xref: /freebsd/sys/kern/vfs_syscalls.c (revision dcf58f92e2c19a32fc171f763698e711c719badc)
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_chflagsat(struct thread *td, int fd, const char *path,
100     enum uio_seg pathseg, u_long flags, int atflag);
101 static int setfflags(struct thread *td, struct vnode *, u_long);
102 static int setutimes(struct thread *td, struct vnode *,
103     const struct timespec *, int, int);
104 static int vn_access(struct vnode *vp, int user_flags, struct ucred *cred,
105     struct thread *td);
106 
107 /*
108  * The module initialization routine for POSIX asynchronous I/O will
109  * set this to the version of AIO that it implements.  (Zero means
110  * that it is not implemented.)  This value is used here by pathconf()
111  * and in kern_descrip.c by fpathconf().
112  */
113 int async_io_version;
114 
115 /*
116  * Sync each mounted filesystem.
117  */
118 #ifndef _SYS_SYSPROTO_H_
119 struct sync_args {
120 	int     dummy;
121 };
122 #endif
123 /* ARGSUSED */
124 int
125 sys_sync(td, uap)
126 	struct thread *td;
127 	struct sync_args *uap;
128 {
129 	struct mount *mp, *nmp;
130 	int save;
131 
132 	mtx_lock(&mountlist_mtx);
133 	for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
134 		if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK)) {
135 			nmp = TAILQ_NEXT(mp, mnt_list);
136 			continue;
137 		}
138 		if ((mp->mnt_flag & MNT_RDONLY) == 0 &&
139 		    vn_start_write(NULL, &mp, V_NOWAIT) == 0) {
140 			save = curthread_pflags_set(TDP_SYNCIO);
141 			vfs_msync(mp, MNT_NOWAIT);
142 			VFS_SYNC(mp, MNT_NOWAIT);
143 			curthread_pflags_restore(save);
144 			vn_finished_write(mp);
145 		}
146 		mtx_lock(&mountlist_mtx);
147 		nmp = TAILQ_NEXT(mp, mnt_list);
148 		vfs_unbusy(mp);
149 	}
150 	mtx_unlock(&mountlist_mtx);
151 	return (0);
152 }
153 
154 /*
155  * Change filesystem quotas.
156  */
157 #ifndef _SYS_SYSPROTO_H_
158 struct quotactl_args {
159 	char *path;
160 	int cmd;
161 	int uid;
162 	caddr_t arg;
163 };
164 #endif
165 int
166 sys_quotactl(td, uap)
167 	struct thread *td;
168 	register struct quotactl_args /* {
169 		char *path;
170 		int cmd;
171 		int uid;
172 		caddr_t arg;
173 	} */ *uap;
174 {
175 	struct mount *mp;
176 	struct nameidata nd;
177 	int error;
178 
179 	AUDIT_ARG_CMD(uap->cmd);
180 	AUDIT_ARG_UID(uap->uid);
181 	if (!prison_allow(td->td_ucred, PR_ALLOW_QUOTAS))
182 		return (EPERM);
183 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
184 	    uap->path, td);
185 	if ((error = namei(&nd)) != 0)
186 		return (error);
187 	NDFREE(&nd, NDF_ONLY_PNBUF);
188 	mp = nd.ni_vp->v_mount;
189 	vfs_ref(mp);
190 	vput(nd.ni_vp);
191 	error = vfs_busy(mp, 0);
192 	vfs_rel(mp);
193 	if (error != 0)
194 		return (error);
195 	error = VFS_QUOTACTL(mp, uap->cmd, uap->uid, uap->arg);
196 
197 	/*
198 	 * Since quota on operation typically needs to open quota
199 	 * file, the Q_QUOTAON handler needs to unbusy the mount point
200 	 * before calling into namei.  Otherwise, unmount might be
201 	 * started between two vfs_busy() invocations (first is our,
202 	 * second is from mount point cross-walk code in lookup()),
203 	 * causing deadlock.
204 	 *
205 	 * Require that Q_QUOTAON handles the vfs_busy() reference on
206 	 * its own, always returning with ubusied mount point.
207 	 */
208 	if ((uap->cmd >> SUBCMDSHIFT) != Q_QUOTAON)
209 		vfs_unbusy(mp);
210 	return (error);
211 }
212 
213 /*
214  * Used by statfs conversion routines to scale the block size up if
215  * necessary so that all of the block counts are <= 'max_size'.  Note
216  * that 'max_size' should be a bitmask, i.e. 2^n - 1 for some non-zero
217  * value of 'n'.
218  */
219 void
220 statfs_scale_blocks(struct statfs *sf, long max_size)
221 {
222 	uint64_t count;
223 	int shift;
224 
225 	KASSERT(powerof2(max_size + 1), ("%s: invalid max_size", __func__));
226 
227 	/*
228 	 * Attempt to scale the block counts to give a more accurate
229 	 * overview to userland of the ratio of free space to used
230 	 * space.  To do this, find the largest block count and compute
231 	 * a divisor that lets it fit into a signed integer <= max_size.
232 	 */
233 	if (sf->f_bavail < 0)
234 		count = -sf->f_bavail;
235 	else
236 		count = sf->f_bavail;
237 	count = MAX(sf->f_blocks, MAX(sf->f_bfree, count));
238 	if (count <= max_size)
239 		return;
240 
241 	count >>= flsl(max_size);
242 	shift = 0;
243 	while (count > 0) {
244 		shift++;
245 		count >>=1;
246 	}
247 
248 	sf->f_bsize <<= shift;
249 	sf->f_blocks >>= shift;
250 	sf->f_bfree >>= shift;
251 	sf->f_bavail >>= shift;
252 }
253 
254 /*
255  * Get filesystem statistics.
256  */
257 #ifndef _SYS_SYSPROTO_H_
258 struct statfs_args {
259 	char *path;
260 	struct statfs *buf;
261 };
262 #endif
263 int
264 sys_statfs(td, uap)
265 	struct thread *td;
266 	register struct statfs_args /* {
267 		char *path;
268 		struct statfs *buf;
269 	} */ *uap;
270 {
271 	struct statfs sf;
272 	int error;
273 
274 	error = kern_statfs(td, uap->path, UIO_USERSPACE, &sf);
275 	if (error == 0)
276 		error = copyout(&sf, uap->buf, sizeof(sf));
277 	return (error);
278 }
279 
280 int
281 kern_statfs(struct thread *td, char *path, enum uio_seg pathseg,
282     struct statfs *buf)
283 {
284 	struct mount *mp;
285 	struct statfs *sp, sb;
286 	struct nameidata nd;
287 	int error;
288 
289 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
290 	    pathseg, path, td);
291 	error = namei(&nd);
292 	if (error != 0)
293 		return (error);
294 	mp = nd.ni_vp->v_mount;
295 	vfs_ref(mp);
296 	NDFREE(&nd, NDF_ONLY_PNBUF);
297 	vput(nd.ni_vp);
298 	error = vfs_busy(mp, 0);
299 	vfs_rel(mp);
300 	if (error != 0)
301 		return (error);
302 #ifdef MAC
303 	error = mac_mount_check_stat(td->td_ucred, mp);
304 	if (error != 0)
305 		goto out;
306 #endif
307 	/*
308 	 * Set these in case the underlying filesystem fails to do so.
309 	 */
310 	sp = &mp->mnt_stat;
311 	sp->f_version = STATFS_VERSION;
312 	sp->f_namemax = NAME_MAX;
313 	sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
314 	error = VFS_STATFS(mp, sp);
315 	if (error != 0)
316 		goto out;
317 	if (priv_check(td, PRIV_VFS_GENERATION)) {
318 		bcopy(sp, &sb, sizeof(sb));
319 		sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
320 		prison_enforce_statfs(td->td_ucred, mp, &sb);
321 		sp = &sb;
322 	}
323 	*buf = *sp;
324 out:
325 	vfs_unbusy(mp);
326 	return (error);
327 }
328 
329 /*
330  * Get filesystem statistics.
331  */
332 #ifndef _SYS_SYSPROTO_H_
333 struct fstatfs_args {
334 	int fd;
335 	struct statfs *buf;
336 };
337 #endif
338 int
339 sys_fstatfs(td, uap)
340 	struct thread *td;
341 	register struct fstatfs_args /* {
342 		int fd;
343 		struct statfs *buf;
344 	} */ *uap;
345 {
346 	struct statfs sf;
347 	int error;
348 
349 	error = kern_fstatfs(td, uap->fd, &sf);
350 	if (error == 0)
351 		error = copyout(&sf, uap->buf, sizeof(sf));
352 	return (error);
353 }
354 
355 int
356 kern_fstatfs(struct thread *td, int fd, struct statfs *buf)
357 {
358 	struct file *fp;
359 	struct mount *mp;
360 	struct statfs *sp, sb;
361 	struct vnode *vp;
362 	cap_rights_t rights;
363 	int error;
364 
365 	AUDIT_ARG_FD(fd);
366 	error = getvnode(td->td_proc->p_fd, fd,
367 	    cap_rights_init(&rights, CAP_FSTATFS), &fp);
368 	if (error != 0)
369 		return (error);
370 	vp = fp->f_vnode;
371 	vn_lock(vp, LK_SHARED | LK_RETRY);
372 #ifdef AUDIT
373 	AUDIT_ARG_VNODE1(vp);
374 #endif
375 	mp = vp->v_mount;
376 	if (mp)
377 		vfs_ref(mp);
378 	VOP_UNLOCK(vp, 0);
379 	fdrop(fp, td);
380 	if (mp == NULL) {
381 		error = EBADF;
382 		goto out;
383 	}
384 	error = vfs_busy(mp, 0);
385 	vfs_rel(mp);
386 	if (error != 0)
387 		return (error);
388 #ifdef MAC
389 	error = mac_mount_check_stat(td->td_ucred, mp);
390 	if (error != 0)
391 		goto out;
392 #endif
393 	/*
394 	 * Set these in case the underlying filesystem fails to do so.
395 	 */
396 	sp = &mp->mnt_stat;
397 	sp->f_version = STATFS_VERSION;
398 	sp->f_namemax = NAME_MAX;
399 	sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
400 	error = VFS_STATFS(mp, sp);
401 	if (error != 0)
402 		goto out;
403 	if (priv_check(td, PRIV_VFS_GENERATION)) {
404 		bcopy(sp, &sb, sizeof(sb));
405 		sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
406 		prison_enforce_statfs(td->td_ucred, mp, &sb);
407 		sp = &sb;
408 	}
409 	*buf = *sp;
410 out:
411 	if (mp)
412 		vfs_unbusy(mp);
413 	return (error);
414 }
415 
416 /*
417  * Get statistics on all filesystems.
418  */
419 #ifndef _SYS_SYSPROTO_H_
420 struct getfsstat_args {
421 	struct statfs *buf;
422 	long bufsize;
423 	int flags;
424 };
425 #endif
426 int
427 sys_getfsstat(td, uap)
428 	struct thread *td;
429 	register struct getfsstat_args /* {
430 		struct statfs *buf;
431 		long bufsize;
432 		int flags;
433 	} */ *uap;
434 {
435 
436 	return (kern_getfsstat(td, &uap->buf, uap->bufsize, UIO_USERSPACE,
437 	    uap->flags));
438 }
439 
440 /*
441  * If (bufsize > 0 && bufseg == UIO_SYSSPACE)
442  *	The caller is responsible for freeing memory which will be allocated
443  *	in '*buf'.
444  */
445 int
446 kern_getfsstat(struct thread *td, struct statfs **buf, size_t bufsize,
447     enum uio_seg bufseg, int flags)
448 {
449 	struct mount *mp, *nmp;
450 	struct statfs *sfsp, *sp, sb;
451 	size_t count, maxcount;
452 	int error;
453 
454 	maxcount = bufsize / sizeof(struct statfs);
455 	if (bufsize == 0)
456 		sfsp = NULL;
457 	else if (bufseg == UIO_USERSPACE)
458 		sfsp = *buf;
459 	else /* if (bufseg == UIO_SYSSPACE) */ {
460 		count = 0;
461 		mtx_lock(&mountlist_mtx);
462 		TAILQ_FOREACH(mp, &mountlist, mnt_list) {
463 			count++;
464 		}
465 		mtx_unlock(&mountlist_mtx);
466 		if (maxcount > count)
467 			maxcount = count;
468 		sfsp = *buf = malloc(maxcount * sizeof(struct statfs), M_TEMP,
469 		    M_WAITOK);
470 	}
471 	count = 0;
472 	mtx_lock(&mountlist_mtx);
473 	for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
474 		if (prison_canseemount(td->td_ucred, mp) != 0) {
475 			nmp = TAILQ_NEXT(mp, mnt_list);
476 			continue;
477 		}
478 #ifdef MAC
479 		if (mac_mount_check_stat(td->td_ucred, mp) != 0) {
480 			nmp = TAILQ_NEXT(mp, mnt_list);
481 			continue;
482 		}
483 #endif
484 		if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK)) {
485 			nmp = TAILQ_NEXT(mp, mnt_list);
486 			continue;
487 		}
488 		if (sfsp && count < maxcount) {
489 			sp = &mp->mnt_stat;
490 			/*
491 			 * Set these in case the underlying filesystem
492 			 * fails to do so.
493 			 */
494 			sp->f_version = STATFS_VERSION;
495 			sp->f_namemax = NAME_MAX;
496 			sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
497 			/*
498 			 * If MNT_NOWAIT or MNT_LAZY is specified, do not
499 			 * refresh the fsstat cache. MNT_NOWAIT or MNT_LAZY
500 			 * overrides MNT_WAIT.
501 			 */
502 			if (((flags & (MNT_LAZY|MNT_NOWAIT)) == 0 ||
503 			    (flags & MNT_WAIT)) &&
504 			    (error = VFS_STATFS(mp, sp))) {
505 				mtx_lock(&mountlist_mtx);
506 				nmp = TAILQ_NEXT(mp, mnt_list);
507 				vfs_unbusy(mp);
508 				continue;
509 			}
510 			if (priv_check(td, PRIV_VFS_GENERATION)) {
511 				bcopy(sp, &sb, sizeof(sb));
512 				sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
513 				prison_enforce_statfs(td->td_ucred, mp, &sb);
514 				sp = &sb;
515 			}
516 			if (bufseg == UIO_SYSSPACE)
517 				bcopy(sp, sfsp, sizeof(*sp));
518 			else /* if (bufseg == UIO_USERSPACE) */ {
519 				error = copyout(sp, sfsp, sizeof(*sp));
520 				if (error != 0) {
521 					vfs_unbusy(mp);
522 					return (error);
523 				}
524 			}
525 			sfsp++;
526 		}
527 		count++;
528 		mtx_lock(&mountlist_mtx);
529 		nmp = TAILQ_NEXT(mp, mnt_list);
530 		vfs_unbusy(mp);
531 	}
532 	mtx_unlock(&mountlist_mtx);
533 	if (sfsp && count > maxcount)
534 		td->td_retval[0] = maxcount;
535 	else
536 		td->td_retval[0] = count;
537 	return (0);
538 }
539 
540 #ifdef COMPAT_FREEBSD4
541 /*
542  * Get old format filesystem statistics.
543  */
544 static void cvtstatfs(struct statfs *, struct ostatfs *);
545 
546 #ifndef _SYS_SYSPROTO_H_
547 struct freebsd4_statfs_args {
548 	char *path;
549 	struct ostatfs *buf;
550 };
551 #endif
552 int
553 freebsd4_statfs(td, uap)
554 	struct thread *td;
555 	struct freebsd4_statfs_args /* {
556 		char *path;
557 		struct ostatfs *buf;
558 	} */ *uap;
559 {
560 	struct ostatfs osb;
561 	struct statfs sf;
562 	int error;
563 
564 	error = kern_statfs(td, uap->path, UIO_USERSPACE, &sf);
565 	if (error != 0)
566 		return (error);
567 	cvtstatfs(&sf, &osb);
568 	return (copyout(&osb, uap->buf, sizeof(osb)));
569 }
570 
571 /*
572  * Get filesystem statistics.
573  */
574 #ifndef _SYS_SYSPROTO_H_
575 struct freebsd4_fstatfs_args {
576 	int fd;
577 	struct ostatfs *buf;
578 };
579 #endif
580 int
581 freebsd4_fstatfs(td, uap)
582 	struct thread *td;
583 	struct freebsd4_fstatfs_args /* {
584 		int fd;
585 		struct ostatfs *buf;
586 	} */ *uap;
587 {
588 	struct ostatfs osb;
589 	struct statfs sf;
590 	int error;
591 
592 	error = kern_fstatfs(td, uap->fd, &sf);
593 	if (error != 0)
594 		return (error);
595 	cvtstatfs(&sf, &osb);
596 	return (copyout(&osb, uap->buf, sizeof(osb)));
597 }
598 
599 /*
600  * Get statistics on all filesystems.
601  */
602 #ifndef _SYS_SYSPROTO_H_
603 struct freebsd4_getfsstat_args {
604 	struct ostatfs *buf;
605 	long bufsize;
606 	int flags;
607 };
608 #endif
609 int
610 freebsd4_getfsstat(td, uap)
611 	struct thread *td;
612 	register struct freebsd4_getfsstat_args /* {
613 		struct ostatfs *buf;
614 		long bufsize;
615 		int flags;
616 	} */ *uap;
617 {
618 	struct statfs *buf, *sp;
619 	struct ostatfs osb;
620 	size_t count, size;
621 	int error;
622 
623 	count = uap->bufsize / sizeof(struct ostatfs);
624 	size = count * sizeof(struct statfs);
625 	error = kern_getfsstat(td, &buf, size, UIO_SYSSPACE, uap->flags);
626 	if (size > 0) {
627 		count = td->td_retval[0];
628 		sp = buf;
629 		while (count > 0 && error == 0) {
630 			cvtstatfs(sp, &osb);
631 			error = copyout(&osb, uap->buf, sizeof(osb));
632 			sp++;
633 			uap->buf++;
634 			count--;
635 		}
636 		free(buf, M_TEMP);
637 	}
638 	return (error);
639 }
640 
641 /*
642  * Implement fstatfs() for (NFS) file handles.
643  */
644 #ifndef _SYS_SYSPROTO_H_
645 struct freebsd4_fhstatfs_args {
646 	struct fhandle *u_fhp;
647 	struct ostatfs *buf;
648 };
649 #endif
650 int
651 freebsd4_fhstatfs(td, uap)
652 	struct thread *td;
653 	struct freebsd4_fhstatfs_args /* {
654 		struct fhandle *u_fhp;
655 		struct ostatfs *buf;
656 	} */ *uap;
657 {
658 	struct ostatfs osb;
659 	struct statfs sf;
660 	fhandle_t fh;
661 	int error;
662 
663 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
664 	if (error != 0)
665 		return (error);
666 	error = kern_fhstatfs(td, fh, &sf);
667 	if (error != 0)
668 		return (error);
669 	cvtstatfs(&sf, &osb);
670 	return (copyout(&osb, uap->buf, sizeof(osb)));
671 }
672 
673 /*
674  * Convert a new format statfs structure to an old format statfs structure.
