
2497 	char	*path;
2498 	int	name;
2499 };
2500 #endif
2501 int
2502 sys_pathconf(struct thread *td, struct pathconf_args *uap)
2503 {
2504 	long value;
2505 	int error;
2506 
2507 	error = kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name, FOLLOW,
2508 	    &value);
2509 	if (error == 0)
2510 		td->td_retval[0] = value;
2511 	return (error);
2512 }
2513 
2514 #ifndef _SYS_SYSPROTO_H_
2515 struct lpathconf_args {
2516 	char	*path;
2517 	int	name;
2518 };
2519 #endif
2520 int
2521 sys_lpathconf(struct thread *td, struct lpathconf_args *uap)
2522 {
2523 	long value;
2524 	int error;
2525 
2526 	error = kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name,
2527 	    NOFOLLOW, &value);
2528 	if (error == 0)
2529 		td->td_retval[0] = value;
2530 	return (error);
2531 }
2532 
2533 int
2534 kern_pathconf(struct thread *td, const char *path, enum uio_seg pathseg,
2535     int name, u_long flags, long *valuep)
2536 {
2537 	struct nameidata nd;
2538 	int error;
2539 
2540 	NDINIT(&nd, LOOKUP, LOCKSHARED | LOCKLEAF | AUDITVNODE1 | flags,
2541 	    pathseg, path, td);
2542 	if ((error = namei(&nd)) != 0)
2543 		return (error);
2544 	NDFREE_NOTHING(&nd);
2545 
2546 	error = VOP_PATHCONF(nd.ni_vp, name, valuep);
2547 	vput(nd.ni_vp);
2548 	return (error);
2549 }
2550 
2551 /*
2552  * Return target name of a symbolic link.
2553  */
2554 #ifndef _SYS_SYSPROTO_H_
2555 struct readlink_args {
2556 	char	*path;
2557 	char	*buf;
2558 	size_t	count;
2559 };
2560 #endif
2561 int
2562 sys_readlink(struct thread *td, struct readlink_args *uap)
2563 {
2564 
2565 	return (kern_readlinkat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2566 	    uap->buf, UIO_USERSPACE, uap->count));
2567 }
2568 #ifndef _SYS_SYSPROTO_H_
2569 struct readlinkat_args {
2570 	int	fd;
2571 	char	*path;
2572 	char	*buf;
2573 	size_t	bufsize;
2574 };
2575 #endif
2576 int
2577 sys_readlinkat(struct thread *td, struct readlinkat_args *uap)
2578 {
2579 
2580 	return (kern_readlinkat(td, uap->fd, uap->path, UIO_USERSPACE,
2581 	    uap->buf, UIO_USERSPACE, uap->bufsize));
2582 }
2583 
2584 int
2585 kern_readlinkat(struct thread *td, int fd, const char *path,
2586     enum uio_seg pathseg, char *buf, enum uio_seg bufseg, size_t count)
2587 {
2588 	struct vnode *vp;
2589 	struct nameidata nd;
2590 	int error;
2591 
2592 	if (count > IOSIZE_MAX)
2593 		return (EINVAL);
2594 
2595 	NDINIT_AT(&nd, LOOKUP, NOFOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
2596 	    pathseg, path, fd, td);
2597 
2598 	if ((error = namei(&nd)) != 0)
2599 		return (error);
2600 	NDFREE_NOTHING(&nd);
2601 	vp = nd.ni_vp;
2602 
2603 	error = kern_readlink_vp(vp, buf, bufseg, count, td);
2604 	vput(vp);
2605 
2606 	return (error);
2607 }
2608 
2609 /*
2610  * Helper function to readlink from a vnode
2611  */
2612 static int
2613 kern_readlink_vp(struct vnode *vp, char *buf, enum uio_seg bufseg, size_t count,
2614     struct thread *td)
2615 {
2616 	struct iovec aiov;
2617 	struct uio auio;
2618 	int error;
2619 
2620 	ASSERT_VOP_LOCKED(vp, "kern_readlink_vp(): vp not locked");
2621 #ifdef MAC
2622 	error = mac_vnode_check_readlink(td->td_ucred, vp);
2623 	if (error != 0)
2624 		return (error);
2625 #endif
2626 	if (vp->v_type != VLNK && (vp->v_vflag & VV_READLINK) == 0)
2627 		return (EINVAL);
2628 
2629 	aiov.iov_base = buf;
2630 	aiov.iov_len = count;
2631 	auio.uio_iov = &aiov;
2632 	auio.uio_iovcnt = 1;
2633 	auio.uio_offset = 0;
2634 	auio.uio_rw = UIO_READ;
2635 	auio.uio_segflg = bufseg;
2636 	auio.uio_td = td;
2637 	auio.uio_resid = count;
2638 	error = VOP_READLINK(vp, &auio, td->td_ucred);
2639 	td->td_retval[0] = count - auio.uio_resid;
2640 	return (error);
2641 }
2642 
2643 /*
2644  * Common implementation code for chflags() and fchflags().
2645  */
2646 static int
2647 setfflags(struct thread *td, struct vnode *vp, u_long flags)
2648 {
2649 	struct mount *mp;
2650 	struct vattr vattr;
2651 	int error;
2652 
2653 	/* We can't support the value matching VNOVAL. */
2654 	if (flags == VNOVAL)
2655 		return (EOPNOTSUPP);
2656 
2657 	/*
2658 	 * Prevent non-root users from setting flags on devices.  When
2659 	 * a device is reused, users can retain ownership of the device
2660 	 * if they are allowed to set flags and programs assume that
2661 	 * chown can't fail when done as root.
2662 	 */
2663 	if (vp->v_type == VCHR || vp->v_type == VBLK) {
2664 		error = priv_check(td, PRIV_VFS_CHFLAGS_DEV);
2665 		if (error != 0)
2666 			return (error);
2667 	}
2668 
2669 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2670 		return (error);
2671 	VATTR_NULL(&vattr);
2672 	vattr.va_flags = flags;
2673 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2674 #ifdef MAC
2675 	error = mac_vnode_check_setflags(td->td_ucred, vp, vattr.va_flags);
2676 	if (error == 0)
2677 #endif
2678 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
2679 	VOP_UNLOCK(vp);
2680 	vn_finished_write(mp);
2681 	return (error);
2682 }
2683 
2684 /*
2685  * Change flags of a file given a path name.
2686  */
2687 #ifndef _SYS_SYSPROTO_H_
2688 struct chflags_args {
2689 	const char *path;
2690 	u_long	flags;
2691 };
2692 #endif
2693 int
2694 sys_chflags(struct thread *td, struct chflags_args *uap)
2695 {
2696 
2697 	return (kern_chflagsat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2698 	    uap->flags, 0));
2699 }
2700 
2701 #ifndef _SYS_SYSPROTO_H_
2702 struct chflagsat_args {
2703 	int	fd;
2704 	const char *path;
2705 	u_long	flags;
2706 	int	atflag;
2707 }
2708 #endif
2709 int
2710 sys_chflagsat(struct thread *td, struct chflagsat_args *uap)
2711 {
2712 
2713 	if ((uap->atflag & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH)) != 0)
2714 		return (EINVAL);
2715 
2716 	return (kern_chflagsat(td, uap->fd, uap->path, UIO_USERSPACE,
2717 	    uap->flags, uap->atflag));
2718 }
2719 
2720 /*
2721  * Same as chflags() but doesn't follow symlinks.
2722  */
2723 #ifndef _SYS_SYSPROTO_H_
2724 struct lchflags_args {
2725 	const char *path;
2726 	u_long flags;
2727 };
2728 #endif
2729 int
2730 sys_lchflags(struct thread *td, struct lchflags_args *uap)
2731 {
2732 
2733 	return (kern_chflagsat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2734 	    uap->flags, AT_SYMLINK_NOFOLLOW));
2735 }
2736 
2737 static int
2738 kern_chflagsat(struct thread *td, int fd, const char *path,
2739     enum uio_seg pathseg, u_long flags, int atflag)
2740 {
2741 	struct nameidata nd;
2742 	int error;
2743 
2744 	AUDIT_ARG_FFLAGS(flags);
2745 	NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(atflag, AT_SYMLINK_NOFOLLOW |
2746 	    AT_RESOLVE_BENEATH) | AUDITVNODE1, pathseg, path, fd,
2747 	    &cap_fchflags_rights, td);
2748 	if ((error = namei(&nd)) != 0)
2749 		return (error);
2750 	NDFREE_NOTHING(&nd);
2751 	error = setfflags(td, nd.ni_vp, flags);
2752 	vrele(nd.ni_vp);
2753 	return (error);
2754 }
2755 
2756 /*
2757  * Change flags of a file given a file descriptor.
2758  */
2759 #ifndef _SYS_SYSPROTO_H_
2760 struct fchflags_args {
2761 	int	fd;
2762 	u_long	flags;
2763 };
2764 #endif
2765 int
2766 sys_fchflags(struct thread *td, struct fchflags_args *uap)
2767 {
2768 	struct file *fp;
2769 	int error;
2770 
2771 	AUDIT_ARG_FD(uap->fd);
2772 	AUDIT_ARG_FFLAGS(uap->flags);
2773 	error = getvnode(td, uap->fd, &cap_fchflags_rights,
2774 	    &fp);
2775 	if (error != 0)
2776 		return (error);
2777 #ifdef AUDIT
2778 	vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
2779 	AUDIT_ARG_VNODE1(fp->f_vnode);
2780 	VOP_UNLOCK(fp->f_vnode);
2781 #endif
2782 	error = setfflags(td, fp->f_vnode, uap->flags);
2783 	fdrop(fp, td);
2784 	return (error);
2785 }
2786 
2787 /*
2788  * Common implementation code for chmod(), lchmod() and fchmod().
2789  */
2790 int
2791 setfmode(struct thread *td, struct ucred *cred, struct vnode *vp, int mode)
2792 {
2793 	struct mount *mp;
2794 	struct vattr vattr;
2795 	int error;
2796 
2797 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2798 		return (error);
2799 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2800 	VATTR_NULL(&vattr);
2801 	vattr.va_mode = mode & ALLPERMS;
2802 #ifdef MAC
2803 	error = mac_vnode_check_setmode(cred, vp, vattr.va_mode);
2804 	if (error == 0)
2805 #endif
2806 		error = VOP_SETATTR(vp, &vattr, cred);
2807 	VOP_UNLOCK(vp);
2808 	vn_finished_write(mp);
2809 	return (error);
2810 }
2811 
2812 /*
2813  * Change mode of a file given path name.
