xref: /freebsd/sys/kern/vfs_syscalls.c (revision a4e4b355f45538a9b9550df744ca43787fd43c93)
1 /*-
2  * Copyright (c) 1989, 1993
3  *	The Regents of the University of California.  All rights reserved.
4  * (c) UNIX System Laboratories, Inc.
5  * All or some portions of this file are derived from material licensed
6  * to the University of California by American Telephone and Telegraph
7  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
8  * the permission of UNIX System Laboratories, Inc.
9  *
10  * Redistribution and use in source and binary forms, with or without
11  * modification, are permitted provided that the following conditions
12  * are met:
13  * 1. Redistributions of source code must retain the above copyright
14  *    notice, this list of conditions and the following disclaimer.
15  * 2. Redistributions in binary form must reproduce the above copyright
16  *    notice, this list of conditions and the following disclaimer in the
17  *    documentation and/or other materials provided with the distribution.
18  * 4. Neither the name of the University nor the names of its contributors
19  *    may be used to endorse or promote products derived from this software
20  *    without specific prior written permission.
21  *
22  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
23  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
24  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
25  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
26  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
27  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
28  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
29  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
30  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
31  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32  * SUCH DAMAGE.
33  *
34  *	@(#)vfs_syscalls.c	8.13 (Berkeley) 4/15/94
35  */
36 
37 #include <sys/cdefs.h>
38 __FBSDID("$FreeBSD$");
39 
40 #include "opt_capsicum.h"
41 #include "opt_compat.h"
42 #include "opt_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 chroot_refuse_vdir_fds(struct filedesc *fdp);
98 static int getutimes(const struct timeval *, enum uio_seg, struct timespec *);
99 static int kern_chflags(struct thread *td, const char *path,
100     enum uio_seg pathseg, u_long flags);
101 static int kern_chflagsat(struct thread *td, int fd, const char *path,
102     enum uio_seg pathseg, u_long flags, int atflag);
103 static int setfflags(struct thread *td, struct vnode *, u_long);
104 static int setutimes(struct thread *td, struct vnode *,
105     const struct timespec *, int, int);
106 static int vn_access(struct vnode *vp, int user_flags, struct ucred *cred,
107     struct thread *td);
108 
109 /*
110  * The module initialization routine for POSIX asynchronous I/O will
111  * set this to the version of AIO that it implements.  (Zero means
112  * that it is not implemented.)  This value is used here by pathconf()
113  * and in kern_descrip.c by fpathconf().
114  */
115 int async_io_version;
116 
117 #ifdef DEBUG
118 static int syncprt = 0;
119 SYSCTL_INT(_debug, OID_AUTO, syncprt, CTLFLAG_RW, &syncprt, 0, "");
120 #endif
121 
122 /*
123  * Sync each mounted filesystem.
124  */
125 #ifndef _SYS_SYSPROTO_H_
126 struct sync_args {
127 	int     dummy;
128 };
129 #endif
130 /* ARGSUSED */
131 int
132 sys_sync(td, uap)
133 	struct thread *td;
134 	struct sync_args *uap;
135 {
136 	struct mount *mp, *nmp;
137 	int save;
138 
139 	mtx_lock(&mountlist_mtx);
140 	for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
141 		if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK)) {
142 			nmp = TAILQ_NEXT(mp, mnt_list);
143 			continue;
144 		}
145 		if ((mp->mnt_flag & MNT_RDONLY) == 0 &&
146 		    vn_start_write(NULL, &mp, V_NOWAIT) == 0) {
147 			save = curthread_pflags_set(TDP_SYNCIO);
148 			vfs_msync(mp, MNT_NOWAIT);
149 			VFS_SYNC(mp, MNT_NOWAIT);
150 			curthread_pflags_restore(save);
151 			vn_finished_write(mp);
152 		}
153 		mtx_lock(&mountlist_mtx);
154 		nmp = TAILQ_NEXT(mp, mnt_list);
155 		vfs_unbusy(mp);
156 	}
157 	mtx_unlock(&mountlist_mtx);
158 	return (0);
159 }
160 
161 /*
162  * Change filesystem quotas.
163  */
164 #ifndef _SYS_SYSPROTO_H_
165 struct quotactl_args {
166 	char *path;
167 	int cmd;
168 	int uid;
169 	caddr_t arg;
170 };
171 #endif
172 int
173 sys_quotactl(td, uap)
174 	struct thread *td;
175 	register struct quotactl_args /* {
176 		char *path;
177 		int cmd;
178 		int uid;
179 		caddr_t arg;
180 	} */ *uap;
181 {
182 	struct mount *mp;
183 	struct nameidata nd;
184 	int error;
185 
186 	AUDIT_ARG_CMD(uap->cmd);
187 	AUDIT_ARG_UID(uap->uid);
188 	if (!prison_allow(td->td_ucred, PR_ALLOW_QUOTAS))
189 		return (EPERM);
190 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
191 	    uap->path, td);
192 	if ((error = namei(&nd)) != 0)
193 		return (error);
194 	NDFREE(&nd, NDF_ONLY_PNBUF);
195 	mp = nd.ni_vp->v_mount;
196 	vfs_ref(mp);
197 	vput(nd.ni_vp);
198 	error = vfs_busy(mp, 0);
199 	vfs_rel(mp);
200 	if (error != 0)
201 		return (error);
202 	error = VFS_QUOTACTL(mp, uap->cmd, uap->uid, uap->arg);
203 
204 	/*
205 	 * Since quota on operation typically needs to open quota
206 	 * file, the Q_QUOTAON handler needs to unbusy the mount point
207 	 * before calling into namei.  Otherwise, unmount might be
208 	 * started between two vfs_busy() invocations (first is our,
209 	 * second is from mount point cross-walk code in lookup()),
210 	 * causing deadlock.
211 	 *
212 	 * Require that Q_QUOTAON handles the vfs_busy() reference on
213 	 * its own, always returning with ubusied mount point.
214 	 */
215 	if ((uap->cmd >> SUBCMDSHIFT) != Q_QUOTAON)
216 		vfs_unbusy(mp);
217 	return (error);
218 }
219 
220 /*
221  * Used by statfs conversion routines to scale the block size up if
222  * necessary so that all of the block counts are <= 'max_size'.  Note
223  * that 'max_size' should be a bitmask, i.e. 2^n - 1 for some non-zero
224  * value of 'n'.
225  */
226 void
227 statfs_scale_blocks(struct statfs *sf, long max_size)
228 {
229 	uint64_t count;
230 	int shift;
231 
232 	KASSERT(powerof2(max_size + 1), ("%s: invalid max_size", __func__));
233 
234 	/*
235 	 * Attempt to scale the block counts to give a more accurate
236 	 * overview to userland of the ratio of free space to used
237 	 * space.  To do this, find the largest block count and compute
238 	 * a divisor that lets it fit into a signed integer <= max_size.
239 	 */
240 	if (sf->f_bavail < 0)
241 		count = -sf->f_bavail;
242 	else
243 		count = sf->f_bavail;
244 	count = MAX(sf->f_blocks, MAX(sf->f_bfree, count));
245 	if (count <= max_size)
246 		return;
247 
248 	count >>= flsl(max_size);
249 	shift = 0;
250 	while (count > 0) {
251 		shift++;
252 		count >>=1;
253 	}
254 
255 	sf->f_bsize <<= shift;
256 	sf->f_blocks >>= shift;
257 	sf->f_bfree >>= shift;
258 	sf->f_bavail >>= shift;
259 }
260 
261 /*
262  * Get filesystem statistics.
263  */
264 #ifndef _SYS_SYSPROTO_H_
265 struct statfs_args {
266 	char *path;
267 	struct statfs *buf;
268 };
269 #endif
270 int
271 sys_statfs(td, uap)
272 	struct thread *td;
273 	register struct statfs_args /* {
274 		char *path;
275 		struct statfs *buf;
276 	} */ *uap;
277 {
278 	struct statfs sf;
279 	int error;
280 
281 	error = kern_statfs(td, uap->path, UIO_USERSPACE, &sf);
282 	if (error == 0)
283 		error = copyout(&sf, uap->buf, sizeof(sf));
284 	return (error);
285 }
286 
287 int
288 kern_statfs(struct thread *td, char *path, enum uio_seg pathseg,
289     struct statfs *buf)
290 {
291 	struct mount *mp;
292 	struct statfs *sp, sb;
293 	struct nameidata nd;
294 	int error;
295 
296 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
297 	    pathseg, path, td);
298 	error = namei(&nd);
299 	if (error != 0)
300 		return (error);
301 	mp = nd.ni_vp->v_mount;
302 	vfs_ref(mp);
303 	NDFREE(&nd, NDF_ONLY_PNBUF);
304 	vput(nd.ni_vp);
305 	error = vfs_busy(mp, 0);
306 	vfs_rel(mp);
307 	if (error != 0)
308 		return (error);
309 #ifdef MAC
310 	error = mac_mount_check_stat(td->td_ucred, mp);
311 	if (error != 0)
312 		goto out;
313 #endif
314 	/*
315 	 * Set these in case the underlying filesystem fails to do so.
316 	 */
317 	sp = &mp->mnt_stat;
318 	sp->f_version = STATFS_VERSION;
319 	sp->f_namemax = NAME_MAX;
320 	sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
321 	error = VFS_STATFS(mp, sp);
322 	if (error != 0)
323 		goto out;
324 	if (priv_check(td, PRIV_VFS_GENERATION)) {
325 		bcopy(sp, &sb, sizeof(sb));
326 		sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
327 		prison_enforce_statfs(td->td_ucred, mp, &sb);
328 		sp = &sb;
329 	}
330 	*buf = *sp;
331 out:
332 	vfs_unbusy(mp);
333 	return (error);
334 }
335 
336 /*
337  * Get filesystem statistics.
338  */
339 #ifndef _SYS_SYSPROTO_H_
340 struct fstatfs_args {
341 	int fd;
342 	struct statfs *buf;
343 };
344 #endif
345 int
346 sys_fstatfs(td, uap)
347 	struct thread *td;
348 	register struct fstatfs_args /* {
349 		int fd;
350 		struct statfs *buf;
351 	} */ *uap;
352 {
353 	struct statfs sf;
354 	int error;
355 
356 	error = kern_fstatfs(td, uap->fd, &sf);
357 	if (error == 0)
358 		error = copyout(&sf, uap->buf, sizeof(sf));
359 	return (error);
360 }
361 
362 int
363 kern_fstatfs(struct thread *td, int fd, struct statfs *buf)
364 {
365 	struct file *fp;
366 	struct mount *mp;
367 	struct statfs *sp, sb;
368 	struct vnode *vp;
369 	cap_rights_t rights;
370 	int error;
371 
372 	AUDIT_ARG_FD(fd);
373 	error = getvnode(td->td_proc->p_fd, fd,
374 	    cap_rights_init(&rights, CAP_FSTATFS), &fp);
375 	if (error != 0)
376 		return (error);
377 	vp = fp->f_vnode;
378 	vn_lock(vp, LK_SHARED | LK_RETRY);
379 #ifdef AUDIT
380 	AUDIT_ARG_VNODE1(vp);
381 #endif
382 	mp = vp->v_mount;
383 	if (mp)
384 		vfs_ref(mp);
385 	VOP_UNLOCK(vp, 0);
386 	fdrop(fp, td);
387 	if (mp == NULL) {
388 		error = EBADF;
389 		goto out;
390 	}
391 	error = vfs_busy(mp, 0);
392 	vfs_rel(mp);
393 	if (error != 0)
394 		return (error);
395 #ifdef MAC
396 	error = mac_mount_check_stat(td->td_ucred, mp);
397 	if (error != 0)
398 		goto out;
399 #endif
400 	/*
401 	 * Set these in case the underlying filesystem fails to do so.
402 	 */
403 	sp = &mp->mnt_stat;
404 	sp->f_version = STATFS_VERSION;
405 	sp->f_namemax = NAME_MAX;
406 	sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
407 	error = VFS_STATFS(mp, sp);
408 	if (error != 0)
409 		goto out;
410 	if (priv_check(td, PRIV_VFS_GENERATION)) {
411 		bcopy(sp, &sb, sizeof(sb));
412 		sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
413 		prison_enforce_statfs(td->td_ucred, mp, &sb);
414 		sp = &sb;
415 	}
416 	*buf = *sp;
417 out:
418 	if (mp)
419 		vfs_unbusy(mp);
420 	return (error);
421 }
422 
423 /*
424  * Get statistics on all filesystems.
425  */
426 #ifndef _SYS_SYSPROTO_H_
427 struct getfsstat_args {
428 	struct statfs *buf;
429 	long bufsize;
430 	int flags;
431 };
432 #endif
433 int
434 sys_getfsstat(td, uap)
435 	struct thread *td;
436 	register struct getfsstat_args /* {
437 		struct statfs *buf;
438 		long bufsize;
439 		int flags;
440 	} */ *uap;
441 {
442 
443 	return (kern_getfsstat(td, &uap->buf, uap->bufsize, UIO_USERSPACE,
444 	    uap->flags));
445 }
446 
447 /*
448  * If (bufsize > 0 && bufseg == UIO_SYSSPACE)
449  *	The caller is responsible for freeing memory which will be allocated
450  *	in '*buf'.
451  */
452 int
453 kern_getfsstat(struct thread *td, struct statfs **buf, size_t bufsize,
454     enum uio_seg bufseg, int flags)
455 {
456 	struct mount *mp, *nmp;
457 	struct statfs *sfsp, *sp, sb;
458 	size_t count, maxcount;
459 	int error;
460 
461 	maxcount = bufsize / sizeof(struct statfs);
462 	if (bufsize == 0)
463 		sfsp = NULL;
464 	else if (bufseg == UIO_USERSPACE)
465 		sfsp = *buf;
466 	else /* if (bufseg == UIO_SYSSPACE) */ {
467 		count = 0;
468 		mtx_lock(&mountlist_mtx);
469 		TAILQ_FOREACH(mp, &mountlist, mnt_list) {
470 			count++;
471 		}
472 		mtx_unlock(&mountlist_mtx);
473 		if (maxcount > count)
474 			maxcount = count;
475 		sfsp = *buf = malloc(maxcount * sizeof(struct statfs), M_TEMP,
476 		    M_WAITOK);
477 	}
478 	count = 0;
479 	mtx_lock(&mountlist_mtx);
480 	for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
481 		if (prison_canseemount(td->td_ucred, mp) != 0) {
482 			nmp = TAILQ_NEXT(mp, mnt_list);
483 			continue;
484 		}
485 #ifdef MAC
486 		if (mac_mount_check_stat(td->td_ucred, mp) != 0) {
487 			nmp = TAILQ_NEXT(mp, mnt_list);
488 			continue;
489 		}
490 #endif
491 		if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK)) {
492 			nmp = TAILQ_NEXT(mp, mnt_list);
493 			continue;
494 		}
495 		if (sfsp && count < maxcount) {
496 			sp = &mp->mnt_stat;
497 			/*
498 			 * Set these in case the underlying filesystem
499 			 * fails to do so.
500 			 */
501 			sp->f_version = STATFS_VERSION;
502 			sp->f_namemax = NAME_MAX;
503 			sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
504 			/*
505 			 * If MNT_NOWAIT or MNT_LAZY is specified, do not
506 			 * refresh the fsstat cache. MNT_NOWAIT or MNT_LAZY
507 			 * overrides MNT_WAIT.
508 			 */
509 			if (((flags & (MNT_LAZY|MNT_NOWAIT)) == 0 ||
510 			    (flags & MNT_WAIT)) &&
511 			    (error = VFS_STATFS(mp, sp))) {
512 				mtx_lock(&mountlist_mtx);
513 				nmp = TAILQ_NEXT(mp, mnt_list);
514 				vfs_unbusy(mp);
515 				continue;
516 			}
517 			if (priv_check(td, PRIV_VFS_GENERATION)) {
518 				bcopy(sp, &sb, sizeof(sb));
519 				sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
520 				prison_enforce_statfs(td->td_ucred, mp, &sb);
521 				sp = &sb;
522 			}
523 			if (bufseg == UIO_SYSSPACE)
524 				bcopy(sp, sfsp, sizeof(*sp));
525 			else /* if (bufseg == UIO_USERSPACE) */ {
526 				error = copyout(sp, sfsp, sizeof(*sp));
527 				if (error != 0) {
528 					vfs_unbusy(mp);
529 					return (error);
530 				}
531 			}
532 			sfsp++;
533 		}
534 		count++;
535 		mtx_lock(&mountlist_mtx);
536 		nmp = TAILQ_NEXT(mp, mnt_list);
537 		vfs_unbusy(mp);
538 	}
539 	mtx_unlock(&mountlist_mtx);
540 	if (sfsp && count > maxcount)
541 		td->td_retval[0] = maxcount;
542 	else
543 		td->td_retval[0] = count;
544 	return (0);
545 }
546 
547 #ifdef COMPAT_FREEBSD4
548 /*
549  * Get old format filesystem statistics.
550  */
551 static void cvtstatfs(struct statfs *, struct ostatfs *);
552 
553 #ifndef _SYS_SYSPROTO_H_
554 struct freebsd4_statfs_args {
555 	char *path;
556 	struct ostatfs *buf;
557 };
558 #endif
559 int
560 freebsd4_statfs(td, uap)
561 	struct thread *td;
562 	struct freebsd4_statfs_args /* {
563 		char *path;
564 		struct ostatfs *buf;
565 	} */ *uap;
566 {
567 	struct ostatfs osb;
568 	struct statfs sf;
569 	int error;
570 
571 	error = kern_statfs(td, uap->path, UIO_USERSPACE, &sf);
572 	if (error != 0)
573 		return (error);
574 	cvtstatfs(&sf, &osb);
575 	return (copyout(&osb, uap->buf, sizeof(osb)));
576 }
577 
578 /*
579  * Get filesystem statistics.
580  */
581 #ifndef _SYS_SYSPROTO_H_
582 struct freebsd4_fstatfs_args {
583 	int fd;
584 	struct ostatfs *buf;
585 };
586 #endif
587 int
588 freebsd4_fstatfs(td, uap)
589 	struct thread *td;
590 	struct freebsd4_fstatfs_args /* {
591 		int fd;
592 		struct ostatfs *buf;
593 	} */ *uap;
594 {
595 	struct ostatfs osb;
596 	struct statfs sf;
597 	int error;
598 
599 	error = kern_fstatfs(td, uap->fd, &sf);
600 	if (error != 0)
601 		return (error);
602 	cvtstatfs(&sf, &osb);
603 	return (copyout(&osb, uap->buf, sizeof(osb)));
604 }
605 
606 /*
607  * Get statistics on all filesystems.
608  */
609 #ifndef _SYS_SYSPROTO_H_
610 struct freebsd4_getfsstat_args {
611 	struct ostatfs *buf;
612 	long bufsize;
613 	int flags;
614 };
615 #endif
616 int
617 freebsd4_getfsstat(td, uap)
618 	struct thread *td;
619 	register struct freebsd4_getfsstat_args /* {
620 		struct ostatfs *buf;
621 		long bufsize;
622 		int flags;
623 	} */ *uap;
624 {
625 	struct statfs *buf, *sp;
626 	struct ostatfs osb;
627 	size_t count, size;
628 	int error;
629 
630 	count = uap->bufsize / sizeof(struct ostatfs);
631 	size = count * sizeof(struct statfs);
632 	error = kern_getfsstat(td, &buf, size, UIO_SYSSPACE, uap->flags);
633 	if (size > 0) {
634 		count = td->td_retval[0];
635 		sp = buf;
636 		while (count > 0 && error == 0) {
637 			cvtstatfs(sp, &osb);
638 			error = copyout(&osb, uap->buf, sizeof(osb));
639 			sp++;
640 			uap->buf++;
641 			count--;
642 		}
643 		free(buf, M_TEMP);
644 	}
645 	return (error);
646 }
647 
648 /*
649  * Implement fstatfs() for (NFS) file handles.
