xref: /freebsd/sys/kern/vfs_syscalls.c (revision 7778ab7e0cc22f0824eb1d1047a7ef8b4785267a)
1 /*-
2  * Copyright (c) 1989, 1993
3  *	The Regents of the University of California.  All rights reserved.
4  * (c) UNIX System Laboratories, Inc.
5  * All or some portions of this file are derived from material licensed
6  * to the University of California by American Telephone and Telegraph
7  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
8  * the permission of UNIX System Laboratories, Inc.
9  *
10  * Redistribution and use in source and binary forms, with or without
11  * modification, are permitted provided that the following conditions
12  * are met:
13  * 1. Redistributions of source code must retain the above copyright
14  *    notice, this list of conditions and the following disclaimer.
15  * 2. Redistributions in binary form must reproduce the above copyright
16  *    notice, this list of conditions and the following disclaimer in the
17  *    documentation and/or other materials provided with the distribution.
18  * 4. Neither the name of the University nor the names of its contributors
19  *    may be used to endorse or promote products derived from this software
20  *    without specific prior written permission.
21  *
22  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
23  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
24  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
25  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
26  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
27  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
28  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
29  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
30  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
31  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32  * SUCH DAMAGE.
33  *
34  *	@(#)vfs_syscalls.c	8.13 (Berkeley) 4/15/94
35  */
36 
37 #include <sys/cdefs.h>
38 __FBSDID("$FreeBSD$");
39 
40 #include "opt_capsicum.h"
41 #include "opt_compat.h"
42 #include "opt_kdtrace.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/capability.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/sdt.h>
65 #include <sys/stat.h>
66 #include <sys/sx.h>
67 #include <sys/unistd.h>
68 #include <sys/vnode.h>
69 #include <sys/priv.h>
70 #include <sys/proc.h>
71 #include <sys/dirent.h>
72 #include <sys/jail.h>
73 #include <sys/syscallsubr.h>
74 #include <sys/sysctl.h>
75 #ifdef KTRACE
76 #include <sys/ktrace.h>
77 #endif
78 
79 #include <machine/stdarg.h>
80 
81 #include <security/audit/audit.h>
82 #include <security/mac/mac_framework.h>
83 
84 #include <vm/vm.h>
85 #include <vm/vm_object.h>
86 #include <vm/vm_page.h>
87 #include <vm/uma.h>
88 
89 SDT_PROVIDER_DEFINE(vfs);
90 SDT_PROBE_DEFINE(vfs, , stat, mode, mode);
91 SDT_PROBE_ARGTYPE(vfs, , stat, mode, 0, "char *");
92 SDT_PROBE_ARGTYPE(vfs, , stat, mode, 1, "int");
93 SDT_PROBE_DEFINE(vfs, , stat, reg, reg);
94 SDT_PROBE_ARGTYPE(vfs, , stat, reg, 0, "char *");
95 SDT_PROBE_ARGTYPE(vfs, , stat, reg, 1, "int");
96 
97 static int chroot_refuse_vdir_fds(struct filedesc *fdp);
98 static int getutimes(const struct timeval *, enum uio_seg, struct timespec *);
99 static int setfflags(struct thread *td, struct vnode *, int);
100 static int setutimes(struct thread *td, struct vnode *,
101     const struct timespec *, int, int);
102 static int vn_access(struct vnode *vp, int user_flags, struct ucred *cred,
103     struct thread *td);
104 
105 /*
106  * The module initialization routine for POSIX asynchronous I/O will
107  * set this to the version of AIO that it implements.  (Zero means
108  * that it is not implemented.)  This value is used here by pathconf()
109  * and in kern_descrip.c by fpathconf().
110  */
111 int async_io_version;
112 
113 #ifdef DEBUG
114 static int syncprt = 0;
115 SYSCTL_INT(_debug, OID_AUTO, syncprt, CTLFLAG_RW, &syncprt, 0, "");
116 #endif
117 
118 /*
119  * Sync each mounted filesystem.
120  */
121 #ifndef _SYS_SYSPROTO_H_
122 struct sync_args {
123 	int     dummy;
124 };
125 #endif
126 /* ARGSUSED */
127 int
128 sys_sync(td, uap)
129 	struct thread *td;
130 	struct sync_args *uap;
131 {
132 	struct mount *mp, *nmp;
133 	int vfslocked;
134 
135 	mtx_lock(&mountlist_mtx);
136 	for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
137 		if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK)) {
138 			nmp = TAILQ_NEXT(mp, mnt_list);
139 			continue;
140 		}
141 		vfslocked = VFS_LOCK_GIANT(mp);
142 		if ((mp->mnt_flag & MNT_RDONLY) == 0 &&
143 		    vn_start_write(NULL, &mp, V_NOWAIT) == 0) {
144 			MNT_ILOCK(mp);
145 			mp->mnt_noasync++;
146 			mp->mnt_kern_flag &= ~MNTK_ASYNC;
147 			MNT_IUNLOCK(mp);
148 			vfs_msync(mp, MNT_NOWAIT);
149 			VFS_SYNC(mp, MNT_NOWAIT);
150 			MNT_ILOCK(mp);
151 			mp->mnt_noasync--;
152 			if ((mp->mnt_flag & MNT_ASYNC) != 0 &&
153 			    mp->mnt_noasync == 0)
154 				mp->mnt_kern_flag |= MNTK_ASYNC;
155 			MNT_IUNLOCK(mp);
156 			vn_finished_write(mp);
157 		}
158 		VFS_UNLOCK_GIANT(vfslocked);
159 		mtx_lock(&mountlist_mtx);
160 		nmp = TAILQ_NEXT(mp, mnt_list);
161 		vfs_unbusy(mp);
162 	}
163 	mtx_unlock(&mountlist_mtx);
164 	return (0);
165 }
166 
167 /*
168  * Change filesystem quotas.
169  */
170 #ifndef _SYS_SYSPROTO_H_
171 struct quotactl_args {
172 	char *path;
173 	int cmd;
174 	int uid;
175 	caddr_t arg;
176 };
177 #endif
178 int
179 sys_quotactl(td, uap)
180 	struct thread *td;
181 	register struct quotactl_args /* {
182 		char *path;
183 		int cmd;
184 		int uid;
185 		caddr_t arg;
186 	} */ *uap;
187 {
188 	struct mount *mp;
189 	int vfslocked;
190 	int error;
191 	struct nameidata nd;
192 
193 	AUDIT_ARG_CMD(uap->cmd);
194 	AUDIT_ARG_UID(uap->uid);
195 	if (!prison_allow(td->td_ucred, PR_ALLOW_QUOTAS))
196 		return (EPERM);
197 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | MPSAFE | AUDITVNODE1,
198 	   UIO_USERSPACE, uap->path, td);
199 	if ((error = namei(&nd)) != 0)
200 		return (error);
201 	vfslocked = NDHASGIANT(&nd);
202 	NDFREE(&nd, NDF_ONLY_PNBUF);
203 	mp = nd.ni_vp->v_mount;
204 	vfs_ref(mp);
205 	vput(nd.ni_vp);
206 	error = vfs_busy(mp, 0);
207 	vfs_rel(mp);
208 	if (error) {
209 		VFS_UNLOCK_GIANT(vfslocked);
210 		return (error);
211 	}
212 	error = VFS_QUOTACTL(mp, uap->cmd, uap->uid, uap->arg);
213 	vfs_unbusy(mp);
214 	VFS_UNLOCK_GIANT(vfslocked);
215 	return (error);
216 }
217 
218 /*
219  * Used by statfs conversion routines to scale the block size up if
220  * necessary so that all of the block counts are <= 'max_size'.  Note
221  * that 'max_size' should be a bitmask, i.e. 2^n - 1 for some non-zero
222  * value of 'n'.
223  */
224 void
225 statfs_scale_blocks(struct statfs *sf, long max_size)
226 {
227 	uint64_t count;
228 	int shift;
229 
230 	KASSERT(powerof2(max_size + 1), ("%s: invalid max_size", __func__));
231 
232 	/*
233 	 * Attempt to scale the block counts to give a more accurate
234 	 * overview to userland of the ratio of free space to used
235 	 * space.  To do this, find the largest block count and compute
236 	 * a divisor that lets it fit into a signed integer <= max_size.
237 	 */
238 	if (sf->f_bavail < 0)
239 		count = -sf->f_bavail;
240 	else
241 		count = sf->f_bavail;
242 	count = MAX(sf->f_blocks, MAX(sf->f_bfree, count));
243 	if (count <= max_size)
244 		return;
245 
246 	count >>= flsl(max_size);
247 	shift = 0;
248 	while (count > 0) {
249 		shift++;
250 		count >>=1;
251 	}
252 
253 	sf->f_bsize <<= shift;
254 	sf->f_blocks >>= shift;
255 	sf->f_bfree >>= shift;
256 	sf->f_bavail >>= shift;
257 }
258 
259 /*
260  * Get filesystem statistics.
261  */
262 #ifndef _SYS_SYSPROTO_H_
263 struct statfs_args {
264 	char *path;
265 	struct statfs *buf;
266 };
267 #endif
268 int
269 sys_statfs(td, uap)
270 	struct thread *td;
271 	register struct statfs_args /* {
272 		char *path;
273 		struct statfs *buf;
274 	} */ *uap;
275 {
276 	struct statfs sf;
277 	int error;
278 
279 	error = kern_statfs(td, uap->path, UIO_USERSPACE, &sf);
280 	if (error == 0)
281 		error = copyout(&sf, uap->buf, sizeof(sf));
282 	return (error);
283 }
284 
285 int
286 kern_statfs(struct thread *td, char *path, enum uio_seg pathseg,
287     struct statfs *buf)
288 {
289 	struct mount *mp;
290 	struct statfs *sp, sb;
291 	int vfslocked;
292 	int error;
293 	struct nameidata nd;
294 
295 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | MPSAFE |
296 	    AUDITVNODE1, pathseg, path, td);
297 	error = namei(&nd);
298 	if (error)
299 		return (error);
300 	vfslocked = NDHASGIANT(&nd);
301 	mp = nd.ni_vp->v_mount;
302 	vfs_ref(mp);
303 	NDFREE(&nd, NDF_ONLY_PNBUF);
304 	vput(nd.ni_vp);
305 	error = vfs_busy(mp, 0);
306 	vfs_rel(mp);
307 	if (error) {
308 		VFS_UNLOCK_GIANT(vfslocked);
309 		return (error);
310 	}
311 #ifdef MAC
312 	error = mac_mount_check_stat(td->td_ucred, mp);
313 	if (error)
314 		goto out;
315 #endif
316 	/*
317 	 * Set these in case the underlying filesystem fails to do so.
318 	 */
319 	sp = &mp->mnt_stat;
320 	sp->f_version = STATFS_VERSION;
321 	sp->f_namemax = NAME_MAX;
322 	sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
323 	error = VFS_STATFS(mp, sp);
324 	if (error)
325 		goto out;
326 	if (priv_check(td, PRIV_VFS_GENERATION)) {
327 		bcopy(sp, &sb, sizeof(sb));
328 		sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
329 		prison_enforce_statfs(td->td_ucred, mp, &sb);
330 		sp = &sb;
331 	}
332 	*buf = *sp;
333 out:
334 	vfs_unbusy(mp);
335 	VFS_UNLOCK_GIANT(vfslocked);
336 	return (error);
337 }
338 
339 /*
340  * Get filesystem statistics.
341  */
342 #ifndef _SYS_SYSPROTO_H_
343 struct fstatfs_args {
344 	int fd;
345 	struct statfs *buf;
346 };
347 #endif
348 int
349 sys_fstatfs(td, uap)
350 	struct thread *td;
351 	register struct fstatfs_args /* {
352 		int fd;
353 		struct statfs *buf;
354 	} */ *uap;
355 {
356 	struct statfs sf;
357 	int error;
358 
359 	error = kern_fstatfs(td, uap->fd, &sf);
360 	if (error == 0)
361 		error = copyout(&sf, uap->buf, sizeof(sf));
362 	return (error);
363 }
364 
365 int
366 kern_fstatfs(struct thread *td, int fd, struct statfs *buf)
367 {
368 	struct file *fp;
369 	struct mount *mp;
370 	struct statfs *sp, sb;
371 	int vfslocked;
372 	struct vnode *vp;
373 	int error;
374 
375 	AUDIT_ARG_FD(fd);
376 	error = getvnode(td->td_proc->p_fd, fd, CAP_FSTATFS, &fp);
377 	if (error)
378 		return (error);
379 	vp = fp->f_vnode;
380 	vfslocked = VFS_LOCK_GIANT(vp->v_mount);
381 	vn_lock(vp, LK_SHARED | LK_RETRY);
382 #ifdef AUDIT
383 	AUDIT_ARG_VNODE1(vp);
384 #endif
385 	mp = vp->v_mount;
386 	if (mp)
387 		vfs_ref(mp);
388 	VOP_UNLOCK(vp, 0);
389 	fdrop(fp, td);
390 	if (mp == NULL) {
391 		error = EBADF;
392 		goto out;
393 	}
394 	error = vfs_busy(mp, 0);
395 	vfs_rel(mp);
396 	if (error) {
397 		VFS_UNLOCK_GIANT(vfslocked);
398 		return (error);
399 	}
400 #ifdef MAC
401 	error = mac_mount_check_stat(td->td_ucred, mp);
402 	if (error)
403 		goto out;
404 #endif
405 	/*
406 	 * Set these in case the underlying filesystem fails to do so.
407 	 */
408 	sp = &mp->mnt_stat;
409 	sp->f_version = STATFS_VERSION;
410 	sp->f_namemax = NAME_MAX;
411 	sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
412 	error = VFS_STATFS(mp, sp);
413 	if (error)
414 		goto out;
415 	if (priv_check(td, PRIV_VFS_GENERATION)) {
416 		bcopy(sp, &sb, sizeof(sb));
417 		sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
418 		prison_enforce_statfs(td->td_ucred, mp, &sb);
419 		sp = &sb;
420 	}
421 	*buf = *sp;
422 out:
423 	if (mp)
424 		vfs_unbusy(mp);
425 	VFS_UNLOCK_GIANT(vfslocked);
426 	return (error);
427 }
428 
429 /*
430  * Get statistics on all filesystems.
431  */
432 #ifndef _SYS_SYSPROTO_H_
433 struct getfsstat_args {
434 	struct statfs *buf;
435 	long bufsize;
436 	int flags;
437 };
438 #endif
439 int
440 sys_getfsstat(td, uap)
441 	struct thread *td;
442 	register struct getfsstat_args /* {
443 		struct statfs *buf;
444 		long bufsize;
445 		int flags;
446 	} */ *uap;
447 {
448 
449 	return (kern_getfsstat(td, &uap->buf, uap->bufsize, UIO_USERSPACE,
450 	    uap->flags));
451 }
452 
453 /*
454  * If (bufsize > 0 && bufseg == UIO_SYSSPACE)
455  * 	The caller is responsible for freeing memory which will be allocated
456  *	in '*buf'.
457  */
458 int
459 kern_getfsstat(struct thread *td, struct statfs **buf, size_t bufsize,
460     enum uio_seg bufseg, int flags)
461 {
462 	struct mount *mp, *nmp;
463 	struct statfs *sfsp, *sp, sb;
464 	size_t count, maxcount;
465 	int vfslocked;
466 	int error;
467 
468 	maxcount = bufsize / sizeof(struct statfs);
469 	if (bufsize == 0)
470 		sfsp = NULL;
471 	else if (bufseg == UIO_USERSPACE)
472 		sfsp = *buf;
473 	else /* if (bufseg == UIO_SYSSPACE) */ {
474 		count = 0;
475 		mtx_lock(&mountlist_mtx);
476 		TAILQ_FOREACH(mp, &mountlist, mnt_list) {
477 			count++;
478 		}
479 		mtx_unlock(&mountlist_mtx);
480 		if (maxcount > count)
481 			maxcount = count;
482 		sfsp = *buf = malloc(maxcount * sizeof(struct statfs), M_TEMP,
483 		    M_WAITOK);
484 	}
485 	count = 0;
486 	mtx_lock(&mountlist_mtx);
487 	for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
488 		if (prison_canseemount(td->td_ucred, mp) != 0) {
489 			nmp = TAILQ_NEXT(mp, mnt_list);
490 			continue;
491 		}
492 #ifdef MAC
493 		if (mac_mount_check_stat(td->td_ucred, mp) != 0) {
494 			nmp = TAILQ_NEXT(mp, mnt_list);
495 			continue;
496 		}
497 #endif
498 		if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK)) {
499 			nmp = TAILQ_NEXT(mp, mnt_list);
500 			continue;
501 		}
502 		vfslocked = VFS_LOCK_GIANT(mp);
503 		if (sfsp && count < maxcount) {
504 			sp = &mp->mnt_stat;
505 			/*
506 			 * Set these in case the underlying filesystem
507 			 * fails to do so.
508 			 */
509 			sp->f_version = STATFS_VERSION;
510 			sp->f_namemax = NAME_MAX;
511 			sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
512 			/*
513 			 * If MNT_NOWAIT or MNT_LAZY is specified, do not
514 			 * refresh the fsstat cache. MNT_NOWAIT or MNT_LAZY
515 			 * overrides MNT_WAIT.
516 			 */
517 			if (((flags & (MNT_LAZY|MNT_NOWAIT)) == 0 ||
518 			    (flags & MNT_WAIT)) &&
519 			    (error = VFS_STATFS(mp, sp))) {
520 				VFS_UNLOCK_GIANT(vfslocked);
521 				mtx_lock(&mountlist_mtx);
522 				nmp = TAILQ_NEXT(mp, mnt_list);
523 				vfs_unbusy(mp);
524 				continue;
525 			}
526 			if (priv_check(td, PRIV_VFS_GENERATION)) {
527 				bcopy(sp, &sb, sizeof(sb));
528 				sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
529 				prison_enforce_statfs(td->td_ucred, mp, &sb);
530 				sp = &sb;
531 			}
532 			if (bufseg == UIO_SYSSPACE)
533 				bcopy(sp, sfsp, sizeof(*sp));
534 			else /* if (bufseg == UIO_USERSPACE) */ {
535 				error = copyout(sp, sfsp, sizeof(*sp));
536 				if (error) {
537 					vfs_unbusy(mp);
538 					VFS_UNLOCK_GIANT(vfslocked);
539 					return (error);
540 				}
541 			}
542 			sfsp++;
543 		}
544 		VFS_UNLOCK_GIANT(vfslocked);
545 		count++;
546 		mtx_lock(&mountlist_mtx);
547 		nmp = TAILQ_NEXT(mp, mnt_list);
548 		vfs_unbusy(mp);
549 	}
550 	mtx_unlock(&mountlist_mtx);
551 	if (sfsp && count > maxcount)
552 		td->td_retval[0] = maxcount;
553 	else
554 		td->td_retval[0] = count;
555 	return (0);
556 }
557 
558 #ifdef COMPAT_FREEBSD4
559 /*
560  * Get old format filesystem statistics.
561  */
562 static void cvtstatfs(struct statfs *, struct ostatfs *);
563 
564 #ifndef _SYS_SYSPROTO_H_
565 struct freebsd4_statfs_args {
566 	char *path;
567 	struct ostatfs *buf;
568 };
569 #endif
570 int
571 freebsd4_statfs(td, uap)
572 	struct thread *td;
573 	struct freebsd4_statfs_args /* {
574 		char *path;
575 		struct ostatfs *buf;
576 	} */ *uap;
577 {
578 	struct ostatfs osb;
579 	struct statfs sf;
580 	int error;
581 
582 	error = kern_statfs(td, uap->path, UIO_USERSPACE, &sf);
583 	if (error)
584 		return (error);
585 	cvtstatfs(&sf, &osb);
586 	return (copyout(&osb, uap->buf, sizeof(osb)));
587 }
588 
589 /*
590  * Get filesystem statistics.
591  */
592 #ifndef _SYS_SYSPROTO_H_
593 struct freebsd4_fstatfs_args {
594 	int fd;
595 	struct ostatfs *buf;
596 };
597 #endif
598 int
599 freebsd4_fstatfs(td, uap)
600 	struct thread *td;
601 	struct freebsd4_fstatfs_args /* {
602 		int fd;
603 		struct ostatfs *buf;
604 	} */ *uap;
605 {
606 	struct ostatfs osb;
607 	struct statfs sf;
608 	int error;
609 
610 	error = kern_fstatfs(td, uap->fd, &sf);
611 	if (error)
612 		return (error);
613 	cvtstatfs(&sf, &osb);
614 	return (copyout(&osb, uap->buf, sizeof(osb)));
615 }
616 
617 /*
618  * Get statistics on all filesystems.