2814  */
2815 #ifndef _SYS_SYSPROTO_H_
2816 struct chmod_args {
2817 	char	*path;
2818 	int	mode;
2819 };
2820 #endif
2821 int
2822 sys_chmod(struct thread *td, struct chmod_args *uap)
2823 {
2824 
2825 	return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2826 	    uap->mode, 0));
2827 }
2828 
2829 #ifndef _SYS_SYSPROTO_H_
2830 struct fchmodat_args {
2831 	int	dirfd;
2832 	char	*path;
2833 	mode_t	mode;
2834 	int	flag;
2835 }
2836 #endif
2837 int
2838 sys_fchmodat(struct thread *td, struct fchmodat_args *uap)
2839 {
2840 
2841 	if ((uap->flag & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH)) != 0)
2842 		return (EINVAL);
2843 
2844 	return (kern_fchmodat(td, uap->fd, uap->path, UIO_USERSPACE,
2845 	    uap->mode, uap->flag));
2846 }
2847 
2848 /*
2849  * Change mode of a file given path name (don't follow links.)
2850  */
2851 #ifndef _SYS_SYSPROTO_H_
2852 struct lchmod_args {
2853 	char	*path;
2854 	int	mode;
2855 };
2856 #endif
2857 int
2858 sys_lchmod(struct thread *td, struct lchmod_args *uap)
2859 {
2860 
2861 	return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2862 	    uap->mode, AT_SYMLINK_NOFOLLOW));
2863 }
2864 
2865 int
2866 kern_fchmodat(struct thread *td, int fd, const char *path,
2867     enum uio_seg pathseg, mode_t mode, int flag)
2868 {
2869 	struct nameidata nd;
2870 	int error;
2871 
2872 	AUDIT_ARG_MODE(mode);
2873 	NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(flag, AT_SYMLINK_NOFOLLOW |
2874 	    AT_RESOLVE_BENEATH) | AUDITVNODE1, pathseg, path, fd,
2875 	    &cap_fchmod_rights, td);
2876 	if ((error = namei(&nd)) != 0)
2877 		return (error);
2878 	NDFREE_NOTHING(&nd);
2879 	error = setfmode(td, td->td_ucred, nd.ni_vp, mode);
2880 	vrele(nd.ni_vp);
2881 	return (error);
2882 }
2883 
2884 /*
2885  * Change mode of a file given a file descriptor.
2886  */
2887 #ifndef _SYS_SYSPROTO_H_
2888 struct fchmod_args {
2889 	int	fd;
2890 	int	mode;
2891 };
2892 #endif
2893 int
2894 sys_fchmod(struct thread *td, struct fchmod_args *uap)
2895 {
2896 	struct file *fp;
2897 	int error;
2898 
2899 	AUDIT_ARG_FD(uap->fd);
2900 	AUDIT_ARG_MODE(uap->mode);
2901 
2902 	error = fget(td, uap->fd, &cap_fchmod_rights, &fp);
2903 	if (error != 0)
2904 		return (error);
2905 	error = fo_chmod(fp, uap->mode, td->td_ucred, td);
2906 	fdrop(fp, td);
2907 	return (error);
2908 }
2909 
2910 /*
2911  * Common implementation for chown(), lchown(), and fchown()
2912  */
2913 int
2914 setfown(struct thread *td, struct ucred *cred, struct vnode *vp, uid_t uid,
2915     gid_t gid)
2916 {
2917 	struct mount *mp;
2918 	struct vattr vattr;
2919 	int error;
2920 
2921 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2922 		return (error);
2923 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2924 	VATTR_NULL(&vattr);
2925 	vattr.va_uid = uid;
2926 	vattr.va_gid = gid;
2927 #ifdef MAC
2928 	error = mac_vnode_check_setowner(cred, vp, vattr.va_uid,
2929 	    vattr.va_gid);
2930 	if (error == 0)
2931 #endif
2932 		error = VOP_SETATTR(vp, &vattr, cred);
2933 	VOP_UNLOCK(vp);
2934 	vn_finished_write(mp);
2935 	return (error);
2936 }
2937 
2938 /*
2939  * Set ownership given a path name.
2940  */
2941 #ifndef _SYS_SYSPROTO_H_
2942 struct chown_args {
2943 	char	*path;
2944 	int	uid;
2945 	int	gid;
2946 };
2947 #endif
2948 int
2949 sys_chown(struct thread *td, struct chown_args *uap)
2950 {
2951 
2952 	return (kern_fchownat(td, AT_FDCWD, uap->path, UIO_USERSPACE, uap->uid,
2953 	    uap->gid, 0));
2954 }
2955 
2956 #ifndef _SYS_SYSPROTO_H_
2957 struct fchownat_args {
2958 	int fd;
2959 	const char * path;
2960 	uid_t uid;
2961 	gid_t gid;
2962 	int flag;
2963 };
2964 #endif
2965 int
2966 sys_fchownat(struct thread *td, struct fchownat_args *uap)
2967 {
2968 
2969 	if ((uap->flag & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH)) != 0)
2970 		return (EINVAL);
2971 
2972 	return (kern_fchownat(td, uap->fd, uap->path, UIO_USERSPACE, uap->uid,
2973 	    uap->gid, uap->flag));
2974 }
2975 
2976 int
2977 kern_fchownat(struct thread *td, int fd, const char *path,
2978     enum uio_seg pathseg, int uid, int gid, int flag)
2979 {
2980 	struct nameidata nd;
2981 	int error;
2982 
2983 	AUDIT_ARG_OWNER(uid, gid);
2984 	NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(flag, AT_SYMLINK_NOFOLLOW |
2985 	    AT_RESOLVE_BENEATH) | AUDITVNODE1, pathseg, path, fd,
2986 	    &cap_fchown_rights, td);
2987 
2988 	if ((error = namei(&nd)) != 0)
2989 		return (error);
2990 	NDFREE_NOTHING(&nd);
2991 	error = setfown(td, td->td_ucred, nd.ni_vp, uid, gid);
2992 	vrele(nd.ni_vp);
2993 	return (error);
2994 }
2995 
2996 /*
2997  * Set ownership given a path name, do not cross symlinks.
2998  */
2999 #ifndef _SYS_SYSPROTO_H_
3000 struct lchown_args {
3001 	char	*path;
3002 	int	uid;
3003 	int	gid;
3004 };
3005 #endif
3006 int
3007 sys_lchown(struct thread *td, struct lchown_args *uap)
3008 {
3009 
3010 	return (kern_fchownat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
3011 	    uap->uid, uap->gid, AT_SYMLINK_NOFOLLOW));
3012 }
3013 
3014 /*
3015  * Set ownership given a file descriptor.
3016  */
3017 #ifndef _SYS_SYSPROTO_H_
3018 struct fchown_args {
3019 	int	fd;
3020 	int	uid;
3021 	int	gid;
3022 };
3023 #endif
3024 int
3025 sys_fchown(struct thread *td, struct fchown_args *uap)
3026 {
3027 	struct file *fp;
3028 	int error;
3029 
3030 	AUDIT_ARG_FD(uap->fd);
3031 	AUDIT_ARG_OWNER(uap->uid, uap->gid);
3032 	error = fget(td, uap->fd, &cap_fchown_rights, &fp);
3033 	if (error != 0)
3034 		return (error);
3035 	error = fo_chown(fp, uap->uid, uap->gid, td->td_ucred, td);
3036 	fdrop(fp, td);
3037 	return (error);
3038 }
3039 
3040 /*
3041  * Common implementation code for utimes(), lutimes(), and futimes().
3042  */
3043 static int
3044 getutimes(const struct timeval *usrtvp, enum uio_seg tvpseg,
3045     struct timespec *tsp)
3046 {
3047 	struct timeval tv[2];
3048 	const struct timeval *tvp;
3049 	int error;
3050 
3051 	if (usrtvp == NULL) {
3052 		vfs_timestamp(&tsp[0]);
3053 		tsp[1] = tsp[0];
3054 	} else {
3055 		if (tvpseg == UIO_SYSSPACE) {
3056 			tvp = usrtvp;
3057 		} else {
3058 			if ((error = copyin(usrtvp, tv, sizeof(tv))) != 0)
3059 				return (error);
3060 			tvp = tv;
3061 		}
3062 
3063 		if (tvp[0].tv_usec < 0 || tvp[0].tv_usec >= 1000000 ||
3064 		    tvp[1].tv_usec < 0 || tvp[1].tv_usec >= 1000000)
3065 			return (EINVAL);
3066 		TIMEVAL_TO_TIMESPEC(&tvp[0], &tsp[0]);
3067 		TIMEVAL_TO_TIMESPEC(&tvp[1], &tsp[1]);
3068 	}
3069 	return (0);
3070 }
3071 
3072 /*
3073  * Common implementation code for futimens(), utimensat().
3074  */
3075 #define	UTIMENS_NULL	0x1
3076 #define	UTIMENS_EXIT	0x2
3077 static int
3078 getutimens(const struct timespec *usrtsp, enum uio_seg tspseg,
3079     struct timespec *tsp, int *retflags)
3080 {
3081 	struct timespec tsnow;
3082 	int error;
3083 
3084 	vfs_timestamp(&tsnow);
3085 	*retflags = 0;
3086 	if (usrtsp == NULL) {
3087 		tsp[0] = tsnow;
3088 		tsp[1] = tsnow;
3089 		*retflags |= UTIMENS_NULL;
3090 		return (0);
3091 	}
3092 	if (tspseg == UIO_SYSSPACE) {
3093 		tsp[0] = usrtsp[0];
3094 		tsp[1] = usrtsp[1];
3095 	} else if ((error = copyin(usrtsp, tsp, sizeof(*tsp) * 2)) != 0)
3096 		return (error);
3097 	if (tsp[0].tv_nsec == UTIME_OMIT && tsp[1].tv_nsec == UTIME_OMIT)
3098 		*retflags |= UTIMENS_EXIT;
3099 	if (tsp[0].tv_nsec == UTIME_NOW && tsp[1].tv_nsec == UTIME_NOW)
3100 		*retflags |= UTIMENS_NULL;
3101 	if (tsp[0].tv_nsec == UTIME_OMIT)
3102 		tsp[0].tv_sec = VNOVAL;
3103 	else if (tsp[0].tv_nsec == UTIME_NOW)
3104 		tsp[0] = tsnow;
3105 	else if (tsp[0].tv_nsec < 0 || tsp[0].tv_nsec >= 1000000000L)
3106 		return (EINVAL);
3107 	if (tsp[1].tv_nsec == UTIME_OMIT)
3108 		tsp[1].tv_sec = VNOVAL;
3109 	else if (tsp[1].tv_nsec == UTIME_NOW)
3110 		tsp[1] = tsnow;
3111 	else if (tsp[1].tv_nsec < 0 || tsp[1].tv_nsec >= 1000000000L)
3112 		return (EINVAL);
3113 
3114 	return (0);
3115 }
3116 
3117 /*
3118  * Common implementation code for utimes(), lutimes(), futimes(), futimens(),
3119  * and utimensat().
