xref: /freebsd/sys/kern/vfs_syscalls.c (revision 08c4a937a6685f05667996228898521fc453f8f3)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (c) 1989, 1993
5  *	The Regents of the University of California.  All rights reserved.
6  * (c) UNIX System Laboratories, Inc.
7  * All or some portions of this file are derived from material licensed
8  * to the University of California by American Telephone and Telegraph
9  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
10  * the permission of UNIX System Laboratories, Inc.
11  *
12  * Redistribution and use in source and binary forms, with or without
13  * modification, are permitted provided that the following conditions
14  * are met:
15  * 1. Redistributions of source code must retain the above copyright
16  *    notice, this list of conditions and the following disclaimer.
17  * 2. Redistributions in binary form must reproduce the above copyright
18  *    notice, this list of conditions and the following disclaimer in the
19  *    documentation and/or other materials provided with the distribution.
20  * 3. Neither the name of the University nor the names of its contributors
21  *    may be used to endorse or promote products derived from this software
22  *    without specific prior written permission.
23  *
24  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
25  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
28  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
29  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
30  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
31  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
32  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
33  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
34  * SUCH DAMAGE.
35  *
36  *	@(#)vfs_syscalls.c	8.13 (Berkeley) 4/15/94
37  */
38 
39 #include <sys/cdefs.h>
40 __FBSDID("$FreeBSD$");
41 
42 #include "opt_capsicum.h"
43 #include "opt_ktrace.h"
44 
45 #include <sys/param.h>
46 #include <sys/systm.h>
47 #include <sys/bio.h>
48 #include <sys/buf.h>
49 #include <sys/capsicum.h>
50 #include <sys/disk.h>
51 #include <sys/sysent.h>
52 #include <sys/malloc.h>
53 #include <sys/mount.h>
54 #include <sys/mutex.h>
55 #include <sys/sysproto.h>
56 #include <sys/namei.h>
57 #include <sys/filedesc.h>
58 #include <sys/kernel.h>
59 #include <sys/fcntl.h>
60 #include <sys/file.h>
61 #include <sys/filio.h>
62 #include <sys/limits.h>
63 #include <sys/linker.h>
64 #include <sys/rwlock.h>
65 #include <sys/sdt.h>
66 #include <sys/stat.h>
67 #include <sys/sx.h>
68 #include <sys/unistd.h>
69 #include <sys/vnode.h>
70 #include <sys/priv.h>
71 #include <sys/proc.h>
72 #include <sys/dirent.h>
73 #include <sys/jail.h>
74 #include <sys/syscallsubr.h>
75 #include <sys/sysctl.h>
76 #ifdef KTRACE
77 #include <sys/ktrace.h>
78 #endif
79 
80 #include <machine/stdarg.h>
81 
82 #include <security/audit/audit.h>
83 #include <security/mac/mac_framework.h>
84 
85 #include <vm/vm.h>
86 #include <vm/vm_object.h>
87 #include <vm/vm_page.h>
88 #include <vm/uma.h>
89 
90 #include <ufs/ufs/quota.h>
91 
92 MALLOC_DEFINE(M_FADVISE, "fadvise", "posix_fadvise(2) information");
93 
94 SDT_PROVIDER_DEFINE(vfs);
95 SDT_PROBE_DEFINE2(vfs, , stat, mode, "char *", "int");
96 SDT_PROBE_DEFINE2(vfs, , stat, reg, "char *", "int");
97 
98 static int kern_chflagsat(struct thread *td, int fd, const char *path,
99     enum uio_seg pathseg, u_long flags, int atflag);
100 static int setfflags(struct thread *td, struct vnode *, u_long);
101 static int getutimes(const struct timeval *, enum uio_seg, struct timespec *);
102 static int getutimens(const struct timespec *, enum uio_seg,
103     struct timespec *, int *);
104 static int setutimes(struct thread *td, struct vnode *,
105     const struct timespec *, int, int);
106 static int vn_access(struct vnode *vp, int user_flags, struct ucred *cred,
107     struct thread *td);
108 static int kern_fhlinkat(struct thread *td, int fd, const char *path,
109     enum uio_seg pathseg, fhandle_t *fhp);
110 static int kern_getfhat(struct thread *td, int flags, int fd,
111     const char *path, enum uio_seg pathseg, fhandle_t *fhp);
112 static int kern_readlink_vp(struct vnode *vp, char *buf, enum uio_seg bufseg,
113     size_t count, struct thread *td);
114 static int kern_linkat_vp(struct thread *td, struct vnode *vp, int fd,
115     const char *path, enum uio_seg segflag);
116 
117 /*
118  * Sync each mounted filesystem.
119  */
120 #ifndef _SYS_SYSPROTO_H_
121 struct sync_args {
122 	int     dummy;
123 };
124 #endif
125 /* ARGSUSED */
126 int
127 sys_sync(struct thread *td, 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(struct thread *td, struct quotactl_args *uap)
167 {
168 	struct mount *mp;
169 	struct nameidata nd;
170 	int error;
171 
172 	AUDIT_ARG_CMD(uap->cmd);
173 	AUDIT_ARG_UID(uap->uid);
174 	if (!prison_allow(td->td_ucred, PR_ALLOW_QUOTAS))
175 		return (EPERM);
176 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
177 	    uap->path, td);
178 	if ((error = namei(&nd)) != 0)
179 		return (error);
180 	NDFREE(&nd, NDF_ONLY_PNBUF);
181 	mp = nd.ni_vp->v_mount;
182 	vfs_ref(mp);
183 	vput(nd.ni_vp);
184 	error = vfs_busy(mp, 0);
185 	vfs_rel(mp);
186 	if (error != 0)
187 		return (error);
188 	error = VFS_QUOTACTL(mp, uap->cmd, uap->uid, uap->arg);
189 
190 	/*
191 	 * Since quota on operation typically needs to open quota
192 	 * file, the Q_QUOTAON handler needs to unbusy the mount point
193 	 * before calling into namei.  Otherwise, unmount might be
194 	 * started between two vfs_busy() invocations (first is our,
195 	 * second is from mount point cross-walk code in lookup()),
196 	 * causing deadlock.
197 	 *
198 	 * Require that Q_QUOTAON handles the vfs_busy() reference on
199 	 * its own, always returning with ubusied mount point.
200 	 */
201 	if ((uap->cmd >> SUBCMDSHIFT) != Q_QUOTAON &&
202 	    (uap->cmd >> SUBCMDSHIFT) != Q_QUOTAOFF)
203 		vfs_unbusy(mp);
204 	return (error);
205 }
206 
207 /*
208  * Used by statfs conversion routines to scale the block size up if
209  * necessary so that all of the block counts are <= 'max_size'.  Note
210  * that 'max_size' should be a bitmask, i.e. 2^n - 1 for some non-zero
211  * value of 'n'.
212  */
213 void
214 statfs_scale_blocks(struct statfs *sf, long max_size)
215 {
216 	uint64_t count;
217 	int shift;
218 
219 	KASSERT(powerof2(max_size + 1), ("%s: invalid max_size", __func__));
220 
221 	/*
222 	 * Attempt to scale the block counts to give a more accurate
223 	 * overview to userland of the ratio of free space to used
224 	 * space.  To do this, find the largest block count and compute
225 	 * a divisor that lets it fit into a signed integer <= max_size.
226 	 */
227 	if (sf->f_bavail < 0)
228 		count = -sf->f_bavail;
229 	else
230 		count = sf->f_bavail;
231 	count = MAX(sf->f_blocks, MAX(sf->f_bfree, count));
232 	if (count <= max_size)
233 		return;
234 
235 	count >>= flsl(max_size);
236 	shift = 0;
237 	while (count > 0) {
238 		shift++;
239 		count >>=1;
240 	}
241 
242 	sf->f_bsize <<= shift;
243 	sf->f_blocks >>= shift;
244 	sf->f_bfree >>= shift;
245 	sf->f_bavail >>= shift;
246 }
247 
248 static int
249 kern_do_statfs(struct thread *td, struct mount *mp, struct statfs *buf)
250 {
251 	struct statfs *sp;
252 	int error;
253 
254 	if (mp == NULL)
255 		return (EBADF);
256 	error = vfs_busy(mp, 0);
257 	vfs_rel(mp);
258 	if (error != 0)
259 		return (error);
260 #ifdef MAC
261 	error = mac_mount_check_stat(td->td_ucred, mp);
262 	if (error != 0)
263 		goto out;
264 #endif
265 	/*
266 	 * Set these in case the underlying filesystem fails to do so.
267 	 */
268 	sp = &mp->mnt_stat;
269 	sp->f_version = STATFS_VERSION;
270 	sp->f_namemax = NAME_MAX;
271 	sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
272 	error = VFS_STATFS(mp, sp);
273 	if (error != 0)
274 		goto out;
275 	*buf = *sp;
276 	if (priv_check(td, PRIV_VFS_GENERATION)) {
277 		buf->f_fsid.val[0] = buf->f_fsid.val[1] = 0;
278 		prison_enforce_statfs(td->td_ucred, mp, buf);
279 	}
280 out:
281 	vfs_unbusy(mp);
282 	return (error);
283 }
284 
285 /*
286  * Get filesystem statistics.
287  */
288 #ifndef _SYS_SYSPROTO_H_
289 struct statfs_args {
290 	char *path;
291 	struct statfs *buf;
292 };
293 #endif
294 int
295 sys_statfs(struct thread *td, struct statfs_args *uap)
296 {
297 	struct statfs *sfp;
298 	int error;
299 
300 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
301 	error = kern_statfs(td, uap->path, UIO_USERSPACE, sfp);
302 	if (error == 0)
303 		error = copyout(sfp, uap->buf, sizeof(struct statfs));
304 	free(sfp, M_STATFS);
305 	return (error);
306 }
307 
308 int
309 kern_statfs(struct thread *td, const char *path, enum uio_seg pathseg,
310     struct statfs *buf)
311 {
312 	struct mount *mp;
313 	struct nameidata nd;
314 	int error;
315 
316 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
317 	    pathseg, path, td);
318 	error = namei(&nd);
319 	if (error != 0)
320 		return (error);
321 	mp = nd.ni_vp->v_mount;
322 	vfs_ref(mp);
323 	NDFREE(&nd, NDF_ONLY_PNBUF);
324 	vput(nd.ni_vp);
325 	return (kern_do_statfs(td, mp, buf));
326 }
327 
328 /*
329  * Get filesystem statistics.
330  */
331 #ifndef _SYS_SYSPROTO_H_
332 struct fstatfs_args {
333 	int fd;
334 	struct statfs *buf;
335 };
336 #endif
337 int
338 sys_fstatfs(struct thread *td, struct fstatfs_args *uap)
339 {
340 	struct statfs *sfp;
341 	int error;
342 
343 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
344 	error = kern_fstatfs(td, uap->fd, sfp);
345 	if (error == 0)
346 		error = copyout(sfp, uap->buf, sizeof(struct statfs));
347 	free(sfp, M_STATFS);
348 	return (error);
349 }
350 
351 int
352 kern_fstatfs(struct thread *td, int fd, struct statfs *buf)
353 {
354 	struct file *fp;
355 	struct mount *mp;
356 	struct vnode *vp;
357 	int error;
358 
359 	AUDIT_ARG_FD(fd);
360 	error = getvnode(td, fd, &cap_fstatfs_rights, &fp);
361 	if (error != 0)
362 		return (error);
363 	vp = fp->f_vnode;
364 	vn_lock(vp, LK_SHARED | LK_RETRY);
365 #ifdef AUDIT
366 	AUDIT_ARG_VNODE1(vp);
367 #endif
368 	mp = vp->v_mount;
369 	if (mp != NULL)
370 		vfs_ref(mp);
371 	VOP_UNLOCK(vp, 0);
372 	fdrop(fp, td);
373 	return (kern_do_statfs(td, mp, buf));
374 }
375 
376 /*
377  * Get statistics on all filesystems.
378  */
379 #ifndef _SYS_SYSPROTO_H_
380 struct getfsstat_args {
381 	struct statfs *buf;
382 	long bufsize;
383 	int mode;
384 };
385 #endif
386 int
387 sys_getfsstat(struct thread *td, struct getfsstat_args *uap)
388 {
389 	size_t count;
390 	int error;
391 
392 	if (uap->bufsize < 0 || uap->bufsize > SIZE_MAX)
393 		return (EINVAL);
394 	error = kern_getfsstat(td, &uap->buf, uap->bufsize, &count,
395 	    UIO_USERSPACE, uap->mode);
396 	if (error == 0)
397 		td->td_retval[0] = count;
398 	return (error);
399 }
400 
401 /*
402  * If (bufsize > 0 && bufseg == UIO_SYSSPACE)
403  *	The caller is responsible for freeing memory which will be allocated
404  *	in '*buf'.
405  */
406 int
407 kern_getfsstat(struct thread *td, struct statfs **buf, size_t bufsize,
408     size_t *countp, enum uio_seg bufseg, int mode)
409 {
410 	struct mount *mp, *nmp;
411 	struct statfs *sfsp, *sp, *sptmp, *tofree;
412 	size_t count, maxcount;
413 	int error;
414 
415 	switch (mode) {
416 	case MNT_WAIT:
417 	case MNT_NOWAIT:
418 		break;
419 	default:
420 		if (bufseg == UIO_SYSSPACE)
421 			*buf = NULL;
422 		return (EINVAL);
423 	}
424 restart:
425 	maxcount = bufsize / sizeof(struct statfs);
426 	if (bufsize == 0) {
427 		sfsp = NULL;
428 		tofree = NULL;
429 	} else if (bufseg == UIO_USERSPACE) {
430 		sfsp = *buf;
431 		tofree = NULL;
432 	} else /* if (bufseg == UIO_SYSSPACE) */ {
433 		count = 0;
434 		mtx_lock(&mountlist_mtx);
435 		TAILQ_FOREACH(mp, &mountlist, mnt_list) {
436 			count++;
437 		}
438 		mtx_unlock(&mountlist_mtx);
439 		if (maxcount > count)
440 			maxcount = count;
441 		tofree = sfsp = *buf = malloc(maxcount * sizeof(struct statfs),
442 		    M_STATFS, M_WAITOK);
443 	}
444 	count = 0;
445 	mtx_lock(&mountlist_mtx);
446 	for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
447 		if (prison_canseemount(td->td_ucred, mp) != 0) {
448 			nmp = TAILQ_NEXT(mp, mnt_list);
449 			continue;
450 		}
451 #ifdef MAC
452 		if (mac_mount_check_stat(td->td_ucred, mp) != 0) {
453 			nmp = TAILQ_NEXT(mp, mnt_list);
454 			continue;
455 		}
456 #endif
457 		if (mode == MNT_WAIT) {
458 			if (vfs_busy(mp, MBF_MNTLSTLOCK) != 0) {
459 				/*
460 				 * If vfs_busy() failed, and MBF_NOWAIT
461 				 * wasn't passed, then the mp is gone.
462 				 * Furthermore, because of MBF_MNTLSTLOCK,
463 				 * the mountlist_mtx was dropped.  We have
464 				 * no other choice than to start over.
465 				 */
466 				mtx_unlock(&mountlist_mtx);
467 				free(tofree, M_STATFS);
468 				goto restart;
469 			}
470 		} else {
471 			if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK) != 0) {
472 				nmp = TAILQ_NEXT(mp, mnt_list);
473 				continue;
474 			}
475 		}
476 		if (sfsp != NULL && count < maxcount) {
477 			sp = &mp->mnt_stat;
478 			/*
479 			 * Set these in case the underlying filesystem
480 			 * fails to do so.
481 			 */
482 			sp->f_version = STATFS_VERSION;
483 			sp->f_namemax = NAME_MAX;
484 			sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
485 			/*
486 			 * If MNT_NOWAIT is specified, do not refresh
487 			 * the fsstat cache.
488 			 */
489 			if (mode != MNT_NOWAIT) {
490 				error = VFS_STATFS(mp, sp);
491 				if (error != 0) {
492 					mtx_lock(&mountlist_mtx);
493 					nmp = TAILQ_NEXT(mp, mnt_list);
494 					vfs_unbusy(mp);
495 					continue;
496 				}
497 			}
498 			if (priv_check(td, PRIV_VFS_GENERATION)) {
499 				sptmp = malloc(sizeof(struct statfs), M_STATFS,
500 				    M_WAITOK);
501 				*sptmp = *sp;
502 				sptmp->f_fsid.val[0] = sptmp->f_fsid.val[1] = 0;
503 				prison_enforce_statfs(td->td_ucred, mp, sptmp);
504 				sp = sptmp;
505 			} else
506 				sptmp = NULL;
507 			if (bufseg == UIO_SYSSPACE) {
508 				bcopy(sp, sfsp, sizeof(*sp));
509 				free(sptmp, M_STATFS);
510 			} else /* if (bufseg == UIO_USERSPACE) */ {
511 				error = copyout(sp, sfsp, sizeof(*sp));
512 				free(sptmp, M_STATFS);
513 				if (error != 0) {
514 					vfs_unbusy(mp);
515 					return (error);
516 				}
517 			}
518 			sfsp++;
519 		}
520 		count++;
521 		mtx_lock(&mountlist_mtx);
522 		nmp = TAILQ_NEXT(mp, mnt_list);
523 		vfs_unbusy(mp);
524 	}
525 	mtx_unlock(&mountlist_mtx);
526 	if (sfsp != NULL && count > maxcount)
527 		*countp = maxcount;
528 	else
529 		*countp = count;
530 	return (0);
531 }
532 
533 #ifdef COMPAT_FREEBSD4
534 /*
535  * Get old format filesystem statistics.
536  */
537 static void freebsd4_cvtstatfs(struct statfs *, struct ostatfs *);
538 
539 #ifndef _SYS_SYSPROTO_H_
540 struct freebsd4_statfs_args {
541 	char *path;
542 	struct ostatfs *buf;
543 };
544 #endif
545 int
546 freebsd4_statfs(struct thread *td, struct freebsd4_statfs_args *uap)
547 {
548 	struct ostatfs osb;
549 	struct statfs *sfp;
550 	int error;
551 
552 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
553 	error = kern_statfs(td, uap->path, UIO_USERSPACE, sfp);
554 	if (error == 0) {
555 		freebsd4_cvtstatfs(sfp, &osb);
556 		error = copyout(&osb, uap->buf, sizeof(osb));
557 	}
558 	free(sfp, M_STATFS);
559 	return (error);
560 }
561 
562 /*
563  * Get filesystem statistics.
564  */
565 #ifndef _SYS_SYSPROTO_H_
566 struct freebsd4_fstatfs_args {
567 	int fd;
568 	struct ostatfs *buf;
569 };
570 #endif
571 int
572 freebsd4_fstatfs(struct thread *td, struct freebsd4_fstatfs_args *uap)
573 {
574 	struct ostatfs osb;
575 	struct statfs *sfp;
576 	int error;
577 
578 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
579 	error = kern_fstatfs(td, uap->fd, sfp);
580 	if (error == 0) {
581 		freebsd4_cvtstatfs(sfp, &osb);
582 		error = copyout(&osb, uap->buf, sizeof(osb));
583 	}
584 	free(sfp, M_STATFS);
585 	return (error);
586 }
587 
588 /*
589  * Get statistics on all filesystems.
590  */
591 #ifndef _SYS_SYSPROTO_H_
592 struct freebsd4_getfsstat_args {
593 	struct ostatfs *buf;
594 	long bufsize;
595 	int mode;
596 };
597 #endif
598 int
599 freebsd4_getfsstat(struct thread *td, struct freebsd4_getfsstat_args *uap)
600 {
601 	struct statfs *buf, *sp;
602 	struct ostatfs osb;
603 	size_t count, size;
604 	int error;
605 
606 	if (uap->bufsize < 0)
607 		return (EINVAL);
608 	count = uap->bufsize / sizeof(struct ostatfs);
609 	if (count > SIZE_MAX / sizeof(struct statfs))
610 		return (EINVAL);
611 	size = count * sizeof(struct statfs);
612 	error = kern_getfsstat(td, &buf, size, &count, UIO_SYSSPACE,
613 	    uap->mode);
614 	if (error == 0)
615 		td->td_retval[0] = count;
616 	if (size != 0) {
617 		sp = buf;
618 		while (count != 0 && error == 0) {
619 			freebsd4_cvtstatfs(sp, &osb);
620 			error = copyout(&osb, uap->buf, sizeof(osb));
621 			sp++;
622 			uap->buf++;
623 			count--;
624 		}
625 		free(buf, M_STATFS);
626 	}
627 	return (error);
628 }
629 
630 /*
631  * Implement fstatfs() for (NFS) file handles.
