xref: /freebsd/sys/kern/vfs_syscalls.c (revision 6132212808e8dccedc9e5d85fea4390c2f38059a)
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 <fs/devfs/devfs.h>
91 
92 #include <ufs/ufs/quota.h>
93 
94 MALLOC_DEFINE(M_FADVISE, "fadvise", "posix_fadvise(2) information");
95 
96 static int kern_chflagsat(struct thread *td, int fd, const char *path,
97     enum uio_seg pathseg, u_long flags, int atflag);
98 static int setfflags(struct thread *td, struct vnode *, u_long);
99 static int getutimes(const struct timeval *, enum uio_seg, struct timespec *);
100 static int getutimens(const struct timespec *, enum uio_seg,
101     struct timespec *, int *);
102 static int setutimes(struct thread *td, struct vnode *,
103     const struct timespec *, int, int);
104 static int vn_access(struct vnode *vp, int user_flags, struct ucred *cred,
105     struct thread *td);
106 static int kern_fhlinkat(struct thread *td, int fd, const char *path,
107     enum uio_seg pathseg, fhandle_t *fhp);
108 static int kern_getfhat(struct thread *td, int flags, int fd,
109     const char *path, enum uio_seg pathseg, fhandle_t *fhp);
110 static int kern_readlink_vp(struct vnode *vp, char *buf, enum uio_seg bufseg,
111     size_t count, struct thread *td);
112 static int kern_linkat_vp(struct thread *td, struct vnode *vp, int fd,
113     const char *path, enum uio_seg segflag);
114 
115 int
116 kern_sync(struct thread *td)
117 {
118 	struct mount *mp, *nmp;
119 	int save;
120 
121 	mtx_lock(&mountlist_mtx);
122 	for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
123 		if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK)) {
124 			nmp = TAILQ_NEXT(mp, mnt_list);
125 			continue;
126 		}
127 		if ((mp->mnt_flag & MNT_RDONLY) == 0 &&
128 		    vn_start_write(NULL, &mp, V_NOWAIT) == 0) {
129 			save = curthread_pflags_set(TDP_SYNCIO);
130 			vfs_periodic(mp, MNT_NOWAIT);
131 			VFS_SYNC(mp, MNT_NOWAIT);
132 			curthread_pflags_restore(save);
133 			vn_finished_write(mp);
134 		}
135 		mtx_lock(&mountlist_mtx);
136 		nmp = TAILQ_NEXT(mp, mnt_list);
137 		vfs_unbusy(mp);
138 	}
139 	mtx_unlock(&mountlist_mtx);
140 	return (0);
141 }
142 
143 /*
144  * Sync each mounted filesystem.
145  */
146 #ifndef _SYS_SYSPROTO_H_
147 struct sync_args {
148 	int     dummy;
149 };
150 #endif
151 /* ARGSUSED */
152 int
153 sys_sync(struct thread *td, struct sync_args *uap)
154 {
155 
156 	return (kern_sync(td));
157 }
158 
159 /*
160  * Change filesystem quotas.
161  */
162 #ifndef _SYS_SYSPROTO_H_
163 struct quotactl_args {
164 	char *path;
165 	int cmd;
166 	int uid;
167 	caddr_t arg;
168 };
169 #endif
170 int
171 sys_quotactl(struct thread *td, struct quotactl_args *uap)
172 {
173 	struct mount *mp;
174 	struct nameidata nd;
175 	int error;
176 
177 	AUDIT_ARG_CMD(uap->cmd);
178 	AUDIT_ARG_UID(uap->uid);
179 	if (!prison_allow(td->td_ucred, PR_ALLOW_QUOTAS))
180 		return (EPERM);
181 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
182 	    uap->path, td);
183 	if ((error = namei(&nd)) != 0)
184 		return (error);
185 	NDFREE(&nd, NDF_ONLY_PNBUF);
186 	mp = nd.ni_vp->v_mount;
187 	vfs_ref(mp);
188 	vput(nd.ni_vp);
189 	error = vfs_busy(mp, 0);
190 	if (error != 0) {
191 		vfs_rel(mp);
192 		return (error);
193 	}
194 	error = VFS_QUOTACTL(mp, uap->cmd, uap->uid, uap->arg);
195 
196 	/*
197 	 * Since quota on operation typically needs to open quota
198 	 * file, the Q_QUOTAON handler needs to unbusy the mount point
199 	 * before calling into namei.  Otherwise, unmount might be
200 	 * started between two vfs_busy() invocations (first is our,
201 	 * second is from mount point cross-walk code in lookup()),
202 	 * causing deadlock.
203 	 *
204 	 * Require that Q_QUOTAON handles the vfs_busy() reference on
205 	 * its own, always returning with ubusied mount point.
206 	 */
207 	if ((uap->cmd >> SUBCMDSHIFT) != Q_QUOTAON &&
208 	    (uap->cmd >> SUBCMDSHIFT) != Q_QUOTAOFF)
209 		vfs_unbusy(mp);
210 	vfs_rel(mp);
211 	return (error);
212 }
213 
214 /*
215  * Used by statfs conversion routines to scale the block size up if
216  * necessary so that all of the block counts are <= 'max_size'.  Note
217  * that 'max_size' should be a bitmask, i.e. 2^n - 1 for some non-zero
218  * value of 'n'.
219  */
220 void
221 statfs_scale_blocks(struct statfs *sf, long max_size)
222 {
223 	uint64_t count;
224 	int shift;
225 
226 	KASSERT(powerof2(max_size + 1), ("%s: invalid max_size", __func__));
227 
228 	/*
229 	 * Attempt to scale the block counts to give a more accurate
230 	 * overview to userland of the ratio of free space to used
231 	 * space.  To do this, find the largest block count and compute
232 	 * a divisor that lets it fit into a signed integer <= max_size.
233 	 */
234 	if (sf->f_bavail < 0)
235 		count = -sf->f_bavail;
236 	else
237 		count = sf->f_bavail;
238 	count = MAX(sf->f_blocks, MAX(sf->f_bfree, count));
239 	if (count <= max_size)
240 		return;
241 
242 	count >>= flsl(max_size);
243 	shift = 0;
244 	while (count > 0) {
245 		shift++;
246 		count >>=1;
247 	}
248 
249 	sf->f_bsize <<= shift;
250 	sf->f_blocks >>= shift;
251 	sf->f_bfree >>= shift;
252 	sf->f_bavail >>= shift;
253 }
254 
255 static int
256 kern_do_statfs(struct thread *td, struct mount *mp, struct statfs *buf)
257 {
258 	int error;
259 
260 	if (mp == NULL)
261 		return (EBADF);
262 	error = vfs_busy(mp, 0);
263 	vfs_rel(mp);
264 	if (error != 0)
265 		return (error);
266 #ifdef MAC
267 	error = mac_mount_check_stat(td->td_ucred, mp);
268 	if (error != 0)
269 		goto out;
270 #endif
271 	error = VFS_STATFS(mp, buf);
272 	if (error != 0)
273 		goto out;
274 	if (priv_check_cred_vfs_generation(td->td_ucred)) {
275 		buf->f_fsid.val[0] = buf->f_fsid.val[1] = 0;
276 		prison_enforce_statfs(td->td_ucred, mp, buf);
277 	}
278 out:
279 	vfs_unbusy(mp);
280 	return (error);
281 }
282 
283 /*
284  * Get filesystem statistics.
285  */
286 #ifndef _SYS_SYSPROTO_H_
287 struct statfs_args {
288 	char *path;
289 	struct statfs *buf;
290 };
291 #endif
292 int
293 sys_statfs(struct thread *td, struct statfs_args *uap)
294 {
295 	struct statfs *sfp;
296 	int error;
297 
298 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
299 	error = kern_statfs(td, uap->path, UIO_USERSPACE, sfp);
300 	if (error == 0)
301 		error = copyout(sfp, uap->buf, sizeof(struct statfs));
302 	free(sfp, M_STATFS);
303 	return (error);
304 }
305 
306 int
307 kern_statfs(struct thread *td, const char *path, enum uio_seg pathseg,
308     struct statfs *buf)
309 {
310 	struct mount *mp;
311 	struct nameidata nd;
312 	int error;
313 
314 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
315 	    pathseg, path, td);
316 	error = namei(&nd);
317 	if (error != 0)
318 		return (error);
319 	mp = nd.ni_vp->v_mount;
320 	vfs_ref(mp);
321 	NDFREE(&nd, NDF_ONLY_PNBUF);
322 	vput(nd.ni_vp);
323 	return (kern_do_statfs(td, mp, buf));
324 }
325 
326 /*
327  * Get filesystem statistics.
328  */
329 #ifndef _SYS_SYSPROTO_H_
330 struct fstatfs_args {
331 	int fd;
332 	struct statfs *buf;
333 };
334 #endif
335 int
336 sys_fstatfs(struct thread *td, struct fstatfs_args *uap)
337 {
338 	struct statfs *sfp;
339 	int error;
340 
341 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
342 	error = kern_fstatfs(td, uap->fd, sfp);
343 	if (error == 0)
344 		error = copyout(sfp, uap->buf, sizeof(struct statfs));
345 	free(sfp, M_STATFS);
346 	return (error);
347 }
348 
349 int
350 kern_fstatfs(struct thread *td, int fd, struct statfs *buf)
351 {
352 	struct file *fp;
353 	struct mount *mp;
354 	struct vnode *vp;
355 	int error;
356 
357 	AUDIT_ARG_FD(fd);
358 	error = getvnode(td, fd, &cap_fstatfs_rights, &fp);
359 	if (error != 0)
360 		return (error);
361 	vp = fp->f_vnode;
362 	vn_lock(vp, LK_SHARED | LK_RETRY);
363 #ifdef AUDIT
364 	AUDIT_ARG_VNODE1(vp);
365 #endif
366 	mp = vp->v_mount;
367 	if (mp != NULL)
368 		vfs_ref(mp);
369 	VOP_UNLOCK(vp);
370 	fdrop(fp, td);
371 	return (kern_do_statfs(td, mp, buf));
372 }
373 
374 /*
375  * Get statistics on all filesystems.
376  */
377 #ifndef _SYS_SYSPROTO_H_
378 struct getfsstat_args {
379 	struct statfs *buf;
380 	long bufsize;
381 	int mode;
382 };
383 #endif
384 int
385 sys_getfsstat(struct thread *td, struct getfsstat_args *uap)
386 {
387 	size_t count;
388 	int error;
389 
390 	if (uap->bufsize < 0 || uap->bufsize > SIZE_MAX)
391 		return (EINVAL);
392 	error = kern_getfsstat(td, &uap->buf, uap->bufsize, &count,
393 	    UIO_USERSPACE, uap->mode);
394 	if (error == 0)
395 		td->td_retval[0] = count;
396 	return (error);
397 }
398 
399 /*
400  * If (bufsize > 0 && bufseg == UIO_SYSSPACE)
401  *	The caller is responsible for freeing memory which will be allocated
402  *	in '*buf'.
403  */
404 int
405 kern_getfsstat(struct thread *td, struct statfs **buf, size_t bufsize,
406     size_t *countp, enum uio_seg bufseg, int mode)
407 {
408 	struct mount *mp, *nmp;
409 	struct statfs *sfsp, *sp, *sptmp, *tofree;
410 	size_t count, maxcount;
411 	int error;
412 
413 	switch (mode) {
414 	case MNT_WAIT:
415 	case MNT_NOWAIT:
416 		break;
417 	default:
418 		if (bufseg == UIO_SYSSPACE)
419 			*buf = NULL;
420 		return (EINVAL);
421 	}
422 restart:
423 	maxcount = bufsize / sizeof(struct statfs);
424 	if (bufsize == 0) {
425 		sfsp = NULL;
426 		tofree = NULL;
427 	} else if (bufseg == UIO_USERSPACE) {
428 		sfsp = *buf;
429 		tofree = NULL;
430 	} else /* if (bufseg == UIO_SYSSPACE) */ {
431 		count = 0;
432 		mtx_lock(&mountlist_mtx);
433 		TAILQ_FOREACH(mp, &mountlist, mnt_list) {
434 			count++;
435 		}
436 		mtx_unlock(&mountlist_mtx);
437 		if (maxcount > count)
438 			maxcount = count;
439 		tofree = sfsp = *buf = malloc(maxcount * sizeof(struct statfs),
440 		    M_STATFS, M_WAITOK);
441 	}
442 
443 	count = 0;
444 
445 	/*
446 	 * If there is no target buffer they only want the count.
447 	 *
448 	 * This could be TAILQ_FOREACH but it is open-coded to match the original
449 	 * code below.
450 	 */
451 	if (sfsp == NULL) {
452 		mtx_lock(&mountlist_mtx);
453 		for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
454 			if (prison_canseemount(td->td_ucred, mp) != 0) {
455 				nmp = TAILQ_NEXT(mp, mnt_list);
456 				continue;
457 			}
458 #ifdef MAC
459 			if (mac_mount_check_stat(td->td_ucred, mp) != 0) {
460 				nmp = TAILQ_NEXT(mp, mnt_list);
461 				continue;
462 			}
463 #endif
464 			count++;
465 			nmp = TAILQ_NEXT(mp, mnt_list);
466 		}
467 		mtx_unlock(&mountlist_mtx);
468 		*countp = count;
469 		return (0);
470 	}
471 
472 	/*
473 	 * They want the entire thing.
474 	 *
475 	 * Short-circuit the corner case of no room for anything, avoids
476 	 * relocking below.
477 	 */
478 	if (maxcount < 1) {
479 		goto out;
480 	}
481 
482 	mtx_lock(&mountlist_mtx);
483 	for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
484 		if (prison_canseemount(td->td_ucred, mp) != 0) {
485 			nmp = TAILQ_NEXT(mp, mnt_list);
486 			continue;
487 		}
488 #ifdef MAC
489 		if (mac_mount_check_stat(td->td_ucred, mp) != 0) {
490 			nmp = TAILQ_NEXT(mp, mnt_list);
491 			continue;
492 		}
493 #endif
494 		if (mode == MNT_WAIT) {
495 			if (vfs_busy(mp, MBF_MNTLSTLOCK) != 0) {
496 				/*
497 				 * If vfs_busy() failed, and MBF_NOWAIT
498 				 * wasn't passed, then the mp is gone.
499 				 * Furthermore, because of MBF_MNTLSTLOCK,
500 				 * the mountlist_mtx was dropped.  We have
501 				 * no other choice than to start over.
502 				 */
503 				mtx_unlock(&mountlist_mtx);
504 				free(tofree, M_STATFS);
505 				goto restart;
506 			}
507 		} else {
508 			if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK) != 0) {
509 				nmp = TAILQ_NEXT(mp, mnt_list);
510 				continue;
511 			}
512 		}
513 		sp = &mp->mnt_stat;
514 		/*
515 		 * If MNT_NOWAIT is specified, do not refresh
516 		 * the fsstat cache.
517 		 */
518 		if (mode != MNT_NOWAIT) {
519 			error = VFS_STATFS(mp, sp);
520 			if (error != 0) {
521 				mtx_lock(&mountlist_mtx);
522 				nmp = TAILQ_NEXT(mp, mnt_list);
523 				vfs_unbusy(mp);
524 				continue;
525 			}
526 		}
527 		if (priv_check_cred_vfs_generation(td->td_ucred)) {
528 			sptmp = malloc(sizeof(struct statfs), M_STATFS,
529 			    M_WAITOK);
530 			*sptmp = *sp;
531 			sptmp->f_fsid.val[0] = sptmp->f_fsid.val[1] = 0;
532 			prison_enforce_statfs(td->td_ucred, mp, sptmp);
533 			sp = sptmp;
534 		} else
535 			sptmp = NULL;
536 		if (bufseg == UIO_SYSSPACE) {
537 			bcopy(sp, sfsp, sizeof(*sp));
538 			free(sptmp, M_STATFS);
539 		} else /* if (bufseg == UIO_USERSPACE) */ {
540 			error = copyout(sp, sfsp, sizeof(*sp));
541 			free(sptmp, M_STATFS);
542 			if (error != 0) {
543 				vfs_unbusy(mp);
544 				return (error);
545 			}
546 		}
547 		sfsp++;
548 		count++;
549 
550 		if (count == maxcount) {
551 			vfs_unbusy(mp);
552 			goto out;
553 		}
554 
555 		mtx_lock(&mountlist_mtx);
556 		nmp = TAILQ_NEXT(mp, mnt_list);
557 		vfs_unbusy(mp);
558 	}
559 	mtx_unlock(&mountlist_mtx);
560 out:
561 	*countp = count;
562 	return (0);
563 }
564 
565 #ifdef COMPAT_FREEBSD4
566 /*
567  * Get old format filesystem statistics.
568  */
569 static void freebsd4_cvtstatfs(struct statfs *, struct ostatfs *);
570 
571 #ifndef _SYS_SYSPROTO_H_
572 struct freebsd4_statfs_args {
573 	char *path;
574 	struct ostatfs *buf;
575 };
576 #endif
577 int
578 freebsd4_statfs(struct thread *td, struct freebsd4_statfs_args *uap)
579 {
580 	struct ostatfs osb;
581 	struct statfs *sfp;
582 	int error;
583 
584 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
585 	error = kern_statfs(td, uap->path, UIO_USERSPACE, sfp);
586 	if (error == 0) {
587 		freebsd4_cvtstatfs(sfp, &osb);
588 		error = copyout(&osb, uap->buf, sizeof(osb));
589 	}
590 	free(sfp, M_STATFS);
591 	return (error);
592 }
593 
594 /*
595  * Get filesystem statistics.
596  */
597 #ifndef _SYS_SYSPROTO_H_
598 struct freebsd4_fstatfs_args {
599 	int fd;
600 	struct ostatfs *buf;
601 };
602 #endif
603 int
604 freebsd4_fstatfs(struct thread *td, struct freebsd4_fstatfs_args *uap)
605 {
606 	struct ostatfs osb;
607 	struct statfs *sfp;
608 	int error;
609 
610 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
611 	error = kern_fstatfs(td, uap->fd, sfp);
612 	if (error == 0) {
613 		freebsd4_cvtstatfs(sfp, &osb);
614 		error = copyout(&osb, uap->buf, sizeof(osb));
615 	}
616 	free(sfp, M_STATFS);
617 	return (error);
618 }
619 
620 /*
621  * Get statistics on all filesystems.
622  */
623 #ifndef _SYS_SYSPROTO_H_
624 struct freebsd4_getfsstat_args {
625 	struct ostatfs *buf;
626 	long bufsize;
627 	int mode;
628 };
629 #endif
630 int
631 freebsd4_getfsstat(struct thread *td, struct freebsd4_getfsstat_args *uap)
632 {
633 	struct statfs *buf, *sp;
634 	struct ostatfs osb;
635 	size_t count, size;
636 	int error;
637 
638 	if (uap->bufsize < 0)
639 		return (EINVAL);
640 	count = uap->bufsize / sizeof(struct ostatfs);
641 	if (count > SIZE_MAX / sizeof(struct statfs))
642 		return (EINVAL);
643 	size = count * sizeof(struct statfs);
644 	error = kern_getfsstat(td, &buf, size, &count, UIO_SYSSPACE,
645 	    uap->mode);
646 	if (error == 0)
647 		td->td_retval[0] = count;
648 	if (size != 0) {
649 		sp = buf;
650 		while (count != 0 && error == 0) {
651 			freebsd4_cvtstatfs(sp, &osb);
652 			error = copyout(&osb, uap->buf, sizeof(osb));
653 			sp++;
654 			uap->buf++;
655 			count--;
656 		}
657 		free(buf, M_STATFS);
658 	}
659 	return (error);
660 }
661 
662 /*
663  * Implement fstatfs() for (NFS) file handles.
664  */
665 #ifndef _SYS_SYSPROTO_H_
666 struct freebsd4_fhstatfs_args {
667 	struct fhandle *u_fhp;
668 	struct ostatfs *buf;
669 };
670 #endif
671 int
672 freebsd4_fhstatfs(struct thread *td, struct freebsd4_fhstatfs_args *uap)
673 {
674 	struct ostatfs osb;
675 	struct statfs *sfp;
676 	fhandle_t fh;
677 	int error;
678 
679 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
680 	if (error != 0)
681 		return (error);
682 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
683 	error = kern_fhstatfs(td, fh, sfp);
684 	if (error == 0) {
685 		freebsd4_cvtstatfs(sfp, &osb);
686 		error = copyout(&osb, uap->buf, sizeof(osb));
687 	}
688 	free(sfp, M_STATFS);
689 	return (error);
690 }
691 
692 /*
693  * Convert a new format statfs structure to an old format statfs structure.
