xref: /freebsd/sys/kern/vfs_syscalls.c (revision 10b59a9b4add0320d52c15ce057dd697261e7dfc)
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 static MALLOC_DEFINE(M_FADVISE, "fadvise", "posix_fadvise(2) information");
90 
91 SDT_PROVIDER_DEFINE(vfs);
92 SDT_PROBE_DEFINE(vfs, , stat, mode, mode);
93 SDT_PROBE_ARGTYPE(vfs, , stat, mode, 0, "char *");
94 SDT_PROBE_ARGTYPE(vfs, , stat, mode, 1, "int");
95 SDT_PROBE_DEFINE(vfs, , stat, reg, reg);
96 SDT_PROBE_ARGTYPE(vfs, , stat, reg, 0, "char *");
97 SDT_PROBE_ARGTYPE(vfs, , stat, reg, 1, "int");
98 
99 static int chroot_refuse_vdir_fds(struct filedesc *fdp);
100 static int getutimes(const struct timeval *, enum uio_seg, struct timespec *);
101 static int setfflags(struct thread *td, struct vnode *, int);
102 static int setutimes(struct thread *td, struct vnode *,
103     const struct timespec *, int, int);
104 static int vn_access(struct vnode *vp, int user_flags, struct ucred *cred,
105     struct thread *td);
106 
107 /*
108  * The module initialization routine for POSIX asynchronous I/O will
109  * set this to the version of AIO that it implements.  (Zero means
110  * that it is not implemented.)  This value is used here by pathconf()
111  * and in kern_descrip.c by fpathconf().
112  */
113 int async_io_version;
114 
115 #ifdef DEBUG
116 static int syncprt = 0;
117 SYSCTL_INT(_debug, OID_AUTO, syncprt, CTLFLAG_RW, &syncprt, 0, "");
118 #endif
119 
120 /*
121  * Sync each mounted filesystem.
122  */
123 #ifndef _SYS_SYSPROTO_H_
124 struct sync_args {
125 	int     dummy;
126 };
127 #endif
128 /* ARGSUSED */
129 int
130 sys_sync(td, uap)
131 	struct thread *td;
132 	struct sync_args *uap;
133 {
134 	struct mount *mp, *nmp;
135 	int vfslocked;
136 
137 	mtx_lock(&mountlist_mtx);
138 	for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
139 		if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK)) {
140 			nmp = TAILQ_NEXT(mp, mnt_list);
141 			continue;
142 		}
143 		vfslocked = VFS_LOCK_GIANT(mp);
144 		if ((mp->mnt_flag & MNT_RDONLY) == 0 &&
145 		    vn_start_write(NULL, &mp, V_NOWAIT) == 0) {
146 			MNT_ILOCK(mp);
147 			mp->mnt_noasync++;
148 			mp->mnt_kern_flag &= ~MNTK_ASYNC;
149 			MNT_IUNLOCK(mp);
150 			vfs_msync(mp, MNT_NOWAIT);
151 			VFS_SYNC(mp, MNT_NOWAIT);
152 			MNT_ILOCK(mp);
153 			mp->mnt_noasync--;
154 			if ((mp->mnt_flag & MNT_ASYNC) != 0 &&
155 			    mp->mnt_noasync == 0)
156 				mp->mnt_kern_flag |= MNTK_ASYNC;
157 			MNT_IUNLOCK(mp);
158 			vn_finished_write(mp);
159 		}
160 		VFS_UNLOCK_GIANT(vfslocked);
161 		mtx_lock(&mountlist_mtx);
162 		nmp = TAILQ_NEXT(mp, mnt_list);
163 		vfs_unbusy(mp);
164 	}
165 	mtx_unlock(&mountlist_mtx);
166 	return (0);
167 }
168 
169 /*
170  * Change filesystem quotas.
171  */
172 #ifndef _SYS_SYSPROTO_H_
173 struct quotactl_args {
174 	char *path;
175 	int cmd;
176 	int uid;
177 	caddr_t arg;
178 };
179 #endif
180 int
181 sys_quotactl(td, uap)
182 	struct thread *td;
183 	register struct quotactl_args /* {
184 		char *path;
185 		int cmd;
186 		int uid;
187 		caddr_t arg;
188 	} */ *uap;
189 {
190 	struct mount *mp;
191 	int vfslocked;
192 	int error;
193 	struct nameidata nd;
194 
195 	AUDIT_ARG_CMD(uap->cmd);
196 	AUDIT_ARG_UID(uap->uid);
197 	if (!prison_allow(td->td_ucred, PR_ALLOW_QUOTAS))
198 		return (EPERM);
199 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | MPSAFE | AUDITVNODE1,
200 	   UIO_USERSPACE, uap->path, td);
201 	if ((error = namei(&nd)) != 0)
202 		return (error);
203 	vfslocked = NDHASGIANT(&nd);
204 	NDFREE(&nd, NDF_ONLY_PNBUF);
205 	mp = nd.ni_vp->v_mount;
206 	vfs_ref(mp);
207 	vput(nd.ni_vp);
208 	error = vfs_busy(mp, 0);
209 	vfs_rel(mp);
210 	if (error) {
211 		VFS_UNLOCK_GIANT(vfslocked);
212 		return (error);
213 	}
214 	error = VFS_QUOTACTL(mp, uap->cmd, uap->uid, uap->arg);
215 	vfs_unbusy(mp);
216 	VFS_UNLOCK_GIANT(vfslocked);
217 	return (error);
218 }
219 
220 /*
221  * Used by statfs conversion routines to scale the block size up if
222  * necessary so that all of the block counts are <= 'max_size'.  Note
223  * that 'max_size' should be a bitmask, i.e. 2^n - 1 for some non-zero
224  * value of 'n'.
225  */
226 void
227 statfs_scale_blocks(struct statfs *sf, long max_size)
228 {
229 	uint64_t count;
230 	int shift;
231 
232 	KASSERT(powerof2(max_size + 1), ("%s: invalid max_size", __func__));
233 
234 	/*
235 	 * Attempt to scale the block counts to give a more accurate
236 	 * overview to userland of the ratio of free space to used
237 	 * space.  To do this, find the largest block count and compute
238 	 * a divisor that lets it fit into a signed integer <= max_size.
239 	 */
240 	if (sf->f_bavail < 0)
241 		count = -sf->f_bavail;
242 	else
243 		count = sf->f_bavail;
244 	count = MAX(sf->f_blocks, MAX(sf->f_bfree, count));
245 	if (count <= max_size)
246 		return;
247 
248 	count >>= flsl(max_size);
249 	shift = 0;
250 	while (count > 0) {
251 		shift++;
252 		count >>=1;
253 	}
254 
255 	sf->f_bsize <<= shift;
256 	sf->f_blocks >>= shift;
257 	sf->f_bfree >>= shift;
258 	sf->f_bavail >>= shift;
259 }
260 
261 /*
262  * Get filesystem statistics.
263  */
264 #ifndef _SYS_SYSPROTO_H_
265 struct statfs_args {
266 	char *path;
267 	struct statfs *buf;
268 };
269 #endif
270 int
271 sys_statfs(td, uap)
272 	struct thread *td;
273 	register struct statfs_args /* {
274 		char *path;
275 		struct statfs *buf;
276 	} */ *uap;
277 {
278 	struct statfs sf;
279 	int error;
280 
281 	error = kern_statfs(td, uap->path, UIO_USERSPACE, &sf);
282 	if (error == 0)
283 		error = copyout(&sf, uap->buf, sizeof(sf));
284 	return (error);
285 }
286 
287 int
288 kern_statfs(struct thread *td, char *path, enum uio_seg pathseg,
289     struct statfs *buf)
290 {
291 	struct mount *mp;
292 	struct statfs *sp, sb;
293 	int vfslocked;
294 	int error;
295 	struct nameidata nd;
296 
297 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | MPSAFE |
298 	    AUDITVNODE1, pathseg, path, td);
299 	error = namei(&nd);
300 	if (error)
301 		return (error);
302 	vfslocked = NDHASGIANT(&nd);
303 	mp = nd.ni_vp->v_mount;
304 	vfs_ref(mp);
305 	NDFREE(&nd, NDF_ONLY_PNBUF);
306 	vput(nd.ni_vp);
307 	error = vfs_busy(mp, 0);
308 	vfs_rel(mp);
309 	if (error) {
310 		VFS_UNLOCK_GIANT(vfslocked);
311 		return (error);
312 	}
313 #ifdef MAC
314 	error = mac_mount_check_stat(td->td_ucred, mp);
315 	if (error)
316 		goto out;
317 #endif
318 	/*
319 	 * Set these in case the underlying filesystem fails to do so.
320 	 */
321 	sp = &mp->mnt_stat;
322 	sp->f_version = STATFS_VERSION;
323 	sp->f_namemax = NAME_MAX;
324 	sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
325 	error = VFS_STATFS(mp, sp);
326 	if (error)
327 		goto out;
328 	if (priv_check(td, PRIV_VFS_GENERATION)) {
329 		bcopy(sp, &sb, sizeof(sb));
330 		sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
331 		prison_enforce_statfs(td->td_ucred, mp, &sb);
332 		sp = &sb;
333 	}
334 	*buf = *sp;
335 out:
336 	vfs_unbusy(mp);
337 	VFS_UNLOCK_GIANT(vfslocked);
338 	return (error);
339 }
340 
341 /*
342  * Get filesystem statistics.
343  */
344 #ifndef _SYS_SYSPROTO_H_
345 struct fstatfs_args {
346 	int fd;
347 	struct statfs *buf;
348 };
349 #endif
350 int
351 sys_fstatfs(td, uap)
352 	struct thread *td;
353 	register struct fstatfs_args /* {
354 		int fd;
355 		struct statfs *buf;
356 	} */ *uap;
357 {
358 	struct statfs sf;
359 	int error;
360 
361 	error = kern_fstatfs(td, uap->fd, &sf);
362 	if (error == 0)
363 		error = copyout(&sf, uap->buf, sizeof(sf));
364 	return (error);
365 }
366 
367 int
368 kern_fstatfs(struct thread *td, int fd, struct statfs *buf)
369 {
370 	struct file *fp;
371 	struct mount *mp;
372 	struct statfs *sp, sb;
373 	int vfslocked;
374 	struct vnode *vp;
375 	int error;
376 
377 	AUDIT_ARG_FD(fd);
378 	error = getvnode(td->td_proc->p_fd, fd, CAP_FSTATFS, &fp);
379 	if (error)
380 		return (error);
381 	vp = fp->f_vnode;
382 	vfslocked = VFS_LOCK_GIANT(vp->v_mount);
383 	vn_lock(vp, LK_SHARED | LK_RETRY);
384 #ifdef AUDIT
385 	AUDIT_ARG_VNODE1(vp);
386 #endif
387 	mp = vp->v_mount;
388 	if (mp)
389 		vfs_ref(mp);
390 	VOP_UNLOCK(vp, 0);
391 	fdrop(fp, td);
392 	if (mp == NULL) {
393 		error = EBADF;
394 		goto out;
395 	}
396 	error = vfs_busy(mp, 0);
397 	vfs_rel(mp);
398 	if (error) {
399 		VFS_UNLOCK_GIANT(vfslocked);
400 		return (error);
401 	}
402 #ifdef MAC
403 	error = mac_mount_check_stat(td->td_ucred, mp);
404 	if (error)
405 		goto out;
406 #endif
407 	/*
408 	 * Set these in case the underlying filesystem fails to do so.
409 	 */
410 	sp = &mp->mnt_stat;
411 	sp->f_version = STATFS_VERSION;
412 	sp->f_namemax = NAME_MAX;
413 	sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
414 	error = VFS_STATFS(mp, sp);
415 	if (error)
416 		goto out;
417 	if (priv_check(td, PRIV_VFS_GENERATION)) {
418 		bcopy(sp, &sb, sizeof(sb));
419 		sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
420 		prison_enforce_statfs(td->td_ucred, mp, &sb);
421 		sp = &sb;
422 	}
423 	*buf = *sp;
424 out:
425 	if (mp)
426 		vfs_unbusy(mp);
427 	VFS_UNLOCK_GIANT(vfslocked);
428 	return (error);
429 }
430 
431 /*
432  * Get statistics on all filesystems.
433  */
434 #ifndef _SYS_SYSPROTO_H_
435 struct getfsstat_args {
436 	struct statfs *buf;
437 	long bufsize;
438 	int flags;
439 };
440 #endif
441 int
442 sys_getfsstat(td, uap)
443 	struct thread *td;
444 	register struct getfsstat_args /* {
445 		struct statfs *buf;
446 		long bufsize;
447 		int flags;
448 	} */ *uap;
449 {
450 
451 	return (kern_getfsstat(td, &uap->buf, uap->bufsize, UIO_USERSPACE,
452 	    uap->flags));
453 }
454 
455 /*
456  * If (bufsize > 0 && bufseg == UIO_SYSSPACE)
457  * 	The caller is responsible for freeing memory which will be allocated
458  *	in '*buf'.
459  */
460 int
461 kern_getfsstat(struct thread *td, struct statfs **buf, size_t bufsize,
462     enum uio_seg bufseg, int flags)
463 {
464 	struct mount *mp, *nmp;
465 	struct statfs *sfsp, *sp, sb;
466 	size_t count, maxcount;
467 	int vfslocked;
468 	int error;
469 
470 	maxcount = bufsize / sizeof(struct statfs);
471 	if (bufsize == 0)
472 		sfsp = NULL;
473 	else if (bufseg == UIO_USERSPACE)
474 		sfsp = *buf;
475 	else /* if (bufseg == UIO_SYSSPACE) */ {
476 		count = 0;
477 		mtx_lock(&mountlist_mtx);
478 		TAILQ_FOREACH(mp, &mountlist, mnt_list) {
479 			count++;
480 		}
481 		mtx_unlock(&mountlist_mtx);
482 		if (maxcount > count)
483 			maxcount = count;
484 		sfsp = *buf = malloc(maxcount * sizeof(struct statfs), M_TEMP,
485 		    M_WAITOK);
486 	}
487 	count = 0;
488 	mtx_lock(&mountlist_mtx);
489 	for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
490 		if (prison_canseemount(td->td_ucred, mp) != 0) {
491 			nmp = TAILQ_NEXT(mp, mnt_list);
492 			continue;
493 		}
494 #ifdef MAC
495 		if (mac_mount_check_stat(td->td_ucred, mp) != 0) {
496 			nmp = TAILQ_NEXT(mp, mnt_list);
497 			continue;
498 		}
499 #endif
500 		if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK)) {
501 			nmp = TAILQ_NEXT(mp, mnt_list);
502 			continue;
503 		}
504 		vfslocked = VFS_LOCK_GIANT(mp);
505 		if (sfsp && count < maxcount) {
506 			sp = &mp->mnt_stat;
507 			/*
508 			 * Set these in case the underlying filesystem
509 			 * fails to do so.
510 			 */
511 			sp->f_version = STATFS_VERSION;
512 			sp->f_namemax = NAME_MAX;
513 			sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
514 			/*
515 			 * If MNT_NOWAIT or MNT_LAZY is specified, do not
516 			 * refresh the fsstat cache. MNT_NOWAIT or MNT_LAZY
517 			 * overrides MNT_WAIT.
518 			 */
519 			if (((flags & (MNT_LAZY|MNT_NOWAIT)) == 0 ||
520 			    (flags & MNT_WAIT)) &&
521 			    (error = VFS_STATFS(mp, sp))) {
522 				VFS_UNLOCK_GIANT(vfslocked);
523 				mtx_lock(&mountlist_mtx);
524 				nmp = TAILQ_NEXT(mp, mnt_list);
525 				vfs_unbusy(mp);
526 				continue;
527 			}
528 			if (priv_check(td, PRIV_VFS_GENERATION)) {
529 				bcopy(sp, &sb, sizeof(sb));
530 				sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
531 				prison_enforce_statfs(td->td_ucred, mp, &sb);
532 				sp = &sb;
533 			}
534 			if (bufseg == UIO_SYSSPACE)
535 				bcopy(sp, sfsp, sizeof(*sp));
536 			else /* if (bufseg == UIO_USERSPACE) */ {
537 				error = copyout(sp, sfsp, sizeof(*sp));
538 				if (error) {
539 					vfs_unbusy(mp);
540 					VFS_UNLOCK_GIANT(vfslocked);
541 					return (error);
542 				}
543 			}
544 			sfsp++;
545 		}
546 		VFS_UNLOCK_GIANT(vfslocked);
547 		count++;
548 		mtx_lock(&mountlist_mtx);
549 		nmp = TAILQ_NEXT(mp, mnt_list);
550 		vfs_unbusy(mp);
551 	}
552 	mtx_unlock(&mountlist_mtx);
553 	if (sfsp && count > maxcount)
554 		td->td_retval[0] = maxcount;
555 	else
556 		td->td_retval[0] = count;
557 	return (0);
558 }
559 
560 #ifdef COMPAT_FREEBSD4
561 /*
562  * Get old format filesystem statistics.
563  */
564 static void cvtstatfs(struct statfs *, struct ostatfs *);
565 
566 #ifndef _SYS_SYSPROTO_H_
567 struct freebsd4_statfs_args {
568 	char *path;
569 	struct ostatfs *buf;
570 };
571 #endif
572 int
573 freebsd4_statfs(td, uap)
574 	struct thread *td;
575 	struct freebsd4_statfs_args /* {
576 		char *path;
577 		struct ostatfs *buf;
578 	} */ *uap;
579 {
580 	struct ostatfs osb;
581 	struct statfs sf;
582 	int error;
583 
584 	error = kern_statfs(td, uap->path, UIO_USERSPACE, &sf);
585 	if (error)
586 		return (error);
587 	cvtstatfs(&sf, &osb);
588 	return (copyout(&osb, uap->buf, sizeof(osb)));
589 }
590 
591 /*
592  * Get filesystem statistics.
593  */
594 #ifndef _SYS_SYSPROTO_H_
595 struct freebsd4_fstatfs_args {
596 	int fd;
597 	struct ostatfs *buf;
598 };
599 #endif
600 int
601 freebsd4_fstatfs(td, uap)
602 	struct thread *td;
603 	struct freebsd4_fstatfs_args /* {
604 		int fd;
605 		struct ostatfs *buf;
606 	} */ *uap;
607 {
608 	struct ostatfs osb;
609 	struct statfs sf;
610 	int error;
611 
612 	error = kern_fstatfs(td, uap->fd, &sf);
613 	if (error)
614 		return (error);
615 	cvtstatfs(&sf, &osb);
616 	return (copyout(&osb, uap->buf, sizeof(osb)));
617 }
618 
619 /*
620  * Get statistics on all filesystems.
621  */
622 #ifndef _SYS_SYSPROTO_H_
623 struct freebsd4_getfsstat_args {
624 	struct ostatfs *buf;
625 	long bufsize;
626 	int flags;
627 };
628 #endif
629 int
630 freebsd4_getfsstat(td, uap)
631 	struct thread *td;
632 	register struct freebsd4_getfsstat_args /* {
633 		struct ostatfs *buf;
634 		long bufsize;
635 		int flags;
636 	} */ *uap;
637 {
638 	struct statfs *buf, *sp;
639 	struct ostatfs osb;
640 	size_t count, size;
641 	int error;
642 
643 	count = uap->bufsize / sizeof(struct ostatfs);
644 	size = count * sizeof(struct statfs);
645 	error = kern_getfsstat(td, &buf, size, UIO_SYSSPACE, uap->flags);
646 	if (size > 0) {
647 		count = td->td_retval[0];
648 		sp = buf;
649 		while (count > 0 && error == 0) {
650 			cvtstatfs(sp, &osb);
651 			error = copyout(&osb, uap->buf, sizeof(osb));
652 			sp++;
653 			uap->buf++;
654 			count--;
655 		}
656 		free(buf, M_TEMP);
657 	}
658 	return (error);
659 }
660 
661 /*
662  * Implement fstatfs() for (NFS) file handles.
663  */
664 #ifndef _SYS_SYSPROTO_H_
665 struct freebsd4_fhstatfs_args {
666 	struct fhandle *u_fhp;
667 	struct ostatfs *buf;
668 };
669 #endif
670 int
671 freebsd4_fhstatfs(td, uap)
672 	struct thread *td;
673 	struct freebsd4_fhstatfs_args /* {
674 		struct fhandle *u_fhp;
675 		struct ostatfs *buf;
676 	} */ *uap;
677 {
678 	struct ostatfs osb;
679 	struct statfs sf;
680 	fhandle_t fh;
681 	int error;
682 
683 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
684 	if (error)
685 		return (error);
686 	error = kern_fhstatfs(td, fh, &sf);
687 	if (error)
688 		return (error);
689 	cvtstatfs(&sf, &osb);
690 	return (copyout(&osb, uap->buf, sizeof(osb)));
691 }
692 
693 /*
694  * Convert a new format statfs structure to an old format statfs structure.
