xref: /freebsd/sys/kern/vfs_syscalls.c (revision 00a5db46de56179184c0f000eaacad695e2b0859)
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_compat.h"
41 #include "opt_kdtrace.h"
42 #include "opt_ktrace.h"
43 #include "opt_mac.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/sysent.h>
50 #include <sys/malloc.h>
51 #include <sys/mount.h>
52 #include <sys/mutex.h>
53 #include <sys/sysproto.h>
54 #include <sys/namei.h>
55 #include <sys/filedesc.h>
56 #include <sys/kernel.h>
57 #include <sys/fcntl.h>
58 #include <sys/file.h>
59 #include <sys/filio.h>
60 #include <sys/limits.h>
61 #include <sys/linker.h>
62 #include <sys/sdt.h>
63 #include <sys/stat.h>
64 #include <sys/sx.h>
65 #include <sys/unistd.h>
66 #include <sys/vnode.h>
67 #include <sys/priv.h>
68 #include <sys/proc.h>
69 #include <sys/dirent.h>
70 #include <sys/jail.h>
71 #include <sys/syscallsubr.h>
72 #include <sys/sysctl.h>
73 #ifdef KTRACE
74 #include <sys/ktrace.h>
75 #endif
76 
77 #include <machine/stdarg.h>
78 
79 #include <security/audit/audit.h>
80 #include <security/mac/mac_framework.h>
81 
82 #include <vm/vm.h>
83 #include <vm/vm_object.h>
84 #include <vm/vm_page.h>
85 #include <vm/uma.h>
86 
87 SDT_PROVIDER_DEFINE(vfs);
88 SDT_PROBE_DEFINE(vfs, , stat, mode);
89 SDT_PROBE_ARGTYPE(vfs, , stat, mode, 0, "char *");
90 SDT_PROBE_ARGTYPE(vfs, , stat, mode, 1, "int");
91 SDT_PROBE_DEFINE(vfs, , stat, reg);
92 SDT_PROBE_ARGTYPE(vfs, , stat, reg, 0, "char *");
93 SDT_PROBE_ARGTYPE(vfs, , stat, reg, 1, "int");
94 
95 static int chroot_refuse_vdir_fds(struct filedesc *fdp);
96 static int getutimes(const struct timeval *, enum uio_seg, struct timespec *);
97 static int setfown(struct thread *td, struct vnode *, uid_t, gid_t);
98 static int setfmode(struct thread *td, struct vnode *, int);
99 static int setfflags(struct thread *td, struct vnode *, int);
100 static int setutimes(struct thread *td, struct vnode *,
101     const struct timespec *, int, int);
102 static int vn_access(struct vnode *vp, int user_flags, struct ucred *cred,
103     struct thread *td);
104 
105 /*
106  * The module initialization routine for POSIX asynchronous I/O will
107  * set this to the version of AIO that it implements.  (Zero means
108  * that it is not implemented.)  This value is used here by pathconf()
109  * and in kern_descrip.c by fpathconf().
110  */
111 int async_io_version;
112 
113 #ifdef DEBUG
114 static int syncprt = 0;
115 SYSCTL_INT(_debug, OID_AUTO, syncprt, CTLFLAG_RW, &syncprt, 0, "");
116 #endif
117 
118 /*
119  * Sync each mounted filesystem.
120  */
121 #ifndef _SYS_SYSPROTO_H_
122 struct sync_args {
123 	int     dummy;
124 };
125 #endif
126 /* ARGSUSED */
127 int
128 sync(td, uap)
129 	struct thread *td;
130 	struct sync_args *uap;
131 {
132 	struct mount *mp, *nmp;
133 	int vfslocked;
134 
135 	mtx_lock(&mountlist_mtx);
136 	for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
137 		if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK)) {
138 			nmp = TAILQ_NEXT(mp, mnt_list);
139 			continue;
140 		}
141 		vfslocked = VFS_LOCK_GIANT(mp);
142 		if ((mp->mnt_flag & MNT_RDONLY) == 0 &&
143 		    vn_start_write(NULL, &mp, V_NOWAIT) == 0) {
144 			MNT_ILOCK(mp);
145 			mp->mnt_noasync++;
146 			mp->mnt_kern_flag &= ~MNTK_ASYNC;
147 			MNT_IUNLOCK(mp);
148 			vfs_msync(mp, MNT_NOWAIT);
149 			VFS_SYNC(mp, MNT_NOWAIT);
150 			MNT_ILOCK(mp);
151 			mp->mnt_noasync--;
152 			if ((mp->mnt_flag & MNT_ASYNC) != 0 &&
153 			    mp->mnt_noasync == 0)
154 				mp->mnt_kern_flag |= MNTK_ASYNC;
155 			MNT_IUNLOCK(mp);
156 			vn_finished_write(mp);
157 		}
158 		VFS_UNLOCK_GIANT(vfslocked);
159 		mtx_lock(&mountlist_mtx);
160 		nmp = TAILQ_NEXT(mp, mnt_list);
161 		vfs_unbusy(mp);
162 	}
163 	mtx_unlock(&mountlist_mtx);
164 	return (0);
165 }
166 
167 /* XXX PRISON: could be per prison flag */
168 static int prison_quotas;
169 #if 0
170 SYSCTL_INT(_kern_prison, OID_AUTO, quotas, CTLFLAG_RW, &prison_quotas, 0, "");
171 #endif
172 
173 /*
174  * Change filesystem quotas.
175  */
176 #ifndef _SYS_SYSPROTO_H_
177 struct quotactl_args {
178 	char *path;
179 	int cmd;
180 	int uid;
181 	caddr_t arg;
182 };
183 #endif
184 int
185 quotactl(td, uap)
186 	struct thread *td;
187 	register struct quotactl_args /* {
188 		char *path;
189 		int cmd;
190 		int uid;
191 		caddr_t arg;
192 	} */ *uap;
193 {
194 	struct mount *mp;
195 	int vfslocked;
196 	int error;
197 	struct nameidata nd;
198 
199 	AUDIT_ARG(cmd, uap->cmd);
200 	AUDIT_ARG(uid, uap->uid);
201 	if (jailed(td->td_ucred) && !prison_quotas)
202 		return (EPERM);
203 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | MPSAFE | AUDITVNODE1,
204 	   UIO_USERSPACE, uap->path, td);
205 	if ((error = namei(&nd)) != 0)
206 		return (error);
207 	vfslocked = NDHASGIANT(&nd);
208 	NDFREE(&nd, NDF_ONLY_PNBUF);
209 	mp = nd.ni_vp->v_mount;
210 	vfs_ref(mp);
211 	vput(nd.ni_vp);
212 	error = vfs_busy(mp, 0);
213 	vfs_rel(mp);
214 	if (error) {
215 		VFS_UNLOCK_GIANT(vfslocked);
216 		return (error);
217 	}
218 	error = VFS_QUOTACTL(mp, uap->cmd, uap->uid, uap->arg);
219 	vfs_unbusy(mp);
220 	VFS_UNLOCK_GIANT(vfslocked);
221 	return (error);
222 }
223 
224 /*
225  * Used by statfs conversion routines to scale the block size up if
226  * necessary so that all of the block counts are <= 'max_size'.  Note
227  * that 'max_size' should be a bitmask, i.e. 2^n - 1 for some non-zero
228  * value of 'n'.
229  */
230 void
231 statfs_scale_blocks(struct statfs *sf, long max_size)
232 {
233 	uint64_t count;
234 	int shift;
235 
236 	KASSERT(powerof2(max_size + 1), ("%s: invalid max_size", __func__));
237 
238 	/*
239 	 * Attempt to scale the block counts to give a more accurate
240 	 * overview to userland of the ratio of free space to used
241 	 * space.  To do this, find the largest block count and compute
242 	 * a divisor that lets it fit into a signed integer <= max_size.
243 	 */
244 	if (sf->f_bavail < 0)
245 		count = -sf->f_bavail;
246 	else
247 		count = sf->f_bavail;
248 	count = MAX(sf->f_blocks, MAX(sf->f_bfree, count));
249 	if (count <= max_size)
250 		return;
251 
252 	count >>= flsl(max_size);
253 	shift = 0;
254 	while (count > 0) {
255 		shift++;
256 		count >>=1;
257 	}
258 
259 	sf->f_bsize <<= shift;
260 	sf->f_blocks >>= shift;
261 	sf->f_bfree >>= shift;
262 	sf->f_bavail >>= shift;
263 }
264 
265 /*
266  * Get filesystem statistics.
267  */
268 #ifndef _SYS_SYSPROTO_H_
269 struct statfs_args {
270 	char *path;
271 	struct statfs *buf;
272 };
273 #endif
274 int
275 statfs(td, uap)
276 	struct thread *td;
277 	register struct statfs_args /* {
278 		char *path;
279 		struct statfs *buf;
280 	} */ *uap;
281 {
282 	struct statfs sf;
283 	int error;
284 
285 	error = kern_statfs(td, uap->path, UIO_USERSPACE, &sf);
286 	if (error == 0)
287 		error = copyout(&sf, uap->buf, sizeof(sf));
288 	return (error);
289 }
290 
291 int
292 kern_statfs(struct thread *td, char *path, enum uio_seg pathseg,
293     struct statfs *buf)
294 {
295 	struct mount *mp;
296 	struct statfs *sp, sb;
297 	int vfslocked;
298 	int error;
299 	struct nameidata nd;
300 
301 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | MPSAFE |
302 	    AUDITVNODE1, pathseg, path, td);
303 	error = namei(&nd);
304 	if (error)
305 		return (error);
306 	vfslocked = NDHASGIANT(&nd);
307 	mp = nd.ni_vp->v_mount;
308 	vfs_ref(mp);
309 	NDFREE(&nd, NDF_ONLY_PNBUF);
310 	vput(nd.ni_vp);
311 	error = vfs_busy(mp, 0);
312 	vfs_rel(mp);
313 	if (error) {
314 		VFS_UNLOCK_GIANT(vfslocked);
315 		return (error);
316 	}
317 #ifdef MAC
318 	error = mac_mount_check_stat(td->td_ucred, mp);
319 	if (error)
320 		goto out;
321 #endif
322 	/*
323 	 * Set these in case the underlying filesystem fails to do so.
324 	 */
325 	sp = &mp->mnt_stat;
326 	sp->f_version = STATFS_VERSION;
327 	sp->f_namemax = NAME_MAX;
328 	sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
329 	error = VFS_STATFS(mp, sp);
330 	if (error)
331 		goto out;
332 	if (priv_check(td, PRIV_VFS_GENERATION)) {
333 		bcopy(sp, &sb, sizeof(sb));
334 		sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
335 		prison_enforce_statfs(td->td_ucred, mp, &sb);
336 		sp = &sb;
337 	}
338 	*buf = *sp;
339 out:
340 	vfs_unbusy(mp);
341 	VFS_UNLOCK_GIANT(vfslocked);
342 	return (error);
343 }
344 
345 /*
346  * Get filesystem statistics.
347  */
348 #ifndef _SYS_SYSPROTO_H_
349 struct fstatfs_args {
350 	int fd;
351 	struct statfs *buf;
352 };
353 #endif
354 int
355 fstatfs(td, uap)
356 	struct thread *td;
357 	register struct fstatfs_args /* {
358 		int fd;
359 		struct statfs *buf;
360 	} */ *uap;
361 {
362 	struct statfs sf;
363 	int error;
364 
365 	error = kern_fstatfs(td, uap->fd, &sf);
366 	if (error == 0)
367 		error = copyout(&sf, uap->buf, sizeof(sf));
368 	return (error);
369 }
370 
371 int
372 kern_fstatfs(struct thread *td, int fd, struct statfs *buf)
373 {
374 	struct file *fp;
375 	struct mount *mp;
376 	struct statfs *sp, sb;
377 	int vfslocked;
378 	struct vnode *vp;
379 	int error;
380 
381 	AUDIT_ARG(fd, fd);
382 	error = getvnode(td->td_proc->p_fd, fd, &fp);
383 	if (error)
384 		return (error);
385 	vp = fp->f_vnode;
386 	vfslocked = VFS_LOCK_GIANT(vp->v_mount);
387 	vn_lock(vp, LK_SHARED | LK_RETRY);
388 #ifdef AUDIT
389 	AUDIT_ARG(vnode, vp, ARG_VNODE1);
390 #endif
391 	mp = vp->v_mount;
392 	if (mp)
393 		vfs_ref(mp);
394 	VOP_UNLOCK(vp, 0);
395 	fdrop(fp, td);
396 	if (mp == NULL) {
397 		error = EBADF;
398 		goto out;
399 	}
400 	error = vfs_busy(mp, 0);
401 	vfs_rel(mp);
402 	if (error) {
403 		VFS_UNLOCK_GIANT(vfslocked);
404 		return (error);
405 	}
406 #ifdef MAC
407 	error = mac_mount_check_stat(td->td_ucred, mp);
408 	if (error)
409 		goto out;
410 #endif
411 	/*
412 	 * Set these in case the underlying filesystem fails to do so.
413 	 */
414 	sp = &mp->mnt_stat;
415 	sp->f_version = STATFS_VERSION;
416 	sp->f_namemax = NAME_MAX;
417 	sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
418 	error = VFS_STATFS(mp, sp);
419 	if (error)
420 		goto out;
421 	if (priv_check(td, PRIV_VFS_GENERATION)) {
422 		bcopy(sp, &sb, sizeof(sb));
423 		sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
424 		prison_enforce_statfs(td->td_ucred, mp, &sb);
425 		sp = &sb;
426 	}
427 	*buf = *sp;
428 out:
429 	if (mp)
430 		vfs_unbusy(mp);
431 	VFS_UNLOCK_GIANT(vfslocked);
432 	return (error);
433 }
434 
435 /*
436  * Get statistics on all filesystems.
437  */
438 #ifndef _SYS_SYSPROTO_H_
439 struct getfsstat_args {
440 	struct statfs *buf;
441 	long bufsize;
442 	int flags;
443 };
444 #endif
445 int
446 getfsstat(td, uap)
447 	struct thread *td;
448 	register struct getfsstat_args /* {
449 		struct statfs *buf;
450 		long bufsize;
451 		int flags;
452 	} */ *uap;
453 {
454 
455 	return (kern_getfsstat(td, &uap->buf, uap->bufsize, UIO_USERSPACE,
456 	    uap->flags));
457 }
458 
459 /*
460  * If (bufsize > 0 && bufseg == UIO_SYSSPACE)
461  * 	The caller is responsible for freeing memory which will be allocated
462  *	in '*buf'.
463  */
464 int
465 kern_getfsstat(struct thread *td, struct statfs **buf, size_t bufsize,
466     enum uio_seg bufseg, int flags)
467 {
468 	struct mount *mp, *nmp;
469 	struct statfs *sfsp, *sp, sb;
470 	size_t count, maxcount;
471 	int vfslocked;
472 	int error;
473 
474 	maxcount = bufsize / sizeof(struct statfs);
475 	if (bufsize == 0)
476 		sfsp = NULL;
477 	else if (bufseg == UIO_USERSPACE)
478 		sfsp = *buf;
479 	else /* if (bufseg == UIO_SYSSPACE) */ {
480 		count = 0;
481 		mtx_lock(&mountlist_mtx);
482 		TAILQ_FOREACH(mp, &mountlist, mnt_list) {
483 			count++;
484 		}
485 		mtx_unlock(&mountlist_mtx);
486 		if (maxcount > count)
487 			maxcount = count;
488 		sfsp = *buf = malloc(maxcount * sizeof(struct statfs), M_TEMP,
489 		    M_WAITOK);
490 	}
491 	count = 0;
492 	mtx_lock(&mountlist_mtx);
493 	for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
494 		if (prison_canseemount(td->td_ucred, mp) != 0) {
495 			nmp = TAILQ_NEXT(mp, mnt_list);
496 			continue;
497 		}
498 #ifdef MAC
499 		if (mac_mount_check_stat(td->td_ucred, mp) != 0) {
500 			nmp = TAILQ_NEXT(mp, mnt_list);
501 			continue;
502 		}
503 #endif
504 		if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK)) {
505 			nmp = TAILQ_NEXT(mp, mnt_list);
506 			continue;
507 		}
508 		vfslocked = VFS_LOCK_GIANT(mp);
509 		if (sfsp && count < maxcount) {
510 			sp = &mp->mnt_stat;
511 			/*
512 			 * Set these in case the underlying filesystem
513 			 * fails to do so.
514 			 */
515 			sp->f_version = STATFS_VERSION;
516 			sp->f_namemax = NAME_MAX;
517 			sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
518 			/*
519 			 * If MNT_NOWAIT or MNT_LAZY is specified, do not
520 			 * refresh the fsstat cache. MNT_NOWAIT or MNT_LAZY
521 			 * overrides MNT_WAIT.
522 			 */
523 			if (((flags & (MNT_LAZY|MNT_NOWAIT)) == 0 ||
524 			    (flags & MNT_WAIT)) &&
525 			    (error = VFS_STATFS(mp, sp))) {
526 				VFS_UNLOCK_GIANT(vfslocked);
527 				mtx_lock(&mountlist_mtx);
528 				nmp = TAILQ_NEXT(mp, mnt_list);
529 				vfs_unbusy(mp);
530 				continue;
531 			}
532 			if (priv_check(td, PRIV_VFS_GENERATION)) {
533 				bcopy(sp, &sb, sizeof(sb));
534 				sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
535 				prison_enforce_statfs(td->td_ucred, mp, &sb);
536 				sp = &sb;
537 			}
538 			if (bufseg == UIO_SYSSPACE)
539 				bcopy(sp, sfsp, sizeof(*sp));
540 			else /* if (bufseg == UIO_USERSPACE) */ {
541 				error = copyout(sp, sfsp, sizeof(*sp));
542 				if (error) {
543 					vfs_unbusy(mp);
544 					VFS_UNLOCK_GIANT(vfslocked);
545 					return (error);
546 				}
547 			}
548 			sfsp++;
549 		}
550 		VFS_UNLOCK_GIANT(vfslocked);
551 		count++;
552 		mtx_lock(&mountlist_mtx);
553 		nmp = TAILQ_NEXT(mp, mnt_list);
554 		vfs_unbusy(mp);
555 	}
556 	mtx_unlock(&mountlist_mtx);
557 	if (sfsp && count > maxcount)
558 		td->td_retval[0] = maxcount;
559 	else
560 		td->td_retval[0] = count;
561 	return (0);
562 }
563 
564 #ifdef COMPAT_FREEBSD4
565 /*
566  * Get old format filesystem statistics.
567  */
568 static void cvtstatfs(struct statfs *, struct ostatfs *);
569 
570 #ifndef _SYS_SYSPROTO_H_
571 struct freebsd4_statfs_args {
572 	char *path;
573 	struct ostatfs *buf;
574 };
575 #endif
576 int
577 freebsd4_statfs(td, uap)
578 	struct thread *td;
579 	struct freebsd4_statfs_args /* {
580 		char *path;
581 		struct ostatfs *buf;
582 	} */ *uap;
583 {
584 	struct ostatfs osb;
585 	struct statfs sf;
586 	int error;
587 
588 	error = kern_statfs(td, uap->path, UIO_USERSPACE, &sf);
589 	if (error)
590 		return (error);
591 	cvtstatfs(&sf, &osb);
592 	return (copyout(&osb, uap->buf, sizeof(osb)));
593 }
594 
595 /*
596  * Get filesystem statistics.
597  */
598 #ifndef _SYS_SYSPROTO_H_
599 struct freebsd4_fstatfs_args {
600 	int fd;
601 	struct ostatfs *buf;
602 };
603 #endif
604 int
605 freebsd4_fstatfs(td, uap)
606 	struct thread *td;
607 	struct freebsd4_fstatfs_args /* {
608 		int fd;
609 		struct ostatfs *buf;
610 	} */ *uap;
611 {
612 	struct ostatfs osb;
613 	struct statfs sf;
614 	int error;
615 
616 	error = kern_fstatfs(td, uap->fd, &sf);
617 	if (error)
618 		return (error);
619 	cvtstatfs(&sf, &osb);
620 	return (copyout(&osb, uap->buf, sizeof(osb)));
621 }
622 
623 /*
624  * Get statistics on all filesystems.