694  */
695 static void
696 freebsd4_cvtstatfs(struct statfs *nsp, struct ostatfs *osp)
697 {
698 
699 	statfs_scale_blocks(nsp, LONG_MAX);
700 	bzero(osp, sizeof(*osp));
701 	osp->f_bsize = nsp->f_bsize;
702 	osp->f_iosize = MIN(nsp->f_iosize, LONG_MAX);
703 	osp->f_blocks = nsp->f_blocks;
704 	osp->f_bfree = nsp->f_bfree;
705 	osp->f_bavail = nsp->f_bavail;
706 	osp->f_files = MIN(nsp->f_files, LONG_MAX);
707 	osp->f_ffree = MIN(nsp->f_ffree, LONG_MAX);
708 	osp->f_owner = nsp->f_owner;
709 	osp->f_type = nsp->f_type;
710 	osp->f_flags = nsp->f_flags;
711 	osp->f_syncwrites = MIN(nsp->f_syncwrites, LONG_MAX);
712 	osp->f_asyncwrites = MIN(nsp->f_asyncwrites, LONG_MAX);
713 	osp->f_syncreads = MIN(nsp->f_syncreads, LONG_MAX);
714 	osp->f_asyncreads = MIN(nsp->f_asyncreads, LONG_MAX);
715 	strlcpy(osp->f_fstypename, nsp->f_fstypename,
716 	    MIN(MFSNAMELEN, OMFSNAMELEN));
717 	strlcpy(osp->f_mntonname, nsp->f_mntonname,
718 	    MIN(MNAMELEN, OMNAMELEN));
719 	strlcpy(osp->f_mntfromname, nsp->f_mntfromname,
720 	    MIN(MNAMELEN, OMNAMELEN));
721 	osp->f_fsid = nsp->f_fsid;
722 }
723 #endif /* COMPAT_FREEBSD4 */
724 
725 #if defined(COMPAT_FREEBSD11)
726 /*
727  * Get old format filesystem statistics.
728  */
729 static void freebsd11_cvtstatfs(struct statfs *, struct freebsd11_statfs *);
730 
731 int
732 freebsd11_statfs(struct thread *td, struct freebsd11_statfs_args *uap)
733 {
734 	struct freebsd11_statfs osb;
735 	struct statfs *sfp;
736 	int error;
737 
738 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
739 	error = kern_statfs(td, uap->path, UIO_USERSPACE, sfp);
740 	if (error == 0) {
741 		freebsd11_cvtstatfs(sfp, &osb);
742 		error = copyout(&osb, uap->buf, sizeof(osb));
743 	}
744 	free(sfp, M_STATFS);
745 	return (error);
746 }
747 
748 /*
749  * Get filesystem statistics.
750  */
751 int
752 freebsd11_fstatfs(struct thread *td, struct freebsd11_fstatfs_args *uap)
753 {
754 	struct freebsd11_statfs osb;
755 	struct statfs *sfp;
756 	int error;
757 
758 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
759 	error = kern_fstatfs(td, uap->fd, sfp);
760 	if (error == 0) {
761 		freebsd11_cvtstatfs(sfp, &osb);
762 		error = copyout(&osb, uap->buf, sizeof(osb));
763 	}
764 	free(sfp, M_STATFS);
765 	return (error);
766 }
767 
768 /*
769  * Get statistics on all filesystems.
770  */
771 int
772 freebsd11_getfsstat(struct thread *td, struct freebsd11_getfsstat_args *uap)
773 {
774 	struct freebsd11_statfs osb;
775 	struct statfs *buf, *sp;
776 	size_t count, size;
777 	int error;
778 
779 	count = uap->bufsize / sizeof(struct ostatfs);
780 	size = count * sizeof(struct statfs);
781 	error = kern_getfsstat(td, &buf, size, &count, UIO_SYSSPACE,
782 	    uap->mode);
783 	if (error == 0)
784 		td->td_retval[0] = count;
785 	if (size > 0) {
786 		sp = buf;
787 		while (count > 0 && error == 0) {
788 			freebsd11_cvtstatfs(sp, &osb);
789 			error = copyout(&osb, uap->buf, sizeof(osb));
790 			sp++;
791 			uap->buf++;
792 			count--;
793 		}
794 		free(buf, M_STATFS);
795 	}
796 	return (error);
797 }
798 
799 /*
800  * Implement fstatfs() for (NFS) file handles.
801  */
802 int
803 freebsd11_fhstatfs(struct thread *td, struct freebsd11_fhstatfs_args *uap)
804 {
805 	struct freebsd11_statfs osb;
806 	struct statfs *sfp;
807 	fhandle_t fh;
808 	int error;
809 
810 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
811 	if (error)
812 		return (error);
813 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
814 	error = kern_fhstatfs(td, fh, sfp);
815 	if (error == 0) {
816 		freebsd11_cvtstatfs(sfp, &osb);
817 		error = copyout(&osb, uap->buf, sizeof(osb));
818 	}
819 	free(sfp, M_STATFS);
820 	return (error);
821 }
822 
823 /*
824  * Convert a new format statfs structure to an old format statfs structure.
825  */
826 static void
827 freebsd11_cvtstatfs(struct statfs *nsp, struct freebsd11_statfs *osp)
828 {
829 
830 	bzero(osp, sizeof(*osp));
831 	osp->f_version = FREEBSD11_STATFS_VERSION;
832 	osp->f_type = nsp->f_type;
833 	osp->f_flags = nsp->f_flags;
834 	osp->f_bsize = nsp->f_bsize;
835 	osp->f_iosize = nsp->f_iosize;
836 	osp->f_blocks = nsp->f_blocks;
837 	osp->f_bfree = nsp->f_bfree;
838 	osp->f_bavail = nsp->f_bavail;
839 	osp->f_files = nsp->f_files;
840 	osp->f_ffree = nsp->f_ffree;
841 	osp->f_syncwrites = nsp->f_syncwrites;
842 	osp->f_asyncwrites = nsp->f_asyncwrites;
843 	osp->f_syncreads = nsp->f_syncreads;
844 	osp->f_asyncreads = nsp->f_asyncreads;
845 	osp->f_namemax = nsp->f_namemax;
846 	osp->f_owner = nsp->f_owner;
847 	osp->f_fsid = nsp->f_fsid;
848 	strlcpy(osp->f_fstypename, nsp->f_fstypename,
849 	    MIN(MFSNAMELEN, sizeof(osp->f_fstypename)));
850 	strlcpy(osp->f_mntonname, nsp->f_mntonname,
851 	    MIN(MNAMELEN, sizeof(osp->f_mntonname)));
852 	strlcpy(osp->f_mntfromname, nsp->f_mntfromname,
853 	    MIN(MNAMELEN, sizeof(osp->f_mntfromname)));
854 }
855 #endif /* COMPAT_FREEBSD11 */
856 
857 /*
858  * Change current working directory to a given file descriptor.
859  */
860 #ifndef _SYS_SYSPROTO_H_
861 struct fchdir_args {
862 	int	fd;
863 };
864 #endif
865 int
866 sys_fchdir(struct thread *td, struct fchdir_args *uap)
867 {
868 	struct vnode *vp, *tdp;
869 	struct mount *mp;
870 	struct file *fp;
871 	int error;
872 
873 	AUDIT_ARG_FD(uap->fd);
874 	error = getvnode(td, uap->fd, &cap_fchdir_rights,
875 	    &fp);
876 	if (error != 0)
877 		return (error);
878 	vp = fp->f_vnode;
879 	vrefact(vp);
880 	fdrop(fp, td);
881 	vn_lock(vp, LK_SHARED | LK_RETRY);
882 	AUDIT_ARG_VNODE1(vp);
883 	error = change_dir(vp, td);
884 	while (!error && (mp = vp->v_mountedhere) != NULL) {
885 		if (vfs_busy(mp, 0))
886 			continue;
887 		error = VFS_ROOT(mp, LK_SHARED, &tdp);
888 		vfs_unbusy(mp);
889 		if (error != 0)
890 			break;
891 		vput(vp);
892 		vp = tdp;
893 	}
894 	if (error != 0) {
895 		vput(vp);
896 		return (error);
897 	}
898 	VOP_UNLOCK(vp);
899 	pwd_chdir(td, vp);
900 	return (0);
901 }
902 
903 /*
904  * Change current working directory (``.'').
905  */
906 #ifndef _SYS_SYSPROTO_H_
907 struct chdir_args {
908 	char	*path;
909 };
910 #endif
911 int
912 sys_chdir(struct thread *td, struct chdir_args *uap)
913 {
914 
915 	return (kern_chdir(td, uap->path, UIO_USERSPACE));
916 }
917 
918 int
919 kern_chdir(struct thread *td, const char *path, enum uio_seg pathseg)
920 {
921 	struct nameidata nd;
922 	int error;
923 
924 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
925 	    pathseg, path, td);
926 	if ((error = namei(&nd)) != 0)
927 		return (error);
928 	if ((error = change_dir(nd.ni_vp, td)) != 0) {
929 		vput(nd.ni_vp);
930 		NDFREE(&nd, NDF_ONLY_PNBUF);
931 		return (error);
932 	}
933 	VOP_UNLOCK(nd.ni_vp);
934 	NDFREE(&nd, NDF_ONLY_PNBUF);
935 	pwd_chdir(td, nd.ni_vp);
936 	return (0);
937 }
938 
939 /*
940  * Change notion of root (``/'') directory.
941  */
942 #ifndef _SYS_SYSPROTO_H_
943 struct chroot_args {
944 	char	*path;
945 };
946 #endif
947 int
948 sys_chroot(struct thread *td, struct chroot_args *uap)
949 {
950 	struct nameidata nd;
951 	int error;
952 
953 	error = priv_check(td, PRIV_VFS_CHROOT);
954 	if (error != 0)
955 		return (error);
956 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
957 	    UIO_USERSPACE, uap->path, td);
958 	error = namei(&nd);
959 	if (error != 0)
960 		goto error;
961 	error = change_dir(nd.ni_vp, td);
962 	if (error != 0)
963 		goto e_vunlock;
964 #ifdef MAC
965 	error = mac_vnode_check_chroot(td->td_ucred, nd.ni_vp);
966 	if (error != 0)
967 		goto e_vunlock;
968 #endif
969 	VOP_UNLOCK(nd.ni_vp);
970 	error = pwd_chroot(td, nd.ni_vp);
971 	vrele(nd.ni_vp);
972 	NDFREE(&nd, NDF_ONLY_PNBUF);
973 	return (error);
974 e_vunlock:
975 	vput(nd.ni_vp);
976 error:
977 	NDFREE(&nd, NDF_ONLY_PNBUF);
978 	return (error);
979 }
980 
981 /*
982  * Common routine for chroot and chdir.  Callers must provide a locked vnode
983  * instance.
984  */
985 int
986 change_dir(struct vnode *vp, struct thread *td)
987 {
988 #ifdef MAC
989 	int error;
990 #endif
991 
992 	ASSERT_VOP_LOCKED(vp, "change_dir(): vp not locked");
993 	if (vp->v_type != VDIR)
994 		return (ENOTDIR);
995 #ifdef MAC
996 	error = mac_vnode_check_chdir(td->td_ucred, vp);
997 	if (error != 0)
998 		return (error);
999 #endif
1000 	return (VOP_ACCESS(vp, VEXEC, td->td_ucred, td));
1001 }
1002 
1003 static __inline void
1004 flags_to_rights(int flags, cap_rights_t *rightsp)
1005 {
1006 
1007 	if (flags & O_EXEC) {
1008 		cap_rights_set_one(rightsp, CAP_FEXECVE);
1009 	} else {
1010 		switch ((flags & O_ACCMODE)) {
1011 		case O_RDONLY:
1012 			cap_rights_set_one(rightsp, CAP_READ);
1013 			break;
1014 		case O_RDWR:
1015 			cap_rights_set_one(rightsp, CAP_READ);
1016 			/* FALLTHROUGH */
1017 		case O_WRONLY:
1018 			cap_rights_set_one(rightsp, CAP_WRITE);
1019 			if (!(flags & (O_APPEND | O_TRUNC)))
1020 				cap_rights_set_one(rightsp, CAP_SEEK);
1021 			break;
1022 		}
1023 	}
1024 
1025 	if (flags & O_CREAT)
1026 		cap_rights_set_one(rightsp, CAP_CREATE);
1027 
1028 	if (flags & O_TRUNC)
1029 		cap_rights_set_one(rightsp, CAP_FTRUNCATE);
1030 
1031 	if (flags & (O_SYNC | O_FSYNC))
1032 		cap_rights_set_one(rightsp, CAP_FSYNC);
1033 
1034 	if (flags & (O_EXLOCK | O_SHLOCK))
1035 		cap_rights_set_one(rightsp, CAP_FLOCK);
1036 }
1037 
1038 /*
1039  * Check permissions, allocate an open file structure, and call the device
1040  * open routine if any.
1041  */
1042 #ifndef _SYS_SYSPROTO_H_
1043 struct open_args {
1044 	char	*path;
1045 	int	flags;
1046 	int	mode;
1047 };
1048 #endif
1049 int
1050 sys_open(struct thread *td, struct open_args *uap)
1051 {
1052 
1053 	return (kern_openat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1054 	    uap->flags, uap->mode));
1055 }
1056 
1057 #ifndef _SYS_SYSPROTO_H_
1058 struct openat_args {
1059 	int	fd;
1060 	char	*path;
1061 	int	flag;
1062 	int	mode;
1063 };
1064 #endif
1065 int
1066 sys_openat(struct thread *td, struct openat_args *uap)
1067 {
1068 
1069 	AUDIT_ARG_FD(uap->fd);
1070 	return (kern_openat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
1071 	    uap->mode));
1072 }
1073 
1074 int
1075 kern_openat(struct thread *td, int fd, const char *path, enum uio_seg pathseg,
1076     int flags, int mode)
1077 {
1078 	struct proc *p = td->td_proc;
1079 	struct filedesc *fdp = p->p_fd;
1080 	struct file *fp;
1081 	struct vnode *vp;
1082 	struct nameidata nd;
1083 	cap_rights_t rights;
1084 	int cmode, error, indx;
1085 
1086 	indx = -1;
1087 
1088 	AUDIT_ARG_FFLAGS(flags);
1089 	AUDIT_ARG_MODE(mode);
1090 	cap_rights_init_one(&rights, CAP_LOOKUP);
1091 	flags_to_rights(flags, &rights);
1092 	/*
1093 	 * Only one of the O_EXEC, O_RDONLY, O_WRONLY and O_RDWR flags
1094 	 * may be specified.
1095 	 */
1096 	if (flags & O_EXEC) {
1097 		if (flags & O_ACCMODE)
1098 			return (EINVAL);
1099 	} else if ((flags & O_ACCMODE) == O_ACCMODE) {
1100 		return (EINVAL);
1101 	} else {
1102 		flags = FFLAGS(flags);
1103 	}
1104 
1105 	/*
1106 	 * Allocate a file structure. The descriptor to reference it
1107 	 * is allocated and set by finstall() below.
1108 	 */
1109 	error = falloc_noinstall(td, &fp);
1110 	if (error != 0)
1111 		return (error);
1112 	/*
1113 	 * An extra reference on `fp' has been held for us by
1114 	 * falloc_noinstall().
1115 	 */
1116 	/* Set the flags early so the finit in devfs can pick them up. */
1117 	fp->f_flag = flags & FMASK;
1118 	cmode = ((mode & ~fdp->fd_cmask) & ALLPERMS) & ~S_ISTXT;
1119 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, fd,
1120 	    &rights, td);
1121 	td->td_dupfd = -1;		/* XXX check for fdopen */
1122 	error = vn_open(&nd, &flags, cmode, fp);
1123 	if (error != 0) {
1124 		/*
1125 		 * If the vn_open replaced the method vector, something
1126 		 * wonderous happened deep below and we just pass it up
1127 		 * pretending we know what we do.
1128 		 */
1129 		if (error == ENXIO && fp->f_ops != &badfileops)
1130 			goto success;
1131 
1132 		/*
1133 		 * Handle special fdopen() case. bleh.
1134 		 *
1135 		 * Don't do this for relative (capability) lookups; we don't
1136 		 * understand exactly what would happen, and we don't think
1137 		 * that it ever should.
1138 		 */
1139 		if ((nd.ni_lcf & NI_LCF_STRICTRELATIVE) == 0 &&
1140 		    (error == ENODEV || error == ENXIO) &&
1141 		    td->td_dupfd >= 0) {
1142 			error = dupfdopen(td, fdp, td->td_dupfd, flags, error,
1143 			    &indx);
1144 			if (error == 0)
1145 				goto success;
1146 		}
1147 
1148 		goto bad;
1149 	}
1150 	td->td_dupfd = 0;
1151 	NDFREE(&nd, NDF_ONLY_PNBUF);
1152 	vp = nd.ni_vp;
1153 
1154 	/*
1155 	 * Store the vnode, for any f_type. Typically, the vnode use
1156 	 * count is decremented by direct call to vn_closefile() for
1157 	 * files that switched type in the cdevsw fdopen() method.
1158 	 */
1159 	fp->f_vnode = vp;
1160 	/*
1161 	 * If the file wasn't claimed by devfs bind it to the normal
1162 	 * vnode operations here.
1163 	 */
1164 	if (fp->f_ops == &badfileops) {
1165 		KASSERT(vp->v_type != VFIFO, ("Unexpected fifo."));
1166 		finit_vnode(fp, flags, NULL, &vnops);
1167 	}
1168 
1169 	VOP_UNLOCK(vp);
1170 	if (flags & O_TRUNC) {
1171 		error = fo_truncate(fp, 0, td->td_ucred, td);
1172 		if (error != 0)
1173 			goto bad;
1174 	}
1175 success:
1176 	/*
1177 	 * If we haven't already installed the FD (for dupfdopen), do so now.
1178 	 */
1179 	if (indx == -1) {
1180 		struct filecaps *fcaps;
1181 
1182 #ifdef CAPABILITIES
1183 		if ((nd.ni_lcf & NI_LCF_STRICTRELATIVE) != 0)
1184 			fcaps = &nd.ni_filecaps;
1185 		else
1186 #endif
1187 			fcaps = NULL;
1188 		error = finstall(td, fp, &indx, flags, fcaps);
1189 		/* On success finstall() consumes fcaps. */
1190 		if (error != 0) {
1191 			filecaps_free(&nd.ni_filecaps);
1192 			goto bad;
1193 		}
1194 	} else {
1195 		filecaps_free(&nd.ni_filecaps);
1196 	}
1197 
1198 	/*
1199 	 * Release our private reference, leaving the one associated with
1200 	 * the descriptor table intact.
1201 	 */
1202 	fdrop(fp, td);
1203 	td->td_retval[0] = indx;
1204 	return (0);
1205 bad:
1206 	KASSERT(indx == -1, ("indx=%d, should be -1", indx));
1207 	fdrop(fp, td);
1208 	return (error);
1209 }
1210 
1211 #ifdef COMPAT_43
1212 /*
1213  * Create a file.
1214  */
1215 #ifndef _SYS_SYSPROTO_H_
1216 struct ocreat_args {
1217 	char	*path;
1218 	int	mode;
1219 };
1220 #endif
1221 int
1222 ocreat(struct thread *td, struct ocreat_args *uap)
1223 {
1224 
1225 	return (kern_openat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1226 	    O_WRONLY | O_CREAT | O_TRUNC, uap->mode));
1227 }
1228 #endif /* COMPAT_43 */
1229 
1230 /*
1231  * Create a special file.