619  */
620 #ifndef _SYS_SYSPROTO_H_
621 struct freebsd4_getfsstat_args {
622 	struct ostatfs *buf;
623 	long bufsize;
624 	int flags;
625 };
626 #endif
627 int
628 freebsd4_getfsstat(td, uap)
629 	struct thread *td;
630 	register struct freebsd4_getfsstat_args /* {
631 		struct ostatfs *buf;
632 		long bufsize;
633 		int flags;
634 	} */ *uap;
635 {
636 	struct statfs *buf, *sp;
637 	struct ostatfs osb;
638 	size_t count, size;
639 	int error;
640 
641 	count = uap->bufsize / sizeof(struct ostatfs);
642 	size = count * sizeof(struct statfs);
643 	error = kern_getfsstat(td, &buf, size, UIO_SYSSPACE, uap->flags);
644 	if (size > 0) {
645 		count = td->td_retval[0];
646 		sp = buf;
647 		while (count > 0 && error == 0) {
648 			cvtstatfs(sp, &osb);
649 			error = copyout(&osb, uap->buf, sizeof(osb));
650 			sp++;
651 			uap->buf++;
652 			count--;
653 		}
654 		free(buf, M_TEMP);
655 	}
656 	return (error);
657 }
658 
659 /*
660  * Implement fstatfs() for (NFS) file handles.
661  */
662 #ifndef _SYS_SYSPROTO_H_
663 struct freebsd4_fhstatfs_args {
664 	struct fhandle *u_fhp;
665 	struct ostatfs *buf;
666 };
667 #endif
668 int
669 freebsd4_fhstatfs(td, uap)
670 	struct thread *td;
671 	struct freebsd4_fhstatfs_args /* {
672 		struct fhandle *u_fhp;
673 		struct ostatfs *buf;
674 	} */ *uap;
675 {
676 	struct ostatfs osb;
677 	struct statfs sf;
678 	fhandle_t fh;
679 	int error;
680 
681 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
682 	if (error)
683 		return (error);
684 	error = kern_fhstatfs(td, fh, &sf);
685 	if (error)
686 		return (error);
687 	cvtstatfs(&sf, &osb);
688 	return (copyout(&osb, uap->buf, sizeof(osb)));
689 }
690 
691 /*
692  * Convert a new format statfs structure to an old format statfs structure.
693  */
694 static void
695 cvtstatfs(nsp, osp)
696 	struct statfs *nsp;
697 	struct ostatfs *osp;
698 {
699 
700 	statfs_scale_blocks(nsp, LONG_MAX);
701 	bzero(osp, sizeof(*osp));
702 	osp->f_bsize = nsp->f_bsize;
703 	osp->f_iosize = MIN(nsp->f_iosize, LONG_MAX);
704 	osp->f_blocks = nsp->f_blocks;
705 	osp->f_bfree = nsp->f_bfree;
706 	osp->f_bavail = nsp->f_bavail;
707 	osp->f_files = MIN(nsp->f_files, LONG_MAX);
708 	osp->f_ffree = MIN(nsp->f_ffree, LONG_MAX);
709 	osp->f_owner = nsp->f_owner;
710 	osp->f_type = nsp->f_type;
711 	osp->f_flags = nsp->f_flags;
712 	osp->f_syncwrites = MIN(nsp->f_syncwrites, LONG_MAX);
713 	osp->f_asyncwrites = MIN(nsp->f_asyncwrites, LONG_MAX);
714 	osp->f_syncreads = MIN(nsp->f_syncreads, LONG_MAX);
715 	osp->f_asyncreads = MIN(nsp->f_asyncreads, LONG_MAX);
716 	strlcpy(osp->f_fstypename, nsp->f_fstypename,
717 	    MIN(MFSNAMELEN, OMFSNAMELEN));
718 	strlcpy(osp->f_mntonname, nsp->f_mntonname,
719 	    MIN(MNAMELEN, OMNAMELEN));
720 	strlcpy(osp->f_mntfromname, nsp->f_mntfromname,
721 	    MIN(MNAMELEN, OMNAMELEN));
722 	osp->f_fsid = nsp->f_fsid;
723 }
724 #endif /* COMPAT_FREEBSD4 */
725 
726 /*
727  * Change current working directory to a given file descriptor.
728  */
729 #ifndef _SYS_SYSPROTO_H_
730 struct fchdir_args {
731 	int	fd;
732 };
733 #endif
734 int
735 sys_fchdir(td, uap)
736 	struct thread *td;
737 	struct fchdir_args /* {
738 		int fd;
739 	} */ *uap;
740 {
741 	register struct filedesc *fdp = td->td_proc->p_fd;
742 	struct vnode *vp, *tdp, *vpold;
743 	struct mount *mp;
744 	struct file *fp;
745 	int vfslocked;
746 	int error;
747 
748 	AUDIT_ARG_FD(uap->fd);
749 	if ((error = getvnode(fdp, uap->fd, CAP_FCHDIR, &fp)) != 0)
750 		return (error);
751 	vp = fp->f_vnode;
752 	VREF(vp);
753 	fdrop(fp, td);
754 	vfslocked = VFS_LOCK_GIANT(vp->v_mount);
755 	vn_lock(vp, LK_SHARED | LK_RETRY);
756 	AUDIT_ARG_VNODE1(vp);
757 	error = change_dir(vp, td);
758 	while (!error && (mp = vp->v_mountedhere) != NULL) {
759 		int tvfslocked;
760 		if (vfs_busy(mp, 0))
761 			continue;
762 		tvfslocked = VFS_LOCK_GIANT(mp);
763 		error = VFS_ROOT(mp, LK_SHARED, &tdp);
764 		vfs_unbusy(mp);
765 		if (error) {
766 			VFS_UNLOCK_GIANT(tvfslocked);
767 			break;
768 		}
769 		vput(vp);
770 		VFS_UNLOCK_GIANT(vfslocked);
771 		vp = tdp;
772 		vfslocked = tvfslocked;
773 	}
774 	if (error) {
775 		vput(vp);
776 		VFS_UNLOCK_GIANT(vfslocked);
777 		return (error);
778 	}
779 	VOP_UNLOCK(vp, 0);
780 	VFS_UNLOCK_GIANT(vfslocked);
781 	FILEDESC_XLOCK(fdp);
782 	vpold = fdp->fd_cdir;
783 	fdp->fd_cdir = vp;
784 	FILEDESC_XUNLOCK(fdp);
785 	vfslocked = VFS_LOCK_GIANT(vpold->v_mount);
786 	vrele(vpold);
787 	VFS_UNLOCK_GIANT(vfslocked);
788 	return (0);
789 }
790 
791 /*
792  * Change current working directory (``.'').
793  */
794 #ifndef _SYS_SYSPROTO_H_
795 struct chdir_args {
796 	char	*path;
797 };
798 #endif
799 int
800 sys_chdir(td, uap)
801 	struct thread *td;
802 	struct chdir_args /* {
803 		char *path;
804 	} */ *uap;
805 {
806 
807 	return (kern_chdir(td, uap->path, UIO_USERSPACE));
808 }
809 
810 int
811 kern_chdir(struct thread *td, char *path, enum uio_seg pathseg)
812 {
813 	register struct filedesc *fdp = td->td_proc->p_fd;
814 	int error;
815 	struct nameidata nd;
816 	struct vnode *vp;
817 	int vfslocked;
818 
819 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1 |
820 	    MPSAFE, pathseg, path, td);
821 	if ((error = namei(&nd)) != 0)
822 		return (error);
823 	vfslocked = NDHASGIANT(&nd);
824 	if ((error = change_dir(nd.ni_vp, td)) != 0) {
825 		vput(nd.ni_vp);
826 		VFS_UNLOCK_GIANT(vfslocked);
827 		NDFREE(&nd, NDF_ONLY_PNBUF);
828 		return (error);
829 	}
830 	VOP_UNLOCK(nd.ni_vp, 0);
831 	VFS_UNLOCK_GIANT(vfslocked);
832 	NDFREE(&nd, NDF_ONLY_PNBUF);
833 	FILEDESC_XLOCK(fdp);
834 	vp = fdp->fd_cdir;
835 	fdp->fd_cdir = nd.ni_vp;
836 	FILEDESC_XUNLOCK(fdp);
837 	vfslocked = VFS_LOCK_GIANT(vp->v_mount);
838 	vrele(vp);
839 	VFS_UNLOCK_GIANT(vfslocked);
840 	return (0);
841 }
842 
843 /*
844  * Helper function for raised chroot(2) security function:  Refuse if
845  * any filedescriptors are open directories.
846  */
847 static int
848 chroot_refuse_vdir_fds(fdp)
849 	struct filedesc *fdp;
850 {
851 	struct vnode *vp;
852 	struct file *fp;
853 	int fd;
854 
855 	FILEDESC_LOCK_ASSERT(fdp);
856 
857 	for (fd = 0; fd < fdp->fd_nfiles ; fd++) {
858 		fp = fget_locked(fdp, fd);
859 		if (fp == NULL)
860 			continue;
861 		if (fp->f_type == DTYPE_VNODE) {
862 			vp = fp->f_vnode;
863 			if (vp->v_type == VDIR)
864 				return (EPERM);
865 		}
866 	}
867 	return (0);
868 }
869 
870 /*
871  * This sysctl determines if we will allow a process to chroot(2) if it
872  * has a directory open:
873  *	0: disallowed for all processes.
874  *	1: allowed for processes that were not already chroot(2)'ed.
875  *	2: allowed for all processes.
876  */
877 
878 static int chroot_allow_open_directories = 1;
879 
880 SYSCTL_INT(_kern, OID_AUTO, chroot_allow_open_directories, CTLFLAG_RW,
881      &chroot_allow_open_directories, 0, "");
882 
883 /*
884  * Change notion of root (``/'') directory.
885  */
886 #ifndef _SYS_SYSPROTO_H_
887 struct chroot_args {
888 	char	*path;
889 };
890 #endif
891 int
892 sys_chroot(td, uap)
893 	struct thread *td;
894 	struct chroot_args /* {
895 		char *path;
896 	} */ *uap;
897 {
898 	int error;
899 	struct nameidata nd;
900 	int vfslocked;
901 
902 	error = priv_check(td, PRIV_VFS_CHROOT);
903 	if (error)
904 		return (error);
905 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | MPSAFE |
906 	    AUDITVNODE1, UIO_USERSPACE, uap->path, td);
907 	error = namei(&nd);
908 	if (error)
909 		goto error;
910 	vfslocked = NDHASGIANT(&nd);
911 	if ((error = change_dir(nd.ni_vp, td)) != 0)
912 		goto e_vunlock;
913 #ifdef MAC
914 	if ((error = mac_vnode_check_chroot(td->td_ucred, nd.ni_vp)))
915 		goto e_vunlock;
916 #endif
917 	VOP_UNLOCK(nd.ni_vp, 0);
918 	error = change_root(nd.ni_vp, td);
919 	vrele(nd.ni_vp);
920 	VFS_UNLOCK_GIANT(vfslocked);
921 	NDFREE(&nd, NDF_ONLY_PNBUF);
922 	return (error);
923 e_vunlock:
924 	vput(nd.ni_vp);
925 	VFS_UNLOCK_GIANT(vfslocked);
926 error:
927 	NDFREE(&nd, NDF_ONLY_PNBUF);
928 	return (error);
929 }
930 
931 /*
932  * Common routine for chroot and chdir.  Callers must provide a locked vnode
933  * instance.
934  */
935 int
936 change_dir(vp, td)
937 	struct vnode *vp;
938 	struct thread *td;
939 {
940 	int error;
941 
942 	ASSERT_VOP_LOCKED(vp, "change_dir(): vp not locked");
943 	if (vp->v_type != VDIR)
944 		return (ENOTDIR);
945 #ifdef MAC
946 	error = mac_vnode_check_chdir(td->td_ucred, vp);
947 	if (error)
948 		return (error);
949 #endif
950 	error = VOP_ACCESS(vp, VEXEC, td->td_ucred, td);
951 	return (error);
952 }
953 
954 /*
955  * Common routine for kern_chroot() and jail_attach().  The caller is
956  * responsible for invoking priv_check() and mac_vnode_check_chroot() to
957  * authorize this operation.
958  */
959 int
960 change_root(vp, td)
961 	struct vnode *vp;
962 	struct thread *td;
963 {
964 	struct filedesc *fdp;
965 	struct vnode *oldvp;
966 	int vfslocked;
967 	int error;
968 
969 	VFS_ASSERT_GIANT(vp->v_mount);
970 	fdp = td->td_proc->p_fd;
971 	FILEDESC_XLOCK(fdp);
972 	if (chroot_allow_open_directories == 0 ||
973 	    (chroot_allow_open_directories == 1 && fdp->fd_rdir != rootvnode)) {
974 		error = chroot_refuse_vdir_fds(fdp);
975 		if (error) {
976 			FILEDESC_XUNLOCK(fdp);
977 			return (error);
978 		}
979 	}
980 	oldvp = fdp->fd_rdir;
981 	fdp->fd_rdir = vp;
982 	VREF(fdp->fd_rdir);
983 	if (!fdp->fd_jdir) {
984 		fdp->fd_jdir = vp;
985 		VREF(fdp->fd_jdir);
986 	}
987 	FILEDESC_XUNLOCK(fdp);
988 	vfslocked = VFS_LOCK_GIANT(oldvp->v_mount);
989 	vrele(oldvp);
990 	VFS_UNLOCK_GIANT(vfslocked);
991 	return (0);
992 }
993 
994 static __inline cap_rights_t
995 flags_to_rights(int flags)
996 {
997 	cap_rights_t rights = 0;
998 
999 	switch ((flags & O_ACCMODE)) {
1000 	case O_RDONLY:
1001 		rights |= CAP_READ;
1002 		break;
1003 
1004 	case O_RDWR:
1005 		rights |= CAP_READ;
1006 		/* fall through */
1007 
1008 	case O_WRONLY:
1009 		rights |= CAP_WRITE;
1010 		break;
1011 
1012 	case O_EXEC:
1013 		rights |= CAP_FEXECVE;
1014 		break;
1015 	}
1016 
1017 	if (flags & O_CREAT)
1018 		rights |= CAP_CREATE;
1019 
1020 	if (flags & O_TRUNC)
1021 		rights |= CAP_FTRUNCATE;
1022 
1023 	if ((flags & O_EXLOCK) || (flags & O_SHLOCK))
1024 		rights |= CAP_FLOCK;
1025 
1026 	return (rights);
1027 }
1028 
1029 /*
1030  * Check permissions, allocate an open file structure, and call the device
1031  * open routine if any.
1032  */
1033 #ifndef _SYS_SYSPROTO_H_
1034 struct open_args {
1035 	char	*path;
1036 	int	flags;
1037 	int	mode;
1038 };
1039 #endif
1040 int
1041 sys_open(td, uap)
1042 	struct thread *td;
1043 	register struct open_args /* {
1044 		char *path;
1045 		int flags;
1046 		int mode;
1047 	} */ *uap;
1048 {
1049 
1050 	return (kern_open(td, uap->path, UIO_USERSPACE, uap->flags, uap->mode));
1051 }
1052 
1053 #ifndef _SYS_SYSPROTO_H_
1054 struct openat_args {
1055 	int	fd;
1056 	char	*path;
1057 	int	flag;
1058 	int	mode;
1059 };
1060 #endif
1061 int
1062 sys_openat(struct thread *td, struct openat_args *uap)
1063 {
1064 
1065 	return (kern_openat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
1066 	    uap->mode));
1067 }
1068 
1069 int
1070 kern_open(struct thread *td, char *path, enum uio_seg pathseg, int flags,
1071     int mode)
1072 {
1073 
1074 	return (kern_openat(td, AT_FDCWD, path, pathseg, flags, mode));
1075 }
1076 
1077 int
1078 kern_openat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1079     int flags, int mode)
1080 {
1081 	struct proc *p = td->td_proc;
1082 	struct filedesc *fdp = p->p_fd;
1083 	struct file *fp;
1084 	struct vnode *vp;
1085 	int cmode;
1086 	struct file *nfp;
1087 	int type, indx = -1, error, error_open;
1088 	struct flock lf;
1089 	struct nameidata nd;
1090 	int vfslocked;
1091 	cap_rights_t rights_needed = CAP_LOOKUP;
1092 
1093 	AUDIT_ARG_FFLAGS(flags);
1094 	AUDIT_ARG_MODE(mode);
1095 	/* XXX: audit dirfd */
1096 	rights_needed |= flags_to_rights(flags);
1097 	/*
1098 	 * Only one of the O_EXEC, O_RDONLY, O_WRONLY and O_RDWR flags
1099 	 * may be specified.
1100 	 */
1101 	if (flags & O_EXEC) {
1102 		if (flags & O_ACCMODE)
1103 			return (EINVAL);
1104 	} else if ((flags & O_ACCMODE) == O_ACCMODE)
1105 		return (EINVAL);
1106 	else
1107 		flags = FFLAGS(flags);
1108 
1109 	/*
1110 	 * allocate the file descriptor, but don't install a descriptor yet
1111 	 */
1112 	error = falloc_noinstall(td, &nfp);
1113 	if (error)
1114 		return (error);
1115 	/* An extra reference on `nfp' has been held for us by falloc_noinstall(). */
1116 	fp = nfp;
1117 	/* Set the flags early so the finit in devfs can pick them up. */
1118 	fp->f_flag = flags & FMASK;
1119 	cmode = ((mode &~ fdp->fd_cmask) & ALLPERMS) &~ S_ISTXT;
1120 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1 | MPSAFE, pathseg,
1121 	    path, fd, rights_needed, td);
1122 	td->td_dupfd = -1;		/* XXX check for fdopen */
1123 	error = vn_open(&nd, &flags, cmode, fp);
1124 	if (error) {
1125 		/*
1126 		 * If the vn_open replaced the method vector, something
1127 		 * wonderous happened deep below and we just pass it up
1128 		 * pretending we know what we do.
1129 		 */
1130 		if (error == ENXIO && fp->f_ops != &badfileops)
1131 			goto success;
1132 
1133 		/*
1134 		 * handle special fdopen() case.  bleh.  dupfdopen() is
1135 		 * responsible for dropping the old contents of ofiles[indx]
1136 		 * if it succeeds.
1137 		 *
1138 		 * Don't do this for relative (capability) lookups; we don't
1139 		 * understand exactly what would happen, and we don't think
1140 		 * that it ever should.
1141 		 */
1142 		if ((nd.ni_strictrelative == 0) &&
1143 		    (error == ENODEV || error == ENXIO) &&
1144 		    (td->td_dupfd >= 0)) {
1145 			/* XXX from fdopen */
1146 			error_open = error;
1147 			if ((error = finstall(td, fp, &indx, flags)) != 0)
1148 				goto bad_unlocked;
1149 			if ((error = dupfdopen(td, fdp, indx, td->td_dupfd,
1150 			    flags, error_open)) == 0)
1151 				goto success;
1152 		}
1153 		/*
1154 		 * Clean up the descriptor, but only if another thread hadn't
1155 		 * replaced or closed it.
1156 		 */
1157 		if (indx != -1)
1158 			fdclose(fdp, fp, indx, td);
1159 		fdrop(fp, td);
1160 
1161 		if (error == ERESTART)
1162 			error = EINTR;
1163 		return (error);
1164 	}
1165 	td->td_dupfd = 0;
1166 	vfslocked = NDHASGIANT(&nd);
1167 	NDFREE(&nd, NDF_ONLY_PNBUF);
1168 	vp = nd.ni_vp;
1169 
1170 	/*
1171 	 * Store the vnode, for any f_type. Typically, the vnode use
1172 	 * count is decremented by direct call to vn_closefile() for
1173 	 * files that switched type in the cdevsw fdopen() method.
1174 	 */
1175 	fp->f_vnode = vp;
1176 	/*
1177 	 * If the file wasn't claimed by devfs bind it to the normal
1178 	 * vnode operations here.
1179 	 */
1180 	if (fp->f_ops == &badfileops) {
1181 		KASSERT(vp->v_type != VFIFO, ("Unexpected fifo."));
1182 		fp->f_seqcount = 1;
1183 		finit(fp, flags & FMASK, DTYPE_VNODE, vp, &vnops);
1184 	}
1185 
1186 	VOP_UNLOCK(vp, 0);
1187 	if (fp->f_type == DTYPE_VNODE && (flags & (O_EXLOCK | O_SHLOCK)) != 0) {
1188 		lf.l_whence = SEEK_SET;
1189 		lf.l_start = 0;
1190 		lf.l_len = 0;
1191 		if (flags & O_EXLOCK)
1192 			lf.l_type = F_WRLCK;
1193 		else
1194 			lf.l_type = F_RDLCK;
1195 		type = F_FLOCK;
1196 		if ((flags & FNONBLOCK) == 0)
1197 			type |= F_WAIT;
1198 		if ((error = VOP_ADVLOCK(vp, (caddr_t)fp, F_SETLK, &lf,
1199 			    type)) != 0)
1200 			goto bad;
1201 		atomic_set_int(&fp->f_flag, FHASLOCK);
1202 	}
1203 	if (flags & O_TRUNC) {
1204 		error = fo_truncate(fp, 0, td->td_ucred, td);
1205 		if (error)
1206 			goto bad;
1207 	}
1208 	VFS_UNLOCK_GIANT(vfslocked);
1209 success:
1210 	/*
1211 	 * If we haven't already installed the FD (for dupfdopen), do so now.