675  */
676 static void
677 cvtstatfs(nsp, osp)
678 	struct statfs *nsp;
679 	struct ostatfs *osp;
680 {
681 
682 	statfs_scale_blocks(nsp, LONG_MAX);
683 	bzero(osp, sizeof(*osp));
684 	osp->f_bsize = nsp->f_bsize;
685 	osp->f_iosize = MIN(nsp->f_iosize, LONG_MAX);
686 	osp->f_blocks = nsp->f_blocks;
687 	osp->f_bfree = nsp->f_bfree;
688 	osp->f_bavail = nsp->f_bavail;
689 	osp->f_files = MIN(nsp->f_files, LONG_MAX);
690 	osp->f_ffree = MIN(nsp->f_ffree, LONG_MAX);
691 	osp->f_owner = nsp->f_owner;
692 	osp->f_type = nsp->f_type;
693 	osp->f_flags = nsp->f_flags;
694 	osp->f_syncwrites = MIN(nsp->f_syncwrites, LONG_MAX);
695 	osp->f_asyncwrites = MIN(nsp->f_asyncwrites, LONG_MAX);
696 	osp->f_syncreads = MIN(nsp->f_syncreads, LONG_MAX);
697 	osp->f_asyncreads = MIN(nsp->f_asyncreads, LONG_MAX);
698 	strlcpy(osp->f_fstypename, nsp->f_fstypename,
699 	    MIN(MFSNAMELEN, OMFSNAMELEN));
700 	strlcpy(osp->f_mntonname, nsp->f_mntonname,
701 	    MIN(MNAMELEN, OMNAMELEN));
702 	strlcpy(osp->f_mntfromname, nsp->f_mntfromname,
703 	    MIN(MNAMELEN, OMNAMELEN));
704 	osp->f_fsid = nsp->f_fsid;
705 }
706 #endif /* COMPAT_FREEBSD4 */
707 
708 /*
709  * Change current working directory to a given file descriptor.
710  */
711 #ifndef _SYS_SYSPROTO_H_
712 struct fchdir_args {
713 	int	fd;
714 };
715 #endif
716 int
717 sys_fchdir(td, uap)
718 	struct thread *td;
719 	struct fchdir_args /* {
720 		int fd;
721 	} */ *uap;
722 {
723 	register struct filedesc *fdp = td->td_proc->p_fd;
724 	struct vnode *vp, *tdp, *vpold;
725 	struct mount *mp;
726 	struct file *fp;
727 	cap_rights_t rights;
728 	int error;
729 
730 	AUDIT_ARG_FD(uap->fd);
731 	error = getvnode(fdp, uap->fd, cap_rights_init(&rights, CAP_FCHDIR),
732 	    &fp);
733 	if (error != 0)
734 		return (error);
735 	vp = fp->f_vnode;
736 	VREF(vp);
737 	fdrop(fp, td);
738 	vn_lock(vp, LK_SHARED | LK_RETRY);
739 	AUDIT_ARG_VNODE1(vp);
740 	error = change_dir(vp, td);
741 	while (!error && (mp = vp->v_mountedhere) != NULL) {
742 		if (vfs_busy(mp, 0))
743 			continue;
744 		error = VFS_ROOT(mp, LK_SHARED, &tdp);
745 		vfs_unbusy(mp);
746 		if (error != 0)
747 			break;
748 		vput(vp);
749 		vp = tdp;
750 	}
751 	if (error != 0) {
752 		vput(vp);
753 		return (error);
754 	}
755 	VOP_UNLOCK(vp, 0);
756 	FILEDESC_XLOCK(fdp);
757 	vpold = fdp->fd_cdir;
758 	fdp->fd_cdir = vp;
759 	FILEDESC_XUNLOCK(fdp);
760 	vrele(vpold);
761 	return (0);
762 }
763 
764 /*
765  * Change current working directory (``.'').
766  */
767 #ifndef _SYS_SYSPROTO_H_
768 struct chdir_args {
769 	char	*path;
770 };
771 #endif
772 int
773 sys_chdir(td, uap)
774 	struct thread *td;
775 	struct chdir_args /* {
776 		char *path;
777 	} */ *uap;
778 {
779 
780 	return (kern_chdir(td, uap->path, UIO_USERSPACE));
781 }
782 
783 int
784 kern_chdir(struct thread *td, char *path, enum uio_seg pathseg)
785 {
786 	register struct filedesc *fdp = td->td_proc->p_fd;
787 	struct nameidata nd;
788 	struct vnode *vp;
789 	int error;
790 
791 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
792 	    pathseg, path, td);
793 	if ((error = namei(&nd)) != 0)
794 		return (error);
795 	if ((error = change_dir(nd.ni_vp, td)) != 0) {
796 		vput(nd.ni_vp);
797 		NDFREE(&nd, NDF_ONLY_PNBUF);
798 		return (error);
799 	}
800 	VOP_UNLOCK(nd.ni_vp, 0);
801 	NDFREE(&nd, NDF_ONLY_PNBUF);
802 	FILEDESC_XLOCK(fdp);
803 	vp = fdp->fd_cdir;
804 	fdp->fd_cdir = nd.ni_vp;
805 	FILEDESC_XUNLOCK(fdp);
806 	vrele(vp);
807 	return (0);
808 }
809 
810 /*
811  * Helper function for raised chroot(2) security function:  Refuse if
812  * any filedescriptors are open directories.
813  */
814 static int
815 chroot_refuse_vdir_fds(fdp)
816 	struct filedesc *fdp;
817 {
818 	struct vnode *vp;
819 	struct file *fp;
820 	int fd;
821 
822 	FILEDESC_LOCK_ASSERT(fdp);
823 
824 	for (fd = 0; fd <= fdp->fd_lastfile; fd++) {
825 		fp = fget_locked(fdp, fd);
826 		if (fp == NULL)
827 			continue;
828 		if (fp->f_type == DTYPE_VNODE) {
829 			vp = fp->f_vnode;
830 			if (vp->v_type == VDIR)
831 				return (EPERM);
832 		}
833 	}
834 	return (0);
835 }
836 
837 /*
838  * This sysctl determines if we will allow a process to chroot(2) if it
839  * has a directory open:
840  *	0: disallowed for all processes.
841  *	1: allowed for processes that were not already chroot(2)'ed.
842  *	2: allowed for all processes.
843  */
844 
845 static int chroot_allow_open_directories = 1;
846 
847 SYSCTL_INT(_kern, OID_AUTO, chroot_allow_open_directories, CTLFLAG_RW,
848      &chroot_allow_open_directories, 0,
849      "Allow a process to chroot(2) if it has a directory open");
850 
851 /*
852  * Change notion of root (``/'') directory.
853  */
854 #ifndef _SYS_SYSPROTO_H_
855 struct chroot_args {
856 	char	*path;
857 };
858 #endif
859 int
860 sys_chroot(td, uap)
861 	struct thread *td;
862 	struct chroot_args /* {
863 		char *path;
864 	} */ *uap;
865 {
866 	struct nameidata nd;
867 	int error;
868 
869 	error = priv_check(td, PRIV_VFS_CHROOT);
870 	if (error != 0)
871 		return (error);
872 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
873 	    UIO_USERSPACE, uap->path, td);
874 	error = namei(&nd);
875 	if (error != 0)
876 		goto error;
877 	error = change_dir(nd.ni_vp, td);
878 	if (error != 0)
879 		goto e_vunlock;
880 #ifdef MAC
881 	error = mac_vnode_check_chroot(td->td_ucred, nd.ni_vp);
882 	if (error != 0)
883 		goto e_vunlock;
884 #endif
885 	VOP_UNLOCK(nd.ni_vp, 0);
886 	error = change_root(nd.ni_vp, td);
887 	vrele(nd.ni_vp);
888 	NDFREE(&nd, NDF_ONLY_PNBUF);
889 	return (error);
890 e_vunlock:
891 	vput(nd.ni_vp);
892 error:
893 	NDFREE(&nd, NDF_ONLY_PNBUF);
894 	return (error);
895 }
896 
897 /*
898  * Common routine for chroot and chdir.  Callers must provide a locked vnode
899  * instance.
900  */
901 int
902 change_dir(vp, td)
903 	struct vnode *vp;
904 	struct thread *td;
905 {
906 #ifdef MAC
907 	int error;
908 #endif
909 
910 	ASSERT_VOP_LOCKED(vp, "change_dir(): vp not locked");
911 	if (vp->v_type != VDIR)
912 		return (ENOTDIR);
913 #ifdef MAC
914 	error = mac_vnode_check_chdir(td->td_ucred, vp);
915 	if (error != 0)
916 		return (error);
917 #endif
918 	return (VOP_ACCESS(vp, VEXEC, td->td_ucred, td));
919 }
920 
921 /*
922  * Common routine for kern_chroot() and jail_attach().  The caller is
923  * responsible for invoking priv_check() and mac_vnode_check_chroot() to
924  * authorize this operation.
925  */
926 int
927 change_root(vp, td)
928 	struct vnode *vp;
929 	struct thread *td;
930 {
931 	struct filedesc *fdp;
932 	struct vnode *oldvp;
933 	int error;
934 
935 	fdp = td->td_proc->p_fd;
936 	FILEDESC_XLOCK(fdp);
937 	if (chroot_allow_open_directories == 0 ||
938 	    (chroot_allow_open_directories == 1 && fdp->fd_rdir != rootvnode)) {
939 		error = chroot_refuse_vdir_fds(fdp);
940 		if (error != 0) {
941 			FILEDESC_XUNLOCK(fdp);
942 			return (error);
943 		}
944 	}
945 	oldvp = fdp->fd_rdir;
946 	fdp->fd_rdir = vp;
947 	VREF(fdp->fd_rdir);
948 	if (!fdp->fd_jdir) {
949 		fdp->fd_jdir = vp;
950 		VREF(fdp->fd_jdir);
951 	}
952 	FILEDESC_XUNLOCK(fdp);
953 	vrele(oldvp);
954 	return (0);
955 }
956 
957 static __inline void
958 flags_to_rights(int flags, cap_rights_t *rightsp)
959 {
960 
961 	if (flags & O_EXEC) {
962 		cap_rights_set(rightsp, CAP_FEXECVE);
963 	} else {
964 		switch ((flags & O_ACCMODE)) {
965 		case O_RDONLY:
966 			cap_rights_set(rightsp, CAP_READ);
967 			break;
968 		case O_RDWR:
969 			cap_rights_set(rightsp, CAP_READ);
970 			/* FALLTHROUGH */
971 		case O_WRONLY:
972 			cap_rights_set(rightsp, CAP_WRITE);
973 			if (!(flags & (O_APPEND | O_TRUNC)))
974 				cap_rights_set(rightsp, CAP_SEEK);
975 			break;
976 		}
977 	}
978 
979 	if (flags & O_CREAT)
980 		cap_rights_set(rightsp, CAP_CREATE);
981 
982 	if (flags & O_TRUNC)
983 		cap_rights_set(rightsp, CAP_FTRUNCATE);
984 
985 	if (flags & (O_SYNC | O_FSYNC))
986 		cap_rights_set(rightsp, CAP_FSYNC);
987 
988 	if (flags & (O_EXLOCK | O_SHLOCK))
989 		cap_rights_set(rightsp, CAP_FLOCK);
990 }
991 
992 /*
993  * Check permissions, allocate an open file structure, and call the device
994  * open routine if any.
995  */
996 #ifndef _SYS_SYSPROTO_H_
997 struct open_args {
998 	char	*path;
999 	int	flags;
1000 	int	mode;
1001 };
1002 #endif
1003 int
1004 sys_open(td, uap)
1005 	struct thread *td;
1006 	register struct open_args /* {
1007 		char *path;
1008 		int flags;
1009 		int mode;
1010 	} */ *uap;
1011 {
1012 
1013 	return (kern_openat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1014 	    uap->flags, uap->mode));
1015 }
1016 
1017 #ifndef _SYS_SYSPROTO_H_
1018 struct openat_args {
1019 	int	fd;
1020 	char	*path;
1021 	int	flag;
1022 	int	mode;
1023 };
1024 #endif
1025 int
1026 sys_openat(struct thread *td, struct openat_args *uap)
1027 {
1028 
1029 	return (kern_openat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
1030 	    uap->mode));
1031 }
1032 
1033 int
1034 kern_openat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1035     int flags, int mode)
1036 {
1037 	struct proc *p = td->td_proc;
1038 	struct filedesc *fdp = p->p_fd;
1039 	struct file *fp;
1040 	struct vnode *vp;
1041 	struct nameidata nd;
1042 	cap_rights_t rights;
1043 	int cmode, error, indx;
1044 
1045 	indx = -1;
1046 
1047 	AUDIT_ARG_FFLAGS(flags);
1048 	AUDIT_ARG_MODE(mode);
1049 	/* XXX: audit dirfd */
1050 	cap_rights_init(&rights, CAP_LOOKUP);
1051 	flags_to_rights(flags, &rights);
1052 	/*
1053 	 * Only one of the O_EXEC, O_RDONLY, O_WRONLY and O_RDWR flags
1054 	 * may be specified.
1055 	 */
1056 	if (flags & O_EXEC) {
1057 		if (flags & O_ACCMODE)
1058 			return (EINVAL);
1059 	} else if ((flags & O_ACCMODE) == O_ACCMODE) {
1060 		return (EINVAL);
1061 	} else {
1062 		flags = FFLAGS(flags);
1063 	}
1064 
1065 	/*
1066 	 * Allocate the file descriptor, but don't install a descriptor yet.
1067 	 */
1068 	error = falloc_noinstall(td, &fp);
1069 	if (error != 0)
1070 		return (error);
1071 	/*
1072 	 * An extra reference on `fp' has been held for us by
1073 	 * falloc_noinstall().
1074 	 */
1075 	/* Set the flags early so the finit in devfs can pick them up. */
1076 	fp->f_flag = flags & FMASK;
1077 	cmode = ((mode & ~fdp->fd_cmask) & ALLPERMS) & ~S_ISTXT;
1078 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, fd,
1079 	    &rights, td);
1080 	td->td_dupfd = -1;		/* XXX check for fdopen */
1081 	error = vn_open(&nd, &flags, cmode, fp);
1082 	if (error != 0) {
1083 		/*
1084 		 * If the vn_open replaced the method vector, something
1085 		 * wonderous happened deep below and we just pass it up
1086 		 * pretending we know what we do.
1087 		 */
1088 		if (error == ENXIO && fp->f_ops != &badfileops)
1089 			goto success;
1090 
1091 		/*
1092 		 * Handle special fdopen() case. bleh.
1093 		 *
1094 		 * Don't do this for relative (capability) lookups; we don't
1095 		 * understand exactly what would happen, and we don't think
1096 		 * that it ever should.
1097 		 */
1098 		if (nd.ni_strictrelative == 0 &&
1099 		    (error == ENODEV || error == ENXIO) &&
1100 		    td->td_dupfd >= 0) {
1101 			error = dupfdopen(td, fdp, td->td_dupfd, flags, error,
1102 			    &indx);
1103 			if (error == 0)
1104 				goto success;
1105 		}
1106 
1107 		goto bad;
1108 	}
1109 	td->td_dupfd = 0;
1110 	NDFREE(&nd, NDF_ONLY_PNBUF);
1111 	vp = nd.ni_vp;
1112 
1113 	/*
1114 	 * Store the vnode, for any f_type. Typically, the vnode use
1115 	 * count is decremented by direct call to vn_closefile() for
1116 	 * files that switched type in the cdevsw fdopen() method.
1117 	 */
1118 	fp->f_vnode = vp;
1119 	/*
1120 	 * If the file wasn't claimed by devfs bind it to the normal
1121 	 * vnode operations here.
1122 	 */
1123 	if (fp->f_ops == &badfileops) {
1124 		KASSERT(vp->v_type != VFIFO, ("Unexpected fifo."));
1125 		fp->f_seqcount = 1;
1126 		finit(fp, (flags & FMASK) | (fp->f_flag & FHASLOCK),
1127 		    DTYPE_VNODE, vp, &vnops);
1128 	}
1129 
1130 	VOP_UNLOCK(vp, 0);
1131 	if (flags & O_TRUNC) {
1132 		error = fo_truncate(fp, 0, td->td_ucred, td);
1133 		if (error != 0)
1134 			goto bad;
1135 	}
1136 success:
1137 	/*
1138 	 * If we haven't already installed the FD (for dupfdopen), do so now.
1139 	 */
1140 	if (indx == -1) {
1141 		struct filecaps *fcaps;
1142 
1143 #ifdef CAPABILITIES
1144 		if (nd.ni_strictrelative == 1)
1145 			fcaps = &nd.ni_filecaps;
1146 		else
1147 #endif
1148 			fcaps = NULL;
1149 		error = finstall(td, fp, &indx, flags, fcaps);
1150 		/* On success finstall() consumes fcaps. */
1151 		if (error != 0) {
1152 			filecaps_free(&nd.ni_filecaps);
1153 			goto bad;
1154 		}
1155 	} else {
1156 		filecaps_free(&nd.ni_filecaps);
1157 	}
1158 
1159 	/*
1160 	 * Release our private reference, leaving the one associated with
1161 	 * the descriptor table intact.
1162 	 */
1163 	fdrop(fp, td);
1164 	td->td_retval[0] = indx;
1165 	return (0);
1166 bad:
1167 	KASSERT(indx == -1, ("indx=%d, should be -1", indx));
1168 	fdrop(fp, td);
1169 	return (error);
1170 }
1171 
1172 #ifdef COMPAT_43
1173 /*
1174  * Create a file.
1175  */
1176 #ifndef _SYS_SYSPROTO_H_
1177 struct ocreat_args {
1178 	char	*path;
1179 	int	mode;
1180 };
1181 #endif
1182 int
1183 ocreat(td, uap)
1184 	struct thread *td;
1185 	register struct ocreat_args /* {
1186 		char *path;
1187 		int mode;
1188 	} */ *uap;
1189 {
1190 
1191 	return (kern_openat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1192 	    O_WRONLY | O_CREAT | O_TRUNC, uap->mode));
1193 }
1194 #endif /* COMPAT_43 */
1195 
1196 /*
1197  * Create a special file.