1232  */
1233 #ifndef _SYS_SYSPROTO_H_
1234 struct mknodat_args {
1235 	int	fd;
1236 	char	*path;
1237 	mode_t	mode;
1238 	dev_t	dev;
1239 };
1240 #endif
1241 int
1242 sys_mknodat(struct thread *td, struct mknodat_args *uap)
1243 {
1244 
1245 	return (kern_mknodat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode,
1246 	    uap->dev));
1247 }
1248 
1249 #if defined(COMPAT_FREEBSD11)
1250 int
1251 freebsd11_mknod(struct thread *td,
1252     struct freebsd11_mknod_args *uap)
1253 {
1254 
1255 	return (kern_mknodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1256 	    uap->mode, uap->dev));
1257 }
1258 
1259 int
1260 freebsd11_mknodat(struct thread *td,
1261     struct freebsd11_mknodat_args *uap)
1262 {
1263 
1264 	return (kern_mknodat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode,
1265 	    uap->dev));
1266 }
1267 #endif /* COMPAT_FREEBSD11 */
1268 
1269 int
1270 kern_mknodat(struct thread *td, int fd, const char *path, enum uio_seg pathseg,
1271     int mode, dev_t dev)
1272 {
1273 	struct vnode *vp;
1274 	struct mount *mp;
1275 	struct vattr vattr;
1276 	struct nameidata nd;
1277 	int error, whiteout = 0;
1278 
1279 	AUDIT_ARG_MODE(mode);
1280 	AUDIT_ARG_DEV(dev);
1281 	switch (mode & S_IFMT) {
1282 	case S_IFCHR:
1283 	case S_IFBLK:
1284 		error = priv_check(td, PRIV_VFS_MKNOD_DEV);
1285 		if (error == 0 && dev == VNOVAL)
1286 			error = EINVAL;
1287 		break;
1288 	case S_IFWHT:
1289 		error = priv_check(td, PRIV_VFS_MKNOD_WHT);
1290 		break;
1291 	case S_IFIFO:
1292 		if (dev == 0)
1293 			return (kern_mkfifoat(td, fd, path, pathseg, mode));
1294 		/* FALLTHROUGH */
1295 	default:
1296 		error = EINVAL;
1297 		break;
1298 	}
1299 	if (error != 0)
1300 		return (error);
1301 restart:
1302 	bwillwrite();
1303 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
1304 	    NOCACHE, pathseg, path, fd, &cap_mknodat_rights,
1305 	    td);
1306 	if ((error = namei(&nd)) != 0)
1307 		return (error);
1308 	vp = nd.ni_vp;
1309 	if (vp != NULL) {
1310 		NDFREE(&nd, NDF_ONLY_PNBUF);
1311 		if (vp == nd.ni_dvp)
1312 			vrele(nd.ni_dvp);
1313 		else
1314 			vput(nd.ni_dvp);
1315 		vrele(vp);
1316 		return (EEXIST);
1317 	} else {
1318 		VATTR_NULL(&vattr);
1319 		vattr.va_mode = (mode & ALLPERMS) &
1320 		    ~td->td_proc->p_fd->fd_cmask;
1321 		vattr.va_rdev = dev;
1322 		whiteout = 0;
1323 
1324 		switch (mode & S_IFMT) {
1325 		case S_IFCHR:
1326 			vattr.va_type = VCHR;
1327 			break;
1328 		case S_IFBLK:
1329 			vattr.va_type = VBLK;
1330 			break;
1331 		case S_IFWHT:
1332 			whiteout = 1;
1333 			break;
1334 		default:
1335 			panic("kern_mknod: invalid mode");
1336 		}
1337 	}
1338 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1339 		NDFREE(&nd, NDF_ONLY_PNBUF);
1340 		vput(nd.ni_dvp);
1341 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1342 			return (error);
1343 		goto restart;
1344 	}
1345 #ifdef MAC
1346 	if (error == 0 && !whiteout)
1347 		error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp,
1348 		    &nd.ni_cnd, &vattr);
1349 #endif
1350 	if (error == 0) {
1351 		if (whiteout)
1352 			error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE);
1353 		else {
1354 			error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp,
1355 						&nd.ni_cnd, &vattr);
1356 			if (error == 0)
1357 				vput(nd.ni_vp);
1358 		}
1359 	}
1360 	NDFREE(&nd, NDF_ONLY_PNBUF);
1361 	vput(nd.ni_dvp);
1362 	vn_finished_write(mp);
1363 	return (error);
1364 }
1365 
1366 /*
1367  * Create a named pipe.
1368  */
1369 #ifndef _SYS_SYSPROTO_H_
1370 struct mkfifo_args {
1371 	char	*path;
1372 	int	mode;
1373 };
1374 #endif
1375 int
1376 sys_mkfifo(struct thread *td, struct mkfifo_args *uap)
1377 {
1378 
1379 	return (kern_mkfifoat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1380 	    uap->mode));
1381 }
1382 
1383 #ifndef _SYS_SYSPROTO_H_
1384 struct mkfifoat_args {
1385 	int	fd;
1386 	char	*path;
1387 	mode_t	mode;
1388 };
1389 #endif
1390 int
1391 sys_mkfifoat(struct thread *td, struct mkfifoat_args *uap)
1392 {
1393 
1394 	return (kern_mkfifoat(td, uap->fd, uap->path, UIO_USERSPACE,
1395 	    uap->mode));
1396 }
1397 
1398 int
1399 kern_mkfifoat(struct thread *td, int fd, const char *path,
1400     enum uio_seg pathseg, int mode)
1401 {
1402 	struct mount *mp;
1403 	struct vattr vattr;
1404 	struct nameidata nd;
1405 	int error;
1406 
1407 	AUDIT_ARG_MODE(mode);
1408 restart:
1409 	bwillwrite();
1410 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
1411 	    NOCACHE, pathseg, path, fd, &cap_mkfifoat_rights,
1412 	    td);
1413 	if ((error = namei(&nd)) != 0)
1414 		return (error);
1415 	if (nd.ni_vp != NULL) {
1416 		NDFREE(&nd, NDF_ONLY_PNBUF);
1417 		if (nd.ni_vp == nd.ni_dvp)
1418 			vrele(nd.ni_dvp);
1419 		else
1420 			vput(nd.ni_dvp);
1421 		vrele(nd.ni_vp);
1422 		return (EEXIST);
1423 	}
1424 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1425 		NDFREE(&nd, NDF_ONLY_PNBUF);
1426 		vput(nd.ni_dvp);
1427 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1428 			return (error);
1429 		goto restart;
1430 	}
1431 	VATTR_NULL(&vattr);
1432 	vattr.va_type = VFIFO;
1433 	vattr.va_mode = (mode & ALLPERMS) & ~td->td_proc->p_fd->fd_cmask;
1434 #ifdef MAC
1435 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1436 	    &vattr);
1437 	if (error != 0)
1438 		goto out;
1439 #endif
1440 	error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
1441 	if (error == 0)
1442 		vput(nd.ni_vp);
1443 #ifdef MAC
1444 out:
1445 #endif
1446 	vput(nd.ni_dvp);
1447 	vn_finished_write(mp);
1448 	NDFREE(&nd, NDF_ONLY_PNBUF);
1449 	return (error);
1450 }
1451 
1452 /*
1453  * Make a hard file link.
1454  */
1455 #ifndef _SYS_SYSPROTO_H_
1456 struct link_args {
1457 	char	*path;
1458 	char	*link;
1459 };
1460 #endif
1461 int
1462 sys_link(struct thread *td, struct link_args *uap)
1463 {
1464 
1465 	return (kern_linkat(td, AT_FDCWD, AT_FDCWD, uap->path, uap->link,
1466 	    UIO_USERSPACE, FOLLOW));
1467 }
1468 
1469 #ifndef _SYS_SYSPROTO_H_
1470 struct linkat_args {
1471 	int	fd1;
1472 	char	*path1;
1473 	int	fd2;
1474 	char	*path2;
1475 	int	flag;
1476 };
1477 #endif
1478 int
1479 sys_linkat(struct thread *td, struct linkat_args *uap)
1480 {
1481 	int flag;
1482 
1483 	flag = uap->flag;
1484 	if ((flag & ~(AT_SYMLINK_FOLLOW | AT_BENEATH)) != 0)
1485 		return (EINVAL);
1486 
1487 	return (kern_linkat(td, uap->fd1, uap->fd2, uap->path1, uap->path2,
1488 	    UIO_USERSPACE, ((flag & AT_SYMLINK_FOLLOW) != 0 ? FOLLOW :
1489 	    NOFOLLOW) | ((flag & AT_BENEATH) != 0 ? BENEATH : 0)));
1490 }
1491 
1492 int hardlink_check_uid = 0;
1493 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_uid, CTLFLAG_RW,
1494     &hardlink_check_uid, 0,
1495     "Unprivileged processes cannot create hard links to files owned by other "
1496     "users");
1497 static int hardlink_check_gid = 0;
1498 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_gid, CTLFLAG_RW,
1499     &hardlink_check_gid, 0,
1500     "Unprivileged processes cannot create hard links to files owned by other "
1501     "groups");
1502 
1503 static int
1504 can_hardlink(struct vnode *vp, struct ucred *cred)
1505 {
1506 	struct vattr va;
1507 	int error;
1508 
1509 	if (!hardlink_check_uid && !hardlink_check_gid)
1510 		return (0);
1511 
1512 	error = VOP_GETATTR(vp, &va, cred);
1513 	if (error != 0)
1514 		return (error);
1515 
1516 	if (hardlink_check_uid && cred->cr_uid != va.va_uid) {
1517 		error = priv_check_cred(cred, PRIV_VFS_LINK);
1518 		if (error != 0)
1519 			return (error);
1520 	}
1521 
1522 	if (hardlink_check_gid && !groupmember(va.va_gid, cred)) {
1523 		error = priv_check_cred(cred, PRIV_VFS_LINK);
1524 		if (error != 0)
1525 			return (error);
1526 	}
1527 
1528 	return (0);
1529 }
1530 
1531 int
1532 kern_linkat(struct thread *td, int fd1, int fd2, const char *path1,
1533     const char *path2, enum uio_seg segflag, int follow)
1534 {
1535 	struct nameidata nd;
1536 	int error;
1537 
1538 	do {
1539 		bwillwrite();
1540 		NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, segflag,
1541 		    path1, fd1, &cap_linkat_source_rights, td);
1542 		if ((error = namei(&nd)) != 0)
1543 			return (error);
1544 		NDFREE(&nd, NDF_ONLY_PNBUF);
1545 		error = kern_linkat_vp(td, nd.ni_vp, fd2, path2, segflag);
1546 	} while (error ==  EAGAIN);
1547 	return (error);
1548 }
1549 
1550 static int
1551 kern_linkat_vp(struct thread *td, struct vnode *vp, int fd, const char *path,
1552     enum uio_seg segflag)
1553 {
1554 	struct nameidata nd;
1555 	struct mount *mp;
1556 	int error;
1557 
1558 	if (vp->v_type == VDIR) {
1559 		vrele(vp);
1560 		return (EPERM);		/* POSIX */
1561 	}
1562 	NDINIT_ATRIGHTS(&nd, CREATE,
1563 	    LOCKPARENT | SAVENAME | AUDITVNODE2 | NOCACHE, segflag, path, fd,
1564 	    &cap_linkat_target_rights, td);
1565 	if ((error = namei(&nd)) == 0) {
1566 		if (nd.ni_vp != NULL) {
1567 			NDFREE(&nd, NDF_ONLY_PNBUF);
1568 			if (nd.ni_dvp == nd.ni_vp)
1569 				vrele(nd.ni_dvp);
1570 			else
1571 				vput(nd.ni_dvp);
1572 			vrele(nd.ni_vp);
1573 			vrele(vp);
1574 			return (EEXIST);
1575 		} else if (nd.ni_dvp->v_mount != vp->v_mount) {
1576 			/*
1577 			 * Cross-device link.  No need to recheck
1578 			 * vp->v_type, since it cannot change, except
1579 			 * to VBAD.
1580 			 */
1581 			NDFREE(&nd, NDF_ONLY_PNBUF);
1582 			vput(nd.ni_dvp);
1583 			vrele(vp);
1584 			return (EXDEV);
1585 		} else if ((error = vn_lock(vp, LK_EXCLUSIVE)) == 0) {
1586 			error = can_hardlink(vp, td->td_ucred);
1587 #ifdef MAC
1588 			if (error == 0)
1589 				error = mac_vnode_check_link(td->td_ucred,
1590 				    nd.ni_dvp, vp, &nd.ni_cnd);
1591 #endif
1592 			if (error != 0) {
1593 				vput(vp);
1594 				vput(nd.ni_dvp);
1595 				NDFREE(&nd, NDF_ONLY_PNBUF);
1596 				return (error);
1597 			}
1598 			error = vn_start_write(vp, &mp, V_NOWAIT);
1599 			if (error != 0) {
1600 				vput(vp);
1601 				vput(nd.ni_dvp);
1602 				NDFREE(&nd, NDF_ONLY_PNBUF);
1603 				error = vn_start_write(NULL, &mp,
1604 				    V_XSLEEP | PCATCH);
1605 				if (error != 0)
1606 					return (error);
1607 				return (EAGAIN);
1608 			}
1609 			error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
1610 			VOP_UNLOCK(vp);
1611 			vput(nd.ni_dvp);
1612 			vn_finished_write(mp);
1613 			NDFREE(&nd, NDF_ONLY_PNBUF);
1614 		} else {
1615 			vput(nd.ni_dvp);
1616 			NDFREE(&nd, NDF_ONLY_PNBUF);
1617 			vrele(vp);
1618 			return (EAGAIN);
1619 		}
1620 	}
1621 	vrele(vp);
1622 	return (error);
1623 }
1624 
1625 /*
1626  * Make a symbolic link.
1627  */
1628 #ifndef _SYS_SYSPROTO_H_
1629 struct symlink_args {
1630 	char	*path;
1631 	char	*link;
1632 };
1633 #endif
1634 int
1635 sys_symlink(struct thread *td, struct symlink_args *uap)
1636 {
1637 
1638 	return (kern_symlinkat(td, uap->path, AT_FDCWD, uap->link,
1639 	    UIO_USERSPACE));
1640 }
1641 
1642 #ifndef _SYS_SYSPROTO_H_
1643 struct symlinkat_args {
1644 	char	*path;
1645 	int	fd;
1646 	char	*path2;
1647 };
1648 #endif
1649 int
1650 sys_symlinkat(struct thread *td, struct symlinkat_args *uap)
1651 {
1652 
1653 	return (kern_symlinkat(td, uap->path1, uap->fd, uap->path2,
1654 	    UIO_USERSPACE));
1655 }
1656 
1657 int
1658 kern_symlinkat(struct thread *td, const char *path1, int fd, const char *path2,
1659     enum uio_seg segflg)
1660 {
1661 	struct mount *mp;
1662 	struct vattr vattr;
1663 	const char *syspath;
1664 	char *tmppath;
1665 	struct nameidata nd;
1666 	int error;
1667 
1668 	if (segflg == UIO_SYSSPACE) {
1669 		syspath = path1;
1670 	} else {
1671 		tmppath = uma_zalloc(namei_zone, M_WAITOK);
1672 		if ((error = copyinstr(path1, tmppath, MAXPATHLEN, NULL)) != 0)
1673 			goto out;
1674 		syspath = tmppath;
1675 	}
1676 	AUDIT_ARG_TEXT(syspath);
1677 restart:
1678 	bwillwrite();
1679 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
1680 	    NOCACHE, segflg, path2, fd, &cap_symlinkat_rights,
1681 	    td);
1682 	if ((error = namei(&nd)) != 0)
1683 		goto out;
1684 	if (nd.ni_vp) {
1685 		NDFREE(&nd, NDF_ONLY_PNBUF);
1686 		if (nd.ni_vp == nd.ni_dvp)
1687 			vrele(nd.ni_dvp);
1688 		else
1689 			vput(nd.ni_dvp);
1690 		vrele(nd.ni_vp);
1691 		error = EEXIST;
1692 		goto out;
1693 	}
1694 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1695 		NDFREE(&nd, NDF_ONLY_PNBUF);
1696 		vput(nd.ni_dvp);
1697 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1698 			goto out;
1699 		goto restart;
1700 	}
1701 	VATTR_NULL(&vattr);
1702 	vattr.va_mode = ACCESSPERMS &~ td->td_proc->p_fd->fd_cmask;
1703 #ifdef MAC
1704 	vattr.va_type = VLNK;
1705 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1706 	    &vattr);
1707 	if (error != 0)
1708 		goto out2;
1709 #endif
1710 	error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, syspath);
1711 	if (error == 0)
1712 		vput(nd.ni_vp);
1713 #ifdef MAC
1714 out2:
1715 #endif
1716 	NDFREE(&nd, NDF_ONLY_PNBUF);
1717 	vput(nd.ni_dvp);
1718 	vn_finished_write(mp);
1719 out:
1720 	if (segflg != UIO_SYSSPACE)
1721 		uma_zfree(namei_zone, tmppath);
1722 	return (error);
1723 }
1724 
1725 /*
1726  * Delete a whiteout from the filesystem.
1727  */
1728 #ifndef _SYS_SYSPROTO_H_
1729 struct undelete_args {
1730 	char *path;
1731 };
1732 #endif
1733 int
1734 sys_undelete(struct thread *td, struct undelete_args *uap)
1735 {
1736 	struct mount *mp;
1737 	struct nameidata nd;
1738 	int error;
1739 
1740 restart:
1741 	bwillwrite();
1742 	NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | AUDITVNODE1,
1743 	    UIO_USERSPACE, uap->path, td);
1744 	error = namei(&nd);
1745 	if (error != 0)
1746 		return (error);
1747 
1748 	if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) {
1749 		NDFREE(&nd, NDF_ONLY_PNBUF);
1750 		if (nd.ni_vp == nd.ni_dvp)
1751 			vrele(nd.ni_dvp);
1752 		else
1753 			vput(nd.ni_dvp);
1754 		if (nd.ni_vp)
1755 			vrele(nd.ni_vp);
1756 		return (EEXIST);
1757 	}
1758 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1759 		NDFREE(&nd, NDF_ONLY_PNBUF);
1760 		vput(nd.ni_dvp);
1761 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1762 			return (error);
1763 		goto restart;
1764 	}
1765 	error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE);
1766 	NDFREE(&nd, NDF_ONLY_PNBUF);
1767 	vput(nd.ni_dvp);
1768 	vn_finished_write(mp);
1769 	return (error);
1770 }
1771 
1772 /*
1773  * Delete a name from the filesystem.
1774  */
1775 #ifndef _SYS_SYSPROTO_H_
1776 struct unlink_args {
1777 	char	*path;
1778 };
1779 #endif
1780 int
1781 sys_unlink(struct thread *td, struct unlink_args *uap)
1782 {
1783 
1784 	return (kern_funlinkat(td, AT_FDCWD, uap->path, FD_NONE, UIO_USERSPACE,
1785 	    0, 0));
1786 }
1787 
1788 static int
1789 kern_funlinkat_ex(struct thread *td, int dfd, const char *path, int fd,
1790     int flag, enum uio_seg pathseg, ino_t oldinum)
1791 {
1792 
1793 	if ((flag & ~AT_REMOVEDIR) != 0)
1794 		return (EINVAL);
1795 
1796 	if ((flag & AT_REMOVEDIR) != 0)
1797 		return (kern_frmdirat(td, dfd, path, fd, UIO_USERSPACE, 0));
1798 
1799 	return (kern_funlinkat(td, dfd, path, fd, UIO_USERSPACE, 0, 0));
1800 }
1801 
1802 #ifndef _SYS_SYSPROTO_H_
1803 struct unlinkat_args {
1804 	int	fd;
1805 	char	*path;
1806 	int	flag;
1807 };
1808 #endif
1809 int
1810 sys_unlinkat(struct thread *td, struct unlinkat_args *uap)
1811 {
1812 
1813 	return (kern_funlinkat_ex(td, uap->fd, uap->path, FD_NONE, uap->flag,
1814 	    UIO_USERSPACE, 0));
1815 }
1816 
1817 #ifndef _SYS_SYSPROTO_H_
1818 struct funlinkat_args {
1819 	int		dfd;
1820 	const char	*path;
1821 	int		fd;
1822 	int		flag;
1823 };
1824 #endif
1825 int
1826 sys_funlinkat(struct thread *td, struct funlinkat_args *uap)
1827 {
1828 
1829 	return (kern_funlinkat_ex(td, uap->dfd, uap->path, uap->fd, uap->flag,
1830 	    UIO_USERSPACE, 0));
1831 }
1832 
1833 int
1834 kern_funlinkat(struct thread *td, int dfd, const char *path, int fd,
1835     enum uio_seg pathseg, int flag, ino_t oldinum)
1836 {
1837 	struct mount *mp;
1838 	struct file *fp;
1839 	struct vnode *vp;
1840 	struct nameidata nd;
1841 	struct stat sb;
1842 	int error;
1843 
1844 	fp = NULL;
1845 	if (fd != FD_NONE) {
1846 		error = getvnode(td, fd, &cap_no_rights, &fp);
1847 		if (error != 0)
1848 			return (error);
1849 	}
1850 
1851 restart:
1852 	bwillwrite();
1853 	NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1 |
1854 	    ((flag & AT_BENEATH) != 0 ? BENEATH : 0),
1855 	    pathseg, path, dfd, &cap_unlinkat_rights, td);
1856 	if ((error = namei(&nd)) != 0) {
1857 		if (error == EINVAL)
1858 			error = EPERM;
1859 		goto fdout;
1860 	}
1861 	vp = nd.ni_vp;
1862 	if (vp->v_type == VDIR && oldinum == 0) {
1863 		error = EPERM;		/* POSIX */
1864 	} else if (oldinum != 0 &&
1865 		  ((error = VOP_STAT(vp, &sb, td->td_ucred, NOCRED, td)) == 0) &&
1866 		  sb.st_ino != oldinum) {
1867 		error = EIDRM;	/* Identifier removed */
1868 	} else if (fp != NULL && fp->f_vnode != vp) {
1869 		if (VN_IS_DOOMED(fp->f_vnode))
1870 			error = EBADF;
1871 		else
1872 			error = EDEADLK;
1873 	} else {
1874 		/*
1875 		 * The root of a mounted filesystem cannot be deleted.
1876 		 *
1877 		 * XXX: can this only be a VDIR case?
