xref: /freebsd/sys/kern/vfs_syscalls.c (revision 10f0bcab61ef441cb5af32fb706688d8cbd55dc0)
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 int
1003 kern_open(struct thread *td, char *path, enum uio_seg pathseg, int flags,
1004     int mode)
1005 {
1006 	struct proc *p = td->td_proc;
1007 	struct filedesc *fdp = p->p_fd;
1008 	struct file *fp;
1009 	struct vnode *vp;
1010 	struct vattr vat;
1011 	struct mount *mp;
1012 	int cmode;
1013 	struct file *nfp;
1014 	int type, indx, error;
1015 	struct flock lf;
1016 	struct nameidata nd;
1017 	int vfslocked;
1018 
1019 	AUDIT_ARG(fflags, flags);
1020 	AUDIT_ARG(mode, mode);
1021 	if ((flags & O_ACCMODE) == O_ACCMODE)
1022 		return (EINVAL);
1023 	flags = FFLAGS(flags);
1024 	error = falloc(td, &nfp, &indx);
1025 	if (error)
1026 		return (error);
1027 	/* An extra reference on `nfp' has been held for us by falloc(). */
1028 	fp = nfp;
1029 	/* Set the flags early so the finit in devfs can pick them up. */
1030 	fp->f_flag = flags & FMASK;
1031 	cmode = ((mode &~ fdp->fd_cmask) & ALLPERMS) &~ S_ISTXT;
1032 	NDINIT(&nd, LOOKUP, FOLLOW | AUDITVNODE1 | MPSAFE, pathseg, path, td);
1033 	td->td_dupfd = -1;		/* XXX check for fdopen */
1034  	error = vn_open(&nd, &flags, cmode, fp);
1035 	if (error) {
1036 		/*
1037 		 * If the vn_open replaced the method vector, something
1038 		 * wonderous happened deep below and we just pass it up
1039 		 * pretending we know what we do.
1040 		 */
1041 		if (error == ENXIO && fp->f_ops != &badfileops) {
1042 			fdrop(fp, td);
1043 			td->td_retval[0] = indx;
1044 			return (0);
1045 		}
1046 
1047 		/*
1048 		 * handle special fdopen() case.  bleh.  dupfdopen() is
1049 		 * responsible for dropping the old contents of ofiles[indx]
1050 		 * if it succeeds.
1051 		 */
1052 		if ((error == ENODEV || error == ENXIO) &&
1053 		    td->td_dupfd >= 0 &&		/* XXX from fdopen */
1054 		    (error =
1055 			dupfdopen(td, fdp, indx, td->td_dupfd, flags, error)) == 0) {
1056 			td->td_retval[0] = indx;
1057 			fdrop(fp, td);
1058 			return (0);
1059 		}
1060 		/*
1061 		 * Clean up the descriptor, but only if another thread hadn't
1062 		 * replaced or closed it.
1063 		 */
1064 		fdclose(fdp, fp, indx, td);
1065 		fdrop(fp, td);
1066 
1067 		if (error == ERESTART)
1068 			error = EINTR;
1069 		return (error);
1070 	}
1071 	td->td_dupfd = 0;
1072 	vfslocked = NDHASGIANT(&nd);
1073 	NDFREE(&nd, NDF_ONLY_PNBUF);
1074 	vp = nd.ni_vp;
1075 
1076 	fp->f_vnode = vp;	/* XXX Does devfs need this? */
1077 	/*
1078 	 * If the file wasn't claimed by devfs bind it to the normal
1079 	 * vnode operations here.
1080 	 */
1081 	if (fp->f_ops == &badfileops) {
1082 		KASSERT(vp->v_type != VFIFO, ("Unexpected fifo."));
1083 		fp->f_seqcount = 1;
1084 		finit(fp, flags & FMASK, DTYPE_VNODE, vp, &vnops);
1085 	}
1086 
1087 	VOP_UNLOCK(vp, 0);
1088 	if (flags & (O_EXLOCK | O_SHLOCK)) {
1089 		lf.l_whence = SEEK_SET;
1090 		lf.l_start = 0;
1091 		lf.l_len = 0;
1092 		if (flags & O_EXLOCK)
1093 			lf.l_type = F_WRLCK;
1094 		else
1095 			lf.l_type = F_RDLCK;
1096 		type = F_FLOCK;
1097 		if ((flags & FNONBLOCK) == 0)
1098 			type |= F_WAIT;
1099 		if ((error = VOP_ADVLOCK(vp, (caddr_t)fp, F_SETLK, &lf,
1100 			    type)) != 0)
1101 			goto bad;
1102 		atomic_set_int(&fp->f_flag, FHASLOCK);
1103 	}
1104 	if (flags & O_TRUNC) {
1105 		if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
1106 			goto bad;
1107 		VOP_LEASE(vp, td, td->td_ucred, LEASE_WRITE);
1108 		VATTR_NULL(&vat);
1109 		vat.va_size = 0;
1110 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1111 #ifdef MAC
1112 		error = mac_vnode_check_write(td->td_ucred, fp->f_cred, vp);
1113 		if (error == 0)
1114 #endif
1115 			error = VOP_SETATTR(vp, &vat, td->td_ucred, td);
1116 		VOP_UNLOCK(vp, 0);
1117 		vn_finished_write(mp);
1118 		if (error)
1119 			goto bad;
1120 	}
1121 	VFS_UNLOCK_GIANT(vfslocked);
1122 	/*
1123 	 * Release our private reference, leaving the one associated with
1124 	 * the descriptor table intact.
1125 	 */
1126 	fdrop(fp, td);
1127 	td->td_retval[0] = indx;
1128 	return (0);
1129 bad:
1130 	VFS_UNLOCK_GIANT(vfslocked);
1131 	fdclose(fdp, fp, indx, td);
1132 	fdrop(fp, td);
1133 	return (error);
1134 }
1135 
1136 #ifdef COMPAT_43
1137 /*
1138  * Create a file.
1139  */
1140 #ifndef _SYS_SYSPROTO_H_
1141 struct ocreat_args {
1142 	char	*path;
1143 	int	mode;
1144 };
1145 #endif
1146 int
1147 ocreat(td, uap)
1148 	struct thread *td;
1149 	register struct ocreat_args /* {
1150 		char *path;
1151 		int mode;
1152 	} */ *uap;
1153 {
1154 
1155 	return (kern_open(td, uap->path, UIO_USERSPACE,
1156 	    O_WRONLY | O_CREAT | O_TRUNC, uap->mode));
1157 }
1158 #endif /* COMPAT_43 */
1159 
1160 /*
1161  * Create a special file.
1162  */
1163 #ifndef _SYS_SYSPROTO_H_
1164 struct mknod_args {
1165 	char	*path;
1166 	int	mode;
1167 	int	dev;
1168 };
1169 #endif
1170 int
1171 mknod(td, uap)
1172 	struct thread *td;
1173 	register struct mknod_args /* {
1174 		char *path;
1175 		int mode;
1176 		int dev;
1177 	} */ *uap;
1178 {
1179 
1180 	return (kern_mknod(td, uap->path, UIO_USERSPACE, uap->mode, uap->dev));
1181 }
1182 
1183 int
1184 kern_mknod(struct thread *td, char *path, enum uio_seg pathseg, int mode,
1185     int dev)
1186 {
1187 	struct vnode *vp;
1188 	struct mount *mp;
1189 	struct vattr vattr;
1190 	int error;
1191 	int whiteout = 0;
1192 	struct nameidata nd;
1193 	int vfslocked;
1194 
1195 	AUDIT_ARG(mode, mode);
1196 	AUDIT_ARG(dev, dev);
1197 	switch (mode & S_IFMT) {
1198 	case S_IFCHR:
1199 	case S_IFBLK:
1200 		error = priv_check(td, PRIV_VFS_MKNOD_DEV);
1201 		break;
1202 	case S_IFMT:
1203 		error = priv_check(td, PRIV_VFS_MKNOD_BAD);
1204 		break;
1205 	case S_IFWHT:
1206 		error = priv_check(td, PRIV_VFS_MKNOD_WHT);
1207 		break;
1208 	default:
1209 		error = EINVAL;
1210 		break;
1211 	}
1212 	if (error)
1213 		return (error);
1214 restart:
1215 	bwillwrite();
1216 	NDINIT(&nd, CREATE, LOCKPARENT | SAVENAME | MPSAFE | AUDITVNODE1,
1217 	    pathseg, path, td);
1218 	if ((error = namei(&nd)) != 0)
1219 		return (error);
1220 	vfslocked = NDHASGIANT(&nd);
1221 	vp = nd.ni_vp;
1222 	if (vp != NULL) {
1223 		NDFREE(&nd, NDF_ONLY_PNBUF);
1224 		if (vp == nd.ni_dvp)
1225 			vrele(nd.ni_dvp);
1226 		else
1227 			vput(nd.ni_dvp);
1228 		vrele(vp);
1229 		VFS_UNLOCK_GIANT(vfslocked);
1230 		return (EEXIST);
1231 	} else {
1232 		VATTR_NULL(&vattr);
1233 		FILEDESC_SLOCK(td->td_proc->p_fd);
1234 		vattr.va_mode = (mode & ALLPERMS) &
1235 		    ~td->td_proc->p_fd->fd_cmask;
1236 		FILEDESC_SUNLOCK(td->td_proc->p_fd);
1237 		vattr.va_rdev = dev;
1238 		whiteout = 0;
1239 
1240 		switch (mode & S_IFMT) {
1241 		case S_IFMT:	/* used by badsect to flag bad sectors */
1242 			vattr.va_type = VBAD;
1243 			break;
1244 		case S_IFCHR:
1245 			vattr.va_type = VCHR;
1246 			break;
1247 		case S_IFBLK:
1248 			vattr.va_type = VBLK;
1249 			break;
1250 		case S_IFWHT:
1251 			whiteout = 1;
1252 			break;
1253 		default:
1254 			panic("kern_mknod: invalid mode");
1255 		}
1256 	}
1257 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1258 		NDFREE(&nd, NDF_ONLY_PNBUF);
1259 		vput(nd.ni_dvp);
1260 		VFS_UNLOCK_GIANT(vfslocked);
1261 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1262 			return (error);
1263 		goto restart;
1264 	}
1265 #ifdef MAC
1266 	if (error == 0 && !whiteout)
1267 		error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp,
1268 		    &nd.ni_cnd, &vattr);
1269 #endif
1270 	if (!error) {
1271 		VOP_LEASE(nd.ni_dvp, td, td->td_ucred, LEASE_WRITE);
1272 		if (whiteout)
1273 			error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE);
1274 		else {
1275 			error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp,
1276 						&nd.ni_cnd, &vattr);
1277 			if (error == 0)
1278 				vput(nd.ni_vp);
1279 		}
1280 	}
1281 	NDFREE(&nd, NDF_ONLY_PNBUF);
1282 	vput(nd.ni_dvp);
1283 	vn_finished_write(mp);
1284 	VFS_UNLOCK_GIANT(vfslocked);
1285 	return (error);
1286 }
1287 
1288 /*
1289  * Create a named pipe.
1290  */
1291 #ifndef _SYS_SYSPROTO_H_
1292 struct mkfifo_args {
1293 	char	*path;
1294 	int	mode;
1295 };
1296 #endif
1297 int
1298 mkfifo(td, uap)
1299 	struct thread *td;
1300 	register struct mkfifo_args /* {
1301 		char *path;
1302 		int mode;
1303 	} */ *uap;
1304 {
1305 
1306 	return (kern_mkfifo(td, uap->path, UIO_USERSPACE, uap->mode));
1307 }
1308 
1309 int
1310 kern_mkfifo(struct thread *td, char *path, enum uio_seg pathseg, int mode)
1311 {
1312 	struct mount *mp;
1313 	struct vattr vattr;
1314 	int error;
1315 	struct nameidata nd;
1316 	int vfslocked;
1317 
1318 	AUDIT_ARG(mode, mode);
1319 restart:
1320 	bwillwrite();
1321 	NDINIT(&nd, CREATE, LOCKPARENT | SAVENAME | MPSAFE | AUDITVNODE1,
1322 	    pathseg, path, td);
1323 	if ((error = namei(&nd)) != 0)
1324 		return (error);
1325 	vfslocked = NDHASGIANT(&nd);
1326 	if (nd.ni_vp != NULL) {
1327 		NDFREE(&nd, NDF_ONLY_PNBUF);
1328 		if (nd.ni_vp == nd.ni_dvp)
1329 			vrele(nd.ni_dvp);
1330 		else
1331 			vput(nd.ni_dvp);
1332 		vrele(nd.ni_vp);
1333 		VFS_UNLOCK_GIANT(vfslocked);
1334 		return (EEXIST);
1335 	}
1336 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1337 		NDFREE(&nd, NDF_ONLY_PNBUF);
1338 		vput(nd.ni_dvp);
1339 		VFS_UNLOCK_GIANT(vfslocked);
1340 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1341 			return (error);
1342 		goto restart;
1343 	}
1344 	VATTR_NULL(&vattr);
1345 	vattr.va_type = VFIFO;
1346 	FILEDESC_SLOCK(td->td_proc->p_fd);
1347 	vattr.va_mode = (mode & ALLPERMS) & ~td->td_proc->p_fd->fd_cmask;
1348 	FILEDESC_SUNLOCK(td->td_proc->p_fd);
1349 #ifdef MAC
1350 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1351 	    &vattr);
1352 	if (error)
1353 		goto out;
1354 #endif
1355 	VOP_LEASE(nd.ni_dvp, td, td->td_ucred, LEASE_WRITE);
1356 	error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
1357 	if (error == 0)
1358 		vput(nd.ni_vp);
1359 #ifdef MAC
1360 out:
1361 #endif
1362 	vput(nd.ni_dvp);
1363 	vn_finished_write(mp);
1364 	VFS_UNLOCK_GIANT(vfslocked);
1365 	NDFREE(&nd, NDF_ONLY_PNBUF);
1366 	return (error);
1367 }
1368 
1369 /*
1370  * Make a hard file link.