1198  */
1199 #ifndef _SYS_SYSPROTO_H_
1200 struct mknod_args {
1201 	char	*path;
1202 	int	mode;
1203 	int	dev;
1204 };
1205 #endif
1206 int
1207 sys_mknod(td, uap)
1208 	struct thread *td;
1209 	register struct mknod_args /* {
1210 		char *path;
1211 		int mode;
1212 		int dev;
1213 	} */ *uap;
1214 {
1215 
1216 	return (kern_mknodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1217 	    uap->mode, uap->dev));
1218 }
1219 
1220 #ifndef _SYS_SYSPROTO_H_
1221 struct mknodat_args {
1222 	int	fd;
1223 	char	*path;
1224 	mode_t	mode;
1225 	dev_t	dev;
1226 };
1227 #endif
1228 int
1229 sys_mknodat(struct thread *td, struct mknodat_args *uap)
1230 {
1231 
1232 	return (kern_mknodat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode,
1233 	    uap->dev));
1234 }
1235 
1236 int
1237 kern_mknodat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1238     int mode, int dev)
1239 {
1240 	struct vnode *vp;
1241 	struct mount *mp;
1242 	struct vattr vattr;
1243 	struct nameidata nd;
1244 	cap_rights_t rights;
1245 	int error, whiteout = 0;
1246 
1247 	AUDIT_ARG_MODE(mode);
1248 	AUDIT_ARG_DEV(dev);
1249 	switch (mode & S_IFMT) {
1250 	case S_IFCHR:
1251 	case S_IFBLK:
1252 		error = priv_check(td, PRIV_VFS_MKNOD_DEV);
1253 		break;
1254 	case S_IFMT:
1255 		error = priv_check(td, PRIV_VFS_MKNOD_BAD);
1256 		break;
1257 	case S_IFWHT:
1258 		error = priv_check(td, PRIV_VFS_MKNOD_WHT);
1259 		break;
1260 	case S_IFIFO:
1261 		if (dev == 0)
1262 			return (kern_mkfifoat(td, fd, path, pathseg, mode));
1263 		/* FALLTHROUGH */
1264 	default:
1265 		error = EINVAL;
1266 		break;
1267 	}
1268 	if (error != 0)
1269 		return (error);
1270 restart:
1271 	bwillwrite();
1272 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1,
1273 	    pathseg, path, fd, cap_rights_init(&rights, CAP_MKNODAT), td);
1274 	if ((error = namei(&nd)) != 0)
1275 		return (error);
1276 	vp = nd.ni_vp;
1277 	if (vp != NULL) {
1278 		NDFREE(&nd, NDF_ONLY_PNBUF);
1279 		if (vp == nd.ni_dvp)
1280 			vrele(nd.ni_dvp);
1281 		else
1282 			vput(nd.ni_dvp);
1283 		vrele(vp);
1284 		return (EEXIST);
1285 	} else {
1286 		VATTR_NULL(&vattr);
1287 		vattr.va_mode = (mode & ALLPERMS) &
1288 		    ~td->td_proc->p_fd->fd_cmask;
1289 		vattr.va_rdev = dev;
1290 		whiteout = 0;
1291 
1292 		switch (mode & S_IFMT) {
1293 		case S_IFMT:	/* used by badsect to flag bad sectors */
1294 			vattr.va_type = VBAD;
1295 			break;
1296 		case S_IFCHR:
1297 			vattr.va_type = VCHR;
1298 			break;
1299 		case S_IFBLK:
1300 			vattr.va_type = VBLK;
1301 			break;
1302 		case S_IFWHT:
1303 			whiteout = 1;
1304 			break;
1305 		default:
1306 			panic("kern_mknod: invalid mode");
1307 		}
1308 	}
1309 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1310 		NDFREE(&nd, NDF_ONLY_PNBUF);
1311 		vput(nd.ni_dvp);
1312 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1313 			return (error);
1314 		goto restart;
1315 	}
1316 #ifdef MAC
1317 	if (error == 0 && !whiteout)
1318 		error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp,
1319 		    &nd.ni_cnd, &vattr);
1320 #endif
1321 	if (error == 0) {
1322 		if (whiteout)
1323 			error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE);
1324 		else {
1325 			error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp,
1326 						&nd.ni_cnd, &vattr);
1327 			if (error == 0)
1328 				vput(nd.ni_vp);
1329 		}
1330 	}
1331 	NDFREE(&nd, NDF_ONLY_PNBUF);
1332 	vput(nd.ni_dvp);
1333 	vn_finished_write(mp);
1334 	return (error);
1335 }
1336 
1337 /*
1338  * Create a named pipe.
1339  */
1340 #ifndef _SYS_SYSPROTO_H_
1341 struct mkfifo_args {
1342 	char	*path;
1343 	int	mode;
1344 };
1345 #endif
1346 int
1347 sys_mkfifo(td, uap)
1348 	struct thread *td;
1349 	register struct mkfifo_args /* {
1350 		char *path;
1351 		int mode;
1352 	} */ *uap;
1353 {
1354 
1355 	return (kern_mkfifoat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1356 	    uap->mode));
1357 }
1358 
1359 #ifndef _SYS_SYSPROTO_H_
1360 struct mkfifoat_args {
1361 	int	fd;
1362 	char	*path;
1363 	mode_t	mode;
1364 };
1365 #endif
1366 int
1367 sys_mkfifoat(struct thread *td, struct mkfifoat_args *uap)
1368 {
1369 
1370 	return (kern_mkfifoat(td, uap->fd, uap->path, UIO_USERSPACE,
1371 	    uap->mode));
1372 }
1373 
1374 int
1375 kern_mkfifoat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1376     int mode)
1377 {
1378 	struct mount *mp;
1379 	struct vattr vattr;
1380 	struct nameidata nd;
1381 	cap_rights_t rights;
1382 	int error;
1383 
1384 	AUDIT_ARG_MODE(mode);
1385 restart:
1386 	bwillwrite();
1387 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1,
1388 	    pathseg, path, fd, cap_rights_init(&rights, CAP_MKFIFOAT), td);
1389 	if ((error = namei(&nd)) != 0)
1390 		return (error);
1391 	if (nd.ni_vp != NULL) {
1392 		NDFREE(&nd, NDF_ONLY_PNBUF);
1393 		if (nd.ni_vp == nd.ni_dvp)
1394 			vrele(nd.ni_dvp);
1395 		else
1396 			vput(nd.ni_dvp);
1397 		vrele(nd.ni_vp);
1398 		return (EEXIST);
1399 	}
1400 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1401 		NDFREE(&nd, NDF_ONLY_PNBUF);
1402 		vput(nd.ni_dvp);
1403 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1404 			return (error);
1405 		goto restart;
1406 	}
1407 	VATTR_NULL(&vattr);
1408 	vattr.va_type = VFIFO;
1409 	vattr.va_mode = (mode & ALLPERMS) & ~td->td_proc->p_fd->fd_cmask;
1410 #ifdef MAC
1411 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1412 	    &vattr);
1413 	if (error != 0)
1414 		goto out;
1415 #endif
1416 	error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
1417 	if (error == 0)
1418 		vput(nd.ni_vp);
1419 #ifdef MAC
1420 out:
1421 #endif
1422 	vput(nd.ni_dvp);
1423 	vn_finished_write(mp);
1424 	NDFREE(&nd, NDF_ONLY_PNBUF);
1425 	return (error);
1426 }
1427 
1428 /*
1429  * Make a hard file link.
1430  */
1431 #ifndef _SYS_SYSPROTO_H_
1432 struct link_args {
1433 	char	*path;
1434 	char	*link;
1435 };
1436 #endif
1437 int
1438 sys_link(td, uap)
1439 	struct thread *td;
1440 	register struct link_args /* {
1441 		char *path;
1442 		char *link;
1443 	} */ *uap;
1444 {
1445 
1446 	return (kern_linkat(td, AT_FDCWD, AT_FDCWD, uap->path, uap->link,
1447 	    UIO_USERSPACE, FOLLOW));
1448 }
1449 
1450 #ifndef _SYS_SYSPROTO_H_
1451 struct linkat_args {
1452 	int	fd1;
1453 	char	*path1;
1454 	int	fd2;
1455 	char	*path2;
1456 	int	flag;
1457 };
1458 #endif
1459 int
1460 sys_linkat(struct thread *td, struct linkat_args *uap)
1461 {
1462 	int flag;
1463 
1464 	flag = uap->flag;
1465 	if (flag & ~AT_SYMLINK_FOLLOW)
1466 		return (EINVAL);
1467 
1468 	return (kern_linkat(td, uap->fd1, uap->fd2, uap->path1, uap->path2,
1469 	    UIO_USERSPACE, (flag & AT_SYMLINK_FOLLOW) ? FOLLOW : NOFOLLOW));
1470 }
1471 
1472 int hardlink_check_uid = 0;
1473 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_uid, CTLFLAG_RW,
1474     &hardlink_check_uid, 0,
1475     "Unprivileged processes cannot create hard links to files owned by other "
1476     "users");
1477 static int hardlink_check_gid = 0;
1478 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_gid, CTLFLAG_RW,
1479     &hardlink_check_gid, 0,
1480     "Unprivileged processes cannot create hard links to files owned by other "
1481     "groups");
1482 
1483 static int
1484 can_hardlink(struct vnode *vp, struct ucred *cred)
1485 {
1486 	struct vattr va;
1487 	int error;
1488 
1489 	if (!hardlink_check_uid && !hardlink_check_gid)
1490 		return (0);
1491 
1492 	error = VOP_GETATTR(vp, &va, cred);
1493 	if (error != 0)
1494 		return (error);
1495 
1496 	if (hardlink_check_uid && cred->cr_uid != va.va_uid) {
1497 		error = priv_check_cred(cred, PRIV_VFS_LINK, 0);
1498 		if (error != 0)
1499 			return (error);
1500 	}
1501 
1502 	if (hardlink_check_gid && !groupmember(va.va_gid, cred)) {
1503 		error = priv_check_cred(cred, PRIV_VFS_LINK, 0);
1504 		if (error != 0)
1505 			return (error);
1506 	}
1507 
1508 	return (0);
1509 }
1510 
1511 int
1512 kern_linkat(struct thread *td, int fd1, int fd2, char *path1, char *path2,
1513     enum uio_seg segflg, int follow)
1514 {
1515 	struct vnode *vp;
1516 	struct mount *mp;
1517 	struct nameidata nd;
1518 	cap_rights_t rights;
1519 	int error;
1520 
1521 again:
1522 	bwillwrite();
1523 	NDINIT_AT(&nd, LOOKUP, follow | AUDITVNODE1, segflg, path1, fd1, td);
1524 
1525 	if ((error = namei(&nd)) != 0)
1526 		return (error);
1527 	NDFREE(&nd, NDF_ONLY_PNBUF);
1528 	vp = nd.ni_vp;
1529 	if (vp->v_type == VDIR) {
1530 		vrele(vp);
1531 		return (EPERM);		/* POSIX */
1532 	}
1533 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE2,
1534 	    segflg, path2, fd2, cap_rights_init(&rights, CAP_LINKAT), td);
1535 	if ((error = namei(&nd)) == 0) {
1536 		if (nd.ni_vp != NULL) {
1537 			NDFREE(&nd, NDF_ONLY_PNBUF);
1538 			if (nd.ni_dvp == nd.ni_vp)
1539 				vrele(nd.ni_dvp);
1540 			else
1541 				vput(nd.ni_dvp);
1542 			vrele(nd.ni_vp);
1543 			vrele(vp);
1544 			return (EEXIST);
1545 		} else if (nd.ni_dvp->v_mount != vp->v_mount) {
1546 			/*
1547 			 * Cross-device link.  No need to recheck
1548 			 * vp->v_type, since it cannot change, except
1549 			 * to VBAD.
1550 			 */
1551 			NDFREE(&nd, NDF_ONLY_PNBUF);
1552 			vput(nd.ni_dvp);
1553 			vrele(vp);
1554 			return (EXDEV);
1555 		} else if ((error = vn_lock(vp, LK_EXCLUSIVE)) == 0) {
1556 			error = can_hardlink(vp, td->td_ucred);
1557 #ifdef MAC
1558 			if (error == 0)
1559 				error = mac_vnode_check_link(td->td_ucred,
1560 				    nd.ni_dvp, vp, &nd.ni_cnd);
1561 #endif
1562 			if (error != 0) {
1563 				vput(vp);
1564 				vput(nd.ni_dvp);
1565 				NDFREE(&nd, NDF_ONLY_PNBUF);
1566 				return (error);
1567 			}
1568 			error = vn_start_write(vp, &mp, V_NOWAIT);
1569 			if (error != 0) {
1570 				vput(vp);
1571 				vput(nd.ni_dvp);
1572 				NDFREE(&nd, NDF_ONLY_PNBUF);
1573 				error = vn_start_write(NULL, &mp,
1574 				    V_XSLEEP | PCATCH);
1575 				if (error != 0)
1576 					return (error);
1577 				goto again;
1578 			}
1579 			error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
1580 			VOP_UNLOCK(vp, 0);
1581 			vput(nd.ni_dvp);
1582 			vn_finished_write(mp);
1583 			NDFREE(&nd, NDF_ONLY_PNBUF);
1584 		} else {
1585 			vput(nd.ni_dvp);
1586 			NDFREE(&nd, NDF_ONLY_PNBUF);
1587 			vrele(vp);
1588 			goto again;
1589 		}
1590 	}
1591 	vrele(vp);
1592 	return (error);
1593 }
1594 
1595 /*
1596  * Make a symbolic link.
1597  */
1598 #ifndef _SYS_SYSPROTO_H_
1599 struct symlink_args {
1600 	char	*path;
1601 	char	*link;
1602 };
1603 #endif
1604 int
1605 sys_symlink(td, uap)
1606 	struct thread *td;
1607 	register struct symlink_args /* {
1608 		char *path;
1609 		char *link;
1610 	} */ *uap;
1611 {
1612 
1613 	return (kern_symlinkat(td, uap->path, AT_FDCWD, uap->link,
1614 	    UIO_USERSPACE));
1615 }
1616 
1617 #ifndef _SYS_SYSPROTO_H_
1618 struct symlinkat_args {
1619 	char	*path;
1620 	int	fd;
1621 	char	*path2;
1622 };
1623 #endif
1624 int
1625 sys_symlinkat(struct thread *td, struct symlinkat_args *uap)
1626 {
1627 
1628 	return (kern_symlinkat(td, uap->path1, uap->fd, uap->path2,
1629 	    UIO_USERSPACE));
1630 }
1631 
1632 int
1633 kern_symlinkat(struct thread *td, char *path1, int fd, char *path2,
1634     enum uio_seg segflg)
1635 {
1636 	struct mount *mp;
1637 	struct vattr vattr;
1638 	char *syspath;
1639 	struct nameidata nd;
1640 	int error;
1641 	cap_rights_t rights;
1642 
1643 	if (segflg == UIO_SYSSPACE) {
1644 		syspath = path1;
1645 	} else {
1646 		syspath = uma_zalloc(namei_zone, M_WAITOK);
1647 		if ((error = copyinstr(path1, syspath, MAXPATHLEN, NULL)) != 0)
1648 			goto out;
1649 	}
1650 	AUDIT_ARG_TEXT(syspath);
1651 restart:
1652 	bwillwrite();
1653 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1,
1654 	    segflg, path2, fd, cap_rights_init(&rights, CAP_SYMLINKAT), td);
1655 	if ((error = namei(&nd)) != 0)
1656 		goto out;
1657 	if (nd.ni_vp) {
1658 		NDFREE(&nd, NDF_ONLY_PNBUF);
1659 		if (nd.ni_vp == nd.ni_dvp)
1660 			vrele(nd.ni_dvp);
1661 		else
1662 			vput(nd.ni_dvp);
1663 		vrele(nd.ni_vp);
1664 		error = EEXIST;
1665 		goto out;
1666 	}
1667 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1668 		NDFREE(&nd, NDF_ONLY_PNBUF);
1669 		vput(nd.ni_dvp);
1670 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1671 			goto out;
1672 		goto restart;
1673 	}
1674 	VATTR_NULL(&vattr);
1675 	vattr.va_mode = ACCESSPERMS &~ td->td_proc->p_fd->fd_cmask;
1676 #ifdef MAC
1677 	vattr.va_type = VLNK;
1678 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1679 	    &vattr);
1680 	if (error != 0)
1681 		goto out2;
1682 #endif
1683 	error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, syspath);
1684 	if (error == 0)
1685 		vput(nd.ni_vp);
1686 #ifdef MAC
1687 out2:
1688 #endif
1689 	NDFREE(&nd, NDF_ONLY_PNBUF);
1690 	vput(nd.ni_dvp);
1691 	vn_finished_write(mp);
1692 out:
1693 	if (segflg != UIO_SYSSPACE)
1694 		uma_zfree(namei_zone, syspath);
1695 	return (error);
1696 }
1697 
1698 /*
1699  * Delete a whiteout from the filesystem.
1700  */
1701 int
1702 sys_undelete(td, uap)
1703 	struct thread *td;
1704 	register struct undelete_args /* {
1705 		char *path;
1706 	} */ *uap;
1707 {
1708 	struct mount *mp;
1709 	struct nameidata nd;
1710 	int error;
1711 
1712 restart:
1713 	bwillwrite();
1714 	NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | AUDITVNODE1,
1715 	    UIO_USERSPACE, uap->path, td);
1716 	error = namei(&nd);
1717 	if (error != 0)
1718 		return (error);
1719 
1720 	if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) {
1721 		NDFREE(&nd, NDF_ONLY_PNBUF);
1722 		if (nd.ni_vp == nd.ni_dvp)
1723 			vrele(nd.ni_dvp);
1724 		else
1725 			vput(nd.ni_dvp);
1726 		if (nd.ni_vp)
1727 			vrele(nd.ni_vp);
1728 		return (EEXIST);
1729 	}
1730 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1731 		NDFREE(&nd, NDF_ONLY_PNBUF);
1732 		vput(nd.ni_dvp);
1733 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1734 			return (error);
1735 		goto restart;
1736 	}
1737 	error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE);
1738 	NDFREE(&nd, NDF_ONLY_PNBUF);
1739 	vput(nd.ni_dvp);
1740 	vn_finished_write(mp);
1741 	return (error);
1742 }
1743 
1744 /*
1745  * Delete a name from the filesystem.
1746  */
1747 #ifndef _SYS_SYSPROTO_H_
1748 struct unlink_args {
1749 	char	*path;
1750 };
1751 #endif
1752 int
1753 sys_unlink(td, uap)
1754 	struct thread *td;
1755 	struct unlink_args /* {
1756 		char *path;
1757 	} */ *uap;
1758 {
1759 
1760 	return (kern_unlinkat(td, AT_FDCWD, uap->path, UIO_USERSPACE, 0));
1761 }
1762 
1763 #ifndef _SYS_SYSPROTO_H_
1764 struct unlinkat_args {
1765 	int	fd;
1766 	char	*path;
1767 	int	flag;
1768 };
1769 #endif
1770 int
1771 sys_unlinkat(struct thread *td, struct unlinkat_args *uap)
1772 {
1773 	int flag = uap->flag;
1774 	int fd = uap->fd;
1775 	char *path = uap->path;
1776 
1777 	if (flag & ~AT_REMOVEDIR)
1778 		return (EINVAL);
1779 
1780 	if (flag & AT_REMOVEDIR)
1781 		return (kern_rmdirat(td, fd, path, UIO_USERSPACE));
1782 	else
1783 		return (kern_unlinkat(td, fd, path, UIO_USERSPACE, 0));
1784 }
1785 
1786 int
1787 kern_unlinkat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1788     ino_t oldinum)
1789 {
1790 	struct mount *mp;
1791 	struct vnode *vp;
1792 	struct nameidata nd;
1793 	struct stat sb;
1794 	cap_rights_t rights;
1795 	int error;
1796 
1797 restart:
1798 	bwillwrite();
1799 	NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1,
1800 	    pathseg, path, fd, cap_rights_init(&rights, CAP_UNLINKAT), td);
1801 	if ((error = namei(&nd)) != 0)
1802 		return (error == EINVAL ? EPERM : error);
1803 	vp = nd.ni_vp;
1804 	if (vp->v_type == VDIR && oldinum == 0) {
1805 		error = EPERM;		/* POSIX */
1806 	} else if (oldinum != 0 &&
1807 		  ((error = vn_stat(vp, &sb, td->td_ucred, NOCRED, td)) == 0) &&
1808 		  sb.st_ino != oldinum) {
1809 			error = EIDRM;	/* Identifier removed */
1810 	} else {
1811 		/*
1812 		 * The root of a mounted filesystem cannot be deleted.