1878 		 */
1879 		if (vp->v_vflag & VV_ROOT)
1880 			error = EBUSY;
1881 	}
1882 	if (error == 0) {
1883 		if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1884 			NDFREE(&nd, NDF_ONLY_PNBUF);
1885 			vput(nd.ni_dvp);
1886 			if (vp == nd.ni_dvp)
1887 				vrele(vp);
1888 			else
1889 				vput(vp);
1890 			if ((error = vn_start_write(NULL, &mp,
1891 			    V_XSLEEP | PCATCH)) != 0) {
1892 				goto fdout;
1893 			}
1894 			goto restart;
1895 		}
1896 #ifdef MAC
1897 		error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
1898 		    &nd.ni_cnd);
1899 		if (error != 0)
1900 			goto out;
1901 #endif
1902 		vfs_notify_upper(vp, VFS_NOTIFY_UPPER_UNLINK);
1903 		error = VOP_REMOVE(nd.ni_dvp, vp, &nd.ni_cnd);
1904 #ifdef MAC
1905 out:
1906 #endif
1907 		vn_finished_write(mp);
1908 	}
1909 	NDFREE(&nd, NDF_ONLY_PNBUF);
1910 	vput(nd.ni_dvp);
1911 	if (vp == nd.ni_dvp)
1912 		vrele(vp);
1913 	else
1914 		vput(vp);
1915 fdout:
1916 	if (fp != NULL)
1917 		fdrop(fp, td);
1918 	return (error);
1919 }
1920 
1921 /*
1922  * Reposition read/write file offset.
1923  */
1924 #ifndef _SYS_SYSPROTO_H_
1925 struct lseek_args {
1926 	int	fd;
1927 	int	pad;
1928 	off_t	offset;
1929 	int	whence;
1930 };
1931 #endif
1932 int
1933 sys_lseek(struct thread *td, struct lseek_args *uap)
1934 {
1935 
1936 	return (kern_lseek(td, uap->fd, uap->offset, uap->whence));
1937 }
1938 
1939 int
1940 kern_lseek(struct thread *td, int fd, off_t offset, int whence)
1941 {
1942 	struct file *fp;
1943 	int error;
1944 
1945 	AUDIT_ARG_FD(fd);
1946 	error = fget(td, fd, &cap_seek_rights, &fp);
1947 	if (error != 0)
1948 		return (error);
1949 	error = (fp->f_ops->fo_flags & DFLAG_SEEKABLE) != 0 ?
1950 	    fo_seek(fp, offset, whence, td) : ESPIPE;
1951 	fdrop(fp, td);
1952 	return (error);
1953 }
1954 
1955 #if defined(COMPAT_43)
1956 /*
1957  * Reposition read/write file offset.
1958  */
1959 #ifndef _SYS_SYSPROTO_H_
1960 struct olseek_args {
1961 	int	fd;
1962 	long	offset;
1963 	int	whence;
1964 };
1965 #endif
1966 int
1967 olseek(struct thread *td, struct olseek_args *uap)
1968 {
1969 
1970 	return (kern_lseek(td, uap->fd, uap->offset, uap->whence));
1971 }
1972 #endif /* COMPAT_43 */
1973 
1974 #if defined(COMPAT_FREEBSD6)
1975 /* Version with the 'pad' argument */
1976 int
1977 freebsd6_lseek(struct thread *td, struct freebsd6_lseek_args *uap)
1978 {
1979 
1980 	return (kern_lseek(td, uap->fd, uap->offset, uap->whence));
1981 }
1982 #endif
1983 
1984 /*
1985  * Check access permissions using passed credentials.
1986  */
1987 static int
1988 vn_access(struct vnode *vp, int user_flags, struct ucred *cred,
1989      struct thread *td)
1990 {
1991 	accmode_t accmode;
1992 	int error;
1993 
1994 	/* Flags == 0 means only check for existence. */
1995 	if (user_flags == 0)
1996 		return (0);
1997 
1998 	accmode = 0;
1999 	if (user_flags & R_OK)
2000 		accmode |= VREAD;
2001 	if (user_flags & W_OK)
2002 		accmode |= VWRITE;
2003 	if (user_flags & X_OK)
2004 		accmode |= VEXEC;
2005 #ifdef MAC
2006 	error = mac_vnode_check_access(cred, vp, accmode);
2007 	if (error != 0)
2008 		return (error);
2009 #endif
2010 	if ((accmode & VWRITE) == 0 || (error = vn_writechk(vp)) == 0)
2011 		error = VOP_ACCESS(vp, accmode, cred, td);
2012 	return (error);
2013 }
2014 
2015 /*
2016  * Check access permissions using "real" credentials.
2017  */
2018 #ifndef _SYS_SYSPROTO_H_
2019 struct access_args {
2020 	char	*path;
2021 	int	amode;
2022 };
2023 #endif
2024 int
2025 sys_access(struct thread *td, struct access_args *uap)
2026 {
2027 
2028 	return (kern_accessat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2029 	    0, uap->amode));
2030 }
2031 
2032 #ifndef _SYS_SYSPROTO_H_
2033 struct faccessat_args {
2034 	int	dirfd;
2035 	char	*path;
2036 	int	amode;
2037 	int	flag;
2038 }
2039 #endif
2040 int
2041 sys_faccessat(struct thread *td, struct faccessat_args *uap)
2042 {
2043 
2044 	return (kern_accessat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
2045 	    uap->amode));
2046 }
2047 
2048 int
2049 kern_accessat(struct thread *td, int fd, const char *path,
2050     enum uio_seg pathseg, int flag, int amode)
2051 {
2052 	struct ucred *cred, *usecred;
2053 	struct vnode *vp;
2054 	struct nameidata nd;
2055 	int error;
2056 
2057 	if ((flag & ~(AT_EACCESS | AT_BENEATH)) != 0)
2058 		return (EINVAL);
2059 	if (amode != F_OK && (amode & ~(R_OK | W_OK | X_OK)) != 0)
2060 		return (EINVAL);
2061 
2062 	/*
2063 	 * Create and modify a temporary credential instead of one that
2064 	 * is potentially shared (if we need one).
2065 	 */
2066 	cred = td->td_ucred;
2067 	if ((flag & AT_EACCESS) == 0 &&
2068 	    ((cred->cr_uid != cred->cr_ruid ||
2069 	    cred->cr_rgid != cred->cr_groups[0]))) {
2070 		usecred = crdup(cred);
2071 		usecred->cr_uid = cred->cr_ruid;
2072 		usecred->cr_groups[0] = cred->cr_rgid;
2073 		td->td_ucred = usecred;
2074 	} else
2075 		usecred = cred;
2076 	AUDIT_ARG_VALUE(amode);
2077 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF |
2078 	    AUDITVNODE1 | ((flag & AT_BENEATH) != 0 ? BENEATH : 0),
2079 	    pathseg, path, fd, &cap_fstat_rights, td);
2080 	if ((error = namei(&nd)) != 0)
2081 		goto out;
2082 	vp = nd.ni_vp;
2083 
2084 	error = vn_access(vp, amode, usecred, td);
2085 	NDFREE(&nd, NDF_ONLY_PNBUF);
2086 	vput(vp);
2087 out:
2088 	if (usecred != cred) {
2089 		td->td_ucred = cred;
2090 		crfree(usecred);
2091 	}
2092 	return (error);
2093 }
2094 
2095 /*
2096  * Check access permissions using "effective" credentials.
2097  */
2098 #ifndef _SYS_SYSPROTO_H_
2099 struct eaccess_args {
2100 	char	*path;
2101 	int	amode;
2102 };
2103 #endif
2104 int
2105 sys_eaccess(struct thread *td, struct eaccess_args *uap)
2106 {
2107 
2108 	return (kern_accessat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2109 	    AT_EACCESS, uap->amode));
2110 }
2111 
2112 #if defined(COMPAT_43)
2113 /*
2114  * Get file status; this version follows links.
2115  */
2116 #ifndef _SYS_SYSPROTO_H_
2117 struct ostat_args {
2118 	char	*path;
2119 	struct ostat *ub;
2120 };
2121 #endif
2122 int
2123 ostat(struct thread *td, struct ostat_args *uap)
2124 {
2125 	struct stat sb;
2126 	struct ostat osb;
2127 	int error;
2128 
2129 	error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE,
2130 	    &sb, NULL);
2131 	if (error != 0)
2132 		return (error);
2133 	cvtstat(&sb, &osb);
2134 	return (copyout(&osb, uap->ub, sizeof (osb)));
2135 }
2136 
2137 /*
2138  * Get file status; this version does not follow links.
2139  */
2140 #ifndef _SYS_SYSPROTO_H_
2141 struct olstat_args {
2142 	char	*path;
2143 	struct ostat *ub;
2144 };
2145 #endif
2146 int
2147 olstat(struct thread *td, struct olstat_args *uap)
2148 {
2149 	struct stat sb;
2150 	struct ostat osb;
2151 	int error;
2152 
2153 	error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2154 	    UIO_USERSPACE, &sb, NULL);
2155 	if (error != 0)
2156 		return (error);
2157 	cvtstat(&sb, &osb);
2158 	return (copyout(&osb, uap->ub, sizeof (osb)));
2159 }
2160 
2161 /*
2162  * Convert from an old to a new stat structure.
2163  * XXX: many values are blindly truncated.
2164  */
2165 void
2166 cvtstat(struct stat *st, struct ostat *ost)
2167 {
2168 
2169 	bzero(ost, sizeof(*ost));
2170 	ost->st_dev = st->st_dev;
2171 	ost->st_ino = st->st_ino;
2172 	ost->st_mode = st->st_mode;
2173 	ost->st_nlink = st->st_nlink;
2174 	ost->st_uid = st->st_uid;
2175 	ost->st_gid = st->st_gid;
2176 	ost->st_rdev = st->st_rdev;
2177 	ost->st_size = MIN(st->st_size, INT32_MAX);
2178 	ost->st_atim = st->st_atim;
2179 	ost->st_mtim = st->st_mtim;
2180 	ost->st_ctim = st->st_ctim;
2181 	ost->st_blksize = st->st_blksize;
2182 	ost->st_blocks = st->st_blocks;
2183 	ost->st_flags = st->st_flags;
2184 	ost->st_gen = st->st_gen;
2185 }
2186 #endif /* COMPAT_43 */
2187 
2188 #if defined(COMPAT_43) || defined(COMPAT_FREEBSD11)
2189 int ino64_trunc_error;
2190 SYSCTL_INT(_vfs, OID_AUTO, ino64_trunc_error, CTLFLAG_RW,
2191     &ino64_trunc_error, 0,
2192     "Error on truncation of device, file or inode number, or link count");
2193 
2194 int
2195 freebsd11_cvtstat(struct stat *st, struct freebsd11_stat *ost)
2196 {
2197 
2198 	ost->st_dev = st->st_dev;
2199 	if (ost->st_dev != st->st_dev) {
2200 		switch (ino64_trunc_error) {
2201 		default:
2202 			/*
2203 			 * Since dev_t is almost raw, don't clamp to the
2204 			 * maximum for case 2, but ignore the error.
2205 			 */
2206 			break;
2207 		case 1:
2208 			return (EOVERFLOW);
2209 		}
2210 	}
2211 	ost->st_ino = st->st_ino;
2212 	if (ost->st_ino != st->st_ino) {
2213 		switch (ino64_trunc_error) {
2214 		default:
2215 		case 0:
2216 			break;
2217 		case 1:
2218 			return (EOVERFLOW);
2219 		case 2:
2220 			ost->st_ino = UINT32_MAX;
2221 			break;
2222 		}
2223 	}
2224 	ost->st_mode = st->st_mode;
2225 	ost->st_nlink = st->st_nlink;
2226 	if (ost->st_nlink != st->st_nlink) {
2227 		switch (ino64_trunc_error) {
2228 		default:
2229 		case 0:
2230 			break;
2231 		case 1:
2232 			return (EOVERFLOW);
2233 		case 2:
2234 			ost->st_nlink = UINT16_MAX;
2235 			break;
2236 		}
2237 	}
2238 	ost->st_uid = st->st_uid;
2239 	ost->st_gid = st->st_gid;
2240 	ost->st_rdev = st->st_rdev;
2241 	if (ost->st_rdev != st->st_rdev) {
2242 		switch (ino64_trunc_error) {
2243 		default:
2244 			break;
2245 		case 1:
2246 			return (EOVERFLOW);
2247 		}
2248 	}
2249 	ost->st_atim = st->st_atim;
2250 	ost->st_mtim = st->st_mtim;
2251 	ost->st_ctim = st->st_ctim;
2252 	ost->st_size = st->st_size;
2253 	ost->st_blocks = st->st_blocks;
2254 	ost->st_blksize = st->st_blksize;
2255 	ost->st_flags = st->st_flags;
2256 	ost->st_gen = st->st_gen;
2257 	ost->st_lspare = 0;
2258 	ost->st_birthtim = st->st_birthtim;
2259 	bzero((char *)&ost->st_birthtim + sizeof(ost->st_birthtim),
2260 	    sizeof(*ost) - offsetof(struct freebsd11_stat,
2261 	    st_birthtim) - sizeof(ost->st_birthtim));
2262 	return (0);
2263 }
2264 
2265 int
2266 freebsd11_stat(struct thread *td, struct freebsd11_stat_args* uap)
2267 {
2268 	struct stat sb;
2269 	struct freebsd11_stat osb;
2270 	int error;
2271 
2272 	error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE,
2273 	    &sb, NULL);
2274 	if (error != 0)
2275 		return (error);
2276 	error = freebsd11_cvtstat(&sb, &osb);
2277 	if (error == 0)
2278 		error = copyout(&osb, uap->ub, sizeof(osb));
2279 	return (error);
2280 }
2281 
2282 int
2283 freebsd11_lstat(struct thread *td, struct freebsd11_lstat_args* uap)
2284 {
2285 	struct stat sb;
2286 	struct freebsd11_stat osb;
2287 	int error;
2288 
2289 	error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2290 	    UIO_USERSPACE, &sb, NULL);
2291 	if (error != 0)
2292 		return (error);
2293 	error = freebsd11_cvtstat(&sb, &osb);
2294 	if (error == 0)
2295 		error = copyout(&osb, uap->ub, sizeof(osb));
2296 	return (error);
2297 }
2298 
2299 int
2300 freebsd11_fhstat(struct thread *td, struct freebsd11_fhstat_args* uap)
2301 {
2302 	struct fhandle fh;
2303 	struct stat sb;
2304 	struct freebsd11_stat osb;
2305 	int error;
2306 
2307 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
2308 	if (error != 0)
2309 		return (error);
2310 	error = kern_fhstat(td, fh, &sb);
2311 	if (error != 0)
2312 		return (error);
2313 	error = freebsd11_cvtstat(&sb, &osb);
2314 	if (error == 0)
2315 		error = copyout(&osb, uap->sb, sizeof(osb));
2316 	return (error);
2317 }
2318 
2319 int
2320 freebsd11_fstatat(struct thread *td, struct freebsd11_fstatat_args* uap)
2321 {
2322 	struct stat sb;
2323 	struct freebsd11_stat osb;
2324 	int error;
2325 
2326 	error = kern_statat(td, uap->flag, uap->fd, uap->path,
2327 	    UIO_USERSPACE, &sb, NULL);
2328 	if (error != 0)
2329 		return (error);
2330 	error = freebsd11_cvtstat(&sb, &osb);
2331 	if (error == 0)
2332 		error = copyout(&osb, uap->buf, sizeof(osb));
2333 	return (error);
2334 }
2335 #endif	/* COMPAT_FREEBSD11 */
2336 
2337 /*
2338  * Get file status
2339  */
2340 #ifndef _SYS_SYSPROTO_H_
2341 struct fstatat_args {
2342 	int	fd;
2343 	char	*path;
2344 	struct stat	*buf;
2345 	int	flag;
2346 }
2347 #endif
2348 int
2349 sys_fstatat(struct thread *td, struct fstatat_args *uap)
2350 {
2351 	struct stat sb;
2352 	int error;
2353 
2354 	error = kern_statat(td, uap->flag, uap->fd, uap->path,
2355 	    UIO_USERSPACE, &sb, NULL);
2356 	if (error == 0)
2357 		error = copyout(&sb, uap->buf, sizeof (sb));
2358 	return (error);
2359 }
2360 
2361 int
2362 kern_statat(struct thread *td, int flag, int fd, const char *path,
2363     enum uio_seg pathseg, struct stat *sbp,
2364     void (*hook)(struct vnode *vp, struct stat *sbp))
2365 {
2366 	struct nameidata nd;
2367 	int error;
2368 
2369 	if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_BENEATH)) != 0)
2370 		return (EINVAL);
2371 
2372 	NDINIT_ATRIGHTS(&nd, LOOKUP, ((flag & AT_SYMLINK_NOFOLLOW) != 0 ?
2373 	    NOFOLLOW : FOLLOW) | ((flag & AT_BENEATH) != 0 ? BENEATH : 0) |
2374 	    LOCKSHARED | LOCKLEAF | AUDITVNODE1, pathseg, path, fd,
2375 	    &cap_fstat_rights, td);
2376 
2377 	if ((error = namei(&nd)) != 0)
2378 		return (error);
2379 	error = VOP_STAT(nd.ni_vp, sbp, td->td_ucred, NOCRED, td);
2380 	if (error == 0) {
2381 		if (__predict_false(hook != NULL))
2382 			hook(nd.ni_vp, sbp);
2383 	}
2384 	NDFREE(&nd, NDF_ONLY_PNBUF);
2385 	vput(nd.ni_vp);
2386 #ifdef __STAT_TIME_T_EXT
2387 	sbp->st_atim_ext = 0;
2388 	sbp->st_mtim_ext = 0;
2389 	sbp->st_ctim_ext = 0;
2390 	sbp->st_btim_ext = 0;
2391 #endif
2392 #ifdef KTRACE
2393 	if (KTRPOINT(td, KTR_STRUCT))
2394 		ktrstat_error(sbp, error);
2395 #endif
2396 	return (error);
2397 }
2398 
2399 #if defined(COMPAT_FREEBSD11)
2400 /*
2401  * Implementation of the NetBSD [l]stat() functions.
2402  */
2403 void
2404 freebsd11_cvtnstat(struct stat *sb, struct nstat *nsb)
2405 {
2406 
2407 	bzero(nsb, sizeof(*nsb));
2408 	nsb->st_dev = sb->st_dev;
2409 	nsb->st_ino = sb->st_ino;
2410 	nsb->st_mode = sb->st_mode;
2411 	nsb->st_nlink = sb->st_nlink;
2412 	nsb->st_uid = sb->st_uid;
2413 	nsb->st_gid = sb->st_gid;
2414 	nsb->st_rdev = sb->st_rdev;
2415 	nsb->st_atim = sb->st_atim;
2416 	nsb->st_mtim = sb->st_mtim;
2417 	nsb->st_ctim = sb->st_ctim;
2418 	nsb->st_size = sb->st_size;
2419 	nsb->st_blocks = sb->st_blocks;
2420 	nsb->st_blksize = sb->st_blksize;
2421 	nsb->st_flags = sb->st_flags;
2422 	nsb->st_gen = sb->st_gen;
2423 	nsb->st_birthtim = sb->st_birthtim;
2424 }
2425 
2426 #ifndef _SYS_SYSPROTO_H_
2427 struct freebsd11_nstat_args {
2428 	char	*path;
2429 	struct nstat *ub;
2430 };
2431 #endif
2432 int
2433 freebsd11_nstat(struct thread *td, struct freebsd11_nstat_args *uap)
2434 {
2435 	struct stat sb;
2436 	struct nstat nsb;
2437 	int error;
2438 
2439 	error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE,
2440 	    &sb, NULL);
2441 	if (error != 0)
2442 		return (error);
2443 	freebsd11_cvtnstat(&sb, &nsb);
2444 	return (copyout(&nsb, uap->ub, sizeof (nsb)));
2445 }
2446 
2447 /*
2448  * NetBSD lstat.  Get file status; this version does not follow links.
2449  */
2450 #ifndef _SYS_SYSPROTO_H_
2451 struct freebsd11_nlstat_args {
2452 	char	*path;
2453 	struct nstat *ub;
2454 };
2455 #endif
2456 int
2457 freebsd11_nlstat(struct thread *td, struct freebsd11_nlstat_args *uap)
2458 {
2459 	struct stat sb;
2460 	struct nstat nsb;
2461 	int error;
2462 
2463 	error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2464 	    UIO_USERSPACE, &sb, NULL);
2465 	if (error != 0)
2466 		return (error);
2467 	freebsd11_cvtnstat(&sb, &nsb);
2468 	return (copyout(&nsb, uap->ub, sizeof (nsb)));
2469 }
2470 #endif /* COMPAT_FREEBSD11 */
2471 
2472 /*
2473  * Get configurable pathname variables.
