xref: /freebsd/sys/kern/vfs_syscalls.c (revision d876124d6ae9d56da5b4ff4c6015efd1d0c9222a)
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_ktrace.h"
42 #include "opt_mac.h"
43 
44 #include <sys/param.h>
45 #include <sys/systm.h>
46 #include <sys/bio.h>
47 #include <sys/buf.h>
48 #include <sys/sysent.h>
49 #include <sys/malloc.h>
50 #include <sys/mount.h>
51 #include <sys/mutex.h>
52 #include <sys/sysproto.h>
53 #include <sys/namei.h>
54 #include <sys/filedesc.h>
55 #include <sys/kernel.h>
56 #include <sys/fcntl.h>
57 #include <sys/file.h>
58 #include <sys/filio.h>
59 #include <sys/limits.h>
60 #include <sys/linker.h>
61 #include <sys/stat.h>
62 #include <sys/sx.h>
63 #include <sys/unistd.h>
64 #include <sys/vnode.h>
65 #include <sys/priv.h>
66 #include <sys/proc.h>
67 #include <sys/dirent.h>
68 #include <sys/jail.h>
69 #include <sys/syscallsubr.h>
70 #include <sys/sysctl.h>
71 #ifdef KTRACE
72 #include <sys/ktrace.h>
73 #endif
74 
75 #include <machine/stdarg.h>
76 
77 #include <security/audit/audit.h>
78 #include <security/mac/mac_framework.h>
79 
80 #include <vm/vm.h>
81 #include <vm/vm_object.h>
82 #include <vm/vm_page.h>
83 #include <vm/uma.h>
84 
85 static int chroot_refuse_vdir_fds(struct filedesc *fdp);
86 static int getutimes(const struct timeval *, enum uio_seg, struct timespec *);
87 static int setfown(struct thread *td, struct vnode *, uid_t, gid_t);
88 static int setfmode(struct thread *td, struct vnode *, int);
89 static int setfflags(struct thread *td, struct vnode *, int);
90 static int setutimes(struct thread *td, struct vnode *,
91     const struct timespec *, int, int);
92 static int vn_access(struct vnode *vp, int user_flags, struct ucred *cred,
93     struct thread *td);
94 
95 /*
96  * The module initialization routine for POSIX asynchronous I/O will
97  * set this to the version of AIO that it implements.  (Zero means
98  * that it is not implemented.)  This value is used here by pathconf()
99  * and in kern_descrip.c by fpathconf().
100  */
101 int async_io_version;
102 
103 #ifdef DEBUG
104 static int syncprt = 0;
105 SYSCTL_INT(_debug, OID_AUTO, syncprt, CTLFLAG_RW, &syncprt, 0, "");
106 #endif
107 
108 /*
109  * Sync each mounted filesystem.
110  */
111 #ifndef _SYS_SYSPROTO_H_
112 struct sync_args {
113 	int     dummy;
114 };
115 #endif
116 /* ARGSUSED */
117 int
118 sync(td, uap)
119 	struct thread *td;
120 	struct sync_args *uap;
121 {
122 	struct mount *mp, *nmp;
123 	int vfslocked;
124 
125 	mtx_lock(&mountlist_mtx);
126 	for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
127 		if (vfs_busy(mp, LK_NOWAIT, &mountlist_mtx, td)) {
128 			nmp = TAILQ_NEXT(mp, mnt_list);
129 			continue;
130 		}
131 		vfslocked = VFS_LOCK_GIANT(mp);
132 		if ((mp->mnt_flag & MNT_RDONLY) == 0 &&
133 		    vn_start_write(NULL, &mp, V_NOWAIT) == 0) {
134 			MNT_ILOCK(mp);
135 			mp->mnt_noasync++;
136 			mp->mnt_kern_flag &= ~MNTK_ASYNC;
137 			MNT_IUNLOCK(mp);
138 			vfs_msync(mp, MNT_NOWAIT);
139 			VFS_SYNC(mp, MNT_NOWAIT, td);
140 			MNT_ILOCK(mp);
141 			mp->mnt_noasync--;
142 			if ((mp->mnt_flag & MNT_ASYNC) != 0 &&
143 			    mp->mnt_noasync == 0)
144 				mp->mnt_kern_flag |= MNTK_ASYNC;
145 			MNT_IUNLOCK(mp);
146 			vn_finished_write(mp);
147 		}
148 		VFS_UNLOCK_GIANT(vfslocked);
149 		mtx_lock(&mountlist_mtx);
150 		nmp = TAILQ_NEXT(mp, mnt_list);
151 		vfs_unbusy(mp, td);
152 	}
153 	mtx_unlock(&mountlist_mtx);
154 	return (0);
155 }
156 
157 /* XXX PRISON: could be per prison flag */
158 static int prison_quotas;
159 #if 0
160 SYSCTL_INT(_kern_prison, OID_AUTO, quotas, CTLFLAG_RW, &prison_quotas, 0, "");
161 #endif
162 
163 /*
164  * Change filesystem quotas.
165  */
166 #ifndef _SYS_SYSPROTO_H_
167 struct quotactl_args {
168 	char *path;
169 	int cmd;
170 	int uid;
171 	caddr_t arg;
172 };
173 #endif
174 int
175 quotactl(td, uap)
176 	struct thread *td;
177 	register struct quotactl_args /* {
178 		char *path;
179 		int cmd;
180 		int uid;
181 		caddr_t arg;
182 	} */ *uap;
183 {
184 	struct mount *mp;
185 	int vfslocked;
186 	int error;
187 	struct nameidata nd;
188 
189 	AUDIT_ARG(cmd, uap->cmd);
190 	AUDIT_ARG(uid, uap->uid);
191 	if (jailed(td->td_ucred) && !prison_quotas)
192 		return (EPERM);
193 	NDINIT(&nd, LOOKUP, FOLLOW | MPSAFE | AUDITVNODE1,
194 	   UIO_USERSPACE, uap->path, td);
195 	if ((error = namei(&nd)) != 0)
196 		return (error);
197 	vfslocked = NDHASGIANT(&nd);
198 	NDFREE(&nd, NDF_ONLY_PNBUF);
199 	mp = nd.ni_vp->v_mount;
200 	if ((error = vfs_busy(mp, 0, NULL, td))) {
201 		vrele(nd.ni_vp);
202 		VFS_UNLOCK_GIANT(vfslocked);
203 		return (error);
204 	}
205 	vrele(nd.ni_vp);
206 	error = VFS_QUOTACTL(mp, uap->cmd, uap->uid, uap->arg, td);
207 	vfs_unbusy(mp, td);
208 	VFS_UNLOCK_GIANT(vfslocked);
209 	return (error);
210 }
211 
212 /*
213  * Used by statfs conversion routines to scale the block size up if
214  * necessary so that all of the block counts are <= 'max_size'.  Note
215  * that 'max_size' should be a bitmask, i.e. 2^n - 1 for some non-zero
216  * value of 'n'.
217  */
218 void
219 statfs_scale_blocks(struct statfs *sf, long max_size)
220 {
221 	uint64_t count;
222 	int shift;
223 
224 	KASSERT(powerof2(max_size + 1), ("%s: invalid max_size", __func__));
225 
226 	/*
227 	 * Attempt to scale the block counts to give a more accurate
228 	 * overview to userland of the ratio of free space to used
229 	 * space.  To do this, find the largest block count and compute
230 	 * a divisor that lets it fit into a signed integer <= max_size.
231 	 */
232 	if (sf->f_bavail < 0)
233 		count = -sf->f_bavail;
234 	else
235 		count = sf->f_bavail;
236 	count = MAX(sf->f_blocks, MAX(sf->f_bfree, count));
237 	if (count <= max_size)
238 		return;
239 
240 	count >>= flsl(max_size);
241 	shift = 0;
242 	while (count > 0) {
243 		shift++;
244 		count >>=1;
245 	}
246 
247 	sf->f_bsize <<= shift;
248 	sf->f_blocks >>= shift;
249 	sf->f_bfree >>= shift;
250 	sf->f_bavail >>= shift;
251 }
252 
253 /*
254  * Get filesystem statistics.
255  */
256 #ifndef _SYS_SYSPROTO_H_
257 struct statfs_args {
258 	char *path;
259 	struct statfs *buf;
260 };
261 #endif
262 int
263 statfs(td, uap)
264 	struct thread *td;
265 	register struct statfs_args /* {
266 		char *path;
267 		struct statfs *buf;
268 	} */ *uap;
269 {
270 	struct statfs sf;
271 	int error;
272 
273 	error = kern_statfs(td, uap->path, UIO_USERSPACE, &sf);
274 	if (error == 0)
275 		error = copyout(&sf, uap->buf, sizeof(sf));
276 	return (error);
277 }
278 
279 int
280 kern_statfs(struct thread *td, char *path, enum uio_seg pathseg,
281     struct statfs *buf)
282 {
283 	struct mount *mp;
284 	struct statfs *sp, sb;
285 	int vfslocked;
286 	int error;
287 	struct nameidata nd;
288 
289 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | MPSAFE | AUDITVNODE1,
290 	    pathseg, path, td);
291 	error = namei(&nd);
292 	if (error)
293 		return (error);
294 	vfslocked = NDHASGIANT(&nd);
295 	mp = nd.ni_vp->v_mount;
296 	vfs_ref(mp);
297 	NDFREE(&nd, NDF_ONLY_PNBUF);
298 	vput(nd.ni_vp);
299 #ifdef MAC
300 	error = mac_mount_check_stat(td->td_ucred, mp);
301 	if (error)
302 		goto out;
303 #endif
304 	/*
305 	 * Set these in case the underlying filesystem fails to do so.
306 	 */
307 	sp = &mp->mnt_stat;
308 	sp->f_version = STATFS_VERSION;
309 	sp->f_namemax = NAME_MAX;
310 	sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
311 	error = VFS_STATFS(mp, sp, td);
312 	if (error)
313 		goto out;
314 	if (priv_check(td, PRIV_VFS_GENERATION)) {
315 		bcopy(sp, &sb, sizeof(sb));
316 		sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
317 		prison_enforce_statfs(td->td_ucred, mp, &sb);
318 		sp = &sb;
319 	}
320 	*buf = *sp;
321 out:
322 	vfs_rel(mp);
323 	VFS_UNLOCK_GIANT(vfslocked);
324 	if (mtx_owned(&Giant))
325 		printf("statfs(%d): %s: %d\n", vfslocked, path, error);
326 	return (error);
327 }
328 
329 /*
330  * Get filesystem statistics.
331  */
332 #ifndef _SYS_SYSPROTO_H_
333 struct fstatfs_args {
334 	int fd;
335 	struct statfs *buf;
336 };
337 #endif
338 int
339 fstatfs(td, uap)
340 	struct thread *td;
341 	register struct fstatfs_args /* {
342 		int fd;
343 		struct statfs *buf;
344 	} */ *uap;
345 {
346 	struct statfs sf;
347 	int error;
348 
349 	error = kern_fstatfs(td, uap->fd, &sf);
350 	if (error == 0)
351 		error = copyout(&sf, uap->buf, sizeof(sf));
352 	return (error);
353 }
354 
355 int
356 kern_fstatfs(struct thread *td, int fd, struct statfs *buf)
357 {
358 	struct file *fp;
359 	struct mount *mp;
360 	struct statfs *sp, sb;
361 	int vfslocked;
362 	struct vnode *vp;
363 	int error;
364 
365 	AUDIT_ARG(fd, fd);
366 	error = getvnode(td->td_proc->p_fd, fd, &fp);
367 	if (error)
368 		return (error);
369 	vp = fp->f_vnode;
370 	vfslocked = VFS_LOCK_GIANT(vp->v_mount);
371 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
372 #ifdef AUDIT
373 	AUDIT_ARG(vnode, vp, ARG_VNODE1);
374 #endif
375 	mp = vp->v_mount;
376 	if (mp)
377 		vfs_ref(mp);
378 	VOP_UNLOCK(vp, 0);
379 	fdrop(fp, td);
380 	if (vp->v_iflag & VI_DOOMED) {
381 		error = EBADF;
382 		goto out;
383 	}
384 #ifdef MAC
385 	error = mac_mount_check_stat(td->td_ucred, mp);
386 	if (error)
387 		goto out;
388 #endif
389 	/*
390 	 * Set these in case the underlying filesystem fails to do so.
391 	 */
392 	sp = &mp->mnt_stat;
393 	sp->f_version = STATFS_VERSION;
394 	sp->f_namemax = NAME_MAX;
395 	sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
396 	error = VFS_STATFS(mp, sp, td);
397 	if (error)
398 		goto out;
399 	if (priv_check(td, PRIV_VFS_GENERATION)) {
400 		bcopy(sp, &sb, sizeof(sb));
401 		sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
402 		prison_enforce_statfs(td->td_ucred, mp, &sb);
403 		sp = &sb;
404 	}
405 	*buf = *sp;
406 out:
407 	if (mp)
408 		vfs_rel(mp);
409 	VFS_UNLOCK_GIANT(vfslocked);
410 	return (error);
411 }
412 
413 /*
414  * Get statistics on all filesystems.
415  */
416 #ifndef _SYS_SYSPROTO_H_
417 struct getfsstat_args {
418 	struct statfs *buf;
419 	long bufsize;
420 	int flags;
421 };
422 #endif
423 int
424 getfsstat(td, uap)
425 	struct thread *td;
426 	register struct getfsstat_args /* {
427 		struct statfs *buf;
428 		long bufsize;
429 		int flags;
430 	} */ *uap;
431 {
432 
433 	return (kern_getfsstat(td, &uap->buf, uap->bufsize, UIO_USERSPACE,
434 	    uap->flags));
435 }
436 
437 /*
438  * If (bufsize > 0 && bufseg == UIO_SYSSPACE)
439  * 	The caller is responsible for freeing memory which will be allocated
440  *	in '*buf'.
441  */
442 int
443 kern_getfsstat(struct thread *td, struct statfs **buf, size_t bufsize,
444     enum uio_seg bufseg, int flags)
445 {
446 	struct mount *mp, *nmp;
447 	struct statfs *sfsp, *sp, sb;
448 	size_t count, maxcount;
449 	int vfslocked;
450 	int error;
451 
452 	maxcount = bufsize / sizeof(struct statfs);
453 	if (bufsize == 0)
454 		sfsp = NULL;
455 	else if (bufseg == UIO_USERSPACE)
456 		sfsp = *buf;
457 	else /* if (bufseg == UIO_SYSSPACE) */ {
458 		count = 0;
459 		mtx_lock(&mountlist_mtx);
460 		TAILQ_FOREACH(mp, &mountlist, mnt_list) {
461 			count++;
462 		}
463 		mtx_unlock(&mountlist_mtx);
464 		if (maxcount > count)
465 			maxcount = count;
466 		sfsp = *buf = malloc(maxcount * sizeof(struct statfs), M_TEMP,
467 		    M_WAITOK);
468 	}
469 	count = 0;
470 	mtx_lock(&mountlist_mtx);
471 	for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
472 		if (prison_canseemount(td->td_ucred, mp) != 0) {
473 			nmp = TAILQ_NEXT(mp, mnt_list);
474 			continue;
475 		}
476 #ifdef MAC
477 		if (mac_mount_check_stat(td->td_ucred, mp) != 0) {
478 			nmp = TAILQ_NEXT(mp, mnt_list);
479 			continue;
480 		}
481 #endif
482 		if (vfs_busy(mp, LK_NOWAIT, &mountlist_mtx, td)) {
483 			nmp = TAILQ_NEXT(mp, mnt_list);
484 			continue;
485 		}
486 		vfslocked = VFS_LOCK_GIANT(mp);
487 		if (sfsp && count < maxcount) {
488 			sp = &mp->mnt_stat;
489 			/*
490 			 * Set these in case the underlying filesystem
491 			 * fails to do so.
492 			 */
493 			sp->f_version = STATFS_VERSION;
494 			sp->f_namemax = NAME_MAX;
495 			sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
496 			/*
497 			 * If MNT_NOWAIT or MNT_LAZY is specified, do not
498 			 * refresh the fsstat cache. MNT_NOWAIT or MNT_LAZY
499 			 * overrides MNT_WAIT.
500 			 */
501 			if (((flags & (MNT_LAZY|MNT_NOWAIT)) == 0 ||
502 			    (flags & MNT_WAIT)) &&
503 			    (error = VFS_STATFS(mp, sp, td))) {
504 				VFS_UNLOCK_GIANT(vfslocked);
505 				mtx_lock(&mountlist_mtx);
506 				nmp = TAILQ_NEXT(mp, mnt_list);
507 				vfs_unbusy(mp, td);
508 				continue;
509 			}
510 			if (priv_check(td, PRIV_VFS_GENERATION)) {
511 				bcopy(sp, &sb, sizeof(sb));
512 				sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
513 				prison_enforce_statfs(td->td_ucred, mp, &sb);
514 				sp = &sb;
515 			}
516 			if (bufseg == UIO_SYSSPACE)
517 				bcopy(sp, sfsp, sizeof(*sp));
518 			else /* if (bufseg == UIO_USERSPACE) */ {
519 				error = copyout(sp, sfsp, sizeof(*sp));
520 				if (error) {
521 					vfs_unbusy(mp, td);
522 					VFS_UNLOCK_GIANT(vfslocked);
523 					return (error);
524 				}
525 			}
526 			sfsp++;
527 		}
528 		VFS_UNLOCK_GIANT(vfslocked);
529 		count++;
530 		mtx_lock(&mountlist_mtx);
531 		nmp = TAILQ_NEXT(mp, mnt_list);
532 		vfs_unbusy(mp, td);
533 	}
534 	mtx_unlock(&mountlist_mtx);
535 	if (sfsp && count > maxcount)
536 		td->td_retval[0] = maxcount;
537 	else
538 		td->td_retval[0] = count;
539 	return (0);
540 }
541 
542 #ifdef COMPAT_FREEBSD4
543 /*
544  * Get old format filesystem statistics.
545  */
546 static void cvtstatfs(struct statfs *, struct ostatfs *);
547 
548 #ifndef _SYS_SYSPROTO_H_
549 struct freebsd4_statfs_args {
550 	char *path;
551 	struct ostatfs *buf;
552 };
553 #endif
554 int
555 freebsd4_statfs(td, uap)
556 	struct thread *td;
557 	struct freebsd4_statfs_args /* {
558 		char *path;
559 		struct ostatfs *buf;
560 	} */ *uap;
561 {
562 	struct ostatfs osb;
563 	struct statfs sf;
564 	int error;
565 
566 	error = kern_statfs(td, uap->path, UIO_USERSPACE, &sf);
567 	if (error)
568 		return (error);
569 	cvtstatfs(&sf, &osb);
570 	return (copyout(&osb, uap->buf, sizeof(osb)));
571 }
572 
573 /*
574  * Get filesystem statistics.
575  */
576 #ifndef _SYS_SYSPROTO_H_
577 struct freebsd4_fstatfs_args {
578 	int fd;
579 	struct ostatfs *buf;
580 };
581 #endif
582 int
583 freebsd4_fstatfs(td, uap)
584 	struct thread *td;
585 	struct freebsd4_fstatfs_args /* {
586 		int fd;
587 		struct ostatfs *buf;
588 	} */ *uap;
589 {
590 	struct ostatfs osb;
591 	struct statfs sf;
592 	int error;
593 
594 	error = kern_fstatfs(td, uap->fd, &sf);
595 	if (error)
596 		return (error);
597 	cvtstatfs(&sf, &osb);
598 	return (copyout(&osb, uap->buf, sizeof(osb)));
599 }
600 
601 /*
602  * Get statistics on all filesystems.
