xref: /freebsd/sys/kern/vfs_syscalls.c (revision 20a4ce8bf215487ce53d30ccae3bb6c86babb5a2)
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  * 3. Neither the name of the University nor the names of its contributors
19  *    may be used to endorse or promote products derived from this software
20  *    without specific prior written permission.
21  *
22  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
23  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
24  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
25  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
26  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
27  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
28  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
29  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
30  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
31  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32  * SUCH DAMAGE.
33  *
34  *	@(#)vfs_syscalls.c	8.13 (Berkeley) 4/15/94
35  */
36 
37 #include <sys/cdefs.h>
38 __FBSDID("$FreeBSD$");
39 
40 #include "opt_capsicum.h"
41 #include "opt_compat.h"
42 #include "opt_ktrace.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/capsicum.h>
49 #include <sys/disk.h>
50 #include <sys/sysent.h>
51 #include <sys/malloc.h>
52 #include <sys/mount.h>
53 #include <sys/mutex.h>
54 #include <sys/sysproto.h>
55 #include <sys/namei.h>
56 #include <sys/filedesc.h>
57 #include <sys/kernel.h>
58 #include <sys/fcntl.h>
59 #include <sys/file.h>
60 #include <sys/filio.h>
61 #include <sys/limits.h>
62 #include <sys/linker.h>
63 #include <sys/rwlock.h>
64 #include <sys/sdt.h>
65 #include <sys/stat.h>
66 #include <sys/sx.h>
67 #include <sys/unistd.h>
68 #include <sys/vnode.h>
69 #include <sys/priv.h>
70 #include <sys/proc.h>
71 #include <sys/dirent.h>
72 #include <sys/jail.h>
73 #include <sys/syscallsubr.h>
74 #include <sys/sysctl.h>
75 #ifdef KTRACE
76 #include <sys/ktrace.h>
77 #endif
78 
79 #include <machine/stdarg.h>
80 
81 #include <security/audit/audit.h>
82 #include <security/mac/mac_framework.h>
83 
84 #include <vm/vm.h>
85 #include <vm/vm_object.h>
86 #include <vm/vm_page.h>
87 #include <vm/uma.h>
88 
89 #include <ufs/ufs/quota.h>
90 
91 MALLOC_DEFINE(M_FADVISE, "fadvise", "posix_fadvise(2) information");
92 
93 SDT_PROVIDER_DEFINE(vfs);
94 SDT_PROBE_DEFINE2(vfs, , stat, mode, "char *", "int");
95 SDT_PROBE_DEFINE2(vfs, , stat, reg, "char *", "int");
96 
97 static int kern_chflagsat(struct thread *td, int fd, const char *path,
98     enum uio_seg pathseg, u_long flags, int atflag);
99 static int setfflags(struct thread *td, struct vnode *, u_long);
100 static int getutimes(const struct timeval *, enum uio_seg, struct timespec *);
101 static int getutimens(const struct timespec *, enum uio_seg,
102     struct timespec *, int *);
103 static int setutimes(struct thread *td, struct vnode *,
104     const struct timespec *, int, int);
105 static int vn_access(struct vnode *vp, int user_flags, struct ucred *cred,
106     struct thread *td);
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 sys_sync(td, uap)
119 	struct thread *td;
120 	struct sync_args *uap;
121 {
122 	struct mount *mp, *nmp;
123 	int save;
124 
125 	mtx_lock(&mountlist_mtx);
126 	for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
127 		if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK)) {
128 			nmp = TAILQ_NEXT(mp, mnt_list);
129 			continue;
130 		}
131 		if ((mp->mnt_flag & MNT_RDONLY) == 0 &&
132 		    vn_start_write(NULL, &mp, V_NOWAIT) == 0) {
133 			save = curthread_pflags_set(TDP_SYNCIO);
134 			vfs_msync(mp, MNT_NOWAIT);
135 			VFS_SYNC(mp, MNT_NOWAIT);
136 			curthread_pflags_restore(save);
137 			vn_finished_write(mp);
138 		}
139 		mtx_lock(&mountlist_mtx);
140 		nmp = TAILQ_NEXT(mp, mnt_list);
141 		vfs_unbusy(mp);
142 	}
143 	mtx_unlock(&mountlist_mtx);
144 	return (0);
145 }
146 
147 /*
148  * Change filesystem quotas.
149  */
150 #ifndef _SYS_SYSPROTO_H_
151 struct quotactl_args {
152 	char *path;
153 	int cmd;
154 	int uid;
155 	caddr_t arg;
156 };
157 #endif
158 int
159 sys_quotactl(td, uap)
160 	struct thread *td;
161 	register struct quotactl_args /* {
162 		char *path;
163 		int cmd;
164 		int uid;
165 		caddr_t arg;
166 	} */ *uap;
167 {
168 	struct mount *mp;
169 	struct nameidata nd;
170 	int error;
171 
172 	AUDIT_ARG_CMD(uap->cmd);
173 	AUDIT_ARG_UID(uap->uid);
174 	if (!prison_allow(td->td_ucred, PR_ALLOW_QUOTAS))
175 		return (EPERM);
176 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
177 	    uap->path, td);
178 	if ((error = namei(&nd)) != 0)
179 		return (error);
180 	NDFREE(&nd, NDF_ONLY_PNBUF);
181 	mp = nd.ni_vp->v_mount;
182 	vfs_ref(mp);
183 	vput(nd.ni_vp);
184 	error = vfs_busy(mp, 0);
185 	vfs_rel(mp);
186 	if (error != 0)
187 		return (error);
188 	error = VFS_QUOTACTL(mp, uap->cmd, uap->uid, uap->arg);
189 
190 	/*
191 	 * Since quota on operation typically needs to open quota
192 	 * file, the Q_QUOTAON handler needs to unbusy the mount point
193 	 * before calling into namei.  Otherwise, unmount might be
194 	 * started between two vfs_busy() invocations (first is our,
195 	 * second is from mount point cross-walk code in lookup()),
196 	 * causing deadlock.
197 	 *
198 	 * Require that Q_QUOTAON handles the vfs_busy() reference on
199 	 * its own, always returning with ubusied mount point.
200 	 */
201 	if ((uap->cmd >> SUBCMDSHIFT) != Q_QUOTAON)
202 		vfs_unbusy(mp);
203 	return (error);
204 }
205 
206 /*
207  * Used by statfs conversion routines to scale the block size up if
208  * necessary so that all of the block counts are <= 'max_size'.  Note
209  * that 'max_size' should be a bitmask, i.e. 2^n - 1 for some non-zero
210  * value of 'n'.
211  */
212 void
213 statfs_scale_blocks(struct statfs *sf, long max_size)
214 {
215 	uint64_t count;
216 	int shift;
217 
218 	KASSERT(powerof2(max_size + 1), ("%s: invalid max_size", __func__));
219 
220 	/*
221 	 * Attempt to scale the block counts to give a more accurate
222 	 * overview to userland of the ratio of free space to used
223 	 * space.  To do this, find the largest block count and compute
224 	 * a divisor that lets it fit into a signed integer <= max_size.
225 	 */
226 	if (sf->f_bavail < 0)
227 		count = -sf->f_bavail;
228 	else
229 		count = sf->f_bavail;
230 	count = MAX(sf->f_blocks, MAX(sf->f_bfree, count));
231 	if (count <= max_size)
232 		return;
233 
234 	count >>= flsl(max_size);
235 	shift = 0;
236 	while (count > 0) {
237 		shift++;
238 		count >>=1;
239 	}
240 
241 	sf->f_bsize <<= shift;
242 	sf->f_blocks >>= shift;
243 	sf->f_bfree >>= shift;
244 	sf->f_bavail >>= shift;
245 }
246 
247 static int
248 kern_do_statfs(struct thread *td, struct mount *mp, struct statfs *buf)
249 {
250 	struct statfs *sp;
251 	int error;
252 
253 	if (mp == NULL)
254 		return (EBADF);
255 	error = vfs_busy(mp, 0);
256 	vfs_rel(mp);
257 	if (error != 0)
258 		return (error);
259 #ifdef MAC
260 	error = mac_mount_check_stat(td->td_ucred, mp);
261 	if (error != 0)
262 		goto out;
263 #endif
264 	/*
265 	 * Set these in case the underlying filesystem fails to do so.
266 	 */
267 	sp = &mp->mnt_stat;
268 	sp->f_version = STATFS_VERSION;
269 	sp->f_namemax = NAME_MAX;
270 	sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
271 	error = VFS_STATFS(mp, sp);
272 	if (error != 0)
273 		goto out;
274 	*buf = *sp;
275 	if (priv_check(td, PRIV_VFS_GENERATION)) {
276 		buf->f_fsid.val[0] = buf->f_fsid.val[1] = 0;
277 		prison_enforce_statfs(td->td_ucred, mp, buf);
278 	}
279 out:
280 	vfs_unbusy(mp);
281 	return (error);
282 }
283 
284 /*
285  * Get filesystem statistics.
286  */
287 #ifndef _SYS_SYSPROTO_H_
288 struct statfs_args {
289 	char *path;
290 	struct statfs *buf;
291 };
292 #endif
293 int
294 sys_statfs(td, uap)
295 	struct thread *td;
296 	register struct statfs_args /* {
297 		char *path;
298 		struct statfs *buf;
299 	} */ *uap;
300 {
301 	struct statfs sf;
302 	int error;
303 
304 	error = kern_statfs(td, uap->path, UIO_USERSPACE, &sf);
305 	if (error == 0)
306 		error = copyout(&sf, uap->buf, sizeof(sf));
307 	return (error);
308 }
309 
310 int
311 kern_statfs(struct thread *td, char *path, enum uio_seg pathseg,
312     struct statfs *buf)
313 {
314 	struct mount *mp;
315 	struct nameidata nd;
316 	int error;
317 
318 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
319 	    pathseg, path, td);
320 	error = namei(&nd);
321 	if (error != 0)
322 		return (error);
323 	mp = nd.ni_vp->v_mount;
324 	vfs_ref(mp);
325 	NDFREE(&nd, NDF_ONLY_PNBUF);
326 	vput(nd.ni_vp);
327 	return (kern_do_statfs(td, mp, buf));
328 }
329 
330 /*
331  * Get filesystem statistics.
332  */
333 #ifndef _SYS_SYSPROTO_H_
334 struct fstatfs_args {
335 	int fd;
336 	struct statfs *buf;
337 };
338 #endif
339 int
340 sys_fstatfs(td, uap)
341 	struct thread *td;
342 	register struct fstatfs_args /* {
343 		int fd;
344 		struct statfs *buf;
345 	} */ *uap;
346 {
347 	struct statfs sf;
348 	int error;
349 
350 	error = kern_fstatfs(td, uap->fd, &sf);
351 	if (error == 0)
352 		error = copyout(&sf, uap->buf, sizeof(sf));
353 	return (error);
354 }
355 
356 int
357 kern_fstatfs(struct thread *td, int fd, struct statfs *buf)
358 {
359 	struct file *fp;
360 	struct mount *mp;
361 	struct vnode *vp;
362 	cap_rights_t rights;
363 	int error;
364 
365 	AUDIT_ARG_FD(fd);
366 	error = getvnode(td, fd, cap_rights_init(&rights, CAP_FSTATFS), &fp);
367 	if (error != 0)
368 		return (error);
369 	vp = fp->f_vnode;
370 	vn_lock(vp, LK_SHARED | LK_RETRY);
371 #ifdef AUDIT
372 	AUDIT_ARG_VNODE1(vp);
373 #endif
374 	mp = vp->v_mount;
375 	if (mp != NULL)
376 		vfs_ref(mp);
377 	VOP_UNLOCK(vp, 0);
378 	fdrop(fp, td);
379 	return (kern_do_statfs(td, mp, buf));
380 }
381 
382 /*
383  * Get statistics on all filesystems.
384  */
385 #ifndef _SYS_SYSPROTO_H_
386 struct getfsstat_args {
387 	struct statfs *buf;
388 	long bufsize;
389 	int mode;
390 };
391 #endif
392 int
393 sys_getfsstat(td, uap)
394 	struct thread *td;
395 	register struct getfsstat_args /* {
396 		struct statfs *buf;
397 		long bufsize;
398 		int mode;
399 	} */ *uap;
400 {
401 	size_t count;
402 	int error;
403 
404 	if (uap->bufsize < 0 || uap->bufsize > SIZE_MAX)
405 		return (EINVAL);
406 	error = kern_getfsstat(td, &uap->buf, uap->bufsize, &count,
407 	    UIO_USERSPACE, uap->mode);
408 	if (error == 0)
409 		td->td_retval[0] = count;
410 	return (error);
411 }
412 
413 /*
414  * If (bufsize > 0 && bufseg == UIO_SYSSPACE)
415  *	The caller is responsible for freeing memory which will be allocated
416  *	in '*buf'.
417  */
418 int
419 kern_getfsstat(struct thread *td, struct statfs **buf, size_t bufsize,
420     size_t *countp, enum uio_seg bufseg, int mode)
421 {
422 	struct mount *mp, *nmp;
423 	struct statfs *sfsp, *sp, sb, *tofree;
424 	size_t count, maxcount;
425 	int error;
426 
427 	switch (mode) {
428 	case MNT_WAIT:
429 	case MNT_NOWAIT:
430 		break;
431 	default:
432 		if (bufseg == UIO_SYSSPACE)
433 			*buf = NULL;
434 		return (EINVAL);
435 	}
436 restart:
437 	maxcount = bufsize / sizeof(struct statfs);
438 	if (bufsize == 0) {
439 		sfsp = NULL;
440 		tofree = NULL;
441 	} else if (bufseg == UIO_USERSPACE) {
442 		sfsp = *buf;
443 		tofree = NULL;
444 	} else /* if (bufseg == UIO_SYSSPACE) */ {
445 		count = 0;
446 		mtx_lock(&mountlist_mtx);
447 		TAILQ_FOREACH(mp, &mountlist, mnt_list) {
448 			count++;
449 		}
450 		mtx_unlock(&mountlist_mtx);
451 		if (maxcount > count)
452 			maxcount = count;
453 		tofree = sfsp = *buf = malloc(maxcount * sizeof(struct statfs),
454 		    M_TEMP, M_WAITOK);
455 	}
456 	count = 0;
457 	mtx_lock(&mountlist_mtx);
458 	for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
459 		if (prison_canseemount(td->td_ucred, mp) != 0) {
460 			nmp = TAILQ_NEXT(mp, mnt_list);
461 			continue;
462 		}
463 #ifdef MAC
464 		if (mac_mount_check_stat(td->td_ucred, mp) != 0) {
465 			nmp = TAILQ_NEXT(mp, mnt_list);
466 			continue;
467 		}
468 #endif
469 		if (mode == MNT_WAIT) {
470 			if (vfs_busy(mp, MBF_MNTLSTLOCK) != 0) {
471 				/*
472 				 * If vfs_busy() failed, and MBF_NOWAIT
473 				 * wasn't passed, then the mp is gone.
474 				 * Furthermore, because of MBF_MNTLSTLOCK,
475 				 * the mountlist_mtx was dropped.  We have
476 				 * no other choice than to start over.
477 				 */
478 				mtx_unlock(&mountlist_mtx);
479 				free(tofree, M_TEMP);
480 				goto restart;
481 			}
482 		} else {
483 			if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK) != 0) {
484 				nmp = TAILQ_NEXT(mp, mnt_list);
485 				continue;
486 			}
487 		}
488 		if (sfsp && count < maxcount) {
489 			sp = &mp->mnt_stat;
490 			/*
491 			 * Set these in case the underlying filesystem
492 			 * fails to do so.
493 			 */
494 			sp->f_version = STATFS_VERSION;
495 			sp->f_namemax = NAME_MAX;
496 			sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
497 			/*
498 			 * If MNT_NOWAIT is specified, do not refresh
499 			 * the fsstat cache.
500 			 */
501 			if (mode != MNT_NOWAIT) {
502 				error = VFS_STATFS(mp, sp);
503 				if (error != 0) {
504 					mtx_lock(&mountlist_mtx);
505 					nmp = TAILQ_NEXT(mp, mnt_list);
506 					vfs_unbusy(mp);
507 					continue;
508 				}
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 != 0) {
521 					vfs_unbusy(mp);
522 					return (error);
523 				}
524 			}
525 			sfsp++;
526 		}
527 		count++;
528 		mtx_lock(&mountlist_mtx);
529 		nmp = TAILQ_NEXT(mp, mnt_list);
530 		vfs_unbusy(mp);
531 	}
532 	mtx_unlock(&mountlist_mtx);
533 	if (sfsp && count > maxcount)
534 		*countp = maxcount;
535 	else
536 		*countp = count;
537 	return (0);
538 }
539 
540 #ifdef COMPAT_FREEBSD4
541 /*
542  * Get old format filesystem statistics.
543  */
544 static void cvtstatfs(struct statfs *, struct ostatfs *);
545 
546 #ifndef _SYS_SYSPROTO_H_
547 struct freebsd4_statfs_args {
548 	char *path;
549 	struct ostatfs *buf;
550 };
551 #endif
552 int
553 freebsd4_statfs(td, uap)
554 	struct thread *td;
555 	struct freebsd4_statfs_args /* {
556 		char *path;
557 		struct ostatfs *buf;
558 	} */ *uap;
559 {
560 	struct ostatfs osb;
561 	struct statfs sf;
562 	int error;
563 
564 	error = kern_statfs(td, uap->path, UIO_USERSPACE, &sf);
565 	if (error != 0)
566 		return (error);
567 	cvtstatfs(&sf, &osb);
568 	return (copyout(&osb, uap->buf, sizeof(osb)));
569 }
570 
571 /*
572  * Get filesystem statistics.
573  */
574 #ifndef _SYS_SYSPROTO_H_
575 struct freebsd4_fstatfs_args {
576 	int fd;
577 	struct ostatfs *buf;
578 };
579 #endif
580 int
581 freebsd4_fstatfs(td, uap)
582 	struct thread *td;
583 	struct freebsd4_fstatfs_args /* {
584 		int fd;
585 		struct ostatfs *buf;
586 	} */ *uap;
587 {
588 	struct ostatfs osb;
589 	struct statfs sf;
590 	int error;
591 
592 	error = kern_fstatfs(td, uap->fd, &sf);
593 	if (error != 0)
594 		return (error);
595 	cvtstatfs(&sf, &osb);
596 	return (copyout(&osb, uap->buf, sizeof(osb)));
597 }
598 
599 /*
600  * Get statistics on all filesystems.
601  */
602 #ifndef _SYS_SYSPROTO_H_
603 struct freebsd4_getfsstat_args {
604 	struct ostatfs *buf;
605 	long bufsize;
606 	int mode;
607 };
608 #endif
609 int
610 freebsd4_getfsstat(td, uap)
611 	struct thread *td;
612 	register struct freebsd4_getfsstat_args /* {
613 		struct ostatfs *buf;
614 		long bufsize;
615 		int mode;
616 	} */ *uap;
617 {
618 	struct statfs *buf, *sp;
619 	struct ostatfs osb;
620 	size_t count, size;
621 	int error;
622 
623 	if (uap->bufsize < 0)
624 		return (EINVAL);
625 	count = uap->bufsize / sizeof(struct ostatfs);
626 	if (count > SIZE_MAX / sizeof(struct statfs))
627 		return (EINVAL);
628 	size = count * sizeof(struct statfs);
629 	error = kern_getfsstat(td, &buf, size, &count, UIO_SYSSPACE,
630 	    uap->mode);
631 	td->td_retval[0] = count;
632 	if (size != 0) {
633 		sp = buf;
634 		while (count != 0 && error == 0) {
635 			cvtstatfs(sp, &osb);
636 			error = copyout(&osb, uap->buf, sizeof(osb));
637 			sp++;
638 			uap->buf++;
639 			count--;
640 		}
641 		free(buf, M_TEMP);
642 	}
643 	return (error);
644 }
645 
646 /*
647  * Implement fstatfs() for (NFS) file handles.