2474  */
2475 #ifndef _SYS_SYSPROTO_H_
2476 struct pathconf_args {
2477 	char	*path;
2478 	int	name;
2479 };
2480 #endif
2481 int
2482 sys_pathconf(struct thread *td, struct pathconf_args *uap)
2483 {
2484 	long value;
2485 	int error;
2486 
2487 	error = kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name, FOLLOW,
2488 	    &value);
2489 	if (error == 0)
2490 		td->td_retval[0] = value;
2491 	return (error);
2492 }
2493 
2494 #ifndef _SYS_SYSPROTO_H_
2495 struct lpathconf_args {
2496 	char	*path;
2497 	int	name;
2498 };
2499 #endif
2500 int
2501 sys_lpathconf(struct thread *td, struct lpathconf_args *uap)
2502 {
2503 	long value;
2504 	int error;
2505 
2506 	error = kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name,
2507 	    NOFOLLOW, &value);
2508 	if (error == 0)
2509 		td->td_retval[0] = value;
2510 	return (error);
2511 }
2512 
2513 int
2514 kern_pathconf(struct thread *td, const char *path, enum uio_seg pathseg,
2515     int name, u_long flags, long *valuep)
2516 {
2517 	struct nameidata nd;
2518 	int error;
2519 
2520 	NDINIT(&nd, LOOKUP, LOCKSHARED | LOCKLEAF | AUDITVNODE1 | flags,
2521 	    pathseg, path, td);
2522 	if ((error = namei(&nd)) != 0)
2523 		return (error);
2524 	NDFREE(&nd, NDF_ONLY_PNBUF);
2525 
2526 	error = VOP_PATHCONF(nd.ni_vp, name, valuep);
2527 	vput(nd.ni_vp);
2528 	return (error);
2529 }
2530 
2531 /*
2532  * Return target name of a symbolic link.
2533  */
2534 #ifndef _SYS_SYSPROTO_H_
2535 struct readlink_args {
2536 	char	*path;
2537 	char	*buf;
2538 	size_t	count;
2539 };
2540 #endif
2541 int
2542 sys_readlink(struct thread *td, struct readlink_args *uap)
2543 {
2544 
2545 	return (kern_readlinkat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2546 	    uap->buf, UIO_USERSPACE, uap->count));
2547 }
2548 #ifndef _SYS_SYSPROTO_H_
2549 struct readlinkat_args {
2550 	int	fd;
2551 	char	*path;
2552 	char	*buf;
2553 	size_t	bufsize;
2554 };
2555 #endif
2556 int
2557 sys_readlinkat(struct thread *td, struct readlinkat_args *uap)
2558 {
2559 
2560 	return (kern_readlinkat(td, uap->fd, uap->path, UIO_USERSPACE,
2561 	    uap->buf, UIO_USERSPACE, uap->bufsize));
2562 }
2563 
2564 int
2565 kern_readlinkat(struct thread *td, int fd, const char *path,
2566     enum uio_seg pathseg, char *buf, enum uio_seg bufseg, size_t count)
2567 {
2568 	struct vnode *vp;
2569 	struct nameidata nd;
2570 	int error;
2571 
2572 	if (count > IOSIZE_MAX)
2573 		return (EINVAL);
2574 
2575 	NDINIT_AT(&nd, LOOKUP, NOFOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
2576 	    pathseg, path, fd, td);
2577 
2578 	if ((error = namei(&nd)) != 0)
2579 		return (error);
2580 	NDFREE(&nd, NDF_ONLY_PNBUF);
2581 	vp = nd.ni_vp;
2582 
2583 	error = kern_readlink_vp(vp, buf, bufseg, count, td);
2584 	vput(vp);
2585 
2586 	return (error);
2587 }
2588 
2589 /*
2590  * Helper function to readlink from a vnode
2591  */
2592 static int
2593 kern_readlink_vp(struct vnode *vp, char *buf, enum uio_seg bufseg, size_t count,
2594     struct thread *td)
2595 {
2596 	struct iovec aiov;
2597 	struct uio auio;
2598 	int error;
2599 
2600 	ASSERT_VOP_LOCKED(vp, "kern_readlink_vp(): vp not locked");
2601 #ifdef MAC
2602 	error = mac_vnode_check_readlink(td->td_ucred, vp);
2603 	if (error != 0)
2604 		return (error);
2605 #endif
2606 	if (vp->v_type != VLNK && (vp->v_vflag & VV_READLINK) == 0)
2607 		return (EINVAL);
2608 
2609 	aiov.iov_base = buf;
2610 	aiov.iov_len = count;
2611 	auio.uio_iov = &aiov;
2612 	auio.uio_iovcnt = 1;
2613 	auio.uio_offset = 0;
2614 	auio.uio_rw = UIO_READ;
2615 	auio.uio_segflg = bufseg;
2616 	auio.uio_td = td;
2617 	auio.uio_resid = count;
2618 	error = VOP_READLINK(vp, &auio, td->td_ucred);
2619 	td->td_retval[0] = count - auio.uio_resid;
2620 	return (error);
2621 }
2622 
2623 /*
2624  * Common implementation code for chflags() and fchflags().
2625  */
2626 static int
2627 setfflags(struct thread *td, struct vnode *vp, u_long flags)
2628 {
2629 	struct mount *mp;
2630 	struct vattr vattr;
2631 	int error;
2632 
2633 	/* We can't support the value matching VNOVAL. */
2634 	if (flags == VNOVAL)
2635 		return (EOPNOTSUPP);
2636 
2637 	/*
2638 	 * Prevent non-root users from setting flags on devices.  When
2639 	 * a device is reused, users can retain ownership of the device
2640 	 * if they are allowed to set flags and programs assume that
2641 	 * chown can't fail when done as root.
2642 	 */
2643 	if (vp->v_type == VCHR || vp->v_type == VBLK) {
2644 		error = priv_check(td, PRIV_VFS_CHFLAGS_DEV);
2645 		if (error != 0)
2646 			return (error);
2647 	}
2648 
2649 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2650 		return (error);
2651 	VATTR_NULL(&vattr);
2652 	vattr.va_flags = flags;
2653 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2654 #ifdef MAC
2655 	error = mac_vnode_check_setflags(td->td_ucred, vp, vattr.va_flags);
2656 	if (error == 0)
2657 #endif
2658 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
2659 	VOP_UNLOCK(vp);
2660 	vn_finished_write(mp);
2661 	return (error);
2662 }
2663 
2664 /*
2665  * Change flags of a file given a path name.
2666  */
2667 #ifndef _SYS_SYSPROTO_H_
2668 struct chflags_args {
2669 	const char *path;
2670 	u_long	flags;
2671 };
2672 #endif
2673 int
2674 sys_chflags(struct thread *td, struct chflags_args *uap)
2675 {
2676 
2677 	return (kern_chflagsat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2678 	    uap->flags, 0));
2679 }
2680 
2681 #ifndef _SYS_SYSPROTO_H_
2682 struct chflagsat_args {
2683 	int	fd;
2684 	const char *path;
2685 	u_long	flags;
2686 	int	atflag;
2687 }
2688 #endif
2689 int
2690 sys_chflagsat(struct thread *td, struct chflagsat_args *uap)
2691 {
2692 
2693 	if ((uap->atflag & ~(AT_SYMLINK_NOFOLLOW | AT_BENEATH)) != 0)
2694 		return (EINVAL);
2695 
2696 	return (kern_chflagsat(td, uap->fd, uap->path, UIO_USERSPACE,
2697 	    uap->flags, uap->atflag));
2698 }
2699 
2700 /*
2701  * Same as chflags() but doesn't follow symlinks.
2702  */
2703 #ifndef _SYS_SYSPROTO_H_
2704 struct lchflags_args {
2705 	const char *path;
2706 	u_long flags;
2707 };
2708 #endif
2709 int
2710 sys_lchflags(struct thread *td, struct lchflags_args *uap)
2711 {
2712 
2713 	return (kern_chflagsat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2714 	    uap->flags, AT_SYMLINK_NOFOLLOW));
2715 }
2716 
2717 static int
2718 kern_chflagsat(struct thread *td, int fd, const char *path,
2719     enum uio_seg pathseg, u_long flags, int atflag)
2720 {
2721 	struct nameidata nd;
2722 	int error, follow;
2723 
2724 	AUDIT_ARG_FFLAGS(flags);
2725 	follow = (atflag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
2726 	follow |= (atflag & AT_BENEATH) != 0 ? BENEATH : 0;
2727 	NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, pathseg, path, fd,
2728 	    &cap_fchflags_rights, td);
2729 	if ((error = namei(&nd)) != 0)
2730 		return (error);
2731 	NDFREE(&nd, NDF_ONLY_PNBUF);
2732 	error = setfflags(td, nd.ni_vp, flags);
2733 	vrele(nd.ni_vp);
2734 	return (error);
2735 }
2736 
2737 /*
2738  * Change flags of a file given a file descriptor.
2739  */
2740 #ifndef _SYS_SYSPROTO_H_
2741 struct fchflags_args {
2742 	int	fd;
2743 	u_long	flags;
2744 };
2745 #endif
2746 int
2747 sys_fchflags(struct thread *td, struct fchflags_args *uap)
2748 {
2749 	struct file *fp;
2750 	int error;
2751 
2752 	AUDIT_ARG_FD(uap->fd);
2753 	AUDIT_ARG_FFLAGS(uap->flags);
2754 	error = getvnode(td, uap->fd, &cap_fchflags_rights,
2755 	    &fp);
2756 	if (error != 0)
2757 		return (error);
2758 #ifdef AUDIT
2759 	vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
2760 	AUDIT_ARG_VNODE1(fp->f_vnode);
2761 	VOP_UNLOCK(fp->f_vnode);
2762 #endif
2763 	error = setfflags(td, fp->f_vnode, uap->flags);
2764 	fdrop(fp, td);
2765 	return (error);
2766 }
2767 
2768 /*
2769  * Common implementation code for chmod(), lchmod() and fchmod().
2770  */
2771 int
2772 setfmode(struct thread *td, struct ucred *cred, struct vnode *vp, int mode)
2773 {
2774 	struct mount *mp;
2775 	struct vattr vattr;
2776 	int error;
2777 
2778 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2779 		return (error);
2780 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2781 	VATTR_NULL(&vattr);
2782 	vattr.va_mode = mode & ALLPERMS;
2783 #ifdef MAC
2784 	error = mac_vnode_check_setmode(cred, vp, vattr.va_mode);
2785 	if (error == 0)
2786 #endif
2787 		error = VOP_SETATTR(vp, &vattr, cred);
2788 	VOP_UNLOCK(vp);
2789 	vn_finished_write(mp);
2790 	return (error);
2791 }
2792 
2793 /*
2794  * Change mode of a file given path name.
2795  */
2796 #ifndef _SYS_SYSPROTO_H_
2797 struct chmod_args {
2798 	char	*path;
2799 	int	mode;
2800 };
2801 #endif
2802 int
2803 sys_chmod(struct thread *td, struct chmod_args *uap)
2804 {
2805 
2806 	return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2807 	    uap->mode, 0));
2808 }
2809 
2810 #ifndef _SYS_SYSPROTO_H_
2811 struct fchmodat_args {
2812 	int	dirfd;
2813 	char	*path;
2814 	mode_t	mode;
2815 	int	flag;
2816 }
2817 #endif
2818 int
2819 sys_fchmodat(struct thread *td, struct fchmodat_args *uap)
2820 {
2821 
2822 	if ((uap->flag & ~(AT_SYMLINK_NOFOLLOW | AT_BENEATH)) != 0)
2823 		return (EINVAL);
2824 
2825 	return (kern_fchmodat(td, uap->fd, uap->path, UIO_USERSPACE,
2826 	    uap->mode, uap->flag));
2827 }
2828 
2829 /*
2830  * Change mode of a file given path name (don't follow links.)
2831  */
2832 #ifndef _SYS_SYSPROTO_H_
2833 struct lchmod_args {
2834 	char	*path;
2835 	int	mode;
2836 };
2837 #endif
2838 int
2839 sys_lchmod(struct thread *td, struct lchmod_args *uap)
2840 {
2841 
2842 	return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2843 	    uap->mode, AT_SYMLINK_NOFOLLOW));
2844 }
2845 
2846 int
2847 kern_fchmodat(struct thread *td, int fd, const char *path,
2848     enum uio_seg pathseg, mode_t mode, int flag)
2849 {
2850 	struct nameidata nd;
2851 	int error, follow;
2852 
2853 	AUDIT_ARG_MODE(mode);
2854 	follow = (flag & AT_SYMLINK_NOFOLLOW) != 0 ? NOFOLLOW : FOLLOW;
2855 	follow |= (flag & AT_BENEATH) != 0 ? BENEATH : 0;
2856 	NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, pathseg, path, fd,
2857 	    &cap_fchmod_rights, td);
2858 	if ((error = namei(&nd)) != 0)
2859 		return (error);
2860 	NDFREE(&nd, NDF_ONLY_PNBUF);
2861 	error = setfmode(td, td->td_ucred, nd.ni_vp, mode);
2862 	vrele(nd.ni_vp);
2863 	return (error);
2864 }
2865 
2866 /*
2867  * Change mode of a file given a file descriptor.
2868  */
2869 #ifndef _SYS_SYSPROTO_H_
2870 struct fchmod_args {
2871 	int	fd;
2872 	int	mode;
2873 };
2874 #endif
2875 int
2876 sys_fchmod(struct thread *td, struct fchmod_args *uap)
2877 {
2878 	struct file *fp;
2879 	int error;
2880 
2881 	AUDIT_ARG_FD(uap->fd);
2882 	AUDIT_ARG_MODE(uap->mode);
2883 
2884 	error = fget(td, uap->fd, &cap_fchmod_rights, &fp);
2885 	if (error != 0)
2886 		return (error);
2887 	error = fo_chmod(fp, uap->mode, td->td_ucred, td);
2888 	fdrop(fp, td);
2889 	return (error);
2890 }
2891 
2892 /*
2893  * Common implementation for chown(), lchown(), and fchown()
2894  */
2895 int
2896 setfown(struct thread *td, struct ucred *cred, struct vnode *vp, uid_t uid,
2897     gid_t gid)
2898 {
2899 	struct mount *mp;
2900 	struct vattr vattr;
2901 	int error;
2902 
2903 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2904 		return (error);
2905 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2906 	VATTR_NULL(&vattr);
2907 	vattr.va_uid = uid;
2908 	vattr.va_gid = gid;
2909 #ifdef MAC
2910 	error = mac_vnode_check_setowner(cred, vp, vattr.va_uid,
2911 	    vattr.va_gid);
2912 	if (error == 0)
2913 #endif
2914 		error = VOP_SETATTR(vp, &vattr, cred);
2915 	VOP_UNLOCK(vp);
2916 	vn_finished_write(mp);
2917 	return (error);
2918 }
2919 
2920 /*
2921  * Set ownership given a path name.
2922  */
2923 #ifndef _SYS_SYSPROTO_H_
2924 struct chown_args {
2925 	char	*path;
2926 	int	uid;
2927 	int	gid;
2928 };
2929 #endif
2930 int
2931 sys_chown(struct thread *td, struct chown_args *uap)
2932 {
2933 
2934 	return (kern_fchownat(td, AT_FDCWD, uap->path, UIO_USERSPACE, uap->uid,
2935 	    uap->gid, 0));
2936 }
2937 
2938 #ifndef _SYS_SYSPROTO_H_
2939 struct fchownat_args {
2940 	int fd;
2941 	const char * path;
2942 	uid_t uid;
2943 	gid_t gid;
2944 	int flag;
2945 };
2946 #endif
2947 int
2948 sys_fchownat(struct thread *td, struct fchownat_args *uap)
2949 {
2950 
2951 	if ((uap->flag & ~(AT_SYMLINK_NOFOLLOW | AT_BENEATH)) != 0)
2952 		return (EINVAL);
2953 
2954 	return (kern_fchownat(td, uap->fd, uap->path, UIO_USERSPACE, uap->uid,
2955 	    uap->gid, uap->flag));
2956 }
2957 
2958 int
2959 kern_fchownat(struct thread *td, int fd, const char *path,
2960     enum uio_seg pathseg, int uid, int gid, int flag)
2961 {
2962 	struct nameidata nd;
2963 	int error, follow;
2964 
2965 	AUDIT_ARG_OWNER(uid, gid);
2966 	follow = (flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
2967 	follow |= (flag & AT_BENEATH) != 0 ? BENEATH : 0;
2968 	NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, pathseg, path, fd,
2969 	    &cap_fchown_rights, td);
2970 
2971 	if ((error = namei(&nd)) != 0)
2972 		return (error);
2973 	NDFREE(&nd, NDF_ONLY_PNBUF);
2974 	error = setfown(td, td->td_ucred, nd.ni_vp, uid, gid);
2975 	vrele(nd.ni_vp);
2976 	return (error);
2977 }
2978 
2979 /*
2980  * Set ownership given a path name, do not cross symlinks.
2981  */
2982 #ifndef _SYS_SYSPROTO_H_
2983 struct lchown_args {
2984 	char	*path;
2985 	int	uid;
2986 	int	gid;
2987 };
2988 #endif
2989 int
2990 sys_lchown(struct thread *td, struct lchown_args *uap)
2991 {
2992 
2993 	return (kern_fchownat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2994 	    uap->uid, uap->gid, AT_SYMLINK_NOFOLLOW));
2995 }
2996 
2997 /*
2998  * Set ownership given a file descriptor.
2999  */
3000 #ifndef _SYS_SYSPROTO_H_
3001 struct fchown_args {
3002 	int	fd;
3003 	int	uid;
3004 	int	gid;
3005 };
3006 #endif
3007 int
3008 sys_fchown(struct thread *td, struct fchown_args *uap)
3009 {
3010 	struct file *fp;
3011 	int error;
3012 
3013 	AUDIT_ARG_FD(uap->fd);
3014 	AUDIT_ARG_OWNER(uap->uid, uap->gid);
3015 	error = fget(td, uap->fd, &cap_fchown_rights, &fp);
3016 	if (error != 0)
3017 		return (error);
3018 	error = fo_chown(fp, uap->uid, uap->gid, td->td_ucred, td);
3019 	fdrop(fp, td);
3020 	return (error);
3021 }
3022 
3023 /*
3024  * Common implementation code for utimes(), lutimes(), and futimes().
3025  */
3026 static int
3027 getutimes(const struct timeval *usrtvp, enum uio_seg tvpseg,
3028     struct timespec *tsp)
3029 {
3030 	struct timeval tv[2];
3031 	const struct timeval *tvp;
3032 	int error;
3033 
3034 	if (usrtvp == NULL) {
3035 		vfs_timestamp(&tsp[0]);
3036 		tsp[1] = tsp[0];
3037 	} else {
3038 		if (tvpseg == UIO_SYSSPACE) {
3039 			tvp = usrtvp;
3040 		} else {
3041 			if ((error = copyin(usrtvp, tv, sizeof(tv))) != 0)
3042 				return (error);
3043 			tvp = tv;
3044 		}
3045 
3046 		if (tvp[0].tv_usec < 0 || tvp[0].tv_usec >= 1000000 ||
3047 		    tvp[1].tv_usec < 0 || tvp[1].tv_usec >= 1000000)
3048 			return (EINVAL);
3049 		TIMEVAL_TO_TIMESPEC(&tvp[0], &tsp[0]);
3050 		TIMEVAL_TO_TIMESPEC(&tvp[1], &tsp[1]);
3051 	}
3052 	return (0);
3053 }
3054 
3055 /*
3056  * Common implementation code for futimens(), utimensat().
3057  */
3058 #define	UTIMENS_NULL	0x1
3059 #define	UTIMENS_EXIT	0x2
3060 static int
3061 getutimens(const struct timespec *usrtsp, enum uio_seg tspseg,
3062     struct timespec *tsp, int *retflags)
3063 {
3064 	struct timespec tsnow;
3065 	int error;
3066 
3067 	vfs_timestamp(&tsnow);
3068 	*retflags = 0;
3069 	if (usrtsp == NULL) {
3070 		tsp[0] = tsnow;
3071 		tsp[1] = tsnow;
3072 		*retflags |= UTIMENS_NULL;
3073 		return (0);
3074 	}
3075 	if (tspseg == UIO_SYSSPACE) {
3076 		tsp[0] = usrtsp[0];
3077 		tsp[1] = usrtsp[1];
3078 	} else if ((error = copyin(usrtsp, tsp, sizeof(*tsp) * 2)) != 0)
3079 		return (error);
3080 	if (tsp[0].tv_nsec == UTIME_OMIT && tsp[1].tv_nsec == UTIME_OMIT)
3081 		*retflags |= UTIMENS_EXIT;
3082 	if (tsp[0].tv_nsec == UTIME_NOW && tsp[1].tv_nsec == UTIME_NOW)
3083 		*retflags |= UTIMENS_NULL;
3084 	if (tsp[0].tv_nsec == UTIME_OMIT)
3085 		tsp[0].tv_sec = VNOVAL;
3086 	else if (tsp[0].tv_nsec == UTIME_NOW)
3087 		tsp[0] = tsnow;
3088 	else if (tsp[0].tv_nsec < 0 || tsp[0].tv_nsec >= 1000000000L)
3089 		return (EINVAL);
3090 	if (tsp[1].tv_nsec == UTIME_OMIT)
3091 		tsp[1].tv_sec = VNOVAL;
3092 	else if (tsp[1].tv_nsec == UTIME_NOW)
3093 		tsp[1] = tsnow;
3094 	else if (tsp[1].tv_nsec < 0 || tsp[1].tv_nsec >= 1000000000L)
3095 		return (EINVAL);
3096 
3097 	return (0);
3098 }
3099 
3100 /*
3101  * Common implementation code for utimes(), lutimes(), futimes(), futimens(),
3102  * and utimensat().