1813 		 *
1814 		 * XXX: can this only be a VDIR case?
1815 		 */
1816 		if (vp->v_vflag & VV_ROOT)
1817 			error = EBUSY;
1818 	}
1819 	if (error == 0) {
1820 		if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1821 			NDFREE(&nd, NDF_ONLY_PNBUF);
1822 			vput(nd.ni_dvp);
1823 			if (vp == nd.ni_dvp)
1824 				vrele(vp);
1825 			else
1826 				vput(vp);
1827 			if ((error = vn_start_write(NULL, &mp,
1828 			    V_XSLEEP | PCATCH)) != 0)
1829 				return (error);
1830 			goto restart;
1831 		}
1832 #ifdef MAC
1833 		error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
1834 		    &nd.ni_cnd);
1835 		if (error != 0)
1836 			goto out;
1837 #endif
1838 		vfs_notify_upper(vp, VFS_NOTIFY_UPPER_UNLINK);
1839 		error = VOP_REMOVE(nd.ni_dvp, vp, &nd.ni_cnd);
1840 #ifdef MAC
1841 out:
1842 #endif
1843 		vn_finished_write(mp);
1844 	}
1845 	NDFREE(&nd, NDF_ONLY_PNBUF);
1846 	vput(nd.ni_dvp);
1847 	if (vp == nd.ni_dvp)
1848 		vrele(vp);
1849 	else
1850 		vput(vp);
1851 	return (error);
1852 }
1853 
1854 /*
1855  * Reposition read/write file offset.
1856  */
1857 #ifndef _SYS_SYSPROTO_H_
1858 struct lseek_args {
1859 	int	fd;
1860 	int	pad;
1861 	off_t	offset;
1862 	int	whence;
1863 };
1864 #endif
1865 int
1866 sys_lseek(td, uap)
1867 	struct thread *td;
1868 	register struct lseek_args /* {
1869 		int fd;
1870 		int pad;
1871 		off_t offset;
1872 		int whence;
1873 	} */ *uap;
1874 {
1875 	struct file *fp;
1876 	cap_rights_t rights;
1877 	int error;
1878 
1879 	AUDIT_ARG_FD(uap->fd);
1880 	error = fget(td, uap->fd, cap_rights_init(&rights, CAP_SEEK), &fp);
1881 	if (error != 0)
1882 		return (error);
1883 	error = (fp->f_ops->fo_flags & DFLAG_SEEKABLE) != 0 ?
1884 	    fo_seek(fp, uap->offset, uap->whence, td) : ESPIPE;
1885 	fdrop(fp, td);
1886 	return (error);
1887 }
1888 
1889 #if defined(COMPAT_43)
1890 /*
1891  * Reposition read/write file offset.
1892  */
1893 #ifndef _SYS_SYSPROTO_H_
1894 struct olseek_args {
1895 	int	fd;
1896 	long	offset;
1897 	int	whence;
1898 };
1899 #endif
1900 int
1901 olseek(td, uap)
1902 	struct thread *td;
1903 	register struct olseek_args /* {
1904 		int fd;
1905 		long offset;
1906 		int whence;
1907 	} */ *uap;
1908 {
1909 	struct lseek_args /* {
1910 		int fd;
1911 		int pad;
1912 		off_t offset;
1913 		int whence;
1914 	} */ nuap;
1915 
1916 	nuap.fd = uap->fd;
1917 	nuap.offset = uap->offset;
1918 	nuap.whence = uap->whence;
1919 	return (sys_lseek(td, &nuap));
1920 }
1921 #endif /* COMPAT_43 */
1922 
1923 /* Version with the 'pad' argument */
1924 int
1925 freebsd6_lseek(td, uap)
1926 	struct thread *td;
1927 	register struct freebsd6_lseek_args *uap;
1928 {
1929 	struct lseek_args ouap;
1930 
1931 	ouap.fd = uap->fd;
1932 	ouap.offset = uap->offset;
1933 	ouap.whence = uap->whence;
1934 	return (sys_lseek(td, &ouap));
1935 }
1936 
1937 /*
1938  * Check access permissions using passed credentials.
1939  */
1940 static int
1941 vn_access(vp, user_flags, cred, td)
1942 	struct vnode	*vp;
1943 	int		user_flags;
1944 	struct ucred	*cred;
1945 	struct thread	*td;
1946 {
1947 	accmode_t accmode;
1948 	int error;
1949 
1950 	/* Flags == 0 means only check for existence. */
1951 	if (user_flags == 0)
1952 		return (0);
1953 
1954 	accmode = 0;
1955 	if (user_flags & R_OK)
1956 		accmode |= VREAD;
1957 	if (user_flags & W_OK)
1958 		accmode |= VWRITE;
1959 	if (user_flags & X_OK)
1960 		accmode |= VEXEC;
1961 #ifdef MAC
1962 	error = mac_vnode_check_access(cred, vp, accmode);
1963 	if (error != 0)
1964 		return (error);
1965 #endif
1966 	if ((accmode & VWRITE) == 0 || (error = vn_writechk(vp)) == 0)
1967 		error = VOP_ACCESS(vp, accmode, cred, td);
1968 	return (error);
1969 }
1970 
1971 /*
1972  * Check access permissions using "real" credentials.
1973  */
1974 #ifndef _SYS_SYSPROTO_H_
1975 struct access_args {
1976 	char	*path;
1977 	int	amode;
1978 };
1979 #endif
1980 int
1981 sys_access(td, uap)
1982 	struct thread *td;
1983 	register struct access_args /* {
1984 		char *path;
1985 		int amode;
1986 	} */ *uap;
1987 {
1988 
1989 	return (kern_accessat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1990 	    0, uap->amode));
1991 }
1992 
1993 #ifndef _SYS_SYSPROTO_H_
1994 struct faccessat_args {
1995 	int	dirfd;
1996 	char	*path;
1997 	int	amode;
1998 	int	flag;
1999 }
2000 #endif
2001 int
2002 sys_faccessat(struct thread *td, struct faccessat_args *uap)
2003 {
2004 
2005 	return (kern_accessat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
2006 	    uap->amode));
2007 }
2008 
2009 int
2010 kern_accessat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2011     int flag, int amode)
2012 {
2013 	struct ucred *cred, *usecred;
2014 	struct vnode *vp;
2015 	struct nameidata nd;
2016 	cap_rights_t rights;
2017 	int error;
2018 
2019 	if (flag & ~AT_EACCESS)
2020 		return (EINVAL);
2021 	if (amode != F_OK && (amode & ~(R_OK | W_OK | X_OK)) != 0)
2022 		return (EINVAL);
2023 
2024 	/*
2025 	 * Create and modify a temporary credential instead of one that
2026 	 * is potentially shared (if we need one).
2027 	 */
2028 	cred = td->td_ucred;
2029 	if ((flag & AT_EACCESS) == 0 &&
2030 	    ((cred->cr_uid != cred->cr_ruid ||
2031 	    cred->cr_rgid != cred->cr_groups[0]))) {
2032 		usecred = crdup(cred);
2033 		usecred->cr_uid = cred->cr_ruid;
2034 		usecred->cr_groups[0] = cred->cr_rgid;
2035 		td->td_ucred = usecred;
2036 	} else
2037 		usecred = cred;
2038 	AUDIT_ARG_VALUE(amode);
2039 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF |
2040 	    AUDITVNODE1, pathseg, path, fd, cap_rights_init(&rights, CAP_FSTAT),
2041 	    td);
2042 	if ((error = namei(&nd)) != 0)
2043 		goto out;
2044 	vp = nd.ni_vp;
2045 
2046 	error = vn_access(vp, amode, usecred, td);
2047 	NDFREE(&nd, NDF_ONLY_PNBUF);
2048 	vput(vp);
2049 out:
2050 	if (usecred != cred) {
2051 		td->td_ucred = cred;
2052 		crfree(usecred);
2053 	}
2054 	return (error);
2055 }
2056 
2057 /*
2058  * Check access permissions using "effective" credentials.
2059  */
2060 #ifndef _SYS_SYSPROTO_H_
2061 struct eaccess_args {
2062 	char	*path;
2063 	int	amode;
2064 };
2065 #endif
2066 int
2067 sys_eaccess(td, uap)
2068 	struct thread *td;
2069 	register struct eaccess_args /* {
2070 		char *path;
2071 		int amode;
2072 	} */ *uap;
2073 {
2074 
2075 	return (kern_accessat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2076 	    AT_EACCESS, uap->amode));
2077 }
2078 
2079 #if defined(COMPAT_43)
2080 /*
2081  * Get file status; this version follows links.
2082  */
2083 #ifndef _SYS_SYSPROTO_H_
2084 struct ostat_args {
2085 	char	*path;
2086 	struct ostat *ub;
2087 };
2088 #endif
2089 int
2090 ostat(td, uap)
2091 	struct thread *td;
2092 	register struct ostat_args /* {
2093 		char *path;
2094 		struct ostat *ub;
2095 	} */ *uap;
2096 {
2097 	struct stat sb;
2098 	struct ostat osb;
2099 	int error;
2100 
2101 	error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE,
2102 	    &sb, NULL);
2103 	if (error != 0)
2104 		return (error);
2105 	cvtstat(&sb, &osb);
2106 	return (copyout(&osb, uap->ub, sizeof (osb)));
2107 }
2108 
2109 /*
2110  * Get file status; this version does not follow links.
2111  */
2112 #ifndef _SYS_SYSPROTO_H_
2113 struct olstat_args {
2114 	char	*path;
2115 	struct ostat *ub;
2116 };
2117 #endif
2118 int
2119 olstat(td, uap)
2120 	struct thread *td;
2121 	register struct olstat_args /* {
2122 		char *path;
2123 		struct ostat *ub;
2124 	} */ *uap;
2125 {
2126 	struct stat sb;
2127 	struct ostat osb;
2128 	int error;
2129 
2130 	error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2131 	    UIO_USERSPACE, &sb, NULL);
2132 	if (error != 0)
2133 		return (error);
2134 	cvtstat(&sb, &osb);
2135 	return (copyout(&osb, uap->ub, sizeof (osb)));
2136 }
2137 
2138 /*
2139  * Convert from an old to a new stat structure.
2140  */
2141 void
2142 cvtstat(st, ost)
2143 	struct stat *st;
2144 	struct ostat *ost;
2145 {
2146 
2147 	ost->st_dev = st->st_dev;
2148 	ost->st_ino = st->st_ino;
2149 	ost->st_mode = st->st_mode;
2150 	ost->st_nlink = st->st_nlink;
2151 	ost->st_uid = st->st_uid;
2152 	ost->st_gid = st->st_gid;
2153 	ost->st_rdev = st->st_rdev;
2154 	if (st->st_size < (quad_t)1 << 32)
2155 		ost->st_size = st->st_size;
2156 	else
2157 		ost->st_size = -2;
2158 	ost->st_atim = st->st_atim;
2159 	ost->st_mtim = st->st_mtim;
2160 	ost->st_ctim = st->st_ctim;
2161 	ost->st_blksize = st->st_blksize;
2162 	ost->st_blocks = st->st_blocks;
2163 	ost->st_flags = st->st_flags;
2164 	ost->st_gen = st->st_gen;
2165 }
2166 #endif /* COMPAT_43 */
2167 
2168 /*
2169  * Get file status; this version follows links.
2170  */
2171 #ifndef _SYS_SYSPROTO_H_
2172 struct stat_args {
2173 	char	*path;
2174 	struct stat *ub;
2175 };
2176 #endif
2177 int
2178 sys_stat(td, uap)
2179 	struct thread *td;
2180 	register struct stat_args /* {
2181 		char *path;
2182 		struct stat *ub;
2183 	} */ *uap;
2184 {
2185 	struct stat sb;
2186 	int error;
2187 
2188 	error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE,
2189 	    &sb, NULL);
2190 	if (error == 0)
2191 		error = copyout(&sb, uap->ub, sizeof (sb));
2192 	return (error);
2193 }
2194 
2195 #ifndef _SYS_SYSPROTO_H_
2196 struct fstatat_args {
2197 	int	fd;
2198 	char	*path;
2199 	struct stat	*buf;
2200 	int	flag;
2201 }
2202 #endif
2203 int
2204 sys_fstatat(struct thread *td, struct fstatat_args *uap)
2205 {
2206 	struct stat sb;
2207 	int error;
2208 
2209 	error = kern_statat(td, uap->flag, uap->fd, uap->path,
2210 	    UIO_USERSPACE, &sb, NULL);
2211 	if (error == 0)
2212 		error = copyout(&sb, uap->buf, sizeof (sb));
2213 	return (error);
2214 }
2215 
2216 int
2217 kern_statat(struct thread *td, int flag, int fd, char *path,
2218     enum uio_seg pathseg, struct stat *sbp,
2219     void (*hook)(struct vnode *vp, struct stat *sbp))
2220 {
2221 	struct nameidata nd;
2222 	struct stat sb;
2223 	cap_rights_t rights;
2224 	int error;
2225 
2226 	if (flag & ~AT_SYMLINK_NOFOLLOW)
2227 		return (EINVAL);
2228 
2229 	NDINIT_ATRIGHTS(&nd, LOOKUP, ((flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW :
2230 	    FOLLOW) | LOCKSHARED | LOCKLEAF | AUDITVNODE1, pathseg, path, fd,
2231 	    cap_rights_init(&rights, CAP_FSTAT), td);
2232 
2233 	if ((error = namei(&nd)) != 0)
2234 		return (error);
2235 	error = vn_stat(nd.ni_vp, &sb, td->td_ucred, NOCRED, td);
2236 	if (error == 0) {
2237 		SDT_PROBE(vfs, , stat, mode, path, sb.st_mode, 0, 0, 0);
2238 		if (S_ISREG(sb.st_mode))
2239 			SDT_PROBE(vfs, , stat, reg, path, pathseg, 0, 0, 0);
2240 		if (__predict_false(hook != NULL))
2241 			hook(nd.ni_vp, &sb);
2242 	}
2243 	NDFREE(&nd, NDF_ONLY_PNBUF);
2244 	vput(nd.ni_vp);
2245 	if (error != 0)
2246 		return (error);
2247 	*sbp = sb;
2248 #ifdef KTRACE
2249 	if (KTRPOINT(td, KTR_STRUCT))
2250 		ktrstat(&sb);
2251 #endif
2252 	return (0);
2253 }
2254 
2255 /*
2256  * Get file status; this version does not follow links.
2257  */
2258 #ifndef _SYS_SYSPROTO_H_
2259 struct lstat_args {
2260 	char	*path;
2261 	struct stat *ub;
2262 };
2263 #endif
2264 int
2265 sys_lstat(td, uap)
2266 	struct thread *td;
2267 	register struct lstat_args /* {
2268 		char *path;
2269 		struct stat *ub;
2270 	} */ *uap;
2271 {
2272 	struct stat sb;
2273 	int error;
2274 
2275 	error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2276 	    UIO_USERSPACE, &sb, NULL);
2277 	if (error == 0)
2278 		error = copyout(&sb, uap->ub, sizeof (sb));
2279 	return (error);
2280 }
2281 
2282 /*
2283  * Implementation of the NetBSD [l]stat() functions.
2284  */
2285 void
2286 cvtnstat(sb, nsb)
2287 	struct stat *sb;
2288 	struct nstat *nsb;
2289 {
2290 
2291 	bzero(nsb, sizeof *nsb);
2292 	nsb->st_dev = sb->st_dev;
2293 	nsb->st_ino = sb->st_ino;
2294 	nsb->st_mode = sb->st_mode;
2295 	nsb->st_nlink = sb->st_nlink;
2296 	nsb->st_uid = sb->st_uid;
2297 	nsb->st_gid = sb->st_gid;
2298 	nsb->st_rdev = sb->st_rdev;
2299 	nsb->st_atim = sb->st_atim;
2300 	nsb->st_mtim = sb->st_mtim;
2301 	nsb->st_ctim = sb->st_ctim;
2302 	nsb->st_size = sb->st_size;
2303 	nsb->st_blocks = sb->st_blocks;
2304 	nsb->st_blksize = sb->st_blksize;
2305 	nsb->st_flags = sb->st_flags;
2306 	nsb->st_gen = sb->st_gen;
2307 	nsb->st_birthtim = sb->st_birthtim;
2308 }
2309 
2310 #ifndef _SYS_SYSPROTO_H_
2311 struct nstat_args {
2312 	char	*path;
2313 	struct nstat *ub;
2314 };
2315 #endif
2316 int
2317 sys_nstat(td, uap)
2318 	struct thread *td;
2319 	register struct nstat_args /* {
2320 		char *path;
2321 		struct nstat *ub;
2322 	} */ *uap;
2323 {
2324 	struct stat sb;
2325 	struct nstat nsb;
2326 	int error;
2327 
2328 	error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE,
2329 	    &sb, NULL);
2330 	if (error != 0)
2331 		return (error);
2332 	cvtnstat(&sb, &nsb);
2333 	return (copyout(&nsb, uap->ub, sizeof (nsb)));
2334 }
2335 
2336 /*
2337  * NetBSD lstat.  Get file status; this version does not follow links.
2338  */
2339 #ifndef _SYS_SYSPROTO_H_
2340 struct lstat_args {
2341 	char	*path;
2342 	struct stat *ub;
2343 };
2344 #endif
2345 int
2346 sys_nlstat(td, uap)
2347 	struct thread *td;
2348 	register struct nlstat_args /* {
2349 		char *path;
2350 		struct nstat *ub;
2351 	} */ *uap;
2352 {
2353 	struct stat sb;
2354 	struct nstat nsb;
2355 	int error;
2356 
2357 	error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2358 	    UIO_USERSPACE, &sb, NULL);
2359 	if (error != 0)
2360 		return (error);
2361 	cvtnstat(&sb, &nsb);
2362 	return (copyout(&nsb, uap->ub, sizeof (nsb)));
2363 }
2364 
2365 /*
2366  * Get configurable pathname variables.