695  */
696 static void
697 cvtstatfs(nsp, osp)
698 	struct statfs *nsp;
699 	struct ostatfs *osp;
700 {
701 
702 	statfs_scale_blocks(nsp, LONG_MAX);
703 	bzero(osp, sizeof(*osp));
704 	osp->f_bsize = nsp->f_bsize;
705 	osp->f_iosize = MIN(nsp->f_iosize, LONG_MAX);
706 	osp->f_blocks = nsp->f_blocks;
707 	osp->f_bfree = nsp->f_bfree;
708 	osp->f_bavail = nsp->f_bavail;
709 	osp->f_files = MIN(nsp->f_files, LONG_MAX);
710 	osp->f_ffree = MIN(nsp->f_ffree, LONG_MAX);
711 	osp->f_owner = nsp->f_owner;
712 	osp->f_type = nsp->f_type;
713 	osp->f_flags = nsp->f_flags;
714 	osp->f_syncwrites = MIN(nsp->f_syncwrites, LONG_MAX);
715 	osp->f_asyncwrites = MIN(nsp->f_asyncwrites, LONG_MAX);
716 	osp->f_syncreads = MIN(nsp->f_syncreads, LONG_MAX);
717 	osp->f_asyncreads = MIN(nsp->f_asyncreads, LONG_MAX);
718 	strlcpy(osp->f_fstypename, nsp->f_fstypename,
719 	    MIN(MFSNAMELEN, OMFSNAMELEN));
720 	strlcpy(osp->f_mntonname, nsp->f_mntonname,
721 	    MIN(MNAMELEN, OMNAMELEN));
722 	strlcpy(osp->f_mntfromname, nsp->f_mntfromname,
723 	    MIN(MNAMELEN, OMNAMELEN));
724 	osp->f_fsid = nsp->f_fsid;
725 }
726 #endif /* COMPAT_FREEBSD4 */
727 
728 /*
729  * Change current working directory to a given file descriptor.
730  */
731 #ifndef _SYS_SYSPROTO_H_
732 struct fchdir_args {
733 	int	fd;
734 };
735 #endif
736 int
737 sys_fchdir(td, uap)
738 	struct thread *td;
739 	struct fchdir_args /* {
740 		int fd;
741 	} */ *uap;
742 {
743 	register struct filedesc *fdp = td->td_proc->p_fd;
744 	struct vnode *vp, *tdp, *vpold;
745 	struct mount *mp;
746 	struct file *fp;
747 	int vfslocked;
748 	int error;
749 
750 	AUDIT_ARG_FD(uap->fd);
751 	if ((error = getvnode(fdp, uap->fd, CAP_FCHDIR, &fp)) != 0)
752 		return (error);
753 	vp = fp->f_vnode;
754 	VREF(vp);
755 	fdrop(fp, td);
756 	vfslocked = VFS_LOCK_GIANT(vp->v_mount);
757 	vn_lock(vp, LK_SHARED | LK_RETRY);
758 	AUDIT_ARG_VNODE1(vp);
759 	error = change_dir(vp, td);
760 	while (!error && (mp = vp->v_mountedhere) != NULL) {
761 		int tvfslocked;
762 		if (vfs_busy(mp, 0))
763 			continue;
764 		tvfslocked = VFS_LOCK_GIANT(mp);
765 		error = VFS_ROOT(mp, LK_SHARED, &tdp);
766 		vfs_unbusy(mp);
767 		if (error) {
768 			VFS_UNLOCK_GIANT(tvfslocked);
769 			break;
770 		}
771 		vput(vp);
772 		VFS_UNLOCK_GIANT(vfslocked);
773 		vp = tdp;
774 		vfslocked = tvfslocked;
775 	}
776 	if (error) {
777 		vput(vp);
778 		VFS_UNLOCK_GIANT(vfslocked);
779 		return (error);
780 	}
781 	VOP_UNLOCK(vp, 0);
782 	VFS_UNLOCK_GIANT(vfslocked);
783 	FILEDESC_XLOCK(fdp);
784 	vpold = fdp->fd_cdir;
785 	fdp->fd_cdir = vp;
786 	FILEDESC_XUNLOCK(fdp);
787 	vfslocked = VFS_LOCK_GIANT(vpold->v_mount);
788 	vrele(vpold);
789 	VFS_UNLOCK_GIANT(vfslocked);
790 	return (0);
791 }
792 
793 /*
794  * Change current working directory (``.'').
795  */
796 #ifndef _SYS_SYSPROTO_H_
797 struct chdir_args {
798 	char	*path;
799 };
800 #endif
801 int
802 sys_chdir(td, uap)
803 	struct thread *td;
804 	struct chdir_args /* {
805 		char *path;
806 	} */ *uap;
807 {
808 
809 	return (kern_chdir(td, uap->path, UIO_USERSPACE));
810 }
811 
812 int
813 kern_chdir(struct thread *td, char *path, enum uio_seg pathseg)
814 {
815 	register struct filedesc *fdp = td->td_proc->p_fd;
816 	int error;
817 	struct nameidata nd;
818 	struct vnode *vp;
819 	int vfslocked;
820 
821 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1 |
822 	    MPSAFE, pathseg, path, td);
823 	if ((error = namei(&nd)) != 0)
824 		return (error);
825 	vfslocked = NDHASGIANT(&nd);
826 	if ((error = change_dir(nd.ni_vp, td)) != 0) {
827 		vput(nd.ni_vp);
828 		VFS_UNLOCK_GIANT(vfslocked);
829 		NDFREE(&nd, NDF_ONLY_PNBUF);
830 		return (error);
831 	}
832 	VOP_UNLOCK(nd.ni_vp, 0);
833 	VFS_UNLOCK_GIANT(vfslocked);
834 	NDFREE(&nd, NDF_ONLY_PNBUF);
835 	FILEDESC_XLOCK(fdp);
836 	vp = fdp->fd_cdir;
837 	fdp->fd_cdir = nd.ni_vp;
838 	FILEDESC_XUNLOCK(fdp);
839 	vfslocked = VFS_LOCK_GIANT(vp->v_mount);
840 	vrele(vp);
841 	VFS_UNLOCK_GIANT(vfslocked);
842 	return (0);
843 }
844 
845 /*
846  * Helper function for raised chroot(2) security function:  Refuse if
847  * any filedescriptors are open directories.
848  */
849 static int
850 chroot_refuse_vdir_fds(fdp)
851 	struct filedesc *fdp;
852 {
853 	struct vnode *vp;
854 	struct file *fp;
855 	int fd;
856 
857 	FILEDESC_LOCK_ASSERT(fdp);
858 
859 	for (fd = 0; fd < fdp->fd_nfiles ; fd++) {
860 		fp = fget_locked(fdp, fd);
861 		if (fp == NULL)
862 			continue;
863 		if (fp->f_type == DTYPE_VNODE) {
864 			vp = fp->f_vnode;
865 			if (vp->v_type == VDIR)
866 				return (EPERM);
867 		}
868 	}
869 	return (0);
870 }
871 
872 /*
873  * This sysctl determines if we will allow a process to chroot(2) if it
874  * has a directory open:
875  *	0: disallowed for all processes.
876  *	1: allowed for processes that were not already chroot(2)'ed.
877  *	2: allowed for all processes.
878  */
879 
880 static int chroot_allow_open_directories = 1;
881 
882 SYSCTL_INT(_kern, OID_AUTO, chroot_allow_open_directories, CTLFLAG_RW,
883      &chroot_allow_open_directories, 0, "");
884 
885 /*
886  * Change notion of root (``/'') directory.
887  */
888 #ifndef _SYS_SYSPROTO_H_
889 struct chroot_args {
890 	char	*path;
891 };
892 #endif
893 int
894 sys_chroot(td, uap)
895 	struct thread *td;
896 	struct chroot_args /* {
897 		char *path;
898 	} */ *uap;
899 {
900 	int error;
901 	struct nameidata nd;
902 	int vfslocked;
903 
904 	error = priv_check(td, PRIV_VFS_CHROOT);
905 	if (error)
906 		return (error);
907 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | MPSAFE |
908 	    AUDITVNODE1, UIO_USERSPACE, uap->path, td);
909 	error = namei(&nd);
910 	if (error)
911 		goto error;
912 	vfslocked = NDHASGIANT(&nd);
913 	if ((error = change_dir(nd.ni_vp, td)) != 0)
914 		goto e_vunlock;
915 #ifdef MAC
916 	if ((error = mac_vnode_check_chroot(td->td_ucred, nd.ni_vp)))
917 		goto e_vunlock;
918 #endif
919 	VOP_UNLOCK(nd.ni_vp, 0);
920 	error = change_root(nd.ni_vp, td);
921 	vrele(nd.ni_vp);
922 	VFS_UNLOCK_GIANT(vfslocked);
923 	NDFREE(&nd, NDF_ONLY_PNBUF);
924 	return (error);
925 e_vunlock:
926 	vput(nd.ni_vp);
927 	VFS_UNLOCK_GIANT(vfslocked);
928 error:
929 	NDFREE(&nd, NDF_ONLY_PNBUF);
930 	return (error);
931 }
932 
933 /*
934  * Common routine for chroot and chdir.  Callers must provide a locked vnode
935  * instance.
936  */
937 int
938 change_dir(vp, td)
939 	struct vnode *vp;
940 	struct thread *td;
941 {
942 	int error;
943 
944 	ASSERT_VOP_LOCKED(vp, "change_dir(): vp not locked");
945 	if (vp->v_type != VDIR)
946 		return (ENOTDIR);
947 #ifdef MAC
948 	error = mac_vnode_check_chdir(td->td_ucred, vp);
949 	if (error)
950 		return (error);
951 #endif
952 	error = VOP_ACCESS(vp, VEXEC, td->td_ucred, td);
953 	return (error);
954 }
955 
956 /*
957  * Common routine for kern_chroot() and jail_attach().  The caller is
958  * responsible for invoking priv_check() and mac_vnode_check_chroot() to
959  * authorize this operation.
960  */
961 int
962 change_root(vp, td)
963 	struct vnode *vp;
964 	struct thread *td;
965 {
966 	struct filedesc *fdp;
967 	struct vnode *oldvp;
968 	int vfslocked;
969 	int error;
970 
971 	VFS_ASSERT_GIANT(vp->v_mount);
972 	fdp = td->td_proc->p_fd;
973 	FILEDESC_XLOCK(fdp);
974 	if (chroot_allow_open_directories == 0 ||
975 	    (chroot_allow_open_directories == 1 && fdp->fd_rdir != rootvnode)) {
976 		error = chroot_refuse_vdir_fds(fdp);
977 		if (error) {
978 			FILEDESC_XUNLOCK(fdp);
979 			return (error);
980 		}
981 	}
982 	oldvp = fdp->fd_rdir;
983 	fdp->fd_rdir = vp;
984 	VREF(fdp->fd_rdir);
985 	if (!fdp->fd_jdir) {
986 		fdp->fd_jdir = vp;
987 		VREF(fdp->fd_jdir);
988 	}
989 	FILEDESC_XUNLOCK(fdp);
990 	vfslocked = VFS_LOCK_GIANT(oldvp->v_mount);
991 	vrele(oldvp);
992 	VFS_UNLOCK_GIANT(vfslocked);
993 	return (0);
994 }
995 
996 static __inline cap_rights_t
997 flags_to_rights(int flags)
998 {
999 	cap_rights_t rights = 0;
1000 
1001 	switch ((flags & O_ACCMODE)) {
1002 	case O_RDONLY:
1003 		rights |= CAP_READ;
1004 		break;
1005 
1006 	case O_RDWR:
1007 		rights |= CAP_READ;
1008 		/* fall through */
1009 
1010 	case O_WRONLY:
1011 		rights |= CAP_WRITE;
1012 		break;
1013 
1014 	case O_EXEC:
1015 		rights |= CAP_FEXECVE;
1016 		break;
1017 	}
1018 
1019 	if (flags & O_CREAT)
1020 		rights |= CAP_CREATE;
1021 
1022 	if (flags & O_TRUNC)
1023 		rights |= CAP_FTRUNCATE;
1024 
1025 	if ((flags & O_EXLOCK) || (flags & O_SHLOCK))
1026 		rights |= CAP_FLOCK;
1027 
1028 	return (rights);
1029 }
1030 
1031 /*
1032  * Check permissions, allocate an open file structure, and call the device
1033  * open routine if any.
1034  */
1035 #ifndef _SYS_SYSPROTO_H_
1036 struct open_args {
1037 	char	*path;
1038 	int	flags;
1039 	int	mode;
1040 };
1041 #endif
1042 int
1043 sys_open(td, uap)
1044 	struct thread *td;
1045 	register struct open_args /* {
1046 		char *path;
1047 		int flags;
1048 		int mode;
1049 	} */ *uap;
1050 {
1051 
1052 	return (kern_open(td, uap->path, UIO_USERSPACE, uap->flags, uap->mode));
1053 }
1054 
1055 #ifndef _SYS_SYSPROTO_H_
1056 struct openat_args {
1057 	int	fd;
1058 	char	*path;
1059 	int	flag;
1060 	int	mode;
1061 };
1062 #endif
1063 int
1064 sys_openat(struct thread *td, struct openat_args *uap)
1065 {
1066 
1067 	return (kern_openat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
1068 	    uap->mode));
1069 }
1070 
1071 int
1072 kern_open(struct thread *td, char *path, enum uio_seg pathseg, int flags,
1073     int mode)
1074 {
1075 
1076 	return (kern_openat(td, AT_FDCWD, path, pathseg, flags, mode));
1077 }
1078 
1079 int
1080 kern_openat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1081     int flags, int mode)
1082 {
1083 	struct proc *p = td->td_proc;
1084 	struct filedesc *fdp = p->p_fd;
1085 	struct file *fp;
1086 	struct vnode *vp;
1087 	int cmode;
1088 	struct file *nfp;
1089 	int type, indx = -1, error, error_open;
1090 	struct flock lf;
1091 	struct nameidata nd;
1092 	int vfslocked;
1093 	cap_rights_t rights_needed = CAP_LOOKUP;
1094 
1095 	AUDIT_ARG_FFLAGS(flags);
1096 	AUDIT_ARG_MODE(mode);
1097 	/* XXX: audit dirfd */
1098 	rights_needed |= flags_to_rights(flags);
1099 	/*
1100 	 * Only one of the O_EXEC, O_RDONLY, O_WRONLY and O_RDWR flags
1101 	 * may be specified.
1102 	 */
1103 	if (flags & O_EXEC) {
1104 		if (flags & O_ACCMODE)
1105 			return (EINVAL);
1106 	} else if ((flags & O_ACCMODE) == O_ACCMODE)
1107 		return (EINVAL);
1108 	else
1109 		flags = FFLAGS(flags);
1110 
1111 	/*
1112 	 * allocate the file descriptor, but don't install a descriptor yet
1113 	 */
1114 	error = falloc_noinstall(td, &nfp);
1115 	if (error)
1116 		return (error);
1117 	/* An extra reference on `nfp' has been held for us by falloc_noinstall(). */
1118 	fp = nfp;
1119 	/* Set the flags early so the finit in devfs can pick them up. */
1120 	fp->f_flag = flags & FMASK;
1121 	cmode = ((mode &~ fdp->fd_cmask) & ALLPERMS) &~ S_ISTXT;
1122 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1 | MPSAFE, pathseg,
1123 	    path, fd, rights_needed, td);
1124 	td->td_dupfd = -1;		/* XXX check for fdopen */
1125 	error = vn_open(&nd, &flags, cmode, fp);
1126 	if (error) {
1127 		/*
1128 		 * If the vn_open replaced the method vector, something
1129 		 * wonderous happened deep below and we just pass it up
1130 		 * pretending we know what we do.
1131 		 */
1132 		if (error == ENXIO && fp->f_ops != &badfileops)
1133 			goto success;
1134 
1135 		/*
1136 		 * handle special fdopen() case.  bleh.  dupfdopen() is
1137 		 * responsible for dropping the old contents of ofiles[indx]
1138 		 * if it succeeds.
1139 		 *
1140 		 * Don't do this for relative (capability) lookups; we don't
1141 		 * understand exactly what would happen, and we don't think
1142 		 * that it ever should.
1143 		 */
1144 		if ((nd.ni_strictrelative == 0) &&
1145 		    (error == ENODEV || error == ENXIO) &&
1146 		    (td->td_dupfd >= 0)) {
1147 			/* XXX from fdopen */
1148 			error_open = error;
1149 			if ((error = finstall(td, fp, &indx, flags)) != 0)
1150 				goto bad_unlocked;
1151 			if ((error = dupfdopen(td, fdp, indx, td->td_dupfd,
1152 			    flags, error_open)) == 0)
1153 				goto success;
1154 		}
1155 		/*
1156 		 * Clean up the descriptor, but only if another thread hadn't
1157 		 * replaced or closed it.
1158 		 */
1159 		if (indx != -1)
1160 			fdclose(fdp, fp, indx, td);
1161 		fdrop(fp, td);
1162 
1163 		if (error == ERESTART)
1164 			error = EINTR;
1165 		return (error);
1166 	}
1167 	td->td_dupfd = 0;
1168 	vfslocked = NDHASGIANT(&nd);
1169 	NDFREE(&nd, NDF_ONLY_PNBUF);
1170 	vp = nd.ni_vp;
1171 
1172 	/*
1173 	 * Store the vnode, for any f_type. Typically, the vnode use
1174 	 * count is decremented by direct call to vn_closefile() for
1175 	 * files that switched type in the cdevsw fdopen() method.
1176 	 */
1177 	fp->f_vnode = vp;
1178 	/*
1179 	 * If the file wasn't claimed by devfs bind it to the normal
1180 	 * vnode operations here.
1181 	 */
1182 	if (fp->f_ops == &badfileops) {
1183 		KASSERT(vp->v_type != VFIFO, ("Unexpected fifo."));
1184 		fp->f_seqcount = 1;
1185 		finit(fp, flags & FMASK, DTYPE_VNODE, vp, &vnops);
1186 	}
1187 
1188 	VOP_UNLOCK(vp, 0);
1189 	if (fp->f_type == DTYPE_VNODE && (flags & (O_EXLOCK | O_SHLOCK)) != 0) {
1190 		lf.l_whence = SEEK_SET;
1191 		lf.l_start = 0;
1192 		lf.l_len = 0;
1193 		if (flags & O_EXLOCK)
1194 			lf.l_type = F_WRLCK;
1195 		else
1196 			lf.l_type = F_RDLCK;
1197 		type = F_FLOCK;
1198 		if ((flags & FNONBLOCK) == 0)
1199 			type |= F_WAIT;
1200 		if ((error = VOP_ADVLOCK(vp, (caddr_t)fp, F_SETLK, &lf,
1201 			    type)) != 0)
1202 			goto bad;
1203 		atomic_set_int(&fp->f_flag, FHASLOCK);
1204 	}
1205 	if (flags & O_TRUNC) {
1206 		error = fo_truncate(fp, 0, td->td_ucred, td);
1207 		if (error)
1208 			goto bad;
1209 	}
1210 	VFS_UNLOCK_GIANT(vfslocked);
1211 success:
1212 	/*
1213 	 * If we haven't already installed the FD (for dupfdopen), do so now.
1214 	 */
1215 	if (indx == -1) {
1216 #ifdef CAPABILITIES
1217 		if (nd.ni_strictrelative == 1) {
1218 			/*
1219 			 * We are doing a strict relative lookup; wrap the
1220 			 * result in a capability.
1221 			 */
1222 			if ((error = kern_capwrap(td, fp, nd.ni_baserights,
1223 			    &indx)) != 0)
1224 				goto bad_unlocked;
1225 		} else
1226 #endif
1227 			if ((error = finstall(td, fp, &indx, flags)) != 0)
1228 				goto bad_unlocked;
1229 
1230 	}
1231 
1232 	/*
1233 	 * Release our private reference, leaving the one associated with
1234 	 * the descriptor table intact.
1235 	 */
1236 	fdrop(fp, td);
1237 	td->td_retval[0] = indx;
1238 	return (0);
1239 bad:
1240 	VFS_UNLOCK_GIANT(vfslocked);
1241 bad_unlocked:
1242 	if (indx != -1)
1243 		fdclose(fdp, fp, indx, td);
1244 	fdrop(fp, td);
1245 	td->td_retval[0] = -1;
1246 	return (error);
1247 }
1248 
1249 #ifdef COMPAT_43
1250 /*
1251  * Create a file.
1252  */
1253 #ifndef _SYS_SYSPROTO_H_
1254 struct ocreat_args {
1255 	char	*path;
1256 	int	mode;
1257 };
1258 #endif
1259 int
1260 ocreat(td, uap)
1261 	struct thread *td;
1262 	register struct ocreat_args /* {
1263 		char *path;
1264 		int mode;
1265 	} */ *uap;
1266 {
1267 
1268 	return (kern_open(td, uap->path, UIO_USERSPACE,
1269 	    O_WRONLY | O_CREAT | O_TRUNC, uap->mode));
1270 }
1271 #endif /* COMPAT_43 */
1272 
1273 /*
1274  * Create a special file.