3103  */
3104 static int
3105 setutimes(struct thread *td, struct vnode *vp, const struct timespec *ts,
3106     int numtimes, int nullflag)
3107 {
3108 	struct mount *mp;
3109 	struct vattr vattr;
3110 	int error, setbirthtime;
3111 
3112 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
3113 		return (error);
3114 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3115 	setbirthtime = 0;
3116 	if (numtimes < 3 && !VOP_GETATTR(vp, &vattr, td->td_ucred) &&
3117 	    timespeccmp(&ts[1], &vattr.va_birthtime, < ))
3118 		setbirthtime = 1;
3119 	VATTR_NULL(&vattr);
3120 	vattr.va_atime = ts[0];
3121 	vattr.va_mtime = ts[1];
3122 	if (setbirthtime)
3123 		vattr.va_birthtime = ts[1];
3124 	if (numtimes > 2)
3125 		vattr.va_birthtime = ts[2];
3126 	if (nullflag)
3127 		vattr.va_vaflags |= VA_UTIMES_NULL;
3128 #ifdef MAC
3129 	error = mac_vnode_check_setutimes(td->td_ucred, vp, vattr.va_atime,
3130 	    vattr.va_mtime);
3131 #endif
3132 	if (error == 0)
3133 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3134 	VOP_UNLOCK(vp);
3135 	vn_finished_write(mp);
3136 	return (error);
3137 }
3138 
3139 /*
3140  * Set the access and modification times of a file.
3141  */
3142 #ifndef _SYS_SYSPROTO_H_
3143 struct utimes_args {
3144 	char	*path;
3145 	struct	timeval *tptr;
3146 };
3147 #endif
3148 int
3149 sys_utimes(struct thread *td, struct utimes_args *uap)
3150 {
3151 
3152 	return (kern_utimesat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
3153 	    uap->tptr, UIO_USERSPACE));
3154 }
3155 
3156 #ifndef _SYS_SYSPROTO_H_
3157 struct futimesat_args {
3158 	int fd;
3159 	const char * path;
3160 	const struct timeval * times;
3161 };
3162 #endif
3163 int
3164 sys_futimesat(struct thread *td, struct futimesat_args *uap)
3165 {
3166 
3167 	return (kern_utimesat(td, uap->fd, uap->path, UIO_USERSPACE,
3168 	    uap->times, UIO_USERSPACE));
3169 }
3170 
3171 int
3172 kern_utimesat(struct thread *td, int fd, const char *path,
3173     enum uio_seg pathseg, struct timeval *tptr, enum uio_seg tptrseg)
3174 {
3175 	struct nameidata nd;
3176 	struct timespec ts[2];
3177 	int error;
3178 
3179 	if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3180 		return (error);
3181 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, fd,
3182 	    &cap_futimes_rights, td);
3183 
3184 	if ((error = namei(&nd)) != 0)
3185 		return (error);
3186 	NDFREE(&nd, NDF_ONLY_PNBUF);
3187 	error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3188 	vrele(nd.ni_vp);
3189 	return (error);
3190 }
3191 
3192 /*
3193  * Set the access and modification times of a file.
3194  */
3195 #ifndef _SYS_SYSPROTO_H_
3196 struct lutimes_args {
3197 	char	*path;
3198 	struct	timeval *tptr;
3199 };
3200 #endif
3201 int
3202 sys_lutimes(struct thread *td, struct lutimes_args *uap)
3203 {
3204 
3205 	return (kern_lutimes(td, uap->path, UIO_USERSPACE, uap->tptr,
3206 	    UIO_USERSPACE));
3207 }
3208 
3209 int
3210 kern_lutimes(struct thread *td, const char *path, enum uio_seg pathseg,
3211     struct timeval *tptr, enum uio_seg tptrseg)
3212 {
3213 	struct timespec ts[2];
3214 	struct nameidata nd;
3215 	int error;
3216 
3217 	if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3218 		return (error);
3219 	NDINIT(&nd, LOOKUP, NOFOLLOW | AUDITVNODE1, pathseg, path, td);
3220 	if ((error = namei(&nd)) != 0)
3221 		return (error);
3222 	NDFREE(&nd, NDF_ONLY_PNBUF);
3223 	error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3224 	vrele(nd.ni_vp);
3225 	return (error);
3226 }
3227 
3228 /*
3229  * Set the access and modification times of a file.
3230  */
3231 #ifndef _SYS_SYSPROTO_H_
3232 struct futimes_args {
3233 	int	fd;
3234 	struct	timeval *tptr;
3235 };
3236 #endif
3237 int
3238 sys_futimes(struct thread *td, struct futimes_args *uap)
3239 {
3240 
3241 	return (kern_futimes(td, uap->fd, uap->tptr, UIO_USERSPACE));
3242 }
3243 
3244 int
3245 kern_futimes(struct thread *td, int fd, struct timeval *tptr,
3246     enum uio_seg tptrseg)
3247 {
3248 	struct timespec ts[2];
3249 	struct file *fp;
3250 	int error;
3251 
3252 	AUDIT_ARG_FD(fd);
3253 	error = getutimes(tptr, tptrseg, ts);
3254 	if (error != 0)
3255 		return (error);
3256 	error = getvnode(td, fd, &cap_futimes_rights, &fp);
3257 	if (error != 0)
3258 		return (error);
3259 #ifdef AUDIT
3260 	vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
3261 	AUDIT_ARG_VNODE1(fp->f_vnode);
3262 	VOP_UNLOCK(fp->f_vnode);
3263 #endif
3264 	error = setutimes(td, fp->f_vnode, ts, 2, tptr == NULL);
3265 	fdrop(fp, td);
3266 	return (error);
3267 }
3268 
3269 int
3270 sys_futimens(struct thread *td, struct futimens_args *uap)
3271 {
3272 
3273 	return (kern_futimens(td, uap->fd, uap->times, UIO_USERSPACE));
3274 }
3275 
3276 int
3277 kern_futimens(struct thread *td, int fd, struct timespec *tptr,
3278     enum uio_seg tptrseg)
3279 {
3280 	struct timespec ts[2];
3281 	struct file *fp;
3282 	int error, flags;
3283 
3284 	AUDIT_ARG_FD(fd);
3285 	error = getutimens(tptr, tptrseg, ts, &flags);
3286 	if (error != 0)
3287 		return (error);
3288 	if (flags & UTIMENS_EXIT)
3289 		return (0);
3290 	error = getvnode(td, fd, &cap_futimes_rights, &fp);
3291 	if (error != 0)
3292 		return (error);
3293 #ifdef AUDIT
3294 	vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
3295 	AUDIT_ARG_VNODE1(fp->f_vnode);
3296 	VOP_UNLOCK(fp->f_vnode);
3297 #endif
3298 	error = setutimes(td, fp->f_vnode, ts, 2, flags & UTIMENS_NULL);
3299 	fdrop(fp, td);
3300 	return (error);
3301 }
3302 
3303 int
3304 sys_utimensat(struct thread *td, struct utimensat_args *uap)
3305 {
3306 
3307 	return (kern_utimensat(td, uap->fd, uap->path, UIO_USERSPACE,
3308 	    uap->times, UIO_USERSPACE, uap->flag));
3309 }
3310 
3311 int
3312 kern_utimensat(struct thread *td, int fd, const char *path,
3313     enum uio_seg pathseg, struct timespec *tptr, enum uio_seg tptrseg,
3314     int flag)
3315 {
3316 	struct nameidata nd;
3317 	struct timespec ts[2];
3318 	int error, flags;
3319 
3320 	if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_BENEATH)) != 0)
3321 		return (EINVAL);
3322 
3323 	if ((error = getutimens(tptr, tptrseg, ts, &flags)) != 0)
3324 		return (error);
3325 	NDINIT_ATRIGHTS(&nd, LOOKUP, ((flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW :
3326 	    FOLLOW) |  ((flag & AT_BENEATH) != 0 ? BENEATH : 0) | AUDITVNODE1,
3327 	    pathseg, path, fd, &cap_futimes_rights, td);
3328 	if ((error = namei(&nd)) != 0)
3329 		return (error);
3330 	/*
3331 	 * We are allowed to call namei() regardless of 2xUTIME_OMIT.
3332 	 * POSIX states:
3333 	 * "If both tv_nsec fields are UTIME_OMIT... EACCESS may be detected."
3334 	 * "Search permission is denied by a component of the path prefix."
3335 	 */
3336 	NDFREE(&nd, NDF_ONLY_PNBUF);
3337 	if ((flags & UTIMENS_EXIT) == 0)
3338 		error = setutimes(td, nd.ni_vp, ts, 2, flags & UTIMENS_NULL);
3339 	vrele(nd.ni_vp);
3340 	return (error);
3341 }
3342 
3343 /*
3344  * Truncate a file given its path name.
3345  */
3346 #ifndef _SYS_SYSPROTO_H_
3347 struct truncate_args {
3348 	char	*path;
3349 	int	pad;
3350 	off_t	length;
3351 };
3352 #endif
3353 int
3354 sys_truncate(struct thread *td, struct truncate_args *uap)
3355 {
3356 
3357 	return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3358 }
3359 
3360 int
3361 kern_truncate(struct thread *td, const char *path, enum uio_seg pathseg,
3362     off_t length)
3363 {
3364 	struct mount *mp;
3365 	struct vnode *vp;
3366 	void *rl_cookie;
3367 	struct vattr vattr;
3368 	struct nameidata nd;
3369 	int error;
3370 
3371 	if (length < 0)
3372 		return(EINVAL);
3373 	NDINIT(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, td);
3374 	if ((error = namei(&nd)) != 0)
3375 		return (error);
3376 	vp = nd.ni_vp;
3377 	rl_cookie = vn_rangelock_wlock(vp, 0, OFF_MAX);
3378 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) {
3379 		vn_rangelock_unlock(vp, rl_cookie);
3380 		vrele(vp);
3381 		return (error);
3382 	}
3383 	NDFREE(&nd, NDF_ONLY_PNBUF);
3384 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3385 	if (vp->v_type == VDIR)
3386 		error = EISDIR;
3387 #ifdef MAC
3388 	else if ((error = mac_vnode_check_write(td->td_ucred, NOCRED, vp))) {
3389 	}
3390 #endif
3391 	else if ((error = vn_writechk(vp)) == 0 &&
3392 	    (error = VOP_ACCESS(vp, VWRITE, td->td_ucred, td)) == 0) {
3393 		VATTR_NULL(&vattr);
3394 		vattr.va_size = length;
3395 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3396 	}
3397 	VOP_UNLOCK(vp);
3398 	vn_finished_write(mp);
3399 	vn_rangelock_unlock(vp, rl_cookie);
3400 	vrele(vp);
3401 	return (error);
3402 }
3403 
3404 #if defined(COMPAT_43)
3405 /*
3406  * Truncate a file given its path name.
3407  */
3408 #ifndef _SYS_SYSPROTO_H_
3409 struct otruncate_args {
3410 	char	*path;
3411 	long	length;
3412 };
3413 #endif
3414 int
3415 otruncate(struct thread *td, struct otruncate_args *uap)
3416 {
3417 
3418 	return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3419 }
3420 #endif /* COMPAT_43 */
3421 
3422 #if defined(COMPAT_FREEBSD6)
3423 /* Versions with the pad argument */
3424 int
3425 freebsd6_truncate(struct thread *td, struct freebsd6_truncate_args *uap)
3426 {
3427 
3428 	return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3429 }
3430 
3431 int
3432 freebsd6_ftruncate(struct thread *td, struct freebsd6_ftruncate_args *uap)
3433 {
3434 
3435 	return (kern_ftruncate(td, uap->fd, uap->length));
3436 }
3437 #endif
3438 
3439 int
3440 kern_fsync(struct thread *td, int fd, bool fullsync)
3441 {
3442 	struct vnode *vp;
3443 	struct mount *mp;
3444 	struct file *fp;
3445 	int error, lock_flags;
3446 
3447 	AUDIT_ARG_FD(fd);
3448 	error = getvnode(td, fd, &cap_fsync_rights, &fp);
3449 	if (error != 0)
3450 		return (error);
3451 	vp = fp->f_vnode;
3452 #if 0
3453 	if (!fullsync)
3454 		/* XXXKIB: compete outstanding aio writes */;
3455 #endif
3456 	error = vn_start_write(vp, &mp, V_WAIT | PCATCH);
3457 	if (error != 0)
3458 		goto drop;
3459 	if (MNT_SHARED_WRITES(mp) ||
3460 	    ((mp == NULL) && MNT_SHARED_WRITES(vp->v_mount))) {
3461 		lock_flags = LK_SHARED;
3462 	} else {
3463 		lock_flags = LK_EXCLUSIVE;
3464 	}
3465 	vn_lock(vp, lock_flags | LK_RETRY);
3466 	AUDIT_ARG_VNODE1(vp);
3467 	if (vp->v_object != NULL) {
3468 		VM_OBJECT_WLOCK(vp->v_object);
3469 		vm_object_page_clean(vp->v_object, 0, 0, 0);
3470 		VM_OBJECT_WUNLOCK(vp->v_object);
3471 	}
3472 	error = fullsync ? VOP_FSYNC(vp, MNT_WAIT, td) : VOP_FDATASYNC(vp, td);
3473 	VOP_UNLOCK(vp);
3474 	vn_finished_write(mp);
3475 drop:
3476 	fdrop(fp, td);
3477 	return (error);
3478 }
3479 
3480 /*
3481  * Sync an open file.
3482  */
3483 #ifndef _SYS_SYSPROTO_H_
3484 struct fsync_args {
3485 	int	fd;
3486 };
3487 #endif
3488 int
3489 sys_fsync(struct thread *td, struct fsync_args *uap)
3490 {
3491 
3492 	return (kern_fsync(td, uap->fd, true));
3493 }
3494 
3495 int
3496 sys_fdatasync(struct thread *td, struct fdatasync_args *uap)
3497 {
3498 
3499 	return (kern_fsync(td, uap->fd, false));
3500 }
3501 
3502 /*
3503  * Rename files.  Source and destination must either both be directories, or
3504  * both not be directories.  If target is a directory, it must be empty.
3505  */
3506 #ifndef _SYS_SYSPROTO_H_
3507 struct rename_args {
3508 	char	*from;
3509 	char	*to;
3510 };
3511 #endif
3512 int
3513 sys_rename(struct thread *td, struct rename_args *uap)
3514 {
3515 
3516 	return (kern_renameat(td, AT_FDCWD, uap->from, AT_FDCWD,
3517 	    uap->to, UIO_USERSPACE));
3518 }
3519 
3520 #ifndef _SYS_SYSPROTO_H_
3521 struct renameat_args {
3522 	int	oldfd;
3523 	char	*old;
3524 	int	newfd;
3525 	char	*new;
3526 };
3527 #endif
3528 int
3529 sys_renameat(struct thread *td, struct renameat_args *uap)
3530 {
3531 
3532 	return (kern_renameat(td, uap->oldfd, uap->old, uap->newfd, uap->new,
3533 	    UIO_USERSPACE));
3534 }
3535 
3536 #ifdef MAC
3537 static int
3538 kern_renameat_mac(struct thread *td, int oldfd, const char *old, int newfd,
3539     const char *new, enum uio_seg pathseg, struct nameidata *fromnd)
3540 {
3541 	int error;
3542 
3543 	NDINIT_ATRIGHTS(fromnd, DELETE, LOCKPARENT | LOCKLEAF | SAVESTART |
3544 	    AUDITVNODE1, pathseg, old, oldfd, &cap_renameat_source_rights, td);
3545 	if ((error = namei(fromnd)) != 0)
3546 		return (error);
3547 	error = mac_vnode_check_rename_from(td->td_ucred, fromnd->ni_dvp,
3548 	    fromnd->ni_vp, &fromnd->ni_cnd);
3549 	VOP_UNLOCK(fromnd->ni_dvp);
3550 	if (fromnd->ni_dvp != fromnd->ni_vp)
3551 		VOP_UNLOCK(fromnd->ni_vp);
3552 	if (error != 0) {
3553 		NDFREE(fromnd, NDF_ONLY_PNBUF);
3554 		vrele(fromnd->ni_dvp);
3555 		vrele(fromnd->ni_vp);
3556 		if (fromnd->ni_startdir)
3557 			vrele(fromnd->ni_startdir);
3558 	}
3559 	return (error);
3560 }
3561 #endif
3562 
3563 int
3564 kern_renameat(struct thread *td, int oldfd, const char *old, int newfd,
3565     const char *new, enum uio_seg pathseg)
3566 {
3567 	struct mount *mp = NULL;
3568 	struct vnode *tvp, *fvp, *tdvp;
3569 	struct nameidata fromnd, tond;
3570 	int error;
3571 
3572 again:
3573 	bwillwrite();
3574 #ifdef MAC
3575 	if (mac_vnode_check_rename_from_enabled()) {
3576 		error = kern_renameat_mac(td, oldfd, old, newfd, new, pathseg,
3577 		    &fromnd);
3578 		if (error != 0)
3579 			return (error);
3580 	} else {
3581 #endif
3582 	NDINIT_ATRIGHTS(&fromnd, DELETE, WANTPARENT | SAVESTART | AUDITVNODE1,
3583 	    pathseg, old, oldfd, &cap_renameat_source_rights, td);
3584 	if ((error = namei(&fromnd)) != 0)
3585 		return (error);
3586 #ifdef MAC
3587 	}
3588 #endif
3589 	fvp = fromnd.ni_vp;
3590 	NDINIT_ATRIGHTS(&tond, RENAME, LOCKPARENT | LOCKLEAF | NOCACHE |
3591 	    SAVESTART | AUDITVNODE2, pathseg, new, newfd,
3592 	    &cap_renameat_target_rights, td);
3593 	if (fromnd.ni_vp->v_type == VDIR)
3594 		tond.ni_cnd.cn_flags |= WILLBEDIR;
3595 	if ((error = namei(&tond)) != 0) {
3596 		/* Translate error code for rename("dir1", "dir2/."). */
3597 		if (error == EISDIR && fvp->v_type == VDIR)
3598 			error = EINVAL;
3599 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3600 		vrele(fromnd.ni_dvp);
3601 		vrele(fvp);
3602 		goto out1;
3603 	}
3604 	tdvp = tond.ni_dvp;
3605 	tvp = tond.ni_vp;
3606 	error = vn_start_write(fvp, &mp, V_NOWAIT);
3607 	if (error != 0) {
3608 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3609 		NDFREE(&tond, NDF_ONLY_PNBUF);
3610 		if (tvp != NULL)
3611 			vput(tvp);
3612 		if (tdvp == tvp)
3613 			vrele(tdvp);
3614 		else
3615 			vput(tdvp);
3616 		vrele(fromnd.ni_dvp);
3617 		vrele(fvp);
3618 		vrele(tond.ni_startdir);
3619 		if (fromnd.ni_startdir != NULL)
3620 			vrele(fromnd.ni_startdir);
3621 		error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH);
3622 		if (error != 0)
3623 			return (error);
3624 		goto again;
3625 	}
3626 	if (tvp != NULL) {
3627 		if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
3628 			error = ENOTDIR;
3629 			goto out;
3630 		} else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
3631 			error = EISDIR;
3632 			goto out;
3633 		}
3634 #ifdef CAPABILITIES
3635 		if (newfd != AT_FDCWD && (tond.ni_resflags & NIRES_ABS) == 0) {
3636 			/*
3637 			 * If the target already exists we require CAP_UNLINKAT
3638 			 * from 'newfd', when newfd was used for the lookup.
3639 			 */
3640 			error = cap_check(&tond.ni_filecaps.fc_rights,
3641 			    &cap_unlinkat_rights);
3642 			if (error != 0)
3643 				goto out;
3644 		}
3645 #endif
3646 	}
3647 	if (fvp == tdvp) {
3648 		error = EINVAL;
3649 		goto out;
3650 	}
3651 	/*
3652 	 * If the source is the same as the destination (that is, if they
3653 	 * are links to the same vnode), then there is nothing to do.