2367  */
2368 #ifndef _SYS_SYSPROTO_H_
2369 struct pathconf_args {
2370 	char	*path;
2371 	int	name;
2372 };
2373 #endif
2374 int
2375 sys_pathconf(td, uap)
2376 	struct thread *td;
2377 	register struct pathconf_args /* {
2378 		char *path;
2379 		int name;
2380 	} */ *uap;
2381 {
2382 
2383 	return (kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name, FOLLOW));
2384 }
2385 
2386 #ifndef _SYS_SYSPROTO_H_
2387 struct lpathconf_args {
2388 	char	*path;
2389 	int	name;
2390 };
2391 #endif
2392 int
2393 sys_lpathconf(td, uap)
2394 	struct thread *td;
2395 	register struct lpathconf_args /* {
2396 		char *path;
2397 		int name;
2398 	} */ *uap;
2399 {
2400 
2401 	return (kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name,
2402 	    NOFOLLOW));
2403 }
2404 
2405 int
2406 kern_pathconf(struct thread *td, char *path, enum uio_seg pathseg, int name,
2407     u_long flags)
2408 {
2409 	struct nameidata nd;
2410 	int error;
2411 
2412 	NDINIT(&nd, LOOKUP, LOCKSHARED | LOCKLEAF | AUDITVNODE1 | flags,
2413 	    pathseg, path, td);
2414 	if ((error = namei(&nd)) != 0)
2415 		return (error);
2416 	NDFREE(&nd, NDF_ONLY_PNBUF);
2417 
2418 	/* If asynchronous I/O is available, it works for all files. */
2419 	if (name == _PC_ASYNC_IO)
2420 		td->td_retval[0] = async_io_version;
2421 	else
2422 		error = VOP_PATHCONF(nd.ni_vp, name, td->td_retval);
2423 	vput(nd.ni_vp);
2424 	return (error);
2425 }
2426 
2427 /*
2428  * Return target name of a symbolic link.
2429  */
2430 #ifndef _SYS_SYSPROTO_H_
2431 struct readlink_args {
2432 	char	*path;
2433 	char	*buf;
2434 	size_t	count;
2435 };
2436 #endif
2437 int
2438 sys_readlink(td, uap)
2439 	struct thread *td;
2440 	register struct readlink_args /* {
2441 		char *path;
2442 		char *buf;
2443 		size_t count;
2444 	} */ *uap;
2445 {
2446 
2447 	return (kern_readlinkat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2448 	    uap->buf, UIO_USERSPACE, uap->count));
2449 }
2450 #ifndef _SYS_SYSPROTO_H_
2451 struct readlinkat_args {
2452 	int	fd;
2453 	char	*path;
2454 	char	*buf;
2455 	size_t	bufsize;
2456 };
2457 #endif
2458 int
2459 sys_readlinkat(struct thread *td, struct readlinkat_args *uap)
2460 {
2461 
2462 	return (kern_readlinkat(td, uap->fd, uap->path, UIO_USERSPACE,
2463 	    uap->buf, UIO_USERSPACE, uap->bufsize));
2464 }
2465 
2466 int
2467 kern_readlinkat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2468     char *buf, enum uio_seg bufseg, size_t count)
2469 {
2470 	struct vnode *vp;
2471 	struct iovec aiov;
2472 	struct uio auio;
2473 	struct nameidata nd;
2474 	int error;
2475 
2476 	if (count > IOSIZE_MAX)
2477 		return (EINVAL);
2478 
2479 	NDINIT_AT(&nd, LOOKUP, NOFOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
2480 	    pathseg, path, fd, td);
2481 
2482 	if ((error = namei(&nd)) != 0)
2483 		return (error);
2484 	NDFREE(&nd, NDF_ONLY_PNBUF);
2485 	vp = nd.ni_vp;
2486 #ifdef MAC
2487 	error = mac_vnode_check_readlink(td->td_ucred, vp);
2488 	if (error != 0) {
2489 		vput(vp);
2490 		return (error);
2491 	}
2492 #endif
2493 	if (vp->v_type != VLNK)
2494 		error = EINVAL;
2495 	else {
2496 		aiov.iov_base = buf;
2497 		aiov.iov_len = count;
2498 		auio.uio_iov = &aiov;
2499 		auio.uio_iovcnt = 1;
2500 		auio.uio_offset = 0;
2501 		auio.uio_rw = UIO_READ;
2502 		auio.uio_segflg = bufseg;
2503 		auio.uio_td = td;
2504 		auio.uio_resid = count;
2505 		error = VOP_READLINK(vp, &auio, td->td_ucred);
2506 		td->td_retval[0] = count - auio.uio_resid;
2507 	}
2508 	vput(vp);
2509 	return (error);
2510 }
2511 
2512 /*
2513  * Common implementation code for chflags() and fchflags().
2514  */
2515 static int
2516 setfflags(td, vp, flags)
2517 	struct thread *td;
2518 	struct vnode *vp;
2519 	u_long flags;
2520 {
2521 	struct mount *mp;
2522 	struct vattr vattr;
2523 	int error;
2524 
2525 	/* We can't support the value matching VNOVAL. */
2526 	if (flags == VNOVAL)
2527 		return (EOPNOTSUPP);
2528 
2529 	/*
2530 	 * Prevent non-root users from setting flags on devices.  When
2531 	 * a device is reused, users can retain ownership of the device
2532 	 * if they are allowed to set flags and programs assume that
2533 	 * chown can't fail when done as root.
2534 	 */
2535 	if (vp->v_type == VCHR || vp->v_type == VBLK) {
2536 		error = priv_check(td, PRIV_VFS_CHFLAGS_DEV);
2537 		if (error != 0)
2538 			return (error);
2539 	}
2540 
2541 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2542 		return (error);
2543 	VATTR_NULL(&vattr);
2544 	vattr.va_flags = flags;
2545 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2546 #ifdef MAC
2547 	error = mac_vnode_check_setflags(td->td_ucred, vp, vattr.va_flags);
2548 	if (error == 0)
2549 #endif
2550 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
2551 	VOP_UNLOCK(vp, 0);
2552 	vn_finished_write(mp);
2553 	return (error);
2554 }
2555 
2556 /*
2557  * Change flags of a file given a path name.
2558  */
2559 #ifndef _SYS_SYSPROTO_H_
2560 struct chflags_args {
2561 	const char *path;
2562 	u_long	flags;
2563 };
2564 #endif
2565 int
2566 sys_chflags(td, uap)
2567 	struct thread *td;
2568 	register struct chflags_args /* {
2569 		const char *path;
2570 		u_long flags;
2571 	} */ *uap;
2572 {
2573 
2574 	return (kern_chflagsat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2575 	    uap->flags, 0));
2576 }
2577 
2578 #ifndef _SYS_SYSPROTO_H_
2579 struct chflagsat_args {
2580 	int	fd;
2581 	const char *path;
2582 	u_long	flags;
2583 	int	atflag;
2584 }
2585 #endif
2586 int
2587 sys_chflagsat(struct thread *td, struct chflagsat_args *uap)
2588 {
2589 	int fd = uap->fd;
2590 	const char *path = uap->path;
2591 	u_long flags = uap->flags;
2592 	int atflag = uap->atflag;
2593 
2594 	if (atflag & ~AT_SYMLINK_NOFOLLOW)
2595 		return (EINVAL);
2596 
2597 	return (kern_chflagsat(td, fd, path, UIO_USERSPACE, flags, atflag));
2598 }
2599 
2600 /*
2601  * Same as chflags() but doesn't follow symlinks.
2602  */
2603 int
2604 sys_lchflags(td, uap)
2605 	struct thread *td;
2606 	register struct lchflags_args /* {
2607 		const char *path;
2608 		u_long flags;
2609 	} */ *uap;
2610 {
2611 
2612 	return (kern_chflagsat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2613 	    uap->flags, AT_SYMLINK_NOFOLLOW));
2614 }
2615 
2616 static int
2617 kern_chflagsat(struct thread *td, int fd, const char *path,
2618     enum uio_seg pathseg, u_long flags, int atflag)
2619 {
2620 	struct nameidata nd;
2621 	cap_rights_t rights;
2622 	int error, follow;
2623 
2624 	AUDIT_ARG_FFLAGS(flags);
2625 	follow = (atflag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
2626 	NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, pathseg, path, fd,
2627 	    cap_rights_init(&rights, CAP_FCHFLAGS), td);
2628 	if ((error = namei(&nd)) != 0)
2629 		return (error);
2630 	NDFREE(&nd, NDF_ONLY_PNBUF);
2631 	error = setfflags(td, nd.ni_vp, flags);
2632 	vrele(nd.ni_vp);
2633 	return (error);
2634 }
2635 
2636 /*
2637  * Change flags of a file given a file descriptor.
2638  */
2639 #ifndef _SYS_SYSPROTO_H_
2640 struct fchflags_args {
2641 	int	fd;
2642 	u_long	flags;
2643 };
2644 #endif
2645 int
2646 sys_fchflags(td, uap)
2647 	struct thread *td;
2648 	register struct fchflags_args /* {
2649 		int fd;
2650 		u_long flags;
2651 	} */ *uap;
2652 {
2653 	struct file *fp;
2654 	cap_rights_t rights;
2655 	int error;
2656 
2657 	AUDIT_ARG_FD(uap->fd);
2658 	AUDIT_ARG_FFLAGS(uap->flags);
2659 	error = getvnode(td->td_proc->p_fd, uap->fd,
2660 	    cap_rights_init(&rights, CAP_FCHFLAGS), &fp);
2661 	if (error != 0)
2662 		return (error);
2663 #ifdef AUDIT
2664 	vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
2665 	AUDIT_ARG_VNODE1(fp->f_vnode);
2666 	VOP_UNLOCK(fp->f_vnode, 0);
2667 #endif
2668 	error = setfflags(td, fp->f_vnode, uap->flags);
2669 	fdrop(fp, td);
2670 	return (error);
2671 }
2672 
2673 /*
2674  * Common implementation code for chmod(), lchmod() and fchmod().
2675  */
2676 int
2677 setfmode(td, cred, vp, mode)
2678 	struct thread *td;
2679 	struct ucred *cred;
2680 	struct vnode *vp;
2681 	int mode;
2682 {
2683 	struct mount *mp;
2684 	struct vattr vattr;
2685 	int error;
2686 
2687 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2688 		return (error);
2689 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2690 	VATTR_NULL(&vattr);
2691 	vattr.va_mode = mode & ALLPERMS;
2692 #ifdef MAC
2693 	error = mac_vnode_check_setmode(cred, vp, vattr.va_mode);
2694 	if (error == 0)
2695 #endif
2696 		error = VOP_SETATTR(vp, &vattr, cred);
2697 	VOP_UNLOCK(vp, 0);
2698 	vn_finished_write(mp);
2699 	return (error);
2700 }
2701 
2702 /*
2703  * Change mode of a file given path name.
2704  */
2705 #ifndef _SYS_SYSPROTO_H_
2706 struct chmod_args {
2707 	char	*path;
2708 	int	mode;
2709 };
2710 #endif
2711 int
2712 sys_chmod(td, uap)
2713 	struct thread *td;
2714 	register struct chmod_args /* {
2715 		char *path;
2716 		int mode;
2717 	} */ *uap;
2718 {
2719 
2720 	return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2721 	    uap->mode, 0));
2722 }
2723 
2724 #ifndef _SYS_SYSPROTO_H_
2725 struct fchmodat_args {
2726 	int	dirfd;
2727 	char	*path;
2728 	mode_t	mode;
2729 	int	flag;
2730 }
2731 #endif
2732 int
2733 sys_fchmodat(struct thread *td, struct fchmodat_args *uap)
2734 {
2735 	int flag = uap->flag;
2736 	int fd = uap->fd;
2737 	char *path = uap->path;
2738 	mode_t mode = uap->mode;
2739 
2740 	if (flag & ~AT_SYMLINK_NOFOLLOW)
2741 		return (EINVAL);
2742 
2743 	return (kern_fchmodat(td, fd, path, UIO_USERSPACE, mode, flag));
2744 }
2745 
2746 /*
2747  * Change mode of a file given path name (don't follow links.)
2748  */
2749 #ifndef _SYS_SYSPROTO_H_
2750 struct lchmod_args {
2751 	char	*path;
2752 	int	mode;
2753 };
2754 #endif
2755 int
2756 sys_lchmod(td, uap)
2757 	struct thread *td;
2758 	register struct lchmod_args /* {
2759 		char *path;
2760 		int mode;
2761 	} */ *uap;
2762 {
2763 
2764 	return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2765 	    uap->mode, AT_SYMLINK_NOFOLLOW));
2766 }
2767 
2768 int
2769 kern_fchmodat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2770     mode_t mode, int flag)
2771 {
2772 	struct nameidata nd;
2773 	cap_rights_t rights;
2774 	int error, follow;
2775 
2776 	AUDIT_ARG_MODE(mode);
2777 	follow = (flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
2778 	NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, pathseg, path, fd,
2779 	    cap_rights_init(&rights, CAP_FCHMOD), td);
2780 	if ((error = namei(&nd)) != 0)
2781 		return (error);
2782 	NDFREE(&nd, NDF_ONLY_PNBUF);
2783 	error = setfmode(td, td->td_ucred, nd.ni_vp, mode);
2784 	vrele(nd.ni_vp);
2785 	return (error);
2786 }
2787 
2788 /*
2789  * Change mode of a file given a file descriptor.
2790  */
2791 #ifndef _SYS_SYSPROTO_H_
2792 struct fchmod_args {
2793 	int	fd;
2794 	int	mode;
2795 };
2796 #endif
2797 int
2798 sys_fchmod(struct thread *td, struct fchmod_args *uap)
2799 {
2800 	struct file *fp;
2801 	cap_rights_t rights;
2802 	int error;
2803 
2804 	AUDIT_ARG_FD(uap->fd);
2805 	AUDIT_ARG_MODE(uap->mode);
2806 
2807 	error = fget(td, uap->fd, cap_rights_init(&rights, CAP_FCHMOD), &fp);
2808 	if (error != 0)
2809 		return (error);
2810 	error = fo_chmod(fp, uap->mode, td->td_ucred, td);
2811 	fdrop(fp, td);
2812 	return (error);
2813 }
2814 
2815 /*
2816  * Common implementation for chown(), lchown(), and fchown()
2817  */
2818 int
2819 setfown(td, cred, vp, uid, gid)
2820 	struct thread *td;
2821 	struct ucred *cred;
2822 	struct vnode *vp;
2823 	uid_t uid;
2824 	gid_t gid;
2825 {
2826 	struct mount *mp;
2827 	struct vattr vattr;
2828 	int error;
2829 
2830 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2831 		return (error);
2832 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2833 	VATTR_NULL(&vattr);
2834 	vattr.va_uid = uid;
2835 	vattr.va_gid = gid;
2836 #ifdef MAC
2837 	error = mac_vnode_check_setowner(cred, vp, vattr.va_uid,
2838 	    vattr.va_gid);
2839 	if (error == 0)
2840 #endif
2841 		error = VOP_SETATTR(vp, &vattr, cred);
2842 	VOP_UNLOCK(vp, 0);
2843 	vn_finished_write(mp);
2844 	return (error);
2845 }
2846 
2847 /*
2848  * Set ownership given a path name.
2849  */
2850 #ifndef _SYS_SYSPROTO_H_
2851 struct chown_args {
2852 	char	*path;
2853 	int	uid;
2854 	int	gid;
2855 };
2856 #endif
2857 int
2858 sys_chown(td, uap)
2859 	struct thread *td;
2860 	register struct chown_args /* {
2861 		char *path;
2862 		int uid;
2863 		int gid;
2864 	} */ *uap;
2865 {
2866 
2867 	return (kern_fchownat(td, AT_FDCWD, uap->path, UIO_USERSPACE, uap->uid,
2868 	    uap->gid, 0));
2869 }
2870 
2871 #ifndef _SYS_SYSPROTO_H_
2872 struct fchownat_args {
2873 	int fd;
2874 	const char * path;
2875 	uid_t uid;
2876 	gid_t gid;
2877 	int flag;
2878 };
2879 #endif
2880 int
2881 sys_fchownat(struct thread *td, struct fchownat_args *uap)
2882 {
2883 	int flag;
2884 
2885 	flag = uap->flag;
2886 	if (flag & ~AT_SYMLINK_NOFOLLOW)
2887 		return (EINVAL);
2888 
2889 	return (kern_fchownat(td, uap->fd, uap->path, UIO_USERSPACE, uap->uid,
2890 	    uap->gid, uap->flag));
2891 }
2892 
2893 int
2894 kern_fchownat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2895     int uid, int gid, int flag)
2896 {
2897 	struct nameidata nd;
2898 	cap_rights_t rights;
2899 	int error, follow;
2900 
2901 	AUDIT_ARG_OWNER(uid, gid);
2902 	follow = (flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
2903 	NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, pathseg, path, fd,
2904 	    cap_rights_init(&rights, CAP_FCHOWN), td);
2905 
2906 	if ((error = namei(&nd)) != 0)
2907 		return (error);
2908 	NDFREE(&nd, NDF_ONLY_PNBUF);
2909 	error = setfown(td, td->td_ucred, nd.ni_vp, uid, gid);
2910 	vrele(nd.ni_vp);
2911 	return (error);
2912 }
2913 
2914 /*
2915  * Set ownership given a path name, do not cross symlinks.
2916  */
2917 #ifndef _SYS_SYSPROTO_H_
2918 struct lchown_args {
2919 	char	*path;
2920 	int	uid;
2921 	int	gid;
2922 };
2923 #endif
2924 int
2925 sys_lchown(td, uap)
2926 	struct thread *td;
2927 	register struct lchown_args /* {
2928 		char *path;
2929 		int uid;
2930 		int gid;
2931 	} */ *uap;
2932 {
2933 
2934 	return (kern_fchownat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2935 	    uap->uid, uap->gid, AT_SYMLINK_NOFOLLOW));
2936 }
2937 
2938 /*
2939  * Set ownership given a file descriptor.
2940  */
2941 #ifndef _SYS_SYSPROTO_H_
2942 struct fchown_args {
2943 	int	fd;
2944 	int	uid;
2945 	int	gid;
2946 };
2947 #endif
2948 int
2949 sys_fchown(td, uap)
2950 	struct thread *td;
2951 	register struct fchown_args /* {
2952 		int fd;
2953 		int uid;
2954 		int gid;
2955 	} */ *uap;
2956 {
2957 	struct file *fp;
2958 	cap_rights_t rights;
2959 	int error;
2960 
2961 	AUDIT_ARG_FD(uap->fd);
2962 	AUDIT_ARG_OWNER(uap->uid, uap->gid);
2963 	error = fget(td, uap->fd, cap_rights_init(&rights, CAP_FCHOWN), &fp);
2964 	if (error != 0)
2965 		return (error);
2966 	error = fo_chown(fp, uap->uid, uap->gid, td->td_ucred, td);
2967 	fdrop(fp, td);
2968 	return (error);
2969 }
2970 
2971 /*
2972  * Common implementation code for utimes(), lutimes(), and futimes().