1275  */
1276 #ifndef _SYS_SYSPROTO_H_
1277 struct mknod_args {
1278 	char	*path;
1279 	int	mode;
1280 	int	dev;
1281 };
1282 #endif
1283 int
1284 sys_mknod(td, uap)
1285 	struct thread *td;
1286 	register struct mknod_args /* {
1287 		char *path;
1288 		int mode;
1289 		int dev;
1290 	} */ *uap;
1291 {
1292 
1293 	return (kern_mknod(td, uap->path, UIO_USERSPACE, uap->mode, uap->dev));
1294 }
1295 
1296 #ifndef _SYS_SYSPROTO_H_
1297 struct mknodat_args {
1298 	int	fd;
1299 	char	*path;
1300 	mode_t	mode;
1301 	dev_t	dev;
1302 };
1303 #endif
1304 int
1305 sys_mknodat(struct thread *td, struct mknodat_args *uap)
1306 {
1307 
1308 	return (kern_mknodat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode,
1309 	    uap->dev));
1310 }
1311 
1312 int
1313 kern_mknod(struct thread *td, char *path, enum uio_seg pathseg, int mode,
1314     int dev)
1315 {
1316 
1317 	return (kern_mknodat(td, AT_FDCWD, path, pathseg, mode, dev));
1318 }
1319 
1320 int
1321 kern_mknodat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1322     int mode, int dev)
1323 {
1324 	struct vnode *vp;
1325 	struct mount *mp;
1326 	struct vattr vattr;
1327 	int error;
1328 	int whiteout = 0;
1329 	struct nameidata nd;
1330 	int vfslocked;
1331 
1332 	AUDIT_ARG_MODE(mode);
1333 	AUDIT_ARG_DEV(dev);
1334 	switch (mode & S_IFMT) {
1335 	case S_IFCHR:
1336 	case S_IFBLK:
1337 		error = priv_check(td, PRIV_VFS_MKNOD_DEV);
1338 		break;
1339 	case S_IFMT:
1340 		error = priv_check(td, PRIV_VFS_MKNOD_BAD);
1341 		break;
1342 	case S_IFWHT:
1343 		error = priv_check(td, PRIV_VFS_MKNOD_WHT);
1344 		break;
1345 	case S_IFIFO:
1346 		if (dev == 0)
1347 			return (kern_mkfifoat(td, fd, path, pathseg, mode));
1348 		/* FALLTHROUGH */
1349 	default:
1350 		error = EINVAL;
1351 		break;
1352 	}
1353 	if (error)
1354 		return (error);
1355 restart:
1356 	bwillwrite();
1357 	NDINIT_ATRIGHTS(&nd, CREATE,
1358 	    LOCKPARENT | SAVENAME | MPSAFE | AUDITVNODE1, pathseg, path, fd,
1359 	    CAP_MKFIFO, td);
1360 	if ((error = namei(&nd)) != 0)
1361 		return (error);
1362 	vfslocked = NDHASGIANT(&nd);
1363 	vp = nd.ni_vp;
1364 	if (vp != NULL) {
1365 		NDFREE(&nd, NDF_ONLY_PNBUF);
1366 		if (vp == nd.ni_dvp)
1367 			vrele(nd.ni_dvp);
1368 		else
1369 			vput(nd.ni_dvp);
1370 		vrele(vp);
1371 		VFS_UNLOCK_GIANT(vfslocked);
1372 		return (EEXIST);
1373 	} else {
1374 		VATTR_NULL(&vattr);
1375 		vattr.va_mode = (mode & ALLPERMS) &
1376 		    ~td->td_proc->p_fd->fd_cmask;
1377 		vattr.va_rdev = dev;
1378 		whiteout = 0;
1379 
1380 		switch (mode & S_IFMT) {
1381 		case S_IFMT:	/* used by badsect to flag bad sectors */
1382 			vattr.va_type = VBAD;
1383 			break;
1384 		case S_IFCHR:
1385 			vattr.va_type = VCHR;
1386 			break;
1387 		case S_IFBLK:
1388 			vattr.va_type = VBLK;
1389 			break;
1390 		case S_IFWHT:
1391 			whiteout = 1;
1392 			break;
1393 		default:
1394 			panic("kern_mknod: invalid mode");
1395 		}
1396 	}
1397 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1398 		NDFREE(&nd, NDF_ONLY_PNBUF);
1399 		vput(nd.ni_dvp);
1400 		VFS_UNLOCK_GIANT(vfslocked);
1401 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1402 			return (error);
1403 		goto restart;
1404 	}
1405 #ifdef MAC
1406 	if (error == 0 && !whiteout)
1407 		error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp,
1408 		    &nd.ni_cnd, &vattr);
1409 #endif
1410 	if (!error) {
1411 		if (whiteout)
1412 			error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE);
1413 		else {
1414 			error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp,
1415 						&nd.ni_cnd, &vattr);
1416 			if (error == 0)
1417 				vput(nd.ni_vp);
1418 		}
1419 	}
1420 	NDFREE(&nd, NDF_ONLY_PNBUF);
1421 	vput(nd.ni_dvp);
1422 	vn_finished_write(mp);
1423 	VFS_UNLOCK_GIANT(vfslocked);
1424 	return (error);
1425 }
1426 
1427 /*
1428  * Create a named pipe.
1429  */
1430 #ifndef _SYS_SYSPROTO_H_
1431 struct mkfifo_args {
1432 	char	*path;
1433 	int	mode;
1434 };
1435 #endif
1436 int
1437 sys_mkfifo(td, uap)
1438 	struct thread *td;
1439 	register struct mkfifo_args /* {
1440 		char *path;
1441 		int mode;
1442 	} */ *uap;
1443 {
1444 
1445 	return (kern_mkfifo(td, uap->path, UIO_USERSPACE, uap->mode));
1446 }
1447 
1448 #ifndef _SYS_SYSPROTO_H_
1449 struct mkfifoat_args {
1450 	int	fd;
1451 	char	*path;
1452 	mode_t	mode;
1453 };
1454 #endif
1455 int
1456 sys_mkfifoat(struct thread *td, struct mkfifoat_args *uap)
1457 {
1458 
1459 	return (kern_mkfifoat(td, uap->fd, uap->path, UIO_USERSPACE,
1460 	    uap->mode));
1461 }
1462 
1463 int
1464 kern_mkfifo(struct thread *td, char *path, enum uio_seg pathseg, int mode)
1465 {
1466 
1467 	return (kern_mkfifoat(td, AT_FDCWD, path, pathseg, mode));
1468 }
1469 
1470 int
1471 kern_mkfifoat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1472     int mode)
1473 {
1474 	struct mount *mp;
1475 	struct vattr vattr;
1476 	int error;
1477 	struct nameidata nd;
1478 	int vfslocked;
1479 
1480 	AUDIT_ARG_MODE(mode);
1481 restart:
1482 	bwillwrite();
1483 	NDINIT_AT(&nd, CREATE, LOCKPARENT | SAVENAME | MPSAFE | AUDITVNODE1,
1484 	    pathseg, path, fd, td);
1485 	if ((error = namei(&nd)) != 0)
1486 		return (error);
1487 	vfslocked = NDHASGIANT(&nd);
1488 	if (nd.ni_vp != NULL) {
1489 		NDFREE(&nd, NDF_ONLY_PNBUF);
1490 		if (nd.ni_vp == nd.ni_dvp)
1491 			vrele(nd.ni_dvp);
1492 		else
1493 			vput(nd.ni_dvp);
1494 		vrele(nd.ni_vp);
1495 		VFS_UNLOCK_GIANT(vfslocked);
1496 		return (EEXIST);
1497 	}
1498 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1499 		NDFREE(&nd, NDF_ONLY_PNBUF);
1500 		vput(nd.ni_dvp);
1501 		VFS_UNLOCK_GIANT(vfslocked);
1502 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1503 			return (error);
1504 		goto restart;
1505 	}
1506 	VATTR_NULL(&vattr);
1507 	vattr.va_type = VFIFO;
1508 	vattr.va_mode = (mode & ALLPERMS) & ~td->td_proc->p_fd->fd_cmask;
1509 #ifdef MAC
1510 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1511 	    &vattr);
1512 	if (error)
1513 		goto out;
1514 #endif
1515 	error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
1516 	if (error == 0)
1517 		vput(nd.ni_vp);
1518 #ifdef MAC
1519 out:
1520 #endif
1521 	vput(nd.ni_dvp);
1522 	vn_finished_write(mp);
1523 	VFS_UNLOCK_GIANT(vfslocked);
1524 	NDFREE(&nd, NDF_ONLY_PNBUF);
1525 	return (error);
1526 }
1527 
1528 /*
1529  * Make a hard file link.
1530  */
1531 #ifndef _SYS_SYSPROTO_H_
1532 struct link_args {
1533 	char	*path;
1534 	char	*link;
1535 };
1536 #endif
1537 int
1538 sys_link(td, uap)
1539 	struct thread *td;
1540 	register struct link_args /* {
1541 		char *path;
1542 		char *link;
1543 	} */ *uap;
1544 {
1545 
1546 	return (kern_link(td, uap->path, uap->link, UIO_USERSPACE));
1547 }
1548 
1549 #ifndef _SYS_SYSPROTO_H_
1550 struct linkat_args {
1551 	int	fd1;
1552 	char	*path1;
1553 	int	fd2;
1554 	char	*path2;
1555 	int	flag;
1556 };
1557 #endif
1558 int
1559 sys_linkat(struct thread *td, struct linkat_args *uap)
1560 {
1561 	int flag;
1562 
1563 	flag = uap->flag;
1564 	if (flag & ~AT_SYMLINK_FOLLOW)
1565 		return (EINVAL);
1566 
1567 	return (kern_linkat(td, uap->fd1, uap->fd2, uap->path1, uap->path2,
1568 	    UIO_USERSPACE, (flag & AT_SYMLINK_FOLLOW) ? FOLLOW : NOFOLLOW));
1569 }
1570 
1571 int hardlink_check_uid = 0;
1572 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_uid, CTLFLAG_RW,
1573     &hardlink_check_uid, 0,
1574     "Unprivileged processes cannot create hard links to files owned by other "
1575     "users");
1576 static int hardlink_check_gid = 0;
1577 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_gid, CTLFLAG_RW,
1578     &hardlink_check_gid, 0,
1579     "Unprivileged processes cannot create hard links to files owned by other "
1580     "groups");
1581 
1582 static int
1583 can_hardlink(struct vnode *vp, struct ucred *cred)
1584 {
1585 	struct vattr va;
1586 	int error;
1587 
1588 	if (!hardlink_check_uid && !hardlink_check_gid)
1589 		return (0);
1590 
1591 	error = VOP_GETATTR(vp, &va, cred);
1592 	if (error != 0)
1593 		return (error);
1594 
1595 	if (hardlink_check_uid && cred->cr_uid != va.va_uid) {
1596 		error = priv_check_cred(cred, PRIV_VFS_LINK, 0);
1597 		if (error)
1598 			return (error);
1599 	}
1600 
1601 	if (hardlink_check_gid && !groupmember(va.va_gid, cred)) {
1602 		error = priv_check_cred(cred, PRIV_VFS_LINK, 0);
1603 		if (error)
1604 			return (error);
1605 	}
1606 
1607 	return (0);
1608 }
1609 
1610 int
1611 kern_link(struct thread *td, char *path, char *link, enum uio_seg segflg)
1612 {
1613 
1614 	return (kern_linkat(td, AT_FDCWD, AT_FDCWD, path,link, segflg, FOLLOW));
1615 }
1616 
1617 int
1618 kern_linkat(struct thread *td, int fd1, int fd2, char *path1, char *path2,
1619     enum uio_seg segflg, int follow)
1620 {
1621 	struct vnode *vp;
1622 	struct mount *mp;
1623 	struct nameidata nd;
1624 	int vfslocked;
1625 	int lvfslocked;
1626 	int error;
1627 
1628 	bwillwrite();
1629 	NDINIT_AT(&nd, LOOKUP, follow | MPSAFE | AUDITVNODE1, segflg, path1,
1630 	    fd1, td);
1631 
1632 	if ((error = namei(&nd)) != 0)
1633 		return (error);
1634 	vfslocked = NDHASGIANT(&nd);
1635 	NDFREE(&nd, NDF_ONLY_PNBUF);
1636 	vp = nd.ni_vp;
1637 	if (vp->v_type == VDIR) {
1638 		vrele(vp);
1639 		VFS_UNLOCK_GIANT(vfslocked);
1640 		return (EPERM);		/* POSIX */
1641 	}
1642 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) {
1643 		vrele(vp);
1644 		VFS_UNLOCK_GIANT(vfslocked);
1645 		return (error);
1646 	}
1647 	NDINIT_AT(&nd, CREATE, LOCKPARENT | SAVENAME | MPSAFE | AUDITVNODE2,
1648 	    segflg, path2, fd2, td);
1649 	if ((error = namei(&nd)) == 0) {
1650 		lvfslocked = NDHASGIANT(&nd);
1651 		if (nd.ni_vp != NULL) {
1652 			if (nd.ni_dvp == nd.ni_vp)
1653 				vrele(nd.ni_dvp);
1654 			else
1655 				vput(nd.ni_dvp);
1656 			vrele(nd.ni_vp);
1657 			error = EEXIST;
1658 		} else if ((error = vn_lock(vp, LK_EXCLUSIVE | LK_RETRY))
1659 		    == 0) {
1660 			error = can_hardlink(vp, td->td_ucred);
1661 			if (error == 0)
1662 #ifdef MAC
1663 				error = mac_vnode_check_link(td->td_ucred,
1664 				    nd.ni_dvp, vp, &nd.ni_cnd);
1665 			if (error == 0)
1666 #endif
1667 				error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
1668 			VOP_UNLOCK(vp, 0);
1669 			vput(nd.ni_dvp);
1670 		}
1671 		NDFREE(&nd, NDF_ONLY_PNBUF);
1672 		VFS_UNLOCK_GIANT(lvfslocked);
1673 	}
1674 	vrele(vp);
1675 	vn_finished_write(mp);
1676 	VFS_UNLOCK_GIANT(vfslocked);
1677 	return (error);
1678 }
1679 
1680 /*
1681  * Make a symbolic link.
1682  */
1683 #ifndef _SYS_SYSPROTO_H_
1684 struct symlink_args {
1685 	char	*path;
1686 	char	*link;
1687 };
1688 #endif
1689 int
1690 sys_symlink(td, uap)
1691 	struct thread *td;
1692 	register struct symlink_args /* {
1693 		char *path;
1694 		char *link;
1695 	} */ *uap;
1696 {
1697 
1698 	return (kern_symlink(td, uap->path, uap->link, UIO_USERSPACE));
1699 }
1700 
1701 #ifndef _SYS_SYSPROTO_H_
1702 struct symlinkat_args {
1703 	char	*path;
1704 	int	fd;
1705 	char	*path2;
1706 };
1707 #endif
1708 int
1709 sys_symlinkat(struct thread *td, struct symlinkat_args *uap)
1710 {
1711 
1712 	return (kern_symlinkat(td, uap->path1, uap->fd, uap->path2,
1713 	    UIO_USERSPACE));
1714 }
1715 
1716 int
1717 kern_symlink(struct thread *td, char *path, char *link, enum uio_seg segflg)
1718 {
1719 
1720 	return (kern_symlinkat(td, path, AT_FDCWD, link, segflg));
1721 }
1722 
1723 int
1724 kern_symlinkat(struct thread *td, char *path1, int fd, char *path2,
1725     enum uio_seg segflg)
1726 {
1727 	struct mount *mp;
1728 	struct vattr vattr;
1729 	char *syspath;
1730 	int error;
1731 	struct nameidata nd;
1732 	int vfslocked;
1733 
1734 	if (segflg == UIO_SYSSPACE) {
1735 		syspath = path1;
1736 	} else {
1737 		syspath = uma_zalloc(namei_zone, M_WAITOK);
1738 		if ((error = copyinstr(path1, syspath, MAXPATHLEN, NULL)) != 0)
1739 			goto out;
1740 	}
1741 	AUDIT_ARG_TEXT(syspath);
1742 restart:
1743 	bwillwrite();
1744 	NDINIT_AT(&nd, CREATE, LOCKPARENT | SAVENAME | MPSAFE | AUDITVNODE1,
1745 	    segflg, path2, fd, td);
1746 	if ((error = namei(&nd)) != 0)
1747 		goto out;
1748 	vfslocked = NDHASGIANT(&nd);
1749 	if (nd.ni_vp) {
1750 		NDFREE(&nd, NDF_ONLY_PNBUF);
1751 		if (nd.ni_vp == nd.ni_dvp)
1752 			vrele(nd.ni_dvp);
1753 		else
1754 			vput(nd.ni_dvp);
1755 		vrele(nd.ni_vp);
1756 		VFS_UNLOCK_GIANT(vfslocked);
1757 		error = EEXIST;
1758 		goto out;
1759 	}
1760 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1761 		NDFREE(&nd, NDF_ONLY_PNBUF);
1762 		vput(nd.ni_dvp);
1763 		VFS_UNLOCK_GIANT(vfslocked);
1764 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1765 			goto out;
1766 		goto restart;
1767 	}
1768 	VATTR_NULL(&vattr);
1769 	vattr.va_mode = ACCESSPERMS &~ td->td_proc->p_fd->fd_cmask;
1770 #ifdef MAC
1771 	vattr.va_type = VLNK;
1772 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1773 	    &vattr);
1774 	if (error)
1775 		goto out2;
1776 #endif
1777 	error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, syspath);
1778 	if (error == 0)
1779 		vput(nd.ni_vp);
1780 #ifdef MAC
1781 out2:
1782 #endif
1783 	NDFREE(&nd, NDF_ONLY_PNBUF);
1784 	vput(nd.ni_dvp);
1785 	vn_finished_write(mp);
1786 	VFS_UNLOCK_GIANT(vfslocked);
1787 out:
1788 	if (segflg != UIO_SYSSPACE)
1789 		uma_zfree(namei_zone, syspath);
1790 	return (error);
1791 }
1792 
1793 /*
1794  * Delete a whiteout from the filesystem.
1795  */
1796 int
1797 sys_undelete(td, uap)
1798 	struct thread *td;
1799 	register struct undelete_args /* {
1800 		char *path;
1801 	} */ *uap;
1802 {
1803 	int error;
1804 	struct mount *mp;
1805 	struct nameidata nd;
1806 	int vfslocked;
1807 
1808 restart:
1809 	bwillwrite();
1810 	NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | MPSAFE | AUDITVNODE1,
1811 	    UIO_USERSPACE, uap->path, td);
1812 	error = namei(&nd);
1813 	if (error)
1814 		return (error);
1815 	vfslocked = NDHASGIANT(&nd);
1816 
1817 	if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) {
1818 		NDFREE(&nd, NDF_ONLY_PNBUF);
1819 		if (nd.ni_vp == nd.ni_dvp)
1820 			vrele(nd.ni_dvp);
1821 		else
1822 			vput(nd.ni_dvp);
1823 		if (nd.ni_vp)
1824 			vrele(nd.ni_vp);
1825 		VFS_UNLOCK_GIANT(vfslocked);
1826 		return (EEXIST);
1827 	}
1828 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1829 		NDFREE(&nd, NDF_ONLY_PNBUF);
1830 		vput(nd.ni_dvp);
1831 		VFS_UNLOCK_GIANT(vfslocked);
1832 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1833 			return (error);
1834 		goto restart;
1835 	}
1836 	error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE);
1837 	NDFREE(&nd, NDF_ONLY_PNBUF);
1838 	vput(nd.ni_dvp);
1839 	vn_finished_write(mp);
1840 	VFS_UNLOCK_GIANT(vfslocked);
1841 	return (error);
1842 }
1843 
1844 /*
1845  * Delete a name from the filesystem.
1846  */
1847 #ifndef _SYS_SYSPROTO_H_
1848 struct unlink_args {
1849 	char	*path;
1850 };
1851 #endif
1852 int
1853 sys_unlink(td, uap)
1854 	struct thread *td;
1855 	struct unlink_args /* {
1856 		char *path;
1857 	} */ *uap;
1858 {
1859 
1860 	return (kern_unlink(td, uap->path, UIO_USERSPACE));
1861 }
1862 
1863 #ifndef _SYS_SYSPROTO_H_
1864 struct unlinkat_args {
1865 	int	fd;
1866 	char	*path;
1867 	int	flag;
1868 };
1869 #endif
1870 int
1871 sys_unlinkat(struct thread *td, struct unlinkat_args *uap)
1872 {
1873 	int flag = uap->flag;
1874 	int fd = uap->fd;
1875 	char *path = uap->path;
1876 
1877 	if (flag & ~AT_REMOVEDIR)
1878 		return (EINVAL);
1879 
1880 	if (flag & AT_REMOVEDIR)
1881 		return (kern_rmdirat(td, fd, path, UIO_USERSPACE));
1882 	else
1883 		return (kern_unlinkat(td, fd, path, UIO_USERSPACE, 0));
1884 }
1885 
1886 int
1887 kern_unlink(struct thread *td, char *path, enum uio_seg pathseg)
1888 {
1889 
1890 	return (kern_unlinkat(td, AT_FDCWD, path, pathseg, 0));
1891 }
1892 
1893 int
1894 kern_unlinkat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1895     ino_t oldinum)
1896 {
1897 	struct mount *mp;
1898 	struct vnode *vp;
1899 	int error;
1900 	struct nameidata nd;
1901 	struct stat sb;
1902 	int vfslocked;
1903 
1904 restart:
1905 	bwillwrite();
1906 	NDINIT_AT(&nd, DELETE, LOCKPARENT | LOCKLEAF | MPSAFE | AUDITVNODE1,
1907 	    pathseg, path, fd, td);
1908 	if ((error = namei(&nd)) != 0)
1909 		return (error == EINVAL ? EPERM : error);
1910 	vfslocked = NDHASGIANT(&nd);
1911 	vp = nd.ni_vp;
1912 	if (vp->v_type == VDIR && oldinum == 0) {
1913 		error = EPERM;		/* POSIX */
1914 	} else if (oldinum != 0 &&
1915 		  ((error = vn_stat(vp, &sb, td->td_ucred, NOCRED, td)) == 0) &&
1916 		  sb.st_ino != oldinum) {
1917 			error = EIDRM;	/* Identifier removed */
1918 	} else {
1919 		/*
1920 		 * The root of a mounted filesystem cannot be deleted.