648  */
649 #ifndef _SYS_SYSPROTO_H_
650 struct freebsd4_fhstatfs_args {
651 	struct fhandle *u_fhp;
652 	struct ostatfs *buf;
653 };
654 #endif
655 int
656 freebsd4_fhstatfs(td, uap)
657 	struct thread *td;
658 	struct freebsd4_fhstatfs_args /* {
659 		struct fhandle *u_fhp;
660 		struct ostatfs *buf;
661 	} */ *uap;
662 {
663 	struct ostatfs osb;
664 	struct statfs sf;
665 	fhandle_t fh;
666 	int error;
667 
668 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
669 	if (error != 0)
670 		return (error);
671 	error = kern_fhstatfs(td, fh, &sf);
672 	if (error != 0)
673 		return (error);
674 	cvtstatfs(&sf, &osb);
675 	return (copyout(&osb, uap->buf, sizeof(osb)));
676 }
677 
678 /*
679  * Convert a new format statfs structure to an old format statfs structure.
680  */
681 static void
682 cvtstatfs(nsp, osp)
683 	struct statfs *nsp;
684 	struct ostatfs *osp;
685 {
686 
687 	statfs_scale_blocks(nsp, LONG_MAX);
688 	bzero(osp, sizeof(*osp));
689 	osp->f_bsize = nsp->f_bsize;
690 	osp->f_iosize = MIN(nsp->f_iosize, LONG_MAX);
691 	osp->f_blocks = nsp->f_blocks;
692 	osp->f_bfree = nsp->f_bfree;
693 	osp->f_bavail = nsp->f_bavail;
694 	osp->f_files = MIN(nsp->f_files, LONG_MAX);
695 	osp->f_ffree = MIN(nsp->f_ffree, LONG_MAX);
696 	osp->f_owner = nsp->f_owner;
697 	osp->f_type = nsp->f_type;
698 	osp->f_flags = nsp->f_flags;
699 	osp->f_syncwrites = MIN(nsp->f_syncwrites, LONG_MAX);
700 	osp->f_asyncwrites = MIN(nsp->f_asyncwrites, LONG_MAX);
701 	osp->f_syncreads = MIN(nsp->f_syncreads, LONG_MAX);
702 	osp->f_asyncreads = MIN(nsp->f_asyncreads, LONG_MAX);
703 	strlcpy(osp->f_fstypename, nsp->f_fstypename,
704 	    MIN(MFSNAMELEN, OMFSNAMELEN));
705 	strlcpy(osp->f_mntonname, nsp->f_mntonname,
706 	    MIN(MNAMELEN, OMNAMELEN));
707 	strlcpy(osp->f_mntfromname, nsp->f_mntfromname,
708 	    MIN(MNAMELEN, OMNAMELEN));
709 	osp->f_fsid = nsp->f_fsid;
710 }
711 #endif /* COMPAT_FREEBSD4 */
712 
713 /*
714  * Change current working directory to a given file descriptor.
715  */
716 #ifndef _SYS_SYSPROTO_H_
717 struct fchdir_args {
718 	int	fd;
719 };
720 #endif
721 int
722 sys_fchdir(td, uap)
723 	struct thread *td;
724 	struct fchdir_args /* {
725 		int fd;
726 	} */ *uap;
727 {
728 	struct vnode *vp, *tdp;
729 	struct mount *mp;
730 	struct file *fp;
731 	cap_rights_t rights;
732 	int error;
733 
734 	AUDIT_ARG_FD(uap->fd);
735 	error = getvnode(td, uap->fd, cap_rights_init(&rights, CAP_FCHDIR),
736 	    &fp);
737 	if (error != 0)
738 		return (error);
739 	vp = fp->f_vnode;
740 	vrefact(vp);
741 	fdrop(fp, td);
742 	vn_lock(vp, LK_SHARED | LK_RETRY);
743 	AUDIT_ARG_VNODE1(vp);
744 	error = change_dir(vp, td);
745 	while (!error && (mp = vp->v_mountedhere) != NULL) {
746 		if (vfs_busy(mp, 0))
747 			continue;
748 		error = VFS_ROOT(mp, LK_SHARED, &tdp);
749 		vfs_unbusy(mp);
750 		if (error != 0)
751 			break;
752 		vput(vp);
753 		vp = tdp;
754 	}
755 	if (error != 0) {
756 		vput(vp);
757 		return (error);
758 	}
759 	VOP_UNLOCK(vp, 0);
760 	pwd_chdir(td, vp);
761 	return (0);
762 }
763 
764 /*
765  * Change current working directory (``.'').
766  */
767 #ifndef _SYS_SYSPROTO_H_
768 struct chdir_args {
769 	char	*path;
770 };
771 #endif
772 int
773 sys_chdir(td, uap)
774 	struct thread *td;
775 	struct chdir_args /* {
776 		char *path;
777 	} */ *uap;
778 {
779 
780 	return (kern_chdir(td, uap->path, UIO_USERSPACE));
781 }
782 
783 int
784 kern_chdir(struct thread *td, char *path, enum uio_seg pathseg)
785 {
786 	struct nameidata nd;
787 	int error;
788 
789 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
790 	    pathseg, path, td);
791 	if ((error = namei(&nd)) != 0)
792 		return (error);
793 	if ((error = change_dir(nd.ni_vp, td)) != 0) {
794 		vput(nd.ni_vp);
795 		NDFREE(&nd, NDF_ONLY_PNBUF);
796 		return (error);
797 	}
798 	VOP_UNLOCK(nd.ni_vp, 0);
799 	NDFREE(&nd, NDF_ONLY_PNBUF);
800 	pwd_chdir(td, nd.ni_vp);
801 	return (0);
802 }
803 
804 /*
805  * Change notion of root (``/'') directory.
806  */
807 #ifndef _SYS_SYSPROTO_H_
808 struct chroot_args {
809 	char	*path;
810 };
811 #endif
812 int
813 sys_chroot(td, uap)
814 	struct thread *td;
815 	struct chroot_args /* {
816 		char *path;
817 	} */ *uap;
818 {
819 	struct nameidata nd;
820 	int error;
821 
822 	error = priv_check(td, PRIV_VFS_CHROOT);
823 	if (error != 0)
824 		return (error);
825 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
826 	    UIO_USERSPACE, uap->path, td);
827 	error = namei(&nd);
828 	if (error != 0)
829 		goto error;
830 	error = change_dir(nd.ni_vp, td);
831 	if (error != 0)
832 		goto e_vunlock;
833 #ifdef MAC
834 	error = mac_vnode_check_chroot(td->td_ucred, nd.ni_vp);
835 	if (error != 0)
836 		goto e_vunlock;
837 #endif
838 	VOP_UNLOCK(nd.ni_vp, 0);
839 	error = pwd_chroot(td, nd.ni_vp);
840 	vrele(nd.ni_vp);
841 	NDFREE(&nd, NDF_ONLY_PNBUF);
842 	return (error);
843 e_vunlock:
844 	vput(nd.ni_vp);
845 error:
846 	NDFREE(&nd, NDF_ONLY_PNBUF);
847 	return (error);
848 }
849 
850 /*
851  * Common routine for chroot and chdir.  Callers must provide a locked vnode
852  * instance.
853  */
854 int
855 change_dir(vp, td)
856 	struct vnode *vp;
857 	struct thread *td;
858 {
859 #ifdef MAC
860 	int error;
861 #endif
862 
863 	ASSERT_VOP_LOCKED(vp, "change_dir(): vp not locked");
864 	if (vp->v_type != VDIR)
865 		return (ENOTDIR);
866 #ifdef MAC
867 	error = mac_vnode_check_chdir(td->td_ucred, vp);
868 	if (error != 0)
869 		return (error);
870 #endif
871 	return (VOP_ACCESS(vp, VEXEC, td->td_ucred, td));
872 }
873 
874 static __inline void
875 flags_to_rights(int flags, cap_rights_t *rightsp)
876 {
877 
878 	if (flags & O_EXEC) {
879 		cap_rights_set(rightsp, CAP_FEXECVE);
880 	} else {
881 		switch ((flags & O_ACCMODE)) {
882 		case O_RDONLY:
883 			cap_rights_set(rightsp, CAP_READ);
884 			break;
885 		case O_RDWR:
886 			cap_rights_set(rightsp, CAP_READ);
887 			/* FALLTHROUGH */
888 		case O_WRONLY:
889 			cap_rights_set(rightsp, CAP_WRITE);
890 			if (!(flags & (O_APPEND | O_TRUNC)))
891 				cap_rights_set(rightsp, CAP_SEEK);
892 			break;
893 		}
894 	}
895 
896 	if (flags & O_CREAT)
897 		cap_rights_set(rightsp, CAP_CREATE);
898 
899 	if (flags & O_TRUNC)
900 		cap_rights_set(rightsp, CAP_FTRUNCATE);
901 
902 	if (flags & (O_SYNC | O_FSYNC))
903 		cap_rights_set(rightsp, CAP_FSYNC);
904 
905 	if (flags & (O_EXLOCK | O_SHLOCK))
906 		cap_rights_set(rightsp, CAP_FLOCK);
907 }
908 
909 /*
910  * Check permissions, allocate an open file structure, and call the device
911  * open routine if any.
912  */
913 #ifndef _SYS_SYSPROTO_H_
914 struct open_args {
915 	char	*path;
916 	int	flags;
917 	int	mode;
918 };
919 #endif
920 int
921 sys_open(td, uap)
922 	struct thread *td;
923 	register struct open_args /* {
924 		char *path;
925 		int flags;
926 		int mode;
927 	} */ *uap;
928 {
929 
930 	return (kern_openat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
931 	    uap->flags, uap->mode));
932 }
933 
934 #ifndef _SYS_SYSPROTO_H_
935 struct openat_args {
936 	int	fd;
937 	char	*path;
938 	int	flag;
939 	int	mode;
940 };
941 #endif
942 int
943 sys_openat(struct thread *td, struct openat_args *uap)
944 {
945 
946 	AUDIT_ARG_FD(uap->fd);
947 	return (kern_openat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
948 	    uap->mode));
949 }
950 
951 int
952 kern_openat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
953     int flags, int mode)
954 {
955 	struct proc *p = td->td_proc;
956 	struct filedesc *fdp = p->p_fd;
957 	struct file *fp;
958 	struct vnode *vp;
959 	struct nameidata nd;
960 	cap_rights_t rights;
961 	int cmode, error, indx;
962 
963 	indx = -1;
964 
965 	AUDIT_ARG_FFLAGS(flags);
966 	AUDIT_ARG_MODE(mode);
967 	cap_rights_init(&rights, CAP_LOOKUP);
968 	flags_to_rights(flags, &rights);
969 	/*
970 	 * Only one of the O_EXEC, O_RDONLY, O_WRONLY and O_RDWR flags
971 	 * may be specified.
972 	 */
973 	if (flags & O_EXEC) {
974 		if (flags & O_ACCMODE)
975 			return (EINVAL);
976 	} else if ((flags & O_ACCMODE) == O_ACCMODE) {
977 		return (EINVAL);
978 	} else {
979 		flags = FFLAGS(flags);
980 	}
981 
982 	/*
983 	 * Allocate a file structure. The descriptor to reference it
984 	 * is allocated and set by finstall() below.
985 	 */
986 	error = falloc_noinstall(td, &fp);
987 	if (error != 0)
988 		return (error);
989 	/*
990 	 * An extra reference on `fp' has been held for us by
991 	 * falloc_noinstall().
992 	 */
993 	/* Set the flags early so the finit in devfs can pick them up. */
994 	fp->f_flag = flags & FMASK;
995 	cmode = ((mode & ~fdp->fd_cmask) & ALLPERMS) & ~S_ISTXT;
996 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, fd,
997 	    &rights, td);
998 	td->td_dupfd = -1;		/* XXX check for fdopen */
999 	error = vn_open(&nd, &flags, cmode, fp);
1000 	if (error != 0) {
1001 		/*
1002 		 * If the vn_open replaced the method vector, something
1003 		 * wonderous happened deep below and we just pass it up
1004 		 * pretending we know what we do.
1005 		 */
1006 		if (error == ENXIO && fp->f_ops != &badfileops)
1007 			goto success;
1008 
1009 		/*
1010 		 * Handle special fdopen() case. bleh.
1011 		 *
1012 		 * Don't do this for relative (capability) lookups; we don't
1013 		 * understand exactly what would happen, and we don't think
1014 		 * that it ever should.
1015 		 */
1016 		if ((nd.ni_lcf & NI_LCF_STRICTRELATIVE) == 0 &&
1017 		    (error == ENODEV || error == ENXIO) &&
1018 		    td->td_dupfd >= 0) {
1019 			error = dupfdopen(td, fdp, td->td_dupfd, flags, error,
1020 			    &indx);
1021 			if (error == 0)
1022 				goto success;
1023 		}
1024 
1025 		goto bad;
1026 	}
1027 	td->td_dupfd = 0;
1028 	NDFREE(&nd, NDF_ONLY_PNBUF);
1029 	vp = nd.ni_vp;
1030 
1031 	/*
1032 	 * Store the vnode, for any f_type. Typically, the vnode use
1033 	 * count is decremented by direct call to vn_closefile() for
1034 	 * files that switched type in the cdevsw fdopen() method.
1035 	 */
1036 	fp->f_vnode = vp;
1037 	/*
1038 	 * If the file wasn't claimed by devfs bind it to the normal
1039 	 * vnode operations here.
1040 	 */
1041 	if (fp->f_ops == &badfileops) {
1042 		KASSERT(vp->v_type != VFIFO, ("Unexpected fifo."));
1043 		fp->f_seqcount = 1;
1044 		finit(fp, (flags & FMASK) | (fp->f_flag & FHASLOCK),
1045 		    DTYPE_VNODE, vp, &vnops);
1046 	}
1047 
1048 	VOP_UNLOCK(vp, 0);
1049 	if (flags & O_TRUNC) {
1050 		error = fo_truncate(fp, 0, td->td_ucred, td);
1051 		if (error != 0)
1052 			goto bad;
1053 	}
1054 success:
1055 	/*
1056 	 * If we haven't already installed the FD (for dupfdopen), do so now.
1057 	 */
1058 	if (indx == -1) {
1059 		struct filecaps *fcaps;
1060 
1061 #ifdef CAPABILITIES
1062 		if ((nd.ni_lcf & NI_LCF_STRICTRELATIVE) != 0)
1063 			fcaps = &nd.ni_filecaps;
1064 		else
1065 #endif
1066 			fcaps = NULL;
1067 		error = finstall(td, fp, &indx, flags, fcaps);
1068 		/* On success finstall() consumes fcaps. */
1069 		if (error != 0) {
1070 			filecaps_free(&nd.ni_filecaps);
1071 			goto bad;
1072 		}
1073 	} else {
1074 		filecaps_free(&nd.ni_filecaps);
1075 	}
1076 
1077 	/*
1078 	 * Release our private reference, leaving the one associated with
1079 	 * the descriptor table intact.
1080 	 */
1081 	fdrop(fp, td);
1082 	td->td_retval[0] = indx;
1083 	return (0);
1084 bad:
1085 	KASSERT(indx == -1, ("indx=%d, should be -1", indx));
1086 	fdrop(fp, td);
1087 	return (error);
1088 }
1089 
1090 #ifdef COMPAT_43
1091 /*
1092  * Create a file.
1093  */
1094 #ifndef _SYS_SYSPROTO_H_
1095 struct ocreat_args {
1096 	char	*path;
1097 	int	mode;
1098 };
1099 #endif
1100 int
1101 ocreat(td, uap)
1102 	struct thread *td;
1103 	register struct ocreat_args /* {
1104 		char *path;
1105 		int mode;
1106 	} */ *uap;
1107 {
1108 
1109 	return (kern_openat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1110 	    O_WRONLY | O_CREAT | O_TRUNC, uap->mode));
1111 }
1112 #endif /* COMPAT_43 */
1113 
1114 /*
1115  * Create a special file.
1116  */
1117 #ifndef _SYS_SYSPROTO_H_
1118 struct mknod_args {
1119 	char	*path;
1120 	int	mode;
1121 	int	dev;
1122 };
1123 #endif
1124 int
1125 sys_mknod(td, uap)
1126 	struct thread *td;
1127 	register struct mknod_args /* {
1128 		char *path;
1129 		int mode;
1130 		int dev;
1131 	} */ *uap;
1132 {
1133 
1134 	return (kern_mknodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1135 	    uap->mode, uap->dev));
1136 }
1137 
1138 #ifndef _SYS_SYSPROTO_H_
1139 struct mknodat_args {
1140 	int	fd;
1141 	char	*path;
1142 	mode_t	mode;
1143 	dev_t	dev;
1144 };
1145 #endif
1146 int
1147 sys_mknodat(struct thread *td, struct mknodat_args *uap)
1148 {
1149 
1150 	return (kern_mknodat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode,
1151 	    uap->dev));
1152 }
1153 
1154 int
1155 kern_mknodat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1156     int mode, int dev)
1157 {
1158 	struct vnode *vp;
1159 	struct mount *mp;
1160 	struct vattr vattr;
1161 	struct nameidata nd;
1162 	cap_rights_t rights;
1163 	int error, whiteout = 0;
1164 
1165 	AUDIT_ARG_MODE(mode);
1166 	AUDIT_ARG_DEV(dev);
1167 	switch (mode & S_IFMT) {
1168 	case S_IFCHR:
1169 	case S_IFBLK:
1170 		error = priv_check(td, PRIV_VFS_MKNOD_DEV);
1171 		if (error == 0 && dev == VNOVAL)
1172 			error = EINVAL;
1173 		break;
1174 	case S_IFMT:
1175 		error = priv_check(td, PRIV_VFS_MKNOD_BAD);
1176 		break;
1177 	case S_IFWHT:
1178 		error = priv_check(td, PRIV_VFS_MKNOD_WHT);
1179 		break;
1180 	case S_IFIFO:
1181 		if (dev == 0)
1182 			return (kern_mkfifoat(td, fd, path, pathseg, mode));
1183 		/* FALLTHROUGH */
1184 	default:
1185 		error = EINVAL;
1186 		break;
1187 	}
1188 	if (error != 0)
1189 		return (error);
1190 restart:
1191 	bwillwrite();
1192 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
1193 	    NOCACHE, pathseg, path, fd, cap_rights_init(&rights, CAP_MKNODAT),
1194 	    td);
1195 	if ((error = namei(&nd)) != 0)
1196 		return (error);
1197 	vp = nd.ni_vp;
1198 	if (vp != NULL) {
1199 		NDFREE(&nd, NDF_ONLY_PNBUF);
1200 		if (vp == nd.ni_dvp)
1201 			vrele(nd.ni_dvp);
1202 		else
1203 			vput(nd.ni_dvp);
1204 		vrele(vp);
1205 		return (EEXIST);
1206 	} else {
1207 		VATTR_NULL(&vattr);
1208 		vattr.va_mode = (mode & ALLPERMS) &
1209 		    ~td->td_proc->p_fd->fd_cmask;
1210 		vattr.va_rdev = dev;
1211 		whiteout = 0;
1212 
1213 		switch (mode & S_IFMT) {
1214 		case S_IFMT:	/* used by badsect to flag bad sectors */
1215 			vattr.va_type = VBAD;
1216 			break;
1217 		case S_IFCHR:
1218 			vattr.va_type = VCHR;
1219 			break;
1220 		case S_IFBLK:
1221 			vattr.va_type = VBLK;
1222 			break;
1223 		case S_IFWHT:
1224 			whiteout = 1;
1225 			break;
1226 		default:
1227 			panic("kern_mknod: invalid mode");
1228 		}
1229 	}
1230 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1231 		NDFREE(&nd, NDF_ONLY_PNBUF);
1232 		vput(nd.ni_dvp);
1233 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1234 			return (error);
1235 		goto restart;
1236 	}
1237 #ifdef MAC
1238 	if (error == 0 && !whiteout)
1239 		error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp,
1240 		    &nd.ni_cnd, &vattr);
1241 #endif
1242 	if (error == 0) {
1243 		if (whiteout)
1244 			error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE);
1245 		else {
1246 			error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp,
1247 						&nd.ni_cnd, &vattr);
1248 			if (error == 0)
1249 				vput(nd.ni_vp);
1250 		}
1251 	}
1252 	NDFREE(&nd, NDF_ONLY_PNBUF);
1253 	vput(nd.ni_dvp);
1254 	vn_finished_write(mp);
1255 	return (error);
1256 }
1257 
1258 /*
1259  * Create a named pipe.