1371  */
1372 #ifndef _SYS_SYSPROTO_H_
1373 struct link_args {
1374 	char	*path;
1375 	char	*link;
1376 };
1377 #endif
1378 int
1379 link(td, uap)
1380 	struct thread *td;
1381 	register struct link_args /* {
1382 		char *path;
1383 		char *link;
1384 	} */ *uap;
1385 {
1386 	int error;
1387 
1388 	error = kern_link(td, uap->path, uap->link, UIO_USERSPACE);
1389 	return (error);
1390 }
1391 
1392 static int hardlink_check_uid = 0;
1393 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_uid, CTLFLAG_RW,
1394     &hardlink_check_uid, 0,
1395     "Unprivileged processes cannot create hard links to files owned by other "
1396     "users");
1397 static int hardlink_check_gid = 0;
1398 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_gid, CTLFLAG_RW,
1399     &hardlink_check_gid, 0,
1400     "Unprivileged processes cannot create hard links to files owned by other "
1401     "groups");
1402 
1403 static int
1404 can_hardlink(struct vnode *vp, struct thread *td, struct ucred *cred)
1405 {
1406 	struct vattr va;
1407 	int error;
1408 
1409 	if (!hardlink_check_uid && !hardlink_check_gid)
1410 		return (0);
1411 
1412 	error = VOP_GETATTR(vp, &va, cred, td);
1413 	if (error != 0)
1414 		return (error);
1415 
1416 	if (hardlink_check_uid && cred->cr_uid != va.va_uid) {
1417 		error = priv_check_cred(cred, PRIV_VFS_LINK, 0);
1418 		if (error)
1419 			return (error);
1420 	}
1421 
1422 	if (hardlink_check_gid && !groupmember(va.va_gid, cred)) {
1423 		error = priv_check_cred(cred, PRIV_VFS_LINK, 0);
1424 		if (error)
1425 			return (error);
1426 	}
1427 
1428 	return (0);
1429 }
1430 
1431 int
1432 kern_link(struct thread *td, char *path, char *link, enum uio_seg segflg)
1433 {
1434 	struct vnode *vp;
1435 	struct mount *mp;
1436 	struct nameidata nd;
1437 	int vfslocked;
1438 	int lvfslocked;
1439 	int error;
1440 
1441 	bwillwrite();
1442 	NDINIT(&nd, LOOKUP, FOLLOW | MPSAFE | AUDITVNODE1, segflg, path, td);
1443 	if ((error = namei(&nd)) != 0)
1444 		return (error);
1445 	vfslocked = NDHASGIANT(&nd);
1446 	NDFREE(&nd, NDF_ONLY_PNBUF);
1447 	vp = nd.ni_vp;
1448 	if (vp->v_type == VDIR) {
1449 		vrele(vp);
1450 		VFS_UNLOCK_GIANT(vfslocked);
1451 		return (EPERM);		/* POSIX */
1452 	}
1453 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) {
1454 		vrele(vp);
1455 		VFS_UNLOCK_GIANT(vfslocked);
1456 		return (error);
1457 	}
1458 	NDINIT(&nd, CREATE, LOCKPARENT | SAVENAME | MPSAFE | AUDITVNODE2,
1459 	    segflg, link, td);
1460 	if ((error = namei(&nd)) == 0) {
1461 		lvfslocked = NDHASGIANT(&nd);
1462 		if (nd.ni_vp != NULL) {
1463 			if (nd.ni_dvp == nd.ni_vp)
1464 				vrele(nd.ni_dvp);
1465 			else
1466 				vput(nd.ni_dvp);
1467 			vrele(nd.ni_vp);
1468 			error = EEXIST;
1469 		} else if ((error = vn_lock(vp, LK_EXCLUSIVE | LK_RETRY))
1470 		    == 0) {
1471 			VOP_LEASE(nd.ni_dvp, td, td->td_ucred, LEASE_WRITE);
1472 			VOP_LEASE(vp, td, td->td_ucred, LEASE_WRITE);
1473 			error = can_hardlink(vp, td, td->td_ucred);
1474 			if (error == 0)
1475 #ifdef MAC
1476 				error = mac_vnode_check_link(td->td_ucred,
1477 				    nd.ni_dvp, vp, &nd.ni_cnd);
1478 			if (error == 0)
1479 #endif
1480 				error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
1481 			VOP_UNLOCK(vp, 0);
1482 			vput(nd.ni_dvp);
1483 		}
1484 		NDFREE(&nd, NDF_ONLY_PNBUF);
1485 		VFS_UNLOCK_GIANT(lvfslocked);
1486 	}
1487 	vrele(vp);
1488 	vn_finished_write(mp);
1489 	VFS_UNLOCK_GIANT(vfslocked);
1490 	return (error);
1491 }
1492 
1493 /*
1494  * Make a symbolic link.
1495  */
1496 #ifndef _SYS_SYSPROTO_H_
1497 struct symlink_args {
1498 	char	*path;
1499 	char	*link;
1500 };
1501 #endif
1502 int
1503 symlink(td, uap)
1504 	struct thread *td;
1505 	register struct symlink_args /* {
1506 		char *path;
1507 		char *link;
1508 	} */ *uap;
1509 {
1510 
1511 	return (kern_symlink(td, uap->path, uap->link, UIO_USERSPACE));
1512 }
1513 
1514 int
1515 kern_symlink(struct thread *td, char *path, char *link, enum uio_seg segflg)
1516 {
1517 	struct mount *mp;
1518 	struct vattr vattr;
1519 	char *syspath;
1520 	int error;
1521 	struct nameidata nd;
1522 	int vfslocked;
1523 
1524 	if (segflg == UIO_SYSSPACE) {
1525 		syspath = path;
1526 	} else {
1527 		syspath = uma_zalloc(namei_zone, M_WAITOK);
1528 		if ((error = copyinstr(path, syspath, MAXPATHLEN, NULL)) != 0)
1529 			goto out;
1530 	}
1531 	AUDIT_ARG(text, syspath);
1532 restart:
1533 	bwillwrite();
1534 	NDINIT(&nd, CREATE, LOCKPARENT | SAVENAME | MPSAFE | AUDITVNODE1,
1535 	    segflg, link, td);
1536 	if ((error = namei(&nd)) != 0)
1537 		goto out;
1538 	vfslocked = NDHASGIANT(&nd);
1539 	if (nd.ni_vp) {
1540 		NDFREE(&nd, NDF_ONLY_PNBUF);
1541 		if (nd.ni_vp == nd.ni_dvp)
1542 			vrele(nd.ni_dvp);
1543 		else
1544 			vput(nd.ni_dvp);
1545 		vrele(nd.ni_vp);
1546 		VFS_UNLOCK_GIANT(vfslocked);
1547 		error = EEXIST;
1548 		goto out;
1549 	}
1550 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1551 		NDFREE(&nd, NDF_ONLY_PNBUF);
1552 		vput(nd.ni_dvp);
1553 		VFS_UNLOCK_GIANT(vfslocked);
1554 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1555 			goto out;
1556 		goto restart;
1557 	}
1558 	VATTR_NULL(&vattr);
1559 	FILEDESC_SLOCK(td->td_proc->p_fd);
1560 	vattr.va_mode = ACCESSPERMS &~ td->td_proc->p_fd->fd_cmask;
1561 	FILEDESC_SUNLOCK(td->td_proc->p_fd);
1562 #ifdef MAC
1563 	vattr.va_type = VLNK;
1564 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1565 	    &vattr);
1566 	if (error)
1567 		goto out2;
1568 #endif
1569 	VOP_LEASE(nd.ni_dvp, td, td->td_ucred, LEASE_WRITE);
1570 	error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, syspath);
1571 	if (error == 0)
1572 		vput(nd.ni_vp);
1573 #ifdef MAC
1574 out2:
1575 #endif
1576 	NDFREE(&nd, NDF_ONLY_PNBUF);
1577 	vput(nd.ni_dvp);
1578 	vn_finished_write(mp);
1579 	VFS_UNLOCK_GIANT(vfslocked);
1580 out:
1581 	if (segflg != UIO_SYSSPACE)
1582 		uma_zfree(namei_zone, syspath);
1583 	return (error);
1584 }
1585 
1586 /*
1587  * Delete a whiteout from the filesystem.
1588  */
1589 int
1590 undelete(td, uap)
1591 	struct thread *td;
1592 	register struct undelete_args /* {
1593 		char *path;
1594 	} */ *uap;
1595 {
1596 	int error;
1597 	struct mount *mp;
1598 	struct nameidata nd;
1599 	int vfslocked;
1600 
1601 restart:
1602 	bwillwrite();
1603 	NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | MPSAFE | AUDITVNODE1,
1604 	    UIO_USERSPACE, uap->path, td);
1605 	error = namei(&nd);
1606 	if (error)
1607 		return (error);
1608 	vfslocked = NDHASGIANT(&nd);
1609 
1610 	if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) {
1611 		NDFREE(&nd, NDF_ONLY_PNBUF);
1612 		if (nd.ni_vp == nd.ni_dvp)
1613 			vrele(nd.ni_dvp);
1614 		else
1615 			vput(nd.ni_dvp);
1616 		if (nd.ni_vp)
1617 			vrele(nd.ni_vp);
1618 		VFS_UNLOCK_GIANT(vfslocked);
1619 		return (EEXIST);
1620 	}
1621 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1622 		NDFREE(&nd, NDF_ONLY_PNBUF);
1623 		vput(nd.ni_dvp);
1624 		VFS_UNLOCK_GIANT(vfslocked);
1625 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1626 			return (error);
1627 		goto restart;
1628 	}
1629 	VOP_LEASE(nd.ni_dvp, td, td->td_ucred, LEASE_WRITE);
1630 	error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE);
1631 	NDFREE(&nd, NDF_ONLY_PNBUF);
1632 	vput(nd.ni_dvp);
1633 	vn_finished_write(mp);
1634 	VFS_UNLOCK_GIANT(vfslocked);
1635 	return (error);
1636 }
1637 
1638 /*
1639  * Delete a name from the filesystem.
1640  */
1641 #ifndef _SYS_SYSPROTO_H_
1642 struct unlink_args {
1643 	char	*path;
1644 };
1645 #endif
1646 int
1647 unlink(td, uap)
1648 	struct thread *td;
1649 	struct unlink_args /* {
1650 		char *path;
1651 	} */ *uap;
1652 {
1653 	int error;
1654 
1655 	error = kern_unlink(td, uap->path, UIO_USERSPACE);
1656 	return (error);
1657 }
1658 
1659 int
1660 kern_unlink(struct thread *td, char *path, enum uio_seg pathseg)
1661 {
1662 	struct mount *mp;
1663 	struct vnode *vp;
1664 	int error;
1665 	struct nameidata nd;
1666 	int vfslocked;
1667 
1668 restart:
1669 	bwillwrite();
1670 	NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF | MPSAFE | AUDITVNODE1,
1671 	    pathseg, path, td);
1672 	if ((error = namei(&nd)) != 0)
1673 		return (error == EINVAL ? EPERM : error);
1674 	vfslocked = NDHASGIANT(&nd);
1675 	vp = nd.ni_vp;
1676 	if (vp->v_type == VDIR)
1677 		error = EPERM;		/* POSIX */
1678 	else {
1679 		/*
1680 		 * The root of a mounted filesystem cannot be deleted.
1681 		 *
1682 		 * XXX: can this only be a VDIR case?
1683 		 */
1684 		if (vp->v_vflag & VV_ROOT)
1685 			error = EBUSY;
1686 	}
1687 	if (error == 0) {
1688 		if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1689 			NDFREE(&nd, NDF_ONLY_PNBUF);
1690 			vput(nd.ni_dvp);
1691 			if (vp == nd.ni_dvp)
1692 				vrele(vp);
1693 			else
1694 				vput(vp);
1695 			VFS_UNLOCK_GIANT(vfslocked);
1696 			if ((error = vn_start_write(NULL, &mp,
1697 			    V_XSLEEP | PCATCH)) != 0)
1698 				return (error);
1699 			goto restart;
1700 		}
1701 #ifdef MAC
1702 		error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
1703 		    &nd.ni_cnd);
1704 		if (error)
1705 			goto out;
1706 #endif
1707 		VOP_LEASE(nd.ni_dvp, td, td->td_ucred, LEASE_WRITE);
1708 		error = VOP_REMOVE(nd.ni_dvp, vp, &nd.ni_cnd);
1709 #ifdef MAC
1710 out:
1711 #endif
1712 		vn_finished_write(mp);
1713 	}
1714 	NDFREE(&nd, NDF_ONLY_PNBUF);
1715 	vput(nd.ni_dvp);
1716 	if (vp == nd.ni_dvp)
1717 		vrele(vp);
1718 	else
1719 		vput(vp);
1720 	VFS_UNLOCK_GIANT(vfslocked);
1721 	return (error);
1722 }
1723 
1724 /*
1725  * Reposition read/write file offset.
1726  */
1727 #ifndef _SYS_SYSPROTO_H_
1728 struct lseek_args {
1729 	int	fd;
1730 	int	pad;
1731 	off_t	offset;
1732 	int	whence;
1733 };
1734 #endif
1735 int
1736 lseek(td, uap)
1737 	struct thread *td;
1738 	register struct lseek_args /* {
1739 		int fd;
1740 		int pad;
1741 		off_t offset;
1742 		int whence;
1743 	} */ *uap;
1744 {
1745 	struct ucred *cred = td->td_ucred;
1746 	struct file *fp;
1747 	struct vnode *vp;
1748 	struct vattr vattr;
1749 	off_t offset;
1750 	int error, noneg;
1751 	int vfslocked;
1752 
1753 	if ((error = fget(td, uap->fd, &fp)) != 0)
1754 		return (error);
1755 	if (!(fp->f_ops->fo_flags & DFLAG_SEEKABLE)) {
1756 		fdrop(fp, td);
1757 		return (ESPIPE);
1758 	}
1759 	vp = fp->f_vnode;
1760 	vfslocked = VFS_LOCK_GIANT(vp->v_mount);
1761 	noneg = (vp->v_type != VCHR);
1762 	offset = uap->offset;
1763 	switch (uap->whence) {
1764 	case L_INCR:
1765 		if (noneg &&
1766 		    (fp->f_offset < 0 ||
1767 		    (offset > 0 && fp->f_offset > OFF_MAX - offset))) {
1768 			error = EOVERFLOW;
1769 			break;
1770 		}
1771 		offset += fp->f_offset;
1772 		break;
1773 	case L_XTND:
1774 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1775 		error = VOP_GETATTR(vp, &vattr, cred, td);
1776 		VOP_UNLOCK(vp, 0);
1777 		if (error)
1778 			break;
1779 		if (noneg &&
1780 		    (vattr.va_size > OFF_MAX ||
1781 		    (offset > 0 && vattr.va_size > OFF_MAX - offset))) {
1782 			error = EOVERFLOW;
1783 			break;
1784 		}
1785 		offset += vattr.va_size;
1786 		break;
1787 	case L_SET:
1788 		break;
1789 	case SEEK_DATA:
1790 		error = fo_ioctl(fp, FIOSEEKDATA, &offset, cred, td);
1791 		break;
1792 	case SEEK_HOLE:
1793 		error = fo_ioctl(fp, FIOSEEKHOLE, &offset, cred, td);
1794 		break;
1795 	default:
1796 		error = EINVAL;
1797 	}
1798 	if (error == 0 && noneg && offset < 0)
1799 		error = EINVAL;
1800 	if (error != 0)
1801 		goto drop;
1802 	fp->f_offset = offset;
1803 	*(off_t *)(td->td_retval) = fp->f_offset;
1804 drop:
1805 	fdrop(fp, td);
1806 	VFS_UNLOCK_GIANT(vfslocked);
1807 	return (error);
1808 }
1809 
1810 #if defined(COMPAT_43)
1811 /*
1812  * Reposition read/write file offset.