1921 		 *
1922 		 * XXX: can this only be a VDIR case?
1923 		 */
1924 		if (vp->v_vflag & VV_ROOT)
1925 			error = EBUSY;
1926 	}
1927 	if (error == 0) {
1928 		if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1929 			NDFREE(&nd, NDF_ONLY_PNBUF);
1930 			vput(nd.ni_dvp);
1931 			if (vp == nd.ni_dvp)
1932 				vrele(vp);
1933 			else
1934 				vput(vp);
1935 			VFS_UNLOCK_GIANT(vfslocked);
1936 			if ((error = vn_start_write(NULL, &mp,
1937 			    V_XSLEEP | PCATCH)) != 0)
1938 				return (error);
1939 			goto restart;
1940 		}
1941 #ifdef MAC
1942 		error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
1943 		    &nd.ni_cnd);
1944 		if (error)
1945 			goto out;
1946 #endif
1947 		error = VOP_REMOVE(nd.ni_dvp, vp, &nd.ni_cnd);
1948 #ifdef MAC
1949 out:
1950 #endif
1951 		vn_finished_write(mp);
1952 	}
1953 	NDFREE(&nd, NDF_ONLY_PNBUF);
1954 	vput(nd.ni_dvp);
1955 	if (vp == nd.ni_dvp)
1956 		vrele(vp);
1957 	else
1958 		vput(vp);
1959 	VFS_UNLOCK_GIANT(vfslocked);
1960 	return (error);
1961 }
1962 
1963 /*
1964  * Reposition read/write file offset.
1965  */
1966 #ifndef _SYS_SYSPROTO_H_
1967 struct lseek_args {
1968 	int	fd;
1969 	int	pad;
1970 	off_t	offset;
1971 	int	whence;
1972 };
1973 #endif
1974 int
1975 sys_lseek(td, uap)
1976 	struct thread *td;
1977 	register struct lseek_args /* {
1978 		int fd;
1979 		int pad;
1980 		off_t offset;
1981 		int whence;
1982 	} */ *uap;
1983 {
1984 	struct ucred *cred = td->td_ucred;
1985 	struct file *fp;
1986 	struct vnode *vp;
1987 	struct vattr vattr;
1988 	off_t offset, size;
1989 	int error, noneg;
1990 	int vfslocked;
1991 
1992 	AUDIT_ARG_FD(uap->fd);
1993 	if ((error = fget(td, uap->fd, CAP_SEEK, &fp)) != 0)
1994 		return (error);
1995 	if (!(fp->f_ops->fo_flags & DFLAG_SEEKABLE)) {
1996 		fdrop(fp, td);
1997 		return (ESPIPE);
1998 	}
1999 	vp = fp->f_vnode;
2000 	vfslocked = VFS_LOCK_GIANT(vp->v_mount);
2001 	noneg = (vp->v_type != VCHR);
2002 	offset = uap->offset;
2003 	switch (uap->whence) {
2004 	case L_INCR:
2005 		if (noneg &&
2006 		    (fp->f_offset < 0 ||
2007 		    (offset > 0 && fp->f_offset > OFF_MAX - offset))) {
2008 			error = EOVERFLOW;
2009 			break;
2010 		}
2011 		offset += fp->f_offset;
2012 		break;
2013 	case L_XTND:
2014 		vn_lock(vp, LK_SHARED | LK_RETRY);
2015 		error = VOP_GETATTR(vp, &vattr, cred);
2016 		VOP_UNLOCK(vp, 0);
2017 		if (error)
2018 			break;
2019 
2020 		/*
2021 		 * If the file references a disk device, then fetch
2022 		 * the media size and use that to determine the ending
2023 		 * offset.
2024 		 */
2025 		if (vattr.va_size == 0 && vp->v_type == VCHR &&
2026 		    fo_ioctl(fp, DIOCGMEDIASIZE, &size, cred, td) == 0)
2027 			vattr.va_size = size;
2028 		if (noneg &&
2029 		    (vattr.va_size > OFF_MAX ||
2030 		    (offset > 0 && vattr.va_size > OFF_MAX - offset))) {
2031 			error = EOVERFLOW;
2032 			break;
2033 		}
2034 		offset += vattr.va_size;
2035 		break;
2036 	case L_SET:
2037 		break;
2038 	case SEEK_DATA:
2039 		error = fo_ioctl(fp, FIOSEEKDATA, &offset, cred, td);
2040 		break;
2041 	case SEEK_HOLE:
2042 		error = fo_ioctl(fp, FIOSEEKHOLE, &offset, cred, td);
2043 		break;
2044 	default:
2045 		error = EINVAL;
2046 	}
2047 	if (error == 0 && noneg && offset < 0)
2048 		error = EINVAL;
2049 	if (error != 0)
2050 		goto drop;
2051 	fp->f_offset = offset;
2052 	*(off_t *)(td->td_retval) = fp->f_offset;
2053 drop:
2054 	fdrop(fp, td);
2055 	VFS_UNLOCK_GIANT(vfslocked);
2056 	return (error);
2057 }
2058 
2059 #if defined(COMPAT_43)
2060 /*
2061  * Reposition read/write file offset.
2062  */
2063 #ifndef _SYS_SYSPROTO_H_
2064 struct olseek_args {
2065 	int	fd;
2066 	long	offset;
2067 	int	whence;
2068 };
2069 #endif
2070 int
2071 olseek(td, uap)
2072 	struct thread *td;
2073 	register struct olseek_args /* {
2074 		int fd;
2075 		long offset;
2076 		int whence;
2077 	} */ *uap;
2078 {
2079 	struct lseek_args /* {
2080 		int fd;
2081 		int pad;
2082 		off_t offset;
2083 		int whence;
2084 	} */ nuap;
2085 
2086 	nuap.fd = uap->fd;
2087 	nuap.offset = uap->offset;
2088 	nuap.whence = uap->whence;
2089 	return (sys_lseek(td, &nuap));
2090 }
2091 #endif /* COMPAT_43 */
2092 
2093 /* Version with the 'pad' argument */
2094 int
2095 freebsd6_lseek(td, uap)
2096 	struct thread *td;
2097 	register struct freebsd6_lseek_args *uap;
2098 {
2099 	struct lseek_args ouap;
2100 
2101 	ouap.fd = uap->fd;
2102 	ouap.offset = uap->offset;
2103 	ouap.whence = uap->whence;
2104 	return (sys_lseek(td, &ouap));
2105 }
2106 
2107 /*
2108  * Check access permissions using passed credentials.
2109  */
2110 static int
2111 vn_access(vp, user_flags, cred, td)
2112 	struct vnode	*vp;
2113 	int		user_flags;
2114 	struct ucred	*cred;
2115 	struct thread	*td;
2116 {
2117 	int error;
2118 	accmode_t accmode;
2119 
2120 	/* Flags == 0 means only check for existence. */
2121 	error = 0;
2122 	if (user_flags) {
2123 		accmode = 0;
2124 		if (user_flags & R_OK)
2125 			accmode |= VREAD;
2126 		if (user_flags & W_OK)
2127 			accmode |= VWRITE;
2128 		if (user_flags & X_OK)
2129 			accmode |= VEXEC;
2130 #ifdef MAC
2131 		error = mac_vnode_check_access(cred, vp, accmode);
2132 		if (error)
2133 			return (error);
2134 #endif
2135 		if ((accmode & VWRITE) == 0 || (error = vn_writechk(vp)) == 0)
2136 			error = VOP_ACCESS(vp, accmode, cred, td);
2137 	}
2138 	return (error);
2139 }
2140 
2141 /*
2142  * Check access permissions using "real" credentials.
2143  */
2144 #ifndef _SYS_SYSPROTO_H_
2145 struct access_args {
2146 	char	*path;
2147 	int	flags;
2148 };
2149 #endif
2150 int
2151 sys_access(td, uap)
2152 	struct thread *td;
2153 	register struct access_args /* {
2154 		char *path;
2155 		int flags;
2156 	} */ *uap;
2157 {
2158 
2159 	return (kern_access(td, uap->path, UIO_USERSPACE, uap->flags));
2160 }
2161 
2162 #ifndef _SYS_SYSPROTO_H_
2163 struct faccessat_args {
2164 	int	dirfd;
2165 	char	*path;
2166 	int	mode;
2167 	int	flag;
2168 }
2169 #endif
2170 int
2171 sys_faccessat(struct thread *td, struct faccessat_args *uap)
2172 {
2173 
2174 	if (uap->flag & ~AT_EACCESS)
2175 		return (EINVAL);
2176 	return (kern_accessat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
2177 	    uap->mode));
2178 }
2179 
2180 int
2181 kern_access(struct thread *td, char *path, enum uio_seg pathseg, int mode)
2182 {
2183 
2184 	return (kern_accessat(td, AT_FDCWD, path, pathseg, 0, mode));
2185 }
2186 
2187 int
2188 kern_accessat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2189     int flags, int mode)
2190 {
2191 	struct ucred *cred, *tmpcred;
2192 	struct vnode *vp;
2193 	struct nameidata nd;
2194 	int vfslocked;
2195 	int error;
2196 
2197 	/*
2198 	 * Create and modify a temporary credential instead of one that
2199 	 * is potentially shared.
2200 	 */
2201 	if (!(flags & AT_EACCESS)) {
2202 		cred = td->td_ucred;
2203 		tmpcred = crdup(cred);
2204 		tmpcred->cr_uid = cred->cr_ruid;
2205 		tmpcred->cr_groups[0] = cred->cr_rgid;
2206 		td->td_ucred = tmpcred;
2207 	} else
2208 		cred = tmpcred = td->td_ucred;
2209 	AUDIT_ARG_VALUE(mode);
2210 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | MPSAFE |
2211 	    AUDITVNODE1, pathseg, path, fd, CAP_FSTAT, td);
2212 	if ((error = namei(&nd)) != 0)
2213 		goto out1;
2214 	vfslocked = NDHASGIANT(&nd);
2215 	vp = nd.ni_vp;
2216 
2217 	error = vn_access(vp, mode, tmpcred, td);
2218 	NDFREE(&nd, NDF_ONLY_PNBUF);
2219 	vput(vp);
2220 	VFS_UNLOCK_GIANT(vfslocked);
2221 out1:
2222 	if (!(flags & AT_EACCESS)) {
2223 		td->td_ucred = cred;
2224 		crfree(tmpcred);
2225 	}
2226 	return (error);
2227 }
2228 
2229 /*
2230  * Check access permissions using "effective" credentials.
2231  */
2232 #ifndef _SYS_SYSPROTO_H_
2233 struct eaccess_args {
2234 	char	*path;
2235 	int	flags;
2236 };
2237 #endif
2238 int
2239 sys_eaccess(td, uap)
2240 	struct thread *td;
2241 	register struct eaccess_args /* {
2242 		char *path;
2243 		int flags;
2244 	} */ *uap;
2245 {
2246 
2247 	return (kern_eaccess(td, uap->path, UIO_USERSPACE, uap->flags));
2248 }
2249 
2250 int
2251 kern_eaccess(struct thread *td, char *path, enum uio_seg pathseg, int flags)
2252 {
2253 
2254 	return (kern_accessat(td, AT_FDCWD, path, pathseg, AT_EACCESS, flags));
2255 }
2256 
2257 #if defined(COMPAT_43)
2258 /*
2259  * Get file status; this version follows links.
2260  */
2261 #ifndef _SYS_SYSPROTO_H_
2262 struct ostat_args {
2263 	char	*path;
2264 	struct ostat *ub;
2265 };
2266 #endif
2267 int
2268 ostat(td, uap)
2269 	struct thread *td;
2270 	register struct ostat_args /* {
2271 		char *path;
2272 		struct ostat *ub;
2273 	} */ *uap;
2274 {
2275 	struct stat sb;
2276 	struct ostat osb;
2277 	int error;
2278 
2279 	error = kern_stat(td, uap->path, UIO_USERSPACE, &sb);
2280 	if (error)
2281 		return (error);
2282 	cvtstat(&sb, &osb);
2283 	error = copyout(&osb, uap->ub, sizeof (osb));
2284 	return (error);
2285 }
2286 
2287 /*
2288  * Get file status; this version does not follow links.
2289  */
2290 #ifndef _SYS_SYSPROTO_H_
2291 struct olstat_args {
2292 	char	*path;
2293 	struct ostat *ub;
2294 };
2295 #endif
2296 int
2297 olstat(td, uap)
2298 	struct thread *td;
2299 	register struct olstat_args /* {
2300 		char *path;
2301 		struct ostat *ub;
2302 	} */ *uap;
2303 {
2304 	struct stat sb;
2305 	struct ostat osb;
2306 	int error;
2307 
2308 	error = kern_lstat(td, uap->path, UIO_USERSPACE, &sb);
2309 	if (error)
2310 		return (error);
2311 	cvtstat(&sb, &osb);
2312 	error = copyout(&osb, uap->ub, sizeof (osb));
2313 	return (error);
2314 }
2315 
2316 /*
2317  * Convert from an old to a new stat structure.
2318  */
2319 void
2320 cvtstat(st, ost)
2321 	struct stat *st;
2322 	struct ostat *ost;
2323 {
2324 
2325 	ost->st_dev = st->st_dev;
2326 	ost->st_ino = st->st_ino;
2327 	ost->st_mode = st->st_mode;
2328 	ost->st_nlink = st->st_nlink;
2329 	ost->st_uid = st->st_uid;
2330 	ost->st_gid = st->st_gid;
2331 	ost->st_rdev = st->st_rdev;
2332 	if (st->st_size < (quad_t)1 << 32)
2333 		ost->st_size = st->st_size;
2334 	else
2335 		ost->st_size = -2;
2336 	ost->st_atim = st->st_atim;
2337 	ost->st_mtim = st->st_mtim;
2338 	ost->st_ctim = st->st_ctim;
2339 	ost->st_blksize = st->st_blksize;
2340 	ost->st_blocks = st->st_blocks;
2341 	ost->st_flags = st->st_flags;
2342 	ost->st_gen = st->st_gen;
2343 }
2344 #endif /* COMPAT_43 */
2345 
2346 /*
2347  * Get file status; this version follows links.
2348  */
2349 #ifndef _SYS_SYSPROTO_H_
2350 struct stat_args {
2351 	char	*path;
2352 	struct stat *ub;
2353 };
2354 #endif
2355 int
2356 sys_stat(td, uap)
2357 	struct thread *td;
2358 	register struct stat_args /* {
2359 		char *path;
2360 		struct stat *ub;
2361 	} */ *uap;
2362 {
2363 	struct stat sb;
2364 	int error;
2365 
2366 	error = kern_stat(td, uap->path, UIO_USERSPACE, &sb);
2367 	if (error == 0)
2368 		error = copyout(&sb, uap->ub, sizeof (sb));
2369 	return (error);
2370 }
2371 
2372 #ifndef _SYS_SYSPROTO_H_
2373 struct fstatat_args {
2374 	int	fd;
2375 	char	*path;
2376 	struct stat	*buf;
2377 	int	flag;
2378 }
2379 #endif
2380 int
2381 sys_fstatat(struct thread *td, struct fstatat_args *uap)
2382 {
2383 	struct stat sb;
2384 	int error;
2385 
2386 	error = kern_statat(td, uap->flag, uap->fd, uap->path,
2387 	    UIO_USERSPACE, &sb);
2388 	if (error == 0)
2389 		error = copyout(&sb, uap->buf, sizeof (sb));
2390 	return (error);
2391 }
2392 
2393 int
2394 kern_stat(struct thread *td, char *path, enum uio_seg pathseg, struct stat *sbp)
2395 {
2396 
2397 	return (kern_statat(td, 0, AT_FDCWD, path, pathseg, sbp));
2398 }
2399 
2400 int
2401 kern_statat(struct thread *td, int flag, int fd, char *path,
2402     enum uio_seg pathseg, struct stat *sbp)
2403 {
2404 
2405 	return (kern_statat_vnhook(td, flag, fd, path, pathseg, sbp, NULL));
2406 }
2407 
2408 int
2409 kern_statat_vnhook(struct thread *td, int flag, int fd, char *path,
2410     enum uio_seg pathseg, struct stat *sbp,
2411     void (*hook)(struct vnode *vp, struct stat *sbp))
2412 {
2413 	struct nameidata nd;
2414 	struct stat sb;
2415 	int error, vfslocked;
2416 
2417 	if (flag & ~AT_SYMLINK_NOFOLLOW)
2418 		return (EINVAL);
2419 
2420 	NDINIT_ATRIGHTS(&nd, LOOKUP, ((flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW :
2421 	    FOLLOW) | LOCKSHARED | LOCKLEAF | AUDITVNODE1 | MPSAFE, pathseg,
2422 	    path, fd, CAP_FSTAT, td);
2423 
2424 	if ((error = namei(&nd)) != 0)
2425 		return (error);
2426 	vfslocked = NDHASGIANT(&nd);
2427 	error = vn_stat(nd.ni_vp, &sb, td->td_ucred, NOCRED, td);
2428 	if (!error) {
2429 		SDT_PROBE(vfs, , stat, mode, path, sb.st_mode, 0, 0, 0);
2430 		if (S_ISREG(sb.st_mode))
2431 			SDT_PROBE(vfs, , stat, reg, path, pathseg, 0, 0, 0);
2432 		if (__predict_false(hook != NULL))
2433 			hook(nd.ni_vp, &sb);
2434 	}
2435 	NDFREE(&nd, NDF_ONLY_PNBUF);
2436 	vput(nd.ni_vp);
2437 	VFS_UNLOCK_GIANT(vfslocked);
2438 	if (error)
2439 		return (error);
2440 	*sbp = sb;
2441 #ifdef KTRACE
2442 	if (KTRPOINT(td, KTR_STRUCT))
2443 		ktrstat(&sb);
2444 #endif
2445 	return (0);
2446 }
2447 
2448 /*
2449  * Get file status; this version does not follow links.
2450  */
2451 #ifndef _SYS_SYSPROTO_H_
2452 struct lstat_args {
2453 	char	*path;
2454 	struct stat *ub;
2455 };
2456 #endif
2457 int
2458 sys_lstat(td, uap)
2459 	struct thread *td;
2460 	register struct lstat_args /* {
2461 		char *path;
2462 		struct stat *ub;
2463 	} */ *uap;
2464 {
2465 	struct stat sb;
2466 	int error;
2467 
2468 	error = kern_lstat(td, uap->path, UIO_USERSPACE, &sb);
2469 	if (error == 0)
2470 		error = copyout(&sb, uap->ub, sizeof (sb));
2471 	return (error);
2472 }
2473 
2474 int
2475 kern_lstat(struct thread *td, char *path, enum uio_seg pathseg, struct stat *sbp)
2476 {
2477 
2478 	return (kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, path, pathseg,
2479 	    sbp));
2480 }
2481 
2482 /*
2483  * Implementation of the NetBSD [l]stat() functions.
2484  */
2485 void
2486 cvtnstat(sb, nsb)
2487 	struct stat *sb;
2488 	struct nstat *nsb;
2489 {
2490 	bzero(nsb, sizeof *nsb);
2491 	nsb->st_dev = sb->st_dev;
2492 	nsb->st_ino = sb->st_ino;
2493 	nsb->st_mode = sb->st_mode;
2494 	nsb->st_nlink = sb->st_nlink;
2495 	nsb->st_uid = sb->st_uid;
2496 	nsb->st_gid = sb->st_gid;
2497 	nsb->st_rdev = sb->st_rdev;
2498 	nsb->st_atim = sb->st_atim;
2499 	nsb->st_mtim = sb->st_mtim;
2500 	nsb->st_ctim = sb->st_ctim;
2501 	nsb->st_size = sb->st_size;
2502 	nsb->st_blocks = sb->st_blocks;
2503 	nsb->st_blksize = sb->st_blksize;
2504 	nsb->st_flags = sb->st_flags;
2505 	nsb->st_gen = sb->st_gen;
2506 	nsb->st_birthtim = sb->st_birthtim;
2507 }
2508 
2509 #ifndef _SYS_SYSPROTO_H_
2510 struct nstat_args {
2511 	char	*path;
2512 	struct nstat *ub;
2513 };
2514 #endif
2515 int
2516 sys_nstat(td, uap)
2517 	struct thread *td;
2518 	register struct nstat_args /* {
2519 		char *path;
2520 		struct nstat *ub;
2521 	} */ *uap;
2522 {
2523 	struct stat sb;
2524 	struct nstat nsb;
2525 	int error;
2526 
2527 	error = kern_stat(td, uap->path, UIO_USERSPACE, &sb);
2528 	if (error)
2529 		return (error);
2530 	cvtnstat(&sb, &nsb);
2531 	error = copyout(&nsb, uap->ub, sizeof (nsb));
2532 	return (error);
2533 }
2534 
2535 /*
2536  * NetBSD lstat.  Get file status; this version does not follow links.