632  */
633 #ifndef _SYS_SYSPROTO_H_
634 struct freebsd4_fhstatfs_args {
635 	struct fhandle *u_fhp;
636 	struct ostatfs *buf;
637 };
638 #endif
639 int
640 freebsd4_fhstatfs(struct thread *td, struct freebsd4_fhstatfs_args *uap)
641 {
642 	struct ostatfs osb;
643 	struct statfs *sfp;
644 	fhandle_t fh;
645 	int error;
646 
647 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
648 	if (error != 0)
649 		return (error);
650 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
651 	error = kern_fhstatfs(td, fh, sfp);
652 	if (error == 0) {
653 		freebsd4_cvtstatfs(sfp, &osb);
654 		error = copyout(&osb, uap->buf, sizeof(osb));
655 	}
656 	free(sfp, M_STATFS);
657 	return (error);
658 }
659 
660 /*
661  * Convert a new format statfs structure to an old format statfs structure.
662  */
663 static void
664 freebsd4_cvtstatfs(struct statfs *nsp, struct ostatfs *osp)
665 {
666 
667 	statfs_scale_blocks(nsp, LONG_MAX);
668 	bzero(osp, sizeof(*osp));
669 	osp->f_bsize = nsp->f_bsize;
670 	osp->f_iosize = MIN(nsp->f_iosize, LONG_MAX);
671 	osp->f_blocks = nsp->f_blocks;
672 	osp->f_bfree = nsp->f_bfree;
673 	osp->f_bavail = nsp->f_bavail;
674 	osp->f_files = MIN(nsp->f_files, LONG_MAX);
675 	osp->f_ffree = MIN(nsp->f_ffree, LONG_MAX);
676 	osp->f_owner = nsp->f_owner;
677 	osp->f_type = nsp->f_type;
678 	osp->f_flags = nsp->f_flags;
679 	osp->f_syncwrites = MIN(nsp->f_syncwrites, LONG_MAX);
680 	osp->f_asyncwrites = MIN(nsp->f_asyncwrites, LONG_MAX);
681 	osp->f_syncreads = MIN(nsp->f_syncreads, LONG_MAX);
682 	osp->f_asyncreads = MIN(nsp->f_asyncreads, LONG_MAX);
683 	strlcpy(osp->f_fstypename, nsp->f_fstypename,
684 	    MIN(MFSNAMELEN, OMFSNAMELEN));
685 	strlcpy(osp->f_mntonname, nsp->f_mntonname,
686 	    MIN(MNAMELEN, OMNAMELEN));
687 	strlcpy(osp->f_mntfromname, nsp->f_mntfromname,
688 	    MIN(MNAMELEN, OMNAMELEN));
689 	osp->f_fsid = nsp->f_fsid;
690 }
691 #endif /* COMPAT_FREEBSD4 */
692 
693 #if defined(COMPAT_FREEBSD11)
694 /*
695  * Get old format filesystem statistics.
696  */
697 static void freebsd11_cvtstatfs(struct statfs *, struct freebsd11_statfs *);
698 
699 int
700 freebsd11_statfs(struct thread *td, struct freebsd11_statfs_args *uap)
701 {
702 	struct freebsd11_statfs osb;
703 	struct statfs *sfp;
704 	int error;
705 
706 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
707 	error = kern_statfs(td, uap->path, UIO_USERSPACE, sfp);
708 	if (error == 0) {
709 		freebsd11_cvtstatfs(sfp, &osb);
710 		error = copyout(&osb, uap->buf, sizeof(osb));
711 	}
712 	free(sfp, M_STATFS);
713 	return (error);
714 }
715 
716 /*
717  * Get filesystem statistics.
718  */
719 int
720 freebsd11_fstatfs(struct thread *td, struct freebsd11_fstatfs_args *uap)
721 {
722 	struct freebsd11_statfs osb;
723 	struct statfs *sfp;
724 	int error;
725 
726 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
727 	error = kern_fstatfs(td, uap->fd, sfp);
728 	if (error == 0) {
729 		freebsd11_cvtstatfs(sfp, &osb);
730 		error = copyout(&osb, uap->buf, sizeof(osb));
731 	}
732 	free(sfp, M_STATFS);
733 	return (error);
734 }
735 
736 /*
737  * Get statistics on all filesystems.
738  */
739 int
740 freebsd11_getfsstat(struct thread *td, struct freebsd11_getfsstat_args *uap)
741 {
742 	struct freebsd11_statfs osb;
743 	struct statfs *buf, *sp;
744 	size_t count, size;
745 	int error;
746 
747 	count = uap->bufsize / sizeof(struct ostatfs);
748 	size = count * sizeof(struct statfs);
749 	error = kern_getfsstat(td, &buf, size, &count, UIO_SYSSPACE,
750 	    uap->mode);
751 	if (error == 0)
752 		td->td_retval[0] = count;
753 	if (size > 0) {
754 		sp = buf;
755 		while (count > 0 && error == 0) {
756 			freebsd11_cvtstatfs(sp, &osb);
757 			error = copyout(&osb, uap->buf, sizeof(osb));
758 			sp++;
759 			uap->buf++;
760 			count--;
761 		}
762 		free(buf, M_STATFS);
763 	}
764 	return (error);
765 }
766 
767 /*
768  * Implement fstatfs() for (NFS) file handles.
769  */
770 int
771 freebsd11_fhstatfs(struct thread *td, struct freebsd11_fhstatfs_args *uap)
772 {
773 	struct freebsd11_statfs osb;
774 	struct statfs *sfp;
775 	fhandle_t fh;
776 	int error;
777 
778 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
779 	if (error)
780 		return (error);
781 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
782 	error = kern_fhstatfs(td, fh, sfp);
783 	if (error == 0) {
784 		freebsd11_cvtstatfs(sfp, &osb);
785 		error = copyout(&osb, uap->buf, sizeof(osb));
786 	}
787 	free(sfp, M_STATFS);
788 	return (error);
789 }
790 
791 /*
792  * Convert a new format statfs structure to an old format statfs structure.
793  */
794 static void
795 freebsd11_cvtstatfs(struct statfs *nsp, struct freebsd11_statfs *osp)
796 {
797 
798 	bzero(osp, sizeof(*osp));
799 	osp->f_version = FREEBSD11_STATFS_VERSION;
800 	osp->f_type = nsp->f_type;
801 	osp->f_flags = nsp->f_flags;
802 	osp->f_bsize = nsp->f_bsize;
803 	osp->f_iosize = nsp->f_iosize;
804 	osp->f_blocks = nsp->f_blocks;
805 	osp->f_bfree = nsp->f_bfree;
806 	osp->f_bavail = nsp->f_bavail;
807 	osp->f_files = nsp->f_files;
808 	osp->f_ffree = nsp->f_ffree;
809 	osp->f_syncwrites = nsp->f_syncwrites;
810 	osp->f_asyncwrites = nsp->f_asyncwrites;
811 	osp->f_syncreads = nsp->f_syncreads;
812 	osp->f_asyncreads = nsp->f_asyncreads;
813 	osp->f_namemax = nsp->f_namemax;
814 	osp->f_owner = nsp->f_owner;
815 	osp->f_fsid = nsp->f_fsid;
816 	strlcpy(osp->f_fstypename, nsp->f_fstypename,
817 	    MIN(MFSNAMELEN, sizeof(osp->f_fstypename)));
818 	strlcpy(osp->f_mntonname, nsp->f_mntonname,
819 	    MIN(MNAMELEN, sizeof(osp->f_mntonname)));
820 	strlcpy(osp->f_mntfromname, nsp->f_mntfromname,
821 	    MIN(MNAMELEN, sizeof(osp->f_mntfromname)));
822 }
823 #endif /* COMPAT_FREEBSD11 */
824 
825 /*
826  * Change current working directory to a given file descriptor.
827  */
828 #ifndef _SYS_SYSPROTO_H_
829 struct fchdir_args {
830 	int	fd;
831 };
832 #endif
833 int
834 sys_fchdir(struct thread *td, struct fchdir_args *uap)
835 {
836 	struct vnode *vp, *tdp;
837 	struct mount *mp;
838 	struct file *fp;
839 	int error;
840 
841 	AUDIT_ARG_FD(uap->fd);
842 	error = getvnode(td, uap->fd, &cap_fchdir_rights,
843 	    &fp);
844 	if (error != 0)
845 		return (error);
846 	vp = fp->f_vnode;
847 	vrefact(vp);
848 	fdrop(fp, td);
849 	vn_lock(vp, LK_SHARED | LK_RETRY);
850 	AUDIT_ARG_VNODE1(vp);
851 	error = change_dir(vp, td);
852 	while (!error && (mp = vp->v_mountedhere) != NULL) {
853 		if (vfs_busy(mp, 0))
854 			continue;
855 		error = VFS_ROOT(mp, LK_SHARED, &tdp);
856 		vfs_unbusy(mp);
857 		if (error != 0)
858 			break;
859 		vput(vp);
860 		vp = tdp;
861 	}
862 	if (error != 0) {
863 		vput(vp);
864 		return (error);
865 	}
866 	VOP_UNLOCK(vp, 0);
867 	pwd_chdir(td, vp);
868 	return (0);
869 }
870 
871 /*
872  * Change current working directory (``.'').
873  */
874 #ifndef _SYS_SYSPROTO_H_
875 struct chdir_args {
876 	char	*path;
877 };
878 #endif
879 int
880 sys_chdir(struct thread *td, struct chdir_args *uap)
881 {
882 
883 	return (kern_chdir(td, uap->path, UIO_USERSPACE));
884 }
885 
886 int
887 kern_chdir(struct thread *td, const char *path, enum uio_seg pathseg)
888 {
889 	struct nameidata nd;
890 	int error;
891 
892 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
893 	    pathseg, path, td);
894 	if ((error = namei(&nd)) != 0)
895 		return (error);
896 	if ((error = change_dir(nd.ni_vp, td)) != 0) {
897 		vput(nd.ni_vp);
898 		NDFREE(&nd, NDF_ONLY_PNBUF);
899 		return (error);
900 	}
901 	VOP_UNLOCK(nd.ni_vp, 0);
902 	NDFREE(&nd, NDF_ONLY_PNBUF);
903 	pwd_chdir(td, nd.ni_vp);
904 	return (0);
905 }
906 
907 /*
908  * Change notion of root (``/'') directory.
909  */
910 #ifndef _SYS_SYSPROTO_H_
911 struct chroot_args {
912 	char	*path;
913 };
914 #endif
915 int
916 sys_chroot(struct thread *td, struct chroot_args *uap)
917 {
918 	struct nameidata nd;
919 	int error;
920 
921 	error = priv_check(td, PRIV_VFS_CHROOT);
922 	if (error != 0)
923 		return (error);
924 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
925 	    UIO_USERSPACE, uap->path, td);
926 	error = namei(&nd);
927 	if (error != 0)
928 		goto error;
929 	error = change_dir(nd.ni_vp, td);
930 	if (error != 0)
931 		goto e_vunlock;
932 #ifdef MAC
933 	error = mac_vnode_check_chroot(td->td_ucred, nd.ni_vp);
934 	if (error != 0)
935 		goto e_vunlock;
936 #endif
937 	VOP_UNLOCK(nd.ni_vp, 0);
938 	error = pwd_chroot(td, nd.ni_vp);
939 	vrele(nd.ni_vp);
940 	NDFREE(&nd, NDF_ONLY_PNBUF);
941 	return (error);
942 e_vunlock:
943 	vput(nd.ni_vp);
944 error:
945 	NDFREE(&nd, NDF_ONLY_PNBUF);
946 	return (error);
947 }
948 
949 /*
950  * Common routine for chroot and chdir.  Callers must provide a locked vnode
951  * instance.
952  */
953 int
954 change_dir(struct vnode *vp, struct thread *td)
955 {
956 #ifdef MAC
957 	int error;
958 #endif
959 
960 	ASSERT_VOP_LOCKED(vp, "change_dir(): vp not locked");
961 	if (vp->v_type != VDIR)
962 		return (ENOTDIR);
963 #ifdef MAC
964 	error = mac_vnode_check_chdir(td->td_ucred, vp);
965 	if (error != 0)
966 		return (error);
967 #endif
968 	return (VOP_ACCESS(vp, VEXEC, td->td_ucred, td));
969 }
970 
971 static __inline void
972 flags_to_rights(int flags, cap_rights_t *rightsp)
973 {
974 
975 	if (flags & O_EXEC) {
976 		cap_rights_set(rightsp, CAP_FEXECVE);
977 	} else {
978 		switch ((flags & O_ACCMODE)) {
979 		case O_RDONLY:
980 			cap_rights_set(rightsp, CAP_READ);
981 			break;
982 		case O_RDWR:
983 			cap_rights_set(rightsp, CAP_READ);
984 			/* FALLTHROUGH */
985 		case O_WRONLY:
986 			cap_rights_set(rightsp, CAP_WRITE);
987 			if (!(flags & (O_APPEND | O_TRUNC)))
988 				cap_rights_set(rightsp, CAP_SEEK);
989 			break;
990 		}
991 	}
992 
993 	if (flags & O_CREAT)
994 		cap_rights_set(rightsp, CAP_CREATE);
995 
996 	if (flags & O_TRUNC)
997 		cap_rights_set(rightsp, CAP_FTRUNCATE);
998 
999 	if (flags & (O_SYNC | O_FSYNC))
1000 		cap_rights_set(rightsp, CAP_FSYNC);
1001 
1002 	if (flags & (O_EXLOCK | O_SHLOCK))
1003 		cap_rights_set(rightsp, CAP_FLOCK);
1004 }
1005 
1006 /*
1007  * Check permissions, allocate an open file structure, and call the device
1008  * open routine if any.
1009  */
1010 #ifndef _SYS_SYSPROTO_H_
1011 struct open_args {
1012 	char	*path;
1013 	int	flags;
1014 	int	mode;
1015 };
1016 #endif
1017 int
1018 sys_open(struct thread *td, struct open_args *uap)
1019 {
1020 
1021 	return (kern_openat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1022 	    uap->flags, uap->mode));
1023 }
1024 
1025 #ifndef _SYS_SYSPROTO_H_
1026 struct openat_args {
1027 	int	fd;
1028 	char	*path;
1029 	int	flag;
1030 	int	mode;
1031 };
1032 #endif
1033 int
1034 sys_openat(struct thread *td, struct openat_args *uap)
1035 {
1036 
1037 	AUDIT_ARG_FD(uap->fd);
1038 	return (kern_openat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
1039 	    uap->mode));
1040 }
1041 
1042 int
1043 kern_openat(struct thread *td, int fd, const char *path, enum uio_seg pathseg,
1044     int flags, int mode)
1045 {
1046 	struct proc *p = td->td_proc;
1047 	struct filedesc *fdp = p->p_fd;
1048 	struct file *fp;
1049 	struct vnode *vp;
1050 	struct nameidata nd;
1051 	cap_rights_t rights;
1052 	int cmode, error, indx;
1053 
1054 	indx = -1;
1055 
1056 	AUDIT_ARG_FFLAGS(flags);
1057 	AUDIT_ARG_MODE(mode);
1058 	cap_rights_init(&rights, CAP_LOOKUP);
1059 	flags_to_rights(flags, &rights);
1060 	/*
1061 	 * Only one of the O_EXEC, O_RDONLY, O_WRONLY and O_RDWR flags
1062 	 * may be specified.
1063 	 */
1064 	if (flags & O_EXEC) {
1065 		if (flags & O_ACCMODE)
1066 			return (EINVAL);
1067 	} else if ((flags & O_ACCMODE) == O_ACCMODE) {
1068 		return (EINVAL);
1069 	} else {
1070 		flags = FFLAGS(flags);
1071 	}
1072 
1073 	/*
1074 	 * Allocate a file structure. The descriptor to reference it
1075 	 * is allocated and set by finstall() below.
1076 	 */
1077 	error = falloc_noinstall(td, &fp);
1078 	if (error != 0)
1079 		return (error);
1080 	/*
1081 	 * An extra reference on `fp' has been held for us by
1082 	 * falloc_noinstall().
1083 	 */
1084 	/* Set the flags early so the finit in devfs can pick them up. */
1085 	fp->f_flag = flags & FMASK;
1086 	cmode = ((mode & ~fdp->fd_cmask) & ALLPERMS) & ~S_ISTXT;
1087 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, fd,
1088 	    &rights, td);
1089 	td->td_dupfd = -1;		/* XXX check for fdopen */
1090 	error = vn_open(&nd, &flags, cmode, fp);
1091 	if (error != 0) {
1092 		/*
1093 		 * If the vn_open replaced the method vector, something
1094 		 * wonderous happened deep below and we just pass it up
1095 		 * pretending we know what we do.
1096 		 */
1097 		if (error == ENXIO && fp->f_ops != &badfileops)
1098 			goto success;
1099 
1100 		/*
1101 		 * Handle special fdopen() case. bleh.
1102 		 *
1103 		 * Don't do this for relative (capability) lookups; we don't
1104 		 * understand exactly what would happen, and we don't think
1105 		 * that it ever should.
1106 		 */
1107 		if ((nd.ni_lcf & NI_LCF_STRICTRELATIVE) == 0 &&
1108 		    (error == ENODEV || error == ENXIO) &&
1109 		    td->td_dupfd >= 0) {
1110 			error = dupfdopen(td, fdp, td->td_dupfd, flags, error,
1111 			    &indx);
1112 			if (error == 0)
1113 				goto success;
1114 		}
1115 
1116 		goto bad;
1117 	}
1118 	td->td_dupfd = 0;
1119 	NDFREE(&nd, NDF_ONLY_PNBUF);
1120 	vp = nd.ni_vp;
1121 
1122 	/*
1123 	 * Store the vnode, for any f_type. Typically, the vnode use
1124 	 * count is decremented by direct call to vn_closefile() for
1125 	 * files that switched type in the cdevsw fdopen() method.
1126 	 */
1127 	fp->f_vnode = vp;
1128 	/*
1129 	 * If the file wasn't claimed by devfs bind it to the normal
1130 	 * vnode operations here.
1131 	 */
1132 	if (fp->f_ops == &badfileops) {
1133 		KASSERT(vp->v_type != VFIFO, ("Unexpected fifo."));
1134 		fp->f_seqcount = 1;
1135 		finit(fp, (flags & FMASK) | (fp->f_flag & FHASLOCK),
1136 		    DTYPE_VNODE, vp, &vnops);
1137 	}
1138 
1139 	VOP_UNLOCK(vp, 0);
1140 	if (flags & O_TRUNC) {
1141 		error = fo_truncate(fp, 0, td->td_ucred, td);
1142 		if (error != 0)
1143 			goto bad;
1144 	}
1145 success:
1146 	/*
1147 	 * If we haven't already installed the FD (for dupfdopen), do so now.
1148 	 */
1149 	if (indx == -1) {
1150 		struct filecaps *fcaps;
1151 
1152 #ifdef CAPABILITIES
1153 		if ((nd.ni_lcf & NI_LCF_STRICTRELATIVE) != 0)
1154 			fcaps = &nd.ni_filecaps;
1155 		else
1156 #endif
1157 			fcaps = NULL;
1158 		error = finstall(td, fp, &indx, flags, fcaps);
1159 		/* On success finstall() consumes fcaps. */
1160 		if (error != 0) {
1161 			filecaps_free(&nd.ni_filecaps);
1162 			goto bad;
1163 		}
1164 	} else {
1165 		filecaps_free(&nd.ni_filecaps);
1166 	}
1167 
1168 	/*
1169 	 * Release our private reference, leaving the one associated with
1170 	 * the descriptor table intact.
1171 	 */
1172 	fdrop(fp, td);
1173 	td->td_retval[0] = indx;
1174 	return (0);
1175 bad:
1176 	KASSERT(indx == -1, ("indx=%d, should be -1", indx));
1177 	fdrop(fp, td);
1178 	return (error);
1179 }
1180 
1181 #ifdef COMPAT_43
1182 /*
1183  * Create a file.
1184  */
1185 #ifndef _SYS_SYSPROTO_H_
1186 struct ocreat_args {
1187 	char	*path;
1188 	int	mode;
1189 };
1190 #endif
1191 int
1192 ocreat(struct thread *td, struct ocreat_args *uap)
1193 {
1194 
1195 	return (kern_openat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1196 	    O_WRONLY | O_CREAT | O_TRUNC, uap->mode));
1197 }
1198 #endif /* COMPAT_43 */
1199 
1200 /*
1201  * Create a special file.