1813  */
1814 #ifndef _SYS_SYSPROTO_H_
1815 struct olseek_args {
1816 	int	fd;
1817 	long	offset;
1818 	int	whence;
1819 };
1820 #endif
1821 int
1822 olseek(td, uap)
1823 	struct thread *td;
1824 	register struct olseek_args /* {
1825 		int fd;
1826 		long offset;
1827 		int whence;
1828 	} */ *uap;
1829 {
1830 	struct lseek_args /* {
1831 		int fd;
1832 		int pad;
1833 		off_t offset;
1834 		int whence;
1835 	} */ nuap;
1836 
1837 	nuap.fd = uap->fd;
1838 	nuap.offset = uap->offset;
1839 	nuap.whence = uap->whence;
1840 	return (lseek(td, &nuap));
1841 }
1842 #endif /* COMPAT_43 */
1843 
1844 /* Version with the 'pad' argument */
1845 int
1846 freebsd6_lseek(td, uap)
1847 	struct thread *td;
1848 	register struct freebsd6_lseek_args *uap;
1849 {
1850 	struct lseek_args ouap;
1851 
1852 	ouap.fd = uap->fd;
1853 	ouap.offset = uap->offset;
1854 	ouap.whence = uap->whence;
1855 	return (lseek(td, &ouap));
1856 }
1857 
1858 /*
1859  * Check access permissions using passed credentials.
1860  */
1861 static int
1862 vn_access(vp, user_flags, cred, td)
1863 	struct vnode	*vp;
1864 	int		user_flags;
1865 	struct ucred	*cred;
1866 	struct thread	*td;
1867 {
1868 	int error, flags;
1869 
1870 	/* Flags == 0 means only check for existence. */
1871 	error = 0;
1872 	if (user_flags) {
1873 		flags = 0;
1874 		if (user_flags & R_OK)
1875 			flags |= VREAD;
1876 		if (user_flags & W_OK)
1877 			flags |= VWRITE;
1878 		if (user_flags & X_OK)
1879 			flags |= VEXEC;
1880 #ifdef MAC
1881 		error = mac_vnode_check_access(cred, vp, flags);
1882 		if (error)
1883 			return (error);
1884 #endif
1885 		if ((flags & VWRITE) == 0 || (error = vn_writechk(vp)) == 0)
1886 			error = VOP_ACCESS(vp, flags, cred, td);
1887 	}
1888 	return (error);
1889 }
1890 
1891 /*
1892  * Check access permissions using "real" credentials.
1893  */
1894 #ifndef _SYS_SYSPROTO_H_
1895 struct access_args {
1896 	char	*path;
1897 	int	flags;
1898 };
1899 #endif
1900 int
1901 access(td, uap)
1902 	struct thread *td;
1903 	register struct access_args /* {
1904 		char *path;
1905 		int flags;
1906 	} */ *uap;
1907 {
1908 
1909 	return (kern_access(td, uap->path, UIO_USERSPACE, uap->flags));
1910 }
1911 
1912 int
1913 kern_access(struct thread *td, char *path, enum uio_seg pathseg, int flags)
1914 {
1915 	struct ucred *cred, *tmpcred;
1916 	register struct vnode *vp;
1917 	struct nameidata nd;
1918 	int vfslocked;
1919 	int error;
1920 
1921 	/*
1922 	 * Create and modify a temporary credential instead of one that
1923 	 * is potentially shared.  This could also mess up socket
1924 	 * buffer accounting which can run in an interrupt context.
1925 	 */
1926 	cred = td->td_ucred;
1927 	tmpcred = crdup(cred);
1928 	tmpcred->cr_uid = cred->cr_ruid;
1929 	tmpcred->cr_groups[0] = cred->cr_rgid;
1930 	td->td_ucred = tmpcred;
1931 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | MPSAFE | AUDITVNODE1,
1932 	    pathseg, path, td);
1933 	if ((error = namei(&nd)) != 0)
1934 		goto out1;
1935 	vfslocked = NDHASGIANT(&nd);
1936 	vp = nd.ni_vp;
1937 
1938 	error = vn_access(vp, flags, tmpcred, td);
1939 	NDFREE(&nd, NDF_ONLY_PNBUF);
1940 	vput(vp);
1941 	VFS_UNLOCK_GIANT(vfslocked);
1942 out1:
1943 	td->td_ucred = cred;
1944 	crfree(tmpcred);
1945 	return (error);
1946 }
1947 
1948 /*
1949  * Check access permissions using "effective" credentials.
1950  */
1951 #ifndef _SYS_SYSPROTO_H_
1952 struct eaccess_args {
1953 	char	*path;
1954 	int	flags;
1955 };
1956 #endif
1957 int
1958 eaccess(td, uap)
1959 	struct thread *td;
1960 	register struct eaccess_args /* {
1961 		char *path;
1962 		int flags;
1963 	} */ *uap;
1964 {
1965 
1966 	return (kern_eaccess(td, uap->path, UIO_USERSPACE, uap->flags));
1967 }
1968 
1969 int
1970 kern_eaccess(struct thread *td, char *path, enum uio_seg pathseg, int flags)
1971 {
1972 	struct nameidata nd;
1973 	struct vnode *vp;
1974 	int vfslocked;
1975 	int error;
1976 
1977 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | MPSAFE | AUDITVNODE1,
1978 	    pathseg, path, td);
1979 	if ((error = namei(&nd)) != 0)
1980 		return (error);
1981 	vp = nd.ni_vp;
1982 	vfslocked = NDHASGIANT(&nd);
1983 	error = vn_access(vp, flags, td->td_ucred, td);
1984 	NDFREE(&nd, NDF_ONLY_PNBUF);
1985 	vput(vp);
1986 	VFS_UNLOCK_GIANT(vfslocked);
1987 	return (error);
1988 }
1989 
1990 #if defined(COMPAT_43)
1991 /*
1992  * Get file status; this version follows links.
1993  */
1994 #ifndef _SYS_SYSPROTO_H_
1995 struct ostat_args {
1996 	char	*path;
1997 	struct ostat *ub;
1998 };
1999 #endif
2000 int
2001 ostat(td, uap)
2002 	struct thread *td;
2003 	register struct ostat_args /* {
2004 		char *path;
2005 		struct ostat *ub;
2006 	} */ *uap;
2007 {
2008 	struct stat sb;
2009 	struct ostat osb;
2010 	int error;
2011 
2012 	error = kern_stat(td, uap->path, UIO_USERSPACE, &sb);
2013 	if (error)
2014 		return (error);
2015 	cvtstat(&sb, &osb);
2016 	error = copyout(&osb, uap->ub, sizeof (osb));
2017 	return (error);
2018 }
2019 
2020 /*
2021  * Get file status; this version does not follow links.
2022  */
2023 #ifndef _SYS_SYSPROTO_H_
2024 struct olstat_args {
2025 	char	*path;
2026 	struct ostat *ub;
2027 };
2028 #endif
2029 int
2030 olstat(td, uap)
2031 	struct thread *td;
2032 	register struct olstat_args /* {
2033 		char *path;
2034 		struct ostat *ub;
2035 	} */ *uap;
2036 {
2037 	struct stat sb;
2038 	struct ostat osb;
2039 	int error;
2040 
2041 	error = kern_lstat(td, uap->path, UIO_USERSPACE, &sb);
2042 	if (error)
2043 		return (error);
2044 	cvtstat(&sb, &osb);
2045 	error = copyout(&osb, uap->ub, sizeof (osb));
2046 	return (error);
2047 }
2048 
2049 /*
2050  * Convert from an old to a new stat structure.
2051  */
2052 void
2053 cvtstat(st, ost)
2054 	struct stat *st;
2055 	struct ostat *ost;
2056 {
2057 
2058 	ost->st_dev = st->st_dev;
2059 	ost->st_ino = st->st_ino;
2060 	ost->st_mode = st->st_mode;
2061 	ost->st_nlink = st->st_nlink;
2062 	ost->st_uid = st->st_uid;
2063 	ost->st_gid = st->st_gid;
2064 	ost->st_rdev = st->st_rdev;
2065 	if (st->st_size < (quad_t)1 << 32)
2066 		ost->st_size = st->st_size;
2067 	else
2068 		ost->st_size = -2;
2069 	ost->st_atime = st->st_atime;
2070 	ost->st_mtime = st->st_mtime;
2071 	ost->st_ctime = st->st_ctime;
2072 	ost->st_blksize = st->st_blksize;
2073 	ost->st_blocks = st->st_blocks;
2074 	ost->st_flags = st->st_flags;
2075 	ost->st_gen = st->st_gen;
2076 }
2077 #endif /* COMPAT_43 */
2078 
2079 /*
2080  * Get file status; this version follows links.
2081  */
2082 #ifndef _SYS_SYSPROTO_H_
2083 struct stat_args {
2084 	char	*path;
2085 	struct stat *ub;
2086 };
2087 #endif
2088 int
2089 stat(td, uap)
2090 	struct thread *td;
2091 	register struct stat_args /* {
2092 		char *path;
2093 		struct stat *ub;
2094 	} */ *uap;
2095 {
2096 	struct stat sb;
2097 	int error;
2098 
2099 	error = kern_stat(td, uap->path, UIO_USERSPACE, &sb);
2100 	if (error == 0)
2101 		error = copyout(&sb, uap->ub, sizeof (sb));
2102 	return (error);
2103 }
2104 
2105 int
2106 kern_stat(struct thread *td, char *path, enum uio_seg pathseg, struct stat *sbp)
2107 {
2108 	struct nameidata nd;
2109 	struct stat sb;
2110 	int error, vfslocked;
2111 
2112 	NDINIT(&nd, LOOKUP,
2113 	    FOLLOW | LOCKSHARED | LOCKLEAF | MPSAFE | AUDITVNODE1,
2114 	    pathseg, path, td);
2115 	if ((error = namei(&nd)) != 0)
2116 		return (error);
2117 	vfslocked = NDHASGIANT(&nd);
2118 	error = vn_stat(nd.ni_vp, &sb, td->td_ucred, NOCRED, td);
2119 	NDFREE(&nd, NDF_ONLY_PNBUF);
2120 	vput(nd.ni_vp);
2121 	VFS_UNLOCK_GIANT(vfslocked);
2122 	if (mtx_owned(&Giant))
2123 		printf("stat(%d): %s\n", vfslocked, path);
2124 	if (error)
2125 		return (error);
2126 	*sbp = sb;
2127 #ifdef KTRACE
2128 	if (KTRPOINT(td, KTR_STRUCT))
2129 		ktrstat(&sb);
2130 #endif
2131 	return (0);
2132 }
2133 
2134 /*
2135  * Get file status; this version does not follow links.
2136  */
2137 #ifndef _SYS_SYSPROTO_H_
2138 struct lstat_args {
2139 	char	*path;
2140 	struct stat *ub;
2141 };
2142 #endif
2143 int
2144 lstat(td, uap)
2145 	struct thread *td;
2146 	register struct lstat_args /* {
2147 		char *path;
2148 		struct stat *ub;
2149 	} */ *uap;
2150 {
2151 	struct stat sb;
2152 	int error;
2153 
2154 	error = kern_lstat(td, uap->path, UIO_USERSPACE, &sb);
2155 	if (error == 0)
2156 		error = copyout(&sb, uap->ub, sizeof (sb));
2157 	return (error);
2158 }
2159 
2160 int
2161 kern_lstat(struct thread *td, char *path, enum uio_seg pathseg, struct stat *sbp)
2162 {
2163 	struct vnode *vp;
2164 	struct stat sb;
2165 	struct nameidata nd;
2166 	int error, vfslocked;
2167 
2168 	NDINIT(&nd, LOOKUP,
2169 	    NOFOLLOW | LOCKLEAF | LOCKSHARED | MPSAFE | AUDITVNODE1,
2170 	    pathseg, path, td);
2171 	if ((error = namei(&nd)) != 0)
2172 		return (error);
2173 	vfslocked = NDHASGIANT(&nd);
2174 	vp = nd.ni_vp;
2175 	error = vn_stat(vp, &sb, td->td_ucred, NOCRED, td);
2176 	NDFREE(&nd, NDF_ONLY_PNBUF);
2177 	vput(vp);
2178 	VFS_UNLOCK_GIANT(vfslocked);
2179 	if (error)
2180 		return (error);
2181 	*sbp = sb;
2182 #ifdef KTRACE
2183 	if (KTRPOINT(td, KTR_STRUCT))
2184 		ktrstat(&sb);
2185 #endif
2186 	return (0);
2187 }
2188 
2189 /*
2190  * Implementation of the NetBSD [l]stat() functions.
2191  */
2192 void
2193 cvtnstat(sb, nsb)
2194 	struct stat *sb;
2195 	struct nstat *nsb;
2196 {
2197 	bzero(nsb, sizeof *nsb);
2198 	nsb->st_dev = sb->st_dev;
2199 	nsb->st_ino = sb->st_ino;
2200 	nsb->st_mode = sb->st_mode;
2201 	nsb->st_nlink = sb->st_nlink;
2202 	nsb->st_uid = sb->st_uid;
2203 	nsb->st_gid = sb->st_gid;
2204 	nsb->st_rdev = sb->st_rdev;
2205 	nsb->st_atimespec = sb->st_atimespec;
2206 	nsb->st_mtimespec = sb->st_mtimespec;
2207 	nsb->st_ctimespec = sb->st_ctimespec;
2208 	nsb->st_size = sb->st_size;
2209 	nsb->st_blocks = sb->st_blocks;
2210 	nsb->st_blksize = sb->st_blksize;
2211 	nsb->st_flags = sb->st_flags;
2212 	nsb->st_gen = sb->st_gen;
2213 	nsb->st_birthtimespec = sb->st_birthtimespec;
2214 }
2215 
2216 #ifndef _SYS_SYSPROTO_H_
2217 struct nstat_args {
2218 	char	*path;
2219 	struct nstat *ub;
2220 };
2221 #endif
2222 int
2223 nstat(td, uap)
2224 	struct thread *td;
2225 	register struct nstat_args /* {
2226 		char *path;
2227 		struct nstat *ub;
2228 	} */ *uap;
2229 {
2230 	struct stat sb;
2231 	struct nstat nsb;
2232 	int error;
2233 
2234 	error = kern_stat(td, uap->path, UIO_USERSPACE, &sb);
2235 	if (error)
2236 		return (error);
2237 	cvtnstat(&sb, &nsb);
2238 	error = copyout(&nsb, uap->ub, sizeof (nsb));
2239 	return (error);
2240 }
2241 
2242 /*
2243  * NetBSD lstat.  Get file status; this version does not follow links.