603  */
604 #ifndef _SYS_SYSPROTO_H_
605 struct freebsd4_getfsstat_args {
606 	struct ostatfs *buf;
607 	long bufsize;
608 	int flags;
609 };
610 #endif
611 int
612 freebsd4_getfsstat(td, uap)
613 	struct thread *td;
614 	register struct freebsd4_getfsstat_args /* {
615 		struct ostatfs *buf;
616 		long bufsize;
617 		int flags;
618 	} */ *uap;
619 {
620 	struct statfs *buf, *sp;
621 	struct ostatfs osb;
622 	size_t count, size;
623 	int error;
624 
625 	count = uap->bufsize / sizeof(struct ostatfs);
626 	size = count * sizeof(struct statfs);
627 	error = kern_getfsstat(td, &buf, size, UIO_SYSSPACE, uap->flags);
628 	if (size > 0) {
629 		count = td->td_retval[0];
630 		sp = buf;
631 		while (count > 0 && error == 0) {
632 			cvtstatfs(sp, &osb);
633 			error = copyout(&osb, uap->buf, sizeof(osb));
634 			sp++;
635 			uap->buf++;
636 			count--;
637 		}
638 		free(buf, M_TEMP);
639 	}
640 	return (error);
641 }
642 
643 /*
644  * Implement fstatfs() for (NFS) file handles.
645  */
646 #ifndef _SYS_SYSPROTO_H_
647 struct freebsd4_fhstatfs_args {
648 	struct fhandle *u_fhp;
649 	struct ostatfs *buf;
650 };
651 #endif
652 int
653 freebsd4_fhstatfs(td, uap)
654 	struct thread *td;
655 	struct freebsd4_fhstatfs_args /* {
656 		struct fhandle *u_fhp;
657 		struct ostatfs *buf;
658 	} */ *uap;
659 {
660 	struct ostatfs osb;
661 	struct statfs sf;
662 	fhandle_t fh;
663 	int error;
664 
665 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
666 	if (error)
667 		return (error);
668 	error = kern_fhstatfs(td, fh, &sf);
669 	if (error)
670 		return (error);
671 	cvtstatfs(&sf, &osb);
672 	return (copyout(&osb, uap->buf, sizeof(osb)));
673 }
674 
675 /*
676  * Convert a new format statfs structure to an old format statfs structure.
677  */
678 static void
679 cvtstatfs(nsp, osp)
680 	struct statfs *nsp;
681 	struct ostatfs *osp;
682 {
683 
684 	statfs_scale_blocks(nsp, LONG_MAX);
685 	bzero(osp, sizeof(*osp));
686 	osp->f_bsize = nsp->f_bsize;
687 	osp->f_iosize = MIN(nsp->f_iosize, LONG_MAX);
688 	osp->f_blocks = nsp->f_blocks;
689 	osp->f_bfree = nsp->f_bfree;
690 	osp->f_bavail = nsp->f_bavail;
691 	osp->f_files = MIN(nsp->f_files, LONG_MAX);
692 	osp->f_ffree = MIN(nsp->f_ffree, LONG_MAX);
693 	osp->f_owner = nsp->f_owner;
694 	osp->f_type = nsp->f_type;
695 	osp->f_flags = nsp->f_flags;
696 	osp->f_syncwrites = MIN(nsp->f_syncwrites, LONG_MAX);
697 	osp->f_asyncwrites = MIN(nsp->f_asyncwrites, LONG_MAX);
698 	osp->f_syncreads = MIN(nsp->f_syncreads, LONG_MAX);
699 	osp->f_asyncreads = MIN(nsp->f_asyncreads, LONG_MAX);
700 	strlcpy(osp->f_fstypename, nsp->f_fstypename,
701 	    MIN(MFSNAMELEN, OMFSNAMELEN));
702 	strlcpy(osp->f_mntonname, nsp->f_mntonname,
703 	    MIN(MNAMELEN, OMNAMELEN));
704 	strlcpy(osp->f_mntfromname, nsp->f_mntfromname,
705 	    MIN(MNAMELEN, OMNAMELEN));
706 	osp->f_fsid = nsp->f_fsid;
707 }
708 #endif /* COMPAT_FREEBSD4 */
709 
710 /*
711  * Change current working directory to a given file descriptor.
712  */
713 #ifndef _SYS_SYSPROTO_H_
714 struct fchdir_args {
715 	int	fd;
716 };
717 #endif
718 int
719 fchdir(td, uap)
720 	struct thread *td;
721 	struct fchdir_args /* {
722 		int fd;
723 	} */ *uap;
724 {
725 	register struct filedesc *fdp = td->td_proc->p_fd;
726 	struct vnode *vp, *tdp, *vpold;
727 	struct mount *mp;
728 	struct file *fp;
729 	int vfslocked;
730 	int error;
731 
732 	AUDIT_ARG(fd, uap->fd);
733 	if ((error = getvnode(fdp, uap->fd, &fp)) != 0)
734 		return (error);
735 	vp = fp->f_vnode;
736 	VREF(vp);
737 	fdrop(fp, td);
738 	vfslocked = VFS_LOCK_GIANT(vp->v_mount);
739 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
740 	AUDIT_ARG(vnode, vp, ARG_VNODE1);
741 	error = change_dir(vp, td);
742 	while (!error && (mp = vp->v_mountedhere) != NULL) {
743 		int tvfslocked;
744 		if (vfs_busy(mp, 0, 0, td))
745 			continue;
746 		tvfslocked = VFS_LOCK_GIANT(mp);
747 		error = VFS_ROOT(mp, LK_EXCLUSIVE, &tdp, td);
748 		vfs_unbusy(mp, td);
749 		if (error) {
750 			VFS_UNLOCK_GIANT(tvfslocked);
751 			break;
752 		}
753 		vput(vp);
754 		VFS_UNLOCK_GIANT(vfslocked);
755 		vp = tdp;
756 		vfslocked = tvfslocked;
757 	}
758 	if (error) {
759 		vput(vp);
760 		VFS_UNLOCK_GIANT(vfslocked);
761 		return (error);
762 	}
763 	VOP_UNLOCK(vp, 0);
764 	VFS_UNLOCK_GIANT(vfslocked);
765 	FILEDESC_XLOCK(fdp);
766 	vpold = fdp->fd_cdir;
767 	fdp->fd_cdir = vp;
768 	FILEDESC_XUNLOCK(fdp);
769 	vfslocked = VFS_LOCK_GIANT(vpold->v_mount);
770 	vrele(vpold);
771 	VFS_UNLOCK_GIANT(vfslocked);
772 	return (0);
773 }
774 
775 /*
776  * Change current working directory (``.'').
777  */
778 #ifndef _SYS_SYSPROTO_H_
779 struct chdir_args {
780 	char	*path;
781 };
782 #endif
783 int
784 chdir(td, uap)
785 	struct thread *td;
786 	struct chdir_args /* {
787 		char *path;
788 	} */ *uap;
789 {
790 
791 	return (kern_chdir(td, uap->path, UIO_USERSPACE));
792 }
793 
794 int
795 kern_chdir(struct thread *td, char *path, enum uio_seg pathseg)
796 {
797 	register struct filedesc *fdp = td->td_proc->p_fd;
798 	int error;
799 	struct nameidata nd;
800 	struct vnode *vp;
801 	int vfslocked;
802 
803 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1 | MPSAFE,
804 	    pathseg, path, td);
805 	if ((error = namei(&nd)) != 0)
806 		return (error);
807 	vfslocked = NDHASGIANT(&nd);
808 	if ((error = change_dir(nd.ni_vp, td)) != 0) {
809 		vput(nd.ni_vp);
810 		VFS_UNLOCK_GIANT(vfslocked);
811 		NDFREE(&nd, NDF_ONLY_PNBUF);
812 		return (error);
813 	}
814 	VOP_UNLOCK(nd.ni_vp, 0);
815 	VFS_UNLOCK_GIANT(vfslocked);
816 	NDFREE(&nd, NDF_ONLY_PNBUF);
817 	FILEDESC_XLOCK(fdp);
818 	vp = fdp->fd_cdir;
819 	fdp->fd_cdir = nd.ni_vp;
820 	FILEDESC_XUNLOCK(fdp);
821 	vfslocked = VFS_LOCK_GIANT(vp->v_mount);
822 	vrele(vp);
823 	VFS_UNLOCK_GIANT(vfslocked);
824 	return (0);
825 }
826 
827 /*
828  * Helper function for raised chroot(2) security function:  Refuse if
829  * any filedescriptors are open directories.
830  */
831 static int
832 chroot_refuse_vdir_fds(fdp)
833 	struct filedesc *fdp;
834 {
835 	struct vnode *vp;
836 	struct file *fp;
837 	int fd;
838 
839 	FILEDESC_LOCK_ASSERT(fdp);
840 
841 	for (fd = 0; fd < fdp->fd_nfiles ; fd++) {
842 		fp = fget_locked(fdp, fd);
843 		if (fp == NULL)
844 			continue;
845 		if (fp->f_type == DTYPE_VNODE) {
846 			vp = fp->f_vnode;
847 			if (vp->v_type == VDIR)
848 				return (EPERM);
849 		}
850 	}
851 	return (0);
852 }
853 
854 /*
855  * This sysctl determines if we will allow a process to chroot(2) if it
856  * has a directory open:
857  *	0: disallowed for all processes.
858  *	1: allowed for processes that were not already chroot(2)'ed.
859  *	2: allowed for all processes.
860  */
861 
862 static int chroot_allow_open_directories = 1;
863 
864 SYSCTL_INT(_kern, OID_AUTO, chroot_allow_open_directories, CTLFLAG_RW,
865      &chroot_allow_open_directories, 0, "");
866 
867 /*
868  * Change notion of root (``/'') directory.
869  */
870 #ifndef _SYS_SYSPROTO_H_
871 struct chroot_args {
872 	char	*path;
873 };
874 #endif
875 int
876 chroot(td, uap)
877 	struct thread *td;
878 	struct chroot_args /* {
879 		char *path;
880 	} */ *uap;
881 {
882 	int error;
883 	struct nameidata nd;
884 	int vfslocked;
885 
886 	error = priv_check(td, PRIV_VFS_CHROOT);
887 	if (error)
888 		return (error);
889 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | MPSAFE | AUDITVNODE1,
890 	    UIO_USERSPACE, uap->path, td);
891 	error = namei(&nd);
892 	if (error)
893 		goto error;
894 	vfslocked = NDHASGIANT(&nd);
895 	if ((error = change_dir(nd.ni_vp, td)) != 0)
896 		goto e_vunlock;
897 #ifdef MAC
898 	if ((error = mac_vnode_check_chroot(td->td_ucred, nd.ni_vp)))
899 		goto e_vunlock;
900 #endif
901 	VOP_UNLOCK(nd.ni_vp, 0);
902 	error = change_root(nd.ni_vp, td);
903 	vrele(nd.ni_vp);
904 	VFS_UNLOCK_GIANT(vfslocked);
905 	NDFREE(&nd, NDF_ONLY_PNBUF);
906 	return (error);
907 e_vunlock:
908 	vput(nd.ni_vp);
909 	VFS_UNLOCK_GIANT(vfslocked);
910 error:
911 	NDFREE(&nd, NDF_ONLY_PNBUF);
912 	return (error);
913 }
914 
915 /*
916  * Common routine for chroot and chdir.  Callers must provide a locked vnode
917  * instance.
918  */
919 int
920 change_dir(vp, td)
921 	struct vnode *vp;
922 	struct thread *td;
923 {
924 	int error;
925 
926 	ASSERT_VOP_LOCKED(vp, "change_dir(): vp not locked");
927 	if (vp->v_type != VDIR)
928 		return (ENOTDIR);
929 #ifdef MAC
930 	error = mac_vnode_check_chdir(td->td_ucred, vp);
931 	if (error)
932 		return (error);
933 #endif
934 	error = VOP_ACCESS(vp, VEXEC, td->td_ucred, td);
935 	return (error);
936 }
937 
938 /*
939  * Common routine for kern_chroot() and jail_attach().  The caller is
940  * responsible for invoking priv_check() and mac_vnode_check_chroot() to
941  * authorize this operation.
942  */
943 int
944 change_root(vp, td)
945 	struct vnode *vp;
946 	struct thread *td;
947 {
948 	struct filedesc *fdp;
949 	struct vnode *oldvp;
950 	int vfslocked;
951 	int error;
952 
953 	VFS_ASSERT_GIANT(vp->v_mount);
954 	fdp = td->td_proc->p_fd;
955 	FILEDESC_XLOCK(fdp);
956 	if (chroot_allow_open_directories == 0 ||
957 	    (chroot_allow_open_directories == 1 && fdp->fd_rdir != rootvnode)) {
958 		error = chroot_refuse_vdir_fds(fdp);
959 		if (error) {
960 			FILEDESC_XUNLOCK(fdp);
961 			return (error);
962 		}
963 	}
964 	oldvp = fdp->fd_rdir;
965 	fdp->fd_rdir = vp;
966 	VREF(fdp->fd_rdir);
967 	if (!fdp->fd_jdir) {
968 		fdp->fd_jdir = vp;
969 		VREF(fdp->fd_jdir);
970 	}
971 	FILEDESC_XUNLOCK(fdp);
972 	vfslocked = VFS_LOCK_GIANT(oldvp->v_mount);
973 	vrele(oldvp);
974 	VFS_UNLOCK_GIANT(vfslocked);
975 	return (0);
976 }
977 
978 /*
979  * Check permissions, allocate an open file structure, and call the device
980  * open routine if any.
981  */
982 #ifndef _SYS_SYSPROTO_H_
983 struct open_args {
984 	char	*path;
985 	int	flags;
986 	int	mode;
987 };
988 #endif
989 int
990 open(td, uap)
991 	struct thread *td;
992 	register struct open_args /* {
993 		char *path;
994 		int flags;
995 		int mode;
996 	} */ *uap;
997 {
998 
999 	return (kern_open(td, uap->path, UIO_USERSPACE, uap->flags, uap->mode));
1000 }
1001 
1002 #ifndef _SYS_SYSPROTO_H_
1003 struct openat_args {
1004 	int	fd;
1005 	char	*path;
1006 	int	flag;
1007 	int	mode;
1008 };
1009 #endif
1010 int
1011 openat(struct thread *td, struct openat_args *uap)
1012 {
1013 
1014 	return (kern_openat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
1015 	    uap->mode));
1016 }
1017 
1018 int
1019 kern_open(struct thread *td, char *path, enum uio_seg pathseg, int flags,
1020     int mode)
1021 {
1022 
1023 	return (kern_openat(td, AT_FDCWD, path, pathseg, flags, mode));
1024 }
1025 
1026 int
1027 kern_openat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1028     int flags, int mode)
1029 {
1030 	struct proc *p = td->td_proc;
1031 	struct filedesc *fdp = p->p_fd;
1032 	struct file *fp;
1033 	struct vnode *vp;
1034 	struct vattr vat;
1035 	struct mount *mp;
1036 	int cmode;
1037 	struct file *nfp;
1038 	int type, indx, error;
1039 	struct flock lf;
1040 	struct nameidata nd;
1041 	int vfslocked;
1042 
1043 	AUDIT_ARG(fflags, flags);
1044 	AUDIT_ARG(mode, mode);
1045 	/* XXX: audit dirfd */
1046 	/*
1047 	 * Only one of the O_EXEC, O_RDONLY, O_WRONLY and O_RDWR may
1048 	 * be specified.
1049 	 */
1050 	if (flags & O_EXEC) {
1051 		if (flags & O_ACCMODE)
1052 			return (EINVAL);
1053 	} else if ((flags & O_ACCMODE) == O_ACCMODE)
1054 		return (EINVAL);
1055 	else
1056 		flags = FFLAGS(flags);
1057 
1058 	error = falloc(td, &nfp, &indx);
1059 	if (error)
1060 		return (error);
1061 	/* An extra reference on `nfp' has been held for us by falloc(). */
1062 	fp = nfp;
1063 	/* Set the flags early so the finit in devfs can pick them up. */
1064 	fp->f_flag = flags & FMASK;
1065 	cmode = ((mode &~ fdp->fd_cmask) & ALLPERMS) &~ S_ISTXT;
1066 	NDINIT_AT(&nd, LOOKUP, FOLLOW | AUDITVNODE1 | MPSAFE, pathseg, path, fd,
1067 	    td);
1068 	td->td_dupfd = -1;		/* XXX check for fdopen */
1069 	error = vn_open(&nd, &flags, cmode, fp);
1070 	if (error) {
1071 		/*
1072 		 * If the vn_open replaced the method vector, something
1073 		 * wonderous happened deep below and we just pass it up
1074 		 * pretending we know what we do.
1075 		 */
1076 		if (error == ENXIO && fp->f_ops != &badfileops) {
1077 			fdrop(fp, td);
1078 			td->td_retval[0] = indx;
1079 			return (0);
1080 		}
1081 
1082 		/*
1083 		 * handle special fdopen() case.  bleh.  dupfdopen() is
1084 		 * responsible for dropping the old contents of ofiles[indx]
1085 		 * if it succeeds.
1086 		 */
1087 		if ((error == ENODEV || error == ENXIO) &&
1088 		    td->td_dupfd >= 0 &&		/* XXX from fdopen */
1089 		    (error =
1090 			dupfdopen(td, fdp, indx, td->td_dupfd, flags, error)) == 0) {
1091 			td->td_retval[0] = indx;
1092 			fdrop(fp, td);
1093 			return (0);
1094 		}
1095 		/*
1096 		 * Clean up the descriptor, but only if another thread hadn't
1097 		 * replaced or closed it.
1098 		 */
1099 		fdclose(fdp, fp, indx, td);
1100 		fdrop(fp, td);
1101 
1102 		if (error == ERESTART)
1103 			error = EINTR;
1104 		return (error);
1105 	}
1106 	td->td_dupfd = 0;
1107 	vfslocked = NDHASGIANT(&nd);
1108 	NDFREE(&nd, NDF_ONLY_PNBUF);
1109 	vp = nd.ni_vp;
1110 
1111 	fp->f_vnode = vp;	/* XXX Does devfs need this? */
1112 	/*
1113 	 * If the file wasn't claimed by devfs bind it to the normal
1114 	 * vnode operations here.
1115 	 */
1116 	if (fp->f_ops == &badfileops) {
1117 		KASSERT(vp->v_type != VFIFO, ("Unexpected fifo."));
1118 		fp->f_seqcount = 1;
1119 		finit(fp, flags & FMASK, DTYPE_VNODE, vp, &vnops);
1120 	}
1121 
1122 	VOP_UNLOCK(vp, 0);
1123 	if (flags & (O_EXLOCK | O_SHLOCK)) {
1124 		lf.l_whence = SEEK_SET;
1125 		lf.l_start = 0;
1126 		lf.l_len = 0;
1127 		if (flags & O_EXLOCK)
1128 			lf.l_type = F_WRLCK;
1129 		else
1130 			lf.l_type = F_RDLCK;
1131 		type = F_FLOCK;
1132 		if ((flags & FNONBLOCK) == 0)
1133 			type |= F_WAIT;
1134 		if ((error = VOP_ADVLOCK(vp, (caddr_t)fp, F_SETLK, &lf,
1135 			    type)) != 0)
1136 			goto bad;
1137 		atomic_set_int(&fp->f_flag, FHASLOCK);
1138 	}
1139 	if (flags & O_TRUNC) {
1140 		if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
1141 			goto bad;
1142 		VOP_LEASE(vp, td, td->td_ucred, LEASE_WRITE);
1143 		VATTR_NULL(&vat);
1144 		vat.va_size = 0;
1145 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1146 #ifdef MAC
1147 		error = mac_vnode_check_write(td->td_ucred, fp->f_cred, vp);
1148 		if (error == 0)
1149 #endif
1150 			error = VOP_SETATTR(vp, &vat, td->td_ucred, td);
1151 		VOP_UNLOCK(vp, 0);
1152 		vn_finished_write(mp);
1153 		if (error)
1154 			goto bad;
1155 	}
1156 	VFS_UNLOCK_GIANT(vfslocked);
1157 	/*
1158 	 * Release our private reference, leaving the one associated with
1159 	 * the descriptor table intact.
1160 	 */
1161 	fdrop(fp, td);
1162 	td->td_retval[0] = indx;
1163 	return (0);
1164 bad:
1165 	VFS_UNLOCK_GIANT(vfslocked);
1166 	fdclose(fdp, fp, indx, td);
1167 	fdrop(fp, td);
1168 	return (error);
1169 }
1170 
1171 #ifdef COMPAT_43
1172 /*
1173  * Create a file.