2973  */
2974 static int
2975 getutimes(usrtvp, tvpseg, tsp)
2976 	const struct timeval *usrtvp;
2977 	enum uio_seg tvpseg;
2978 	struct timespec *tsp;
2979 {
2980 	struct timeval tv[2];
2981 	const struct timeval *tvp;
2982 	int error;
2983 
2984 	if (usrtvp == NULL) {
2985 		vfs_timestamp(&tsp[0]);
2986 		tsp[1] = tsp[0];
2987 	} else {
2988 		if (tvpseg == UIO_SYSSPACE) {
2989 			tvp = usrtvp;
2990 		} else {
2991 			if ((error = copyin(usrtvp, tv, sizeof(tv))) != 0)
2992 				return (error);
2993 			tvp = tv;
2994 		}
2995 
2996 		if (tvp[0].tv_usec < 0 || tvp[0].tv_usec >= 1000000 ||
2997 		    tvp[1].tv_usec < 0 || tvp[1].tv_usec >= 1000000)
2998 			return (EINVAL);
2999 		TIMEVAL_TO_TIMESPEC(&tvp[0], &tsp[0]);
3000 		TIMEVAL_TO_TIMESPEC(&tvp[1], &tsp[1]);
3001 	}
3002 	return (0);
3003 }
3004 
3005 /*
3006  * Common implementation code for utimes(), lutimes(), and futimes().
3007  */
3008 static int
3009 setutimes(td, vp, ts, numtimes, nullflag)
3010 	struct thread *td;
3011 	struct vnode *vp;
3012 	const struct timespec *ts;
3013 	int numtimes;
3014 	int nullflag;
3015 {
3016 	struct mount *mp;
3017 	struct vattr vattr;
3018 	int error, setbirthtime;
3019 
3020 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
3021 		return (error);
3022 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3023 	setbirthtime = 0;
3024 	if (numtimes < 3 && !VOP_GETATTR(vp, &vattr, td->td_ucred) &&
3025 	    timespeccmp(&ts[1], &vattr.va_birthtime, < ))
3026 		setbirthtime = 1;
3027 	VATTR_NULL(&vattr);
3028 	vattr.va_atime = ts[0];
3029 	vattr.va_mtime = ts[1];
3030 	if (setbirthtime)
3031 		vattr.va_birthtime = ts[1];
3032 	if (numtimes > 2)
3033 		vattr.va_birthtime = ts[2];
3034 	if (nullflag)
3035 		vattr.va_vaflags |= VA_UTIMES_NULL;
3036 #ifdef MAC
3037 	error = mac_vnode_check_setutimes(td->td_ucred, vp, vattr.va_atime,
3038 	    vattr.va_mtime);
3039 #endif
3040 	if (error == 0)
3041 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3042 	VOP_UNLOCK(vp, 0);
3043 	vn_finished_write(mp);
3044 	return (error);
3045 }
3046 
3047 /*
3048  * Set the access and modification times of a file.
3049  */
3050 #ifndef _SYS_SYSPROTO_H_
3051 struct utimes_args {
3052 	char	*path;
3053 	struct	timeval *tptr;
3054 };
3055 #endif
3056 int
3057 sys_utimes(td, uap)
3058 	struct thread *td;
3059 	register struct utimes_args /* {
3060 		char *path;
3061 		struct timeval *tptr;
3062 	} */ *uap;
3063 {
3064 
3065 	return (kern_utimesat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
3066 	    uap->tptr, UIO_USERSPACE));
3067 }
3068 
3069 #ifndef _SYS_SYSPROTO_H_
3070 struct futimesat_args {
3071 	int fd;
3072 	const char * path;
3073 	const struct timeval * times;
3074 };
3075 #endif
3076 int
3077 sys_futimesat(struct thread *td, struct futimesat_args *uap)
3078 {
3079 
3080 	return (kern_utimesat(td, uap->fd, uap->path, UIO_USERSPACE,
3081 	    uap->times, UIO_USERSPACE));
3082 }
3083 
3084 int
3085 kern_utimesat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
3086     struct timeval *tptr, enum uio_seg tptrseg)
3087 {
3088 	struct nameidata nd;
3089 	struct timespec ts[2];
3090 	cap_rights_t rights;
3091 	int error;
3092 
3093 	if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3094 		return (error);
3095 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, fd,
3096 	    cap_rights_init(&rights, CAP_FUTIMES), td);
3097 
3098 	if ((error = namei(&nd)) != 0)
3099 		return (error);
3100 	NDFREE(&nd, NDF_ONLY_PNBUF);
3101 	error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3102 	vrele(nd.ni_vp);
3103 	return (error);
3104 }
3105 
3106 /*
3107  * Set the access and modification times of a file.
3108  */
3109 #ifndef _SYS_SYSPROTO_H_
3110 struct lutimes_args {
3111 	char	*path;
3112 	struct	timeval *tptr;
3113 };
3114 #endif
3115 int
3116 sys_lutimes(td, uap)
3117 	struct thread *td;
3118 	register struct lutimes_args /* {
3119 		char *path;
3120 		struct timeval *tptr;
3121 	} */ *uap;
3122 {
3123 
3124 	return (kern_lutimes(td, uap->path, UIO_USERSPACE, uap->tptr,
3125 	    UIO_USERSPACE));
3126 }
3127 
3128 int
3129 kern_lutimes(struct thread *td, char *path, enum uio_seg pathseg,
3130     struct timeval *tptr, enum uio_seg tptrseg)
3131 {
3132 	struct timespec ts[2];
3133 	struct nameidata nd;
3134 	int error;
3135 
3136 	if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3137 		return (error);
3138 	NDINIT(&nd, LOOKUP, NOFOLLOW | AUDITVNODE1, pathseg, path, td);
3139 	if ((error = namei(&nd)) != 0)
3140 		return (error);
3141 	NDFREE(&nd, NDF_ONLY_PNBUF);
3142 	error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3143 	vrele(nd.ni_vp);
3144 	return (error);
3145 }
3146 
3147 /*
3148  * Set the access and modification times of a file.
3149  */
3150 #ifndef _SYS_SYSPROTO_H_
3151 struct futimes_args {
3152 	int	fd;
3153 	struct	timeval *tptr;
3154 };
3155 #endif
3156 int
3157 sys_futimes(td, uap)
3158 	struct thread *td;
3159 	register struct futimes_args /* {
3160 		int  fd;
3161 		struct timeval *tptr;
3162 	} */ *uap;
3163 {
3164 
3165 	return (kern_futimes(td, uap->fd, uap->tptr, UIO_USERSPACE));
3166 }
3167 
3168 int
3169 kern_futimes(struct thread *td, int fd, struct timeval *tptr,
3170     enum uio_seg tptrseg)
3171 {
3172 	struct timespec ts[2];
3173 	struct file *fp;
3174 	cap_rights_t rights;
3175 	int error;
3176 
3177 	AUDIT_ARG_FD(fd);
3178 	error = getutimes(tptr, tptrseg, ts);
3179 	if (error != 0)
3180 		return (error);
3181 	error = getvnode(td->td_proc->p_fd, fd,
3182 	    cap_rights_init(&rights, CAP_FUTIMES), &fp);
3183 	if (error != 0)
3184 		return (error);
3185 #ifdef AUDIT
3186 	vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
3187 	AUDIT_ARG_VNODE1(fp->f_vnode);
3188 	VOP_UNLOCK(fp->f_vnode, 0);
3189 #endif
3190 	error = setutimes(td, fp->f_vnode, ts, 2, tptr == NULL);
3191 	fdrop(fp, td);
3192 	return (error);
3193 }
3194 
3195 /*
3196  * Truncate a file given its path name.
3197  */
3198 #ifndef _SYS_SYSPROTO_H_
3199 struct truncate_args {
3200 	char	*path;
3201 	int	pad;
3202 	off_t	length;
3203 };
3204 #endif
3205 int
3206 sys_truncate(td, uap)
3207 	struct thread *td;
3208 	register struct truncate_args /* {
3209 		char *path;
3210 		int pad;
3211 		off_t length;
3212 	} */ *uap;
3213 {
3214 
3215 	return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3216 }
3217 
3218 int
3219 kern_truncate(struct thread *td, char *path, enum uio_seg pathseg, off_t length)
3220 {
3221 	struct mount *mp;
3222 	struct vnode *vp;
3223 	void *rl_cookie;
3224 	struct vattr vattr;
3225 	struct nameidata nd;
3226 	int error;
3227 
3228 	if (length < 0)
3229 		return(EINVAL);
3230 	NDINIT(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, td);
3231 	if ((error = namei(&nd)) != 0)
3232 		return (error);
3233 	vp = nd.ni_vp;
3234 	rl_cookie = vn_rangelock_wlock(vp, 0, OFF_MAX);
3235 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) {
3236 		vn_rangelock_unlock(vp, rl_cookie);
3237 		vrele(vp);
3238 		return (error);
3239 	}
3240 	NDFREE(&nd, NDF_ONLY_PNBUF);
3241 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3242 	if (vp->v_type == VDIR)
3243 		error = EISDIR;
3244 #ifdef MAC
3245 	else if ((error = mac_vnode_check_write(td->td_ucred, NOCRED, vp))) {
3246 	}
3247 #endif
3248 	else if ((error = vn_writechk(vp)) == 0 &&
3249 	    (error = VOP_ACCESS(vp, VWRITE, td->td_ucred, td)) == 0) {
3250 		VATTR_NULL(&vattr);
3251 		vattr.va_size = length;
3252 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3253 	}
3254 	VOP_UNLOCK(vp, 0);
3255 	vn_finished_write(mp);
3256 	vn_rangelock_unlock(vp, rl_cookie);
3257 	vrele(vp);
3258 	return (error);
3259 }
3260 
3261 #if defined(COMPAT_43)
3262 /*
3263  * Truncate a file given its path name.
3264  */
3265 #ifndef _SYS_SYSPROTO_H_
3266 struct otruncate_args {
3267 	char	*path;
3268 	long	length;
3269 };
3270 #endif
3271 int
3272 otruncate(td, uap)
3273 	struct thread *td;
3274 	register struct otruncate_args /* {
3275 		char *path;
3276 		long length;
3277 	} */ *uap;
3278 {
3279 	struct truncate_args /* {
3280 		char *path;
3281 		int pad;
3282 		off_t length;
3283 	} */ nuap;
3284 
3285 	nuap.path = uap->path;
3286 	nuap.length = uap->length;
3287 	return (sys_truncate(td, &nuap));
3288 }
3289 #endif /* COMPAT_43 */
3290 
3291 /* Versions with the pad argument */
3292 int
3293 freebsd6_truncate(struct thread *td, struct freebsd6_truncate_args *uap)
3294 {
3295 	struct truncate_args ouap;
3296 
3297 	ouap.path = uap->path;
3298 	ouap.length = uap->length;
3299 	return (sys_truncate(td, &ouap));
3300 }
3301 
3302 int
3303 freebsd6_ftruncate(struct thread *td, struct freebsd6_ftruncate_args *uap)
3304 {
3305 	struct ftruncate_args ouap;
3306 
3307 	ouap.fd = uap->fd;
3308 	ouap.length = uap->length;
3309 	return (sys_ftruncate(td, &ouap));
3310 }
3311 
3312 /*
3313  * Sync an open file.
3314  */
3315 #ifndef _SYS_SYSPROTO_H_
3316 struct fsync_args {
3317 	int	fd;
3318 };
3319 #endif
3320 int
3321 sys_fsync(td, uap)
3322 	struct thread *td;
3323 	struct fsync_args /* {
3324 		int fd;
3325 	} */ *uap;
3326 {
3327 	struct vnode *vp;
3328 	struct mount *mp;
3329 	struct file *fp;
3330 	cap_rights_t rights;
3331 	int error, lock_flags;
3332 
3333 	AUDIT_ARG_FD(uap->fd);
3334 	error = getvnode(td->td_proc->p_fd, uap->fd,
3335 	    cap_rights_init(&rights, CAP_FSYNC), &fp);
3336 	if (error != 0)
3337 		return (error);
3338 	vp = fp->f_vnode;
3339 	error = vn_start_write(vp, &mp, V_WAIT | PCATCH);
3340 	if (error != 0)
3341 		goto drop;
3342 	if (MNT_SHARED_WRITES(mp) ||
3343 	    ((mp == NULL) && MNT_SHARED_WRITES(vp->v_mount))) {
3344 		lock_flags = LK_SHARED;
3345 	} else {
3346 		lock_flags = LK_EXCLUSIVE;
3347 	}
3348 	vn_lock(vp, lock_flags | LK_RETRY);
3349 	AUDIT_ARG_VNODE1(vp);
3350 	if (vp->v_object != NULL) {
3351 		VM_OBJECT_WLOCK(vp->v_object);
3352 		vm_object_page_clean(vp->v_object, 0, 0, 0);
3353 		VM_OBJECT_WUNLOCK(vp->v_object);
3354 	}
3355 	error = VOP_FSYNC(vp, MNT_WAIT, td);
3356 
3357 	VOP_UNLOCK(vp, 0);
3358 	vn_finished_write(mp);
3359 drop:
3360 	fdrop(fp, td);
3361 	return (error);
3362 }
3363 
3364 /*
3365  * Rename files.  Source and destination must either both be directories, or
3366  * both not be directories.  If target is a directory, it must be empty.
3367  */
3368 #ifndef _SYS_SYSPROTO_H_
3369 struct rename_args {
3370 	char	*from;
3371 	char	*to;
3372 };
3373 #endif
3374 int
3375 sys_rename(td, uap)
3376 	struct thread *td;
3377 	register struct rename_args /* {
3378 		char *from;
3379 		char *to;
3380 	} */ *uap;
3381 {
3382 
3383 	return (kern_renameat(td, AT_FDCWD, uap->from, AT_FDCWD,
3384 	    uap->to, UIO_USERSPACE));
3385 }
3386 
3387 #ifndef _SYS_SYSPROTO_H_
3388 struct renameat_args {
3389 	int	oldfd;
3390 	char	*old;
3391 	int	newfd;
3392 	char	*new;
3393 };
3394 #endif
3395 int
3396 sys_renameat(struct thread *td, struct renameat_args *uap)
3397 {
3398 
3399 	return (kern_renameat(td, uap->oldfd, uap->old, uap->newfd, uap->new,
3400 	    UIO_USERSPACE));
3401 }
3402 
3403 int
3404 kern_renameat(struct thread *td, int oldfd, char *old, int newfd, char *new,
3405     enum uio_seg pathseg)
3406 {
3407 	struct mount *mp = NULL;
3408 	struct vnode *tvp, *fvp, *tdvp;
3409 	struct nameidata fromnd, tond;
3410 	cap_rights_t rights;
3411 	int error;
3412 
3413 again:
3414 	bwillwrite();
3415 #ifdef MAC
3416 	NDINIT_ATRIGHTS(&fromnd, DELETE, LOCKPARENT | LOCKLEAF | SAVESTART |
3417 	    AUDITVNODE1, pathseg, old, oldfd,
3418 	    cap_rights_init(&rights, CAP_RENAMEAT), td);
3419 #else
3420 	NDINIT_ATRIGHTS(&fromnd, DELETE, WANTPARENT | SAVESTART | AUDITVNODE1,
3421 	    pathseg, old, oldfd, cap_rights_init(&rights, CAP_RENAMEAT), td);
3422 #endif
3423 
3424 	if ((error = namei(&fromnd)) != 0)
3425 		return (error);
3426 #ifdef MAC
3427 	error = mac_vnode_check_rename_from(td->td_ucred, fromnd.ni_dvp,
3428 	    fromnd.ni_vp, &fromnd.ni_cnd);
3429 	VOP_UNLOCK(fromnd.ni_dvp, 0);
3430 	if (fromnd.ni_dvp != fromnd.ni_vp)
3431 		VOP_UNLOCK(fromnd.ni_vp, 0);
3432 #endif
3433 	fvp = fromnd.ni_vp;
3434 	NDINIT_ATRIGHTS(&tond, RENAME, LOCKPARENT | LOCKLEAF | NOCACHE |
3435 	    SAVESTART | AUDITVNODE2, pathseg, new, newfd,
3436 	    cap_rights_init(&rights, CAP_LINKAT), td);
3437 	if (fromnd.ni_vp->v_type == VDIR)
3438 		tond.ni_cnd.cn_flags |= WILLBEDIR;
3439 	if ((error = namei(&tond)) != 0) {
3440 		/* Translate error code for rename("dir1", "dir2/."). */
3441 		if (error == EISDIR && fvp->v_type == VDIR)
3442 			error = EINVAL;
3443 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3444 		vrele(fromnd.ni_dvp);
3445 		vrele(fvp);
3446 		goto out1;
3447 	}
3448 	tdvp = tond.ni_dvp;
3449 	tvp = tond.ni_vp;
3450 	error = vn_start_write(fvp, &mp, V_NOWAIT);
3451 	if (error != 0) {
3452 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3453 		NDFREE(&tond, NDF_ONLY_PNBUF);
3454 		if (tvp != NULL)
3455 			vput(tvp);
3456 		if (tdvp == tvp)
3457 			vrele(tdvp);
3458 		else
3459 			vput(tdvp);
3460 		vrele(fromnd.ni_dvp);
3461 		vrele(fvp);
3462 		vrele(tond.ni_startdir);
3463 		if (fromnd.ni_startdir != NULL)
3464 			vrele(fromnd.ni_startdir);
3465 		error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH);
3466 		if (error != 0)
3467 			return (error);
3468 		goto again;
3469 	}
3470 	if (tvp != NULL) {
3471 		if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
3472 			error = ENOTDIR;
3473 			goto out;
3474 		} else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
3475 			error = EISDIR;
3476 			goto out;
3477 		}
3478 #ifdef CAPABILITIES
3479 		if (newfd != AT_FDCWD) {
3480 			/*
3481 			 * If the target already exists we require CAP_UNLINKAT
3482 			 * from 'newfd'.
3483 			 */
3484 			error = cap_check(&tond.ni_filecaps.fc_rights,
3485 			    cap_rights_init(&rights, CAP_UNLINKAT));
3486 			if (error != 0)
3487 				goto out;
3488 		}
3489 #endif
3490 	}
3491 	if (fvp == tdvp) {
3492 		error = EINVAL;
3493 		goto out;
3494 	}
3495 	/*
3496 	 * If the source is the same as the destination (that is, if they
3497 	 * are links to the same vnode), then there is nothing to do.
3498 	 */
3499 	if (fvp == tvp)
3500 		error = -1;
3501 #ifdef MAC
3502 	else
3503 		error = mac_vnode_check_rename_to(td->td_ucred, tdvp,
3504 		    tond.ni_vp, fromnd.ni_dvp == tdvp, &tond.ni_cnd);
3505 #endif
3506 out:
3507 	if (error == 0) {
3508 		error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
3509 		    tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
3510 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3511 		NDFREE(&tond, NDF_ONLY_PNBUF);
3512 	} else {
3513 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3514 		NDFREE(&tond, NDF_ONLY_PNBUF);
3515 		if (tvp != NULL)
3516 			vput(tvp);
3517 		if (tdvp == tvp)
3518 			vrele(tdvp);
3519 		else
3520 			vput(tdvp);
3521 		vrele(fromnd.ni_dvp);
3522 		vrele(fvp);
3523 	}
3524 	vrele(tond.ni_startdir);
3525 	vn_finished_write(mp);
3526 out1:
3527 	if (fromnd.ni_startdir)
3528 		vrele(fromnd.ni_startdir);
3529 	if (error == -1)
3530 		return (0);
3531 	return (error);
3532 }
3533 
3534 /*
3535  * Make a directory file.