1202  */
1203 #ifndef _SYS_SYSPROTO_H_
1204 struct mknodat_args {
1205 	int	fd;
1206 	char	*path;
1207 	mode_t	mode;
1208 	dev_t	dev;
1209 };
1210 #endif
1211 int
1212 sys_mknodat(struct thread *td, struct mknodat_args *uap)
1213 {
1214 
1215 	return (kern_mknodat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode,
1216 	    uap->dev));
1217 }
1218 
1219 #if defined(COMPAT_FREEBSD11)
1220 int
1221 freebsd11_mknod(struct thread *td,
1222     struct freebsd11_mknod_args *uap)
1223 {
1224 
1225 	return (kern_mknodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1226 	    uap->mode, uap->dev));
1227 }
1228 
1229 int
1230 freebsd11_mknodat(struct thread *td,
1231     struct freebsd11_mknodat_args *uap)
1232 {
1233 
1234 	return (kern_mknodat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode,
1235 	    uap->dev));
1236 }
1237 #endif /* COMPAT_FREEBSD11 */
1238 
1239 int
1240 kern_mknodat(struct thread *td, int fd, const char *path, enum uio_seg pathseg,
1241     int mode, dev_t dev)
1242 {
1243 	struct vnode *vp;
1244 	struct mount *mp;
1245 	struct vattr vattr;
1246 	struct nameidata nd;
1247 	int error, whiteout = 0;
1248 
1249 	AUDIT_ARG_MODE(mode);
1250 	AUDIT_ARG_DEV(dev);
1251 	switch (mode & S_IFMT) {
1252 	case S_IFCHR:
1253 	case S_IFBLK:
1254 		error = priv_check(td, PRIV_VFS_MKNOD_DEV);
1255 		if (error == 0 && dev == VNOVAL)
1256 			error = EINVAL;
1257 		break;
1258 	case S_IFWHT:
1259 		error = priv_check(td, PRIV_VFS_MKNOD_WHT);
1260 		break;
1261 	case S_IFIFO:
1262 		if (dev == 0)
1263 			return (kern_mkfifoat(td, fd, path, pathseg, mode));
1264 		/* FALLTHROUGH */
1265 	default:
1266 		error = EINVAL;
1267 		break;
1268 	}
1269 	if (error != 0)
1270 		return (error);
1271 restart:
1272 	bwillwrite();
1273 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
1274 	    NOCACHE, pathseg, path, fd, &cap_mknodat_rights,
1275 	    td);
1276 	if ((error = namei(&nd)) != 0)
1277 		return (error);
1278 	vp = nd.ni_vp;
1279 	if (vp != NULL) {
1280 		NDFREE(&nd, NDF_ONLY_PNBUF);
1281 		if (vp == nd.ni_dvp)
1282 			vrele(nd.ni_dvp);
1283 		else
1284 			vput(nd.ni_dvp);
1285 		vrele(vp);
1286 		return (EEXIST);
1287 	} else {
1288 		VATTR_NULL(&vattr);
1289 		vattr.va_mode = (mode & ALLPERMS) &
1290 		    ~td->td_proc->p_fd->fd_cmask;
1291 		vattr.va_rdev = dev;
1292 		whiteout = 0;
1293 
1294 		switch (mode & S_IFMT) {
1295 		case S_IFCHR:
1296 			vattr.va_type = VCHR;
1297 			break;
1298 		case S_IFBLK:
1299 			vattr.va_type = VBLK;
1300 			break;
1301 		case S_IFWHT:
1302 			whiteout = 1;
1303 			break;
1304 		default:
1305 			panic("kern_mknod: invalid mode");
1306 		}
1307 	}
1308 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1309 		NDFREE(&nd, NDF_ONLY_PNBUF);
1310 		vput(nd.ni_dvp);
1311 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1312 			return (error);
1313 		goto restart;
1314 	}
1315 #ifdef MAC
1316 	if (error == 0 && !whiteout)
1317 		error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp,
1318 		    &nd.ni_cnd, &vattr);
1319 #endif
1320 	if (error == 0) {
1321 		if (whiteout)
1322 			error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE);
1323 		else {
1324 			error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp,
1325 						&nd.ni_cnd, &vattr);
1326 			if (error == 0)
1327 				vput(nd.ni_vp);
1328 		}
1329 	}
1330 	NDFREE(&nd, NDF_ONLY_PNBUF);
1331 	vput(nd.ni_dvp);
1332 	vn_finished_write(mp);
1333 	return (error);
1334 }
1335 
1336 /*
1337  * Create a named pipe.
1338  */
1339 #ifndef _SYS_SYSPROTO_H_
1340 struct mkfifo_args {
1341 	char	*path;
1342 	int	mode;
1343 };
1344 #endif
1345 int
1346 sys_mkfifo(struct thread *td, struct mkfifo_args *uap)
1347 {
1348 
1349 	return (kern_mkfifoat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1350 	    uap->mode));
1351 }
1352 
1353 #ifndef _SYS_SYSPROTO_H_
1354 struct mkfifoat_args {
1355 	int	fd;
1356 	char	*path;
1357 	mode_t	mode;
1358 };
1359 #endif
1360 int
1361 sys_mkfifoat(struct thread *td, struct mkfifoat_args *uap)
1362 {
1363 
1364 	return (kern_mkfifoat(td, uap->fd, uap->path, UIO_USERSPACE,
1365 	    uap->mode));
1366 }
1367 
1368 int
1369 kern_mkfifoat(struct thread *td, int fd, const char *path,
1370     enum uio_seg pathseg, int mode)
1371 {
1372 	struct mount *mp;
1373 	struct vattr vattr;
1374 	struct nameidata nd;
1375 	int error;
1376 
1377 	AUDIT_ARG_MODE(mode);
1378 restart:
1379 	bwillwrite();
1380 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
1381 	    NOCACHE, pathseg, path, fd, &cap_mkfifoat_rights,
1382 	    td);
1383 	if ((error = namei(&nd)) != 0)
1384 		return (error);
1385 	if (nd.ni_vp != NULL) {
1386 		NDFREE(&nd, NDF_ONLY_PNBUF);
1387 		if (nd.ni_vp == nd.ni_dvp)
1388 			vrele(nd.ni_dvp);
1389 		else
1390 			vput(nd.ni_dvp);
1391 		vrele(nd.ni_vp);
1392 		return (EEXIST);
1393 	}
1394 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1395 		NDFREE(&nd, NDF_ONLY_PNBUF);
1396 		vput(nd.ni_dvp);
1397 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1398 			return (error);
1399 		goto restart;
1400 	}
1401 	VATTR_NULL(&vattr);
1402 	vattr.va_type = VFIFO;
1403 	vattr.va_mode = (mode & ALLPERMS) & ~td->td_proc->p_fd->fd_cmask;
1404 #ifdef MAC
1405 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1406 	    &vattr);
1407 	if (error != 0)
1408 		goto out;
1409 #endif
1410 	error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
1411 	if (error == 0)
1412 		vput(nd.ni_vp);
1413 #ifdef MAC
1414 out:
1415 #endif
1416 	vput(nd.ni_dvp);
1417 	vn_finished_write(mp);
1418 	NDFREE(&nd, NDF_ONLY_PNBUF);
1419 	return (error);
1420 }
1421 
1422 /*
1423  * Make a hard file link.
1424  */
1425 #ifndef _SYS_SYSPROTO_H_
1426 struct link_args {
1427 	char	*path;
1428 	char	*link;
1429 };
1430 #endif
1431 int
1432 sys_link(struct thread *td, struct link_args *uap)
1433 {
1434 
1435 	return (kern_linkat(td, AT_FDCWD, AT_FDCWD, uap->path, uap->link,
1436 	    UIO_USERSPACE, FOLLOW));
1437 }
1438 
1439 #ifndef _SYS_SYSPROTO_H_
1440 struct linkat_args {
1441 	int	fd1;
1442 	char	*path1;
1443 	int	fd2;
1444 	char	*path2;
1445 	int	flag;
1446 };
1447 #endif
1448 int
1449 sys_linkat(struct thread *td, struct linkat_args *uap)
1450 {
1451 	int flag;
1452 
1453 	flag = uap->flag;
1454 	if ((flag & ~(AT_SYMLINK_FOLLOW | AT_BENEATH)) != 0)
1455 		return (EINVAL);
1456 
1457 	return (kern_linkat(td, uap->fd1, uap->fd2, uap->path1, uap->path2,
1458 	    UIO_USERSPACE, ((flag & AT_SYMLINK_FOLLOW) != 0 ? FOLLOW :
1459 	    NOFOLLOW) | ((flag & AT_BENEATH) != 0 ? BENEATH : 0)));
1460 }
1461 
1462 int hardlink_check_uid = 0;
1463 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_uid, CTLFLAG_RW,
1464     &hardlink_check_uid, 0,
1465     "Unprivileged processes cannot create hard links to files owned by other "
1466     "users");
1467 static int hardlink_check_gid = 0;
1468 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_gid, CTLFLAG_RW,
1469     &hardlink_check_gid, 0,
1470     "Unprivileged processes cannot create hard links to files owned by other "
1471     "groups");
1472 
1473 static int
1474 can_hardlink(struct vnode *vp, struct ucred *cred)
1475 {
1476 	struct vattr va;
1477 	int error;
1478 
1479 	if (!hardlink_check_uid && !hardlink_check_gid)
1480 		return (0);
1481 
1482 	error = VOP_GETATTR(vp, &va, cred);
1483 	if (error != 0)
1484 		return (error);
1485 
1486 	if (hardlink_check_uid && cred->cr_uid != va.va_uid) {
1487 		error = priv_check_cred(cred, PRIV_VFS_LINK, 0);
1488 		if (error != 0)
1489 			return (error);
1490 	}
1491 
1492 	if (hardlink_check_gid && !groupmember(va.va_gid, cred)) {
1493 		error = priv_check_cred(cred, PRIV_VFS_LINK, 0);
1494 		if (error != 0)
1495 			return (error);
1496 	}
1497 
1498 	return (0);
1499 }
1500 
1501 int
1502 kern_linkat(struct thread *td, int fd1, int fd2, const char *path1,
1503     const char *path2, enum uio_seg segflag, int follow)
1504 {
1505 	struct vnode *vp;
1506 	struct nameidata nd;
1507 	int error;
1508 
1509 	do {
1510 		bwillwrite();
1511 		NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, segflag, path1, fd1,
1512 		    &cap_linkat_source_rights, td);
1513 		if ((error = namei(&nd)) != 0)
1514 			return (error);
1515 		NDFREE(&nd, NDF_ONLY_PNBUF);
1516 		vp = nd.ni_vp;
1517 	} while ((error = kern_linkat_vp(td, vp, fd2, path2, segflag) == EAGAIN));
1518 	return (error);
1519 }
1520 
1521 static int
1522 kern_linkat_vp(struct thread *td, struct vnode *vp, int fd, const char *path,
1523     enum uio_seg segflag)
1524 {
1525 	struct nameidata nd;
1526 	struct mount *mp;
1527 	int error;
1528 
1529 	if (vp->v_type == VDIR) {
1530 		vrele(vp);
1531 		return (EPERM);		/* POSIX */
1532 	}
1533 	NDINIT_ATRIGHTS(&nd, CREATE,
1534 	    LOCKPARENT | SAVENAME | AUDITVNODE2 | NOCACHE, segflag, path, fd,
1535 	    &cap_linkat_target_rights, td);
1536 	if ((error = namei(&nd)) == 0) {
1537 		if (nd.ni_vp != NULL) {
1538 			NDFREE(&nd, NDF_ONLY_PNBUF);
1539 			if (nd.ni_dvp == nd.ni_vp)
1540 				vrele(nd.ni_dvp);
1541 			else
1542 				vput(nd.ni_dvp);
1543 			vrele(nd.ni_vp);
1544 			vrele(vp);
1545 			return (EEXIST);
1546 		} else if (nd.ni_dvp->v_mount != vp->v_mount) {
1547 			/*
1548 			 * Cross-device link.  No need to recheck
1549 			 * vp->v_type, since it cannot change, except
1550 			 * to VBAD.
1551 			 */
1552 			NDFREE(&nd, NDF_ONLY_PNBUF);
1553 			vput(nd.ni_dvp);
1554 			vrele(vp);
1555 			return (EXDEV);
1556 		} else if ((error = vn_lock(vp, LK_EXCLUSIVE)) == 0) {
1557 			error = can_hardlink(vp, td->td_ucred);
1558 #ifdef MAC
1559 			if (error == 0)
1560 				error = mac_vnode_check_link(td->td_ucred,
1561 				    nd.ni_dvp, vp, &nd.ni_cnd);
1562 #endif
1563 			if (error != 0) {
1564 				vput(vp);
1565 				vput(nd.ni_dvp);
1566 				NDFREE(&nd, NDF_ONLY_PNBUF);
1567 				return (error);
1568 			}
1569 			error = vn_start_write(vp, &mp, V_NOWAIT);
1570 			if (error != 0) {
1571 				vput(vp);
1572 				vput(nd.ni_dvp);
1573 				NDFREE(&nd, NDF_ONLY_PNBUF);
1574 				error = vn_start_write(NULL, &mp,
1575 				    V_XSLEEP | PCATCH);
1576 				if (error != 0)
1577 					return (error);
1578 				return (EAGAIN);
1579 			}
1580 			error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
1581 			VOP_UNLOCK(vp, 0);
1582 			vput(nd.ni_dvp);
1583 			vn_finished_write(mp);
1584 			NDFREE(&nd, NDF_ONLY_PNBUF);
1585 		} else {
1586 			vput(nd.ni_dvp);
1587 			NDFREE(&nd, NDF_ONLY_PNBUF);
1588 			vrele(vp);
1589 			return (EAGAIN);
1590 		}
1591 	}
1592 	vrele(vp);
1593 	return (error);
1594 }
1595 
1596 /*
1597  * Make a symbolic link.
1598  */
1599 #ifndef _SYS_SYSPROTO_H_
1600 struct symlink_args {
1601 	char	*path;
1602 	char	*link;
1603 };
1604 #endif
1605 int
1606 sys_symlink(struct thread *td, struct symlink_args *uap)
1607 {
1608 
1609 	return (kern_symlinkat(td, uap->path, AT_FDCWD, uap->link,
1610 	    UIO_USERSPACE));
1611 }
1612 
1613 #ifndef _SYS_SYSPROTO_H_
1614 struct symlinkat_args {
1615 	char	*path;
1616 	int	fd;
1617 	char	*path2;
1618 };
1619 #endif
1620 int
1621 sys_symlinkat(struct thread *td, struct symlinkat_args *uap)
1622 {
1623 
1624 	return (kern_symlinkat(td, uap->path1, uap->fd, uap->path2,
1625 	    UIO_USERSPACE));
1626 }
1627 
1628 int
1629 kern_symlinkat(struct thread *td, const char *path1, int fd, const char *path2,
1630     enum uio_seg segflg)
1631 {
1632 	struct mount *mp;
1633 	struct vattr vattr;
1634 	const char *syspath;
1635 	char *tmppath;
1636 	struct nameidata nd;
1637 	int error;
1638 
1639 	if (segflg == UIO_SYSSPACE) {
1640 		syspath = path1;
1641 	} else {
1642 		tmppath = uma_zalloc(namei_zone, M_WAITOK);
1643 		if ((error = copyinstr(path1, tmppath, MAXPATHLEN, NULL)) != 0)
1644 			goto out;
1645 		syspath = tmppath;
1646 	}
1647 	AUDIT_ARG_TEXT(syspath);
1648 restart:
1649 	bwillwrite();
1650 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
1651 	    NOCACHE, segflg, path2, fd, &cap_symlinkat_rights,
1652 	    td);
1653 	if ((error = namei(&nd)) != 0)
1654 		goto out;
1655 	if (nd.ni_vp) {
1656 		NDFREE(&nd, NDF_ONLY_PNBUF);
1657 		if (nd.ni_vp == nd.ni_dvp)
1658 			vrele(nd.ni_dvp);
1659 		else
1660 			vput(nd.ni_dvp);
1661 		vrele(nd.ni_vp);
1662 		error = EEXIST;
1663 		goto out;
1664 	}
1665 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1666 		NDFREE(&nd, NDF_ONLY_PNBUF);
1667 		vput(nd.ni_dvp);
1668 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1669 			goto out;
1670 		goto restart;
1671 	}
1672 	VATTR_NULL(&vattr);
1673 	vattr.va_mode = ACCESSPERMS &~ td->td_proc->p_fd->fd_cmask;
1674 #ifdef MAC
1675 	vattr.va_type = VLNK;
1676 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1677 	    &vattr);
1678 	if (error != 0)
1679 		goto out2;
1680 #endif
1681 	error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, syspath);
1682 	if (error == 0)
1683 		vput(nd.ni_vp);
1684 #ifdef MAC
1685 out2:
1686 #endif
1687 	NDFREE(&nd, NDF_ONLY_PNBUF);
1688 	vput(nd.ni_dvp);
1689 	vn_finished_write(mp);
1690 out:
1691 	if (segflg != UIO_SYSSPACE)
1692 		uma_zfree(namei_zone, tmppath);
1693 	return (error);
1694 }
1695 
1696 /*
1697  * Delete a whiteout from the filesystem.
1698  */
1699 #ifndef _SYS_SYSPROTO_H_
1700 struct undelete_args {
1701 	char *path;
1702 };
1703 #endif
1704 int
1705 sys_undelete(struct thread *td, struct undelete_args *uap)
1706 {
1707 	struct mount *mp;
1708 	struct nameidata nd;
1709 	int error;
1710 
1711 restart:
1712 	bwillwrite();
1713 	NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | AUDITVNODE1,
1714 	    UIO_USERSPACE, uap->path, td);
1715 	error = namei(&nd);
1716 	if (error != 0)
1717 		return (error);
1718 
1719 	if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) {
1720 		NDFREE(&nd, NDF_ONLY_PNBUF);
1721 		if (nd.ni_vp == nd.ni_dvp)
1722 			vrele(nd.ni_dvp);
1723 		else
1724 			vput(nd.ni_dvp);
1725 		if (nd.ni_vp)
1726 			vrele(nd.ni_vp);
1727 		return (EEXIST);
1728 	}
1729 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1730 		NDFREE(&nd, NDF_ONLY_PNBUF);
1731 		vput(nd.ni_dvp);
1732 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1733 			return (error);
1734 		goto restart;
1735 	}
1736 	error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE);
1737 	NDFREE(&nd, NDF_ONLY_PNBUF);
1738 	vput(nd.ni_dvp);
1739 	vn_finished_write(mp);
1740 	return (error);
1741 }
1742 
1743 /*
1744  * Delete a name from the filesystem.
1745  */
1746 #ifndef _SYS_SYSPROTO_H_
1747 struct unlink_args {
1748 	char	*path;
1749 };
1750 #endif
1751 int
1752 sys_unlink(struct thread *td, struct unlink_args *uap)
1753 {
1754 
1755 	return (kern_unlinkat(td, AT_FDCWD, uap->path, UIO_USERSPACE, 0, 0));
1756 }
1757 
1758 #ifndef _SYS_SYSPROTO_H_
1759 struct unlinkat_args {
1760 	int	fd;
1761 	char	*path;
1762 	int	flag;
1763 };
1764 #endif
1765 int
1766 sys_unlinkat(struct thread *td, struct unlinkat_args *uap)
1767 {
1768 	int fd, flag;
1769 	const char *path;
1770 
1771 	flag = uap->flag;
1772 	fd = uap->fd;
1773 	path = uap->path;
1774 
1775 	if ((flag & ~(AT_REMOVEDIR | AT_BENEATH)) != 0)
1776 		return (EINVAL);
1777 
1778 	if ((uap->flag & AT_REMOVEDIR) != 0)
1779 		return (kern_rmdirat(td, fd, path, UIO_USERSPACE, flag));
1780 	else
1781 		return (kern_unlinkat(td, fd, path, UIO_USERSPACE, flag, 0));
1782 }
1783 
1784 int
1785 kern_unlinkat(struct thread *td, int fd, const char *path,
1786     enum uio_seg pathseg, int flag, ino_t oldinum)
1787 {
1788 	struct mount *mp;
1789 	struct vnode *vp;
1790 	struct nameidata nd;
1791 	struct stat sb;
1792 	int error;
1793 
1794 restart:
1795 	bwillwrite();
1796 	NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1 |
1797 	    ((flag & AT_BENEATH) != 0 ? BENEATH : 0),
1798 	    pathseg, path, fd, &cap_unlinkat_rights, td);
1799 	if ((error = namei(&nd)) != 0)
1800 		return (error == EINVAL ? EPERM : error);
1801 	vp = nd.ni_vp;
1802 	if (vp->v_type == VDIR && oldinum == 0) {
1803 		error = EPERM;		/* POSIX */
1804 	} else if (oldinum != 0 &&
1805 		  ((error = vn_stat(vp, &sb, td->td_ucred, NOCRED, td)) == 0) &&
1806 		  sb.st_ino != oldinum) {
1807 			error = EIDRM;	/* Identifier removed */
1808 	} else {
1809 		/*
1810 		 * The root of a mounted filesystem cannot be deleted.
1811 		 *
1812 		 * XXX: can this only be a VDIR case?