650  */
651 #ifndef _SYS_SYSPROTO_H_
652 struct freebsd4_fhstatfs_args {
653 	struct fhandle *u_fhp;
654 	struct ostatfs *buf;
655 };
656 #endif
657 int
658 freebsd4_fhstatfs(td, uap)
659 	struct thread *td;
660 	struct freebsd4_fhstatfs_args /* {
661 		struct fhandle *u_fhp;
662 		struct ostatfs *buf;
663 	} */ *uap;
664 {
665 	struct ostatfs osb;
666 	struct statfs sf;
667 	fhandle_t fh;
668 	int error;
669 
670 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
671 	if (error != 0)
672 		return (error);
673 	error = kern_fhstatfs(td, fh, &sf);
674 	if (error != 0)
675 		return (error);
676 	cvtstatfs(&sf, &osb);
677 	return (copyout(&osb, uap->buf, sizeof(osb)));
678 }
679 
680 /*
681  * Convert a new format statfs structure to an old format statfs structure.
682  */
683 static void
684 cvtstatfs(nsp, osp)
685 	struct statfs *nsp;
686 	struct ostatfs *osp;
687 {
688 
689 	statfs_scale_blocks(nsp, LONG_MAX);
690 	bzero(osp, sizeof(*osp));
691 	osp->f_bsize = nsp->f_bsize;
692 	osp->f_iosize = MIN(nsp->f_iosize, LONG_MAX);
693 	osp->f_blocks = nsp->f_blocks;
694 	osp->f_bfree = nsp->f_bfree;
695 	osp->f_bavail = nsp->f_bavail;
696 	osp->f_files = MIN(nsp->f_files, LONG_MAX);
697 	osp->f_ffree = MIN(nsp->f_ffree, LONG_MAX);
698 	osp->f_owner = nsp->f_owner;
699 	osp->f_type = nsp->f_type;
700 	osp->f_flags = nsp->f_flags;
701 	osp->f_syncwrites = MIN(nsp->f_syncwrites, LONG_MAX);
702 	osp->f_asyncwrites = MIN(nsp->f_asyncwrites, LONG_MAX);
703 	osp->f_syncreads = MIN(nsp->f_syncreads, LONG_MAX);
704 	osp->f_asyncreads = MIN(nsp->f_asyncreads, LONG_MAX);
705 	strlcpy(osp->f_fstypename, nsp->f_fstypename,
706 	    MIN(MFSNAMELEN, OMFSNAMELEN));
707 	strlcpy(osp->f_mntonname, nsp->f_mntonname,
708 	    MIN(MNAMELEN, OMNAMELEN));
709 	strlcpy(osp->f_mntfromname, nsp->f_mntfromname,
710 	    MIN(MNAMELEN, OMNAMELEN));
711 	osp->f_fsid = nsp->f_fsid;
712 }
713 #endif /* COMPAT_FREEBSD4 */
714 
715 /*
716  * Change current working directory to a given file descriptor.
717  */
718 #ifndef _SYS_SYSPROTO_H_
719 struct fchdir_args {
720 	int	fd;
721 };
722 #endif
723 int
724 sys_fchdir(td, uap)
725 	struct thread *td;
726 	struct fchdir_args /* {
727 		int fd;
728 	} */ *uap;
729 {
730 	register struct filedesc *fdp = td->td_proc->p_fd;
731 	struct vnode *vp, *tdp, *vpold;
732 	struct mount *mp;
733 	struct file *fp;
734 	cap_rights_t rights;
735 	int error;
736 
737 	AUDIT_ARG_FD(uap->fd);
738 	error = getvnode(fdp, uap->fd, cap_rights_init(&rights, CAP_FCHDIR),
739 	    &fp);
740 	if (error != 0)
741 		return (error);
742 	vp = fp->f_vnode;
743 	VREF(vp);
744 	fdrop(fp, td);
745 	vn_lock(vp, LK_SHARED | LK_RETRY);
746 	AUDIT_ARG_VNODE1(vp);
747 	error = change_dir(vp, td);
748 	while (!error && (mp = vp->v_mountedhere) != NULL) {
749 		if (vfs_busy(mp, 0))
750 			continue;
751 		error = VFS_ROOT(mp, LK_SHARED, &tdp);
752 		vfs_unbusy(mp);
753 		if (error != 0)
754 			break;
755 		vput(vp);
756 		vp = tdp;
757 	}
758 	if (error != 0) {
759 		vput(vp);
760 		return (error);
761 	}
762 	VOP_UNLOCK(vp, 0);
763 	FILEDESC_XLOCK(fdp);
764 	vpold = fdp->fd_cdir;
765 	fdp->fd_cdir = vp;
766 	FILEDESC_XUNLOCK(fdp);
767 	vrele(vpold);
768 	return (0);
769 }
770 
771 /*
772  * Change current working directory (``.'').
773  */
774 #ifndef _SYS_SYSPROTO_H_
775 struct chdir_args {
776 	char	*path;
777 };
778 #endif
779 int
780 sys_chdir(td, uap)
781 	struct thread *td;
782 	struct chdir_args /* {
783 		char *path;
784 	} */ *uap;
785 {
786 
787 	return (kern_chdir(td, uap->path, UIO_USERSPACE));
788 }
789 
790 int
791 kern_chdir(struct thread *td, char *path, enum uio_seg pathseg)
792 {
793 	register struct filedesc *fdp = td->td_proc->p_fd;
794 	struct nameidata nd;
795 	struct vnode *vp;
796 	int error;
797 
798 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
799 	    pathseg, path, td);
800 	if ((error = namei(&nd)) != 0)
801 		return (error);
802 	if ((error = change_dir(nd.ni_vp, td)) != 0) {
803 		vput(nd.ni_vp);
804 		NDFREE(&nd, NDF_ONLY_PNBUF);
805 		return (error);
806 	}
807 	VOP_UNLOCK(nd.ni_vp, 0);
808 	NDFREE(&nd, NDF_ONLY_PNBUF);
809 	FILEDESC_XLOCK(fdp);
810 	vp = fdp->fd_cdir;
811 	fdp->fd_cdir = nd.ni_vp;
812 	FILEDESC_XUNLOCK(fdp);
813 	vrele(vp);
814 	return (0);
815 }
816 
817 /*
818  * Helper function for raised chroot(2) security function:  Refuse if
819  * any filedescriptors are open directories.
820  */
821 static int
822 chroot_refuse_vdir_fds(fdp)
823 	struct filedesc *fdp;
824 {
825 	struct vnode *vp;
826 	struct file *fp;
827 	int fd;
828 
829 	FILEDESC_LOCK_ASSERT(fdp);
830 
831 	for (fd = 0; fd <= fdp->fd_lastfile; fd++) {
832 		fp = fget_locked(fdp, fd);
833 		if (fp == NULL)
834 			continue;
835 		if (fp->f_type == DTYPE_VNODE) {
836 			vp = fp->f_vnode;
837 			if (vp->v_type == VDIR)
838 				return (EPERM);
839 		}
840 	}
841 	return (0);
842 }
843 
844 /*
845  * This sysctl determines if we will allow a process to chroot(2) if it
846  * has a directory open:
847  *	0: disallowed for all processes.
848  *	1: allowed for processes that were not already chroot(2)'ed.
849  *	2: allowed for all processes.
850  */
851 
852 static int chroot_allow_open_directories = 1;
853 
854 SYSCTL_INT(_kern, OID_AUTO, chroot_allow_open_directories, CTLFLAG_RW,
855      &chroot_allow_open_directories, 0,
856      "Allow a process to chroot(2) if it has a directory open");
857 
858 /*
859  * Change notion of root (``/'') directory.
860  */
861 #ifndef _SYS_SYSPROTO_H_
862 struct chroot_args {
863 	char	*path;
864 };
865 #endif
866 int
867 sys_chroot(td, uap)
868 	struct thread *td;
869 	struct chroot_args /* {
870 		char *path;
871 	} */ *uap;
872 {
873 	struct nameidata nd;
874 	int error;
875 
876 	error = priv_check(td, PRIV_VFS_CHROOT);
877 	if (error != 0)
878 		return (error);
879 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
880 	    UIO_USERSPACE, uap->path, td);
881 	error = namei(&nd);
882 	if (error != 0)
883 		goto error;
884 	error = change_dir(nd.ni_vp, td);
885 	if (error != 0)
886 		goto e_vunlock;
887 #ifdef MAC
888 	error = mac_vnode_check_chroot(td->td_ucred, nd.ni_vp);
889 	if (error != 0)
890 		goto e_vunlock;
891 #endif
892 	VOP_UNLOCK(nd.ni_vp, 0);
893 	error = change_root(nd.ni_vp, td);
894 	vrele(nd.ni_vp);
895 	NDFREE(&nd, NDF_ONLY_PNBUF);
896 	return (error);
897 e_vunlock:
898 	vput(nd.ni_vp);
899 error:
900 	NDFREE(&nd, NDF_ONLY_PNBUF);
901 	return (error);
902 }
903 
904 /*
905  * Common routine for chroot and chdir.  Callers must provide a locked vnode
906  * instance.
907  */
908 int
909 change_dir(vp, td)
910 	struct vnode *vp;
911 	struct thread *td;
912 {
913 #ifdef MAC
914 	int error;
915 #endif
916 
917 	ASSERT_VOP_LOCKED(vp, "change_dir(): vp not locked");
918 	if (vp->v_type != VDIR)
919 		return (ENOTDIR);
920 #ifdef MAC
921 	error = mac_vnode_check_chdir(td->td_ucred, vp);
922 	if (error != 0)
923 		return (error);
924 #endif
925 	return (VOP_ACCESS(vp, VEXEC, td->td_ucred, td));
926 }
927 
928 /*
929  * Common routine for kern_chroot() and jail_attach().  The caller is
930  * responsible for invoking priv_check() and mac_vnode_check_chroot() to
931  * authorize this operation.
932  */
933 int
934 change_root(vp, td)
935 	struct vnode *vp;
936 	struct thread *td;
937 {
938 	struct filedesc *fdp;
939 	struct vnode *oldvp;
940 	int error;
941 
942 	fdp = td->td_proc->p_fd;
943 	FILEDESC_XLOCK(fdp);
944 	if (chroot_allow_open_directories == 0 ||
945 	    (chroot_allow_open_directories == 1 && fdp->fd_rdir != rootvnode)) {
946 		error = chroot_refuse_vdir_fds(fdp);
947 		if (error != 0) {
948 			FILEDESC_XUNLOCK(fdp);
949 			return (error);
950 		}
951 	}
952 	oldvp = fdp->fd_rdir;
953 	fdp->fd_rdir = vp;
954 	VREF(fdp->fd_rdir);
955 	if (!fdp->fd_jdir) {
956 		fdp->fd_jdir = vp;
957 		VREF(fdp->fd_jdir);
958 	}
959 	FILEDESC_XUNLOCK(fdp);
960 	vrele(oldvp);
961 	return (0);
962 }
963 
964 static __inline void
965 flags_to_rights(int flags, cap_rights_t *rightsp)
966 {
967 
968 	if (flags & O_EXEC) {
969 		cap_rights_set(rightsp, CAP_FEXECVE);
970 	} else {
971 		switch ((flags & O_ACCMODE)) {
972 		case O_RDONLY:
973 			cap_rights_set(rightsp, CAP_READ);
974 			break;
975 		case O_RDWR:
976 			cap_rights_set(rightsp, CAP_READ);
977 			/* FALLTHROUGH */
978 		case O_WRONLY:
979 			cap_rights_set(rightsp, CAP_WRITE);
980 			if (!(flags & (O_APPEND | O_TRUNC)))
981 				cap_rights_set(rightsp, CAP_SEEK);
982 			break;
983 		}
984 	}
985 
986 	if (flags & O_CREAT)
987 		cap_rights_set(rightsp, CAP_CREATE);
988 
989 	if (flags & O_TRUNC)
990 		cap_rights_set(rightsp, CAP_FTRUNCATE);
991 
992 	if (flags & (O_SYNC | O_FSYNC))
993 		cap_rights_set(rightsp, CAP_FSYNC);
994 
995 	if (flags & (O_EXLOCK | O_SHLOCK))
996 		cap_rights_set(rightsp, CAP_FLOCK);
997 }
998 
999 /*
1000  * Check permissions, allocate an open file structure, and call the device
1001  * open routine if any.
1002  */
1003 #ifndef _SYS_SYSPROTO_H_
1004 struct open_args {
1005 	char	*path;
1006 	int	flags;
1007 	int	mode;
1008 };
1009 #endif
1010 int
1011 sys_open(td, uap)
1012 	struct thread *td;
1013 	register struct open_args /* {
1014 		char *path;
1015 		int flags;
1016 		int mode;
1017 	} */ *uap;
1018 {
1019 
1020 	return (kern_open(td, uap->path, UIO_USERSPACE, uap->flags, uap->mode));
1021 }
1022 
1023 #ifndef _SYS_SYSPROTO_H_
1024 struct openat_args {
1025 	int	fd;
1026 	char	*path;
1027 	int	flag;
1028 	int	mode;
1029 };
1030 #endif
1031 int
1032 sys_openat(struct thread *td, struct openat_args *uap)
1033 {
1034 
1035 	return (kern_openat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
1036 	    uap->mode));
1037 }
1038 
1039 int
1040 kern_open(struct thread *td, char *path, enum uio_seg pathseg, int flags,
1041     int mode)
1042 {
1043 
1044 	return (kern_openat(td, AT_FDCWD, path, pathseg, flags, mode));
1045 }
1046 
1047 int
1048 kern_openat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1049     int flags, int mode)
1050 {
1051 	struct proc *p = td->td_proc;
1052 	struct filedesc *fdp = p->p_fd;
1053 	struct file *fp;
1054 	struct vnode *vp;
1055 	struct nameidata nd;
1056 	cap_rights_t rights;
1057 	int cmode, error, indx;
1058 
1059 	indx = -1;
1060 
1061 	AUDIT_ARG_FFLAGS(flags);
1062 	AUDIT_ARG_MODE(mode);
1063 	/* XXX: audit dirfd */
1064 	cap_rights_init(&rights, CAP_LOOKUP);
1065 	flags_to_rights(flags, &rights);
1066 	/*
1067 	 * Only one of the O_EXEC, O_RDONLY, O_WRONLY and O_RDWR flags
1068 	 * may be specified.
1069 	 */
1070 	if (flags & O_EXEC) {
1071 		if (flags & O_ACCMODE)
1072 			return (EINVAL);
1073 	} else if ((flags & O_ACCMODE) == O_ACCMODE) {
1074 		return (EINVAL);
1075 	} else {
1076 		flags = FFLAGS(flags);
1077 	}
1078 
1079 	/*
1080 	 * Allocate the file descriptor, but don't install a descriptor yet.
1081 	 */
1082 	error = falloc_noinstall(td, &fp);
1083 	if (error != 0)
1084 		return (error);
1085 	/*
1086 	 * An extra reference on `fp' has been held for us by
1087 	 * falloc_noinstall().
1088 	 */
1089 	/* Set the flags early so the finit in devfs can pick them up. */
1090 	fp->f_flag = flags & FMASK;
1091 	cmode = ((mode & ~fdp->fd_cmask) & ALLPERMS) & ~S_ISTXT;
1092 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, fd,
1093 	    &rights, td);
1094 	td->td_dupfd = -1;		/* XXX check for fdopen */
1095 	error = vn_open(&nd, &flags, cmode, fp);
1096 	if (error != 0) {
1097 		/*
1098 		 * If the vn_open replaced the method vector, something
1099 		 * wonderous happened deep below and we just pass it up
1100 		 * pretending we know what we do.
1101 		 */
1102 		if (error == ENXIO && fp->f_ops != &badfileops)
1103 			goto success;
1104 
1105 		/*
1106 		 * Handle special fdopen() case. bleh.
1107 		 *
1108 		 * Don't do this for relative (capability) lookups; we don't
1109 		 * understand exactly what would happen, and we don't think
1110 		 * that it ever should.
1111 		 */
1112 		if (nd.ni_strictrelative == 0 &&
1113 		    (error == ENODEV || error == ENXIO) &&
1114 		    td->td_dupfd >= 0) {
1115 			error = dupfdopen(td, fdp, td->td_dupfd, flags, error,
1116 			    &indx);
1117 			if (error == 0)
1118 				goto success;
1119 		}
1120 
1121 		goto bad;
1122 	}
1123 	td->td_dupfd = 0;
1124 	NDFREE(&nd, NDF_ONLY_PNBUF);
1125 	vp = nd.ni_vp;
1126 
1127 	/*
1128 	 * Store the vnode, for any f_type. Typically, the vnode use
1129 	 * count is decremented by direct call to vn_closefile() for
1130 	 * files that switched type in the cdevsw fdopen() method.
1131 	 */
1132 	fp->f_vnode = vp;
1133 	/*
1134 	 * If the file wasn't claimed by devfs bind it to the normal
1135 	 * vnode operations here.
1136 	 */
1137 	if (fp->f_ops == &badfileops) {
1138 		KASSERT(vp->v_type != VFIFO, ("Unexpected fifo."));
1139 		fp->f_seqcount = 1;
1140 		finit(fp, (flags & FMASK) | (fp->f_flag & FHASLOCK),
1141 		    DTYPE_VNODE, vp, &vnops);
1142 	}
1143 
1144 	VOP_UNLOCK(vp, 0);
1145 	if (flags & O_TRUNC) {
1146 		error = fo_truncate(fp, 0, td->td_ucred, td);
1147 		if (error != 0)
1148 			goto bad;
1149 	}
1150 success:
1151 	/*
1152 	 * If we haven't already installed the FD (for dupfdopen), do so now.
1153 	 */
1154 	if (indx == -1) {
1155 		struct filecaps *fcaps;
1156 
1157 #ifdef CAPABILITIES
1158 		if (nd.ni_strictrelative == 1)
1159 			fcaps = &nd.ni_filecaps;
1160 		else
1161 #endif
1162 			fcaps = NULL;
1163 		error = finstall(td, fp, &indx, flags, fcaps);
1164 		/* On success finstall() consumes fcaps. */
1165 		if (error != 0) {
1166 			filecaps_free(&nd.ni_filecaps);
1167 			goto bad;
1168 		}
1169 	} else {
1170 		filecaps_free(&nd.ni_filecaps);
1171 	}
1172 
1173 	/*
1174 	 * Release our private reference, leaving the one associated with
1175 	 * the descriptor table intact.
1176 	 */
1177 	fdrop(fp, td);
1178 	td->td_retval[0] = indx;
1179 	return (0);
1180 bad:
1181 	KASSERT(indx == -1, ("indx=%d, should be -1", indx));
1182 	fdrop(fp, td);
1183 	return (error);
1184 }
1185 
1186 #ifdef COMPAT_43
1187 /*
1188  * Create a file.
1189  */
1190 #ifndef _SYS_SYSPROTO_H_
1191 struct ocreat_args {
1192 	char	*path;
1193 	int	mode;
1194 };
1195 #endif
1196 int
1197 ocreat(td, uap)
1198 	struct thread *td;
1199 	register struct ocreat_args /* {
1200 		char *path;
1201 		int mode;
1202 	} */ *uap;
1203 {
1204 
1205 	return (kern_open(td, uap->path, UIO_USERSPACE,
1206 	    O_WRONLY | O_CREAT | O_TRUNC, uap->mode));
1207 }
1208 #endif /* COMPAT_43 */
1209 
1210 /*
1211  * Create a special file.