3120  */
3121 static int
3122 setutimes(struct thread *td, struct vnode *vp, const struct timespec *ts,
3123     int numtimes, int nullflag)
3124 {
3125 	struct mount *mp;
3126 	struct vattr vattr;
3127 	int error, setbirthtime;
3128 
3129 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
3130 		return (error);
3131 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3132 	setbirthtime = 0;
3133 	if (numtimes < 3 && !VOP_GETATTR(vp, &vattr, td->td_ucred) &&
3134 	    timespeccmp(&ts[1], &vattr.va_birthtime, < ))
3135 		setbirthtime = 1;
3136 	VATTR_NULL(&vattr);
3137 	vattr.va_atime = ts[0];
3138 	vattr.va_mtime = ts[1];
3139 	if (setbirthtime)
3140 		vattr.va_birthtime = ts[1];
3141 	if (numtimes > 2)
3142 		vattr.va_birthtime = ts[2];
3143 	if (nullflag)
3144 		vattr.va_vaflags |= VA_UTIMES_NULL;
3145 #ifdef MAC
3146 	error = mac_vnode_check_setutimes(td->td_ucred, vp, vattr.va_atime,
3147 	    vattr.va_mtime);
3148 #endif
3149 	if (error == 0)
3150 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3151 	VOP_UNLOCK(vp);
3152 	vn_finished_write(mp);
3153 	return (error);
3154 }
3155 
3156 /*
3157  * Set the access and modification times of a file.
3158  */
3159 #ifndef _SYS_SYSPROTO_H_
3160 struct utimes_args {
3161 	char	*path;
3162 	struct	timeval *tptr;
3163 };
3164 #endif
3165 int
3166 sys_utimes(struct thread *td, struct utimes_args *uap)
3167 {
3168 
3169 	return (kern_utimesat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
3170 	    uap->tptr, UIO_USERSPACE));
3171 }
3172 
3173 #ifndef _SYS_SYSPROTO_H_
3174 struct futimesat_args {
3175 	int fd;
3176 	const char * path;
3177 	const struct timeval * times;
3178 };
3179 #endif
3180 int
3181 sys_futimesat(struct thread *td, struct futimesat_args *uap)
3182 {
3183 
3184 	return (kern_utimesat(td, uap->fd, uap->path, UIO_USERSPACE,
3185 	    uap->times, UIO_USERSPACE));
3186 }
3187 
3188 int
3189 kern_utimesat(struct thread *td, int fd, const char *path,
3190     enum uio_seg pathseg, struct timeval *tptr, enum uio_seg tptrseg)
3191 {
3192 	struct nameidata nd;
3193 	struct timespec ts[2];
3194 	int error;
3195 
3196 	if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3197 		return (error);
3198 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, fd,
3199 	    &cap_futimes_rights, td);
3200 
3201 	if ((error = namei(&nd)) != 0)
3202 		return (error);
3203 	NDFREE_NOTHING(&nd);
3204 	error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3205 	vrele(nd.ni_vp);
3206 	return (error);
3207 }
3208 
3209 /*
3210  * Set the access and modification times of a file.
3211  */
3212 #ifndef _SYS_SYSPROTO_H_
3213 struct lutimes_args {
3214 	char	*path;
3215 	struct	timeval *tptr;
3216 };
3217 #endif
3218 int
3219 sys_lutimes(struct thread *td, struct lutimes_args *uap)
3220 {
3221 
3222 	return (kern_lutimes(td, uap->path, UIO_USERSPACE, uap->tptr,
3223 	    UIO_USERSPACE));
3224 }
3225 
3226 int
3227 kern_lutimes(struct thread *td, const char *path, enum uio_seg pathseg,
3228     struct timeval *tptr, enum uio_seg tptrseg)
3229 {
3230 	struct timespec ts[2];
3231 	struct nameidata nd;
3232 	int error;
3233 
3234 	if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3235 		return (error);
3236 	NDINIT(&nd, LOOKUP, NOFOLLOW | AUDITVNODE1, pathseg, path, td);
3237 	if ((error = namei(&nd)) != 0)
3238 		return (error);
3239 	NDFREE_NOTHING(&nd);
3240 	error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3241 	vrele(nd.ni_vp);
3242 	return (error);
3243 }
3244 
3245 /*
3246  * Set the access and modification times of a file.
3247  */
3248 #ifndef _SYS_SYSPROTO_H_
3249 struct futimes_args {
3250 	int	fd;
3251 	struct	timeval *tptr;
3252 };
3253 #endif
3254 int
3255 sys_futimes(struct thread *td, struct futimes_args *uap)
3256 {
3257 
3258 	return (kern_futimes(td, uap->fd, uap->tptr, UIO_USERSPACE));
3259 }
3260 
3261 int
3262 kern_futimes(struct thread *td, int fd, struct timeval *tptr,
3263     enum uio_seg tptrseg)
3264 {
3265 	struct timespec ts[2];
3266 	struct file *fp;
3267 	int error;
3268 
3269 	AUDIT_ARG_FD(fd);
3270 	error = getutimes(tptr, tptrseg, ts);
3271 	if (error != 0)
3272 		return (error);
3273 	error = getvnode(td, fd, &cap_futimes_rights, &fp);
3274 	if (error != 0)
3275 		return (error);
3276 #ifdef AUDIT
3277 	vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
3278 	AUDIT_ARG_VNODE1(fp->f_vnode);
3279 	VOP_UNLOCK(fp->f_vnode);
3280 #endif
3281 	error = setutimes(td, fp->f_vnode, ts, 2, tptr == NULL);
3282 	fdrop(fp, td);
3283 	return (error);
3284 }
3285 
3286 int
3287 sys_futimens(struct thread *td, struct futimens_args *uap)
3288 {
3289 
3290 	return (kern_futimens(td, uap->fd, uap->times, UIO_USERSPACE));
3291 }
3292 
3293 int
3294 kern_futimens(struct thread *td, int fd, struct timespec *tptr,
3295     enum uio_seg tptrseg)
3296 {
3297 	struct timespec ts[2];
3298 	struct file *fp;
3299 	int error, flags;
3300 
3301 	AUDIT_ARG_FD(fd);
3302 	error = getutimens(tptr, tptrseg, ts, &flags);
3303 	if (error != 0)
3304 		return (error);
3305 	if (flags & UTIMENS_EXIT)
3306 		return (0);
3307 	error = getvnode(td, fd, &cap_futimes_rights, &fp);
3308 	if (error != 0)
3309 		return (error);
3310 #ifdef AUDIT
3311 	vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
3312 	AUDIT_ARG_VNODE1(fp->f_vnode);
3313 	VOP_UNLOCK(fp->f_vnode);
3314 #endif
3315 	error = setutimes(td, fp->f_vnode, ts, 2, flags & UTIMENS_NULL);
3316 	fdrop(fp, td);
3317 	return (error);
3318 }
3319 
3320 int
3321 sys_utimensat(struct thread *td, struct utimensat_args *uap)
3322 {
3323 
3324 	return (kern_utimensat(td, uap->fd, uap->path, UIO_USERSPACE,
3325 	    uap->times, UIO_USERSPACE, uap->flag));
3326 }
3327 
3328 int
3329 kern_utimensat(struct thread *td, int fd, const char *path,
3330     enum uio_seg pathseg, struct timespec *tptr, enum uio_seg tptrseg,
3331     int flag)
3332 {
3333 	struct nameidata nd;
3334 	struct timespec ts[2];
3335 	int error, flags;
3336 
3337 	if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH)) != 0)
3338 		return (EINVAL);
3339 
3340 	if ((error = getutimens(tptr, tptrseg, ts, &flags)) != 0)
3341 		return (error);
3342 	NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(flag, AT_SYMLINK_NOFOLLOW |
3343 	    AT_RESOLVE_BENEATH) | AUDITVNODE1,
3344 	    pathseg, path, fd, &cap_futimes_rights, td);
3345 	if ((error = namei(&nd)) != 0)
3346 		return (error);
3347 	/*
3348 	 * We are allowed to call namei() regardless of 2xUTIME_OMIT.
3349 	 * POSIX states:
3350 	 * "If both tv_nsec fields are UTIME_OMIT... EACCESS may be detected."
3351 	 * "Search permission is denied by a component of the path prefix."
3352 	 */
3353 	NDFREE_NOTHING(&nd);
3354 	if ((flags & UTIMENS_EXIT) == 0)
3355 		error = setutimes(td, nd.ni_vp, ts, 2, flags & UTIMENS_NULL);
3356 	vrele(nd.ni_vp);
3357 	return (error);
3358 }
3359 
3360 /*
3361  * Truncate a file given its path name.
3362  */
3363 #ifndef _SYS_SYSPROTO_H_
3364 struct truncate_args {
3365 	char	*path;
3366 	int	pad;
3367 	off_t	length;
3368 };
3369 #endif
3370 int
3371 sys_truncate(struct thread *td, struct truncate_args *uap)
3372 {
3373 
3374 	return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3375 }
3376 
3377 int
3378 kern_truncate(struct thread *td, const char *path, enum uio_seg pathseg,
3379     off_t length)
3380 {
3381 	struct mount *mp;
3382 	struct vnode *vp;
3383 	void *rl_cookie;
3384 	struct vattr vattr;
3385 	struct nameidata nd;
3386 	int error;
3387 
3388 	if (length < 0)
3389 		return (EINVAL);
3390 retry:
3391 	NDINIT(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, td);
3392 	if ((error = namei(&nd)) != 0)
3393 		return (error);
3394 	vp = nd.ni_vp;
3395 	rl_cookie = vn_rangelock_wlock(vp, 0, OFF_MAX);
3396 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) {
3397 		vn_rangelock_unlock(vp, rl_cookie);
3398 		vrele(vp);
3399 		return (error);
3400 	}
3401 	NDFREE(&nd, NDF_ONLY_PNBUF);
3402 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3403 	if (vp->v_type == VDIR)
3404 		error = EISDIR;
3405 #ifdef MAC
3406 	else if ((error = mac_vnode_check_write(td->td_ucred, NOCRED, vp))) {
3407 	}
3408 #endif
3409 	else if ((error = vn_writechk(vp)) == 0 &&
3410 	    (error = VOP_ACCESS(vp, VWRITE, td->td_ucred, td)) == 0) {
3411 		VATTR_NULL(&vattr);
3412 		vattr.va_size = length;
3413 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3414 	}
3415 	VOP_UNLOCK(vp);
3416 	vn_finished_write(mp);
3417 	vn_rangelock_unlock(vp, rl_cookie);
3418 	vrele(vp);
3419 	if (error == ERELOOKUP)
3420 		goto retry;
3421 	return (error);
3422 }
3423 
3424 #if defined(COMPAT_43)
3425 /*
3426  * Truncate a file given its path name.