1260  */
1261 #ifndef _SYS_SYSPROTO_H_
1262 struct mkfifo_args {
1263 	char	*path;
1264 	int	mode;
1265 };
1266 #endif
1267 int
1268 sys_mkfifo(td, uap)
1269 	struct thread *td;
1270 	register struct mkfifo_args /* {
1271 		char *path;
1272 		int mode;
1273 	} */ *uap;
1274 {
1275 
1276 	return (kern_mkfifoat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1277 	    uap->mode));
1278 }
1279 
1280 #ifndef _SYS_SYSPROTO_H_
1281 struct mkfifoat_args {
1282 	int	fd;
1283 	char	*path;
1284 	mode_t	mode;
1285 };
1286 #endif
1287 int
1288 sys_mkfifoat(struct thread *td, struct mkfifoat_args *uap)
1289 {
1290 
1291 	return (kern_mkfifoat(td, uap->fd, uap->path, UIO_USERSPACE,
1292 	    uap->mode));
1293 }
1294 
1295 int
1296 kern_mkfifoat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1297     int mode)
1298 {
1299 	struct mount *mp;
1300 	struct vattr vattr;
1301 	struct nameidata nd;
1302 	cap_rights_t rights;
1303 	int error;
1304 
1305 	AUDIT_ARG_MODE(mode);
1306 restart:
1307 	bwillwrite();
1308 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
1309 	    NOCACHE, pathseg, path, fd, cap_rights_init(&rights, CAP_MKFIFOAT),
1310 	    td);
1311 	if ((error = namei(&nd)) != 0)
1312 		return (error);
1313 	if (nd.ni_vp != NULL) {
1314 		NDFREE(&nd, NDF_ONLY_PNBUF);
1315 		if (nd.ni_vp == nd.ni_dvp)
1316 			vrele(nd.ni_dvp);
1317 		else
1318 			vput(nd.ni_dvp);
1319 		vrele(nd.ni_vp);
1320 		return (EEXIST);
1321 	}
1322 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1323 		NDFREE(&nd, NDF_ONLY_PNBUF);
1324 		vput(nd.ni_dvp);
1325 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1326 			return (error);
1327 		goto restart;
1328 	}
1329 	VATTR_NULL(&vattr);
1330 	vattr.va_type = VFIFO;
1331 	vattr.va_mode = (mode & ALLPERMS) & ~td->td_proc->p_fd->fd_cmask;
1332 #ifdef MAC
1333 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1334 	    &vattr);
1335 	if (error != 0)
1336 		goto out;
1337 #endif
1338 	error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
1339 	if (error == 0)
1340 		vput(nd.ni_vp);
1341 #ifdef MAC
1342 out:
1343 #endif
1344 	vput(nd.ni_dvp);
1345 	vn_finished_write(mp);
1346 	NDFREE(&nd, NDF_ONLY_PNBUF);
1347 	return (error);
1348 }
1349 
1350 /*
1351  * Make a hard file link.
1352  */
1353 #ifndef _SYS_SYSPROTO_H_
1354 struct link_args {
1355 	char	*path;
1356 	char	*link;
1357 };
1358 #endif
1359 int
1360 sys_link(td, uap)
1361 	struct thread *td;
1362 	register struct link_args /* {
1363 		char *path;
1364 		char *link;
1365 	} */ *uap;
1366 {
1367 
1368 	return (kern_linkat(td, AT_FDCWD, AT_FDCWD, uap->path, uap->link,
1369 	    UIO_USERSPACE, FOLLOW));
1370 }
1371 
1372 #ifndef _SYS_SYSPROTO_H_
1373 struct linkat_args {
1374 	int	fd1;
1375 	char	*path1;
1376 	int	fd2;
1377 	char	*path2;
1378 	int	flag;
1379 };
1380 #endif
1381 int
1382 sys_linkat(struct thread *td, struct linkat_args *uap)
1383 {
1384 	int flag;
1385 
1386 	flag = uap->flag;
1387 	if (flag & ~AT_SYMLINK_FOLLOW)
1388 		return (EINVAL);
1389 
1390 	return (kern_linkat(td, uap->fd1, uap->fd2, uap->path1, uap->path2,
1391 	    UIO_USERSPACE, (flag & AT_SYMLINK_FOLLOW) ? FOLLOW : NOFOLLOW));
1392 }
1393 
1394 int hardlink_check_uid = 0;
1395 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_uid, CTLFLAG_RW,
1396     &hardlink_check_uid, 0,
1397     "Unprivileged processes cannot create hard links to files owned by other "
1398     "users");
1399 static int hardlink_check_gid = 0;
1400 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_gid, CTLFLAG_RW,
1401     &hardlink_check_gid, 0,
1402     "Unprivileged processes cannot create hard links to files owned by other "
1403     "groups");
1404 
1405 static int
1406 can_hardlink(struct vnode *vp, struct ucred *cred)
1407 {
1408 	struct vattr va;
1409 	int error;
1410 
1411 	if (!hardlink_check_uid && !hardlink_check_gid)
1412 		return (0);
1413 
1414 	error = VOP_GETATTR(vp, &va, cred);
1415 	if (error != 0)
1416 		return (error);
1417 
1418 	if (hardlink_check_uid && cred->cr_uid != va.va_uid) {
1419 		error = priv_check_cred(cred, PRIV_VFS_LINK, 0);
1420 		if (error != 0)
1421 			return (error);
1422 	}
1423 
1424 	if (hardlink_check_gid && !groupmember(va.va_gid, cred)) {
1425 		error = priv_check_cred(cred, PRIV_VFS_LINK, 0);
1426 		if (error != 0)
1427 			return (error);
1428 	}
1429 
1430 	return (0);
1431 }
1432 
1433 int
1434 kern_linkat(struct thread *td, int fd1, int fd2, char *path1, char *path2,
1435     enum uio_seg segflg, int follow)
1436 {
1437 	struct vnode *vp;
1438 	struct mount *mp;
1439 	struct nameidata nd;
1440 	cap_rights_t rights;
1441 	int error;
1442 
1443 again:
1444 	bwillwrite();
1445 	NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, segflg, path1, fd1,
1446 	    cap_rights_init(&rights, CAP_LINKAT_SOURCE), td);
1447 
1448 	if ((error = namei(&nd)) != 0)
1449 		return (error);
1450 	NDFREE(&nd, NDF_ONLY_PNBUF);
1451 	vp = nd.ni_vp;
1452 	if (vp->v_type == VDIR) {
1453 		vrele(vp);
1454 		return (EPERM);		/* POSIX */
1455 	}
1456 	NDINIT_ATRIGHTS(&nd, CREATE,
1457 	    LOCKPARENT | SAVENAME | AUDITVNODE2 | NOCACHE, segflg, path2, fd2,
1458 	    cap_rights_init(&rights, CAP_LINKAT_TARGET), td);
1459 	if ((error = namei(&nd)) == 0) {
1460 		if (nd.ni_vp != NULL) {
1461 			NDFREE(&nd, NDF_ONLY_PNBUF);
1462 			if (nd.ni_dvp == nd.ni_vp)
1463 				vrele(nd.ni_dvp);
1464 			else
1465 				vput(nd.ni_dvp);
1466 			vrele(nd.ni_vp);
1467 			vrele(vp);
1468 			return (EEXIST);
1469 		} else if (nd.ni_dvp->v_mount != vp->v_mount) {
1470 			/*
1471 			 * Cross-device link.  No need to recheck
1472 			 * vp->v_type, since it cannot change, except
1473 			 * to VBAD.
1474 			 */
1475 			NDFREE(&nd, NDF_ONLY_PNBUF);
1476 			vput(nd.ni_dvp);
1477 			vrele(vp);
1478 			return (EXDEV);
1479 		} else if ((error = vn_lock(vp, LK_EXCLUSIVE)) == 0) {
1480 			error = can_hardlink(vp, td->td_ucred);
1481 #ifdef MAC
1482 			if (error == 0)
1483 				error = mac_vnode_check_link(td->td_ucred,
1484 				    nd.ni_dvp, vp, &nd.ni_cnd);
1485 #endif
1486 			if (error != 0) {
1487 				vput(vp);
1488 				vput(nd.ni_dvp);
1489 				NDFREE(&nd, NDF_ONLY_PNBUF);
1490 				return (error);
1491 			}
1492 			error = vn_start_write(vp, &mp, V_NOWAIT);
1493 			if (error != 0) {
1494 				vput(vp);
1495 				vput(nd.ni_dvp);
1496 				NDFREE(&nd, NDF_ONLY_PNBUF);
1497 				error = vn_start_write(NULL, &mp,
1498 				    V_XSLEEP | PCATCH);
1499 				if (error != 0)
1500 					return (error);
1501 				goto again;
1502 			}
1503 			error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
1504 			VOP_UNLOCK(vp, 0);
1505 			vput(nd.ni_dvp);
1506 			vn_finished_write(mp);
1507 			NDFREE(&nd, NDF_ONLY_PNBUF);
1508 		} else {
1509 			vput(nd.ni_dvp);
1510 			NDFREE(&nd, NDF_ONLY_PNBUF);
1511 			vrele(vp);
1512 			goto again;
1513 		}
1514 	}
1515 	vrele(vp);
1516 	return (error);
1517 }
1518 
1519 /*
1520  * Make a symbolic link.
1521  */
1522 #ifndef _SYS_SYSPROTO_H_
1523 struct symlink_args {
1524 	char	*path;
1525 	char	*link;
1526 };
1527 #endif
1528 int
1529 sys_symlink(td, uap)
1530 	struct thread *td;
1531 	register struct symlink_args /* {
1532 		char *path;
1533 		char *link;
1534 	} */ *uap;
1535 {
1536 
1537 	return (kern_symlinkat(td, uap->path, AT_FDCWD, uap->link,
1538 	    UIO_USERSPACE));
1539 }
1540 
1541 #ifndef _SYS_SYSPROTO_H_
1542 struct symlinkat_args {
1543 	char	*path;
1544 	int	fd;
1545 	char	*path2;
1546 };
1547 #endif
1548 int
1549 sys_symlinkat(struct thread *td, struct symlinkat_args *uap)
1550 {
1551 
1552 	return (kern_symlinkat(td, uap->path1, uap->fd, uap->path2,
1553 	    UIO_USERSPACE));
1554 }
1555 
1556 int
1557 kern_symlinkat(struct thread *td, char *path1, int fd, char *path2,
1558     enum uio_seg segflg)
1559 {
1560 	struct mount *mp;
1561 	struct vattr vattr;
1562 	char *syspath;
1563 	struct nameidata nd;
1564 	int error;
1565 	cap_rights_t rights;
1566 
1567 	if (segflg == UIO_SYSSPACE) {
1568 		syspath = path1;
1569 	} else {
1570 		syspath = uma_zalloc(namei_zone, M_WAITOK);
1571 		if ((error = copyinstr(path1, syspath, MAXPATHLEN, NULL)) != 0)
1572 			goto out;
1573 	}
1574 	AUDIT_ARG_TEXT(syspath);
1575 restart:
1576 	bwillwrite();
1577 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
1578 	    NOCACHE, segflg, path2, fd, cap_rights_init(&rights, CAP_SYMLINKAT),
1579 	    td);
1580 	if ((error = namei(&nd)) != 0)
1581 		goto out;
1582 	if (nd.ni_vp) {
1583 		NDFREE(&nd, NDF_ONLY_PNBUF);
1584 		if (nd.ni_vp == nd.ni_dvp)
1585 			vrele(nd.ni_dvp);
1586 		else
1587 			vput(nd.ni_dvp);
1588 		vrele(nd.ni_vp);
1589 		error = EEXIST;
1590 		goto out;
1591 	}
1592 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1593 		NDFREE(&nd, NDF_ONLY_PNBUF);
1594 		vput(nd.ni_dvp);
1595 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1596 			goto out;
1597 		goto restart;
1598 	}
1599 	VATTR_NULL(&vattr);
1600 	vattr.va_mode = ACCESSPERMS &~ td->td_proc->p_fd->fd_cmask;
1601 #ifdef MAC
1602 	vattr.va_type = VLNK;
1603 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1604 	    &vattr);
1605 	if (error != 0)
1606 		goto out2;
1607 #endif
1608 	error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, syspath);
1609 	if (error == 0)
1610 		vput(nd.ni_vp);
1611 #ifdef MAC
1612 out2:
1613 #endif
1614 	NDFREE(&nd, NDF_ONLY_PNBUF);
1615 	vput(nd.ni_dvp);
1616 	vn_finished_write(mp);
1617 out:
1618 	if (segflg != UIO_SYSSPACE)
1619 		uma_zfree(namei_zone, syspath);
1620 	return (error);
1621 }
1622 
1623 /*
1624  * Delete a whiteout from the filesystem.
1625  */
1626 int
1627 sys_undelete(td, uap)
1628 	struct thread *td;
1629 	register struct undelete_args /* {
1630 		char *path;
1631 	} */ *uap;
1632 {
1633 	struct mount *mp;
1634 	struct nameidata nd;
1635 	int error;
1636 
1637 restart:
1638 	bwillwrite();
1639 	NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | AUDITVNODE1,
1640 	    UIO_USERSPACE, uap->path, td);
1641 	error = namei(&nd);
1642 	if (error != 0)
1643 		return (error);
1644 
1645 	if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) {
1646 		NDFREE(&nd, NDF_ONLY_PNBUF);
1647 		if (nd.ni_vp == nd.ni_dvp)
1648 			vrele(nd.ni_dvp);
1649 		else
1650 			vput(nd.ni_dvp);
1651 		if (nd.ni_vp)
1652 			vrele(nd.ni_vp);
1653 		return (EEXIST);
1654 	}
1655 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1656 		NDFREE(&nd, NDF_ONLY_PNBUF);
1657 		vput(nd.ni_dvp);
1658 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1659 			return (error);
1660 		goto restart;
1661 	}
1662 	error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE);
1663 	NDFREE(&nd, NDF_ONLY_PNBUF);
1664 	vput(nd.ni_dvp);
1665 	vn_finished_write(mp);
1666 	return (error);
1667 }
1668 
1669 /*
1670  * Delete a name from the filesystem.
1671  */
1672 #ifndef _SYS_SYSPROTO_H_
1673 struct unlink_args {
1674 	char	*path;
1675 };
1676 #endif
1677 int
1678 sys_unlink(td, uap)
1679 	struct thread *td;
1680 	struct unlink_args /* {
1681 		char *path;
1682 	} */ *uap;
1683 {
1684 
1685 	return (kern_unlinkat(td, AT_FDCWD, uap->path, UIO_USERSPACE, 0));
1686 }
1687 
1688 #ifndef _SYS_SYSPROTO_H_
1689 struct unlinkat_args {
1690 	int	fd;
1691 	char	*path;
1692 	int	flag;
1693 };
1694 #endif
1695 int
1696 sys_unlinkat(struct thread *td, struct unlinkat_args *uap)
1697 {
1698 	int flag = uap->flag;
1699 	int fd = uap->fd;
1700 	char *path = uap->path;
1701 
1702 	if (flag & ~AT_REMOVEDIR)
1703 		return (EINVAL);
1704 
1705 	if (flag & AT_REMOVEDIR)
1706 		return (kern_rmdirat(td, fd, path, UIO_USERSPACE));
1707 	else
1708 		return (kern_unlinkat(td, fd, path, UIO_USERSPACE, 0));
1709 }
1710 
1711 int
1712 kern_unlinkat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1713     ino_t oldinum)
1714 {
1715 	struct mount *mp;
1716 	struct vnode *vp;
1717 	struct nameidata nd;
1718 	struct stat sb;
1719 	cap_rights_t rights;
1720 	int error;
1721 
1722 restart:
1723 	bwillwrite();
1724 	NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1,
1725 	    pathseg, path, fd, cap_rights_init(&rights, CAP_UNLINKAT), td);
1726 	if ((error = namei(&nd)) != 0)
1727 		return (error == EINVAL ? EPERM : error);
1728 	vp = nd.ni_vp;
1729 	if (vp->v_type == VDIR && oldinum == 0) {
1730 		error = EPERM;		/* POSIX */
1731 	} else if (oldinum != 0 &&
1732 		  ((error = vn_stat(vp, &sb, td->td_ucred, NOCRED, td)) == 0) &&
1733 		  sb.st_ino != oldinum) {
1734 			error = EIDRM;	/* Identifier removed */
1735 	} else {
1736 		/*
1737 		 * The root of a mounted filesystem cannot be deleted.
1738 		 *
1739 		 * XXX: can this only be a VDIR case?
1740 		 */
1741 		if (vp->v_vflag & VV_ROOT)
1742 			error = EBUSY;
1743 	}
1744 	if (error == 0) {
1745 		if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1746 			NDFREE(&nd, NDF_ONLY_PNBUF);
1747 			vput(nd.ni_dvp);
1748 			if (vp == nd.ni_dvp)
1749 				vrele(vp);
1750 			else
1751 				vput(vp);
1752 			if ((error = vn_start_write(NULL, &mp,
1753 			    V_XSLEEP | PCATCH)) != 0)
1754 				return (error);
1755 			goto restart;
1756 		}
1757 #ifdef MAC
1758 		error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
1759 		    &nd.ni_cnd);
1760 		if (error != 0)
1761 			goto out;
1762 #endif
1763 		vfs_notify_upper(vp, VFS_NOTIFY_UPPER_UNLINK);
1764 		error = VOP_REMOVE(nd.ni_dvp, vp, &nd.ni_cnd);
1765 #ifdef MAC
1766 out:
1767 #endif
1768 		vn_finished_write(mp);
1769 	}
1770 	NDFREE(&nd, NDF_ONLY_PNBUF);
1771 	vput(nd.ni_dvp);
1772 	if (vp == nd.ni_dvp)
1773 		vrele(vp);
1774 	else
1775 		vput(vp);
1776 	return (error);
1777 }
1778 
1779 /*
1780  * Reposition read/write file offset.
1781  */
1782 #ifndef _SYS_SYSPROTO_H_
1783 struct lseek_args {
1784 	int	fd;
1785 	int	pad;
1786 	off_t	offset;
1787 	int	whence;
1788 };
1789 #endif
1790 int
1791 sys_lseek(td, uap)
1792 	struct thread *td;
1793 	register struct lseek_args /* {
1794 		int fd;
1795 		int pad;
1796 		off_t offset;
1797 		int whence;
1798 	} */ *uap;
1799 {
1800 	struct file *fp;
1801 	cap_rights_t rights;
1802 	int error;
1803 
1804 	AUDIT_ARG_FD(uap->fd);
1805 	error = fget(td, uap->fd, cap_rights_init(&rights, CAP_SEEK), &fp);
1806 	if (error != 0)
1807 		return (error);
1808 	error = (fp->f_ops->fo_flags & DFLAG_SEEKABLE) != 0 ?