1212  */
1213 #ifndef _SYS_SYSPROTO_H_
1214 struct mknod_args {
1215 	char	*path;
1216 	int	mode;
1217 	int	dev;
1218 };
1219 #endif
1220 int
1221 sys_mknod(td, uap)
1222 	struct thread *td;
1223 	register struct mknod_args /* {
1224 		char *path;
1225 		int mode;
1226 		int dev;
1227 	} */ *uap;
1228 {
1229 
1230 	return (kern_mknod(td, uap->path, UIO_USERSPACE, uap->mode, uap->dev));
1231 }
1232 
1233 #ifndef _SYS_SYSPROTO_H_
1234 struct mknodat_args {
1235 	int	fd;
1236 	char	*path;
1237 	mode_t	mode;
1238 	dev_t	dev;
1239 };
1240 #endif
1241 int
1242 sys_mknodat(struct thread *td, struct mknodat_args *uap)
1243 {
1244 
1245 	return (kern_mknodat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode,
1246 	    uap->dev));
1247 }
1248 
1249 int
1250 kern_mknod(struct thread *td, char *path, enum uio_seg pathseg, int mode,
1251     int dev)
1252 {
1253 
1254 	return (kern_mknodat(td, AT_FDCWD, path, pathseg, mode, dev));
1255 }
1256 
1257 int
1258 kern_mknodat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1259     int mode, int dev)
1260 {
1261 	struct vnode *vp;
1262 	struct mount *mp;
1263 	struct vattr vattr;
1264 	struct nameidata nd;
1265 	cap_rights_t rights;
1266 	int error, whiteout = 0;
1267 
1268 	AUDIT_ARG_MODE(mode);
1269 	AUDIT_ARG_DEV(dev);
1270 	switch (mode & S_IFMT) {
1271 	case S_IFCHR:
1272 	case S_IFBLK:
1273 		error = priv_check(td, PRIV_VFS_MKNOD_DEV);
1274 		break;
1275 	case S_IFMT:
1276 		error = priv_check(td, PRIV_VFS_MKNOD_BAD);
1277 		break;
1278 	case S_IFWHT:
1279 		error = priv_check(td, PRIV_VFS_MKNOD_WHT);
1280 		break;
1281 	case S_IFIFO:
1282 		if (dev == 0)
1283 			return (kern_mkfifoat(td, fd, path, pathseg, mode));
1284 		/* FALLTHROUGH */
1285 	default:
1286 		error = EINVAL;
1287 		break;
1288 	}
1289 	if (error != 0)
1290 		return (error);
1291 restart:
1292 	bwillwrite();
1293 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1,
1294 	    pathseg, path, fd, cap_rights_init(&rights, CAP_MKNODAT), td);
1295 	if ((error = namei(&nd)) != 0)
1296 		return (error);
1297 	vp = nd.ni_vp;
1298 	if (vp != NULL) {
1299 		NDFREE(&nd, NDF_ONLY_PNBUF);
1300 		if (vp == nd.ni_dvp)
1301 			vrele(nd.ni_dvp);
1302 		else
1303 			vput(nd.ni_dvp);
1304 		vrele(vp);
1305 		return (EEXIST);
1306 	} else {
1307 		VATTR_NULL(&vattr);
1308 		vattr.va_mode = (mode & ALLPERMS) &
1309 		    ~td->td_proc->p_fd->fd_cmask;
1310 		vattr.va_rdev = dev;
1311 		whiteout = 0;
1312 
1313 		switch (mode & S_IFMT) {
1314 		case S_IFMT:	/* used by badsect to flag bad sectors */
1315 			vattr.va_type = VBAD;
1316 			break;
1317 		case S_IFCHR:
1318 			vattr.va_type = VCHR;
1319 			break;
1320 		case S_IFBLK:
1321 			vattr.va_type = VBLK;
1322 			break;
1323 		case S_IFWHT:
1324 			whiteout = 1;
1325 			break;
1326 		default:
1327 			panic("kern_mknod: invalid mode");
1328 		}
1329 	}
1330 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1331 		NDFREE(&nd, NDF_ONLY_PNBUF);
1332 		vput(nd.ni_dvp);
1333 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1334 			return (error);
1335 		goto restart;
1336 	}
1337 #ifdef MAC
1338 	if (error == 0 && !whiteout)
1339 		error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp,
1340 		    &nd.ni_cnd, &vattr);
1341 #endif
1342 	if (error == 0) {
1343 		if (whiteout)
1344 			error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE);
1345 		else {
1346 			error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp,
1347 						&nd.ni_cnd, &vattr);
1348 			if (error == 0)
1349 				vput(nd.ni_vp);
1350 		}
1351 	}
1352 	NDFREE(&nd, NDF_ONLY_PNBUF);
1353 	vput(nd.ni_dvp);
1354 	vn_finished_write(mp);
1355 	return (error);
1356 }
1357 
1358 /*
1359  * Create a named pipe.
1360  */
1361 #ifndef _SYS_SYSPROTO_H_
1362 struct mkfifo_args {
1363 	char	*path;
1364 	int	mode;
1365 };
1366 #endif
1367 int
1368 sys_mkfifo(td, uap)
1369 	struct thread *td;
1370 	register struct mkfifo_args /* {
1371 		char *path;
1372 		int mode;
1373 	} */ *uap;
1374 {
1375 
1376 	return (kern_mkfifo(td, uap->path, UIO_USERSPACE, uap->mode));
1377 }
1378 
1379 #ifndef _SYS_SYSPROTO_H_
1380 struct mkfifoat_args {
1381 	int	fd;
1382 	char	*path;
1383 	mode_t	mode;
1384 };
1385 #endif
1386 int
1387 sys_mkfifoat(struct thread *td, struct mkfifoat_args *uap)
1388 {
1389 
1390 	return (kern_mkfifoat(td, uap->fd, uap->path, UIO_USERSPACE,
1391 	    uap->mode));
1392 }
1393 
1394 int
1395 kern_mkfifo(struct thread *td, char *path, enum uio_seg pathseg, int mode)
1396 {
1397 
1398 	return (kern_mkfifoat(td, AT_FDCWD, path, pathseg, mode));
1399 }
1400 
1401 int
1402 kern_mkfifoat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1403     int mode)
1404 {
1405 	struct mount *mp;
1406 	struct vattr vattr;
1407 	struct nameidata nd;
1408 	cap_rights_t rights;
1409 	int error;
1410 
1411 	AUDIT_ARG_MODE(mode);
1412 restart:
1413 	bwillwrite();
1414 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1,
1415 	    pathseg, path, fd, cap_rights_init(&rights, CAP_MKFIFOAT), td);
1416 	if ((error = namei(&nd)) != 0)
1417 		return (error);
1418 	if (nd.ni_vp != NULL) {
1419 		NDFREE(&nd, NDF_ONLY_PNBUF);
1420 		if (nd.ni_vp == nd.ni_dvp)
1421 			vrele(nd.ni_dvp);
1422 		else
1423 			vput(nd.ni_dvp);
1424 		vrele(nd.ni_vp);
1425 		return (EEXIST);
1426 	}
1427 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1428 		NDFREE(&nd, NDF_ONLY_PNBUF);
1429 		vput(nd.ni_dvp);
1430 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1431 			return (error);
1432 		goto restart;
1433 	}
1434 	VATTR_NULL(&vattr);
1435 	vattr.va_type = VFIFO;
1436 	vattr.va_mode = (mode & ALLPERMS) & ~td->td_proc->p_fd->fd_cmask;
1437 #ifdef MAC
1438 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1439 	    &vattr);
1440 	if (error != 0)
1441 		goto out;
1442 #endif
1443 	error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
1444 	if (error == 0)
1445 		vput(nd.ni_vp);
1446 #ifdef MAC
1447 out:
1448 #endif
1449 	vput(nd.ni_dvp);
1450 	vn_finished_write(mp);
1451 	NDFREE(&nd, NDF_ONLY_PNBUF);
1452 	return (error);
1453 }
1454 
1455 /*
1456  * Make a hard file link.
1457  */
1458 #ifndef _SYS_SYSPROTO_H_
1459 struct link_args {
1460 	char	*path;
1461 	char	*link;
1462 };
1463 #endif
1464 int
1465 sys_link(td, uap)
1466 	struct thread *td;
1467 	register struct link_args /* {
1468 		char *path;
1469 		char *link;
1470 	} */ *uap;
1471 {
1472 
1473 	return (kern_link(td, uap->path, uap->link, UIO_USERSPACE));
1474 }
1475 
1476 #ifndef _SYS_SYSPROTO_H_
1477 struct linkat_args {
1478 	int	fd1;
1479 	char	*path1;
1480 	int	fd2;
1481 	char	*path2;
1482 	int	flag;
1483 };
1484 #endif
1485 int
1486 sys_linkat(struct thread *td, struct linkat_args *uap)
1487 {
1488 	int flag;
1489 
1490 	flag = uap->flag;
1491 	if (flag & ~AT_SYMLINK_FOLLOW)
1492 		return (EINVAL);
1493 
1494 	return (kern_linkat(td, uap->fd1, uap->fd2, uap->path1, uap->path2,
1495 	    UIO_USERSPACE, (flag & AT_SYMLINK_FOLLOW) ? FOLLOW : NOFOLLOW));
1496 }
1497 
1498 int hardlink_check_uid = 0;
1499 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_uid, CTLFLAG_RW,
1500     &hardlink_check_uid, 0,
1501     "Unprivileged processes cannot create hard links to files owned by other "
1502     "users");
1503 static int hardlink_check_gid = 0;
1504 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_gid, CTLFLAG_RW,
1505     &hardlink_check_gid, 0,
1506     "Unprivileged processes cannot create hard links to files owned by other "
1507     "groups");
1508 
1509 static int
1510 can_hardlink(struct vnode *vp, struct ucred *cred)
1511 {
1512 	struct vattr va;
1513 	int error;
1514 
1515 	if (!hardlink_check_uid && !hardlink_check_gid)
1516 		return (0);
1517 
1518 	error = VOP_GETATTR(vp, &va, cred);
1519 	if (error != 0)
1520 		return (error);
1521 
1522 	if (hardlink_check_uid && cred->cr_uid != va.va_uid) {
1523 		error = priv_check_cred(cred, PRIV_VFS_LINK, 0);
1524 		if (error != 0)
1525 			return (error);
1526 	}
1527 
1528 	if (hardlink_check_gid && !groupmember(va.va_gid, cred)) {
1529 		error = priv_check_cred(cred, PRIV_VFS_LINK, 0);
1530 		if (error != 0)
1531 			return (error);
1532 	}
1533 
1534 	return (0);
1535 }
1536 
1537 int
1538 kern_link(struct thread *td, char *path, char *link, enum uio_seg segflg)
1539 {
1540 
1541 	return (kern_linkat(td, AT_FDCWD, AT_FDCWD, path,link, segflg, FOLLOW));
1542 }
1543 
1544 int
1545 kern_linkat(struct thread *td, int fd1, int fd2, char *path1, char *path2,
1546     enum uio_seg segflg, int follow)
1547 {
1548 	struct vnode *vp;
1549 	struct mount *mp;
1550 	struct nameidata nd;
1551 	cap_rights_t rights;
1552 	int error;
1553 
1554 again:
1555 	bwillwrite();
1556 	NDINIT_AT(&nd, LOOKUP, follow | AUDITVNODE1, segflg, path1, fd1, td);
1557 
1558 	if ((error = namei(&nd)) != 0)
1559 		return (error);
1560 	NDFREE(&nd, NDF_ONLY_PNBUF);
1561 	vp = nd.ni_vp;
1562 	if (vp->v_type == VDIR) {
1563 		vrele(vp);
1564 		return (EPERM);		/* POSIX */
1565 	}
1566 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE2,
1567 	    segflg, path2, fd2, cap_rights_init(&rights, CAP_LINKAT), td);
1568 	if ((error = namei(&nd)) == 0) {
1569 		if (nd.ni_vp != NULL) {
1570 			NDFREE(&nd, NDF_ONLY_PNBUF);
1571 			if (nd.ni_dvp == nd.ni_vp)
1572 				vrele(nd.ni_dvp);
1573 			else
1574 				vput(nd.ni_dvp);
1575 			vrele(nd.ni_vp);
1576 			vrele(vp);
1577 			return (EEXIST);
1578 		} else if (nd.ni_dvp->v_mount != vp->v_mount) {
1579 			/*
1580 			 * Cross-device link.  No need to recheck
1581 			 * vp->v_type, since it cannot change, except
1582 			 * to VBAD.
1583 			 */
1584 			NDFREE(&nd, NDF_ONLY_PNBUF);
1585 			vput(nd.ni_dvp);
1586 			vrele(vp);
1587 			return (EXDEV);
1588 		} else if ((error = vn_lock(vp, LK_EXCLUSIVE)) == 0) {
1589 			error = can_hardlink(vp, td->td_ucred);
1590 #ifdef MAC
1591 			if (error == 0)
1592 				error = mac_vnode_check_link(td->td_ucred,
1593 				    nd.ni_dvp, vp, &nd.ni_cnd);
1594 #endif
1595 			if (error != 0) {
1596 				vput(vp);
1597 				vput(nd.ni_dvp);
1598 				NDFREE(&nd, NDF_ONLY_PNBUF);
1599 				return (error);
1600 			}
1601 			error = vn_start_write(vp, &mp, V_NOWAIT);
1602 			if (error != 0) {
1603 				vput(vp);
1604 				vput(nd.ni_dvp);
1605 				NDFREE(&nd, NDF_ONLY_PNBUF);
1606 				error = vn_start_write(NULL, &mp,
1607 				    V_XSLEEP | PCATCH);
1608 				if (error != 0)
1609 					return (error);
1610 				goto again;
1611 			}
1612 			error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
1613 			VOP_UNLOCK(vp, 0);
1614 			vput(nd.ni_dvp);
1615 			vn_finished_write(mp);
1616 			NDFREE(&nd, NDF_ONLY_PNBUF);
1617 		} else {
1618 			vput(nd.ni_dvp);
1619 			NDFREE(&nd, NDF_ONLY_PNBUF);
1620 			vrele(vp);
1621 			goto again;
1622 		}
1623 	}
1624 	vrele(vp);
1625 	return (error);
1626 }
1627 
1628 /*
1629  * Make a symbolic link.
1630  */
1631 #ifndef _SYS_SYSPROTO_H_
1632 struct symlink_args {
1633 	char	*path;
1634 	char	*link;
1635 };
1636 #endif
1637 int
1638 sys_symlink(td, uap)
1639 	struct thread *td;
1640 	register struct symlink_args /* {
1641 		char *path;
1642 		char *link;
1643 	} */ *uap;
1644 {
1645 
1646 	return (kern_symlink(td, uap->path, uap->link, UIO_USERSPACE));
1647 }
1648 
1649 #ifndef _SYS_SYSPROTO_H_
1650 struct symlinkat_args {
1651 	char	*path;
1652 	int	fd;
1653 	char	*path2;
1654 };
1655 #endif
1656 int
1657 sys_symlinkat(struct thread *td, struct symlinkat_args *uap)
1658 {
1659 
1660 	return (kern_symlinkat(td, uap->path1, uap->fd, uap->path2,
1661 	    UIO_USERSPACE));
1662 }
1663 
1664 int
1665 kern_symlink(struct thread *td, char *path, char *link, enum uio_seg segflg)
1666 {
1667 
1668 	return (kern_symlinkat(td, path, AT_FDCWD, link, segflg));
1669 }
1670 
1671 int
1672 kern_symlinkat(struct thread *td, char *path1, int fd, char *path2,
1673     enum uio_seg segflg)
1674 {
1675 	struct mount *mp;
1676 	struct vattr vattr;
1677 	char *syspath;
1678 	struct nameidata nd;
1679 	int error;
1680 	cap_rights_t rights;
1681 
1682 	if (segflg == UIO_SYSSPACE) {
1683 		syspath = path1;
1684 	} else {
1685 		syspath = uma_zalloc(namei_zone, M_WAITOK);
1686 		if ((error = copyinstr(path1, syspath, MAXPATHLEN, NULL)) != 0)
1687 			goto out;
1688 	}
1689 	AUDIT_ARG_TEXT(syspath);
1690 restart:
1691 	bwillwrite();
1692 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1,
1693 	    segflg, path2, fd, cap_rights_init(&rights, CAP_SYMLINKAT), td);
1694 	if ((error = namei(&nd)) != 0)
1695 		goto out;
1696 	if (nd.ni_vp) {
1697 		NDFREE(&nd, NDF_ONLY_PNBUF);
1698 		if (nd.ni_vp == nd.ni_dvp)
1699 			vrele(nd.ni_dvp);
1700 		else
1701 			vput(nd.ni_dvp);
1702 		vrele(nd.ni_vp);
1703 		error = EEXIST;
1704 		goto out;
1705 	}
1706 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1707 		NDFREE(&nd, NDF_ONLY_PNBUF);
1708 		vput(nd.ni_dvp);
1709 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1710 			goto out;
1711 		goto restart;
1712 	}
1713 	VATTR_NULL(&vattr);
1714 	vattr.va_mode = ACCESSPERMS &~ td->td_proc->p_fd->fd_cmask;
1715 #ifdef MAC
1716 	vattr.va_type = VLNK;
1717 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1718 	    &vattr);
1719 	if (error != 0)
1720 		goto out2;
1721 #endif
1722 	error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, syspath);
1723 	if (error == 0)
1724 		vput(nd.ni_vp);
1725 #ifdef MAC
1726 out2:
1727 #endif
1728 	NDFREE(&nd, NDF_ONLY_PNBUF);
1729 	vput(nd.ni_dvp);
1730 	vn_finished_write(mp);
1731 out:
1732 	if (segflg != UIO_SYSSPACE)
1733 		uma_zfree(namei_zone, syspath);
1734 	return (error);
1735 }
1736 
1737 /*
1738  * Delete a whiteout from the filesystem.
1739  */
1740 int
1741 sys_undelete(td, uap)
1742 	struct thread *td;
1743 	register struct undelete_args /* {
1744 		char *path;
1745 	} */ *uap;
1746 {
1747 	struct mount *mp;
1748 	struct nameidata nd;
1749 	int error;
1750 
1751 restart:
1752 	bwillwrite();
1753 	NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | AUDITVNODE1,
1754 	    UIO_USERSPACE, uap->path, td);
1755 	error = namei(&nd);
1756 	if (error != 0)
1757 		return (error);
1758 
1759 	if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) {
1760 		NDFREE(&nd, NDF_ONLY_PNBUF);
1761 		if (nd.ni_vp == nd.ni_dvp)
1762 			vrele(nd.ni_dvp);
1763 		else
1764 			vput(nd.ni_dvp);
1765 		if (nd.ni_vp)
1766 			vrele(nd.ni_vp);
1767 		return (EEXIST);
1768 	}
1769 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1770 		NDFREE(&nd, NDF_ONLY_PNBUF);
1771 		vput(nd.ni_dvp);
1772 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1773 			return (error);
1774 		goto restart;
1775 	}
1776 	error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE);
1777 	NDFREE(&nd, NDF_ONLY_PNBUF);
1778 	vput(nd.ni_dvp);
1779 	vn_finished_write(mp);
1780 	return (error);
1781 }
1782 
1783 /*
1784  * Delete a name from the filesystem.
1785  */
1786 #ifndef _SYS_SYSPROTO_H_
1787 struct unlink_args {
1788 	char	*path;
1789 };
1790 #endif
1791 int
1792 sys_unlink(td, uap)
1793 	struct thread *td;
1794 	struct unlink_args /* {
1795 		char *path;
1796 	} */ *uap;
1797 {
1798 
1799 	return (kern_unlink(td, uap->path, UIO_USERSPACE));
1800 }
1801 
1802 #ifndef _SYS_SYSPROTO_H_
1803 struct unlinkat_args {
1804 	int	fd;
1805 	char	*path;
1806 	int	flag;
1807 };
1808 #endif
1809 int
1810 sys_unlinkat(struct thread *td, struct unlinkat_args *uap)
1811 {
1812 	int flag = uap->flag;
1813 	int fd = uap->fd;
1814 	char *path = uap->path;
1815 
1816 	if (flag & ~AT_REMOVEDIR)
1817 		return (EINVAL);
1818 
1819 	if (flag & AT_REMOVEDIR)
1820 		return (kern_rmdirat(td, fd, path, UIO_USERSPACE));
1821 	else
1822 		return (kern_unlinkat(td, fd, path, UIO_USERSPACE, 0));
1823 }
1824 
1825 int
1826 kern_unlink(struct thread *td, char *path, enum uio_seg pathseg)
1827 {
1828 
1829 	return (kern_unlinkat(td, AT_FDCWD, path, pathseg, 0));
1830 }
1831 
1832 int
1833 kern_unlinkat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1834     ino_t oldinum)
1835 {
1836 	struct mount *mp;
1837 	struct vnode *vp;
1838 	struct nameidata nd;
1839 	struct stat sb;
1840 	cap_rights_t rights;
1841 	int error;
1842 
1843 restart:
1844 	bwillwrite();
1845 	NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1,
1846 	    pathseg, path, fd, cap_rights_init(&rights, CAP_UNLINKAT), td);
1847 	if ((error = namei(&nd)) != 0)
1848 		return (error == EINVAL ? EPERM : error);
1849 	vp = nd.ni_vp;
1850 	if (vp->v_type == VDIR && oldinum == 0) {
1851 		error = EPERM;		/* POSIX */
1852 	} else if (oldinum != 0 &&
1853 		  ((error = vn_stat(vp, &sb, td->td_ucred, NOCRED, td)) == 0) &&
1854 		  sb.st_ino != oldinum) {
1855 			error = EIDRM;	/* Identifier removed */
1856 	} else {
1857 		/*
1858 		 * The root of a mounted filesystem cannot be deleted.