3536  */
3537 #ifndef _SYS_SYSPROTO_H_
3538 struct mkdir_args {
3539 	char	*path;
3540 	int	mode;
3541 };
3542 #endif
3543 int
3544 sys_mkdir(td, uap)
3545 	struct thread *td;
3546 	register struct mkdir_args /* {
3547 		char *path;
3548 		int mode;
3549 	} */ *uap;
3550 {
3551 
3552 	return (kern_mkdirat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
3553 	    uap->mode));
3554 }
3555 
3556 #ifndef _SYS_SYSPROTO_H_
3557 struct mkdirat_args {
3558 	int	fd;
3559 	char	*path;
3560 	mode_t	mode;
3561 };
3562 #endif
3563 int
3564 sys_mkdirat(struct thread *td, struct mkdirat_args *uap)
3565 {
3566 
3567 	return (kern_mkdirat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode));
3568 }
3569 
3570 int
3571 kern_mkdirat(struct thread *td, int fd, char *path, enum uio_seg segflg,
3572     int mode)
3573 {
3574 	struct mount *mp;
3575 	struct vnode *vp;
3576 	struct vattr vattr;
3577 	struct nameidata nd;
3578 	cap_rights_t rights;
3579 	int error;
3580 
3581 	AUDIT_ARG_MODE(mode);
3582 restart:
3583 	bwillwrite();
3584 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1,
3585 	    segflg, path, fd, cap_rights_init(&rights, CAP_MKDIRAT), td);
3586 	nd.ni_cnd.cn_flags |= WILLBEDIR;
3587 	if ((error = namei(&nd)) != 0)
3588 		return (error);
3589 	vp = nd.ni_vp;
3590 	if (vp != NULL) {
3591 		NDFREE(&nd, NDF_ONLY_PNBUF);
3592 		/*
3593 		 * XXX namei called with LOCKPARENT but not LOCKLEAF has
3594 		 * the strange behaviour of leaving the vnode unlocked
3595 		 * if the target is the same vnode as the parent.
3596 		 */
3597 		if (vp == nd.ni_dvp)
3598 			vrele(nd.ni_dvp);
3599 		else
3600 			vput(nd.ni_dvp);
3601 		vrele(vp);
3602 		return (EEXIST);
3603 	}
3604 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3605 		NDFREE(&nd, NDF_ONLY_PNBUF);
3606 		vput(nd.ni_dvp);
3607 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3608 			return (error);
3609 		goto restart;
3610 	}
3611 	VATTR_NULL(&vattr);
3612 	vattr.va_type = VDIR;
3613 	vattr.va_mode = (mode & ACCESSPERMS) &~ td->td_proc->p_fd->fd_cmask;
3614 #ifdef MAC
3615 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
3616 	    &vattr);
3617 	if (error != 0)
3618 		goto out;
3619 #endif
3620 	error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
3621 #ifdef MAC
3622 out:
3623 #endif
3624 	NDFREE(&nd, NDF_ONLY_PNBUF);
3625 	vput(nd.ni_dvp);
3626 	if (error == 0)
3627 		vput(nd.ni_vp);
3628 	vn_finished_write(mp);
3629 	return (error);
3630 }
3631 
3632 /*
3633  * Remove a directory file.
3634  */
3635 #ifndef _SYS_SYSPROTO_H_
3636 struct rmdir_args {
3637 	char	*path;
3638 };
3639 #endif
3640 int
3641 sys_rmdir(td, uap)
3642 	struct thread *td;
3643 	struct rmdir_args /* {
3644 		char *path;
3645 	} */ *uap;
3646 {
3647 
3648 	return (kern_rmdirat(td, AT_FDCWD, uap->path, UIO_USERSPACE));
3649 }
3650 
3651 int
3652 kern_rmdirat(struct thread *td, int fd, char *path, enum uio_seg pathseg)
3653 {
3654 	struct mount *mp;
3655 	struct vnode *vp;
3656 	struct nameidata nd;
3657 	cap_rights_t rights;
3658 	int error;
3659 
3660 restart:
3661 	bwillwrite();
3662 	NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1,
3663 	    pathseg, path, fd, cap_rights_init(&rights, CAP_UNLINKAT), td);
3664 	if ((error = namei(&nd)) != 0)
3665 		return (error);
3666 	vp = nd.ni_vp;
3667 	if (vp->v_type != VDIR) {
3668 		error = ENOTDIR;
3669 		goto out;
3670 	}
3671 	/*
3672 	 * No rmdir "." please.
3673 	 */
3674 	if (nd.ni_dvp == vp) {
3675 		error = EINVAL;
3676 		goto out;
3677 	}
3678 	/*
3679 	 * The root of a mounted filesystem cannot be deleted.
3680 	 */
3681 	if (vp->v_vflag & VV_ROOT) {
3682 		error = EBUSY;
3683 		goto out;
3684 	}
3685 #ifdef MAC
3686 	error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
3687 	    &nd.ni_cnd);
3688 	if (error != 0)
3689 		goto out;
3690 #endif
3691 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3692 		NDFREE(&nd, NDF_ONLY_PNBUF);
3693 		vput(vp);
3694 		if (nd.ni_dvp == vp)
3695 			vrele(nd.ni_dvp);
3696 		else
3697 			vput(nd.ni_dvp);
3698 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3699 			return (error);
3700 		goto restart;
3701 	}
3702 	vfs_notify_upper(vp, VFS_NOTIFY_UPPER_UNLINK);
3703 	error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
3704 	vn_finished_write(mp);
3705 out:
3706 	NDFREE(&nd, NDF_ONLY_PNBUF);
3707 	vput(vp);
3708 	if (nd.ni_dvp == vp)
3709 		vrele(nd.ni_dvp);
3710 	else
3711 		vput(nd.ni_dvp);
3712 	return (error);
3713 }
3714 
3715 #ifdef COMPAT_43
3716 /*
3717  * Read a block of directory entries in a filesystem independent format.
3718  */
3719 #ifndef _SYS_SYSPROTO_H_
3720 struct ogetdirentries_args {
3721 	int	fd;
3722 	char	*buf;
3723 	u_int	count;
3724 	long	*basep;
3725 };
3726 #endif
3727 int
3728 ogetdirentries(struct thread *td, struct ogetdirentries_args *uap)
3729 {
3730 	long loff;
3731 	int error;
3732 
3733 	error = kern_ogetdirentries(td, uap, &loff);
3734 	if (error == 0)
3735 		error = copyout(&loff, uap->basep, sizeof(long));
3736 	return (error);
3737 }
3738 
3739 int
3740 kern_ogetdirentries(struct thread *td, struct ogetdirentries_args *uap,
3741     long *ploff)
3742 {
3743 	struct vnode *vp;
3744 	struct file *fp;
3745 	struct uio auio, kuio;
3746 	struct iovec aiov, kiov;
3747 	struct dirent *dp, *edp;
3748 	cap_rights_t rights;
3749 	caddr_t dirbuf;
3750 	int error, eofflag, readcnt;
3751 	long loff;
3752 	off_t foffset;
3753 
3754 	/* XXX arbitrary sanity limit on `count'. */
3755 	if (uap->count > 64 * 1024)
3756 		return (EINVAL);
3757 	error = getvnode(td->td_proc->p_fd, uap->fd,
3758 	    cap_rights_init(&rights, CAP_READ), &fp);
3759 	if (error != 0)
3760 		return (error);
3761 	if ((fp->f_flag & FREAD) == 0) {
3762 		fdrop(fp, td);
3763 		return (EBADF);
3764 	}
3765 	vp = fp->f_vnode;
3766 	foffset = foffset_lock(fp, 0);
3767 unionread:
3768 	if (vp->v_type != VDIR) {
3769 		foffset_unlock(fp, foffset, 0);
3770 		fdrop(fp, td);
3771 		return (EINVAL);
3772 	}
3773 	aiov.iov_base = uap->buf;
3774 	aiov.iov_len = uap->count;
3775 	auio.uio_iov = &aiov;
3776 	auio.uio_iovcnt = 1;
3777 	auio.uio_rw = UIO_READ;
3778 	auio.uio_segflg = UIO_USERSPACE;
3779 	auio.uio_td = td;
3780 	auio.uio_resid = uap->count;
3781 	vn_lock(vp, LK_SHARED | LK_RETRY);
3782 	loff = auio.uio_offset = foffset;
3783 #ifdef MAC
3784 	error = mac_vnode_check_readdir(td->td_ucred, vp);
3785 	if (error != 0) {
3786 		VOP_UNLOCK(vp, 0);
3787 		foffset_unlock(fp, foffset, FOF_NOUPDATE);
3788 		fdrop(fp, td);
3789 		return (error);
3790 	}
3791 #endif
3792 #	if (BYTE_ORDER != LITTLE_ENDIAN)
3793 		if (vp->v_mount->mnt_maxsymlinklen <= 0) {
3794 			error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag,
3795 			    NULL, NULL);
3796 			foffset = auio.uio_offset;
3797 		} else
3798 #	endif
3799 	{
3800 		kuio = auio;
3801 		kuio.uio_iov = &kiov;
3802 		kuio.uio_segflg = UIO_SYSSPACE;
3803 		kiov.iov_len = uap->count;
3804 		dirbuf = malloc(uap->count, M_TEMP, M_WAITOK);
3805 		kiov.iov_base = dirbuf;
3806 		error = VOP_READDIR(vp, &kuio, fp->f_cred, &eofflag,
3807 			    NULL, NULL);
3808 		foffset = kuio.uio_offset;
3809 		if (error == 0) {
3810 			readcnt = uap->count - kuio.uio_resid;
3811 			edp = (struct dirent *)&dirbuf[readcnt];
3812 			for (dp = (struct dirent *)dirbuf; dp < edp; ) {
3813 #				if (BYTE_ORDER == LITTLE_ENDIAN)
3814 					/*
3815 					 * The expected low byte of
3816 					 * dp->d_namlen is our dp->d_type.
3817 					 * The high MBZ byte of dp->d_namlen
3818 					 * is our dp->d_namlen.
3819 					 */
3820 					dp->d_type = dp->d_namlen;
3821 					dp->d_namlen = 0;
3822 #				else
3823 					/*
3824 					 * The dp->d_type is the high byte
3825 					 * of the expected dp->d_namlen,
3826 					 * so must be zero'ed.
3827 					 */
3828 					dp->d_type = 0;
3829 #				endif
3830 				if (dp->d_reclen > 0) {
3831 					dp = (struct dirent *)
3832 					    ((char *)dp + dp->d_reclen);
3833 				} else {
3834 					error = EIO;
3835 					break;
3836 				}
3837 			}
3838 			if (dp >= edp)
3839 				error = uiomove(dirbuf, readcnt, &auio);
3840 		}
3841 		free(dirbuf, M_TEMP);
3842 	}
3843 	if (error != 0) {
3844 		VOP_UNLOCK(vp, 0);
3845 		foffset_unlock(fp, foffset, 0);
3846 		fdrop(fp, td);
3847 		return (error);
3848 	}
3849 	if (uap->count == auio.uio_resid &&
3850 	    (vp->v_vflag & VV_ROOT) &&
3851 	    (vp->v_mount->mnt_flag & MNT_UNION)) {
3852 		struct vnode *tvp = vp;
3853 		vp = vp->v_mount->mnt_vnodecovered;
3854 		VREF(vp);
3855 		fp->f_vnode = vp;
3856 		fp->f_data = vp;
3857 		foffset = 0;
3858 		vput(tvp);
3859 		goto unionread;
3860 	}
3861 	VOP_UNLOCK(vp, 0);
3862 	foffset_unlock(fp, foffset, 0);
3863 	fdrop(fp, td);
3864 	td->td_retval[0] = uap->count - auio.uio_resid;
3865 	if (error == 0)
3866 		*ploff = loff;
3867 	return (error);
3868 }
3869 #endif /* COMPAT_43 */
3870 
3871 /*
3872  * Read a block of directory entries in a filesystem independent format.
3873  */
3874 #ifndef _SYS_SYSPROTO_H_
3875 struct getdirentries_args {
3876 	int	fd;
3877 	char	*buf;
3878 	u_int	count;
3879 	long	*basep;
3880 };
3881 #endif
3882 int
3883 sys_getdirentries(td, uap)
3884 	struct thread *td;
3885 	register struct getdirentries_args /* {
3886 		int fd;
3887 		char *buf;
3888 		u_int count;
3889 		long *basep;
3890 	} */ *uap;
3891 {
3892 	long base;
3893 	int error;
3894 
3895 	error = kern_getdirentries(td, uap->fd, uap->buf, uap->count, &base,
3896 	    NULL, UIO_USERSPACE);
3897 	if (error != 0)
3898 		return (error);
3899 	if (uap->basep != NULL)
3900 		error = copyout(&base, uap->basep, sizeof(long));
3901 	return (error);
3902 }
3903 
3904 int
3905 kern_getdirentries(struct thread *td, int fd, char *buf, u_int count,
3906     long *basep, ssize_t *residp, enum uio_seg bufseg)
3907 {
3908 	struct vnode *vp;
3909 	struct file *fp;
3910 	struct uio auio;
3911 	struct iovec aiov;
3912 	cap_rights_t rights;
3913 	long loff;
3914 	int error, eofflag;
3915 	off_t foffset;
3916 
3917 	AUDIT_ARG_FD(fd);
3918 	if (count > IOSIZE_MAX)
3919 		return (EINVAL);
3920 	auio.uio_resid = count;
3921 	error = getvnode(td->td_proc->p_fd, fd,
3922 	    cap_rights_init(&rights, CAP_READ), &fp);
3923 	if (error != 0)
3924 		return (error);
3925 	if ((fp->f_flag & FREAD) == 0) {
3926 		fdrop(fp, td);
3927 		return (EBADF);
3928 	}
3929 	vp = fp->f_vnode;
3930 	foffset = foffset_lock(fp, 0);
3931 unionread:
3932 	if (vp->v_type != VDIR) {
3933 		error = EINVAL;
3934 		goto fail;
3935 	}
3936 	aiov.iov_base = buf;
3937 	aiov.iov_len = count;
3938 	auio.uio_iov = &aiov;
3939 	auio.uio_iovcnt = 1;
3940 	auio.uio_rw = UIO_READ;
3941 	auio.uio_segflg = bufseg;
3942 	auio.uio_td = td;
3943 	vn_lock(vp, LK_SHARED | LK_RETRY);
3944 	AUDIT_ARG_VNODE1(vp);
3945 	loff = auio.uio_offset = foffset;
3946 #ifdef MAC
3947 	error = mac_vnode_check_readdir(td->td_ucred, vp);
3948 	if (error == 0)
3949 #endif
3950 		error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, NULL,
3951 		    NULL);
3952 	foffset = auio.uio_offset;
3953 	if (error != 0) {
3954 		VOP_UNLOCK(vp, 0);
3955 		goto fail;
3956 	}
3957 	if (count == auio.uio_resid &&
3958 	    (vp->v_vflag & VV_ROOT) &&
3959 	    (vp->v_mount->mnt_flag & MNT_UNION)) {
3960 		struct vnode *tvp = vp;
3961 
3962 		vp = vp->v_mount->mnt_vnodecovered;
3963 		VREF(vp);
3964 		fp->f_vnode = vp;
3965 		fp->f_data = vp;
3966 		foffset = 0;
3967 		vput(tvp);
3968 		goto unionread;
3969 	}
3970 	VOP_UNLOCK(vp, 0);
3971 	*basep = loff;
3972 	if (residp != NULL)
3973 		*residp = auio.uio_resid;
3974 	td->td_retval[0] = count - auio.uio_resid;
3975 fail:
3976 	foffset_unlock(fp, foffset, 0);
3977 	fdrop(fp, td);
3978 	return (error);
3979 }
3980 
3981 #ifndef _SYS_SYSPROTO_H_
3982 struct getdents_args {
3983 	int fd;
3984 	char *buf;
3985 	size_t count;
3986 };
3987 #endif
3988 int
3989 sys_getdents(td, uap)
3990 	struct thread *td;
3991 	register struct getdents_args /* {
3992 		int fd;
3993 		char *buf;
3994 		u_int count;
3995 	} */ *uap;
3996 {
3997 	struct getdirentries_args ap;
3998 
3999 	ap.fd = uap->fd;
4000 	ap.buf = uap->buf;
4001 	ap.count = uap->count;
4002 	ap.basep = NULL;
4003 	return (sys_getdirentries(td, &ap));
4004 }
4005 
4006 /*
4007  * Set the mode mask for creation of filesystem nodes.
4008  */
4009 #ifndef _SYS_SYSPROTO_H_
4010 struct umask_args {
4011 	int	newmask;
4012 };
4013 #endif
4014 int
4015 sys_umask(td, uap)
4016 	struct thread *td;
4017 	struct umask_args /* {
4018 		int newmask;
4019 	} */ *uap;
4020 {
4021 	register struct filedesc *fdp;
4022 
4023 	FILEDESC_XLOCK(td->td_proc->p_fd);
4024 	fdp = td->td_proc->p_fd;
4025 	td->td_retval[0] = fdp->fd_cmask;
4026 	fdp->fd_cmask = uap->newmask & ALLPERMS;
4027 	FILEDESC_XUNLOCK(td->td_proc->p_fd);
4028 	return (0);
4029 }
4030 
4031 /*
4032  * Void all references to file by ripping underlying filesystem away from
4033  * vnode.
4034  */
4035 #ifndef _SYS_SYSPROTO_H_
4036 struct revoke_args {
4037 	char	*path;
4038 };
4039 #endif
4040 int
4041 sys_revoke(td, uap)
4042 	struct thread *td;
4043 	register struct revoke_args /* {
4044 		char *path;
4045 	} */ *uap;
4046 {
4047 	struct vnode *vp;
4048 	struct vattr vattr;
4049 	struct nameidata nd;
4050 	int error;
4051 
4052 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
4053 	    uap->path, td);
4054 	if ((error = namei(&nd)) != 0)
4055 		return (error);
4056 	vp = nd.ni_vp;
4057 	NDFREE(&nd, NDF_ONLY_PNBUF);
4058 	if (vp->v_type != VCHR || vp->v_rdev == NULL) {
4059 		error = EINVAL;
4060 		goto out;
4061 	}
4062 #ifdef MAC
4063 	error = mac_vnode_check_revoke(td->td_ucred, vp);
4064 	if (error != 0)
4065 		goto out;
4066 #endif
4067 	error = VOP_GETATTR(vp, &vattr, td->td_ucred);
4068 	if (error != 0)
4069 		goto out;
4070 	if (td->td_ucred->cr_uid != vattr.va_uid) {
4071 		error = priv_check(td, PRIV_VFS_ADMIN);
4072 		if (error != 0)
4073 			goto out;
4074 	}
4075 	if (vcount(vp) > 1)
4076 		VOP_REVOKE(vp, REVOKEALL);
4077 out:
4078 	vput(vp);
4079 	return (error);
4080 }
4081 
4082 /*
4083  * Convert a user file descriptor to a kernel file entry and check that, if it
4084  * is a capability, the correct rights are present. A reference on the file
4085  * entry is held upon returning.