2537  */
2538 #ifndef _SYS_SYSPROTO_H_
2539 struct lstat_args {
2540 	char	*path;
2541 	struct stat *ub;
2542 };
2543 #endif
2544 int
2545 sys_nlstat(td, uap)
2546 	struct thread *td;
2547 	register struct nlstat_args /* {
2548 		char *path;
2549 		struct nstat *ub;
2550 	} */ *uap;
2551 {
2552 	struct stat sb;
2553 	struct nstat nsb;
2554 	int error;
2555 
2556 	error = kern_lstat(td, uap->path, UIO_USERSPACE, &sb);
2557 	if (error)
2558 		return (error);
2559 	cvtnstat(&sb, &nsb);
2560 	error = copyout(&nsb, uap->ub, sizeof (nsb));
2561 	return (error);
2562 }
2563 
2564 /*
2565  * Get configurable pathname variables.
2566  */
2567 #ifndef _SYS_SYSPROTO_H_
2568 struct pathconf_args {
2569 	char	*path;
2570 	int	name;
2571 };
2572 #endif
2573 int
2574 sys_pathconf(td, uap)
2575 	struct thread *td;
2576 	register struct pathconf_args /* {
2577 		char *path;
2578 		int name;
2579 	} */ *uap;
2580 {
2581 
2582 	return (kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name, FOLLOW));
2583 }
2584 
2585 #ifndef _SYS_SYSPROTO_H_
2586 struct lpathconf_args {
2587 	char	*path;
2588 	int	name;
2589 };
2590 #endif
2591 int
2592 sys_lpathconf(td, uap)
2593 	struct thread *td;
2594 	register struct lpathconf_args /* {
2595 		char *path;
2596 		int name;
2597 	} */ *uap;
2598 {
2599 
2600 	return (kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name, NOFOLLOW));
2601 }
2602 
2603 int
2604 kern_pathconf(struct thread *td, char *path, enum uio_seg pathseg, int name,
2605     u_long flags)
2606 {
2607 	struct nameidata nd;
2608 	int error, vfslocked;
2609 
2610 	NDINIT(&nd, LOOKUP, LOCKSHARED | LOCKLEAF | MPSAFE | AUDITVNODE1 |
2611 	    flags, pathseg, path, td);
2612 	if ((error = namei(&nd)) != 0)
2613 		return (error);
2614 	vfslocked = NDHASGIANT(&nd);
2615 	NDFREE(&nd, NDF_ONLY_PNBUF);
2616 
2617 	/* If asynchronous I/O is available, it works for all files. */
2618 	if (name == _PC_ASYNC_IO)
2619 		td->td_retval[0] = async_io_version;
2620 	else
2621 		error = VOP_PATHCONF(nd.ni_vp, name, td->td_retval);
2622 	vput(nd.ni_vp);
2623 	VFS_UNLOCK_GIANT(vfslocked);
2624 	return (error);
2625 }
2626 
2627 /*
2628  * Return target name of a symbolic link.
2629  */
2630 #ifndef _SYS_SYSPROTO_H_
2631 struct readlink_args {
2632 	char	*path;
2633 	char	*buf;
2634 	size_t	count;
2635 };
2636 #endif
2637 int
2638 sys_readlink(td, uap)
2639 	struct thread *td;
2640 	register struct readlink_args /* {
2641 		char *path;
2642 		char *buf;
2643 		size_t count;
2644 	} */ *uap;
2645 {
2646 
2647 	return (kern_readlink(td, uap->path, UIO_USERSPACE, uap->buf,
2648 	    UIO_USERSPACE, uap->count));
2649 }
2650 #ifndef _SYS_SYSPROTO_H_
2651 struct readlinkat_args {
2652 	int	fd;
2653 	char	*path;
2654 	char	*buf;
2655 	size_t	bufsize;
2656 };
2657 #endif
2658 int
2659 sys_readlinkat(struct thread *td, struct readlinkat_args *uap)
2660 {
2661 
2662 	return (kern_readlinkat(td, uap->fd, uap->path, UIO_USERSPACE,
2663 	    uap->buf, UIO_USERSPACE, uap->bufsize));
2664 }
2665 
2666 int
2667 kern_readlink(struct thread *td, char *path, enum uio_seg pathseg, char *buf,
2668     enum uio_seg bufseg, size_t count)
2669 {
2670 
2671 	return (kern_readlinkat(td, AT_FDCWD, path, pathseg, buf, bufseg,
2672 	    count));
2673 }
2674 
2675 int
2676 kern_readlinkat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2677     char *buf, enum uio_seg bufseg, size_t count)
2678 {
2679 	struct vnode *vp;
2680 	struct iovec aiov;
2681 	struct uio auio;
2682 	int error;
2683 	struct nameidata nd;
2684 	int vfslocked;
2685 
2686 	if (count > INT_MAX)
2687 		return (EINVAL);
2688 
2689 	NDINIT_AT(&nd, LOOKUP, NOFOLLOW | LOCKSHARED | LOCKLEAF | MPSAFE |
2690 	    AUDITVNODE1, pathseg, path, fd, td);
2691 
2692 	if ((error = namei(&nd)) != 0)
2693 		return (error);
2694 	NDFREE(&nd, NDF_ONLY_PNBUF);
2695 	vfslocked = NDHASGIANT(&nd);
2696 	vp = nd.ni_vp;
2697 #ifdef MAC
2698 	error = mac_vnode_check_readlink(td->td_ucred, vp);
2699 	if (error) {
2700 		vput(vp);
2701 		VFS_UNLOCK_GIANT(vfslocked);
2702 		return (error);
2703 	}
2704 #endif
2705 	if (vp->v_type != VLNK)
2706 		error = EINVAL;
2707 	else {
2708 		aiov.iov_base = buf;
2709 		aiov.iov_len = count;
2710 		auio.uio_iov = &aiov;
2711 		auio.uio_iovcnt = 1;
2712 		auio.uio_offset = 0;
2713 		auio.uio_rw = UIO_READ;
2714 		auio.uio_segflg = bufseg;
2715 		auio.uio_td = td;
2716 		auio.uio_resid = count;
2717 		error = VOP_READLINK(vp, &auio, td->td_ucred);
2718 	}
2719 	vput(vp);
2720 	VFS_UNLOCK_GIANT(vfslocked);
2721 	td->td_retval[0] = count - auio.uio_resid;
2722 	return (error);
2723 }
2724 
2725 /*
2726  * Common implementation code for chflags() and fchflags().
2727  */
2728 static int
2729 setfflags(td, vp, flags)
2730 	struct thread *td;
2731 	struct vnode *vp;
2732 	int flags;
2733 {
2734 	int error;
2735 	struct mount *mp;
2736 	struct vattr vattr;
2737 
2738 	/*
2739 	 * Prevent non-root users from setting flags on devices.  When
2740 	 * a device is reused, users can retain ownership of the device
2741 	 * if they are allowed to set flags and programs assume that
2742 	 * chown can't fail when done as root.
2743 	 */
2744 	if (vp->v_type == VCHR || vp->v_type == VBLK) {
2745 		error = priv_check(td, PRIV_VFS_CHFLAGS_DEV);
2746 		if (error)
2747 			return (error);
2748 	}
2749 
2750 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2751 		return (error);
2752 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2753 	VATTR_NULL(&vattr);
2754 	vattr.va_flags = flags;
2755 #ifdef MAC
2756 	error = mac_vnode_check_setflags(td->td_ucred, vp, vattr.va_flags);
2757 	if (error == 0)
2758 #endif
2759 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
2760 	VOP_UNLOCK(vp, 0);
2761 	vn_finished_write(mp);
2762 	return (error);
2763 }
2764 
2765 /*
2766  * Change flags of a file given a path name.
2767  */
2768 #ifndef _SYS_SYSPROTO_H_
2769 struct chflags_args {
2770 	char	*path;
2771 	int	flags;
2772 };
2773 #endif
2774 int
2775 sys_chflags(td, uap)
2776 	struct thread *td;
2777 	register struct chflags_args /* {
2778 		char *path;
2779 		int flags;
2780 	} */ *uap;
2781 {
2782 	int error;
2783 	struct nameidata nd;
2784 	int vfslocked;
2785 
2786 	AUDIT_ARG_FFLAGS(uap->flags);
2787 	NDINIT(&nd, LOOKUP, FOLLOW | MPSAFE | AUDITVNODE1, UIO_USERSPACE,
2788 	    uap->path, td);
2789 	if ((error = namei(&nd)) != 0)
2790 		return (error);
2791 	NDFREE(&nd, NDF_ONLY_PNBUF);
2792 	vfslocked = NDHASGIANT(&nd);
2793 	error = setfflags(td, nd.ni_vp, uap->flags);
2794 	vrele(nd.ni_vp);
2795 	VFS_UNLOCK_GIANT(vfslocked);
2796 	return (error);
2797 }
2798 
2799 /*
2800  * Same as chflags() but doesn't follow symlinks.
2801  */
2802 int
2803 sys_lchflags(td, uap)
2804 	struct thread *td;
2805 	register struct lchflags_args /* {
2806 		char *path;
2807 		int flags;
2808 	} */ *uap;
2809 {
2810 	int error;
2811 	struct nameidata nd;
2812 	int vfslocked;
2813 
2814 	AUDIT_ARG_FFLAGS(uap->flags);
2815 	NDINIT(&nd, LOOKUP, NOFOLLOW | MPSAFE | AUDITVNODE1, UIO_USERSPACE,
2816 	    uap->path, td);
2817 	if ((error = namei(&nd)) != 0)
2818 		return (error);
2819 	vfslocked = NDHASGIANT(&nd);
2820 	NDFREE(&nd, NDF_ONLY_PNBUF);
2821 	error = setfflags(td, nd.ni_vp, uap->flags);
2822 	vrele(nd.ni_vp);
2823 	VFS_UNLOCK_GIANT(vfslocked);
2824 	return (error);
2825 }
2826 
2827 /*
2828  * Change flags of a file given a file descriptor.
2829  */
2830 #ifndef _SYS_SYSPROTO_H_
2831 struct fchflags_args {
2832 	int	fd;
2833 	int	flags;
2834 };
2835 #endif
2836 int
2837 sys_fchflags(td, uap)
2838 	struct thread *td;
2839 	register struct fchflags_args /* {
2840 		int fd;
2841 		int flags;
2842 	} */ *uap;
2843 {
2844 	struct file *fp;
2845 	int vfslocked;
2846 	int error;
2847 
2848 	AUDIT_ARG_FD(uap->fd);
2849 	AUDIT_ARG_FFLAGS(uap->flags);
2850 	if ((error = getvnode(td->td_proc->p_fd, uap->fd, CAP_FCHFLAGS,
2851 	    &fp)) != 0)
2852 		return (error);
2853 	vfslocked = VFS_LOCK_GIANT(fp->f_vnode->v_mount);
2854 #ifdef AUDIT
2855 	vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
2856 	AUDIT_ARG_VNODE1(fp->f_vnode);
2857 	VOP_UNLOCK(fp->f_vnode, 0);
2858 #endif
2859 	error = setfflags(td, fp->f_vnode, uap->flags);
2860 	VFS_UNLOCK_GIANT(vfslocked);
2861 	fdrop(fp, td);
2862 	return (error);
2863 }
2864 
2865 /*
2866  * Common implementation code for chmod(), lchmod() and fchmod().
2867  */
2868 int
2869 setfmode(td, cred, vp, mode)
2870 	struct thread *td;
2871 	struct ucred *cred;
2872 	struct vnode *vp;
2873 	int mode;
2874 {
2875 	int error;
2876 	struct mount *mp;
2877 	struct vattr vattr;
2878 
2879 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2880 		return (error);
2881 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2882 	VATTR_NULL(&vattr);
2883 	vattr.va_mode = mode & ALLPERMS;
2884 #ifdef MAC
2885 	error = mac_vnode_check_setmode(cred, vp, vattr.va_mode);
2886 	if (error == 0)
2887 #endif
2888 		error = VOP_SETATTR(vp, &vattr, cred);
2889 	VOP_UNLOCK(vp, 0);
2890 	vn_finished_write(mp);
2891 	return (error);
2892 }
2893 
2894 /*
2895  * Change mode of a file given path name.
2896  */
2897 #ifndef _SYS_SYSPROTO_H_
2898 struct chmod_args {
2899 	char	*path;
2900 	int	mode;
2901 };
2902 #endif
2903 int
2904 sys_chmod(td, uap)
2905 	struct thread *td;
2906 	register struct chmod_args /* {
2907 		char *path;
2908 		int mode;
2909 	} */ *uap;
2910 {
2911 
2912 	return (kern_chmod(td, uap->path, UIO_USERSPACE, uap->mode));
2913 }
2914 
2915 #ifndef _SYS_SYSPROTO_H_
2916 struct fchmodat_args {
2917 	int	dirfd;
2918 	char	*path;
2919 	mode_t	mode;
2920 	int	flag;
2921 }
2922 #endif
2923 int
2924 sys_fchmodat(struct thread *td, struct fchmodat_args *uap)
2925 {
2926 	int flag = uap->flag;
2927 	int fd = uap->fd;
2928 	char *path = uap->path;
2929 	mode_t mode = uap->mode;
2930 
2931 	if (flag & ~AT_SYMLINK_NOFOLLOW)
2932 		return (EINVAL);
2933 
2934 	return (kern_fchmodat(td, fd, path, UIO_USERSPACE, mode, flag));
2935 }
2936 
2937 int
2938 kern_chmod(struct thread *td, char *path, enum uio_seg pathseg, int mode)
2939 {
2940 
2941 	return (kern_fchmodat(td, AT_FDCWD, path, pathseg, mode, 0));
2942 }
2943 
2944 /*
2945  * Change mode of a file given path name (don't follow links.)
2946  */
2947 #ifndef _SYS_SYSPROTO_H_
2948 struct lchmod_args {
2949 	char	*path;
2950 	int	mode;
2951 };
2952 #endif
2953 int
2954 sys_lchmod(td, uap)
2955 	struct thread *td;
2956 	register struct lchmod_args /* {
2957 		char *path;
2958 		int mode;
2959 	} */ *uap;
2960 {
2961 
2962 	return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2963 	    uap->mode, AT_SYMLINK_NOFOLLOW));
2964 }
2965 
2966 
2967 int
2968 kern_fchmodat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2969     mode_t mode, int flag)
2970 {
2971 	int error;
2972 	struct nameidata nd;
2973 	int vfslocked;
2974 	int follow;
2975 
2976 	AUDIT_ARG_MODE(mode);
2977 	follow = (flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
2978 	NDINIT_ATRIGHTS(&nd, LOOKUP,  follow | MPSAFE | AUDITVNODE1, pathseg,
2979 	    path, fd, CAP_FCHMOD, td);
2980 	if ((error = namei(&nd)) != 0)
2981 		return (error);
2982 	vfslocked = NDHASGIANT(&nd);
2983 	NDFREE(&nd, NDF_ONLY_PNBUF);
2984 	error = setfmode(td, td->td_ucred, nd.ni_vp, mode);
2985 	vrele(nd.ni_vp);
2986 	VFS_UNLOCK_GIANT(vfslocked);
2987 	return (error);
2988 }
2989 
2990 /*
2991  * Change mode of a file given a file descriptor.
2992  */
2993 #ifndef _SYS_SYSPROTO_H_
2994 struct fchmod_args {
2995 	int	fd;
2996 	int	mode;
2997 };
2998 #endif
2999 int
3000 sys_fchmod(struct thread *td, struct fchmod_args *uap)
3001 {
3002 	struct file *fp;
3003 	int error;
3004 
3005 	AUDIT_ARG_FD(uap->fd);
3006 	AUDIT_ARG_MODE(uap->mode);
3007 
3008 	error = fget(td, uap->fd, CAP_FCHMOD, &fp);
3009 	if (error != 0)
3010 		return (error);
3011 	error = fo_chmod(fp, uap->mode, td->td_ucred, td);
3012 	fdrop(fp, td);
3013 	return (error);
3014 }
3015 
3016 /*
3017  * Common implementation for chown(), lchown(), and fchown()
3018  */
3019 int
3020 setfown(td, cred, vp, uid, gid)
3021 	struct thread *td;
3022 	struct ucred *cred;
3023 	struct vnode *vp;
3024 	uid_t uid;
3025 	gid_t gid;
3026 {
3027 	int error;
3028 	struct mount *mp;
3029 	struct vattr vattr;
3030 
3031 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
3032 		return (error);
3033 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3034 	VATTR_NULL(&vattr);
3035 	vattr.va_uid = uid;
3036 	vattr.va_gid = gid;
3037 #ifdef MAC
3038 	error = mac_vnode_check_setowner(cred, vp, vattr.va_uid,
3039 	    vattr.va_gid);
3040 	if (error == 0)
3041 #endif
3042 		error = VOP_SETATTR(vp, &vattr, cred);
3043 	VOP_UNLOCK(vp, 0);
3044 	vn_finished_write(mp);
3045 	return (error);
3046 }
3047 
3048 /*
3049  * Set ownership given a path name.
3050  */
3051 #ifndef _SYS_SYSPROTO_H_
3052 struct chown_args {
3053 	char	*path;
3054 	int	uid;
3055 	int	gid;
3056 };
3057 #endif
3058 int
3059 sys_chown(td, uap)
3060 	struct thread *td;
3061 	register struct chown_args /* {
3062 		char *path;
3063 		int uid;
3064 		int gid;
3065 	} */ *uap;
3066 {
3067 
3068 	return (kern_chown(td, uap->path, UIO_USERSPACE, uap->uid, uap->gid));
3069 }
3070 
3071 #ifndef _SYS_SYSPROTO_H_
3072 struct fchownat_args {
3073 	int fd;
3074 	const char * path;
3075 	uid_t uid;
3076 	gid_t gid;
3077 	int flag;
3078 };
3079 #endif
3080 int
3081 sys_fchownat(struct thread *td, struct fchownat_args *uap)
3082 {
3083 	int flag;
3084 
3085 	flag = uap->flag;
3086 	if (flag & ~AT_SYMLINK_NOFOLLOW)
3087 		return (EINVAL);
3088 
3089 	return (kern_fchownat(td, uap->fd, uap->path, UIO_USERSPACE, uap->uid,
3090 	    uap->gid, uap->flag));
3091 }
3092 
3093 int
3094 kern_chown(struct thread *td, char *path, enum uio_seg pathseg, int uid,
3095     int gid)
3096 {
3097 
3098 	return (kern_fchownat(td, AT_FDCWD, path, pathseg, uid, gid, 0));
3099 }
3100 
3101 int
3102 kern_fchownat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
3103     int uid, int gid, int flag)
3104 {
3105 	struct nameidata nd;
3106 	int error, vfslocked, follow;
3107 
3108 	AUDIT_ARG_OWNER(uid, gid);
3109 	follow = (flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
3110 	NDINIT_ATRIGHTS(&nd, LOOKUP, follow | MPSAFE | AUDITVNODE1, pathseg,
3111 	    path, fd, CAP_FCHOWN, td);
3112 
3113 	if ((error = namei(&nd)) != 0)
3114 		return (error);
3115 	vfslocked = NDHASGIANT(&nd);
3116 	NDFREE(&nd, NDF_ONLY_PNBUF);
3117 	error = setfown(td, td->td_ucred, nd.ni_vp, uid, gid);
3118 	vrele(nd.ni_vp);
3119 	VFS_UNLOCK_GIANT(vfslocked);
3120 	return (error);
3121 }
3122 
3123 /*
3124  * Set ownership given a path name, do not cross symlinks.
3125  */
3126 #ifndef _SYS_SYSPROTO_H_
3127 struct lchown_args {
3128 	char	*path;
3129 	int	uid;
3130 	int	gid;
3131 };
3132 #endif
3133 int
3134 sys_lchown(td, uap)
3135 	struct thread *td;
3136 	register struct lchown_args /* {
3137 		char *path;
3138 		int uid;
3139 		int gid;
3140 	} */ *uap;
3141 {
3142 
3143 	return (kern_lchown(td, uap->path, UIO_USERSPACE, uap->uid, uap->gid));
3144 }
3145 
3146 int
3147 kern_lchown(struct thread *td, char *path, enum uio_seg pathseg, int uid,
3148     int gid)
3149 {
3150 
3151 	return (kern_fchownat(td, AT_FDCWD, path, pathseg, uid, gid,
3152 	    AT_SYMLINK_NOFOLLOW));
3153 }
3154 
3155 /*
3156  * Set ownership given a file descriptor.
3157  */
3158 #ifndef _SYS_SYSPROTO_H_
3159 struct fchown_args {
3160 	int	fd;
3161 	int	uid;
3162 	int	gid;
3163 };
3164 #endif
3165 int
3166 sys_fchown(td, uap)
3167 	struct thread *td;
3168 	register struct fchown_args /* {
3169 		int fd;
3170 		int uid;
3171 		int gid;
3172 	} */ *uap;
3173 {
3174 	struct file *fp;
3175 	int error;
3176 
3177 	AUDIT_ARG_FD(uap->fd);
3178 	AUDIT_ARG_OWNER(uap->uid, uap->gid);
3179 	error = fget(td, uap->fd, CAP_FCHOWN, &fp);
3180 	if (error != 0)
3181 		return (error);
3182 	error = fo_chown(fp, uap->uid, uap->gid, td->td_ucred, td);
3183 	fdrop(fp, td);
3184 	return (error);
3185 }
3186 
3187 /*
3188  * Common implementation code for utimes(), lutimes(), and futimes().