1212 	 */
1213 	if (indx == -1) {
1214 #ifdef CAPABILITIES
1215 		if (nd.ni_strictrelative == 1) {
1216 			/*
1217 			 * We are doing a strict relative lookup; wrap the
1218 			 * result in a capability.
1219 			 */
1220 			if ((error = kern_capwrap(td, fp, nd.ni_baserights,
1221 			    &indx)) != 0)
1222 				goto bad_unlocked;
1223 		} else
1224 #endif
1225 			if ((error = finstall(td, fp, &indx, flags)) != 0)
1226 				goto bad_unlocked;
1227 
1228 	}
1229 
1230 	/*
1231 	 * Release our private reference, leaving the one associated with
1232 	 * the descriptor table intact.
1233 	 */
1234 	fdrop(fp, td);
1235 	td->td_retval[0] = indx;
1236 	return (0);
1237 bad:
1238 	VFS_UNLOCK_GIANT(vfslocked);
1239 bad_unlocked:
1240 	if (indx != -1)
1241 		fdclose(fdp, fp, indx, td);
1242 	fdrop(fp, td);
1243 	td->td_retval[0] = -1;
1244 	return (error);
1245 }
1246 
1247 #ifdef COMPAT_43
1248 /*
1249  * Create a file.
1250  */
1251 #ifndef _SYS_SYSPROTO_H_
1252 struct ocreat_args {
1253 	char	*path;
1254 	int	mode;
1255 };
1256 #endif
1257 int
1258 ocreat(td, uap)
1259 	struct thread *td;
1260 	register struct ocreat_args /* {
1261 		char *path;
1262 		int mode;
1263 	} */ *uap;
1264 {
1265 
1266 	return (kern_open(td, uap->path, UIO_USERSPACE,
1267 	    O_WRONLY | O_CREAT | O_TRUNC, uap->mode));
1268 }
1269 #endif /* COMPAT_43 */
1270 
1271 /*
1272  * Create a special file.
1273  */
1274 #ifndef _SYS_SYSPROTO_H_
1275 struct mknod_args {
1276 	char	*path;
1277 	int	mode;
1278 	int	dev;
1279 };
1280 #endif
1281 int
1282 sys_mknod(td, uap)
1283 	struct thread *td;
1284 	register struct mknod_args /* {
1285 		char *path;
1286 		int mode;
1287 		int dev;
1288 	} */ *uap;
1289 {
1290 
1291 	return (kern_mknod(td, uap->path, UIO_USERSPACE, uap->mode, uap->dev));
1292 }
1293 
1294 #ifndef _SYS_SYSPROTO_H_
1295 struct mknodat_args {
1296 	int	fd;
1297 	char	*path;
1298 	mode_t	mode;
1299 	dev_t	dev;
1300 };
1301 #endif
1302 int
1303 sys_mknodat(struct thread *td, struct mknodat_args *uap)
1304 {
1305 
1306 	return (kern_mknodat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode,
1307 	    uap->dev));
1308 }
1309 
1310 int
1311 kern_mknod(struct thread *td, char *path, enum uio_seg pathseg, int mode,
1312     int dev)
1313 {
1314 
1315 	return (kern_mknodat(td, AT_FDCWD, path, pathseg, mode, dev));
1316 }
1317 
1318 int
1319 kern_mknodat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1320     int mode, int dev)
1321 {
1322 	struct vnode *vp;
1323 	struct mount *mp;
1324 	struct vattr vattr;
1325 	int error;
1326 	int whiteout = 0;
1327 	struct nameidata nd;
1328 	int vfslocked;
1329 
1330 	AUDIT_ARG_MODE(mode);
1331 	AUDIT_ARG_DEV(dev);
1332 	switch (mode & S_IFMT) {
1333 	case S_IFCHR:
1334 	case S_IFBLK:
1335 		error = priv_check(td, PRIV_VFS_MKNOD_DEV);
1336 		break;
1337 	case S_IFMT:
1338 		error = priv_check(td, PRIV_VFS_MKNOD_BAD);
1339 		break;
1340 	case S_IFWHT:
1341 		error = priv_check(td, PRIV_VFS_MKNOD_WHT);
1342 		break;
1343 	case S_IFIFO:
1344 		if (dev == 0)
1345 			return (kern_mkfifoat(td, fd, path, pathseg, mode));
1346 		/* FALLTHROUGH */
1347 	default:
1348 		error = EINVAL;
1349 		break;
1350 	}
1351 	if (error)
1352 		return (error);
1353 restart:
1354 	bwillwrite();
1355 	NDINIT_ATRIGHTS(&nd, CREATE,
1356 	    LOCKPARENT | SAVENAME | MPSAFE | AUDITVNODE1, pathseg, path, fd,
1357 	    CAP_MKFIFO, td);
1358 	if ((error = namei(&nd)) != 0)
1359 		return (error);
1360 	vfslocked = NDHASGIANT(&nd);
1361 	vp = nd.ni_vp;
1362 	if (vp != NULL) {
1363 		NDFREE(&nd, NDF_ONLY_PNBUF);
1364 		if (vp == nd.ni_dvp)
1365 			vrele(nd.ni_dvp);
1366 		else
1367 			vput(nd.ni_dvp);
1368 		vrele(vp);
1369 		VFS_UNLOCK_GIANT(vfslocked);
1370 		return (EEXIST);
1371 	} else {
1372 		VATTR_NULL(&vattr);
1373 		vattr.va_mode = (mode & ALLPERMS) &
1374 		    ~td->td_proc->p_fd->fd_cmask;
1375 		vattr.va_rdev = dev;
1376 		whiteout = 0;
1377 
1378 		switch (mode & S_IFMT) {
1379 		case S_IFMT:	/* used by badsect to flag bad sectors */
1380 			vattr.va_type = VBAD;
1381 			break;
1382 		case S_IFCHR:
1383 			vattr.va_type = VCHR;
1384 			break;
1385 		case S_IFBLK:
1386 			vattr.va_type = VBLK;
1387 			break;
1388 		case S_IFWHT:
1389 			whiteout = 1;
1390 			break;
1391 		default:
1392 			panic("kern_mknod: invalid mode");
1393 		}
1394 	}
1395 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1396 		NDFREE(&nd, NDF_ONLY_PNBUF);
1397 		vput(nd.ni_dvp);
1398 		VFS_UNLOCK_GIANT(vfslocked);
1399 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1400 			return (error);
1401 		goto restart;
1402 	}
1403 #ifdef MAC
1404 	if (error == 0 && !whiteout)
1405 		error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp,
1406 		    &nd.ni_cnd, &vattr);
1407 #endif
1408 	if (!error) {
1409 		if (whiteout)
1410 			error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE);
1411 		else {
1412 			error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp,
1413 						&nd.ni_cnd, &vattr);
1414 			if (error == 0)
1415 				vput(nd.ni_vp);
1416 		}
1417 	}
1418 	NDFREE(&nd, NDF_ONLY_PNBUF);
1419 	vput(nd.ni_dvp);
1420 	vn_finished_write(mp);
1421 	VFS_UNLOCK_GIANT(vfslocked);
1422 	return (error);
1423 }
1424 
1425 /*
1426  * Create a named pipe.
1427  */
1428 #ifndef _SYS_SYSPROTO_H_
1429 struct mkfifo_args {
1430 	char	*path;
1431 	int	mode;
1432 };
1433 #endif
1434 int
1435 sys_mkfifo(td, uap)
1436 	struct thread *td;
1437 	register struct mkfifo_args /* {
1438 		char *path;
1439 		int mode;
1440 	} */ *uap;
1441 {
1442 
1443 	return (kern_mkfifo(td, uap->path, UIO_USERSPACE, uap->mode));
1444 }
1445 
1446 #ifndef _SYS_SYSPROTO_H_
1447 struct mkfifoat_args {
1448 	int	fd;
1449 	char	*path;
1450 	mode_t	mode;
1451 };
1452 #endif
1453 int
1454 sys_mkfifoat(struct thread *td, struct mkfifoat_args *uap)
1455 {
1456 
1457 	return (kern_mkfifoat(td, uap->fd, uap->path, UIO_USERSPACE,
1458 	    uap->mode));
1459 }
1460 
1461 int
1462 kern_mkfifo(struct thread *td, char *path, enum uio_seg pathseg, int mode)
1463 {
1464 
1465 	return (kern_mkfifoat(td, AT_FDCWD, path, pathseg, mode));
1466 }
1467 
1468 int
1469 kern_mkfifoat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1470     int mode)
1471 {
1472 	struct mount *mp;
1473 	struct vattr vattr;
1474 	int error;
1475 	struct nameidata nd;
1476 	int vfslocked;
1477 
1478 	AUDIT_ARG_MODE(mode);
1479 restart:
1480 	bwillwrite();
1481 	NDINIT_AT(&nd, CREATE, LOCKPARENT | SAVENAME | MPSAFE | AUDITVNODE1,
1482 	    pathseg, path, fd, td);
1483 	if ((error = namei(&nd)) != 0)
1484 		return (error);
1485 	vfslocked = NDHASGIANT(&nd);
1486 	if (nd.ni_vp != NULL) {
1487 		NDFREE(&nd, NDF_ONLY_PNBUF);
1488 		if (nd.ni_vp == nd.ni_dvp)
1489 			vrele(nd.ni_dvp);
1490 		else
1491 			vput(nd.ni_dvp);
1492 		vrele(nd.ni_vp);
1493 		VFS_UNLOCK_GIANT(vfslocked);
1494 		return (EEXIST);
1495 	}
1496 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1497 		NDFREE(&nd, NDF_ONLY_PNBUF);
1498 		vput(nd.ni_dvp);
1499 		VFS_UNLOCK_GIANT(vfslocked);
1500 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1501 			return (error);
1502 		goto restart;
1503 	}
1504 	VATTR_NULL(&vattr);
1505 	vattr.va_type = VFIFO;
1506 	vattr.va_mode = (mode & ALLPERMS) & ~td->td_proc->p_fd->fd_cmask;
1507 #ifdef MAC
1508 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1509 	    &vattr);
1510 	if (error)
1511 		goto out;
1512 #endif
1513 	error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
1514 	if (error == 0)
1515 		vput(nd.ni_vp);
1516 #ifdef MAC
1517 out:
1518 #endif
1519 	vput(nd.ni_dvp);
1520 	vn_finished_write(mp);
1521 	VFS_UNLOCK_GIANT(vfslocked);
1522 	NDFREE(&nd, NDF_ONLY_PNBUF);
1523 	return (error);
1524 }
1525 
1526 /*
1527  * Make a hard file link.
1528  */
1529 #ifndef _SYS_SYSPROTO_H_
1530 struct link_args {
1531 	char	*path;
1532 	char	*link;
1533 };
1534 #endif
1535 int
1536 sys_link(td, uap)
1537 	struct thread *td;
1538 	register struct link_args /* {
1539 		char *path;
1540 		char *link;
1541 	} */ *uap;
1542 {
1543 
1544 	return (kern_link(td, uap->path, uap->link, UIO_USERSPACE));
1545 }
1546 
1547 #ifndef _SYS_SYSPROTO_H_
1548 struct linkat_args {
1549 	int	fd1;
1550 	char	*path1;
1551 	int	fd2;
1552 	char	*path2;
1553 	int	flag;
1554 };
1555 #endif
1556 int
1557 sys_linkat(struct thread *td, struct linkat_args *uap)
1558 {
1559 	int flag;
1560 
1561 	flag = uap->flag;
1562 	if (flag & ~AT_SYMLINK_FOLLOW)
1563 		return (EINVAL);
1564 
1565 	return (kern_linkat(td, uap->fd1, uap->fd2, uap->path1, uap->path2,
1566 	    UIO_USERSPACE, (flag & AT_SYMLINK_FOLLOW) ? FOLLOW : NOFOLLOW));
1567 }
1568 
1569 int hardlink_check_uid = 0;
1570 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_uid, CTLFLAG_RW,
1571     &hardlink_check_uid, 0,
1572     "Unprivileged processes cannot create hard links to files owned by other "
1573     "users");
1574 static int hardlink_check_gid = 0;
1575 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_gid, CTLFLAG_RW,
1576     &hardlink_check_gid, 0,
1577     "Unprivileged processes cannot create hard links to files owned by other "
1578     "groups");
1579 
1580 static int
1581 can_hardlink(struct vnode *vp, struct ucred *cred)
1582 {
1583 	struct vattr va;
1584 	int error;
1585 
1586 	if (!hardlink_check_uid && !hardlink_check_gid)
1587 		return (0);
1588 
1589 	error = VOP_GETATTR(vp, &va, cred);
1590 	if (error != 0)
1591 		return (error);
1592 
1593 	if (hardlink_check_uid && cred->cr_uid != va.va_uid) {
1594 		error = priv_check_cred(cred, PRIV_VFS_LINK, 0);
1595 		if (error)
1596 			return (error);
1597 	}
1598 
1599 	if (hardlink_check_gid && !groupmember(va.va_gid, cred)) {
1600 		error = priv_check_cred(cred, PRIV_VFS_LINK, 0);
1601 		if (error)
1602 			return (error);
1603 	}
1604 
1605 	return (0);
1606 }
1607 
1608 int
1609 kern_link(struct thread *td, char *path, char *link, enum uio_seg segflg)
1610 {
1611 
1612 	return (kern_linkat(td, AT_FDCWD, AT_FDCWD, path,link, segflg, FOLLOW));
1613 }
1614 
1615 int
1616 kern_linkat(struct thread *td, int fd1, int fd2, char *path1, char *path2,
1617     enum uio_seg segflg, int follow)
1618 {
1619 	struct vnode *vp;
1620 	struct mount *mp;
1621 	struct nameidata nd;
1622 	int vfslocked;
1623 	int lvfslocked;
1624 	int error;
1625 
1626 	bwillwrite();
1627 	NDINIT_AT(&nd, LOOKUP, follow | MPSAFE | AUDITVNODE1, segflg, path1,
1628 	    fd1, td);
1629 
1630 	if ((error = namei(&nd)) != 0)
1631 		return (error);
1632 	vfslocked = NDHASGIANT(&nd);
1633 	NDFREE(&nd, NDF_ONLY_PNBUF);
1634 	vp = nd.ni_vp;
1635 	if (vp->v_type == VDIR) {
1636 		vrele(vp);
1637 		VFS_UNLOCK_GIANT(vfslocked);
1638 		return (EPERM);		/* POSIX */
1639 	}
1640 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) {
1641 		vrele(vp);
1642 		VFS_UNLOCK_GIANT(vfslocked);
1643 		return (error);
1644 	}
1645 	NDINIT_AT(&nd, CREATE, LOCKPARENT | SAVENAME | MPSAFE | AUDITVNODE2,
1646 	    segflg, path2, fd2, td);
1647 	if ((error = namei(&nd)) == 0) {
1648 		lvfslocked = NDHASGIANT(&nd);
1649 		if (nd.ni_vp != NULL) {
1650 			if (nd.ni_dvp == nd.ni_vp)
1651 				vrele(nd.ni_dvp);
1652 			else
1653 				vput(nd.ni_dvp);
1654 			vrele(nd.ni_vp);
1655 			error = EEXIST;
1656 		} else if ((error = vn_lock(vp, LK_EXCLUSIVE | LK_RETRY))
1657 		    == 0) {
1658 			error = can_hardlink(vp, td->td_ucred);
1659 			if (error == 0)
1660 #ifdef MAC
1661 				error = mac_vnode_check_link(td->td_ucred,
1662 				    nd.ni_dvp, vp, &nd.ni_cnd);
1663 			if (error == 0)
1664 #endif
1665 				error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
1666 			VOP_UNLOCK(vp, 0);
1667 			vput(nd.ni_dvp);
1668 		}
1669 		NDFREE(&nd, NDF_ONLY_PNBUF);
1670 		VFS_UNLOCK_GIANT(lvfslocked);
1671 	}
1672 	vrele(vp);
1673 	vn_finished_write(mp);
1674 	VFS_UNLOCK_GIANT(vfslocked);
1675 	return (error);
1676 }
1677 
1678 /*
1679  * Make a symbolic link.
1680  */
1681 #ifndef _SYS_SYSPROTO_H_
1682 struct symlink_args {
1683 	char	*path;
1684 	char	*link;
1685 };
1686 #endif
1687 int
1688 sys_symlink(td, uap)
1689 	struct thread *td;
1690 	register struct symlink_args /* {
1691 		char *path;
1692 		char *link;
1693 	} */ *uap;
1694 {
1695 
1696 	return (kern_symlink(td, uap->path, uap->link, UIO_USERSPACE));
1697 }
1698 
1699 #ifndef _SYS_SYSPROTO_H_
1700 struct symlinkat_args {
1701 	char	*path;
1702 	int	fd;
1703 	char	*path2;
1704 };
1705 #endif
1706 int
1707 sys_symlinkat(struct thread *td, struct symlinkat_args *uap)
1708 {
1709 
1710 	return (kern_symlinkat(td, uap->path1, uap->fd, uap->path2,
1711 	    UIO_USERSPACE));
1712 }
1713 
1714 int
1715 kern_symlink(struct thread *td, char *path, char *link, enum uio_seg segflg)
1716 {
1717 
1718 	return (kern_symlinkat(td, path, AT_FDCWD, link, segflg));
1719 }
1720 
1721 int
1722 kern_symlinkat(struct thread *td, char *path1, int fd, char *path2,
1723     enum uio_seg segflg)
1724 {
1725 	struct mount *mp;
1726 	struct vattr vattr;
1727 	char *syspath;
1728 	int error;
1729 	struct nameidata nd;
1730 	int vfslocked;
1731 
1732 	if (segflg == UIO_SYSSPACE) {
1733 		syspath = path1;
1734 	} else {
1735 		syspath = uma_zalloc(namei_zone, M_WAITOK);
1736 		if ((error = copyinstr(path1, syspath, MAXPATHLEN, NULL)) != 0)
1737 			goto out;
1738 	}
1739 	AUDIT_ARG_TEXT(syspath);
1740 restart:
1741 	bwillwrite();
1742 	NDINIT_AT(&nd, CREATE, LOCKPARENT | SAVENAME | MPSAFE | AUDITVNODE1,
1743 	    segflg, path2, fd, td);
1744 	if ((error = namei(&nd)) != 0)
1745 		goto out;
1746 	vfslocked = NDHASGIANT(&nd);
1747 	if (nd.ni_vp) {
1748 		NDFREE(&nd, NDF_ONLY_PNBUF);
1749 		if (nd.ni_vp == nd.ni_dvp)
1750 			vrele(nd.ni_dvp);
1751 		else
1752 			vput(nd.ni_dvp);
1753 		vrele(nd.ni_vp);
1754 		VFS_UNLOCK_GIANT(vfslocked);
1755 		error = EEXIST;
1756 		goto out;
1757 	}
1758 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1759 		NDFREE(&nd, NDF_ONLY_PNBUF);
1760 		vput(nd.ni_dvp);
1761 		VFS_UNLOCK_GIANT(vfslocked);
1762 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1763 			goto out;
1764 		goto restart;
1765 	}
1766 	VATTR_NULL(&vattr);
1767 	vattr.va_mode = ACCESSPERMS &~ td->td_proc->p_fd->fd_cmask;
1768 #ifdef MAC
1769 	vattr.va_type = VLNK;
1770 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1771 	    &vattr);
1772 	if (error)
1773 		goto out2;
1774 #endif
1775 	error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, syspath);
1776 	if (error == 0)
1777 		vput(nd.ni_vp);
1778 #ifdef MAC
1779 out2:
1780 #endif
1781 	NDFREE(&nd, NDF_ONLY_PNBUF);
1782 	vput(nd.ni_dvp);
1783 	vn_finished_write(mp);
1784 	VFS_UNLOCK_GIANT(vfslocked);
1785 out:
1786 	if (segflg != UIO_SYSSPACE)
1787 		uma_zfree(namei_zone, syspath);
1788 	return (error);
1789 }
1790 
1791 /*
1792  * Delete a whiteout from the filesystem.