1813 		 */
1814 		if (vp->v_vflag & VV_ROOT)
1815 			error = EBUSY;
1816 	}
1817 	if (error == 0) {
1818 		if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1819 			NDFREE(&nd, NDF_ONLY_PNBUF);
1820 			vput(nd.ni_dvp);
1821 			if (vp == nd.ni_dvp)
1822 				vrele(vp);
1823 			else
1824 				vput(vp);
1825 			if ((error = vn_start_write(NULL, &mp,
1826 			    V_XSLEEP | PCATCH)) != 0)
1827 				return (error);
1828 			goto restart;
1829 		}
1830 #ifdef MAC
1831 		error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
1832 		    &nd.ni_cnd);
1833 		if (error != 0)
1834 			goto out;
1835 #endif
1836 		vfs_notify_upper(vp, VFS_NOTIFY_UPPER_UNLINK);
1837 		error = VOP_REMOVE(nd.ni_dvp, vp, &nd.ni_cnd);
1838 #ifdef MAC
1839 out:
1840 #endif
1841 		vn_finished_write(mp);
1842 	}
1843 	NDFREE(&nd, NDF_ONLY_PNBUF);
1844 	vput(nd.ni_dvp);
1845 	if (vp == nd.ni_dvp)
1846 		vrele(vp);
1847 	else
1848 		vput(vp);
1849 	return (error);
1850 }
1851 
1852 /*
1853  * Reposition read/write file offset.
1854  */
1855 #ifndef _SYS_SYSPROTO_H_
1856 struct lseek_args {
1857 	int	fd;
1858 	int	pad;
1859 	off_t	offset;
1860 	int	whence;
1861 };
1862 #endif
1863 int
1864 sys_lseek(struct thread *td, struct lseek_args *uap)
1865 {
1866 
1867 	return (kern_lseek(td, uap->fd, uap->offset, uap->whence));
1868 }
1869 
1870 int
1871 kern_lseek(struct thread *td, int fd, off_t offset, int whence)
1872 {
1873 	struct file *fp;
1874 	int error;
1875 
1876 	AUDIT_ARG_FD(fd);
1877 	error = fget(td, fd, &cap_seek_rights, &fp);
1878 	if (error != 0)
1879 		return (error);
1880 	error = (fp->f_ops->fo_flags & DFLAG_SEEKABLE) != 0 ?
1881 	    fo_seek(fp, offset, whence, td) : ESPIPE;
1882 	fdrop(fp, td);
1883 	return (error);
1884 }
1885 
1886 #if defined(COMPAT_43)
1887 /*
1888  * Reposition read/write file offset.
1889  */
1890 #ifndef _SYS_SYSPROTO_H_
1891 struct olseek_args {
1892 	int	fd;
1893 	long	offset;
1894 	int	whence;
1895 };
1896 #endif
1897 int
1898 olseek(struct thread *td, struct olseek_args *uap)
1899 {
1900 
1901 	return (kern_lseek(td, uap->fd, uap->offset, uap->whence));
1902 }
1903 #endif /* COMPAT_43 */
1904 
1905 #if defined(COMPAT_FREEBSD6)
1906 /* Version with the 'pad' argument */
1907 int
1908 freebsd6_lseek(struct thread *td, struct freebsd6_lseek_args *uap)
1909 {
1910 
1911 	return (kern_lseek(td, uap->fd, uap->offset, uap->whence));
1912 }
1913 #endif
1914 
1915 /*
1916  * Check access permissions using passed credentials.
1917  */
1918 static int
1919 vn_access(struct vnode *vp, int user_flags, struct ucred *cred,
1920      struct thread *td)
1921 {
1922 	accmode_t accmode;
1923 	int error;
1924 
1925 	/* Flags == 0 means only check for existence. */
1926 	if (user_flags == 0)
1927 		return (0);
1928 
1929 	accmode = 0;
1930 	if (user_flags & R_OK)
1931 		accmode |= VREAD;
1932 	if (user_flags & W_OK)
1933 		accmode |= VWRITE;
1934 	if (user_flags & X_OK)
1935 		accmode |= VEXEC;
1936 #ifdef MAC
1937 	error = mac_vnode_check_access(cred, vp, accmode);
1938 	if (error != 0)
1939 		return (error);
1940 #endif
1941 	if ((accmode & VWRITE) == 0 || (error = vn_writechk(vp)) == 0)
1942 		error = VOP_ACCESS(vp, accmode, cred, td);
1943 	return (error);
1944 }
1945 
1946 /*
1947  * Check access permissions using "real" credentials.
1948  */
1949 #ifndef _SYS_SYSPROTO_H_
1950 struct access_args {
1951 	char	*path;
1952 	int	amode;
1953 };
1954 #endif
1955 int
1956 sys_access(struct thread *td, struct access_args *uap)
1957 {
1958 
1959 	return (kern_accessat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1960 	    0, uap->amode));
1961 }
1962 
1963 #ifndef _SYS_SYSPROTO_H_
1964 struct faccessat_args {
1965 	int	dirfd;
1966 	char	*path;
1967 	int	amode;
1968 	int	flag;
1969 }
1970 #endif
1971 int
1972 sys_faccessat(struct thread *td, struct faccessat_args *uap)
1973 {
1974 
1975 	return (kern_accessat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
1976 	    uap->amode));
1977 }
1978 
1979 int
1980 kern_accessat(struct thread *td, int fd, const char *path,
1981     enum uio_seg pathseg, int flag, int amode)
1982 {
1983 	struct ucred *cred, *usecred;
1984 	struct vnode *vp;
1985 	struct nameidata nd;
1986 	int error;
1987 
1988 	if ((flag & ~(AT_EACCESS | AT_BENEATH)) != 0)
1989 		return (EINVAL);
1990 	if (amode != F_OK && (amode & ~(R_OK | W_OK | X_OK)) != 0)
1991 		return (EINVAL);
1992 
1993 	/*
1994 	 * Create and modify a temporary credential instead of one that
1995 	 * is potentially shared (if we need one).
1996 	 */
1997 	cred = td->td_ucred;
1998 	if ((flag & AT_EACCESS) == 0 &&
1999 	    ((cred->cr_uid != cred->cr_ruid ||
2000 	    cred->cr_rgid != cred->cr_groups[0]))) {
2001 		usecred = crdup(cred);
2002 		usecred->cr_uid = cred->cr_ruid;
2003 		usecred->cr_groups[0] = cred->cr_rgid;
2004 		td->td_ucred = usecred;
2005 	} else
2006 		usecred = cred;
2007 	AUDIT_ARG_VALUE(amode);
2008 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF |
2009 	    AUDITVNODE1 | ((flag & AT_BENEATH) != 0 ? BENEATH : 0),
2010 	    pathseg, path, fd, &cap_fstat_rights, td);
2011 	if ((error = namei(&nd)) != 0)
2012 		goto out;
2013 	vp = nd.ni_vp;
2014 
2015 	error = vn_access(vp, amode, usecred, td);
2016 	NDFREE(&nd, NDF_ONLY_PNBUF);
2017 	vput(vp);
2018 out:
2019 	if (usecred != cred) {
2020 		td->td_ucred = cred;
2021 		crfree(usecred);
2022 	}
2023 	return (error);
2024 }
2025 
2026 /*
2027  * Check access permissions using "effective" credentials.
2028  */
2029 #ifndef _SYS_SYSPROTO_H_
2030 struct eaccess_args {
2031 	char	*path;
2032 	int	amode;
2033 };
2034 #endif
2035 int
2036 sys_eaccess(struct thread *td, struct eaccess_args *uap)
2037 {
2038 
2039 	return (kern_accessat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2040 	    AT_EACCESS, uap->amode));
2041 }
2042 
2043 #if defined(COMPAT_43)
2044 /*
2045  * Get file status; this version follows links.
2046  */
2047 #ifndef _SYS_SYSPROTO_H_
2048 struct ostat_args {
2049 	char	*path;
2050 	struct ostat *ub;
2051 };
2052 #endif
2053 int
2054 ostat(struct thread *td, struct ostat_args *uap)
2055 {
2056 	struct stat sb;
2057 	struct ostat osb;
2058 	int error;
2059 
2060 	error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE,
2061 	    &sb, NULL);
2062 	if (error != 0)
2063 		return (error);
2064 	cvtstat(&sb, &osb);
2065 	return (copyout(&osb, uap->ub, sizeof (osb)));
2066 }
2067 
2068 /*
2069  * Get file status; this version does not follow links.
2070  */
2071 #ifndef _SYS_SYSPROTO_H_
2072 struct olstat_args {
2073 	char	*path;
2074 	struct ostat *ub;
2075 };
2076 #endif
2077 int
2078 olstat(struct thread *td, struct olstat_args *uap)
2079 {
2080 	struct stat sb;
2081 	struct ostat osb;
2082 	int error;
2083 
2084 	error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2085 	    UIO_USERSPACE, &sb, NULL);
2086 	if (error != 0)
2087 		return (error);
2088 	cvtstat(&sb, &osb);
2089 	return (copyout(&osb, uap->ub, sizeof (osb)));
2090 }
2091 
2092 /*
2093  * Convert from an old to a new stat structure.
2094  * XXX: many values are blindly truncated.
2095  */
2096 void
2097 cvtstat(struct stat *st, struct ostat *ost)
2098 {
2099 
2100 	bzero(ost, sizeof(*ost));
2101 	ost->st_dev = st->st_dev;
2102 	ost->st_ino = st->st_ino;
2103 	ost->st_mode = st->st_mode;
2104 	ost->st_nlink = st->st_nlink;
2105 	ost->st_uid = st->st_uid;
2106 	ost->st_gid = st->st_gid;
2107 	ost->st_rdev = st->st_rdev;
2108 	ost->st_size = MIN(st->st_size, INT32_MAX);
2109 	ost->st_atim = st->st_atim;
2110 	ost->st_mtim = st->st_mtim;
2111 	ost->st_ctim = st->st_ctim;
2112 	ost->st_blksize = st->st_blksize;
2113 	ost->st_blocks = st->st_blocks;
2114 	ost->st_flags = st->st_flags;
2115 	ost->st_gen = st->st_gen;
2116 }
2117 #endif /* COMPAT_43 */
2118 
2119 #if defined(COMPAT_43) || defined(COMPAT_FREEBSD11)
2120 int ino64_trunc_error;
2121 SYSCTL_INT(_vfs, OID_AUTO, ino64_trunc_error, CTLFLAG_RW,
2122     &ino64_trunc_error, 0,
2123     "Error on truncation of device, file or inode number, or link count");
2124 
2125 int
2126 freebsd11_cvtstat(struct stat *st, struct freebsd11_stat *ost)
2127 {
2128 
2129 	ost->st_dev = st->st_dev;
2130 	if (ost->st_dev != st->st_dev) {
2131 		switch (ino64_trunc_error) {
2132 		default:
2133 			/*
2134 			 * Since dev_t is almost raw, don't clamp to the
2135 			 * maximum for case 2, but ignore the error.
2136 			 */
2137 			break;
2138 		case 1:
2139 			return (EOVERFLOW);
2140 		}
2141 	}
2142 	ost->st_ino = st->st_ino;
2143 	if (ost->st_ino != st->st_ino) {
2144 		switch (ino64_trunc_error) {
2145 		default:
2146 		case 0:
2147 			break;
2148 		case 1:
2149 			return (EOVERFLOW);
2150 		case 2:
2151 			ost->st_ino = UINT32_MAX;
2152 			break;
2153 		}
2154 	}
2155 	ost->st_mode = st->st_mode;
2156 	ost->st_nlink = st->st_nlink;
2157 	if (ost->st_nlink != st->st_nlink) {
2158 		switch (ino64_trunc_error) {
2159 		default:
2160 		case 0:
2161 			break;
2162 		case 1:
2163 			return (EOVERFLOW);
2164 		case 2:
2165 			ost->st_nlink = UINT16_MAX;
2166 			break;
2167 		}
2168 	}
2169 	ost->st_uid = st->st_uid;
2170 	ost->st_gid = st->st_gid;
2171 	ost->st_rdev = st->st_rdev;
2172 	if (ost->st_rdev != st->st_rdev) {
2173 		switch (ino64_trunc_error) {
2174 		default:
2175 			break;
2176 		case 1:
2177 			return (EOVERFLOW);
2178 		}
2179 	}
2180 	ost->st_atim = st->st_atim;
2181 	ost->st_mtim = st->st_mtim;
2182 	ost->st_ctim = st->st_ctim;
2183 	ost->st_size = st->st_size;
2184 	ost->st_blocks = st->st_blocks;
2185 	ost->st_blksize = st->st_blksize;
2186 	ost->st_flags = st->st_flags;
2187 	ost->st_gen = st->st_gen;
2188 	ost->st_lspare = 0;
2189 	ost->st_birthtim = st->st_birthtim;
2190 	bzero((char *)&ost->st_birthtim + sizeof(ost->st_birthtim),
2191 	    sizeof(*ost) - offsetof(struct freebsd11_stat,
2192 	    st_birthtim) - sizeof(ost->st_birthtim));
2193 	return (0);
2194 }
2195 
2196 int
2197 freebsd11_stat(struct thread *td, struct freebsd11_stat_args* uap)
2198 {
2199 	struct stat sb;
2200 	struct freebsd11_stat osb;
2201 	int error;
2202 
2203 	error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE,
2204 	    &sb, NULL);
2205 	if (error != 0)
2206 		return (error);
2207 	error = freebsd11_cvtstat(&sb, &osb);
2208 	if (error == 0)
2209 		error = copyout(&osb, uap->ub, sizeof(osb));
2210 	return (error);
2211 }
2212 
2213 int
2214 freebsd11_lstat(struct thread *td, struct freebsd11_lstat_args* uap)
2215 {
2216 	struct stat sb;
2217 	struct freebsd11_stat osb;
2218 	int error;
2219 
2220 	error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2221 	    UIO_USERSPACE, &sb, NULL);
2222 	if (error != 0)
2223 		return (error);
2224 	error = freebsd11_cvtstat(&sb, &osb);
2225 	if (error == 0)
2226 		error = copyout(&osb, uap->ub, sizeof(osb));
2227 	return (error);
2228 }
2229 
2230 int
2231 freebsd11_fhstat(struct thread *td, struct freebsd11_fhstat_args* uap)
2232 {
2233 	struct fhandle fh;
2234 	struct stat sb;
2235 	struct freebsd11_stat osb;
2236 	int error;
2237 
2238 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
2239 	if (error != 0)
2240 		return (error);
2241 	error = kern_fhstat(td, fh, &sb);
2242 	if (error != 0)
2243 		return (error);
2244 	error = freebsd11_cvtstat(&sb, &osb);
2245 	if (error == 0)
2246 		error = copyout(&osb, uap->sb, sizeof(osb));
2247 	return (error);
2248 }
2249 
2250 int
2251 freebsd11_fstatat(struct thread *td, struct freebsd11_fstatat_args* uap)
2252 {
2253 	struct stat sb;
2254 	struct freebsd11_stat osb;
2255 	int error;
2256 
2257 	error = kern_statat(td, uap->flag, uap->fd, uap->path,
2258 	    UIO_USERSPACE, &sb, NULL);
2259 	if (error != 0)
2260 		return (error);
2261 	error = freebsd11_cvtstat(&sb, &osb);
2262 	if (error == 0)
2263 		error = copyout(&osb, uap->buf, sizeof(osb));
2264 	return (error);
2265 }
2266 #endif	/* COMPAT_FREEBSD11 */
2267 
2268 /*
2269  * Get file status
2270  */
2271 #ifndef _SYS_SYSPROTO_H_
2272 struct fstatat_args {
2273 	int	fd;
2274 	char	*path;
2275 	struct stat	*buf;
2276 	int	flag;
2277 }
2278 #endif
2279 int
2280 sys_fstatat(struct thread *td, struct fstatat_args *uap)
2281 {
2282 	struct stat sb;
2283 	int error;
2284 
2285 	error = kern_statat(td, uap->flag, uap->fd, uap->path,
2286 	    UIO_USERSPACE, &sb, NULL);
2287 	if (error == 0)
2288 		error = copyout(&sb, uap->buf, sizeof (sb));
2289 	return (error);
2290 }
2291 
2292 int
2293 kern_statat(struct thread *td, int flag, int fd, const char *path,
2294     enum uio_seg pathseg, struct stat *sbp,
2295     void (*hook)(struct vnode *vp, struct stat *sbp))
2296 {
2297 	struct nameidata nd;
2298 	int error;
2299 
2300 	if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_BENEATH)) != 0)
2301 		return (EINVAL);
2302 
2303 	NDINIT_ATRIGHTS(&nd, LOOKUP, ((flag & AT_SYMLINK_NOFOLLOW) != 0 ?
2304 	    NOFOLLOW : FOLLOW) | ((flag & AT_BENEATH) != 0 ? BENEATH : 0) |
2305 	    LOCKSHARED | LOCKLEAF | AUDITVNODE1, pathseg, path, fd,
2306 	    &cap_fstat_rights, td);
2307 
2308 	if ((error = namei(&nd)) != 0)
2309 		return (error);
2310 	error = vn_stat(nd.ni_vp, sbp, td->td_ucred, NOCRED, td);
2311 	if (error == 0) {
2312 		SDT_PROBE2(vfs, , stat, mode, path, sbp->st_mode);
2313 		if (S_ISREG(sbp->st_mode))
2314 			SDT_PROBE2(vfs, , stat, reg, path, pathseg);
2315 		if (__predict_false(hook != NULL))
2316 			hook(nd.ni_vp, sbp);
2317 	}
2318 	NDFREE(&nd, NDF_ONLY_PNBUF);
2319 	vput(nd.ni_vp);
2320 	if (error != 0)
2321 		return (error);
2322 #ifdef __STAT_TIME_T_EXT
2323 	sbp->st_atim_ext = 0;
2324 	sbp->st_mtim_ext = 0;
2325 	sbp->st_ctim_ext = 0;
2326 	sbp->st_btim_ext = 0;
2327 #endif
2328 #ifdef KTRACE
2329 	if (KTRPOINT(td, KTR_STRUCT))
2330 		ktrstat(sbp);
2331 #endif
2332 	return (0);
2333 }
2334 
2335 #if defined(COMPAT_FREEBSD11)
2336 /*
2337  * Implementation of the NetBSD [l]stat() functions.
2338  */
2339 void
2340 freebsd11_cvtnstat(struct stat *sb, struct nstat *nsb)
2341 {
2342 
2343 	bzero(nsb, sizeof(*nsb));
2344 	nsb->st_dev = sb->st_dev;
2345 	nsb->st_ino = sb->st_ino;
2346 	nsb->st_mode = sb->st_mode;
2347 	nsb->st_nlink = sb->st_nlink;
2348 	nsb->st_uid = sb->st_uid;
2349 	nsb->st_gid = sb->st_gid;
2350 	nsb->st_rdev = sb->st_rdev;
2351 	nsb->st_atim = sb->st_atim;
2352 	nsb->st_mtim = sb->st_mtim;
2353 	nsb->st_ctim = sb->st_ctim;
2354 	nsb->st_size = sb->st_size;
2355 	nsb->st_blocks = sb->st_blocks;
2356 	nsb->st_blksize = sb->st_blksize;
2357 	nsb->st_flags = sb->st_flags;
2358 	nsb->st_gen = sb->st_gen;
2359 	nsb->st_birthtim = sb->st_birthtim;
2360 }
2361 
2362 #ifndef _SYS_SYSPROTO_H_
2363 struct freebsd11_nstat_args {
2364 	char	*path;
2365 	struct nstat *ub;
2366 };
2367 #endif
2368 int
2369 freebsd11_nstat(struct thread *td, struct freebsd11_nstat_args *uap)
2370 {
2371 	struct stat sb;
2372 	struct nstat nsb;
2373 	int error;
2374 
2375 	error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE,
2376 	    &sb, NULL);
2377 	if (error != 0)
2378 		return (error);
2379 	freebsd11_cvtnstat(&sb, &nsb);
2380 	return (copyout(&nsb, uap->ub, sizeof (nsb)));
2381 }
2382 
2383 /*
2384  * NetBSD lstat.  Get file status; this version does not follow links.
2385  */
2386 #ifndef _SYS_SYSPROTO_H_
2387 struct freebsd11_nlstat_args {
2388 	char	*path;
2389 	struct nstat *ub;
2390 };
2391 #endif
2392 int
2393 freebsd11_nlstat(struct thread *td, struct freebsd11_nlstat_args *uap)
2394 {
2395 	struct stat sb;
2396 	struct nstat nsb;
2397 	int error;
2398 
2399 	error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2400 	    UIO_USERSPACE, &sb, NULL);
2401 	if (error != 0)
2402 		return (error);
2403 	freebsd11_cvtnstat(&sb, &nsb);
2404 	return (copyout(&nsb, uap->ub, sizeof (nsb)));
2405 }
2406 #endif /* COMPAT_FREEBSD11 */
2407 
2408 /*
2409  * Get configurable pathname variables.