3189  */
3190 static int
3191 getutimes(usrtvp, tvpseg, tsp)
3192 	const struct timeval *usrtvp;
3193 	enum uio_seg tvpseg;
3194 	struct timespec *tsp;
3195 {
3196 	struct timeval tv[2];
3197 	const struct timeval *tvp;
3198 	int error;
3199 
3200 	if (usrtvp == NULL) {
3201 		vfs_timestamp(&tsp[0]);
3202 		tsp[1] = tsp[0];
3203 	} else {
3204 		if (tvpseg == UIO_SYSSPACE) {
3205 			tvp = usrtvp;
3206 		} else {
3207 			if ((error = copyin(usrtvp, tv, sizeof(tv))) != 0)
3208 				return (error);
3209 			tvp = tv;
3210 		}
3211 
3212 		if (tvp[0].tv_usec < 0 || tvp[0].tv_usec >= 1000000 ||
3213 		    tvp[1].tv_usec < 0 || tvp[1].tv_usec >= 1000000)
3214 			return (EINVAL);
3215 		TIMEVAL_TO_TIMESPEC(&tvp[0], &tsp[0]);
3216 		TIMEVAL_TO_TIMESPEC(&tvp[1], &tsp[1]);
3217 	}
3218 	return (0);
3219 }
3220 
3221 /*
3222  * Common implementation code for utimes(), lutimes(), and futimes().
3223  */
3224 static int
3225 setutimes(td, vp, ts, numtimes, nullflag)
3226 	struct thread *td;
3227 	struct vnode *vp;
3228 	const struct timespec *ts;
3229 	int numtimes;
3230 	int nullflag;
3231 {
3232 	int error, setbirthtime;
3233 	struct mount *mp;
3234 	struct vattr vattr;
3235 
3236 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
3237 		return (error);
3238 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3239 	setbirthtime = 0;
3240 	if (numtimes < 3 && !VOP_GETATTR(vp, &vattr, td->td_ucred) &&
3241 	    timespeccmp(&ts[1], &vattr.va_birthtime, < ))
3242 		setbirthtime = 1;
3243 	VATTR_NULL(&vattr);
3244 	vattr.va_atime = ts[0];
3245 	vattr.va_mtime = ts[1];
3246 	if (setbirthtime)
3247 		vattr.va_birthtime = ts[1];
3248 	if (numtimes > 2)
3249 		vattr.va_birthtime = ts[2];
3250 	if (nullflag)
3251 		vattr.va_vaflags |= VA_UTIMES_NULL;
3252 #ifdef MAC
3253 	error = mac_vnode_check_setutimes(td->td_ucred, vp, vattr.va_atime,
3254 	    vattr.va_mtime);
3255 #endif
3256 	if (error == 0)
3257 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3258 	VOP_UNLOCK(vp, 0);
3259 	vn_finished_write(mp);
3260 	return (error);
3261 }
3262 
3263 /*
3264  * Set the access and modification times of a file.
3265  */
3266 #ifndef _SYS_SYSPROTO_H_
3267 struct utimes_args {
3268 	char	*path;
3269 	struct	timeval *tptr;
3270 };
3271 #endif
3272 int
3273 sys_utimes(td, uap)
3274 	struct thread *td;
3275 	register struct utimes_args /* {
3276 		char *path;
3277 		struct timeval *tptr;
3278 	} */ *uap;
3279 {
3280 
3281 	return (kern_utimes(td, uap->path, UIO_USERSPACE, uap->tptr,
3282 	    UIO_USERSPACE));
3283 }
3284 
3285 #ifndef _SYS_SYSPROTO_H_
3286 struct futimesat_args {
3287 	int fd;
3288 	const char * path;
3289 	const struct timeval * times;
3290 };
3291 #endif
3292 int
3293 sys_futimesat(struct thread *td, struct futimesat_args *uap)
3294 {
3295 
3296 	return (kern_utimesat(td, uap->fd, uap->path, UIO_USERSPACE,
3297 	    uap->times, UIO_USERSPACE));
3298 }
3299 
3300 int
3301 kern_utimes(struct thread *td, char *path, enum uio_seg pathseg,
3302     struct timeval *tptr, enum uio_seg tptrseg)
3303 {
3304 
3305 	return (kern_utimesat(td, AT_FDCWD, path, pathseg, tptr, tptrseg));
3306 }
3307 
3308 int
3309 kern_utimesat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
3310     struct timeval *tptr, enum uio_seg tptrseg)
3311 {
3312 	struct nameidata nd;
3313 	struct timespec ts[2];
3314 	int error, vfslocked;
3315 
3316 	if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3317 		return (error);
3318 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | MPSAFE | AUDITVNODE1, pathseg,
3319 	    path, fd, CAP_FUTIMES, td);
3320 
3321 	if ((error = namei(&nd)) != 0)
3322 		return (error);
3323 	vfslocked = NDHASGIANT(&nd);
3324 	NDFREE(&nd, NDF_ONLY_PNBUF);
3325 	error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3326 	vrele(nd.ni_vp);
3327 	VFS_UNLOCK_GIANT(vfslocked);
3328 	return (error);
3329 }
3330 
3331 /*
3332  * Set the access and modification times of a file.
3333  */
3334 #ifndef _SYS_SYSPROTO_H_
3335 struct lutimes_args {
3336 	char	*path;
3337 	struct	timeval *tptr;
3338 };
3339 #endif
3340 int
3341 sys_lutimes(td, uap)
3342 	struct thread *td;
3343 	register struct lutimes_args /* {
3344 		char *path;
3345 		struct timeval *tptr;
3346 	} */ *uap;
3347 {
3348 
3349 	return (kern_lutimes(td, uap->path, UIO_USERSPACE, uap->tptr,
3350 	    UIO_USERSPACE));
3351 }
3352 
3353 int
3354 kern_lutimes(struct thread *td, char *path, enum uio_seg pathseg,
3355     struct timeval *tptr, enum uio_seg tptrseg)
3356 {
3357 	struct timespec ts[2];
3358 	int error;
3359 	struct nameidata nd;
3360 	int vfslocked;
3361 
3362 	if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3363 		return (error);
3364 	NDINIT(&nd, LOOKUP, NOFOLLOW | MPSAFE | AUDITVNODE1, pathseg, path, td);
3365 	if ((error = namei(&nd)) != 0)
3366 		return (error);
3367 	vfslocked = NDHASGIANT(&nd);
3368 	NDFREE(&nd, NDF_ONLY_PNBUF);
3369 	error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3370 	vrele(nd.ni_vp);
3371 	VFS_UNLOCK_GIANT(vfslocked);
3372 	return (error);
3373 }
3374 
3375 /*
3376  * Set the access and modification times of a file.
3377  */
3378 #ifndef _SYS_SYSPROTO_H_
3379 struct futimes_args {
3380 	int	fd;
3381 	struct	timeval *tptr;
3382 };
3383 #endif
3384 int
3385 sys_futimes(td, uap)
3386 	struct thread *td;
3387 	register struct futimes_args /* {
3388 		int  fd;
3389 		struct timeval *tptr;
3390 	} */ *uap;
3391 {
3392 
3393 	return (kern_futimes(td, uap->fd, uap->tptr, UIO_USERSPACE));
3394 }
3395 
3396 int
3397 kern_futimes(struct thread *td, int fd, struct timeval *tptr,
3398     enum uio_seg tptrseg)
3399 {
3400 	struct timespec ts[2];
3401 	struct file *fp;
3402 	int vfslocked;
3403 	int error;
3404 
3405 	AUDIT_ARG_FD(fd);
3406 	if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3407 		return (error);
3408 	if ((error = getvnode(td->td_proc->p_fd, fd, CAP_FUTIMES, &fp))
3409 	    != 0)
3410 		return (error);
3411 	vfslocked = VFS_LOCK_GIANT(fp->f_vnode->v_mount);
3412 #ifdef AUDIT
3413 	vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
3414 	AUDIT_ARG_VNODE1(fp->f_vnode);
3415 	VOP_UNLOCK(fp->f_vnode, 0);
3416 #endif
3417 	error = setutimes(td, fp->f_vnode, ts, 2, tptr == NULL);
3418 	VFS_UNLOCK_GIANT(vfslocked);
3419 	fdrop(fp, td);
3420 	return (error);
3421 }
3422 
3423 /*
3424  * Truncate a file given its path name.
3425  */
3426 #ifndef _SYS_SYSPROTO_H_
3427 struct truncate_args {
3428 	char	*path;
3429 	int	pad;
3430 	off_t	length;
3431 };
3432 #endif
3433 int
3434 sys_truncate(td, uap)
3435 	struct thread *td;
3436 	register struct truncate_args /* {
3437 		char *path;
3438 		int pad;
3439 		off_t length;
3440 	} */ *uap;
3441 {
3442 
3443 	return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3444 }
3445 
3446 int
3447 kern_truncate(struct thread *td, char *path, enum uio_seg pathseg, off_t length)
3448 {
3449 	struct mount *mp;
3450 	struct vnode *vp;
3451 	struct vattr vattr;
3452 	int error;
3453 	struct nameidata nd;
3454 	int vfslocked;
3455 
3456 	if (length < 0)
3457 		return(EINVAL);
3458 	NDINIT(&nd, LOOKUP, FOLLOW | MPSAFE | AUDITVNODE1, pathseg, path, td);
3459 	if ((error = namei(&nd)) != 0)
3460 		return (error);
3461 	vfslocked = NDHASGIANT(&nd);
3462 	vp = nd.ni_vp;
3463 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) {
3464 		vrele(vp);
3465 		VFS_UNLOCK_GIANT(vfslocked);
3466 		return (error);
3467 	}
3468 	NDFREE(&nd, NDF_ONLY_PNBUF);
3469 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3470 	if (vp->v_type == VDIR)
3471 		error = EISDIR;
3472 #ifdef MAC
3473 	else if ((error = mac_vnode_check_write(td->td_ucred, NOCRED, vp))) {
3474 	}
3475 #endif
3476 	else if ((error = vn_writechk(vp)) == 0 &&
3477 	    (error = VOP_ACCESS(vp, VWRITE, td->td_ucred, td)) == 0) {
3478 		VATTR_NULL(&vattr);
3479 		vattr.va_size = length;
3480 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3481 	}
3482 	vput(vp);
3483 	vn_finished_write(mp);
3484 	VFS_UNLOCK_GIANT(vfslocked);
3485 	return (error);
3486 }
3487 
3488 #if defined(COMPAT_43)
3489 /*
3490  * Truncate a file given its path name.
3491  */
3492 #ifndef _SYS_SYSPROTO_H_
3493 struct otruncate_args {
3494 	char	*path;
3495 	long	length;
3496 };
3497 #endif
3498 int
3499 otruncate(td, uap)
3500 	struct thread *td;
3501 	register struct otruncate_args /* {
3502 		char *path;
3503 		long length;
3504 	} */ *uap;
3505 {
3506 	struct truncate_args /* {
3507 		char *path;
3508 		int pad;
3509 		off_t length;
3510 	} */ nuap;
3511 
3512 	nuap.path = uap->path;
3513 	nuap.length = uap->length;
3514 	return (sys_truncate(td, &nuap));
3515 }
3516 #endif /* COMPAT_43 */
3517 
3518 /* Versions with the pad argument */
3519 int
3520 freebsd6_truncate(struct thread *td, struct freebsd6_truncate_args *uap)
3521 {
3522 	struct truncate_args ouap;
3523 
3524 	ouap.path = uap->path;
3525 	ouap.length = uap->length;
3526 	return (sys_truncate(td, &ouap));
3527 }
3528 
3529 int
3530 freebsd6_ftruncate(struct thread *td, struct freebsd6_ftruncate_args *uap)
3531 {
3532 	struct ftruncate_args ouap;
3533 
3534 	ouap.fd = uap->fd;
3535 	ouap.length = uap->length;
3536 	return (sys_ftruncate(td, &ouap));
3537 }
3538 
3539 /*
3540  * Sync an open file.
3541  */
3542 #ifndef _SYS_SYSPROTO_H_
3543 struct fsync_args {
3544 	int	fd;
3545 };
3546 #endif
3547 int
3548 sys_fsync(td, uap)
3549 	struct thread *td;
3550 	struct fsync_args /* {
3551 		int fd;
3552 	} */ *uap;
3553 {
3554 	struct vnode *vp;
3555 	struct mount *mp;
3556 	struct file *fp;
3557 	int vfslocked;
3558 	int error, lock_flags;
3559 
3560 	AUDIT_ARG_FD(uap->fd);
3561 	if ((error = getvnode(td->td_proc->p_fd, uap->fd, CAP_FSYNC,
3562 	    &fp)) != 0)
3563 		return (error);
3564 	vp = fp->f_vnode;
3565 	vfslocked = VFS_LOCK_GIANT(vp->v_mount);
3566 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
3567 		goto drop;
3568 	if (MNT_SHARED_WRITES(mp) ||
3569 	    ((mp == NULL) && MNT_SHARED_WRITES(vp->v_mount))) {
3570 		lock_flags = LK_SHARED;
3571 	} else {
3572 		lock_flags = LK_EXCLUSIVE;
3573 	}
3574 	vn_lock(vp, lock_flags | LK_RETRY);
3575 	AUDIT_ARG_VNODE1(vp);
3576 	if (vp->v_object != NULL) {
3577 		VM_OBJECT_LOCK(vp->v_object);
3578 		vm_object_page_clean(vp->v_object, 0, 0, 0);
3579 		VM_OBJECT_UNLOCK(vp->v_object);
3580 	}
3581 	error = VOP_FSYNC(vp, MNT_WAIT, td);
3582 
3583 	VOP_UNLOCK(vp, 0);
3584 	vn_finished_write(mp);
3585 drop:
3586 	VFS_UNLOCK_GIANT(vfslocked);
3587 	fdrop(fp, td);
3588 	return (error);
3589 }
3590 
3591 /*
3592  * Rename files.  Source and destination must either both be directories, or
3593  * both not be directories.  If target is a directory, it must be empty.
3594  */
3595 #ifndef _SYS_SYSPROTO_H_
3596 struct rename_args {
3597 	char	*from;
3598 	char	*to;
3599 };
3600 #endif
3601 int
3602 sys_rename(td, uap)
3603 	struct thread *td;
3604 	register struct rename_args /* {
3605 		char *from;
3606 		char *to;
3607 	} */ *uap;
3608 {
3609 
3610 	return (kern_rename(td, uap->from, uap->to, UIO_USERSPACE));
3611 }
3612 
3613 #ifndef _SYS_SYSPROTO_H_
3614 struct renameat_args {
3615 	int	oldfd;
3616 	char	*old;
3617 	int	newfd;
3618 	char	*new;
3619 };
3620 #endif
3621 int
3622 sys_renameat(struct thread *td, struct renameat_args *uap)
3623 {
3624 
3625 	return (kern_renameat(td, uap->oldfd, uap->old, uap->newfd, uap->new,
3626 	    UIO_USERSPACE));
3627 }
3628 
3629 int
3630 kern_rename(struct thread *td, char *from, char *to, enum uio_seg pathseg)
3631 {
3632 
3633 	return (kern_renameat(td, AT_FDCWD, from, AT_FDCWD, to, pathseg));
3634 }
3635 
3636 int
3637 kern_renameat(struct thread *td, int oldfd, char *old, int newfd, char *new,
3638     enum uio_seg pathseg)
3639 {
3640 	struct mount *mp = NULL;
3641 	struct vnode *tvp, *fvp, *tdvp;
3642 	struct nameidata fromnd, tond;
3643 	int tvfslocked;
3644 	int fvfslocked;
3645 	int error;
3646 
3647 	bwillwrite();
3648 #ifdef MAC
3649 	NDINIT_ATRIGHTS(&fromnd, DELETE, LOCKPARENT | LOCKLEAF | SAVESTART |
3650 	    MPSAFE | AUDITVNODE1, pathseg, old, oldfd, CAP_DELETE, td);
3651 #else
3652 	NDINIT_ATRIGHTS(&fromnd, DELETE, WANTPARENT | SAVESTART | MPSAFE |
3653 	    AUDITVNODE1, pathseg, old, oldfd, CAP_DELETE, td);
3654 #endif
3655 
3656 	if ((error = namei(&fromnd)) != 0)
3657 		return (error);
3658 	fvfslocked = NDHASGIANT(&fromnd);
3659 	tvfslocked = 0;
3660 #ifdef MAC
3661 	error = mac_vnode_check_rename_from(td->td_ucred, fromnd.ni_dvp,
3662 	    fromnd.ni_vp, &fromnd.ni_cnd);
3663 	VOP_UNLOCK(fromnd.ni_dvp, 0);
3664 	if (fromnd.ni_dvp != fromnd.ni_vp)
3665 		VOP_UNLOCK(fromnd.ni_vp, 0);
3666 #endif
3667 	fvp = fromnd.ni_vp;
3668 	if (error == 0)
3669 		error = vn_start_write(fvp, &mp, V_WAIT | PCATCH);
3670 	if (error != 0) {
3671 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3672 		vrele(fromnd.ni_dvp);
3673 		vrele(fvp);
3674 		goto out1;
3675 	}
3676 	NDINIT_ATRIGHTS(&tond, RENAME, LOCKPARENT | LOCKLEAF | NOCACHE |
3677 	    SAVESTART | MPSAFE | AUDITVNODE2, pathseg, new, newfd, CAP_CREATE,
3678 	    td);
3679 	if (fromnd.ni_vp->v_type == VDIR)
3680 		tond.ni_cnd.cn_flags |= WILLBEDIR;
3681 	if ((error = namei(&tond)) != 0) {
3682 		/* Translate error code for rename("dir1", "dir2/."). */
3683 		if (error == EISDIR && fvp->v_type == VDIR)
3684 			error = EINVAL;
3685 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3686 		vrele(fromnd.ni_dvp);
3687 		vrele(fvp);
3688 		vn_finished_write(mp);
3689 		goto out1;
3690 	}
3691 	tvfslocked = NDHASGIANT(&tond);
3692 	tdvp = tond.ni_dvp;
3693 	tvp = tond.ni_vp;
3694 	if (tvp != NULL) {
3695 		if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
3696 			error = ENOTDIR;
3697 			goto out;
3698 		} else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
3699 			error = EISDIR;
3700 			goto out;
3701 		}
3702 	}
3703 	if (fvp == tdvp) {
3704 		error = EINVAL;
3705 		goto out;
3706 	}
3707 	/*
3708 	 * If the source is the same as the destination (that is, if they
3709 	 * are links to the same vnode), then there is nothing to do.
3710 	 */
3711 	if (fvp == tvp)
3712 		error = -1;
3713 #ifdef MAC
3714 	else
3715 		error = mac_vnode_check_rename_to(td->td_ucred, tdvp,
3716 		    tond.ni_vp, fromnd.ni_dvp == tdvp, &tond.ni_cnd);
3717 #endif
3718 out:
3719 	if (!error) {
3720 		error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
3721 				   tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
3722 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3723 		NDFREE(&tond, NDF_ONLY_PNBUF);
3724 	} else {
3725 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3726 		NDFREE(&tond, NDF_ONLY_PNBUF);
3727 		if (tvp)
3728 			vput(tvp);
3729 		if (tdvp == tvp)
3730 			vrele(tdvp);
3731 		else
3732 			vput(tdvp);
3733 		vrele(fromnd.ni_dvp);
3734 		vrele(fvp);
3735 	}
3736 	vrele(tond.ni_startdir);
3737 	vn_finished_write(mp);
3738 out1:
3739 	if (fromnd.ni_startdir)
3740 		vrele(fromnd.ni_startdir);
3741 	VFS_UNLOCK_GIANT(fvfslocked);
3742 	VFS_UNLOCK_GIANT(tvfslocked);
3743 	if (error == -1)
3744 		return (0);
3745 	return (error);
3746 }
3747 
3748 /*
3749  * Make a directory file.
3750  */
3751 #ifndef _SYS_SYSPROTO_H_
3752 struct mkdir_args {
3753 	char	*path;
3754 	int	mode;
3755 };
3756 #endif
3757 int
3758 sys_mkdir(td, uap)
3759 	struct thread *td;
3760 	register struct mkdir_args /* {
3761 		char *path;
3762 		int mode;
3763 	} */ *uap;
3764 {
3765 
3766 	return (kern_mkdir(td, uap->path, UIO_USERSPACE, uap->mode));
3767 }
3768 
3769 #ifndef _SYS_SYSPROTO_H_
3770 struct mkdirat_args {
3771 	int	fd;
3772 	char	*path;
3773 	mode_t	mode;
3774 };
3775 #endif
3776 int
3777 sys_mkdirat(struct thread *td, struct mkdirat_args *uap)
3778 {
3779 
3780 	return (kern_mkdirat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode));
3781 }
3782 
3783 int
3784 kern_mkdir(struct thread *td, char *path, enum uio_seg segflg, int mode)
3785 {
3786 
3787 	return (kern_mkdirat(td, AT_FDCWD, path, segflg, mode));
3788 }
3789 
3790 int
3791 kern_mkdirat(struct thread *td, int fd, char *path, enum uio_seg segflg,
3792     int mode)
3793 {
3794 	struct mount *mp;
3795 	struct vnode *vp;
3796 	struct vattr vattr;
3797 	int error;
3798 	struct nameidata nd;
3799 	int vfslocked;
3800 
3801 	AUDIT_ARG_MODE(mode);
3802 restart:
3803 	bwillwrite();
3804 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | MPSAFE |
3805 	    AUDITVNODE1, segflg, path, fd, CAP_MKDIR, td);
3806 	nd.ni_cnd.cn_flags |= WILLBEDIR;
3807 	if ((error = namei(&nd)) != 0)
3808 		return (error);
3809 	vfslocked = NDHASGIANT(&nd);
3810 	vp = nd.ni_vp;
3811 	if (vp != NULL) {
3812 		NDFREE(&nd, NDF_ONLY_PNBUF);
3813 		/*
3814 		 * XXX namei called with LOCKPARENT but not LOCKLEAF has
3815 		 * the strange behaviour of leaving the vnode unlocked
3816 		 * if the target is the same vnode as the parent.