1793  */
1794 int
1795 sys_undelete(td, uap)
1796 	struct thread *td;
1797 	register struct undelete_args /* {
1798 		char *path;
1799 	} */ *uap;
1800 {
1801 	int error;
1802 	struct mount *mp;
1803 	struct nameidata nd;
1804 	int vfslocked;
1805 
1806 restart:
1807 	bwillwrite();
1808 	NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | MPSAFE | AUDITVNODE1,
1809 	    UIO_USERSPACE, uap->path, td);
1810 	error = namei(&nd);
1811 	if (error)
1812 		return (error);
1813 	vfslocked = NDHASGIANT(&nd);
1814 
1815 	if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) {
1816 		NDFREE(&nd, NDF_ONLY_PNBUF);
1817 		if (nd.ni_vp == nd.ni_dvp)
1818 			vrele(nd.ni_dvp);
1819 		else
1820 			vput(nd.ni_dvp);
1821 		if (nd.ni_vp)
1822 			vrele(nd.ni_vp);
1823 		VFS_UNLOCK_GIANT(vfslocked);
1824 		return (EEXIST);
1825 	}
1826 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1827 		NDFREE(&nd, NDF_ONLY_PNBUF);
1828 		vput(nd.ni_dvp);
1829 		VFS_UNLOCK_GIANT(vfslocked);
1830 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1831 			return (error);
1832 		goto restart;
1833 	}
1834 	error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE);
1835 	NDFREE(&nd, NDF_ONLY_PNBUF);
1836 	vput(nd.ni_dvp);
1837 	vn_finished_write(mp);
1838 	VFS_UNLOCK_GIANT(vfslocked);
1839 	return (error);
1840 }
1841 
1842 /*
1843  * Delete a name from the filesystem.
1844  */
1845 #ifndef _SYS_SYSPROTO_H_
1846 struct unlink_args {
1847 	char	*path;
1848 };
1849 #endif
1850 int
1851 sys_unlink(td, uap)
1852 	struct thread *td;
1853 	struct unlink_args /* {
1854 		char *path;
1855 	} */ *uap;
1856 {
1857 
1858 	return (kern_unlink(td, uap->path, UIO_USERSPACE));
1859 }
1860 
1861 #ifndef _SYS_SYSPROTO_H_
1862 struct unlinkat_args {
1863 	int	fd;
1864 	char	*path;
1865 	int	flag;
1866 };
1867 #endif
1868 int
1869 sys_unlinkat(struct thread *td, struct unlinkat_args *uap)
1870 {
1871 	int flag = uap->flag;
1872 	int fd = uap->fd;
1873 	char *path = uap->path;
1874 
1875 	if (flag & ~AT_REMOVEDIR)
1876 		return (EINVAL);
1877 
1878 	if (flag & AT_REMOVEDIR)
1879 		return (kern_rmdirat(td, fd, path, UIO_USERSPACE));
1880 	else
1881 		return (kern_unlinkat(td, fd, path, UIO_USERSPACE, 0));
1882 }
1883 
1884 int
1885 kern_unlink(struct thread *td, char *path, enum uio_seg pathseg)
1886 {
1887 
1888 	return (kern_unlinkat(td, AT_FDCWD, path, pathseg, 0));
1889 }
1890 
1891 int
1892 kern_unlinkat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1893     ino_t oldinum)
1894 {
1895 	struct mount *mp;
1896 	struct vnode *vp;
1897 	int error;
1898 	struct nameidata nd;
1899 	struct stat sb;
1900 	int vfslocked;
1901 
1902 restart:
1903 	bwillwrite();
1904 	NDINIT_AT(&nd, DELETE, LOCKPARENT | LOCKLEAF | MPSAFE | AUDITVNODE1,
1905 	    pathseg, path, fd, td);
1906 	if ((error = namei(&nd)) != 0)
1907 		return (error == EINVAL ? EPERM : error);
1908 	vfslocked = NDHASGIANT(&nd);
1909 	vp = nd.ni_vp;
1910 	if (vp->v_type == VDIR && oldinum == 0) {
1911 		error = EPERM;		/* POSIX */
1912 	} else if (oldinum != 0 &&
1913 		  ((error = vn_stat(vp, &sb, td->td_ucred, NOCRED, td)) == 0) &&
1914 		  sb.st_ino != oldinum) {
1915 			error = EIDRM;	/* Identifier removed */
1916 	} else {
1917 		/*
1918 		 * The root of a mounted filesystem cannot be deleted.
1919 		 *
1920 		 * XXX: can this only be a VDIR case?
1921 		 */
1922 		if (vp->v_vflag & VV_ROOT)
1923 			error = EBUSY;
1924 	}
1925 	if (error == 0) {
1926 		if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1927 			NDFREE(&nd, NDF_ONLY_PNBUF);
1928 			vput(nd.ni_dvp);
1929 			if (vp == nd.ni_dvp)
1930 				vrele(vp);
1931 			else
1932 				vput(vp);
1933 			VFS_UNLOCK_GIANT(vfslocked);
1934 			if ((error = vn_start_write(NULL, &mp,
1935 			    V_XSLEEP | PCATCH)) != 0)
1936 				return (error);
1937 			goto restart;
1938 		}
1939 #ifdef MAC
1940 		error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
1941 		    &nd.ni_cnd);
1942 		if (error)
1943 			goto out;
1944 #endif
1945 		error = VOP_REMOVE(nd.ni_dvp, vp, &nd.ni_cnd);
1946 #ifdef MAC
1947 out:
1948 #endif
1949 		vn_finished_write(mp);
1950 	}
1951 	NDFREE(&nd, NDF_ONLY_PNBUF);
1952 	vput(nd.ni_dvp);
1953 	if (vp == nd.ni_dvp)
1954 		vrele(vp);
1955 	else
1956 		vput(vp);
1957 	VFS_UNLOCK_GIANT(vfslocked);
1958 	return (error);
1959 }
1960 
1961 /*
1962  * Reposition read/write file offset.
1963  */
1964 #ifndef _SYS_SYSPROTO_H_
1965 struct lseek_args {
1966 	int	fd;
1967 	int	pad;
1968 	off_t	offset;
1969 	int	whence;
1970 };
1971 #endif
1972 int
1973 sys_lseek(td, uap)
1974 	struct thread *td;
1975 	register struct lseek_args /* {
1976 		int fd;
1977 		int pad;
1978 		off_t offset;
1979 		int whence;
1980 	} */ *uap;
1981 {
1982 	struct ucred *cred = td->td_ucred;
1983 	struct file *fp;
1984 	struct vnode *vp;
1985 	struct vattr vattr;
1986 	off_t offset, size;
1987 	int error, noneg;
1988 	int vfslocked;
1989 
1990 	AUDIT_ARG_FD(uap->fd);
1991 	if ((error = fget(td, uap->fd, CAP_SEEK, &fp)) != 0)
1992 		return (error);
1993 	if (!(fp->f_ops->fo_flags & DFLAG_SEEKABLE)) {
1994 		fdrop(fp, td);
1995 		return (ESPIPE);
1996 	}
1997 	vp = fp->f_vnode;
1998 	vfslocked = VFS_LOCK_GIANT(vp->v_mount);
1999 	noneg = (vp->v_type != VCHR);
2000 	offset = uap->offset;
2001 	switch (uap->whence) {
2002 	case L_INCR:
2003 		if (noneg &&
2004 		    (fp->f_offset < 0 ||
2005 		    (offset > 0 && fp->f_offset > OFF_MAX - offset))) {
2006 			error = EOVERFLOW;
2007 			break;
2008 		}
2009 		offset += fp->f_offset;
2010 		break;
2011 	case L_XTND:
2012 		vn_lock(vp, LK_SHARED | LK_RETRY);
2013 		error = VOP_GETATTR(vp, &vattr, cred);
2014 		VOP_UNLOCK(vp, 0);
2015 		if (error)
2016 			break;
2017 
2018 		/*
2019 		 * If the file references a disk device, then fetch
2020 		 * the media size and use that to determine the ending
2021 		 * offset.
2022 		 */
2023 		if (vattr.va_size == 0 && vp->v_type == VCHR &&
2024 		    fo_ioctl(fp, DIOCGMEDIASIZE, &size, cred, td) == 0)
2025 			vattr.va_size = size;
2026 		if (noneg &&
2027 		    (vattr.va_size > OFF_MAX ||
2028 		    (offset > 0 && vattr.va_size > OFF_MAX - offset))) {
2029 			error = EOVERFLOW;
2030 			break;
2031 		}
2032 		offset += vattr.va_size;
2033 		break;
2034 	case L_SET:
2035 		break;
2036 	case SEEK_DATA:
2037 		error = fo_ioctl(fp, FIOSEEKDATA, &offset, cred, td);
2038 		break;
2039 	case SEEK_HOLE:
2040 		error = fo_ioctl(fp, FIOSEEKHOLE, &offset, cred, td);
2041 		break;
2042 	default:
2043 		error = EINVAL;
2044 	}
2045 	if (error == 0 && noneg && offset < 0)
2046 		error = EINVAL;
2047 	if (error != 0)
2048 		goto drop;
2049 	fp->f_offset = offset;
2050 	*(off_t *)(td->td_retval) = fp->f_offset;
2051 drop:
2052 	fdrop(fp, td);
2053 	VFS_UNLOCK_GIANT(vfslocked);
2054 	return (error);
2055 }
2056 
2057 #if defined(COMPAT_43)
2058 /*
2059  * Reposition read/write file offset.
2060  */
2061 #ifndef _SYS_SYSPROTO_H_
2062 struct olseek_args {
2063 	int	fd;
2064 	long	offset;
2065 	int	whence;
2066 };
2067 #endif
2068 int
2069 olseek(td, uap)
2070 	struct thread *td;
2071 	register struct olseek_args /* {
2072 		int fd;
2073 		long offset;
2074 		int whence;
2075 	} */ *uap;
2076 {
2077 	struct lseek_args /* {
2078 		int fd;
2079 		int pad;
2080 		off_t offset;
2081 		int whence;
2082 	} */ nuap;
2083 
2084 	nuap.fd = uap->fd;
2085 	nuap.offset = uap->offset;
2086 	nuap.whence = uap->whence;
2087 	return (sys_lseek(td, &nuap));
2088 }
2089 #endif /* COMPAT_43 */
2090 
2091 /* Version with the 'pad' argument */
2092 int
2093 freebsd6_lseek(td, uap)
2094 	struct thread *td;
2095 	register struct freebsd6_lseek_args *uap;
2096 {
2097 	struct lseek_args ouap;
2098 
2099 	ouap.fd = uap->fd;
2100 	ouap.offset = uap->offset;
2101 	ouap.whence = uap->whence;
2102 	return (sys_lseek(td, &ouap));
2103 }
2104 
2105 /*
2106  * Check access permissions using passed credentials.
2107  */
2108 static int
2109 vn_access(vp, user_flags, cred, td)
2110 	struct vnode	*vp;
2111 	int		user_flags;
2112 	struct ucred	*cred;
2113 	struct thread	*td;
2114 {
2115 	int error;
2116 	accmode_t accmode;
2117 
2118 	/* Flags == 0 means only check for existence. */
2119 	error = 0;
2120 	if (user_flags) {
2121 		accmode = 0;
2122 		if (user_flags & R_OK)
2123 			accmode |= VREAD;
2124 		if (user_flags & W_OK)
2125 			accmode |= VWRITE;
2126 		if (user_flags & X_OK)
2127 			accmode |= VEXEC;
2128 #ifdef MAC
2129 		error = mac_vnode_check_access(cred, vp, accmode);
2130 		if (error)
2131 			return (error);
2132 #endif
2133 		if ((accmode & VWRITE) == 0 || (error = vn_writechk(vp)) == 0)
2134 			error = VOP_ACCESS(vp, accmode, cred, td);
2135 	}
2136 	return (error);
2137 }
2138 
2139 /*
2140  * Check access permissions using "real" credentials.
2141  */
2142 #ifndef _SYS_SYSPROTO_H_
2143 struct access_args {
2144 	char	*path;
2145 	int	flags;
2146 };
2147 #endif
2148 int
2149 sys_access(td, uap)
2150 	struct thread *td;
2151 	register struct access_args /* {
2152 		char *path;
2153 		int flags;
2154 	} */ *uap;
2155 {
2156 
2157 	return (kern_access(td, uap->path, UIO_USERSPACE, uap->flags));
2158 }
2159 
2160 #ifndef _SYS_SYSPROTO_H_
2161 struct faccessat_args {
2162 	int	dirfd;
2163 	char	*path;
2164 	int	mode;
2165 	int	flag;
2166 }
2167 #endif
2168 int
2169 sys_faccessat(struct thread *td, struct faccessat_args *uap)
2170 {
2171 
2172 	if (uap->flag & ~AT_EACCESS)
2173 		return (EINVAL);
2174 	return (kern_accessat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
2175 	    uap->mode));
2176 }
2177 
2178 int
2179 kern_access(struct thread *td, char *path, enum uio_seg pathseg, int mode)
2180 {
2181 
2182 	return (kern_accessat(td, AT_FDCWD, path, pathseg, 0, mode));
2183 }
2184 
2185 int
2186 kern_accessat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2187     int flags, int mode)
2188 {
2189 	struct ucred *cred, *tmpcred;
2190 	struct vnode *vp;
2191 	struct nameidata nd;
2192 	int vfslocked;
2193 	int error;
2194 
2195 	/*
2196 	 * Create and modify a temporary credential instead of one that
2197 	 * is potentially shared.
2198 	 */
2199 	if (!(flags & AT_EACCESS)) {
2200 		cred = td->td_ucred;
2201 		tmpcred = crdup(cred);
2202 		tmpcred->cr_uid = cred->cr_ruid;
2203 		tmpcred->cr_groups[0] = cred->cr_rgid;
2204 		td->td_ucred = tmpcred;
2205 	} else
2206 		cred = tmpcred = td->td_ucred;
2207 	AUDIT_ARG_VALUE(mode);
2208 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | MPSAFE |
2209 	    AUDITVNODE1, pathseg, path, fd, CAP_FSTAT, td);
2210 	if ((error = namei(&nd)) != 0)
2211 		goto out1;
2212 	vfslocked = NDHASGIANT(&nd);
2213 	vp = nd.ni_vp;
2214 
2215 	error = vn_access(vp, mode, tmpcred, td);
2216 	NDFREE(&nd, NDF_ONLY_PNBUF);
2217 	vput(vp);
2218 	VFS_UNLOCK_GIANT(vfslocked);
2219 out1:
2220 	if (!(flags & AT_EACCESS)) {
2221 		td->td_ucred = cred;
2222 		crfree(tmpcred);
2223 	}
2224 	return (error);
2225 }
2226 
2227 /*
2228  * Check access permissions using "effective" credentials.
2229  */
2230 #ifndef _SYS_SYSPROTO_H_
2231 struct eaccess_args {
2232 	char	*path;
2233 	int	flags;
2234 };
2235 #endif
2236 int
2237 sys_eaccess(td, uap)
2238 	struct thread *td;
2239 	register struct eaccess_args /* {
2240 		char *path;
2241 		int flags;
2242 	} */ *uap;
2243 {
2244 
2245 	return (kern_eaccess(td, uap->path, UIO_USERSPACE, uap->flags));
2246 }
2247 
2248 int
2249 kern_eaccess(struct thread *td, char *path, enum uio_seg pathseg, int flags)
2250 {
2251 
2252 	return (kern_accessat(td, AT_FDCWD, path, pathseg, AT_EACCESS, flags));
2253 }
2254 
2255 #if defined(COMPAT_43)
2256 /*
2257  * Get file status; this version follows links.
2258  */
2259 #ifndef _SYS_SYSPROTO_H_
2260 struct ostat_args {
2261 	char	*path;
2262 	struct ostat *ub;
2263 };
2264 #endif
2265 int
2266 ostat(td, uap)
2267 	struct thread *td;
2268 	register struct ostat_args /* {
2269 		char *path;
2270 		struct ostat *ub;
2271 	} */ *uap;
2272 {
2273 	struct stat sb;
2274 	struct ostat osb;
2275 	int error;
2276 
2277 	error = kern_stat(td, uap->path, UIO_USERSPACE, &sb);
2278 	if (error)
2279 		return (error);
2280 	cvtstat(&sb, &osb);
2281 	error = copyout(&osb, uap->ub, sizeof (osb));
2282 	return (error);
2283 }
2284 
2285 /*
2286  * Get file status; this version does not follow links.
2287  */
2288 #ifndef _SYS_SYSPROTO_H_
2289 struct olstat_args {
2290 	char	*path;
2291 	struct ostat *ub;
2292 };
2293 #endif
2294 int
2295 olstat(td, uap)
2296 	struct thread *td;
2297 	register struct olstat_args /* {
2298 		char *path;
2299 		struct ostat *ub;
2300 	} */ *uap;
2301 {
2302 	struct stat sb;
2303 	struct ostat osb;
2304 	int error;
2305 
2306 	error = kern_lstat(td, uap->path, UIO_USERSPACE, &sb);
2307 	if (error)
2308 		return (error);
2309 	cvtstat(&sb, &osb);
2310 	error = copyout(&osb, uap->ub, sizeof (osb));
2311 	return (error);
2312 }
2313 
2314 /*
2315  * Convert from an old to a new stat structure.
2316  */
2317 void
2318 cvtstat(st, ost)
2319 	struct stat *st;
2320 	struct ostat *ost;
2321 {
2322 
2323 	ost->st_dev = st->st_dev;
2324 	ost->st_ino = st->st_ino;
2325 	ost->st_mode = st->st_mode;
2326 	ost->st_nlink = st->st_nlink;
2327 	ost->st_uid = st->st_uid;
2328 	ost->st_gid = st->st_gid;
2329 	ost->st_rdev = st->st_rdev;
2330 	if (st->st_size < (quad_t)1 << 32)
2331 		ost->st_size = st->st_size;
2332 	else
2333 		ost->st_size = -2;
2334 	ost->st_atim = st->st_atim;
2335 	ost->st_mtim = st->st_mtim;
2336 	ost->st_ctim = st->st_ctim;
2337 	ost->st_blksize = st->st_blksize;
2338 	ost->st_blocks = st->st_blocks;
2339 	ost->st_flags = st->st_flags;
2340 	ost->st_gen = st->st_gen;
2341 }
2342 #endif /* COMPAT_43 */
2343 
2344 /*
2345  * Get file status; this version follows links.
2346  */
2347 #ifndef _SYS_SYSPROTO_H_
2348 struct stat_args {
2349 	char	*path;
2350 	struct stat *ub;
2351 };
2352 #endif
2353 int
2354 sys_stat(td, uap)
2355 	struct thread *td;
2356 	register struct stat_args /* {
2357 		char *path;
2358 		struct stat *ub;
2359 	} */ *uap;
2360 {
2361 	struct stat sb;
2362 	int error;
2363 
2364 	error = kern_stat(td, uap->path, UIO_USERSPACE, &sb);
2365 	if (error == 0)
2366 		error = copyout(&sb, uap->ub, sizeof (sb));
2367 	return (error);
2368 }
2369 
2370 #ifndef _SYS_SYSPROTO_H_
2371 struct fstatat_args {
2372 	int	fd;
2373 	char	*path;
2374 	struct stat	*buf;
2375 	int	flag;
2376 }
2377 #endif
2378 int
2379 sys_fstatat(struct thread *td, struct fstatat_args *uap)
2380 {
2381 	struct stat sb;
2382 	int error;
2383 
2384 	error = kern_statat(td, uap->flag, uap->fd, uap->path,
2385 	    UIO_USERSPACE, &sb);
2386 	if (error == 0)
2387 		error = copyout(&sb, uap->buf, sizeof (sb));
2388 	return (error);
2389 }
2390 
2391 int
2392 kern_stat(struct thread *td, char *path, enum uio_seg pathseg, struct stat *sbp)
2393 {
2394 
2395 	return (kern_statat(td, 0, AT_FDCWD, path, pathseg, sbp));
2396 }
2397 
2398 int
2399 kern_statat(struct thread *td, int flag, int fd, char *path,
2400     enum uio_seg pathseg, struct stat *sbp)
2401 {
2402 
2403 	return (kern_statat_vnhook(td, flag, fd, path, pathseg, sbp, NULL));
2404 }
2405 
2406 int
2407 kern_statat_vnhook(struct thread *td, int flag, int fd, char *path,
2408     enum uio_seg pathseg, struct stat *sbp,
2409     void (*hook)(struct vnode *vp, struct stat *sbp))
2410 {
2411 	struct nameidata nd;
2412 	struct stat sb;
2413 	int error, vfslocked;
2414 
2415 	if (flag & ~AT_SYMLINK_NOFOLLOW)
2416 		return (EINVAL);
2417 
2418 	NDINIT_ATRIGHTS(&nd, LOOKUP, ((flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW :
2419 	    FOLLOW) | LOCKSHARED | LOCKLEAF | AUDITVNODE1 | MPSAFE, pathseg,
2420 	    path, fd, CAP_FSTAT, td);
2421 
2422 	if ((error = namei(&nd)) != 0)
2423 		return (error);
2424 	vfslocked = NDHASGIANT(&nd);
2425 	error = vn_stat(nd.ni_vp, &sb, td->td_ucred, NOCRED, td);
2426 	if (!error) {
2427 		SDT_PROBE(vfs, , stat, mode, path, sb.st_mode, 0, 0, 0);
2428 		if (S_ISREG(sb.st_mode))
2429 			SDT_PROBE(vfs, , stat, reg, path, pathseg, 0, 0, 0);
2430 		if (__predict_false(hook != NULL))
2431 			hook(nd.ni_vp, &sb);
2432 	}
2433 	NDFREE(&nd, NDF_ONLY_PNBUF);
2434 	vput(nd.ni_vp);
2435 	VFS_UNLOCK_GIANT(vfslocked);
2436 	if (error)
2437 		return (error);
2438 	*sbp = sb;
2439 #ifdef KTRACE
2440 	if (KTRPOINT(td, KTR_STRUCT))
2441 		ktrstat(&sb);
2442 #endif
2443 	return (0);
2444 }
2445 
2446 /*
2447  * Get file status; this version does not follow links.