1174  */
1175 #ifndef _SYS_SYSPROTO_H_
1176 struct ocreat_args {
1177 	char	*path;
1178 	int	mode;
1179 };
1180 #endif
1181 int
1182 ocreat(td, uap)
1183 	struct thread *td;
1184 	register struct ocreat_args /* {
1185 		char *path;
1186 		int mode;
1187 	} */ *uap;
1188 {
1189 
1190 	return (kern_open(td, uap->path, UIO_USERSPACE,
1191 	    O_WRONLY | O_CREAT | O_TRUNC, uap->mode));
1192 }
1193 #endif /* COMPAT_43 */
1194 
1195 /*
1196  * Create a special file.
1197  */
1198 #ifndef _SYS_SYSPROTO_H_
1199 struct mknod_args {
1200 	char	*path;
1201 	int	mode;
1202 	int	dev;
1203 };
1204 #endif
1205 int
1206 mknod(td, uap)
1207 	struct thread *td;
1208 	register struct mknod_args /* {
1209 		char *path;
1210 		int mode;
1211 		int dev;
1212 	} */ *uap;
1213 {
1214 
1215 	return (kern_mknod(td, uap->path, UIO_USERSPACE, uap->mode, uap->dev));
1216 }
1217 
1218 #ifndef _SYS_SYSPROTO_H_
1219 struct mknodat_args {
1220 	int	fd;
1221 	char	*path;
1222 	mode_t	mode;
1223 	dev_t	dev;
1224 };
1225 #endif
1226 int
1227 mknodat(struct thread *td, struct mknodat_args *uap)
1228 {
1229 
1230 	return (kern_mknodat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode,
1231 	    uap->dev));
1232 }
1233 
1234 int
1235 kern_mknod(struct thread *td, char *path, enum uio_seg pathseg, int mode,
1236     int dev)
1237 {
1238 
1239 	return (kern_mknodat(td, AT_FDCWD, path, pathseg, mode, dev));
1240 }
1241 
1242 int
1243 kern_mknodat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1244     int mode, int dev)
1245 {
1246 	struct vnode *vp;
1247 	struct mount *mp;
1248 	struct vattr vattr;
1249 	int error;
1250 	int whiteout = 0;
1251 	struct nameidata nd;
1252 	int vfslocked;
1253 
1254 	AUDIT_ARG(mode, mode);
1255 	AUDIT_ARG(dev, dev);
1256 	switch (mode & S_IFMT) {
1257 	case S_IFCHR:
1258 	case S_IFBLK:
1259 		error = priv_check(td, PRIV_VFS_MKNOD_DEV);
1260 		break;
1261 	case S_IFMT:
1262 		error = priv_check(td, PRIV_VFS_MKNOD_BAD);
1263 		break;
1264 	case S_IFWHT:
1265 		error = priv_check(td, PRIV_VFS_MKNOD_WHT);
1266 		break;
1267 	default:
1268 		error = EINVAL;
1269 		break;
1270 	}
1271 	if (error)
1272 		return (error);
1273 restart:
1274 	bwillwrite();
1275 	NDINIT_AT(&nd, CREATE, LOCKPARENT | SAVENAME | MPSAFE | AUDITVNODE1,
1276 	    pathseg, path, fd, td);
1277 	if ((error = namei(&nd)) != 0)
1278 		return (error);
1279 	vfslocked = NDHASGIANT(&nd);
1280 	vp = nd.ni_vp;
1281 	if (vp != NULL) {
1282 		NDFREE(&nd, NDF_ONLY_PNBUF);
1283 		if (vp == nd.ni_dvp)
1284 			vrele(nd.ni_dvp);
1285 		else
1286 			vput(nd.ni_dvp);
1287 		vrele(vp);
1288 		VFS_UNLOCK_GIANT(vfslocked);
1289 		return (EEXIST);
1290 	} else {
1291 		VATTR_NULL(&vattr);
1292 		FILEDESC_SLOCK(td->td_proc->p_fd);
1293 		vattr.va_mode = (mode & ALLPERMS) &
1294 		    ~td->td_proc->p_fd->fd_cmask;
1295 		FILEDESC_SUNLOCK(td->td_proc->p_fd);
1296 		vattr.va_rdev = dev;
1297 		whiteout = 0;
1298 
1299 		switch (mode & S_IFMT) {
1300 		case S_IFMT:	/* used by badsect to flag bad sectors */
1301 			vattr.va_type = VBAD;
1302 			break;
1303 		case S_IFCHR:
1304 			vattr.va_type = VCHR;
1305 			break;
1306 		case S_IFBLK:
1307 			vattr.va_type = VBLK;
1308 			break;
1309 		case S_IFWHT:
1310 			whiteout = 1;
1311 			break;
1312 		default:
1313 			panic("kern_mknod: invalid mode");
1314 		}
1315 	}
1316 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1317 		NDFREE(&nd, NDF_ONLY_PNBUF);
1318 		vput(nd.ni_dvp);
1319 		VFS_UNLOCK_GIANT(vfslocked);
1320 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1321 			return (error);
1322 		goto restart;
1323 	}
1324 #ifdef MAC
1325 	if (error == 0 && !whiteout)
1326 		error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp,
1327 		    &nd.ni_cnd, &vattr);
1328 #endif
1329 	if (!error) {
1330 		VOP_LEASE(nd.ni_dvp, td, td->td_ucred, LEASE_WRITE);
1331 		if (whiteout)
1332 			error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE);
1333 		else {
1334 			error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp,
1335 						&nd.ni_cnd, &vattr);
1336 			if (error == 0)
1337 				vput(nd.ni_vp);
1338 		}
1339 	}
1340 	NDFREE(&nd, NDF_ONLY_PNBUF);
1341 	vput(nd.ni_dvp);
1342 	vn_finished_write(mp);
1343 	VFS_UNLOCK_GIANT(vfslocked);
1344 	return (error);
1345 }
1346 
1347 /*
1348  * Create a named pipe.
1349  */
1350 #ifndef _SYS_SYSPROTO_H_
1351 struct mkfifo_args {
1352 	char	*path;
1353 	int	mode;
1354 };
1355 #endif
1356 int
1357 mkfifo(td, uap)
1358 	struct thread *td;
1359 	register struct mkfifo_args /* {
1360 		char *path;
1361 		int mode;
1362 	} */ *uap;
1363 {
1364 
1365 	return (kern_mkfifo(td, uap->path, UIO_USERSPACE, uap->mode));
1366 }
1367 
1368 #ifndef _SYS_SYSPROTO_H_
1369 struct mkfifoat_args {
1370 	int	fd;
1371 	char	*path;
1372 	mode_t	mode;
1373 };
1374 #endif
1375 int
1376 mkfifoat(struct thread *td, struct mkfifoat_args *uap)
1377 {
1378 
1379 	return (kern_mkfifoat(td, uap->fd, uap->path, UIO_USERSPACE,
1380 	    uap->mode));
1381 }
1382 
1383 int
1384 kern_mkfifo(struct thread *td, char *path, enum uio_seg pathseg, int mode)
1385 {
1386 
1387 	return (kern_mkfifoat(td, AT_FDCWD, path, pathseg, mode));
1388 }
1389 
1390 int
1391 kern_mkfifoat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1392     int mode)
1393 {
1394 	struct mount *mp;
1395 	struct vattr vattr;
1396 	int error;
1397 	struct nameidata nd;
1398 	int vfslocked;
1399 
1400 	AUDIT_ARG(mode, mode);
1401 restart:
1402 	bwillwrite();
1403 	NDINIT_AT(&nd, CREATE, LOCKPARENT | SAVENAME | MPSAFE | AUDITVNODE1,
1404 	    pathseg, path, fd, td);
1405 	if ((error = namei(&nd)) != 0)
1406 		return (error);
1407 	vfslocked = NDHASGIANT(&nd);
1408 	if (nd.ni_vp != NULL) {
1409 		NDFREE(&nd, NDF_ONLY_PNBUF);
1410 		if (nd.ni_vp == nd.ni_dvp)
1411 			vrele(nd.ni_dvp);
1412 		else
1413 			vput(nd.ni_dvp);
1414 		vrele(nd.ni_vp);
1415 		VFS_UNLOCK_GIANT(vfslocked);
1416 		return (EEXIST);
1417 	}
1418 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1419 		NDFREE(&nd, NDF_ONLY_PNBUF);
1420 		vput(nd.ni_dvp);
1421 		VFS_UNLOCK_GIANT(vfslocked);
1422 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1423 			return (error);
1424 		goto restart;
1425 	}
1426 	VATTR_NULL(&vattr);
1427 	vattr.va_type = VFIFO;
1428 	FILEDESC_SLOCK(td->td_proc->p_fd);
1429 	vattr.va_mode = (mode & ALLPERMS) & ~td->td_proc->p_fd->fd_cmask;
1430 	FILEDESC_SUNLOCK(td->td_proc->p_fd);
1431 #ifdef MAC
1432 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1433 	    &vattr);
1434 	if (error)
1435 		goto out;
1436 #endif
1437 	VOP_LEASE(nd.ni_dvp, td, td->td_ucred, LEASE_WRITE);
1438 	error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
1439 	if (error == 0)
1440 		vput(nd.ni_vp);
1441 #ifdef MAC
1442 out:
1443 #endif
1444 	vput(nd.ni_dvp);
1445 	vn_finished_write(mp);
1446 	VFS_UNLOCK_GIANT(vfslocked);
1447 	NDFREE(&nd, NDF_ONLY_PNBUF);
1448 	return (error);
1449 }
1450 
1451 /*
1452  * Make a hard file link.
1453  */
1454 #ifndef _SYS_SYSPROTO_H_
1455 struct link_args {
1456 	char	*path;
1457 	char	*link;
1458 };
1459 #endif
1460 int
1461 link(td, uap)
1462 	struct thread *td;
1463 	register struct link_args /* {
1464 		char *path;
1465 		char *link;
1466 	} */ *uap;
1467 {
1468 
1469 	return (kern_link(td, uap->path, uap->link, UIO_USERSPACE));
1470 }
1471 
1472 #ifndef _SYS_SYSPROTO_H_
1473 struct linkat_args {
1474 	int	fd1;
1475 	char	*path1;
1476 	int	fd2;
1477 	char	*path2;
1478 	int	flag;
1479 };
1480 #endif
1481 int
1482 linkat(struct thread *td, struct linkat_args *uap)
1483 {
1484 	int flag;
1485 
1486 	flag = uap->flag;
1487 	if (flag & ~AT_SYMLINK_FOLLOW)
1488 		return (EINVAL);
1489 
1490 	return (kern_linkat(td, uap->fd1, uap->fd2, uap->path1, uap->path2,
1491 	    UIO_USERSPACE, (flag & AT_SYMLINK_FOLLOW) ? FOLLOW : NOFOLLOW));
1492 }
1493 
1494 static int hardlink_check_uid = 0;
1495 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_uid, CTLFLAG_RW,
1496     &hardlink_check_uid, 0,
1497     "Unprivileged processes cannot create hard links to files owned by other "
1498     "users");
1499 static int hardlink_check_gid = 0;
1500 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_gid, CTLFLAG_RW,
1501     &hardlink_check_gid, 0,
1502     "Unprivileged processes cannot create hard links to files owned by other "
1503     "groups");
1504 
1505 static int
1506 can_hardlink(struct vnode *vp, struct thread *td, struct ucred *cred)
1507 {
1508 	struct vattr va;
1509 	int error;
1510 
1511 	if (!hardlink_check_uid && !hardlink_check_gid)
1512 		return (0);
1513 
1514 	error = VOP_GETATTR(vp, &va, cred, td);
1515 	if (error != 0)
1516 		return (error);
1517 
1518 	if (hardlink_check_uid && cred->cr_uid != va.va_uid) {
1519 		error = priv_check_cred(cred, PRIV_VFS_LINK, 0);
1520 		if (error)
1521 			return (error);
1522 	}
1523 
1524 	if (hardlink_check_gid && !groupmember(va.va_gid, cred)) {
1525 		error = priv_check_cred(cred, PRIV_VFS_LINK, 0);
1526 		if (error)
1527 			return (error);
1528 	}
1529 
1530 	return (0);
1531 }
1532 
1533 int
1534 kern_link(struct thread *td, char *path, char *link, enum uio_seg segflg)
1535 {
1536 
1537 	return (kern_linkat(td, AT_FDCWD, AT_FDCWD, path,link, segflg, FOLLOW));
1538 }
1539 
1540 int
1541 kern_linkat(struct thread *td, int fd1, int fd2, char *path1, char *path2,
1542     enum uio_seg segflg, int follow)
1543 {
1544 	struct vnode *vp;
1545 	struct mount *mp;
1546 	struct nameidata nd;
1547 	int vfslocked;
1548 	int lvfslocked;
1549 	int error;
1550 
1551 	bwillwrite();
1552 	NDINIT_AT(&nd, LOOKUP, follow | MPSAFE | AUDITVNODE1, segflg, path1,
1553 	    fd1, td);
1554 
1555 	if ((error = namei(&nd)) != 0)
1556 		return (error);
1557 	vfslocked = NDHASGIANT(&nd);
1558 	NDFREE(&nd, NDF_ONLY_PNBUF);
1559 	vp = nd.ni_vp;
1560 	if (vp->v_type == VDIR) {
1561 		vrele(vp);
1562 		VFS_UNLOCK_GIANT(vfslocked);
1563 		return (EPERM);		/* POSIX */
1564 	}
1565 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) {
1566 		vrele(vp);
1567 		VFS_UNLOCK_GIANT(vfslocked);
1568 		return (error);
1569 	}
1570 	NDINIT_AT(&nd, CREATE, LOCKPARENT | SAVENAME | MPSAFE | AUDITVNODE1,
1571 	    segflg, path2, fd2, td);
1572 	if ((error = namei(&nd)) == 0) {
1573 		lvfslocked = NDHASGIANT(&nd);
1574 		if (nd.ni_vp != NULL) {
1575 			if (nd.ni_dvp == nd.ni_vp)
1576 				vrele(nd.ni_dvp);
1577 			else
1578 				vput(nd.ni_dvp);
1579 			vrele(nd.ni_vp);
1580 			error = EEXIST;
1581 		} else if ((error = vn_lock(vp, LK_EXCLUSIVE | LK_RETRY))
1582 		    == 0) {
1583 			VOP_LEASE(nd.ni_dvp, td, td->td_ucred, LEASE_WRITE);
1584 			VOP_LEASE(vp, td, td->td_ucred, LEASE_WRITE);
1585 			error = can_hardlink(vp, td, td->td_ucred);
1586 			if (error == 0)
1587 #ifdef MAC
1588 				error = mac_vnode_check_link(td->td_ucred,
1589 				    nd.ni_dvp, vp, &nd.ni_cnd);
1590 			if (error == 0)
1591 #endif
1592 				error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
1593 			VOP_UNLOCK(vp, 0);
1594 			vput(nd.ni_dvp);
1595 		}
1596 		NDFREE(&nd, NDF_ONLY_PNBUF);
1597 		VFS_UNLOCK_GIANT(lvfslocked);
1598 	}
1599 	vrele(vp);
1600 	vn_finished_write(mp);
1601 	VFS_UNLOCK_GIANT(vfslocked);
1602 	return (error);
1603 }
1604 
1605 /*
1606  * Make a symbolic link.
1607  */
1608 #ifndef _SYS_SYSPROTO_H_
1609 struct symlink_args {
1610 	char	*path;
1611 	char	*link;
1612 };
1613 #endif
1614 int
1615 symlink(td, uap)
1616 	struct thread *td;
1617 	register struct symlink_args /* {
1618 		char *path;
1619 		char *link;
1620 	} */ *uap;
1621 {
1622 
1623 	return (kern_symlink(td, uap->path, uap->link, UIO_USERSPACE));
1624 }
1625 
1626 #ifndef _SYS_SYSPROTO_H_
1627 struct symlinkat_args {
1628 	char	*path;
1629 	int	fd;
1630 	char	*path2;
1631 };
1632 #endif
1633 int
1634 symlinkat(struct thread *td, struct symlinkat_args *uap)
1635 {
1636 
1637 	return (kern_symlinkat(td, uap->path1, uap->fd, uap->path2,
1638 	    UIO_USERSPACE));
1639 }
1640 
1641 int
1642 kern_symlink(struct thread *td, char *path, char *link, enum uio_seg segflg)
1643 {
1644 
1645 	return (kern_symlinkat(td, path, AT_FDCWD, link, segflg));
1646 }
1647 
1648 int
1649 kern_symlinkat(struct thread *td, char *path1, int fd, char *path2,
1650     enum uio_seg segflg)
1651 {
1652 	struct mount *mp;
1653 	struct vattr vattr;
1654 	char *syspath;
1655 	int error;
1656 	struct nameidata nd;
1657 	int vfslocked;
1658 
1659 	if (segflg == UIO_SYSSPACE) {
1660 		syspath = path1;
1661 	} else {
1662 		syspath = uma_zalloc(namei_zone, M_WAITOK);
1663 		if ((error = copyinstr(path1, syspath, MAXPATHLEN, NULL)) != 0)
1664 			goto out;
1665 	}
1666 	AUDIT_ARG(text, syspath);
1667 restart:
1668 	bwillwrite();
1669 	NDINIT_AT(&nd, CREATE, LOCKPARENT | SAVENAME | MPSAFE | AUDITVNODE1,
1670 	    segflg, path2, fd, td);
1671 	if ((error = namei(&nd)) != 0)
1672 		goto out;
1673 	vfslocked = NDHASGIANT(&nd);
1674 	if (nd.ni_vp) {
1675 		NDFREE(&nd, NDF_ONLY_PNBUF);
1676 		if (nd.ni_vp == nd.ni_dvp)
1677 			vrele(nd.ni_dvp);
1678 		else
1679 			vput(nd.ni_dvp);
1680 		vrele(nd.ni_vp);
1681 		VFS_UNLOCK_GIANT(vfslocked);
1682 		error = EEXIST;
1683 		goto out;
1684 	}
1685 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1686 		NDFREE(&nd, NDF_ONLY_PNBUF);
1687 		vput(nd.ni_dvp);
1688 		VFS_UNLOCK_GIANT(vfslocked);
1689 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1690 			goto out;
1691 		goto restart;
1692 	}
1693 	VATTR_NULL(&vattr);
1694 	FILEDESC_SLOCK(td->td_proc->p_fd);
1695 	vattr.va_mode = ACCESSPERMS &~ td->td_proc->p_fd->fd_cmask;
1696 	FILEDESC_SUNLOCK(td->td_proc->p_fd);
1697 #ifdef MAC
1698 	vattr.va_type = VLNK;
1699 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1700 	    &vattr);
1701 	if (error)
1702 		goto out2;
1703 #endif
1704 	VOP_LEASE(nd.ni_dvp, td, td->td_ucred, LEASE_WRITE);
1705 	error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, syspath);
1706 	if (error == 0)
1707 		vput(nd.ni_vp);
1708 #ifdef MAC
1709 out2:
1710 #endif
1711 	NDFREE(&nd, NDF_ONLY_PNBUF);
1712 	vput(nd.ni_dvp);
1713 	vn_finished_write(mp);
1714 	VFS_UNLOCK_GIANT(vfslocked);
1715 out:
1716 	if (segflg != UIO_SYSSPACE)
1717 		uma_zfree(namei_zone, syspath);
1718 	return (error);
1719 }
1720 
1721 /*
1722  * Delete a whiteout from the filesystem.