1809 	    fo_seek(fp, uap->offset, uap->whence, td) : ESPIPE;
1810 	fdrop(fp, td);
1811 	return (error);
1812 }
1813 
1814 #if defined(COMPAT_43)
1815 /*
1816  * Reposition read/write file offset.
1817  */
1818 #ifndef _SYS_SYSPROTO_H_
1819 struct olseek_args {
1820 	int	fd;
1821 	long	offset;
1822 	int	whence;
1823 };
1824 #endif
1825 int
1826 olseek(td, uap)
1827 	struct thread *td;
1828 	register struct olseek_args /* {
1829 		int fd;
1830 		long offset;
1831 		int whence;
1832 	} */ *uap;
1833 {
1834 	struct lseek_args /* {
1835 		int fd;
1836 		int pad;
1837 		off_t offset;
1838 		int whence;
1839 	} */ nuap;
1840 
1841 	nuap.fd = uap->fd;
1842 	nuap.offset = uap->offset;
1843 	nuap.whence = uap->whence;
1844 	return (sys_lseek(td, &nuap));
1845 }
1846 #endif /* COMPAT_43 */
1847 
1848 #if defined(COMPAT_FREEBSD6)
1849 /* Version with the 'pad' argument */
1850 int
1851 freebsd6_lseek(td, uap)
1852 	struct thread *td;
1853 	register struct freebsd6_lseek_args *uap;
1854 {
1855 	struct lseek_args ouap;
1856 
1857 	ouap.fd = uap->fd;
1858 	ouap.offset = uap->offset;
1859 	ouap.whence = uap->whence;
1860 	return (sys_lseek(td, &ouap));
1861 }
1862 #endif
1863 
1864 /*
1865  * Check access permissions using passed credentials.
1866  */
1867 static int
1868 vn_access(vp, user_flags, cred, td)
1869 	struct vnode	*vp;
1870 	int		user_flags;
1871 	struct ucred	*cred;
1872 	struct thread	*td;
1873 {
1874 	accmode_t accmode;
1875 	int error;
1876 
1877 	/* Flags == 0 means only check for existence. */
1878 	if (user_flags == 0)
1879 		return (0);
1880 
1881 	accmode = 0;
1882 	if (user_flags & R_OK)
1883 		accmode |= VREAD;
1884 	if (user_flags & W_OK)
1885 		accmode |= VWRITE;
1886 	if (user_flags & X_OK)
1887 		accmode |= VEXEC;
1888 #ifdef MAC
1889 	error = mac_vnode_check_access(cred, vp, accmode);
1890 	if (error != 0)
1891 		return (error);
1892 #endif
1893 	if ((accmode & VWRITE) == 0 || (error = vn_writechk(vp)) == 0)
1894 		error = VOP_ACCESS(vp, accmode, cred, td);
1895 	return (error);
1896 }
1897 
1898 /*
1899  * Check access permissions using "real" credentials.
1900  */
1901 #ifndef _SYS_SYSPROTO_H_
1902 struct access_args {
1903 	char	*path;
1904 	int	amode;
1905 };
1906 #endif
1907 int
1908 sys_access(td, uap)
1909 	struct thread *td;
1910 	register struct access_args /* {
1911 		char *path;
1912 		int amode;
1913 	} */ *uap;
1914 {
1915 
1916 	return (kern_accessat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1917 	    0, uap->amode));
1918 }
1919 
1920 #ifndef _SYS_SYSPROTO_H_
1921 struct faccessat_args {
1922 	int	dirfd;
1923 	char	*path;
1924 	int	amode;
1925 	int	flag;
1926 }
1927 #endif
1928 int
1929 sys_faccessat(struct thread *td, struct faccessat_args *uap)
1930 {
1931 
1932 	return (kern_accessat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
1933 	    uap->amode));
1934 }
1935 
1936 int
1937 kern_accessat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1938     int flag, int amode)
1939 {
1940 	struct ucred *cred, *usecred;
1941 	struct vnode *vp;
1942 	struct nameidata nd;
1943 	cap_rights_t rights;
1944 	int error;
1945 
1946 	if (flag & ~AT_EACCESS)
1947 		return (EINVAL);
1948 	if (amode != F_OK && (amode & ~(R_OK | W_OK | X_OK)) != 0)
1949 		return (EINVAL);
1950 
1951 	/*
1952 	 * Create and modify a temporary credential instead of one that
1953 	 * is potentially shared (if we need one).
1954 	 */
1955 	cred = td->td_ucred;
1956 	if ((flag & AT_EACCESS) == 0 &&
1957 	    ((cred->cr_uid != cred->cr_ruid ||
1958 	    cred->cr_rgid != cred->cr_groups[0]))) {
1959 		usecred = crdup(cred);
1960 		usecred->cr_uid = cred->cr_ruid;
1961 		usecred->cr_groups[0] = cred->cr_rgid;
1962 		td->td_ucred = usecred;
1963 	} else
1964 		usecred = cred;
1965 	AUDIT_ARG_VALUE(amode);
1966 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF |
1967 	    AUDITVNODE1, pathseg, path, fd, cap_rights_init(&rights, CAP_FSTAT),
1968 	    td);
1969 	if ((error = namei(&nd)) != 0)
1970 		goto out;
1971 	vp = nd.ni_vp;
1972 
1973 	error = vn_access(vp, amode, usecred, td);
1974 	NDFREE(&nd, NDF_ONLY_PNBUF);
1975 	vput(vp);
1976 out:
1977 	if (usecred != cred) {
1978 		td->td_ucred = cred;
1979 		crfree(usecred);
1980 	}
1981 	return (error);
1982 }
1983 
1984 /*
1985  * Check access permissions using "effective" credentials.
1986  */
1987 #ifndef _SYS_SYSPROTO_H_
1988 struct eaccess_args {
1989 	char	*path;
1990 	int	amode;
1991 };
1992 #endif
1993 int
1994 sys_eaccess(td, uap)
1995 	struct thread *td;
1996 	register struct eaccess_args /* {
1997 		char *path;
1998 		int amode;
1999 	} */ *uap;
2000 {
2001 
2002 	return (kern_accessat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2003 	    AT_EACCESS, uap->amode));
2004 }
2005 
2006 #if defined(COMPAT_43)
2007 /*
2008  * Get file status; this version follows links.
2009  */
2010 #ifndef _SYS_SYSPROTO_H_
2011 struct ostat_args {
2012 	char	*path;
2013 	struct ostat *ub;
2014 };
2015 #endif
2016 int
2017 ostat(td, uap)
2018 	struct thread *td;
2019 	register struct ostat_args /* {
2020 		char *path;
2021 		struct ostat *ub;
2022 	} */ *uap;
2023 {
2024 	struct stat sb;
2025 	struct ostat osb;
2026 	int error;
2027 
2028 	error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE,
2029 	    &sb, NULL);
2030 	if (error != 0)
2031 		return (error);
2032 	cvtstat(&sb, &osb);
2033 	return (copyout(&osb, uap->ub, sizeof (osb)));
2034 }
2035 
2036 /*
2037  * Get file status; this version does not follow links.
2038  */
2039 #ifndef _SYS_SYSPROTO_H_
2040 struct olstat_args {
2041 	char	*path;
2042 	struct ostat *ub;
2043 };
2044 #endif
2045 int
2046 olstat(td, uap)
2047 	struct thread *td;
2048 	register struct olstat_args /* {
2049 		char *path;
2050 		struct ostat *ub;
2051 	} */ *uap;
2052 {
2053 	struct stat sb;
2054 	struct ostat osb;
2055 	int error;
2056 
2057 	error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2058 	    UIO_USERSPACE, &sb, NULL);
2059 	if (error != 0)
2060 		return (error);
2061 	cvtstat(&sb, &osb);
2062 	return (copyout(&osb, uap->ub, sizeof (osb)));
2063 }
2064 
2065 /*
2066  * Convert from an old to a new stat structure.
2067  */
2068 void
2069 cvtstat(st, ost)
2070 	struct stat *st;
2071 	struct ostat *ost;
2072 {
2073 
2074 	bzero(ost, sizeof(*ost));
2075 	ost->st_dev = st->st_dev;
2076 	ost->st_ino = st->st_ino;
2077 	ost->st_mode = st->st_mode;
2078 	ost->st_nlink = st->st_nlink;
2079 	ost->st_uid = st->st_uid;
2080 	ost->st_gid = st->st_gid;
2081 	ost->st_rdev = st->st_rdev;
2082 	if (st->st_size < (quad_t)1 << 32)
2083 		ost->st_size = st->st_size;
2084 	else
2085 		ost->st_size = -2;
2086 	ost->st_atim = st->st_atim;
2087 	ost->st_mtim = st->st_mtim;
2088 	ost->st_ctim = st->st_ctim;
2089 	ost->st_blksize = st->st_blksize;
2090 	ost->st_blocks = st->st_blocks;
2091 	ost->st_flags = st->st_flags;
2092 	ost->st_gen = st->st_gen;
2093 }
2094 #endif /* COMPAT_43 */
2095 
2096 /*
2097  * Get file status; this version follows links.
2098  */
2099 #ifndef _SYS_SYSPROTO_H_
2100 struct stat_args {
2101 	char	*path;
2102 	struct stat *ub;
2103 };
2104 #endif
2105 int
2106 sys_stat(td, uap)
2107 	struct thread *td;
2108 	register struct stat_args /* {
2109 		char *path;
2110 		struct stat *ub;
2111 	} */ *uap;
2112 {
2113 	struct stat sb;
2114 	int error;
2115 
2116 	error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE,
2117 	    &sb, NULL);
2118 	if (error == 0)
2119 		error = copyout(&sb, uap->ub, sizeof (sb));
2120 	return (error);
2121 }
2122 
2123 #ifndef _SYS_SYSPROTO_H_
2124 struct fstatat_args {
2125 	int	fd;
2126 	char	*path;
2127 	struct stat	*buf;
2128 	int	flag;
2129 }
2130 #endif
2131 int
2132 sys_fstatat(struct thread *td, struct fstatat_args *uap)
2133 {
2134 	struct stat sb;
2135 	int error;
2136 
2137 	error = kern_statat(td, uap->flag, uap->fd, uap->path,
2138 	    UIO_USERSPACE, &sb, NULL);
2139 	if (error == 0)
2140 		error = copyout(&sb, uap->buf, sizeof (sb));
2141 	return (error);
2142 }
2143 
2144 int
2145 kern_statat(struct thread *td, int flag, int fd, char *path,
2146     enum uio_seg pathseg, struct stat *sbp,
2147     void (*hook)(struct vnode *vp, struct stat *sbp))
2148 {
2149 	struct nameidata nd;
2150 	struct stat sb;
2151 	cap_rights_t rights;
2152 	int error;
2153 
2154 	if (flag & ~AT_SYMLINK_NOFOLLOW)
2155 		return (EINVAL);
2156 
2157 	NDINIT_ATRIGHTS(&nd, LOOKUP, ((flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW :
2158 	    FOLLOW) | LOCKSHARED | LOCKLEAF | AUDITVNODE1, pathseg, path, fd,
2159 	    cap_rights_init(&rights, CAP_FSTAT), td);
2160 
2161 	if ((error = namei(&nd)) != 0)
2162 		return (error);
2163 	error = vn_stat(nd.ni_vp, &sb, td->td_ucred, NOCRED, td);
2164 	if (error == 0) {
2165 		SDT_PROBE2(vfs, , stat, mode, path, sb.st_mode);
2166 		if (S_ISREG(sb.st_mode))
2167 			SDT_PROBE2(vfs, , stat, reg, path, pathseg);
2168 		if (__predict_false(hook != NULL))
2169 			hook(nd.ni_vp, &sb);
2170 	}
2171 	NDFREE(&nd, NDF_ONLY_PNBUF);
2172 	vput(nd.ni_vp);
2173 	if (error != 0)
2174 		return (error);
2175 	*sbp = sb;
2176 #ifdef KTRACE
2177 	if (KTRPOINT(td, KTR_STRUCT))
2178 		ktrstat(&sb);
2179 #endif
2180 	return (0);
2181 }
2182 
2183 /*
2184  * Get file status; this version does not follow links.
2185  */
2186 #ifndef _SYS_SYSPROTO_H_
2187 struct lstat_args {
2188 	char	*path;
2189 	struct stat *ub;
2190 };
2191 #endif
2192 int
2193 sys_lstat(td, uap)
2194 	struct thread *td;
2195 	register struct lstat_args /* {
2196 		char *path;
2197 		struct stat *ub;
2198 	} */ *uap;
2199 {
2200 	struct stat sb;
2201 	int error;
2202 
2203 	error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2204 	    UIO_USERSPACE, &sb, NULL);
2205 	if (error == 0)
2206 		error = copyout(&sb, uap->ub, sizeof (sb));
2207 	return (error);
2208 }
2209 
2210 /*
2211  * Implementation of the NetBSD [l]stat() functions.
2212  */
2213 void
2214 cvtnstat(sb, nsb)
2215 	struct stat *sb;
2216 	struct nstat *nsb;
2217 {
2218 
2219 	bzero(nsb, sizeof *nsb);
2220 	nsb->st_dev = sb->st_dev;
2221 	nsb->st_ino = sb->st_ino;
2222 	nsb->st_mode = sb->st_mode;
2223 	nsb->st_nlink = sb->st_nlink;
2224 	nsb->st_uid = sb->st_uid;
2225 	nsb->st_gid = sb->st_gid;
2226 	nsb->st_rdev = sb->st_rdev;
2227 	nsb->st_atim = sb->st_atim;
2228 	nsb->st_mtim = sb->st_mtim;
2229 	nsb->st_ctim = sb->st_ctim;
2230 	nsb->st_size = sb->st_size;
2231 	nsb->st_blocks = sb->st_blocks;
2232 	nsb->st_blksize = sb->st_blksize;
2233 	nsb->st_flags = sb->st_flags;
2234 	nsb->st_gen = sb->st_gen;
2235 	nsb->st_birthtim = sb->st_birthtim;
2236 }
2237 
2238 #ifndef _SYS_SYSPROTO_H_
2239 struct nstat_args {
2240 	char	*path;
2241 	struct nstat *ub;
2242 };
2243 #endif
2244 int
2245 sys_nstat(td, uap)
2246 	struct thread *td;
2247 	register struct nstat_args /* {
2248 		char *path;
2249 		struct nstat *ub;
2250 	} */ *uap;
2251 {
2252 	struct stat sb;
2253 	struct nstat nsb;
2254 	int error;
2255 
2256 	error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE,
2257 	    &sb, NULL);
2258 	if (error != 0)
2259 		return (error);
2260 	cvtnstat(&sb, &nsb);
2261 	return (copyout(&nsb, uap->ub, sizeof (nsb)));
2262 }
2263 
2264 /*
2265  * NetBSD lstat.  Get file status; this version does not follow links.
2266  */
2267 #ifndef _SYS_SYSPROTO_H_
2268 struct lstat_args {
2269 	char	*path;
2270 	struct stat *ub;
2271 };
2272 #endif
2273 int
2274 sys_nlstat(td, uap)
2275 	struct thread *td;
2276 	register struct nlstat_args /* {
2277 		char *path;
2278 		struct nstat *ub;
2279 	} */ *uap;
2280 {
2281 	struct stat sb;
2282 	struct nstat nsb;
2283 	int error;
2284 
2285 	error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2286 	    UIO_USERSPACE, &sb, NULL);
2287 	if (error != 0)
2288 		return (error);
2289 	cvtnstat(&sb, &nsb);
2290 	return (copyout(&nsb, uap->ub, sizeof (nsb)));
2291 }
2292 
2293 /*
2294  * Get configurable pathname variables.
2295  */
2296 #ifndef _SYS_SYSPROTO_H_
2297 struct pathconf_args {
2298 	char	*path;
2299 	int	name;
2300 };
2301 #endif
2302 int
2303 sys_pathconf(td, uap)
2304 	struct thread *td;
2305 	register struct pathconf_args /* {
2306 		char *path;
2307 		int name;
2308 	} */ *uap;
2309 {
2310 
2311 	return (kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name, FOLLOW));
2312 }
2313 
2314 #ifndef _SYS_SYSPROTO_H_
2315 struct lpathconf_args {
2316 	char	*path;
2317 	int	name;
2318 };
2319 #endif
2320 int
2321 sys_lpathconf(td, uap)
2322 	struct thread *td;
2323 	register struct lpathconf_args /* {
2324 		char *path;
2325 		int name;
2326 	} */ *uap;
2327 {
2328 
2329 	return (kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name,
2330 	    NOFOLLOW));
2331 }
2332 
2333 int
2334 kern_pathconf(struct thread *td, char *path, enum uio_seg pathseg, int name,
2335     u_long flags)
2336 {
2337 	struct nameidata nd;
2338 	int error;
2339 
2340 	NDINIT(&nd, LOOKUP, LOCKSHARED | LOCKLEAF | AUDITVNODE1 | flags,
2341 	    pathseg, path, td);
2342 	if ((error = namei(&nd)) != 0)
2343 		return (error);
2344 	NDFREE(&nd, NDF_ONLY_PNBUF);
2345 
2346 	error = VOP_PATHCONF(nd.ni_vp, name, td->td_retval);
2347 	vput(nd.ni_vp);
2348 	return (error);
2349 }
2350 
2351 /*
2352  * Return target name of a symbolic link.
2353  */
2354 #ifndef _SYS_SYSPROTO_H_
2355 struct readlink_args {
2356 	char	*path;
2357 	char	*buf;
2358 	size_t	count;
2359 };
2360 #endif
2361 int
2362 sys_readlink(td, uap)
2363 	struct thread *td;
2364 	register struct readlink_args /* {
2365 		char *path;
2366 		char *buf;
2367 		size_t count;
2368 	} */ *uap;
2369 {
2370 
2371 	return (kern_readlinkat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2372 	    uap->buf, UIO_USERSPACE, uap->count));
2373 }
2374 #ifndef _SYS_SYSPROTO_H_
2375 struct readlinkat_args {
2376 	int	fd;
2377 	char	*path;
2378 	char	*buf;
2379 	size_t	bufsize;
2380 };
2381 #endif
2382 int
2383 sys_readlinkat(struct thread *td, struct readlinkat_args *uap)
2384 {
2385 
2386 	return (kern_readlinkat(td, uap->fd, uap->path, UIO_USERSPACE,
2387 	    uap->buf, UIO_USERSPACE, uap->bufsize));
2388 }
2389 
2390 int
2391 kern_readlinkat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2392     char *buf, enum uio_seg bufseg, size_t count)
2393 {
2394 	struct vnode *vp;
2395 	struct iovec aiov;
2396 	struct uio auio;
2397 	struct nameidata nd;
2398 	int error;
2399 
2400 	if (count > IOSIZE_MAX)
2401 		return (EINVAL);
2402 
2403 	NDINIT_AT(&nd, LOOKUP, NOFOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
2404 	    pathseg, path, fd, td);
2405 
2406 	if ((error = namei(&nd)) != 0)
2407 		return (error);
2408 	NDFREE(&nd, NDF_ONLY_PNBUF);
2409 	vp = nd.ni_vp;
2410 #ifdef MAC
2411 	error = mac_vnode_check_readlink(td->td_ucred, vp);
2412 	if (error != 0) {
2413 		vput(vp);
2414 		return (error);
2415 	}
2416 #endif
2417 	if (vp->v_type != VLNK)
2418 		error = EINVAL;
2419 	else {
2420 		aiov.iov_base = buf;
2421 		aiov.iov_len = count;
2422 		auio.uio_iov = &aiov;
2423 		auio.uio_iovcnt = 1;
2424 		auio.uio_offset = 0;
2425 		auio.uio_rw = UIO_READ;
2426 		auio.uio_segflg = bufseg;
2427 		auio.uio_td = td;
2428 		auio.uio_resid = count;
2429 		error = VOP_READLINK(vp, &auio, td->td_ucred);
2430 		td->td_retval[0] = count - auio.uio_resid;
2431 	}
2432 	vput(vp);
2433 	return (error);
2434 }
2435 
2436 /*
2437  * Common implementation code for chflags() and fchflags().