625  */
626 #ifndef _SYS_SYSPROTO_H_
627 struct freebsd4_getfsstat_args {
628 	struct ostatfs *buf;
629 	long bufsize;
630 	int flags;
631 };
632 #endif
633 int
634 freebsd4_getfsstat(td, uap)
635 	struct thread *td;
636 	register struct freebsd4_getfsstat_args /* {
637 		struct ostatfs *buf;
638 		long bufsize;
639 		int flags;
640 	} */ *uap;
641 {
642 	struct statfs *buf, *sp;
643 	struct ostatfs osb;
644 	size_t count, size;
645 	int error;
646 
647 	count = uap->bufsize / sizeof(struct ostatfs);
648 	size = count * sizeof(struct statfs);
649 	error = kern_getfsstat(td, &buf, size, UIO_SYSSPACE, uap->flags);
650 	if (size > 0) {
651 		count = td->td_retval[0];
652 		sp = buf;
653 		while (count > 0 && error == 0) {
654 			cvtstatfs(sp, &osb);
655 			error = copyout(&osb, uap->buf, sizeof(osb));
656 			sp++;
657 			uap->buf++;
658 			count--;
659 		}
660 		free(buf, M_TEMP);
661 	}
662 	return (error);
663 }
664 
665 /*
666  * Implement fstatfs() for (NFS) file handles.
667  */
668 #ifndef _SYS_SYSPROTO_H_
669 struct freebsd4_fhstatfs_args {
670 	struct fhandle *u_fhp;
671 	struct ostatfs *buf;
672 };
673 #endif
674 int
675 freebsd4_fhstatfs(td, uap)
676 	struct thread *td;
677 	struct freebsd4_fhstatfs_args /* {
678 		struct fhandle *u_fhp;
679 		struct ostatfs *buf;
680 	} */ *uap;
681 {
682 	struct ostatfs osb;
683 	struct statfs sf;
684 	fhandle_t fh;
685 	int error;
686 
687 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
688 	if (error)
689 		return (error);
690 	error = kern_fhstatfs(td, fh, &sf);
691 	if (error)
692 		return (error);
693 	cvtstatfs(&sf, &osb);
694 	return (copyout(&osb, uap->buf, sizeof(osb)));
695 }
696 
697 /*
698  * Convert a new format statfs structure to an old format statfs structure.
699  */
700 static void
701 cvtstatfs(nsp, osp)
702 	struct statfs *nsp;
703 	struct ostatfs *osp;
704 {
705 
706 	statfs_scale_blocks(nsp, LONG_MAX);
707 	bzero(osp, sizeof(*osp));
708 	osp->f_bsize = nsp->f_bsize;
709 	osp->f_iosize = MIN(nsp->f_iosize, LONG_MAX);
710 	osp->f_blocks = nsp->f_blocks;
711 	osp->f_bfree = nsp->f_bfree;
712 	osp->f_bavail = nsp->f_bavail;
713 	osp->f_files = MIN(nsp->f_files, LONG_MAX);
714 	osp->f_ffree = MIN(nsp->f_ffree, LONG_MAX);
715 	osp->f_owner = nsp->f_owner;
716 	osp->f_type = nsp->f_type;
717 	osp->f_flags = nsp->f_flags;
718 	osp->f_syncwrites = MIN(nsp->f_syncwrites, LONG_MAX);
719 	osp->f_asyncwrites = MIN(nsp->f_asyncwrites, LONG_MAX);
720 	osp->f_syncreads = MIN(nsp->f_syncreads, LONG_MAX);
721 	osp->f_asyncreads = MIN(nsp->f_asyncreads, LONG_MAX);
722 	strlcpy(osp->f_fstypename, nsp->f_fstypename,
723 	    MIN(MFSNAMELEN, OMFSNAMELEN));
724 	strlcpy(osp->f_mntonname, nsp->f_mntonname,
725 	    MIN(MNAMELEN, OMNAMELEN));
726 	strlcpy(osp->f_mntfromname, nsp->f_mntfromname,
727 	    MIN(MNAMELEN, OMNAMELEN));
728 	osp->f_fsid = nsp->f_fsid;
729 }
730 #endif /* COMPAT_FREEBSD4 */
731 
732 /*
733  * Change current working directory to a given file descriptor.
734  */
735 #ifndef _SYS_SYSPROTO_H_
736 struct fchdir_args {
737 	int	fd;
738 };
739 #endif
740 int
741 fchdir(td, uap)
742 	struct thread *td;
743 	struct fchdir_args /* {
744 		int fd;
745 	} */ *uap;
746 {
747 	register struct filedesc *fdp = td->td_proc->p_fd;
748 	struct vnode *vp, *tdp, *vpold;
749 	struct mount *mp;
750 	struct file *fp;
751 	int vfslocked;
752 	int error;
753 
754 	AUDIT_ARG(fd, uap->fd);
755 	if ((error = getvnode(fdp, uap->fd, &fp)) != 0)
756 		return (error);
757 	vp = fp->f_vnode;
758 	VREF(vp);
759 	fdrop(fp, td);
760 	vfslocked = VFS_LOCK_GIANT(vp->v_mount);
761 	vn_lock(vp, LK_SHARED | LK_RETRY);
762 	AUDIT_ARG(vnode, vp, ARG_VNODE1);
763 	error = change_dir(vp, td);
764 	while (!error && (mp = vp->v_mountedhere) != NULL) {
765 		int tvfslocked;
766 		if (vfs_busy(mp, 0))
767 			continue;
768 		tvfslocked = VFS_LOCK_GIANT(mp);
769 		error = VFS_ROOT(mp, LK_SHARED, &tdp);
770 		vfs_unbusy(mp);
771 		if (error) {
772 			VFS_UNLOCK_GIANT(tvfslocked);
773 			break;
774 		}
775 		vput(vp);
776 		VFS_UNLOCK_GIANT(vfslocked);
777 		vp = tdp;
778 		vfslocked = tvfslocked;
779 	}
780 	if (error) {
781 		vput(vp);
782 		VFS_UNLOCK_GIANT(vfslocked);
783 		return (error);
784 	}
785 	VOP_UNLOCK(vp, 0);
786 	VFS_UNLOCK_GIANT(vfslocked);
787 	FILEDESC_XLOCK(fdp);
788 	vpold = fdp->fd_cdir;
789 	fdp->fd_cdir = vp;
790 	FILEDESC_XUNLOCK(fdp);
791 	vfslocked = VFS_LOCK_GIANT(vpold->v_mount);
792 	vrele(vpold);
793 	VFS_UNLOCK_GIANT(vfslocked);
794 	return (0);
795 }
796 
797 /*
798  * Change current working directory (``.'').
799  */
800 #ifndef _SYS_SYSPROTO_H_
801 struct chdir_args {
802 	char	*path;
803 };
804 #endif
805 int
806 chdir(td, uap)
807 	struct thread *td;
808 	struct chdir_args /* {
809 		char *path;
810 	} */ *uap;
811 {
812 
813 	return (kern_chdir(td, uap->path, UIO_USERSPACE));
814 }
815 
816 int
817 kern_chdir(struct thread *td, char *path, enum uio_seg pathseg)
818 {
819 	register struct filedesc *fdp = td->td_proc->p_fd;
820 	int error;
821 	struct nameidata nd;
822 	struct vnode *vp;
823 	int vfslocked;
824 
825 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1 |
826 	    MPSAFE, pathseg, path, td);
827 	if ((error = namei(&nd)) != 0)
828 		return (error);
829 	vfslocked = NDHASGIANT(&nd);
830 	if ((error = change_dir(nd.ni_vp, td)) != 0) {
831 		vput(nd.ni_vp);
832 		VFS_UNLOCK_GIANT(vfslocked);
833 		NDFREE(&nd, NDF_ONLY_PNBUF);
834 		return (error);
835 	}
836 	VOP_UNLOCK(nd.ni_vp, 0);
837 	VFS_UNLOCK_GIANT(vfslocked);
838 	NDFREE(&nd, NDF_ONLY_PNBUF);
839 	FILEDESC_XLOCK(fdp);
840 	vp = fdp->fd_cdir;
841 	fdp->fd_cdir = nd.ni_vp;
842 	FILEDESC_XUNLOCK(fdp);
843 	vfslocked = VFS_LOCK_GIANT(vp->v_mount);
844 	vrele(vp);
845 	VFS_UNLOCK_GIANT(vfslocked);
846 	return (0);
847 }
848 
849 /*
850  * Helper function for raised chroot(2) security function:  Refuse if
851  * any filedescriptors are open directories.
852  */
853 static int
854 chroot_refuse_vdir_fds(fdp)
855 	struct filedesc *fdp;
856 {
857 	struct vnode *vp;
858 	struct file *fp;
859 	int fd;
860 
861 	FILEDESC_LOCK_ASSERT(fdp);
862 
863 	for (fd = 0; fd < fdp->fd_nfiles ; fd++) {
864 		fp = fget_locked(fdp, fd);
865 		if (fp == NULL)
866 			continue;
867 		if (fp->f_type == DTYPE_VNODE) {
868 			vp = fp->f_vnode;
869 			if (vp->v_type == VDIR)
870 				return (EPERM);
871 		}
872 	}
873 	return (0);
874 }
875 
876 /*
877  * This sysctl determines if we will allow a process to chroot(2) if it
878  * has a directory open:
879  *	0: disallowed for all processes.
880  *	1: allowed for processes that were not already chroot(2)'ed.
881  *	2: allowed for all processes.
882  */
883 
884 static int chroot_allow_open_directories = 1;
885 
886 SYSCTL_INT(_kern, OID_AUTO, chroot_allow_open_directories, CTLFLAG_RW,
887      &chroot_allow_open_directories, 0, "");
888 
889 /*
890  * Change notion of root (``/'') directory.
891  */
892 #ifndef _SYS_SYSPROTO_H_
893 struct chroot_args {
894 	char	*path;
895 };
896 #endif
897 int
898 chroot(td, uap)
899 	struct thread *td;
900 	struct chroot_args /* {
901 		char *path;
902 	} */ *uap;
903 {
904 	int error;
905 	struct nameidata nd;
906 	int vfslocked;
907 
908 	error = priv_check(td, PRIV_VFS_CHROOT);
909 	if (error)
910 		return (error);
911 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | MPSAFE |
912 	    AUDITVNODE1, UIO_USERSPACE, uap->path, td);
913 	error = namei(&nd);
914 	if (error)
915 		goto error;
916 	vfslocked = NDHASGIANT(&nd);
917 	if ((error = change_dir(nd.ni_vp, td)) != 0)
918 		goto e_vunlock;
919 #ifdef MAC
920 	if ((error = mac_vnode_check_chroot(td->td_ucred, nd.ni_vp)))
921 		goto e_vunlock;
922 #endif
923 	VOP_UNLOCK(nd.ni_vp, 0);
924 	error = change_root(nd.ni_vp, td);
925 	vrele(nd.ni_vp);
926 	VFS_UNLOCK_GIANT(vfslocked);
927 	NDFREE(&nd, NDF_ONLY_PNBUF);
928 	return (error);
929 e_vunlock:
930 	vput(nd.ni_vp);
931 	VFS_UNLOCK_GIANT(vfslocked);
932 error:
933 	NDFREE(&nd, NDF_ONLY_PNBUF);
934 	return (error);
935 }
936 
937 /*
938  * Common routine for chroot and chdir.  Callers must provide a locked vnode
939  * instance.
940  */
941 int
942 change_dir(vp, td)
943 	struct vnode *vp;
944 	struct thread *td;
945 {
946 	int error;
947 
948 	ASSERT_VOP_LOCKED(vp, "change_dir(): vp not locked");
949 	if (vp->v_type != VDIR)
950 		return (ENOTDIR);
951 #ifdef MAC
952 	error = mac_vnode_check_chdir(td->td_ucred, vp);
953 	if (error)
954 		return (error);
955 #endif
956 	error = VOP_ACCESS(vp, VEXEC, td->td_ucred, td);
957 	return (error);
958 }
959 
960 /*
961  * Common routine for kern_chroot() and jail_attach().  The caller is
962  * responsible for invoking priv_check() and mac_vnode_check_chroot() to
963  * authorize this operation.
964  */
965 int
966 change_root(vp, td)
967 	struct vnode *vp;
968 	struct thread *td;
969 {
970 	struct filedesc *fdp;
971 	struct vnode *oldvp;
972 	int vfslocked;
973 	int error;
974 
975 	VFS_ASSERT_GIANT(vp->v_mount);
976 	fdp = td->td_proc->p_fd;
977 	FILEDESC_XLOCK(fdp);
978 	if (chroot_allow_open_directories == 0 ||
979 	    (chroot_allow_open_directories == 1 && fdp->fd_rdir != rootvnode)) {
980 		error = chroot_refuse_vdir_fds(fdp);
981 		if (error) {
982 			FILEDESC_XUNLOCK(fdp);
983 			return (error);
984 		}
985 	}
986 	oldvp = fdp->fd_rdir;
987 	fdp->fd_rdir = vp;
988 	VREF(fdp->fd_rdir);
989 	if (!fdp->fd_jdir) {
990 		fdp->fd_jdir = vp;
991 		VREF(fdp->fd_jdir);
992 	}
993 	FILEDESC_XUNLOCK(fdp);
994 	vfslocked = VFS_LOCK_GIANT(oldvp->v_mount);
995 	vrele(oldvp);
996 	VFS_UNLOCK_GIANT(vfslocked);
997 	return (0);
998 }
999 
1000 /*
1001  * Check permissions, allocate an open file structure, and call the device
1002  * open routine if any.
1003  */
1004 #ifndef _SYS_SYSPROTO_H_
1005 struct open_args {
1006 	char	*path;
1007 	int	flags;
1008 	int	mode;
1009 };
1010 #endif
1011 int
1012 open(td, uap)
1013 	struct thread *td;
1014 	register struct open_args /* {
1015 		char *path;
1016 		int flags;
1017 		int mode;
1018 	} */ *uap;
1019 {
1020 
1021 	return (kern_open(td, uap->path, UIO_USERSPACE, uap->flags, uap->mode));
1022 }
1023 
1024 #ifndef _SYS_SYSPROTO_H_
1025 struct openat_args {
1026 	int	fd;
1027 	char	*path;
1028 	int	flag;
1029 	int	mode;
1030 };
1031 #endif
1032 int
1033 openat(struct thread *td, struct openat_args *uap)
1034 {
1035 
1036 	return (kern_openat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
1037 	    uap->mode));
1038 }
1039 
1040 int
1041 kern_open(struct thread *td, char *path, enum uio_seg pathseg, int flags,
1042     int mode)
1043 {
1044 
1045 	return (kern_openat(td, AT_FDCWD, path, pathseg, flags, mode));
1046 }
1047 
1048 int
1049 kern_openat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1050     int flags, int mode)
1051 {
1052 	struct proc *p = td->td_proc;
1053 	struct filedesc *fdp = p->p_fd;
1054 	struct file *fp;
1055 	struct vnode *vp;
1056 	struct vattr vat;
1057 	struct mount *mp;
1058 	int cmode;
1059 	struct file *nfp;
1060 	int type, indx, error;
1061 	struct flock lf;
1062 	struct nameidata nd;
1063 	int vfslocked;
1064 
1065 	AUDIT_ARG(fflags, flags);
1066 	AUDIT_ARG(mode, mode);
1067 	/* XXX: audit dirfd */
1068 	/*
1069 	 * Only one of the O_EXEC, O_RDONLY, O_WRONLY and O_RDWR may
1070 	 * be specified.
1071 	 */
1072 	if (flags & O_EXEC) {
1073 		if (flags & O_ACCMODE)
1074 			return (EINVAL);
1075 	} else if ((flags & O_ACCMODE) == O_ACCMODE)
1076 		return (EINVAL);
1077 	else
1078 		flags = FFLAGS(flags);
1079 
1080 	error = falloc(td, &nfp, &indx);
1081 	if (error)
1082 		return (error);
1083 	/* An extra reference on `nfp' has been held for us by falloc(). */
1084 	fp = nfp;
1085 	/* Set the flags early so the finit in devfs can pick them up. */
1086 	fp->f_flag = flags & FMASK;
1087 	cmode = ((mode &~ fdp->fd_cmask) & ALLPERMS) &~ S_ISTXT;
1088 	NDINIT_AT(&nd, LOOKUP, FOLLOW | AUDITVNODE1 | MPSAFE, pathseg, path, fd,
1089 	    td);
1090 	td->td_dupfd = -1;		/* XXX check for fdopen */
1091 	error = vn_open(&nd, &flags, cmode, fp);
1092 	if (error) {
1093 		/*
1094 		 * If the vn_open replaced the method vector, something
1095 		 * wonderous happened deep below and we just pass it up
1096 		 * pretending we know what we do.
1097 		 */
1098 		if (error == ENXIO && fp->f_ops != &badfileops) {
1099 			fdrop(fp, td);
1100 			td->td_retval[0] = indx;
1101 			return (0);
1102 		}
1103 
1104 		/*
1105 		 * handle special fdopen() case.  bleh.  dupfdopen() is
1106 		 * responsible for dropping the old contents of ofiles[indx]
1107 		 * if it succeeds.
1108 		 */
1109 		if ((error == ENODEV || error == ENXIO) &&
1110 		    td->td_dupfd >= 0 &&		/* XXX from fdopen */
1111 		    (error =
1112 			dupfdopen(td, fdp, indx, td->td_dupfd, flags, error)) == 0) {
1113 			td->td_retval[0] = indx;
1114 			fdrop(fp, td);
1115 			return (0);
1116 		}
1117 		/*
1118 		 * Clean up the descriptor, but only if another thread hadn't
1119 		 * replaced or closed it.
1120 		 */
1121 		fdclose(fdp, fp, indx, td);
1122 		fdrop(fp, td);
1123 
1124 		if (error == ERESTART)
1125 			error = EINTR;
1126 		return (error);
1127 	}
1128 	td->td_dupfd = 0;
1129 	vfslocked = NDHASGIANT(&nd);
1130 	NDFREE(&nd, NDF_ONLY_PNBUF);
1131 	vp = nd.ni_vp;
1132 
1133 	fp->f_vnode = vp;	/* XXX Does devfs need this? */
1134 	/*
1135 	 * If the file wasn't claimed by devfs bind it to the normal
1136 	 * vnode operations here.
1137 	 */
1138 	if (fp->f_ops == &badfileops) {
1139 		KASSERT(vp->v_type != VFIFO, ("Unexpected fifo."));
1140 		fp->f_seqcount = 1;
1141 		finit(fp, flags & FMASK, DTYPE_VNODE, vp, &vnops);
1142 	}
1143 
1144 	VOP_UNLOCK(vp, 0);
1145 	if (flags & (O_EXLOCK | O_SHLOCK)) {
1146 		lf.l_whence = SEEK_SET;
1147 		lf.l_start = 0;
1148 		lf.l_len = 0;
1149 		if (flags & O_EXLOCK)
1150 			lf.l_type = F_WRLCK;
1151 		else
1152 			lf.l_type = F_RDLCK;
1153 		type = F_FLOCK;
1154 		if ((flags & FNONBLOCK) == 0)
1155 			type |= F_WAIT;
1156 		if ((error = VOP_ADVLOCK(vp, (caddr_t)fp, F_SETLK, &lf,
1157 			    type)) != 0)
1158 			goto bad;
1159 		atomic_set_int(&fp->f_flag, FHASLOCK);
1160 	}
1161 	if (flags & O_TRUNC) {
1162 		if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
1163 			goto bad;
1164 		VATTR_NULL(&vat);
1165 		vat.va_size = 0;
1166 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1167 #ifdef MAC
1168 		error = mac_vnode_check_write(td->td_ucred, fp->f_cred, vp);
1169 		if (error == 0)
1170 #endif
1171 			error = VOP_SETATTR(vp, &vat, td->td_ucred);
1172 		VOP_UNLOCK(vp, 0);
1173 		vn_finished_write(mp);
1174 		if (error)
1175 			goto bad;
1176 	}
1177 	VFS_UNLOCK_GIANT(vfslocked);
1178 	/*
1179 	 * Release our private reference, leaving the one associated with
1180 	 * the descriptor table intact.