2410  */
2411 #ifndef _SYS_SYSPROTO_H_
2412 struct pathconf_args {
2413 	char	*path;
2414 	int	name;
2415 };
2416 #endif
2417 int
2418 sys_pathconf(struct thread *td, struct pathconf_args *uap)
2419 {
2420 	long value;
2421 	int error;
2422 
2423 	error = kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name, FOLLOW,
2424 	    &value);
2425 	if (error == 0)
2426 		td->td_retval[0] = value;
2427 	return (error);
2428 }
2429 
2430 #ifndef _SYS_SYSPROTO_H_
2431 struct lpathconf_args {
2432 	char	*path;
2433 	int	name;
2434 };
2435 #endif
2436 int
2437 sys_lpathconf(struct thread *td, struct lpathconf_args *uap)
2438 {
2439 	long value;
2440 	int error;
2441 
2442 	error = kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name,
2443 	    NOFOLLOW, &value);
2444 	if (error == 0)
2445 		td->td_retval[0] = value;
2446 	return (error);
2447 }
2448 
2449 int
2450 kern_pathconf(struct thread *td, const char *path, enum uio_seg pathseg,
2451     int name, u_long flags, long *valuep)
2452 {
2453 	struct nameidata nd;
2454 	int error;
2455 
2456 	NDINIT(&nd, LOOKUP, LOCKSHARED | LOCKLEAF | AUDITVNODE1 | flags,
2457 	    pathseg, path, td);
2458 	if ((error = namei(&nd)) != 0)
2459 		return (error);
2460 	NDFREE(&nd, NDF_ONLY_PNBUF);
2461 
2462 	error = VOP_PATHCONF(nd.ni_vp, name, valuep);
2463 	vput(nd.ni_vp);
2464 	return (error);
2465 }
2466 
2467 /*
2468  * Return target name of a symbolic link.
2469  */
2470 #ifndef _SYS_SYSPROTO_H_
2471 struct readlink_args {
2472 	char	*path;
2473 	char	*buf;
2474 	size_t	count;
2475 };
2476 #endif
2477 int
2478 sys_readlink(struct thread *td, struct readlink_args *uap)
2479 {
2480 
2481 	return (kern_readlinkat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2482 	    uap->buf, UIO_USERSPACE, uap->count));
2483 }
2484 #ifndef _SYS_SYSPROTO_H_
2485 struct readlinkat_args {
2486 	int	fd;
2487 	char	*path;
2488 	char	*buf;
2489 	size_t	bufsize;
2490 };
2491 #endif
2492 int
2493 sys_readlinkat(struct thread *td, struct readlinkat_args *uap)
2494 {
2495 
2496 	return (kern_readlinkat(td, uap->fd, uap->path, UIO_USERSPACE,
2497 	    uap->buf, UIO_USERSPACE, uap->bufsize));
2498 }
2499 
2500 int
2501 kern_readlinkat(struct thread *td, int fd, const char *path,
2502     enum uio_seg pathseg, char *buf, enum uio_seg bufseg, size_t count)
2503 {
2504 	struct vnode *vp;
2505 	struct nameidata nd;
2506 	int error;
2507 
2508 	if (count > IOSIZE_MAX)
2509 		return (EINVAL);
2510 
2511 	NDINIT_AT(&nd, LOOKUP, NOFOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
2512 	    pathseg, path, fd, td);
2513 
2514 	if ((error = namei(&nd)) != 0)
2515 		return (error);
2516 	NDFREE(&nd, NDF_ONLY_PNBUF);
2517 	vp = nd.ni_vp;
2518 
2519 	error = kern_readlink_vp(vp, buf, bufseg, count, td);
2520 	vput(vp);
2521 
2522 	return (error);
2523 }
2524 
2525 /*
2526  * Helper function to readlink from a vnode
2527  */
2528 static int
2529 kern_readlink_vp(struct vnode *vp, char *buf, enum uio_seg bufseg, size_t count,
2530     struct thread *td)
2531 {
2532 	struct iovec aiov;
2533 	struct uio auio;
2534 	int error;
2535 
2536 	ASSERT_VOP_LOCKED(vp, "kern_readlink_vp(): vp not locked");
2537 #ifdef MAC
2538 	error = mac_vnode_check_readlink(td->td_ucred, vp);
2539 	if (error != 0)
2540 		return (error);
2541 #endif
2542 	if (vp->v_type != VLNK && (vp->v_vflag & VV_READLINK) == 0)
2543 		error = EINVAL;
2544 	else {
2545 		aiov.iov_base = buf;
2546 		aiov.iov_len = count;
2547 		auio.uio_iov = &aiov;
2548 		auio.uio_iovcnt = 1;
2549 		auio.uio_offset = 0;
2550 		auio.uio_rw = UIO_READ;
2551 		auio.uio_segflg = bufseg;
2552 		auio.uio_td = td;
2553 		auio.uio_resid = count;
2554 		error = VOP_READLINK(vp, &auio, td->td_ucred);
2555 		td->td_retval[0] = count - auio.uio_resid;
2556 	}
2557 	return (error);
2558 }
2559 
2560 /*
2561  * Common implementation code for chflags() and fchflags().
2562  */
2563 static int
2564 setfflags(struct thread *td, struct vnode *vp, u_long flags)
2565 {
2566 	struct mount *mp;
2567 	struct vattr vattr;
2568 	int error;
2569 
2570 	/* We can't support the value matching VNOVAL. */
2571 	if (flags == VNOVAL)
2572 		return (EOPNOTSUPP);
2573 
2574 	/*
2575 	 * Prevent non-root users from setting flags on devices.  When
2576 	 * a device is reused, users can retain ownership of the device
2577 	 * if they are allowed to set flags and programs assume that
2578 	 * chown can't fail when done as root.
2579 	 */
2580 	if (vp->v_type == VCHR || vp->v_type == VBLK) {
2581 		error = priv_check(td, PRIV_VFS_CHFLAGS_DEV);
2582 		if (error != 0)
2583 			return (error);
2584 	}
2585 
2586 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2587 		return (error);
2588 	VATTR_NULL(&vattr);
2589 	vattr.va_flags = flags;
2590 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2591 #ifdef MAC
2592 	error = mac_vnode_check_setflags(td->td_ucred, vp, vattr.va_flags);
2593 	if (error == 0)
2594 #endif
2595 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
2596 	VOP_UNLOCK(vp, 0);
2597 	vn_finished_write(mp);
2598 	return (error);
2599 }
2600 
2601 /*
2602  * Change flags of a file given a path name.
2603  */
2604 #ifndef _SYS_SYSPROTO_H_
2605 struct chflags_args {
2606 	const char *path;
2607 	u_long	flags;
2608 };
2609 #endif
2610 int
2611 sys_chflags(struct thread *td, struct chflags_args *uap)
2612 {
2613 
2614 	return (kern_chflagsat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2615 	    uap->flags, 0));
2616 }
2617 
2618 #ifndef _SYS_SYSPROTO_H_
2619 struct chflagsat_args {
2620 	int	fd;
2621 	const char *path;
2622 	u_long	flags;
2623 	int	atflag;
2624 }
2625 #endif
2626 int
2627 sys_chflagsat(struct thread *td, struct chflagsat_args *uap)
2628 {
2629 
2630 	if ((uap->atflag & ~(AT_SYMLINK_NOFOLLOW | AT_BENEATH)) != 0)
2631 		return (EINVAL);
2632 
2633 	return (kern_chflagsat(td, uap->fd, uap->path, UIO_USERSPACE,
2634 	    uap->flags, uap->atflag));
2635 }
2636 
2637 /*
2638  * Same as chflags() but doesn't follow symlinks.
2639  */
2640 #ifndef _SYS_SYSPROTO_H_
2641 struct lchflags_args {
2642 	const char *path;
2643 	u_long flags;
2644 };
2645 #endif
2646 int
2647 sys_lchflags(struct thread *td, struct lchflags_args *uap)
2648 {
2649 
2650 	return (kern_chflagsat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2651 	    uap->flags, AT_SYMLINK_NOFOLLOW));
2652 }
2653 
2654 static int
2655 kern_chflagsat(struct thread *td, int fd, const char *path,
2656     enum uio_seg pathseg, u_long flags, int atflag)
2657 {
2658 	struct nameidata nd;
2659 	int error, follow;
2660 
2661 	AUDIT_ARG_FFLAGS(flags);
2662 	follow = (atflag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
2663 	follow |= (atflag & AT_BENEATH) != 0 ? BENEATH : 0;
2664 	NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, pathseg, path, fd,
2665 	    &cap_fchflags_rights, td);
2666 	if ((error = namei(&nd)) != 0)
2667 		return (error);
2668 	NDFREE(&nd, NDF_ONLY_PNBUF);
2669 	error = setfflags(td, nd.ni_vp, flags);
2670 	vrele(nd.ni_vp);
2671 	return (error);
2672 }
2673 
2674 /*
2675  * Change flags of a file given a file descriptor.
2676  */
2677 #ifndef _SYS_SYSPROTO_H_
2678 struct fchflags_args {
2679 	int	fd;
2680 	u_long	flags;
2681 };
2682 #endif
2683 int
2684 sys_fchflags(struct thread *td, struct fchflags_args *uap)
2685 {
2686 	struct file *fp;
2687 	int error;
2688 
2689 	AUDIT_ARG_FD(uap->fd);
2690 	AUDIT_ARG_FFLAGS(uap->flags);
2691 	error = getvnode(td, uap->fd, &cap_fchflags_rights,
2692 	    &fp);
2693 	if (error != 0)
2694 		return (error);
2695 #ifdef AUDIT
2696 	vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
2697 	AUDIT_ARG_VNODE1(fp->f_vnode);
2698 	VOP_UNLOCK(fp->f_vnode, 0);
2699 #endif
2700 	error = setfflags(td, fp->f_vnode, uap->flags);
2701 	fdrop(fp, td);
2702 	return (error);
2703 }
2704 
2705 /*
2706  * Common implementation code for chmod(), lchmod() and fchmod().
2707  */
2708 int
2709 setfmode(struct thread *td, struct ucred *cred, struct vnode *vp, int mode)
2710 {
2711 	struct mount *mp;
2712 	struct vattr vattr;
2713 	int error;
2714 
2715 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2716 		return (error);
2717 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2718 	VATTR_NULL(&vattr);
2719 	vattr.va_mode = mode & ALLPERMS;
2720 #ifdef MAC
2721 	error = mac_vnode_check_setmode(cred, vp, vattr.va_mode);
2722 	if (error == 0)
2723 #endif
2724 		error = VOP_SETATTR(vp, &vattr, cred);
2725 	VOP_UNLOCK(vp, 0);
2726 	vn_finished_write(mp);
2727 	return (error);
2728 }
2729 
2730 /*
2731  * Change mode of a file given path name.
2732  */
2733 #ifndef _SYS_SYSPROTO_H_
2734 struct chmod_args {
2735 	char	*path;
2736 	int	mode;
2737 };
2738 #endif
2739 int
2740 sys_chmod(struct thread *td, struct chmod_args *uap)
2741 {
2742 
2743 	return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2744 	    uap->mode, 0));
2745 }
2746 
2747 #ifndef _SYS_SYSPROTO_H_
2748 struct fchmodat_args {
2749 	int	dirfd;
2750 	char	*path;
2751 	mode_t	mode;
2752 	int	flag;
2753 }
2754 #endif
2755 int
2756 sys_fchmodat(struct thread *td, struct fchmodat_args *uap)
2757 {
2758 
2759 	if ((uap->flag & ~(AT_SYMLINK_NOFOLLOW | AT_BENEATH)) != 0)
2760 		return (EINVAL);
2761 
2762 	return (kern_fchmodat(td, uap->fd, uap->path, UIO_USERSPACE,
2763 	    uap->mode, uap->flag));
2764 }
2765 
2766 /*
2767  * Change mode of a file given path name (don't follow links.)
2768  */
2769 #ifndef _SYS_SYSPROTO_H_
2770 struct lchmod_args {
2771 	char	*path;
2772 	int	mode;
2773 };
2774 #endif
2775 int
2776 sys_lchmod(struct thread *td, struct lchmod_args *uap)
2777 {
2778 
2779 	return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2780 	    uap->mode, AT_SYMLINK_NOFOLLOW));
2781 }
2782 
2783 int
2784 kern_fchmodat(struct thread *td, int fd, const char *path,
2785     enum uio_seg pathseg, mode_t mode, int flag)
2786 {
2787 	struct nameidata nd;
2788 	int error, follow;
2789 
2790 	AUDIT_ARG_MODE(mode);
2791 	follow = (flag & AT_SYMLINK_NOFOLLOW) != 0 ? NOFOLLOW : FOLLOW;
2792 	follow |= (flag & AT_BENEATH) != 0 ? BENEATH : 0;
2793 	NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, pathseg, path, fd,
2794 	    &cap_fchmod_rights, td);
2795 	if ((error = namei(&nd)) != 0)
2796 		return (error);
2797 	NDFREE(&nd, NDF_ONLY_PNBUF);
2798 	error = setfmode(td, td->td_ucred, nd.ni_vp, mode);
2799 	vrele(nd.ni_vp);
2800 	return (error);
2801 }
2802 
2803 /*
2804  * Change mode of a file given a file descriptor.
2805  */
2806 #ifndef _SYS_SYSPROTO_H_
2807 struct fchmod_args {
2808 	int	fd;
2809 	int	mode;
2810 };
2811 #endif
2812 int
2813 sys_fchmod(struct thread *td, struct fchmod_args *uap)
2814 {
2815 	struct file *fp;
2816 	int error;
2817 
2818 	AUDIT_ARG_FD(uap->fd);
2819 	AUDIT_ARG_MODE(uap->mode);
2820 
2821 	error = fget(td, uap->fd, &cap_fchmod_rights, &fp);
2822 	if (error != 0)
2823 		return (error);
2824 	error = fo_chmod(fp, uap->mode, td->td_ucred, td);
2825 	fdrop(fp, td);
2826 	return (error);
2827 }
2828 
2829 /*
2830  * Common implementation for chown(), lchown(), and fchown()
2831  */
2832 int
2833 setfown(struct thread *td, struct ucred *cred, struct vnode *vp, uid_t uid,
2834     gid_t gid)
2835 {
2836 	struct mount *mp;
2837 	struct vattr vattr;
2838 	int error;
2839 
2840 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2841 		return (error);
2842 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2843 	VATTR_NULL(&vattr);
2844 	vattr.va_uid = uid;
2845 	vattr.va_gid = gid;
2846 #ifdef MAC
2847 	error = mac_vnode_check_setowner(cred, vp, vattr.va_uid,
2848 	    vattr.va_gid);
2849 	if (error == 0)
2850 #endif
2851 		error = VOP_SETATTR(vp, &vattr, cred);
2852 	VOP_UNLOCK(vp, 0);
2853 	vn_finished_write(mp);
2854 	return (error);
2855 }
2856 
2857 /*
2858  * Set ownership given a path name.
2859  */
2860 #ifndef _SYS_SYSPROTO_H_
2861 struct chown_args {
2862 	char	*path;
2863 	int	uid;
2864 	int	gid;
2865 };
2866 #endif
2867 int
2868 sys_chown(struct thread *td, struct chown_args *uap)
2869 {
2870 
2871 	return (kern_fchownat(td, AT_FDCWD, uap->path, UIO_USERSPACE, uap->uid,
2872 	    uap->gid, 0));
2873 }
2874 
2875 #ifndef _SYS_SYSPROTO_H_
2876 struct fchownat_args {
2877 	int fd;
2878 	const char * path;
2879 	uid_t uid;
2880 	gid_t gid;
2881 	int flag;
2882 };
2883 #endif
2884 int
2885 sys_fchownat(struct thread *td, struct fchownat_args *uap)
2886 {
2887 
2888 	if ((uap->flag & ~(AT_SYMLINK_NOFOLLOW | AT_BENEATH)) != 0)
2889 		return (EINVAL);
2890 
2891 	return (kern_fchownat(td, uap->fd, uap->path, UIO_USERSPACE, uap->uid,
2892 	    uap->gid, uap->flag));
2893 }
2894 
2895 int
2896 kern_fchownat(struct thread *td, int fd, const char *path,
2897     enum uio_seg pathseg, int uid, int gid, int flag)
2898 {
2899 	struct nameidata nd;
2900 	int error, follow;
2901 
2902 	AUDIT_ARG_OWNER(uid, gid);
2903 	follow = (flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
2904 	follow |= (flag & AT_BENEATH) != 0 ? BENEATH : 0;
2905 	NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, pathseg, path, fd,
2906 	    &cap_fchown_rights, td);
2907 
2908 	if ((error = namei(&nd)) != 0)
2909 		return (error);
2910 	NDFREE(&nd, NDF_ONLY_PNBUF);
2911 	error = setfown(td, td->td_ucred, nd.ni_vp, uid, gid);
2912 	vrele(nd.ni_vp);
2913 	return (error);
2914 }
2915 
2916 /*
2917  * Set ownership given a path name, do not cross symlinks.
2918  */
2919 #ifndef _SYS_SYSPROTO_H_
2920 struct lchown_args {
2921 	char	*path;
2922 	int	uid;
2923 	int	gid;
2924 };
2925 #endif
2926 int
2927 sys_lchown(struct thread *td, struct lchown_args *uap)
2928 {
2929 
2930 	return (kern_fchownat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2931 	    uap->uid, uap->gid, AT_SYMLINK_NOFOLLOW));
2932 }
2933 
2934 /*
2935  * Set ownership given a file descriptor.
2936  */
2937 #ifndef _SYS_SYSPROTO_H_
2938 struct fchown_args {
2939 	int	fd;
2940 	int	uid;
2941 	int	gid;
2942 };
2943 #endif
2944 int
2945 sys_fchown(struct thread *td, struct fchown_args *uap)
2946 {
2947 	struct file *fp;
2948 	int error;
2949 
2950 	AUDIT_ARG_FD(uap->fd);
2951 	AUDIT_ARG_OWNER(uap->uid, uap->gid);
2952 	error = fget(td, uap->fd, &cap_fchown_rights, &fp);
2953 	if (error != 0)
2954 		return (error);
2955 	error = fo_chown(fp, uap->uid, uap->gid, td->td_ucred, td);
2956 	fdrop(fp, td);
2957 	return (error);
2958 }
2959 
2960 /*
2961  * Common implementation code for utimes(), lutimes(), and futimes().
2962  */
2963 static int
2964 getutimes(const struct timeval *usrtvp, enum uio_seg tvpseg,
2965     struct timespec *tsp)
2966 {
2967 	struct timeval tv[2];
2968 	const struct timeval *tvp;
2969 	int error;
2970 
2971 	if (usrtvp == NULL) {
2972 		vfs_timestamp(&tsp[0]);
2973 		tsp[1] = tsp[0];
2974 	} else {
2975 		if (tvpseg == UIO_SYSSPACE) {
2976 			tvp = usrtvp;
2977 		} else {
2978 			if ((error = copyin(usrtvp, tv, sizeof(tv))) != 0)
2979 				return (error);
2980 			tvp = tv;
2981 		}
2982 
2983 		if (tvp[0].tv_usec < 0 || tvp[0].tv_usec >= 1000000 ||
2984 		    tvp[1].tv_usec < 0 || tvp[1].tv_usec >= 1000000)
2985 			return (EINVAL);
2986 		TIMEVAL_TO_TIMESPEC(&tvp[0], &tsp[0]);
2987 		TIMEVAL_TO_TIMESPEC(&tvp[1], &tsp[1]);
2988 	}
2989 	return (0);
2990 }
2991 
2992 /*
2993  * Common implementation code for futimens(), utimensat().
2994  */
2995 #define	UTIMENS_NULL	0x1
2996 #define	UTIMENS_EXIT	0x2
2997 static int
2998 getutimens(const struct timespec *usrtsp, enum uio_seg tspseg,
2999     struct timespec *tsp, int *retflags)
3000 {
3001 	struct timespec tsnow;
3002 	int error;
3003 
3004 	vfs_timestamp(&tsnow);
3005 	*retflags = 0;
3006 	if (usrtsp == NULL) {
3007 		tsp[0] = tsnow;
3008 		tsp[1] = tsnow;
3009 		*retflags |= UTIMENS_NULL;
3010 		return (0);
3011 	}
3012 	if (tspseg == UIO_SYSSPACE) {
3013 		tsp[0] = usrtsp[0];
3014 		tsp[1] = usrtsp[1];
3015 	} else if ((error = copyin(usrtsp, tsp, sizeof(*tsp) * 2)) != 0)
3016 		return (error);
3017 	if (tsp[0].tv_nsec == UTIME_OMIT && tsp[1].tv_nsec == UTIME_OMIT)
3018 		*retflags |= UTIMENS_EXIT;
3019 	if (tsp[0].tv_nsec == UTIME_NOW && tsp[1].tv_nsec == UTIME_NOW)
3020 		*retflags |= UTIMENS_NULL;
3021 	if (tsp[0].tv_nsec == UTIME_OMIT)
3022 		tsp[0].tv_sec = VNOVAL;
3023 	else if (tsp[0].tv_nsec == UTIME_NOW)
3024 		tsp[0] = tsnow;
3025 	else if (tsp[0].tv_nsec < 0 || tsp[0].tv_nsec >= 1000000000L)
3026 		return (EINVAL);
3027 	if (tsp[1].tv_nsec == UTIME_OMIT)
3028 		tsp[1].tv_sec = VNOVAL;
3029 	else if (tsp[1].tv_nsec == UTIME_NOW)
3030 		tsp[1] = tsnow;
3031 	else if (tsp[1].tv_nsec < 0 || tsp[1].tv_nsec >= 1000000000L)
3032 		return (EINVAL);
3033 
3034 	return (0);
3035 }
3036 
3037 /*
3038  * Common implementation code for utimes(), lutimes(), futimes(), futimens(),
3039  * and utimensat().