1859 		 *
1860 		 * XXX: can this only be a VDIR case?
1861 		 */
1862 		if (vp->v_vflag & VV_ROOT)
1863 			error = EBUSY;
1864 	}
1865 	if (error == 0) {
1866 		if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1867 			NDFREE(&nd, NDF_ONLY_PNBUF);
1868 			vput(nd.ni_dvp);
1869 			if (vp == nd.ni_dvp)
1870 				vrele(vp);
1871 			else
1872 				vput(vp);
1873 			if ((error = vn_start_write(NULL, &mp,
1874 			    V_XSLEEP | PCATCH)) != 0)
1875 				return (error);
1876 			goto restart;
1877 		}
1878 #ifdef MAC
1879 		error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
1880 		    &nd.ni_cnd);
1881 		if (error != 0)
1882 			goto out;
1883 #endif
1884 		vfs_notify_upper(vp, VFS_NOTIFY_UPPER_UNLINK);
1885 		error = VOP_REMOVE(nd.ni_dvp, vp, &nd.ni_cnd);
1886 #ifdef MAC
1887 out:
1888 #endif
1889 		vn_finished_write(mp);
1890 	}
1891 	NDFREE(&nd, NDF_ONLY_PNBUF);
1892 	vput(nd.ni_dvp);
1893 	if (vp == nd.ni_dvp)
1894 		vrele(vp);
1895 	else
1896 		vput(vp);
1897 	return (error);
1898 }
1899 
1900 /*
1901  * Reposition read/write file offset.
1902  */
1903 #ifndef _SYS_SYSPROTO_H_
1904 struct lseek_args {
1905 	int	fd;
1906 	int	pad;
1907 	off_t	offset;
1908 	int	whence;
1909 };
1910 #endif
1911 int
1912 sys_lseek(td, uap)
1913 	struct thread *td;
1914 	register struct lseek_args /* {
1915 		int fd;
1916 		int pad;
1917 		off_t offset;
1918 		int whence;
1919 	} */ *uap;
1920 {
1921 	struct file *fp;
1922 	cap_rights_t rights;
1923 	int error;
1924 
1925 	AUDIT_ARG_FD(uap->fd);
1926 	error = fget(td, uap->fd, cap_rights_init(&rights, CAP_SEEK), &fp);
1927 	if (error != 0)
1928 		return (error);
1929 	error = (fp->f_ops->fo_flags & DFLAG_SEEKABLE) != 0 ?
1930 	    fo_seek(fp, uap->offset, uap->whence, td) : ESPIPE;
1931 	fdrop(fp, td);
1932 	return (error);
1933 }
1934 
1935 #if defined(COMPAT_43)
1936 /*
1937  * Reposition read/write file offset.
1938  */
1939 #ifndef _SYS_SYSPROTO_H_
1940 struct olseek_args {
1941 	int	fd;
1942 	long	offset;
1943 	int	whence;
1944 };
1945 #endif
1946 int
1947 olseek(td, uap)
1948 	struct thread *td;
1949 	register struct olseek_args /* {
1950 		int fd;
1951 		long offset;
1952 		int whence;
1953 	} */ *uap;
1954 {
1955 	struct lseek_args /* {
1956 		int fd;
1957 		int pad;
1958 		off_t offset;
1959 		int whence;
1960 	} */ nuap;
1961 
1962 	nuap.fd = uap->fd;
1963 	nuap.offset = uap->offset;
1964 	nuap.whence = uap->whence;
1965 	return (sys_lseek(td, &nuap));
1966 }
1967 #endif /* COMPAT_43 */
1968 
1969 /* Version with the 'pad' argument */
1970 int
1971 freebsd6_lseek(td, uap)
1972 	struct thread *td;
1973 	register struct freebsd6_lseek_args *uap;
1974 {
1975 	struct lseek_args ouap;
1976 
1977 	ouap.fd = uap->fd;
1978 	ouap.offset = uap->offset;
1979 	ouap.whence = uap->whence;
1980 	return (sys_lseek(td, &ouap));
1981 }
1982 
1983 /*
1984  * Check access permissions using passed credentials.
1985  */
1986 static int
1987 vn_access(vp, user_flags, cred, td)
1988 	struct vnode	*vp;
1989 	int		user_flags;
1990 	struct ucred	*cred;
1991 	struct thread	*td;
1992 {
1993 	accmode_t accmode;
1994 	int error;
1995 
1996 	/* Flags == 0 means only check for existence. */
1997 	error = 0;
1998 	if (user_flags) {
1999 		accmode = 0;
2000 		if (user_flags & R_OK)
2001 			accmode |= VREAD;
2002 		if (user_flags & W_OK)
2003 			accmode |= VWRITE;
2004 		if (user_flags & X_OK)
2005 			accmode |= VEXEC;
2006 #ifdef MAC
2007 		error = mac_vnode_check_access(cred, vp, accmode);
2008 		if (error != 0)
2009 			return (error);
2010 #endif
2011 		if ((accmode & VWRITE) == 0 || (error = vn_writechk(vp)) == 0)
2012 			error = VOP_ACCESS(vp, accmode, cred, td);
2013 	}
2014 	return (error);
2015 }
2016 
2017 /*
2018  * Check access permissions using "real" credentials.
2019  */
2020 #ifndef _SYS_SYSPROTO_H_
2021 struct access_args {
2022 	char	*path;
2023 	int	amode;
2024 };
2025 #endif
2026 int
2027 sys_access(td, uap)
2028 	struct thread *td;
2029 	register struct access_args /* {
2030 		char *path;
2031 		int amode;
2032 	} */ *uap;
2033 {
2034 
2035 	return (kern_access(td, uap->path, UIO_USERSPACE, uap->amode));
2036 }
2037 
2038 #ifndef _SYS_SYSPROTO_H_
2039 struct faccessat_args {
2040 	int	dirfd;
2041 	char	*path;
2042 	int	amode;
2043 	int	flag;
2044 }
2045 #endif
2046 int
2047 sys_faccessat(struct thread *td, struct faccessat_args *uap)
2048 {
2049 
2050 	if (uap->flag & ~AT_EACCESS)
2051 		return (EINVAL);
2052 	return (kern_accessat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
2053 	    uap->amode));
2054 }
2055 
2056 int
2057 kern_access(struct thread *td, char *path, enum uio_seg pathseg, int amode)
2058 {
2059 
2060 	return (kern_accessat(td, AT_FDCWD, path, pathseg, 0, amode));
2061 }
2062 
2063 int
2064 kern_accessat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2065     int flag, int amode)
2066 {
2067 	struct ucred *cred, *tmpcred;
2068 	struct vnode *vp;
2069 	struct nameidata nd;
2070 	cap_rights_t rights;
2071 	int error;
2072 
2073 	if (amode != F_OK && (amode & ~(R_OK | W_OK | X_OK)) != 0)
2074 		return (EINVAL);
2075 
2076 	/*
2077 	 * Create and modify a temporary credential instead of one that
2078 	 * is potentially shared.
2079 	 */
2080 	if (!(flag & AT_EACCESS)) {
2081 		cred = td->td_ucred;
2082 		tmpcred = crdup(cred);
2083 		tmpcred->cr_uid = cred->cr_ruid;
2084 		tmpcred->cr_groups[0] = cred->cr_rgid;
2085 		td->td_ucred = tmpcred;
2086 	} else
2087 		cred = tmpcred = td->td_ucred;
2088 	AUDIT_ARG_VALUE(amode);
2089 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF |
2090 	    AUDITVNODE1, pathseg, path, fd, cap_rights_init(&rights, CAP_FSTAT),
2091 	    td);
2092 	if ((error = namei(&nd)) != 0)
2093 		goto out1;
2094 	vp = nd.ni_vp;
2095 
2096 	error = vn_access(vp, amode, tmpcred, td);
2097 	NDFREE(&nd, NDF_ONLY_PNBUF);
2098 	vput(vp);
2099 out1:
2100 	if (!(flag & AT_EACCESS)) {
2101 		td->td_ucred = cred;
2102 		crfree(tmpcred);
2103 	}
2104 	return (error);
2105 }
2106 
2107 /*
2108  * Check access permissions using "effective" credentials.
2109  */
2110 #ifndef _SYS_SYSPROTO_H_
2111 struct eaccess_args {
2112 	char	*path;
2113 	int	amode;
2114 };
2115 #endif
2116 int
2117 sys_eaccess(td, uap)
2118 	struct thread *td;
2119 	register struct eaccess_args /* {
2120 		char *path;
2121 		int amode;
2122 	} */ *uap;
2123 {
2124 
2125 	return (kern_eaccess(td, uap->path, UIO_USERSPACE, uap->amode));
2126 }
2127 
2128 int
2129 kern_eaccess(struct thread *td, char *path, enum uio_seg pathseg, int amode)
2130 {
2131 
2132 	return (kern_accessat(td, AT_FDCWD, path, pathseg, AT_EACCESS, amode));
2133 }
2134 
2135 #if defined(COMPAT_43)
2136 /*
2137  * Get file status; this version follows links.
2138  */
2139 #ifndef _SYS_SYSPROTO_H_
2140 struct ostat_args {
2141 	char	*path;
2142 	struct ostat *ub;
2143 };
2144 #endif
2145 int
2146 ostat(td, uap)
2147 	struct thread *td;
2148 	register struct ostat_args /* {
2149 		char *path;
2150 		struct ostat *ub;
2151 	} */ *uap;
2152 {
2153 	struct stat sb;
2154 	struct ostat osb;
2155 	int error;
2156 
2157 	error = kern_stat(td, uap->path, UIO_USERSPACE, &sb);
2158 	if (error != 0)
2159 		return (error);
2160 	cvtstat(&sb, &osb);
2161 	return (copyout(&osb, uap->ub, sizeof (osb)));
2162 }
2163 
2164 /*
2165  * Get file status; this version does not follow links.
2166  */
2167 #ifndef _SYS_SYSPROTO_H_
2168 struct olstat_args {
2169 	char	*path;
2170 	struct ostat *ub;
2171 };
2172 #endif
2173 int
2174 olstat(td, uap)
2175 	struct thread *td;
2176 	register struct olstat_args /* {
2177 		char *path;
2178 		struct ostat *ub;
2179 	} */ *uap;
2180 {
2181 	struct stat sb;
2182 	struct ostat osb;
2183 	int error;
2184 
2185 	error = kern_lstat(td, uap->path, UIO_USERSPACE, &sb);
2186 	if (error != 0)
2187 		return (error);
2188 	cvtstat(&sb, &osb);
2189 	return (copyout(&osb, uap->ub, sizeof (osb)));
2190 }
2191 
2192 /*
2193  * Convert from an old to a new stat structure.
2194  */
2195 void
2196 cvtstat(st, ost)
2197 	struct stat *st;
2198 	struct ostat *ost;
2199 {
2200 
2201 	ost->st_dev = st->st_dev;
2202 	ost->st_ino = st->st_ino;
2203 	ost->st_mode = st->st_mode;
2204 	ost->st_nlink = st->st_nlink;
2205 	ost->st_uid = st->st_uid;
2206 	ost->st_gid = st->st_gid;
2207 	ost->st_rdev = st->st_rdev;
2208 	if (st->st_size < (quad_t)1 << 32)
2209 		ost->st_size = st->st_size;
2210 	else
2211 		ost->st_size = -2;
2212 	ost->st_atim = st->st_atim;
2213 	ost->st_mtim = st->st_mtim;
2214 	ost->st_ctim = st->st_ctim;
2215 	ost->st_blksize = st->st_blksize;
2216 	ost->st_blocks = st->st_blocks;
2217 	ost->st_flags = st->st_flags;
2218 	ost->st_gen = st->st_gen;
2219 }
2220 #endif /* COMPAT_43 */
2221 
2222 /*
2223  * Get file status; this version follows links.
2224  */
2225 #ifndef _SYS_SYSPROTO_H_
2226 struct stat_args {
2227 	char	*path;
2228 	struct stat *ub;
2229 };
2230 #endif
2231 int
2232 sys_stat(td, uap)
2233 	struct thread *td;
2234 	register struct stat_args /* {
2235 		char *path;
2236 		struct stat *ub;
2237 	} */ *uap;
2238 {
2239 	struct stat sb;
2240 	int error;
2241 
2242 	error = kern_stat(td, uap->path, UIO_USERSPACE, &sb);
2243 	if (error == 0)
2244 		error = copyout(&sb, uap->ub, sizeof (sb));
2245 	return (error);
2246 }
2247 
2248 #ifndef _SYS_SYSPROTO_H_
2249 struct fstatat_args {
2250 	int	fd;
2251 	char	*path;
2252 	struct stat	*buf;
2253 	int	flag;
2254 }
2255 #endif
2256 int
2257 sys_fstatat(struct thread *td, struct fstatat_args *uap)
2258 {
2259 	struct stat sb;
2260 	int error;
2261 
2262 	error = kern_statat(td, uap->flag, uap->fd, uap->path,
2263 	    UIO_USERSPACE, &sb);
2264 	if (error == 0)
2265 		error = copyout(&sb, uap->buf, sizeof (sb));
2266 	return (error);
2267 }
2268 
2269 int
2270 kern_stat(struct thread *td, char *path, enum uio_seg pathseg, struct stat *sbp)
2271 {
2272 
2273 	return (kern_statat(td, 0, AT_FDCWD, path, pathseg, sbp));
2274 }
2275 
2276 int
2277 kern_statat(struct thread *td, int flag, int fd, char *path,
2278     enum uio_seg pathseg, struct stat *sbp)
2279 {
2280 
2281 	return (kern_statat_vnhook(td, flag, fd, path, pathseg, sbp, NULL));
2282 }
2283 
2284 int
2285 kern_statat_vnhook(struct thread *td, int flag, int fd, char *path,
2286     enum uio_seg pathseg, struct stat *sbp,
2287     void (*hook)(struct vnode *vp, struct stat *sbp))
2288 {
2289 	struct nameidata nd;
2290 	struct stat sb;
2291 	cap_rights_t rights;
2292 	int error;
2293 
2294 	if (flag & ~AT_SYMLINK_NOFOLLOW)
2295 		return (EINVAL);
2296 
2297 	NDINIT_ATRIGHTS(&nd, LOOKUP, ((flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW :
2298 	    FOLLOW) | LOCKSHARED | LOCKLEAF | AUDITVNODE1, pathseg, path, fd,
2299 	    cap_rights_init(&rights, CAP_FSTAT), td);
2300 
2301 	if ((error = namei(&nd)) != 0)
2302 		return (error);
2303 	error = vn_stat(nd.ni_vp, &sb, td->td_ucred, NOCRED, td);
2304 	if (error == 0) {
2305 		SDT_PROBE(vfs, , stat, mode, path, sb.st_mode, 0, 0, 0);
2306 		if (S_ISREG(sb.st_mode))
2307 			SDT_PROBE(vfs, , stat, reg, path, pathseg, 0, 0, 0);
2308 		if (__predict_false(hook != NULL))
2309 			hook(nd.ni_vp, &sb);
2310 	}
2311 	NDFREE(&nd, NDF_ONLY_PNBUF);
2312 	vput(nd.ni_vp);
2313 	if (error != 0)
2314 		return (error);
2315 	*sbp = sb;
2316 #ifdef KTRACE
2317 	if (KTRPOINT(td, KTR_STRUCT))
2318 		ktrstat(&sb);
2319 #endif
2320 	return (0);
2321 }
2322 
2323 /*
2324  * Get file status; this version does not follow links.
2325  */
2326 #ifndef _SYS_SYSPROTO_H_
2327 struct lstat_args {
2328 	char	*path;
2329 	struct stat *ub;
2330 };
2331 #endif
2332 int
2333 sys_lstat(td, uap)
2334 	struct thread *td;
2335 	register struct lstat_args /* {
2336 		char *path;
2337 		struct stat *ub;
2338 	} */ *uap;
2339 {
2340 	struct stat sb;
2341 	int error;
2342 
2343 	error = kern_lstat(td, uap->path, UIO_USERSPACE, &sb);
2344 	if (error == 0)
2345 		error = copyout(&sb, uap->ub, sizeof (sb));
2346 	return (error);
2347 }
2348 
2349 int
2350 kern_lstat(struct thread *td, char *path, enum uio_seg pathseg, struct stat *sbp)
2351 {
2352 
2353 	return (kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, path, pathseg,
2354 	    sbp));
2355 }
2356 
2357 /*
2358  * Implementation of the NetBSD [l]stat() functions.
2359  */
2360 void
2361 cvtnstat(sb, nsb)
2362 	struct stat *sb;
2363 	struct nstat *nsb;
2364 {
2365 
2366 	bzero(nsb, sizeof *nsb);
2367 	nsb->st_dev = sb->st_dev;
2368 	nsb->st_ino = sb->st_ino;
2369 	nsb->st_mode = sb->st_mode;
2370 	nsb->st_nlink = sb->st_nlink;
2371 	nsb->st_uid = sb->st_uid;
2372 	nsb->st_gid = sb->st_gid;
2373 	nsb->st_rdev = sb->st_rdev;
2374 	nsb->st_atim = sb->st_atim;
2375 	nsb->st_mtim = sb->st_mtim;
2376 	nsb->st_ctim = sb->st_ctim;
2377 	nsb->st_size = sb->st_size;
2378 	nsb->st_blocks = sb->st_blocks;
2379 	nsb->st_blksize = sb->st_blksize;
2380 	nsb->st_flags = sb->st_flags;
2381 	nsb->st_gen = sb->st_gen;
2382 	nsb->st_birthtim = sb->st_birthtim;
2383 }
2384 
2385 #ifndef _SYS_SYSPROTO_H_
2386 struct nstat_args {
2387 	char	*path;
2388 	struct nstat *ub;
2389 };
2390 #endif
2391 int
2392 sys_nstat(td, uap)
2393 	struct thread *td;
2394 	register struct nstat_args /* {
2395 		char *path;
2396 		struct nstat *ub;
2397 	} */ *uap;
2398 {
2399 	struct stat sb;
2400 	struct nstat nsb;
2401 	int error;
2402 
2403 	error = kern_stat(td, uap->path, UIO_USERSPACE, &sb);
2404 	if (error != 0)
2405 		return (error);
2406 	cvtnstat(&sb, &nsb);
2407 	return (copyout(&nsb, uap->ub, sizeof (nsb)));
2408 }
2409 
2410 /*
2411  * NetBSD lstat.  Get file status; this version does not follow links.