2244  */
2245 #ifndef _SYS_SYSPROTO_H_
2246 struct lstat_args {
2247 	char	*path;
2248 	struct stat *ub;
2249 };
2250 #endif
2251 int
2252 nlstat(td, uap)
2253 	struct thread *td;
2254 	register struct nlstat_args /* {
2255 		char *path;
2256 		struct nstat *ub;
2257 	} */ *uap;
2258 {
2259 	struct stat sb;
2260 	struct nstat nsb;
2261 	int error;
2262 
2263 	error = kern_lstat(td, uap->path, UIO_USERSPACE, &sb);
2264 	if (error)
2265 		return (error);
2266 	cvtnstat(&sb, &nsb);
2267 	error = copyout(&nsb, uap->ub, sizeof (nsb));
2268 	return (error);
2269 }
2270 
2271 /*
2272  * Get configurable pathname variables.
2273  */
2274 #ifndef _SYS_SYSPROTO_H_
2275 struct pathconf_args {
2276 	char	*path;
2277 	int	name;
2278 };
2279 #endif
2280 int
2281 pathconf(td, uap)
2282 	struct thread *td;
2283 	register struct pathconf_args /* {
2284 		char *path;
2285 		int name;
2286 	} */ *uap;
2287 {
2288 
2289 	return (kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name));
2290 }
2291 
2292 int
2293 kern_pathconf(struct thread *td, char *path, enum uio_seg pathseg, int name)
2294 {
2295 	struct nameidata nd;
2296 	int error, vfslocked;
2297 
2298 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | MPSAFE | AUDITVNODE1,
2299 	    pathseg, path, td);
2300 	if ((error = namei(&nd)) != 0)
2301 		return (error);
2302 	vfslocked = NDHASGIANT(&nd);
2303 	NDFREE(&nd, NDF_ONLY_PNBUF);
2304 
2305 	/* If asynchronous I/O is available, it works for all files. */
2306 	if (name == _PC_ASYNC_IO)
2307 		td->td_retval[0] = async_io_version;
2308 	else
2309 		error = VOP_PATHCONF(nd.ni_vp, name, td->td_retval);
2310 	vput(nd.ni_vp);
2311 	VFS_UNLOCK_GIANT(vfslocked);
2312 	return (error);
2313 }
2314 
2315 /*
2316  * Return target name of a symbolic link.
2317  */
2318 #ifndef _SYS_SYSPROTO_H_
2319 struct readlink_args {
2320 	char	*path;
2321 	char	*buf;
2322 	size_t	count;
2323 };
2324 #endif
2325 int
2326 readlink(td, uap)
2327 	struct thread *td;
2328 	register struct readlink_args /* {
2329 		char *path;
2330 		char *buf;
2331 		size_t count;
2332 	} */ *uap;
2333 {
2334 
2335 	return (kern_readlink(td, uap->path, UIO_USERSPACE, uap->buf,
2336 	    UIO_USERSPACE, uap->count));
2337 }
2338 
2339 int
2340 kern_readlink(struct thread *td, char *path, enum uio_seg pathseg, char *buf,
2341     enum uio_seg bufseg, size_t count)
2342 {
2343 	register struct vnode *vp;
2344 	struct iovec aiov;
2345 	struct uio auio;
2346 	int error;
2347 	struct nameidata nd;
2348 	int vfslocked;
2349 
2350 	NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | MPSAFE | AUDITVNODE1,
2351 	    pathseg, path, td);
2352 	if ((error = namei(&nd)) != 0)
2353 		return (error);
2354 	NDFREE(&nd, NDF_ONLY_PNBUF);
2355 	vfslocked = NDHASGIANT(&nd);
2356 	vp = nd.ni_vp;
2357 #ifdef MAC
2358 	error = mac_vnode_check_readlink(td->td_ucred, vp);
2359 	if (error) {
2360 		vput(vp);
2361 		VFS_UNLOCK_GIANT(vfslocked);
2362 		return (error);
2363 	}
2364 #endif
2365 	if (vp->v_type != VLNK)
2366 		error = EINVAL;
2367 	else {
2368 		aiov.iov_base = buf;
2369 		aiov.iov_len = count;
2370 		auio.uio_iov = &aiov;
2371 		auio.uio_iovcnt = 1;
2372 		auio.uio_offset = 0;
2373 		auio.uio_rw = UIO_READ;
2374 		auio.uio_segflg = bufseg;
2375 		auio.uio_td = td;
2376 		auio.uio_resid = count;
2377 		error = VOP_READLINK(vp, &auio, td->td_ucred);
2378 	}
2379 	vput(vp);
2380 	VFS_UNLOCK_GIANT(vfslocked);
2381 	td->td_retval[0] = count - auio.uio_resid;
2382 	return (error);
2383 }
2384 
2385 /*
2386  * Common implementation code for chflags() and fchflags().
2387  */
2388 static int
2389 setfflags(td, vp, flags)
2390 	struct thread *td;
2391 	struct vnode *vp;
2392 	int flags;
2393 {
2394 	int error;
2395 	struct mount *mp;
2396 	struct vattr vattr;
2397 
2398 	/*
2399 	 * Prevent non-root users from setting flags on devices.  When
2400 	 * a device is reused, users can retain ownership of the device
2401 	 * if they are allowed to set flags and programs assume that
2402 	 * chown can't fail when done as root.
2403 	 */
2404 	if (vp->v_type == VCHR || vp->v_type == VBLK) {
2405 		error = priv_check(td, PRIV_VFS_CHFLAGS_DEV);
2406 		if (error)
2407 			return (error);
2408 	}
2409 
2410 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2411 		return (error);
2412 	VOP_LEASE(vp, td, td->td_ucred, LEASE_WRITE);
2413 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2414 	VATTR_NULL(&vattr);
2415 	vattr.va_flags = flags;
2416 #ifdef MAC
2417 	error = mac_vnode_check_setflags(td->td_ucred, vp, vattr.va_flags);
2418 	if (error == 0)
2419 #endif
2420 		error = VOP_SETATTR(vp, &vattr, td->td_ucred, td);
2421 	VOP_UNLOCK(vp, 0);
2422 	vn_finished_write(mp);
2423 	return (error);
2424 }
2425 
2426 /*
2427  * Change flags of a file given a path name.
2428  */
2429 #ifndef _SYS_SYSPROTO_H_
2430 struct chflags_args {
2431 	char	*path;
2432 	int	flags;
2433 };
2434 #endif
2435 int
2436 chflags(td, uap)
2437 	struct thread *td;
2438 	register struct chflags_args /* {
2439 		char *path;
2440 		int flags;
2441 	} */ *uap;
2442 {
2443 	int error;
2444 	struct nameidata nd;
2445 	int vfslocked;
2446 
2447 	AUDIT_ARG(fflags, uap->flags);
2448 	NDINIT(&nd, LOOKUP, FOLLOW | MPSAFE | AUDITVNODE1, UIO_USERSPACE,
2449 	    uap->path, td);
2450 	if ((error = namei(&nd)) != 0)
2451 		return (error);
2452 	NDFREE(&nd, NDF_ONLY_PNBUF);
2453 	vfslocked = NDHASGIANT(&nd);
2454 	error = setfflags(td, nd.ni_vp, uap->flags);
2455 	vrele(nd.ni_vp);
2456 	VFS_UNLOCK_GIANT(vfslocked);
2457 	return (error);
2458 }
2459 
2460 /*
2461  * Same as chflags() but doesn't follow symlinks.
2462  */
2463 int
2464 lchflags(td, uap)
2465 	struct thread *td;
2466 	register struct lchflags_args /* {
2467 		char *path;
2468 		int flags;
2469 	} */ *uap;
2470 {
2471 	int error;
2472 	struct nameidata nd;
2473 	int vfslocked;
2474 
2475 	AUDIT_ARG(fflags, uap->flags);
2476 	NDINIT(&nd, LOOKUP, NOFOLLOW | MPSAFE | AUDITVNODE1, UIO_USERSPACE,
2477 	    uap->path, td);
2478 	if ((error = namei(&nd)) != 0)
2479 		return (error);
2480 	vfslocked = NDHASGIANT(&nd);
2481 	NDFREE(&nd, NDF_ONLY_PNBUF);
2482 	error = setfflags(td, nd.ni_vp, uap->flags);
2483 	vrele(nd.ni_vp);
2484 	VFS_UNLOCK_GIANT(vfslocked);
2485 	return (error);
2486 }
2487 
2488 /*
2489  * Change flags of a file given a file descriptor.
2490  */
2491 #ifndef _SYS_SYSPROTO_H_
2492 struct fchflags_args {
2493 	int	fd;
2494 	int	flags;
2495 };
2496 #endif
2497 int
2498 fchflags(td, uap)
2499 	struct thread *td;
2500 	register struct fchflags_args /* {
2501 		int fd;
2502 		int flags;
2503 	} */ *uap;
2504 {
2505 	struct file *fp;
2506 	int vfslocked;
2507 	int error;
2508 
2509 	AUDIT_ARG(fd, uap->fd);
2510 	AUDIT_ARG(fflags, uap->flags);
2511 	if ((error = getvnode(td->td_proc->p_fd, uap->fd, &fp)) != 0)
2512 		return (error);
2513 	vfslocked = VFS_LOCK_GIANT(fp->f_vnode->v_mount);
2514 #ifdef AUDIT
2515 	vn_lock(fp->f_vnode, LK_EXCLUSIVE | LK_RETRY);
2516 	AUDIT_ARG(vnode, fp->f_vnode, ARG_VNODE1);
2517 	VOP_UNLOCK(fp->f_vnode, 0);
2518 #endif
2519 	error = setfflags(td, fp->f_vnode, uap->flags);
2520 	VFS_UNLOCK_GIANT(vfslocked);
2521 	fdrop(fp, td);
2522 	return (error);
2523 }
2524 
2525 /*
2526  * Common implementation code for chmod(), lchmod() and fchmod().
2527  */
2528 static int
2529 setfmode(td, vp, mode)
2530 	struct thread *td;
2531 	struct vnode *vp;
2532 	int mode;
2533 {
2534 	int error;
2535 	struct mount *mp;
2536 	struct vattr vattr;
2537 
2538 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2539 		return (error);
2540 	VOP_LEASE(vp, td, td->td_ucred, LEASE_WRITE);
2541 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2542 	VATTR_NULL(&vattr);
2543 	vattr.va_mode = mode & ALLPERMS;
2544 #ifdef MAC
2545 	error = mac_vnode_check_setmode(td->td_ucred, vp, vattr.va_mode);
2546 	if (error == 0)
2547 #endif
2548 		error = VOP_SETATTR(vp, &vattr, td->td_ucred, td);
2549 	VOP_UNLOCK(vp, 0);
2550 	vn_finished_write(mp);
2551 	return (error);
2552 }
2553 
2554 /*
2555  * Change mode of a file given path name.
2556  */
2557 #ifndef _SYS_SYSPROTO_H_
2558 struct chmod_args {
2559 	char	*path;
2560 	int	mode;
2561 };
2562 #endif
2563 int
2564 chmod(td, uap)
2565 	struct thread *td;
2566 	register struct chmod_args /* {
2567 		char *path;
2568 		int mode;
2569 	} */ *uap;
2570 {
2571 
2572 	return (kern_chmod(td, uap->path, UIO_USERSPACE, uap->mode));
2573 }
2574 
2575 int
2576 kern_chmod(struct thread *td, char *path, enum uio_seg pathseg, int mode)
2577 {
2578 	int error;
2579 	struct nameidata nd;
2580 	int vfslocked;
2581 
2582 	AUDIT_ARG(mode, mode);
2583 	NDINIT(&nd, LOOKUP, FOLLOW | MPSAFE | AUDITVNODE1, pathseg, path, td);
2584 	if ((error = namei(&nd)) != 0)
2585 		return (error);
2586 	vfslocked = NDHASGIANT(&nd);
2587 	NDFREE(&nd, NDF_ONLY_PNBUF);
2588 	error = setfmode(td, nd.ni_vp, mode);
2589 	vrele(nd.ni_vp);
2590 	VFS_UNLOCK_GIANT(vfslocked);
2591 	return (error);
2592 }
2593 
2594 /*
2595  * Change mode of a file given path name (don't follow links.)
2596  */
2597 #ifndef _SYS_SYSPROTO_H_
2598 struct lchmod_args {
2599 	char	*path;
2600 	int	mode;
2601 };
2602 #endif
2603 int
2604 lchmod(td, uap)
2605 	struct thread *td;
2606 	register struct lchmod_args /* {
2607 		char *path;
2608 		int mode;
2609 	} */ *uap;
2610 {
2611 	int error;
2612 	struct nameidata nd;
2613 	int vfslocked;
2614 
2615 	AUDIT_ARG(mode, (mode_t)uap->mode);
2616 	NDINIT(&nd, LOOKUP, NOFOLLOW | MPSAFE | AUDITVNODE1, UIO_USERSPACE,
2617 	    uap->path, td);
2618 	if ((error = namei(&nd)) != 0)
2619 		return (error);
2620 	vfslocked = NDHASGIANT(&nd);
2621 	NDFREE(&nd, NDF_ONLY_PNBUF);
2622 	error = setfmode(td, nd.ni_vp, uap->mode);
2623 	vrele(nd.ni_vp);
2624 	VFS_UNLOCK_GIANT(vfslocked);
2625 	return (error);
2626 }
2627 
2628 /*
2629  * Change mode of a file given a file descriptor.