3040  */
3041 static int
3042 setutimes(struct thread *td, struct vnode *vp, const struct timespec *ts,
3043     int numtimes, int nullflag)
3044 {
3045 	struct mount *mp;
3046 	struct vattr vattr;
3047 	int error, setbirthtime;
3048 
3049 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
3050 		return (error);
3051 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3052 	setbirthtime = 0;
3053 	if (numtimes < 3 && !VOP_GETATTR(vp, &vattr, td->td_ucred) &&
3054 	    timespeccmp(&ts[1], &vattr.va_birthtime, < ))
3055 		setbirthtime = 1;
3056 	VATTR_NULL(&vattr);
3057 	vattr.va_atime = ts[0];
3058 	vattr.va_mtime = ts[1];
3059 	if (setbirthtime)
3060 		vattr.va_birthtime = ts[1];
3061 	if (numtimes > 2)
3062 		vattr.va_birthtime = ts[2];
3063 	if (nullflag)
3064 		vattr.va_vaflags |= VA_UTIMES_NULL;
3065 #ifdef MAC
3066 	error = mac_vnode_check_setutimes(td->td_ucred, vp, vattr.va_atime,
3067 	    vattr.va_mtime);
3068 #endif
3069 	if (error == 0)
3070 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3071 	VOP_UNLOCK(vp, 0);
3072 	vn_finished_write(mp);
3073 	return (error);
3074 }
3075 
3076 /*
3077  * Set the access and modification times of a file.
3078  */
3079 #ifndef _SYS_SYSPROTO_H_
3080 struct utimes_args {
3081 	char	*path;
3082 	struct	timeval *tptr;
3083 };
3084 #endif
3085 int
3086 sys_utimes(struct thread *td, struct utimes_args *uap)
3087 {
3088 
3089 	return (kern_utimesat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
3090 	    uap->tptr, UIO_USERSPACE));
3091 }
3092 
3093 #ifndef _SYS_SYSPROTO_H_
3094 struct futimesat_args {
3095 	int fd;
3096 	const char * path;
3097 	const struct timeval * times;
3098 };
3099 #endif
3100 int
3101 sys_futimesat(struct thread *td, struct futimesat_args *uap)
3102 {
3103 
3104 	return (kern_utimesat(td, uap->fd, uap->path, UIO_USERSPACE,
3105 	    uap->times, UIO_USERSPACE));
3106 }
3107 
3108 int
3109 kern_utimesat(struct thread *td, int fd, const char *path,
3110     enum uio_seg pathseg, struct timeval *tptr, enum uio_seg tptrseg)
3111 {
3112 	struct nameidata nd;
3113 	struct timespec ts[2];
3114 	int error;
3115 
3116 	if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3117 		return (error);
3118 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, fd,
3119 	    &cap_futimes_rights, td);
3120 
3121 	if ((error = namei(&nd)) != 0)
3122 		return (error);
3123 	NDFREE(&nd, NDF_ONLY_PNBUF);
3124 	error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3125 	vrele(nd.ni_vp);
3126 	return (error);
3127 }
3128 
3129 /*
3130  * Set the access and modification times of a file.
3131  */
3132 #ifndef _SYS_SYSPROTO_H_
3133 struct lutimes_args {
3134 	char	*path;
3135 	struct	timeval *tptr;
3136 };
3137 #endif
3138 int
3139 sys_lutimes(struct thread *td, struct lutimes_args *uap)
3140 {
3141 
3142 	return (kern_lutimes(td, uap->path, UIO_USERSPACE, uap->tptr,
3143 	    UIO_USERSPACE));
3144 }
3145 
3146 int
3147 kern_lutimes(struct thread *td, const char *path, enum uio_seg pathseg,
3148     struct timeval *tptr, enum uio_seg tptrseg)
3149 {
3150 	struct timespec ts[2];
3151 	struct nameidata nd;
3152 	int error;
3153 
3154 	if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3155 		return (error);
3156 	NDINIT(&nd, LOOKUP, NOFOLLOW | AUDITVNODE1, pathseg, path, td);
3157 	if ((error = namei(&nd)) != 0)
3158 		return (error);
3159 	NDFREE(&nd, NDF_ONLY_PNBUF);
3160 	error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3161 	vrele(nd.ni_vp);
3162 	return (error);
3163 }
3164 
3165 /*
3166  * Set the access and modification times of a file.
3167  */
3168 #ifndef _SYS_SYSPROTO_H_
3169 struct futimes_args {
3170 	int	fd;
3171 	struct	timeval *tptr;
3172 };
3173 #endif
3174 int
3175 sys_futimes(struct thread *td, struct futimes_args *uap)
3176 {
3177 
3178 	return (kern_futimes(td, uap->fd, uap->tptr, UIO_USERSPACE));
3179 }
3180 
3181 int
3182 kern_futimes(struct thread *td, int fd, struct timeval *tptr,
3183     enum uio_seg tptrseg)
3184 {
3185 	struct timespec ts[2];
3186 	struct file *fp;
3187 	int error;
3188 
3189 	AUDIT_ARG_FD(fd);
3190 	error = getutimes(tptr, tptrseg, ts);
3191 	if (error != 0)
3192 		return (error);
3193 	error = getvnode(td, fd, &cap_futimes_rights, &fp);
3194 	if (error != 0)
3195 		return (error);
3196 #ifdef AUDIT
3197 	vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
3198 	AUDIT_ARG_VNODE1(fp->f_vnode);
3199 	VOP_UNLOCK(fp->f_vnode, 0);
3200 #endif
3201 	error = setutimes(td, fp->f_vnode, ts, 2, tptr == NULL);
3202 	fdrop(fp, td);
3203 	return (error);
3204 }
3205 
3206 int
3207 sys_futimens(struct thread *td, struct futimens_args *uap)
3208 {
3209 
3210 	return (kern_futimens(td, uap->fd, uap->times, UIO_USERSPACE));
3211 }
3212 
3213 int
3214 kern_futimens(struct thread *td, int fd, struct timespec *tptr,
3215     enum uio_seg tptrseg)
3216 {
3217 	struct timespec ts[2];
3218 	struct file *fp;
3219 	int error, flags;
3220 
3221 	AUDIT_ARG_FD(fd);
3222 	error = getutimens(tptr, tptrseg, ts, &flags);
3223 	if (error != 0)
3224 		return (error);
3225 	if (flags & UTIMENS_EXIT)
3226 		return (0);
3227 	error = getvnode(td, fd, &cap_futimes_rights, &fp);
3228 	if (error != 0)
3229 		return (error);
3230 #ifdef AUDIT
3231 	vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
3232 	AUDIT_ARG_VNODE1(fp->f_vnode);
3233 	VOP_UNLOCK(fp->f_vnode, 0);
3234 #endif
3235 	error = setutimes(td, fp->f_vnode, ts, 2, flags & UTIMENS_NULL);
3236 	fdrop(fp, td);
3237 	return (error);
3238 }
3239 
3240 int
3241 sys_utimensat(struct thread *td, struct utimensat_args *uap)
3242 {
3243 
3244 	return (kern_utimensat(td, uap->fd, uap->path, UIO_USERSPACE,
3245 	    uap->times, UIO_USERSPACE, uap->flag));
3246 }
3247 
3248 int
3249 kern_utimensat(struct thread *td, int fd, const char *path,
3250     enum uio_seg pathseg, struct timespec *tptr, enum uio_seg tptrseg,
3251     int flag)
3252 {
3253 	struct nameidata nd;
3254 	struct timespec ts[2];
3255 	int error, flags;
3256 
3257 	if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_BENEATH)) != 0)
3258 		return (EINVAL);
3259 
3260 	if ((error = getutimens(tptr, tptrseg, ts, &flags)) != 0)
3261 		return (error);
3262 	NDINIT_ATRIGHTS(&nd, LOOKUP, ((flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW :
3263 	    FOLLOW) |  ((flag & AT_BENEATH) != 0 ? BENEATH : 0) | AUDITVNODE1,
3264 	    pathseg, path, fd, &cap_futimes_rights, td);
3265 	if ((error = namei(&nd)) != 0)
3266 		return (error);
3267 	/*
3268 	 * We are allowed to call namei() regardless of 2xUTIME_OMIT.
3269 	 * POSIX states:
3270 	 * "If both tv_nsec fields are UTIME_OMIT... EACCESS may be detected."
3271 	 * "Search permission is denied by a component of the path prefix."
3272 	 */
3273 	NDFREE(&nd, NDF_ONLY_PNBUF);
3274 	if ((flags & UTIMENS_EXIT) == 0)
3275 		error = setutimes(td, nd.ni_vp, ts, 2, flags & UTIMENS_NULL);
3276 	vrele(nd.ni_vp);
3277 	return (error);
3278 }
3279 
3280 /*
3281  * Truncate a file given its path name.
3282  */
3283 #ifndef _SYS_SYSPROTO_H_
3284 struct truncate_args {
3285 	char	*path;
3286 	int	pad;
3287 	off_t	length;
3288 };
3289 #endif
3290 int
3291 sys_truncate(struct thread *td, struct truncate_args *uap)
3292 {
3293 
3294 	return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3295 }
3296 
3297 int
3298 kern_truncate(struct thread *td, const char *path, enum uio_seg pathseg,
3299     off_t length)
3300 {
3301 	struct mount *mp;
3302 	struct vnode *vp;
3303 	void *rl_cookie;
3304 	struct vattr vattr;
3305 	struct nameidata nd;
3306 	int error;
3307 
3308 	if (length < 0)
3309 		return(EINVAL);
3310 	NDINIT(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, td);
3311 	if ((error = namei(&nd)) != 0)
3312 		return (error);
3313 	vp = nd.ni_vp;
3314 	rl_cookie = vn_rangelock_wlock(vp, 0, OFF_MAX);
3315 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) {
3316 		vn_rangelock_unlock(vp, rl_cookie);
3317 		vrele(vp);
3318 		return (error);
3319 	}
3320 	NDFREE(&nd, NDF_ONLY_PNBUF);
3321 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3322 	if (vp->v_type == VDIR)
3323 		error = EISDIR;
3324 #ifdef MAC
3325 	else if ((error = mac_vnode_check_write(td->td_ucred, NOCRED, vp))) {
3326 	}
3327 #endif
3328 	else if ((error = vn_writechk(vp)) == 0 &&
3329 	    (error = VOP_ACCESS(vp, VWRITE, td->td_ucred, td)) == 0) {
3330 		VATTR_NULL(&vattr);
3331 		vattr.va_size = length;
3332 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3333 	}
3334 	VOP_UNLOCK(vp, 0);
3335 	vn_finished_write(mp);
3336 	vn_rangelock_unlock(vp, rl_cookie);
3337 	vrele(vp);
3338 	return (error);
3339 }
3340 
3341 #if defined(COMPAT_43)
3342 /*
3343  * Truncate a file given its path name.
3344  */
3345 #ifndef _SYS_SYSPROTO_H_
3346 struct otruncate_args {
3347 	char	*path;
3348 	long	length;
3349 };
3350 #endif
3351 int
3352 otruncate(struct thread *td, struct otruncate_args *uap)
3353 {
3354 
3355 	return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3356 }
3357 #endif /* COMPAT_43 */
3358 
3359 #if defined(COMPAT_FREEBSD6)
3360 /* Versions with the pad argument */
3361 int
3362 freebsd6_truncate(struct thread *td, struct freebsd6_truncate_args *uap)
3363 {
3364 
3365 	return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3366 }
3367 
3368 int
3369 freebsd6_ftruncate(struct thread *td, struct freebsd6_ftruncate_args *uap)
3370 {
3371 
3372 	return (kern_ftruncate(td, uap->fd, uap->length));
3373 }
3374 #endif
3375 
3376 int
3377 kern_fsync(struct thread *td, int fd, bool fullsync)
3378 {
3379 	struct vnode *vp;
3380 	struct mount *mp;
3381 	struct file *fp;
3382 	int error, lock_flags;
3383 
3384 	AUDIT_ARG_FD(fd);
3385 	error = getvnode(td, fd, &cap_fsync_rights, &fp);
3386 	if (error != 0)
3387 		return (error);
3388 	vp = fp->f_vnode;
3389 #if 0
3390 	if (!fullsync)
3391 		/* XXXKIB: compete outstanding aio writes */;
3392 #endif
3393 	error = vn_start_write(vp, &mp, V_WAIT | PCATCH);
3394 	if (error != 0)
3395 		goto drop;
3396 	if (MNT_SHARED_WRITES(mp) ||
3397 	    ((mp == NULL) && MNT_SHARED_WRITES(vp->v_mount))) {
3398 		lock_flags = LK_SHARED;
3399 	} else {
3400 		lock_flags = LK_EXCLUSIVE;
3401 	}
3402 	vn_lock(vp, lock_flags | LK_RETRY);
3403 	AUDIT_ARG_VNODE1(vp);
3404 	if (vp->v_object != NULL) {
3405 		VM_OBJECT_WLOCK(vp->v_object);
3406 		vm_object_page_clean(vp->v_object, 0, 0, 0);
3407 		VM_OBJECT_WUNLOCK(vp->v_object);
3408 	}
3409 	error = fullsync ? VOP_FSYNC(vp, MNT_WAIT, td) : VOP_FDATASYNC(vp, td);
3410 	VOP_UNLOCK(vp, 0);
3411 	vn_finished_write(mp);
3412 drop:
3413 	fdrop(fp, td);
3414 	return (error);
3415 }
3416 
3417 /*
3418  * Sync an open file.
3419  */
3420 #ifndef _SYS_SYSPROTO_H_
3421 struct fsync_args {
3422 	int	fd;
3423 };
3424 #endif
3425 int
3426 sys_fsync(struct thread *td, struct fsync_args *uap)
3427 {
3428 
3429 	return (kern_fsync(td, uap->fd, true));
3430 }
3431 
3432 int
3433 sys_fdatasync(struct thread *td, struct fdatasync_args *uap)
3434 {
3435 
3436 	return (kern_fsync(td, uap->fd, false));
3437 }
3438 
3439 /*
3440  * Rename files.  Source and destination must either both be directories, or
3441  * both not be directories.  If target is a directory, it must be empty.
3442  */
3443 #ifndef _SYS_SYSPROTO_H_
3444 struct rename_args {
3445 	char	*from;
3446 	char	*to;
3447 };
3448 #endif
3449 int
3450 sys_rename(struct thread *td, struct rename_args *uap)
3451 {
3452 
3453 	return (kern_renameat(td, AT_FDCWD, uap->from, AT_FDCWD,
3454 	    uap->to, UIO_USERSPACE));
3455 }
3456 
3457 #ifndef _SYS_SYSPROTO_H_
3458 struct renameat_args {
3459 	int	oldfd;
3460 	char	*old;
3461 	int	newfd;
3462 	char	*new;
3463 };
3464 #endif
3465 int
3466 sys_renameat(struct thread *td, struct renameat_args *uap)
3467 {
3468 
3469 	return (kern_renameat(td, uap->oldfd, uap->old, uap->newfd, uap->new,
3470 	    UIO_USERSPACE));
3471 }
3472 
3473 int
3474 kern_renameat(struct thread *td, int oldfd, const char *old, int newfd,
3475     const char *new, enum uio_seg pathseg)
3476 {
3477 	struct mount *mp = NULL;
3478 	struct vnode *tvp, *fvp, *tdvp;
3479 	struct nameidata fromnd, tond;
3480 	int error;
3481 
3482 again:
3483 	bwillwrite();
3484 #ifdef MAC
3485 	NDINIT_ATRIGHTS(&fromnd, DELETE, LOCKPARENT | LOCKLEAF | SAVESTART |
3486 	    AUDITVNODE1, pathseg, old, oldfd,
3487 	    &cap_renameat_source_rights, td);
3488 #else
3489 	NDINIT_ATRIGHTS(&fromnd, DELETE, WANTPARENT | SAVESTART | AUDITVNODE1,
3490 	    pathseg, old, oldfd,
3491 	    &cap_renameat_source_rights, td);
3492 #endif
3493 
3494 	if ((error = namei(&fromnd)) != 0)
3495 		return (error);
3496 #ifdef MAC
3497 	error = mac_vnode_check_rename_from(td->td_ucred, fromnd.ni_dvp,
3498 	    fromnd.ni_vp, &fromnd.ni_cnd);
3499 	VOP_UNLOCK(fromnd.ni_dvp, 0);
3500 	if (fromnd.ni_dvp != fromnd.ni_vp)
3501 		VOP_UNLOCK(fromnd.ni_vp, 0);
3502 #endif
3503 	fvp = fromnd.ni_vp;
3504 	NDINIT_ATRIGHTS(&tond, RENAME, LOCKPARENT | LOCKLEAF | NOCACHE |
3505 	    SAVESTART | AUDITVNODE2, pathseg, new, newfd,
3506 	    &cap_renameat_target_rights, td);
3507 	if (fromnd.ni_vp->v_type == VDIR)
3508 		tond.ni_cnd.cn_flags |= WILLBEDIR;
3509 	if ((error = namei(&tond)) != 0) {
3510 		/* Translate error code for rename("dir1", "dir2/."). */
3511 		if (error == EISDIR && fvp->v_type == VDIR)
3512 			error = EINVAL;
3513 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3514 		vrele(fromnd.ni_dvp);
3515 		vrele(fvp);
3516 		goto out1;
3517 	}
3518 	tdvp = tond.ni_dvp;
3519 	tvp = tond.ni_vp;
3520 	error = vn_start_write(fvp, &mp, V_NOWAIT);
3521 	if (error != 0) {
3522 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3523 		NDFREE(&tond, NDF_ONLY_PNBUF);
3524 		if (tvp != NULL)
3525 			vput(tvp);
3526 		if (tdvp == tvp)
3527 			vrele(tdvp);
3528 		else
3529 			vput(tdvp);
3530 		vrele(fromnd.ni_dvp);
3531 		vrele(fvp);
3532 		vrele(tond.ni_startdir);
3533 		if (fromnd.ni_startdir != NULL)
3534 			vrele(fromnd.ni_startdir);
3535 		error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH);
3536 		if (error != 0)
3537 			return (error);
3538 		goto again;
3539 	}
3540 	if (tvp != NULL) {
3541 		if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
3542 			error = ENOTDIR;
3543 			goto out;
3544 		} else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
3545 			error = EISDIR;
3546 			goto out;
3547 		}
3548 #ifdef CAPABILITIES
3549 		if (newfd != AT_FDCWD) {
3550 			/*
3551 			 * If the target already exists we require CAP_UNLINKAT
3552 			 * from 'newfd'.
3553 			 */
3554 			error = cap_check(&tond.ni_filecaps.fc_rights,
3555 			    &cap_unlinkat_rights);
3556 			if (error != 0)
3557 				goto out;
3558 		}
3559 #endif
3560 	}
3561 	if (fvp == tdvp) {
3562 		error = EINVAL;
3563 		goto out;
3564 	}
3565 	/*
3566 	 * If the source is the same as the destination (that is, if they
3567 	 * are links to the same vnode), then there is nothing to do.
3568 	 */
3569 	if (fvp == tvp)
3570 		error = -1;
3571 #ifdef MAC
3572 	else
3573 		error = mac_vnode_check_rename_to(td->td_ucred, tdvp,
3574 		    tond.ni_vp, fromnd.ni_dvp == tdvp, &tond.ni_cnd);
3575 #endif
3576 out:
3577 	if (error == 0) {
3578 		error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
3579 		    tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
3580 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3581 		NDFREE(&tond, NDF_ONLY_PNBUF);
3582 	} else {
3583 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3584 		NDFREE(&tond, NDF_ONLY_PNBUF);
3585 		if (tvp != NULL)
3586 			vput(tvp);
3587 		if (tdvp == tvp)
3588 			vrele(tdvp);
3589 		else
3590 			vput(tdvp);
3591 		vrele(fromnd.ni_dvp);
3592 		vrele(fvp);
3593 	}
3594 	vrele(tond.ni_startdir);
3595 	vn_finished_write(mp);
3596 out1:
3597 	if (fromnd.ni_startdir)
3598 		vrele(fromnd.ni_startdir);
3599 	if (error == -1)
3600 		return (0);
3601 	return (error);
3602 }
3603 
3604 /*
3605  * Make a directory file.