2448  */
2449 #ifndef _SYS_SYSPROTO_H_
2450 struct lstat_args {
2451 	char	*path;
2452 	struct stat *ub;
2453 };
2454 #endif
2455 int
2456 sys_lstat(td, uap)
2457 	struct thread *td;
2458 	register struct lstat_args /* {
2459 		char *path;
2460 		struct stat *ub;
2461 	} */ *uap;
2462 {
2463 	struct stat sb;
2464 	int error;
2465 
2466 	error = kern_lstat(td, uap->path, UIO_USERSPACE, &sb);
2467 	if (error == 0)
2468 		error = copyout(&sb, uap->ub, sizeof (sb));
2469 	return (error);
2470 }
2471 
2472 int
2473 kern_lstat(struct thread *td, char *path, enum uio_seg pathseg, struct stat *sbp)
2474 {
2475 
2476 	return (kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, path, pathseg,
2477 	    sbp));
2478 }
2479 
2480 /*
2481  * Implementation of the NetBSD [l]stat() functions.
2482  */
2483 void
2484 cvtnstat(sb, nsb)
2485 	struct stat *sb;
2486 	struct nstat *nsb;
2487 {
2488 	bzero(nsb, sizeof *nsb);
2489 	nsb->st_dev = sb->st_dev;
2490 	nsb->st_ino = sb->st_ino;
2491 	nsb->st_mode = sb->st_mode;
2492 	nsb->st_nlink = sb->st_nlink;
2493 	nsb->st_uid = sb->st_uid;
2494 	nsb->st_gid = sb->st_gid;
2495 	nsb->st_rdev = sb->st_rdev;
2496 	nsb->st_atim = sb->st_atim;
2497 	nsb->st_mtim = sb->st_mtim;
2498 	nsb->st_ctim = sb->st_ctim;
2499 	nsb->st_size = sb->st_size;
2500 	nsb->st_blocks = sb->st_blocks;
2501 	nsb->st_blksize = sb->st_blksize;
2502 	nsb->st_flags = sb->st_flags;
2503 	nsb->st_gen = sb->st_gen;
2504 	nsb->st_birthtim = sb->st_birthtim;
2505 }
2506 
2507 #ifndef _SYS_SYSPROTO_H_
2508 struct nstat_args {
2509 	char	*path;
2510 	struct nstat *ub;
2511 };
2512 #endif
2513 int
2514 sys_nstat(td, uap)
2515 	struct thread *td;
2516 	register struct nstat_args /* {
2517 		char *path;
2518 		struct nstat *ub;
2519 	} */ *uap;
2520 {
2521 	struct stat sb;
2522 	struct nstat nsb;
2523 	int error;
2524 
2525 	error = kern_stat(td, uap->path, UIO_USERSPACE, &sb);
2526 	if (error)
2527 		return (error);
2528 	cvtnstat(&sb, &nsb);
2529 	error = copyout(&nsb, uap->ub, sizeof (nsb));
2530 	return (error);
2531 }
2532 
2533 /*
2534  * NetBSD lstat.  Get file status; this version does not follow links.
2535  */
2536 #ifndef _SYS_SYSPROTO_H_
2537 struct lstat_args {
2538 	char	*path;
2539 	struct stat *ub;
2540 };
2541 #endif
2542 int
2543 sys_nlstat(td, uap)
2544 	struct thread *td;
2545 	register struct nlstat_args /* {
2546 		char *path;
2547 		struct nstat *ub;
2548 	} */ *uap;
2549 {
2550 	struct stat sb;
2551 	struct nstat nsb;
2552 	int error;
2553 
2554 	error = kern_lstat(td, uap->path, UIO_USERSPACE, &sb);
2555 	if (error)
2556 		return (error);
2557 	cvtnstat(&sb, &nsb);
2558 	error = copyout(&nsb, uap->ub, sizeof (nsb));
2559 	return (error);
2560 }
2561 
2562 /*
2563  * Get configurable pathname variables.
2564  */
2565 #ifndef _SYS_SYSPROTO_H_
2566 struct pathconf_args {
2567 	char	*path;
2568 	int	name;
2569 };
2570 #endif
2571 int
2572 sys_pathconf(td, uap)
2573 	struct thread *td;
2574 	register struct pathconf_args /* {
2575 		char *path;
2576 		int name;
2577 	} */ *uap;
2578 {
2579 
2580 	return (kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name, FOLLOW));
2581 }
2582 
2583 #ifndef _SYS_SYSPROTO_H_
2584 struct lpathconf_args {
2585 	char	*path;
2586 	int	name;
2587 };
2588 #endif
2589 int
2590 sys_lpathconf(td, uap)
2591 	struct thread *td;
2592 	register struct lpathconf_args /* {
2593 		char *path;
2594 		int name;
2595 	} */ *uap;
2596 {
2597 
2598 	return (kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name, NOFOLLOW));
2599 }
2600 
2601 int
2602 kern_pathconf(struct thread *td, char *path, enum uio_seg pathseg, int name,
2603     u_long flags)
2604 {
2605 	struct nameidata nd;
2606 	int error, vfslocked;
2607 
2608 	NDINIT(&nd, LOOKUP, LOCKSHARED | LOCKLEAF | MPSAFE | AUDITVNODE1 |
2609 	    flags, pathseg, path, td);
2610 	if ((error = namei(&nd)) != 0)
2611 		return (error);
2612 	vfslocked = NDHASGIANT(&nd);
2613 	NDFREE(&nd, NDF_ONLY_PNBUF);
2614 
2615 	/* If asynchronous I/O is available, it works for all files. */
2616 	if (name == _PC_ASYNC_IO)
2617 		td->td_retval[0] = async_io_version;
2618 	else
2619 		error = VOP_PATHCONF(nd.ni_vp, name, td->td_retval);
2620 	vput(nd.ni_vp);
2621 	VFS_UNLOCK_GIANT(vfslocked);
2622 	return (error);
2623 }
2624 
2625 /*
2626  * Return target name of a symbolic link.
2627  */
2628 #ifndef _SYS_SYSPROTO_H_
2629 struct readlink_args {
2630 	char	*path;
2631 	char	*buf;
2632 	size_t	count;
2633 };
2634 #endif
2635 int
2636 sys_readlink(td, uap)
2637 	struct thread *td;
2638 	register struct readlink_args /* {
2639 		char *path;
2640 		char *buf;
2641 		size_t count;
2642 	} */ *uap;
2643 {
2644 
2645 	return (kern_readlink(td, uap->path, UIO_USERSPACE, uap->buf,
2646 	    UIO_USERSPACE, uap->count));
2647 }
2648 #ifndef _SYS_SYSPROTO_H_
2649 struct readlinkat_args {
2650 	int	fd;
2651 	char	*path;
2652 	char	*buf;
2653 	size_t	bufsize;
2654 };
2655 #endif
2656 int
2657 sys_readlinkat(struct thread *td, struct readlinkat_args *uap)
2658 {
2659 
2660 	return (kern_readlinkat(td, uap->fd, uap->path, UIO_USERSPACE,
2661 	    uap->buf, UIO_USERSPACE, uap->bufsize));
2662 }
2663 
2664 int
2665 kern_readlink(struct thread *td, char *path, enum uio_seg pathseg, char *buf,
2666     enum uio_seg bufseg, size_t count)
2667 {
2668 
2669 	return (kern_readlinkat(td, AT_FDCWD, path, pathseg, buf, bufseg,
2670 	    count));
2671 }
2672 
2673 int
2674 kern_readlinkat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2675     char *buf, enum uio_seg bufseg, size_t count)
2676 {
2677 	struct vnode *vp;
2678 	struct iovec aiov;
2679 	struct uio auio;
2680 	int error;
2681 	struct nameidata nd;
2682 	int vfslocked;
2683 
2684 	if (count > INT_MAX)
2685 		return (EINVAL);
2686 
2687 	NDINIT_AT(&nd, LOOKUP, NOFOLLOW | LOCKSHARED | LOCKLEAF | MPSAFE |
2688 	    AUDITVNODE1, pathseg, path, fd, td);
2689 
2690 	if ((error = namei(&nd)) != 0)
2691 		return (error);
2692 	NDFREE(&nd, NDF_ONLY_PNBUF);
2693 	vfslocked = NDHASGIANT(&nd);
2694 	vp = nd.ni_vp;
2695 #ifdef MAC
2696 	error = mac_vnode_check_readlink(td->td_ucred, vp);
2697 	if (error) {
2698 		vput(vp);
2699 		VFS_UNLOCK_GIANT(vfslocked);
2700 		return (error);
2701 	}
2702 #endif
2703 	if (vp->v_type != VLNK)
2704 		error = EINVAL;
2705 	else {
2706 		aiov.iov_base = buf;
2707 		aiov.iov_len = count;
2708 		auio.uio_iov = &aiov;
2709 		auio.uio_iovcnt = 1;
2710 		auio.uio_offset = 0;
2711 		auio.uio_rw = UIO_READ;
2712 		auio.uio_segflg = bufseg;
2713 		auio.uio_td = td;
2714 		auio.uio_resid = count;
2715 		error = VOP_READLINK(vp, &auio, td->td_ucred);
2716 	}
2717 	vput(vp);
2718 	VFS_UNLOCK_GIANT(vfslocked);
2719 	td->td_retval[0] = count - auio.uio_resid;
2720 	return (error);
2721 }
2722 
2723 /*
2724  * Common implementation code for chflags() and fchflags().
2725  */
2726 static int
2727 setfflags(td, vp, flags)
2728 	struct thread *td;
2729 	struct vnode *vp;
2730 	int flags;
2731 {
2732 	int error;
2733 	struct mount *mp;
2734 	struct vattr vattr;
2735 
2736 	/*
2737 	 * Prevent non-root users from setting flags on devices.  When
2738 	 * a device is reused, users can retain ownership of the device
2739 	 * if they are allowed to set flags and programs assume that
2740 	 * chown can't fail when done as root.
2741 	 */
2742 	if (vp->v_type == VCHR || vp->v_type == VBLK) {
2743 		error = priv_check(td, PRIV_VFS_CHFLAGS_DEV);
2744 		if (error)
2745 			return (error);
2746 	}
2747 
2748 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2749 		return (error);
2750 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2751 	VATTR_NULL(&vattr);
2752 	vattr.va_flags = flags;
2753 #ifdef MAC
2754 	error = mac_vnode_check_setflags(td->td_ucred, vp, vattr.va_flags);
2755 	if (error == 0)
2756 #endif
2757 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
2758 	VOP_UNLOCK(vp, 0);
2759 	vn_finished_write(mp);
2760 	return (error);
2761 }
2762 
2763 /*
2764  * Change flags of a file given a path name.
2765  */
2766 #ifndef _SYS_SYSPROTO_H_
2767 struct chflags_args {
2768 	char	*path;
2769 	int	flags;
2770 };
2771 #endif
2772 int
2773 sys_chflags(td, uap)
2774 	struct thread *td;
2775 	register struct chflags_args /* {
2776 		char *path;
2777 		int flags;
2778 	} */ *uap;
2779 {
2780 	int error;
2781 	struct nameidata nd;
2782 	int vfslocked;
2783 
2784 	AUDIT_ARG_FFLAGS(uap->flags);
2785 	NDINIT(&nd, LOOKUP, FOLLOW | MPSAFE | AUDITVNODE1, UIO_USERSPACE,
2786 	    uap->path, td);
2787 	if ((error = namei(&nd)) != 0)
2788 		return (error);
2789 	NDFREE(&nd, NDF_ONLY_PNBUF);
2790 	vfslocked = NDHASGIANT(&nd);
2791 	error = setfflags(td, nd.ni_vp, uap->flags);
2792 	vrele(nd.ni_vp);
2793 	VFS_UNLOCK_GIANT(vfslocked);
2794 	return (error);
2795 }
2796 
2797 /*
2798  * Same as chflags() but doesn't follow symlinks.
2799  */
2800 int
2801 sys_lchflags(td, uap)
2802 	struct thread *td;
2803 	register struct lchflags_args /* {
2804 		char *path;
2805 		int flags;
2806 	} */ *uap;
2807 {
2808 	int error;
2809 	struct nameidata nd;
2810 	int vfslocked;
2811 
2812 	AUDIT_ARG_FFLAGS(uap->flags);
2813 	NDINIT(&nd, LOOKUP, NOFOLLOW | MPSAFE | AUDITVNODE1, UIO_USERSPACE,
2814 	    uap->path, td);
2815 	if ((error = namei(&nd)) != 0)
2816 		return (error);
2817 	vfslocked = NDHASGIANT(&nd);
2818 	NDFREE(&nd, NDF_ONLY_PNBUF);
2819 	error = setfflags(td, nd.ni_vp, uap->flags);
2820 	vrele(nd.ni_vp);
2821 	VFS_UNLOCK_GIANT(vfslocked);
2822 	return (error);
2823 }
2824 
2825 /*
2826  * Change flags of a file given a file descriptor.
2827  */
2828 #ifndef _SYS_SYSPROTO_H_
2829 struct fchflags_args {
2830 	int	fd;
2831 	int	flags;
2832 };
2833 #endif
2834 int
2835 sys_fchflags(td, uap)
2836 	struct thread *td;
2837 	register struct fchflags_args /* {
2838 		int fd;
2839 		int flags;
2840 	} */ *uap;
2841 {
2842 	struct file *fp;
2843 	int vfslocked;
2844 	int error;
2845 
2846 	AUDIT_ARG_FD(uap->fd);
2847 	AUDIT_ARG_FFLAGS(uap->flags);
2848 	if ((error = getvnode(td->td_proc->p_fd, uap->fd, CAP_FCHFLAGS,
2849 	    &fp)) != 0)
2850 		return (error);
2851 	vfslocked = VFS_LOCK_GIANT(fp->f_vnode->v_mount);
2852 #ifdef AUDIT
2853 	vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
2854 	AUDIT_ARG_VNODE1(fp->f_vnode);
2855 	VOP_UNLOCK(fp->f_vnode, 0);
2856 #endif
2857 	error = setfflags(td, fp->f_vnode, uap->flags);
2858 	VFS_UNLOCK_GIANT(vfslocked);
2859 	fdrop(fp, td);
2860 	return (error);
2861 }
2862 
2863 /*
2864  * Common implementation code for chmod(), lchmod() and fchmod().
2865  */
2866 int
2867 setfmode(td, cred, vp, mode)
2868 	struct thread *td;
2869 	struct ucred *cred;
2870 	struct vnode *vp;
2871 	int mode;
2872 {
2873 	int error;
2874 	struct mount *mp;
2875 	struct vattr vattr;
2876 
2877 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2878 		return (error);
2879 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2880 	VATTR_NULL(&vattr);
2881 	vattr.va_mode = mode & ALLPERMS;
2882 #ifdef MAC
2883 	error = mac_vnode_check_setmode(cred, vp, vattr.va_mode);
2884 	if (error == 0)
2885 #endif
2886 		error = VOP_SETATTR(vp, &vattr, cred);
2887 	VOP_UNLOCK(vp, 0);
2888 	vn_finished_write(mp);
2889 	return (error);
2890 }
2891 
2892 /*
2893  * Change mode of a file given path name.
2894  */
2895 #ifndef _SYS_SYSPROTO_H_
2896 struct chmod_args {
2897 	char	*path;
2898 	int	mode;
2899 };
2900 #endif
2901 int
2902 sys_chmod(td, uap)
2903 	struct thread *td;
2904 	register struct chmod_args /* {
2905 		char *path;
2906 		int mode;
2907 	} */ *uap;
2908 {
2909 
2910 	return (kern_chmod(td, uap->path, UIO_USERSPACE, uap->mode));
2911 }
2912 
2913 #ifndef _SYS_SYSPROTO_H_
2914 struct fchmodat_args {
2915 	int	dirfd;
2916 	char	*path;
2917 	mode_t	mode;
2918 	int	flag;
2919 }
2920 #endif
2921 int
2922 sys_fchmodat(struct thread *td, struct fchmodat_args *uap)
2923 {
2924 	int flag = uap->flag;
2925 	int fd = uap->fd;
2926 	char *path = uap->path;
2927 	mode_t mode = uap->mode;
2928 
2929 	if (flag & ~AT_SYMLINK_NOFOLLOW)
2930 		return (EINVAL);
2931 
2932 	return (kern_fchmodat(td, fd, path, UIO_USERSPACE, mode, flag));
2933 }
2934 
2935 int
2936 kern_chmod(struct thread *td, char *path, enum uio_seg pathseg, int mode)
2937 {
2938 
2939 	return (kern_fchmodat(td, AT_FDCWD, path, pathseg, mode, 0));
2940 }
2941 
2942 /*
2943  * Change mode of a file given path name (don't follow links.)
2944  */
2945 #ifndef _SYS_SYSPROTO_H_
2946 struct lchmod_args {
2947 	char	*path;
2948 	int	mode;
2949 };
2950 #endif
2951 int
2952 sys_lchmod(td, uap)
2953 	struct thread *td;
2954 	register struct lchmod_args /* {
2955 		char *path;
2956 		int mode;
2957 	} */ *uap;
2958 {
2959 
2960 	return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2961 	    uap->mode, AT_SYMLINK_NOFOLLOW));
2962 }
2963 
2964 
2965 int
2966 kern_fchmodat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2967     mode_t mode, int flag)
2968 {
2969 	int error;
2970 	struct nameidata nd;
2971 	int vfslocked;
2972 	int follow;
2973 
2974 	AUDIT_ARG_MODE(mode);
2975 	follow = (flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
2976 	NDINIT_ATRIGHTS(&nd, LOOKUP,  follow | MPSAFE | AUDITVNODE1, pathseg,
2977 	    path, fd, CAP_FCHMOD, td);
2978 	if ((error = namei(&nd)) != 0)
2979 		return (error);
2980 	vfslocked = NDHASGIANT(&nd);
2981 	NDFREE(&nd, NDF_ONLY_PNBUF);
2982 	error = setfmode(td, td->td_ucred, nd.ni_vp, mode);
2983 	vrele(nd.ni_vp);
2984 	VFS_UNLOCK_GIANT(vfslocked);
2985 	return (error);
2986 }
2987 
2988 /*
2989  * Change mode of a file given a file descriptor.
2990  */
2991 #ifndef _SYS_SYSPROTO_H_
2992 struct fchmod_args {
2993 	int	fd;
2994 	int	mode;
2995 };
2996 #endif
2997 int
2998 sys_fchmod(struct thread *td, struct fchmod_args *uap)
2999 {
3000 	struct file *fp;
3001 	int error;
3002 
3003 	AUDIT_ARG_FD(uap->fd);
3004 	AUDIT_ARG_MODE(uap->mode);
3005 
3006 	error = fget(td, uap->fd, CAP_FCHMOD, &fp);
3007 	if (error != 0)
3008 		return (error);
3009 	error = fo_chmod(fp, uap->mode, td->td_ucred, td);
3010 	fdrop(fp, td);
3011 	return (error);
3012 }
3013 
3014 /*
3015  * Common implementation for chown(), lchown(), and fchown()
3016  */
3017 int
3018 setfown(td, cred, vp, uid, gid)
3019 	struct thread *td;
3020 	struct ucred *cred;
3021 	struct vnode *vp;
3022 	uid_t uid;
3023 	gid_t gid;
3024 {
3025 	int error;
3026 	struct mount *mp;
3027 	struct vattr vattr;
3028 
3029 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
3030 		return (error);
3031 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3032 	VATTR_NULL(&vattr);
3033 	vattr.va_uid = uid;
3034 	vattr.va_gid = gid;
3035 #ifdef MAC
3036 	error = mac_vnode_check_setowner(cred, vp, vattr.va_uid,
3037 	    vattr.va_gid);
3038 	if (error == 0)
3039 #endif
3040 		error = VOP_SETATTR(vp, &vattr, cred);
3041 	VOP_UNLOCK(vp, 0);
3042 	vn_finished_write(mp);
3043 	return (error);
3044 }
3045 
3046 /*
3047  * Set ownership given a path name.