2438  */
2439 static int
2440 setfflags(td, vp, flags)
2441 	struct thread *td;
2442 	struct vnode *vp;
2443 	u_long flags;
2444 {
2445 	struct mount *mp;
2446 	struct vattr vattr;
2447 	int error;
2448 
2449 	/* We can't support the value matching VNOVAL. */
2450 	if (flags == VNOVAL)
2451 		return (EOPNOTSUPP);
2452 
2453 	/*
2454 	 * Prevent non-root users from setting flags on devices.  When
2455 	 * a device is reused, users can retain ownership of the device
2456 	 * if they are allowed to set flags and programs assume that
2457 	 * chown can't fail when done as root.
2458 	 */
2459 	if (vp->v_type == VCHR || vp->v_type == VBLK) {
2460 		error = priv_check(td, PRIV_VFS_CHFLAGS_DEV);
2461 		if (error != 0)
2462 			return (error);
2463 	}
2464 
2465 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2466 		return (error);
2467 	VATTR_NULL(&vattr);
2468 	vattr.va_flags = flags;
2469 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2470 #ifdef MAC
2471 	error = mac_vnode_check_setflags(td->td_ucred, vp, vattr.va_flags);
2472 	if (error == 0)
2473 #endif
2474 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
2475 	VOP_UNLOCK(vp, 0);
2476 	vn_finished_write(mp);
2477 	return (error);
2478 }
2479 
2480 /*
2481  * Change flags of a file given a path name.
2482  */
2483 #ifndef _SYS_SYSPROTO_H_
2484 struct chflags_args {
2485 	const char *path;
2486 	u_long	flags;
2487 };
2488 #endif
2489 int
2490 sys_chflags(td, uap)
2491 	struct thread *td;
2492 	register struct chflags_args /* {
2493 		const char *path;
2494 		u_long flags;
2495 	} */ *uap;
2496 {
2497 
2498 	return (kern_chflagsat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2499 	    uap->flags, 0));
2500 }
2501 
2502 #ifndef _SYS_SYSPROTO_H_
2503 struct chflagsat_args {
2504 	int	fd;
2505 	const char *path;
2506 	u_long	flags;
2507 	int	atflag;
2508 }
2509 #endif
2510 int
2511 sys_chflagsat(struct thread *td, struct chflagsat_args *uap)
2512 {
2513 	int fd = uap->fd;
2514 	const char *path = uap->path;
2515 	u_long flags = uap->flags;
2516 	int atflag = uap->atflag;
2517 
2518 	if (atflag & ~AT_SYMLINK_NOFOLLOW)
2519 		return (EINVAL);
2520 
2521 	return (kern_chflagsat(td, fd, path, UIO_USERSPACE, flags, atflag));
2522 }
2523 
2524 /*
2525  * Same as chflags() but doesn't follow symlinks.
2526  */
2527 int
2528 sys_lchflags(td, uap)
2529 	struct thread *td;
2530 	register struct lchflags_args /* {
2531 		const char *path;
2532 		u_long flags;
2533 	} */ *uap;
2534 {
2535 
2536 	return (kern_chflagsat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2537 	    uap->flags, AT_SYMLINK_NOFOLLOW));
2538 }
2539 
2540 static int
2541 kern_chflagsat(struct thread *td, int fd, const char *path,
2542     enum uio_seg pathseg, u_long flags, int atflag)
2543 {
2544 	struct nameidata nd;
2545 	cap_rights_t rights;
2546 	int error, follow;
2547 
2548 	AUDIT_ARG_FFLAGS(flags);
2549 	follow = (atflag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
2550 	NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, pathseg, path, fd,
2551 	    cap_rights_init(&rights, CAP_FCHFLAGS), td);
2552 	if ((error = namei(&nd)) != 0)
2553 		return (error);
2554 	NDFREE(&nd, NDF_ONLY_PNBUF);
2555 	error = setfflags(td, nd.ni_vp, flags);
2556 	vrele(nd.ni_vp);
2557 	return (error);
2558 }
2559 
2560 /*
2561  * Change flags of a file given a file descriptor.
2562  */
2563 #ifndef _SYS_SYSPROTO_H_
2564 struct fchflags_args {
2565 	int	fd;
2566 	u_long	flags;
2567 };
2568 #endif
2569 int
2570 sys_fchflags(td, uap)
2571 	struct thread *td;
2572 	register struct fchflags_args /* {
2573 		int fd;
2574 		u_long flags;
2575 	} */ *uap;
2576 {
2577 	struct file *fp;
2578 	cap_rights_t rights;
2579 	int error;
2580 
2581 	AUDIT_ARG_FD(uap->fd);
2582 	AUDIT_ARG_FFLAGS(uap->flags);
2583 	error = getvnode(td, uap->fd, cap_rights_init(&rights, CAP_FCHFLAGS),
2584 	    &fp);
2585 	if (error != 0)
2586 		return (error);
2587 #ifdef AUDIT
2588 	vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
2589 	AUDIT_ARG_VNODE1(fp->f_vnode);
2590 	VOP_UNLOCK(fp->f_vnode, 0);
2591 #endif
2592 	error = setfflags(td, fp->f_vnode, uap->flags);
2593 	fdrop(fp, td);
2594 	return (error);
2595 }
2596 
2597 /*
2598  * Common implementation code for chmod(), lchmod() and fchmod().
2599  */
2600 int
2601 setfmode(td, cred, vp, mode)
2602 	struct thread *td;
2603 	struct ucred *cred;
2604 	struct vnode *vp;
2605 	int mode;
2606 {
2607 	struct mount *mp;
2608 	struct vattr vattr;
2609 	int error;
2610 
2611 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2612 		return (error);
2613 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2614 	VATTR_NULL(&vattr);
2615 	vattr.va_mode = mode & ALLPERMS;
2616 #ifdef MAC
2617 	error = mac_vnode_check_setmode(cred, vp, vattr.va_mode);
2618 	if (error == 0)
2619 #endif
2620 		error = VOP_SETATTR(vp, &vattr, cred);
2621 	VOP_UNLOCK(vp, 0);
2622 	vn_finished_write(mp);
2623 	return (error);
2624 }
2625 
2626 /*
2627  * Change mode of a file given path name.
2628  */
2629 #ifndef _SYS_SYSPROTO_H_
2630 struct chmod_args {
2631 	char	*path;
2632 	int	mode;
2633 };
2634 #endif
2635 int
2636 sys_chmod(td, uap)
2637 	struct thread *td;
2638 	register struct chmod_args /* {
2639 		char *path;
2640 		int mode;
2641 	} */ *uap;
2642 {
2643 
2644 	return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2645 	    uap->mode, 0));
2646 }
2647 
2648 #ifndef _SYS_SYSPROTO_H_
2649 struct fchmodat_args {
2650 	int	dirfd;
2651 	char	*path;
2652 	mode_t	mode;
2653 	int	flag;
2654 }
2655 #endif
2656 int
2657 sys_fchmodat(struct thread *td, struct fchmodat_args *uap)
2658 {
2659 	int flag = uap->flag;
2660 	int fd = uap->fd;
2661 	char *path = uap->path;
2662 	mode_t mode = uap->mode;
2663 
2664 	if (flag & ~AT_SYMLINK_NOFOLLOW)
2665 		return (EINVAL);
2666 
2667 	return (kern_fchmodat(td, fd, path, UIO_USERSPACE, mode, flag));
2668 }
2669 
2670 /*
2671  * Change mode of a file given path name (don't follow links.)
2672  */
2673 #ifndef _SYS_SYSPROTO_H_
2674 struct lchmod_args {
2675 	char	*path;
2676 	int	mode;
2677 };
2678 #endif
2679 int
2680 sys_lchmod(td, uap)
2681 	struct thread *td;
2682 	register struct lchmod_args /* {
2683 		char *path;
2684 		int mode;
2685 	} */ *uap;
2686 {
2687 
2688 	return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2689 	    uap->mode, AT_SYMLINK_NOFOLLOW));
2690 }
2691 
2692 int
2693 kern_fchmodat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2694     mode_t mode, int flag)
2695 {
2696 	struct nameidata nd;
2697 	cap_rights_t rights;
2698 	int error, follow;
2699 
2700 	AUDIT_ARG_MODE(mode);
2701 	follow = (flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
2702 	NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, pathseg, path, fd,
2703 	    cap_rights_init(&rights, CAP_FCHMOD), td);
2704 	if ((error = namei(&nd)) != 0)
2705 		return (error);
2706 	NDFREE(&nd, NDF_ONLY_PNBUF);
2707 	error = setfmode(td, td->td_ucred, nd.ni_vp, mode);
2708 	vrele(nd.ni_vp);
2709 	return (error);
2710 }
2711 
2712 /*
2713  * Change mode of a file given a file descriptor.
2714  */
2715 #ifndef _SYS_SYSPROTO_H_
2716 struct fchmod_args {
2717 	int	fd;
2718 	int	mode;
2719 };
2720 #endif
2721 int
2722 sys_fchmod(struct thread *td, struct fchmod_args *uap)
2723 {
2724 	struct file *fp;
2725 	cap_rights_t rights;
2726 	int error;
2727 
2728 	AUDIT_ARG_FD(uap->fd);
2729 	AUDIT_ARG_MODE(uap->mode);
2730 
2731 	error = fget(td, uap->fd, cap_rights_init(&rights, CAP_FCHMOD), &fp);
2732 	if (error != 0)
2733 		return (error);
2734 	error = fo_chmod(fp, uap->mode, td->td_ucred, td);
2735 	fdrop(fp, td);
2736 	return (error);
2737 }
2738 
2739 /*
2740  * Common implementation for chown(), lchown(), and fchown()
2741  */
2742 int
2743 setfown(td, cred, vp, uid, gid)
2744 	struct thread *td;
2745 	struct ucred *cred;
2746 	struct vnode *vp;
2747 	uid_t uid;
2748 	gid_t gid;
2749 {
2750 	struct mount *mp;
2751 	struct vattr vattr;
2752 	int error;
2753 
2754 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2755 		return (error);
2756 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2757 	VATTR_NULL(&vattr);
2758 	vattr.va_uid = uid;
2759 	vattr.va_gid = gid;
2760 #ifdef MAC
2761 	error = mac_vnode_check_setowner(cred, vp, vattr.va_uid,
2762 	    vattr.va_gid);
2763 	if (error == 0)
2764 #endif
2765 		error = VOP_SETATTR(vp, &vattr, cred);
2766 	VOP_UNLOCK(vp, 0);
2767 	vn_finished_write(mp);
2768 	return (error);
2769 }
2770 
2771 /*
2772  * Set ownership given a path name.
2773  */
2774 #ifndef _SYS_SYSPROTO_H_
2775 struct chown_args {
2776 	char	*path;
2777 	int	uid;
2778 	int	gid;
2779 };
2780 #endif
2781 int
2782 sys_chown(td, uap)
2783 	struct thread *td;
2784 	register struct chown_args /* {
2785 		char *path;
2786 		int uid;
2787 		int gid;
2788 	} */ *uap;
2789 {
2790 
2791 	return (kern_fchownat(td, AT_FDCWD, uap->path, UIO_USERSPACE, uap->uid,
2792 	    uap->gid, 0));
2793 }
2794 
2795 #ifndef _SYS_SYSPROTO_H_
2796 struct fchownat_args {
2797 	int fd;
2798 	const char * path;
2799 	uid_t uid;
2800 	gid_t gid;
2801 	int flag;
2802 };
2803 #endif
2804 int
2805 sys_fchownat(struct thread *td, struct fchownat_args *uap)
2806 {
2807 	int flag;
2808 
2809 	flag = uap->flag;
2810 	if (flag & ~AT_SYMLINK_NOFOLLOW)
2811 		return (EINVAL);
2812 
2813 	return (kern_fchownat(td, uap->fd, uap->path, UIO_USERSPACE, uap->uid,
2814 	    uap->gid, uap->flag));
2815 }
2816 
2817 int
2818 kern_fchownat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2819     int uid, int gid, int flag)
2820 {
2821 	struct nameidata nd;
2822 	cap_rights_t rights;
2823 	int error, follow;
2824 
2825 	AUDIT_ARG_OWNER(uid, gid);
2826 	follow = (flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
2827 	NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, pathseg, path, fd,
2828 	    cap_rights_init(&rights, CAP_FCHOWN), td);
2829 
2830 	if ((error = namei(&nd)) != 0)
2831 		return (error);
2832 	NDFREE(&nd, NDF_ONLY_PNBUF);
2833 	error = setfown(td, td->td_ucred, nd.ni_vp, uid, gid);
2834 	vrele(nd.ni_vp);
2835 	return (error);
2836 }
2837 
2838 /*
2839  * Set ownership given a path name, do not cross symlinks.
2840  */
2841 #ifndef _SYS_SYSPROTO_H_
2842 struct lchown_args {
2843 	char	*path;
2844 	int	uid;
2845 	int	gid;
2846 };
2847 #endif
2848 int
2849 sys_lchown(td, uap)
2850 	struct thread *td;
2851 	register struct lchown_args /* {
2852 		char *path;
2853 		int uid;
2854 		int gid;
2855 	} */ *uap;
2856 {
2857 
2858 	return (kern_fchownat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2859 	    uap->uid, uap->gid, AT_SYMLINK_NOFOLLOW));
2860 }
2861 
2862 /*
2863  * Set ownership given a file descriptor.
2864  */
2865 #ifndef _SYS_SYSPROTO_H_
2866 struct fchown_args {
2867 	int	fd;
2868 	int	uid;
2869 	int	gid;
2870 };
2871 #endif
2872 int
2873 sys_fchown(td, uap)
2874 	struct thread *td;
2875 	register struct fchown_args /* {
2876 		int fd;
2877 		int uid;
2878 		int gid;
2879 	} */ *uap;
2880 {
2881 	struct file *fp;
2882 	cap_rights_t rights;
2883 	int error;
2884 
2885 	AUDIT_ARG_FD(uap->fd);
2886 	AUDIT_ARG_OWNER(uap->uid, uap->gid);
2887 	error = fget(td, uap->fd, cap_rights_init(&rights, CAP_FCHOWN), &fp);
2888 	if (error != 0)
2889 		return (error);
2890 	error = fo_chown(fp, uap->uid, uap->gid, td->td_ucred, td);
2891 	fdrop(fp, td);
2892 	return (error);
2893 }
2894 
2895 /*
2896  * Common implementation code for utimes(), lutimes(), and futimes().
2897  */
2898 static int
2899 getutimes(usrtvp, tvpseg, tsp)
2900 	const struct timeval *usrtvp;
2901 	enum uio_seg tvpseg;
2902 	struct timespec *tsp;
2903 {
2904 	struct timeval tv[2];
2905 	const struct timeval *tvp;
2906 	int error;
2907 
2908 	if (usrtvp == NULL) {
2909 		vfs_timestamp(&tsp[0]);
2910 		tsp[1] = tsp[0];
2911 	} else {
2912 		if (tvpseg == UIO_SYSSPACE) {
2913 			tvp = usrtvp;
2914 		} else {
2915 			if ((error = copyin(usrtvp, tv, sizeof(tv))) != 0)
2916 				return (error);
2917 			tvp = tv;
2918 		}
2919 
2920 		if (tvp[0].tv_usec < 0 || tvp[0].tv_usec >= 1000000 ||
2921 		    tvp[1].tv_usec < 0 || tvp[1].tv_usec >= 1000000)
2922 			return (EINVAL);
2923 		TIMEVAL_TO_TIMESPEC(&tvp[0], &tsp[0]);
2924 		TIMEVAL_TO_TIMESPEC(&tvp[1], &tsp[1]);
2925 	}
2926 	return (0);
2927 }
2928 
2929 /*
2930  * Common implementation code for futimens(), utimensat().
2931  */
2932 #define	UTIMENS_NULL	0x1
2933 #define	UTIMENS_EXIT	0x2
2934 static int
2935 getutimens(const struct timespec *usrtsp, enum uio_seg tspseg,
2936     struct timespec *tsp, int *retflags)
2937 {
2938 	struct timespec tsnow;
2939 	int error;
2940 
2941 	vfs_timestamp(&tsnow);
2942 	*retflags = 0;
2943 	if (usrtsp == NULL) {
2944 		tsp[0] = tsnow;
2945 		tsp[1] = tsnow;
2946 		*retflags |= UTIMENS_NULL;
2947 		return (0);
2948 	}
2949 	if (tspseg == UIO_SYSSPACE) {
2950 		tsp[0] = usrtsp[0];
2951 		tsp[1] = usrtsp[1];
2952 	} else if ((error = copyin(usrtsp, tsp, sizeof(*tsp) * 2)) != 0)
2953 		return (error);
2954 	if (tsp[0].tv_nsec == UTIME_OMIT && tsp[1].tv_nsec == UTIME_OMIT)
2955 		*retflags |= UTIMENS_EXIT;
2956 	if (tsp[0].tv_nsec == UTIME_NOW && tsp[1].tv_nsec == UTIME_NOW)
2957 		*retflags |= UTIMENS_NULL;
2958 	if (tsp[0].tv_nsec == UTIME_OMIT)
2959 		tsp[0].tv_sec = VNOVAL;
2960 	else if (tsp[0].tv_nsec == UTIME_NOW)
2961 		tsp[0] = tsnow;
2962 	else if (tsp[0].tv_nsec < 0 || tsp[0].tv_nsec >= 1000000000L)
2963 		return (EINVAL);
2964 	if (tsp[1].tv_nsec == UTIME_OMIT)
2965 		tsp[1].tv_sec = VNOVAL;
2966 	else if (tsp[1].tv_nsec == UTIME_NOW)
2967 		tsp[1] = tsnow;
2968 	else if (tsp[1].tv_nsec < 0 || tsp[1].tv_nsec >= 1000000000L)
2969 		return (EINVAL);
2970 
2971 	return (0);
2972 }
2973 
2974 /*
2975  * Common implementation code for utimes(), lutimes(), futimes(), futimens(),
2976  * and utimensat().
2977  */
2978 static int
2979 setutimes(td, vp, ts, numtimes, nullflag)
2980 	struct thread *td;
2981 	struct vnode *vp;
2982 	const struct timespec *ts;
2983 	int numtimes;
2984 	int nullflag;
2985 {
2986 	struct mount *mp;
2987 	struct vattr vattr;
2988 	int error, setbirthtime;
2989 
2990 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2991 		return (error);
2992 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2993 	setbirthtime = 0;
2994 	if (numtimes < 3 && !VOP_GETATTR(vp, &vattr, td->td_ucred) &&
2995 	    timespeccmp(&ts[1], &vattr.va_birthtime, < ))
2996 		setbirthtime = 1;
2997 	VATTR_NULL(&vattr);
2998 	vattr.va_atime = ts[0];
2999 	vattr.va_mtime = ts[1];
3000 	if (setbirthtime)
3001 		vattr.va_birthtime = ts[1];
3002 	if (numtimes > 2)
3003 		vattr.va_birthtime = ts[2];
3004 	if (nullflag)
3005 		vattr.va_vaflags |= VA_UTIMES_NULL;
3006 #ifdef MAC
3007 	error = mac_vnode_check_setutimes(td->td_ucred, vp, vattr.va_atime,
3008 	    vattr.va_mtime);
3009 #endif
3010 	if (error == 0)
3011 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3012 	VOP_UNLOCK(vp, 0);
3013 	vn_finished_write(mp);
3014 	return (error);
3015 }
3016 
3017 /*
3018  * Set the access and modification times of a file.