3606  */
3607 #ifndef _SYS_SYSPROTO_H_
3608 struct mkdir_args {
3609 	char	*path;
3610 	int	mode;
3611 };
3612 #endif
3613 int
3614 sys_mkdir(struct thread *td, struct mkdir_args *uap)
3615 {
3616 
3617 	return (kern_mkdirat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
3618 	    uap->mode));
3619 }
3620 
3621 #ifndef _SYS_SYSPROTO_H_
3622 struct mkdirat_args {
3623 	int	fd;
3624 	char	*path;
3625 	mode_t	mode;
3626 };
3627 #endif
3628 int
3629 sys_mkdirat(struct thread *td, struct mkdirat_args *uap)
3630 {
3631 
3632 	return (kern_mkdirat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode));
3633 }
3634 
3635 int
3636 kern_mkdirat(struct thread *td, int fd, const char *path, enum uio_seg segflg,
3637     int mode)
3638 {
3639 	struct mount *mp;
3640 	struct vnode *vp;
3641 	struct vattr vattr;
3642 	struct nameidata nd;
3643 	int error;
3644 
3645 	AUDIT_ARG_MODE(mode);
3646 restart:
3647 	bwillwrite();
3648 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
3649 	    NOCACHE, segflg, path, fd, &cap_mkdirat_rights,
3650 	    td);
3651 	nd.ni_cnd.cn_flags |= WILLBEDIR;
3652 	if ((error = namei(&nd)) != 0)
3653 		return (error);
3654 	vp = nd.ni_vp;
3655 	if (vp != NULL) {
3656 		NDFREE(&nd, NDF_ONLY_PNBUF);
3657 		/*
3658 		 * XXX namei called with LOCKPARENT but not LOCKLEAF has
3659 		 * the strange behaviour of leaving the vnode unlocked
3660 		 * if the target is the same vnode as the parent.
3661 		 */
3662 		if (vp == nd.ni_dvp)
3663 			vrele(nd.ni_dvp);
3664 		else
3665 			vput(nd.ni_dvp);
3666 		vrele(vp);
3667 		return (EEXIST);
3668 	}
3669 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3670 		NDFREE(&nd, NDF_ONLY_PNBUF);
3671 		vput(nd.ni_dvp);
3672 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3673 			return (error);
3674 		goto restart;
3675 	}
3676 	VATTR_NULL(&vattr);
3677 	vattr.va_type = VDIR;
3678 	vattr.va_mode = (mode & ACCESSPERMS) &~ td->td_proc->p_fd->fd_cmask;
3679 #ifdef MAC
3680 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
3681 	    &vattr);
3682 	if (error != 0)
3683 		goto out;
3684 #endif
3685 	error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
3686 #ifdef MAC
3687 out:
3688 #endif
3689 	NDFREE(&nd, NDF_ONLY_PNBUF);
3690 	vput(nd.ni_dvp);
3691 	if (error == 0)
3692 		vput(nd.ni_vp);
3693 	vn_finished_write(mp);
3694 	return (error);
3695 }
3696 
3697 /*
3698  * Remove a directory file.
3699  */
3700 #ifndef _SYS_SYSPROTO_H_
3701 struct rmdir_args {
3702 	char	*path;
3703 };
3704 #endif
3705 int
3706 sys_rmdir(struct thread *td, struct rmdir_args *uap)
3707 {
3708 
3709 	return (kern_rmdirat(td, AT_FDCWD, uap->path, UIO_USERSPACE, 0));
3710 }
3711 
3712 int
3713 kern_rmdirat(struct thread *td, int fd, const char *path, enum uio_seg pathseg,
3714     int flag)
3715 {
3716 	struct mount *mp;
3717 	struct vnode *vp;
3718 	struct nameidata nd;
3719 	int error;
3720 
3721 restart:
3722 	bwillwrite();
3723 	NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1 |
3724 	    ((flag & AT_BENEATH) != 0 ? BENEATH : 0),
3725 	    pathseg, path, fd, &cap_unlinkat_rights, td);
3726 	if ((error = namei(&nd)) != 0)
3727 		return (error);
3728 	vp = nd.ni_vp;
3729 	if (vp->v_type != VDIR) {
3730 		error = ENOTDIR;
3731 		goto out;
3732 	}
3733 	/*
3734 	 * No rmdir "." please.
3735 	 */
3736 	if (nd.ni_dvp == vp) {
3737 		error = EINVAL;
3738 		goto out;
3739 	}
3740 	/*
3741 	 * The root of a mounted filesystem cannot be deleted.
3742 	 */
3743 	if (vp->v_vflag & VV_ROOT) {
3744 		error = EBUSY;
3745 		goto out;
3746 	}
3747 #ifdef MAC
3748 	error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
3749 	    &nd.ni_cnd);
3750 	if (error != 0)
3751 		goto out;
3752 #endif
3753 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3754 		NDFREE(&nd, NDF_ONLY_PNBUF);
3755 		vput(vp);
3756 		if (nd.ni_dvp == vp)
3757 			vrele(nd.ni_dvp);
3758 		else
3759 			vput(nd.ni_dvp);
3760 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3761 			return (error);
3762 		goto restart;
3763 	}
3764 	vfs_notify_upper(vp, VFS_NOTIFY_UPPER_UNLINK);
3765 	error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
3766 	vn_finished_write(mp);
3767 out:
3768 	NDFREE(&nd, NDF_ONLY_PNBUF);
3769 	vput(vp);
3770 	if (nd.ni_dvp == vp)
3771 		vrele(nd.ni_dvp);
3772 	else
3773 		vput(nd.ni_dvp);
3774 	return (error);
3775 }
3776 
3777 #if defined(COMPAT_43) || defined(COMPAT_FREEBSD11)
3778 int
3779 freebsd11_kern_getdirentries(struct thread *td, int fd, char *ubuf, u_int count,
3780     long *basep, void (*func)(struct freebsd11_dirent *))
3781 {
3782 	struct freebsd11_dirent dstdp;
3783 	struct dirent *dp, *edp;
3784 	char *dirbuf;
3785 	off_t base;
3786 	ssize_t resid, ucount;
3787 	int error;
3788 
3789 	/* XXX arbitrary sanity limit on `count'. */
3790 	count = min(count, 64 * 1024);
3791 
3792 	dirbuf = malloc(count, M_TEMP, M_WAITOK);
3793 
3794 	error = kern_getdirentries(td, fd, dirbuf, count, &base, &resid,
3795 	    UIO_SYSSPACE);
3796 	if (error != 0)
3797 		goto done;
3798 	if (basep != NULL)
3799 		*basep = base;
3800 
3801 	ucount = 0;
3802 	for (dp = (struct dirent *)dirbuf,
3803 	    edp = (struct dirent *)&dirbuf[count - resid];
3804 	    ucount < count && dp < edp; ) {
3805 		if (dp->d_reclen == 0)
3806 			break;
3807 		MPASS(dp->d_reclen >= _GENERIC_DIRLEN(0));
3808 		if (dp->d_namlen >= sizeof(dstdp.d_name))
3809 			continue;
3810 		dstdp.d_type = dp->d_type;
3811 		dstdp.d_namlen = dp->d_namlen;
3812 		dstdp.d_fileno = dp->d_fileno;		/* truncate */
3813 		if (dstdp.d_fileno != dp->d_fileno) {
3814 			switch (ino64_trunc_error) {
3815 			default:
3816 			case 0:
3817 				break;
3818 			case 1:
3819 				error = EOVERFLOW;
3820 				goto done;
3821 			case 2:
3822 				dstdp.d_fileno = UINT32_MAX;
3823 				break;
3824 			}
3825 		}
3826 		dstdp.d_reclen = sizeof(dstdp) - sizeof(dstdp.d_name) +
3827 		    ((dp->d_namlen + 1 + 3) &~ 3);
3828 		bcopy(dp->d_name, dstdp.d_name, dstdp.d_namlen);
3829 		bzero(dstdp.d_name + dstdp.d_namlen,
3830 		    dstdp.d_reclen - offsetof(struct freebsd11_dirent, d_name) -
3831 		    dstdp.d_namlen);
3832 		MPASS(dstdp.d_reclen <= dp->d_reclen);
3833 		MPASS(ucount + dstdp.d_reclen <= count);
3834 		if (func != NULL)
3835 			func(&dstdp);
3836 		error = copyout(&dstdp, ubuf + ucount, dstdp.d_reclen);
3837 		if (error != 0)
3838 			break;
3839 		dp = (struct dirent *)((char *)dp + dp->d_reclen);
3840 		ucount += dstdp.d_reclen;
3841 	}
3842 
3843 done:
3844 	free(dirbuf, M_TEMP);
3845 	if (error == 0)
3846 		td->td_retval[0] = ucount;
3847 	return (error);
3848 }
3849 #endif /* COMPAT */
3850 
3851 #ifdef COMPAT_43
3852 static void
3853 ogetdirentries_cvt(struct freebsd11_dirent *dp)
3854 {
3855 #if (BYTE_ORDER == LITTLE_ENDIAN)
3856 	/*
3857 	 * The expected low byte of dp->d_namlen is our dp->d_type.
3858 	 * The high MBZ byte of dp->d_namlen is our dp->d_namlen.
3859 	 */
3860 	dp->d_type = dp->d_namlen;
3861 	dp->d_namlen = 0;
3862 #else
3863 	/*
3864 	 * The dp->d_type is the high byte of the expected dp->d_namlen,
3865 	 * so must be zero'ed.
3866 	 */
3867 	dp->d_type = 0;
3868 #endif
3869 }
3870 
3871 /*
3872  * Read a block of directory entries in a filesystem independent format.
3873  */
3874 #ifndef _SYS_SYSPROTO_H_
3875 struct ogetdirentries_args {
3876 	int	fd;
3877 	char	*buf;
3878 	u_int	count;
3879 	long	*basep;
3880 };
3881 #endif
3882 int
3883 ogetdirentries(struct thread *td, struct ogetdirentries_args *uap)
3884 {
3885 	long loff;
3886 	int error;
3887 
3888 	error = kern_ogetdirentries(td, uap, &loff);
3889 	if (error == 0)
3890 		error = copyout(&loff, uap->basep, sizeof(long));
3891 	return (error);
3892 }
3893 
3894 int
3895 kern_ogetdirentries(struct thread *td, struct ogetdirentries_args *uap,
3896     long *ploff)
3897 {
3898 	long base;
3899 	int error;
3900 
3901 	/* XXX arbitrary sanity limit on `count'. */
3902 	if (uap->count > 64 * 1024)
3903 		return (EINVAL);
3904 
3905 	error = freebsd11_kern_getdirentries(td, uap->fd, uap->buf, uap->count,
3906 	    &base, ogetdirentries_cvt);
3907 
3908 	if (error == 0 && uap->basep != NULL)
3909 		error = copyout(&base, uap->basep, sizeof(long));
3910 
3911 	return (error);
3912 }
3913 #endif /* COMPAT_43 */
3914 
3915 #if defined(COMPAT_FREEBSD11)
3916 #ifndef _SYS_SYSPROTO_H_
3917 struct freebsd11_getdirentries_args {
3918 	int	fd;
3919 	char	*buf;
3920 	u_int	count;
3921 	long	*basep;
3922 };
3923 #endif
3924 int
3925 freebsd11_getdirentries(struct thread *td,
3926     struct freebsd11_getdirentries_args *uap)
3927 {
3928 	long base;
3929 	int error;
3930 
3931 	error = freebsd11_kern_getdirentries(td, uap->fd, uap->buf, uap->count,
3932 	    &base, NULL);
3933 
3934 	if (error == 0 && uap->basep != NULL)
3935 		error = copyout(&base, uap->basep, sizeof(long));
3936 	return (error);
3937 }
3938 
3939 int
3940 freebsd11_getdents(struct thread *td, struct freebsd11_getdents_args *uap)
3941 {
3942 	struct freebsd11_getdirentries_args ap;
3943 
3944 	ap.fd = uap->fd;
3945 	ap.buf = uap->buf;
3946 	ap.count = uap->count;
3947 	ap.basep = NULL;
3948 	return (freebsd11_getdirentries(td, &ap));
3949 }
3950 #endif /* COMPAT_FREEBSD11 */
3951 
3952 /*
3953  * Read a block of directory entries in a filesystem independent format.
3954  */
3955 int
3956 sys_getdirentries(struct thread *td, struct getdirentries_args *uap)
3957 {
3958 	off_t base;
3959 	int error;
3960 
3961 	error = kern_getdirentries(td, uap->fd, uap->buf, uap->count, &base,
3962 	    NULL, UIO_USERSPACE);
3963 	if (error != 0)
3964 		return (error);
3965 	if (uap->basep != NULL)
3966 		error = copyout(&base, uap->basep, sizeof(off_t));
3967 	return (error);
3968 }
3969 
3970 int
3971 kern_getdirentries(struct thread *td, int fd, char *buf, size_t count,
3972     off_t *basep, ssize_t *residp, enum uio_seg bufseg)
3973 {
3974 	struct vnode *vp;
3975 	struct file *fp;
3976 	struct uio auio;
3977 	struct iovec aiov;
3978 	off_t loff;
3979 	int error, eofflag;
3980 	off_t foffset;
3981 
3982 	AUDIT_ARG_FD(fd);
3983 	if (count > IOSIZE_MAX)
3984 		return (EINVAL);
3985 	auio.uio_resid = count;
3986 	error = getvnode(td, fd, &cap_read_rights, &fp);
3987 	if (error != 0)
3988 		return (error);
3989 	if ((fp->f_flag & FREAD) == 0) {
3990 		fdrop(fp, td);
3991 		return (EBADF);
3992 	}
3993 	vp = fp->f_vnode;
3994 	foffset = foffset_lock(fp, 0);
3995 unionread:
3996 	if (vp->v_type != VDIR) {
3997 		error = EINVAL;
3998 		goto fail;
3999 	}
4000 	aiov.iov_base = buf;
4001 	aiov.iov_len = count;
4002 	auio.uio_iov = &aiov;
4003 	auio.uio_iovcnt = 1;
4004 	auio.uio_rw = UIO_READ;
4005 	auio.uio_segflg = bufseg;
4006 	auio.uio_td = td;
4007 	vn_lock(vp, LK_SHARED | LK_RETRY);
4008 	AUDIT_ARG_VNODE1(vp);
4009 	loff = auio.uio_offset = foffset;
4010 #ifdef MAC
4011 	error = mac_vnode_check_readdir(td->td_ucred, vp);
4012 	if (error == 0)
4013 #endif
4014 		error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, NULL,
4015 		    NULL);
4016 	foffset = auio.uio_offset;
4017 	if (error != 0) {
4018 		VOP_UNLOCK(vp, 0);
4019 		goto fail;
4020 	}
4021 	if (count == auio.uio_resid &&
4022 	    (vp->v_vflag & VV_ROOT) &&
4023 	    (vp->v_mount->mnt_flag & MNT_UNION)) {
4024 		struct vnode *tvp = vp;
4025 
4026 		vp = vp->v_mount->mnt_vnodecovered;
4027 		VREF(vp);
4028 		fp->f_vnode = vp;
4029 		fp->f_data = vp;
4030 		foffset = 0;
4031 		vput(tvp);
4032 		goto unionread;
4033 	}
4034 	VOP_UNLOCK(vp, 0);
4035 	*basep = loff;
4036 	if (residp != NULL)
4037 		*residp = auio.uio_resid;
4038 	td->td_retval[0] = count - auio.uio_resid;
4039 fail:
4040 	foffset_unlock(fp, foffset, 0);
4041 	fdrop(fp, td);
4042 	return (error);
4043 }
4044 
4045 /*
4046  * Set the mode mask for creation of filesystem nodes.
4047  */
4048 #ifndef _SYS_SYSPROTO_H_
4049 struct umask_args {
4050 	int	newmask;
4051 };
4052 #endif
4053 int
4054 sys_umask(struct thread *td, struct umask_args *uap)
4055 {
4056 	struct filedesc *fdp;
4057 
4058 	fdp = td->td_proc->p_fd;
4059 	FILEDESC_XLOCK(fdp);
4060 	td->td_retval[0] = fdp->fd_cmask;
4061 	fdp->fd_cmask = uap->newmask & ALLPERMS;
4062 	FILEDESC_XUNLOCK(fdp);
4063 	return (0);
4064 }
4065 
4066 /*
4067  * Void all references to file by ripping underlying filesystem away from
4068  * vnode.
4069  */
4070 #ifndef _SYS_SYSPROTO_H_
4071 struct revoke_args {
4072 	char	*path;
4073 };
4074 #endif
4075 int
4076 sys_revoke(struct thread *td, struct revoke_args *uap)
4077 {
4078 	struct vnode *vp;
4079 	struct vattr vattr;
4080 	struct nameidata nd;
4081 	int error;
4082 
4083 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
4084 	    uap->path, td);
4085 	if ((error = namei(&nd)) != 0)
4086 		return (error);
4087 	vp = nd.ni_vp;
4088 	NDFREE(&nd, NDF_ONLY_PNBUF);
4089 	if (vp->v_type != VCHR || vp->v_rdev == NULL) {
4090 		error = EINVAL;
4091 		goto out;
4092 	}
4093 #ifdef MAC
4094 	error = mac_vnode_check_revoke(td->td_ucred, vp);
4095 	if (error != 0)
4096 		goto out;
4097 #endif
4098 	error = VOP_GETATTR(vp, &vattr, td->td_ucred);
4099 	if (error != 0)
4100 		goto out;
4101 	if (td->td_ucred->cr_uid != vattr.va_uid) {
4102 		error = priv_check(td, PRIV_VFS_ADMIN);
4103 		if (error != 0)
4104 			goto out;
4105 	}
4106 	if (vcount(vp) > 1)
4107 		VOP_REVOKE(vp, REVOKEALL);
4108 out:
4109 	vput(vp);
4110 	return (error);
4111 }
4112 
4113 /*
4114  * Convert a user file descriptor to a kernel file entry and check that, if it
4115  * is a capability, the correct rights are present. A reference on the file
4116  * entry is held upon returning.
4117  */
4118 int
4119 getvnode(struct thread *td, int fd, cap_rights_t *rightsp, struct file **fpp)
4120 {
4121 	struct file *fp;
4122 	int error;
4123 
4124 	error = fget_unlocked(td->td_proc->p_fd, fd, rightsp, &fp, NULL);
4125 	if (error != 0)
4126 		return (error);
4127 
4128 	/*
4129 	 * The file could be not of the vnode type, or it may be not
4130 	 * yet fully initialized, in which case the f_vnode pointer
4131 	 * may be set, but f_ops is still badfileops.  E.g.,
4132 	 * devfs_open() transiently create such situation to
4133 	 * facilitate csw d_fdopen().
4134 	 *
4135 	 * Dupfdopen() handling in kern_openat() installs the
4136 	 * half-baked file into the process descriptor table, allowing
4137 	 * other thread to dereference it. Guard against the race by
4138 	 * checking f_ops.
4139 	 */
4140 	if (fp->f_vnode == NULL || fp->f_ops == &badfileops) {
4141 		fdrop(fp, td);
4142 		return (EINVAL);
4143 	}
4144 	*fpp = fp;
4145 	return (0);
4146 }
4147 
4148 
4149 /*
4150  * Get an (NFS) file handle.
4151  */
4152 #ifndef _SYS_SYSPROTO_H_
4153 struct lgetfh_args {
4154 	char *fname;
4155 	fhandle_t *fhp;
4156 };
4157 #endif
4158 int
4159 sys_lgetfh(struct thread *td, struct lgetfh_args *uap)
4160 {
4161 
4162 	return (kern_getfhat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->fname,
4163 	    UIO_USERSPACE, uap->fhp));
4164 }
4165 
4166 #ifndef _SYS_SYSPROTO_H_
4167 struct getfh_args {
4168 	char *fname;
4169 	fhandle_t *fhp;
4170 };
4171 #endif
4172 int
4173 sys_getfh(struct thread *td, struct getfh_args *uap)
4174 {
4175 
4176 	return (kern_getfhat(td, 0, AT_FDCWD, uap->fname, UIO_USERSPACE,
4177 	    uap->fhp));
4178 }
4179 
4180 /*
4181  * syscall for the rpc.lockd to use to translate an open descriptor into
4182  * a NFS file handle.
4183  *
4184  * warning: do not remove the priv_check() call or this becomes one giant
4185  * security hole.