1723  */
1724 int
1725 undelete(td, uap)
1726 	struct thread *td;
1727 	register struct undelete_args /* {
1728 		char *path;
1729 	} */ *uap;
1730 {
1731 	int error;
1732 	struct mount *mp;
1733 	struct nameidata nd;
1734 	int vfslocked;
1735 
1736 restart:
1737 	bwillwrite();
1738 	NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | MPSAFE | AUDITVNODE1,
1739 	    UIO_USERSPACE, uap->path, td);
1740 	error = namei(&nd);
1741 	if (error)
1742 		return (error);
1743 	vfslocked = NDHASGIANT(&nd);
1744 
1745 	if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) {
1746 		NDFREE(&nd, NDF_ONLY_PNBUF);
1747 		if (nd.ni_vp == nd.ni_dvp)
1748 			vrele(nd.ni_dvp);
1749 		else
1750 			vput(nd.ni_dvp);
1751 		if (nd.ni_vp)
1752 			vrele(nd.ni_vp);
1753 		VFS_UNLOCK_GIANT(vfslocked);
1754 		return (EEXIST);
1755 	}
1756 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1757 		NDFREE(&nd, NDF_ONLY_PNBUF);
1758 		vput(nd.ni_dvp);
1759 		VFS_UNLOCK_GIANT(vfslocked);
1760 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1761 			return (error);
1762 		goto restart;
1763 	}
1764 	VOP_LEASE(nd.ni_dvp, td, td->td_ucred, LEASE_WRITE);
1765 	error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE);
1766 	NDFREE(&nd, NDF_ONLY_PNBUF);
1767 	vput(nd.ni_dvp);
1768 	vn_finished_write(mp);
1769 	VFS_UNLOCK_GIANT(vfslocked);
1770 	return (error);
1771 }
1772 
1773 /*
1774  * Delete a name from the filesystem.
1775  */
1776 #ifndef _SYS_SYSPROTO_H_
1777 struct unlink_args {
1778 	char	*path;
1779 };
1780 #endif
1781 int
1782 unlink(td, uap)
1783 	struct thread *td;
1784 	struct unlink_args /* {
1785 		char *path;
1786 	} */ *uap;
1787 {
1788 
1789 	return (kern_unlink(td, uap->path, UIO_USERSPACE));
1790 }
1791 
1792 #ifndef _SYS_SYSPROTO_H_
1793 struct unlinkat_args {
1794 	int	fd;
1795 	char	*path;
1796 	int	flag;
1797 };
1798 #endif
1799 int
1800 unlinkat(struct thread *td, struct unlinkat_args *uap)
1801 {
1802 	int flag = uap->flag;
1803 	int fd = uap->fd;
1804 	char *path = uap->path;
1805 
1806 	if (flag & ~AT_REMOVEDIR)
1807 		return (EINVAL);
1808 
1809 	if (flag & AT_REMOVEDIR)
1810 		return (kern_rmdirat(td, fd, path, UIO_USERSPACE));
1811 	else
1812 		return (kern_unlinkat(td, fd, path, UIO_USERSPACE));
1813 }
1814 
1815 int
1816 kern_unlink(struct thread *td, char *path, enum uio_seg pathseg)
1817 {
1818 
1819 	return (kern_unlinkat(td, AT_FDCWD, path, pathseg));
1820 }
1821 
1822 int
1823 kern_unlinkat(struct thread *td, int fd, char *path, enum uio_seg pathseg)
1824 {
1825 	struct mount *mp;
1826 	struct vnode *vp;
1827 	int error;
1828 	struct nameidata nd;
1829 	int vfslocked;
1830 
1831 restart:
1832 	bwillwrite();
1833 	NDINIT_AT(&nd, DELETE, LOCKPARENT | LOCKLEAF | MPSAFE | AUDITVNODE1,
1834 	    pathseg, path, fd, td);
1835 	if ((error = namei(&nd)) != 0)
1836 		return (error == EINVAL ? EPERM : error);
1837 	vfslocked = NDHASGIANT(&nd);
1838 	vp = nd.ni_vp;
1839 	if (vp->v_type == VDIR)
1840 		error = EPERM;		/* POSIX */
1841 	else {
1842 		/*
1843 		 * The root of a mounted filesystem cannot be deleted.
1844 		 *
1845 		 * XXX: can this only be a VDIR case?
1846 		 */
1847 		if (vp->v_vflag & VV_ROOT)
1848 			error = EBUSY;
1849 	}
1850 	if (error == 0) {
1851 		if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1852 			NDFREE(&nd, NDF_ONLY_PNBUF);
1853 			vput(nd.ni_dvp);
1854 			if (vp == nd.ni_dvp)
1855 				vrele(vp);
1856 			else
1857 				vput(vp);
1858 			VFS_UNLOCK_GIANT(vfslocked);
1859 			if ((error = vn_start_write(NULL, &mp,
1860 			    V_XSLEEP | PCATCH)) != 0)
1861 				return (error);
1862 			goto restart;
1863 		}
1864 #ifdef MAC
1865 		error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
1866 		    &nd.ni_cnd);
1867 		if (error)
1868 			goto out;
1869 #endif
1870 		VOP_LEASE(nd.ni_dvp, td, td->td_ucred, LEASE_WRITE);
1871 		error = VOP_REMOVE(nd.ni_dvp, vp, &nd.ni_cnd);
1872 #ifdef MAC
1873 out:
1874 #endif
1875 		vn_finished_write(mp);
1876 	}
1877 	NDFREE(&nd, NDF_ONLY_PNBUF);
1878 	vput(nd.ni_dvp);
1879 	if (vp == nd.ni_dvp)
1880 		vrele(vp);
1881 	else
1882 		vput(vp);
1883 	VFS_UNLOCK_GIANT(vfslocked);
1884 	return (error);
1885 }
1886 
1887 /*
1888  * Reposition read/write file offset.
1889  */
1890 #ifndef _SYS_SYSPROTO_H_
1891 struct lseek_args {
1892 	int	fd;
1893 	int	pad;
1894 	off_t	offset;
1895 	int	whence;
1896 };
1897 #endif
1898 int
1899 lseek(td, uap)
1900 	struct thread *td;
1901 	register struct lseek_args /* {
1902 		int fd;
1903 		int pad;
1904 		off_t offset;
1905 		int whence;
1906 	} */ *uap;
1907 {
1908 	struct ucred *cred = td->td_ucred;
1909 	struct file *fp;
1910 	struct vnode *vp;
1911 	struct vattr vattr;
1912 	off_t offset;
1913 	int error, noneg;
1914 	int vfslocked;
1915 
1916 	if ((error = fget(td, uap->fd, &fp)) != 0)
1917 		return (error);
1918 	if (!(fp->f_ops->fo_flags & DFLAG_SEEKABLE)) {
1919 		fdrop(fp, td);
1920 		return (ESPIPE);
1921 	}
1922 	vp = fp->f_vnode;
1923 	vfslocked = VFS_LOCK_GIANT(vp->v_mount);
1924 	noneg = (vp->v_type != VCHR);
1925 	offset = uap->offset;
1926 	switch (uap->whence) {
1927 	case L_INCR:
1928 		if (noneg &&
1929 		    (fp->f_offset < 0 ||
1930 		    (offset > 0 && fp->f_offset > OFF_MAX - offset))) {
1931 			error = EOVERFLOW;
1932 			break;
1933 		}
1934 		offset += fp->f_offset;
1935 		break;
1936 	case L_XTND:
1937 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1938 		error = VOP_GETATTR(vp, &vattr, cred, td);
1939 		VOP_UNLOCK(vp, 0);
1940 		if (error)
1941 			break;
1942 		if (noneg &&
1943 		    (vattr.va_size > OFF_MAX ||
1944 		    (offset > 0 && vattr.va_size > OFF_MAX - offset))) {
1945 			error = EOVERFLOW;
1946 			break;
1947 		}
1948 		offset += vattr.va_size;
1949 		break;
1950 	case L_SET:
1951 		break;
1952 	case SEEK_DATA:
1953 		error = fo_ioctl(fp, FIOSEEKDATA, &offset, cred, td);
1954 		break;
1955 	case SEEK_HOLE:
1956 		error = fo_ioctl(fp, FIOSEEKHOLE, &offset, cred, td);
1957 		break;
1958 	default:
1959 		error = EINVAL;
1960 	}
1961 	if (error == 0 && noneg && offset < 0)
1962 		error = EINVAL;
1963 	if (error != 0)
1964 		goto drop;
1965 	fp->f_offset = offset;
1966 	*(off_t *)(td->td_retval) = fp->f_offset;
1967 drop:
1968 	fdrop(fp, td);
1969 	VFS_UNLOCK_GIANT(vfslocked);
1970 	return (error);
1971 }
1972 
1973 #if defined(COMPAT_43)
1974 /*
1975  * Reposition read/write file offset.
1976  */
1977 #ifndef _SYS_SYSPROTO_H_
1978 struct olseek_args {
1979 	int	fd;
1980 	long	offset;
1981 	int	whence;
1982 };
1983 #endif
1984 int
1985 olseek(td, uap)
1986 	struct thread *td;
1987 	register struct olseek_args /* {
1988 		int fd;
1989 		long offset;
1990 		int whence;
1991 	} */ *uap;
1992 {
1993 	struct lseek_args /* {
1994 		int fd;
1995 		int pad;
1996 		off_t offset;
1997 		int whence;
1998 	} */ nuap;
1999 
2000 	nuap.fd = uap->fd;
2001 	nuap.offset = uap->offset;
2002 	nuap.whence = uap->whence;
2003 	return (lseek(td, &nuap));
2004 }
2005 #endif /* COMPAT_43 */
2006 
2007 /* Version with the 'pad' argument */
2008 int
2009 freebsd6_lseek(td, uap)
2010 	struct thread *td;
2011 	register struct freebsd6_lseek_args *uap;
2012 {
2013 	struct lseek_args ouap;
2014 
2015 	ouap.fd = uap->fd;
2016 	ouap.offset = uap->offset;
2017 	ouap.whence = uap->whence;
2018 	return (lseek(td, &ouap));
2019 }
2020 
2021 /*
2022  * Check access permissions using passed credentials.
2023  */
2024 static int
2025 vn_access(vp, user_flags, cred, td)
2026 	struct vnode	*vp;
2027 	int		user_flags;
2028 	struct ucred	*cred;
2029 	struct thread	*td;
2030 {
2031 	int error, flags;
2032 
2033 	/* Flags == 0 means only check for existence. */
2034 	error = 0;
2035 	if (user_flags) {
2036 		flags = 0;
2037 		if (user_flags & R_OK)
2038 			flags |= VREAD;
2039 		if (user_flags & W_OK)
2040 			flags |= VWRITE;
2041 		if (user_flags & X_OK)
2042 			flags |= VEXEC;
2043 #ifdef MAC
2044 		error = mac_vnode_check_access(cred, vp, flags);
2045 		if (error)
2046 			return (error);
2047 #endif
2048 		if ((flags & VWRITE) == 0 || (error = vn_writechk(vp)) == 0)
2049 			error = VOP_ACCESS(vp, flags, cred, td);
2050 	}
2051 	return (error);
2052 }
2053 
2054 /*
2055  * Check access permissions using "real" credentials.
2056  */
2057 #ifndef _SYS_SYSPROTO_H_
2058 struct access_args {
2059 	char	*path;
2060 	int	flags;
2061 };
2062 #endif
2063 int
2064 access(td, uap)
2065 	struct thread *td;
2066 	register struct access_args /* {
2067 		char *path;
2068 		int flags;
2069 	} */ *uap;
2070 {
2071 
2072 	return (kern_access(td, uap->path, UIO_USERSPACE, uap->flags));
2073 }
2074 
2075 #ifndef _SYS_SYSPROTO_H_
2076 struct faccessat_args {
2077 	int	dirfd;
2078 	char	*path;
2079 	int	mode;
2080 	int	flag;
2081 }
2082 #endif
2083 int faccessat(struct thread *td, struct faccessat_args *uap)
2084 {
2085 
2086 	if (uap->flag & ~AT_EACCESS)
2087 		return (EINVAL);
2088 	return (kern_accessat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
2089 	    uap->mode));
2090 }
2091 
2092 int
2093 kern_access(struct thread *td, char *path, enum uio_seg pathseg, int mode)
2094 {
2095 
2096 	return (kern_accessat(td, AT_FDCWD, path, pathseg, 0, mode));
2097 }
2098 
2099 int
2100 kern_accessat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2101     int flags, int mode)
2102 {
2103 	struct ucred *cred, *tmpcred;
2104 	struct vnode *vp;
2105 	struct nameidata nd;
2106 	int vfslocked;
2107 	int error;
2108 
2109 	/*
2110 	 * Create and modify a temporary credential instead of one that
2111 	 * is potentially shared.  This could also mess up socket
2112 	 * buffer accounting which can run in an interrupt context.
2113 	 */
2114 	if (!(flags & AT_EACCESS)) {
2115 		cred = td->td_ucred;
2116 		tmpcred = crdup(cred);
2117 		tmpcred->cr_uid = cred->cr_ruid;
2118 		tmpcred->cr_groups[0] = cred->cr_rgid;
2119 		td->td_ucred = tmpcred;
2120 	} else
2121 		cred = tmpcred = td->td_ucred;
2122 	NDINIT_AT(&nd, LOOKUP, FOLLOW | LOCKLEAF | MPSAFE | AUDITVNODE1,
2123 	    pathseg, path, fd, td);
2124 	if ((error = namei(&nd)) != 0)
2125 		goto out1;
2126 	vfslocked = NDHASGIANT(&nd);
2127 	vp = nd.ni_vp;
2128 
2129 	error = vn_access(vp, mode, tmpcred, td);
2130 	NDFREE(&nd, NDF_ONLY_PNBUF);
2131 	vput(vp);
2132 	VFS_UNLOCK_GIANT(vfslocked);
2133 out1:
2134 	if (!(flags & AT_EACCESS)) {
2135 		td->td_ucred = cred;
2136 		crfree(tmpcred);
2137 	}
2138 	return (error);
2139 }
2140 
2141 /*
2142  * Check access permissions using "effective" credentials.
2143  */
2144 #ifndef _SYS_SYSPROTO_H_
2145 struct eaccess_args {
2146 	char	*path;
2147 	int	flags;
2148 };
2149 #endif
2150 int
2151 eaccess(td, uap)
2152 	struct thread *td;
2153 	register struct eaccess_args /* {
2154 		char *path;
2155 		int flags;
2156 	} */ *uap;
2157 {
2158 
2159 	return (kern_eaccess(td, uap->path, UIO_USERSPACE, uap->flags));
2160 }
2161 
2162 int
2163 kern_eaccess(struct thread *td, char *path, enum uio_seg pathseg, int flags)
2164 {
2165 
2166 	return (kern_accessat(td, AT_FDCWD, path, pathseg, AT_EACCESS, flags));
2167 }
2168 
2169 #if defined(COMPAT_43)
2170 /*
2171  * Get file status; this version follows links.
2172  */
2173 #ifndef _SYS_SYSPROTO_H_
2174 struct ostat_args {
2175 	char	*path;
2176 	struct ostat *ub;
2177 };
2178 #endif
2179 int
2180 ostat(td, uap)
2181 	struct thread *td;
2182 	register struct ostat_args /* {
2183 		char *path;
2184 		struct ostat *ub;
2185 	} */ *uap;
2186 {
2187 	struct stat sb;
2188 	struct ostat osb;
2189 	int error;
2190 
2191 	error = kern_stat(td, uap->path, UIO_USERSPACE, &sb);
2192 	if (error)
2193 		return (error);
2194 	cvtstat(&sb, &osb);
2195 	error = copyout(&osb, uap->ub, sizeof (osb));
2196 	return (error);
2197 }
2198 
2199 /*
2200  * Get file status; this version does not follow links.
2201  */
2202 #ifndef _SYS_SYSPROTO_H_
2203 struct olstat_args {
2204 	char	*path;
2205 	struct ostat *ub;
2206 };
2207 #endif
2208 int
2209 olstat(td, uap)
2210 	struct thread *td;
2211 	register struct olstat_args /* {
2212 		char *path;
2213 		struct ostat *ub;
2214 	} */ *uap;
2215 {
2216 	struct stat sb;
2217 	struct ostat osb;
2218 	int error;
2219 
2220 	error = kern_lstat(td, uap->path, UIO_USERSPACE, &sb);
2221 	if (error)
2222 		return (error);
2223 	cvtstat(&sb, &osb);
2224 	error = copyout(&osb, uap->ub, sizeof (osb));
2225 	return (error);
2226 }
2227 
2228 /*
2229  * Convert from an old to a new stat structure.
2230  */
2231 void
2232 cvtstat(st, ost)
2233 	struct stat *st;
2234 	struct ostat *ost;
2235 {
2236 
2237 	ost->st_dev = st->st_dev;
2238 	ost->st_ino = st->st_ino;
2239 	ost->st_mode = st->st_mode;
2240 	ost->st_nlink = st->st_nlink;
2241 	ost->st_uid = st->st_uid;
2242 	ost->st_gid = st->st_gid;
2243 	ost->st_rdev = st->st_rdev;
2244 	if (st->st_size < (quad_t)1 << 32)
2245 		ost->st_size = st->st_size;
2246 	else
2247 		ost->st_size = -2;
2248 	ost->st_atime = st->st_atime;
2249 	ost->st_mtime = st->st_mtime;
2250 	ost->st_ctime = st->st_ctime;
2251 	ost->st_blksize = st->st_blksize;
2252 	ost->st_blocks = st->st_blocks;
2253 	ost->st_flags = st->st_flags;
2254 	ost->st_gen = st->st_gen;
2255 }
2256 #endif /* COMPAT_43 */
2257 
2258 /*
2259  * Get file status; this version follows links.
2260  */
2261 #ifndef _SYS_SYSPROTO_H_
2262 struct stat_args {
2263 	char	*path;
2264 	struct stat *ub;
2265 };
2266 #endif
2267 int
2268 stat(td, uap)
2269 	struct thread *td;
2270 	register struct stat_args /* {
2271 		char *path;
2272 		struct stat *ub;
2273 	} */ *uap;
2274 {
2275 	struct stat sb;
2276 	int error;
2277 
2278 	error = kern_stat(td, uap->path, UIO_USERSPACE, &sb);
2279 	if (error == 0)
2280 		error = copyout(&sb, uap->ub, sizeof (sb));
2281 	return (error);
2282 }
2283 
2284 #ifndef _SYS_SYSPROTO_H_
2285 struct fstatat_args {
2286 	int	fd;
2287 	char	*path;
2288 	struct stat	*buf;
2289 	int	flag;
2290 }
2291 #endif
2292 int
2293 fstatat(struct thread *td, struct fstatat_args *uap)
2294 {
2295 	struct stat sb;
2296 	int error;
2297 
2298 	error = kern_statat(td, uap->flag, uap->fd, uap->path,
2299 	    UIO_USERSPACE, &sb);
2300 	if (error == 0)
2301 		error = copyout(&sb, uap->buf, sizeof (sb));
2302 	return (error);
2303 }
2304 
2305 int
2306 kern_stat(struct thread *td, char *path, enum uio_seg pathseg, struct stat *sbp)
2307 {
2308 
2309 	return (kern_statat(td, 0, AT_FDCWD, path, pathseg, sbp));
2310 }
2311 
2312 int
2313 kern_statat(struct thread *td, int flag, int fd, char *path,
2314     enum uio_seg pathseg, struct stat *sbp)
2315 {
2316 	struct nameidata nd;
2317 	struct stat sb;
2318 	int error, vfslocked;
2319 
2320 	if (flag & ~AT_SYMLINK_NOFOLLOW)
2321 		return (EINVAL);
2322 
2323 	NDINIT_AT(&nd, LOOKUP, ((flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW :
2324 	    FOLLOW) | LOCKSHARED | LOCKLEAF | AUDITVNODE1 | MPSAFE, pathseg,
2325 	    path, fd, td);
2326 
2327 	if ((error = namei(&nd)) != 0)
2328 		return (error);
2329 	vfslocked = NDHASGIANT(&nd);
2330 	error = vn_stat(nd.ni_vp, &sb, td->td_ucred, NOCRED, td);
2331 	NDFREE(&nd, NDF_ONLY_PNBUF);
2332 	vput(nd.ni_vp);
2333 	VFS_UNLOCK_GIANT(vfslocked);
2334 	if (mtx_owned(&Giant))
2335 		printf("stat(%d): %s\n", vfslocked, path);
2336 	if (error)
2337 		return (error);
2338 	*sbp = sb;
2339 #ifdef KTRACE
2340 	if (KTRPOINT(td, KTR_STRUCT))
2341 		ktrstat(&sb);
2342 #endif
2343 	return (0);
2344 }
2345 
2346 /*
2347  * Get file status; this version does not follow links.