1181 	 */
1182 	fdrop(fp, td);
1183 	td->td_retval[0] = indx;
1184 	return (0);
1185 bad:
1186 	VFS_UNLOCK_GIANT(vfslocked);
1187 	fdclose(fdp, fp, indx, td);
1188 	fdrop(fp, td);
1189 	return (error);
1190 }
1191 
1192 #ifdef COMPAT_43
1193 /*
1194  * Create a file.
1195  */
1196 #ifndef _SYS_SYSPROTO_H_
1197 struct ocreat_args {
1198 	char	*path;
1199 	int	mode;
1200 };
1201 #endif
1202 int
1203 ocreat(td, uap)
1204 	struct thread *td;
1205 	register struct ocreat_args /* {
1206 		char *path;
1207 		int mode;
1208 	} */ *uap;
1209 {
1210 
1211 	return (kern_open(td, uap->path, UIO_USERSPACE,
1212 	    O_WRONLY | O_CREAT | O_TRUNC, uap->mode));
1213 }
1214 #endif /* COMPAT_43 */
1215 
1216 /*
1217  * Create a special file.
1218  */
1219 #ifndef _SYS_SYSPROTO_H_
1220 struct mknod_args {
1221 	char	*path;
1222 	int	mode;
1223 	int	dev;
1224 };
1225 #endif
1226 int
1227 mknod(td, uap)
1228 	struct thread *td;
1229 	register struct mknod_args /* {
1230 		char *path;
1231 		int mode;
1232 		int dev;
1233 	} */ *uap;
1234 {
1235 
1236 	return (kern_mknod(td, uap->path, UIO_USERSPACE, uap->mode, uap->dev));
1237 }
1238 
1239 #ifndef _SYS_SYSPROTO_H_
1240 struct mknodat_args {
1241 	int	fd;
1242 	char	*path;
1243 	mode_t	mode;
1244 	dev_t	dev;
1245 };
1246 #endif
1247 int
1248 mknodat(struct thread *td, struct mknodat_args *uap)
1249 {
1250 
1251 	return (kern_mknodat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode,
1252 	    uap->dev));
1253 }
1254 
1255 int
1256 kern_mknod(struct thread *td, char *path, enum uio_seg pathseg, int mode,
1257     int dev)
1258 {
1259 
1260 	return (kern_mknodat(td, AT_FDCWD, path, pathseg, mode, dev));
1261 }
1262 
1263 int
1264 kern_mknodat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1265     int mode, int dev)
1266 {
1267 	struct vnode *vp;
1268 	struct mount *mp;
1269 	struct vattr vattr;
1270 	int error;
1271 	int whiteout = 0;
1272 	struct nameidata nd;
1273 	int vfslocked;
1274 
1275 	AUDIT_ARG(mode, mode);
1276 	AUDIT_ARG(dev, dev);
1277 	switch (mode & S_IFMT) {
1278 	case S_IFCHR:
1279 	case S_IFBLK:
1280 		error = priv_check(td, PRIV_VFS_MKNOD_DEV);
1281 		break;
1282 	case S_IFMT:
1283 		error = priv_check(td, PRIV_VFS_MKNOD_BAD);
1284 		break;
1285 	case S_IFWHT:
1286 		error = priv_check(td, PRIV_VFS_MKNOD_WHT);
1287 		break;
1288 	case S_IFIFO:
1289 		if (dev == 0)
1290 			return (kern_mkfifoat(td, fd, path, pathseg, mode));
1291 		/* FALLTHROUGH */
1292 	default:
1293 		error = EINVAL;
1294 		break;
1295 	}
1296 	if (error)
1297 		return (error);
1298 restart:
1299 	bwillwrite();
1300 	NDINIT_AT(&nd, CREATE, LOCKPARENT | SAVENAME | MPSAFE | AUDITVNODE1,
1301 	    pathseg, path, fd, td);
1302 	if ((error = namei(&nd)) != 0)
1303 		return (error);
1304 	vfslocked = NDHASGIANT(&nd);
1305 	vp = nd.ni_vp;
1306 	if (vp != NULL) {
1307 		NDFREE(&nd, NDF_ONLY_PNBUF);
1308 		if (vp == nd.ni_dvp)
1309 			vrele(nd.ni_dvp);
1310 		else
1311 			vput(nd.ni_dvp);
1312 		vrele(vp);
1313 		VFS_UNLOCK_GIANT(vfslocked);
1314 		return (EEXIST);
1315 	} else {
1316 		VATTR_NULL(&vattr);
1317 		FILEDESC_SLOCK(td->td_proc->p_fd);
1318 		vattr.va_mode = (mode & ALLPERMS) &
1319 		    ~td->td_proc->p_fd->fd_cmask;
1320 		FILEDESC_SUNLOCK(td->td_proc->p_fd);
1321 		vattr.va_rdev = dev;
1322 		whiteout = 0;
1323 
1324 		switch (mode & S_IFMT) {
1325 		case S_IFMT:	/* used by badsect to flag bad sectors */
1326 			vattr.va_type = VBAD;
1327 			break;
1328 		case S_IFCHR:
1329 			vattr.va_type = VCHR;
1330 			break;
1331 		case S_IFBLK:
1332 			vattr.va_type = VBLK;
1333 			break;
1334 		case S_IFWHT:
1335 			whiteout = 1;
1336 			break;
1337 		default:
1338 			panic("kern_mknod: invalid mode");
1339 		}
1340 	}
1341 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1342 		NDFREE(&nd, NDF_ONLY_PNBUF);
1343 		vput(nd.ni_dvp);
1344 		VFS_UNLOCK_GIANT(vfslocked);
1345 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1346 			return (error);
1347 		goto restart;
1348 	}
1349 #ifdef MAC
1350 	if (error == 0 && !whiteout)
1351 		error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp,
1352 		    &nd.ni_cnd, &vattr);
1353 #endif
1354 	if (!error) {
1355 		if (whiteout)
1356 			error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE);
1357 		else {
1358 			error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp,
1359 						&nd.ni_cnd, &vattr);
1360 			if (error == 0)
1361 				vput(nd.ni_vp);
1362 		}
1363 	}
1364 	NDFREE(&nd, NDF_ONLY_PNBUF);
1365 	vput(nd.ni_dvp);
1366 	vn_finished_write(mp);
1367 	VFS_UNLOCK_GIANT(vfslocked);
1368 	return (error);
1369 }
1370 
1371 /*
1372  * Create a named pipe.
1373  */
1374 #ifndef _SYS_SYSPROTO_H_
1375 struct mkfifo_args {
1376 	char	*path;
1377 	int	mode;
1378 };
1379 #endif
1380 int
1381 mkfifo(td, uap)
1382 	struct thread *td;
1383 	register struct mkfifo_args /* {
1384 		char *path;
1385 		int mode;
1386 	} */ *uap;
1387 {
1388 
1389 	return (kern_mkfifo(td, uap->path, UIO_USERSPACE, uap->mode));
1390 }
1391 
1392 #ifndef _SYS_SYSPROTO_H_
1393 struct mkfifoat_args {
1394 	int	fd;
1395 	char	*path;
1396 	mode_t	mode;
1397 };
1398 #endif
1399 int
1400 mkfifoat(struct thread *td, struct mkfifoat_args *uap)
1401 {
1402 
1403 	return (kern_mkfifoat(td, uap->fd, uap->path, UIO_USERSPACE,
1404 	    uap->mode));
1405 }
1406 
1407 int
1408 kern_mkfifo(struct thread *td, char *path, enum uio_seg pathseg, int mode)
1409 {
1410 
1411 	return (kern_mkfifoat(td, AT_FDCWD, path, pathseg, mode));
1412 }
1413 
1414 int
1415 kern_mkfifoat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1416     int mode)
1417 {
1418 	struct mount *mp;
1419 	struct vattr vattr;
1420 	int error;
1421 	struct nameidata nd;
1422 	int vfslocked;
1423 
1424 	AUDIT_ARG(mode, mode);
1425 restart:
1426 	bwillwrite();
1427 	NDINIT_AT(&nd, CREATE, LOCKPARENT | SAVENAME | MPSAFE | AUDITVNODE1,
1428 	    pathseg, path, fd, td);
1429 	if ((error = namei(&nd)) != 0)
1430 		return (error);
1431 	vfslocked = NDHASGIANT(&nd);
1432 	if (nd.ni_vp != NULL) {
1433 		NDFREE(&nd, NDF_ONLY_PNBUF);
1434 		if (nd.ni_vp == nd.ni_dvp)
1435 			vrele(nd.ni_dvp);
1436 		else
1437 			vput(nd.ni_dvp);
1438 		vrele(nd.ni_vp);
1439 		VFS_UNLOCK_GIANT(vfslocked);
1440 		return (EEXIST);
1441 	}
1442 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1443 		NDFREE(&nd, NDF_ONLY_PNBUF);
1444 		vput(nd.ni_dvp);
1445 		VFS_UNLOCK_GIANT(vfslocked);
1446 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1447 			return (error);
1448 		goto restart;
1449 	}
1450 	VATTR_NULL(&vattr);
1451 	vattr.va_type = VFIFO;
1452 	FILEDESC_SLOCK(td->td_proc->p_fd);
1453 	vattr.va_mode = (mode & ALLPERMS) & ~td->td_proc->p_fd->fd_cmask;
1454 	FILEDESC_SUNLOCK(td->td_proc->p_fd);
1455 #ifdef MAC
1456 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1457 	    &vattr);
1458 	if (error)
1459 		goto out;
1460 #endif
1461 	error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
1462 	if (error == 0)
1463 		vput(nd.ni_vp);
1464 #ifdef MAC
1465 out:
1466 #endif
1467 	vput(nd.ni_dvp);
1468 	vn_finished_write(mp);
1469 	VFS_UNLOCK_GIANT(vfslocked);
1470 	NDFREE(&nd, NDF_ONLY_PNBUF);
1471 	return (error);
1472 }
1473 
1474 /*
1475  * Make a hard file link.
1476  */
1477 #ifndef _SYS_SYSPROTO_H_
1478 struct link_args {
1479 	char	*path;
1480 	char	*link;
1481 };
1482 #endif
1483 int
1484 link(td, uap)
1485 	struct thread *td;
1486 	register struct link_args /* {
1487 		char *path;
1488 		char *link;
1489 	} */ *uap;
1490 {
1491 
1492 	return (kern_link(td, uap->path, uap->link, UIO_USERSPACE));
1493 }
1494 
1495 #ifndef _SYS_SYSPROTO_H_
1496 struct linkat_args {
1497 	int	fd1;
1498 	char	*path1;
1499 	int	fd2;
1500 	char	*path2;
1501 	int	flag;
1502 };
1503 #endif
1504 int
1505 linkat(struct thread *td, struct linkat_args *uap)
1506 {
1507 	int flag;
1508 
1509 	flag = uap->flag;
1510 	if (flag & ~AT_SYMLINK_FOLLOW)
1511 		return (EINVAL);
1512 
1513 	return (kern_linkat(td, uap->fd1, uap->fd2, uap->path1, uap->path2,
1514 	    UIO_USERSPACE, (flag & AT_SYMLINK_FOLLOW) ? FOLLOW : NOFOLLOW));
1515 }
1516 
1517 static int hardlink_check_uid = 0;
1518 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_uid, CTLFLAG_RW,
1519     &hardlink_check_uid, 0,
1520     "Unprivileged processes cannot create hard links to files owned by other "
1521     "users");
1522 static int hardlink_check_gid = 0;
1523 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_gid, CTLFLAG_RW,
1524     &hardlink_check_gid, 0,
1525     "Unprivileged processes cannot create hard links to files owned by other "
1526     "groups");
1527 
1528 static int
1529 can_hardlink(struct vnode *vp, struct ucred *cred)
1530 {
1531 	struct vattr va;
1532 	int error;
1533 
1534 	if (!hardlink_check_uid && !hardlink_check_gid)
1535 		return (0);
1536 
1537 	error = VOP_GETATTR(vp, &va, cred);
1538 	if (error != 0)
1539 		return (error);
1540 
1541 	if (hardlink_check_uid && cred->cr_uid != va.va_uid) {
1542 		error = priv_check_cred(cred, PRIV_VFS_LINK, 0);
1543 		if (error)
1544 			return (error);
1545 	}
1546 
1547 	if (hardlink_check_gid && !groupmember(va.va_gid, cred)) {
1548 		error = priv_check_cred(cred, PRIV_VFS_LINK, 0);
1549 		if (error)
1550 			return (error);
1551 	}
1552 
1553 	return (0);
1554 }
1555 
1556 int
1557 kern_link(struct thread *td, char *path, char *link, enum uio_seg segflg)
1558 {
1559 
1560 	return (kern_linkat(td, AT_FDCWD, AT_FDCWD, path,link, segflg, FOLLOW));
1561 }
1562 
1563 int
1564 kern_linkat(struct thread *td, int fd1, int fd2, char *path1, char *path2,
1565     enum uio_seg segflg, int follow)
1566 {
1567 	struct vnode *vp;
1568 	struct mount *mp;
1569 	struct nameidata nd;
1570 	int vfslocked;
1571 	int lvfslocked;
1572 	int error;
1573 
1574 	bwillwrite();
1575 	NDINIT_AT(&nd, LOOKUP, follow | MPSAFE | AUDITVNODE1, segflg, path1,
1576 	    fd1, td);
1577 
1578 	if ((error = namei(&nd)) != 0)
1579 		return (error);
1580 	vfslocked = NDHASGIANT(&nd);
1581 	NDFREE(&nd, NDF_ONLY_PNBUF);
1582 	vp = nd.ni_vp;
1583 	if (vp->v_type == VDIR) {
1584 		vrele(vp);
1585 		VFS_UNLOCK_GIANT(vfslocked);
1586 		return (EPERM);		/* POSIX */
1587 	}
1588 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) {
1589 		vrele(vp);
1590 		VFS_UNLOCK_GIANT(vfslocked);
1591 		return (error);
1592 	}
1593 	NDINIT_AT(&nd, CREATE, LOCKPARENT | SAVENAME | MPSAFE | AUDITVNODE1,
1594 	    segflg, path2, fd2, td);
1595 	if ((error = namei(&nd)) == 0) {
1596 		lvfslocked = NDHASGIANT(&nd);
1597 		if (nd.ni_vp != NULL) {
1598 			if (nd.ni_dvp == nd.ni_vp)
1599 				vrele(nd.ni_dvp);
1600 			else
1601 				vput(nd.ni_dvp);
1602 			vrele(nd.ni_vp);
1603 			error = EEXIST;
1604 		} else if ((error = vn_lock(vp, LK_EXCLUSIVE | LK_RETRY))
1605 		    == 0) {
1606 			error = can_hardlink(vp, td->td_ucred);
1607 			if (error == 0)
1608 #ifdef MAC
1609 				error = mac_vnode_check_link(td->td_ucred,
1610 				    nd.ni_dvp, vp, &nd.ni_cnd);
1611 			if (error == 0)
1612 #endif
1613 				error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
1614 			VOP_UNLOCK(vp, 0);
1615 			vput(nd.ni_dvp);
1616 		}
1617 		NDFREE(&nd, NDF_ONLY_PNBUF);
1618 		VFS_UNLOCK_GIANT(lvfslocked);
1619 	}
1620 	vrele(vp);
1621 	vn_finished_write(mp);
1622 	VFS_UNLOCK_GIANT(vfslocked);
1623 	return (error);
1624 }
1625 
1626 /*
1627  * Make a symbolic link.
1628  */
1629 #ifndef _SYS_SYSPROTO_H_
1630 struct symlink_args {
1631 	char	*path;
1632 	char	*link;
1633 };
1634 #endif
1635 int
1636 symlink(td, uap)
1637 	struct thread *td;
1638 	register struct symlink_args /* {
1639 		char *path;
1640 		char *link;
1641 	} */ *uap;
1642 {
1643 
1644 	return (kern_symlink(td, uap->path, uap->link, UIO_USERSPACE));
1645 }
1646 
1647 #ifndef _SYS_SYSPROTO_H_
1648 struct symlinkat_args {
1649 	char	*path;
1650 	int	fd;
1651 	char	*path2;
1652 };
1653 #endif
1654 int
1655 symlinkat(struct thread *td, struct symlinkat_args *uap)
1656 {
1657 
1658 	return (kern_symlinkat(td, uap->path1, uap->fd, uap->path2,
1659 	    UIO_USERSPACE));
1660 }
1661 
1662 int
1663 kern_symlink(struct thread *td, char *path, char *link, enum uio_seg segflg)
1664 {
1665 
1666 	return (kern_symlinkat(td, path, AT_FDCWD, link, segflg));
1667 }
1668 
1669 int
1670 kern_symlinkat(struct thread *td, char *path1, int fd, char *path2,
1671     enum uio_seg segflg)
1672 {
1673 	struct mount *mp;
1674 	struct vattr vattr;
1675 	char *syspath;
1676 	int error;
1677 	struct nameidata nd;
1678 	int vfslocked;
1679 
1680 	if (segflg == UIO_SYSSPACE) {
1681 		syspath = path1;
1682 	} else {
1683 		syspath = uma_zalloc(namei_zone, M_WAITOK);
1684 		if ((error = copyinstr(path1, syspath, MAXPATHLEN, NULL)) != 0)
1685 			goto out;
1686 	}
1687 	AUDIT_ARG(text, syspath);
1688 restart:
1689 	bwillwrite();
1690 	NDINIT_AT(&nd, CREATE, LOCKPARENT | SAVENAME | MPSAFE | AUDITVNODE1,
1691 	    segflg, path2, fd, td);
1692 	if ((error = namei(&nd)) != 0)
1693 		goto out;
1694 	vfslocked = NDHASGIANT(&nd);
1695 	if (nd.ni_vp) {
1696 		NDFREE(&nd, NDF_ONLY_PNBUF);
1697 		if (nd.ni_vp == nd.ni_dvp)
1698 			vrele(nd.ni_dvp);
1699 		else
1700 			vput(nd.ni_dvp);
1701 		vrele(nd.ni_vp);
1702 		VFS_UNLOCK_GIANT(vfslocked);
1703 		error = EEXIST;
1704 		goto out;
1705 	}
1706 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1707 		NDFREE(&nd, NDF_ONLY_PNBUF);
1708 		vput(nd.ni_dvp);
1709 		VFS_UNLOCK_GIANT(vfslocked);
1710 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1711 			goto out;
1712 		goto restart;
1713 	}
1714 	VATTR_NULL(&vattr);
1715 	FILEDESC_SLOCK(td->td_proc->p_fd);
1716 	vattr.va_mode = ACCESSPERMS &~ td->td_proc->p_fd->fd_cmask;
1717 	FILEDESC_SUNLOCK(td->td_proc->p_fd);
1718 #ifdef MAC
1719 	vattr.va_type = VLNK;
1720 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1721 	    &vattr);
1722 	if (error)
1723 		goto out2;
1724 #endif
1725 	error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, syspath);
1726 	if (error == 0)
1727 		vput(nd.ni_vp);
1728 #ifdef MAC
1729 out2:
1730 #endif
1731 	NDFREE(&nd, NDF_ONLY_PNBUF);
1732 	vput(nd.ni_dvp);
1733 	vn_finished_write(mp);
1734 	VFS_UNLOCK_GIANT(vfslocked);
1735 out:
1736 	if (segflg != UIO_SYSSPACE)
1737 		uma_zfree(namei_zone, syspath);
1738 	return (error);
1739 }
1740 
1741 /*
1742  * Delete a whiteout from the filesystem.