3427  */
3428 #ifndef _SYS_SYSPROTO_H_
3429 struct otruncate_args {
3430 	char	*path;
3431 	long	length;
3432 };
3433 #endif
3434 int
3435 otruncate(struct thread *td, struct otruncate_args *uap)
3436 {
3437 
3438 	return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3439 }
3440 #endif /* COMPAT_43 */
3441 
3442 #if defined(COMPAT_FREEBSD6)
3443 /* Versions with the pad argument */
3444 int
3445 freebsd6_truncate(struct thread *td, struct freebsd6_truncate_args *uap)
3446 {
3447 
3448 	return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3449 }
3450 
3451 int
3452 freebsd6_ftruncate(struct thread *td, struct freebsd6_ftruncate_args *uap)
3453 {
3454 
3455 	return (kern_ftruncate(td, uap->fd, uap->length));
3456 }
3457 #endif
3458 
3459 int
3460 kern_fsync(struct thread *td, int fd, bool fullsync)
3461 {
3462 	struct vnode *vp;
3463 	struct mount *mp;
3464 	struct file *fp;
3465 	int error, lock_flags;
3466 
3467 	AUDIT_ARG_FD(fd);
3468 	error = getvnode(td, fd, &cap_fsync_rights, &fp);
3469 	if (error != 0)
3470 		return (error);
3471 	vp = fp->f_vnode;
3472 #if 0
3473 	if (!fullsync)
3474 		/* XXXKIB: compete outstanding aio writes */;
3475 #endif
3476 retry:
3477 	error = vn_start_write(vp, &mp, V_WAIT | PCATCH);
3478 	if (error != 0)
3479 		goto drop;
3480 	if (MNT_SHARED_WRITES(mp) ||
3481 	    ((mp == NULL) && MNT_SHARED_WRITES(vp->v_mount))) {
3482 		lock_flags = LK_SHARED;
3483 	} else {
3484 		lock_flags = LK_EXCLUSIVE;
3485 	}
3486 	vn_lock(vp, lock_flags | LK_RETRY);
3487 	AUDIT_ARG_VNODE1(vp);
3488 	if (vp->v_object != NULL) {
3489 		VM_OBJECT_WLOCK(vp->v_object);
3490 		vm_object_page_clean(vp->v_object, 0, 0, 0);
3491 		VM_OBJECT_WUNLOCK(vp->v_object);
3492 	}
3493 	error = fullsync ? VOP_FSYNC(vp, MNT_WAIT, td) : VOP_FDATASYNC(vp, td);
3494 	VOP_UNLOCK(vp);
3495 	vn_finished_write(mp);
3496 	if (error == ERELOOKUP)
3497 		goto retry;
3498 drop:
3499 	fdrop(fp, td);
3500 	return (error);
3501 }
3502 
3503 /*
3504  * Sync an open file.
3505  */
3506 #ifndef _SYS_SYSPROTO_H_
3507 struct fsync_args {
3508 	int	fd;
3509 };
3510 #endif
3511 int
3512 sys_fsync(struct thread *td, struct fsync_args *uap)
3513 {
3514 
3515 	return (kern_fsync(td, uap->fd, true));
3516 }
3517 
3518 int
3519 sys_fdatasync(struct thread *td, struct fdatasync_args *uap)
3520 {
3521 
3522 	return (kern_fsync(td, uap->fd, false));
3523 }
3524 
3525 /*
3526  * Rename files.  Source and destination must either both be directories, or
3527  * both not be directories.  If target is a directory, it must be empty.
3528  */
3529 #ifndef _SYS_SYSPROTO_H_
3530 struct rename_args {
3531 	char	*from;
3532 	char	*to;
3533 };
3534 #endif
3535 int
3536 sys_rename(struct thread *td, struct rename_args *uap)
3537 {
3538 
3539 	return (kern_renameat(td, AT_FDCWD, uap->from, AT_FDCWD,
3540 	    uap->to, UIO_USERSPACE));
3541 }
3542 
3543 #ifndef _SYS_SYSPROTO_H_
3544 struct renameat_args {
3545 	int	oldfd;
3546 	char	*old;
3547 	int	newfd;
3548 	char	*new;
3549 };
3550 #endif
3551 int
3552 sys_renameat(struct thread *td, struct renameat_args *uap)
3553 {
3554 
3555 	return (kern_renameat(td, uap->oldfd, uap->old, uap->newfd, uap->new,
3556 	    UIO_USERSPACE));
3557 }
3558 
3559 #ifdef MAC
3560 static int
3561 kern_renameat_mac(struct thread *td, int oldfd, const char *old, int newfd,
3562     const char *new, enum uio_seg pathseg, struct nameidata *fromnd)
3563 {
3564 	int error;
3565 
3566 	NDINIT_ATRIGHTS(fromnd, DELETE, LOCKPARENT | LOCKLEAF | SAVESTART |
3567 	    AUDITVNODE1, pathseg, old, oldfd, &cap_renameat_source_rights, td);
3568 	if ((error = namei(fromnd)) != 0)
3569 		return (error);
3570 	error = mac_vnode_check_rename_from(td->td_ucred, fromnd->ni_dvp,
3571 	    fromnd->ni_vp, &fromnd->ni_cnd);
3572 	VOP_UNLOCK(fromnd->ni_dvp);
3573 	if (fromnd->ni_dvp != fromnd->ni_vp)
3574 		VOP_UNLOCK(fromnd->ni_vp);
3575 	if (error != 0) {
3576 		NDFREE(fromnd, NDF_ONLY_PNBUF);
3577 		vrele(fromnd->ni_dvp);
3578 		vrele(fromnd->ni_vp);
3579 		if (fromnd->ni_startdir)
3580 			vrele(fromnd->ni_startdir);
3581 	}
3582 	return (error);
3583 }
3584 #endif
3585 
3586 int
3587 kern_renameat(struct thread *td, int oldfd, const char *old, int newfd,
3588     const char *new, enum uio_seg pathseg)
3589 {
3590 	struct mount *mp = NULL;
3591 	struct vnode *tvp, *fvp, *tdvp;
3592 	struct nameidata fromnd, tond;
3593 	u_int64_t tondflags;
3594 	int error;
3595 
3596 again:
3597 	bwillwrite();
3598 #ifdef MAC
3599 	if (mac_vnode_check_rename_from_enabled()) {
3600 		error = kern_renameat_mac(td, oldfd, old, newfd, new, pathseg,
3601 		    &fromnd);
3602 		if (error != 0)
3603 			return (error);
3604 	} else {
3605 #endif
3606 	NDINIT_ATRIGHTS(&fromnd, DELETE, WANTPARENT | SAVESTART | AUDITVNODE1,
3607 	    pathseg, old, oldfd, &cap_renameat_source_rights, td);
3608 	if ((error = namei(&fromnd)) != 0)
3609 		return (error);
3610 #ifdef MAC
3611 	}
3612 #endif
3613 	fvp = fromnd.ni_vp;
3614 	tondflags = LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART | AUDITVNODE2;
3615 	if (fromnd.ni_vp->v_type == VDIR)
3616 		tondflags |= WILLBEDIR;
3617 	NDINIT_ATRIGHTS(&tond, RENAME, tondflags, pathseg, new, newfd,
3618 	    &cap_renameat_target_rights, td);
3619 	if ((error = namei(&tond)) != 0) {
3620 		/* Translate error code for rename("dir1", "dir2/."). */
3621 		if (error == EISDIR && fvp->v_type == VDIR)
3622 			error = EINVAL;
3623 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3624 		vrele(fromnd.ni_dvp);
3625 		vrele(fvp);
3626 		goto out1;
3627 	}
3628 	tdvp = tond.ni_dvp;
3629 	tvp = tond.ni_vp;
3630 	error = vn_start_write(fvp, &mp, V_NOWAIT);
3631 	if (error != 0) {
3632 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3633 		NDFREE(&tond, NDF_ONLY_PNBUF);
3634 		if (tvp != NULL)
3635 			vput(tvp);
3636 		if (tdvp == tvp)
3637 			vrele(tdvp);
3638 		else
3639 			vput(tdvp);
3640 		vrele(fromnd.ni_dvp);
3641 		vrele(fvp);
3642 		vrele(tond.ni_startdir);
3643 		if (fromnd.ni_startdir != NULL)
3644 			vrele(fromnd.ni_startdir);
3645 		error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH);
3646 		if (error != 0)
3647 			return (error);
3648 		goto again;
3649 	}
3650 	if (tvp != NULL) {
3651 		if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
3652 			error = ENOTDIR;
3653 			goto out;
3654 		} else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
3655 			error = EISDIR;
3656 			goto out;
3657 		}
3658 #ifdef CAPABILITIES
3659 		if (newfd != AT_FDCWD && (tond.ni_resflags & NIRES_ABS) == 0) {
3660 			/*
3661 			 * If the target already exists we require CAP_UNLINKAT
3662 			 * from 'newfd', when newfd was used for the lookup.
3663 			 */
3664 			error = cap_check(&tond.ni_filecaps.fc_rights,
3665 			    &cap_unlinkat_rights);
3666 			if (error != 0)
3667 				goto out;
3668 		}
3669 #endif
3670 	}
3671 	if (fvp == tdvp) {
3672 		error = EINVAL;
3673 		goto out;
3674 	}
3675 	/*
3676 	 * If the source is the same as the destination (that is, if they
3677 	 * are links to the same vnode), then there is nothing to do.
3678 	 */
3679 	if (fvp == tvp)
3680 		error = ERESTART;
3681 #ifdef MAC
3682 	else
3683 		error = mac_vnode_check_rename_to(td->td_ucred, tdvp,
3684 		    tond.ni_vp, fromnd.ni_dvp == tdvp, &tond.ni_cnd);
3685 #endif
3686 out:
3687 	if (error == 0) {
3688 		error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
3689 		    tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
3690 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3691 		NDFREE(&tond, NDF_ONLY_PNBUF);
3692 	} else {
3693 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3694 		NDFREE(&tond, NDF_ONLY_PNBUF);
3695 		if (tvp != NULL)
3696 			vput(tvp);
3697 		if (tdvp == tvp)
3698 			vrele(tdvp);
3699 		else
3700 			vput(tdvp);
3701 		vrele(fromnd.ni_dvp);
3702 		vrele(fvp);
3703 	}
3704 	vrele(tond.ni_startdir);
3705 	vn_finished_write(mp);
3706 out1:
3707 	if (fromnd.ni_startdir)
3708 		vrele(fromnd.ni_startdir);
3709 	if (error == ERESTART)
3710 		return (0);
3711 	if (error == ERELOOKUP)
3712 		goto again;
3713 	return (error);
3714 }
3715 
3716 /*
3717  * Make a directory file.