2630  */
2631 #ifndef _SYS_SYSPROTO_H_
2632 struct fchmod_args {
2633 	int	fd;
2634 	int	mode;
2635 };
2636 #endif
2637 int
2638 fchmod(td, uap)
2639 	struct thread *td;
2640 	register struct fchmod_args /* {
2641 		int fd;
2642 		int mode;
2643 	} */ *uap;
2644 {
2645 	struct file *fp;
2646 	int vfslocked;
2647 	int error;
2648 
2649 	AUDIT_ARG(fd, uap->fd);
2650 	AUDIT_ARG(mode, uap->mode);
2651 	if ((error = getvnode(td->td_proc->p_fd, uap->fd, &fp)) != 0)
2652 		return (error);
2653 	vfslocked = VFS_LOCK_GIANT(fp->f_vnode->v_mount);
2654 #ifdef AUDIT
2655 	vn_lock(fp->f_vnode, LK_EXCLUSIVE | LK_RETRY);
2656 	AUDIT_ARG(vnode, fp->f_vnode, ARG_VNODE1);
2657 	VOP_UNLOCK(fp->f_vnode, 0);
2658 #endif
2659 	error = setfmode(td, fp->f_vnode, uap->mode);
2660 	VFS_UNLOCK_GIANT(vfslocked);
2661 	fdrop(fp, td);
2662 	return (error);
2663 }
2664 
2665 /*
2666  * Common implementation for chown(), lchown(), and fchown()
2667  */
2668 static int
2669 setfown(td, vp, uid, gid)
2670 	struct thread *td;
2671 	struct vnode *vp;
2672 	uid_t uid;
2673 	gid_t gid;
2674 {
2675 	int error;
2676 	struct mount *mp;
2677 	struct vattr vattr;
2678 
2679 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2680 		return (error);
2681 	VOP_LEASE(vp, td, td->td_ucred, LEASE_WRITE);
2682 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2683 	VATTR_NULL(&vattr);
2684 	vattr.va_uid = uid;
2685 	vattr.va_gid = gid;
2686 #ifdef MAC
2687 	error = mac_vnode_check_setowner(td->td_ucred, vp, vattr.va_uid,
2688 	    vattr.va_gid);
2689 	if (error == 0)
2690 #endif
2691 		error = VOP_SETATTR(vp, &vattr, td->td_ucred, td);
2692 	VOP_UNLOCK(vp, 0);
2693 	vn_finished_write(mp);
2694 	return (error);
2695 }
2696 
2697 /*
2698  * Set ownership given a path name.
2699  */
2700 #ifndef _SYS_SYSPROTO_H_
2701 struct chown_args {
2702 	char	*path;
2703 	int	uid;
2704 	int	gid;
2705 };
2706 #endif
2707 int
2708 chown(td, uap)
2709 	struct thread *td;
2710 	register struct chown_args /* {
2711 		char *path;
2712 		int uid;
2713 		int gid;
2714 	} */ *uap;
2715 {
2716 
2717 	return (kern_chown(td, uap->path, UIO_USERSPACE, uap->uid, uap->gid));
2718 }
2719 
2720 int
2721 kern_chown(struct thread *td, char *path, enum uio_seg pathseg, int uid,
2722     int gid)
2723 {
2724 	int error;
2725 	struct nameidata nd;
2726 	int vfslocked;
2727 
2728 	AUDIT_ARG(owner, uid, gid);
2729 	NDINIT(&nd, LOOKUP, FOLLOW | MPSAFE | AUDITVNODE1, pathseg, path, td);
2730 	if ((error = namei(&nd)) != 0)
2731 		return (error);
2732 	vfslocked = NDHASGIANT(&nd);
2733 	NDFREE(&nd, NDF_ONLY_PNBUF);
2734 	error = setfown(td, nd.ni_vp, uid, gid);
2735 	vrele(nd.ni_vp);
2736 	VFS_UNLOCK_GIANT(vfslocked);
2737 	return (error);
2738 }
2739 
2740 /*
2741  * Set ownership given a path name, do not cross symlinks.
2742  */
2743 #ifndef _SYS_SYSPROTO_H_
2744 struct lchown_args {
2745 	char	*path;
2746 	int	uid;
2747 	int	gid;
2748 };
2749 #endif
2750 int
2751 lchown(td, uap)
2752 	struct thread *td;
2753 	register struct lchown_args /* {
2754 		char *path;
2755 		int uid;
2756 		int gid;
2757 	} */ *uap;
2758 {
2759 
2760 	return (kern_lchown(td, uap->path, UIO_USERSPACE, uap->uid, uap->gid));
2761 }
2762 
2763 int
2764 kern_lchown(struct thread *td, char *path, enum uio_seg pathseg, int uid,
2765     int gid)
2766 {
2767 	int error;
2768 	struct nameidata nd;
2769 	int vfslocked;
2770 
2771 	AUDIT_ARG(owner, uid, gid);
2772 	NDINIT(&nd, LOOKUP, NOFOLLOW | MPSAFE | AUDITVNODE1, pathseg, path, td);
2773 	if ((error = namei(&nd)) != 0)
2774 		return (error);
2775 	vfslocked = NDHASGIANT(&nd);
2776 	NDFREE(&nd, NDF_ONLY_PNBUF);
2777 	error = setfown(td, nd.ni_vp, uid, gid);
2778 	vrele(nd.ni_vp);
2779 	VFS_UNLOCK_GIANT(vfslocked);
2780 	return (error);
2781 }
2782 
2783 /*
2784  * Set ownership given a file descriptor.
2785  */
2786 #ifndef _SYS_SYSPROTO_H_
2787 struct fchown_args {
2788 	int	fd;
2789 	int	uid;
2790 	int	gid;
2791 };
2792 #endif
2793 int
2794 fchown(td, uap)
2795 	struct thread *td;
2796 	register struct fchown_args /* {
2797 		int fd;
2798 		int uid;
2799 		int gid;
2800 	} */ *uap;
2801 {
2802 	struct file *fp;
2803 	int vfslocked;
2804 	int error;
2805 
2806 	AUDIT_ARG(fd, uap->fd);
2807 	AUDIT_ARG(owner, uap->uid, uap->gid);
2808 	if ((error = getvnode(td->td_proc->p_fd, uap->fd, &fp)) != 0)
2809 		return (error);
2810 	vfslocked = VFS_LOCK_GIANT(fp->f_vnode->v_mount);
2811 #ifdef AUDIT
2812 	vn_lock(fp->f_vnode, LK_EXCLUSIVE | LK_RETRY);
2813 	AUDIT_ARG(vnode, fp->f_vnode, ARG_VNODE1);
2814 	VOP_UNLOCK(fp->f_vnode, 0);
2815 #endif
2816 	error = setfown(td, fp->f_vnode, uap->uid, uap->gid);
2817 	VFS_UNLOCK_GIANT(vfslocked);
2818 	fdrop(fp, td);
2819 	return (error);
2820 }
2821 
2822 /*
2823  * Common implementation code for utimes(), lutimes(), and futimes().
2824  */
2825 static int
2826 getutimes(usrtvp, tvpseg, tsp)
2827 	const struct timeval *usrtvp;
2828 	enum uio_seg tvpseg;
2829 	struct timespec *tsp;
2830 {
2831 	struct timeval tv[2];
2832 	const struct timeval *tvp;
2833 	int error;
2834 
2835 	if (usrtvp == NULL) {
2836 		microtime(&tv[0]);
2837 		TIMEVAL_TO_TIMESPEC(&tv[0], &tsp[0]);
2838 		tsp[1] = tsp[0];
2839 	} else {
2840 		if (tvpseg == UIO_SYSSPACE) {
2841 			tvp = usrtvp;
2842 		} else {
2843 			if ((error = copyin(usrtvp, tv, sizeof(tv))) != 0)
2844 				return (error);
2845 			tvp = tv;
2846 		}
2847 
2848 		if (tvp[0].tv_usec < 0 || tvp[0].tv_usec >= 1000000 ||
2849 		    tvp[1].tv_usec < 0 || tvp[1].tv_usec >= 1000000)
2850 			return (EINVAL);
2851 		TIMEVAL_TO_TIMESPEC(&tvp[0], &tsp[0]);
2852 		TIMEVAL_TO_TIMESPEC(&tvp[1], &tsp[1]);
2853 	}
2854 	return (0);
2855 }
2856 
2857 /*
2858  * Common implementation code for utimes(), lutimes(), and futimes().
2859  */
2860 static int
2861 setutimes(td, vp, ts, numtimes, nullflag)
2862 	struct thread *td;
2863 	struct vnode *vp;
2864 	const struct timespec *ts;
2865 	int numtimes;
2866 	int nullflag;
2867 {
2868 	int error, setbirthtime;
2869 	struct mount *mp;
2870 	struct vattr vattr;
2871 
2872 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2873 		return (error);
2874 	VOP_LEASE(vp, td, td->td_ucred, LEASE_WRITE);
2875 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2876 	setbirthtime = 0;
2877 	if (numtimes < 3 && VOP_GETATTR(vp, &vattr, td->td_ucred, td) == 0 &&
2878 	    timespeccmp(&ts[1], &vattr.va_birthtime, < ))
2879 		setbirthtime = 1;
2880 	VATTR_NULL(&vattr);
2881 	vattr.va_atime = ts[0];
2882 	vattr.va_mtime = ts[1];
2883 	if (setbirthtime)
2884 		vattr.va_birthtime = ts[1];
2885 	if (numtimes > 2)
2886 		vattr.va_birthtime = ts[2];
2887 	if (nullflag)
2888 		vattr.va_vaflags |= VA_UTIMES_NULL;
2889 #ifdef MAC
2890 	error = mac_vnode_check_setutimes(td->td_ucred, vp, vattr.va_atime,
2891 	    vattr.va_mtime);
2892 #endif
2893 	if (error == 0)
2894 		error = VOP_SETATTR(vp, &vattr, td->td_ucred, td);
2895 	VOP_UNLOCK(vp, 0);
2896 	vn_finished_write(mp);
2897 	return (error);
2898 }
2899 
2900 /*
2901  * Set the access and modification times of a file.
2902  */
2903 #ifndef _SYS_SYSPROTO_H_
2904 struct utimes_args {
2905 	char	*path;
2906 	struct	timeval *tptr;
2907 };
2908 #endif
2909 int
2910 utimes(td, uap)
2911 	struct thread *td;
2912 	register struct utimes_args /* {
2913 		char *path;
2914 		struct timeval *tptr;
2915 	} */ *uap;
2916 {
2917 
2918 	return (kern_utimes(td, uap->path, UIO_USERSPACE, uap->tptr,
2919 	    UIO_USERSPACE));
2920 }
2921 
2922 int
2923 kern_utimes(struct thread *td, char *path, enum uio_seg pathseg,
2924     struct timeval *tptr, enum uio_seg tptrseg)
2925 {
2926 	struct timespec ts[2];
2927 	int error;
2928 	struct nameidata nd;
2929 	int vfslocked;
2930 
2931 	if ((error = getutimes(tptr, tptrseg, ts)) != 0)
2932 		return (error);
2933 	NDINIT(&nd, LOOKUP, FOLLOW | MPSAFE | AUDITVNODE1, pathseg, path, td);
2934 	if ((error = namei(&nd)) != 0)
2935 		return (error);
2936 	vfslocked = NDHASGIANT(&nd);
2937 	NDFREE(&nd, NDF_ONLY_PNBUF);
2938 	error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
2939 	vrele(nd.ni_vp);
2940 	VFS_UNLOCK_GIANT(vfslocked);
2941 	return (error);
2942 }
2943 
2944 /*
2945  * Set the access and modification times of a file.
2946  */
2947 #ifndef _SYS_SYSPROTO_H_
2948 struct lutimes_args {
2949 	char	*path;
2950 	struct	timeval *tptr;
2951 };
2952 #endif
2953 int
2954 lutimes(td, uap)
2955 	struct thread *td;
2956 	register struct lutimes_args /* {
2957 		char *path;
2958 		struct timeval *tptr;
2959 	} */ *uap;
2960 {
2961 
2962 	return (kern_lutimes(td, uap->path, UIO_USERSPACE, uap->tptr,
2963 	    UIO_USERSPACE));
2964 }
2965 
2966 int
2967 kern_lutimes(struct thread *td, char *path, enum uio_seg pathseg,
2968     struct timeval *tptr, enum uio_seg tptrseg)
2969 {
2970 	struct timespec ts[2];
2971 	int error;
2972 	struct nameidata nd;
2973 	int vfslocked;
2974 
2975 	if ((error = getutimes(tptr, tptrseg, ts)) != 0)
2976 		return (error);
2977 	NDINIT(&nd, LOOKUP, NOFOLLOW | MPSAFE | AUDITVNODE1, pathseg, path, td);
2978 	if ((error = namei(&nd)) != 0)
2979 		return (error);
2980 	vfslocked = NDHASGIANT(&nd);
2981 	NDFREE(&nd, NDF_ONLY_PNBUF);
2982 	error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
2983 	vrele(nd.ni_vp);
2984 	VFS_UNLOCK_GIANT(vfslocked);
2985 	return (error);
2986 }
2987 
2988 /*
2989  * Set the access and modification times of a file.
2990  */
2991 #ifndef _SYS_SYSPROTO_H_
2992 struct futimes_args {
2993 	int	fd;
2994 	struct	timeval *tptr;
2995 };
2996 #endif
2997 int
2998 futimes(td, uap)
2999 	struct thread *td;
3000 	register struct futimes_args /* {
3001 		int  fd;
3002 		struct timeval *tptr;
3003 	} */ *uap;
3004 {
3005 
3006 	return (kern_futimes(td, uap->fd, uap->tptr, UIO_USERSPACE));
3007 }
3008 
3009 int
3010 kern_futimes(struct thread *td, int fd, struct timeval *tptr,
3011     enum uio_seg tptrseg)
3012 {
3013 	struct timespec ts[2];
3014 	struct file *fp;
3015 	int vfslocked;
3016 	int error;
3017 
3018 	AUDIT_ARG(fd, fd);
3019 	if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3020 		return (error);
3021 	if ((error = getvnode(td->td_proc->p_fd, fd, &fp)) != 0)
3022 		return (error);
3023 	vfslocked = VFS_LOCK_GIANT(fp->f_vnode->v_mount);
3024 #ifdef AUDIT
3025 	vn_lock(fp->f_vnode, LK_EXCLUSIVE | LK_RETRY);
3026 	AUDIT_ARG(vnode, fp->f_vnode, ARG_VNODE1);
3027 	VOP_UNLOCK(fp->f_vnode, 0);
3028 #endif
3029 	error = setutimes(td, fp->f_vnode, ts, 2, tptr == NULL);
3030 	VFS_UNLOCK_GIANT(vfslocked);
3031 	fdrop(fp, td);
3032 	return (error);
3033 }
3034 
3035 /*
3036  * Truncate a file given its path name.