1743  */
1744 int
1745 undelete(td, uap)
1746 	struct thread *td;
1747 	register struct undelete_args /* {
1748 		char *path;
1749 	} */ *uap;
1750 {
1751 	int error;
1752 	struct mount *mp;
1753 	struct nameidata nd;
1754 	int vfslocked;
1755 
1756 restart:
1757 	bwillwrite();
1758 	NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | MPSAFE | AUDITVNODE1,
1759 	    UIO_USERSPACE, uap->path, td);
1760 	error = namei(&nd);
1761 	if (error)
1762 		return (error);
1763 	vfslocked = NDHASGIANT(&nd);
1764 
1765 	if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) {
1766 		NDFREE(&nd, NDF_ONLY_PNBUF);
1767 		if (nd.ni_vp == nd.ni_dvp)
1768 			vrele(nd.ni_dvp);
1769 		else
1770 			vput(nd.ni_dvp);
1771 		if (nd.ni_vp)
1772 			vrele(nd.ni_vp);
1773 		VFS_UNLOCK_GIANT(vfslocked);
1774 		return (EEXIST);
1775 	}
1776 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1777 		NDFREE(&nd, NDF_ONLY_PNBUF);
1778 		vput(nd.ni_dvp);
1779 		VFS_UNLOCK_GIANT(vfslocked);
1780 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1781 			return (error);
1782 		goto restart;
1783 	}
1784 	error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE);
1785 	NDFREE(&nd, NDF_ONLY_PNBUF);
1786 	vput(nd.ni_dvp);
1787 	vn_finished_write(mp);
1788 	VFS_UNLOCK_GIANT(vfslocked);
1789 	return (error);
1790 }
1791 
1792 /*
1793  * Delete a name from the filesystem.
1794  */
1795 #ifndef _SYS_SYSPROTO_H_
1796 struct unlink_args {
1797 	char	*path;
1798 };
1799 #endif
1800 int
1801 unlink(td, uap)
1802 	struct thread *td;
1803 	struct unlink_args /* {
1804 		char *path;
1805 	} */ *uap;
1806 {
1807 
1808 	return (kern_unlink(td, uap->path, UIO_USERSPACE));
1809 }
1810 
1811 #ifndef _SYS_SYSPROTO_H_
1812 struct unlinkat_args {
1813 	int	fd;
1814 	char	*path;
1815 	int	flag;
1816 };
1817 #endif
1818 int
1819 unlinkat(struct thread *td, struct unlinkat_args *uap)
1820 {
1821 	int flag = uap->flag;
1822 	int fd = uap->fd;
1823 	char *path = uap->path;
1824 
1825 	if (flag & ~AT_REMOVEDIR)
1826 		return (EINVAL);
1827 
1828 	if (flag & AT_REMOVEDIR)
1829 		return (kern_rmdirat(td, fd, path, UIO_USERSPACE));
1830 	else
1831 		return (kern_unlinkat(td, fd, path, UIO_USERSPACE));
1832 }
1833 
1834 int
1835 kern_unlink(struct thread *td, char *path, enum uio_seg pathseg)
1836 {
1837 
1838 	return (kern_unlinkat(td, AT_FDCWD, path, pathseg));
1839 }
1840 
1841 int
1842 kern_unlinkat(struct thread *td, int fd, char *path, enum uio_seg pathseg)
1843 {
1844 	struct mount *mp;
1845 	struct vnode *vp;
1846 	int error;
1847 	struct nameidata nd;
1848 	int vfslocked;
1849 
1850 restart:
1851 	bwillwrite();
1852 	NDINIT_AT(&nd, DELETE, LOCKPARENT | LOCKLEAF | MPSAFE | AUDITVNODE1,
1853 	    pathseg, path, fd, td);
1854 	if ((error = namei(&nd)) != 0)
1855 		return (error == EINVAL ? EPERM : error);
1856 	vfslocked = NDHASGIANT(&nd);
1857 	vp = nd.ni_vp;
1858 	if (vp->v_type == VDIR)
1859 		error = EPERM;		/* POSIX */
1860 	else {
1861 		/*
1862 		 * The root of a mounted filesystem cannot be deleted.
1863 		 *
1864 		 * XXX: can this only be a VDIR case?
1865 		 */
1866 		if (vp->v_vflag & VV_ROOT)
1867 			error = EBUSY;
1868 	}
1869 	if (error == 0) {
1870 		if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1871 			NDFREE(&nd, NDF_ONLY_PNBUF);
1872 			vput(nd.ni_dvp);
1873 			if (vp == nd.ni_dvp)
1874 				vrele(vp);
1875 			else
1876 				vput(vp);
1877 			VFS_UNLOCK_GIANT(vfslocked);
1878 			if ((error = vn_start_write(NULL, &mp,
1879 			    V_XSLEEP | PCATCH)) != 0)
1880 				return (error);
1881 			goto restart;
1882 		}
1883 #ifdef MAC
1884 		error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
1885 		    &nd.ni_cnd);
1886 		if (error)
1887 			goto out;
1888 #endif
1889 		error = VOP_REMOVE(nd.ni_dvp, vp, &nd.ni_cnd);
1890 #ifdef MAC
1891 out:
1892 #endif
1893 		vn_finished_write(mp);
1894 	}
1895 	NDFREE(&nd, NDF_ONLY_PNBUF);
1896 	vput(nd.ni_dvp);
1897 	if (vp == nd.ni_dvp)
1898 		vrele(vp);
1899 	else
1900 		vput(vp);
1901 	VFS_UNLOCK_GIANT(vfslocked);
1902 	return (error);
1903 }
1904 
1905 /*
1906  * Reposition read/write file offset.
1907  */
1908 #ifndef _SYS_SYSPROTO_H_
1909 struct lseek_args {
1910 	int	fd;
1911 	int	pad;
1912 	off_t	offset;
1913 	int	whence;
1914 };
1915 #endif
1916 int
1917 lseek(td, uap)
1918 	struct thread *td;
1919 	register struct lseek_args /* {
1920 		int fd;
1921 		int pad;
1922 		off_t offset;
1923 		int whence;
1924 	} */ *uap;
1925 {
1926 	struct ucred *cred = td->td_ucred;
1927 	struct file *fp;
1928 	struct vnode *vp;
1929 	struct vattr vattr;
1930 	off_t offset;
1931 	int error, noneg;
1932 	int vfslocked;
1933 
1934 	if ((error = fget(td, uap->fd, &fp)) != 0)
1935 		return (error);
1936 	if (!(fp->f_ops->fo_flags & DFLAG_SEEKABLE)) {
1937 		fdrop(fp, td);
1938 		return (ESPIPE);
1939 	}
1940 	vp = fp->f_vnode;
1941 	vfslocked = VFS_LOCK_GIANT(vp->v_mount);
1942 	noneg = (vp->v_type != VCHR);
1943 	offset = uap->offset;
1944 	switch (uap->whence) {
1945 	case L_INCR:
1946 		if (noneg &&
1947 		    (fp->f_offset < 0 ||
1948 		    (offset > 0 && fp->f_offset > OFF_MAX - offset))) {
1949 			error = EOVERFLOW;
1950 			break;
1951 		}
1952 		offset += fp->f_offset;
1953 		break;
1954 	case L_XTND:
1955 		vn_lock(vp, LK_SHARED | LK_RETRY);
1956 		error = VOP_GETATTR(vp, &vattr, cred);
1957 		VOP_UNLOCK(vp, 0);
1958 		if (error)
1959 			break;
1960 		if (noneg &&
1961 		    (vattr.va_size > OFF_MAX ||
1962 		    (offset > 0 && vattr.va_size > OFF_MAX - offset))) {
1963 			error = EOVERFLOW;
1964 			break;
1965 		}
1966 		offset += vattr.va_size;
1967 		break;
1968 	case L_SET:
1969 		break;
1970 	case SEEK_DATA:
1971 		error = fo_ioctl(fp, FIOSEEKDATA, &offset, cred, td);
1972 		break;
1973 	case SEEK_HOLE:
1974 		error = fo_ioctl(fp, FIOSEEKHOLE, &offset, cred, td);
1975 		break;
1976 	default:
1977 		error = EINVAL;
1978 	}
1979 	if (error == 0 && noneg && offset < 0)
1980 		error = EINVAL;
1981 	if (error != 0)
1982 		goto drop;
1983 	fp->f_offset = offset;
1984 	*(off_t *)(td->td_retval) = fp->f_offset;
1985 drop:
1986 	fdrop(fp, td);
1987 	VFS_UNLOCK_GIANT(vfslocked);
1988 	return (error);
1989 }
1990 
1991 #if defined(COMPAT_43)
1992 /*
1993  * Reposition read/write file offset.
1994  */
1995 #ifndef _SYS_SYSPROTO_H_
1996 struct olseek_args {
1997 	int	fd;
1998 	long	offset;
1999 	int	whence;
2000 };
2001 #endif
2002 int
2003 olseek(td, uap)
2004 	struct thread *td;
2005 	register struct olseek_args /* {
2006 		int fd;
2007 		long offset;
2008 		int whence;
2009 	} */ *uap;
2010 {
2011 	struct lseek_args /* {
2012 		int fd;
2013 		int pad;
2014 		off_t offset;
2015 		int whence;
2016 	} */ nuap;
2017 
2018 	nuap.fd = uap->fd;
2019 	nuap.offset = uap->offset;
2020 	nuap.whence = uap->whence;
2021 	return (lseek(td, &nuap));
2022 }
2023 #endif /* COMPAT_43 */
2024 
2025 /* Version with the 'pad' argument */
2026 int
2027 freebsd6_lseek(td, uap)
2028 	struct thread *td;
2029 	register struct freebsd6_lseek_args *uap;
2030 {
2031 	struct lseek_args ouap;
2032 
2033 	ouap.fd = uap->fd;
2034 	ouap.offset = uap->offset;
2035 	ouap.whence = uap->whence;
2036 	return (lseek(td, &ouap));
2037 }
2038 
2039 /*
2040  * Check access permissions using passed credentials.
2041  */
2042 static int
2043 vn_access(vp, user_flags, cred, td)
2044 	struct vnode	*vp;
2045 	int		user_flags;
2046 	struct ucred	*cred;
2047 	struct thread	*td;
2048 {
2049 	int error;
2050 	accmode_t accmode;
2051 
2052 	/* Flags == 0 means only check for existence. */
2053 	error = 0;
2054 	if (user_flags) {
2055 		accmode = 0;
2056 		if (user_flags & R_OK)
2057 			accmode |= VREAD;
2058 		if (user_flags & W_OK)
2059 			accmode |= VWRITE;
2060 		if (user_flags & X_OK)
2061 			accmode |= VEXEC;
2062 #ifdef MAC
2063 		error = mac_vnode_check_access(cred, vp, accmode);
2064 		if (error)
2065 			return (error);
2066 #endif
2067 		if ((accmode & VWRITE) == 0 || (error = vn_writechk(vp)) == 0)
2068 			error = VOP_ACCESS(vp, accmode, cred, td);
2069 	}
2070 	return (error);
2071 }
2072 
2073 /*
2074  * Check access permissions using "real" credentials.
2075  */
2076 #ifndef _SYS_SYSPROTO_H_
2077 struct access_args {
2078 	char	*path;
2079 	int	flags;
2080 };
2081 #endif
2082 int
2083 access(td, uap)
2084 	struct thread *td;
2085 	register struct access_args /* {
2086 		char *path;
2087 		int flags;
2088 	} */ *uap;
2089 {
2090 
2091 	return (kern_access(td, uap->path, UIO_USERSPACE, uap->flags));
2092 }
2093 
2094 #ifndef _SYS_SYSPROTO_H_
2095 struct faccessat_args {
2096 	int	dirfd;
2097 	char	*path;
2098 	int	mode;
2099 	int	flag;
2100 }
2101 #endif
2102 int
2103 faccessat(struct thread *td, struct faccessat_args *uap)
2104 {
2105 
2106 	if (uap->flag & ~AT_EACCESS)
2107 		return (EINVAL);
2108 	return (kern_accessat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
2109 	    uap->mode));
2110 }
2111 
2112 int
2113 kern_access(struct thread *td, char *path, enum uio_seg pathseg, int mode)
2114 {
2115 
2116 	return (kern_accessat(td, AT_FDCWD, path, pathseg, 0, mode));
2117 }
2118 
2119 int
2120 kern_accessat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2121     int flags, int mode)
2122 {
2123 	struct ucred *cred, *tmpcred;
2124 	struct vnode *vp;
2125 	struct nameidata nd;
2126 	int vfslocked;
2127 	int error;
2128 
2129 	/*
2130 	 * Create and modify a temporary credential instead of one that
2131 	 * is potentially shared.  This could also mess up socket
2132 	 * buffer accounting which can run in an interrupt context.
2133 	 */
2134 	if (!(flags & AT_EACCESS)) {
2135 		cred = td->td_ucred;
2136 		tmpcred = crdup(cred);
2137 		tmpcred->cr_uid = cred->cr_ruid;
2138 		tmpcred->cr_groups[0] = cred->cr_rgid;
2139 		td->td_ucred = tmpcred;
2140 	} else
2141 		cred = tmpcred = td->td_ucred;
2142 	NDINIT_AT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | MPSAFE |
2143 	    AUDITVNODE1, pathseg, path, fd, td);
2144 	if ((error = namei(&nd)) != 0)
2145 		goto out1;
2146 	vfslocked = NDHASGIANT(&nd);
2147 	vp = nd.ni_vp;
2148 
2149 	error = vn_access(vp, mode, tmpcred, td);
2150 	NDFREE(&nd, NDF_ONLY_PNBUF);
2151 	vput(vp);
2152 	VFS_UNLOCK_GIANT(vfslocked);
2153 out1:
2154 	if (!(flags & AT_EACCESS)) {
2155 		td->td_ucred = cred;
2156 		crfree(tmpcred);
2157 	}
2158 	return (error);
2159 }
2160 
2161 /*
2162  * Check access permissions using "effective" credentials.
2163  */
2164 #ifndef _SYS_SYSPROTO_H_
2165 struct eaccess_args {
2166 	char	*path;
2167 	int	flags;
2168 };
2169 #endif
2170 int
2171 eaccess(td, uap)
2172 	struct thread *td;
2173 	register struct eaccess_args /* {
2174 		char *path;
2175 		int flags;
2176 	} */ *uap;
2177 {
2178 
2179 	return (kern_eaccess(td, uap->path, UIO_USERSPACE, uap->flags));
2180 }
2181 
2182 int
2183 kern_eaccess(struct thread *td, char *path, enum uio_seg pathseg, int flags)
2184 {
2185 
2186 	return (kern_accessat(td, AT_FDCWD, path, pathseg, AT_EACCESS, flags));
2187 }
2188 
2189 #if defined(COMPAT_43)
2190 /*
2191  * Get file status; this version follows links.
2192  */
2193 #ifndef _SYS_SYSPROTO_H_
2194 struct ostat_args {
2195 	char	*path;
2196 	struct ostat *ub;
2197 };
2198 #endif
2199 int
2200 ostat(td, uap)
2201 	struct thread *td;
2202 	register struct ostat_args /* {
2203 		char *path;
2204 		struct ostat *ub;
2205 	} */ *uap;
2206 {
2207 	struct stat sb;
2208 	struct ostat osb;
2209 	int error;
2210 
2211 	error = kern_stat(td, uap->path, UIO_USERSPACE, &sb);
2212 	if (error)
2213 		return (error);
2214 	cvtstat(&sb, &osb);
2215 	error = copyout(&osb, uap->ub, sizeof (osb));
2216 	return (error);
2217 }
2218 
2219 /*
2220  * Get file status; this version does not follow links.
2221  */
2222 #ifndef _SYS_SYSPROTO_H_
2223 struct olstat_args {
2224 	char	*path;
2225 	struct ostat *ub;
2226 };
2227 #endif
2228 int
2229 olstat(td, uap)
2230 	struct thread *td;
2231 	register struct olstat_args /* {
2232 		char *path;
2233 		struct ostat *ub;
2234 	} */ *uap;
2235 {
2236 	struct stat sb;
2237 	struct ostat osb;
2238 	int error;
2239 
2240 	error = kern_lstat(td, uap->path, UIO_USERSPACE, &sb);
2241 	if (error)
2242 		return (error);
2243 	cvtstat(&sb, &osb);
2244 	error = copyout(&osb, uap->ub, sizeof (osb));
2245 	return (error);
2246 }
2247 
2248 /*
2249  * Convert from an old to a new stat structure.
2250  */
2251 void
2252 cvtstat(st, ost)
2253 	struct stat *st;
2254 	struct ostat *ost;
2255 {
2256 
2257 	ost->st_dev = st->st_dev;
2258 	ost->st_ino = st->st_ino;
2259 	ost->st_mode = st->st_mode;
2260 	ost->st_nlink = st->st_nlink;
2261 	ost->st_uid = st->st_uid;
2262 	ost->st_gid = st->st_gid;
2263 	ost->st_rdev = st->st_rdev;
2264 	if (st->st_size < (quad_t)1 << 32)
2265 		ost->st_size = st->st_size;
2266 	else
2267 		ost->st_size = -2;
2268 	ost->st_atime = st->st_atime;
2269 	ost->st_mtime = st->st_mtime;
2270 	ost->st_ctime = st->st_ctime;
2271 	ost->st_blksize = st->st_blksize;
2272 	ost->st_blocks = st->st_blocks;
2273 	ost->st_flags = st->st_flags;
2274 	ost->st_gen = st->st_gen;
2275 }
2276 #endif /* COMPAT_43 */
2277 
2278 /*
2279  * Get file status; this version follows links.
2280  */
2281 #ifndef _SYS_SYSPROTO_H_
2282 struct stat_args {
2283 	char	*path;
2284 	struct stat *ub;
2285 };
2286 #endif
2287 int
2288 stat(td, uap)
2289 	struct thread *td;
2290 	register struct stat_args /* {
2291 		char *path;
2292 		struct stat *ub;
2293 	} */ *uap;
2294 {
2295 	struct stat sb;
2296 	int error;
2297 
2298 	error = kern_stat(td, uap->path, UIO_USERSPACE, &sb);
2299 	if (error == 0)
2300 		error = copyout(&sb, uap->ub, sizeof (sb));
2301 	return (error);
2302 }
2303 
2304 #ifndef _SYS_SYSPROTO_H_
2305 struct fstatat_args {
2306 	int	fd;
2307 	char	*path;
2308 	struct stat	*buf;
2309 	int	flag;
2310 }
2311 #endif
2312 int
2313 fstatat(struct thread *td, struct fstatat_args *uap)
2314 {
2315 	struct stat sb;
2316 	int error;
2317 
2318 	error = kern_statat(td, uap->flag, uap->fd, uap->path,
2319 	    UIO_USERSPACE, &sb);
2320 	if (error == 0)
2321 		error = copyout(&sb, uap->buf, sizeof (sb));
2322 	return (error);
2323 }
2324 
2325 int
2326 kern_stat(struct thread *td, char *path, enum uio_seg pathseg, struct stat *sbp)
2327 {
2328 
2329 	return (kern_statat(td, 0, AT_FDCWD, path, pathseg, sbp));
2330 }
2331 
2332 int
2333 kern_statat(struct thread *td, int flag, int fd, char *path,
2334     enum uio_seg pathseg, struct stat *sbp)
2335 {
2336 
2337 	return (kern_statat_vnhook(td, flag, fd, path, pathseg, sbp, NULL));
2338 }
2339 
2340 int
2341 kern_statat_vnhook(struct thread *td, int flag, int fd, char *path,
2342     enum uio_seg pathseg, struct stat *sbp,
2343     void (*hook)(struct vnode *vp, struct stat *sbp))
2344 {
2345 	struct nameidata nd;
2346 	struct stat sb;
2347 	int error, vfslocked;
2348 
2349 	if (flag & ~AT_SYMLINK_NOFOLLOW)
2350 		return (EINVAL);
2351 
2352 	NDINIT_AT(&nd, LOOKUP, ((flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW :
2353 	    FOLLOW) | LOCKSHARED | LOCKLEAF | AUDITVNODE1 | MPSAFE, pathseg,
2354 	    path, fd, td);
2355 
2356 	if ((error = namei(&nd)) != 0)
2357 		return (error);
2358 	vfslocked = NDHASGIANT(&nd);
2359 	error = vn_stat(nd.ni_vp, &sb, td->td_ucred, NOCRED, td);
2360 	if (!error) {
2361 		SDT_PROBE(vfs, , stat, mode, path, sb.st_mode, 0, 0, 0);
2362 		if (S_ISREG(sb.st_mode))
2363 			SDT_PROBE(vfs, , stat, reg, path, pathseg, 0, 0, 0);
2364 		if (__predict_false(hook != NULL))
2365 			hook(nd.ni_vp, &sb);
2366 	}
2367 	NDFREE(&nd, NDF_ONLY_PNBUF);
2368 	vput(nd.ni_vp);
2369 	VFS_UNLOCK_GIANT(vfslocked);
2370 	if (error)
2371 		return (error);
2372 	*sbp = sb;
2373 #ifdef KTRACE
2374 	if (KTRPOINT(td, KTR_STRUCT))
2375 		ktrstat(&sb);
2376 #endif
2377 	return (0);
2378 }
2379 
2380 /*
2381  * Get file status; this version does not follow links.