3718  */
3719 #ifndef _SYS_SYSPROTO_H_
3720 struct mkdir_args {
3721 	char	*path;
3722 	int	mode;
3723 };
3724 #endif
3725 int
3726 sys_mkdir(struct thread *td, struct mkdir_args *uap)
3727 {
3728 
3729 	return (kern_mkdirat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
3730 	    uap->mode));
3731 }
3732 
3733 #ifndef _SYS_SYSPROTO_H_
3734 struct mkdirat_args {
3735 	int	fd;
3736 	char	*path;
3737 	mode_t	mode;
3738 };
3739 #endif
3740 int
3741 sys_mkdirat(struct thread *td, struct mkdirat_args *uap)
3742 {
3743 
3744 	return (kern_mkdirat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode));
3745 }
3746 
3747 int
3748 kern_mkdirat(struct thread *td, int fd, const char *path, enum uio_seg segflg,
3749     int mode)
3750 {
3751 	struct mount *mp;
3752 	struct vattr vattr;
3753 	struct nameidata nd;
3754 	int error;
3755 
3756 	AUDIT_ARG_MODE(mode);
3757 restart:
3758 	bwillwrite();
3759 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
3760 	    NC_NOMAKEENTRY | NC_KEEPPOSENTRY | FAILIFEXISTS | WILLBEDIR,
3761 	    segflg, path, fd, &cap_mkdirat_rights, td);
3762 	if ((error = namei(&nd)) != 0)
3763 		return (error);
3764 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3765 		NDFREE(&nd, NDF_ONLY_PNBUF);
3766 		vput(nd.ni_dvp);
3767 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3768 			return (error);
3769 		goto restart;
3770 	}
3771 	VATTR_NULL(&vattr);
3772 	vattr.va_type = VDIR;
3773 	vattr.va_mode = (mode & ACCESSPERMS) &~ td->td_proc->p_pd->pd_cmask;
3774 #ifdef MAC
3775 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
3776 	    &vattr);
3777 	if (error != 0)
3778 		goto out;
3779 #endif
3780 	error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
3781 #ifdef MAC
3782 out:
3783 #endif
3784 	NDFREE(&nd, NDF_ONLY_PNBUF);
3785 	VOP_VPUT_PAIR(nd.ni_dvp, error == 0 ? &nd.ni_vp : NULL, true);
3786 	vn_finished_write(mp);
3787 	if (error == ERELOOKUP)
3788 		goto restart;
3789 	return (error);
3790 }
3791 
3792 /*
3793  * Remove a directory file.
3794  */
3795 #ifndef _SYS_SYSPROTO_H_
3796 struct rmdir_args {
3797 	char	*path;
3798 };
3799 #endif
3800 int
3801 sys_rmdir(struct thread *td, struct rmdir_args *uap)
3802 {
3803 
3804 	return (kern_frmdirat(td, AT_FDCWD, uap->path, FD_NONE, UIO_USERSPACE,
3805 	    0));
3806 }
3807 
3808 int
3809 kern_frmdirat(struct thread *td, int dfd, const char *path, int fd,
3810     enum uio_seg pathseg, int flag)
3811 {
3812 	struct mount *mp;
3813 	struct vnode *vp;
3814 	struct file *fp;
3815 	struct nameidata nd;
3816 	cap_rights_t rights;
3817 	int error;
3818 
3819 	fp = NULL;
3820 	if (fd != FD_NONE) {
3821 		error = getvnode(td, fd, cap_rights_init_one(&rights, CAP_LOOKUP),
3822 		    &fp);
3823 		if (error != 0)
3824 			return (error);
3825 	}
3826 
3827 restart:
3828 	bwillwrite();
3829 	NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1 |
3830 	    at2cnpflags(flag, AT_RESOLVE_BENEATH),
3831 	    pathseg, path, dfd, &cap_unlinkat_rights, td);
3832 	if ((error = namei(&nd)) != 0)
3833 		goto fdout;
3834 	vp = nd.ni_vp;
3835 	if (vp->v_type != VDIR) {
3836 		error = ENOTDIR;
3837 		goto out;
3838 	}
3839 	/*
3840 	 * No rmdir "." please.
3841 	 */
3842 	if (nd.ni_dvp == vp) {
3843 		error = EINVAL;
3844 		goto out;
3845 	}
3846 	/*
3847 	 * The root of a mounted filesystem cannot be deleted.
3848 	 */
3849 	if (vp->v_vflag & VV_ROOT) {
3850 		error = EBUSY;
3851 		goto out;
3852 	}
3853 
3854 	if (fp != NULL && fp->f_vnode != vp) {
3855 		if (VN_IS_DOOMED(fp->f_vnode))
3856 			error = EBADF;
3857 		else
3858 			error = EDEADLK;
3859 		goto out;
3860 	}
3861 
3862 #ifdef MAC
3863 	error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
3864 	    &nd.ni_cnd);
3865 	if (error != 0)
3866 		goto out;
3867 #endif
3868 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3869 		NDFREE(&nd, NDF_ONLY_PNBUF);
3870 		vput(vp);
3871 		if (nd.ni_dvp == vp)
3872 			vrele(nd.ni_dvp);
3873 		else
3874 			vput(nd.ni_dvp);
3875 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3876 			goto fdout;
3877 		goto restart;
3878 	}
3879 	vfs_notify_upper(vp, VFS_NOTIFY_UPPER_UNLINK);
3880 	error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
3881 	vn_finished_write(mp);
3882 out:
3883 	NDFREE(&nd, NDF_ONLY_PNBUF);
3884 	vput(vp);
3885 	if (nd.ni_dvp == vp)
3886 		vrele(nd.ni_dvp);
3887 	else
3888 		vput(nd.ni_dvp);
3889 	if (error == ERELOOKUP)
3890 		goto restart;
3891 fdout:
3892 	if (fp != NULL)
3893 		fdrop(fp, td);
3894 	return (error);
3895 }
3896 
3897 #if defined(COMPAT_43) || defined(COMPAT_FREEBSD11)
3898 int
3899 freebsd11_kern_getdirentries(struct thread *td, int fd, char *ubuf, u_int count,
3900     long *basep, void (*func)(struct freebsd11_dirent *))
3901 {
3902 	struct freebsd11_dirent dstdp;
3903 	struct dirent *dp, *edp;
3904 	char *dirbuf;
3905 	off_t base;
3906 	ssize_t resid, ucount;
3907 	int error;
3908 
3909 	/* XXX arbitrary sanity limit on `count'. */
3910 	count = min(count, 64 * 1024);
3911 
3912 	dirbuf = malloc(count, M_TEMP, M_WAITOK);
3913 
3914 	error = kern_getdirentries(td, fd, dirbuf, count, &base, &resid,
3915 	    UIO_SYSSPACE);
3916 	if (error != 0)
3917 		goto done;
3918 	if (basep != NULL)
3919 		*basep = base;
3920 
3921 	ucount = 0;
3922 	for (dp = (struct dirent *)dirbuf,
3923 	    edp = (struct dirent *)&dirbuf[count - resid];
3924 	    ucount < count && dp < edp; ) {
3925 		if (dp->d_reclen == 0)
3926 			break;
3927 		MPASS(dp->d_reclen >= _GENERIC_DIRLEN(0));
3928 		if (dp->d_namlen >= sizeof(dstdp.d_name))
3929 			continue;
3930 		dstdp.d_type = dp->d_type;
3931 		dstdp.d_namlen = dp->d_namlen;
3932 		dstdp.d_fileno = dp->d_fileno;		/* truncate */
3933 		if (dstdp.d_fileno != dp->d_fileno) {
3934 			switch (ino64_trunc_error) {
3935 			default:
3936 			case 0:
3937 				break;
3938 			case 1:
3939 				error = EOVERFLOW;
3940 				goto done;
3941 			case 2:
3942 				dstdp.d_fileno = UINT32_MAX;
3943 				break;
3944 			}
3945 		}
3946 		dstdp.d_reclen = sizeof(dstdp) - sizeof(dstdp.d_name) +
3947 		    ((dp->d_namlen + 1 + 3) &~ 3);
3948 		bcopy(dp->d_name, dstdp.d_name, dstdp.d_namlen);
3949 		bzero(dstdp.d_name + dstdp.d_namlen,
3950 		    dstdp.d_reclen - offsetof(struct freebsd11_dirent, d_name) -
3951 		    dstdp.d_namlen);
3952 		MPASS(dstdp.d_reclen <= dp->d_reclen);
3953 		MPASS(ucount + dstdp.d_reclen <= count);
3954 		if (func != NULL)
3955 			func(&dstdp);
3956 		error = copyout(&dstdp, ubuf + ucount, dstdp.d_reclen);
3957 		if (error != 0)
3958 			break;
3959 		dp = (struct dirent *)((char *)dp + dp->d_reclen);
3960 		ucount += dstdp.d_reclen;
3961 	}
3962 
3963 done:
3964 	free(dirbuf, M_TEMP);
3965 	if (error == 0)
3966 		td->td_retval[0] = ucount;
3967 	return (error);
3968 }
3969 #endif /* COMPAT */
3970 
3971 #ifdef COMPAT_43
3972 static void
3973 ogetdirentries_cvt(struct freebsd11_dirent *dp)
3974 {
3975 #if (BYTE_ORDER == LITTLE_ENDIAN)
3976 	/*
3977 	 * The expected low byte of dp->d_namlen is our dp->d_type.
3978 	 * The high MBZ byte of dp->d_namlen is our dp->d_namlen.
3979 	 */
3980 	dp->d_type = dp->d_namlen;
3981 	dp->d_namlen = 0;
3982 #else
3983 	/*
3984 	 * The dp->d_type is the high byte of the expected dp->d_namlen,
3985 	 * so must be zero'ed.
3986 	 */
3987 	dp->d_type = 0;
3988 #endif
3989 }
3990 
3991 /*
3992  * Read a block of directory entries in a filesystem independent format.
3993  */
3994 #ifndef _SYS_SYSPROTO_H_
3995 struct ogetdirentries_args {
3996 	int	fd;
3997 	char	*buf;
3998 	u_int	count;
3999 	long	*basep;
4000 };
4001 #endif
4002 int
4003 ogetdirentries(struct thread *td, struct ogetdirentries_args *uap)
4004 {
4005 	long loff;
4006 	int error;
4007 
4008 	error = kern_ogetdirentries(td, uap, &loff);
4009 	if (error == 0)
4010 		error = copyout(&loff, uap->basep, sizeof(long));
4011 	return (error);
4012 }
4013 
4014 int
4015 kern_ogetdirentries(struct thread *td, struct ogetdirentries_args *uap,
4016     long *ploff)
4017 {
4018 	long base;
4019 	int error;
4020 
4021 	/* XXX arbitrary sanity limit on `count'. */
4022 	if (uap->count > 64 * 1024)
4023 		return (EINVAL);
4024 
4025 	error = freebsd11_kern_getdirentries(td, uap->fd, uap->buf, uap->count,
4026 	    &base, ogetdirentries_cvt);
4027 
4028 	if (error == 0 && uap->basep != NULL)
4029 		error = copyout(&base, uap->basep, sizeof(long));
4030 
4031 	return (error);
4032 }
4033 #endif /* COMPAT_43 */
4034 
4035 #if defined(COMPAT_FREEBSD11)
4036 #ifndef _SYS_SYSPROTO_H_
4037 struct freebsd11_getdirentries_args {
4038 	int	fd;
4039 	char	*buf;
4040 	u_int	count;
4041 	long	*basep;
4042 };
4043 #endif
4044 int
4045 freebsd11_getdirentries(struct thread *td,
4046     struct freebsd11_getdirentries_args *uap)
4047 {
4048 	long base;
4049 	int error;
4050 
4051 	error = freebsd11_kern_getdirentries(td, uap->fd, uap->buf, uap->count,
4052 	    &base, NULL);
4053 
4054 	if (error == 0 && uap->basep != NULL)
4055 		error = copyout(&base, uap->basep, sizeof(long));
4056 	return (error);
4057 }
4058 
4059 int
4060 freebsd11_getdents(struct thread *td, struct freebsd11_getdents_args *uap)
4061 {
4062 	struct freebsd11_getdirentries_args ap;
4063 
4064 	ap.fd = uap->fd;
4065 	ap.buf = uap->buf;
4066 	ap.count = uap->count;
4067 	ap.basep = NULL;
4068 	return (freebsd11_getdirentries(td, &ap));
4069 }
4070 #endif /* COMPAT_FREEBSD11 */
4071 
4072 /*
4073  * Read a block of directory entries in a filesystem independent format.