2348  */
2349 #ifndef _SYS_SYSPROTO_H_
2350 struct lstat_args {
2351 	char	*path;
2352 	struct stat *ub;
2353 };
2354 #endif
2355 int
2356 lstat(td, uap)
2357 	struct thread *td;
2358 	register struct lstat_args /* {
2359 		char *path;
2360 		struct stat *ub;
2361 	} */ *uap;
2362 {
2363 	struct stat sb;
2364 	int error;
2365 
2366 	error = kern_lstat(td, uap->path, UIO_USERSPACE, &sb);
2367 	if (error == 0)
2368 		error = copyout(&sb, uap->ub, sizeof (sb));
2369 	return (error);
2370 }
2371 
2372 int
2373 kern_lstat(struct thread *td, char *path, enum uio_seg pathseg, struct stat *sbp)
2374 {
2375 
2376 	return (kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, path, pathseg,
2377 	    sbp));
2378 }
2379 
2380 /*
2381  * Implementation of the NetBSD [l]stat() functions.
2382  */
2383 void
2384 cvtnstat(sb, nsb)
2385 	struct stat *sb;
2386 	struct nstat *nsb;
2387 {
2388 	bzero(nsb, sizeof *nsb);
2389 	nsb->st_dev = sb->st_dev;
2390 	nsb->st_ino = sb->st_ino;
2391 	nsb->st_mode = sb->st_mode;
2392 	nsb->st_nlink = sb->st_nlink;
2393 	nsb->st_uid = sb->st_uid;
2394 	nsb->st_gid = sb->st_gid;
2395 	nsb->st_rdev = sb->st_rdev;
2396 	nsb->st_atimespec = sb->st_atimespec;
2397 	nsb->st_mtimespec = sb->st_mtimespec;
2398 	nsb->st_ctimespec = sb->st_ctimespec;
2399 	nsb->st_size = sb->st_size;
2400 	nsb->st_blocks = sb->st_blocks;
2401 	nsb->st_blksize = sb->st_blksize;
2402 	nsb->st_flags = sb->st_flags;
2403 	nsb->st_gen = sb->st_gen;
2404 	nsb->st_birthtimespec = sb->st_birthtimespec;
2405 }
2406 
2407 #ifndef _SYS_SYSPROTO_H_
2408 struct nstat_args {
2409 	char	*path;
2410 	struct nstat *ub;
2411 };
2412 #endif
2413 int
2414 nstat(td, uap)
2415 	struct thread *td;
2416 	register struct nstat_args /* {
2417 		char *path;
2418 		struct nstat *ub;
2419 	} */ *uap;
2420 {
2421 	struct stat sb;
2422 	struct nstat nsb;
2423 	int error;
2424 
2425 	error = kern_stat(td, uap->path, UIO_USERSPACE, &sb);
2426 	if (error)
2427 		return (error);
2428 	cvtnstat(&sb, &nsb);
2429 	error = copyout(&nsb, uap->ub, sizeof (nsb));
2430 	return (error);
2431 }
2432 
2433 /*
2434  * NetBSD lstat.  Get file status; this version does not follow links.
2435  */
2436 #ifndef _SYS_SYSPROTO_H_
2437 struct lstat_args {
2438 	char	*path;
2439 	struct stat *ub;
2440 };
2441 #endif
2442 int
2443 nlstat(td, uap)
2444 	struct thread *td;
2445 	register struct nlstat_args /* {
2446 		char *path;
2447 		struct nstat *ub;
2448 	} */ *uap;
2449 {
2450 	struct stat sb;
2451 	struct nstat nsb;
2452 	int error;
2453 
2454 	error = kern_lstat(td, uap->path, UIO_USERSPACE, &sb);
2455 	if (error)
2456 		return (error);
2457 	cvtnstat(&sb, &nsb);
2458 	error = copyout(&nsb, uap->ub, sizeof (nsb));
2459 	return (error);
2460 }
2461 
2462 /*
2463  * Get configurable pathname variables.
2464  */
2465 #ifndef _SYS_SYSPROTO_H_
2466 struct pathconf_args {
2467 	char	*path;
2468 	int	name;
2469 };
2470 #endif
2471 int
2472 pathconf(td, uap)
2473 	struct thread *td;
2474 	register struct pathconf_args /* {
2475 		char *path;
2476 		int name;
2477 	} */ *uap;
2478 {
2479 
2480 	return (kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name));
2481 }
2482 
2483 int
2484 kern_pathconf(struct thread *td, char *path, enum uio_seg pathseg, int name)
2485 {
2486 	struct nameidata nd;
2487 	int error, vfslocked;
2488 
2489 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | MPSAFE | AUDITVNODE1,
2490 	    pathseg, path, td);
2491 	if ((error = namei(&nd)) != 0)
2492 		return (error);
2493 	vfslocked = NDHASGIANT(&nd);
2494 	NDFREE(&nd, NDF_ONLY_PNBUF);
2495 
2496 	/* If asynchronous I/O is available, it works for all files. */
2497 	if (name == _PC_ASYNC_IO)
2498 		td->td_retval[0] = async_io_version;
2499 	else
2500 		error = VOP_PATHCONF(nd.ni_vp, name, td->td_retval);
2501 	vput(nd.ni_vp);
2502 	VFS_UNLOCK_GIANT(vfslocked);
2503 	return (error);
2504 }
2505 
2506 /*
2507  * Return target name of a symbolic link.
2508  */
2509 #ifndef _SYS_SYSPROTO_H_
2510 struct readlink_args {
2511 	char	*path;
2512 	char	*buf;
2513 	size_t	count;
2514 };
2515 #endif
2516 int
2517 readlink(td, uap)
2518 	struct thread *td;
2519 	register struct readlink_args /* {
2520 		char *path;
2521 		char *buf;
2522 		size_t count;
2523 	} */ *uap;
2524 {
2525 
2526 	return (kern_readlink(td, uap->path, UIO_USERSPACE, uap->buf,
2527 	    UIO_USERSPACE, uap->count));
2528 }
2529 #ifndef _SYS_SYSPROTO_H_
2530 struct readlinkat_args {
2531 	int	fd;
2532 	char	*path;
2533 	char	*buf;
2534 	size_t	bufsize;
2535 };
2536 #endif
2537 int
2538 readlinkat(struct thread *td, struct readlinkat_args *uap)
2539 {
2540 
2541 	return (kern_readlinkat(td, uap->fd, uap->path, UIO_USERSPACE,
2542 	    uap->buf, UIO_USERSPACE, uap->bufsize));
2543 }
2544 
2545 int
2546 kern_readlink(struct thread *td, char *path, enum uio_seg pathseg, char *buf,
2547     enum uio_seg bufseg, size_t count)
2548 {
2549 
2550 	return (kern_readlinkat(td, AT_FDCWD, path, pathseg, buf, bufseg,
2551 	    count));
2552 }
2553 
2554 int
2555 kern_readlinkat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2556     char *buf, enum uio_seg bufseg, size_t count)
2557 {
2558 	struct vnode *vp;
2559 	struct iovec aiov;
2560 	struct uio auio;
2561 	int error;
2562 	struct nameidata nd;
2563 	int vfslocked;
2564 
2565 	NDINIT_AT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | MPSAFE | AUDITVNODE1,
2566 	    pathseg, path, fd, td);
2567 
2568 	if ((error = namei(&nd)) != 0)
2569 		return (error);
2570 	NDFREE(&nd, NDF_ONLY_PNBUF);
2571 	vfslocked = NDHASGIANT(&nd);
2572 	vp = nd.ni_vp;
2573 #ifdef MAC
2574 	error = mac_vnode_check_readlink(td->td_ucred, vp);
2575 	if (error) {
2576 		vput(vp);
2577 		VFS_UNLOCK_GIANT(vfslocked);
2578 		return (error);
2579 	}
2580 #endif
2581 	if (vp->v_type != VLNK)
2582 		error = EINVAL;
2583 	else {
2584 		aiov.iov_base = buf;
2585 		aiov.iov_len = count;
2586 		auio.uio_iov = &aiov;
2587 		auio.uio_iovcnt = 1;
2588 		auio.uio_offset = 0;
2589 		auio.uio_rw = UIO_READ;
2590 		auio.uio_segflg = bufseg;
2591 		auio.uio_td = td;
2592 		auio.uio_resid = count;
2593 		error = VOP_READLINK(vp, &auio, td->td_ucred);
2594 	}
2595 	vput(vp);
2596 	VFS_UNLOCK_GIANT(vfslocked);
2597 	td->td_retval[0] = count - auio.uio_resid;
2598 	return (error);
2599 }
2600 
2601 /*
2602  * Common implementation code for chflags() and fchflags().
2603  */
2604 static int
2605 setfflags(td, vp, flags)
2606 	struct thread *td;
2607 	struct vnode *vp;
2608 	int flags;
2609 {
2610 	int error;
2611 	struct mount *mp;
2612 	struct vattr vattr;
2613 
2614 	/*
2615 	 * Prevent non-root users from setting flags on devices.  When
2616 	 * a device is reused, users can retain ownership of the device
2617 	 * if they are allowed to set flags and programs assume that
2618 	 * chown can't fail when done as root.
2619 	 */
2620 	if (vp->v_type == VCHR || vp->v_type == VBLK) {
2621 		error = priv_check(td, PRIV_VFS_CHFLAGS_DEV);
2622 		if (error)
2623 			return (error);
2624 	}
2625 
2626 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2627 		return (error);
2628 	VOP_LEASE(vp, td, td->td_ucred, LEASE_WRITE);
2629 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2630 	VATTR_NULL(&vattr);
2631 	vattr.va_flags = flags;
2632 #ifdef MAC
2633 	error = mac_vnode_check_setflags(td->td_ucred, vp, vattr.va_flags);
2634 	if (error == 0)
2635 #endif
2636 		error = VOP_SETATTR(vp, &vattr, td->td_ucred, td);
2637 	VOP_UNLOCK(vp, 0);
2638 	vn_finished_write(mp);
2639 	return (error);
2640 }
2641 
2642 /*
2643  * Change flags of a file given a path name.
2644  */
2645 #ifndef _SYS_SYSPROTO_H_
2646 struct chflags_args {
2647 	char	*path;
2648 	int	flags;
2649 };
2650 #endif
2651 int
2652 chflags(td, uap)
2653 	struct thread *td;
2654 	register struct chflags_args /* {
2655 		char *path;
2656 		int flags;
2657 	} */ *uap;
2658 {
2659 	int error;
2660 	struct nameidata nd;
2661 	int vfslocked;
2662 
2663 	AUDIT_ARG(fflags, uap->flags);
2664 	NDINIT(&nd, LOOKUP, FOLLOW | MPSAFE | AUDITVNODE1, UIO_USERSPACE,
2665 	    uap->path, td);
2666 	if ((error = namei(&nd)) != 0)
2667 		return (error);
2668 	NDFREE(&nd, NDF_ONLY_PNBUF);
2669 	vfslocked = NDHASGIANT(&nd);
2670 	error = setfflags(td, nd.ni_vp, uap->flags);
2671 	vrele(nd.ni_vp);
2672 	VFS_UNLOCK_GIANT(vfslocked);
2673 	return (error);
2674 }
2675 
2676 /*
2677  * Same as chflags() but doesn't follow symlinks.
2678  */
2679 int
2680 lchflags(td, uap)
2681 	struct thread *td;
2682 	register struct lchflags_args /* {
2683 		char *path;
2684 		int flags;
2685 	} */ *uap;
2686 {
2687 	int error;
2688 	struct nameidata nd;
2689 	int vfslocked;
2690 
2691 	AUDIT_ARG(fflags, uap->flags);
2692 	NDINIT(&nd, LOOKUP, NOFOLLOW | MPSAFE | AUDITVNODE1, UIO_USERSPACE,
2693 	    uap->path, td);
2694 	if ((error = namei(&nd)) != 0)
2695 		return (error);
2696 	vfslocked = NDHASGIANT(&nd);
2697 	NDFREE(&nd, NDF_ONLY_PNBUF);
2698 	error = setfflags(td, nd.ni_vp, uap->flags);
2699 	vrele(nd.ni_vp);
2700 	VFS_UNLOCK_GIANT(vfslocked);
2701 	return (error);
2702 }
2703 
2704 /*
2705  * Change flags of a file given a file descriptor.
2706  */
2707 #ifndef _SYS_SYSPROTO_H_
2708 struct fchflags_args {
2709 	int	fd;
2710 	int	flags;
2711 };
2712 #endif
2713 int
2714 fchflags(td, uap)
2715 	struct thread *td;
2716 	register struct fchflags_args /* {
2717 		int fd;
2718 		int flags;
2719 	} */ *uap;
2720 {
2721 	struct file *fp;
2722 	int vfslocked;
2723 	int error;
2724 
2725 	AUDIT_ARG(fd, uap->fd);
2726 	AUDIT_ARG(fflags, uap->flags);
2727 	if ((error = getvnode(td->td_proc->p_fd, uap->fd, &fp)) != 0)
2728 		return (error);
2729 	vfslocked = VFS_LOCK_GIANT(fp->f_vnode->v_mount);
2730 #ifdef AUDIT
2731 	vn_lock(fp->f_vnode, LK_EXCLUSIVE | LK_RETRY);
2732 	AUDIT_ARG(vnode, fp->f_vnode, ARG_VNODE1);
2733 	VOP_UNLOCK(fp->f_vnode, 0);
2734 #endif
2735 	error = setfflags(td, fp->f_vnode, uap->flags);
2736 	VFS_UNLOCK_GIANT(vfslocked);
2737 	fdrop(fp, td);
2738 	return (error);
2739 }
2740 
2741 /*
2742  * Common implementation code for chmod(), lchmod() and fchmod().
2743  */
2744 static int
2745 setfmode(td, vp, mode)
2746 	struct thread *td;
2747 	struct vnode *vp;
2748 	int mode;
2749 {
2750 	int error;
2751 	struct mount *mp;
2752 	struct vattr vattr;
2753 
2754 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2755 		return (error);
2756 	VOP_LEASE(vp, td, td->td_ucred, LEASE_WRITE);
2757 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2758 	VATTR_NULL(&vattr);
2759 	vattr.va_mode = mode & ALLPERMS;
2760 #ifdef MAC
2761 	error = mac_vnode_check_setmode(td->td_ucred, vp, vattr.va_mode);
2762 	if (error == 0)
2763 #endif
2764 		error = VOP_SETATTR(vp, &vattr, td->td_ucred, td);
2765 	VOP_UNLOCK(vp, 0);
2766 	vn_finished_write(mp);
2767 	return (error);
2768 }
2769 
2770 /*
2771  * Change mode of a file given path name.
2772  */
2773 #ifndef _SYS_SYSPROTO_H_
2774 struct chmod_args {
2775 	char	*path;
2776 	int	mode;
2777 };
2778 #endif
2779 int
2780 chmod(td, uap)
2781 	struct thread *td;
2782 	register struct chmod_args /* {
2783 		char *path;
2784 		int mode;
2785 	} */ *uap;
2786 {
2787 
2788 	return (kern_chmod(td, uap->path, UIO_USERSPACE, uap->mode));
2789 }
2790 
2791 #ifndef _SYS_SYSPROTO_H_
2792 struct fchmodat_args {
2793 	int	dirfd;
2794 	char	*path;
2795 	mode_t	mode;
2796 	int	flag;
2797 }
2798 #endif
2799 int
2800 fchmodat(struct thread *td, struct fchmodat_args *uap)
2801 {
2802 	int flag = uap->flag;
2803 	int fd = uap->fd;
2804 	char *path = uap->path;
2805 	mode_t mode = uap->mode;
2806 
2807 	if (flag & ~AT_SYMLINK_NOFOLLOW)
2808 		return (EINVAL);
2809 
2810 	return (kern_fchmodat(td, fd, path, UIO_USERSPACE, mode, flag));
2811 }
2812 
2813 int
2814 kern_chmod(struct thread *td, char *path, enum uio_seg pathseg, int mode)
2815 {
2816 
2817 	return (kern_fchmodat(td, AT_FDCWD, path, pathseg, mode, 0));
2818 }
2819 
2820 /*
2821  * Change mode of a file given path name (don't follow links.)
2822  */
2823 #ifndef _SYS_SYSPROTO_H_
2824 struct lchmod_args {
2825 	char	*path;
2826 	int	mode;
2827 };
2828 #endif
2829 int
2830 lchmod(td, uap)
2831 	struct thread *td;
2832 	register struct lchmod_args /* {
2833 		char *path;
2834 		int mode;
2835 	} */ *uap;
2836 {
2837 
2838 	return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2839 	    uap->mode, AT_SYMLINK_NOFOLLOW));
2840 }
2841 
2842 
2843 int
2844 kern_fchmodat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2845     mode_t mode, int flag)
2846 {
2847 	int error;
2848 	struct nameidata nd;
2849 	int vfslocked;
2850 	int follow;
2851 
2852 	AUDIT_ARG(mode, mode);
2853 	follow = (flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
2854 	NDINIT_AT(&nd, LOOKUP,  follow | MPSAFE | AUDITVNODE1, pathseg, path,
2855 	    fd, td);
2856 	if ((error = namei(&nd)) != 0)
2857 		return (error);
2858 	vfslocked = NDHASGIANT(&nd);
2859 	NDFREE(&nd, NDF_ONLY_PNBUF);
2860 	error = setfmode(td, nd.ni_vp, mode);
2861 	vrele(nd.ni_vp);
2862 	VFS_UNLOCK_GIANT(vfslocked);
2863 	return (error);
2864 }
2865 
2866 /*
2867  * Change mode of a file given a file descriptor.
2868  */
2869 #ifndef _SYS_SYSPROTO_H_
2870 struct fchmod_args {
2871 	int	fd;
2872 	int	mode;
2873 };
2874 #endif
2875 int
2876 fchmod(td, uap)
2877 	struct thread *td;
2878 	register struct fchmod_args /* {
2879 		int fd;
2880 		int mode;
2881 	} */ *uap;
2882 {
2883 	struct file *fp;
2884 	int vfslocked;
2885 	int error;
2886 
2887 	AUDIT_ARG(fd, uap->fd);
2888 	AUDIT_ARG(mode, uap->mode);
2889 	if ((error = getvnode(td->td_proc->p_fd, uap->fd, &fp)) != 0)
2890 		return (error);
2891 	vfslocked = VFS_LOCK_GIANT(fp->f_vnode->v_mount);
2892 #ifdef AUDIT
2893 	vn_lock(fp->f_vnode, LK_EXCLUSIVE | LK_RETRY);
2894 	AUDIT_ARG(vnode, fp->f_vnode, ARG_VNODE1);
2895 	VOP_UNLOCK(fp->f_vnode, 0);
2896 #endif
2897 	error = setfmode(td, fp->f_vnode, uap->mode);
2898 	VFS_UNLOCK_GIANT(vfslocked);
2899 	fdrop(fp, td);
2900 	return (error);
2901 }
2902 
2903 /*
2904  * Common implementation for chown(), lchown(), and fchown()
2905  */
2906 static int
2907 setfown(td, vp, uid, gid)
2908 	struct thread *td;
2909 	struct vnode *vp;
2910 	uid_t uid;
2911 	gid_t gid;
2912 {
2913 	int error;
2914 	struct mount *mp;
2915 	struct vattr vattr;
2916 
2917 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2918 		return (error);
2919 	VOP_LEASE(vp, td, td->td_ucred, LEASE_WRITE);
2920 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2921 	VATTR_NULL(&vattr);
2922 	vattr.va_uid = uid;
2923 	vattr.va_gid = gid;
2924 #ifdef MAC
2925 	error = mac_vnode_check_setowner(td->td_ucred, vp, vattr.va_uid,
2926 	    vattr.va_gid);
2927 	if (error == 0)
2928 #endif
2929 		error = VOP_SETATTR(vp, &vattr, td->td_ucred, td);
2930 	VOP_UNLOCK(vp, 0);
2931 	vn_finished_write(mp);
2932 	return (error);
2933 }
2934 
2935 /*
2936  * Set ownership given a path name.