2382  */
2383 #ifndef _SYS_SYSPROTO_H_
2384 struct lstat_args {
2385 	char	*path;
2386 	struct stat *ub;
2387 };
2388 #endif
2389 int
2390 lstat(td, uap)
2391 	struct thread *td;
2392 	register struct lstat_args /* {
2393 		char *path;
2394 		struct stat *ub;
2395 	} */ *uap;
2396 {
2397 	struct stat sb;
2398 	int error;
2399 
2400 	error = kern_lstat(td, uap->path, UIO_USERSPACE, &sb);
2401 	if (error == 0)
2402 		error = copyout(&sb, uap->ub, sizeof (sb));
2403 	return (error);
2404 }
2405 
2406 int
2407 kern_lstat(struct thread *td, char *path, enum uio_seg pathseg, struct stat *sbp)
2408 {
2409 
2410 	return (kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, path, pathseg,
2411 	    sbp));
2412 }
2413 
2414 /*
2415  * Implementation of the NetBSD [l]stat() functions.
2416  */
2417 void
2418 cvtnstat(sb, nsb)
2419 	struct stat *sb;
2420 	struct nstat *nsb;
2421 {
2422 	bzero(nsb, sizeof *nsb);
2423 	nsb->st_dev = sb->st_dev;
2424 	nsb->st_ino = sb->st_ino;
2425 	nsb->st_mode = sb->st_mode;
2426 	nsb->st_nlink = sb->st_nlink;
2427 	nsb->st_uid = sb->st_uid;
2428 	nsb->st_gid = sb->st_gid;
2429 	nsb->st_rdev = sb->st_rdev;
2430 	nsb->st_atimespec = sb->st_atimespec;
2431 	nsb->st_mtimespec = sb->st_mtimespec;
2432 	nsb->st_ctimespec = sb->st_ctimespec;
2433 	nsb->st_size = sb->st_size;
2434 	nsb->st_blocks = sb->st_blocks;
2435 	nsb->st_blksize = sb->st_blksize;
2436 	nsb->st_flags = sb->st_flags;
2437 	nsb->st_gen = sb->st_gen;
2438 	nsb->st_birthtimespec = sb->st_birthtimespec;
2439 }
2440 
2441 #ifndef _SYS_SYSPROTO_H_
2442 struct nstat_args {
2443 	char	*path;
2444 	struct nstat *ub;
2445 };
2446 #endif
2447 int
2448 nstat(td, uap)
2449 	struct thread *td;
2450 	register struct nstat_args /* {
2451 		char *path;
2452 		struct nstat *ub;
2453 	} */ *uap;
2454 {
2455 	struct stat sb;
2456 	struct nstat nsb;
2457 	int error;
2458 
2459 	error = kern_stat(td, uap->path, UIO_USERSPACE, &sb);
2460 	if (error)
2461 		return (error);
2462 	cvtnstat(&sb, &nsb);
2463 	error = copyout(&nsb, uap->ub, sizeof (nsb));
2464 	return (error);
2465 }
2466 
2467 /*
2468  * NetBSD lstat.  Get file status; this version does not follow links.
2469  */
2470 #ifndef _SYS_SYSPROTO_H_
2471 struct lstat_args {
2472 	char	*path;
2473 	struct stat *ub;
2474 };
2475 #endif
2476 int
2477 nlstat(td, uap)
2478 	struct thread *td;
2479 	register struct nlstat_args /* {
2480 		char *path;
2481 		struct nstat *ub;
2482 	} */ *uap;
2483 {
2484 	struct stat sb;
2485 	struct nstat nsb;
2486 	int error;
2487 
2488 	error = kern_lstat(td, uap->path, UIO_USERSPACE, &sb);
2489 	if (error)
2490 		return (error);
2491 	cvtnstat(&sb, &nsb);
2492 	error = copyout(&nsb, uap->ub, sizeof (nsb));
2493 	return (error);
2494 }
2495 
2496 /*
2497  * Get configurable pathname variables.
2498  */
2499 #ifndef _SYS_SYSPROTO_H_
2500 struct pathconf_args {
2501 	char	*path;
2502 	int	name;
2503 };
2504 #endif
2505 int
2506 pathconf(td, uap)
2507 	struct thread *td;
2508 	register struct pathconf_args /* {
2509 		char *path;
2510 		int name;
2511 	} */ *uap;
2512 {
2513 
2514 	return (kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name));
2515 }
2516 
2517 int
2518 kern_pathconf(struct thread *td, char *path, enum uio_seg pathseg, int name)
2519 {
2520 	struct nameidata nd;
2521 	int error, vfslocked;
2522 
2523 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | MPSAFE |
2524 	    AUDITVNODE1, pathseg, path, td);
2525 	if ((error = namei(&nd)) != 0)
2526 		return (error);
2527 	vfslocked = NDHASGIANT(&nd);
2528 	NDFREE(&nd, NDF_ONLY_PNBUF);
2529 
2530 	/* If asynchronous I/O is available, it works for all files. */
2531 	if (name == _PC_ASYNC_IO)
2532 		td->td_retval[0] = async_io_version;
2533 	else
2534 		error = VOP_PATHCONF(nd.ni_vp, name, td->td_retval);
2535 	vput(nd.ni_vp);
2536 	VFS_UNLOCK_GIANT(vfslocked);
2537 	return (error);
2538 }
2539 
2540 /*
2541  * Return target name of a symbolic link.
2542  */
2543 #ifndef _SYS_SYSPROTO_H_
2544 struct readlink_args {
2545 	char	*path;
2546 	char	*buf;
2547 	size_t	count;
2548 };
2549 #endif
2550 int
2551 readlink(td, uap)
2552 	struct thread *td;
2553 	register struct readlink_args /* {
2554 		char *path;
2555 		char *buf;
2556 		size_t count;
2557 	} */ *uap;
2558 {
2559 
2560 	return (kern_readlink(td, uap->path, UIO_USERSPACE, uap->buf,
2561 	    UIO_USERSPACE, uap->count));
2562 }
2563 #ifndef _SYS_SYSPROTO_H_
2564 struct readlinkat_args {
2565 	int	fd;
2566 	char	*path;
2567 	char	*buf;
2568 	size_t	bufsize;
2569 };
2570 #endif
2571 int
2572 readlinkat(struct thread *td, struct readlinkat_args *uap)
2573 {
2574 
2575 	return (kern_readlinkat(td, uap->fd, uap->path, UIO_USERSPACE,
2576 	    uap->buf, UIO_USERSPACE, uap->bufsize));
2577 }
2578 
2579 int
2580 kern_readlink(struct thread *td, char *path, enum uio_seg pathseg, char *buf,
2581     enum uio_seg bufseg, size_t count)
2582 {
2583 
2584 	return (kern_readlinkat(td, AT_FDCWD, path, pathseg, buf, bufseg,
2585 	    count));
2586 }
2587 
2588 int
2589 kern_readlinkat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2590     char *buf, enum uio_seg bufseg, size_t count)
2591 {
2592 	struct vnode *vp;
2593 	struct iovec aiov;
2594 	struct uio auio;
2595 	int error;
2596 	struct nameidata nd;
2597 	int vfslocked;
2598 
2599 	if (count > INT_MAX)
2600 		return (EINVAL);
2601 
2602 	NDINIT_AT(&nd, LOOKUP, NOFOLLOW | LOCKSHARED | LOCKLEAF | MPSAFE |
2603 	    AUDITVNODE1, pathseg, path, fd, td);
2604 
2605 	if ((error = namei(&nd)) != 0)
2606 		return (error);
2607 	NDFREE(&nd, NDF_ONLY_PNBUF);
2608 	vfslocked = NDHASGIANT(&nd);
2609 	vp = nd.ni_vp;
2610 #ifdef MAC
2611 	error = mac_vnode_check_readlink(td->td_ucred, vp);
2612 	if (error) {
2613 		vput(vp);
2614 		VFS_UNLOCK_GIANT(vfslocked);
2615 		return (error);
2616 	}
2617 #endif
2618 	if (vp->v_type != VLNK)
2619 		error = EINVAL;
2620 	else {
2621 		aiov.iov_base = buf;
2622 		aiov.iov_len = count;
2623 		auio.uio_iov = &aiov;
2624 		auio.uio_iovcnt = 1;
2625 		auio.uio_offset = 0;
2626 		auio.uio_rw = UIO_READ;
2627 		auio.uio_segflg = bufseg;
2628 		auio.uio_td = td;
2629 		auio.uio_resid = count;
2630 		error = VOP_READLINK(vp, &auio, td->td_ucred);
2631 	}
2632 	vput(vp);
2633 	VFS_UNLOCK_GIANT(vfslocked);
2634 	td->td_retval[0] = count - auio.uio_resid;
2635 	return (error);
2636 }
2637 
2638 /*
2639  * Common implementation code for chflags() and fchflags().
2640  */
2641 static int
2642 setfflags(td, vp, flags)
2643 	struct thread *td;
2644 	struct vnode *vp;
2645 	int flags;
2646 {
2647 	int error;
2648 	struct mount *mp;
2649 	struct vattr vattr;
2650 
2651 	/*
2652 	 * Prevent non-root users from setting flags on devices.  When
2653 	 * a device is reused, users can retain ownership of the device
2654 	 * if they are allowed to set flags and programs assume that
2655 	 * chown can't fail when done as root.
2656 	 */
2657 	if (vp->v_type == VCHR || vp->v_type == VBLK) {
2658 		error = priv_check(td, PRIV_VFS_CHFLAGS_DEV);
2659 		if (error)
2660 			return (error);
2661 	}
2662 
2663 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2664 		return (error);
2665 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2666 	VATTR_NULL(&vattr);
2667 	vattr.va_flags = flags;
2668 #ifdef MAC
2669 	error = mac_vnode_check_setflags(td->td_ucred, vp, vattr.va_flags);
2670 	if (error == 0)
2671 #endif
2672 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
2673 	VOP_UNLOCK(vp, 0);
2674 	vn_finished_write(mp);
2675 	return (error);
2676 }
2677 
2678 /*
2679  * Change flags of a file given a path name.
2680  */
2681 #ifndef _SYS_SYSPROTO_H_
2682 struct chflags_args {
2683 	char	*path;
2684 	int	flags;
2685 };
2686 #endif
2687 int
2688 chflags(td, uap)
2689 	struct thread *td;
2690 	register struct chflags_args /* {
2691 		char *path;
2692 		int flags;
2693 	} */ *uap;
2694 {
2695 	int error;
2696 	struct nameidata nd;
2697 	int vfslocked;
2698 
2699 	AUDIT_ARG(fflags, uap->flags);
2700 	NDINIT(&nd, LOOKUP, FOLLOW | MPSAFE | AUDITVNODE1, UIO_USERSPACE,
2701 	    uap->path, td);
2702 	if ((error = namei(&nd)) != 0)
2703 		return (error);
2704 	NDFREE(&nd, NDF_ONLY_PNBUF);
2705 	vfslocked = NDHASGIANT(&nd);
2706 	error = setfflags(td, nd.ni_vp, uap->flags);
2707 	vrele(nd.ni_vp);
2708 	VFS_UNLOCK_GIANT(vfslocked);
2709 	return (error);
2710 }
2711 
2712 /*
2713  * Same as chflags() but doesn't follow symlinks.
2714  */
2715 int
2716 lchflags(td, uap)
2717 	struct thread *td;
2718 	register struct lchflags_args /* {
2719 		char *path;
2720 		int flags;
2721 	} */ *uap;
2722 {
2723 	int error;
2724 	struct nameidata nd;
2725 	int vfslocked;
2726 
2727 	AUDIT_ARG(fflags, uap->flags);
2728 	NDINIT(&nd, LOOKUP, NOFOLLOW | MPSAFE | AUDITVNODE1, UIO_USERSPACE,
2729 	    uap->path, td);
2730 	if ((error = namei(&nd)) != 0)
2731 		return (error);
2732 	vfslocked = NDHASGIANT(&nd);
2733 	NDFREE(&nd, NDF_ONLY_PNBUF);
2734 	error = setfflags(td, nd.ni_vp, uap->flags);
2735 	vrele(nd.ni_vp);
2736 	VFS_UNLOCK_GIANT(vfslocked);
2737 	return (error);
2738 }
2739 
2740 /*
2741  * Change flags of a file given a file descriptor.
2742  */
2743 #ifndef _SYS_SYSPROTO_H_
2744 struct fchflags_args {
2745 	int	fd;
2746 	int	flags;
2747 };
2748 #endif
2749 int
2750 fchflags(td, uap)
2751 	struct thread *td;
2752 	register struct fchflags_args /* {
2753 		int fd;
2754 		int flags;
2755 	} */ *uap;
2756 {
2757 	struct file *fp;
2758 	int vfslocked;
2759 	int error;
2760 
2761 	AUDIT_ARG(fd, uap->fd);
2762 	AUDIT_ARG(fflags, uap->flags);
2763 	if ((error = getvnode(td->td_proc->p_fd, uap->fd, &fp)) != 0)
2764 		return (error);
2765 	vfslocked = VFS_LOCK_GIANT(fp->f_vnode->v_mount);
2766 #ifdef AUDIT
2767 	vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
2768 	AUDIT_ARG(vnode, fp->f_vnode, ARG_VNODE1);
2769 	VOP_UNLOCK(fp->f_vnode, 0);
2770 #endif
2771 	error = setfflags(td, fp->f_vnode, uap->flags);
2772 	VFS_UNLOCK_GIANT(vfslocked);
2773 	fdrop(fp, td);
2774 	return (error);
2775 }
2776 
2777 /*
2778  * Common implementation code for chmod(), lchmod() and fchmod().
2779  */
2780 static int
2781 setfmode(td, vp, mode)
2782 	struct thread *td;
2783 	struct vnode *vp;
2784 	int mode;
2785 {
2786 	int error;
2787 	struct mount *mp;
2788 	struct vattr vattr;
2789 
2790 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2791 		return (error);
2792 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2793 	VATTR_NULL(&vattr);
2794 	vattr.va_mode = mode & ALLPERMS;
2795 #ifdef MAC
2796 	error = mac_vnode_check_setmode(td->td_ucred, vp, vattr.va_mode);
2797 	if (error == 0)
2798 #endif
2799 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
2800 	VOP_UNLOCK(vp, 0);
2801 	vn_finished_write(mp);
2802 	return (error);
2803 }
2804 
2805 /*
2806  * Change mode of a file given path name.
2807  */
2808 #ifndef _SYS_SYSPROTO_H_
2809 struct chmod_args {
2810 	char	*path;
2811 	int	mode;
2812 };
2813 #endif
2814 int
2815 chmod(td, uap)
2816 	struct thread *td;
2817 	register struct chmod_args /* {
2818 		char *path;
2819 		int mode;
2820 	} */ *uap;
2821 {
2822 
2823 	return (kern_chmod(td, uap->path, UIO_USERSPACE, uap->mode));
2824 }
2825 
2826 #ifndef _SYS_SYSPROTO_H_
2827 struct fchmodat_args {
2828 	int	dirfd;
2829 	char	*path;
2830 	mode_t	mode;
2831 	int	flag;
2832 }
2833 #endif
2834 int
2835 fchmodat(struct thread *td, struct fchmodat_args *uap)
2836 {
2837 	int flag = uap->flag;
2838 	int fd = uap->fd;
2839 	char *path = uap->path;
2840 	mode_t mode = uap->mode;
2841 
2842 	if (flag & ~AT_SYMLINK_NOFOLLOW)
2843 		return (EINVAL);
2844 
2845 	return (kern_fchmodat(td, fd, path, UIO_USERSPACE, mode, flag));
2846 }
2847 
2848 int
2849 kern_chmod(struct thread *td, char *path, enum uio_seg pathseg, int mode)
2850 {
2851 
2852 	return (kern_fchmodat(td, AT_FDCWD, path, pathseg, mode, 0));
2853 }
2854 
2855 /*
2856  * Change mode of a file given path name (don't follow links.)
2857  */
2858 #ifndef _SYS_SYSPROTO_H_
2859 struct lchmod_args {
2860 	char	*path;
2861 	int	mode;
2862 };
2863 #endif
2864 int
2865 lchmod(td, uap)
2866 	struct thread *td;
2867 	register struct lchmod_args /* {
2868 		char *path;
2869 		int mode;
2870 	} */ *uap;
2871 {
2872 
2873 	return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2874 	    uap->mode, AT_SYMLINK_NOFOLLOW));
2875 }
2876 
2877 
2878 int
2879 kern_fchmodat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2880     mode_t mode, int flag)
2881 {
2882 	int error;
2883 	struct nameidata nd;
2884 	int vfslocked;
2885 	int follow;
2886 
2887 	AUDIT_ARG(mode, mode);
2888 	follow = (flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
2889 	NDINIT_AT(&nd, LOOKUP,  follow | MPSAFE | AUDITVNODE1, pathseg, path,
2890 	    fd, td);
2891 	if ((error = namei(&nd)) != 0)
2892 		return (error);
2893 	vfslocked = NDHASGIANT(&nd);
2894 	NDFREE(&nd, NDF_ONLY_PNBUF);
2895 	error = setfmode(td, nd.ni_vp, mode);
2896 	vrele(nd.ni_vp);
2897 	VFS_UNLOCK_GIANT(vfslocked);
2898 	return (error);
2899 }
2900 
2901 /*
2902  * Change mode of a file given a file descriptor.