2412  */
2413 #ifndef _SYS_SYSPROTO_H_
2414 struct lstat_args {
2415 	char	*path;
2416 	struct stat *ub;
2417 };
2418 #endif
2419 int
2420 sys_nlstat(td, uap)
2421 	struct thread *td;
2422 	register struct nlstat_args /* {
2423 		char *path;
2424 		struct nstat *ub;
2425 	} */ *uap;
2426 {
2427 	struct stat sb;
2428 	struct nstat nsb;
2429 	int error;
2430 
2431 	error = kern_lstat(td, uap->path, UIO_USERSPACE, &sb);
2432 	if (error != 0)
2433 		return (error);
2434 	cvtnstat(&sb, &nsb);
2435 	return (copyout(&nsb, uap->ub, sizeof (nsb)));
2436 }
2437 
2438 /*
2439  * Get configurable pathname variables.
2440  */
2441 #ifndef _SYS_SYSPROTO_H_
2442 struct pathconf_args {
2443 	char	*path;
2444 	int	name;
2445 };
2446 #endif
2447 int
2448 sys_pathconf(td, uap)
2449 	struct thread *td;
2450 	register struct pathconf_args /* {
2451 		char *path;
2452 		int name;
2453 	} */ *uap;
2454 {
2455 
2456 	return (kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name, FOLLOW));
2457 }
2458 
2459 #ifndef _SYS_SYSPROTO_H_
2460 struct lpathconf_args {
2461 	char	*path;
2462 	int	name;
2463 };
2464 #endif
2465 int
2466 sys_lpathconf(td, uap)
2467 	struct thread *td;
2468 	register struct lpathconf_args /* {
2469 		char *path;
2470 		int name;
2471 	} */ *uap;
2472 {
2473 
2474 	return (kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name,
2475 	    NOFOLLOW));
2476 }
2477 
2478 int
2479 kern_pathconf(struct thread *td, char *path, enum uio_seg pathseg, int name,
2480     u_long flags)
2481 {
2482 	struct nameidata nd;
2483 	int error;
2484 
2485 	NDINIT(&nd, LOOKUP, LOCKSHARED | LOCKLEAF | AUDITVNODE1 | flags,
2486 	    pathseg, path, td);
2487 	if ((error = namei(&nd)) != 0)
2488 		return (error);
2489 	NDFREE(&nd, NDF_ONLY_PNBUF);
2490 
2491 	/* If asynchronous I/O is available, it works for all files. */
2492 	if (name == _PC_ASYNC_IO)
2493 		td->td_retval[0] = async_io_version;
2494 	else
2495 		error = VOP_PATHCONF(nd.ni_vp, name, td->td_retval);
2496 	vput(nd.ni_vp);
2497 	return (error);
2498 }
2499 
2500 /*
2501  * Return target name of a symbolic link.
2502  */
2503 #ifndef _SYS_SYSPROTO_H_
2504 struct readlink_args {
2505 	char	*path;
2506 	char	*buf;
2507 	size_t	count;
2508 };
2509 #endif
2510 int
2511 sys_readlink(td, uap)
2512 	struct thread *td;
2513 	register struct readlink_args /* {
2514 		char *path;
2515 		char *buf;
2516 		size_t count;
2517 	} */ *uap;
2518 {
2519 
2520 	return (kern_readlink(td, uap->path, UIO_USERSPACE, uap->buf,
2521 	    UIO_USERSPACE, uap->count));
2522 }
2523 #ifndef _SYS_SYSPROTO_H_
2524 struct readlinkat_args {
2525 	int	fd;
2526 	char	*path;
2527 	char	*buf;
2528 	size_t	bufsize;
2529 };
2530 #endif
2531 int
2532 sys_readlinkat(struct thread *td, struct readlinkat_args *uap)
2533 {
2534 
2535 	return (kern_readlinkat(td, uap->fd, uap->path, UIO_USERSPACE,
2536 	    uap->buf, UIO_USERSPACE, uap->bufsize));
2537 }
2538 
2539 int
2540 kern_readlink(struct thread *td, char *path, enum uio_seg pathseg, char *buf,
2541     enum uio_seg bufseg, size_t count)
2542 {
2543 
2544 	return (kern_readlinkat(td, AT_FDCWD, path, pathseg, buf, bufseg,
2545 	    count));
2546 }
2547 
2548 int
2549 kern_readlinkat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2550     char *buf, enum uio_seg bufseg, size_t count)
2551 {
2552 	struct vnode *vp;
2553 	struct iovec aiov;
2554 	struct uio auio;
2555 	struct nameidata nd;
2556 	int error;
2557 
2558 	if (count > IOSIZE_MAX)
2559 		return (EINVAL);
2560 
2561 	NDINIT_AT(&nd, LOOKUP, NOFOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
2562 	    pathseg, path, fd, td);
2563 
2564 	if ((error = namei(&nd)) != 0)
2565 		return (error);
2566 	NDFREE(&nd, NDF_ONLY_PNBUF);
2567 	vp = nd.ni_vp;
2568 #ifdef MAC
2569 	error = mac_vnode_check_readlink(td->td_ucred, vp);
2570 	if (error != 0) {
2571 		vput(vp);
2572 		return (error);
2573 	}
2574 #endif
2575 	if (vp->v_type != VLNK)
2576 		error = EINVAL;
2577 	else {
2578 		aiov.iov_base = buf;
2579 		aiov.iov_len = count;
2580 		auio.uio_iov = &aiov;
2581 		auio.uio_iovcnt = 1;
2582 		auio.uio_offset = 0;
2583 		auio.uio_rw = UIO_READ;
2584 		auio.uio_segflg = bufseg;
2585 		auio.uio_td = td;
2586 		auio.uio_resid = count;
2587 		error = VOP_READLINK(vp, &auio, td->td_ucred);
2588 		td->td_retval[0] = count - auio.uio_resid;
2589 	}
2590 	vput(vp);
2591 	return (error);
2592 }
2593 
2594 /*
2595  * Common implementation code for chflags() and fchflags().
2596  */
2597 static int
2598 setfflags(td, vp, flags)
2599 	struct thread *td;
2600 	struct vnode *vp;
2601 	u_long flags;
2602 {
2603 	struct mount *mp;
2604 	struct vattr vattr;
2605 	int error;
2606 
2607 	/* We can't support the value matching VNOVAL. */
2608 	if (flags == VNOVAL)
2609 		return (EOPNOTSUPP);
2610 
2611 	/*
2612 	 * Prevent non-root users from setting flags on devices.  When
2613 	 * a device is reused, users can retain ownership of the device
2614 	 * if they are allowed to set flags and programs assume that
2615 	 * chown can't fail when done as root.
2616 	 */
2617 	if (vp->v_type == VCHR || vp->v_type == VBLK) {
2618 		error = priv_check(td, PRIV_VFS_CHFLAGS_DEV);
2619 		if (error != 0)
2620 			return (error);
2621 	}
2622 
2623 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2624 		return (error);
2625 	VATTR_NULL(&vattr);
2626 	vattr.va_flags = flags;
2627 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2628 #ifdef MAC
2629 	error = mac_vnode_check_setflags(td->td_ucred, vp, vattr.va_flags);
2630 	if (error == 0)
2631 #endif
2632 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
2633 	VOP_UNLOCK(vp, 0);
2634 	vn_finished_write(mp);
2635 	return (error);
2636 }
2637 
2638 /*
2639  * Change flags of a file given a path name.
2640  */
2641 #ifndef _SYS_SYSPROTO_H_
2642 struct chflags_args {
2643 	const char *path;
2644 	u_long	flags;
2645 };
2646 #endif
2647 int
2648 sys_chflags(td, uap)
2649 	struct thread *td;
2650 	register struct chflags_args /* {
2651 		const char *path;
2652 		u_long flags;
2653 	} */ *uap;
2654 {
2655 
2656 	return (kern_chflags(td, uap->path, UIO_USERSPACE, uap->flags));
2657 }
2658 
2659 #ifndef _SYS_SYSPROTO_H_
2660 struct chflagsat_args {
2661 	int	fd;
2662 	const char *path;
2663 	u_long	flags;
2664 	int	atflag;
2665 }
2666 #endif
2667 int
2668 sys_chflagsat(struct thread *td, struct chflagsat_args *uap)
2669 {
2670 	int fd = uap->fd;
2671 	const char *path = uap->path;
2672 	u_long flags = uap->flags;
2673 	int atflag = uap->atflag;
2674 
2675 	if (atflag & ~AT_SYMLINK_NOFOLLOW)
2676 		return (EINVAL);
2677 
2678 	return (kern_chflagsat(td, fd, path, UIO_USERSPACE, flags, atflag));
2679 }
2680 
2681 static int
2682 kern_chflags(struct thread *td, const char *path, enum uio_seg pathseg,
2683     u_long flags)
2684 {
2685 
2686 	return (kern_chflagsat(td, AT_FDCWD, path, pathseg, flags, 0));
2687 }
2688 
2689 /*
2690  * Same as chflags() but doesn't follow symlinks.
2691  */
2692 int
2693 sys_lchflags(td, uap)
2694 	struct thread *td;
2695 	register struct lchflags_args /* {
2696 		const char *path;
2697 		u_long flags;
2698 	} */ *uap;
2699 {
2700 
2701 	return (kern_chflagsat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2702 	    uap->flags, AT_SYMLINK_NOFOLLOW));
2703 }
2704 
2705 static int
2706 kern_chflagsat(struct thread *td, int fd, const char *path,
2707     enum uio_seg pathseg, u_long flags, int atflag)
2708 {
2709 	struct nameidata nd;
2710 	cap_rights_t rights;
2711 	int error, follow;
2712 
2713 	AUDIT_ARG_FFLAGS(flags);
2714 	follow = (atflag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
2715 	NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, pathseg, path, fd,
2716 	    cap_rights_init(&rights, CAP_FCHFLAGS), td);
2717 	if ((error = namei(&nd)) != 0)
2718 		return (error);
2719 	NDFREE(&nd, NDF_ONLY_PNBUF);
2720 	error = setfflags(td, nd.ni_vp, flags);
2721 	vrele(nd.ni_vp);
2722 	return (error);
2723 }
2724 
2725 /*
2726  * Change flags of a file given a file descriptor.
2727  */
2728 #ifndef _SYS_SYSPROTO_H_
2729 struct fchflags_args {
2730 	int	fd;
2731 	u_long	flags;
2732 };
2733 #endif
2734 int
2735 sys_fchflags(td, uap)
2736 	struct thread *td;
2737 	register struct fchflags_args /* {
2738 		int fd;
2739 		u_long flags;
2740 	} */ *uap;
2741 {
2742 	struct file *fp;
2743 	cap_rights_t rights;
2744 	int error;
2745 
2746 	AUDIT_ARG_FD(uap->fd);
2747 	AUDIT_ARG_FFLAGS(uap->flags);
2748 	error = getvnode(td->td_proc->p_fd, uap->fd,
2749 	    cap_rights_init(&rights, CAP_FCHFLAGS), &fp);
2750 	if (error != 0)
2751 		return (error);
2752 #ifdef AUDIT
2753 	vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
2754 	AUDIT_ARG_VNODE1(fp->f_vnode);
2755 	VOP_UNLOCK(fp->f_vnode, 0);
2756 #endif
2757 	error = setfflags(td, fp->f_vnode, uap->flags);
2758 	fdrop(fp, td);
2759 	return (error);
2760 }
2761 
2762 /*
2763  * Common implementation code for chmod(), lchmod() and fchmod().
2764  */
2765 int
2766 setfmode(td, cred, vp, mode)
2767 	struct thread *td;
2768 	struct ucred *cred;
2769 	struct vnode *vp;
2770 	int mode;
2771 {
2772 	struct mount *mp;
2773 	struct vattr vattr;
2774 	int error;
2775 
2776 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2777 		return (error);
2778 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2779 	VATTR_NULL(&vattr);
2780 	vattr.va_mode = mode & ALLPERMS;
2781 #ifdef MAC
2782 	error = mac_vnode_check_setmode(cred, vp, vattr.va_mode);
2783 	if (error == 0)
2784 #endif
2785 		error = VOP_SETATTR(vp, &vattr, cred);
2786 	VOP_UNLOCK(vp, 0);
2787 	vn_finished_write(mp);
2788 	return (error);
2789 }
2790 
2791 /*
2792  * Change mode of a file given path name.
2793  */
2794 #ifndef _SYS_SYSPROTO_H_
2795 struct chmod_args {
2796 	char	*path;
2797 	int	mode;
2798 };
2799 #endif
2800 int
2801 sys_chmod(td, uap)
2802 	struct thread *td;
2803 	register struct chmod_args /* {
2804 		char *path;
2805 		int mode;
2806 	} */ *uap;
2807 {
2808 
2809 	return (kern_chmod(td, uap->path, UIO_USERSPACE, uap->mode));
2810 }
2811 
2812 #ifndef _SYS_SYSPROTO_H_
2813 struct fchmodat_args {
2814 	int	dirfd;
2815 	char	*path;
2816 	mode_t	mode;
2817 	int	flag;
2818 }
2819 #endif
2820 int
2821 sys_fchmodat(struct thread *td, struct fchmodat_args *uap)
2822 {
2823 	int flag = uap->flag;
2824 	int fd = uap->fd;
2825 	char *path = uap->path;
2826 	mode_t mode = uap->mode;
2827 
2828 	if (flag & ~AT_SYMLINK_NOFOLLOW)
2829 		return (EINVAL);
2830 
2831 	return (kern_fchmodat(td, fd, path, UIO_USERSPACE, mode, flag));
2832 }
2833 
2834 int
2835 kern_chmod(struct thread *td, char *path, enum uio_seg pathseg, int mode)
2836 {
2837 
2838 	return (kern_fchmodat(td, AT_FDCWD, path, pathseg, mode, 0));
2839 }
2840 
2841 /*
2842  * Change mode of a file given path name (don't follow links.)
2843  */
2844 #ifndef _SYS_SYSPROTO_H_
2845 struct lchmod_args {
2846 	char	*path;
2847 	int	mode;
2848 };
2849 #endif
2850 int
2851 sys_lchmod(td, uap)
2852 	struct thread *td;
2853 	register struct lchmod_args /* {
2854 		char *path;
2855 		int mode;
2856 	} */ *uap;
2857 {
2858 
2859 	return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2860 	    uap->mode, AT_SYMLINK_NOFOLLOW));
2861 }
2862 
2863 int
2864 kern_fchmodat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2865     mode_t mode, int flag)
2866 {
2867 	struct nameidata nd;
2868 	cap_rights_t rights;
2869 	int error, follow;
2870 
2871 	AUDIT_ARG_MODE(mode);
2872 	follow = (flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
2873 	NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, pathseg, path, fd,
2874 	    cap_rights_init(&rights, CAP_FCHMOD), td);
2875 	if ((error = namei(&nd)) != 0)
2876 		return (error);
2877 	NDFREE(&nd, NDF_ONLY_PNBUF);
2878 	error = setfmode(td, td->td_ucred, nd.ni_vp, mode);
2879 	vrele(nd.ni_vp);
2880 	return (error);
2881 }
2882 
2883 /*
2884  * Change mode of a file given a file descriptor.
2885  */
2886 #ifndef _SYS_SYSPROTO_H_
2887 struct fchmod_args {
2888 	int	fd;
2889 	int	mode;
2890 };
2891 #endif
2892 int
2893 sys_fchmod(struct thread *td, struct fchmod_args *uap)
2894 {
2895 	struct file *fp;
2896 	cap_rights_t rights;
2897 	int error;
2898 
2899 	AUDIT_ARG_FD(uap->fd);
2900 	AUDIT_ARG_MODE(uap->mode);
2901 
2902 	error = fget(td, uap->fd, cap_rights_init(&rights, CAP_FCHMOD), &fp);
2903 	if (error != 0)
2904 		return (error);
2905 	error = fo_chmod(fp, uap->mode, td->td_ucred, td);
2906 	fdrop(fp, td);
2907 	return (error);
2908 }
2909 
2910 /*
2911  * Common implementation for chown(), lchown(), and fchown()
2912  */
2913 int
2914 setfown(td, cred, vp, uid, gid)
2915 	struct thread *td;
2916 	struct ucred *cred;
2917 	struct vnode *vp;
2918 	uid_t uid;
2919 	gid_t gid;
2920 {
2921 	struct mount *mp;
2922 	struct vattr vattr;
2923 	int error;
2924 
2925 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2926 		return (error);
2927 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2928 	VATTR_NULL(&vattr);
2929 	vattr.va_uid = uid;
2930 	vattr.va_gid = gid;
2931 #ifdef MAC
2932 	error = mac_vnode_check_setowner(cred, vp, vattr.va_uid,
2933 	    vattr.va_gid);
2934 	if (error == 0)
2935 #endif
2936 		error = VOP_SETATTR(vp, &vattr, cred);
2937 	VOP_UNLOCK(vp, 0);
2938 	vn_finished_write(mp);
2939 	return (error);
2940 }
2941 
2942 /*
2943  * Set ownership given a path name.
2944  */
2945 #ifndef _SYS_SYSPROTO_H_
2946 struct chown_args {
2947 	char	*path;
2948 	int	uid;
2949 	int	gid;
2950 };
2951 #endif
2952 int
2953 sys_chown(td, uap)
2954 	struct thread *td;
2955 	register struct chown_args /* {
2956 		char *path;
2957 		int uid;
2958 		int gid;
2959 	} */ *uap;
2960 {
2961 
2962 	return (kern_chown(td, uap->path, UIO_USERSPACE, uap->uid, uap->gid));
2963 }
2964 
2965 #ifndef _SYS_SYSPROTO_H_
2966 struct fchownat_args {
2967 	int fd;
2968 	const char * path;
2969 	uid_t uid;
2970 	gid_t gid;
2971 	int flag;
2972 };
2973 #endif
2974 int
2975 sys_fchownat(struct thread *td, struct fchownat_args *uap)
2976 {
2977 	int flag;
2978 
2979 	flag = uap->flag;
2980 	if (flag & ~AT_SYMLINK_NOFOLLOW)
2981 		return (EINVAL);
2982 
2983 	return (kern_fchownat(td, uap->fd, uap->path, UIO_USERSPACE, uap->uid,
2984 	    uap->gid, uap->flag));
2985 }
2986 
2987 int
2988 kern_chown(struct thread *td, char *path, enum uio_seg pathseg, int uid,
2989     int gid)
2990 {
2991 
2992 	return (kern_fchownat(td, AT_FDCWD, path, pathseg, uid, gid, 0));
2993 }
2994 
2995 int
2996 kern_fchownat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2997     int uid, int gid, int flag)
2998 {
2999 	struct nameidata nd;
3000 	cap_rights_t rights;
3001 	int error, follow;
3002 
3003 	AUDIT_ARG_OWNER(uid, gid);
3004 	follow = (flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
3005 	NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, pathseg, path, fd,
3006 	    cap_rights_init(&rights, CAP_FCHOWN), td);
3007 
3008 	if ((error = namei(&nd)) != 0)
3009 		return (error);
3010 	NDFREE(&nd, NDF_ONLY_PNBUF);
3011 	error = setfown(td, td->td_ucred, nd.ni_vp, uid, gid);
3012 	vrele(nd.ni_vp);
3013 	return (error);
3014 }
3015 
3016 /*
3017  * Set ownership given a path name, do not cross symlinks.
3018  */
3019 #ifndef _SYS_SYSPROTO_H_
3020 struct lchown_args {
3021 	char	*path;
3022 	int	uid;
3023 	int	gid;
3024 };
3025 #endif
3026 int
3027 sys_lchown(td, uap)
3028 	struct thread *td;
3029 	register struct lchown_args /* {
3030 		char *path;
3031 		int uid;
3032 		int gid;
3033 	} */ *uap;
3034 {
3035 
3036 	return (kern_lchown(td, uap->path, UIO_USERSPACE, uap->uid, uap->gid));
3037 }
3038 
3039 int
3040 kern_lchown(struct thread *td, char *path, enum uio_seg pathseg, int uid,
3041     int gid)
3042 {
3043 
3044 	return (kern_fchownat(td, AT_FDCWD, path, pathseg, uid, gid,
3045 	    AT_SYMLINK_NOFOLLOW));
3046 }
3047 
3048 /*
3049  * Set ownership given a file descriptor.