2937  */
2938 #ifndef _SYS_SYSPROTO_H_
2939 struct chown_args {
2940 	char	*path;
2941 	int	uid;
2942 	int	gid;
2943 };
2944 #endif
2945 int
2946 chown(td, uap)
2947 	struct thread *td;
2948 	register struct chown_args /* {
2949 		char *path;
2950 		int uid;
2951 		int gid;
2952 	} */ *uap;
2953 {
2954 
2955 	return (kern_chown(td, uap->path, UIO_USERSPACE, uap->uid, uap->gid));
2956 }
2957 
2958 #ifndef _SYS_SYSPROTO_H_
2959 struct fchownat_args {
2960 	int fd;
2961 	const char * path;
2962 	uid_t uid;
2963 	gid_t gid;
2964 	int flag;
2965 };
2966 #endif
2967 int
2968 fchownat(struct thread *td, struct fchownat_args *uap)
2969 {
2970 	int flag;
2971 
2972 	flag = uap->flag;
2973 	if (flag & ~AT_SYMLINK_NOFOLLOW)
2974 		return (EINVAL);
2975 
2976 	return (kern_fchownat(td, uap->fd, uap->path, UIO_USERSPACE, uap->uid,
2977 	    uap->gid, uap->flag));
2978 }
2979 
2980 int
2981 kern_chown(struct thread *td, char *path, enum uio_seg pathseg, int uid,
2982     int gid)
2983 {
2984 
2985 	return (kern_fchownat(td, AT_FDCWD, path, pathseg, uid, gid, 0));
2986 }
2987 
2988 int
2989 kern_fchownat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2990     int uid, int gid, int flag)
2991 {
2992 	struct nameidata nd;
2993 	int error, vfslocked, follow;
2994 
2995 	AUDIT_ARG(owner, uid, gid);
2996 	follow = (flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
2997 	NDINIT_AT(&nd, LOOKUP, follow | MPSAFE | AUDITVNODE1, pathseg, path,
2998 	    fd, td);
2999 
3000 	if ((error = namei(&nd)) != 0)
3001 		return (error);
3002 	vfslocked = NDHASGIANT(&nd);
3003 	NDFREE(&nd, NDF_ONLY_PNBUF);
3004 	error = setfown(td, nd.ni_vp, uid, gid);
3005 	vrele(nd.ni_vp);
3006 	VFS_UNLOCK_GIANT(vfslocked);
3007 	return (error);
3008 }
3009 
3010 /*
3011  * Set ownership given a path name, do not cross symlinks.
3012  */
3013 #ifndef _SYS_SYSPROTO_H_
3014 struct lchown_args {
3015 	char	*path;
3016 	int	uid;
3017 	int	gid;
3018 };
3019 #endif
3020 int
3021 lchown(td, uap)
3022 	struct thread *td;
3023 	register struct lchown_args /* {
3024 		char *path;
3025 		int uid;
3026 		int gid;
3027 	} */ *uap;
3028 {
3029 
3030 	return (kern_lchown(td, uap->path, UIO_USERSPACE, uap->uid, uap->gid));
3031 }
3032 
3033 int
3034 kern_lchown(struct thread *td, char *path, enum uio_seg pathseg, int uid,
3035     int gid)
3036 {
3037 
3038 	return (kern_fchownat(td, AT_FDCWD, path, pathseg, uid, gid,
3039 	    AT_SYMLINK_NOFOLLOW));
3040 }
3041 
3042 /*
3043  * Set ownership given a file descriptor.
3044  */
3045 #ifndef _SYS_SYSPROTO_H_
3046 struct fchown_args {
3047 	int	fd;
3048 	int	uid;
3049 	int	gid;
3050 };
3051 #endif
3052 int
3053 fchown(td, uap)
3054 	struct thread *td;
3055 	register struct fchown_args /* {
3056 		int fd;
3057 		int uid;
3058 		int gid;
3059 	} */ *uap;
3060 {
3061 	struct file *fp;
3062 	int vfslocked;
3063 	int error;
3064 
3065 	AUDIT_ARG(fd, uap->fd);
3066 	AUDIT_ARG(owner, uap->uid, uap->gid);
3067 	if ((error = getvnode(td->td_proc->p_fd, uap->fd, &fp)) != 0)
3068 		return (error);
3069 	vfslocked = VFS_LOCK_GIANT(fp->f_vnode->v_mount);
3070 #ifdef AUDIT
3071 	vn_lock(fp->f_vnode, LK_EXCLUSIVE | LK_RETRY);
3072 	AUDIT_ARG(vnode, fp->f_vnode, ARG_VNODE1);
3073 	VOP_UNLOCK(fp->f_vnode, 0);
3074 #endif
3075 	error = setfown(td, fp->f_vnode, uap->uid, uap->gid);
3076 	VFS_UNLOCK_GIANT(vfslocked);
3077 	fdrop(fp, td);
3078 	return (error);
3079 }
3080 
3081 /*
3082  * Common implementation code for utimes(), lutimes(), and futimes().
3083  */
3084 static int
3085 getutimes(usrtvp, tvpseg, tsp)
3086 	const struct timeval *usrtvp;
3087 	enum uio_seg tvpseg;
3088 	struct timespec *tsp;
3089 {
3090 	struct timeval tv[2];
3091 	const struct timeval *tvp;
3092 	int error;
3093 
3094 	if (usrtvp == NULL) {
3095 		microtime(&tv[0]);
3096 		TIMEVAL_TO_TIMESPEC(&tv[0], &tsp[0]);
3097 		tsp[1] = tsp[0];
3098 	} else {
3099 		if (tvpseg == UIO_SYSSPACE) {
3100 			tvp = usrtvp;
3101 		} else {
3102 			if ((error = copyin(usrtvp, tv, sizeof(tv))) != 0)
3103 				return (error);
3104 			tvp = tv;
3105 		}
3106 
3107 		if (tvp[0].tv_usec < 0 || tvp[0].tv_usec >= 1000000 ||
3108 		    tvp[1].tv_usec < 0 || tvp[1].tv_usec >= 1000000)
3109 			return (EINVAL);
3110 		TIMEVAL_TO_TIMESPEC(&tvp[0], &tsp[0]);
3111 		TIMEVAL_TO_TIMESPEC(&tvp[1], &tsp[1]);
3112 	}
3113 	return (0);
3114 }
3115 
3116 /*
3117  * Common implementation code for utimes(), lutimes(), and futimes().
3118  */
3119 static int
3120 setutimes(td, vp, ts, numtimes, nullflag)
3121 	struct thread *td;
3122 	struct vnode *vp;
3123 	const struct timespec *ts;
3124 	int numtimes;
3125 	int nullflag;
3126 {
3127 	int error, setbirthtime;
3128 	struct mount *mp;
3129 	struct vattr vattr;
3130 
3131 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
3132 		return (error);
3133 	VOP_LEASE(vp, td, td->td_ucred, LEASE_WRITE);
3134 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3135 	setbirthtime = 0;
3136 	if (numtimes < 3 && VOP_GETATTR(vp, &vattr, td->td_ucred, td) == 0 &&
3137 	    timespeccmp(&ts[1], &vattr.va_birthtime, < ))
3138 		setbirthtime = 1;
3139 	VATTR_NULL(&vattr);
3140 	vattr.va_atime = ts[0];
3141 	vattr.va_mtime = ts[1];
3142 	if (setbirthtime)
3143 		vattr.va_birthtime = ts[1];
3144 	if (numtimes > 2)
3145 		vattr.va_birthtime = ts[2];
3146 	if (nullflag)
3147 		vattr.va_vaflags |= VA_UTIMES_NULL;
3148 #ifdef MAC
3149 	error = mac_vnode_check_setutimes(td->td_ucred, vp, vattr.va_atime,
3150 	    vattr.va_mtime);
3151 #endif
3152 	if (error == 0)
3153 		error = VOP_SETATTR(vp, &vattr, td->td_ucred, td);
3154 	VOP_UNLOCK(vp, 0);
3155 	vn_finished_write(mp);
3156 	return (error);
3157 }
3158 
3159 /*
3160  * Set the access and modification times of a file.
3161  */
3162 #ifndef _SYS_SYSPROTO_H_
3163 struct utimes_args {
3164 	char	*path;
3165 	struct	timeval *tptr;
3166 };
3167 #endif
3168 int
3169 utimes(td, uap)
3170 	struct thread *td;
3171 	register struct utimes_args /* {
3172 		char *path;
3173 		struct timeval *tptr;
3174 	} */ *uap;
3175 {
3176 
3177 	return (kern_utimes(td, uap->path, UIO_USERSPACE, uap->tptr,
3178 	    UIO_USERSPACE));
3179 }
3180 
3181 #ifndef _SYS_SYSPROTO_H_
3182 struct futimesat_args {
3183 	int fd;
3184 	const char * path;
3185 	const struct timeval * times;
3186 };
3187 #endif
3188 int
3189 futimesat(struct thread *td, struct futimesat_args *uap)
3190 {
3191 
3192 	return (kern_utimesat(td, uap->fd, uap->path, UIO_USERSPACE,
3193 	    uap->times, UIO_USERSPACE));
3194 }
3195 
3196 int
3197 kern_utimes(struct thread *td, char *path, enum uio_seg pathseg,
3198     struct timeval *tptr, enum uio_seg tptrseg)
3199 {
3200 
3201 	return (kern_utimesat(td, AT_FDCWD, path, pathseg, tptr, tptrseg));
3202 }
3203 
3204 int
3205 kern_utimesat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
3206     struct timeval *tptr, enum uio_seg tptrseg)
3207 {
3208 	struct nameidata nd;
3209 	struct timespec ts[2];
3210 	int error, vfslocked;
3211 
3212 	if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3213 		return (error);
3214 	NDINIT_AT(&nd, LOOKUP, FOLLOW | MPSAFE | AUDITVNODE1, pathseg, path,
3215 	    fd, td);
3216 
3217 	if ((error = namei(&nd)) != 0)
3218 		return (error);
3219 	vfslocked = NDHASGIANT(&nd);
3220 	NDFREE(&nd, NDF_ONLY_PNBUF);
3221 	error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3222 	vrele(nd.ni_vp);
3223 	VFS_UNLOCK_GIANT(vfslocked);
3224 	return (error);
3225 }
3226 
3227 /*
3228  * Set the access and modification times of a file.
3229  */
3230 #ifndef _SYS_SYSPROTO_H_
3231 struct lutimes_args {
3232 	char	*path;
3233 	struct	timeval *tptr;
3234 };
3235 #endif
3236 int
3237 lutimes(td, uap)
3238 	struct thread *td;
3239 	register struct lutimes_args /* {
3240 		char *path;
3241 		struct timeval *tptr;
3242 	} */ *uap;
3243 {
3244 
3245 	return (kern_lutimes(td, uap->path, UIO_USERSPACE, uap->tptr,
3246 	    UIO_USERSPACE));
3247 }
3248 
3249 int
3250 kern_lutimes(struct thread *td, char *path, enum uio_seg pathseg,
3251     struct timeval *tptr, enum uio_seg tptrseg)
3252 {
3253 	struct timespec ts[2];
3254 	int error;
3255 	struct nameidata nd;
3256 	int vfslocked;
3257 
3258 	if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3259 		return (error);
3260 	NDINIT(&nd, LOOKUP, NOFOLLOW | MPSAFE | AUDITVNODE1, pathseg, path, td);
3261 	if ((error = namei(&nd)) != 0)
3262 		return (error);
3263 	vfslocked = NDHASGIANT(&nd);
3264 	NDFREE(&nd, NDF_ONLY_PNBUF);
3265 	error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3266 	vrele(nd.ni_vp);
3267 	VFS_UNLOCK_GIANT(vfslocked);
3268 	return (error);
3269 }
3270 
3271 /*
3272  * Set the access and modification times of a file.
3273  */
3274 #ifndef _SYS_SYSPROTO_H_
3275 struct futimes_args {
3276 	int	fd;
3277 	struct	timeval *tptr;
3278 };
3279 #endif
3280 int
3281 futimes(td, uap)
3282 	struct thread *td;
3283 	register struct futimes_args /* {
3284 		int  fd;
3285 		struct timeval *tptr;
3286 	} */ *uap;
3287 {
3288 
3289 	return (kern_futimes(td, uap->fd, uap->tptr, UIO_USERSPACE));
3290 }
3291 
3292 int
3293 kern_futimes(struct thread *td, int fd, struct timeval *tptr,
3294     enum uio_seg tptrseg)
3295 {
3296 	struct timespec ts[2];
3297 	struct file *fp;
3298 	int vfslocked;
3299 	int error;
3300 
3301 	AUDIT_ARG(fd, fd);
3302 	if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3303 		return (error);
3304 	if ((error = getvnode(td->td_proc->p_fd, fd, &fp)) != 0)
3305 		return (error);
3306 	vfslocked = VFS_LOCK_GIANT(fp->f_vnode->v_mount);
3307 #ifdef AUDIT
3308 	vn_lock(fp->f_vnode, LK_EXCLUSIVE | LK_RETRY);
3309 	AUDIT_ARG(vnode, fp->f_vnode, ARG_VNODE1);
3310 	VOP_UNLOCK(fp->f_vnode, 0);
3311 #endif
3312 	error = setutimes(td, fp->f_vnode, ts, 2, tptr == NULL);
3313 	VFS_UNLOCK_GIANT(vfslocked);
3314 	fdrop(fp, td);
3315 	return (error);
3316 }
3317 
3318 /*
3319  * Truncate a file given its path name.
3320  */
3321 #ifndef _SYS_SYSPROTO_H_
3322 struct truncate_args {
3323 	char	*path;
3324 	int	pad;
3325 	off_t	length;
3326 };
3327 #endif
3328 int
3329 truncate(td, uap)
3330 	struct thread *td;
3331 	register struct truncate_args /* {
3332 		char *path;
3333 		int pad;
3334 		off_t length;
3335 	} */ *uap;
3336 {
3337 
3338 	return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3339 }
3340 
3341 int
3342 kern_truncate(struct thread *td, char *path, enum uio_seg pathseg, off_t length)
3343 {
3344 	struct mount *mp;
3345 	struct vnode *vp;
3346 	struct vattr vattr;
3347 	int error;
3348 	struct nameidata nd;
3349 	int vfslocked;
3350 
3351 	if (length < 0)
3352 		return(EINVAL);
3353 	NDINIT(&nd, LOOKUP, FOLLOW | MPSAFE | AUDITVNODE1, pathseg, path, td);
3354 	if ((error = namei(&nd)) != 0)
3355 		return (error);
3356 	vfslocked = NDHASGIANT(&nd);
3357 	vp = nd.ni_vp;
3358 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) {
3359 		vrele(vp);
3360 		VFS_UNLOCK_GIANT(vfslocked);
3361 		return (error);
3362 	}
3363 	NDFREE(&nd, NDF_ONLY_PNBUF);
3364 	VOP_LEASE(vp, td, td->td_ucred, LEASE_WRITE);
3365 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3366 	if (vp->v_type == VDIR)
3367 		error = EISDIR;
3368 #ifdef MAC
3369 	else if ((error = mac_vnode_check_write(td->td_ucred, NOCRED, vp))) {
3370 	}
3371 #endif
3372 	else if ((error = vn_writechk(vp)) == 0 &&
3373 	    (error = VOP_ACCESS(vp, VWRITE, td->td_ucred, td)) == 0) {
3374 		VATTR_NULL(&vattr);
3375 		vattr.va_size = length;
3376 		error = VOP_SETATTR(vp, &vattr, td->td_ucred, td);
3377 	}
3378 	vput(vp);
3379 	vn_finished_write(mp);
3380 	VFS_UNLOCK_GIANT(vfslocked);
3381 	return (error);
3382 }
3383 
3384 #if defined(COMPAT_43)
3385 /*
3386  * Truncate a file given its path name.
3387  */
3388 #ifndef _SYS_SYSPROTO_H_
3389 struct otruncate_args {
3390 	char	*path;
3391 	long	length;
3392 };
3393 #endif
3394 int
3395 otruncate(td, uap)
3396 	struct thread *td;
3397 	register struct otruncate_args /* {
3398 		char *path;
3399 		long length;
3400 	} */ *uap;
3401 {
3402 	struct truncate_args /* {
3403 		char *path;
3404 		int pad;
3405 		off_t length;
3406 	} */ nuap;
3407 
3408 	nuap.path = uap->path;
3409 	nuap.length = uap->length;
3410 	return (truncate(td, &nuap));
3411 }
3412 #endif /* COMPAT_43 */
3413 
3414 /* Versions with the pad argument */
3415 int
3416 freebsd6_truncate(struct thread *td, struct freebsd6_truncate_args *uap)
3417 {
3418 	struct truncate_args ouap;
3419 
3420 	ouap.path = uap->path;
3421 	ouap.length = uap->length;
3422 	return (truncate(td, &ouap));
3423 }
3424 
3425 int
3426 freebsd6_ftruncate(struct thread *td, struct freebsd6_ftruncate_args *uap)
3427 {
3428 	struct ftruncate_args ouap;
3429 
3430 	ouap.fd = uap->fd;
3431 	ouap.length = uap->length;
3432 	return (ftruncate(td, &ouap));
3433 }
3434 
3435 /*
3436  * Sync an open file.
3437  */
3438 #ifndef _SYS_SYSPROTO_H_
3439 struct fsync_args {
3440 	int	fd;
3441 };
3442 #endif
3443 int
3444 fsync(td, uap)
3445 	struct thread *td;
3446 	struct fsync_args /* {
3447 		int fd;
3448 	} */ *uap;
3449 {
3450 	struct vnode *vp;
3451 	struct mount *mp;
3452 	struct file *fp;
3453 	int vfslocked;
3454 	int error;
3455 
3456 	AUDIT_ARG(fd, uap->fd);
3457 	if ((error = getvnode(td->td_proc->p_fd, uap->fd, &fp)) != 0)
3458 		return (error);
3459 	vp = fp->f_vnode;
3460 	vfslocked = VFS_LOCK_GIANT(vp->v_mount);
3461 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
3462 		goto drop;
3463 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3464 	AUDIT_ARG(vnode, vp, ARG_VNODE1);
3465 	if (vp->v_object != NULL) {
3466 		VM_OBJECT_LOCK(vp->v_object);
3467 		vm_object_page_clean(vp->v_object, 0, 0, 0);
3468 		VM_OBJECT_UNLOCK(vp->v_object);
3469 	}
3470 	error = VOP_FSYNC(vp, MNT_WAIT, td);
3471 
3472 	VOP_UNLOCK(vp, 0);
3473 	vn_finished_write(mp);
3474 drop:
3475 	VFS_UNLOCK_GIANT(vfslocked);
3476 	fdrop(fp, td);
3477 	return (error);
3478 }
3479 
3480 /*
3481  * Rename files.  Source and destination must either both be directories, or
3482  * both not be directories.  If target is a directory, it must be empty.