3019  */
3020 #ifndef _SYS_SYSPROTO_H_
3021 struct utimes_args {
3022 	char	*path;
3023 	struct	timeval *tptr;
3024 };
3025 #endif
3026 int
3027 sys_utimes(td, uap)
3028 	struct thread *td;
3029 	register struct utimes_args /* {
3030 		char *path;
3031 		struct timeval *tptr;
3032 	} */ *uap;
3033 {
3034 
3035 	return (kern_utimesat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
3036 	    uap->tptr, UIO_USERSPACE));
3037 }
3038 
3039 #ifndef _SYS_SYSPROTO_H_
3040 struct futimesat_args {
3041 	int fd;
3042 	const char * path;
3043 	const struct timeval * times;
3044 };
3045 #endif
3046 int
3047 sys_futimesat(struct thread *td, struct futimesat_args *uap)
3048 {
3049 
3050 	return (kern_utimesat(td, uap->fd, uap->path, UIO_USERSPACE,
3051 	    uap->times, UIO_USERSPACE));
3052 }
3053 
3054 int
3055 kern_utimesat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
3056     struct timeval *tptr, enum uio_seg tptrseg)
3057 {
3058 	struct nameidata nd;
3059 	struct timespec ts[2];
3060 	cap_rights_t rights;
3061 	int error;
3062 
3063 	if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3064 		return (error);
3065 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, fd,
3066 	    cap_rights_init(&rights, CAP_FUTIMES), td);
3067 
3068 	if ((error = namei(&nd)) != 0)
3069 		return (error);
3070 	NDFREE(&nd, NDF_ONLY_PNBUF);
3071 	error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3072 	vrele(nd.ni_vp);
3073 	return (error);
3074 }
3075 
3076 /*
3077  * Set the access and modification times of a file.
3078  */
3079 #ifndef _SYS_SYSPROTO_H_
3080 struct lutimes_args {
3081 	char	*path;
3082 	struct	timeval *tptr;
3083 };
3084 #endif
3085 int
3086 sys_lutimes(td, uap)
3087 	struct thread *td;
3088 	register struct lutimes_args /* {
3089 		char *path;
3090 		struct timeval *tptr;
3091 	} */ *uap;
3092 {
3093 
3094 	return (kern_lutimes(td, uap->path, UIO_USERSPACE, uap->tptr,
3095 	    UIO_USERSPACE));
3096 }
3097 
3098 int
3099 kern_lutimes(struct thread *td, char *path, enum uio_seg pathseg,
3100     struct timeval *tptr, enum uio_seg tptrseg)
3101 {
3102 	struct timespec ts[2];
3103 	struct nameidata nd;
3104 	int error;
3105 
3106 	if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3107 		return (error);
3108 	NDINIT(&nd, LOOKUP, NOFOLLOW | AUDITVNODE1, pathseg, path, td);
3109 	if ((error = namei(&nd)) != 0)
3110 		return (error);
3111 	NDFREE(&nd, NDF_ONLY_PNBUF);
3112 	error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3113 	vrele(nd.ni_vp);
3114 	return (error);
3115 }
3116 
3117 /*
3118  * Set the access and modification times of a file.
3119  */
3120 #ifndef _SYS_SYSPROTO_H_
3121 struct futimes_args {
3122 	int	fd;
3123 	struct	timeval *tptr;
3124 };
3125 #endif
3126 int
3127 sys_futimes(td, uap)
3128 	struct thread *td;
3129 	register struct futimes_args /* {
3130 		int  fd;
3131 		struct timeval *tptr;
3132 	} */ *uap;
3133 {
3134 
3135 	return (kern_futimes(td, uap->fd, uap->tptr, UIO_USERSPACE));
3136 }
3137 
3138 int
3139 kern_futimes(struct thread *td, int fd, struct timeval *tptr,
3140     enum uio_seg tptrseg)
3141 {
3142 	struct timespec ts[2];
3143 	struct file *fp;
3144 	cap_rights_t rights;
3145 	int error;
3146 
3147 	AUDIT_ARG_FD(fd);
3148 	error = getutimes(tptr, tptrseg, ts);
3149 	if (error != 0)
3150 		return (error);
3151 	error = getvnode(td, fd, cap_rights_init(&rights, CAP_FUTIMES), &fp);
3152 	if (error != 0)
3153 		return (error);
3154 #ifdef AUDIT
3155 	vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
3156 	AUDIT_ARG_VNODE1(fp->f_vnode);
3157 	VOP_UNLOCK(fp->f_vnode, 0);
3158 #endif
3159 	error = setutimes(td, fp->f_vnode, ts, 2, tptr == NULL);
3160 	fdrop(fp, td);
3161 	return (error);
3162 }
3163 
3164 int
3165 sys_futimens(struct thread *td, struct futimens_args *uap)
3166 {
3167 
3168 	return (kern_futimens(td, uap->fd, uap->times, UIO_USERSPACE));
3169 }
3170 
3171 int
3172 kern_futimens(struct thread *td, int fd, struct timespec *tptr,
3173     enum uio_seg tptrseg)
3174 {
3175 	struct timespec ts[2];
3176 	struct file *fp;
3177 	cap_rights_t rights;
3178 	int error, flags;
3179 
3180 	AUDIT_ARG_FD(fd);
3181 	error = getutimens(tptr, tptrseg, ts, &flags);
3182 	if (error != 0)
3183 		return (error);
3184 	if (flags & UTIMENS_EXIT)
3185 		return (0);
3186 	error = getvnode(td, fd, cap_rights_init(&rights, CAP_FUTIMES), &fp);
3187 	if (error != 0)
3188 		return (error);
3189 #ifdef AUDIT
3190 	vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
3191 	AUDIT_ARG_VNODE1(fp->f_vnode);
3192 	VOP_UNLOCK(fp->f_vnode, 0);
3193 #endif
3194 	error = setutimes(td, fp->f_vnode, ts, 2, flags & UTIMENS_NULL);
3195 	fdrop(fp, td);
3196 	return (error);
3197 }
3198 
3199 int
3200 sys_utimensat(struct thread *td, struct utimensat_args *uap)
3201 {
3202 
3203 	return (kern_utimensat(td, uap->fd, uap->path, UIO_USERSPACE,
3204 	    uap->times, UIO_USERSPACE, uap->flag));
3205 }
3206 
3207 int
3208 kern_utimensat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
3209     struct timespec *tptr, enum uio_seg tptrseg, int flag)
3210 {
3211 	struct nameidata nd;
3212 	struct timespec ts[2];
3213 	cap_rights_t rights;
3214 	int error, flags;
3215 
3216 	if (flag & ~AT_SYMLINK_NOFOLLOW)
3217 		return (EINVAL);
3218 
3219 	if ((error = getutimens(tptr, tptrseg, ts, &flags)) != 0)
3220 		return (error);
3221 	NDINIT_ATRIGHTS(&nd, LOOKUP, ((flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW :
3222 	    FOLLOW) | AUDITVNODE1, pathseg, path, fd,
3223 	    cap_rights_init(&rights, CAP_FUTIMES), td);
3224 	if ((error = namei(&nd)) != 0)
3225 		return (error);
3226 	/*
3227 	 * We are allowed to call namei() regardless of 2xUTIME_OMIT.
3228 	 * POSIX states:
3229 	 * "If both tv_nsec fields are UTIME_OMIT... EACCESS may be detected."
3230 	 * "Search permission is denied by a component of the path prefix."
3231 	 */
3232 	NDFREE(&nd, NDF_ONLY_PNBUF);
3233 	if ((flags & UTIMENS_EXIT) == 0)
3234 		error = setutimes(td, nd.ni_vp, ts, 2, flags & UTIMENS_NULL);
3235 	vrele(nd.ni_vp);
3236 	return (error);
3237 }
3238 
3239 /*
3240  * Truncate a file given its path name.
3241  */
3242 #ifndef _SYS_SYSPROTO_H_
3243 struct truncate_args {
3244 	char	*path;
3245 	int	pad;
3246 	off_t	length;
3247 };
3248 #endif
3249 int
3250 sys_truncate(td, uap)
3251 	struct thread *td;
3252 	register struct truncate_args /* {
3253 		char *path;
3254 		int pad;
3255 		off_t length;
3256 	} */ *uap;
3257 {
3258 
3259 	return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3260 }
3261 
3262 int
3263 kern_truncate(struct thread *td, char *path, enum uio_seg pathseg, off_t length)
3264 {
3265 	struct mount *mp;
3266 	struct vnode *vp;
3267 	void *rl_cookie;
3268 	struct vattr vattr;
3269 	struct nameidata nd;
3270 	int error;
3271 
3272 	if (length < 0)
3273 		return(EINVAL);
3274 	NDINIT(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, td);
3275 	if ((error = namei(&nd)) != 0)
3276 		return (error);
3277 	vp = nd.ni_vp;
3278 	rl_cookie = vn_rangelock_wlock(vp, 0, OFF_MAX);
3279 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) {
3280 		vn_rangelock_unlock(vp, rl_cookie);
3281 		vrele(vp);
3282 		return (error);
3283 	}
3284 	NDFREE(&nd, NDF_ONLY_PNBUF);
3285 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3286 	if (vp->v_type == VDIR)
3287 		error = EISDIR;
3288 #ifdef MAC
3289 	else if ((error = mac_vnode_check_write(td->td_ucred, NOCRED, vp))) {
3290 	}
3291 #endif
3292 	else if ((error = vn_writechk(vp)) == 0 &&
3293 	    (error = VOP_ACCESS(vp, VWRITE, td->td_ucred, td)) == 0) {
3294 		VATTR_NULL(&vattr);
3295 		vattr.va_size = length;
3296 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3297 	}
3298 	VOP_UNLOCK(vp, 0);
3299 	vn_finished_write(mp);
3300 	vn_rangelock_unlock(vp, rl_cookie);
3301 	vrele(vp);
3302 	return (error);
3303 }
3304 
3305 #if defined(COMPAT_43)
3306 /*
3307  * Truncate a file given its path name.
3308  */
3309 #ifndef _SYS_SYSPROTO_H_
3310 struct otruncate_args {
3311 	char	*path;
3312 	long	length;
3313 };
3314 #endif
3315 int
3316 otruncate(td, uap)
3317 	struct thread *td;
3318 	register struct otruncate_args /* {
3319 		char *path;
3320 		long length;
3321 	} */ *uap;
3322 {
3323 	struct truncate_args /* {
3324 		char *path;
3325 		int pad;
3326 		off_t length;
3327 	} */ nuap;
3328 
3329 	nuap.path = uap->path;
3330 	nuap.length = uap->length;
3331 	return (sys_truncate(td, &nuap));
3332 }
3333 #endif /* COMPAT_43 */
3334 
3335 #if defined(COMPAT_FREEBSD6)
3336 /* Versions with the pad argument */
3337 int
3338 freebsd6_truncate(struct thread *td, struct freebsd6_truncate_args *uap)
3339 {
3340 	struct truncate_args ouap;
3341 
3342 	ouap.path = uap->path;
3343 	ouap.length = uap->length;
3344 	return (sys_truncate(td, &ouap));
3345 }
3346 
3347 int
3348 freebsd6_ftruncate(struct thread *td, struct freebsd6_ftruncate_args *uap)
3349 {
3350 	struct ftruncate_args ouap;
3351 
3352 	ouap.fd = uap->fd;
3353 	ouap.length = uap->length;
3354 	return (sys_ftruncate(td, &ouap));
3355 }
3356 #endif
3357 
3358 int
3359 kern_fsync(struct thread *td, int fd, bool fullsync)
3360 {
3361 	struct vnode *vp;
3362 	struct mount *mp;
3363 	struct file *fp;
3364 	cap_rights_t rights;
3365 	int error, lock_flags;
3366 
3367 	AUDIT_ARG_FD(fd);
3368 	error = getvnode(td, fd, cap_rights_init(&rights, CAP_FSYNC), &fp);
3369 	if (error != 0)
3370 		return (error);
3371 	vp = fp->f_vnode;
3372 #if 0
3373 	if (!fullsync)
3374 		/* XXXKIB: compete outstanding aio writes */;
3375 #endif
3376 	error = vn_start_write(vp, &mp, V_WAIT | PCATCH);
3377 	if (error != 0)
3378 		goto drop;
3379 	if (MNT_SHARED_WRITES(mp) ||
3380 	    ((mp == NULL) && MNT_SHARED_WRITES(vp->v_mount))) {
3381 		lock_flags = LK_SHARED;
3382 	} else {
3383 		lock_flags = LK_EXCLUSIVE;
3384 	}
3385 	vn_lock(vp, lock_flags | LK_RETRY);
3386 	AUDIT_ARG_VNODE1(vp);
3387 	if (vp->v_object != NULL) {
3388 		VM_OBJECT_WLOCK(vp->v_object);
3389 		vm_object_page_clean(vp->v_object, 0, 0, 0);
3390 		VM_OBJECT_WUNLOCK(vp->v_object);
3391 	}
3392 	error = fullsync ? VOP_FSYNC(vp, MNT_WAIT, td) : VOP_FDATASYNC(vp, td);
3393 	VOP_UNLOCK(vp, 0);
3394 	vn_finished_write(mp);
3395 drop:
3396 	fdrop(fp, td);
3397 	return (error);
3398 }
3399 
3400 /*
3401  * Sync an open file.
3402  */
3403 #ifndef _SYS_SYSPROTO_H_
3404 struct fsync_args {
3405 	int	fd;
3406 };
3407 #endif
3408 int
3409 sys_fsync(struct thread *td, struct fsync_args *uap)
3410 {
3411 
3412 	return (kern_fsync(td, uap->fd, true));
3413 }
3414 
3415 int
3416 sys_fdatasync(struct thread *td, struct fdatasync_args *uap)
3417 {
3418 
3419 	return (kern_fsync(td, uap->fd, false));
3420 }
3421 
3422 /*
3423  * Rename files.  Source and destination must either both be directories, or
3424  * both not be directories.  If target is a directory, it must be empty.
3425  */
3426 #ifndef _SYS_SYSPROTO_H_
3427 struct rename_args {
3428 	char	*from;
3429 	char	*to;
3430 };
3431 #endif
3432 int
3433 sys_rename(td, uap)
3434 	struct thread *td;
3435 	register struct rename_args /* {
3436 		char *from;
3437 		char *to;
3438 	} */ *uap;
3439 {
3440 
3441 	return (kern_renameat(td, AT_FDCWD, uap->from, AT_FDCWD,
3442 	    uap->to, UIO_USERSPACE));
3443 }
3444 
3445 #ifndef _SYS_SYSPROTO_H_
3446 struct renameat_args {
3447 	int	oldfd;
3448 	char	*old;
3449 	int	newfd;
3450 	char	*new;
3451 };
3452 #endif
3453 int
3454 sys_renameat(struct thread *td, struct renameat_args *uap)
3455 {
3456 
3457 	return (kern_renameat(td, uap->oldfd, uap->old, uap->newfd, uap->new,
3458 	    UIO_USERSPACE));
3459 }
3460 
3461 int
3462 kern_renameat(struct thread *td, int oldfd, char *old, int newfd, char *new,
3463     enum uio_seg pathseg)
3464 {
3465 	struct mount *mp = NULL;
3466 	struct vnode *tvp, *fvp, *tdvp;
3467 	struct nameidata fromnd, tond;
3468 	cap_rights_t rights;
3469 	int error;
3470 
3471 again:
3472 	bwillwrite();
3473 #ifdef MAC
3474 	NDINIT_ATRIGHTS(&fromnd, DELETE, LOCKPARENT | LOCKLEAF | SAVESTART |
3475 	    AUDITVNODE1, pathseg, old, oldfd,
3476 	    cap_rights_init(&rights, CAP_RENAMEAT_SOURCE), td);
3477 #else
3478 	NDINIT_ATRIGHTS(&fromnd, DELETE, WANTPARENT | SAVESTART | AUDITVNODE1,
3479 	    pathseg, old, oldfd,
3480 	    cap_rights_init(&rights, CAP_RENAMEAT_SOURCE), td);
3481 #endif
3482 
3483 	if ((error = namei(&fromnd)) != 0)
3484 		return (error);
3485 #ifdef MAC
3486 	error = mac_vnode_check_rename_from(td->td_ucred, fromnd.ni_dvp,
3487 	    fromnd.ni_vp, &fromnd.ni_cnd);
3488 	VOP_UNLOCK(fromnd.ni_dvp, 0);
3489 	if (fromnd.ni_dvp != fromnd.ni_vp)
3490 		VOP_UNLOCK(fromnd.ni_vp, 0);
3491 #endif
3492 	fvp = fromnd.ni_vp;
3493 	NDINIT_ATRIGHTS(&tond, RENAME, LOCKPARENT | LOCKLEAF | NOCACHE |
3494 	    SAVESTART | AUDITVNODE2, pathseg, new, newfd,
3495 	    cap_rights_init(&rights, CAP_RENAMEAT_TARGET), td);
3496 	if (fromnd.ni_vp->v_type == VDIR)
3497 		tond.ni_cnd.cn_flags |= WILLBEDIR;
3498 	if ((error = namei(&tond)) != 0) {
3499 		/* Translate error code for rename("dir1", "dir2/."). */
3500 		if (error == EISDIR && fvp->v_type == VDIR)
3501 			error = EINVAL;
3502 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3503 		vrele(fromnd.ni_dvp);
3504 		vrele(fvp);
3505 		goto out1;
3506 	}
3507 	tdvp = tond.ni_dvp;
3508 	tvp = tond.ni_vp;
3509 	error = vn_start_write(fvp, &mp, V_NOWAIT);
3510 	if (error != 0) {
3511 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3512 		NDFREE(&tond, NDF_ONLY_PNBUF);
3513 		if (tvp != NULL)
3514 			vput(tvp);
3515 		if (tdvp == tvp)
3516 			vrele(tdvp);
3517 		else
3518 			vput(tdvp);
3519 		vrele(fromnd.ni_dvp);
3520 		vrele(fvp);
3521 		vrele(tond.ni_startdir);
3522 		if (fromnd.ni_startdir != NULL)
3523 			vrele(fromnd.ni_startdir);
3524 		error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH);
3525 		if (error != 0)
3526 			return (error);
3527 		goto again;
3528 	}
3529 	if (tvp != NULL) {
3530 		if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
3531 			error = ENOTDIR;
3532 			goto out;
3533 		} else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
3534 			error = EISDIR;
3535 			goto out;
3536 		}
3537 #ifdef CAPABILITIES
3538 		if (newfd != AT_FDCWD) {
3539 			/*
3540 			 * If the target already exists we require CAP_UNLINKAT
3541 			 * from 'newfd'.
3542 			 */
3543 			error = cap_check(&tond.ni_filecaps.fc_rights,
3544 			    cap_rights_init(&rights, CAP_UNLINKAT));
3545 			if (error != 0)
3546 				goto out;
3547 		}
3548 #endif
3549 	}
3550 	if (fvp == tdvp) {
3551 		error = EINVAL;
3552 		goto out;
3553 	}
3554 	/*
3555 	 * If the source is the same as the destination (that is, if they
3556 	 * are links to the same vnode), then there is nothing to do.