3050  */
3051 #ifndef _SYS_SYSPROTO_H_
3052 struct fchown_args {
3053 	int	fd;
3054 	int	uid;
3055 	int	gid;
3056 };
3057 #endif
3058 int
3059 sys_fchown(td, uap)
3060 	struct thread *td;
3061 	register struct fchown_args /* {
3062 		int fd;
3063 		int uid;
3064 		int gid;
3065 	} */ *uap;
3066 {
3067 	struct file *fp;
3068 	cap_rights_t rights;
3069 	int error;
3070 
3071 	AUDIT_ARG_FD(uap->fd);
3072 	AUDIT_ARG_OWNER(uap->uid, uap->gid);
3073 	error = fget(td, uap->fd, cap_rights_init(&rights, CAP_FCHOWN), &fp);
3074 	if (error != 0)
3075 		return (error);
3076 	error = fo_chown(fp, uap->uid, uap->gid, td->td_ucred, td);
3077 	fdrop(fp, td);
3078 	return (error);
3079 }
3080 
3081 /*
3082  * Common implementation code for utimes(), lutimes(), and futimes().
3083  */
3084 static int
3085 getutimes(usrtvp, tvpseg, tsp)
3086 	const struct timeval *usrtvp;
3087 	enum uio_seg tvpseg;
3088 	struct timespec *tsp;
3089 {
3090 	struct timeval tv[2];
3091 	const struct timeval *tvp;
3092 	int error;
3093 
3094 	if (usrtvp == NULL) {
3095 		vfs_timestamp(&tsp[0]);
3096 		tsp[1] = tsp[0];
3097 	} else {
3098 		if (tvpseg == UIO_SYSSPACE) {
3099 			tvp = usrtvp;
3100 		} else {
3101 			if ((error = copyin(usrtvp, tv, sizeof(tv))) != 0)
3102 				return (error);
3103 			tvp = tv;
3104 		}
3105 
3106 		if (tvp[0].tv_usec < 0 || tvp[0].tv_usec >= 1000000 ||
3107 		    tvp[1].tv_usec < 0 || tvp[1].tv_usec >= 1000000)
3108 			return (EINVAL);
3109 		TIMEVAL_TO_TIMESPEC(&tvp[0], &tsp[0]);
3110 		TIMEVAL_TO_TIMESPEC(&tvp[1], &tsp[1]);
3111 	}
3112 	return (0);
3113 }
3114 
3115 /*
3116  * Common implementation code for utimes(), lutimes(), and futimes().
3117  */
3118 static int
3119 setutimes(td, vp, ts, numtimes, nullflag)
3120 	struct thread *td;
3121 	struct vnode *vp;
3122 	const struct timespec *ts;
3123 	int numtimes;
3124 	int nullflag;
3125 {
3126 	struct mount *mp;
3127 	struct vattr vattr;
3128 	int error, setbirthtime;
3129 
3130 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
3131 		return (error);
3132 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3133 	setbirthtime = 0;
3134 	if (numtimes < 3 && !VOP_GETATTR(vp, &vattr, td->td_ucred) &&
3135 	    timespeccmp(&ts[1], &vattr.va_birthtime, < ))
3136 		setbirthtime = 1;
3137 	VATTR_NULL(&vattr);
3138 	vattr.va_atime = ts[0];
3139 	vattr.va_mtime = ts[1];
3140 	if (setbirthtime)
3141 		vattr.va_birthtime = ts[1];
3142 	if (numtimes > 2)
3143 		vattr.va_birthtime = ts[2];
3144 	if (nullflag)
3145 		vattr.va_vaflags |= VA_UTIMES_NULL;
3146 #ifdef MAC
3147 	error = mac_vnode_check_setutimes(td->td_ucred, vp, vattr.va_atime,
3148 	    vattr.va_mtime);
3149 #endif
3150 	if (error == 0)
3151 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3152 	VOP_UNLOCK(vp, 0);
3153 	vn_finished_write(mp);
3154 	return (error);
3155 }
3156 
3157 /*
3158  * Set the access and modification times of a file.
3159  */
3160 #ifndef _SYS_SYSPROTO_H_
3161 struct utimes_args {
3162 	char	*path;
3163 	struct	timeval *tptr;
3164 };
3165 #endif
3166 int
3167 sys_utimes(td, uap)
3168 	struct thread *td;
3169 	register struct utimes_args /* {
3170 		char *path;
3171 		struct timeval *tptr;
3172 	} */ *uap;
3173 {
3174 
3175 	return (kern_utimes(td, uap->path, UIO_USERSPACE, uap->tptr,
3176 	    UIO_USERSPACE));
3177 }
3178 
3179 #ifndef _SYS_SYSPROTO_H_
3180 struct futimesat_args {
3181 	int fd;
3182 	const char * path;
3183 	const struct timeval * times;
3184 };
3185 #endif
3186 int
3187 sys_futimesat(struct thread *td, struct futimesat_args *uap)
3188 {
3189 
3190 	return (kern_utimesat(td, uap->fd, uap->path, UIO_USERSPACE,
3191 	    uap->times, UIO_USERSPACE));
3192 }
3193 
3194 int
3195 kern_utimes(struct thread *td, char *path, enum uio_seg pathseg,
3196     struct timeval *tptr, enum uio_seg tptrseg)
3197 {
3198 
3199 	return (kern_utimesat(td, AT_FDCWD, path, pathseg, tptr, tptrseg));
3200 }
3201 
3202 int
3203 kern_utimesat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
3204     struct timeval *tptr, enum uio_seg tptrseg)
3205 {
3206 	struct nameidata nd;
3207 	struct timespec ts[2];
3208 	cap_rights_t rights;
3209 	int error;
3210 
3211 	if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3212 		return (error);
3213 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, fd,
3214 	    cap_rights_init(&rights, CAP_FUTIMES), td);
3215 
3216 	if ((error = namei(&nd)) != 0)
3217 		return (error);
3218 	NDFREE(&nd, NDF_ONLY_PNBUF);
3219 	error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3220 	vrele(nd.ni_vp);
3221 	return (error);
3222 }
3223 
3224 /*
3225  * Set the access and modification times of a file.
3226  */
3227 #ifndef _SYS_SYSPROTO_H_
3228 struct lutimes_args {
3229 	char	*path;
3230 	struct	timeval *tptr;
3231 };
3232 #endif
3233 int
3234 sys_lutimes(td, uap)
3235 	struct thread *td;
3236 	register struct lutimes_args /* {
3237 		char *path;
3238 		struct timeval *tptr;
3239 	} */ *uap;
3240 {
3241 
3242 	return (kern_lutimes(td, uap->path, UIO_USERSPACE, uap->tptr,
3243 	    UIO_USERSPACE));
3244 }
3245 
3246 int
3247 kern_lutimes(struct thread *td, char *path, enum uio_seg pathseg,
3248     struct timeval *tptr, enum uio_seg tptrseg)
3249 {
3250 	struct timespec ts[2];
3251 	struct nameidata nd;
3252 	int error;
3253 
3254 	if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3255 		return (error);
3256 	NDINIT(&nd, LOOKUP, NOFOLLOW | AUDITVNODE1, pathseg, path, td);
3257 	if ((error = namei(&nd)) != 0)
3258 		return (error);
3259 	NDFREE(&nd, NDF_ONLY_PNBUF);
3260 	error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3261 	vrele(nd.ni_vp);
3262 	return (error);
3263 }
3264 
3265 /*
3266  * Set the access and modification times of a file.
3267  */
3268 #ifndef _SYS_SYSPROTO_H_
3269 struct futimes_args {
3270 	int	fd;
3271 	struct	timeval *tptr;
3272 };
3273 #endif
3274 int
3275 sys_futimes(td, uap)
3276 	struct thread *td;
3277 	register struct futimes_args /* {
3278 		int  fd;
3279 		struct timeval *tptr;
3280 	} */ *uap;
3281 {
3282 
3283 	return (kern_futimes(td, uap->fd, uap->tptr, UIO_USERSPACE));
3284 }
3285 
3286 int
3287 kern_futimes(struct thread *td, int fd, struct timeval *tptr,
3288     enum uio_seg tptrseg)
3289 {
3290 	struct timespec ts[2];
3291 	struct file *fp;
3292 	cap_rights_t rights;
3293 	int error;
3294 
3295 	AUDIT_ARG_FD(fd);
3296 	error = getutimes(tptr, tptrseg, ts);
3297 	if (error != 0)
3298 		return (error);
3299 	error = getvnode(td->td_proc->p_fd, fd,
3300 	    cap_rights_init(&rights, CAP_FUTIMES), &fp);
3301 	if (error != 0)
3302 		return (error);
3303 #ifdef AUDIT
3304 	vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
3305 	AUDIT_ARG_VNODE1(fp->f_vnode);
3306 	VOP_UNLOCK(fp->f_vnode, 0);
3307 #endif
3308 	error = setutimes(td, fp->f_vnode, ts, 2, tptr == NULL);
3309 	fdrop(fp, td);
3310 	return (error);
3311 }
3312 
3313 /*
3314  * Truncate a file given its path name.
3315  */
3316 #ifndef _SYS_SYSPROTO_H_
3317 struct truncate_args {
3318 	char	*path;
3319 	int	pad;
3320 	off_t	length;
3321 };
3322 #endif
3323 int
3324 sys_truncate(td, uap)
3325 	struct thread *td;
3326 	register struct truncate_args /* {
3327 		char *path;
3328 		int pad;
3329 		off_t length;
3330 	} */ *uap;
3331 {
3332 
3333 	return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3334 }
3335 
3336 int
3337 kern_truncate(struct thread *td, char *path, enum uio_seg pathseg, off_t length)
3338 {
3339 	struct mount *mp;
3340 	struct vnode *vp;
3341 	void *rl_cookie;
3342 	struct vattr vattr;
3343 	struct nameidata nd;
3344 	int error;
3345 
3346 	if (length < 0)
3347 		return(EINVAL);
3348 	NDINIT(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, td);
3349 	if ((error = namei(&nd)) != 0)
3350 		return (error);
3351 	vp = nd.ni_vp;
3352 	rl_cookie = vn_rangelock_wlock(vp, 0, OFF_MAX);
3353 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) {
3354 		vn_rangelock_unlock(vp, rl_cookie);
3355 		vrele(vp);
3356 		return (error);
3357 	}
3358 	NDFREE(&nd, NDF_ONLY_PNBUF);
3359 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3360 	if (vp->v_type == VDIR)
3361 		error = EISDIR;
3362 #ifdef MAC
3363 	else if ((error = mac_vnode_check_write(td->td_ucred, NOCRED, vp))) {
3364 	}
3365 #endif
3366 	else if ((error = vn_writechk(vp)) == 0 &&
3367 	    (error = VOP_ACCESS(vp, VWRITE, td->td_ucred, td)) == 0) {
3368 		VATTR_NULL(&vattr);
3369 		vattr.va_size = length;
3370 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3371 	}
3372 	VOP_UNLOCK(vp, 0);
3373 	vn_finished_write(mp);
3374 	vn_rangelock_unlock(vp, rl_cookie);
3375 	vrele(vp);
3376 	return (error);
3377 }
3378 
3379 #if defined(COMPAT_43)
3380 /*
3381  * Truncate a file given its path name.
3382  */
3383 #ifndef _SYS_SYSPROTO_H_
3384 struct otruncate_args {
3385 	char	*path;
3386 	long	length;
3387 };
3388 #endif
3389 int
3390 otruncate(td, uap)
3391 	struct thread *td;
3392 	register struct otruncate_args /* {
3393 		char *path;
3394 		long length;
3395 	} */ *uap;
3396 {
3397 	struct truncate_args /* {
3398 		char *path;
3399 		int pad;
3400 		off_t length;
3401 	} */ nuap;
3402 
3403 	nuap.path = uap->path;
3404 	nuap.length = uap->length;
3405 	return (sys_truncate(td, &nuap));
3406 }
3407 #endif /* COMPAT_43 */
3408 
3409 /* Versions with the pad argument */
3410 int
3411 freebsd6_truncate(struct thread *td, struct freebsd6_truncate_args *uap)
3412 {
3413 	struct truncate_args ouap;
3414 
3415 	ouap.path = uap->path;
3416 	ouap.length = uap->length;
3417 	return (sys_truncate(td, &ouap));
3418 }
3419 
3420 int
3421 freebsd6_ftruncate(struct thread *td, struct freebsd6_ftruncate_args *uap)
3422 {
3423 	struct ftruncate_args ouap;
3424 
3425 	ouap.fd = uap->fd;
3426 	ouap.length = uap->length;
3427 	return (sys_ftruncate(td, &ouap));
3428 }
3429 
3430 /*
3431  * Sync an open file.
3432  */
3433 #ifndef _SYS_SYSPROTO_H_
3434 struct fsync_args {
3435 	int	fd;
3436 };
3437 #endif
3438 int
3439 sys_fsync(td, uap)
3440 	struct thread *td;
3441 	struct fsync_args /* {
3442 		int fd;
3443 	} */ *uap;
3444 {
3445 	struct vnode *vp;
3446 	struct mount *mp;
3447 	struct file *fp;
3448 	cap_rights_t rights;
3449 	int error, lock_flags;
3450 
3451 	AUDIT_ARG_FD(uap->fd);
3452 	error = getvnode(td->td_proc->p_fd, uap->fd,
3453 	    cap_rights_init(&rights, CAP_FSYNC), &fp);
3454 	if (error != 0)
3455 		return (error);
3456 	vp = fp->f_vnode;
3457 	error = vn_start_write(vp, &mp, V_WAIT | PCATCH);
3458 	if (error != 0)
3459 		goto drop;
3460 	if (MNT_SHARED_WRITES(mp) ||
3461 	    ((mp == NULL) && MNT_SHARED_WRITES(vp->v_mount))) {
3462 		lock_flags = LK_SHARED;
3463 	} else {
3464 		lock_flags = LK_EXCLUSIVE;
3465 	}
3466 	vn_lock(vp, lock_flags | LK_RETRY);
3467 	AUDIT_ARG_VNODE1(vp);
3468 	if (vp->v_object != NULL) {
3469 		VM_OBJECT_WLOCK(vp->v_object);
3470 		vm_object_page_clean(vp->v_object, 0, 0, 0);
3471 		VM_OBJECT_WUNLOCK(vp->v_object);
3472 	}
3473 	error = VOP_FSYNC(vp, MNT_WAIT, td);
3474 
3475 	VOP_UNLOCK(vp, 0);
3476 	vn_finished_write(mp);
3477 drop:
3478 	fdrop(fp, td);
3479 	return (error);
3480 }
3481 
3482 /*
3483  * Rename files.  Source and destination must either both be directories, or
3484  * both not be directories.  If target is a directory, it must be empty.
3485  */
3486 #ifndef _SYS_SYSPROTO_H_
3487 struct rename_args {
3488 	char	*from;
3489 	char	*to;
3490 };
3491 #endif
3492 int
3493 sys_rename(td, uap)
3494 	struct thread *td;
3495 	register struct rename_args /* {
3496 		char *from;
3497 		char *to;
3498 	} */ *uap;
3499 {
3500 
3501 	return (kern_rename(td, uap->from, uap->to, UIO_USERSPACE));
3502 }
3503 
3504 #ifndef _SYS_SYSPROTO_H_
3505 struct renameat_args {
3506 	int	oldfd;
3507 	char	*old;
3508 	int	newfd;
3509 	char	*new;
3510 };
3511 #endif
3512 int
3513 sys_renameat(struct thread *td, struct renameat_args *uap)
3514 {
3515 
3516 	return (kern_renameat(td, uap->oldfd, uap->old, uap->newfd, uap->new,
3517 	    UIO_USERSPACE));
3518 }
3519 
3520 int
3521 kern_rename(struct thread *td, char *from, char *to, enum uio_seg pathseg)
3522 {
3523 
3524 	return (kern_renameat(td, AT_FDCWD, from, AT_FDCWD, to, pathseg));
3525 }
3526 
3527 int
3528 kern_renameat(struct thread *td, int oldfd, char *old, int newfd, char *new,
3529     enum uio_seg pathseg)
3530 {
3531 	struct mount *mp = NULL;
3532 	struct vnode *tvp, *fvp, *tdvp;
3533 	struct nameidata fromnd, tond;
3534 	cap_rights_t rights;
3535 	int error;
3536 
3537 again:
3538 	bwillwrite();
3539 #ifdef MAC
3540 	NDINIT_ATRIGHTS(&fromnd, DELETE, LOCKPARENT | LOCKLEAF | SAVESTART |
3541 	    AUDITVNODE1, pathseg, old, oldfd,
3542 	    cap_rights_init(&rights, CAP_RENAMEAT), td);
3543 #else
3544 	NDINIT_ATRIGHTS(&fromnd, DELETE, WANTPARENT | SAVESTART | AUDITVNODE1,
3545 	    pathseg, old, oldfd, cap_rights_init(&rights, CAP_RENAMEAT), td);
3546 #endif
3547 
3548 	if ((error = namei(&fromnd)) != 0)
3549 		return (error);
3550 #ifdef MAC
3551 	error = mac_vnode_check_rename_from(td->td_ucred, fromnd.ni_dvp,
3552 	    fromnd.ni_vp, &fromnd.ni_cnd);
3553 	VOP_UNLOCK(fromnd.ni_dvp, 0);
3554 	if (fromnd.ni_dvp != fromnd.ni_vp)
3555 		VOP_UNLOCK(fromnd.ni_vp, 0);
3556 #endif
3557 	fvp = fromnd.ni_vp;
3558 	NDINIT_ATRIGHTS(&tond, RENAME, LOCKPARENT | LOCKLEAF | NOCACHE |
3559 	    SAVESTART | AUDITVNODE2, pathseg, new, newfd,
3560 	    cap_rights_init(&rights, CAP_LINKAT), td);
3561 	if (fromnd.ni_vp->v_type == VDIR)
3562 		tond.ni_cnd.cn_flags |= WILLBEDIR;
3563 	if ((error = namei(&tond)) != 0) {
3564 		/* Translate error code for rename("dir1", "dir2/."). */
3565 		if (error == EISDIR && fvp->v_type == VDIR)
3566 			error = EINVAL;
3567 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3568 		vrele(fromnd.ni_dvp);
3569 		vrele(fvp);
3570 		goto out1;
3571 	}
3572 	tdvp = tond.ni_dvp;
3573 	tvp = tond.ni_vp;
3574 	error = vn_start_write(fvp, &mp, V_NOWAIT);
3575 	if (error != 0) {
3576 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3577 		NDFREE(&tond, NDF_ONLY_PNBUF);
3578 		if (tvp != NULL)
3579 			vput(tvp);
3580 		if (tdvp == tvp)
3581 			vrele(tdvp);
3582 		else
3583 			vput(tdvp);
3584 		vrele(fromnd.ni_dvp);
3585 		vrele(fvp);
3586 		vrele(tond.ni_startdir);
3587 		if (fromnd.ni_startdir != NULL)
3588 			vrele(fromnd.ni_startdir);
3589 		error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH);
3590 		if (error != 0)
3591 			return (error);
3592 		goto again;
3593 	}
3594 	if (tvp != NULL) {
3595 		if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
3596 			error = ENOTDIR;
3597 			goto out;
3598 		} else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
3599 			error = EISDIR;
3600 			goto out;
3601 		}
3602 #ifdef CAPABILITIES
3603 		if (newfd != AT_FDCWD) {
3604 			/*
3605 			 * If the target already exists we require CAP_UNLINKAT
3606 			 * from 'newfd'.
3607 			 */
3608 			error = cap_check(&tond.ni_filecaps.fc_rights,
3609 			    cap_rights_init(&rights, CAP_UNLINKAT));
3610 			if (error != 0)
3611 				goto out;
3612 		}
3613 #endif
3614 	}
3615 	if (fvp == tdvp) {
3616 		error = EINVAL;
3617 		goto out;
3618 	}
3619 	/*
3620 	 * If the source is the same as the destination (that is, if they
3621 	 * are links to the same vnode), then there is nothing to do.