3817 		 */
3818 		if (vp == nd.ni_dvp)
3819 			vrele(nd.ni_dvp);
3820 		else
3821 			vput(nd.ni_dvp);
3822 		vrele(vp);
3823 		VFS_UNLOCK_GIANT(vfslocked);
3824 		return (EEXIST);
3825 	}
3826 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3827 		NDFREE(&nd, NDF_ONLY_PNBUF);
3828 		vput(nd.ni_dvp);
3829 		VFS_UNLOCK_GIANT(vfslocked);
3830 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3831 			return (error);
3832 		goto restart;
3833 	}
3834 	VATTR_NULL(&vattr);
3835 	vattr.va_type = VDIR;
3836 	vattr.va_mode = (mode & ACCESSPERMS) &~ td->td_proc->p_fd->fd_cmask;
3837 #ifdef MAC
3838 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
3839 	    &vattr);
3840 	if (error)
3841 		goto out;
3842 #endif
3843 	error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
3844 #ifdef MAC
3845 out:
3846 #endif
3847 	NDFREE(&nd, NDF_ONLY_PNBUF);
3848 	vput(nd.ni_dvp);
3849 	if (!error)
3850 		vput(nd.ni_vp);
3851 	vn_finished_write(mp);
3852 	VFS_UNLOCK_GIANT(vfslocked);
3853 	return (error);
3854 }
3855 
3856 /*
3857  * Remove a directory file.
3858  */
3859 #ifndef _SYS_SYSPROTO_H_
3860 struct rmdir_args {
3861 	char	*path;
3862 };
3863 #endif
3864 int
3865 sys_rmdir(td, uap)
3866 	struct thread *td;
3867 	struct rmdir_args /* {
3868 		char *path;
3869 	} */ *uap;
3870 {
3871 
3872 	return (kern_rmdir(td, uap->path, UIO_USERSPACE));
3873 }
3874 
3875 int
3876 kern_rmdir(struct thread *td, char *path, enum uio_seg pathseg)
3877 {
3878 
3879 	return (kern_rmdirat(td, AT_FDCWD, path, pathseg));
3880 }
3881 
3882 int
3883 kern_rmdirat(struct thread *td, int fd, char *path, enum uio_seg pathseg)
3884 {
3885 	struct mount *mp;
3886 	struct vnode *vp;
3887 	int error;
3888 	struct nameidata nd;
3889 	int vfslocked;
3890 
3891 restart:
3892 	bwillwrite();
3893 	NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | MPSAFE |
3894 	    AUDITVNODE1, pathseg, path, fd, CAP_RMDIR, td);
3895 	if ((error = namei(&nd)) != 0)
3896 		return (error);
3897 	vfslocked = NDHASGIANT(&nd);
3898 	vp = nd.ni_vp;
3899 	if (vp->v_type != VDIR) {
3900 		error = ENOTDIR;
3901 		goto out;
3902 	}
3903 	/*
3904 	 * No rmdir "." please.
3905 	 */
3906 	if (nd.ni_dvp == vp) {
3907 		error = EINVAL;
3908 		goto out;
3909 	}
3910 	/*
3911 	 * The root of a mounted filesystem cannot be deleted.
3912 	 */
3913 	if (vp->v_vflag & VV_ROOT) {
3914 		error = EBUSY;
3915 		goto out;
3916 	}
3917 #ifdef MAC
3918 	error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
3919 	    &nd.ni_cnd);
3920 	if (error)
3921 		goto out;
3922 #endif
3923 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3924 		NDFREE(&nd, NDF_ONLY_PNBUF);
3925 		vput(vp);
3926 		if (nd.ni_dvp == vp)
3927 			vrele(nd.ni_dvp);
3928 		else
3929 			vput(nd.ni_dvp);
3930 		VFS_UNLOCK_GIANT(vfslocked);
3931 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3932 			return (error);
3933 		goto restart;
3934 	}
3935 	error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
3936 	vn_finished_write(mp);
3937 out:
3938 	NDFREE(&nd, NDF_ONLY_PNBUF);
3939 	vput(vp);
3940 	if (nd.ni_dvp == vp)
3941 		vrele(nd.ni_dvp);
3942 	else
3943 		vput(nd.ni_dvp);
3944 	VFS_UNLOCK_GIANT(vfslocked);
3945 	return (error);
3946 }
3947 
3948 #ifdef COMPAT_43
3949 /*
3950  * Read a block of directory entries in a filesystem independent format.
3951  */
3952 #ifndef _SYS_SYSPROTO_H_
3953 struct ogetdirentries_args {
3954 	int	fd;
3955 	char	*buf;
3956 	u_int	count;
3957 	long	*basep;
3958 };
3959 #endif
3960 int
3961 ogetdirentries(struct thread *td, struct ogetdirentries_args *uap)
3962 {
3963 	long loff;
3964 	int error;
3965 
3966 	error = kern_ogetdirentries(td, uap, &loff);
3967 	if (error == 0)
3968 		error = copyout(&loff, uap->basep, sizeof(long));
3969 	return (error);
3970 }
3971 
3972 int
3973 kern_ogetdirentries(struct thread *td, struct ogetdirentries_args *uap,
3974     long *ploff)
3975 {
3976 	struct vnode *vp;
3977 	struct file *fp;
3978 	struct uio auio, kuio;
3979 	struct iovec aiov, kiov;
3980 	struct dirent *dp, *edp;
3981 	caddr_t dirbuf;
3982 	int error, eofflag, readcnt, vfslocked;
3983 	long loff;
3984 
3985 	/* XXX arbitrary sanity limit on `count'. */
3986 	if (uap->count > 64 * 1024)
3987 		return (EINVAL);
3988 	if ((error = getvnode(td->td_proc->p_fd, uap->fd, CAP_READ,
3989 	    &fp)) != 0)
3990 		return (error);
3991 	if ((fp->f_flag & FREAD) == 0) {
3992 		fdrop(fp, td);
3993 		return (EBADF);
3994 	}
3995 	vp = fp->f_vnode;
3996 unionread:
3997 	vfslocked = VFS_LOCK_GIANT(vp->v_mount);
3998 	if (vp->v_type != VDIR) {
3999 		VFS_UNLOCK_GIANT(vfslocked);
4000 		fdrop(fp, td);
4001 		return (EINVAL);
4002 	}
4003 	aiov.iov_base = uap->buf;
4004 	aiov.iov_len = uap->count;
4005 	auio.uio_iov = &aiov;
4006 	auio.uio_iovcnt = 1;
4007 	auio.uio_rw = UIO_READ;
4008 	auio.uio_segflg = UIO_USERSPACE;
4009 	auio.uio_td = td;
4010 	auio.uio_resid = uap->count;
4011 	vn_lock(vp, LK_SHARED | LK_RETRY);
4012 	loff = auio.uio_offset = fp->f_offset;
4013 #ifdef MAC
4014 	error = mac_vnode_check_readdir(td->td_ucred, vp);
4015 	if (error) {
4016 		VOP_UNLOCK(vp, 0);
4017 		VFS_UNLOCK_GIANT(vfslocked);
4018 		fdrop(fp, td);
4019 		return (error);
4020 	}
4021 #endif
4022 #	if (BYTE_ORDER != LITTLE_ENDIAN)
4023 		if (vp->v_mount->mnt_maxsymlinklen <= 0) {
4024 			error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag,
4025 			    NULL, NULL);
4026 			fp->f_offset = auio.uio_offset;
4027 		} else
4028 #	endif
4029 	{
4030 		kuio = auio;
4031 		kuio.uio_iov = &kiov;
4032 		kuio.uio_segflg = UIO_SYSSPACE;
4033 		kiov.iov_len = uap->count;
4034 		dirbuf = malloc(uap->count, M_TEMP, M_WAITOK);
4035 		kiov.iov_base = dirbuf;
4036 		error = VOP_READDIR(vp, &kuio, fp->f_cred, &eofflag,
4037 			    NULL, NULL);
4038 		fp->f_offset = kuio.uio_offset;
4039 		if (error == 0) {
4040 			readcnt = uap->count - kuio.uio_resid;
4041 			edp = (struct dirent *)&dirbuf[readcnt];
4042 			for (dp = (struct dirent *)dirbuf; dp < edp; ) {
4043 #				if (BYTE_ORDER == LITTLE_ENDIAN)
4044 					/*
4045 					 * The expected low byte of
4046 					 * dp->d_namlen is our dp->d_type.
4047 					 * The high MBZ byte of dp->d_namlen
4048 					 * is our dp->d_namlen.
4049 					 */
4050 					dp->d_type = dp->d_namlen;
4051 					dp->d_namlen = 0;
4052 #				else
4053 					/*
4054 					 * The dp->d_type is the high byte
4055 					 * of the expected dp->d_namlen,
4056 					 * so must be zero'ed.
4057 					 */
4058 					dp->d_type = 0;
4059 #				endif
4060 				if (dp->d_reclen > 0) {
4061 					dp = (struct dirent *)
4062 					    ((char *)dp + dp->d_reclen);
4063 				} else {
4064 					error = EIO;
4065 					break;
4066 				}
4067 			}
4068 			if (dp >= edp)
4069 				error = uiomove(dirbuf, readcnt, &auio);
4070 		}
4071 		free(dirbuf, M_TEMP);
4072 	}
4073 	if (error) {
4074 		VOP_UNLOCK(vp, 0);
4075 		VFS_UNLOCK_GIANT(vfslocked);
4076 		fdrop(fp, td);
4077 		return (error);
4078 	}
4079 	if (uap->count == auio.uio_resid &&
4080 	    (vp->v_vflag & VV_ROOT) &&
4081 	    (vp->v_mount->mnt_flag & MNT_UNION)) {
4082 		struct vnode *tvp = vp;
4083 		vp = vp->v_mount->mnt_vnodecovered;
4084 		VREF(vp);
4085 		fp->f_vnode = vp;
4086 		fp->f_data = vp;
4087 		fp->f_offset = 0;
4088 		vput(tvp);
4089 		VFS_UNLOCK_GIANT(vfslocked);
4090 		goto unionread;
4091 	}
4092 	VOP_UNLOCK(vp, 0);
4093 	VFS_UNLOCK_GIANT(vfslocked);
4094 	fdrop(fp, td);
4095 	td->td_retval[0] = uap->count - auio.uio_resid;
4096 	if (error == 0)
4097 		*ploff = loff;
4098 	return (error);
4099 }
4100 #endif /* COMPAT_43 */
4101 
4102 /*
4103  * Read a block of directory entries in a filesystem independent format.
4104  */
4105 #ifndef _SYS_SYSPROTO_H_
4106 struct getdirentries_args {
4107 	int	fd;
4108 	char	*buf;
4109 	u_int	count;
4110 	long	*basep;
4111 };
4112 #endif
4113 int
4114 sys_getdirentries(td, uap)
4115 	struct thread *td;
4116 	register struct getdirentries_args /* {
4117 		int fd;
4118 		char *buf;
4119 		u_int count;
4120 		long *basep;
4121 	} */ *uap;
4122 {
4123 	long base;
4124 	int error;
4125 
4126 	error = kern_getdirentries(td, uap->fd, uap->buf, uap->count, &base);
4127 	if (error)
4128 		return (error);
4129 	if (uap->basep != NULL)
4130 		error = copyout(&base, uap->basep, sizeof(long));
4131 	return (error);
4132 }
4133 
4134 int
4135 kern_getdirentries(struct thread *td, int fd, char *buf, u_int count,
4136     long *basep)
4137 {
4138 	struct vnode *vp;
4139 	struct file *fp;
4140 	struct uio auio;
4141 	struct iovec aiov;
4142 	int vfslocked;
4143 	long loff;
4144 	int error, eofflag;
4145 
4146 	AUDIT_ARG_FD(fd);
4147 	if (count > INT_MAX)
4148 		return (EINVAL);
4149 	if ((error = getvnode(td->td_proc->p_fd, fd, CAP_READ | CAP_SEEK,
4150 	    &fp)) != 0)
4151 		return (error);
4152 	if ((fp->f_flag & FREAD) == 0) {
4153 		fdrop(fp, td);
4154 		return (EBADF);
4155 	}
4156 	vp = fp->f_vnode;
4157 unionread:
4158 	vfslocked = VFS_LOCK_GIANT(vp->v_mount);
4159 	if (vp->v_type != VDIR) {
4160 		VFS_UNLOCK_GIANT(vfslocked);
4161 		error = EINVAL;
4162 		goto fail;
4163 	}
4164 	aiov.iov_base = buf;
4165 	aiov.iov_len = count;
4166 	auio.uio_iov = &aiov;
4167 	auio.uio_iovcnt = 1;
4168 	auio.uio_rw = UIO_READ;
4169 	auio.uio_segflg = UIO_USERSPACE;
4170 	auio.uio_td = td;
4171 	auio.uio_resid = count;
4172 	vn_lock(vp, LK_SHARED | LK_RETRY);
4173 	AUDIT_ARG_VNODE1(vp);
4174 	loff = auio.uio_offset = fp->f_offset;
4175 #ifdef MAC
4176 	error = mac_vnode_check_readdir(td->td_ucred, vp);
4177 	if (error == 0)
4178 #endif
4179 		error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, NULL,
4180 		    NULL);
4181 	fp->f_offset = auio.uio_offset;
4182 	if (error) {
4183 		VOP_UNLOCK(vp, 0);
4184 		VFS_UNLOCK_GIANT(vfslocked);
4185 		goto fail;
4186 	}
4187 	if (count == auio.uio_resid &&
4188 	    (vp->v_vflag & VV_ROOT) &&
4189 	    (vp->v_mount->mnt_flag & MNT_UNION)) {
4190 		struct vnode *tvp = vp;
4191 		vp = vp->v_mount->mnt_vnodecovered;
4192 		VREF(vp);
4193 		fp->f_vnode = vp;
4194 		fp->f_data = vp;
4195 		fp->f_offset = 0;
4196 		vput(tvp);
4197 		VFS_UNLOCK_GIANT(vfslocked);
4198 		goto unionread;
4199 	}
4200 	VOP_UNLOCK(vp, 0);
4201 	VFS_UNLOCK_GIANT(vfslocked);
4202 	*basep = loff;
4203 	td->td_retval[0] = count - auio.uio_resid;
4204 fail:
4205 	fdrop(fp, td);
4206 	return (error);
4207 }
4208 
4209 #ifndef _SYS_SYSPROTO_H_
4210 struct getdents_args {
4211 	int fd;
4212 	char *buf;
4213 	size_t count;
4214 };
4215 #endif
4216 int
4217 sys_getdents(td, uap)
4218 	struct thread *td;
4219 	register struct getdents_args /* {
4220 		int fd;
4221 		char *buf;
4222 		u_int count;
4223 	} */ *uap;
4224 {
4225 	struct getdirentries_args ap;
4226 	ap.fd = uap->fd;
4227 	ap.buf = uap->buf;
4228 	ap.count = uap->count;
4229 	ap.basep = NULL;
4230 	return (sys_getdirentries(td, &ap));
4231 }
4232 
4233 /*
4234  * Set the mode mask for creation of filesystem nodes.
4235  */
4236 #ifndef _SYS_SYSPROTO_H_
4237 struct umask_args {
4238 	int	newmask;
4239 };
4240 #endif
4241 int
4242 sys_umask(td, uap)
4243 	struct thread *td;
4244 	struct umask_args /* {
4245 		int newmask;
4246 	} */ *uap;
4247 {
4248 	register struct filedesc *fdp;
4249 
4250 	FILEDESC_XLOCK(td->td_proc->p_fd);
4251 	fdp = td->td_proc->p_fd;
4252 	td->td_retval[0] = fdp->fd_cmask;
4253 	fdp->fd_cmask = uap->newmask & ALLPERMS;
4254 	FILEDESC_XUNLOCK(td->td_proc->p_fd);
4255 	return (0);
4256 }
4257 
4258 /*
4259  * Void all references to file by ripping underlying filesystem away from
4260  * vnode.
4261  */
4262 #ifndef _SYS_SYSPROTO_H_
4263 struct revoke_args {
4264 	char	*path;
4265 };
4266 #endif
4267 int
4268 sys_revoke(td, uap)
4269 	struct thread *td;
4270 	register struct revoke_args /* {
4271 		char *path;
4272 	} */ *uap;
4273 {
4274 	struct vnode *vp;
4275 	struct vattr vattr;
4276 	int error;
4277 	struct nameidata nd;
4278 	int vfslocked;
4279 
4280 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | MPSAFE | AUDITVNODE1,
4281 	    UIO_USERSPACE, uap->path, td);
4282 	if ((error = namei(&nd)) != 0)
4283 		return (error);
4284 	vfslocked = NDHASGIANT(&nd);
4285 	vp = nd.ni_vp;
4286 	NDFREE(&nd, NDF_ONLY_PNBUF);
4287 	if (vp->v_type != VCHR || vp->v_rdev == NULL) {
4288 		error = EINVAL;
4289 		goto out;
4290 	}
4291 #ifdef MAC
4292 	error = mac_vnode_check_revoke(td->td_ucred, vp);
4293 	if (error)
4294 		goto out;
4295 #endif
4296 	error = VOP_GETATTR(vp, &vattr, td->td_ucred);
4297 	if (error)
4298 		goto out;
4299 	if (td->td_ucred->cr_uid != vattr.va_uid) {
4300 		error = priv_check(td, PRIV_VFS_ADMIN);
4301 		if (error)
4302 			goto out;
4303 	}
4304 	if (vcount(vp) > 1)
4305 		VOP_REVOKE(vp, REVOKEALL);
4306 out:
4307 	vput(vp);
4308 	VFS_UNLOCK_GIANT(vfslocked);
4309 	return (error);
4310 }
4311 
4312 /*
4313  * Convert a user file descriptor to a kernel file entry and check that, if it
4314  * is a capability, the correct rights are present. A reference on the file
4315  * entry is held upon returning.
4316  */
4317 int
4318 getvnode(struct filedesc *fdp, int fd, cap_rights_t rights,
4319     struct file **fpp)
4320 {
4321 	struct file *fp;
4322 #ifdef CAPABILITIES
4323 	struct file *fp_fromcap;
4324 #endif
4325 	int error;
4326 
4327 	error = 0;
4328 	fp = NULL;
4329 	if ((fdp == NULL) || (fp = fget_unlocked(fdp, fd)) == NULL)
4330 		return (EBADF);
4331 #ifdef CAPABILITIES
4332 	/*
4333 	 * If the file descriptor is for a capability, test rights and use the
4334 	 * file descriptor referenced by the capability.
4335 	 */
4336 	error = cap_funwrap(fp, rights, &fp_fromcap);
4337 	if (error) {
4338 		fdrop(fp, curthread);
4339 		return (error);
4340 	}
4341 	if (fp != fp_fromcap) {
4342 		fhold(fp_fromcap);
4343 		fdrop(fp, curthread);
4344 		fp = fp_fromcap;
4345 	}
4346 #endif /* CAPABILITIES */
4347 	if (fp->f_vnode == NULL) {
4348 		fdrop(fp, curthread);
4349 		return (EINVAL);
4350 	}
4351 	*fpp = fp;
4352 	return (0);
4353 }
4354 
4355 
4356 /*
4357  * Get an (NFS) file handle.