4074  */
4075 int
4076 sys_getdirentries(struct thread *td, struct getdirentries_args *uap)
4077 {
4078 	off_t base;
4079 	int error;
4080 
4081 	error = kern_getdirentries(td, uap->fd, uap->buf, uap->count, &base,
4082 	    NULL, UIO_USERSPACE);
4083 	if (error != 0)
4084 		return (error);
4085 	if (uap->basep != NULL)
4086 		error = copyout(&base, uap->basep, sizeof(off_t));
4087 	return (error);
4088 }
4089 
4090 int
4091 kern_getdirentries(struct thread *td, int fd, char *buf, size_t count,
4092     off_t *basep, ssize_t *residp, enum uio_seg bufseg)
4093 {
4094 	struct vnode *vp;
4095 	struct file *fp;
4096 	struct uio auio;
4097 	struct iovec aiov;
4098 	off_t loff;
4099 	int error, eofflag;
4100 	off_t foffset;
4101 
4102 	AUDIT_ARG_FD(fd);
4103 	if (count > IOSIZE_MAX)
4104 		return (EINVAL);
4105 	auio.uio_resid = count;
4106 	error = getvnode(td, fd, &cap_read_rights, &fp);
4107 	if (error != 0)
4108 		return (error);
4109 	if ((fp->f_flag & FREAD) == 0) {
4110 		fdrop(fp, td);
4111 		return (EBADF);
4112 	}
4113 	vp = fp->f_vnode;
4114 	foffset = foffset_lock(fp, 0);
4115 unionread:
4116 	if (vp->v_type != VDIR) {
4117 		error = EINVAL;
4118 		goto fail;
4119 	}
4120 	aiov.iov_base = buf;
4121 	aiov.iov_len = count;
4122 	auio.uio_iov = &aiov;
4123 	auio.uio_iovcnt = 1;
4124 	auio.uio_rw = UIO_READ;
4125 	auio.uio_segflg = bufseg;
4126 	auio.uio_td = td;
4127 	vn_lock(vp, LK_SHARED | LK_RETRY);
4128 	AUDIT_ARG_VNODE1(vp);
4129 	loff = auio.uio_offset = foffset;
4130 #ifdef MAC
4131 	error = mac_vnode_check_readdir(td->td_ucred, vp);
4132 	if (error == 0)
4133 #endif
4134 		error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, NULL,
4135 		    NULL);
4136 	foffset = auio.uio_offset;
4137 	if (error != 0) {
4138 		VOP_UNLOCK(vp);
4139 		goto fail;
4140 	}
4141 	if (count == auio.uio_resid &&
4142 	    (vp->v_vflag & VV_ROOT) &&
4143 	    (vp->v_mount->mnt_flag & MNT_UNION)) {
4144 		struct vnode *tvp = vp;
4145 
4146 		vp = vp->v_mount->mnt_vnodecovered;
4147 		VREF(vp);
4148 		fp->f_vnode = vp;
4149 		foffset = 0;
4150 		vput(tvp);
4151 		goto unionread;
4152 	}
4153 	VOP_UNLOCK(vp);
4154 	*basep = loff;
4155 	if (residp != NULL)
4156 		*residp = auio.uio_resid;
4157 	td->td_retval[0] = count - auio.uio_resid;
4158 fail:
4159 	foffset_unlock(fp, foffset, 0);
4160 	fdrop(fp, td);
4161 	return (error);
4162 }
4163 
4164 /*
4165  * Set the mode mask for creation of filesystem nodes.
4166  */
4167 #ifndef _SYS_SYSPROTO_H_
4168 struct umask_args {
4169 	int	newmask;
4170 };
4171 #endif
4172 int
4173 sys_umask(struct thread *td, struct umask_args *uap)
4174 {
4175 	struct pwddesc *pdp;
4176 
4177 	pdp = td->td_proc->p_pd;
4178 	PWDDESC_XLOCK(pdp);
4179 	td->td_retval[0] = pdp->pd_cmask;
4180 	pdp->pd_cmask = uap->newmask & ALLPERMS;
4181 	PWDDESC_XUNLOCK(pdp);
4182 	return (0);
4183 }
4184 
4185 /*
4186  * Void all references to file by ripping underlying filesystem away from
4187  * vnode.
4188  */
4189 #ifndef _SYS_SYSPROTO_H_
4190 struct revoke_args {
4191 	char	*path;
4192 };
4193 #endif
4194 int
4195 sys_revoke(struct thread *td, struct revoke_args *uap)
4196 {
4197 	struct vnode *vp;
4198 	struct vattr vattr;
4199 	struct nameidata nd;
4200 	int error;
4201 
4202 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
4203 	    uap->path, td);
4204 	if ((error = namei(&nd)) != 0)
4205 		return (error);
4206 	vp = nd.ni_vp;
4207 	NDFREE_NOTHING(&nd);
4208 	if (vp->v_type != VCHR || vp->v_rdev == NULL) {
4209 		error = EINVAL;
4210 		goto out;
4211 	}
4212 #ifdef MAC
4213 	error = mac_vnode_check_revoke(td->td_ucred, vp);
4214 	if (error != 0)
4215 		goto out;
4216 #endif
4217 	error = VOP_GETATTR(vp, &vattr, td->td_ucred);
4218 	if (error != 0)
4219 		goto out;
4220 	if (td->td_ucred->cr_uid != vattr.va_uid) {
4221 		error = priv_check(td, PRIV_VFS_ADMIN);
4222 		if (error != 0)
4223 			goto out;
4224 	}
4225 	if (devfs_usecount(vp) > 0)
4226 		VOP_REVOKE(vp, REVOKEALL);
4227 out:
4228 	vput(vp);
4229 	return (error);
4230 }
4231 
4232 /*
4233  * Convert a user file descriptor to a kernel file entry and check that, if it
4234  * is a capability, the correct rights are present. A reference on the file
4235  * entry is held upon returning.
4236  */
4237 int
4238 getvnode(struct thread *td, int fd, cap_rights_t *rightsp, struct file **fpp)
4239 {
4240 	struct file *fp;
4241 	int error;
4242 
4243 	error = fget_unlocked(td->td_proc->p_fd, fd, rightsp, &fp);
4244 	if (error != 0)
4245 		return (error);
4246 
4247 	/*
4248 	 * The file could be not of the vnode type, or it may be not
4249 	 * yet fully initialized, in which case the f_vnode pointer
4250 	 * may be set, but f_ops is still badfileops.  E.g.,
4251 	 * devfs_open() transiently create such situation to
4252 	 * facilitate csw d_fdopen().
4253 	 *
4254 	 * Dupfdopen() handling in kern_openat() installs the
4255 	 * half-baked file into the process descriptor table, allowing
4256 	 * other thread to dereference it. Guard against the race by
4257 	 * checking f_ops.
4258 	 */
4259 	if (fp->f_vnode == NULL || fp->f_ops == &badfileops) {
4260 		fdrop(fp, td);
4261 		return (EINVAL);
4262 	}
4263 	*fpp = fp;
4264 	return (0);
4265 }
4266 
4267 /*
4268  * Get an (NFS) file handle.
4269  */
4270 #ifndef _SYS_SYSPROTO_H_
4271 struct lgetfh_args {
4272 	char *fname;
4273 	fhandle_t *fhp;
4274 };
4275 #endif
4276 int
4277 sys_lgetfh(struct thread *td, struct lgetfh_args *uap)
4278 {
4279 
4280 	return (kern_getfhat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->fname,
4281 	    UIO_USERSPACE, uap->fhp, UIO_USERSPACE));
4282 }
4283 
4284 #ifndef _SYS_SYSPROTO_H_
4285 struct getfh_args {
4286 	char *fname;
4287 	fhandle_t *fhp;
4288 };
4289 #endif
4290 int
4291 sys_getfh(struct thread *td, struct getfh_args *uap)
4292 {
4293 
4294 	return (kern_getfhat(td, 0, AT_FDCWD, uap->fname, UIO_USERSPACE,
4295 	    uap->fhp, UIO_USERSPACE));
4296 }
4297 
4298 /*
4299  * syscall for the rpc.lockd to use to translate an open descriptor into
4300  * a NFS file handle.
4301  *
4302  * warning: do not remove the priv_check() call or this becomes one giant
4303  * security hole.