3048  */
3049 #ifndef _SYS_SYSPROTO_H_
3050 struct chown_args {
3051 	char	*path;
3052 	int	uid;
3053 	int	gid;
3054 };
3055 #endif
3056 int
3057 sys_chown(td, uap)
3058 	struct thread *td;
3059 	register struct chown_args /* {
3060 		char *path;
3061 		int uid;
3062 		int gid;
3063 	} */ *uap;
3064 {
3065 
3066 	return (kern_chown(td, uap->path, UIO_USERSPACE, uap->uid, uap->gid));
3067 }
3068 
3069 #ifndef _SYS_SYSPROTO_H_
3070 struct fchownat_args {
3071 	int fd;
3072 	const char * path;
3073 	uid_t uid;
3074 	gid_t gid;
3075 	int flag;
3076 };
3077 #endif
3078 int
3079 sys_fchownat(struct thread *td, struct fchownat_args *uap)
3080 {
3081 	int flag;
3082 
3083 	flag = uap->flag;
3084 	if (flag & ~AT_SYMLINK_NOFOLLOW)
3085 		return (EINVAL);
3086 
3087 	return (kern_fchownat(td, uap->fd, uap->path, UIO_USERSPACE, uap->uid,
3088 	    uap->gid, uap->flag));
3089 }
3090 
3091 int
3092 kern_chown(struct thread *td, char *path, enum uio_seg pathseg, int uid,
3093     int gid)
3094 {
3095 
3096 	return (kern_fchownat(td, AT_FDCWD, path, pathseg, uid, gid, 0));
3097 }
3098 
3099 int
3100 kern_fchownat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
3101     int uid, int gid, int flag)
3102 {
3103 	struct nameidata nd;
3104 	int error, vfslocked, follow;
3105 
3106 	AUDIT_ARG_OWNER(uid, gid);
3107 	follow = (flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
3108 	NDINIT_ATRIGHTS(&nd, LOOKUP, follow | MPSAFE | AUDITVNODE1, pathseg,
3109 	    path, fd, CAP_FCHOWN, td);
3110 
3111 	if ((error = namei(&nd)) != 0)
3112 		return (error);
3113 	vfslocked = NDHASGIANT(&nd);
3114 	NDFREE(&nd, NDF_ONLY_PNBUF);
3115 	error = setfown(td, td->td_ucred, nd.ni_vp, uid, gid);
3116 	vrele(nd.ni_vp);
3117 	VFS_UNLOCK_GIANT(vfslocked);
3118 	return (error);
3119 }
3120 
3121 /*
3122  * Set ownership given a path name, do not cross symlinks.
3123  */
3124 #ifndef _SYS_SYSPROTO_H_
3125 struct lchown_args {
3126 	char	*path;
3127 	int	uid;
3128 	int	gid;
3129 };
3130 #endif
3131 int
3132 sys_lchown(td, uap)
3133 	struct thread *td;
3134 	register struct lchown_args /* {
3135 		char *path;
3136 		int uid;
3137 		int gid;
3138 	} */ *uap;
3139 {
3140 
3141 	return (kern_lchown(td, uap->path, UIO_USERSPACE, uap->uid, uap->gid));
3142 }
3143 
3144 int
3145 kern_lchown(struct thread *td, char *path, enum uio_seg pathseg, int uid,
3146     int gid)
3147 {
3148 
3149 	return (kern_fchownat(td, AT_FDCWD, path, pathseg, uid, gid,
3150 	    AT_SYMLINK_NOFOLLOW));
3151 }
3152 
3153 /*
3154  * Set ownership given a file descriptor.
3155  */
3156 #ifndef _SYS_SYSPROTO_H_
3157 struct fchown_args {
3158 	int	fd;
3159 	int	uid;
3160 	int	gid;
3161 };
3162 #endif
3163 int
3164 sys_fchown(td, uap)
3165 	struct thread *td;
3166 	register struct fchown_args /* {
3167 		int fd;
3168 		int uid;
3169 		int gid;
3170 	} */ *uap;
3171 {
3172 	struct file *fp;
3173 	int error;
3174 
3175 	AUDIT_ARG_FD(uap->fd);
3176 	AUDIT_ARG_OWNER(uap->uid, uap->gid);
3177 	error = fget(td, uap->fd, CAP_FCHOWN, &fp);
3178 	if (error != 0)
3179 		return (error);
3180 	error = fo_chown(fp, uap->uid, uap->gid, td->td_ucred, td);
3181 	fdrop(fp, td);
3182 	return (error);
3183 }
3184 
3185 /*
3186  * Common implementation code for utimes(), lutimes(), and futimes().
3187  */
3188 static int
3189 getutimes(usrtvp, tvpseg, tsp)
3190 	const struct timeval *usrtvp;
3191 	enum uio_seg tvpseg;
3192 	struct timespec *tsp;
3193 {
3194 	struct timeval tv[2];
3195 	const struct timeval *tvp;
3196 	int error;
3197 
3198 	if (usrtvp == NULL) {
3199 		vfs_timestamp(&tsp[0]);
3200 		tsp[1] = tsp[0];
3201 	} else {
3202 		if (tvpseg == UIO_SYSSPACE) {
3203 			tvp = usrtvp;
3204 		} else {
3205 			if ((error = copyin(usrtvp, tv, sizeof(tv))) != 0)
3206 				return (error);
3207 			tvp = tv;
3208 		}
3209 
3210 		if (tvp[0].tv_usec < 0 || tvp[0].tv_usec >= 1000000 ||
3211 		    tvp[1].tv_usec < 0 || tvp[1].tv_usec >= 1000000)
3212 			return (EINVAL);
3213 		TIMEVAL_TO_TIMESPEC(&tvp[0], &tsp[0]);
3214 		TIMEVAL_TO_TIMESPEC(&tvp[1], &tsp[1]);
3215 	}
3216 	return (0);
3217 }
3218 
3219 /*
3220  * Common implementation code for utimes(), lutimes(), and futimes().
3221  */
3222 static int
3223 setutimes(td, vp, ts, numtimes, nullflag)
3224 	struct thread *td;
3225 	struct vnode *vp;
3226 	const struct timespec *ts;
3227 	int numtimes;
3228 	int nullflag;
3229 {
3230 	int error, setbirthtime;
3231 	struct mount *mp;
3232 	struct vattr vattr;
3233 
3234 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
3235 		return (error);
3236 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3237 	setbirthtime = 0;
3238 	if (numtimes < 3 && !VOP_GETATTR(vp, &vattr, td->td_ucred) &&
3239 	    timespeccmp(&ts[1], &vattr.va_birthtime, < ))
3240 		setbirthtime = 1;
3241 	VATTR_NULL(&vattr);
3242 	vattr.va_atime = ts[0];
3243 	vattr.va_mtime = ts[1];
3244 	if (setbirthtime)
3245 		vattr.va_birthtime = ts[1];
3246 	if (numtimes > 2)
3247 		vattr.va_birthtime = ts[2];
3248 	if (nullflag)
3249 		vattr.va_vaflags |= VA_UTIMES_NULL;
3250 #ifdef MAC
3251 	error = mac_vnode_check_setutimes(td->td_ucred, vp, vattr.va_atime,
3252 	    vattr.va_mtime);
3253 #endif
3254 	if (error == 0)
3255 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3256 	VOP_UNLOCK(vp, 0);
3257 	vn_finished_write(mp);
3258 	return (error);
3259 }
3260 
3261 /*
3262  * Set the access and modification times of a file.
3263  */
3264 #ifndef _SYS_SYSPROTO_H_
3265 struct utimes_args {
3266 	char	*path;
3267 	struct	timeval *tptr;
3268 };
3269 #endif
3270 int
3271 sys_utimes(td, uap)
3272 	struct thread *td;
3273 	register struct utimes_args /* {
3274 		char *path;
3275 		struct timeval *tptr;
3276 	} */ *uap;
3277 {
3278 
3279 	return (kern_utimes(td, uap->path, UIO_USERSPACE, uap->tptr,
3280 	    UIO_USERSPACE));
3281 }
3282 
3283 #ifndef _SYS_SYSPROTO_H_
3284 struct futimesat_args {
3285 	int fd;
3286 	const char * path;
3287 	const struct timeval * times;
3288 };
3289 #endif
3290 int
3291 sys_futimesat(struct thread *td, struct futimesat_args *uap)
3292 {
3293 
3294 	return (kern_utimesat(td, uap->fd, uap->path, UIO_USERSPACE,
3295 	    uap->times, UIO_USERSPACE));
3296 }
3297 
3298 int
3299 kern_utimes(struct thread *td, char *path, enum uio_seg pathseg,
3300     struct timeval *tptr, enum uio_seg tptrseg)
3301 {
3302 
3303 	return (kern_utimesat(td, AT_FDCWD, path, pathseg, tptr, tptrseg));
3304 }
3305 
3306 int
3307 kern_utimesat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
3308     struct timeval *tptr, enum uio_seg tptrseg)
3309 {
3310 	struct nameidata nd;
3311 	struct timespec ts[2];
3312 	int error, vfslocked;
3313 
3314 	if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3315 		return (error);
3316 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | MPSAFE | AUDITVNODE1, pathseg,
3317 	    path, fd, CAP_FUTIMES, td);
3318 
3319 	if ((error = namei(&nd)) != 0)
3320 		return (error);
3321 	vfslocked = NDHASGIANT(&nd);
3322 	NDFREE(&nd, NDF_ONLY_PNBUF);
3323 	error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3324 	vrele(nd.ni_vp);
3325 	VFS_UNLOCK_GIANT(vfslocked);
3326 	return (error);
3327 }
3328 
3329 /*
3330  * Set the access and modification times of a file.
3331  */
3332 #ifndef _SYS_SYSPROTO_H_
3333 struct lutimes_args {
3334 	char	*path;
3335 	struct	timeval *tptr;
3336 };
3337 #endif
3338 int
3339 sys_lutimes(td, uap)
3340 	struct thread *td;
3341 	register struct lutimes_args /* {
3342 		char *path;
3343 		struct timeval *tptr;
3344 	} */ *uap;
3345 {
3346 
3347 	return (kern_lutimes(td, uap->path, UIO_USERSPACE, uap->tptr,
3348 	    UIO_USERSPACE));
3349 }
3350 
3351 int
3352 kern_lutimes(struct thread *td, char *path, enum uio_seg pathseg,
3353     struct timeval *tptr, enum uio_seg tptrseg)
3354 {
3355 	struct timespec ts[2];
3356 	int error;
3357 	struct nameidata nd;
3358 	int vfslocked;
3359 
3360 	if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3361 		return (error);
3362 	NDINIT(&nd, LOOKUP, NOFOLLOW | MPSAFE | AUDITVNODE1, pathseg, path, td);
3363 	if ((error = namei(&nd)) != 0)
3364 		return (error);
3365 	vfslocked = NDHASGIANT(&nd);
3366 	NDFREE(&nd, NDF_ONLY_PNBUF);
3367 	error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3368 	vrele(nd.ni_vp);
3369 	VFS_UNLOCK_GIANT(vfslocked);
3370 	return (error);
3371 }
3372 
3373 /*
3374  * Set the access and modification times of a file.
3375  */
3376 #ifndef _SYS_SYSPROTO_H_
3377 struct futimes_args {
3378 	int	fd;
3379 	struct	timeval *tptr;
3380 };
3381 #endif
3382 int
3383 sys_futimes(td, uap)
3384 	struct thread *td;
3385 	register struct futimes_args /* {
3386 		int  fd;
3387 		struct timeval *tptr;
3388 	} */ *uap;
3389 {
3390 
3391 	return (kern_futimes(td, uap->fd, uap->tptr, UIO_USERSPACE));
3392 }
3393 
3394 int
3395 kern_futimes(struct thread *td, int fd, struct timeval *tptr,
3396     enum uio_seg tptrseg)
3397 {
3398 	struct timespec ts[2];
3399 	struct file *fp;
3400 	int vfslocked;
3401 	int error;
3402 
3403 	AUDIT_ARG_FD(fd);
3404 	if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3405 		return (error);
3406 	if ((error = getvnode(td->td_proc->p_fd, fd, CAP_FUTIMES, &fp))
3407 	    != 0)
3408 		return (error);
3409 	vfslocked = VFS_LOCK_GIANT(fp->f_vnode->v_mount);
3410 #ifdef AUDIT
3411 	vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
3412 	AUDIT_ARG_VNODE1(fp->f_vnode);
3413 	VOP_UNLOCK(fp->f_vnode, 0);
3414 #endif
3415 	error = setutimes(td, fp->f_vnode, ts, 2, tptr == NULL);
3416 	VFS_UNLOCK_GIANT(vfslocked);
3417 	fdrop(fp, td);
3418 	return (error);
3419 }
3420 
3421 /*
3422  * Truncate a file given its path name.
3423  */
3424 #ifndef _SYS_SYSPROTO_H_
3425 struct truncate_args {
3426 	char	*path;
3427 	int	pad;
3428 	off_t	length;
3429 };
3430 #endif
3431 int
3432 sys_truncate(td, uap)
3433 	struct thread *td;
3434 	register struct truncate_args /* {
3435 		char *path;
3436 		int pad;
3437 		off_t length;
3438 	} */ *uap;
3439 {
3440 
3441 	return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3442 }
3443 
3444 int
3445 kern_truncate(struct thread *td, char *path, enum uio_seg pathseg, off_t length)
3446 {
3447 	struct mount *mp;
3448 	struct vnode *vp;
3449 	struct vattr vattr;
3450 	int error;
3451 	struct nameidata nd;
3452 	int vfslocked;
3453 
3454 	if (length < 0)
3455 		return(EINVAL);
3456 	NDINIT(&nd, LOOKUP, FOLLOW | MPSAFE | AUDITVNODE1, pathseg, path, td);
3457 	if ((error = namei(&nd)) != 0)
3458 		return (error);
3459 	vfslocked = NDHASGIANT(&nd);
3460 	vp = nd.ni_vp;
3461 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) {
3462 		vrele(vp);
3463 		VFS_UNLOCK_GIANT(vfslocked);
3464 		return (error);
3465 	}
3466 	NDFREE(&nd, NDF_ONLY_PNBUF);
3467 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3468 	if (vp->v_type == VDIR)
3469 		error = EISDIR;
3470 #ifdef MAC
3471 	else if ((error = mac_vnode_check_write(td->td_ucred, NOCRED, vp))) {
3472 	}
3473 #endif
3474 	else if ((error = vn_writechk(vp)) == 0 &&
3475 	    (error = VOP_ACCESS(vp, VWRITE, td->td_ucred, td)) == 0) {
3476 		VATTR_NULL(&vattr);
3477 		vattr.va_size = length;
3478 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3479 	}
3480 	vput(vp);
3481 	vn_finished_write(mp);
3482 	VFS_UNLOCK_GIANT(vfslocked);
3483 	return (error);
3484 }
3485 
3486 #if defined(COMPAT_43)
3487 /*
3488  * Truncate a file given its path name.
3489  */
3490 #ifndef _SYS_SYSPROTO_H_
3491 struct otruncate_args {
3492 	char	*path;
3493 	long	length;
3494 };
3495 #endif
3496 int
3497 otruncate(td, uap)
3498 	struct thread *td;
3499 	register struct otruncate_args /* {
3500 		char *path;
3501 		long length;
3502 	} */ *uap;
3503 {
3504 	struct truncate_args /* {
3505 		char *path;
3506 		int pad;
3507 		off_t length;
3508 	} */ nuap;
3509 
3510 	nuap.path = uap->path;
3511 	nuap.length = uap->length;
3512 	return (sys_truncate(td, &nuap));
3513 }
3514 #endif /* COMPAT_43 */
3515 
3516 /* Versions with the pad argument */
3517 int
3518 freebsd6_truncate(struct thread *td, struct freebsd6_truncate_args *uap)
3519 {
3520 	struct truncate_args ouap;
3521 
3522 	ouap.path = uap->path;
3523 	ouap.length = uap->length;
3524 	return (sys_truncate(td, &ouap));
3525 }
3526 
3527 int
3528 freebsd6_ftruncate(struct thread *td, struct freebsd6_ftruncate_args *uap)
3529 {
3530 	struct ftruncate_args ouap;
3531 
3532 	ouap.fd = uap->fd;
3533 	ouap.length = uap->length;
3534 	return (sys_ftruncate(td, &ouap));
3535 }
3536 
3537 /*
3538  * Sync an open file.
3539  */
3540 #ifndef _SYS_SYSPROTO_H_
3541 struct fsync_args {
3542 	int	fd;
3543 };
3544 #endif
3545 int
3546 sys_fsync(td, uap)
3547 	struct thread *td;
3548 	struct fsync_args /* {
3549 		int fd;
3550 	} */ *uap;
3551 {
3552 	struct vnode *vp;
3553 	struct mount *mp;
3554 	struct file *fp;
3555 	int vfslocked;
3556 	int error, lock_flags;
3557 
3558 	AUDIT_ARG_FD(uap->fd);
3559 	if ((error = getvnode(td->td_proc->p_fd, uap->fd, CAP_FSYNC,
3560 	    &fp)) != 0)
3561 		return (error);
3562 	vp = fp->f_vnode;
3563 	vfslocked = VFS_LOCK_GIANT(vp->v_mount);
3564 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
3565 		goto drop;
3566 	if (MNT_SHARED_WRITES(mp) ||
3567 	    ((mp == NULL) && MNT_SHARED_WRITES(vp->v_mount))) {
3568 		lock_flags = LK_SHARED;
3569 	} else {
3570 		lock_flags = LK_EXCLUSIVE;
3571 	}
3572 	vn_lock(vp, lock_flags | LK_RETRY);
3573 	AUDIT_ARG_VNODE1(vp);
3574 	if (vp->v_object != NULL) {
3575 		VM_OBJECT_LOCK(vp->v_object);
3576 		vm_object_page_clean(vp->v_object, 0, 0, 0);
3577 		VM_OBJECT_UNLOCK(vp->v_object);
3578 	}
3579 	error = VOP_FSYNC(vp, MNT_WAIT, td);
3580 
3581 	VOP_UNLOCK(vp, 0);
3582 	vn_finished_write(mp);
3583 drop:
3584 	VFS_UNLOCK_GIANT(vfslocked);
3585 	fdrop(fp, td);
3586 	return (error);
3587 }
3588 
3589 /*
3590  * Rename files.  Source and destination must either both be directories, or
3591  * both not be directories.  If target is a directory, it must be empty.
3592  */
3593 #ifndef _SYS_SYSPROTO_H_
3594 struct rename_args {
3595 	char	*from;
3596 	char	*to;
3597 };
3598 #endif
3599 int
3600 sys_rename(td, uap)
3601 	struct thread *td;
3602 	register struct rename_args /* {
3603 		char *from;
3604 		char *to;
3605 	} */ *uap;
3606 {
3607 
3608 	return (kern_rename(td, uap->from, uap->to, UIO_USERSPACE));
3609 }
3610 
3611 #ifndef _SYS_SYSPROTO_H_
3612 struct renameat_args {
3613 	int	oldfd;
3614 	char	*old;
3615 	int	newfd;
3616 	char	*new;
3617 };
3618 #endif
3619 int
3620 sys_renameat(struct thread *td, struct renameat_args *uap)
3621 {
3622 
3623 	return (kern_renameat(td, uap->oldfd, uap->old, uap->newfd, uap->new,
3624 	    UIO_USERSPACE));
3625 }
3626 
3627 int
3628 kern_rename(struct thread *td, char *from, char *to, enum uio_seg pathseg)
3629 {
3630 
3631 	return (kern_renameat(td, AT_FDCWD, from, AT_FDCWD, to, pathseg));
3632 }
3633 
3634 int
3635 kern_renameat(struct thread *td, int oldfd, char *old, int newfd, char *new,
3636     enum uio_seg pathseg)
3637 {
3638 	struct mount *mp = NULL;
3639 	struct vnode *tvp, *fvp, *tdvp;
3640 	struct nameidata fromnd, tond;
3641 	int tvfslocked;
3642 	int fvfslocked;
3643 	int error;
3644 
3645 	bwillwrite();
3646 #ifdef MAC
3647 	NDINIT_ATRIGHTS(&fromnd, DELETE, LOCKPARENT | LOCKLEAF | SAVESTART |
3648 	    MPSAFE | AUDITVNODE1, pathseg, old, oldfd, CAP_DELETE, td);
3649 #else
3650 	NDINIT_ATRIGHTS(&fromnd, DELETE, WANTPARENT | SAVESTART | MPSAFE |
3651 	    AUDITVNODE1, pathseg, old, oldfd, CAP_DELETE, td);
3652 #endif
3653 
3654 	if ((error = namei(&fromnd)) != 0)
3655 		return (error);
3656 	fvfslocked = NDHASGIANT(&fromnd);
3657 	tvfslocked = 0;
3658 #ifdef MAC
3659 	error = mac_vnode_check_rename_from(td->td_ucred, fromnd.ni_dvp,
3660 	    fromnd.ni_vp, &fromnd.ni_cnd);
3661 	VOP_UNLOCK(fromnd.ni_dvp, 0);
3662 	if (fromnd.ni_dvp != fromnd.ni_vp)
3663 		VOP_UNLOCK(fromnd.ni_vp, 0);
3664 #endif
3665 	fvp = fromnd.ni_vp;
3666 	if (error == 0)
3667 		error = vn_start_write(fvp, &mp, V_WAIT | PCATCH);
3668 	if (error != 0) {
3669 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3670 		vrele(fromnd.ni_dvp);
3671 		vrele(fvp);
3672 		goto out1;
3673 	}
3674 	NDINIT_ATRIGHTS(&tond, RENAME, LOCKPARENT | LOCKLEAF | NOCACHE |
3675 	    SAVESTART | MPSAFE | AUDITVNODE2, pathseg, new, newfd, CAP_CREATE,
3676 	    td);
3677 	if (fromnd.ni_vp->v_type == VDIR)
3678 		tond.ni_cnd.cn_flags |= WILLBEDIR;
3679 	if ((error = namei(&tond)) != 0) {
3680 		/* Translate error code for rename("dir1", "dir2/."). */
3681 		if (error == EISDIR && fvp->v_type == VDIR)
3682 			error = EINVAL;
3683 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3684 		vrele(fromnd.ni_dvp);
3685 		vrele(fvp);
3686 		vn_finished_write(mp);
3687 		goto out1;
3688 	}
3689 	tvfslocked = NDHASGIANT(&tond);
3690 	tdvp = tond.ni_dvp;
3691 	tvp = tond.ni_vp;
3692 	if (tvp != NULL) {
3693 		if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
3694 			error = ENOTDIR;
3695 			goto out;
3696 		} else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
3697 			error = EISDIR;
3698 			goto out;
3699 		}
3700 	}
3701 	if (fvp == tdvp) {
3702 		error = EINVAL;
3703 		goto out;
3704 	}
3705 	/*
3706 	 * If the source is the same as the destination (that is, if they
3707 	 * are links to the same vnode), then there is nothing to do.