3557 	 */
3558 	if (fvp == tvp)
3559 		error = -1;
3560 #ifdef MAC
3561 	else
3562 		error = mac_vnode_check_rename_to(td->td_ucred, tdvp,
3563 		    tond.ni_vp, fromnd.ni_dvp == tdvp, &tond.ni_cnd);
3564 #endif
3565 out:
3566 	if (error == 0) {
3567 		error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
3568 		    tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
3569 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3570 		NDFREE(&tond, NDF_ONLY_PNBUF);
3571 	} else {
3572 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3573 		NDFREE(&tond, NDF_ONLY_PNBUF);
3574 		if (tvp != NULL)
3575 			vput(tvp);
3576 		if (tdvp == tvp)
3577 			vrele(tdvp);
3578 		else
3579 			vput(tdvp);
3580 		vrele(fromnd.ni_dvp);
3581 		vrele(fvp);
3582 	}
3583 	vrele(tond.ni_startdir);
3584 	vn_finished_write(mp);
3585 out1:
3586 	if (fromnd.ni_startdir)
3587 		vrele(fromnd.ni_startdir);
3588 	if (error == -1)
3589 		return (0);
3590 	return (error);
3591 }
3592 
3593 /*
3594  * Make a directory file.
3595  */
3596 #ifndef _SYS_SYSPROTO_H_
3597 struct mkdir_args {
3598 	char	*path;
3599 	int	mode;
3600 };
3601 #endif
3602 int
3603 sys_mkdir(td, uap)
3604 	struct thread *td;
3605 	register struct mkdir_args /* {
3606 		char *path;
3607 		int mode;
3608 	} */ *uap;
3609 {
3610 
3611 	return (kern_mkdirat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
3612 	    uap->mode));
3613 }
3614 
3615 #ifndef _SYS_SYSPROTO_H_
3616 struct mkdirat_args {
3617 	int	fd;
3618 	char	*path;
3619 	mode_t	mode;
3620 };
3621 #endif
3622 int
3623 sys_mkdirat(struct thread *td, struct mkdirat_args *uap)
3624 {
3625 
3626 	return (kern_mkdirat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode));
3627 }
3628 
3629 int
3630 kern_mkdirat(struct thread *td, int fd, char *path, enum uio_seg segflg,
3631     int mode)
3632 {
3633 	struct mount *mp;
3634 	struct vnode *vp;
3635 	struct vattr vattr;
3636 	struct nameidata nd;
3637 	cap_rights_t rights;
3638 	int error;
3639 
3640 	AUDIT_ARG_MODE(mode);
3641 restart:
3642 	bwillwrite();
3643 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
3644 	    NOCACHE, segflg, path, fd, cap_rights_init(&rights, CAP_MKDIRAT),
3645 	    td);
3646 	nd.ni_cnd.cn_flags |= WILLBEDIR;
3647 	if ((error = namei(&nd)) != 0)
3648 		return (error);
3649 	vp = nd.ni_vp;
3650 	if (vp != NULL) {
3651 		NDFREE(&nd, NDF_ONLY_PNBUF);
3652 		/*
3653 		 * XXX namei called with LOCKPARENT but not LOCKLEAF has
3654 		 * the strange behaviour of leaving the vnode unlocked
3655 		 * if the target is the same vnode as the parent.
3656 		 */
3657 		if (vp == nd.ni_dvp)
3658 			vrele(nd.ni_dvp);
3659 		else
3660 			vput(nd.ni_dvp);
3661 		vrele(vp);
3662 		return (EEXIST);
3663 	}
3664 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3665 		NDFREE(&nd, NDF_ONLY_PNBUF);
3666 		vput(nd.ni_dvp);
3667 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3668 			return (error);
3669 		goto restart;
3670 	}
3671 	VATTR_NULL(&vattr);
3672 	vattr.va_type = VDIR;
3673 	vattr.va_mode = (mode & ACCESSPERMS) &~ td->td_proc->p_fd->fd_cmask;
3674 #ifdef MAC
3675 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
3676 	    &vattr);
3677 	if (error != 0)
3678 		goto out;
3679 #endif
3680 	error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
3681 #ifdef MAC
3682 out:
3683 #endif
3684 	NDFREE(&nd, NDF_ONLY_PNBUF);
3685 	vput(nd.ni_dvp);
3686 	if (error == 0)
3687 		vput(nd.ni_vp);
3688 	vn_finished_write(mp);
3689 	return (error);
3690 }
3691 
3692 /*
3693  * Remove a directory file.
3694  */
3695 #ifndef _SYS_SYSPROTO_H_
3696 struct rmdir_args {
3697 	char	*path;
3698 };
3699 #endif
3700 int
3701 sys_rmdir(td, uap)
3702 	struct thread *td;
3703 	struct rmdir_args /* {
3704 		char *path;
3705 	} */ *uap;
3706 {
3707 
3708 	return (kern_rmdirat(td, AT_FDCWD, uap->path, UIO_USERSPACE));
3709 }
3710 
3711 int
3712 kern_rmdirat(struct thread *td, int fd, char *path, enum uio_seg pathseg)
3713 {
3714 	struct mount *mp;
3715 	struct vnode *vp;
3716 	struct nameidata nd;
3717 	cap_rights_t rights;
3718 	int error;
3719 
3720 restart:
3721 	bwillwrite();
3722 	NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1,
3723 	    pathseg, path, fd, cap_rights_init(&rights, CAP_UNLINKAT), td);
3724 	if ((error = namei(&nd)) != 0)
3725 		return (error);
3726 	vp = nd.ni_vp;
3727 	if (vp->v_type != VDIR) {
3728 		error = ENOTDIR;
3729 		goto out;
3730 	}
3731 	/*
3732 	 * No rmdir "." please.
3733 	 */
3734 	if (nd.ni_dvp == vp) {
3735 		error = EINVAL;
3736 		goto out;
3737 	}
3738 	/*
3739 	 * The root of a mounted filesystem cannot be deleted.
3740 	 */
3741 	if (vp->v_vflag & VV_ROOT) {
3742 		error = EBUSY;
3743 		goto out;
3744 	}
3745 #ifdef MAC
3746 	error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
3747 	    &nd.ni_cnd);
3748 	if (error != 0)
3749 		goto out;
3750 #endif
3751 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3752 		NDFREE(&nd, NDF_ONLY_PNBUF);
3753 		vput(vp);
3754 		if (nd.ni_dvp == vp)
3755 			vrele(nd.ni_dvp);
3756 		else
3757 			vput(nd.ni_dvp);
3758 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3759 			return (error);
3760 		goto restart;
3761 	}
3762 	vfs_notify_upper(vp, VFS_NOTIFY_UPPER_UNLINK);
3763 	error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
3764 	vn_finished_write(mp);
3765 out:
3766 	NDFREE(&nd, NDF_ONLY_PNBUF);
3767 	vput(vp);
3768 	if (nd.ni_dvp == vp)
3769 		vrele(nd.ni_dvp);
3770 	else
3771 		vput(nd.ni_dvp);
3772 	return (error);
3773 }
3774 
3775 #ifdef COMPAT_43
3776 /*
3777  * Read a block of directory entries in a filesystem independent format.
3778  */
3779 #ifndef _SYS_SYSPROTO_H_
3780 struct ogetdirentries_args {
3781 	int	fd;
3782 	char	*buf;
3783 	u_int	count;
3784 	long	*basep;
3785 };
3786 #endif
3787 int
3788 ogetdirentries(struct thread *td, struct ogetdirentries_args *uap)
3789 {
3790 	long loff;
3791 	int error;
3792 
3793 	error = kern_ogetdirentries(td, uap, &loff);
3794 	if (error == 0)
3795 		error = copyout(&loff, uap->basep, sizeof(long));
3796 	return (error);
3797 }
3798 
3799 int
3800 kern_ogetdirentries(struct thread *td, struct ogetdirentries_args *uap,
3801     long *ploff)
3802 {
3803 	struct vnode *vp;
3804 	struct file *fp;
3805 	struct uio auio, kuio;
3806 	struct iovec aiov, kiov;
3807 	struct dirent *dp, *edp;
3808 	cap_rights_t rights;
3809 	caddr_t dirbuf;
3810 	int error, eofflag, readcnt;
3811 	long loff;
3812 	off_t foffset;
3813 
3814 	/* XXX arbitrary sanity limit on `count'. */
3815 	if (uap->count > 64 * 1024)
3816 		return (EINVAL);
3817 	error = getvnode(td, uap->fd, cap_rights_init(&rights, CAP_READ), &fp);
3818 	if (error != 0)
3819 		return (error);
3820 	if ((fp->f_flag & FREAD) == 0) {
3821 		fdrop(fp, td);
3822 		return (EBADF);
3823 	}
3824 	vp = fp->f_vnode;
3825 	foffset = foffset_lock(fp, 0);
3826 unionread:
3827 	if (vp->v_type != VDIR) {
3828 		foffset_unlock(fp, foffset, 0);
3829 		fdrop(fp, td);
3830 		return (EINVAL);
3831 	}
3832 	aiov.iov_base = uap->buf;
3833 	aiov.iov_len = uap->count;
3834 	auio.uio_iov = &aiov;
3835 	auio.uio_iovcnt = 1;
3836 	auio.uio_rw = UIO_READ;
3837 	auio.uio_segflg = UIO_USERSPACE;
3838 	auio.uio_td = td;
3839 	auio.uio_resid = uap->count;
3840 	vn_lock(vp, LK_SHARED | LK_RETRY);
3841 	loff = auio.uio_offset = foffset;
3842 #ifdef MAC
3843 	error = mac_vnode_check_readdir(td->td_ucred, vp);
3844 	if (error != 0) {
3845 		VOP_UNLOCK(vp, 0);
3846 		foffset_unlock(fp, foffset, FOF_NOUPDATE);
3847 		fdrop(fp, td);
3848 		return (error);
3849 	}
3850 #endif
3851 #	if (BYTE_ORDER != LITTLE_ENDIAN)
3852 		if (vp->v_mount->mnt_maxsymlinklen <= 0) {
3853 			error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag,
3854 			    NULL, NULL);
3855 			foffset = auio.uio_offset;
3856 		} else
3857 #	endif
3858 	{
3859 		kuio = auio;
3860 		kuio.uio_iov = &kiov;
3861 		kuio.uio_segflg = UIO_SYSSPACE;
3862 		kiov.iov_len = uap->count;
3863 		dirbuf = malloc(uap->count, M_TEMP, M_WAITOK);
3864 		kiov.iov_base = dirbuf;
3865 		error = VOP_READDIR(vp, &kuio, fp->f_cred, &eofflag,
3866 			    NULL, NULL);
3867 		foffset = kuio.uio_offset;
3868 		if (error == 0) {
3869 			readcnt = uap->count - kuio.uio_resid;
3870 			edp = (struct dirent *)&dirbuf[readcnt];
3871 			for (dp = (struct dirent *)dirbuf; dp < edp; ) {
3872 #				if (BYTE_ORDER == LITTLE_ENDIAN)
3873 					/*
3874 					 * The expected low byte of
3875 					 * dp->d_namlen is our dp->d_type.
3876 					 * The high MBZ byte of dp->d_namlen
3877 					 * is our dp->d_namlen.
3878 					 */
3879 					dp->d_type = dp->d_namlen;
3880 					dp->d_namlen = 0;
3881 #				else
3882 					/*
3883 					 * The dp->d_type is the high byte
3884 					 * of the expected dp->d_namlen,
3885 					 * so must be zero'ed.
3886 					 */
3887 					dp->d_type = 0;
3888 #				endif
3889 				if (dp->d_reclen > 0) {
3890 					dp = (struct dirent *)
3891 					    ((char *)dp + dp->d_reclen);
3892 				} else {
3893 					error = EIO;
3894 					break;
3895 				}
3896 			}
3897 			if (dp >= edp)
3898 				error = uiomove(dirbuf, readcnt, &auio);
3899 		}
3900 		free(dirbuf, M_TEMP);
3901 	}
3902 	if (error != 0) {
3903 		VOP_UNLOCK(vp, 0);
3904 		foffset_unlock(fp, foffset, 0);
3905 		fdrop(fp, td);
3906 		return (error);
3907 	}
3908 	if (uap->count == auio.uio_resid &&
3909 	    (vp->v_vflag & VV_ROOT) &&
3910 	    (vp->v_mount->mnt_flag & MNT_UNION)) {
3911 		struct vnode *tvp = vp;
3912 		vp = vp->v_mount->mnt_vnodecovered;
3913 		VREF(vp);
3914 		fp->f_vnode = vp;
3915 		fp->f_data = vp;
3916 		foffset = 0;
3917 		vput(tvp);
3918 		goto unionread;
3919 	}
3920 	VOP_UNLOCK(vp, 0);
3921 	foffset_unlock(fp, foffset, 0);
3922 	fdrop(fp, td);
3923 	td->td_retval[0] = uap->count - auio.uio_resid;
3924 	if (error == 0)
3925 		*ploff = loff;
3926 	return (error);
3927 }
3928 #endif /* COMPAT_43 */
3929 
3930 /*
3931  * Read a block of directory entries in a filesystem independent format.
3932  */
3933 #ifndef _SYS_SYSPROTO_H_
3934 struct getdirentries_args {
3935 	int	fd;
3936 	char	*buf;
3937 	u_int	count;
3938 	long	*basep;
3939 };
3940 #endif
3941 int
3942 sys_getdirentries(td, uap)
3943 	struct thread *td;
3944 	register struct getdirentries_args /* {
3945 		int fd;
3946 		char *buf;
3947 		u_int count;
3948 		long *basep;
3949 	} */ *uap;
3950 {
3951 	long base;
3952 	int error;
3953 
3954 	error = kern_getdirentries(td, uap->fd, uap->buf, uap->count, &base,
3955 	    NULL, UIO_USERSPACE);
3956 	if (error != 0)
3957 		return (error);
3958 	if (uap->basep != NULL)
3959 		error = copyout(&base, uap->basep, sizeof(long));
3960 	return (error);
3961 }
3962 
3963 int
3964 kern_getdirentries(struct thread *td, int fd, char *buf, u_int count,
3965     long *basep, ssize_t *residp, enum uio_seg bufseg)
3966 {
3967 	struct vnode *vp;
3968 	struct file *fp;
3969 	struct uio auio;
3970 	struct iovec aiov;
3971 	cap_rights_t rights;
3972 	long loff;
3973 	int error, eofflag;
3974 	off_t foffset;
3975 
3976 	AUDIT_ARG_FD(fd);
3977 	if (count > IOSIZE_MAX)
3978 		return (EINVAL);
3979 	auio.uio_resid = count;
3980 	error = getvnode(td, fd, cap_rights_init(&rights, CAP_READ), &fp);
3981 	if (error != 0)
3982 		return (error);
3983 	if ((fp->f_flag & FREAD) == 0) {
3984 		fdrop(fp, td);
3985 		return (EBADF);
3986 	}
3987 	vp = fp->f_vnode;
3988 	foffset = foffset_lock(fp, 0);
3989 unionread:
3990 	if (vp->v_type != VDIR) {
3991 		error = EINVAL;
3992 		goto fail;
3993 	}
3994 	aiov.iov_base = buf;
3995 	aiov.iov_len = count;
3996 	auio.uio_iov = &aiov;
3997 	auio.uio_iovcnt = 1;
3998 	auio.uio_rw = UIO_READ;
3999 	auio.uio_segflg = bufseg;
4000 	auio.uio_td = td;
4001 	vn_lock(vp, LK_SHARED | LK_RETRY);
4002 	AUDIT_ARG_VNODE1(vp);
4003 	loff = auio.uio_offset = foffset;
4004 #ifdef MAC
4005 	error = mac_vnode_check_readdir(td->td_ucred, vp);
4006 	if (error == 0)
4007 #endif
4008 		error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, NULL,
4009 		    NULL);
4010 	foffset = auio.uio_offset;
4011 	if (error != 0) {
4012 		VOP_UNLOCK(vp, 0);
4013 		goto fail;
4014 	}
4015 	if (count == auio.uio_resid &&
4016 	    (vp->v_vflag & VV_ROOT) &&
4017 	    (vp->v_mount->mnt_flag & MNT_UNION)) {
4018 		struct vnode *tvp = vp;
4019 
4020 		vp = vp->v_mount->mnt_vnodecovered;
4021 		VREF(vp);
4022 		fp->f_vnode = vp;
4023 		fp->f_data = vp;
4024 		foffset = 0;
4025 		vput(tvp);
4026 		goto unionread;
4027 	}
4028 	VOP_UNLOCK(vp, 0);
4029 	*basep = loff;
4030 	if (residp != NULL)
4031 		*residp = auio.uio_resid;
4032 	td->td_retval[0] = count - auio.uio_resid;
4033 fail:
4034 	foffset_unlock(fp, foffset, 0);
4035 	fdrop(fp, td);
4036 	return (error);
4037 }
4038 
4039 #ifndef _SYS_SYSPROTO_H_
4040 struct getdents_args {
4041 	int fd;
4042 	char *buf;
4043 	size_t count;
4044 };
4045 #endif
4046 int
4047 sys_getdents(td, uap)
4048 	struct thread *td;
4049 	register struct getdents_args /* {
4050 		int fd;
4051 		char *buf;
4052 		u_int count;
4053 	} */ *uap;
4054 {
4055 	struct getdirentries_args ap;
4056 
4057 	ap.fd = uap->fd;
4058 	ap.buf = uap->buf;
4059 	ap.count = uap->count;
4060 	ap.basep = NULL;
4061 	return (sys_getdirentries(td, &ap));
4062 }
4063 
4064 /*
4065  * Set the mode mask for creation of filesystem nodes.
4066  */
4067 #ifndef _SYS_SYSPROTO_H_
4068 struct umask_args {
4069 	int	newmask;
4070 };
4071 #endif
4072 int
4073 sys_umask(td, uap)
4074 	struct thread *td;
4075 	struct umask_args /* {
4076 		int newmask;
4077 	} */ *uap;
4078 {
4079 	struct filedesc *fdp;
4080 
4081 	fdp = td->td_proc->p_fd;
4082 	FILEDESC_XLOCK(fdp);
4083 	td->td_retval[0] = fdp->fd_cmask;
4084 	fdp->fd_cmask = uap->newmask & ALLPERMS;
4085 	FILEDESC_XUNLOCK(fdp);
4086 	return (0);
4087 }
4088 
4089 /*
4090  * Void all references to file by ripping underlying filesystem away from
4091  * vnode.
4092  */
4093 #ifndef _SYS_SYSPROTO_H_
4094 struct revoke_args {
4095 	char	*path;
4096 };
4097 #endif
4098 int
4099 sys_revoke(td, uap)
4100 	struct thread *td;
4101 	register struct revoke_args /* {
4102 		char *path;
4103 	} */ *uap;
4104 {
4105 	struct vnode *vp;
4106 	struct vattr vattr;
4107 	struct nameidata nd;
4108 	int error;
4109 
4110 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
4111 	    uap->path, td);
4112 	if ((error = namei(&nd)) != 0)
4113 		return (error);
4114 	vp = nd.ni_vp;
4115 	NDFREE(&nd, NDF_ONLY_PNBUF);
4116 	if (vp->v_type != VCHR || vp->v_rdev == NULL) {
4117 		error = EINVAL;
4118 		goto out;
4119 	}
4120 #ifdef MAC
4121 	error = mac_vnode_check_revoke(td->td_ucred, vp);
4122 	if (error != 0)
4123 		goto out;
4124 #endif
4125 	error = VOP_GETATTR(vp, &vattr, td->td_ucred);
4126 	if (error != 0)
4127 		goto out;
4128 	if (td->td_ucred->cr_uid != vattr.va_uid) {
4129 		error = priv_check(td, PRIV_VFS_ADMIN);
4130 		if (error != 0)
4131 			goto out;
4132 	}
4133 	if (vcount(vp) > 1)
4134 		VOP_REVOKE(vp, REVOKEALL);
4135 out:
4136 	vput(vp);
4137 	return (error);
4138 }
4139 
4140 /*
4141  * Convert a user file descriptor to a kernel file entry and check that, if it
4142  * is a capability, the correct rights are present. A reference on the file
4143  * entry is held upon returning.