3037  */
3038 #ifndef _SYS_SYSPROTO_H_
3039 struct truncate_args {
3040 	char	*path;
3041 	int	pad;
3042 	off_t	length;
3043 };
3044 #endif
3045 int
3046 truncate(td, uap)
3047 	struct thread *td;
3048 	register struct truncate_args /* {
3049 		char *path;
3050 		int pad;
3051 		off_t length;
3052 	} */ *uap;
3053 {
3054 
3055 	return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3056 }
3057 
3058 int
3059 kern_truncate(struct thread *td, char *path, enum uio_seg pathseg, off_t length)
3060 {
3061 	struct mount *mp;
3062 	struct vnode *vp;
3063 	struct vattr vattr;
3064 	int error;
3065 	struct nameidata nd;
3066 	int vfslocked;
3067 
3068 	if (length < 0)
3069 		return(EINVAL);
3070 	NDINIT(&nd, LOOKUP, FOLLOW | MPSAFE | AUDITVNODE1, pathseg, path, td);
3071 	if ((error = namei(&nd)) != 0)
3072 		return (error);
3073 	vfslocked = NDHASGIANT(&nd);
3074 	vp = nd.ni_vp;
3075 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) {
3076 		vrele(vp);
3077 		VFS_UNLOCK_GIANT(vfslocked);
3078 		return (error);
3079 	}
3080 	NDFREE(&nd, NDF_ONLY_PNBUF);
3081 	VOP_LEASE(vp, td, td->td_ucred, LEASE_WRITE);
3082 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3083 	if (vp->v_type == VDIR)
3084 		error = EISDIR;
3085 #ifdef MAC
3086 	else if ((error = mac_vnode_check_write(td->td_ucred, NOCRED, vp))) {
3087 	}
3088 #endif
3089 	else if ((error = vn_writechk(vp)) == 0 &&
3090 	    (error = VOP_ACCESS(vp, VWRITE, td->td_ucred, td)) == 0) {
3091 		VATTR_NULL(&vattr);
3092 		vattr.va_size = length;
3093 		error = VOP_SETATTR(vp, &vattr, td->td_ucred, td);
3094 	}
3095 	vput(vp);
3096 	vn_finished_write(mp);
3097 	VFS_UNLOCK_GIANT(vfslocked);
3098 	return (error);
3099 }
3100 
3101 #if defined(COMPAT_43)
3102 /*
3103  * Truncate a file given its path name.
3104  */
3105 #ifndef _SYS_SYSPROTO_H_
3106 struct otruncate_args {
3107 	char	*path;
3108 	long	length;
3109 };
3110 #endif
3111 int
3112 otruncate(td, uap)
3113 	struct thread *td;
3114 	register struct otruncate_args /* {
3115 		char *path;
3116 		long length;
3117 	} */ *uap;
3118 {
3119 	struct truncate_args /* {
3120 		char *path;
3121 		int pad;
3122 		off_t length;
3123 	} */ nuap;
3124 
3125 	nuap.path = uap->path;
3126 	nuap.length = uap->length;
3127 	return (truncate(td, &nuap));
3128 }
3129 #endif /* COMPAT_43 */
3130 
3131 /* Versions with the pad argument */
3132 int
3133 freebsd6_truncate(struct thread *td, struct freebsd6_truncate_args *uap)
3134 {
3135 	struct truncate_args ouap;
3136 
3137 	ouap.path = uap->path;
3138 	ouap.length = uap->length;
3139 	return (truncate(td, &ouap));
3140 }
3141 
3142 int
3143 freebsd6_ftruncate(struct thread *td, struct freebsd6_ftruncate_args *uap)
3144 {
3145 	struct ftruncate_args ouap;
3146 
3147 	ouap.fd = uap->fd;
3148 	ouap.length = uap->length;
3149 	return (ftruncate(td, &ouap));
3150 }
3151 
3152 /*
3153  * Sync an open file.
3154  */
3155 #ifndef _SYS_SYSPROTO_H_
3156 struct fsync_args {
3157 	int	fd;
3158 };
3159 #endif
3160 int
3161 fsync(td, uap)
3162 	struct thread *td;
3163 	struct fsync_args /* {
3164 		int fd;
3165 	} */ *uap;
3166 {
3167 	struct vnode *vp;
3168 	struct mount *mp;
3169 	struct file *fp;
3170 	int vfslocked;
3171 	int error;
3172 
3173 	AUDIT_ARG(fd, uap->fd);
3174 	if ((error = getvnode(td->td_proc->p_fd, uap->fd, &fp)) != 0)
3175 		return (error);
3176 	vp = fp->f_vnode;
3177 	vfslocked = VFS_LOCK_GIANT(vp->v_mount);
3178 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
3179 		goto drop;
3180 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3181 	AUDIT_ARG(vnode, vp, ARG_VNODE1);
3182 	if (vp->v_object != NULL) {
3183 		VM_OBJECT_LOCK(vp->v_object);
3184 		vm_object_page_clean(vp->v_object, 0, 0, 0);
3185 		VM_OBJECT_UNLOCK(vp->v_object);
3186 	}
3187 	error = VOP_FSYNC(vp, MNT_WAIT, td);
3188 
3189 	VOP_UNLOCK(vp, 0);
3190 	vn_finished_write(mp);
3191 drop:
3192 	VFS_UNLOCK_GIANT(vfslocked);
3193 	fdrop(fp, td);
3194 	return (error);
3195 }
3196 
3197 /*
3198  * Rename files.  Source and destination must either both be directories, or
3199  * both not be directories.  If target is a directory, it must be empty.
3200  */
3201 #ifndef _SYS_SYSPROTO_H_
3202 struct rename_args {
3203 	char	*from;
3204 	char	*to;
3205 };
3206 #endif
3207 int
3208 rename(td, uap)
3209 	struct thread *td;
3210 	register struct rename_args /* {
3211 		char *from;
3212 		char *to;
3213 	} */ *uap;
3214 {
3215 
3216 	return (kern_rename(td, uap->from, uap->to, UIO_USERSPACE));
3217 }
3218 
3219 int
3220 kern_rename(struct thread *td, char *from, char *to, enum uio_seg pathseg)
3221 {
3222 	struct mount *mp = NULL;
3223 	struct vnode *tvp, *fvp, *tdvp;
3224 	struct nameidata fromnd, tond;
3225 	int tvfslocked;
3226 	int fvfslocked;
3227 	int error;
3228 
3229 	bwillwrite();
3230 #ifdef MAC
3231 	NDINIT(&fromnd, DELETE, LOCKPARENT | LOCKLEAF | SAVESTART | MPSAFE |
3232 	    AUDITVNODE1, pathseg, from, td);
3233 #else
3234 	NDINIT(&fromnd, DELETE, WANTPARENT | SAVESTART | MPSAFE |
3235 	    AUDITVNODE1, pathseg, from, td);
3236 #endif
3237 	if ((error = namei(&fromnd)) != 0)
3238 		return (error);
3239 	fvfslocked = NDHASGIANT(&fromnd);
3240 	tvfslocked = 0;
3241 #ifdef MAC
3242 	error = mac_vnode_check_rename_from(td->td_ucred, fromnd.ni_dvp,
3243 	    fromnd.ni_vp, &fromnd.ni_cnd);
3244 	VOP_UNLOCK(fromnd.ni_dvp, 0);
3245 	if (fromnd.ni_dvp != fromnd.ni_vp)
3246 		VOP_UNLOCK(fromnd.ni_vp, 0);
3247 #endif
3248 	fvp = fromnd.ni_vp;
3249 	if (error == 0)
3250 		error = vn_start_write(fvp, &mp, V_WAIT | PCATCH);
3251 	if (error != 0) {
3252 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3253 		vrele(fromnd.ni_dvp);
3254 		vrele(fvp);
3255 		goto out1;
3256 	}
3257 	NDINIT(&tond, RENAME, LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART |
3258 	    MPSAFE | AUDITVNODE2, pathseg, to, td);
3259 	if (fromnd.ni_vp->v_type == VDIR)
3260 		tond.ni_cnd.cn_flags |= WILLBEDIR;
3261 	if ((error = namei(&tond)) != 0) {
3262 		/* Translate error code for rename("dir1", "dir2/."). */
3263 		if (error == EISDIR && fvp->v_type == VDIR)
3264 			error = EINVAL;
3265 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3266 		vrele(fromnd.ni_dvp);
3267 		vrele(fvp);
3268 		vn_finished_write(mp);
3269 		goto out1;
3270 	}
3271 	tvfslocked = NDHASGIANT(&tond);
3272 	tdvp = tond.ni_dvp;
3273 	tvp = tond.ni_vp;
3274 	if (tvp != NULL) {
3275 		if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
3276 			error = ENOTDIR;
3277 			goto out;
3278 		} else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
3279 			error = EISDIR;
3280 			goto out;
3281 		}
3282 	}
3283 	if (fvp == tdvp)
3284 		error = EINVAL;
3285 	/*
3286 	 * If the source is the same as the destination (that is, if they
3287 	 * are links to the same vnode), then there is nothing to do.
3288 	 */
3289 	if (fvp == tvp)
3290 		error = -1;
3291 #ifdef MAC
3292 	else
3293 		error = mac_vnode_check_rename_to(td->td_ucred, tdvp,
3294 		    tond.ni_vp, fromnd.ni_dvp == tdvp, &tond.ni_cnd);
3295 #endif
3296 out:
3297 	if (!error) {
3298 		VOP_LEASE(tdvp, td, td->td_ucred, LEASE_WRITE);
3299 		if (fromnd.ni_dvp != tdvp) {
3300 			VOP_LEASE(fromnd.ni_dvp, td, td->td_ucred, LEASE_WRITE);
3301 		}
3302 		if (tvp) {
3303 			VOP_LEASE(tvp, td, td->td_ucred, LEASE_WRITE);
3304 		}
3305 		error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
3306 				   tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
3307 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3308 		NDFREE(&tond, NDF_ONLY_PNBUF);
3309 	} else {
3310 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3311 		NDFREE(&tond, NDF_ONLY_PNBUF);
3312 		if (tvp)
3313 			vput(tvp);
3314 		if (tdvp == tvp)
3315 			vrele(tdvp);
3316 		else
3317 			vput(tdvp);
3318 		vrele(fromnd.ni_dvp);
3319 		vrele(fvp);
3320 	}
3321 	vrele(tond.ni_startdir);
3322 	vn_finished_write(mp);
3323 out1:
3324 	if (fromnd.ni_startdir)
3325 		vrele(fromnd.ni_startdir);
3326 	VFS_UNLOCK_GIANT(fvfslocked);
3327 	VFS_UNLOCK_GIANT(tvfslocked);
3328 	if (error == -1)
3329 		return (0);
3330 	return (error);
3331 }
3332 
3333 /*
3334  * Make a directory file.
3335  */
3336 #ifndef _SYS_SYSPROTO_H_
3337 struct mkdir_args {
3338 	char	*path;
3339 	int	mode;
3340 };
3341 #endif
3342 int
3343 mkdir(td, uap)
3344 	struct thread *td;
3345 	register struct mkdir_args /* {
3346 		char *path;
3347 		int mode;
3348 	} */ *uap;
3349 {
3350 
3351 	return (kern_mkdir(td, uap->path, UIO_USERSPACE, uap->mode));
3352 }
3353 
3354 int
3355 kern_mkdir(struct thread *td, char *path, enum uio_seg segflg, int mode)
3356 {
3357 	struct mount *mp;
3358 	struct vnode *vp;
3359 	struct vattr vattr;
3360 	int error;
3361 	struct nameidata nd;
3362 	int vfslocked;
3363 
3364 	AUDIT_ARG(mode, mode);
3365 restart:
3366 	bwillwrite();
3367 	NDINIT(&nd, CREATE, LOCKPARENT | SAVENAME | MPSAFE | AUDITVNODE1,
3368 	    segflg, path, td);
3369 	nd.ni_cnd.cn_flags |= WILLBEDIR;
3370 	if ((error = namei(&nd)) != 0)
3371 		return (error);
3372 	vfslocked = NDHASGIANT(&nd);
3373 	vp = nd.ni_vp;
3374 	if (vp != NULL) {
3375 		NDFREE(&nd, NDF_ONLY_PNBUF);
3376 		/*
3377 		 * XXX namei called with LOCKPARENT but not LOCKLEAF has
3378 		 * the strange behaviour of leaving the vnode unlocked
3379 		 * if the target is the same vnode as the parent.
3380 		 */
3381 		if (vp == nd.ni_dvp)
3382 			vrele(nd.ni_dvp);
3383 		else
3384 			vput(nd.ni_dvp);
3385 		vrele(vp);
3386 		VFS_UNLOCK_GIANT(vfslocked);
3387 		return (EEXIST);
3388 	}
3389 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3390 		NDFREE(&nd, NDF_ONLY_PNBUF);
3391 		vput(nd.ni_dvp);
3392 		VFS_UNLOCK_GIANT(vfslocked);
3393 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3394 			return (error);
3395 		goto restart;
3396 	}
3397 	VATTR_NULL(&vattr);
3398 	vattr.va_type = VDIR;
3399 	FILEDESC_SLOCK(td->td_proc->p_fd);
3400 	vattr.va_mode = (mode & ACCESSPERMS) &~ td->td_proc->p_fd->fd_cmask;
3401 	FILEDESC_SUNLOCK(td->td_proc->p_fd);
3402 #ifdef MAC
3403 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
3404 	    &vattr);
3405 	if (error)
3406 		goto out;
3407 #endif
3408 	VOP_LEASE(nd.ni_dvp, td, td->td_ucred, LEASE_WRITE);
3409 	error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
3410 #ifdef MAC
3411 out:
3412 #endif
3413 	NDFREE(&nd, NDF_ONLY_PNBUF);
3414 	vput(nd.ni_dvp);
3415 	if (!error)
3416 		vput(nd.ni_vp);
3417 	vn_finished_write(mp);
3418 	VFS_UNLOCK_GIANT(vfslocked);
3419 	return (error);
3420 }
3421 
3422 /*
3423  * Remove a directory file.
3424  */
3425 #ifndef _SYS_SYSPROTO_H_
3426 struct rmdir_args {
3427 	char	*path;
3428 };
3429 #endif
3430 int
3431 rmdir(td, uap)
3432 	struct thread *td;
3433 	struct rmdir_args /* {
3434 		char *path;
3435 	} */ *uap;
3436 {
3437 
3438 	return (kern_rmdir(td, uap->path, UIO_USERSPACE));
3439 }
3440 
3441 int
3442 kern_rmdir(struct thread *td, char *path, enum uio_seg pathseg)
3443 {
3444 	struct mount *mp;
3445 	struct vnode *vp;
3446 	int error;
3447 	struct nameidata nd;
3448 	int vfslocked;
3449 
3450 restart:
3451 	bwillwrite();
3452 	NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF | MPSAFE | AUDITVNODE1,
3453 	    pathseg, path, td);
3454 	if ((error = namei(&nd)) != 0)
3455 		return (error);
3456 	vfslocked = NDHASGIANT(&nd);
3457 	vp = nd.ni_vp;
3458 	if (vp->v_type != VDIR) {
3459 		error = ENOTDIR;
3460 		goto out;
3461 	}
3462 	/*
3463 	 * No rmdir "." please.