3654 	 */
3655 	if (fvp == tvp)
3656 		error = -1;
3657 #ifdef MAC
3658 	else
3659 		error = mac_vnode_check_rename_to(td->td_ucred, tdvp,
3660 		    tond.ni_vp, fromnd.ni_dvp == tdvp, &tond.ni_cnd);
3661 #endif
3662 out:
3663 	if (error == 0) {
3664 		error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
3665 		    tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
3666 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3667 		NDFREE(&tond, NDF_ONLY_PNBUF);
3668 	} else {
3669 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3670 		NDFREE(&tond, NDF_ONLY_PNBUF);
3671 		if (tvp != NULL)
3672 			vput(tvp);
3673 		if (tdvp == tvp)
3674 			vrele(tdvp);
3675 		else
3676 			vput(tdvp);
3677 		vrele(fromnd.ni_dvp);
3678 		vrele(fvp);
3679 	}
3680 	vrele(tond.ni_startdir);
3681 	vn_finished_write(mp);
3682 out1:
3683 	if (fromnd.ni_startdir)
3684 		vrele(fromnd.ni_startdir);
3685 	if (error == -1)
3686 		return (0);
3687 	return (error);
3688 }
3689 
3690 /*
3691  * Make a directory file.
3692  */
3693 #ifndef _SYS_SYSPROTO_H_
3694 struct mkdir_args {
3695 	char	*path;
3696 	int	mode;
3697 };
3698 #endif
3699 int
3700 sys_mkdir(struct thread *td, struct mkdir_args *uap)
3701 {
3702 
3703 	return (kern_mkdirat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
3704 	    uap->mode));
3705 }
3706 
3707 #ifndef _SYS_SYSPROTO_H_
3708 struct mkdirat_args {
3709 	int	fd;
3710 	char	*path;
3711 	mode_t	mode;
3712 };
3713 #endif
3714 int
3715 sys_mkdirat(struct thread *td, struct mkdirat_args *uap)
3716 {
3717 
3718 	return (kern_mkdirat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode));
3719 }
3720 
3721 int
3722 kern_mkdirat(struct thread *td, int fd, const char *path, enum uio_seg segflg,
3723     int mode)
3724 {
3725 	struct mount *mp;
3726 	struct vnode *vp;
3727 	struct vattr vattr;
3728 	struct nameidata nd;
3729 	int error;
3730 
3731 	AUDIT_ARG_MODE(mode);
3732 restart:
3733 	bwillwrite();
3734 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
3735 	    NOCACHE, segflg, path, fd, &cap_mkdirat_rights,
3736 	    td);
3737 	nd.ni_cnd.cn_flags |= WILLBEDIR;
3738 	if ((error = namei(&nd)) != 0)
3739 		return (error);
3740 	vp = nd.ni_vp;
3741 	if (vp != NULL) {
3742 		NDFREE(&nd, NDF_ONLY_PNBUF);
3743 		/*
3744 		 * XXX namei called with LOCKPARENT but not LOCKLEAF has
3745 		 * the strange behaviour of leaving the vnode unlocked
3746 		 * if the target is the same vnode as the parent.
3747 		 */
3748 		if (vp == nd.ni_dvp)
3749 			vrele(nd.ni_dvp);
3750 		else
3751 			vput(nd.ni_dvp);
3752 		vrele(vp);
3753 		return (EEXIST);
3754 	}
3755 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3756 		NDFREE(&nd, NDF_ONLY_PNBUF);
3757 		vput(nd.ni_dvp);
3758 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3759 			return (error);
3760 		goto restart;
3761 	}
3762 	VATTR_NULL(&vattr);
3763 	vattr.va_type = VDIR;
3764 	vattr.va_mode = (mode & ACCESSPERMS) &~ td->td_proc->p_fd->fd_cmask;
3765 #ifdef MAC
3766 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
3767 	    &vattr);
3768 	if (error != 0)
3769 		goto out;
3770 #endif
3771 	error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
3772 #ifdef MAC
3773 out:
3774 #endif
3775 	NDFREE(&nd, NDF_ONLY_PNBUF);
3776 	vput(nd.ni_dvp);
3777 	if (error == 0)
3778 		vput(nd.ni_vp);
3779 	vn_finished_write(mp);
3780 	return (error);
3781 }
3782 
3783 /*
3784  * Remove a directory file.
3785  */
3786 #ifndef _SYS_SYSPROTO_H_
3787 struct rmdir_args {
3788 	char	*path;
3789 };
3790 #endif
3791 int
3792 sys_rmdir(struct thread *td, struct rmdir_args *uap)
3793 {
3794 
3795 	return (kern_frmdirat(td, AT_FDCWD, uap->path, FD_NONE, UIO_USERSPACE,
3796 	    0));
3797 }
3798 
3799 int
3800 kern_frmdirat(struct thread *td, int dfd, const char *path, int fd,
3801     enum uio_seg pathseg, int flag)
3802 {
3803 	struct mount *mp;
3804 	struct vnode *vp;
3805 	struct file *fp;
3806 	struct nameidata nd;
3807 	cap_rights_t rights;
3808 	int error;
3809 
3810 	fp = NULL;
3811 	if (fd != FD_NONE) {
3812 		error = getvnode(td, fd, cap_rights_init_one(&rights, CAP_LOOKUP),
3813 		    &fp);
3814 		if (error != 0)
3815 			return (error);
3816 	}
3817 
3818 restart:
3819 	bwillwrite();
3820 	NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1 |
3821 	    ((flag & AT_BENEATH) != 0 ? BENEATH : 0),
3822 	    pathseg, path, dfd, &cap_unlinkat_rights, td);
3823 	if ((error = namei(&nd)) != 0)
3824 		goto fdout;
3825 	vp = nd.ni_vp;
3826 	if (vp->v_type != VDIR) {
3827 		error = ENOTDIR;
3828 		goto out;
3829 	}
3830 	/*
3831 	 * No rmdir "." please.
3832 	 */
3833 	if (nd.ni_dvp == vp) {
3834 		error = EINVAL;
3835 		goto out;
3836 	}
3837 	/*
3838 	 * The root of a mounted filesystem cannot be deleted.
3839 	 */
3840 	if (vp->v_vflag & VV_ROOT) {
3841 		error = EBUSY;
3842 		goto out;
3843 	}
3844 
3845 	if (fp != NULL && fp->f_vnode != vp) {
3846 		if (VN_IS_DOOMED(fp->f_vnode))
3847 			error = EBADF;
3848 		else
3849 			error = EDEADLK;
3850 		goto out;
3851 	}
3852 
3853 #ifdef MAC
3854 	error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
3855 	    &nd.ni_cnd);
3856 	if (error != 0)
3857 		goto out;
3858 #endif
3859 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3860 		NDFREE(&nd, NDF_ONLY_PNBUF);
3861 		vput(vp);
3862 		if (nd.ni_dvp == vp)
3863 			vrele(nd.ni_dvp);
3864 		else
3865 			vput(nd.ni_dvp);
3866 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3867 			goto fdout;
3868 		goto restart;
3869 	}
3870 	vfs_notify_upper(vp, VFS_NOTIFY_UPPER_UNLINK);
3871 	error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
3872 	vn_finished_write(mp);
3873 out:
3874 	NDFREE(&nd, NDF_ONLY_PNBUF);
3875 	vput(vp);
3876 	if (nd.ni_dvp == vp)
3877 		vrele(nd.ni_dvp);
3878 	else
3879 		vput(nd.ni_dvp);
3880 fdout:
3881 	if (fp != NULL)
3882 		fdrop(fp, td);
3883 	return (error);
3884 }
3885 
3886 #if defined(COMPAT_43) || defined(COMPAT_FREEBSD11)
3887 int
3888 freebsd11_kern_getdirentries(struct thread *td, int fd, char *ubuf, u_int count,
3889     long *basep, void (*func)(struct freebsd11_dirent *))
3890 {
3891 	struct freebsd11_dirent dstdp;
3892 	struct dirent *dp, *edp;
3893 	char *dirbuf;
3894 	off_t base;
3895 	ssize_t resid, ucount;
3896 	int error;
3897 
3898 	/* XXX arbitrary sanity limit on `count'. */
3899 	count = min(count, 64 * 1024);
3900 
3901 	dirbuf = malloc(count, M_TEMP, M_WAITOK);
3902 
3903 	error = kern_getdirentries(td, fd, dirbuf, count, &base, &resid,
3904 	    UIO_SYSSPACE);
3905 	if (error != 0)
3906 		goto done;
3907 	if (basep != NULL)
3908 		*basep = base;
3909 
3910 	ucount = 0;
3911 	for (dp = (struct dirent *)dirbuf,
3912 	    edp = (struct dirent *)&dirbuf[count - resid];
3913 	    ucount < count && dp < edp; ) {
3914 		if (dp->d_reclen == 0)
3915 			break;
3916 		MPASS(dp->d_reclen >= _GENERIC_DIRLEN(0));
3917 		if (dp->d_namlen >= sizeof(dstdp.d_name))
3918 			continue;
3919 		dstdp.d_type = dp->d_type;
3920 		dstdp.d_namlen = dp->d_namlen;
3921 		dstdp.d_fileno = dp->d_fileno;		/* truncate */
3922 		if (dstdp.d_fileno != dp->d_fileno) {
3923 			switch (ino64_trunc_error) {
3924 			default:
3925 			case 0:
3926 				break;
3927 			case 1:
3928 				error = EOVERFLOW;
3929 				goto done;
3930 			case 2:
3931 				dstdp.d_fileno = UINT32_MAX;
3932 				break;
3933 			}
3934 		}
3935 		dstdp.d_reclen = sizeof(dstdp) - sizeof(dstdp.d_name) +
3936 		    ((dp->d_namlen + 1 + 3) &~ 3);
3937 		bcopy(dp->d_name, dstdp.d_name, dstdp.d_namlen);
3938 		bzero(dstdp.d_name + dstdp.d_namlen,
3939 		    dstdp.d_reclen - offsetof(struct freebsd11_dirent, d_name) -
3940 		    dstdp.d_namlen);
3941 		MPASS(dstdp.d_reclen <= dp->d_reclen);
3942 		MPASS(ucount + dstdp.d_reclen <= count);
3943 		if (func != NULL)
3944 			func(&dstdp);
3945 		error = copyout(&dstdp, ubuf + ucount, dstdp.d_reclen);
3946 		if (error != 0)
3947 			break;
3948 		dp = (struct dirent *)((char *)dp + dp->d_reclen);
3949 		ucount += dstdp.d_reclen;
3950 	}
3951 
3952 done:
3953 	free(dirbuf, M_TEMP);
3954 	if (error == 0)
3955 		td->td_retval[0] = ucount;
3956 	return (error);
3957 }
3958 #endif /* COMPAT */
3959 
3960 #ifdef COMPAT_43
3961 static void
3962 ogetdirentries_cvt(struct freebsd11_dirent *dp)
3963 {
3964 #if (BYTE_ORDER == LITTLE_ENDIAN)
3965 	/*
3966 	 * The expected low byte of dp->d_namlen is our dp->d_type.
3967 	 * The high MBZ byte of dp->d_namlen is our dp->d_namlen.
3968 	 */
3969 	dp->d_type = dp->d_namlen;
3970 	dp->d_namlen = 0;
3971 #else
3972 	/*
3973 	 * The dp->d_type is the high byte of the expected dp->d_namlen,
3974 	 * so must be zero'ed.
3975 	 */
3976 	dp->d_type = 0;
3977 #endif
3978 }
3979 
3980 /*
3981  * Read a block of directory entries in a filesystem independent format.
3982  */
3983 #ifndef _SYS_SYSPROTO_H_
3984 struct ogetdirentries_args {
3985 	int	fd;
3986 	char	*buf;
3987 	u_int	count;
3988 	long	*basep;
3989 };
3990 #endif
3991 int
3992 ogetdirentries(struct thread *td, struct ogetdirentries_args *uap)
3993 {
3994 	long loff;
3995 	int error;
3996 
3997 	error = kern_ogetdirentries(td, uap, &loff);
3998 	if (error == 0)
3999 		error = copyout(&loff, uap->basep, sizeof(long));
4000 	return (error);
4001 }
4002 
4003 int
4004 kern_ogetdirentries(struct thread *td, struct ogetdirentries_args *uap,
4005     long *ploff)
4006 {
4007 	long base;
4008 	int error;
4009 
4010 	/* XXX arbitrary sanity limit on `count'. */
4011 	if (uap->count > 64 * 1024)
4012 		return (EINVAL);
4013 
4014 	error = freebsd11_kern_getdirentries(td, uap->fd, uap->buf, uap->count,
4015 	    &base, ogetdirentries_cvt);
4016 
4017 	if (error == 0 && uap->basep != NULL)
4018 		error = copyout(&base, uap->basep, sizeof(long));
4019 
4020 	return (error);
4021 }
4022 #endif /* COMPAT_43 */
4023 
4024 #if defined(COMPAT_FREEBSD11)
4025 #ifndef _SYS_SYSPROTO_H_
4026 struct freebsd11_getdirentries_args {
4027 	int	fd;
4028 	char	*buf;
4029 	u_int	count;
4030 	long	*basep;
4031 };
4032 #endif
4033 int
4034 freebsd11_getdirentries(struct thread *td,
4035     struct freebsd11_getdirentries_args *uap)
4036 {
4037 	long base;
4038 	int error;
4039 
4040 	error = freebsd11_kern_getdirentries(td, uap->fd, uap->buf, uap->count,
4041 	    &base, NULL);
4042 
4043 	if (error == 0 && uap->basep != NULL)
4044 		error = copyout(&base, uap->basep, sizeof(long));
4045 	return (error);
4046 }
4047 
4048 int
4049 freebsd11_getdents(struct thread *td, struct freebsd11_getdents_args *uap)
4050 {
4051 	struct freebsd11_getdirentries_args ap;
4052 
4053 	ap.fd = uap->fd;
4054 	ap.buf = uap->buf;
4055 	ap.count = uap->count;
4056 	ap.basep = NULL;
4057 	return (freebsd11_getdirentries(td, &ap));
4058 }
4059 #endif /* COMPAT_FREEBSD11 */
4060 
4061 /*
4062  * Read a block of directory entries in a filesystem independent format.
4063  */
4064 int
4065 sys_getdirentries(struct thread *td, struct getdirentries_args *uap)
4066 {
4067 	off_t base;
4068 	int error;
4069 
4070 	error = kern_getdirentries(td, uap->fd, uap->buf, uap->count, &base,
4071 	    NULL, UIO_USERSPACE);
4072 	if (error != 0)
4073 		return (error);
4074 	if (uap->basep != NULL)
4075 		error = copyout(&base, uap->basep, sizeof(off_t));
4076 	return (error);
4077 }
4078 
4079 int
4080 kern_getdirentries(struct thread *td, int fd, char *buf, size_t count,
4081     off_t *basep, ssize_t *residp, enum uio_seg bufseg)
4082 {
4083 	struct vnode *vp;
4084 	struct file *fp;
4085 	struct uio auio;
4086 	struct iovec aiov;
4087 	off_t loff;
4088 	int error, eofflag;
4089 	off_t foffset;
4090 
4091 	AUDIT_ARG_FD(fd);
4092 	if (count > IOSIZE_MAX)
4093 		return (EINVAL);
4094 	auio.uio_resid = count;
4095 	error = getvnode(td, fd, &cap_read_rights, &fp);
4096 	if (error != 0)
4097 		return (error);
4098 	if ((fp->f_flag & FREAD) == 0) {
4099 		fdrop(fp, td);
4100 		return (EBADF);
4101 	}
4102 	vp = fp->f_vnode;
4103 	foffset = foffset_lock(fp, 0);
4104 unionread:
4105 	if (vp->v_type != VDIR) {
4106 		error = EINVAL;
4107 		goto fail;
4108 	}
4109 	aiov.iov_base = buf;
4110 	aiov.iov_len = count;
4111 	auio.uio_iov = &aiov;
4112 	auio.uio_iovcnt = 1;
4113 	auio.uio_rw = UIO_READ;
4114 	auio.uio_segflg = bufseg;
4115 	auio.uio_td = td;
4116 	vn_lock(vp, LK_SHARED | LK_RETRY);
4117 	AUDIT_ARG_VNODE1(vp);
4118 	loff = auio.uio_offset = foffset;
4119 #ifdef MAC
4120 	error = mac_vnode_check_readdir(td->td_ucred, vp);
4121 	if (error == 0)
4122 #endif
4123 		error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, NULL,
4124 		    NULL);
4125 	foffset = auio.uio_offset;
4126 	if (error != 0) {
4127 		VOP_UNLOCK(vp);
4128 		goto fail;
4129 	}
4130 	if (count == auio.uio_resid &&
4131 	    (vp->v_vflag & VV_ROOT) &&
4132 	    (vp->v_mount->mnt_flag & MNT_UNION)) {
4133 		struct vnode *tvp = vp;
4134 
4135 		vp = vp->v_mount->mnt_vnodecovered;
4136 		VREF(vp);
4137 		fp->f_vnode = vp;
4138 		foffset = 0;
4139 		vput(tvp);
4140 		goto unionread;
4141 	}
4142 	VOP_UNLOCK(vp);
4143 	*basep = loff;
4144 	if (residp != NULL)
4145 		*residp = auio.uio_resid;
4146 	td->td_retval[0] = count - auio.uio_resid;
4147 fail:
4148 	foffset_unlock(fp, foffset, 0);
4149 	fdrop(fp, td);
4150 	return (error);
4151 }
4152 
4153 /*
4154  * Set the mode mask for creation of filesystem nodes.
4155  */
4156 #ifndef _SYS_SYSPROTO_H_
4157 struct umask_args {
4158 	int	newmask;
4159 };
4160 #endif
4161 int
4162 sys_umask(struct thread *td, struct umask_args *uap)
4163 {
4164 	struct filedesc *fdp;
4165 
4166 	fdp = td->td_proc->p_fd;
4167 	FILEDESC_XLOCK(fdp);
4168 	td->td_retval[0] = fdp->fd_cmask;
4169 	fdp->fd_cmask = uap->newmask & ALLPERMS;
4170 	FILEDESC_XUNLOCK(fdp);
4171 	return (0);
4172 }
4173 
4174 /*
4175  * Void all references to file by ripping underlying filesystem away from
4176  * vnode.
4177  */
4178 #ifndef _SYS_SYSPROTO_H_
4179 struct revoke_args {
4180 	char	*path;
4181 };
4182 #endif
4183 int
4184 sys_revoke(struct thread *td, struct revoke_args *uap)
4185 {
4186 	struct vnode *vp;
4187 	struct vattr vattr;
4188 	struct nameidata nd;
4189 	int error;
4190 
4191 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
4192 	    uap->path, td);
4193 	if ((error = namei(&nd)) != 0)
4194 		return (error);
4195 	vp = nd.ni_vp;
4196 	NDFREE(&nd, NDF_ONLY_PNBUF);
4197 	if (vp->v_type != VCHR || vp->v_rdev == NULL) {
4198 		error = EINVAL;
4199 		goto out;
4200 	}
4201 #ifdef MAC
4202 	error = mac_vnode_check_revoke(td->td_ucred, vp);
4203 	if (error != 0)
4204 		goto out;
4205 #endif
4206 	error = VOP_GETATTR(vp, &vattr, td->td_ucred);
4207 	if (error != 0)
4208 		goto out;
4209 	if (td->td_ucred->cr_uid != vattr.va_uid) {
4210 		error = priv_check(td, PRIV_VFS_ADMIN);
4211 		if (error != 0)
4212 			goto out;
4213 	}
4214 	if (devfs_usecount(vp) > 0)
4215 		VOP_REVOKE(vp, REVOKEALL);
4216 out:
4217 	vput(vp);
4218 	return (error);
4219 }
4220 
4221 /*
4222  * Convert a user file descriptor to a kernel file entry and check that, if it
4223  * is a capability, the correct rights are present. A reference on the file
4224  * entry is held upon returning.
4225  */
4226 int
4227 getvnode(struct thread *td, int fd, cap_rights_t *rightsp, struct file **fpp)
4228 {
4229 	struct file *fp;
4230 	int error;
4231 
4232 	error = fget_unlocked(td->td_proc->p_fd, fd, rightsp, &fp);
4233 	if (error != 0)
4234 		return (error);
4235 
4236 	/*
4237 	 * The file could be not of the vnode type, or it may be not
4238 	 * yet fully initialized, in which case the f_vnode pointer
4239 	 * may be set, but f_ops is still badfileops.  E.g.,
4240 	 * devfs_open() transiently create such situation to
4241 	 * facilitate csw d_fdopen().
4242 	 *
4243 	 * Dupfdopen() handling in kern_openat() installs the
4244 	 * half-baked file into the process descriptor table, allowing
4245 	 * other thread to dereference it. Guard against the race by
4246 	 * checking f_ops.
4247 	 */
4248 	if (fp->f_vnode == NULL || fp->f_ops == &badfileops) {
4249 		fdrop(fp, td);
4250 		return (EINVAL);
4251 	}
4252 	*fpp = fp;
4253 	return (0);
4254 }
4255 
4256 /*
4257  * Get an (NFS) file handle.