4304  */
4305 #ifndef _SYS_SYSPROTO_H_
4306 struct getfhat_args {
4307 	int fd;
4308 	char *path;
4309 	fhandle_t *fhp;
4310 	int flags;
4311 };
4312 #endif
4313 int
4314 sys_getfhat(struct thread *td, struct getfhat_args *uap)
4315 {
4316 
4317 	if ((uap->flags & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH)) != 0)
4318 		return (EINVAL);
4319 	return (kern_getfhat(td, uap->flags, uap->fd, uap->path, UIO_USERSPACE,
4320 	    uap->fhp, UIO_USERSPACE));
4321 }
4322 
4323 int
4324 kern_getfhat(struct thread *td, int flags, int fd, const char *path,
4325     enum uio_seg pathseg, fhandle_t *fhp, enum uio_seg fhseg)
4326 {
4327 	struct nameidata nd;
4328 	fhandle_t fh;
4329 	struct vnode *vp;
4330 	int error;
4331 
4332 	error = priv_check(td, PRIV_VFS_GETFH);
4333 	if (error != 0)
4334 		return (error);
4335 	NDINIT_AT(&nd, LOOKUP, at2cnpflags(flags, AT_SYMLINK_NOFOLLOW |
4336 	    AT_RESOLVE_BENEATH) | LOCKLEAF | AUDITVNODE1, pathseg, path,
4337 	    fd, td);
4338 	error = namei(&nd);
4339 	if (error != 0)
4340 		return (error);
4341 	NDFREE_NOTHING(&nd);
4342 	vp = nd.ni_vp;
4343 	bzero(&fh, sizeof(fh));
4344 	fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
4345 	error = VOP_VPTOFH(vp, &fh.fh_fid);
4346 	vput(vp);
4347 	if (error == 0) {
4348 		if (fhseg == UIO_USERSPACE)
4349 			error = copyout(&fh, fhp, sizeof (fh));
4350 		else
4351 			memcpy(fhp, &fh, sizeof(fh));
4352 	}
4353 	return (error);
4354 }
4355 
4356 #ifndef _SYS_SYSPROTO_H_
4357 struct fhlink_args {
4358 	fhandle_t *fhp;
4359 	const char *to;
4360 };
4361 #endif
4362 int
4363 sys_fhlink(struct thread *td, struct fhlink_args *uap)
4364 {
4365 
4366 	return (kern_fhlinkat(td, AT_FDCWD, uap->to, UIO_USERSPACE, uap->fhp));
4367 }
4368 
4369 #ifndef _SYS_SYSPROTO_H_
4370 struct fhlinkat_args {
4371 	fhandle_t *fhp;
4372 	int tofd;
4373 	const char *to;
4374 };
4375 #endif
4376 int
4377 sys_fhlinkat(struct thread *td, struct fhlinkat_args *uap)
4378 {
4379 
4380 	return (kern_fhlinkat(td, uap->tofd, uap->to, UIO_USERSPACE, uap->fhp));
4381 }
4382 
4383 static int
4384 kern_fhlinkat(struct thread *td, int fd, const char *path,
4385     enum uio_seg pathseg, fhandle_t *fhp)
4386 {
4387 	fhandle_t fh;
4388 	struct mount *mp;
4389 	struct vnode *vp;
4390 	int error;
4391 
4392 	error = priv_check(td, PRIV_VFS_GETFH);
4393 	if (error != 0)
4394 		return (error);
4395 	error = copyin(fhp, &fh, sizeof(fh));
4396 	if (error != 0)
4397 		return (error);
4398 	do {
4399 		bwillwrite();
4400 		if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4401 			return (ESTALE);
4402 		error = VFS_FHTOVP(mp, &fh.fh_fid, LK_SHARED, &vp);
4403 		vfs_unbusy(mp);
4404 		if (error != 0)
4405 			return (error);
4406 		VOP_UNLOCK(vp);
4407 		error = kern_linkat_vp(td, vp, fd, path, pathseg);
4408 	} while (error == EAGAIN || error == ERELOOKUP);
4409 	return (error);
4410 }
4411 
4412 #ifndef _SYS_SYSPROTO_H_
4413 struct fhreadlink_args {
4414 	fhandle_t *fhp;
4415 	char *buf;
4416 	size_t bufsize;
4417 };
4418 #endif
4419 int
4420 sys_fhreadlink(struct thread *td, struct fhreadlink_args *uap)
4421 {
4422 	fhandle_t fh;
4423 	struct mount *mp;
4424 	struct vnode *vp;
4425 	int error;
4426 
4427 	error = priv_check(td, PRIV_VFS_GETFH);
4428 	if (error != 0)
4429 		return (error);
4430 	if (uap->bufsize > IOSIZE_MAX)
4431 		return (EINVAL);
4432 	error = copyin(uap->fhp, &fh, sizeof(fh));
4433 	if (error != 0)
4434 		return (error);
4435 	if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4436 		return (ESTALE);
4437 	error = VFS_FHTOVP(mp, &fh.fh_fid, LK_SHARED, &vp);
4438 	vfs_unbusy(mp);
4439 	if (error != 0)
4440 		return (error);
4441 	error = kern_readlink_vp(vp, uap->buf, UIO_USERSPACE, uap->bufsize, td);
4442 	vput(vp);
4443 	return (error);
4444 }
4445 
4446 /*
4447  * syscall for the rpc.lockd to use to translate a NFS file handle into an
4448  * open descriptor.
4449  *
4450  * warning: do not remove the priv_check() call or this becomes one giant
4451  * security hole.
4452  */
4453 #ifndef _SYS_SYSPROTO_H_
4454 struct fhopen_args {
4455 	const struct fhandle *u_fhp;
4456 	int flags;
4457 };
4458 #endif
4459 int
4460 sys_fhopen(struct thread *td, struct fhopen_args *uap)
4461 {
4462 	return (kern_fhopen(td, uap->u_fhp, uap->flags));
4463 }
4464 
4465 int
4466 kern_fhopen(struct thread *td, const struct fhandle *u_fhp, int flags)
4467 {
4468 	struct mount *mp;
4469 	struct vnode *vp;
4470 	struct fhandle fhp;
4471 	struct file *fp;
4472 	int fmode, error;
4473 	int indx;
4474 
4475 	error = priv_check(td, PRIV_VFS_FHOPEN);
4476 	if (error != 0)
4477 		return (error);
4478 	indx = -1;
4479 	fmode = FFLAGS(flags);
4480 	/* why not allow a non-read/write open for our lockd? */
4481 	if (((fmode & (FREAD | FWRITE)) == 0) || (fmode & O_CREAT))
4482 		return (EINVAL);
4483 	error = copyin(u_fhp, &fhp, sizeof(fhp));
4484 	if (error != 0)
4485 		return(error);
4486 	/* find the mount point */
4487 	mp = vfs_busyfs(&fhp.fh_fsid);
4488 	if (mp == NULL)
4489 		return (ESTALE);
4490 	/* now give me my vnode, it gets returned to me locked */
4491 	error = VFS_FHTOVP(mp, &fhp.fh_fid, LK_EXCLUSIVE, &vp);
4492 	vfs_unbusy(mp);
4493 	if (error != 0)
4494 		return (error);
4495 
4496 	error = falloc_noinstall(td, &fp);
4497 	if (error != 0) {
4498 		vput(vp);
4499 		return (error);
4500 	}
4501 	/*
4502 	 * An extra reference on `fp' has been held for us by
4503 	 * falloc_noinstall().
4504 	 */
4505 
4506 #ifdef INVARIANTS
4507 	td->td_dupfd = -1;
4508 #endif
4509 	error = vn_open_vnode(vp, fmode, td->td_ucred, td, fp);
4510 	if (error != 0) {
4511 		KASSERT(fp->f_ops == &badfileops,
4512 		    ("VOP_OPEN in fhopen() set f_ops"));
4513 		KASSERT(td->td_dupfd < 0,
4514 		    ("fhopen() encountered fdopen()"));
4515 
4516 		vput(vp);
4517 		goto bad;
4518 	}
4519 #ifdef INVARIANTS
4520 	td->td_dupfd = 0;
4521 #endif
4522 	fp->f_vnode = vp;
4523 	finit_vnode(fp, fmode, NULL, &vnops);
4524 	VOP_UNLOCK(vp);
4525 	if ((fmode & O_TRUNC) != 0) {
4526 		error = fo_truncate(fp, 0, td->td_ucred, td);
4527 		if (error != 0)
4528 			goto bad;
4529 	}
4530 
4531 	error = finstall(td, fp, &indx, fmode, NULL);
4532 bad:
4533 	fdrop(fp, td);
4534 	td->td_retval[0] = indx;
4535 	return (error);
4536 }
4537 
4538 /*
4539  * Stat an (NFS) file handle.
4540  */
4541 #ifndef _SYS_SYSPROTO_H_
4542 struct fhstat_args {
4543 	struct fhandle *u_fhp;
4544 	struct stat *sb;
4545 };
4546 #endif
4547 int
4548 sys_fhstat(struct thread *td, struct fhstat_args *uap)
4549 {
4550 	struct stat sb;
4551 	struct fhandle fh;
4552 	int error;
4553 
4554 	error = copyin(uap->u_fhp, &fh, sizeof(fh));
4555 	if (error != 0)
4556 		return (error);
4557 	error = kern_fhstat(td, fh, &sb);
4558 	if (error == 0)
4559 		error = copyout(&sb, uap->sb, sizeof(sb));
4560 	return (error);
4561 }
4562 
4563 int
4564 kern_fhstat(struct thread *td, struct fhandle fh, struct stat *sb)
4565 {
4566 	struct mount *mp;
4567 	struct vnode *vp;
4568 	int error;
4569 
4570 	error = priv_check(td, PRIV_VFS_FHSTAT);
4571 	if (error != 0)
4572 		return (error);
4573 	if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4574 		return (ESTALE);
4575 	error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4576 	vfs_unbusy(mp);
4577 	if (error != 0)
4578 		return (error);
4579 	error = VOP_STAT(vp, sb, td->td_ucred, NOCRED, td);
4580 	vput(vp);
4581 	return (error);
4582 }
4583 
4584 /*
4585  * Implement fstatfs() for (NFS) file handles.
4586  */
4587 #ifndef _SYS_SYSPROTO_H_
4588 struct fhstatfs_args {
4589 	struct fhandle *u_fhp;
4590 	struct statfs *buf;
4591 };
4592 #endif
4593 int
4594 sys_fhstatfs(struct thread *td, struct fhstatfs_args *uap)
4595 {
4596 	struct statfs *sfp;
4597 	fhandle_t fh;
4598 	int error;
4599 
4600 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
4601 	if (error != 0)
4602 		return (error);
4603 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
4604 	error = kern_fhstatfs(td, fh, sfp);
4605 	if (error == 0)
4606 		error = copyout(sfp, uap->buf, sizeof(*sfp));
4607 	free(sfp, M_STATFS);
4608 	return (error);
4609 }
4610 
4611 int
4612 kern_fhstatfs(struct thread *td, fhandle_t fh, struct statfs *buf)
4613 {
4614 	struct mount *mp;
4615 	struct vnode *vp;
4616 	int error;
4617 
4618 	error = priv_check(td, PRIV_VFS_FHSTATFS);
4619 	if (error != 0)
4620 		return (error);
4621 	if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4622 		return (ESTALE);
4623 	error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4624 	if (error != 0) {
4625 		vfs_unbusy(mp);
4626 		return (error);
4627 	}
4628 	vput(vp);
4629 	error = prison_canseemount(td->td_ucred, mp);
4630 	if (error != 0)
4631 		goto out;
4632 #ifdef MAC
4633 	error = mac_mount_check_stat(td->td_ucred, mp);
4634 	if (error != 0)
4635 		goto out;
4636 #endif
4637 	error = VFS_STATFS(mp, buf);
4638 out:
4639 	vfs_unbusy(mp);
4640 	return (error);
4641 }
4642 
4643 /*
4644  * Unlike madvise(2), we do not make a best effort to remember every
4645  * possible caching hint.  Instead, we remember the last setting with
4646  * the exception that we will allow POSIX_FADV_NORMAL to adjust the
4647  * region of any current setting.