3622 	 */
3623 	if (fvp == tvp)
3624 		error = -1;
3625 #ifdef MAC
3626 	else
3627 		error = mac_vnode_check_rename_to(td->td_ucred, tdvp,
3628 		    tond.ni_vp, fromnd.ni_dvp == tdvp, &tond.ni_cnd);
3629 #endif
3630 out:
3631 	if (error == 0) {
3632 		error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
3633 		    tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
3634 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3635 		NDFREE(&tond, NDF_ONLY_PNBUF);
3636 	} else {
3637 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3638 		NDFREE(&tond, NDF_ONLY_PNBUF);
3639 		if (tvp != NULL)
3640 			vput(tvp);
3641 		if (tdvp == tvp)
3642 			vrele(tdvp);
3643 		else
3644 			vput(tdvp);
3645 		vrele(fromnd.ni_dvp);
3646 		vrele(fvp);
3647 	}
3648 	vrele(tond.ni_startdir);
3649 	vn_finished_write(mp);
3650 out1:
3651 	if (fromnd.ni_startdir)
3652 		vrele(fromnd.ni_startdir);
3653 	if (error == -1)
3654 		return (0);
3655 	return (error);
3656 }
3657 
3658 /*
3659  * Make a directory file.
3660  */
3661 #ifndef _SYS_SYSPROTO_H_
3662 struct mkdir_args {
3663 	char	*path;
3664 	int	mode;
3665 };
3666 #endif
3667 int
3668 sys_mkdir(td, uap)
3669 	struct thread *td;
3670 	register struct mkdir_args /* {
3671 		char *path;
3672 		int mode;
3673 	} */ *uap;
3674 {
3675 
3676 	return (kern_mkdir(td, uap->path, UIO_USERSPACE, uap->mode));
3677 }
3678 
3679 #ifndef _SYS_SYSPROTO_H_
3680 struct mkdirat_args {
3681 	int	fd;
3682 	char	*path;
3683 	mode_t	mode;
3684 };
3685 #endif
3686 int
3687 sys_mkdirat(struct thread *td, struct mkdirat_args *uap)
3688 {
3689 
3690 	return (kern_mkdirat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode));
3691 }
3692 
3693 int
3694 kern_mkdir(struct thread *td, char *path, enum uio_seg segflg, int mode)
3695 {
3696 
3697 	return (kern_mkdirat(td, AT_FDCWD, path, segflg, mode));
3698 }
3699 
3700 int
3701 kern_mkdirat(struct thread *td, int fd, char *path, enum uio_seg segflg,
3702     int mode)
3703 {
3704 	struct mount *mp;
3705 	struct vnode *vp;
3706 	struct vattr vattr;
3707 	struct nameidata nd;
3708 	cap_rights_t rights;
3709 	int error;
3710 
3711 	AUDIT_ARG_MODE(mode);
3712 restart:
3713 	bwillwrite();
3714 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1,
3715 	    segflg, path, fd, cap_rights_init(&rights, CAP_MKDIRAT), td);
3716 	nd.ni_cnd.cn_flags |= WILLBEDIR;
3717 	if ((error = namei(&nd)) != 0)
3718 		return (error);
3719 	vp = nd.ni_vp;
3720 	if (vp != NULL) {
3721 		NDFREE(&nd, NDF_ONLY_PNBUF);
3722 		/*
3723 		 * XXX namei called with LOCKPARENT but not LOCKLEAF has
3724 		 * the strange behaviour of leaving the vnode unlocked
3725 		 * if the target is the same vnode as the parent.
3726 		 */
3727 		if (vp == nd.ni_dvp)
3728 			vrele(nd.ni_dvp);
3729 		else
3730 			vput(nd.ni_dvp);
3731 		vrele(vp);
3732 		return (EEXIST);
3733 	}
3734 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3735 		NDFREE(&nd, NDF_ONLY_PNBUF);
3736 		vput(nd.ni_dvp);
3737 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3738 			return (error);
3739 		goto restart;
3740 	}
3741 	VATTR_NULL(&vattr);
3742 	vattr.va_type = VDIR;
3743 	vattr.va_mode = (mode & ACCESSPERMS) &~ td->td_proc->p_fd->fd_cmask;
3744 #ifdef MAC
3745 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
3746 	    &vattr);
3747 	if (error != 0)
3748 		goto out;
3749 #endif
3750 	error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
3751 #ifdef MAC
3752 out:
3753 #endif
3754 	NDFREE(&nd, NDF_ONLY_PNBUF);
3755 	vput(nd.ni_dvp);
3756 	if (error == 0)
3757 		vput(nd.ni_vp);
3758 	vn_finished_write(mp);
3759 	return (error);
3760 }
3761 
3762 /*
3763  * Remove a directory file.
3764  */
3765 #ifndef _SYS_SYSPROTO_H_
3766 struct rmdir_args {
3767 	char	*path;
3768 };
3769 #endif
3770 int
3771 sys_rmdir(td, uap)
3772 	struct thread *td;
3773 	struct rmdir_args /* {
3774 		char *path;
3775 	} */ *uap;
3776 {
3777 
3778 	return (kern_rmdir(td, uap->path, UIO_USERSPACE));
3779 }
3780 
3781 int
3782 kern_rmdir(struct thread *td, char *path, enum uio_seg pathseg)
3783 {
3784 
3785 	return (kern_rmdirat(td, AT_FDCWD, path, pathseg));
3786 }
3787 
3788 int
3789 kern_rmdirat(struct thread *td, int fd, char *path, enum uio_seg pathseg)
3790 {
3791 	struct mount *mp;
3792 	struct vnode *vp;
3793 	struct nameidata nd;
3794 	cap_rights_t rights;
3795 	int error;
3796 
3797 restart:
3798 	bwillwrite();
3799 	NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1,
3800 	    pathseg, path, fd, cap_rights_init(&rights, CAP_UNLINKAT), td);
3801 	if ((error = namei(&nd)) != 0)
3802 		return (error);
3803 	vp = nd.ni_vp;
3804 	if (vp->v_type != VDIR) {
3805 		error = ENOTDIR;
3806 		goto out;
3807 	}
3808 	/*
3809 	 * No rmdir "." please.
3810 	 */
3811 	if (nd.ni_dvp == vp) {
3812 		error = EINVAL;
3813 		goto out;
3814 	}
3815 	/*
3816 	 * The root of a mounted filesystem cannot be deleted.
3817 	 */
3818 	if (vp->v_vflag & VV_ROOT) {
3819 		error = EBUSY;
3820 		goto out;
3821 	}
3822 #ifdef MAC
3823 	error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
3824 	    &nd.ni_cnd);
3825 	if (error != 0)
3826 		goto out;
3827 #endif
3828 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3829 		NDFREE(&nd, NDF_ONLY_PNBUF);
3830 		vput(vp);
3831 		if (nd.ni_dvp == vp)
3832 			vrele(nd.ni_dvp);
3833 		else
3834 			vput(nd.ni_dvp);
3835 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3836 			return (error);
3837 		goto restart;
3838 	}
3839 	vfs_notify_upper(vp, VFS_NOTIFY_UPPER_UNLINK);
3840 	error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
3841 	vn_finished_write(mp);
3842 out:
3843 	NDFREE(&nd, NDF_ONLY_PNBUF);
3844 	vput(vp);
3845 	if (nd.ni_dvp == vp)
3846 		vrele(nd.ni_dvp);
3847 	else
3848 		vput(nd.ni_dvp);
3849 	return (error);
3850 }
3851 
3852 #ifdef COMPAT_43
3853 /*
3854  * Read a block of directory entries in a filesystem independent format.
3855  */
3856 #ifndef _SYS_SYSPROTO_H_
3857 struct ogetdirentries_args {
3858 	int	fd;
3859 	char	*buf;
3860 	u_int	count;
3861 	long	*basep;
3862 };
3863 #endif
3864 int
3865 ogetdirentries(struct thread *td, struct ogetdirentries_args *uap)
3866 {
3867 	long loff;
3868 	int error;
3869 
3870 	error = kern_ogetdirentries(td, uap, &loff);
3871 	if (error == 0)
3872 		error = copyout(&loff, uap->basep, sizeof(long));
3873 	return (error);
3874 }
3875 
3876 int
3877 kern_ogetdirentries(struct thread *td, struct ogetdirentries_args *uap,
3878     long *ploff)
3879 {
3880 	struct vnode *vp;
3881 	struct file *fp;
3882 	struct uio auio, kuio;
3883 	struct iovec aiov, kiov;
3884 	struct dirent *dp, *edp;
3885 	cap_rights_t rights;
3886 	caddr_t dirbuf;
3887 	int error, eofflag, readcnt;
3888 	long loff;
3889 	off_t foffset;
3890 
3891 	/* XXX arbitrary sanity limit on `count'. */
3892 	if (uap->count > 64 * 1024)
3893 		return (EINVAL);
3894 	error = getvnode(td->td_proc->p_fd, uap->fd,
3895 	    cap_rights_init(&rights, CAP_READ), &fp);
3896 	if (error != 0)
3897 		return (error);
3898 	if ((fp->f_flag & FREAD) == 0) {
3899 		fdrop(fp, td);
3900 		return (EBADF);
3901 	}
3902 	vp = fp->f_vnode;
3903 	foffset = foffset_lock(fp, 0);
3904 unionread:
3905 	if (vp->v_type != VDIR) {
3906 		foffset_unlock(fp, foffset, 0);
3907 		fdrop(fp, td);
3908 		return (EINVAL);
3909 	}
3910 	aiov.iov_base = uap->buf;
3911 	aiov.iov_len = uap->count;
3912 	auio.uio_iov = &aiov;
3913 	auio.uio_iovcnt = 1;
3914 	auio.uio_rw = UIO_READ;
3915 	auio.uio_segflg = UIO_USERSPACE;
3916 	auio.uio_td = td;
3917 	auio.uio_resid = uap->count;
3918 	vn_lock(vp, LK_SHARED | LK_RETRY);
3919 	loff = auio.uio_offset = foffset;
3920 #ifdef MAC
3921 	error = mac_vnode_check_readdir(td->td_ucred, vp);
3922 	if (error != 0) {
3923 		VOP_UNLOCK(vp, 0);
3924 		foffset_unlock(fp, foffset, FOF_NOUPDATE);
3925 		fdrop(fp, td);
3926 		return (error);
3927 	}
3928 #endif
3929 #	if (BYTE_ORDER != LITTLE_ENDIAN)
3930 		if (vp->v_mount->mnt_maxsymlinklen <= 0) {
3931 			error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag,
3932 			    NULL, NULL);
3933 			foffset = auio.uio_offset;
3934 		} else
3935 #	endif
3936 	{
3937 		kuio = auio;
3938 		kuio.uio_iov = &kiov;
3939 		kuio.uio_segflg = UIO_SYSSPACE;
3940 		kiov.iov_len = uap->count;
3941 		dirbuf = malloc(uap->count, M_TEMP, M_WAITOK);
3942 		kiov.iov_base = dirbuf;
3943 		error = VOP_READDIR(vp, &kuio, fp->f_cred, &eofflag,
3944 			    NULL, NULL);
3945 		foffset = kuio.uio_offset;
3946 		if (error == 0) {
3947 			readcnt = uap->count - kuio.uio_resid;
3948 			edp = (struct dirent *)&dirbuf[readcnt];
3949 			for (dp = (struct dirent *)dirbuf; dp < edp; ) {
3950 #				if (BYTE_ORDER == LITTLE_ENDIAN)
3951 					/*
3952 					 * The expected low byte of
3953 					 * dp->d_namlen is our dp->d_type.
3954 					 * The high MBZ byte of dp->d_namlen
3955 					 * is our dp->d_namlen.
3956 					 */
3957 					dp->d_type = dp->d_namlen;
3958 					dp->d_namlen = 0;
3959 #				else
3960 					/*
3961 					 * The dp->d_type is the high byte
3962 					 * of the expected dp->d_namlen,
3963 					 * so must be zero'ed.
3964 					 */
3965 					dp->d_type = 0;
3966 #				endif
3967 				if (dp->d_reclen > 0) {
3968 					dp = (struct dirent *)
3969 					    ((char *)dp + dp->d_reclen);
3970 				} else {
3971 					error = EIO;
3972 					break;
3973 				}
3974 			}
3975 			if (dp >= edp)
3976 				error = uiomove(dirbuf, readcnt, &auio);
3977 		}
3978 		free(dirbuf, M_TEMP);
3979 	}
3980 	if (error != 0) {
3981 		VOP_UNLOCK(vp, 0);
3982 		foffset_unlock(fp, foffset, 0);
3983 		fdrop(fp, td);
3984 		return (error);
3985 	}
3986 	if (uap->count == auio.uio_resid &&
3987 	    (vp->v_vflag & VV_ROOT) &&
3988 	    (vp->v_mount->mnt_flag & MNT_UNION)) {
3989 		struct vnode *tvp = vp;
3990 		vp = vp->v_mount->mnt_vnodecovered;
3991 		VREF(vp);
3992 		fp->f_vnode = vp;
3993 		fp->f_data = vp;
3994 		foffset = 0;
3995 		vput(tvp);
3996 		goto unionread;
3997 	}
3998 	VOP_UNLOCK(vp, 0);
3999 	foffset_unlock(fp, foffset, 0);
4000 	fdrop(fp, td);
4001 	td->td_retval[0] = uap->count - auio.uio_resid;
4002 	if (error == 0)
4003 		*ploff = loff;
4004 	return (error);
4005 }
4006 #endif /* COMPAT_43 */
4007 
4008 /*
4009  * Read a block of directory entries in a filesystem independent format.
4010  */
4011 #ifndef _SYS_SYSPROTO_H_
4012 struct getdirentries_args {
4013 	int	fd;
4014 	char	*buf;
4015 	u_int	count;
4016 	long	*basep;
4017 };
4018 #endif
4019 int
4020 sys_getdirentries(td, uap)
4021 	struct thread *td;
4022 	register struct getdirentries_args /* {
4023 		int fd;
4024 		char *buf;
4025 		u_int count;
4026 		long *basep;
4027 	} */ *uap;
4028 {
4029 	long base;
4030 	int error;
4031 
4032 	error = kern_getdirentries(td, uap->fd, uap->buf, uap->count, &base,
4033 	    NULL, UIO_USERSPACE);
4034 	if (error != 0)
4035 		return (error);
4036 	if (uap->basep != NULL)
4037 		error = copyout(&base, uap->basep, sizeof(long));
4038 	return (error);
4039 }
4040 
4041 int
4042 kern_getdirentries(struct thread *td, int fd, char *buf, u_int count,
4043     long *basep, ssize_t *residp, enum uio_seg bufseg)
4044 {
4045 	struct vnode *vp;
4046 	struct file *fp;
4047 	struct uio auio;
4048 	struct iovec aiov;
4049 	cap_rights_t rights;
4050 	long loff;
4051 	int error, eofflag;
4052 	off_t foffset;
4053 
4054 	AUDIT_ARG_FD(fd);
4055 	if (count > IOSIZE_MAX)
4056 		return (EINVAL);
4057 	auio.uio_resid = count;
4058 	error = getvnode(td->td_proc->p_fd, fd,
4059 	    cap_rights_init(&rights, CAP_READ), &fp);
4060 	if (error != 0)
4061 		return (error);
4062 	if ((fp->f_flag & FREAD) == 0) {
4063 		fdrop(fp, td);
4064 		return (EBADF);
4065 	}
4066 	vp = fp->f_vnode;
4067 	foffset = foffset_lock(fp, 0);
4068 unionread:
4069 	if (vp->v_type != VDIR) {
4070 		error = EINVAL;
4071 		goto fail;
4072 	}
4073 	aiov.iov_base = buf;
4074 	aiov.iov_len = count;
4075 	auio.uio_iov = &aiov;
4076 	auio.uio_iovcnt = 1;
4077 	auio.uio_rw = UIO_READ;
4078 	auio.uio_segflg = bufseg;
4079 	auio.uio_td = td;
4080 	vn_lock(vp, LK_SHARED | LK_RETRY);
4081 	AUDIT_ARG_VNODE1(vp);
4082 	loff = auio.uio_offset = foffset;
4083 #ifdef MAC
4084 	error = mac_vnode_check_readdir(td->td_ucred, vp);
4085 	if (error == 0)
4086 #endif
4087 		error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, NULL,
4088 		    NULL);
4089 	foffset = auio.uio_offset;
4090 	if (error != 0) {
4091 		VOP_UNLOCK(vp, 0);
4092 		goto fail;
4093 	}
4094 	if (count == auio.uio_resid &&
4095 	    (vp->v_vflag & VV_ROOT) &&
4096 	    (vp->v_mount->mnt_flag & MNT_UNION)) {
4097 		struct vnode *tvp = vp;
4098 
4099 		vp = vp->v_mount->mnt_vnodecovered;
4100 		VREF(vp);
4101 		fp->f_vnode = vp;
4102 		fp->f_data = vp;
4103 		foffset = 0;
4104 		vput(tvp);
4105 		goto unionread;
4106 	}
4107 	VOP_UNLOCK(vp, 0);
4108 	*basep = loff;
4109 	if (residp != NULL)
4110 		*residp = auio.uio_resid;
4111 	td->td_retval[0] = count - auio.uio_resid;
4112 fail:
4113 	foffset_unlock(fp, foffset, 0);
4114 	fdrop(fp, td);
4115 	return (error);
4116 }
4117 
4118 #ifndef _SYS_SYSPROTO_H_
4119 struct getdents_args {
4120 	int fd;
4121 	char *buf;
4122 	size_t count;
4123 };
4124 #endif
4125 int
4126 sys_getdents(td, uap)
4127 	struct thread *td;
4128 	register struct getdents_args /* {
4129 		int fd;
4130 		char *buf;
4131 		u_int count;
4132 	} */ *uap;
4133 {
4134 	struct getdirentries_args ap;
4135 
4136 	ap.fd = uap->fd;
4137 	ap.buf = uap->buf;
4138 	ap.count = uap->count;
4139 	ap.basep = NULL;
4140 	return (sys_getdirentries(td, &ap));
4141 }
4142 
4143 /*
4144  * Set the mode mask for creation of filesystem nodes.
4145  */
4146 #ifndef _SYS_SYSPROTO_H_
4147 struct umask_args {
4148 	int	newmask;
4149 };
4150 #endif
4151 int
4152 sys_umask(td, uap)
4153 	struct thread *td;
4154 	struct umask_args /* {
4155 		int newmask;
4156 	} */ *uap;
4157 {
4158 	register struct filedesc *fdp;
4159 
4160 	FILEDESC_XLOCK(td->td_proc->p_fd);
4161 	fdp = td->td_proc->p_fd;
4162 	td->td_retval[0] = fdp->fd_cmask;
4163 	fdp->fd_cmask = uap->newmask & ALLPERMS;
4164 	FILEDESC_XUNLOCK(td->td_proc->p_fd);
4165 	return (0);
4166 }
4167 
4168 /*
4169  * Void all references to file by ripping underlying filesystem away from
4170  * vnode.
4171  */
4172 #ifndef _SYS_SYSPROTO_H_
4173 struct revoke_args {
4174 	char	*path;
4175 };
4176 #endif
4177 int
4178 sys_revoke(td, uap)
4179 	struct thread *td;
4180 	register struct revoke_args /* {
4181 		char *path;
4182 	} */ *uap;
4183 {
4184 	struct vnode *vp;
4185 	struct vattr vattr;
4186 	struct nameidata nd;
4187 	int error;
4188 
4189 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
4190 	    uap->path, td);
4191 	if ((error = namei(&nd)) != 0)
4192 		return (error);
4193 	vp = nd.ni_vp;
4194 	NDFREE(&nd, NDF_ONLY_PNBUF);
4195 	if (vp->v_type != VCHR || vp->v_rdev == NULL) {
4196 		error = EINVAL;
4197 		goto out;
4198 	}
4199 #ifdef MAC
4200 	error = mac_vnode_check_revoke(td->td_ucred, vp);
4201 	if (error != 0)
4202 		goto out;
4203 #endif
4204 	error = VOP_GETATTR(vp, &vattr, td->td_ucred);
4205 	if (error != 0)
4206 		goto out;
4207 	if (td->td_ucred->cr_uid != vattr.va_uid) {
4208 		error = priv_check(td, PRIV_VFS_ADMIN);
4209 		if (error != 0)
4210 			goto out;
4211 	}
4212 	if (vcount(vp) > 1)
4213 		VOP_REVOKE(vp, REVOKEALL);
4214 out:
4215 	vput(vp);
4216 	return (error);
4217 }
4218 
4219 /*
4220  * Convert a user file descriptor to a kernel file entry and check that, if it
4221  * is a capability, the correct rights are present. A reference on the file
4222  * entry is held upon returning.