2903  */
2904 #ifndef _SYS_SYSPROTO_H_
2905 struct fchmod_args {
2906 	int	fd;
2907 	int	mode;
2908 };
2909 #endif
2910 int
2911 fchmod(td, uap)
2912 	struct thread *td;
2913 	register struct fchmod_args /* {
2914 		int fd;
2915 		int mode;
2916 	} */ *uap;
2917 {
2918 	struct file *fp;
2919 	int vfslocked;
2920 	int error;
2921 
2922 	AUDIT_ARG(fd, uap->fd);
2923 	AUDIT_ARG(mode, uap->mode);
2924 	if ((error = getvnode(td->td_proc->p_fd, uap->fd, &fp)) != 0)
2925 		return (error);
2926 	vfslocked = VFS_LOCK_GIANT(fp->f_vnode->v_mount);
2927 #ifdef AUDIT
2928 	vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
2929 	AUDIT_ARG(vnode, fp->f_vnode, ARG_VNODE1);
2930 	VOP_UNLOCK(fp->f_vnode, 0);
2931 #endif
2932 	error = setfmode(td, fp->f_vnode, uap->mode);
2933 	VFS_UNLOCK_GIANT(vfslocked);
2934 	fdrop(fp, td);
2935 	return (error);
2936 }
2937 
2938 /*
2939  * Common implementation for chown(), lchown(), and fchown()
2940  */
2941 static int
2942 setfown(td, vp, uid, gid)
2943 	struct thread *td;
2944 	struct vnode *vp;
2945 	uid_t uid;
2946 	gid_t gid;
2947 {
2948 	int error;
2949 	struct mount *mp;
2950 	struct vattr vattr;
2951 
2952 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2953 		return (error);
2954 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2955 	VATTR_NULL(&vattr);
2956 	vattr.va_uid = uid;
2957 	vattr.va_gid = gid;
2958 #ifdef MAC
2959 	error = mac_vnode_check_setowner(td->td_ucred, vp, vattr.va_uid,
2960 	    vattr.va_gid);
2961 	if (error == 0)
2962 #endif
2963 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
2964 	VOP_UNLOCK(vp, 0);
2965 	vn_finished_write(mp);
2966 	return (error);
2967 }
2968 
2969 /*
2970  * Set ownership given a path name.
2971  */
2972 #ifndef _SYS_SYSPROTO_H_
2973 struct chown_args {
2974 	char	*path;
2975 	int	uid;
2976 	int	gid;
2977 };
2978 #endif
2979 int
2980 chown(td, uap)
2981 	struct thread *td;
2982 	register struct chown_args /* {
2983 		char *path;
2984 		int uid;
2985 		int gid;
2986 	} */ *uap;
2987 {
2988 
2989 	return (kern_chown(td, uap->path, UIO_USERSPACE, uap->uid, uap->gid));
2990 }
2991 
2992 #ifndef _SYS_SYSPROTO_H_
2993 struct fchownat_args {
2994 	int fd;
2995 	const char * path;
2996 	uid_t uid;
2997 	gid_t gid;
2998 	int flag;
2999 };
3000 #endif
3001 int
3002 fchownat(struct thread *td, struct fchownat_args *uap)
3003 {
3004 	int flag;
3005 
3006 	flag = uap->flag;
3007 	if (flag & ~AT_SYMLINK_NOFOLLOW)
3008 		return (EINVAL);
3009 
3010 	return (kern_fchownat(td, uap->fd, uap->path, UIO_USERSPACE, uap->uid,
3011 	    uap->gid, uap->flag));
3012 }
3013 
3014 int
3015 kern_chown(struct thread *td, char *path, enum uio_seg pathseg, int uid,
3016     int gid)
3017 {
3018 
3019 	return (kern_fchownat(td, AT_FDCWD, path, pathseg, uid, gid, 0));
3020 }
3021 
3022 int
3023 kern_fchownat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
3024     int uid, int gid, int flag)
3025 {
3026 	struct nameidata nd;
3027 	int error, vfslocked, follow;
3028 
3029 	AUDIT_ARG(owner, uid, gid);
3030 	follow = (flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
3031 	NDINIT_AT(&nd, LOOKUP, follow | MPSAFE | AUDITVNODE1, pathseg, path,
3032 	    fd, td);
3033 
3034 	if ((error = namei(&nd)) != 0)
3035 		return (error);
3036 	vfslocked = NDHASGIANT(&nd);
3037 	NDFREE(&nd, NDF_ONLY_PNBUF);
3038 	error = setfown(td, nd.ni_vp, uid, gid);
3039 	vrele(nd.ni_vp);
3040 	VFS_UNLOCK_GIANT(vfslocked);
3041 	return (error);
3042 }
3043 
3044 /*
3045  * Set ownership given a path name, do not cross symlinks.
3046  */
3047 #ifndef _SYS_SYSPROTO_H_
3048 struct lchown_args {
3049 	char	*path;
3050 	int	uid;
3051 	int	gid;
3052 };
3053 #endif
3054 int
3055 lchown(td, uap)
3056 	struct thread *td;
3057 	register struct lchown_args /* {
3058 		char *path;
3059 		int uid;
3060 		int gid;
3061 	} */ *uap;
3062 {
3063 
3064 	return (kern_lchown(td, uap->path, UIO_USERSPACE, uap->uid, uap->gid));
3065 }
3066 
3067 int
3068 kern_lchown(struct thread *td, char *path, enum uio_seg pathseg, int uid,
3069     int gid)
3070 {
3071 
3072 	return (kern_fchownat(td, AT_FDCWD, path, pathseg, uid, gid,
3073 	    AT_SYMLINK_NOFOLLOW));
3074 }
3075 
3076 /*
3077  * Set ownership given a file descriptor.
3078  */
3079 #ifndef _SYS_SYSPROTO_H_
3080 struct fchown_args {
3081 	int	fd;
3082 	int	uid;
3083 	int	gid;
3084 };
3085 #endif
3086 int
3087 fchown(td, uap)
3088 	struct thread *td;
3089 	register struct fchown_args /* {
3090 		int fd;
3091 		int uid;
3092 		int gid;
3093 	} */ *uap;
3094 {
3095 	struct file *fp;
3096 	int vfslocked;
3097 	int error;
3098 
3099 	AUDIT_ARG(fd, uap->fd);
3100 	AUDIT_ARG(owner, uap->uid, uap->gid);
3101 	if ((error = getvnode(td->td_proc->p_fd, uap->fd, &fp)) != 0)
3102 		return (error);
3103 	vfslocked = VFS_LOCK_GIANT(fp->f_vnode->v_mount);
3104 #ifdef AUDIT
3105 	vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
3106 	AUDIT_ARG(vnode, fp->f_vnode, ARG_VNODE1);
3107 	VOP_UNLOCK(fp->f_vnode, 0);
3108 #endif
3109 	error = setfown(td, fp->f_vnode, uap->uid, uap->gid);
3110 	VFS_UNLOCK_GIANT(vfslocked);
3111 	fdrop(fp, td);
3112 	return (error);
3113 }
3114 
3115 /*
3116  * Common implementation code for utimes(), lutimes(), and futimes().
3117  */
3118 static int
3119 getutimes(usrtvp, tvpseg, tsp)
3120 	const struct timeval *usrtvp;
3121 	enum uio_seg tvpseg;
3122 	struct timespec *tsp;
3123 {
3124 	struct timeval tv[2];
3125 	const struct timeval *tvp;
3126 	int error;
3127 
3128 	if (usrtvp == NULL) {
3129 		microtime(&tv[0]);
3130 		TIMEVAL_TO_TIMESPEC(&tv[0], &tsp[0]);
3131 		tsp[1] = tsp[0];
3132 	} else {
3133 		if (tvpseg == UIO_SYSSPACE) {
3134 			tvp = usrtvp;
3135 		} else {
3136 			if ((error = copyin(usrtvp, tv, sizeof(tv))) != 0)
3137 				return (error);
3138 			tvp = tv;
3139 		}
3140 
3141 		if (tvp[0].tv_usec < 0 || tvp[0].tv_usec >= 1000000 ||
3142 		    tvp[1].tv_usec < 0 || tvp[1].tv_usec >= 1000000)
3143 			return (EINVAL);
3144 		TIMEVAL_TO_TIMESPEC(&tvp[0], &tsp[0]);
3145 		TIMEVAL_TO_TIMESPEC(&tvp[1], &tsp[1]);
3146 	}
3147 	return (0);
3148 }
3149 
3150 /*
3151  * Common implementation code for utimes(), lutimes(), and futimes().
3152  */
3153 static int
3154 setutimes(td, vp, ts, numtimes, nullflag)
3155 	struct thread *td;
3156 	struct vnode *vp;
3157 	const struct timespec *ts;
3158 	int numtimes;
3159 	int nullflag;
3160 {
3161 	int error, setbirthtime;
3162 	struct mount *mp;
3163 	struct vattr vattr;
3164 
3165 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
3166 		return (error);
3167 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3168 	setbirthtime = 0;
3169 	if (numtimes < 3 && !VOP_GETATTR(vp, &vattr, td->td_ucred) &&
3170 	    timespeccmp(&ts[1], &vattr.va_birthtime, < ))
3171 		setbirthtime = 1;
3172 	VATTR_NULL(&vattr);
3173 	vattr.va_atime = ts[0];
3174 	vattr.va_mtime = ts[1];
3175 	if (setbirthtime)
3176 		vattr.va_birthtime = ts[1];
3177 	if (numtimes > 2)
3178 		vattr.va_birthtime = ts[2];
3179 	if (nullflag)
3180 		vattr.va_vaflags |= VA_UTIMES_NULL;
3181 #ifdef MAC
3182 	error = mac_vnode_check_setutimes(td->td_ucred, vp, vattr.va_atime,
3183 	    vattr.va_mtime);
3184 #endif
3185 	if (error == 0)
3186 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3187 	VOP_UNLOCK(vp, 0);
3188 	vn_finished_write(mp);
3189 	return (error);
3190 }
3191 
3192 /*
3193  * Set the access and modification times of a file.
3194  */
3195 #ifndef _SYS_SYSPROTO_H_
3196 struct utimes_args {
3197 	char	*path;
3198 	struct	timeval *tptr;
3199 };
3200 #endif
3201 int
3202 utimes(td, uap)
3203 	struct thread *td;
3204 	register struct utimes_args /* {
3205 		char *path;
3206 		struct timeval *tptr;
3207 	} */ *uap;
3208 {
3209 
3210 	return (kern_utimes(td, uap->path, UIO_USERSPACE, uap->tptr,
3211 	    UIO_USERSPACE));
3212 }
3213 
3214 #ifndef _SYS_SYSPROTO_H_
3215 struct futimesat_args {
3216 	int fd;
3217 	const char * path;
3218 	const struct timeval * times;
3219 };
3220 #endif
3221 int
3222 futimesat(struct thread *td, struct futimesat_args *uap)
3223 {
3224 
3225 	return (kern_utimesat(td, uap->fd, uap->path, UIO_USERSPACE,
3226 	    uap->times, UIO_USERSPACE));
3227 }
3228 
3229 int
3230 kern_utimes(struct thread *td, char *path, enum uio_seg pathseg,
3231     struct timeval *tptr, enum uio_seg tptrseg)
3232 {
3233 
3234 	return (kern_utimesat(td, AT_FDCWD, path, pathseg, tptr, tptrseg));
3235 }
3236 
3237 int
3238 kern_utimesat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
3239     struct timeval *tptr, enum uio_seg tptrseg)
3240 {
3241 	struct nameidata nd;
3242 	struct timespec ts[2];
3243 	int error, vfslocked;
3244 
3245 	if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3246 		return (error);
3247 	NDINIT_AT(&nd, LOOKUP, FOLLOW | MPSAFE | AUDITVNODE1, pathseg, path,
3248 	    fd, td);
3249 
3250 	if ((error = namei(&nd)) != 0)
3251 		return (error);
3252 	vfslocked = NDHASGIANT(&nd);
3253 	NDFREE(&nd, NDF_ONLY_PNBUF);
3254 	error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3255 	vrele(nd.ni_vp);
3256 	VFS_UNLOCK_GIANT(vfslocked);
3257 	return (error);
3258 }
3259 
3260 /*
3261  * Set the access and modification times of a file.
3262  */
3263 #ifndef _SYS_SYSPROTO_H_
3264 struct lutimes_args {
3265 	char	*path;
3266 	struct	timeval *tptr;
3267 };
3268 #endif
3269 int
3270 lutimes(td, uap)
3271 	struct thread *td;
3272 	register struct lutimes_args /* {
3273 		char *path;
3274 		struct timeval *tptr;
3275 	} */ *uap;
3276 {
3277 
3278 	return (kern_lutimes(td, uap->path, UIO_USERSPACE, uap->tptr,
3279 	    UIO_USERSPACE));
3280 }
3281 
3282 int
3283 kern_lutimes(struct thread *td, char *path, enum uio_seg pathseg,
3284     struct timeval *tptr, enum uio_seg tptrseg)
3285 {
3286 	struct timespec ts[2];
3287 	int error;
3288 	struct nameidata nd;
3289 	int vfslocked;
3290 
3291 	if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3292 		return (error);
3293 	NDINIT(&nd, LOOKUP, NOFOLLOW | MPSAFE | AUDITVNODE1, pathseg, path, td);
3294 	if ((error = namei(&nd)) != 0)
3295 		return (error);
3296 	vfslocked = NDHASGIANT(&nd);
3297 	NDFREE(&nd, NDF_ONLY_PNBUF);
3298 	error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3299 	vrele(nd.ni_vp);
3300 	VFS_UNLOCK_GIANT(vfslocked);
3301 	return (error);
3302 }
3303 
3304 /*
3305  * Set the access and modification times of a file.
3306  */
3307 #ifndef _SYS_SYSPROTO_H_
3308 struct futimes_args {
3309 	int	fd;
3310 	struct	timeval *tptr;
3311 };
3312 #endif
3313 int
3314 futimes(td, uap)
3315 	struct thread *td;
3316 	register struct futimes_args /* {
3317 		int  fd;
3318 		struct timeval *tptr;
3319 	} */ *uap;
3320 {
3321 
3322 	return (kern_futimes(td, uap->fd, uap->tptr, UIO_USERSPACE));
3323 }
3324 
3325 int
3326 kern_futimes(struct thread *td, int fd, struct timeval *tptr,
3327     enum uio_seg tptrseg)
3328 {
3329 	struct timespec ts[2];
3330 	struct file *fp;
3331 	int vfslocked;
3332 	int error;
3333 
3334 	AUDIT_ARG(fd, fd);
3335 	if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3336 		return (error);
3337 	if ((error = getvnode(td->td_proc->p_fd, fd, &fp)) != 0)
3338 		return (error);
3339 	vfslocked = VFS_LOCK_GIANT(fp->f_vnode->v_mount);
3340 #ifdef AUDIT
3341 	vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
3342 	AUDIT_ARG(vnode, fp->f_vnode, ARG_VNODE1);
3343 	VOP_UNLOCK(fp->f_vnode, 0);
3344 #endif
3345 	error = setutimes(td, fp->f_vnode, ts, 2, tptr == NULL);
3346 	VFS_UNLOCK_GIANT(vfslocked);
3347 	fdrop(fp, td);
3348 	return (error);
3349 }
3350 
3351 /*
3352  * Truncate a file given its path name.
3353  */
3354 #ifndef _SYS_SYSPROTO_H_
3355 struct truncate_args {
3356 	char	*path;
3357 	int	pad;
3358 	off_t	length;
3359 };
3360 #endif
3361 int
3362 truncate(td, uap)
3363 	struct thread *td;
3364 	register struct truncate_args /* {
3365 		char *path;
3366 		int pad;
3367 		off_t length;
3368 	} */ *uap;
3369 {
3370 
3371 	return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3372 }
3373 
3374 int
3375 kern_truncate(struct thread *td, char *path, enum uio_seg pathseg, off_t length)
3376 {
3377 	struct mount *mp;
3378 	struct vnode *vp;
3379 	struct vattr vattr;
3380 	int error;
3381 	struct nameidata nd;
3382 	int vfslocked;
3383 
3384 	if (length < 0)
3385 		return(EINVAL);
3386 	NDINIT(&nd, LOOKUP, FOLLOW | MPSAFE | AUDITVNODE1, pathseg, path, td);
3387 	if ((error = namei(&nd)) != 0)
3388 		return (error);
3389 	vfslocked = NDHASGIANT(&nd);
3390 	vp = nd.ni_vp;
3391 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) {
3392 		vrele(vp);
3393 		VFS_UNLOCK_GIANT(vfslocked);
3394 		return (error);
3395 	}
3396 	NDFREE(&nd, NDF_ONLY_PNBUF);
3397 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3398 	if (vp->v_type == VDIR)
3399 		error = EISDIR;
3400 #ifdef MAC
3401 	else if ((error = mac_vnode_check_write(td->td_ucred, NOCRED, vp))) {
3402 	}
3403 #endif
3404 	else if ((error = vn_writechk(vp)) == 0 &&
3405 	    (error = VOP_ACCESS(vp, VWRITE, td->td_ucred, td)) == 0) {
3406 		VATTR_NULL(&vattr);
3407 		vattr.va_size = length;
3408 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3409 	}
3410 	vput(vp);
3411 	vn_finished_write(mp);
3412 	VFS_UNLOCK_GIANT(vfslocked);
3413 	return (error);
3414 }
3415 
3416 #if defined(COMPAT_43)
3417 /*
3418  * Truncate a file given its path name.
3419  */
3420 #ifndef _SYS_SYSPROTO_H_
3421 struct otruncate_args {
3422 	char	*path;
3423 	long	length;
3424 };
3425 #endif
3426 int
3427 otruncate(td, uap)
3428 	struct thread *td;
3429 	register struct otruncate_args /* {
3430 		char *path;
3431 		long length;
3432 	} */ *uap;
3433 {
3434 	struct truncate_args /* {
3435 		char *path;
3436 		int pad;
3437 		off_t length;
3438 	} */ nuap;
3439 
3440 	nuap.path = uap->path;
3441 	nuap.length = uap->length;
3442 	return (truncate(td, &nuap));
3443 }
3444 #endif /* COMPAT_43 */
3445 
3446 /* Versions with the pad argument */
3447 int
3448 freebsd6_truncate(struct thread *td, struct freebsd6_truncate_args *uap)
3449 {
3450 	struct truncate_args ouap;
3451 
3452 	ouap.path = uap->path;
3453 	ouap.length = uap->length;
3454 	return (truncate(td, &ouap));
3455 }
3456 
3457 int
3458 freebsd6_ftruncate(struct thread *td, struct freebsd6_ftruncate_args *uap)
3459 {
3460 	struct ftruncate_args ouap;
3461 
3462 	ouap.fd = uap->fd;
3463 	ouap.length = uap->length;
3464 	return (ftruncate(td, &ouap));
3465 }
3466 
3467 /*
3468  * Sync an open file.
3469  */
3470 #ifndef _SYS_SYSPROTO_H_
3471 struct fsync_args {
3472 	int	fd;
3473 };
3474 #endif
3475 int
3476 fsync(td, uap)
3477 	struct thread *td;
3478 	struct fsync_args /* {
3479 		int fd;
3480 	} */ *uap;
3481 {
3482 	struct vnode *vp;
3483 	struct mount *mp;
3484 	struct file *fp;
3485 	int vfslocked;
3486 	int error;
3487 
3488 	AUDIT_ARG(fd, uap->fd);
3489 	if ((error = getvnode(td->td_proc->p_fd, uap->fd, &fp)) != 0)
3490 		return (error);
3491 	vp = fp->f_vnode;
3492 	vfslocked = VFS_LOCK_GIANT(vp->v_mount);
3493 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
3494 		goto drop;
3495 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3496 	AUDIT_ARG(vnode, vp, ARG_VNODE1);
3497 	if (vp->v_object != NULL) {
3498 		VM_OBJECT_LOCK(vp->v_object);
3499 		vm_object_page_clean(vp->v_object, 0, 0, 0);
3500 		VM_OBJECT_UNLOCK(vp->v_object);
3501 	}
3502 	error = VOP_FSYNC(vp, MNT_WAIT, td);
3503 
3504 	VOP_UNLOCK(vp, 0);
3505 	vn_finished_write(mp);
3506 drop:
3507 	VFS_UNLOCK_GIANT(vfslocked);
3508 	fdrop(fp, td);
3509 	return (error);
3510 }
3511 
3512 /*
3513  * Rename files.  Source and destination must either both be directories, or
3514  * both not be directories.  If target is a directory, it must be empty.