4086  */
4087 int
4088 getvnode(struct filedesc *fdp, int fd, cap_rights_t *rightsp, struct file **fpp)
4089 {
4090 	struct file *fp;
4091 	int error;
4092 
4093 	error = fget_unlocked(fdp, fd, rightsp, 0, &fp, NULL);
4094 	if (error != 0)
4095 		return (error);
4096 
4097 	/*
4098 	 * The file could be not of the vnode type, or it may be not
4099 	 * yet fully initialized, in which case the f_vnode pointer
4100 	 * may be set, but f_ops is still badfileops.  E.g.,
4101 	 * devfs_open() transiently create such situation to
4102 	 * facilitate csw d_fdopen().
4103 	 *
4104 	 * Dupfdopen() handling in kern_openat() installs the
4105 	 * half-baked file into the process descriptor table, allowing
4106 	 * other thread to dereference it. Guard against the race by
4107 	 * checking f_ops.
4108 	 */
4109 	if (fp->f_vnode == NULL || fp->f_ops == &badfileops) {
4110 		fdrop(fp, curthread);
4111 		return (EINVAL);
4112 	}
4113 	*fpp = fp;
4114 	return (0);
4115 }
4116 
4117 
4118 /*
4119  * Get an (NFS) file handle.
4120  */
4121 #ifndef _SYS_SYSPROTO_H_
4122 struct lgetfh_args {
4123 	char	*fname;
4124 	fhandle_t *fhp;
4125 };
4126 #endif
4127 int
4128 sys_lgetfh(td, uap)
4129 	struct thread *td;
4130 	register struct lgetfh_args *uap;
4131 {
4132 	struct nameidata nd;
4133 	fhandle_t fh;
4134 	register struct vnode *vp;
4135 	int error;
4136 
4137 	error = priv_check(td, PRIV_VFS_GETFH);
4138 	if (error != 0)
4139 		return (error);
4140 	NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
4141 	    uap->fname, td);
4142 	error = namei(&nd);
4143 	if (error != 0)
4144 		return (error);
4145 	NDFREE(&nd, NDF_ONLY_PNBUF);
4146 	vp = nd.ni_vp;
4147 	bzero(&fh, sizeof(fh));
4148 	fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
4149 	error = VOP_VPTOFH(vp, &fh.fh_fid);
4150 	vput(vp);
4151 	if (error == 0)
4152 		error = copyout(&fh, uap->fhp, sizeof (fh));
4153 	return (error);
4154 }
4155 
4156 #ifndef _SYS_SYSPROTO_H_
4157 struct getfh_args {
4158 	char	*fname;
4159 	fhandle_t *fhp;
4160 };
4161 #endif
4162 int
4163 sys_getfh(td, uap)
4164 	struct thread *td;
4165 	register struct getfh_args *uap;
4166 {
4167 	struct nameidata nd;
4168 	fhandle_t fh;
4169 	register struct vnode *vp;
4170 	int error;
4171 
4172 	error = priv_check(td, PRIV_VFS_GETFH);
4173 	if (error != 0)
4174 		return (error);
4175 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
4176 	    uap->fname, td);
4177 	error = namei(&nd);
4178 	if (error != 0)
4179 		return (error);
4180 	NDFREE(&nd, NDF_ONLY_PNBUF);
4181 	vp = nd.ni_vp;
4182 	bzero(&fh, sizeof(fh));
4183 	fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
4184 	error = VOP_VPTOFH(vp, &fh.fh_fid);
4185 	vput(vp);
4186 	if (error == 0)
4187 		error = copyout(&fh, uap->fhp, sizeof (fh));
4188 	return (error);
4189 }
4190 
4191 /*
4192  * syscall for the rpc.lockd to use to translate a NFS file handle into an
4193  * open descriptor.
4194  *
4195  * warning: do not remove the priv_check() call or this becomes one giant
4196  * security hole.
4197  */
4198 #ifndef _SYS_SYSPROTO_H_
4199 struct fhopen_args {
4200 	const struct fhandle *u_fhp;
4201 	int flags;
4202 };
4203 #endif
4204 int
4205 sys_fhopen(td, uap)
4206 	struct thread *td;
4207 	struct fhopen_args /* {
4208 		const struct fhandle *u_fhp;
4209 		int flags;
4210 	} */ *uap;
4211 {
4212 	struct mount *mp;
4213 	struct vnode *vp;
4214 	struct fhandle fhp;
4215 	struct file *fp;
4216 	int fmode, error;
4217 	int indx;
4218 
4219 	error = priv_check(td, PRIV_VFS_FHOPEN);
4220 	if (error != 0)
4221 		return (error);
4222 	indx = -1;
4223 	fmode = FFLAGS(uap->flags);
4224 	/* why not allow a non-read/write open for our lockd? */
4225 	if (((fmode & (FREAD | FWRITE)) == 0) || (fmode & O_CREAT))
4226 		return (EINVAL);
4227 	error = copyin(uap->u_fhp, &fhp, sizeof(fhp));
4228 	if (error != 0)
4229 		return(error);
4230 	/* find the mount point */
4231 	mp = vfs_busyfs(&fhp.fh_fsid);
4232 	if (mp == NULL)
4233 		return (ESTALE);
4234 	/* now give me my vnode, it gets returned to me locked */
4235 	error = VFS_FHTOVP(mp, &fhp.fh_fid, LK_EXCLUSIVE, &vp);
4236 	vfs_unbusy(mp);
4237 	if (error != 0)
4238 		return (error);
4239 
4240 	error = falloc_noinstall(td, &fp);
4241 	if (error != 0) {
4242 		vput(vp);
4243 		return (error);
4244 	}
4245 	/*
4246 	 * An extra reference on `fp' has been held for us by
4247 	 * falloc_noinstall().
4248 	 */
4249 
4250 #ifdef INVARIANTS
4251 	td->td_dupfd = -1;
4252 #endif
4253 	error = vn_open_vnode(vp, fmode, td->td_ucred, td, fp);
4254 	if (error != 0) {
4255 		KASSERT(fp->f_ops == &badfileops,
4256 		    ("VOP_OPEN in fhopen() set f_ops"));
4257 		KASSERT(td->td_dupfd < 0,
4258 		    ("fhopen() encountered fdopen()"));
4259 
4260 		vput(vp);
4261 		goto bad;
4262 	}
4263 #ifdef INVARIANTS
4264 	td->td_dupfd = 0;
4265 #endif
4266 	fp->f_vnode = vp;
4267 	fp->f_seqcount = 1;
4268 	finit(fp, (fmode & FMASK) | (fp->f_flag & FHASLOCK), DTYPE_VNODE, vp,
4269 	    &vnops);
4270 	VOP_UNLOCK(vp, 0);
4271 	if ((fmode & O_TRUNC) != 0) {
4272 		error = fo_truncate(fp, 0, td->td_ucred, td);
4273 		if (error != 0)
4274 			goto bad;
4275 	}
4276 
4277 	error = finstall(td, fp, &indx, fmode, NULL);
4278 bad:
4279 	fdrop(fp, td);
4280 	td->td_retval[0] = indx;
4281 	return (error);
4282 }
4283 
4284 /*
4285  * Stat an (NFS) file handle.
4286  */
4287 #ifndef _SYS_SYSPROTO_H_
4288 struct fhstat_args {
4289 	struct fhandle *u_fhp;
4290 	struct stat *sb;
4291 };
4292 #endif
4293 int
4294 sys_fhstat(td, uap)
4295 	struct thread *td;
4296 	register struct fhstat_args /* {
4297 		struct fhandle *u_fhp;
4298 		struct stat *sb;
4299 	} */ *uap;
4300 {
4301 	struct stat sb;
4302 	struct fhandle fh;
4303 	int error;
4304 
4305 	error = copyin(uap->u_fhp, &fh, sizeof(fh));
4306 	if (error != 0)
4307 		return (error);
4308 	error = kern_fhstat(td, fh, &sb);
4309 	if (error == 0)
4310 		error = copyout(&sb, uap->sb, sizeof(sb));
4311 	return (error);
4312 }
4313 
4314 int
4315 kern_fhstat(struct thread *td, struct fhandle fh, struct stat *sb)
4316 {
4317 	struct mount *mp;
4318 	struct vnode *vp;
4319 	int error;
4320 
4321 	error = priv_check(td, PRIV_VFS_FHSTAT);
4322 	if (error != 0)
4323 		return (error);
4324 	if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4325 		return (ESTALE);
4326 	error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4327 	vfs_unbusy(mp);
4328 	if (error != 0)
4329 		return (error);
4330 	error = vn_stat(vp, sb, td->td_ucred, NOCRED, td);
4331 	vput(vp);
4332 	return (error);
4333 }
4334 
4335 /*
4336  * Implement fstatfs() for (NFS) file handles.
4337  */
4338 #ifndef _SYS_SYSPROTO_H_
4339 struct fhstatfs_args {
4340 	struct fhandle *u_fhp;
4341 	struct statfs *buf;
4342 };
4343 #endif
4344 int
4345 sys_fhstatfs(td, uap)
4346 	struct thread *td;
4347 	struct fhstatfs_args /* {
4348 		struct fhandle *u_fhp;
4349 		struct statfs *buf;
4350 	} */ *uap;
4351 {
4352 	struct statfs sf;
4353 	fhandle_t fh;
4354 	int error;
4355 
4356 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
4357 	if (error != 0)
4358 		return (error);
4359 	error = kern_fhstatfs(td, fh, &sf);
4360 	if (error != 0)
4361 		return (error);
4362 	return (copyout(&sf, uap->buf, sizeof(sf)));
4363 }
4364 
4365 int
4366 kern_fhstatfs(struct thread *td, fhandle_t fh, struct statfs *buf)
4367 {
4368 	struct statfs *sp;
4369 	struct mount *mp;
4370 	struct vnode *vp;
4371 	int error;
4372 
4373 	error = priv_check(td, PRIV_VFS_FHSTATFS);
4374 	if (error != 0)
4375 		return (error);
4376 	if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4377 		return (ESTALE);
4378 	error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4379 	if (error != 0) {
4380 		vfs_unbusy(mp);
4381 		return (error);
4382 	}
4383 	vput(vp);
4384 	error = prison_canseemount(td->td_ucred, mp);
4385 	if (error != 0)
4386 		goto out;
4387 #ifdef MAC
4388 	error = mac_mount_check_stat(td->td_ucred, mp);
4389 	if (error != 0)
4390 		goto out;
4391 #endif
4392 	/*
4393 	 * Set these in case the underlying filesystem fails to do so.
4394 	 */
4395 	sp = &mp->mnt_stat;
4396 	sp->f_version = STATFS_VERSION;
4397 	sp->f_namemax = NAME_MAX;
4398 	sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
4399 	error = VFS_STATFS(mp, sp);
4400 	if (error == 0)
4401 		*buf = *sp;
4402 out:
4403 	vfs_unbusy(mp);
4404 	return (error);
4405 }
4406 
4407 int
4408 kern_posix_fallocate(struct thread *td, int fd, off_t offset, off_t len)
4409 {
4410 	struct file *fp;
4411 	struct mount *mp;
4412 	struct vnode *vp;
4413 	cap_rights_t rights;
4414 	off_t olen, ooffset;
4415 	int error;
4416 
4417 	if (offset < 0 || len <= 0)
4418 		return (EINVAL);
4419 	/* Check for wrap. */
4420 	if (offset > OFF_MAX - len)
4421 		return (EFBIG);
4422 	error = fget(td, fd, cap_rights_init(&rights, CAP_WRITE), &fp);
4423 	if (error != 0)
4424 		return (error);
4425 	if ((fp->f_ops->fo_flags & DFLAG_SEEKABLE) == 0) {
4426 		error = ESPIPE;
4427 		goto out;
4428 	}
4429 	if ((fp->f_flag & FWRITE) == 0) {
4430 		error = EBADF;
4431 		goto out;
4432 	}
4433 	if (fp->f_type != DTYPE_VNODE) {
4434 		error = ENODEV;
4435 		goto out;
4436 	}
4437 	vp = fp->f_vnode;
4438 	if (vp->v_type != VREG) {
4439 		error = ENODEV;
4440 		goto out;
4441 	}
4442 
4443 	/* Allocating blocks may take a long time, so iterate. */
4444 	for (;;) {
4445 		olen = len;
4446 		ooffset = offset;
4447 
4448 		bwillwrite();
4449 		mp = NULL;
4450 		error = vn_start_write(vp, &mp, V_WAIT | PCATCH);
4451 		if (error != 0)
4452 			break;
4453 		error = vn_lock(vp, LK_EXCLUSIVE);
4454 		if (error != 0) {
4455 			vn_finished_write(mp);
4456 			break;
4457 		}
4458 #ifdef MAC
4459 		error = mac_vnode_check_write(td->td_ucred, fp->f_cred, vp);
4460 		if (error == 0)
4461 #endif
4462 			error = VOP_ALLOCATE(vp, &offset, &len);
4463 		VOP_UNLOCK(vp, 0);
4464 		vn_finished_write(mp);
4465 
4466 		if (olen + ooffset != offset + len) {
4467 			panic("offset + len changed from %jx/%jx to %jx/%jx",
4468 			    ooffset, olen, offset, len);
4469 		}
4470 		if (error != 0 || len == 0)
4471 			break;
4472 		KASSERT(olen > len, ("Iteration did not make progress?"));
4473 		maybe_yield();
4474 	}
4475  out:
4476 	fdrop(fp, td);
4477 	return (error);
4478 }
4479 
4480 int
4481 sys_posix_fallocate(struct thread *td, struct posix_fallocate_args *uap)
4482 {
4483 
4484 	td->td_retval[0] = kern_posix_fallocate(td, uap->fd, uap->offset,
4485 	    uap->len);
4486 	return (0);
4487 }
4488 
4489 /*
4490  * Unlike madvise(2), we do not make a best effort to remember every
4491  * possible caching hint.  Instead, we remember the last setting with
4492  * the exception that we will allow POSIX_FADV_NORMAL to adjust the
4493  * region of any current setting.
4494  */
4495 int
4496 kern_posix_fadvise(struct thread *td, int fd, off_t offset, off_t len,
4497     int advice)
4498 {
4499 	struct fadvise_info *fa, *new;
4500 	struct file *fp;
4501 	struct vnode *vp;
4502 	cap_rights_t rights;
4503 	off_t end;
4504 	int error;
4505 
4506 	if (offset < 0 || len < 0 || offset > OFF_MAX - len)
4507 		return (EINVAL);
4508 	switch (advice) {
4509 	case POSIX_FADV_SEQUENTIAL:
4510 	case POSIX_FADV_RANDOM:
4511 	case POSIX_FADV_NOREUSE:
4512 		new = malloc(sizeof(*fa), M_FADVISE, M_WAITOK);
4513 		break;
4514 	case POSIX_FADV_NORMAL:
4515 	case POSIX_FADV_WILLNEED:
4516 	case POSIX_FADV_DONTNEED:
4517 		new = NULL;
4518 		break;
4519 	default:
4520 		return (EINVAL);
4521 	}
4522 	/* XXX: CAP_POSIX_FADVISE? */
4523 	error = fget(td, fd, cap_rights_init(&rights), &fp);
4524 	if (error != 0)
4525 		goto out;
4526 	if ((fp->f_ops->fo_flags & DFLAG_SEEKABLE) == 0) {
4527 		error = ESPIPE;
4528 		goto out;
4529 	}
4530 	if (fp->f_type != DTYPE_VNODE) {
4531 		error = ENODEV;
4532 		goto out;
4533 	}
4534 	vp = fp->f_vnode;
4535 	if (vp->v_type != VREG) {
4536 		error = ENODEV;
4537 		goto out;
4538 	}
4539 	if (len == 0)
4540 		end = OFF_MAX;
4541 	else
4542 		end = offset + len - 1;
4543 	switch (advice) {
4544 	case POSIX_FADV_SEQUENTIAL:
4545 	case POSIX_FADV_RANDOM:
4546 	case POSIX_FADV_NOREUSE:
4547 		/*
4548 		 * Try to merge any existing non-standard region with
4549 		 * this new region if possible, otherwise create a new
4550 		 * non-standard region for this request.
4551 		 */
4552 		mtx_pool_lock(mtxpool_sleep, fp);
4553 		fa = fp->f_advice;
4554 		if (fa != NULL && fa->fa_advice == advice &&
4555 		    ((fa->fa_start <= end && fa->fa_end >= offset) ||
4556 		    (end != OFF_MAX && fa->fa_start == end + 1) ||
4557 		    (fa->fa_end != OFF_MAX && fa->fa_end + 1 == offset))) {
4558 			if (offset < fa->fa_start)
4559 				fa->fa_start = offset;
4560 			if (end > fa->fa_end)
4561 				fa->fa_end = end;
4562 		} else {
4563 			new->fa_advice = advice;
4564 			new->fa_start = offset;
4565 			new->fa_end = end;
4566 			new->fa_prevstart = 0;
4567 			new->fa_prevend = 0;
4568 			fp->f_advice = new;
4569 			new = fa;
4570 		}
4571 		mtx_pool_unlock(mtxpool_sleep, fp);
4572 		break;
4573 	case POSIX_FADV_NORMAL:
4574 		/*
4575 		 * If a the "normal" region overlaps with an existing
4576 		 * non-standard region, trim or remove the
4577 		 * non-standard region.
4578 		 */
4579 		mtx_pool_lock(mtxpool_sleep, fp);
4580 		fa = fp->f_advice;
4581 		if (fa != NULL) {
4582 			if (offset <= fa->fa_start && end >= fa->fa_end) {
4583 				new = fa;
4584 				fp->f_advice = NULL;
4585 			} else if (offset <= fa->fa_start &&
4586 			    end >= fa->fa_start)
4587 				fa->fa_start = end + 1;
4588 			else if (offset <= fa->fa_end && end >= fa->fa_end)
4589 				fa->fa_end = offset - 1;
4590 			else if (offset >= fa->fa_start && end <= fa->fa_end) {
4591 				/*
4592 				 * If the "normal" region is a middle
4593 				 * portion of the existing
4594 				 * non-standard region, just remove
4595 				 * the whole thing rather than picking
4596 				 * one side or the other to
4597 				 * preserve.
4598 				 */
4599 				new = fa;
4600 				fp->f_advice = NULL;
4601 			}
4602 		}
4603 		mtx_pool_unlock(mtxpool_sleep, fp);
4604 		break;
4605 	case POSIX_FADV_WILLNEED:
4606 	case POSIX_FADV_DONTNEED:
4607 		error = VOP_ADVISE(vp, offset, end, advice);
4608 		break;
4609 	}
4610 out:
4611 	if (fp != NULL)
4612 		fdrop(fp, td);
4613 	free(new, M_FADVISE);
4614 	return (error);
4615 }
4616 
4617 int
4618 sys_posix_fadvise(struct thread *td, struct posix_fadvise_args *uap)
4619 {
4620 
4621 	td->td_retval[0] = kern_posix_fadvise(td, uap->fd, uap->offset,
4622 	    uap->len, uap->advice);
4623 	return (0);
4624 }
4625