3483  */
3484 #ifndef _SYS_SYSPROTO_H_
3485 struct rename_args {
3486 	char	*from;
3487 	char	*to;
3488 };
3489 #endif
3490 int
3491 rename(td, uap)
3492 	struct thread *td;
3493 	register struct rename_args /* {
3494 		char *from;
3495 		char *to;
3496 	} */ *uap;
3497 {
3498 
3499 	return (kern_rename(td, uap->from, uap->to, UIO_USERSPACE));
3500 }
3501 
3502 #ifndef _SYS_SYSPROTO_H_
3503 struct renameat_args {
3504 	int	oldfd;
3505 	char	*old;
3506 	int	newfd;
3507 	char	*new;
3508 };
3509 #endif
3510 int
3511 renameat(struct thread *td, struct renameat_args *uap)
3512 {
3513 
3514 	return (kern_renameat(td, uap->oldfd, uap->old, uap->newfd, uap->new,
3515 	    UIO_USERSPACE));
3516 }
3517 
3518 int
3519 kern_rename(struct thread *td, char *from, char *to, enum uio_seg pathseg)
3520 {
3521 
3522 	return (kern_renameat(td, AT_FDCWD, from, AT_FDCWD, to, pathseg));
3523 }
3524 
3525 int
3526 kern_renameat(struct thread *td, int oldfd, char *old, int newfd, char *new,
3527     enum uio_seg pathseg)
3528 {
3529 	struct mount *mp = NULL;
3530 	struct vnode *tvp, *fvp, *tdvp;
3531 	struct nameidata fromnd, tond;
3532 	int tvfslocked;
3533 	int fvfslocked;
3534 	int error;
3535 
3536 	bwillwrite();
3537 #ifdef MAC
3538 	NDINIT_AT(&fromnd, DELETE, LOCKPARENT | LOCKLEAF | SAVESTART | MPSAFE |
3539 	    AUDITVNODE1, pathseg, old, oldfd, td);
3540 #else
3541 	NDINIT_AT(&fromnd, DELETE, WANTPARENT | SAVESTART | MPSAFE |
3542 	    AUDITVNODE1, pathseg, old, oldfd, td);
3543 #endif
3544 
3545 	if ((error = namei(&fromnd)) != 0)
3546 		return (error);
3547 	fvfslocked = NDHASGIANT(&fromnd);
3548 	tvfslocked = 0;
3549 #ifdef MAC
3550 	error = mac_vnode_check_rename_from(td->td_ucred, fromnd.ni_dvp,
3551 	    fromnd.ni_vp, &fromnd.ni_cnd);
3552 	VOP_UNLOCK(fromnd.ni_dvp, 0);
3553 	if (fromnd.ni_dvp != fromnd.ni_vp)
3554 		VOP_UNLOCK(fromnd.ni_vp, 0);
3555 #endif
3556 	fvp = fromnd.ni_vp;
3557 	if (error == 0)
3558 		error = vn_start_write(fvp, &mp, V_WAIT | PCATCH);
3559 	if (error != 0) {
3560 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3561 		vrele(fromnd.ni_dvp);
3562 		vrele(fvp);
3563 		goto out1;
3564 	}
3565 	NDINIT_AT(&tond, RENAME, LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART |
3566 	    MPSAFE | AUDITVNODE2, pathseg, new, newfd, td);
3567 	if (fromnd.ni_vp->v_type == VDIR)
3568 		tond.ni_cnd.cn_flags |= WILLBEDIR;
3569 	if ((error = namei(&tond)) != 0) {
3570 		/* Translate error code for rename("dir1", "dir2/."). */
3571 		if (error == EISDIR && fvp->v_type == VDIR)
3572 			error = EINVAL;
3573 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3574 		vrele(fromnd.ni_dvp);
3575 		vrele(fvp);
3576 		vn_finished_write(mp);
3577 		goto out1;
3578 	}
3579 	tvfslocked = NDHASGIANT(&tond);
3580 	tdvp = tond.ni_dvp;
3581 	tvp = tond.ni_vp;
3582 	if (tvp != NULL) {
3583 		if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
3584 			error = ENOTDIR;
3585 			goto out;
3586 		} else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
3587 			error = EISDIR;
3588 			goto out;
3589 		}
3590 	}
3591 	if (fvp == tdvp)
3592 		error = EINVAL;
3593 	/*
3594 	 * If the source is the same as the destination (that is, if they
3595 	 * are links to the same vnode), then there is nothing to do.
3596 	 */
3597 	if (fvp == tvp)
3598 		error = -1;
3599 #ifdef MAC
3600 	else
3601 		error = mac_vnode_check_rename_to(td->td_ucred, tdvp,
3602 		    tond.ni_vp, fromnd.ni_dvp == tdvp, &tond.ni_cnd);
3603 #endif
3604 out:
3605 	if (!error) {
3606 		VOP_LEASE(tdvp, td, td->td_ucred, LEASE_WRITE);
3607 		if (fromnd.ni_dvp != tdvp) {
3608 			VOP_LEASE(fromnd.ni_dvp, td, td->td_ucred, LEASE_WRITE);
3609 		}
3610 		if (tvp) {
3611 			VOP_LEASE(tvp, td, td->td_ucred, LEASE_WRITE);
3612 		}
3613 		error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
3614 				   tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
3615 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3616 		NDFREE(&tond, NDF_ONLY_PNBUF);
3617 	} else {
3618 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3619 		NDFREE(&tond, NDF_ONLY_PNBUF);
3620 		if (tvp)
3621 			vput(tvp);
3622 		if (tdvp == tvp)
3623 			vrele(tdvp);
3624 		else
3625 			vput(tdvp);
3626 		vrele(fromnd.ni_dvp);
3627 		vrele(fvp);
3628 	}
3629 	vrele(tond.ni_startdir);
3630 	vn_finished_write(mp);
3631 out1:
3632 	if (fromnd.ni_startdir)
3633 		vrele(fromnd.ni_startdir);
3634 	VFS_UNLOCK_GIANT(fvfslocked);
3635 	VFS_UNLOCK_GIANT(tvfslocked);
3636 	if (error == -1)
3637 		return (0);
3638 	return (error);
3639 }
3640 
3641 /*
3642  * Make a directory file.
3643  */
3644 #ifndef _SYS_SYSPROTO_H_
3645 struct mkdir_args {
3646 	char	*path;
3647 	int	mode;
3648 };
3649 #endif
3650 int
3651 mkdir(td, uap)
3652 	struct thread *td;
3653 	register struct mkdir_args /* {
3654 		char *path;
3655 		int mode;
3656 	} */ *uap;
3657 {
3658 
3659 	return (kern_mkdir(td, uap->path, UIO_USERSPACE, uap->mode));
3660 }
3661 
3662 #ifndef _SYS_SYSPROTO_H_
3663 struct mkdirat_args {
3664 	int	fd;
3665 	char	*path;
3666 	mode_t	mode;
3667 };
3668 #endif
3669 int
3670 mkdirat(struct thread *td, struct mkdirat_args *uap)
3671 {
3672 
3673 	return (kern_mkdirat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode));
3674 }
3675 
3676 int
3677 kern_mkdir(struct thread *td, char *path, enum uio_seg segflg, int mode)
3678 {
3679 
3680 	return (kern_mkdirat(td, AT_FDCWD, path, segflg, mode));
3681 }
3682 
3683 int
3684 kern_mkdirat(struct thread *td, int fd, char *path, enum uio_seg segflg,
3685     int mode)
3686 {
3687 	struct mount *mp;
3688 	struct vnode *vp;
3689 	struct vattr vattr;
3690 	int error;
3691 	struct nameidata nd;
3692 	int vfslocked;
3693 
3694 	AUDIT_ARG(mode, mode);
3695 restart:
3696 	bwillwrite();
3697 	NDINIT_AT(&nd, CREATE, LOCKPARENT | SAVENAME | MPSAFE | AUDITVNODE1,
3698 	    segflg, path, fd, td);
3699 	nd.ni_cnd.cn_flags |= WILLBEDIR;
3700 	if ((error = namei(&nd)) != 0)
3701 		return (error);
3702 	vfslocked = NDHASGIANT(&nd);
3703 	vp = nd.ni_vp;
3704 	if (vp != NULL) {
3705 		NDFREE(&nd, NDF_ONLY_PNBUF);
3706 		/*
3707 		 * XXX namei called with LOCKPARENT but not LOCKLEAF has
3708 		 * the strange behaviour of leaving the vnode unlocked
3709 		 * if the target is the same vnode as the parent.
3710 		 */
3711 		if (vp == nd.ni_dvp)
3712 			vrele(nd.ni_dvp);
3713 		else
3714 			vput(nd.ni_dvp);
3715 		vrele(vp);
3716 		VFS_UNLOCK_GIANT(vfslocked);
3717 		return (EEXIST);
3718 	}
3719 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3720 		NDFREE(&nd, NDF_ONLY_PNBUF);
3721 		vput(nd.ni_dvp);
3722 		VFS_UNLOCK_GIANT(vfslocked);
3723 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3724 			return (error);
3725 		goto restart;
3726 	}
3727 	VATTR_NULL(&vattr);
3728 	vattr.va_type = VDIR;
3729 	FILEDESC_SLOCK(td->td_proc->p_fd);
3730 	vattr.va_mode = (mode & ACCESSPERMS) &~ td->td_proc->p_fd->fd_cmask;
3731 	FILEDESC_SUNLOCK(td->td_proc->p_fd);
3732 #ifdef MAC
3733 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
3734 	    &vattr);
3735 	if (error)
3736 		goto out;
3737 #endif
3738 	VOP_LEASE(nd.ni_dvp, td, td->td_ucred, LEASE_WRITE);
3739 	error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
3740 #ifdef MAC
3741 out:
3742 #endif
3743 	NDFREE(&nd, NDF_ONLY_PNBUF);
3744 	vput(nd.ni_dvp);
3745 	if (!error)
3746 		vput(nd.ni_vp);
3747 	vn_finished_write(mp);
3748 	VFS_UNLOCK_GIANT(vfslocked);
3749 	return (error);
3750 }
3751 
3752 /*
3753  * Remove a directory file.
3754  */
3755 #ifndef _SYS_SYSPROTO_H_
3756 struct rmdir_args {
3757 	char	*path;
3758 };
3759 #endif
3760 int
3761 rmdir(td, uap)
3762 	struct thread *td;
3763 	struct rmdir_args /* {
3764 		char *path;
3765 	} */ *uap;
3766 {
3767 
3768 	return (kern_rmdir(td, uap->path, UIO_USERSPACE));
3769 }
3770 
3771 int
3772 kern_rmdir(struct thread *td, char *path, enum uio_seg pathseg)
3773 {
3774 
3775 	return (kern_rmdirat(td, AT_FDCWD, path, pathseg));
3776 }
3777 
3778 int
3779 kern_rmdirat(struct thread *td, int fd, char *path, enum uio_seg pathseg)
3780 {
3781 	struct mount *mp;
3782 	struct vnode *vp;
3783 	int error;
3784 	struct nameidata nd;
3785 	int vfslocked;
3786 
3787 restart:
3788 	bwillwrite();
3789 	NDINIT_AT(&nd, DELETE, LOCKPARENT | LOCKLEAF | MPSAFE | AUDITVNODE1,
3790 	    pathseg, path, fd, td);
3791 	if ((error = namei(&nd)) != 0)
3792 		return (error);
3793 	vfslocked = NDHASGIANT(&nd);
3794 	vp = nd.ni_vp;
3795 	if (vp->v_type != VDIR) {
3796 		error = ENOTDIR;
3797 		goto out;
3798 	}
3799 	/*
3800 	 * No rmdir "." please.
3801 	 */
3802 	if (nd.ni_dvp == vp) {
3803 		error = EINVAL;
3804 		goto out;
3805 	}
3806 	/*
3807 	 * The root of a mounted filesystem cannot be deleted.
3808 	 */
3809 	if (vp->v_vflag & VV_ROOT) {
3810 		error = EBUSY;
3811 		goto out;
3812 	}
3813 #ifdef MAC
3814 	error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
3815 	    &nd.ni_cnd);
3816 	if (error)
3817 		goto out;
3818 #endif
3819 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3820 		NDFREE(&nd, NDF_ONLY_PNBUF);
3821 		vput(vp);
3822 		if (nd.ni_dvp == vp)
3823 			vrele(nd.ni_dvp);
3824 		else
3825 			vput(nd.ni_dvp);
3826 		VFS_UNLOCK_GIANT(vfslocked);
3827 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3828 			return (error);
3829 		goto restart;
3830 	}
3831 	VOP_LEASE(nd.ni_dvp, td, td->td_ucred, LEASE_WRITE);
3832 	VOP_LEASE(vp, td, td->td_ucred, LEASE_WRITE);
3833 	error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
3834 	vn_finished_write(mp);
3835 out:
3836 	NDFREE(&nd, NDF_ONLY_PNBUF);
3837 	vput(vp);
3838 	if (nd.ni_dvp == vp)
3839 		vrele(nd.ni_dvp);
3840 	else
3841 		vput(nd.ni_dvp);
3842 	VFS_UNLOCK_GIANT(vfslocked);
3843 	return (error);
3844 }
3845 
3846 #ifdef COMPAT_43
3847 /*
3848  * Read a block of directory entries in a filesystem independent format.
3849  */
3850 #ifndef _SYS_SYSPROTO_H_
3851 struct ogetdirentries_args {
3852 	int	fd;
3853 	char	*buf;
3854 	u_int	count;
3855 	long	*basep;
3856 };
3857 #endif
3858 int
3859 ogetdirentries(td, uap)
3860 	struct thread *td;
3861 	register struct ogetdirentries_args /* {
3862 		int fd;
3863 		char *buf;
3864 		u_int count;
3865 		long *basep;
3866 	} */ *uap;
3867 {
3868 	struct vnode *vp;
3869 	struct file *fp;
3870 	struct uio auio, kuio;
3871 	struct iovec aiov, kiov;
3872 	struct dirent *dp, *edp;
3873 	caddr_t dirbuf;
3874 	int error, eofflag, readcnt, vfslocked;
3875 	long loff;
3876 
3877 	/* XXX arbitrary sanity limit on `count'. */
3878 	if (uap->count > 64 * 1024)
3879 		return (EINVAL);
3880 	if ((error = getvnode(td->td_proc->p_fd, uap->fd, &fp)) != 0)
3881 		return (error);
3882 	if ((fp->f_flag & FREAD) == 0) {
3883 		fdrop(fp, td);
3884 		return (EBADF);
3885 	}
3886 	vp = fp->f_vnode;
3887 unionread:
3888 	vfslocked = VFS_LOCK_GIANT(vp->v_mount);
3889 	if (vp->v_type != VDIR) {
3890 		VFS_UNLOCK_GIANT(vfslocked);
3891 		fdrop(fp, td);
3892 		return (EINVAL);
3893 	}
3894 	aiov.iov_base = uap->buf;
3895 	aiov.iov_len = uap->count;
3896 	auio.uio_iov = &aiov;
3897 	auio.uio_iovcnt = 1;
3898 	auio.uio_rw = UIO_READ;
3899 	auio.uio_segflg = UIO_USERSPACE;
3900 	auio.uio_td = td;
3901 	auio.uio_resid = uap->count;
3902 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3903 	loff = auio.uio_offset = fp->f_offset;
3904 #ifdef MAC
3905 	error = mac_vnode_check_readdir(td->td_ucred, vp);
3906 	if (error) {
3907 		VOP_UNLOCK(vp, 0);
3908 		VFS_UNLOCK_GIANT(vfslocked);
3909 		fdrop(fp, td);
3910 		return (error);
3911 	}
3912 #endif
3913 #	if (BYTE_ORDER != LITTLE_ENDIAN)
3914 		if (vp->v_mount->mnt_maxsymlinklen <= 0) {
3915 			error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag,
3916 			    NULL, NULL);
3917 			fp->f_offset = auio.uio_offset;
3918 		} else
3919 #	endif
3920 	{
3921 		kuio = auio;
3922 		kuio.uio_iov = &kiov;
3923 		kuio.uio_segflg = UIO_SYSSPACE;
3924 		kiov.iov_len = uap->count;
3925 		MALLOC(dirbuf, caddr_t, uap->count, M_TEMP, M_WAITOK);
3926 		kiov.iov_base = dirbuf;
3927 		error = VOP_READDIR(vp, &kuio, fp->f_cred, &eofflag,
3928 			    NULL, NULL);
3929 		fp->f_offset = kuio.uio_offset;
3930 		if (error == 0) {
3931 			readcnt = uap->count - kuio.uio_resid;
3932 			edp = (struct dirent *)&dirbuf[readcnt];
3933 			for (dp = (struct dirent *)dirbuf; dp < edp; ) {
3934 #				if (BYTE_ORDER == LITTLE_ENDIAN)
3935 					/*
3936 					 * The expected low byte of
3937 					 * dp->d_namlen is our dp->d_type.
3938 					 * The high MBZ byte of dp->d_namlen
3939 					 * is our dp->d_namlen.
3940 					 */
3941 					dp->d_type = dp->d_namlen;
3942 					dp->d_namlen = 0;
3943 #				else
3944 					/*
3945 					 * The dp->d_type is the high byte
3946 					 * of the expected dp->d_namlen,
3947 					 * so must be zero'ed.
3948 					 */
3949 					dp->d_type = 0;
3950 #				endif
3951 				if (dp->d_reclen > 0) {
3952 					dp = (struct dirent *)
3953 					    ((char *)dp + dp->d_reclen);
3954 				} else {
3955 					error = EIO;
3956 					break;
3957 				}
3958 			}
3959 			if (dp >= edp)
3960 				error = uiomove(dirbuf, readcnt, &auio);
3961 		}
3962 		FREE(dirbuf, M_TEMP);
3963 	}
3964 	if (error) {
3965 		VOP_UNLOCK(vp, 0);
3966 		VFS_UNLOCK_GIANT(vfslocked);
3967 		fdrop(fp, td);
3968 		return (error);
3969 	}
3970 	if (uap->count == auio.uio_resid &&
3971 	    (vp->v_vflag & VV_ROOT) &&
3972 	    (vp->v_mount->mnt_flag & MNT_UNION)) {
3973 		struct vnode *tvp = vp;
3974 		vp = vp->v_mount->mnt_vnodecovered;
3975 		VREF(vp);
3976 		fp->f_vnode = vp;
3977 		fp->f_data = vp;
3978 		fp->f_offset = 0;
3979 		vput(tvp);
3980 		VFS_UNLOCK_GIANT(vfslocked);
3981 		goto unionread;
3982 	}
3983 	VOP_UNLOCK(vp, 0);
3984 	VFS_UNLOCK_GIANT(vfslocked);
3985 	error = copyout(&loff, uap->basep, sizeof(long));
3986 	fdrop(fp, td);
3987 	td->td_retval[0] = uap->count - auio.uio_resid;
3988 	return (error);
3989 }
3990 #endif /* COMPAT_43 */
3991 
3992 /*
3993  * Read a block of directory entries in a filesystem independent format.