3515  */
3516 #ifndef _SYS_SYSPROTO_H_
3517 struct rename_args {
3518 	char	*from;
3519 	char	*to;
3520 };
3521 #endif
3522 int
3523 rename(td, uap)
3524 	struct thread *td;
3525 	register struct rename_args /* {
3526 		char *from;
3527 		char *to;
3528 	} */ *uap;
3529 {
3530 
3531 	return (kern_rename(td, uap->from, uap->to, UIO_USERSPACE));
3532 }
3533 
3534 #ifndef _SYS_SYSPROTO_H_
3535 struct renameat_args {
3536 	int	oldfd;
3537 	char	*old;
3538 	int	newfd;
3539 	char	*new;
3540 };
3541 #endif
3542 int
3543 renameat(struct thread *td, struct renameat_args *uap)
3544 {
3545 
3546 	return (kern_renameat(td, uap->oldfd, uap->old, uap->newfd, uap->new,
3547 	    UIO_USERSPACE));
3548 }
3549 
3550 int
3551 kern_rename(struct thread *td, char *from, char *to, enum uio_seg pathseg)
3552 {
3553 
3554 	return (kern_renameat(td, AT_FDCWD, from, AT_FDCWD, to, pathseg));
3555 }
3556 
3557 int
3558 kern_renameat(struct thread *td, int oldfd, char *old, int newfd, char *new,
3559     enum uio_seg pathseg)
3560 {
3561 	struct mount *mp = NULL;
3562 	struct vnode *tvp, *fvp, *tdvp;
3563 	struct nameidata fromnd, tond;
3564 	int tvfslocked;
3565 	int fvfslocked;
3566 	int error;
3567 
3568 	bwillwrite();
3569 #ifdef MAC
3570 	NDINIT_AT(&fromnd, DELETE, LOCKPARENT | LOCKLEAF | SAVESTART | MPSAFE |
3571 	    AUDITVNODE1, pathseg, old, oldfd, td);
3572 #else
3573 	NDINIT_AT(&fromnd, DELETE, WANTPARENT | SAVESTART | MPSAFE |
3574 	    AUDITVNODE1, pathseg, old, oldfd, td);
3575 #endif
3576 
3577 	if ((error = namei(&fromnd)) != 0)
3578 		return (error);
3579 	fvfslocked = NDHASGIANT(&fromnd);
3580 	tvfslocked = 0;
3581 #ifdef MAC
3582 	error = mac_vnode_check_rename_from(td->td_ucred, fromnd.ni_dvp,
3583 	    fromnd.ni_vp, &fromnd.ni_cnd);
3584 	VOP_UNLOCK(fromnd.ni_dvp, 0);
3585 	if (fromnd.ni_dvp != fromnd.ni_vp)
3586 		VOP_UNLOCK(fromnd.ni_vp, 0);
3587 #endif
3588 	fvp = fromnd.ni_vp;
3589 	if (error == 0)
3590 		error = vn_start_write(fvp, &mp, V_WAIT | PCATCH);
3591 	if (error != 0) {
3592 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3593 		vrele(fromnd.ni_dvp);
3594 		vrele(fvp);
3595 		goto out1;
3596 	}
3597 	NDINIT_AT(&tond, RENAME, LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART |
3598 	    MPSAFE | AUDITVNODE2, pathseg, new, newfd, td);
3599 	if (fromnd.ni_vp->v_type == VDIR)
3600 		tond.ni_cnd.cn_flags |= WILLBEDIR;
3601 	if ((error = namei(&tond)) != 0) {
3602 		/* Translate error code for rename("dir1", "dir2/."). */
3603 		if (error == EISDIR && fvp->v_type == VDIR)
3604 			error = EINVAL;
3605 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3606 		vrele(fromnd.ni_dvp);
3607 		vrele(fvp);
3608 		vn_finished_write(mp);
3609 		goto out1;
3610 	}
3611 	tvfslocked = NDHASGIANT(&tond);
3612 	tdvp = tond.ni_dvp;
3613 	tvp = tond.ni_vp;
3614 	if (tvp != NULL) {
3615 		if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
3616 			error = ENOTDIR;
3617 			goto out;
3618 		} else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
3619 			error = EISDIR;
3620 			goto out;
3621 		}
3622 	}
3623 	if (fvp == tdvp) {
3624 		error = EINVAL;
3625 		goto out;
3626 	}
3627 	/*
3628 	 * If the source is the same as the destination (that is, if they
3629 	 * are links to the same vnode), then there is nothing to do.
3630 	 */
3631 	if (fvp == tvp)
3632 		error = -1;
3633 #ifdef MAC
3634 	else
3635 		error = mac_vnode_check_rename_to(td->td_ucred, tdvp,
3636 		    tond.ni_vp, fromnd.ni_dvp == tdvp, &tond.ni_cnd);
3637 #endif
3638 out:
3639 	if (!error) {
3640 		error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
3641 				   tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
3642 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3643 		NDFREE(&tond, NDF_ONLY_PNBUF);
3644 	} else {
3645 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3646 		NDFREE(&tond, NDF_ONLY_PNBUF);
3647 		if (tvp)
3648 			vput(tvp);
3649 		if (tdvp == tvp)
3650 			vrele(tdvp);
3651 		else
3652 			vput(tdvp);
3653 		vrele(fromnd.ni_dvp);
3654 		vrele(fvp);
3655 	}
3656 	vrele(tond.ni_startdir);
3657 	vn_finished_write(mp);
3658 out1:
3659 	if (fromnd.ni_startdir)
3660 		vrele(fromnd.ni_startdir);
3661 	VFS_UNLOCK_GIANT(fvfslocked);
3662 	VFS_UNLOCK_GIANT(tvfslocked);
3663 	if (error == -1)
3664 		return (0);
3665 	return (error);
3666 }
3667 
3668 /*
3669  * Make a directory file.
3670  */
3671 #ifndef _SYS_SYSPROTO_H_
3672 struct mkdir_args {
3673 	char	*path;
3674 	int	mode;
3675 };
3676 #endif
3677 int
3678 mkdir(td, uap)
3679 	struct thread *td;
3680 	register struct mkdir_args /* {
3681 		char *path;
3682 		int mode;
3683 	} */ *uap;
3684 {
3685 
3686 	return (kern_mkdir(td, uap->path, UIO_USERSPACE, uap->mode));
3687 }
3688 
3689 #ifndef _SYS_SYSPROTO_H_
3690 struct mkdirat_args {
3691 	int	fd;
3692 	char	*path;
3693 	mode_t	mode;
3694 };
3695 #endif
3696 int
3697 mkdirat(struct thread *td, struct mkdirat_args *uap)
3698 {
3699 
3700 	return (kern_mkdirat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode));
3701 }
3702 
3703 int
3704 kern_mkdir(struct thread *td, char *path, enum uio_seg segflg, int mode)
3705 {
3706 
3707 	return (kern_mkdirat(td, AT_FDCWD, path, segflg, mode));
3708 }
3709 
3710 int
3711 kern_mkdirat(struct thread *td, int fd, char *path, enum uio_seg segflg,
3712     int mode)
3713 {
3714 	struct mount *mp;
3715 	struct vnode *vp;
3716 	struct vattr vattr;
3717 	int error;
3718 	struct nameidata nd;
3719 	int vfslocked;
3720 
3721 	AUDIT_ARG(mode, mode);
3722 restart:
3723 	bwillwrite();
3724 	NDINIT_AT(&nd, CREATE, LOCKPARENT | SAVENAME | MPSAFE | AUDITVNODE1,
3725 	    segflg, path, fd, td);
3726 	nd.ni_cnd.cn_flags |= WILLBEDIR;
3727 	if ((error = namei(&nd)) != 0)
3728 		return (error);
3729 	vfslocked = NDHASGIANT(&nd);
3730 	vp = nd.ni_vp;
3731 	if (vp != NULL) {
3732 		NDFREE(&nd, NDF_ONLY_PNBUF);
3733 		/*
3734 		 * XXX namei called with LOCKPARENT but not LOCKLEAF has
3735 		 * the strange behaviour of leaving the vnode unlocked
3736 		 * if the target is the same vnode as the parent.
3737 		 */
3738 		if (vp == nd.ni_dvp)
3739 			vrele(nd.ni_dvp);
3740 		else
3741 			vput(nd.ni_dvp);
3742 		vrele(vp);
3743 		VFS_UNLOCK_GIANT(vfslocked);
3744 		return (EEXIST);
3745 	}
3746 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3747 		NDFREE(&nd, NDF_ONLY_PNBUF);
3748 		vput(nd.ni_dvp);
3749 		VFS_UNLOCK_GIANT(vfslocked);
3750 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3751 			return (error);
3752 		goto restart;
3753 	}
3754 	VATTR_NULL(&vattr);
3755 	vattr.va_type = VDIR;
3756 	FILEDESC_SLOCK(td->td_proc->p_fd);
3757 	vattr.va_mode = (mode & ACCESSPERMS) &~ td->td_proc->p_fd->fd_cmask;
3758 	FILEDESC_SUNLOCK(td->td_proc->p_fd);
3759 #ifdef MAC
3760 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
3761 	    &vattr);
3762 	if (error)
3763 		goto out;
3764 #endif
3765 	error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
3766 #ifdef MAC
3767 out:
3768 #endif
3769 	NDFREE(&nd, NDF_ONLY_PNBUF);
3770 	vput(nd.ni_dvp);
3771 	if (!error)
3772 		vput(nd.ni_vp);
3773 	vn_finished_write(mp);
3774 	VFS_UNLOCK_GIANT(vfslocked);
3775 	return (error);
3776 }
3777 
3778 /*
3779  * Remove a directory file.
3780  */
3781 #ifndef _SYS_SYSPROTO_H_
3782 struct rmdir_args {
3783 	char	*path;
3784 };
3785 #endif
3786 int
3787 rmdir(td, uap)
3788 	struct thread *td;
3789 	struct rmdir_args /* {
3790 		char *path;
3791 	} */ *uap;
3792 {
3793 
3794 	return (kern_rmdir(td, uap->path, UIO_USERSPACE));
3795 }
3796 
3797 int
3798 kern_rmdir(struct thread *td, char *path, enum uio_seg pathseg)
3799 {
3800 
3801 	return (kern_rmdirat(td, AT_FDCWD, path, pathseg));
3802 }
3803 
3804 int
3805 kern_rmdirat(struct thread *td, int fd, char *path, enum uio_seg pathseg)
3806 {
3807 	struct mount *mp;
3808 	struct vnode *vp;
3809 	int error;
3810 	struct nameidata nd;
3811 	int vfslocked;
3812 
3813 restart:
3814 	bwillwrite();
3815 	NDINIT_AT(&nd, DELETE, LOCKPARENT | LOCKLEAF | MPSAFE | AUDITVNODE1,
3816 	    pathseg, path, fd, td);
3817 	if ((error = namei(&nd)) != 0)
3818 		return (error);
3819 	vfslocked = NDHASGIANT(&nd);
3820 	vp = nd.ni_vp;
3821 	if (vp->v_type != VDIR) {
3822 		error = ENOTDIR;
3823 		goto out;
3824 	}
3825 	/*
3826 	 * No rmdir "." please.
3827 	 */
3828 	if (nd.ni_dvp == vp) {
3829 		error = EINVAL;
3830 		goto out;
3831 	}
3832 	/*
3833 	 * The root of a mounted filesystem cannot be deleted.
3834 	 */
3835 	if (vp->v_vflag & VV_ROOT) {
3836 		error = EBUSY;
3837 		goto out;
3838 	}
3839 #ifdef MAC
3840 	error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
3841 	    &nd.ni_cnd);
3842 	if (error)
3843 		goto out;
3844 #endif
3845 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3846 		NDFREE(&nd, NDF_ONLY_PNBUF);
3847 		vput(vp);
3848 		if (nd.ni_dvp == vp)
3849 			vrele(nd.ni_dvp);
3850 		else
3851 			vput(nd.ni_dvp);
3852 		VFS_UNLOCK_GIANT(vfslocked);
3853 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3854 			return (error);
3855 		goto restart;
3856 	}
3857 	error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
3858 	vn_finished_write(mp);
3859 out:
3860 	NDFREE(&nd, NDF_ONLY_PNBUF);
3861 	vput(vp);
3862 	if (nd.ni_dvp == vp)
3863 		vrele(nd.ni_dvp);
3864 	else
3865 		vput(nd.ni_dvp);
3866 	VFS_UNLOCK_GIANT(vfslocked);
3867 	return (error);
3868 }
3869 
3870 #ifdef COMPAT_43
3871 /*
3872  * Read a block of directory entries in a filesystem independent format.
3873  */
3874 #ifndef _SYS_SYSPROTO_H_
3875 struct ogetdirentries_args {
3876 	int	fd;
3877 	char	*buf;
3878 	u_int	count;
3879 	long	*basep;
3880 };
3881 #endif
3882 int
3883 ogetdirentries(td, uap)
3884 	struct thread *td;
3885 	register struct ogetdirentries_args /* {
3886 		int fd;
3887 		char *buf;
3888 		u_int count;
3889 		long *basep;
3890 	} */ *uap;
3891 {
3892 	struct vnode *vp;
3893 	struct file *fp;
3894 	struct uio auio, kuio;
3895 	struct iovec aiov, kiov;
3896 	struct dirent *dp, *edp;
3897 	caddr_t dirbuf;
3898 	int error, eofflag, readcnt, vfslocked;
3899 	long loff;
3900 
3901 	/* XXX arbitrary sanity limit on `count'. */
3902 	if (uap->count > 64 * 1024)
3903 		return (EINVAL);
3904 	if ((error = getvnode(td->td_proc->p_fd, uap->fd, &fp)) != 0)
3905 		return (error);
3906 	if ((fp->f_flag & FREAD) == 0) {
3907 		fdrop(fp, td);
3908 		return (EBADF);
3909 	}
3910 	vp = fp->f_vnode;
3911 unionread:
3912 	vfslocked = VFS_LOCK_GIANT(vp->v_mount);
3913 	if (vp->v_type != VDIR) {
3914 		VFS_UNLOCK_GIANT(vfslocked);
3915 		fdrop(fp, td);
3916 		return (EINVAL);
3917 	}
3918 	aiov.iov_base = uap->buf;
3919 	aiov.iov_len = uap->count;
3920 	auio.uio_iov = &aiov;
3921 	auio.uio_iovcnt = 1;
3922 	auio.uio_rw = UIO_READ;
3923 	auio.uio_segflg = UIO_USERSPACE;
3924 	auio.uio_td = td;
3925 	auio.uio_resid = uap->count;
3926 	vn_lock(vp, LK_SHARED | LK_RETRY);
3927 	loff = auio.uio_offset = fp->f_offset;
3928 #ifdef MAC
3929 	error = mac_vnode_check_readdir(td->td_ucred, vp);
3930 	if (error) {
3931 		VOP_UNLOCK(vp, 0);
3932 		VFS_UNLOCK_GIANT(vfslocked);
3933 		fdrop(fp, td);
3934 		return (error);
3935 	}
3936 #endif
3937 #	if (BYTE_ORDER != LITTLE_ENDIAN)
3938 		if (vp->v_mount->mnt_maxsymlinklen <= 0) {
3939 			error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag,
3940 			    NULL, NULL);
3941 			fp->f_offset = auio.uio_offset;
3942 		} else
3943 #	endif
3944 	{
3945 		kuio = auio;
3946 		kuio.uio_iov = &kiov;
3947 		kuio.uio_segflg = UIO_SYSSPACE;
3948 		kiov.iov_len = uap->count;
3949 		dirbuf = malloc(uap->count, M_TEMP, M_WAITOK);
3950 		kiov.iov_base = dirbuf;
3951 		error = VOP_READDIR(vp, &kuio, fp->f_cred, &eofflag,
3952 			    NULL, NULL);
3953 		fp->f_offset = kuio.uio_offset;
3954 		if (error == 0) {
3955 			readcnt = uap->count - kuio.uio_resid;
3956 			edp = (struct dirent *)&dirbuf[readcnt];
3957 			for (dp = (struct dirent *)dirbuf; dp < edp; ) {
3958 #				if (BYTE_ORDER == LITTLE_ENDIAN)
3959 					/*
3960 					 * The expected low byte of
3961 					 * dp->d_namlen is our dp->d_type.
3962 					 * The high MBZ byte of dp->d_namlen
3963 					 * is our dp->d_namlen.
3964 					 */
3965 					dp->d_type = dp->d_namlen;
3966 					dp->d_namlen = 0;
3967 #				else
3968 					/*
3969 					 * The dp->d_type is the high byte
3970 					 * of the expected dp->d_namlen,
3971 					 * so must be zero'ed.
3972 					 */
3973 					dp->d_type = 0;
3974 #				endif
3975 				if (dp->d_reclen > 0) {
3976 					dp = (struct dirent *)
3977 					    ((char *)dp + dp->d_reclen);
3978 				} else {
3979 					error = EIO;
3980 					break;
3981 				}
3982 			}
3983 			if (dp >= edp)
3984 				error = uiomove(dirbuf, readcnt, &auio);
3985 		}
3986 		free(dirbuf, M_TEMP);
3987 	}
3988 	if (error) {
3989 		VOP_UNLOCK(vp, 0);
3990 		VFS_UNLOCK_GIANT(vfslocked);
3991 		fdrop(fp, td);
3992 		return (error);
3993 	}
3994 	if (uap->count == auio.uio_resid &&
3995 	    (vp->v_vflag & VV_ROOT) &&
3996 	    (vp->v_mount->mnt_flag & MNT_UNION)) {
3997 		struct vnode *tvp = vp;
3998 		vp = vp->v_mount->mnt_vnodecovered;
3999 		VREF(vp);
4000 		fp->f_vnode = vp;
4001 		fp->f_data = vp;
4002 		fp->f_offset = 0;
4003 		vput(tvp);
4004 		VFS_UNLOCK_GIANT(vfslocked);
4005 		goto unionread;
4006 	}
4007 	VOP_UNLOCK(vp, 0);
4008 	VFS_UNLOCK_GIANT(vfslocked);
4009 	error = copyout(&loff, uap->basep, sizeof(long));
4010 	fdrop(fp, td);
4011 	td->td_retval[0] = uap->count - auio.uio_resid;
4012 	return (error);
4013 }
4014 #endif /* COMPAT_43 */
4015 
4016 /*
4017  * Read a block of directory entries in a filesystem independent format.