4648  */
4649 int
4650 kern_posix_fadvise(struct thread *td, int fd, off_t offset, off_t len,
4651     int advice)
4652 {
4653 	struct fadvise_info *fa, *new;
4654 	struct file *fp;
4655 	struct vnode *vp;
4656 	off_t end;
4657 	int error;
4658 
4659 	if (offset < 0 || len < 0 || offset > OFF_MAX - len)
4660 		return (EINVAL);
4661 	AUDIT_ARG_VALUE(advice);
4662 	switch (advice) {
4663 	case POSIX_FADV_SEQUENTIAL:
4664 	case POSIX_FADV_RANDOM:
4665 	case POSIX_FADV_NOREUSE:
4666 		new = malloc(sizeof(*fa), M_FADVISE, M_WAITOK);
4667 		break;
4668 	case POSIX_FADV_NORMAL:
4669 	case POSIX_FADV_WILLNEED:
4670 	case POSIX_FADV_DONTNEED:
4671 		new = NULL;
4672 		break;
4673 	default:
4674 		return (EINVAL);
4675 	}
4676 	/* XXX: CAP_POSIX_FADVISE? */
4677 	AUDIT_ARG_FD(fd);
4678 	error = fget(td, fd, &cap_no_rights, &fp);
4679 	if (error != 0)
4680 		goto out;
4681 	AUDIT_ARG_FILE(td->td_proc, fp);
4682 	if ((fp->f_ops->fo_flags & DFLAG_SEEKABLE) == 0) {
4683 		error = ESPIPE;
4684 		goto out;
4685 	}
4686 	if (fp->f_type != DTYPE_VNODE) {
4687 		error = ENODEV;
4688 		goto out;
4689 	}
4690 	vp = fp->f_vnode;
4691 	if (vp->v_type != VREG) {
4692 		error = ENODEV;
4693 		goto out;
4694 	}
4695 	if (len == 0)
4696 		end = OFF_MAX;
4697 	else
4698 		end = offset + len - 1;
4699 	switch (advice) {
4700 	case POSIX_FADV_SEQUENTIAL:
4701 	case POSIX_FADV_RANDOM:
4702 	case POSIX_FADV_NOREUSE:
4703 		/*
4704 		 * Try to merge any existing non-standard region with
4705 		 * this new region if possible, otherwise create a new
4706 		 * non-standard region for this request.
4707 		 */
4708 		mtx_pool_lock(mtxpool_sleep, fp);
4709 		fa = fp->f_advice;
4710 		if (fa != NULL && fa->fa_advice == advice &&
4711 		    ((fa->fa_start <= end && fa->fa_end >= offset) ||
4712 		    (end != OFF_MAX && fa->fa_start == end + 1) ||
4713 		    (fa->fa_end != OFF_MAX && fa->fa_end + 1 == offset))) {
4714 			if (offset < fa->fa_start)
4715 				fa->fa_start = offset;
4716 			if (end > fa->fa_end)
4717 				fa->fa_end = end;
4718 		} else {
4719 			new->fa_advice = advice;
4720 			new->fa_start = offset;
4721 			new->fa_end = end;
4722 			fp->f_advice = new;
4723 			new = fa;
4724 		}
4725 		mtx_pool_unlock(mtxpool_sleep, fp);
4726 		break;
4727 	case POSIX_FADV_NORMAL:
4728 		/*
4729 		 * If a the "normal" region overlaps with an existing
4730 		 * non-standard region, trim or remove the
4731 		 * non-standard region.
4732 		 */
4733 		mtx_pool_lock(mtxpool_sleep, fp);
4734 		fa = fp->f_advice;
4735 		if (fa != NULL) {
4736 			if (offset <= fa->fa_start && end >= fa->fa_end) {
4737 				new = fa;
4738 				fp->f_advice = NULL;
4739 			} else if (offset <= fa->fa_start &&
4740 			    end >= fa->fa_start)
4741 				fa->fa_start = end + 1;
4742 			else if (offset <= fa->fa_end && end >= fa->fa_end)
4743 				fa->fa_end = offset - 1;
4744 			else if (offset >= fa->fa_start && end <= fa->fa_end) {
4745 				/*
4746 				 * If the "normal" region is a middle
4747 				 * portion of the existing
4748 				 * non-standard region, just remove
4749 				 * the whole thing rather than picking
4750 				 * one side or the other to
4751 				 * preserve.
4752 				 */
4753 				new = fa;
4754 				fp->f_advice = NULL;
4755 			}
4756 		}
4757 		mtx_pool_unlock(mtxpool_sleep, fp);
4758 		break;
4759 	case POSIX_FADV_WILLNEED:
4760 	case POSIX_FADV_DONTNEED:
4761 		error = VOP_ADVISE(vp, offset, end, advice);
4762 		break;
4763 	}
4764 out:
4765 	if (fp != NULL)
4766 		fdrop(fp, td);
4767 	free(new, M_FADVISE);
4768 	return (error);
4769 }
4770 
4771 int
4772 sys_posix_fadvise(struct thread *td, struct posix_fadvise_args *uap)
4773 {
4774 	int error;
4775 
4776 	error = kern_posix_fadvise(td, uap->fd, uap->offset, uap->len,
4777 	    uap->advice);
4778 	return (kern_posix_error(td, error));
4779 }
4780 
4781 int
4782 kern_copy_file_range(struct thread *td, int infd, off_t *inoffp, int outfd,
4783     off_t *outoffp, size_t len, unsigned int flags)
4784 {
4785 	struct file *infp, *outfp;
4786 	struct vnode *invp, *outvp;
4787 	int error;
4788 	size_t retlen;
4789 	void *rl_rcookie, *rl_wcookie;
4790 	off_t savinoff, savoutoff;
4791 
4792 	infp = outfp = NULL;
4793 	rl_rcookie = rl_wcookie = NULL;
4794 	savinoff = -1;
4795 	error = 0;
4796 	retlen = 0;
4797 
4798 	if (flags != 0) {
4799 		error = EINVAL;
4800 		goto out;
4801 	}
4802 	if (len > SSIZE_MAX)
4803 		/*
4804 		 * Although the len argument is size_t, the return argument
4805 		 * is ssize_t (which is signed).  Therefore a size that won't
4806 		 * fit in ssize_t can't be returned.
4807 		 */
4808 		len = SSIZE_MAX;
4809 
4810 	/* Get the file structures for the file descriptors. */
4811 	error = fget_read(td, infd, &cap_read_rights, &infp);
4812 	if (error != 0)
4813 		goto out;
4814 	if (infp->f_ops == &badfileops) {
4815 		error = EBADF;
4816 		goto out;
4817 	}
4818 	if (infp->f_vnode == NULL) {
4819 		error = EINVAL;
4820 		goto out;
4821 	}
4822 	error = fget_write(td, outfd, &cap_write_rights, &outfp);
4823 	if (error != 0)
4824 		goto out;
4825 	if (outfp->f_ops == &badfileops) {
4826 		error = EBADF;
4827 		goto out;
4828 	}
4829 	if (outfp->f_vnode == NULL) {
4830 		error = EINVAL;
4831 		goto out;
4832 	}
4833 
4834 	/* Set the offset pointers to the correct place. */
4835 	if (inoffp == NULL)
4836 		inoffp = &infp->f_offset;
4837 	if (outoffp == NULL)
4838 		outoffp = &outfp->f_offset;
4839 	savinoff = *inoffp;
4840 	savoutoff = *outoffp;
4841 
4842 	invp = infp->f_vnode;
4843 	outvp = outfp->f_vnode;
4844 	/* Sanity check the f_flag bits. */
4845 	if ((outfp->f_flag & (FWRITE | FAPPEND)) != FWRITE ||
4846 	    (infp->f_flag & FREAD) == 0) {
4847 		error = EBADF;
4848 		goto out;
4849 	}
4850 
4851 	/* If len == 0, just return 0. */
4852 	if (len == 0)
4853 		goto out;
4854 
4855 	/*
4856 	 * If infp and outfp refer to the same file, the byte ranges cannot
4857 	 * overlap.
4858 	 */
4859 	if (invp == outvp && ((savinoff <= savoutoff && savinoff + len >
4860 	    savoutoff) || (savinoff > savoutoff && savoutoff + len >
4861 	    savinoff))) {
4862 		error = EINVAL;
4863 		goto out;
4864 	}
4865 
4866 	/* Range lock the byte ranges for both invp and outvp. */
4867 	for (;;) {
4868 		rl_wcookie = vn_rangelock_wlock(outvp, *outoffp, *outoffp +
4869 		    len);
4870 		rl_rcookie = vn_rangelock_tryrlock(invp, *inoffp, *inoffp +
4871 		    len);
4872 		if (rl_rcookie != NULL)
4873 			break;
4874 		vn_rangelock_unlock(outvp, rl_wcookie);
4875 		rl_rcookie = vn_rangelock_rlock(invp, *inoffp, *inoffp + len);
4876 		vn_rangelock_unlock(invp, rl_rcookie);
4877 	}
4878 
4879 	retlen = len;
4880 	error = vn_copy_file_range(invp, inoffp, outvp, outoffp, &retlen,
4881 	    flags, infp->f_cred, outfp->f_cred, td);
4882 out:
4883 	if (rl_rcookie != NULL)
4884 		vn_rangelock_unlock(invp, rl_rcookie);
4885 	if (rl_wcookie != NULL)
4886 		vn_rangelock_unlock(outvp, rl_wcookie);
4887 	if (savinoff != -1 && (error == EINTR || error == ERESTART)) {
4888 		*inoffp = savinoff;
4889 		*outoffp = savoutoff;
4890 	}
4891 	if (outfp != NULL)
4892 		fdrop(outfp, td);
4893 	if (infp != NULL)
4894 		fdrop(infp, td);
4895 	td->td_retval[0] = retlen;
4896 	return (error);
4897 }
4898 
4899 int
4900 sys_copy_file_range(struct thread *td, struct copy_file_range_args *uap)
4901 {
4902 	off_t inoff, outoff, *inoffp, *outoffp;
4903 	int error;
4904 
4905 	inoffp = outoffp = NULL;
4906 	if (uap->inoffp != NULL) {
4907 		error = copyin(uap->inoffp, &inoff, sizeof(off_t));
4908 		if (error != 0)
4909 			return (error);
4910 		inoffp = &inoff;
4911 	}
4912 	if (uap->outoffp != NULL) {
4913 		error = copyin(uap->outoffp, &outoff, sizeof(off_t));
4914 		if (error != 0)
4915 			return (error);
4916 		outoffp = &outoff;
4917 	}
4918 	error = kern_copy_file_range(td, uap->infd, inoffp, uap->outfd,
4919 	    outoffp, uap->len, uap->flags);
4920 	if (error == 0 && uap->inoffp != NULL)
4921 		error = copyout(inoffp, uap->inoffp, sizeof(off_t));
4922 	if (error == 0 && uap->outoffp != NULL)
4923 		error = copyout(outoffp, uap->outoffp, sizeof(off_t));
4924 	return (error);
4925 }
4926