4223  */
4224 int
4225 getvnode(struct filedesc *fdp, int fd, cap_rights_t *rightsp, struct file **fpp)
4226 {
4227 	struct file *fp;
4228 	int error;
4229 
4230 	error = fget_unlocked(fdp, fd, rightsp, 0, &fp, NULL);
4231 	if (error != 0)
4232 		return (error);
4233 
4234 	/*
4235 	 * The file could be not of the vnode type, or it may be not
4236 	 * yet fully initialized, in which case the f_vnode pointer
4237 	 * may be set, but f_ops is still badfileops.  E.g.,
4238 	 * devfs_open() transiently create such situation to
4239 	 * facilitate csw d_fdopen().
4240 	 *
4241 	 * Dupfdopen() handling in kern_openat() installs the
4242 	 * half-baked file into the process descriptor table, allowing
4243 	 * other thread to dereference it. Guard against the race by
4244 	 * checking f_ops.
4245 	 */
4246 	if (fp->f_vnode == NULL || fp->f_ops == &badfileops) {
4247 		fdrop(fp, curthread);
4248 		return (EINVAL);
4249 	}
4250 	*fpp = fp;
4251 	return (0);
4252 }
4253 
4254 
4255 /*
4256  * Get an (NFS) file handle.
4257  */
4258 #ifndef _SYS_SYSPROTO_H_
4259 struct lgetfh_args {
4260 	char	*fname;
4261 	fhandle_t *fhp;
4262 };
4263 #endif
4264 int
4265 sys_lgetfh(td, uap)
4266 	struct thread *td;
4267 	register struct lgetfh_args *uap;
4268 {
4269 	struct nameidata nd;
4270 	fhandle_t fh;
4271 	register struct vnode *vp;
4272 	int error;
4273 
4274 	error = priv_check(td, PRIV_VFS_GETFH);
4275 	if (error != 0)
4276 		return (error);
4277 	NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
4278 	    uap->fname, td);
4279 	error = namei(&nd);
4280 	if (error != 0)
4281 		return (error);
4282 	NDFREE(&nd, NDF_ONLY_PNBUF);
4283 	vp = nd.ni_vp;
4284 	bzero(&fh, sizeof(fh));
4285 	fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
4286 	error = VOP_VPTOFH(vp, &fh.fh_fid);
4287 	vput(vp);
4288 	if (error == 0)
4289 		error = copyout(&fh, uap->fhp, sizeof (fh));
4290 	return (error);
4291 }
4292 
4293 #ifndef _SYS_SYSPROTO_H_
4294 struct getfh_args {
4295 	char	*fname;
4296 	fhandle_t *fhp;
4297 };
4298 #endif
4299 int
4300 sys_getfh(td, uap)
4301 	struct thread *td;
4302 	register struct getfh_args *uap;
4303 {
4304 	struct nameidata nd;
4305 	fhandle_t fh;
4306 	register struct vnode *vp;
4307 	int error;
4308 
4309 	error = priv_check(td, PRIV_VFS_GETFH);
4310 	if (error != 0)
4311 		return (error);
4312 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
4313 	    uap->fname, td);
4314 	error = namei(&nd);
4315 	if (error != 0)
4316 		return (error);
4317 	NDFREE(&nd, NDF_ONLY_PNBUF);
4318 	vp = nd.ni_vp;
4319 	bzero(&fh, sizeof(fh));
4320 	fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
4321 	error = VOP_VPTOFH(vp, &fh.fh_fid);
4322 	vput(vp);
4323 	if (error == 0)
4324 		error = copyout(&fh, uap->fhp, sizeof (fh));
4325 	return (error);
4326 }
4327 
4328 /*
4329  * syscall for the rpc.lockd to use to translate a NFS file handle into an
4330  * open descriptor.
4331  *
4332  * warning: do not remove the priv_check() call or this becomes one giant
4333  * security hole.
4334  */
4335 #ifndef _SYS_SYSPROTO_H_
4336 struct fhopen_args {
4337 	const struct fhandle *u_fhp;
4338 	int flags;
4339 };
4340 #endif
4341 int
4342 sys_fhopen(td, uap)
4343 	struct thread *td;
4344 	struct fhopen_args /* {
4345 		const struct fhandle *u_fhp;
4346 		int flags;
4347 	} */ *uap;
4348 {
4349 	struct mount *mp;
4350 	struct vnode *vp;
4351 	struct fhandle fhp;
4352 	struct file *fp;
4353 	int fmode, error;
4354 	int indx;
4355 
4356 	error = priv_check(td, PRIV_VFS_FHOPEN);
4357 	if (error != 0)
4358 		return (error);
4359 	indx = -1;
4360 	fmode = FFLAGS(uap->flags);
4361 	/* why not allow a non-read/write open for our lockd? */
4362 	if (((fmode & (FREAD | FWRITE)) == 0) || (fmode & O_CREAT))
4363 		return (EINVAL);
4364 	error = copyin(uap->u_fhp, &fhp, sizeof(fhp));
4365 	if (error != 0)
4366 		return(error);
4367 	/* find the mount point */
4368 	mp = vfs_busyfs(&fhp.fh_fsid);
4369 	if (mp == NULL)
4370 		return (ESTALE);
4371 	/* now give me my vnode, it gets returned to me locked */
4372 	error = VFS_FHTOVP(mp, &fhp.fh_fid, LK_EXCLUSIVE, &vp);
4373 	vfs_unbusy(mp);
4374 	if (error != 0)
4375 		return (error);
4376 
4377 	error = falloc_noinstall(td, &fp);
4378 	if (error != 0) {
4379 		vput(vp);
4380 		return (error);
4381 	}
4382 	/*
4383 	 * An extra reference on `fp' has been held for us by
4384 	 * falloc_noinstall().
4385 	 */
4386 
4387 #ifdef INVARIANTS
4388 	td->td_dupfd = -1;
4389 #endif
4390 	error = vn_open_vnode(vp, fmode, td->td_ucred, td, fp);
4391 	if (error != 0) {
4392 		KASSERT(fp->f_ops == &badfileops,
4393 		    ("VOP_OPEN in fhopen() set f_ops"));
4394 		KASSERT(td->td_dupfd < 0,
4395 		    ("fhopen() encountered fdopen()"));
4396 
4397 		vput(vp);
4398 		goto bad;
4399 	}
4400 #ifdef INVARIANTS
4401 	td->td_dupfd = 0;
4402 #endif
4403 	fp->f_vnode = vp;
4404 	fp->f_seqcount = 1;
4405 	finit(fp, (fmode & FMASK) | (fp->f_flag & FHASLOCK), DTYPE_VNODE, vp,
4406 	    &vnops);
4407 	VOP_UNLOCK(vp, 0);
4408 	if ((fmode & O_TRUNC) != 0) {
4409 		error = fo_truncate(fp, 0, td->td_ucred, td);
4410 		if (error != 0)
4411 			goto bad;
4412 	}
4413 
4414 	error = finstall(td, fp, &indx, fmode, NULL);
4415 bad:
4416 	fdrop(fp, td);
4417 	td->td_retval[0] = indx;
4418 	return (error);
4419 }
4420 
4421 /*
4422  * Stat an (NFS) file handle.
4423  */
4424 #ifndef _SYS_SYSPROTO_H_
4425 struct fhstat_args {
4426 	struct fhandle *u_fhp;
4427 	struct stat *sb;
4428 };
4429 #endif
4430 int
4431 sys_fhstat(td, uap)
4432 	struct thread *td;
4433 	register struct fhstat_args /* {
4434 		struct fhandle *u_fhp;
4435 		struct stat *sb;
4436 	} */ *uap;
4437 {
4438 	struct stat sb;
4439 	struct fhandle fh;
4440 	int error;
4441 
4442 	error = copyin(uap->u_fhp, &fh, sizeof(fh));
4443 	if (error != 0)
4444 		return (error);
4445 	error = kern_fhstat(td, fh, &sb);
4446 	if (error == 0)
4447 		error = copyout(&sb, uap->sb, sizeof(sb));
4448 	return (error);
4449 }
4450 
4451 int
4452 kern_fhstat(struct thread *td, struct fhandle fh, struct stat *sb)
4453 {
4454 	struct mount *mp;
4455 	struct vnode *vp;
4456 	int error;
4457 
4458 	error = priv_check(td, PRIV_VFS_FHSTAT);
4459 	if (error != 0)
4460 		return (error);
4461 	if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4462 		return (ESTALE);
4463 	error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4464 	vfs_unbusy(mp);
4465 	if (error != 0)
4466 		return (error);
4467 	error = vn_stat(vp, sb, td->td_ucred, NOCRED, td);
4468 	vput(vp);
4469 	return (error);
4470 }
4471 
4472 /*
4473  * Implement fstatfs() for (NFS) file handles.
4474  */
4475 #ifndef _SYS_SYSPROTO_H_
4476 struct fhstatfs_args {
4477 	struct fhandle *u_fhp;
4478 	struct statfs *buf;
4479 };
4480 #endif
4481 int
4482 sys_fhstatfs(td, uap)
4483 	struct thread *td;
4484 	struct fhstatfs_args /* {
4485 		struct fhandle *u_fhp;
4486 		struct statfs *buf;
4487 	} */ *uap;
4488 {
4489 	struct statfs sf;
4490 	fhandle_t fh;
4491 	int error;
4492 
4493 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
4494 	if (error != 0)
4495 		return (error);
4496 	error = kern_fhstatfs(td, fh, &sf);
4497 	if (error != 0)
4498 		return (error);
4499 	return (copyout(&sf, uap->buf, sizeof(sf)));
4500 }
4501 
4502 int
4503 kern_fhstatfs(struct thread *td, fhandle_t fh, struct statfs *buf)
4504 {
4505 	struct statfs *sp;
4506 	struct mount *mp;
4507 	struct vnode *vp;
4508 	int error;
4509 
4510 	error = priv_check(td, PRIV_VFS_FHSTATFS);
4511 	if (error != 0)
4512 		return (error);
4513 	if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4514 		return (ESTALE);
4515 	error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4516 	if (error != 0) {
4517 		vfs_unbusy(mp);
4518 		return (error);
4519 	}
4520 	vput(vp);
4521 	error = prison_canseemount(td->td_ucred, mp);
4522 	if (error != 0)
4523 		goto out;
4524 #ifdef MAC
4525 	error = mac_mount_check_stat(td->td_ucred, mp);
4526 	if (error != 0)
4527 		goto out;
4528 #endif
4529 	/*
4530 	 * Set these in case the underlying filesystem fails to do so.
4531 	 */
4532 	sp = &mp->mnt_stat;
4533 	sp->f_version = STATFS_VERSION;
4534 	sp->f_namemax = NAME_MAX;
4535 	sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
4536 	error = VFS_STATFS(mp, sp);
4537 	if (error == 0)
4538 		*buf = *sp;
4539 out:
4540 	vfs_unbusy(mp);
4541 	return (error);
4542 }
4543 
4544 int
4545 kern_posix_fallocate(struct thread *td, int fd, off_t offset, off_t len)
4546 {
4547 	struct file *fp;
4548 	struct mount *mp;
4549 	struct vnode *vp;
4550 	cap_rights_t rights;
4551 	off_t olen, ooffset;
4552 	int error;
4553 
4554 	fp = NULL;
4555 	error = fget(td, fd, cap_rights_init(&rights, CAP_WRITE), &fp);
4556 	if (error != 0)
4557 		goto out;
4558 
4559 	switch (fp->f_type) {
4560 	case DTYPE_VNODE:
4561 		break;
4562 	case DTYPE_PIPE:
4563 	case DTYPE_FIFO:
4564 		error = ESPIPE;
4565 		goto out;
4566 	default:
4567 		error = ENODEV;
4568 		goto out;
4569 	}
4570 	if ((fp->f_flag & FWRITE) == 0) {
4571 		error = EBADF;
4572 		goto out;
4573 	}
4574 	vp = fp->f_vnode;
4575 	if (vp->v_type != VREG) {
4576 		error = ENODEV;
4577 		goto out;
4578 	}
4579 	if (offset < 0 || len <= 0) {
4580 		error = EINVAL;
4581 		goto out;
4582 	}
4583 	/* Check for wrap. */
4584 	if (offset > OFF_MAX - len) {
4585 		error = EFBIG;
4586 		goto out;
4587 	}
4588 
4589 	/* Allocating blocks may take a long time, so iterate. */
4590 	for (;;) {
4591 		olen = len;
4592 		ooffset = offset;
4593 
4594 		bwillwrite();
4595 		mp = NULL;
4596 		error = vn_start_write(vp, &mp, V_WAIT | PCATCH);
4597 		if (error != 0)
4598 			break;
4599 		error = vn_lock(vp, LK_EXCLUSIVE);
4600 		if (error != 0) {
4601 			vn_finished_write(mp);
4602 			break;
4603 		}
4604 #ifdef MAC
4605 		error = mac_vnode_check_write(td->td_ucred, fp->f_cred, vp);
4606 		if (error == 0)
4607 #endif
4608 			error = VOP_ALLOCATE(vp, &offset, &len);
4609 		VOP_UNLOCK(vp, 0);
4610 		vn_finished_write(mp);
4611 
4612 		if (olen + ooffset != offset + len) {
4613 			panic("offset + len changed from %jx/%jx to %jx/%jx",
4614 			    ooffset, olen, offset, len);
4615 		}
4616 		if (error != 0 || len == 0)
4617 			break;
4618 		KASSERT(olen > len, ("Iteration did not make progress?"));
4619 		maybe_yield();
4620 	}
4621  out:
4622 	if (fp != NULL)
4623 		fdrop(fp, td);
4624 	return (error);
4625 }
4626 
4627 int
4628 sys_posix_fallocate(struct thread *td, struct posix_fallocate_args *uap)
4629 {
4630 
4631 	td->td_retval[0] = kern_posix_fallocate(td, uap->fd, uap->offset,
4632 	    uap->len);
4633 	return (0);
4634 }
4635 
4636 /*
4637  * Unlike madvise(2), we do not make a best effort to remember every
4638  * possible caching hint.  Instead, we remember the last setting with
4639  * the exception that we will allow POSIX_FADV_NORMAL to adjust the
4640  * region of any current setting.
4641  */
4642 int
4643 kern_posix_fadvise(struct thread *td, int fd, off_t offset, off_t len,
4644     int advice)
4645 {
4646 	struct fadvise_info *fa, *new;
4647 	struct file *fp;
4648 	struct vnode *vp;
4649 	cap_rights_t rights;
4650 	off_t end;
4651 	int error;
4652 
4653 	if (offset < 0 || len < 0 || offset > OFF_MAX - len)
4654 		return (EINVAL);
4655 	switch (advice) {
4656 	case POSIX_FADV_SEQUENTIAL:
4657 	case POSIX_FADV_RANDOM:
4658 	case POSIX_FADV_NOREUSE:
4659 		new = malloc(sizeof(*fa), M_FADVISE, M_WAITOK);
4660 		break;
4661 	case POSIX_FADV_NORMAL:
4662 	case POSIX_FADV_WILLNEED:
4663 	case POSIX_FADV_DONTNEED:
4664 		new = NULL;
4665 		break;
4666 	default:
4667 		return (EINVAL);
4668 	}
4669 	/* XXX: CAP_POSIX_FADVISE? */
4670 	error = fget(td, fd, cap_rights_init(&rights), &fp);
4671 	if (error != 0)
4672 		goto out;
4673 
4674 	switch (fp->f_type) {
4675 	case DTYPE_VNODE:
4676 		break;
4677 	case DTYPE_PIPE:
4678 	case DTYPE_FIFO:
4679 		error = ESPIPE;
4680 		goto out;
4681 	default:
4682 		error = ENODEV;
4683 		goto out;
4684 	}
4685 	vp = fp->f_vnode;
4686 	if (vp->v_type != VREG) {
4687 		error = ENODEV;
4688 		goto out;
4689 	}
4690 	if (len == 0)
4691 		end = OFF_MAX;
4692 	else
4693 		end = offset + len - 1;
4694 	switch (advice) {
4695 	case POSIX_FADV_SEQUENTIAL:
4696 	case POSIX_FADV_RANDOM:
4697 	case POSIX_FADV_NOREUSE:
4698 		/*
4699 		 * Try to merge any existing non-standard region with
4700 		 * this new region if possible, otherwise create a new
4701 		 * non-standard region for this request.
4702 		 */
4703 		mtx_pool_lock(mtxpool_sleep, fp);
4704 		fa = fp->f_advice;
4705 		if (fa != NULL && fa->fa_advice == advice &&
4706 		    ((fa->fa_start <= end && fa->fa_end >= offset) ||
4707 		    (end != OFF_MAX && fa->fa_start == end + 1) ||
4708 		    (fa->fa_end != OFF_MAX && fa->fa_end + 1 == offset))) {
4709 			if (offset < fa->fa_start)
4710 				fa->fa_start = offset;
4711 			if (end > fa->fa_end)
4712 				fa->fa_end = end;
4713 		} else {
4714 			new->fa_advice = advice;
4715 			new->fa_start = offset;
4716 			new->fa_end = end;
4717 			new->fa_prevstart = 0;
4718 			new->fa_prevend = 0;
4719 			fp->f_advice = new;
4720 			new = fa;
4721 		}
4722 		mtx_pool_unlock(mtxpool_sleep, fp);
4723 		break;
4724 	case POSIX_FADV_NORMAL:
4725 		/*
4726 		 * If a the "normal" region overlaps with an existing
4727 		 * non-standard region, trim or remove the
4728 		 * non-standard region.
4729 		 */
4730 		mtx_pool_lock(mtxpool_sleep, fp);
4731 		fa = fp->f_advice;
4732 		if (fa != NULL) {
4733 			if (offset <= fa->fa_start && end >= fa->fa_end) {
4734 				new = fa;
4735 				fp->f_advice = NULL;
4736 			} else if (offset <= fa->fa_start &&
4737 			    end >= fa->fa_start)
4738 				fa->fa_start = end + 1;
4739 			else if (offset <= fa->fa_end && end >= fa->fa_end)
4740 				fa->fa_end = offset - 1;
4741 			else if (offset >= fa->fa_start && end <= fa->fa_end) {
4742 				/*
4743 				 * If the "normal" region is a middle
4744 				 * portion of the existing
4745 				 * non-standard region, just remove
4746 				 * the whole thing rather than picking
4747 				 * one side or the other to
4748 				 * preserve.
4749 				 */
4750 				new = fa;
4751 				fp->f_advice = NULL;
4752 			}
4753 		}
4754 		mtx_pool_unlock(mtxpool_sleep, fp);
4755 		break;
4756 	case POSIX_FADV_WILLNEED:
4757 	case POSIX_FADV_DONTNEED:
4758 		error = VOP_ADVISE(vp, offset, end, advice);
4759 		break;
4760 	}
4761 out:
4762 	if (fp != NULL)
4763 		fdrop(fp, td);
4764 	free(new, M_FADVISE);
4765 	return (error);
4766 }
4767 
4768 int
4769 sys_posix_fadvise(struct thread *td, struct posix_fadvise_args *uap)
4770 {
4771 
4772 	td->td_retval[0] = kern_posix_fadvise(td, uap->fd, uap->offset,
4773 	    uap->len, uap->advice);
4774 	return (0);
4775 }
4776