4186  */
4187 #ifndef _SYS_SYSPROTO_H_
4188 struct getfhat_args {
4189 	int fd;
4190 	char *path;
4191 	fhandle_t *fhp;
4192 	int flags;
4193 };
4194 #endif
4195 int
4196 sys_getfhat(struct thread *td, struct getfhat_args *uap)
4197 {
4198 
4199 	if ((uap->flags & ~(AT_SYMLINK_NOFOLLOW | AT_BENEATH)) != 0)
4200 	    return (EINVAL);
4201 	return (kern_getfhat(td, uap->flags, uap->fd, uap->path ? uap->path : ".",
4202 	    UIO_USERSPACE, uap->fhp));
4203 }
4204 
4205 static int
4206 kern_getfhat(struct thread *td, int flags, int fd, const char *path,
4207     enum uio_seg pathseg, fhandle_t *fhp)
4208 {
4209 	struct nameidata nd;
4210 	fhandle_t fh;
4211 	struct vnode *vp;
4212 	int error;
4213 
4214 	error = priv_check(td, PRIV_VFS_GETFH);
4215 	if (error != 0)
4216 		return (error);
4217 	NDINIT_AT(&nd, LOOKUP, ((flags & AT_SYMLINK_NOFOLLOW) != 0 ? NOFOLLOW :
4218 	    FOLLOW) | ((flags & AT_BENEATH) != 0 ? BENEATH : 0) | LOCKLEAF |
4219 	    AUDITVNODE1, pathseg, path, fd, td);
4220 	error = namei(&nd);
4221 	if (error != 0)
4222 		return (error);
4223 	NDFREE(&nd, NDF_ONLY_PNBUF);
4224 	vp = nd.ni_vp;
4225 	bzero(&fh, sizeof(fh));
4226 	fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
4227 	error = VOP_VPTOFH(vp, &fh.fh_fid);
4228 	vput(vp);
4229 	if (error == 0)
4230 		error = copyout(&fh, fhp, sizeof (fh));
4231 	return (error);
4232 }
4233 
4234 #ifndef _SYS_SYSPROTO_H_
4235 struct fhlink_args {
4236 	fhandle_t *fhp;
4237 	const char *to;
4238 };
4239 #endif
4240 int
4241 sys_fhlink(struct thread *td, struct fhlink_args *uap)
4242 {
4243 
4244 	return (kern_fhlinkat(td, AT_FDCWD, uap->to, UIO_USERSPACE, uap->fhp));
4245 }
4246 
4247 #ifndef _SYS_SYSPROTO_H_
4248 struct fhlinkat_args {
4249 	fhandle_t *fhp;
4250 	int tofd;
4251 	const char *to;
4252 };
4253 #endif
4254 int
4255 sys_fhlinkat(struct thread *td, struct fhlinkat_args *uap)
4256 {
4257 
4258 	return (kern_fhlinkat(td, uap->tofd, uap->to, UIO_USERSPACE, uap->fhp));
4259 }
4260 
4261 static int
4262 kern_fhlinkat(struct thread *td, int fd, const char *path,
4263     enum uio_seg pathseg, fhandle_t *fhp)
4264 {
4265 	fhandle_t fh;
4266 	struct mount *mp;
4267 	struct vnode *vp;
4268 	int error;
4269 
4270 	error = priv_check(td, PRIV_VFS_GETFH);
4271 	if (error != 0)
4272 		return (error);
4273 	error = copyin(fhp, &fh, sizeof(fh));
4274 	if (error != 0)
4275 		return (error);
4276 	do {
4277 		bwillwrite();
4278 		if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4279 			return (ESTALE);
4280 		error = VFS_FHTOVP(mp, &fh.fh_fid, LK_SHARED, &vp);
4281 		vfs_unbusy(mp);
4282 		if (error != 0)
4283 			return (error);
4284 		VOP_UNLOCK(vp, 0);
4285 	} while ((error = kern_linkat_vp(td, vp, fd, path, pathseg)) == EAGAIN);
4286 	return (error);
4287 }
4288 
4289 #ifndef _SYS_SYSPROTO_H_
4290 struct fhreadlink_args {
4291 	fhandle_t *fhp;
4292 	char *buf;
4293 	size_t bufsize;
4294 };
4295 #endif
4296 int
4297 sys_fhreadlink(struct thread *td, struct fhreadlink_args *uap)
4298 {
4299 	fhandle_t fh;
4300 	struct mount *mp;
4301 	struct vnode *vp;
4302 	int error;
4303 
4304 	error = priv_check(td, PRIV_VFS_GETFH);
4305 	if (error != 0)
4306 		return (error);
4307 	if (uap->bufsize > IOSIZE_MAX)
4308 		return (EINVAL);
4309 	error = copyin(uap->fhp, &fh, sizeof(fh));
4310 	if (error != 0)
4311 		return (error);
4312 	if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4313 		return (ESTALE);
4314 	error = VFS_FHTOVP(mp, &fh.fh_fid, LK_SHARED, &vp);
4315 	vfs_unbusy(mp);
4316 	if (error != 0)
4317 		return (error);
4318 	error = kern_readlink_vp(vp, uap->buf, UIO_USERSPACE, uap->bufsize, td);
4319 	vput(vp);
4320 	return (error);
4321 }
4322 
4323 /*
4324  * syscall for the rpc.lockd to use to translate a NFS file handle into an
4325  * open descriptor.
4326  *
4327  * warning: do not remove the priv_check() call or this becomes one giant
4328  * security hole.
4329  */
4330 #ifndef _SYS_SYSPROTO_H_
4331 struct fhopen_args {
4332 	const struct fhandle *u_fhp;
4333 	int flags;
4334 };
4335 #endif
4336 int
4337 sys_fhopen(struct thread *td, struct fhopen_args *uap)
4338 {
4339 	struct mount *mp;
4340 	struct vnode *vp;
4341 	struct fhandle fhp;
4342 	struct file *fp;
4343 	int fmode, error;
4344 	int indx;
4345 
4346 	error = priv_check(td, PRIV_VFS_FHOPEN);
4347 	if (error != 0)
4348 		return (error);
4349 	indx = -1;
4350 	fmode = FFLAGS(uap->flags);
4351 	/* why not allow a non-read/write open for our lockd? */
4352 	if (((fmode & (FREAD | FWRITE)) == 0) || (fmode & O_CREAT))
4353 		return (EINVAL);
4354 	error = copyin(uap->u_fhp, &fhp, sizeof(fhp));
4355 	if (error != 0)
4356 		return(error);
4357 	/* find the mount point */
4358 	mp = vfs_busyfs(&fhp.fh_fsid);
4359 	if (mp == NULL)
4360 		return (ESTALE);
4361 	/* now give me my vnode, it gets returned to me locked */
4362 	error = VFS_FHTOVP(mp, &fhp.fh_fid, LK_EXCLUSIVE, &vp);
4363 	vfs_unbusy(mp);
4364 	if (error != 0)
4365 		return (error);
4366 
4367 	error = falloc_noinstall(td, &fp);
4368 	if (error != 0) {
4369 		vput(vp);
4370 		return (error);
4371 	}
4372 	/*
4373 	 * An extra reference on `fp' has been held for us by
4374 	 * falloc_noinstall().
4375 	 */
4376 
4377 #ifdef INVARIANTS
4378 	td->td_dupfd = -1;
4379 #endif
4380 	error = vn_open_vnode(vp, fmode, td->td_ucred, td, fp);
4381 	if (error != 0) {
4382 		KASSERT(fp->f_ops == &badfileops,
4383 		    ("VOP_OPEN in fhopen() set f_ops"));
4384 		KASSERT(td->td_dupfd < 0,
4385 		    ("fhopen() encountered fdopen()"));
4386 
4387 		vput(vp);
4388 		goto bad;
4389 	}
4390 #ifdef INVARIANTS
4391 	td->td_dupfd = 0;
4392 #endif
4393 	fp->f_vnode = vp;
4394 	fp->f_seqcount = 1;
4395 	finit(fp, (fmode & FMASK) | (fp->f_flag & FHASLOCK), DTYPE_VNODE, vp,
4396 	    &vnops);
4397 	VOP_UNLOCK(vp, 0);
4398 	if ((fmode & O_TRUNC) != 0) {
4399 		error = fo_truncate(fp, 0, td->td_ucred, td);
4400 		if (error != 0)
4401 			goto bad;
4402 	}
4403 
4404 	error = finstall(td, fp, &indx, fmode, NULL);
4405 bad:
4406 	fdrop(fp, td);
4407 	td->td_retval[0] = indx;
4408 	return (error);
4409 }
4410 
4411 /*
4412  * Stat an (NFS) file handle.
4413  */
4414 #ifndef _SYS_SYSPROTO_H_
4415 struct fhstat_args {
4416 	struct fhandle *u_fhp;
4417 	struct stat *sb;
4418 };
4419 #endif
4420 int
4421 sys_fhstat(struct thread *td, struct fhstat_args *uap)
4422 {
4423 	struct stat sb;
4424 	struct fhandle fh;
4425 	int error;
4426 
4427 	error = copyin(uap->u_fhp, &fh, sizeof(fh));
4428 	if (error != 0)
4429 		return (error);
4430 	error = kern_fhstat(td, fh, &sb);
4431 	if (error == 0)
4432 		error = copyout(&sb, uap->sb, sizeof(sb));
4433 	return (error);
4434 }
4435 
4436 int
4437 kern_fhstat(struct thread *td, struct fhandle fh, struct stat *sb)
4438 {
4439 	struct mount *mp;
4440 	struct vnode *vp;
4441 	int error;
4442 
4443 	error = priv_check(td, PRIV_VFS_FHSTAT);
4444 	if (error != 0)
4445 		return (error);
4446 	if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4447 		return (ESTALE);
4448 	error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4449 	vfs_unbusy(mp);
4450 	if (error != 0)
4451 		return (error);
4452 	error = vn_stat(vp, sb, td->td_ucred, NOCRED, td);
4453 	vput(vp);
4454 	return (error);
4455 }
4456 
4457 /*
4458  * Implement fstatfs() for (NFS) file handles.
4459  */
4460 #ifndef _SYS_SYSPROTO_H_
4461 struct fhstatfs_args {
4462 	struct fhandle *u_fhp;
4463 	struct statfs *buf;
4464 };
4465 #endif
4466 int
4467 sys_fhstatfs(struct thread *td, struct fhstatfs_args *uap)
4468 {
4469 	struct statfs *sfp;
4470 	fhandle_t fh;
4471 	int error;
4472 
4473 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
4474 	if (error != 0)
4475 		return (error);
4476 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
4477 	error = kern_fhstatfs(td, fh, sfp);
4478 	if (error == 0)
4479 		error = copyout(sfp, uap->buf, sizeof(*sfp));
4480 	free(sfp, M_STATFS);
4481 	return (error);
4482 }
4483 
4484 int
4485 kern_fhstatfs(struct thread *td, fhandle_t fh, struct statfs *buf)
4486 {
4487 	struct statfs *sp;
4488 	struct mount *mp;
4489 	struct vnode *vp;
4490 	int error;
4491 
4492 	error = priv_check(td, PRIV_VFS_FHSTATFS);
4493 	if (error != 0)
4494 		return (error);
4495 	if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4496 		return (ESTALE);
4497 	error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4498 	if (error != 0) {
4499 		vfs_unbusy(mp);
4500 		return (error);
4501 	}
4502 	vput(vp);
4503 	error = prison_canseemount(td->td_ucred, mp);
4504 	if (error != 0)
4505 		goto out;
4506 #ifdef MAC
4507 	error = mac_mount_check_stat(td->td_ucred, mp);
4508 	if (error != 0)
4509 		goto out;
4510 #endif
4511 	/*
4512 	 * Set these in case the underlying filesystem fails to do so.
4513 	 */
4514 	sp = &mp->mnt_stat;
4515 	sp->f_version = STATFS_VERSION;
4516 	sp->f_namemax = NAME_MAX;
4517 	sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
4518 	error = VFS_STATFS(mp, sp);
4519 	if (error == 0)
4520 		*buf = *sp;
4521 out:
4522 	vfs_unbusy(mp);
4523 	return (error);
4524 }
4525 
4526 int
4527 kern_posix_fallocate(struct thread *td, int fd, off_t offset, off_t len)
4528 {
4529 	struct file *fp;
4530 	struct mount *mp;
4531 	struct vnode *vp;
4532 	off_t olen, ooffset;
4533 	int error;
4534 #ifdef AUDIT
4535 	int audited_vnode1 = 0;
4536 #endif
4537 
4538 	AUDIT_ARG_FD(fd);
4539 	if (offset < 0 || len <= 0)
4540 		return (EINVAL);
4541 	/* Check for wrap. */
4542 	if (offset > OFF_MAX - len)
4543 		return (EFBIG);
4544 	AUDIT_ARG_FD(fd);
4545 	error = fget(td, fd, &cap_pwrite_rights, &fp);
4546 	if (error != 0)
4547 		return (error);
4548 	AUDIT_ARG_FILE(td->td_proc, fp);
4549 	if ((fp->f_ops->fo_flags & DFLAG_SEEKABLE) == 0) {
4550 		error = ESPIPE;
4551 		goto out;
4552 	}
4553 	if ((fp->f_flag & FWRITE) == 0) {
4554 		error = EBADF;
4555 		goto out;
4556 	}
4557 	if (fp->f_type != DTYPE_VNODE) {
4558 		error = ENODEV;
4559 		goto out;
4560 	}
4561 	vp = fp->f_vnode;
4562 	if (vp->v_type != VREG) {
4563 		error = ENODEV;
4564 		goto out;
4565 	}
4566 
4567 	/* Allocating blocks may take a long time, so iterate. */
4568 	for (;;) {
4569 		olen = len;
4570 		ooffset = offset;
4571 
4572 		bwillwrite();
4573 		mp = NULL;
4574 		error = vn_start_write(vp, &mp, V_WAIT | PCATCH);
4575 		if (error != 0)
4576 			break;
4577 		error = vn_lock(vp, LK_EXCLUSIVE);
4578 		if (error != 0) {
4579 			vn_finished_write(mp);
4580 			break;
4581 		}
4582 #ifdef AUDIT
4583 		if (!audited_vnode1) {
4584 			AUDIT_ARG_VNODE1(vp);
4585 			audited_vnode1 = 1;
4586 		}
4587 #endif
4588 #ifdef MAC
4589 		error = mac_vnode_check_write(td->td_ucred, fp->f_cred, vp);
4590 		if (error == 0)
4591 #endif
4592 			error = VOP_ALLOCATE(vp, &offset, &len);
4593 		VOP_UNLOCK(vp, 0);
4594 		vn_finished_write(mp);
4595 
4596 		if (olen + ooffset != offset + len) {
4597 			panic("offset + len changed from %jx/%jx to %jx/%jx",
4598 			    ooffset, olen, offset, len);
4599 		}
4600 		if (error != 0 || len == 0)
4601 			break;
4602 		KASSERT(olen > len, ("Iteration did not make progress?"));
4603 		maybe_yield();
4604 	}
4605  out:
4606 	fdrop(fp, td);
4607 	return (error);
4608 }
4609 
4610 int
4611 sys_posix_fallocate(struct thread *td, struct posix_fallocate_args *uap)
4612 {
4613 	int error;
4614 
4615 	error = kern_posix_fallocate(td, uap->fd, uap->offset, uap->len);
4616 	return (kern_posix_error(td, error));
4617 }
4618 
4619 /*
4620  * Unlike madvise(2), we do not make a best effort to remember every
4621  * possible caching hint.  Instead, we remember the last setting with
4622  * the exception that we will allow POSIX_FADV_NORMAL to adjust the
4623  * region of any current setting.
4624  */
4625 int
4626 kern_posix_fadvise(struct thread *td, int fd, off_t offset, off_t len,
4627     int advice)
4628 {
4629 	struct fadvise_info *fa, *new;
4630 	struct file *fp;
4631 	struct vnode *vp;
4632 	off_t end;
4633 	int error;
4634 
4635 	if (offset < 0 || len < 0 || offset > OFF_MAX - len)
4636 		return (EINVAL);
4637 	AUDIT_ARG_VALUE(advice);
4638 	switch (advice) {
4639 	case POSIX_FADV_SEQUENTIAL:
4640 	case POSIX_FADV_RANDOM:
4641 	case POSIX_FADV_NOREUSE:
4642 		new = malloc(sizeof(*fa), M_FADVISE, M_WAITOK);
4643 		break;
4644 	case POSIX_FADV_NORMAL:
4645 	case POSIX_FADV_WILLNEED:
4646 	case POSIX_FADV_DONTNEED:
4647 		new = NULL;
4648 		break;
4649 	default:
4650 		return (EINVAL);
4651 	}
4652 	/* XXX: CAP_POSIX_FADVISE? */
4653 	AUDIT_ARG_FD(fd);
4654 	error = fget(td, fd, &cap_no_rights, &fp);
4655 	if (error != 0)
4656 		goto out;
4657 	AUDIT_ARG_FILE(td->td_proc, fp);
4658 	if ((fp->f_ops->fo_flags & DFLAG_SEEKABLE) == 0) {
4659 		error = ESPIPE;
4660 		goto out;
4661 	}
4662 	if (fp->f_type != DTYPE_VNODE) {
4663 		error = ENODEV;
4664 		goto out;
4665 	}
4666 	vp = fp->f_vnode;
4667 	if (vp->v_type != VREG) {
4668 		error = ENODEV;
4669 		goto out;
4670 	}
4671 	if (len == 0)
4672 		end = OFF_MAX;
4673 	else
4674 		end = offset + len - 1;
4675 	switch (advice) {
4676 	case POSIX_FADV_SEQUENTIAL:
4677 	case POSIX_FADV_RANDOM:
4678 	case POSIX_FADV_NOREUSE:
4679 		/*
4680 		 * Try to merge any existing non-standard region with
4681 		 * this new region if possible, otherwise create a new
4682 		 * non-standard region for this request.
4683 		 */
4684 		mtx_pool_lock(mtxpool_sleep, fp);
4685 		fa = fp->f_advice;
4686 		if (fa != NULL && fa->fa_advice == advice &&
4687 		    ((fa->fa_start <= end && fa->fa_end >= offset) ||
4688 		    (end != OFF_MAX && fa->fa_start == end + 1) ||
4689 		    (fa->fa_end != OFF_MAX && fa->fa_end + 1 == offset))) {
4690 			if (offset < fa->fa_start)
4691 				fa->fa_start = offset;
4692 			if (end > fa->fa_end)
4693 				fa->fa_end = end;
4694 		} else {
4695 			new->fa_advice = advice;
4696 			new->fa_start = offset;
4697 			new->fa_end = end;
4698 			fp->f_advice = new;
4699 			new = fa;
4700 		}
4701 		mtx_pool_unlock(mtxpool_sleep, fp);
4702 		break;
4703 	case POSIX_FADV_NORMAL:
4704 		/*
4705 		 * If a the "normal" region overlaps with an existing
4706 		 * non-standard region, trim or remove the
4707 		 * non-standard region.
4708 		 */
4709 		mtx_pool_lock(mtxpool_sleep, fp);
4710 		fa = fp->f_advice;
4711 		if (fa != NULL) {
4712 			if (offset <= fa->fa_start && end >= fa->fa_end) {
4713 				new = fa;
4714 				fp->f_advice = NULL;
4715 			} else if (offset <= fa->fa_start &&
4716 			    end >= fa->fa_start)
4717 				fa->fa_start = end + 1;
4718 			else if (offset <= fa->fa_end && end >= fa->fa_end)
4719 				fa->fa_end = offset - 1;
4720 			else if (offset >= fa->fa_start && end <= fa->fa_end) {
4721 				/*
4722 				 * If the "normal" region is a middle
4723 				 * portion of the existing
4724 				 * non-standard region, just remove
4725 				 * the whole thing rather than picking
4726 				 * one side or the other to
4727 				 * preserve.
4728 				 */
4729 				new = fa;
4730 				fp->f_advice = NULL;
4731 			}
4732 		}
4733 		mtx_pool_unlock(mtxpool_sleep, fp);
4734 		break;
4735 	case POSIX_FADV_WILLNEED:
4736 	case POSIX_FADV_DONTNEED:
4737 		error = VOP_ADVISE(vp, offset, end, advice);
4738 		break;
4739 	}
4740 out:
4741 	if (fp != NULL)
4742 		fdrop(fp, td);
4743 	free(new, M_FADVISE);
4744 	return (error);
4745 }
4746 
4747 int
4748 sys_posix_fadvise(struct thread *td, struct posix_fadvise_args *uap)
4749 {
4750 	int error;
4751 
4752 	error = kern_posix_fadvise(td, uap->fd, uap->offset, uap->len,
4753 	    uap->advice);
4754 	return (kern_posix_error(td, error));
4755 }
4756