3464 	 */
3465 	if (nd.ni_dvp == vp) {
3466 		error = EINVAL;
3467 		goto out;
3468 	}
3469 	/*
3470 	 * The root of a mounted filesystem cannot be deleted.
3471 	 */
3472 	if (vp->v_vflag & VV_ROOT) {
3473 		error = EBUSY;
3474 		goto out;
3475 	}
3476 #ifdef MAC
3477 	error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
3478 	    &nd.ni_cnd);
3479 	if (error)
3480 		goto out;
3481 #endif
3482 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3483 		NDFREE(&nd, NDF_ONLY_PNBUF);
3484 		vput(vp);
3485 		if (nd.ni_dvp == vp)
3486 			vrele(nd.ni_dvp);
3487 		else
3488 			vput(nd.ni_dvp);
3489 		VFS_UNLOCK_GIANT(vfslocked);
3490 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3491 			return (error);
3492 		goto restart;
3493 	}
3494 	VOP_LEASE(nd.ni_dvp, td, td->td_ucred, LEASE_WRITE);
3495 	VOP_LEASE(vp, td, td->td_ucred, LEASE_WRITE);
3496 	error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
3497 	vn_finished_write(mp);
3498 out:
3499 	NDFREE(&nd, NDF_ONLY_PNBUF);
3500 	vput(vp);
3501 	if (nd.ni_dvp == vp)
3502 		vrele(nd.ni_dvp);
3503 	else
3504 		vput(nd.ni_dvp);
3505 	VFS_UNLOCK_GIANT(vfslocked);
3506 	return (error);
3507 }
3508 
3509 #ifdef COMPAT_43
3510 /*
3511  * Read a block of directory entries in a filesystem independent format.
3512  */
3513 #ifndef _SYS_SYSPROTO_H_
3514 struct ogetdirentries_args {
3515 	int	fd;
3516 	char	*buf;
3517 	u_int	count;
3518 	long	*basep;
3519 };
3520 #endif
3521 int
3522 ogetdirentries(td, uap)
3523 	struct thread *td;
3524 	register struct ogetdirentries_args /* {
3525 		int fd;
3526 		char *buf;
3527 		u_int count;
3528 		long *basep;
3529 	} */ *uap;
3530 {
3531 	struct vnode *vp;
3532 	struct file *fp;
3533 	struct uio auio, kuio;
3534 	struct iovec aiov, kiov;
3535 	struct dirent *dp, *edp;
3536 	caddr_t dirbuf;
3537 	int error, eofflag, readcnt, vfslocked;
3538 	long loff;
3539 
3540 	/* XXX arbitrary sanity limit on `count'. */
3541 	if (uap->count > 64 * 1024)
3542 		return (EINVAL);
3543 	if ((error = getvnode(td->td_proc->p_fd, uap->fd, &fp)) != 0)
3544 		return (error);
3545 	if ((fp->f_flag & FREAD) == 0) {
3546 		fdrop(fp, td);
3547 		return (EBADF);
3548 	}
3549 	vp = fp->f_vnode;
3550 unionread:
3551 	vfslocked = VFS_LOCK_GIANT(vp->v_mount);
3552 	if (vp->v_type != VDIR) {
3553 		VFS_UNLOCK_GIANT(vfslocked);
3554 		fdrop(fp, td);
3555 		return (EINVAL);
3556 	}
3557 	aiov.iov_base = uap->buf;
3558 	aiov.iov_len = uap->count;
3559 	auio.uio_iov = &aiov;
3560 	auio.uio_iovcnt = 1;
3561 	auio.uio_rw = UIO_READ;
3562 	auio.uio_segflg = UIO_USERSPACE;
3563 	auio.uio_td = td;
3564 	auio.uio_resid = uap->count;
3565 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3566 	loff = auio.uio_offset = fp->f_offset;
3567 #ifdef MAC
3568 	error = mac_vnode_check_readdir(td->td_ucred, vp);
3569 	if (error) {
3570 		VOP_UNLOCK(vp, 0);
3571 		VFS_UNLOCK_GIANT(vfslocked);
3572 		fdrop(fp, td);
3573 		return (error);
3574 	}
3575 #endif
3576 #	if (BYTE_ORDER != LITTLE_ENDIAN)
3577 		if (vp->v_mount->mnt_maxsymlinklen <= 0) {
3578 			error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag,
3579 			    NULL, NULL);
3580 			fp->f_offset = auio.uio_offset;
3581 		} else
3582 #	endif
3583 	{
3584 		kuio = auio;
3585 		kuio.uio_iov = &kiov;
3586 		kuio.uio_segflg = UIO_SYSSPACE;
3587 		kiov.iov_len = uap->count;
3588 		MALLOC(dirbuf, caddr_t, uap->count, M_TEMP, M_WAITOK);
3589 		kiov.iov_base = dirbuf;
3590 		error = VOP_READDIR(vp, &kuio, fp->f_cred, &eofflag,
3591 			    NULL, NULL);
3592 		fp->f_offset = kuio.uio_offset;
3593 		if (error == 0) {
3594 			readcnt = uap->count - kuio.uio_resid;
3595 			edp = (struct dirent *)&dirbuf[readcnt];
3596 			for (dp = (struct dirent *)dirbuf; dp < edp; ) {
3597 #				if (BYTE_ORDER == LITTLE_ENDIAN)
3598 					/*
3599 					 * The expected low byte of
3600 					 * dp->d_namlen is our dp->d_type.
3601 					 * The high MBZ byte of dp->d_namlen
3602 					 * is our dp->d_namlen.
3603 					 */
3604 					dp->d_type = dp->d_namlen;
3605 					dp->d_namlen = 0;
3606 #				else
3607 					/*
3608 					 * The dp->d_type is the high byte
3609 					 * of the expected dp->d_namlen,
3610 					 * so must be zero'ed.
3611 					 */
3612 					dp->d_type = 0;
3613 #				endif
3614 				if (dp->d_reclen > 0) {
3615 					dp = (struct dirent *)
3616 					    ((char *)dp + dp->d_reclen);
3617 				} else {
3618 					error = EIO;
3619 					break;
3620 				}
3621 			}
3622 			if (dp >= edp)
3623 				error = uiomove(dirbuf, readcnt, &auio);
3624 		}
3625 		FREE(dirbuf, M_TEMP);
3626 	}
3627 	if (error) {
3628 		VOP_UNLOCK(vp, 0);
3629 		VFS_UNLOCK_GIANT(vfslocked);
3630 		fdrop(fp, td);
3631 		return (error);
3632 	}
3633 	if (uap->count == auio.uio_resid &&
3634 	    (vp->v_vflag & VV_ROOT) &&
3635 	    (vp->v_mount->mnt_flag & MNT_UNION)) {
3636 		struct vnode *tvp = vp;
3637 		vp = vp->v_mount->mnt_vnodecovered;
3638 		VREF(vp);
3639 		fp->f_vnode = vp;
3640 		fp->f_data = vp;
3641 		fp->f_offset = 0;
3642 		vput(tvp);
3643 		VFS_UNLOCK_GIANT(vfslocked);
3644 		goto unionread;
3645 	}
3646 	VOP_UNLOCK(vp, 0);
3647 	VFS_UNLOCK_GIANT(vfslocked);
3648 	error = copyout(&loff, uap->basep, sizeof(long));
3649 	fdrop(fp, td);
3650 	td->td_retval[0] = uap->count - auio.uio_resid;
3651 	return (error);
3652 }
3653 #endif /* COMPAT_43 */
3654 
3655 /*
3656  * Read a block of directory entries in a filesystem independent format.
3657  */
3658 #ifndef _SYS_SYSPROTO_H_
3659 struct getdirentries_args {
3660 	int	fd;
3661 	char	*buf;
3662 	u_int	count;
3663 	long	*basep;
3664 };
3665 #endif
3666 int
3667 getdirentries(td, uap)
3668 	struct thread *td;
3669 	register struct getdirentries_args /* {
3670 		int fd;
3671 		char *buf;
3672 		u_int count;
3673 		long *basep;
3674 	} */ *uap;
3675 {
3676 	struct vnode *vp;
3677 	struct file *fp;
3678 	struct uio auio;
3679 	struct iovec aiov;
3680 	int vfslocked;
3681 	long loff;
3682 	int error, eofflag;
3683 
3684 	AUDIT_ARG(fd, uap->fd);
3685 	if ((error = getvnode(td->td_proc->p_fd, uap->fd, &fp)) != 0)
3686 		return (error);
3687 	if ((fp->f_flag & FREAD) == 0) {
3688 		fdrop(fp, td);
3689 		return (EBADF);
3690 	}
3691 	vp = fp->f_vnode;
3692 unionread:
3693 	vfslocked = VFS_LOCK_GIANT(vp->v_mount);
3694 	if (vp->v_type != VDIR) {
3695 		VFS_UNLOCK_GIANT(vfslocked);
3696 		error = EINVAL;
3697 		goto fail;
3698 	}
3699 	aiov.iov_base = uap->buf;
3700 	aiov.iov_len = uap->count;
3701 	auio.uio_iov = &aiov;
3702 	auio.uio_iovcnt = 1;
3703 	auio.uio_rw = UIO_READ;
3704 	auio.uio_segflg = UIO_USERSPACE;
3705 	auio.uio_td = td;
3706 	auio.uio_resid = uap->count;
3707 	/* vn_lock(vp, LK_SHARED | LK_RETRY); */
3708 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3709 	AUDIT_ARG(vnode, vp, ARG_VNODE1);
3710 	loff = auio.uio_offset = fp->f_offset;
3711 #ifdef MAC
3712 	error = mac_vnode_check_readdir(td->td_ucred, vp);
3713 	if (error == 0)
3714 #endif
3715 		error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, NULL,
3716 		    NULL);
3717 	fp->f_offset = auio.uio_offset;
3718 	if (error) {
3719 		VOP_UNLOCK(vp, 0);
3720 		VFS_UNLOCK_GIANT(vfslocked);
3721 		goto fail;
3722 	}
3723 	if (uap->count == auio.uio_resid &&
3724 	    (vp->v_vflag & VV_ROOT) &&
3725 	    (vp->v_mount->mnt_flag & MNT_UNION)) {
3726 		struct vnode *tvp = vp;
3727 		vp = vp->v_mount->mnt_vnodecovered;
3728 		VREF(vp);
3729 		fp->f_vnode = vp;
3730 		fp->f_data = vp;
3731 		fp->f_offset = 0;
3732 		vput(tvp);
3733 		VFS_UNLOCK_GIANT(vfslocked);
3734 		goto unionread;
3735 	}
3736 	VOP_UNLOCK(vp, 0);
3737 	VFS_UNLOCK_GIANT(vfslocked);
3738 	if (uap->basep != NULL) {
3739 		error = copyout(&loff, uap->basep, sizeof(long));
3740 	}
3741 	td->td_retval[0] = uap->count - auio.uio_resid;
3742 fail:
3743 	fdrop(fp, td);
3744 	return (error);
3745 }
3746 
3747 #ifndef _SYS_SYSPROTO_H_
3748 struct getdents_args {
3749 	int fd;
3750 	char *buf;
3751 	size_t count;
3752 };
3753 #endif
3754 int
3755 getdents(td, uap)
3756 	struct thread *td;
3757 	register struct getdents_args /* {
3758 		int fd;
3759 		char *buf;
3760 		u_int count;
3761 	} */ *uap;
3762 {
3763 	struct getdirentries_args ap;
3764 	ap.fd = uap->fd;
3765 	ap.buf = uap->buf;
3766 	ap.count = uap->count;
3767 	ap.basep = NULL;
3768 	return (getdirentries(td, &ap));
3769 }
3770 
3771 /*
3772  * Set the mode mask for creation of filesystem nodes.
3773  */
3774 #ifndef _SYS_SYSPROTO_H_
3775 struct umask_args {
3776 	int	newmask;
3777 };
3778 #endif
3779 int
3780 umask(td, uap)
3781 	struct thread *td;
3782 	struct umask_args /* {
3783 		int newmask;
3784 	} */ *uap;
3785 {
3786 	register struct filedesc *fdp;
3787 
3788 	FILEDESC_XLOCK(td->td_proc->p_fd);
3789 	fdp = td->td_proc->p_fd;
3790 	td->td_retval[0] = fdp->fd_cmask;
3791 	fdp->fd_cmask = uap->newmask & ALLPERMS;
3792 	FILEDESC_XUNLOCK(td->td_proc->p_fd);
3793 	return (0);
3794 }
3795 
3796 /*
3797  * Void all references to file by ripping underlying filesystem away from
3798  * vnode.
3799  */
3800 #ifndef _SYS_SYSPROTO_H_
3801 struct revoke_args {
3802 	char	*path;
3803 };
3804 #endif
3805 int
3806 revoke(td, uap)
3807 	struct thread *td;
3808 	register struct revoke_args /* {
3809 		char *path;
3810 	} */ *uap;
3811 {
3812 	struct vnode *vp;
3813 	struct vattr vattr;
3814 	int error;
3815 	struct nameidata nd;
3816 	int vfslocked;
3817 
3818 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | MPSAFE | AUDITVNODE1,
3819 	    UIO_USERSPACE, uap->path, td);
3820 	if ((error = namei(&nd)) != 0)
3821 		return (error);
3822 	vfslocked = NDHASGIANT(&nd);
3823 	vp = nd.ni_vp;
3824 	NDFREE(&nd, NDF_ONLY_PNBUF);
3825 	if (vp->v_type != VCHR) {
3826 		error = EINVAL;
3827 		goto out;
3828 	}
3829 #ifdef MAC
3830 	error = mac_vnode_check_revoke(td->td_ucred, vp);
3831 	if (error)
3832 		goto out;
3833 #endif
3834 	error = VOP_GETATTR(vp, &vattr, td->td_ucred, td);
3835 	if (error)
3836 		goto out;
3837 	if (td->td_ucred->cr_uid != vattr.va_uid) {
3838 		error = priv_check(td, PRIV_VFS_ADMIN);
3839 		if (error)
3840 			goto out;
3841 	}
3842 	if (vcount(vp) > 1)
3843 		VOP_REVOKE(vp, REVOKEALL);
3844 out:
3845 	vput(vp);
3846 	VFS_UNLOCK_GIANT(vfslocked);
3847 	return (error);
3848 }
3849 
3850 /*
3851  * Convert a user file descriptor to a kernel file entry.
3852  * A reference on the file entry is held upon returning.