3994  */
3995 #ifndef _SYS_SYSPROTO_H_
3996 struct getdirentries_args {
3997 	int	fd;
3998 	char	*buf;
3999 	u_int	count;
4000 	long	*basep;
4001 };
4002 #endif
4003 int
4004 getdirentries(td, uap)
4005 	struct thread *td;
4006 	register struct getdirentries_args /* {
4007 		int fd;
4008 		char *buf;
4009 		u_int count;
4010 		long *basep;
4011 	} */ *uap;
4012 {
4013 	struct vnode *vp;
4014 	struct file *fp;
4015 	struct uio auio;
4016 	struct iovec aiov;
4017 	int vfslocked;
4018 	long loff;
4019 	int error, eofflag;
4020 
4021 	AUDIT_ARG(fd, uap->fd);
4022 	if ((error = getvnode(td->td_proc->p_fd, uap->fd, &fp)) != 0)
4023 		return (error);
4024 	if ((fp->f_flag & FREAD) == 0) {
4025 		fdrop(fp, td);
4026 		return (EBADF);
4027 	}
4028 	vp = fp->f_vnode;
4029 unionread:
4030 	vfslocked = VFS_LOCK_GIANT(vp->v_mount);
4031 	if (vp->v_type != VDIR) {
4032 		VFS_UNLOCK_GIANT(vfslocked);
4033 		error = EINVAL;
4034 		goto fail;
4035 	}
4036 	aiov.iov_base = uap->buf;
4037 	aiov.iov_len = uap->count;
4038 	auio.uio_iov = &aiov;
4039 	auio.uio_iovcnt = 1;
4040 	auio.uio_rw = UIO_READ;
4041 	auio.uio_segflg = UIO_USERSPACE;
4042 	auio.uio_td = td;
4043 	auio.uio_resid = uap->count;
4044 	/* vn_lock(vp, LK_SHARED | LK_RETRY); */
4045 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
4046 	AUDIT_ARG(vnode, vp, ARG_VNODE1);
4047 	loff = auio.uio_offset = fp->f_offset;
4048 #ifdef MAC
4049 	error = mac_vnode_check_readdir(td->td_ucred, vp);
4050 	if (error == 0)
4051 #endif
4052 		error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, NULL,
4053 		    NULL);
4054 	fp->f_offset = auio.uio_offset;
4055 	if (error) {
4056 		VOP_UNLOCK(vp, 0);
4057 		VFS_UNLOCK_GIANT(vfslocked);
4058 		goto fail;
4059 	}
4060 	if (uap->count == auio.uio_resid &&
4061 	    (vp->v_vflag & VV_ROOT) &&
4062 	    (vp->v_mount->mnt_flag & MNT_UNION)) {
4063 		struct vnode *tvp = vp;
4064 		vp = vp->v_mount->mnt_vnodecovered;
4065 		VREF(vp);
4066 		fp->f_vnode = vp;
4067 		fp->f_data = vp;
4068 		fp->f_offset = 0;
4069 		vput(tvp);
4070 		VFS_UNLOCK_GIANT(vfslocked);
4071 		goto unionread;
4072 	}
4073 	VOP_UNLOCK(vp, 0);
4074 	VFS_UNLOCK_GIANT(vfslocked);
4075 	if (uap->basep != NULL) {
4076 		error = copyout(&loff, uap->basep, sizeof(long));
4077 	}
4078 	td->td_retval[0] = uap->count - auio.uio_resid;
4079 fail:
4080 	fdrop(fp, td);
4081 	return (error);
4082 }
4083 
4084 #ifndef _SYS_SYSPROTO_H_
4085 struct getdents_args {
4086 	int fd;
4087 	char *buf;
4088 	size_t count;
4089 };
4090 #endif
4091 int
4092 getdents(td, uap)
4093 	struct thread *td;
4094 	register struct getdents_args /* {
4095 		int fd;
4096 		char *buf;
4097 		u_int count;
4098 	} */ *uap;
4099 {
4100 	struct getdirentries_args ap;
4101 	ap.fd = uap->fd;
4102 	ap.buf = uap->buf;
4103 	ap.count = uap->count;
4104 	ap.basep = NULL;
4105 	return (getdirentries(td, &ap));
4106 }
4107 
4108 /*
4109  * Set the mode mask for creation of filesystem nodes.
4110  */
4111 #ifndef _SYS_SYSPROTO_H_
4112 struct umask_args {
4113 	int	newmask;
4114 };
4115 #endif
4116 int
4117 umask(td, uap)
4118 	struct thread *td;
4119 	struct umask_args /* {
4120 		int newmask;
4121 	} */ *uap;
4122 {
4123 	register struct filedesc *fdp;
4124 
4125 	FILEDESC_XLOCK(td->td_proc->p_fd);
4126 	fdp = td->td_proc->p_fd;
4127 	td->td_retval[0] = fdp->fd_cmask;
4128 	fdp->fd_cmask = uap->newmask & ALLPERMS;
4129 	FILEDESC_XUNLOCK(td->td_proc->p_fd);
4130 	return (0);
4131 }
4132 
4133 /*
4134  * Void all references to file by ripping underlying filesystem away from
4135  * vnode.
4136  */
4137 #ifndef _SYS_SYSPROTO_H_
4138 struct revoke_args {
4139 	char	*path;
4140 };
4141 #endif
4142 int
4143 revoke(td, uap)
4144 	struct thread *td;
4145 	register struct revoke_args /* {
4146 		char *path;
4147 	} */ *uap;
4148 {
4149 	struct vnode *vp;
4150 	struct vattr vattr;
4151 	int error;
4152 	struct nameidata nd;
4153 	int vfslocked;
4154 
4155 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | MPSAFE | AUDITVNODE1,
4156 	    UIO_USERSPACE, uap->path, td);
4157 	if ((error = namei(&nd)) != 0)
4158 		return (error);
4159 	vfslocked = NDHASGIANT(&nd);
4160 	vp = nd.ni_vp;
4161 	NDFREE(&nd, NDF_ONLY_PNBUF);
4162 	if (vp->v_type != VCHR) {
4163 		error = EINVAL;
4164 		goto out;
4165 	}
4166 #ifdef MAC
4167 	error = mac_vnode_check_revoke(td->td_ucred, vp);
4168 	if (error)
4169 		goto out;
4170 #endif
4171 	error = VOP_GETATTR(vp, &vattr, td->td_ucred, td);
4172 	if (error)
4173 		goto out;
4174 	if (td->td_ucred->cr_uid != vattr.va_uid) {
4175 		error = priv_check(td, PRIV_VFS_ADMIN);
4176 		if (error)
4177 			goto out;
4178 	}
4179 	if (vcount(vp) > 1)
4180 		VOP_REVOKE(vp, REVOKEALL);
4181 out:
4182 	vput(vp);
4183 	VFS_UNLOCK_GIANT(vfslocked);
4184 	return (error);
4185 }
4186 
4187 /*
4188  * Convert a user file descriptor to a kernel file entry.
4189  * A reference on the file entry is held upon returning.
4190  */
4191 int
4192 getvnode(fdp, fd, fpp)
4193 	struct filedesc *fdp;
4194 	int fd;
4195 	struct file **fpp;
4196 {
4197 	int error;
4198 	struct file *fp;
4199 
4200 	fp = NULL;
4201 	if (fdp == NULL)
4202 		error = EBADF;
4203 	else {
4204 		FILEDESC_SLOCK(fdp);
4205 		if ((u_int)fd >= fdp->fd_nfiles ||
4206 		    (fp = fdp->fd_ofiles[fd]) == NULL)
4207 			error = EBADF;
4208 		else if (fp->f_vnode == NULL) {
4209 			fp = NULL;
4210 			error = EINVAL;
4211 		} else {
4212 			fhold(fp);
4213 			error = 0;
4214 		}
4215 		FILEDESC_SUNLOCK(fdp);
4216 	}
4217 	*fpp = fp;
4218 	return (error);
4219 }
4220 
4221 /*
4222  * Get an (NFS) file handle.
4223  */
4224 #ifndef _SYS_SYSPROTO_H_
4225 struct lgetfh_args {
4226 	char	*fname;
4227 	fhandle_t *fhp;
4228 };
4229 #endif
4230 int
4231 lgetfh(td, uap)
4232 	struct thread *td;
4233 	register struct lgetfh_args *uap;
4234 {
4235 	struct nameidata nd;
4236 	fhandle_t fh;
4237 	register struct vnode *vp;
4238 	int vfslocked;
4239 	int error;
4240 
4241 	error = priv_check(td, PRIV_VFS_GETFH);
4242 	if (error)
4243 		return (error);
4244 	NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | MPSAFE | AUDITVNODE1,
4245 	    UIO_USERSPACE, uap->fname, td);
4246 	error = namei(&nd);
4247 	if (error)
4248 		return (error);
4249 	vfslocked = NDHASGIANT(&nd);
4250 	NDFREE(&nd, NDF_ONLY_PNBUF);
4251 	vp = nd.ni_vp;
4252 	bzero(&fh, sizeof(fh));
4253 	fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
4254 	error = VOP_VPTOFH(vp, &fh.fh_fid);
4255 	vput(vp);
4256 	VFS_UNLOCK_GIANT(vfslocked);
4257 	if (error)
4258 		return (error);
4259 	error = copyout(&fh, uap->fhp, sizeof (fh));
4260 	return (error);
4261 }
4262 
4263 #ifndef _SYS_SYSPROTO_H_
4264 struct getfh_args {
4265 	char	*fname;
4266 	fhandle_t *fhp;
4267 };
4268 #endif
4269 int
4270 getfh(td, uap)
4271 	struct thread *td;
4272 	register struct getfh_args *uap;
4273 {
4274 	struct nameidata nd;
4275 	fhandle_t fh;
4276 	register struct vnode *vp;
4277 	int vfslocked;
4278 	int error;
4279 
4280 	error = priv_check(td, PRIV_VFS_GETFH);
4281 	if (error)
4282 		return (error);
4283 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | MPSAFE | AUDITVNODE1,
4284 	    UIO_USERSPACE, uap->fname, td);
4285 	error = namei(&nd);
4286 	if (error)
4287 		return (error);
4288 	vfslocked = NDHASGIANT(&nd);
4289 	NDFREE(&nd, NDF_ONLY_PNBUF);
4290 	vp = nd.ni_vp;
4291 	bzero(&fh, sizeof(fh));
4292 	fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
4293 	error = VOP_VPTOFH(vp, &fh.fh_fid);
4294 	vput(vp);
4295 	VFS_UNLOCK_GIANT(vfslocked);
4296 	if (error)
4297 		return (error);
4298 	error = copyout(&fh, uap->fhp, sizeof (fh));
4299 	return (error);
4300 }
4301 
4302 /*
4303  * syscall for the rpc.lockd to use to translate a NFS file handle into an
4304  * open descriptor.
4305  *
4306  * warning: do not remove the priv_check() call or this becomes one giant
4307  * security hole.
4308  */
4309 #ifndef _SYS_SYSPROTO_H_
4310 struct fhopen_args {
4311 	const struct fhandle *u_fhp;
4312 	int flags;
4313 };
4314 #endif
4315 int
4316 fhopen(td, uap)
4317 	struct thread *td;
4318 	struct fhopen_args /* {
4319 		const struct fhandle *u_fhp;
4320 		int flags;
4321 	} */ *uap;
4322 {
4323 	struct proc *p = td->td_proc;
4324 	struct mount *mp;
4325 	struct vnode *vp;
4326 	struct fhandle fhp;
4327 	struct vattr vat;
4328 	struct vattr *vap = &vat;
4329 	struct flock lf;
4330 	struct file *fp;
4331 	register struct filedesc *fdp = p->p_fd;
4332 	int fmode, mode, error, type;
4333 	struct file *nfp;
4334 	int vfslocked;
4335 	int indx;
4336 
4337 	error = priv_check(td, PRIV_VFS_FHOPEN);
4338 	if (error)
4339 		return (error);
4340 	fmode = FFLAGS(uap->flags);
4341 	/* why not allow a non-read/write open for our lockd? */
4342 	if (((fmode & (FREAD | FWRITE)) == 0) || (fmode & O_CREAT))
4343 		return (EINVAL);
4344 	error = copyin(uap->u_fhp, &fhp, sizeof(fhp));
4345 	if (error)
4346 		return(error);
4347 	/* find the mount point */
4348 	mp = vfs_getvfs(&fhp.fh_fsid);
4349 	if (mp == NULL)
4350 		return (ESTALE);
4351 	vfslocked = VFS_LOCK_GIANT(mp);
4352 	/* now give me my vnode, it gets returned to me locked */
4353 	error = VFS_FHTOVP(mp, &fhp.fh_fid, &vp);
4354 	if (error)
4355 		goto out;
4356 	/*
4357 	 * from now on we have to make sure not
4358 	 * to forget about the vnode
4359 	 * any error that causes an abort must vput(vp)
4360 	 * just set error = err and 'goto bad;'.
4361 	 */
4362 
4363 	/*
4364 	 * from vn_open
4365 	 */
4366 	if (vp->v_type == VLNK) {
4367 		error = EMLINK;
4368 		goto bad;
4369 	}
4370 	if (vp->v_type == VSOCK) {
4371 		error = EOPNOTSUPP;
4372 		goto bad;
4373 	}
4374 	mode = 0;
4375 	if (fmode & (FWRITE | O_TRUNC)) {
4376 		if (vp->v_type == VDIR) {
4377 			error = EISDIR;
4378 			goto bad;
4379 		}
4380 		error = vn_writechk(vp);
4381 		if (error)
4382 			goto bad;
4383 		mode |= VWRITE;
4384 	}
4385 	if (fmode & FREAD)
4386 		mode |= VREAD;
4387 	if (fmode & O_APPEND)
4388 		mode |= VAPPEND;
4389 #ifdef MAC
4390 	error = mac_vnode_check_open(td->td_ucred, vp, mode);
4391 	if (error)
4392 		goto bad;
4393 #endif
4394 	if (mode) {
4395 		error = VOP_ACCESS(vp, mode, td->td_ucred, td);
4396 		if (error)
4397 			goto bad;
4398 	}
4399 	if (fmode & O_TRUNC) {
4400 		VOP_UNLOCK(vp, 0);				/* XXX */
4401 		if ((error = vn_start_write(NULL, &mp, V_WAIT | PCATCH)) != 0) {
4402 			vrele(vp);
4403 			goto out;
4404 		}
4405 		VOP_LEASE(vp, td, td->td_ucred, LEASE_WRITE);
4406 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);	/* XXX */
4407 #ifdef MAC
4408 		/*
4409 		 * We don't yet have fp->f_cred, so use td->td_ucred, which
4410 		 * should be right.
4411 		 */
4412 		error = mac_vnode_check_write(td->td_ucred, td->td_ucred, vp);
4413 		if (error == 0) {
4414 #endif
4415 			VATTR_NULL(vap);
4416 			vap->va_size = 0;
4417 			error = VOP_SETATTR(vp, vap, td->td_ucred, td);
4418 #ifdef MAC
4419 		}
4420 #endif
4421 		vn_finished_write(mp);
4422 		if (error)
4423 			goto bad;
4424 	}
4425 	error = VOP_OPEN(vp, fmode, td->td_ucred, td, NULL);
4426 	if (error)
4427 		goto bad;
4428 
4429 	if (fmode & FWRITE)
4430 		vp->v_writecount++;
4431 
4432 	/*
4433 	 * end of vn_open code
4434 	 */
4435 
4436 	if ((error = falloc(td, &nfp, &indx)) != 0) {
4437 		if (fmode & FWRITE)
4438 			vp->v_writecount--;
4439 		goto bad;
4440 	}
4441 	/* An extra reference on `nfp' has been held for us by falloc(). */
4442 	fp = nfp;
4443 	nfp->f_vnode = vp;
4444 	finit(nfp, fmode & FMASK, DTYPE_VNODE, vp, &vnops);
4445 	if (fmode & (O_EXLOCK | O_SHLOCK)) {
4446 		lf.l_whence = SEEK_SET;
4447 		lf.l_start = 0;
4448 		lf.l_len = 0;
4449 		if (fmode & O_EXLOCK)
4450 			lf.l_type = F_WRLCK;
4451 		else
4452 			lf.l_type = F_RDLCK;
4453 		type = F_FLOCK;
4454 		if ((fmode & FNONBLOCK) == 0)
4455 			type |= F_WAIT;
4456 		VOP_UNLOCK(vp, 0);
4457 		if ((error = VOP_ADVLOCK(vp, (caddr_t)fp, F_SETLK, &lf,
4458 			    type)) != 0) {
4459 			/*
4460 			 * The lock request failed.  Normally close the
4461 			 * descriptor but handle the case where someone might
4462 			 * have dup()d or close()d it when we weren't looking.
4463 			 */
4464 			fdclose(fdp, fp, indx, td);
4465 
4466 			/*
4467 			 * release our private reference
4468 			 */
4469 			fdrop(fp, td);
4470 			goto out;
4471 		}
4472 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
4473 		atomic_set_int(&fp->f_flag, FHASLOCK);
4474 	}
4475 
4476 	VOP_UNLOCK(vp, 0);
4477 	fdrop(fp, td);
4478 	vfs_rel(mp);
4479 	VFS_UNLOCK_GIANT(vfslocked);
4480 	td->td_retval[0] = indx;
4481 	return (0);
4482 
4483 bad:
4484 	vput(vp);
4485 out:
4486 	vfs_rel(mp);
4487 	VFS_UNLOCK_GIANT(vfslocked);
4488 	return (error);
4489 }
4490 
4491 /*
4492  * Stat an (NFS) file handle.
4493  */
4494 #ifndef _SYS_SYSPROTO_H_
4495 struct fhstat_args {
4496 	struct fhandle *u_fhp;
4497 	struct stat *sb;
4498 };
4499 #endif
4500 int
4501 fhstat(td, uap)
4502 	struct thread *td;
4503 	register struct fhstat_args /* {
4504 		struct fhandle *u_fhp;
4505 		struct stat *sb;
4506 	} */ *uap;
4507 {
4508 	struct stat sb;
4509 	fhandle_t fh;
4510 	struct mount *mp;
4511 	struct vnode *vp;
4512 	int vfslocked;
4513 	int error;
4514 
4515 	error = priv_check(td, PRIV_VFS_FHSTAT);
4516 	if (error)
4517 		return (error);
4518 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
4519 	if (error)
4520 		return (error);
4521 	if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL)
4522 		return (ESTALE);
4523 	vfslocked = VFS_LOCK_GIANT(mp);
4524 	if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp))) {
4525 		vfs_rel(mp);
4526 		VFS_UNLOCK_GIANT(vfslocked);
4527 		return (error);
4528 	}
4529 	error = vn_stat(vp, &sb, td->td_ucred, NOCRED, td);
4530 	vput(vp);
4531 	vfs_rel(mp);
4532 	VFS_UNLOCK_GIANT(vfslocked);
4533 	if (error)
4534 		return (error);
4535 	error = copyout(&sb, uap->sb, sizeof(sb));
4536 	return (error);
4537 }
4538 
4539 /*
4540  * Implement fstatfs() for (NFS) file handles.
4541  */
4542 #ifndef _SYS_SYSPROTO_H_
4543 struct fhstatfs_args {
4544 	struct fhandle *u_fhp;
4545 	struct statfs *buf;
4546 };
4547 #endif
4548 int
4549 fhstatfs(td, uap)
4550 	struct thread *td;
4551 	struct fhstatfs_args /* {
4552 		struct fhandle *u_fhp;
4553 		struct statfs *buf;
4554 	} */ *uap;
4555 {
4556 	struct statfs sf;
4557 	fhandle_t fh;
4558 	int error;
4559 
4560 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
4561 	if (error)
4562 		return (error);
4563 	error = kern_fhstatfs(td, fh, &sf);
4564 	if (error)
4565 		return (error);
4566 	return (copyout(&sf, uap->buf, sizeof(sf)));
4567 }
4568 
4569 int
4570 kern_fhstatfs(struct thread *td, fhandle_t fh, struct statfs *buf)
4571 {
4572 	struct statfs *sp;
4573 	struct mount *mp;
4574 	struct vnode *vp;
4575 	int vfslocked;
4576 	int error;
4577 
4578 	error = priv_check(td, PRIV_VFS_FHSTATFS);
4579 	if (error)
4580 		return (error);
4581 	if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL)
4582 		return (ESTALE);
4583 	vfslocked = VFS_LOCK_GIANT(mp);
4584 	error = VFS_FHTOVP(mp, &fh.fh_fid, &vp);
4585 	if (error) {
4586 		VFS_UNLOCK_GIANT(vfslocked);
4587 		vfs_rel(mp);
4588 		return (error);
4589 	}
4590 	vput(vp);
4591 	error = prison_canseemount(td->td_ucred, mp);
4592 	if (error)
4593 		goto out;
4594 #ifdef MAC
4595 	error = mac_mount_check_stat(td->td_ucred, mp);
4596 	if (error)
4597 		goto out;
4598 #endif
4599 	/*
4600 	 * Set these in case the underlying filesystem fails to do so.
4601 	 */
4602 	sp = &mp->mnt_stat;
4603 	sp->f_version = STATFS_VERSION;
4604 	sp->f_namemax = NAME_MAX;
4605 	sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
4606 	error = VFS_STATFS(mp, sp, td);
4607 	if (error == 0)
4608 		*buf = *sp;
4609 out:
4610 	vfs_rel(mp);
4611 	VFS_UNLOCK_GIANT(vfslocked);
4612 	return (error);
4613 }
4614