4144  */
4145 int
4146 getvnode(struct thread *td, int fd, cap_rights_t *rightsp, struct file **fpp)
4147 {
4148 	struct file *fp;
4149 	int error;
4150 
4151 	error = fget_unlocked(td->td_proc->p_fd, fd, rightsp, &fp, NULL);
4152 	if (error != 0)
4153 		return (error);
4154 
4155 	/*
4156 	 * The file could be not of the vnode type, or it may be not
4157 	 * yet fully initialized, in which case the f_vnode pointer
4158 	 * may be set, but f_ops is still badfileops.  E.g.,
4159 	 * devfs_open() transiently create such situation to
4160 	 * facilitate csw d_fdopen().
4161 	 *
4162 	 * Dupfdopen() handling in kern_openat() installs the
4163 	 * half-baked file into the process descriptor table, allowing
4164 	 * other thread to dereference it. Guard against the race by
4165 	 * checking f_ops.
4166 	 */
4167 	if (fp->f_vnode == NULL || fp->f_ops == &badfileops) {
4168 		fdrop(fp, td);
4169 		return (EINVAL);
4170 	}
4171 	*fpp = fp;
4172 	return (0);
4173 }
4174 
4175 
4176 /*
4177  * Get an (NFS) file handle.
4178  */
4179 #ifndef _SYS_SYSPROTO_H_
4180 struct lgetfh_args {
4181 	char	*fname;
4182 	fhandle_t *fhp;
4183 };
4184 #endif
4185 int
4186 sys_lgetfh(td, uap)
4187 	struct thread *td;
4188 	register struct lgetfh_args *uap;
4189 {
4190 	struct nameidata nd;
4191 	fhandle_t fh;
4192 	register struct vnode *vp;
4193 	int error;
4194 
4195 	error = priv_check(td, PRIV_VFS_GETFH);
4196 	if (error != 0)
4197 		return (error);
4198 	NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
4199 	    uap->fname, td);
4200 	error = namei(&nd);
4201 	if (error != 0)
4202 		return (error);
4203 	NDFREE(&nd, NDF_ONLY_PNBUF);
4204 	vp = nd.ni_vp;
4205 	bzero(&fh, sizeof(fh));
4206 	fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
4207 	error = VOP_VPTOFH(vp, &fh.fh_fid);
4208 	vput(vp);
4209 	if (error == 0)
4210 		error = copyout(&fh, uap->fhp, sizeof (fh));
4211 	return (error);
4212 }
4213 
4214 #ifndef _SYS_SYSPROTO_H_
4215 struct getfh_args {
4216 	char	*fname;
4217 	fhandle_t *fhp;
4218 };
4219 #endif
4220 int
4221 sys_getfh(td, uap)
4222 	struct thread *td;
4223 	register struct getfh_args *uap;
4224 {
4225 	struct nameidata nd;
4226 	fhandle_t fh;
4227 	register struct vnode *vp;
4228 	int error;
4229 
4230 	error = priv_check(td, PRIV_VFS_GETFH);
4231 	if (error != 0)
4232 		return (error);
4233 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
4234 	    uap->fname, td);
4235 	error = namei(&nd);
4236 	if (error != 0)
4237 		return (error);
4238 	NDFREE(&nd, NDF_ONLY_PNBUF);
4239 	vp = nd.ni_vp;
4240 	bzero(&fh, sizeof(fh));
4241 	fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
4242 	error = VOP_VPTOFH(vp, &fh.fh_fid);
4243 	vput(vp);
4244 	if (error == 0)
4245 		error = copyout(&fh, uap->fhp, sizeof (fh));
4246 	return (error);
4247 }
4248 
4249 /*
4250  * syscall for the rpc.lockd to use to translate a NFS file handle into an
4251  * open descriptor.
4252  *
4253  * warning: do not remove the priv_check() call or this becomes one giant
4254  * security hole.
4255  */
4256 #ifndef _SYS_SYSPROTO_H_
4257 struct fhopen_args {
4258 	const struct fhandle *u_fhp;
4259 	int flags;
4260 };
4261 #endif
4262 int
4263 sys_fhopen(td, uap)
4264 	struct thread *td;
4265 	struct fhopen_args /* {
4266 		const struct fhandle *u_fhp;
4267 		int flags;
4268 	} */ *uap;
4269 {
4270 	struct mount *mp;
4271 	struct vnode *vp;
4272 	struct fhandle fhp;
4273 	struct file *fp;
4274 	int fmode, error;
4275 	int indx;
4276 
4277 	error = priv_check(td, PRIV_VFS_FHOPEN);
4278 	if (error != 0)
4279 		return (error);
4280 	indx = -1;
4281 	fmode = FFLAGS(uap->flags);
4282 	/* why not allow a non-read/write open for our lockd? */
4283 	if (((fmode & (FREAD | FWRITE)) == 0) || (fmode & O_CREAT))
4284 		return (EINVAL);
4285 	error = copyin(uap->u_fhp, &fhp, sizeof(fhp));
4286 	if (error != 0)
4287 		return(error);
4288 	/* find the mount point */
4289 	mp = vfs_busyfs(&fhp.fh_fsid);
4290 	if (mp == NULL)
4291 		return (ESTALE);
4292 	/* now give me my vnode, it gets returned to me locked */
4293 	error = VFS_FHTOVP(mp, &fhp.fh_fid, LK_EXCLUSIVE, &vp);
4294 	vfs_unbusy(mp);
4295 	if (error != 0)
4296 		return (error);
4297 
4298 	error = falloc_noinstall(td, &fp);
4299 	if (error != 0) {
4300 		vput(vp);
4301 		return (error);
4302 	}
4303 	/*
4304 	 * An extra reference on `fp' has been held for us by
4305 	 * falloc_noinstall().
4306 	 */
4307 
4308 #ifdef INVARIANTS
4309 	td->td_dupfd = -1;
4310 #endif
4311 	error = vn_open_vnode(vp, fmode, td->td_ucred, td, fp);
4312 	if (error != 0) {
4313 		KASSERT(fp->f_ops == &badfileops,
4314 		    ("VOP_OPEN in fhopen() set f_ops"));
4315 		KASSERT(td->td_dupfd < 0,
4316 		    ("fhopen() encountered fdopen()"));
4317 
4318 		vput(vp);
4319 		goto bad;
4320 	}
4321 #ifdef INVARIANTS
4322 	td->td_dupfd = 0;
4323 #endif
4324 	fp->f_vnode = vp;
4325 	fp->f_seqcount = 1;
4326 	finit(fp, (fmode & FMASK) | (fp->f_flag & FHASLOCK), DTYPE_VNODE, vp,
4327 	    &vnops);
4328 	VOP_UNLOCK(vp, 0);
4329 	if ((fmode & O_TRUNC) != 0) {
4330 		error = fo_truncate(fp, 0, td->td_ucred, td);
4331 		if (error != 0)
4332 			goto bad;
4333 	}
4334 
4335 	error = finstall(td, fp, &indx, fmode, NULL);
4336 bad:
4337 	fdrop(fp, td);
4338 	td->td_retval[0] = indx;
4339 	return (error);
4340 }
4341 
4342 /*
4343  * Stat an (NFS) file handle.
4344  */
4345 #ifndef _SYS_SYSPROTO_H_
4346 struct fhstat_args {
4347 	struct fhandle *u_fhp;
4348 	struct stat *sb;
4349 };
4350 #endif
4351 int
4352 sys_fhstat(td, uap)
4353 	struct thread *td;
4354 	register struct fhstat_args /* {
4355 		struct fhandle *u_fhp;
4356 		struct stat *sb;
4357 	} */ *uap;
4358 {
4359 	struct stat sb;
4360 	struct fhandle fh;
4361 	int error;
4362 
4363 	error = copyin(uap->u_fhp, &fh, sizeof(fh));
4364 	if (error != 0)
4365 		return (error);
4366 	error = kern_fhstat(td, fh, &sb);
4367 	if (error == 0)
4368 		error = copyout(&sb, uap->sb, sizeof(sb));
4369 	return (error);
4370 }
4371 
4372 int
4373 kern_fhstat(struct thread *td, struct fhandle fh, struct stat *sb)
4374 {
4375 	struct mount *mp;
4376 	struct vnode *vp;
4377 	int error;
4378 
4379 	error = priv_check(td, PRIV_VFS_FHSTAT);
4380 	if (error != 0)
4381 		return (error);
4382 	if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4383 		return (ESTALE);
4384 	error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4385 	vfs_unbusy(mp);
4386 	if (error != 0)
4387 		return (error);
4388 	error = vn_stat(vp, sb, td->td_ucred, NOCRED, td);
4389 	vput(vp);
4390 	return (error);
4391 }
4392 
4393 /*
4394  * Implement fstatfs() for (NFS) file handles.
4395  */
4396 #ifndef _SYS_SYSPROTO_H_
4397 struct fhstatfs_args {
4398 	struct fhandle *u_fhp;
4399 	struct statfs *buf;
4400 };
4401 #endif
4402 int
4403 sys_fhstatfs(td, uap)
4404 	struct thread *td;
4405 	struct fhstatfs_args /* {
4406 		struct fhandle *u_fhp;
4407 		struct statfs *buf;
4408 	} */ *uap;
4409 {
4410 	struct statfs sf;
4411 	fhandle_t fh;
4412 	int error;
4413 
4414 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
4415 	if (error != 0)
4416 		return (error);
4417 	error = kern_fhstatfs(td, fh, &sf);
4418 	if (error != 0)
4419 		return (error);
4420 	return (copyout(&sf, uap->buf, sizeof(sf)));
4421 }
4422 
4423 int
4424 kern_fhstatfs(struct thread *td, fhandle_t fh, struct statfs *buf)
4425 {
4426 	struct statfs *sp;
4427 	struct mount *mp;
4428 	struct vnode *vp;
4429 	int error;
4430 
4431 	error = priv_check(td, PRIV_VFS_FHSTATFS);
4432 	if (error != 0)
4433 		return (error);
4434 	if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4435 		return (ESTALE);
4436 	error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4437 	if (error != 0) {
4438 		vfs_unbusy(mp);
4439 		return (error);
4440 	}
4441 	vput(vp);
4442 	error = prison_canseemount(td->td_ucred, mp);
4443 	if (error != 0)
4444 		goto out;
4445 #ifdef MAC
4446 	error = mac_mount_check_stat(td->td_ucred, mp);
4447 	if (error != 0)
4448 		goto out;
4449 #endif
4450 	/*
4451 	 * Set these in case the underlying filesystem fails to do so.
4452 	 */
4453 	sp = &mp->mnt_stat;
4454 	sp->f_version = STATFS_VERSION;
4455 	sp->f_namemax = NAME_MAX;
4456 	sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
4457 	error = VFS_STATFS(mp, sp);
4458 	if (error == 0)
4459 		*buf = *sp;
4460 out:
4461 	vfs_unbusy(mp);
4462 	return (error);
4463 }
4464 
4465 int
4466 kern_posix_fallocate(struct thread *td, int fd, off_t offset, off_t len)
4467 {
4468 	struct file *fp;
4469 	struct mount *mp;
4470 	struct vnode *vp;
4471 	cap_rights_t rights;
4472 	off_t olen, ooffset;
4473 	int error;
4474 
4475 	if (offset < 0 || len <= 0)
4476 		return (EINVAL);
4477 	/* Check for wrap. */
4478 	if (offset > OFF_MAX - len)
4479 		return (EFBIG);
4480 	error = fget(td, fd, cap_rights_init(&rights, CAP_WRITE), &fp);
4481 	if (error != 0)
4482 		return (error);
4483 	if ((fp->f_ops->fo_flags & DFLAG_SEEKABLE) == 0) {
4484 		error = ESPIPE;
4485 		goto out;
4486 	}
4487 	if ((fp->f_flag & FWRITE) == 0) {
4488 		error = EBADF;
4489 		goto out;
4490 	}
4491 	if (fp->f_type != DTYPE_VNODE) {
4492 		error = ENODEV;
4493 		goto out;
4494 	}
4495 	vp = fp->f_vnode;
4496 	if (vp->v_type != VREG) {
4497 		error = ENODEV;
4498 		goto out;
4499 	}
4500 
4501 	/* Allocating blocks may take a long time, so iterate. */
4502 	for (;;) {
4503 		olen = len;
4504 		ooffset = offset;
4505 
4506 		bwillwrite();
4507 		mp = NULL;
4508 		error = vn_start_write(vp, &mp, V_WAIT | PCATCH);
4509 		if (error != 0)
4510 			break;
4511 		error = vn_lock(vp, LK_EXCLUSIVE);
4512 		if (error != 0) {
4513 			vn_finished_write(mp);
4514 			break;
4515 		}
4516 #ifdef MAC
4517 		error = mac_vnode_check_write(td->td_ucred, fp->f_cred, vp);
4518 		if (error == 0)
4519 #endif
4520 			error = VOP_ALLOCATE(vp, &offset, &len);
4521 		VOP_UNLOCK(vp, 0);
4522 		vn_finished_write(mp);
4523 
4524 		if (olen + ooffset != offset + len) {
4525 			panic("offset + len changed from %jx/%jx to %jx/%jx",
4526 			    ooffset, olen, offset, len);
4527 		}
4528 		if (error != 0 || len == 0)
4529 			break;
4530 		KASSERT(olen > len, ("Iteration did not make progress?"));
4531 		maybe_yield();
4532 	}
4533  out:
4534 	fdrop(fp, td);
4535 	return (error);
4536 }
4537 
4538 int
4539 sys_posix_fallocate(struct thread *td, struct posix_fallocate_args *uap)
4540 {
4541 	int error;
4542 
4543 	error = kern_posix_fallocate(td, uap->fd, uap->offset, uap->len);
4544 	return (kern_posix_error(td, error));
4545 }
4546 
4547 /*
4548  * Unlike madvise(2), we do not make a best effort to remember every
4549  * possible caching hint.  Instead, we remember the last setting with
4550  * the exception that we will allow POSIX_FADV_NORMAL to adjust the
4551  * region of any current setting.
4552  */
4553 int
4554 kern_posix_fadvise(struct thread *td, int fd, off_t offset, off_t len,
4555     int advice)
4556 {
4557 	struct fadvise_info *fa, *new;
4558 	struct file *fp;
4559 	struct vnode *vp;
4560 	cap_rights_t rights;
4561 	off_t end;
4562 	int error;
4563 
4564 	if (offset < 0 || len < 0 || offset > OFF_MAX - len)
4565 		return (EINVAL);
4566 	switch (advice) {
4567 	case POSIX_FADV_SEQUENTIAL:
4568 	case POSIX_FADV_RANDOM:
4569 	case POSIX_FADV_NOREUSE:
4570 		new = malloc(sizeof(*fa), M_FADVISE, M_WAITOK);
4571 		break;
4572 	case POSIX_FADV_NORMAL:
4573 	case POSIX_FADV_WILLNEED:
4574 	case POSIX_FADV_DONTNEED:
4575 		new = NULL;
4576 		break;
4577 	default:
4578 		return (EINVAL);
4579 	}
4580 	/* XXX: CAP_POSIX_FADVISE? */
4581 	error = fget(td, fd, cap_rights_init(&rights), &fp);
4582 	if (error != 0)
4583 		goto out;
4584 	if ((fp->f_ops->fo_flags & DFLAG_SEEKABLE) == 0) {
4585 		error = ESPIPE;
4586 		goto out;
4587 	}
4588 	if (fp->f_type != DTYPE_VNODE) {
4589 		error = ENODEV;
4590 		goto out;
4591 	}
4592 	vp = fp->f_vnode;
4593 	if (vp->v_type != VREG) {
4594 		error = ENODEV;
4595 		goto out;
4596 	}
4597 	if (len == 0)
4598 		end = OFF_MAX;
4599 	else
4600 		end = offset + len - 1;
4601 	switch (advice) {
4602 	case POSIX_FADV_SEQUENTIAL:
4603 	case POSIX_FADV_RANDOM:
4604 	case POSIX_FADV_NOREUSE:
4605 		/*
4606 		 * Try to merge any existing non-standard region with
4607 		 * this new region if possible, otherwise create a new
4608 		 * non-standard region for this request.
4609 		 */
4610 		mtx_pool_lock(mtxpool_sleep, fp);
4611 		fa = fp->f_advice;
4612 		if (fa != NULL && fa->fa_advice == advice &&
4613 		    ((fa->fa_start <= end && fa->fa_end >= offset) ||
4614 		    (end != OFF_MAX && fa->fa_start == end + 1) ||
4615 		    (fa->fa_end != OFF_MAX && fa->fa_end + 1 == offset))) {
4616 			if (offset < fa->fa_start)
4617 				fa->fa_start = offset;
4618 			if (end > fa->fa_end)
4619 				fa->fa_end = end;
4620 		} else {
4621 			new->fa_advice = advice;
4622 			new->fa_start = offset;
4623 			new->fa_end = end;
4624 			fp->f_advice = new;
4625 			new = fa;
4626 		}
4627 		mtx_pool_unlock(mtxpool_sleep, fp);
4628 		break;
4629 	case POSIX_FADV_NORMAL:
4630 		/*
4631 		 * If a the "normal" region overlaps with an existing
4632 		 * non-standard region, trim or remove the
4633 		 * non-standard region.
4634 		 */
4635 		mtx_pool_lock(mtxpool_sleep, fp);
4636 		fa = fp->f_advice;
4637 		if (fa != NULL) {
4638 			if (offset <= fa->fa_start && end >= fa->fa_end) {
4639 				new = fa;
4640 				fp->f_advice = NULL;
4641 			} else if (offset <= fa->fa_start &&
4642 			    end >= fa->fa_start)
4643 				fa->fa_start = end + 1;
4644 			else if (offset <= fa->fa_end && end >= fa->fa_end)
4645 				fa->fa_end = offset - 1;
4646 			else if (offset >= fa->fa_start && end <= fa->fa_end) {
4647 				/*
4648 				 * If the "normal" region is a middle
4649 				 * portion of the existing
4650 				 * non-standard region, just remove
4651 				 * the whole thing rather than picking
4652 				 * one side or the other to
4653 				 * preserve.
4654 				 */
4655 				new = fa;
4656 				fp->f_advice = NULL;
4657 			}
4658 		}
4659 		mtx_pool_unlock(mtxpool_sleep, fp);
4660 		break;
4661 	case POSIX_FADV_WILLNEED:
4662 	case POSIX_FADV_DONTNEED:
4663 		error = VOP_ADVISE(vp, offset, end, advice);
4664 		break;
4665 	}
4666 out:
4667 	if (fp != NULL)
4668 		fdrop(fp, td);
4669 	free(new, M_FADVISE);
4670 	return (error);
4671 }
4672 
4673 int
4674 sys_posix_fadvise(struct thread *td, struct posix_fadvise_args *uap)
4675 {
4676 	int error;
4677 
4678 	error = kern_posix_fadvise(td, uap->fd, uap->offset, uap->len,
4679 	    uap->advice);
4680 	return (kern_posix_error(td, error));
4681 }
4682