4358  */
4359 #ifndef _SYS_SYSPROTO_H_
4360 struct lgetfh_args {
4361 	char	*fname;
4362 	fhandle_t *fhp;
4363 };
4364 #endif
4365 int
4366 sys_lgetfh(td, uap)
4367 	struct thread *td;
4368 	register struct lgetfh_args *uap;
4369 {
4370 	struct nameidata nd;
4371 	fhandle_t fh;
4372 	register struct vnode *vp;
4373 	int vfslocked;
4374 	int error;
4375 
4376 	error = priv_check(td, PRIV_VFS_GETFH);
4377 	if (error)
4378 		return (error);
4379 	NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | MPSAFE | AUDITVNODE1,
4380 	    UIO_USERSPACE, uap->fname, td);
4381 	error = namei(&nd);
4382 	if (error)
4383 		return (error);
4384 	vfslocked = NDHASGIANT(&nd);
4385 	NDFREE(&nd, NDF_ONLY_PNBUF);
4386 	vp = nd.ni_vp;
4387 	bzero(&fh, sizeof(fh));
4388 	fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
4389 	error = VOP_VPTOFH(vp, &fh.fh_fid);
4390 	vput(vp);
4391 	VFS_UNLOCK_GIANT(vfslocked);
4392 	if (error)
4393 		return (error);
4394 	error = copyout(&fh, uap->fhp, sizeof (fh));
4395 	return (error);
4396 }
4397 
4398 #ifndef _SYS_SYSPROTO_H_
4399 struct getfh_args {
4400 	char	*fname;
4401 	fhandle_t *fhp;
4402 };
4403 #endif
4404 int
4405 sys_getfh(td, uap)
4406 	struct thread *td;
4407 	register struct getfh_args *uap;
4408 {
4409 	struct nameidata nd;
4410 	fhandle_t fh;
4411 	register struct vnode *vp;
4412 	int vfslocked;
4413 	int error;
4414 
4415 	error = priv_check(td, PRIV_VFS_GETFH);
4416 	if (error)
4417 		return (error);
4418 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | MPSAFE | AUDITVNODE1,
4419 	    UIO_USERSPACE, uap->fname, td);
4420 	error = namei(&nd);
4421 	if (error)
4422 		return (error);
4423 	vfslocked = NDHASGIANT(&nd);
4424 	NDFREE(&nd, NDF_ONLY_PNBUF);
4425 	vp = nd.ni_vp;
4426 	bzero(&fh, sizeof(fh));
4427 	fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
4428 	error = VOP_VPTOFH(vp, &fh.fh_fid);
4429 	vput(vp);
4430 	VFS_UNLOCK_GIANT(vfslocked);
4431 	if (error)
4432 		return (error);
4433 	error = copyout(&fh, uap->fhp, sizeof (fh));
4434 	return (error);
4435 }
4436 
4437 /*
4438  * syscall for the rpc.lockd to use to translate a NFS file handle into an
4439  * open descriptor.
4440  *
4441  * warning: do not remove the priv_check() call or this becomes one giant
4442  * security hole.
4443  */
4444 #ifndef _SYS_SYSPROTO_H_
4445 struct fhopen_args {
4446 	const struct fhandle *u_fhp;
4447 	int flags;
4448 };
4449 #endif
4450 int
4451 sys_fhopen(td, uap)
4452 	struct thread *td;
4453 	struct fhopen_args /* {
4454 		const struct fhandle *u_fhp;
4455 		int flags;
4456 	} */ *uap;
4457 {
4458 	struct proc *p = td->td_proc;
4459 	struct mount *mp;
4460 	struct vnode *vp;
4461 	struct fhandle fhp;
4462 	struct vattr vat;
4463 	struct vattr *vap = &vat;
4464 	struct flock lf;
4465 	struct file *fp;
4466 	register struct filedesc *fdp = p->p_fd;
4467 	int fmode, error, type;
4468 	accmode_t accmode;
4469 	struct file *nfp;
4470 	int vfslocked;
4471 	int indx;
4472 
4473 	error = priv_check(td, PRIV_VFS_FHOPEN);
4474 	if (error)
4475 		return (error);
4476 	fmode = FFLAGS(uap->flags);
4477 	/* why not allow a non-read/write open for our lockd? */
4478 	if (((fmode & (FREAD | FWRITE)) == 0) || (fmode & O_CREAT))
4479 		return (EINVAL);
4480 	error = copyin(uap->u_fhp, &fhp, sizeof(fhp));
4481 	if (error)
4482 		return(error);
4483 	/* find the mount point */
4484 	mp = vfs_busyfs(&fhp.fh_fsid);
4485 	if (mp == NULL)
4486 		return (ESTALE);
4487 	vfslocked = VFS_LOCK_GIANT(mp);
4488 	/* now give me my vnode, it gets returned to me locked */
4489 	error = VFS_FHTOVP(mp, &fhp.fh_fid, LK_EXCLUSIVE, &vp);
4490 	vfs_unbusy(mp);
4491 	if (error)
4492 		goto out;
4493 	/*
4494 	 * from now on we have to make sure not
4495 	 * to forget about the vnode
4496 	 * any error that causes an abort must vput(vp)
4497 	 * just set error = err and 'goto bad;'.
4498 	 */
4499 
4500 	/*
4501 	 * from vn_open
4502 	 */
4503 	if (vp->v_type == VLNK) {
4504 		error = EMLINK;
4505 		goto bad;
4506 	}
4507 	if (vp->v_type == VSOCK) {
4508 		error = EOPNOTSUPP;
4509 		goto bad;
4510 	}
4511 	if (vp->v_type != VDIR && fmode & O_DIRECTORY) {
4512 		error = ENOTDIR;
4513 		goto bad;
4514 	}
4515 	accmode = 0;
4516 	if (fmode & (FWRITE | O_TRUNC)) {
4517 		if (vp->v_type == VDIR) {
4518 			error = EISDIR;
4519 			goto bad;
4520 		}
4521 		error = vn_writechk(vp);
4522 		if (error)
4523 			goto bad;
4524 		accmode |= VWRITE;
4525 	}
4526 	if (fmode & FREAD)
4527 		accmode |= VREAD;
4528 	if ((fmode & O_APPEND) && (fmode & FWRITE))
4529 		accmode |= VAPPEND;
4530 #ifdef MAC
4531 	error = mac_vnode_check_open(td->td_ucred, vp, accmode);
4532 	if (error)
4533 		goto bad;
4534 #endif
4535 	if (accmode) {
4536 		error = VOP_ACCESS(vp, accmode, td->td_ucred, td);
4537 		if (error)
4538 			goto bad;
4539 	}
4540 	if (fmode & O_TRUNC) {
4541 		vfs_ref(mp);
4542 		VOP_UNLOCK(vp, 0);				/* XXX */
4543 		if ((error = vn_start_write(NULL, &mp, V_WAIT | PCATCH)) != 0) {
4544 			vrele(vp);
4545 			vfs_rel(mp);
4546 			goto out;
4547 		}
4548 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);	/* XXX */
4549 		vfs_rel(mp);
4550 #ifdef MAC
4551 		/*
4552 		 * We don't yet have fp->f_cred, so use td->td_ucred, which
4553 		 * should be right.
4554 		 */
4555 		error = mac_vnode_check_write(td->td_ucred, td->td_ucred, vp);
4556 		if (error == 0) {
4557 #endif
4558 			VATTR_NULL(vap);
4559 			vap->va_size = 0;
4560 			error = VOP_SETATTR(vp, vap, td->td_ucred);
4561 #ifdef MAC
4562 		}
4563 #endif
4564 		vn_finished_write(mp);
4565 		if (error)
4566 			goto bad;
4567 	}
4568 	error = VOP_OPEN(vp, fmode, td->td_ucred, td, NULL);
4569 	if (error)
4570 		goto bad;
4571 
4572 	if (fmode & FWRITE)
4573 		vp->v_writecount++;
4574 
4575 	/*
4576 	 * end of vn_open code
4577 	 */
4578 
4579 	if ((error = falloc(td, &nfp, &indx, fmode)) != 0) {
4580 		if (fmode & FWRITE)
4581 			vp->v_writecount--;
4582 		goto bad;
4583 	}
4584 	/* An extra reference on `nfp' has been held for us by falloc(). */
4585 	fp = nfp;
4586 	nfp->f_vnode = vp;
4587 	finit(nfp, fmode & FMASK, DTYPE_VNODE, vp, &vnops);
4588 	if (fmode & (O_EXLOCK | O_SHLOCK)) {
4589 		lf.l_whence = SEEK_SET;
4590 		lf.l_start = 0;
4591 		lf.l_len = 0;
4592 		if (fmode & O_EXLOCK)
4593 			lf.l_type = F_WRLCK;
4594 		else
4595 			lf.l_type = F_RDLCK;
4596 		type = F_FLOCK;
4597 		if ((fmode & FNONBLOCK) == 0)
4598 			type |= F_WAIT;
4599 		VOP_UNLOCK(vp, 0);
4600 		if ((error = VOP_ADVLOCK(vp, (caddr_t)fp, F_SETLK, &lf,
4601 			    type)) != 0) {
4602 			/*
4603 			 * The lock request failed.  Normally close the
4604 			 * descriptor but handle the case where someone might
4605 			 * have dup()d or close()d it when we weren't looking.
4606 			 */
4607 			fdclose(fdp, fp, indx, td);
4608 
4609 			/*
4610 			 * release our private reference
4611 			 */
4612 			fdrop(fp, td);
4613 			goto out;
4614 		}
4615 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
4616 		atomic_set_int(&fp->f_flag, FHASLOCK);
4617 	}
4618 
4619 	VOP_UNLOCK(vp, 0);
4620 	fdrop(fp, td);
4621 	VFS_UNLOCK_GIANT(vfslocked);
4622 	td->td_retval[0] = indx;
4623 	return (0);
4624 
4625 bad:
4626 	vput(vp);
4627 out:
4628 	VFS_UNLOCK_GIANT(vfslocked);
4629 	return (error);
4630 }
4631 
4632 /*
4633  * Stat an (NFS) file handle.
4634  */
4635 #ifndef _SYS_SYSPROTO_H_
4636 struct fhstat_args {
4637 	struct fhandle *u_fhp;
4638 	struct stat *sb;
4639 };
4640 #endif
4641 int
4642 sys_fhstat(td, uap)
4643 	struct thread *td;
4644 	register struct fhstat_args /* {
4645 		struct fhandle *u_fhp;
4646 		struct stat *sb;
4647 	} */ *uap;
4648 {
4649 	struct stat sb;
4650 	fhandle_t fh;
4651 	struct mount *mp;
4652 	struct vnode *vp;
4653 	int vfslocked;
4654 	int error;
4655 
4656 	error = priv_check(td, PRIV_VFS_FHSTAT);
4657 	if (error)
4658 		return (error);
4659 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
4660 	if (error)
4661 		return (error);
4662 	if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4663 		return (ESTALE);
4664 	vfslocked = VFS_LOCK_GIANT(mp);
4665 	error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4666 	vfs_unbusy(mp);
4667 	if (error) {
4668 		VFS_UNLOCK_GIANT(vfslocked);
4669 		return (error);
4670 	}
4671 	error = vn_stat(vp, &sb, td->td_ucred, NOCRED, td);
4672 	vput(vp);
4673 	VFS_UNLOCK_GIANT(vfslocked);
4674 	if (error)
4675 		return (error);
4676 	error = copyout(&sb, uap->sb, sizeof(sb));
4677 	return (error);
4678 }
4679 
4680 /*
4681  * Implement fstatfs() for (NFS) file handles.
4682  */
4683 #ifndef _SYS_SYSPROTO_H_
4684 struct fhstatfs_args {
4685 	struct fhandle *u_fhp;
4686 	struct statfs *buf;
4687 };
4688 #endif
4689 int
4690 sys_fhstatfs(td, uap)
4691 	struct thread *td;
4692 	struct fhstatfs_args /* {
4693 		struct fhandle *u_fhp;
4694 		struct statfs *buf;
4695 	} */ *uap;
4696 {
4697 	struct statfs sf;
4698 	fhandle_t fh;
4699 	int error;
4700 
4701 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
4702 	if (error)
4703 		return (error);
4704 	error = kern_fhstatfs(td, fh, &sf);
4705 	if (error)
4706 		return (error);
4707 	return (copyout(&sf, uap->buf, sizeof(sf)));
4708 }
4709 
4710 int
4711 kern_fhstatfs(struct thread *td, fhandle_t fh, struct statfs *buf)
4712 {
4713 	struct statfs *sp;
4714 	struct mount *mp;
4715 	struct vnode *vp;
4716 	int vfslocked;
4717 	int error;
4718 
4719 	error = priv_check(td, PRIV_VFS_FHSTATFS);
4720 	if (error)
4721 		return (error);
4722 	if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4723 		return (ESTALE);
4724 	vfslocked = VFS_LOCK_GIANT(mp);
4725 	error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4726 	if (error) {
4727 		vfs_unbusy(mp);
4728 		VFS_UNLOCK_GIANT(vfslocked);
4729 		return (error);
4730 	}
4731 	vput(vp);
4732 	error = prison_canseemount(td->td_ucred, mp);
4733 	if (error)
4734 		goto out;
4735 #ifdef MAC
4736 	error = mac_mount_check_stat(td->td_ucred, mp);
4737 	if (error)
4738 		goto out;
4739 #endif
4740 	/*
4741 	 * Set these in case the underlying filesystem fails to do so.
4742 	 */
4743 	sp = &mp->mnt_stat;
4744 	sp->f_version = STATFS_VERSION;
4745 	sp->f_namemax = NAME_MAX;
4746 	sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
4747 	error = VFS_STATFS(mp, sp);
4748 	if (error == 0)
4749 		*buf = *sp;
4750 out:
4751 	vfs_unbusy(mp);
4752 	VFS_UNLOCK_GIANT(vfslocked);
4753 	return (error);
4754 }
4755 
4756 int
4757 kern_posix_fallocate(struct thread *td, int fd, off_t offset, off_t len)
4758 {
4759 	struct file *fp;
4760 	struct mount *mp;
4761 	struct vnode *vp;
4762 	off_t olen, ooffset;
4763 	int error, vfslocked;
4764 
4765 	fp = NULL;
4766 	vfslocked = 0;
4767 	error = fget(td, fd, CAP_WRITE, &fp);
4768 	if (error != 0)
4769 		goto out;
4770 
4771 	switch (fp->f_type) {
4772 	case DTYPE_VNODE:
4773 		break;
4774 	case DTYPE_PIPE:
4775 	case DTYPE_FIFO:
4776 		error = ESPIPE;
4777 		goto out;
4778 	default:
4779 		error = ENODEV;
4780 		goto out;
4781 	}
4782 	if ((fp->f_flag & FWRITE) == 0) {
4783 		error = EBADF;
4784 		goto out;
4785 	}
4786 	vp = fp->f_vnode;
4787 	if (vp->v_type != VREG) {
4788 		error = ENODEV;
4789 		goto out;
4790 	}
4791 	if (offset < 0 || len <= 0) {
4792 		error = EINVAL;
4793 		goto out;
4794 	}
4795 	/* Check for wrap. */
4796 	if (offset > OFF_MAX - len) {
4797 		error = EFBIG;
4798 		goto out;
4799 	}
4800 
4801 	/* Allocating blocks may take a long time, so iterate. */
4802 	for (;;) {
4803 		olen = len;
4804 		ooffset = offset;
4805 
4806 		bwillwrite();
4807 		vfslocked = VFS_LOCK_GIANT(vp->v_mount);
4808 		mp = NULL;
4809 		error = vn_start_write(vp, &mp, V_WAIT | PCATCH);
4810 		if (error != 0) {
4811 			VFS_UNLOCK_GIANT(vfslocked);
4812 			break;
4813 		}
4814 		error = vn_lock(vp, LK_EXCLUSIVE);
4815 		if (error != 0) {
4816 			vn_finished_write(mp);
4817 			VFS_UNLOCK_GIANT(vfslocked);
4818 			break;
4819 		}
4820 #ifdef MAC
4821 		error = mac_vnode_check_write(td->td_ucred, fp->f_cred, vp);
4822 		if (error == 0)
4823 #endif
4824 			error = VOP_ALLOCATE(vp, &offset, &len);
4825 		VOP_UNLOCK(vp, 0);
4826 		vn_finished_write(mp);
4827 		VFS_UNLOCK_GIANT(vfslocked);
4828 
4829 		if (olen + ooffset != offset + len) {
4830 			panic("offset + len changed from %jx/%jx to %jx/%jx",
4831 			    ooffset, olen, offset, len);
4832 		}
4833 		if (error != 0 || len == 0)
4834 			break;
4835 		KASSERT(olen > len, ("Iteration did not make progress?"));
4836 		maybe_yield();
4837 	}
4838  out:
4839 	if (fp != NULL)
4840 		fdrop(fp, td);
4841 	return (error);
4842 }
4843 
4844 int
4845 sys_posix_fallocate(struct thread *td, struct posix_fallocate_args *uap)
4846 {
4847 
4848 	return (kern_posix_fallocate(td, uap->fd, uap->offset, uap->len));
4849 }
4850 
4851 /*
4852  * Unlike madvise(2), we do not make a best effort to remember every
4853  * possible caching hint.  Instead, we remember the last setting with
4854  * the exception that we will allow POSIX_FADV_NORMAL to adjust the
4855  * region of any current setting.
4856  */
4857 int
4858 kern_posix_fadvise(struct thread *td, int fd, off_t offset, off_t len,
4859     int advice)
4860 {
4861 	struct fadvise_info *fa, *new;
4862 	struct file *fp;
4863 	struct vnode *vp;
4864 	off_t end;
4865 	int error;
4866 
4867 	if (offset < 0 || len < 0 || offset > OFF_MAX - len)
4868 		return (EINVAL);
4869 	switch (advice) {
4870 	case POSIX_FADV_SEQUENTIAL:
4871 	case POSIX_FADV_RANDOM:
4872 	case POSIX_FADV_NOREUSE:
4873 		new = malloc(sizeof(*fa), M_FADVISE, M_WAITOK);
4874 		break;
4875 	case POSIX_FADV_NORMAL:
4876 	case POSIX_FADV_WILLNEED:
4877 	case POSIX_FADV_DONTNEED:
4878 		new = NULL;
4879 		break;
4880 	default:
4881 		return (EINVAL);
4882 	}
4883 	/* XXX: CAP_POSIX_FADVISE? */
4884 	error = fget(td, fd, 0, &fp);
4885 	if (error != 0)
4886 		goto out;
4887 
4888 	switch (fp->f_type) {
4889 	case DTYPE_VNODE:
4890 		break;
4891 	case DTYPE_PIPE:
4892 	case DTYPE_FIFO:
4893 		error = ESPIPE;
4894 		goto out;
4895 	default:
4896 		error = ENODEV;
4897 		goto out;
4898 	}
4899 	vp = fp->f_vnode;
4900 	if (vp->v_type != VREG) {
4901 		error = ENODEV;
4902 		goto out;
4903 	}
4904 	if (len == 0)
4905 		end = OFF_MAX;
4906 	else
4907 		end = offset + len - 1;
4908 	switch (advice) {
4909 	case POSIX_FADV_SEQUENTIAL:
4910 	case POSIX_FADV_RANDOM:
4911 	case POSIX_FADV_NOREUSE:
4912 		/*
4913 		 * Try to merge any existing non-standard region with
4914 		 * this new region if possible, otherwise create a new
4915 		 * non-standard region for this request.
4916 		 */
4917 		mtx_pool_lock(mtxpool_sleep, fp);
4918 		fa = fp->f_advice;
4919 		if (fa != NULL && fa->fa_advice == advice &&
4920 		    ((fa->fa_start <= end && fa->fa_end >= offset) ||
4921 		    (end != OFF_MAX && fa->fa_start == end + 1) ||
4922 		    (fa->fa_end != OFF_MAX && fa->fa_end + 1 == offset))) {
4923 			if (offset < fa->fa_start)
4924 				fa->fa_start = offset;
4925 			if (end > fa->fa_end)
4926 				fa->fa_end = end;
4927 		} else {
4928 			new->fa_advice = advice;
4929 			new->fa_start = offset;
4930 			new->fa_end = end;
4931 			fp->f_advice = new;
4932 			new = fa;
4933 		}
4934 		mtx_pool_unlock(mtxpool_sleep, fp);
4935 		break;
4936 	case POSIX_FADV_NORMAL:
4937 		/*
4938 		 * If a the "normal" region overlaps with an existing
4939 		 * non-standard region, trim or remove the
4940 		 * non-standard region.
4941 		 */
4942 		mtx_pool_lock(mtxpool_sleep, fp);
4943 		fa = fp->f_advice;
4944 		if (fa != NULL) {
4945 			if (offset <= fa->fa_start && end >= fa->fa_end) {
4946 				new = fa;
4947 				fp->f_advice = NULL;
4948 			} else if (offset <= fa->fa_start &&
4949  			    end >= fa->fa_start)
4950 				fa->fa_start = end + 1;
4951 			else if (offset <= fa->fa_end && end >= fa->fa_end)
4952 				fa->fa_end = offset - 1;
4953 			else if (offset >= fa->fa_start && end <= fa->fa_end) {
4954 				/*
4955 				 * If the "normal" region is a middle
4956 				 * portion of the existing
4957 				 * non-standard region, just remove
4958 				 * the whole thing rather than picking
4959 				 * one side or the other to
4960 				 * preserve.
4961 				 */
4962 				new = fa;
4963 				fp->f_advice = NULL;
4964 			}
4965 		}
4966 		mtx_pool_unlock(mtxpool_sleep, fp);
4967 		break;
4968 	case POSIX_FADV_WILLNEED:
4969 	case POSIX_FADV_DONTNEED:
4970 		error = VOP_ADVISE(vp, offset, end, advice);
4971 		break;
4972 	}
4973 out:
4974 	if (fp != NULL)
4975 		fdrop(fp, td);
4976 	free(new, M_FADVISE);
4977 	return (error);
4978 }
4979 
4980 int
4981 sys_posix_fadvise(struct thread *td, struct posix_fadvise_args *uap)
4982 {
4983 
4984 	return (kern_posix_fadvise(td, uap->fd, uap->offset, uap->len,
4985 	    uap->advice));
4986 }
4987