3708 	 */
3709 	if (fvp == tvp)
3710 		error = -1;
3711 #ifdef MAC
3712 	else
3713 		error = mac_vnode_check_rename_to(td->td_ucred, tdvp,
3714 		    tond.ni_vp, fromnd.ni_dvp == tdvp, &tond.ni_cnd);
3715 #endif
3716 out:
3717 	if (!error) {
3718 		error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
3719 				   tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
3720 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3721 		NDFREE(&tond, NDF_ONLY_PNBUF);
3722 	} else {
3723 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3724 		NDFREE(&tond, NDF_ONLY_PNBUF);
3725 		if (tvp)
3726 			vput(tvp);
3727 		if (tdvp == tvp)
3728 			vrele(tdvp);
3729 		else
3730 			vput(tdvp);
3731 		vrele(fromnd.ni_dvp);
3732 		vrele(fvp);
3733 	}
3734 	vrele(tond.ni_startdir);
3735 	vn_finished_write(mp);
3736 out1:
3737 	if (fromnd.ni_startdir)
3738 		vrele(fromnd.ni_startdir);
3739 	VFS_UNLOCK_GIANT(fvfslocked);
3740 	VFS_UNLOCK_GIANT(tvfslocked);
3741 	if (error == -1)
3742 		return (0);
3743 	return (error);
3744 }
3745 
3746 /*
3747  * Make a directory file.
3748  */
3749 #ifndef _SYS_SYSPROTO_H_
3750 struct mkdir_args {
3751 	char	*path;
3752 	int	mode;
3753 };
3754 #endif
3755 int
3756 sys_mkdir(td, uap)
3757 	struct thread *td;
3758 	register struct mkdir_args /* {
3759 		char *path;
3760 		int mode;
3761 	} */ *uap;
3762 {
3763 
3764 	return (kern_mkdir(td, uap->path, UIO_USERSPACE, uap->mode));
3765 }
3766 
3767 #ifndef _SYS_SYSPROTO_H_
3768 struct mkdirat_args {
3769 	int	fd;
3770 	char	*path;
3771 	mode_t	mode;
3772 };
3773 #endif
3774 int
3775 sys_mkdirat(struct thread *td, struct mkdirat_args *uap)
3776 {
3777 
3778 	return (kern_mkdirat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode));
3779 }
3780 
3781 int
3782 kern_mkdir(struct thread *td, char *path, enum uio_seg segflg, int mode)
3783 {
3784 
3785 	return (kern_mkdirat(td, AT_FDCWD, path, segflg, mode));
3786 }
3787 
3788 int
3789 kern_mkdirat(struct thread *td, int fd, char *path, enum uio_seg segflg,
3790     int mode)
3791 {
3792 	struct mount *mp;
3793 	struct vnode *vp;
3794 	struct vattr vattr;
3795 	int error;
3796 	struct nameidata nd;
3797 	int vfslocked;
3798 
3799 	AUDIT_ARG_MODE(mode);
3800 restart:
3801 	bwillwrite();
3802 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | MPSAFE |
3803 	    AUDITVNODE1, segflg, path, fd, CAP_MKDIR, td);
3804 	nd.ni_cnd.cn_flags |= WILLBEDIR;
3805 	if ((error = namei(&nd)) != 0)
3806 		return (error);
3807 	vfslocked = NDHASGIANT(&nd);
3808 	vp = nd.ni_vp;
3809 	if (vp != NULL) {
3810 		NDFREE(&nd, NDF_ONLY_PNBUF);
3811 		/*
3812 		 * XXX namei called with LOCKPARENT but not LOCKLEAF has
3813 		 * the strange behaviour of leaving the vnode unlocked
3814 		 * if the target is the same vnode as the parent.
3815 		 */
3816 		if (vp == nd.ni_dvp)
3817 			vrele(nd.ni_dvp);
3818 		else
3819 			vput(nd.ni_dvp);
3820 		vrele(vp);
3821 		VFS_UNLOCK_GIANT(vfslocked);
3822 		return (EEXIST);
3823 	}
3824 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3825 		NDFREE(&nd, NDF_ONLY_PNBUF);
3826 		vput(nd.ni_dvp);
3827 		VFS_UNLOCK_GIANT(vfslocked);
3828 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3829 			return (error);
3830 		goto restart;
3831 	}
3832 	VATTR_NULL(&vattr);
3833 	vattr.va_type = VDIR;
3834 	vattr.va_mode = (mode & ACCESSPERMS) &~ td->td_proc->p_fd->fd_cmask;
3835 #ifdef MAC
3836 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
3837 	    &vattr);
3838 	if (error)
3839 		goto out;
3840 #endif
3841 	error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
3842 #ifdef MAC
3843 out:
3844 #endif
3845 	NDFREE(&nd, NDF_ONLY_PNBUF);
3846 	vput(nd.ni_dvp);
3847 	if (!error)
3848 		vput(nd.ni_vp);
3849 	vn_finished_write(mp);
3850 	VFS_UNLOCK_GIANT(vfslocked);
3851 	return (error);
3852 }
3853 
3854 /*
3855  * Remove a directory file.
3856  */
3857 #ifndef _SYS_SYSPROTO_H_
3858 struct rmdir_args {
3859 	char	*path;
3860 };
3861 #endif
3862 int
3863 sys_rmdir(td, uap)
3864 	struct thread *td;
3865 	struct rmdir_args /* {
3866 		char *path;
3867 	} */ *uap;
3868 {
3869 
3870 	return (kern_rmdir(td, uap->path, UIO_USERSPACE));
3871 }
3872 
3873 int
3874 kern_rmdir(struct thread *td, char *path, enum uio_seg pathseg)
3875 {
3876 
3877 	return (kern_rmdirat(td, AT_FDCWD, path, pathseg));
3878 }
3879 
3880 int
3881 kern_rmdirat(struct thread *td, int fd, char *path, enum uio_seg pathseg)
3882 {
3883 	struct mount *mp;
3884 	struct vnode *vp;
3885 	int error;
3886 	struct nameidata nd;
3887 	int vfslocked;
3888 
3889 restart:
3890 	bwillwrite();
3891 	NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | MPSAFE |
3892 	    AUDITVNODE1, pathseg, path, fd, CAP_RMDIR, td);
3893 	if ((error = namei(&nd)) != 0)
3894 		return (error);
3895 	vfslocked = NDHASGIANT(&nd);
3896 	vp = nd.ni_vp;
3897 	if (vp->v_type != VDIR) {
3898 		error = ENOTDIR;
3899 		goto out;
3900 	}
3901 	/*
3902 	 * No rmdir "." please.
3903 	 */
3904 	if (nd.ni_dvp == vp) {
3905 		error = EINVAL;
3906 		goto out;
3907 	}
3908 	/*
3909 	 * The root of a mounted filesystem cannot be deleted.
3910 	 */
3911 	if (vp->v_vflag & VV_ROOT) {
3912 		error = EBUSY;
3913 		goto out;
3914 	}
3915 #ifdef MAC
3916 	error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
3917 	    &nd.ni_cnd);
3918 	if (error)
3919 		goto out;
3920 #endif
3921 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3922 		NDFREE(&nd, NDF_ONLY_PNBUF);
3923 		vput(vp);
3924 		if (nd.ni_dvp == vp)
3925 			vrele(nd.ni_dvp);
3926 		else
3927 			vput(nd.ni_dvp);
3928 		VFS_UNLOCK_GIANT(vfslocked);
3929 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3930 			return (error);
3931 		goto restart;
3932 	}
3933 	error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
3934 	vn_finished_write(mp);
3935 out:
3936 	NDFREE(&nd, NDF_ONLY_PNBUF);
3937 	vput(vp);
3938 	if (nd.ni_dvp == vp)
3939 		vrele(nd.ni_dvp);
3940 	else
3941 		vput(nd.ni_dvp);
3942 	VFS_UNLOCK_GIANT(vfslocked);
3943 	return (error);
3944 }
3945 
3946 #ifdef COMPAT_43
3947 /*
3948  * Read a block of directory entries in a filesystem independent format.
3949  */
3950 #ifndef _SYS_SYSPROTO_H_
3951 struct ogetdirentries_args {
3952 	int	fd;
3953 	char	*buf;
3954 	u_int	count;
3955 	long	*basep;
3956 };
3957 #endif
3958 int
3959 ogetdirentries(struct thread *td, struct ogetdirentries_args *uap)
3960 {
3961 	long loff;
3962 	int error;
3963 
3964 	error = kern_ogetdirentries(td, uap, &loff);
3965 	if (error == 0)
3966 		error = copyout(&loff, uap->basep, sizeof(long));
3967 	return (error);
3968 }
3969 
3970 int
3971 kern_ogetdirentries(struct thread *td, struct ogetdirentries_args *uap,
3972     long *ploff)
3973 {
3974 	struct vnode *vp;
3975 	struct file *fp;
3976 	struct uio auio, kuio;
3977 	struct iovec aiov, kiov;
3978 	struct dirent *dp, *edp;
3979 	caddr_t dirbuf;
3980 	int error, eofflag, readcnt, vfslocked;
3981 	long loff;
3982 
3983 	/* XXX arbitrary sanity limit on `count'. */
3984 	if (uap->count > 64 * 1024)
3985 		return (EINVAL);
3986 	if ((error = getvnode(td->td_proc->p_fd, uap->fd, CAP_READ,
3987 	    &fp)) != 0)
3988 		return (error);
3989 	if ((fp->f_flag & FREAD) == 0) {
3990 		fdrop(fp, td);
3991 		return (EBADF);
3992 	}
3993 	vp = fp->f_vnode;
3994 unionread:
3995 	vfslocked = VFS_LOCK_GIANT(vp->v_mount);
3996 	if (vp->v_type != VDIR) {
3997 		VFS_UNLOCK_GIANT(vfslocked);
3998 		fdrop(fp, td);
3999 		return (EINVAL);
4000 	}
4001 	aiov.iov_base = uap->buf;
4002 	aiov.iov_len = uap->count;
4003 	auio.uio_iov = &aiov;
4004 	auio.uio_iovcnt = 1;
4005 	auio.uio_rw = UIO_READ;
4006 	auio.uio_segflg = UIO_USERSPACE;
4007 	auio.uio_td = td;
4008 	auio.uio_resid = uap->count;
4009 	vn_lock(vp, LK_SHARED | LK_RETRY);
4010 	loff = auio.uio_offset = fp->f_offset;
4011 #ifdef MAC
4012 	error = mac_vnode_check_readdir(td->td_ucred, vp);
4013 	if (error) {
4014 		VOP_UNLOCK(vp, 0);
4015 		VFS_UNLOCK_GIANT(vfslocked);
4016 		fdrop(fp, td);
4017 		return (error);
4018 	}
4019 #endif
4020 #	if (BYTE_ORDER != LITTLE_ENDIAN)
4021 		if (vp->v_mount->mnt_maxsymlinklen <= 0) {
4022 			error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag,
4023 			    NULL, NULL);
4024 			fp->f_offset = auio.uio_offset;
4025 		} else
4026 #	endif
4027 	{
4028 		kuio = auio;
4029 		kuio.uio_iov = &kiov;
4030 		kuio.uio_segflg = UIO_SYSSPACE;
4031 		kiov.iov_len = uap->count;
4032 		dirbuf = malloc(uap->count, M_TEMP, M_WAITOK);
4033 		kiov.iov_base = dirbuf;
4034 		error = VOP_READDIR(vp, &kuio, fp->f_cred, &eofflag,
4035 			    NULL, NULL);
4036 		fp->f_offset = kuio.uio_offset;
4037 		if (error == 0) {
4038 			readcnt = uap->count - kuio.uio_resid;
4039 			edp = (struct dirent *)&dirbuf[readcnt];
4040 			for (dp = (struct dirent *)dirbuf; dp < edp; ) {
4041 #				if (BYTE_ORDER == LITTLE_ENDIAN)
4042 					/*
4043 					 * The expected low byte of
4044 					 * dp->d_namlen is our dp->d_type.
4045 					 * The high MBZ byte of dp->d_namlen
4046 					 * is our dp->d_namlen.
4047 					 */
4048 					dp->d_type = dp->d_namlen;
4049 					dp->d_namlen = 0;
4050 #				else
4051 					/*
4052 					 * The dp->d_type is the high byte
4053 					 * of the expected dp->d_namlen,
4054 					 * so must be zero'ed.
4055 					 */
4056 					dp->d_type = 0;
4057 #				endif
4058 				if (dp->d_reclen > 0) {
4059 					dp = (struct dirent *)
4060 					    ((char *)dp + dp->d_reclen);
4061 				} else {
4062 					error = EIO;
4063 					break;
4064 				}
4065 			}
4066 			if (dp >= edp)
4067 				error = uiomove(dirbuf, readcnt, &auio);
4068 		}
4069 		free(dirbuf, M_TEMP);
4070 	}
4071 	if (error) {
4072 		VOP_UNLOCK(vp, 0);
4073 		VFS_UNLOCK_GIANT(vfslocked);
4074 		fdrop(fp, td);
4075 		return (error);
4076 	}
4077 	if (uap->count == auio.uio_resid &&
4078 	    (vp->v_vflag & VV_ROOT) &&
4079 	    (vp->v_mount->mnt_flag & MNT_UNION)) {
4080 		struct vnode *tvp = vp;
4081 		vp = vp->v_mount->mnt_vnodecovered;
4082 		VREF(vp);
4083 		fp->f_vnode = vp;
4084 		fp->f_data = vp;
4085 		fp->f_offset = 0;
4086 		vput(tvp);
4087 		VFS_UNLOCK_GIANT(vfslocked);
4088 		goto unionread;
4089 	}
4090 	VOP_UNLOCK(vp, 0);
4091 	VFS_UNLOCK_GIANT(vfslocked);
4092 	fdrop(fp, td);
4093 	td->td_retval[0] = uap->count - auio.uio_resid;
4094 	if (error == 0)
4095 		*ploff = loff;
4096 	return (error);
4097 }
4098 #endif /* COMPAT_43 */
4099 
4100 /*
4101  * Read a block of directory entries in a filesystem independent format.
4102  */
4103 #ifndef _SYS_SYSPROTO_H_
4104 struct getdirentries_args {
4105 	int	fd;
4106 	char	*buf;
4107 	u_int	count;
4108 	long	*basep;
4109 };
4110 #endif
4111 int
4112 sys_getdirentries(td, uap)
4113 	struct thread *td;
4114 	register struct getdirentries_args /* {
4115 		int fd;
4116 		char *buf;
4117 		u_int count;
4118 		long *basep;
4119 	} */ *uap;
4120 {
4121 	long base;
4122 	int error;
4123 
4124 	error = kern_getdirentries(td, uap->fd, uap->buf, uap->count, &base);
4125 	if (error)
4126 		return (error);
4127 	if (uap->basep != NULL)
4128 		error = copyout(&base, uap->basep, sizeof(long));
4129 	return (error);
4130 }
4131 
4132 int
4133 kern_getdirentries(struct thread *td, int fd, char *buf, u_int count,
4134     long *basep)
4135 {
4136 	struct vnode *vp;
4137 	struct file *fp;
4138 	struct uio auio;
4139 	struct iovec aiov;
4140 	int vfslocked;
4141 	long loff;
4142 	int error, eofflag;
4143 
4144 	AUDIT_ARG_FD(fd);
4145 	if (count > INT_MAX)
4146 		return (EINVAL);
4147 	if ((error = getvnode(td->td_proc->p_fd, fd, CAP_READ | CAP_SEEK,
4148 	    &fp)) != 0)
4149 		return (error);
4150 	if ((fp->f_flag & FREAD) == 0) {
4151 		fdrop(fp, td);
4152 		return (EBADF);
4153 	}
4154 	vp = fp->f_vnode;
4155 unionread:
4156 	vfslocked = VFS_LOCK_GIANT(vp->v_mount);
4157 	if (vp->v_type != VDIR) {
4158 		VFS_UNLOCK_GIANT(vfslocked);
4159 		error = EINVAL;
4160 		goto fail;
4161 	}
4162 	aiov.iov_base = buf;
4163 	aiov.iov_len = count;
4164 	auio.uio_iov = &aiov;
4165 	auio.uio_iovcnt = 1;
4166 	auio.uio_rw = UIO_READ;
4167 	auio.uio_segflg = UIO_USERSPACE;
4168 	auio.uio_td = td;
4169 	auio.uio_resid = count;
4170 	vn_lock(vp, LK_SHARED | LK_RETRY);
4171 	AUDIT_ARG_VNODE1(vp);
4172 	loff = auio.uio_offset = fp->f_offset;
4173 #ifdef MAC
4174 	error = mac_vnode_check_readdir(td->td_ucred, vp);
4175 	if (error == 0)
4176 #endif
4177 		error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, NULL,
4178 		    NULL);
4179 	fp->f_offset = auio.uio_offset;
4180 	if (error) {
4181 		VOP_UNLOCK(vp, 0);
4182 		VFS_UNLOCK_GIANT(vfslocked);
4183 		goto fail;
4184 	}
4185 	if (count == auio.uio_resid &&
4186 	    (vp->v_vflag & VV_ROOT) &&
4187 	    (vp->v_mount->mnt_flag & MNT_UNION)) {
4188 		struct vnode *tvp = vp;
4189 		vp = vp->v_mount->mnt_vnodecovered;
4190 		VREF(vp);
4191 		fp->f_vnode = vp;
4192 		fp->f_data = vp;
4193 		fp->f_offset = 0;
4194 		vput(tvp);
4195 		VFS_UNLOCK_GIANT(vfslocked);
4196 		goto unionread;
4197 	}
4198 	VOP_UNLOCK(vp, 0);
4199 	VFS_UNLOCK_GIANT(vfslocked);
4200 	*basep = loff;
4201 	td->td_retval[0] = count - auio.uio_resid;
4202 fail:
4203 	fdrop(fp, td);
4204 	return (error);
4205 }
4206 
4207 #ifndef _SYS_SYSPROTO_H_
4208 struct getdents_args {
4209 	int fd;
4210 	char *buf;
4211 	size_t count;
4212 };
4213 #endif
4214 int
4215 sys_getdents(td, uap)
4216 	struct thread *td;
4217 	register struct getdents_args /* {
4218 		int fd;
4219 		char *buf;
4220 		u_int count;
4221 	} */ *uap;
4222 {
4223 	struct getdirentries_args ap;
4224 	ap.fd = uap->fd;
4225 	ap.buf = uap->buf;
4226 	ap.count = uap->count;
4227 	ap.basep = NULL;
4228 	return (sys_getdirentries(td, &ap));
4229 }
4230 
4231 /*
4232  * Set the mode mask for creation of filesystem nodes.