3853  */
3854 int
3855 getvnode(fdp, fd, fpp)
3856 	struct filedesc *fdp;
3857 	int fd;
3858 	struct file **fpp;
3859 {
3860 	int error;
3861 	struct file *fp;
3862 
3863 	fp = NULL;
3864 	if (fdp == NULL)
3865 		error = EBADF;
3866 	else {
3867 		FILEDESC_SLOCK(fdp);
3868 		if ((u_int)fd >= fdp->fd_nfiles ||
3869 		    (fp = fdp->fd_ofiles[fd]) == NULL)
3870 			error = EBADF;
3871 		else if (fp->f_vnode == NULL) {
3872 			fp = NULL;
3873 			error = EINVAL;
3874 		} else {
3875 			fhold(fp);
3876 			error = 0;
3877 		}
3878 		FILEDESC_SUNLOCK(fdp);
3879 	}
3880 	*fpp = fp;
3881 	return (error);
3882 }
3883 
3884 /*
3885  * Get an (NFS) file handle.
3886  */
3887 #ifndef _SYS_SYSPROTO_H_
3888 struct lgetfh_args {
3889 	char	*fname;
3890 	fhandle_t *fhp;
3891 };
3892 #endif
3893 int
3894 lgetfh(td, uap)
3895 	struct thread *td;
3896 	register struct lgetfh_args *uap;
3897 {
3898 	struct nameidata nd;
3899 	fhandle_t fh;
3900 	register struct vnode *vp;
3901 	int vfslocked;
3902 	int error;
3903 
3904 	error = priv_check(td, PRIV_VFS_GETFH);
3905 	if (error)
3906 		return (error);
3907 	NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | MPSAFE | AUDITVNODE1,
3908 	    UIO_USERSPACE, uap->fname, td);
3909 	error = namei(&nd);
3910 	if (error)
3911 		return (error);
3912 	vfslocked = NDHASGIANT(&nd);
3913 	NDFREE(&nd, NDF_ONLY_PNBUF);
3914 	vp = nd.ni_vp;
3915 	bzero(&fh, sizeof(fh));
3916 	fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
3917 	error = VOP_VPTOFH(vp, &fh.fh_fid);
3918 	vput(vp);
3919 	VFS_UNLOCK_GIANT(vfslocked);
3920 	if (error)
3921 		return (error);
3922 	error = copyout(&fh, uap->fhp, sizeof (fh));
3923 	return (error);
3924 }
3925 
3926 #ifndef _SYS_SYSPROTO_H_
3927 struct getfh_args {
3928 	char	*fname;
3929 	fhandle_t *fhp;
3930 };
3931 #endif
3932 int
3933 getfh(td, uap)
3934 	struct thread *td;
3935 	register struct getfh_args *uap;
3936 {
3937 	struct nameidata nd;
3938 	fhandle_t fh;
3939 	register struct vnode *vp;
3940 	int vfslocked;
3941 	int error;
3942 
3943 	error = priv_check(td, PRIV_VFS_GETFH);
3944 	if (error)
3945 		return (error);
3946 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | MPSAFE | AUDITVNODE1,
3947 	    UIO_USERSPACE, uap->fname, td);
3948 	error = namei(&nd);
3949 	if (error)
3950 		return (error);
3951 	vfslocked = NDHASGIANT(&nd);
3952 	NDFREE(&nd, NDF_ONLY_PNBUF);
3953 	vp = nd.ni_vp;
3954 	bzero(&fh, sizeof(fh));
3955 	fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
3956 	error = VOP_VPTOFH(vp, &fh.fh_fid);
3957 	vput(vp);
3958 	VFS_UNLOCK_GIANT(vfslocked);
3959 	if (error)
3960 		return (error);
3961 	error = copyout(&fh, uap->fhp, sizeof (fh));
3962 	return (error);
3963 }
3964 
3965 /*
3966  * syscall for the rpc.lockd to use to translate a NFS file handle into an
3967  * open descriptor.
3968  *
3969  * warning: do not remove the priv_check() call or this becomes one giant
3970  * security hole.
3971  */
3972 #ifndef _SYS_SYSPROTO_H_
3973 struct fhopen_args {
3974 	const struct fhandle *u_fhp;
3975 	int flags;
3976 };
3977 #endif
3978 int
3979 fhopen(td, uap)
3980 	struct thread *td;
3981 	struct fhopen_args /* {
3982 		const struct fhandle *u_fhp;
3983 		int flags;
3984 	} */ *uap;
3985 {
3986 	struct proc *p = td->td_proc;
3987 	struct mount *mp;
3988 	struct vnode *vp;
3989 	struct fhandle fhp;
3990 	struct vattr vat;
3991 	struct vattr *vap = &vat;
3992 	struct flock lf;
3993 	struct file *fp;
3994 	register struct filedesc *fdp = p->p_fd;
3995 	int fmode, mode, error, type;
3996 	struct file *nfp;
3997 	int vfslocked;
3998 	int indx;
3999 
4000 	error = priv_check(td, PRIV_VFS_FHOPEN);
4001 	if (error)
4002 		return (error);
4003 	fmode = FFLAGS(uap->flags);
4004 	/* why not allow a non-read/write open for our lockd? */
4005 	if (((fmode & (FREAD | FWRITE)) == 0) || (fmode & O_CREAT))
4006 		return (EINVAL);
4007 	error = copyin(uap->u_fhp, &fhp, sizeof(fhp));
4008 	if (error)
4009 		return(error);
4010 	/* find the mount point */
4011 	mp = vfs_getvfs(&fhp.fh_fsid);
4012 	if (mp == NULL)
4013 		return (ESTALE);
4014 	vfslocked = VFS_LOCK_GIANT(mp);
4015 	/* now give me my vnode, it gets returned to me locked */
4016 	error = VFS_FHTOVP(mp, &fhp.fh_fid, &vp);
4017 	if (error)
4018 		goto out;
4019 	/*
4020 	 * from now on we have to make sure not
4021 	 * to forget about the vnode
4022 	 * any error that causes an abort must vput(vp)
4023 	 * just set error = err and 'goto bad;'.
4024 	 */
4025 
4026 	/*
4027 	 * from vn_open
4028 	 */
4029 	if (vp->v_type == VLNK) {
4030 		error = EMLINK;
4031 		goto bad;
4032 	}
4033 	if (vp->v_type == VSOCK) {
4034 		error = EOPNOTSUPP;
4035 		goto bad;
4036 	}
4037 	mode = 0;
4038 	if (fmode & (FWRITE | O_TRUNC)) {
4039 		if (vp->v_type == VDIR) {
4040 			error = EISDIR;
4041 			goto bad;
4042 		}
4043 		error = vn_writechk(vp);
4044 		if (error)
4045 			goto bad;
4046 		mode |= VWRITE;
4047 	}
4048 	if (fmode & FREAD)
4049 		mode |= VREAD;
4050 	if (fmode & O_APPEND)
4051 		mode |= VAPPEND;
4052 #ifdef MAC
4053 	error = mac_vnode_check_open(td->td_ucred, vp, mode);
4054 	if (error)
4055 		goto bad;
4056 #endif
4057 	if (mode) {
4058 		error = VOP_ACCESS(vp, mode, td->td_ucred, td);
4059 		if (error)
4060 			goto bad;
4061 	}
4062 	if (fmode & O_TRUNC) {
4063 		VOP_UNLOCK(vp, 0);				/* XXX */
4064 		if ((error = vn_start_write(NULL, &mp, V_WAIT | PCATCH)) != 0) {
4065 			vrele(vp);
4066 			goto out;
4067 		}
4068 		VOP_LEASE(vp, td, td->td_ucred, LEASE_WRITE);
4069 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);	/* XXX */
4070 #ifdef MAC
4071 		/*
4072 		 * We don't yet have fp->f_cred, so use td->td_ucred, which
4073 		 * should be right.
4074 		 */
4075 		error = mac_vnode_check_write(td->td_ucred, td->td_ucred, vp);
4076 		if (error == 0) {
4077 #endif
4078 			VATTR_NULL(vap);
4079 			vap->va_size = 0;
4080 			error = VOP_SETATTR(vp, vap, td->td_ucred, td);
4081 #ifdef MAC
4082 		}
4083 #endif
4084 		vn_finished_write(mp);
4085 		if (error)
4086 			goto bad;
4087 	}
4088 	error = VOP_OPEN(vp, fmode, td->td_ucred, td, NULL);
4089 	if (error)
4090 		goto bad;
4091 
4092 	if (fmode & FWRITE)
4093 		vp->v_writecount++;
4094 
4095 	/*
4096 	 * end of vn_open code
4097 	 */
4098 
4099 	if ((error = falloc(td, &nfp, &indx)) != 0) {
4100 		if (fmode & FWRITE)
4101 			vp->v_writecount--;
4102 		goto bad;
4103 	}
4104 	/* An extra reference on `nfp' has been held for us by falloc(). */
4105 	fp = nfp;
4106 	nfp->f_vnode = vp;
4107 	finit(nfp, fmode & FMASK, DTYPE_VNODE, vp, &vnops);
4108 	if (fmode & (O_EXLOCK | O_SHLOCK)) {
4109 		lf.l_whence = SEEK_SET;
4110 		lf.l_start = 0;
4111 		lf.l_len = 0;
4112 		if (fmode & O_EXLOCK)
4113 			lf.l_type = F_WRLCK;
4114 		else
4115 			lf.l_type = F_RDLCK;
4116 		type = F_FLOCK;
4117 		if ((fmode & FNONBLOCK) == 0)
4118 			type |= F_WAIT;
4119 		VOP_UNLOCK(vp, 0);
4120 		if ((error = VOP_ADVLOCK(vp, (caddr_t)fp, F_SETLK, &lf,
4121 			    type)) != 0) {
4122 			/*
4123 			 * The lock request failed.  Normally close the
4124 			 * descriptor but handle the case where someone might
4125 			 * have dup()d or close()d it when we weren't looking.
4126 			 */
4127 			fdclose(fdp, fp, indx, td);
4128 
4129 			/*
4130 			 * release our private reference
4131 			 */
4132 			fdrop(fp, td);
4133 			goto out;
4134 		}
4135 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
4136 		atomic_set_int(&fp->f_flag, FHASLOCK);
4137 	}
4138 
4139 	VOP_UNLOCK(vp, 0);
4140 	fdrop(fp, td);
4141 	vfs_rel(mp);
4142 	VFS_UNLOCK_GIANT(vfslocked);
4143 	td->td_retval[0] = indx;
4144 	return (0);
4145 
4146 bad:
4147 	vput(vp);
4148 out:
4149 	vfs_rel(mp);
4150 	VFS_UNLOCK_GIANT(vfslocked);
4151 	return (error);
4152 }
4153 
4154 /*
4155  * Stat an (NFS) file handle.
4156  */
4157 #ifndef _SYS_SYSPROTO_H_
4158 struct fhstat_args {
4159 	struct fhandle *u_fhp;
4160 	struct stat *sb;
4161 };
4162 #endif
4163 int
4164 fhstat(td, uap)
4165 	struct thread *td;
4166 	register struct fhstat_args /* {
4167 		struct fhandle *u_fhp;
4168 		struct stat *sb;
4169 	} */ *uap;
4170 {
4171 	struct stat sb;
4172 	fhandle_t fh;
4173 	struct mount *mp;
4174 	struct vnode *vp;
4175 	int vfslocked;
4176 	int error;
4177 
4178 	error = priv_check(td, PRIV_VFS_FHSTAT);
4179 	if (error)
4180 		return (error);
4181 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
4182 	if (error)
4183 		return (error);
4184 	if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL)
4185 		return (ESTALE);
4186 	vfslocked = VFS_LOCK_GIANT(mp);
4187 	if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp))) {
4188 		vfs_rel(mp);
4189 		VFS_UNLOCK_GIANT(vfslocked);
4190 		return (error);
4191 	}
4192 	error = vn_stat(vp, &sb, td->td_ucred, NOCRED, td);
4193 	vput(vp);
4194 	vfs_rel(mp);
4195 	VFS_UNLOCK_GIANT(vfslocked);
4196 	if (error)
4197 		return (error);
4198 	error = copyout(&sb, uap->sb, sizeof(sb));
4199 	return (error);
4200 }
4201 
4202 /*
4203  * Implement fstatfs() for (NFS) file handles.
4204  */
4205 #ifndef _SYS_SYSPROTO_H_
4206 struct fhstatfs_args {
4207 	struct fhandle *u_fhp;
4208 	struct statfs *buf;
4209 };
4210 #endif
4211 int
4212 fhstatfs(td, uap)
4213 	struct thread *td;
4214 	struct fhstatfs_args /* {
4215 		struct fhandle *u_fhp;
4216 		struct statfs *buf;
4217 	} */ *uap;
4218 {
4219 	struct statfs sf;
4220 	fhandle_t fh;
4221 	int error;
4222 
4223 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
4224 	if (error)
4225 		return (error);
4226 	error = kern_fhstatfs(td, fh, &sf);
4227 	if (error)
4228 		return (error);
4229 	return (copyout(&sf, uap->buf, sizeof(sf)));
4230 }
4231 
4232 int
4233 kern_fhstatfs(struct thread *td, fhandle_t fh, struct statfs *buf)
4234 {
4235 	struct statfs *sp;
4236 	struct mount *mp;
4237 	struct vnode *vp;
4238 	int vfslocked;
4239 	int error;
4240 
4241 	error = priv_check(td, PRIV_VFS_FHSTATFS);
4242 	if (error)
4243 		return (error);
4244 	if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL)
4245 		return (ESTALE);
4246 	vfslocked = VFS_LOCK_GIANT(mp);
4247 	error = VFS_FHTOVP(mp, &fh.fh_fid, &vp);
4248 	if (error) {
4249 		VFS_UNLOCK_GIANT(vfslocked);
4250 		vfs_rel(mp);
4251 		return (error);
4252 	}
4253 	vput(vp);
4254 	error = prison_canseemount(td->td_ucred, mp);
4255 	if (error)
4256 		goto out;
4257 #ifdef MAC
4258 	error = mac_mount_check_stat(td->td_ucred, mp);
4259 	if (error)
4260 		goto out;
4261 #endif
4262 	/*
4263 	 * Set these in case the underlying filesystem fails to do so.
4264 	 */
4265 	sp = &mp->mnt_stat;
4266 	sp->f_version = STATFS_VERSION;
4267 	sp->f_namemax = NAME_MAX;
4268 	sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
4269 	error = VFS_STATFS(mp, sp, td);
4270 	if (error == 0)
4271 		*buf = *sp;
4272 out:
4273 	vfs_rel(mp);
4274 	VFS_UNLOCK_GIANT(vfslocked);
4275 	return (error);
4276 }
4277