4018  */
4019 #ifndef _SYS_SYSPROTO_H_
4020 struct getdirentries_args {
4021 	int	fd;
4022 	char	*buf;
4023 	u_int	count;
4024 	long	*basep;
4025 };
4026 #endif
4027 int
4028 getdirentries(td, uap)
4029 	struct thread *td;
4030 	register struct getdirentries_args /* {
4031 		int fd;
4032 		char *buf;
4033 		u_int count;
4034 		long *basep;
4035 	} */ *uap;
4036 {
4037 	long base;
4038 	int error;
4039 
4040 	error = kern_getdirentries(td, uap->fd, uap->buf, uap->count, &base);
4041 	if (error)
4042 		return (error);
4043 	if (uap->basep != NULL)
4044 		error = copyout(&base, uap->basep, sizeof(long));
4045 	return (error);
4046 }
4047 
4048 int
4049 kern_getdirentries(struct thread *td, int fd, char *buf, u_int count,
4050     long *basep)
4051 {
4052 	struct vnode *vp;
4053 	struct file *fp;
4054 	struct uio auio;
4055 	struct iovec aiov;
4056 	int vfslocked;
4057 	long loff;
4058 	int error, eofflag;
4059 
4060 	AUDIT_ARG(fd, fd);
4061 	if (count > INT_MAX)
4062 		return (EINVAL);
4063 	if ((error = getvnode(td->td_proc->p_fd, fd, &fp)) != 0)
4064 		return (error);
4065 	if ((fp->f_flag & FREAD) == 0) {
4066 		fdrop(fp, td);
4067 		return (EBADF);
4068 	}
4069 	vp = fp->f_vnode;
4070 unionread:
4071 	vfslocked = VFS_LOCK_GIANT(vp->v_mount);
4072 	if (vp->v_type != VDIR) {
4073 		VFS_UNLOCK_GIANT(vfslocked);
4074 		error = EINVAL;
4075 		goto fail;
4076 	}
4077 	aiov.iov_base = buf;
4078 	aiov.iov_len = count;
4079 	auio.uio_iov = &aiov;
4080 	auio.uio_iovcnt = 1;
4081 	auio.uio_rw = UIO_READ;
4082 	auio.uio_segflg = UIO_USERSPACE;
4083 	auio.uio_td = td;
4084 	auio.uio_resid = count;
4085 	vn_lock(vp, LK_SHARED | LK_RETRY);
4086 	AUDIT_ARG(vnode, vp, ARG_VNODE1);
4087 	loff = auio.uio_offset = fp->f_offset;
4088 #ifdef MAC
4089 	error = mac_vnode_check_readdir(td->td_ucred, vp);
4090 	if (error == 0)
4091 #endif
4092 		error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, NULL,
4093 		    NULL);
4094 	fp->f_offset = auio.uio_offset;
4095 	if (error) {
4096 		VOP_UNLOCK(vp, 0);
4097 		VFS_UNLOCK_GIANT(vfslocked);
4098 		goto fail;
4099 	}
4100 	if (count == auio.uio_resid &&
4101 	    (vp->v_vflag & VV_ROOT) &&
4102 	    (vp->v_mount->mnt_flag & MNT_UNION)) {
4103 		struct vnode *tvp = vp;
4104 		vp = vp->v_mount->mnt_vnodecovered;
4105 		VREF(vp);
4106 		fp->f_vnode = vp;
4107 		fp->f_data = vp;
4108 		fp->f_offset = 0;
4109 		vput(tvp);
4110 		VFS_UNLOCK_GIANT(vfslocked);
4111 		goto unionread;
4112 	}
4113 	VOP_UNLOCK(vp, 0);
4114 	VFS_UNLOCK_GIANT(vfslocked);
4115 	*basep = loff;
4116 	td->td_retval[0] = count - auio.uio_resid;
4117 fail:
4118 	fdrop(fp, td);
4119 	return (error);
4120 }
4121 
4122 #ifndef _SYS_SYSPROTO_H_
4123 struct getdents_args {
4124 	int fd;
4125 	char *buf;
4126 	size_t count;
4127 };
4128 #endif
4129 int
4130 getdents(td, uap)
4131 	struct thread *td;
4132 	register struct getdents_args /* {
4133 		int fd;
4134 		char *buf;
4135 		u_int count;
4136 	} */ *uap;
4137 {
4138 	struct getdirentries_args ap;
4139 	ap.fd = uap->fd;
4140 	ap.buf = uap->buf;
4141 	ap.count = uap->count;
4142 	ap.basep = NULL;
4143 	return (getdirentries(td, &ap));
4144 }
4145 
4146 /*
4147  * Set the mode mask for creation of filesystem nodes.
4148  */
4149 #ifndef _SYS_SYSPROTO_H_
4150 struct umask_args {
4151 	int	newmask;
4152 };
4153 #endif
4154 int
4155 umask(td, uap)
4156 	struct thread *td;
4157 	struct umask_args /* {
4158 		int newmask;
4159 	} */ *uap;
4160 {
4161 	register struct filedesc *fdp;
4162 
4163 	FILEDESC_XLOCK(td->td_proc->p_fd);
4164 	fdp = td->td_proc->p_fd;
4165 	td->td_retval[0] = fdp->fd_cmask;
4166 	fdp->fd_cmask = uap->newmask & ALLPERMS;
4167 	FILEDESC_XUNLOCK(td->td_proc->p_fd);
4168 	return (0);
4169 }
4170 
4171 /*
4172  * Void all references to file by ripping underlying filesystem away from
4173  * vnode.
4174  */
4175 #ifndef _SYS_SYSPROTO_H_
4176 struct revoke_args {
4177 	char	*path;
4178 };
4179 #endif
4180 int
4181 revoke(td, uap)
4182 	struct thread *td;
4183 	register struct revoke_args /* {
4184 		char *path;
4185 	} */ *uap;
4186 {
4187 	struct vnode *vp;
4188 	struct vattr vattr;
4189 	int error;
4190 	struct nameidata nd;
4191 	int vfslocked;
4192 
4193 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | MPSAFE | AUDITVNODE1,
4194 	    UIO_USERSPACE, uap->path, td);
4195 	if ((error = namei(&nd)) != 0)
4196 		return (error);
4197 	vfslocked = NDHASGIANT(&nd);
4198 	vp = nd.ni_vp;
4199 	NDFREE(&nd, NDF_ONLY_PNBUF);
4200 	if (vp->v_type != VCHR) {
4201 		error = EINVAL;
4202 		goto out;
4203 	}
4204 #ifdef MAC
4205 	error = mac_vnode_check_revoke(td->td_ucred, vp);
4206 	if (error)
4207 		goto out;
4208 #endif
4209 	error = VOP_GETATTR(vp, &vattr, td->td_ucred);
4210 	if (error)
4211 		goto out;
4212 	if (td->td_ucred->cr_uid != vattr.va_uid) {
4213 		error = priv_check(td, PRIV_VFS_ADMIN);
4214 		if (error)
4215 			goto out;
4216 	}
4217 	if (vcount(vp) > 1)
4218 		VOP_REVOKE(vp, REVOKEALL);
4219 out:
4220 	vput(vp);
4221 	VFS_UNLOCK_GIANT(vfslocked);
4222 	return (error);
4223 }
4224 
4225 /*
4226  * Convert a user file descriptor to a kernel file entry.
4227  * A reference on the file entry is held upon returning.
4228  */
4229 int
4230 getvnode(fdp, fd, fpp)
4231 	struct filedesc *fdp;
4232 	int fd;
4233 	struct file **fpp;
4234 {
4235 	int error;
4236 	struct file *fp;
4237 
4238 	error = 0;
4239 	fp = NULL;
4240 	if (fdp == NULL || (fp = fget_unlocked(fdp, fd)) == NULL)
4241 		error = EBADF;
4242 	else if (fp->f_vnode == NULL) {
4243 		error = EINVAL;
4244 		fdrop(fp, curthread);
4245 	}
4246 	*fpp = fp;
4247 	return (error);
4248 }
4249 
4250 /*
4251  * Get an (NFS) file handle.
4252  */
4253 #ifndef _SYS_SYSPROTO_H_
4254 struct lgetfh_args {
4255 	char	*fname;
4256 	fhandle_t *fhp;
4257 };
4258 #endif
4259 int
4260 lgetfh(td, uap)
4261 	struct thread *td;
4262 	register struct lgetfh_args *uap;
4263 {
4264 	struct nameidata nd;
4265 	fhandle_t fh;
4266 	register struct vnode *vp;
4267 	int vfslocked;
4268 	int error;
4269 
4270 	error = priv_check(td, PRIV_VFS_GETFH);
4271 	if (error)
4272 		return (error);
4273 	NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | MPSAFE | AUDITVNODE1,
4274 	    UIO_USERSPACE, uap->fname, td);
4275 	error = namei(&nd);
4276 	if (error)
4277 		return (error);
4278 	vfslocked = NDHASGIANT(&nd);
4279 	NDFREE(&nd, NDF_ONLY_PNBUF);
4280 	vp = nd.ni_vp;
4281 	bzero(&fh, sizeof(fh));
4282 	fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
4283 	error = VOP_VPTOFH(vp, &fh.fh_fid);
4284 	vput(vp);
4285 	VFS_UNLOCK_GIANT(vfslocked);
4286 	if (error)
4287 		return (error);
4288 	error = copyout(&fh, uap->fhp, sizeof (fh));
4289 	return (error);
4290 }
4291 
4292 #ifndef _SYS_SYSPROTO_H_
4293 struct getfh_args {
4294 	char	*fname;
4295 	fhandle_t *fhp;
4296 };
4297 #endif
4298 int
4299 getfh(td, uap)
4300 	struct thread *td;
4301 	register struct getfh_args *uap;
4302 {
4303 	struct nameidata nd;
4304 	fhandle_t fh;
4305 	register struct vnode *vp;
4306 	int vfslocked;
4307 	int error;
4308 
4309 	error = priv_check(td, PRIV_VFS_GETFH);
4310 	if (error)
4311 		return (error);
4312 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | MPSAFE | AUDITVNODE1,
4313 	    UIO_USERSPACE, uap->fname, td);
4314 	error = namei(&nd);
4315 	if (error)
4316 		return (error);
4317 	vfslocked = NDHASGIANT(&nd);
4318 	NDFREE(&nd, NDF_ONLY_PNBUF);
4319 	vp = nd.ni_vp;
4320 	bzero(&fh, sizeof(fh));
4321 	fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
4322 	error = VOP_VPTOFH(vp, &fh.fh_fid);
4323 	vput(vp);
4324 	VFS_UNLOCK_GIANT(vfslocked);
4325 	if (error)
4326 		return (error);
4327 	error = copyout(&fh, uap->fhp, sizeof (fh));
4328 	return (error);
4329 }
4330 
4331 /*
4332  * syscall for the rpc.lockd to use to translate a NFS file handle into an
4333  * open descriptor.
4334  *
4335  * warning: do not remove the priv_check() call or this becomes one giant
4336  * security hole.
4337  */
4338 #ifndef _SYS_SYSPROTO_H_
4339 struct fhopen_args {
4340 	const struct fhandle *u_fhp;
4341 	int flags;
4342 };
4343 #endif
4344 int
4345 fhopen(td, uap)
4346 	struct thread *td;
4347 	struct fhopen_args /* {
4348 		const struct fhandle *u_fhp;
4349 		int flags;
4350 	} */ *uap;
4351 {
4352 	struct proc *p = td->td_proc;
4353 	struct mount *mp;
4354 	struct vnode *vp;
4355 	struct fhandle fhp;
4356 	struct vattr vat;
4357 	struct vattr *vap = &vat;
4358 	struct flock lf;
4359 	struct file *fp;
4360 	register struct filedesc *fdp = p->p_fd;
4361 	int fmode, error, type;
4362 	accmode_t accmode;
4363 	struct file *nfp;
4364 	int vfslocked;
4365 	int indx;
4366 
4367 	error = priv_check(td, PRIV_VFS_FHOPEN);
4368 	if (error)
4369 		return (error);
4370 	fmode = FFLAGS(uap->flags);
4371 	/* why not allow a non-read/write open for our lockd? */
4372 	if (((fmode & (FREAD | FWRITE)) == 0) || (fmode & O_CREAT))
4373 		return (EINVAL);
4374 	error = copyin(uap->u_fhp, &fhp, sizeof(fhp));
4375 	if (error)
4376 		return(error);
4377 	/* find the mount point */
4378 	mp = vfs_busyfs(&fhp.fh_fsid);
4379 	if (mp == NULL)
4380 		return (ESTALE);
4381 	vfslocked = VFS_LOCK_GIANT(mp);
4382 	/* now give me my vnode, it gets returned to me locked */
4383 	error = VFS_FHTOVP(mp, &fhp.fh_fid, &vp);
4384 	vfs_unbusy(mp);
4385 	if (error)
4386 		goto out;
4387 	/*
4388 	 * from now on we have to make sure not
4389 	 * to forget about the vnode
4390 	 * any error that causes an abort must vput(vp)
4391 	 * just set error = err and 'goto bad;'.
4392 	 */
4393 
4394 	/*
4395 	 * from vn_open
4396 	 */
4397 	if (vp->v_type == VLNK) {
4398 		error = EMLINK;
4399 		goto bad;
4400 	}
4401 	if (vp->v_type == VSOCK) {
4402 		error = EOPNOTSUPP;
4403 		goto bad;
4404 	}
4405 	accmode = 0;
4406 	if (fmode & (FWRITE | O_TRUNC)) {
4407 		if (vp->v_type == VDIR) {
4408 			error = EISDIR;
4409 			goto bad;
4410 		}
4411 		error = vn_writechk(vp);
4412 		if (error)
4413 			goto bad;
4414 		accmode |= VWRITE;
4415 	}
4416 	if (fmode & FREAD)
4417 		accmode |= VREAD;
4418 	if (fmode & O_APPEND)
4419 		accmode |= VAPPEND;
4420 #ifdef MAC
4421 	error = mac_vnode_check_open(td->td_ucred, vp, accmode);
4422 	if (error)
4423 		goto bad;
4424 #endif
4425 	if (accmode) {
4426 		error = VOP_ACCESS(vp, accmode, td->td_ucred, td);
4427 		if (error)
4428 			goto bad;
4429 	}
4430 	if (fmode & O_TRUNC) {
4431 		VOP_UNLOCK(vp, 0);				/* XXX */
4432 		if ((error = vn_start_write(NULL, &mp, V_WAIT | PCATCH)) != 0) {
4433 			vrele(vp);
4434 			goto out;
4435 		}
4436 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);	/* XXX */
4437 #ifdef MAC
4438 		/*
4439 		 * We don't yet have fp->f_cred, so use td->td_ucred, which
4440 		 * should be right.
4441 		 */
4442 		error = mac_vnode_check_write(td->td_ucred, td->td_ucred, vp);
4443 		if (error == 0) {
4444 #endif
4445 			VATTR_NULL(vap);
4446 			vap->va_size = 0;
4447 			error = VOP_SETATTR(vp, vap, td->td_ucred);
4448 #ifdef MAC
4449 		}
4450 #endif
4451 		vn_finished_write(mp);
4452 		if (error)
4453 			goto bad;
4454 	}
4455 	error = VOP_OPEN(vp, fmode, td->td_ucred, td, NULL);
4456 	if (error)
4457 		goto bad;
4458 
4459 	if (fmode & FWRITE)
4460 		vp->v_writecount++;
4461 
4462 	/*
4463 	 * end of vn_open code
4464 	 */
4465 
4466 	if ((error = falloc(td, &nfp, &indx)) != 0) {
4467 		if (fmode & FWRITE)
4468 			vp->v_writecount--;
4469 		goto bad;
4470 	}
4471 	/* An extra reference on `nfp' has been held for us by falloc(). */
4472 	fp = nfp;
4473 	nfp->f_vnode = vp;
4474 	finit(nfp, fmode & FMASK, DTYPE_VNODE, vp, &vnops);
4475 	if (fmode & (O_EXLOCK | O_SHLOCK)) {
4476 		lf.l_whence = SEEK_SET;
4477 		lf.l_start = 0;
4478 		lf.l_len = 0;
4479 		if (fmode & O_EXLOCK)
4480 			lf.l_type = F_WRLCK;
4481 		else
4482 			lf.l_type = F_RDLCK;
4483 		type = F_FLOCK;
4484 		if ((fmode & FNONBLOCK) == 0)
4485 			type |= F_WAIT;
4486 		VOP_UNLOCK(vp, 0);
4487 		if ((error = VOP_ADVLOCK(vp, (caddr_t)fp, F_SETLK, &lf,
4488 			    type)) != 0) {
4489 			/*
4490 			 * The lock request failed.  Normally close the
4491 			 * descriptor but handle the case where someone might
4492 			 * have dup()d or close()d it when we weren't looking.
4493 			 */
4494 			fdclose(fdp, fp, indx, td);
4495 
4496 			/*
4497 			 * release our private reference
4498 			 */
4499 			fdrop(fp, td);
4500 			goto out;
4501 		}
4502 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
4503 		atomic_set_int(&fp->f_flag, FHASLOCK);
4504 	}
4505 
4506 	VOP_UNLOCK(vp, 0);
4507 	fdrop(fp, td);
4508 	vfs_rel(mp);
4509 	VFS_UNLOCK_GIANT(vfslocked);
4510 	td->td_retval[0] = indx;
4511 	return (0);
4512 
4513 bad:
4514 	vput(vp);
4515 out:
4516 	VFS_UNLOCK_GIANT(vfslocked);
4517 	return (error);
4518 }
4519 
4520 /*
4521  * Stat an (NFS) file handle.
4522  */
4523 #ifndef _SYS_SYSPROTO_H_
4524 struct fhstat_args {
4525 	struct fhandle *u_fhp;
4526 	struct stat *sb;
4527 };
4528 #endif
4529 int
4530 fhstat(td, uap)
4531 	struct thread *td;
4532 	register struct fhstat_args /* {
4533 		struct fhandle *u_fhp;
4534 		struct stat *sb;
4535 	} */ *uap;
4536 {
4537 	struct stat sb;
4538 	fhandle_t fh;
4539 	struct mount *mp;
4540 	struct vnode *vp;
4541 	int vfslocked;
4542 	int error;
4543 
4544 	error = priv_check(td, PRIV_VFS_FHSTAT);
4545 	if (error)
4546 		return (error);
4547 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
4548 	if (error)
4549 		return (error);
4550 	if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4551 		return (ESTALE);
4552 	vfslocked = VFS_LOCK_GIANT(mp);
4553 	error = VFS_FHTOVP(mp, &fh.fh_fid, &vp);
4554 	vfs_unbusy(mp);
4555 	if (error) {
4556 		VFS_UNLOCK_GIANT(vfslocked);
4557 		return (error);
4558 	}
4559 	error = vn_stat(vp, &sb, td->td_ucred, NOCRED, td);
4560 	vput(vp);
4561 	VFS_UNLOCK_GIANT(vfslocked);
4562 	if (error)
4563 		return (error);
4564 	error = copyout(&sb, uap->sb, sizeof(sb));
4565 	return (error);
4566 }
4567 
4568 /*
4569  * Implement fstatfs() for (NFS) file handles.
4570  */
4571 #ifndef _SYS_SYSPROTO_H_
4572 struct fhstatfs_args {
4573 	struct fhandle *u_fhp;
4574 	struct statfs *buf;
4575 };
4576 #endif
4577 int
4578 fhstatfs(td, uap)
4579 	struct thread *td;
4580 	struct fhstatfs_args /* {
4581 		struct fhandle *u_fhp;
4582 		struct statfs *buf;
4583 	} */ *uap;
4584 {
4585 	struct statfs sf;
4586 	fhandle_t fh;
4587 	int error;
4588 
4589 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
4590 	if (error)
4591 		return (error);
4592 	error = kern_fhstatfs(td, fh, &sf);
4593 	if (error)
4594 		return (error);
4595 	return (copyout(&sf, uap->buf, sizeof(sf)));
4596 }
4597 
4598 int
4599 kern_fhstatfs(struct thread *td, fhandle_t fh, struct statfs *buf)
4600 {
4601 	struct statfs *sp;
4602 	struct mount *mp;
4603 	struct vnode *vp;
4604 	int vfslocked;
4605 	int error;
4606 
4607 	error = priv_check(td, PRIV_VFS_FHSTATFS);
4608 	if (error)
4609 		return (error);
4610 	if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4611 		return (ESTALE);
4612 	vfslocked = VFS_LOCK_GIANT(mp);
4613 	error = VFS_FHTOVP(mp, &fh.fh_fid, &vp);
4614 	if (error) {
4615 		vfs_unbusy(mp);
4616 		VFS_UNLOCK_GIANT(vfslocked);
4617 		return (error);
4618 	}
4619 	vput(vp);
4620 	error = prison_canseemount(td->td_ucred, mp);
4621 	if (error)
4622 		goto out;
4623 #ifdef MAC
4624 	error = mac_mount_check_stat(td->td_ucred, mp);
4625 	if (error)
4626 		goto out;
4627 #endif
4628 	/*
4629 	 * Set these in case the underlying filesystem fails to do so.
4630 	 */
4631 	sp = &mp->mnt_stat;
4632 	sp->f_version = STATFS_VERSION;
4633 	sp->f_namemax = NAME_MAX;
4634 	sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
4635 	error = VFS_STATFS(mp, sp);
4636 	if (error == 0)
4637 		*buf = *sp;
4638 out:
4639 	vfs_unbusy(mp);
4640 	VFS_UNLOCK_GIANT(vfslocked);
4641 	return (error);
4642 }
4643