4233  */
4234 #ifndef _SYS_SYSPROTO_H_
4235 struct umask_args {
4236 	int	newmask;
4237 };
4238 #endif
4239 int
4240 sys_umask(td, uap)
4241 	struct thread *td;
4242 	struct umask_args /* {
4243 		int newmask;
4244 	} */ *uap;
4245 {
4246 	register struct filedesc *fdp;
4247 
4248 	FILEDESC_XLOCK(td->td_proc->p_fd);
4249 	fdp = td->td_proc->p_fd;
4250 	td->td_retval[0] = fdp->fd_cmask;
4251 	fdp->fd_cmask = uap->newmask & ALLPERMS;
4252 	FILEDESC_XUNLOCK(td->td_proc->p_fd);
4253 	return (0);
4254 }
4255 
4256 /*
4257  * Void all references to file by ripping underlying filesystem away from
4258  * vnode.
4259  */
4260 #ifndef _SYS_SYSPROTO_H_
4261 struct revoke_args {
4262 	char	*path;
4263 };
4264 #endif
4265 int
4266 sys_revoke(td, uap)
4267 	struct thread *td;
4268 	register struct revoke_args /* {
4269 		char *path;
4270 	} */ *uap;
4271 {
4272 	struct vnode *vp;
4273 	struct vattr vattr;
4274 	int error;
4275 	struct nameidata nd;
4276 	int vfslocked;
4277 
4278 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | MPSAFE | AUDITVNODE1,
4279 	    UIO_USERSPACE, uap->path, td);
4280 	if ((error = namei(&nd)) != 0)
4281 		return (error);
4282 	vfslocked = NDHASGIANT(&nd);
4283 	vp = nd.ni_vp;
4284 	NDFREE(&nd, NDF_ONLY_PNBUF);
4285 	if (vp->v_type != VCHR || vp->v_rdev == NULL) {
4286 		error = EINVAL;
4287 		goto out;
4288 	}
4289 #ifdef MAC
4290 	error = mac_vnode_check_revoke(td->td_ucred, vp);
4291 	if (error)
4292 		goto out;
4293 #endif
4294 	error = VOP_GETATTR(vp, &vattr, td->td_ucred);
4295 	if (error)
4296 		goto out;
4297 	if (td->td_ucred->cr_uid != vattr.va_uid) {
4298 		error = priv_check(td, PRIV_VFS_ADMIN);
4299 		if (error)
4300 			goto out;
4301 	}
4302 	if (vcount(vp) > 1)
4303 		VOP_REVOKE(vp, REVOKEALL);
4304 out:
4305 	vput(vp);
4306 	VFS_UNLOCK_GIANT(vfslocked);
4307 	return (error);
4308 }
4309 
4310 /*
4311  * Convert a user file descriptor to a kernel file entry and check that, if it
4312  * is a capability, the correct rights are present. A reference on the file
4313  * entry is held upon returning.
4314  */
4315 int
4316 getvnode(struct filedesc *fdp, int fd, cap_rights_t rights,
4317     struct file **fpp)
4318 {
4319 	struct file *fp;
4320 #ifdef CAPABILITIES
4321 	struct file *fp_fromcap;
4322 #endif
4323 	int error;
4324 
4325 	error = 0;
4326 	fp = NULL;
4327 	if ((fdp == NULL) || (fp = fget_unlocked(fdp, fd)) == NULL)
4328 		return (EBADF);
4329 #ifdef CAPABILITIES
4330 	/*
4331 	 * If the file descriptor is for a capability, test rights and use the
4332 	 * file descriptor referenced by the capability.
4333 	 */
4334 	error = cap_funwrap(fp, rights, &fp_fromcap);
4335 	if (error) {
4336 		fdrop(fp, curthread);
4337 		return (error);
4338 	}
4339 	if (fp != fp_fromcap) {
4340 		fhold(fp_fromcap);
4341 		fdrop(fp, curthread);
4342 		fp = fp_fromcap;
4343 	}
4344 #endif /* CAPABILITIES */
4345 	if (fp->f_vnode == NULL) {
4346 		fdrop(fp, curthread);
4347 		return (EINVAL);
4348 	}
4349 	*fpp = fp;
4350 	return (0);
4351 }
4352 
4353 
4354 /*
4355  * Get an (NFS) file handle.
4356  */
4357 #ifndef _SYS_SYSPROTO_H_
4358 struct lgetfh_args {
4359 	char	*fname;
4360 	fhandle_t *fhp;
4361 };
4362 #endif
4363 int
4364 sys_lgetfh(td, uap)
4365 	struct thread *td;
4366 	register struct lgetfh_args *uap;
4367 {
4368 	struct nameidata nd;
4369 	fhandle_t fh;
4370 	register struct vnode *vp;
4371 	int vfslocked;
4372 	int error;
4373 
4374 	error = priv_check(td, PRIV_VFS_GETFH);
4375 	if (error)
4376 		return (error);
4377 	NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | MPSAFE | AUDITVNODE1,
4378 	    UIO_USERSPACE, uap->fname, td);
4379 	error = namei(&nd);
4380 	if (error)
4381 		return (error);
4382 	vfslocked = NDHASGIANT(&nd);
4383 	NDFREE(&nd, NDF_ONLY_PNBUF);
4384 	vp = nd.ni_vp;
4385 	bzero(&fh, sizeof(fh));
4386 	fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
4387 	error = VOP_VPTOFH(vp, &fh.fh_fid);
4388 	vput(vp);
4389 	VFS_UNLOCK_GIANT(vfslocked);
4390 	if (error)
4391 		return (error);
4392 	error = copyout(&fh, uap->fhp, sizeof (fh));
4393 	return (error);
4394 }
4395 
4396 #ifndef _SYS_SYSPROTO_H_
4397 struct getfh_args {
4398 	char	*fname;
4399 	fhandle_t *fhp;
4400 };
4401 #endif
4402 int
4403 sys_getfh(td, uap)
4404 	struct thread *td;
4405 	register struct getfh_args *uap;
4406 {
4407 	struct nameidata nd;
4408 	fhandle_t fh;
4409 	register struct vnode *vp;
4410 	int vfslocked;
4411 	int error;
4412 
4413 	error = priv_check(td, PRIV_VFS_GETFH);
4414 	if (error)
4415 		return (error);
4416 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | MPSAFE | AUDITVNODE1,
4417 	    UIO_USERSPACE, uap->fname, td);
4418 	error = namei(&nd);
4419 	if (error)
4420 		return (error);
4421 	vfslocked = NDHASGIANT(&nd);
4422 	NDFREE(&nd, NDF_ONLY_PNBUF);
4423 	vp = nd.ni_vp;
4424 	bzero(&fh, sizeof(fh));
4425 	fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
4426 	error = VOP_VPTOFH(vp, &fh.fh_fid);
4427 	vput(vp);
4428 	VFS_UNLOCK_GIANT(vfslocked);
4429 	if (error)
4430 		return (error);
4431 	error = copyout(&fh, uap->fhp, sizeof (fh));
4432 	return (error);
4433 }
4434 
4435 /*
4436  * syscall for the rpc.lockd to use to translate a NFS file handle into an
4437  * open descriptor.
4438  *
4439  * warning: do not remove the priv_check() call or this becomes one giant
4440  * security hole.
4441  */
4442 #ifndef _SYS_SYSPROTO_H_
4443 struct fhopen_args {
4444 	const struct fhandle *u_fhp;
4445 	int flags;
4446 };
4447 #endif
4448 int
4449 sys_fhopen(td, uap)
4450 	struct thread *td;
4451 	struct fhopen_args /* {
4452 		const struct fhandle *u_fhp;
4453 		int flags;
4454 	} */ *uap;
4455 {
4456 	struct proc *p = td->td_proc;
4457 	struct mount *mp;
4458 	struct vnode *vp;
4459 	struct fhandle fhp;
4460 	struct vattr vat;
4461 	struct vattr *vap = &vat;
4462 	struct flock lf;
4463 	struct file *fp;
4464 	register struct filedesc *fdp = p->p_fd;
4465 	int fmode, error, type;
4466 	accmode_t accmode;
4467 	struct file *nfp;
4468 	int vfslocked;
4469 	int indx;
4470 
4471 	error = priv_check(td, PRIV_VFS_FHOPEN);
4472 	if (error)
4473 		return (error);
4474 	fmode = FFLAGS(uap->flags);
4475 	/* why not allow a non-read/write open for our lockd? */
4476 	if (((fmode & (FREAD | FWRITE)) == 0) || (fmode & O_CREAT))
4477 		return (EINVAL);
4478 	error = copyin(uap->u_fhp, &fhp, sizeof(fhp));
4479 	if (error)
4480 		return(error);
4481 	/* find the mount point */
4482 	mp = vfs_busyfs(&fhp.fh_fsid);
4483 	if (mp == NULL)
4484 		return (ESTALE);
4485 	vfslocked = VFS_LOCK_GIANT(mp);
4486 	/* now give me my vnode, it gets returned to me locked */
4487 	error = VFS_FHTOVP(mp, &fhp.fh_fid, LK_EXCLUSIVE, &vp);
4488 	vfs_unbusy(mp);
4489 	if (error)
4490 		goto out;
4491 	/*
4492 	 * from now on we have to make sure not
4493 	 * to forget about the vnode
4494 	 * any error that causes an abort must vput(vp)
4495 	 * just set error = err and 'goto bad;'.
4496 	 */
4497 
4498 	/*
4499 	 * from vn_open
4500 	 */
4501 	if (vp->v_type == VLNK) {
4502 		error = EMLINK;
4503 		goto bad;
4504 	}
4505 	if (vp->v_type == VSOCK) {
4506 		error = EOPNOTSUPP;
4507 		goto bad;
4508 	}
4509 	if (vp->v_type != VDIR && fmode & O_DIRECTORY) {
4510 		error = ENOTDIR;
4511 		goto bad;
4512 	}
4513 	accmode = 0;
4514 	if (fmode & (FWRITE | O_TRUNC)) {
4515 		if (vp->v_type == VDIR) {
4516 			error = EISDIR;
4517 			goto bad;
4518 		}
4519 		error = vn_writechk(vp);
4520 		if (error)
4521 			goto bad;
4522 		accmode |= VWRITE;
4523 	}
4524 	if (fmode & FREAD)
4525 		accmode |= VREAD;
4526 	if ((fmode & O_APPEND) && (fmode & FWRITE))
4527 		accmode |= VAPPEND;
4528 #ifdef MAC
4529 	error = mac_vnode_check_open(td->td_ucred, vp, accmode);
4530 	if (error)
4531 		goto bad;
4532 #endif
4533 	if (accmode) {
4534 		error = VOP_ACCESS(vp, accmode, td->td_ucred, td);
4535 		if (error)
4536 			goto bad;
4537 	}
4538 	if (fmode & O_TRUNC) {
4539 		vfs_ref(mp);
4540 		VOP_UNLOCK(vp, 0);				/* XXX */
4541 		if ((error = vn_start_write(NULL, &mp, V_WAIT | PCATCH)) != 0) {
4542 			vrele(vp);
4543 			vfs_rel(mp);
4544 			goto out;
4545 		}
4546 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);	/* XXX */
4547 		vfs_rel(mp);
4548 #ifdef MAC
4549 		/*
4550 		 * We don't yet have fp->f_cred, so use td->td_ucred, which
4551 		 * should be right.
4552 		 */
4553 		error = mac_vnode_check_write(td->td_ucred, td->td_ucred, vp);
4554 		if (error == 0) {
4555 #endif
4556 			VATTR_NULL(vap);
4557 			vap->va_size = 0;
4558 			error = VOP_SETATTR(vp, vap, td->td_ucred);
4559 #ifdef MAC
4560 		}
4561 #endif
4562 		vn_finished_write(mp);
4563 		if (error)
4564 			goto bad;
4565 	}
4566 	error = VOP_OPEN(vp, fmode, td->td_ucred, td, NULL);
4567 	if (error)
4568 		goto bad;
4569 
4570 	if (fmode & FWRITE)
4571 		vp->v_writecount++;
4572 
4573 	/*
4574 	 * end of vn_open code
4575 	 */
4576 
4577 	if ((error = falloc(td, &nfp, &indx, fmode)) != 0) {
4578 		if (fmode & FWRITE)
4579 			vp->v_writecount--;
4580 		goto bad;
4581 	}
4582 	/* An extra reference on `nfp' has been held for us by falloc(). */
4583 	fp = nfp;
4584 	nfp->f_vnode = vp;
4585 	finit(nfp, fmode & FMASK, DTYPE_VNODE, vp, &vnops);
4586 	if (fmode & (O_EXLOCK | O_SHLOCK)) {
4587 		lf.l_whence = SEEK_SET;
4588 		lf.l_start = 0;
4589 		lf.l_len = 0;
4590 		if (fmode & O_EXLOCK)
4591 			lf.l_type = F_WRLCK;
4592 		else
4593 			lf.l_type = F_RDLCK;
4594 		type = F_FLOCK;
4595 		if ((fmode & FNONBLOCK) == 0)
4596 			type |= F_WAIT;
4597 		VOP_UNLOCK(vp, 0);
4598 		if ((error = VOP_ADVLOCK(vp, (caddr_t)fp, F_SETLK, &lf,
4599 			    type)) != 0) {
4600 			/*
4601 			 * The lock request failed.  Normally close the
4602 			 * descriptor but handle the case where someone might
4603 			 * have dup()d or close()d it when we weren't looking.
4604 			 */
4605 			fdclose(fdp, fp, indx, td);
4606 
4607 			/*
4608 			 * release our private reference
4609 			 */
4610 			fdrop(fp, td);
4611 			goto out;
4612 		}
4613 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
4614 		atomic_set_int(&fp->f_flag, FHASLOCK);
4615 	}
4616 
4617 	VOP_UNLOCK(vp, 0);
4618 	fdrop(fp, td);
4619 	VFS_UNLOCK_GIANT(vfslocked);
4620 	td->td_retval[0] = indx;
4621 	return (0);
4622 
4623 bad:
4624 	vput(vp);
4625 out:
4626 	VFS_UNLOCK_GIANT(vfslocked);
4627 	return (error);
4628 }
4629 
4630 /*
4631  * Stat an (NFS) file handle.
4632  */
4633 #ifndef _SYS_SYSPROTO_H_
4634 struct fhstat_args {
4635 	struct fhandle *u_fhp;
4636 	struct stat *sb;
4637 };
4638 #endif
4639 int
4640 sys_fhstat(td, uap)
4641 	struct thread *td;
4642 	register struct fhstat_args /* {
4643 		struct fhandle *u_fhp;
4644 		struct stat *sb;
4645 	} */ *uap;
4646 {
4647 	struct stat sb;
4648 	fhandle_t fh;
4649 	struct mount *mp;
4650 	struct vnode *vp;
4651 	int vfslocked;
4652 	int error;
4653 
4654 	error = priv_check(td, PRIV_VFS_FHSTAT);
4655 	if (error)
4656 		return (error);
4657 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
4658 	if (error)
4659 		return (error);
4660 	if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4661 		return (ESTALE);
4662 	vfslocked = VFS_LOCK_GIANT(mp);
4663 	error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4664 	vfs_unbusy(mp);
4665 	if (error) {
4666 		VFS_UNLOCK_GIANT(vfslocked);
4667 		return (error);
4668 	}
4669 	error = vn_stat(vp, &sb, td->td_ucred, NOCRED, td);
4670 	vput(vp);
4671 	VFS_UNLOCK_GIANT(vfslocked);
4672 	if (error)
4673 		return (error);
4674 	error = copyout(&sb, uap->sb, sizeof(sb));
4675 	return (error);
4676 }
4677 
4678 /*
4679  * Implement fstatfs() for (NFS) file handles.
4680  */
4681 #ifndef _SYS_SYSPROTO_H_
4682 struct fhstatfs_args {
4683 	struct fhandle *u_fhp;
4684 	struct statfs *buf;
4685 };
4686 #endif
4687 int
4688 sys_fhstatfs(td, uap)
4689 	struct thread *td;
4690 	struct fhstatfs_args /* {
4691 		struct fhandle *u_fhp;
4692 		struct statfs *buf;
4693 	} */ *uap;
4694 {
4695 	struct statfs sf;
4696 	fhandle_t fh;
4697 	int error;
4698 
4699 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
4700 	if (error)
4701 		return (error);
4702 	error = kern_fhstatfs(td, fh, &sf);
4703 	if (error)
4704 		return (error);
4705 	return (copyout(&sf, uap->buf, sizeof(sf)));
4706 }
4707 
4708 int
4709 kern_fhstatfs(struct thread *td, fhandle_t fh, struct statfs *buf)
4710 {
4711 	struct statfs *sp;
4712 	struct mount *mp;
4713 	struct vnode *vp;
4714 	int vfslocked;
4715 	int error;
4716 
4717 	error = priv_check(td, PRIV_VFS_FHSTATFS);
4718 	if (error)
4719 		return (error);
4720 	if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4721 		return (ESTALE);
4722 	vfslocked = VFS_LOCK_GIANT(mp);
4723 	error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4724 	if (error) {
4725 		vfs_unbusy(mp);
4726 		VFS_UNLOCK_GIANT(vfslocked);
4727 		return (error);
4728 	}
4729 	vput(vp);
4730 	error = prison_canseemount(td->td_ucred, mp);
4731 	if (error)
4732 		goto out;
4733 #ifdef MAC
4734 	error = mac_mount_check_stat(td->td_ucred, mp);
4735 	if (error)
4736 		goto out;
4737 #endif
4738 	/*
4739 	 * Set these in case the underlying filesystem fails to do so.
4740 	 */
4741 	sp = &mp->mnt_stat;
4742 	sp->f_version = STATFS_VERSION;
4743 	sp->f_namemax = NAME_MAX;
4744 	sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
4745 	error = VFS_STATFS(mp, sp);
4746 	if (error == 0)
4747 		*buf = *sp;
4748 out:
4749 	vfs_unbusy(mp);
4750 	VFS_UNLOCK_GIANT(vfslocked);
4751 	return (error);
4752 }
4753 
4754 static int
4755 kern_posix_fallocate(struct thread *td, int fd, off_t offset, off_t len)
4756 {
4757 	struct file *fp;
4758 	struct mount *mp;
4759 	struct vnode *vp;
4760 	off_t olen, ooffset;
4761 	int error, vfslocked;
4762 
4763 	fp = NULL;
4764 	vfslocked = 0;
4765 	error = fget(td, fd, CAP_WRITE, &fp);
4766 	if (error != 0)
4767 		goto out;
4768 
4769 	switch (fp->f_type) {
4770 	case DTYPE_VNODE:
4771 		break;
4772 	case DTYPE_PIPE:
4773 	case DTYPE_FIFO:
4774 		error = ESPIPE;
4775 		goto out;
4776 	default:
4777 		error = ENODEV;
4778 		goto out;
4779 	}
4780 	if ((fp->f_flag & FWRITE) == 0) {
4781 		error = EBADF;
4782 		goto out;
4783 	}
4784 	vp = fp->f_vnode;
4785 	if (vp->v_type != VREG) {
4786 		error = ENODEV;
4787 		goto out;
4788 	}
4789 	if (offset < 0 || len <= 0) {
4790 		error = EINVAL;
4791 		goto out;
4792 	}
4793 	/* Check for wrap. */
4794 	if (offset > OFF_MAX - len) {
4795 		error = EFBIG;
4796 		goto out;
4797 	}
4798 
4799 	/* Allocating blocks may take a long time, so iterate. */
4800 	for (;;) {
4801 		olen = len;
4802 		ooffset = offset;
4803 
4804 		bwillwrite();
4805 		vfslocked = VFS_LOCK_GIANT(vp->v_mount);
4806 		mp = NULL;
4807 		error = vn_start_write(vp, &mp, V_WAIT | PCATCH);
4808 		if (error != 0) {
4809 			VFS_UNLOCK_GIANT(vfslocked);
4810 			break;
4811 		}
4812 		error = vn_lock(vp, LK_EXCLUSIVE);
4813 		if (error != 0) {
4814 			vn_finished_write(mp);
4815 			VFS_UNLOCK_GIANT(vfslocked);
4816 			break;
4817 		}
4818 #ifdef MAC
4819 		error = mac_vnode_check_write(td->td_ucred, fp->f_cred, vp);
4820 		if (error == 0)
4821 #endif
4822 			error = VOP_ALLOCATE(vp, &offset, &len);
4823 		VOP_UNLOCK(vp, 0);
4824 		vn_finished_write(mp);
4825 		VFS_UNLOCK_GIANT(vfslocked);
4826 
4827 		if (olen + ooffset != offset + len) {
4828 			panic("offset + len changed from %jx/%jx to %jx/%jx",
4829 			    ooffset, olen, offset, len);
4830 		}
4831 		if (error != 0 || len == 0)
4832 			break;
4833 		KASSERT(olen > len, ("Iteration did not make progress?"));
4834 		maybe_yield();
4835 	}
4836  out:
4837 	if (fp != NULL)
4838 		fdrop(fp, td);
4839 	return (error);
4840 }
4841 
4842 int
4843 sys_posix_fallocate(struct thread *td, struct posix_fallocate_args *uap)
4844 {
4845 
4846 	return (kern_posix_fallocate(td, uap->fd, uap->offset, uap->len));
4847 }
4848