4258  */
4259 #ifndef _SYS_SYSPROTO_H_
4260 struct lgetfh_args {
4261 	char *fname;
4262 	fhandle_t *fhp;
4263 };
4264 #endif
4265 int
4266 sys_lgetfh(struct thread *td, struct lgetfh_args *uap)
4267 {
4268 
4269 	return (kern_getfhat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->fname,
4270 	    UIO_USERSPACE, uap->fhp));
4271 }
4272 
4273 #ifndef _SYS_SYSPROTO_H_
4274 struct getfh_args {
4275 	char *fname;
4276 	fhandle_t *fhp;
4277 };
4278 #endif
4279 int
4280 sys_getfh(struct thread *td, struct getfh_args *uap)
4281 {
4282 
4283 	return (kern_getfhat(td, 0, AT_FDCWD, uap->fname, UIO_USERSPACE,
4284 	    uap->fhp));
4285 }
4286 
4287 /*
4288  * syscall for the rpc.lockd to use to translate an open descriptor into
4289  * a NFS file handle.
4290  *
4291  * warning: do not remove the priv_check() call or this becomes one giant
4292  * security hole.
4293  */
4294 #ifndef _SYS_SYSPROTO_H_
4295 struct getfhat_args {
4296 	int fd;
4297 	char *path;
4298 	fhandle_t *fhp;
4299 	int flags;
4300 };
4301 #endif
4302 int
4303 sys_getfhat(struct thread *td, struct getfhat_args *uap)
4304 {
4305 
4306 	if ((uap->flags & ~(AT_SYMLINK_NOFOLLOW | AT_BENEATH)) != 0)
4307 		return (EINVAL);
4308 	return (kern_getfhat(td, uap->flags, uap->fd, uap->path, UIO_USERSPACE,
4309 	    uap->fhp));
4310 }
4311 
4312 static int
4313 kern_getfhat(struct thread *td, int flags, int fd, const char *path,
4314     enum uio_seg pathseg, fhandle_t *fhp)
4315 {
4316 	struct nameidata nd;
4317 	fhandle_t fh;
4318 	struct vnode *vp;
4319 	int error;
4320 
4321 	error = priv_check(td, PRIV_VFS_GETFH);
4322 	if (error != 0)
4323 		return (error);
4324 	NDINIT_AT(&nd, LOOKUP, ((flags & AT_SYMLINK_NOFOLLOW) != 0 ? NOFOLLOW :
4325 	    FOLLOW) | ((flags & AT_BENEATH) != 0 ? BENEATH : 0) | LOCKLEAF |
4326 	    AUDITVNODE1, pathseg, path, fd, td);
4327 	error = namei(&nd);
4328 	if (error != 0)
4329 		return (error);
4330 	NDFREE(&nd, NDF_ONLY_PNBUF);
4331 	vp = nd.ni_vp;
4332 	bzero(&fh, sizeof(fh));
4333 	fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
4334 	error = VOP_VPTOFH(vp, &fh.fh_fid);
4335 	vput(vp);
4336 	if (error == 0)
4337 		error = copyout(&fh, fhp, sizeof (fh));
4338 	return (error);
4339 }
4340 
4341 #ifndef _SYS_SYSPROTO_H_
4342 struct fhlink_args {
4343 	fhandle_t *fhp;
4344 	const char *to;
4345 };
4346 #endif
4347 int
4348 sys_fhlink(struct thread *td, struct fhlink_args *uap)
4349 {
4350 
4351 	return (kern_fhlinkat(td, AT_FDCWD, uap->to, UIO_USERSPACE, uap->fhp));
4352 }
4353 
4354 #ifndef _SYS_SYSPROTO_H_
4355 struct fhlinkat_args {
4356 	fhandle_t *fhp;
4357 	int tofd;
4358 	const char *to;
4359 };
4360 #endif
4361 int
4362 sys_fhlinkat(struct thread *td, struct fhlinkat_args *uap)
4363 {
4364 
4365 	return (kern_fhlinkat(td, uap->tofd, uap->to, UIO_USERSPACE, uap->fhp));
4366 }
4367 
4368 static int
4369 kern_fhlinkat(struct thread *td, int fd, const char *path,
4370     enum uio_seg pathseg, fhandle_t *fhp)
4371 {
4372 	fhandle_t fh;
4373 	struct mount *mp;
4374 	struct vnode *vp;
4375 	int error;
4376 
4377 	error = priv_check(td, PRIV_VFS_GETFH);
4378 	if (error != 0)
4379 		return (error);
4380 	error = copyin(fhp, &fh, sizeof(fh));
4381 	if (error != 0)
4382 		return (error);
4383 	do {
4384 		bwillwrite();
4385 		if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4386 			return (ESTALE);
4387 		error = VFS_FHTOVP(mp, &fh.fh_fid, LK_SHARED, &vp);
4388 		vfs_unbusy(mp);
4389 		if (error != 0)
4390 			return (error);
4391 		VOP_UNLOCK(vp);
4392 	} while ((error = kern_linkat_vp(td, vp, fd, path, pathseg)) == EAGAIN);
4393 	return (error);
4394 }
4395 
4396 #ifndef _SYS_SYSPROTO_H_
4397 struct fhreadlink_args {
4398 	fhandle_t *fhp;
4399 	char *buf;
4400 	size_t bufsize;
4401 };
4402 #endif
4403 int
4404 sys_fhreadlink(struct thread *td, struct fhreadlink_args *uap)
4405 {
4406 	fhandle_t fh;
4407 	struct mount *mp;
4408 	struct vnode *vp;
4409 	int error;
4410 
4411 	error = priv_check(td, PRIV_VFS_GETFH);
4412 	if (error != 0)
4413 		return (error);
4414 	if (uap->bufsize > IOSIZE_MAX)
4415 		return (EINVAL);
4416 	error = copyin(uap->fhp, &fh, sizeof(fh));
4417 	if (error != 0)
4418 		return (error);
4419 	if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4420 		return (ESTALE);
4421 	error = VFS_FHTOVP(mp, &fh.fh_fid, LK_SHARED, &vp);
4422 	vfs_unbusy(mp);
4423 	if (error != 0)
4424 		return (error);
4425 	error = kern_readlink_vp(vp, uap->buf, UIO_USERSPACE, uap->bufsize, td);
4426 	vput(vp);
4427 	return (error);
4428 }
4429 
4430 /*
4431  * syscall for the rpc.lockd to use to translate a NFS file handle into an
4432  * open descriptor.
4433  *
4434  * warning: do not remove the priv_check() call or this becomes one giant
4435  * security hole.
4436  */
4437 #ifndef _SYS_SYSPROTO_H_
4438 struct fhopen_args {
4439 	const struct fhandle *u_fhp;
4440 	int flags;
4441 };
4442 #endif
4443 int
4444 sys_fhopen(struct thread *td, struct fhopen_args *uap)
4445 {
4446 	struct mount *mp;
4447 	struct vnode *vp;
4448 	struct fhandle fhp;
4449 	struct file *fp;
4450 	int fmode, error;
4451 	int indx;
4452 
4453 	error = priv_check(td, PRIV_VFS_FHOPEN);
4454 	if (error != 0)
4455 		return (error);
4456 	indx = -1;
4457 	fmode = FFLAGS(uap->flags);
4458 	/* why not allow a non-read/write open for our lockd? */
4459 	if (((fmode & (FREAD | FWRITE)) == 0) || (fmode & O_CREAT))
4460 		return (EINVAL);
4461 	error = copyin(uap->u_fhp, &fhp, sizeof(fhp));
4462 	if (error != 0)
4463 		return(error);
4464 	/* find the mount point */
4465 	mp = vfs_busyfs(&fhp.fh_fsid);
4466 	if (mp == NULL)
4467 		return (ESTALE);
4468 	/* now give me my vnode, it gets returned to me locked */
4469 	error = VFS_FHTOVP(mp, &fhp.fh_fid, LK_EXCLUSIVE, &vp);
4470 	vfs_unbusy(mp);
4471 	if (error != 0)
4472 		return (error);
4473 
4474 	error = falloc_noinstall(td, &fp);
4475 	if (error != 0) {
4476 		vput(vp);
4477 		return (error);
4478 	}
4479 	/*
4480 	 * An extra reference on `fp' has been held for us by
4481 	 * falloc_noinstall().
4482 	 */
4483 
4484 #ifdef INVARIANTS
4485 	td->td_dupfd = -1;
4486 #endif
4487 	error = vn_open_vnode(vp, fmode, td->td_ucred, td, fp);
4488 	if (error != 0) {
4489 		KASSERT(fp->f_ops == &badfileops,
4490 		    ("VOP_OPEN in fhopen() set f_ops"));
4491 		KASSERT(td->td_dupfd < 0,
4492 		    ("fhopen() encountered fdopen()"));
4493 
4494 		vput(vp);
4495 		goto bad;
4496 	}
4497 #ifdef INVARIANTS
4498 	td->td_dupfd = 0;
4499 #endif
4500 	fp->f_vnode = vp;
4501 	finit_vnode(fp, fmode, NULL, &vnops);
4502 	VOP_UNLOCK(vp);
4503 	if ((fmode & O_TRUNC) != 0) {
4504 		error = fo_truncate(fp, 0, td->td_ucred, td);
4505 		if (error != 0)
4506 			goto bad;
4507 	}
4508 
4509 	error = finstall(td, fp, &indx, fmode, NULL);
4510 bad:
4511 	fdrop(fp, td);
4512 	td->td_retval[0] = indx;
4513 	return (error);
4514 }
4515 
4516 /*
4517  * Stat an (NFS) file handle.
4518  */
4519 #ifndef _SYS_SYSPROTO_H_
4520 struct fhstat_args {
4521 	struct fhandle *u_fhp;
4522 	struct stat *sb;
4523 };
4524 #endif
4525 int
4526 sys_fhstat(struct thread *td, struct fhstat_args *uap)
4527 {
4528 	struct stat sb;
4529 	struct fhandle fh;
4530 	int error;
4531 
4532 	error = copyin(uap->u_fhp, &fh, sizeof(fh));
4533 	if (error != 0)
4534 		return (error);
4535 	error = kern_fhstat(td, fh, &sb);
4536 	if (error == 0)
4537 		error = copyout(&sb, uap->sb, sizeof(sb));
4538 	return (error);
4539 }
4540 
4541 int
4542 kern_fhstat(struct thread *td, struct fhandle fh, struct stat *sb)
4543 {
4544 	struct mount *mp;
4545 	struct vnode *vp;
4546 	int error;
4547 
4548 	error = priv_check(td, PRIV_VFS_FHSTAT);
4549 	if (error != 0)
4550 		return (error);
4551 	if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4552 		return (ESTALE);
4553 	error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4554 	vfs_unbusy(mp);
4555 	if (error != 0)
4556 		return (error);
4557 	error = VOP_STAT(vp, sb, td->td_ucred, NOCRED, td);
4558 	vput(vp);
4559 	return (error);
4560 }
4561 
4562 /*
4563  * Implement fstatfs() for (NFS) file handles.
4564  */
4565 #ifndef _SYS_SYSPROTO_H_
4566 struct fhstatfs_args {
4567 	struct fhandle *u_fhp;
4568 	struct statfs *buf;
4569 };
4570 #endif
4571 int
4572 sys_fhstatfs(struct thread *td, struct fhstatfs_args *uap)
4573 {
4574 	struct statfs *sfp;
4575 	fhandle_t fh;
4576 	int error;
4577 
4578 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
4579 	if (error != 0)
4580 		return (error);
4581 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
4582 	error = kern_fhstatfs(td, fh, sfp);
4583 	if (error == 0)
4584 		error = copyout(sfp, uap->buf, sizeof(*sfp));
4585 	free(sfp, M_STATFS);
4586 	return (error);
4587 }
4588 
4589 int
4590 kern_fhstatfs(struct thread *td, fhandle_t fh, struct statfs *buf)
4591 {
4592 	struct mount *mp;
4593 	struct vnode *vp;
4594 	int error;
4595 
4596 	error = priv_check(td, PRIV_VFS_FHSTATFS);
4597 	if (error != 0)
4598 		return (error);
4599 	if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4600 		return (ESTALE);
4601 	error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4602 	if (error != 0) {
4603 		vfs_unbusy(mp);
4604 		return (error);
4605 	}
4606 	vput(vp);
4607 	error = prison_canseemount(td->td_ucred, mp);
4608 	if (error != 0)
4609 		goto out;
4610 #ifdef MAC
4611 	error = mac_mount_check_stat(td->td_ucred, mp);
4612 	if (error != 0)
4613 		goto out;
4614 #endif
4615 	error = VFS_STATFS(mp, buf);
4616 out:
4617 	vfs_unbusy(mp);
4618 	return (error);
4619 }
4620 
4621 /*
4622  * Unlike madvise(2), we do not make a best effort to remember every
4623  * possible caching hint.  Instead, we remember the last setting with
4624  * the exception that we will allow POSIX_FADV_NORMAL to adjust the
4625  * region of any current setting.
4626  */
4627 int
4628 kern_posix_fadvise(struct thread *td, int fd, off_t offset, off_t len,
4629     int advice)
4630 {
4631 	struct fadvise_info *fa, *new;
4632 	struct file *fp;
4633 	struct vnode *vp;
4634 	off_t end;
4635 	int error;
4636 
4637 	if (offset < 0 || len < 0 || offset > OFF_MAX - len)
4638 		return (EINVAL);
4639 	AUDIT_ARG_VALUE(advice);
4640 	switch (advice) {
4641 	case POSIX_FADV_SEQUENTIAL:
4642 	case POSIX_FADV_RANDOM:
4643 	case POSIX_FADV_NOREUSE:
4644 		new = malloc(sizeof(*fa), M_FADVISE, M_WAITOK);
4645 		break;
4646 	case POSIX_FADV_NORMAL:
4647 	case POSIX_FADV_WILLNEED:
4648 	case POSIX_FADV_DONTNEED:
4649 		new = NULL;
4650 		break;
4651 	default:
4652 		return (EINVAL);
4653 	}
4654 	/* XXX: CAP_POSIX_FADVISE? */
4655 	AUDIT_ARG_FD(fd);
4656 	error = fget(td, fd, &cap_no_rights, &fp);
4657 	if (error != 0)
4658 		goto out;
4659 	AUDIT_ARG_FILE(td->td_proc, fp);
4660 	if ((fp->f_ops->fo_flags & DFLAG_SEEKABLE) == 0) {
4661 		error = ESPIPE;
4662 		goto out;
4663 	}
4664 	if (fp->f_type != DTYPE_VNODE) {
4665 		error = ENODEV;
4666 		goto out;
4667 	}
4668 	vp = fp->f_vnode;
4669 	if (vp->v_type != VREG) {
4670 		error = ENODEV;
4671 		goto out;
4672 	}
4673 	if (len == 0)
4674 		end = OFF_MAX;
4675 	else
4676 		end = offset + len - 1;
4677 	switch (advice) {
4678 	case POSIX_FADV_SEQUENTIAL:
4679 	case POSIX_FADV_RANDOM:
4680 	case POSIX_FADV_NOREUSE:
4681 		/*
4682 		 * Try to merge any existing non-standard region with
4683 		 * this new region if possible, otherwise create a new
4684 		 * non-standard region for this request.
4685 		 */
4686 		mtx_pool_lock(mtxpool_sleep, fp);
4687 		fa = fp->f_advice;
4688 		if (fa != NULL && fa->fa_advice == advice &&
4689 		    ((fa->fa_start <= end && fa->fa_end >= offset) ||
4690 		    (end != OFF_MAX && fa->fa_start == end + 1) ||
4691 		    (fa->fa_end != OFF_MAX && fa->fa_end + 1 == offset))) {
4692 			if (offset < fa->fa_start)
4693 				fa->fa_start = offset;
4694 			if (end > fa->fa_end)
4695 				fa->fa_end = end;
4696 		} else {
4697 			new->fa_advice = advice;
4698 			new->fa_start = offset;
4699 			new->fa_end = end;
4700 			fp->f_advice = new;
4701 			new = fa;
4702 		}
4703 		mtx_pool_unlock(mtxpool_sleep, fp);
4704 		break;
4705 	case POSIX_FADV_NORMAL:
4706 		/*
4707 		 * If a the "normal" region overlaps with an existing
4708 		 * non-standard region, trim or remove the
4709 		 * non-standard region.
4710 		 */
4711 		mtx_pool_lock(mtxpool_sleep, fp);
4712 		fa = fp->f_advice;
4713 		if (fa != NULL) {
4714 			if (offset <= fa->fa_start && end >= fa->fa_end) {
4715 				new = fa;
4716 				fp->f_advice = NULL;
4717 			} else if (offset <= fa->fa_start &&
4718 			    end >= fa->fa_start)
4719 				fa->fa_start = end + 1;
4720 			else if (offset <= fa->fa_end && end >= fa->fa_end)
4721 				fa->fa_end = offset - 1;
4722 			else if (offset >= fa->fa_start && end <= fa->fa_end) {
4723 				/*
4724 				 * If the "normal" region is a middle
4725 				 * portion of the existing
4726 				 * non-standard region, just remove
4727 				 * the whole thing rather than picking
4728 				 * one side or the other to
4729 				 * preserve.
4730 				 */
4731 				new = fa;
4732 				fp->f_advice = NULL;
4733 			}
4734 		}
4735 		mtx_pool_unlock(mtxpool_sleep, fp);
4736 		break;
4737 	case POSIX_FADV_WILLNEED:
4738 	case POSIX_FADV_DONTNEED:
4739 		error = VOP_ADVISE(vp, offset, end, advice);
4740 		break;
4741 	}
4742 out:
4743 	if (fp != NULL)
4744 		fdrop(fp, td);
4745 	free(new, M_FADVISE);
4746 	return (error);
4747 }
4748 
4749 int
4750 sys_posix_fadvise(struct thread *td, struct posix_fadvise_args *uap)
4751 {
4752 	int error;
4753 
4754 	error = kern_posix_fadvise(td, uap->fd, uap->offset, uap->len,
4755 	    uap->advice);
4756 	return (kern_posix_error(td, error));
4757 }
4758 
4759 int
4760 kern_copy_file_range(struct thread *td, int infd, off_t *inoffp, int outfd,
4761     off_t *outoffp, size_t len, unsigned int flags)
4762 {
4763 	struct file *infp, *outfp;
4764 	struct vnode *invp, *outvp;
4765 	int error;
4766 	size_t retlen;
4767 	void *rl_rcookie, *rl_wcookie;
4768 	off_t savinoff, savoutoff;
4769 
4770 	infp = outfp = NULL;
4771 	rl_rcookie = rl_wcookie = NULL;
4772 	savinoff = -1;
4773 	error = 0;
4774 	retlen = 0;
4775 
4776 	if (flags != 0) {
4777 		error = EINVAL;
4778 		goto out;
4779 	}
4780 	if (len > SSIZE_MAX)
4781 		/*
4782 		 * Although the len argument is size_t, the return argument
4783 		 * is ssize_t (which is signed).  Therefore a size that won't
4784 		 * fit in ssize_t can't be returned.
4785 		 */
4786 		len = SSIZE_MAX;
4787 
4788 	/* Get the file structures for the file descriptors. */
4789 	error = fget_read(td, infd, &cap_read_rights, &infp);
4790 	if (error != 0)
4791 		goto out;
4792 	if (infp->f_ops == &badfileops) {
4793 		error = EBADF;
4794 		goto out;
4795 	}
4796 	if (infp->f_vnode == NULL) {
4797 		error = EINVAL;
4798 		goto out;
4799 	}
4800 	error = fget_write(td, outfd, &cap_write_rights, &outfp);
4801 	if (error != 0)
4802 		goto out;
4803 	if (outfp->f_ops == &badfileops) {
4804 		error = EBADF;
4805 		goto out;
4806 	}
4807 	if (outfp->f_vnode == NULL) {
4808 		error = EINVAL;
4809 		goto out;
4810 	}
4811 
4812 	/* Set the offset pointers to the correct place. */
4813 	if (inoffp == NULL)
4814 		inoffp = &infp->f_offset;
4815 	if (outoffp == NULL)
4816 		outoffp = &outfp->f_offset;
4817 	savinoff = *inoffp;
4818 	savoutoff = *outoffp;
4819 
4820 	invp = infp->f_vnode;
4821 	outvp = outfp->f_vnode;
4822 	/* Sanity check the f_flag bits. */
4823 	if ((outfp->f_flag & (FWRITE | FAPPEND)) != FWRITE ||
4824 	    (infp->f_flag & FREAD) == 0) {
4825 		error = EBADF;
4826 		goto out;
4827 	}
4828 
4829 	/* If len == 0, just return 0. */
4830 	if (len == 0)
4831 		goto out;
4832 
4833 	/*
4834 	 * If infp and outfp refer to the same file, the byte ranges cannot
4835 	 * overlap.
4836 	 */
4837 	if (invp == outvp && ((savinoff <= savoutoff && savinoff + len >
4838 	    savoutoff) || (savinoff > savoutoff && savoutoff + len >
4839 	    savinoff))) {
4840 		error = EINVAL;
4841 		goto out;
4842 	}
4843 
4844 	/* Range lock the byte ranges for both invp and outvp. */
4845 	for (;;) {
4846 		rl_wcookie = vn_rangelock_wlock(outvp, *outoffp, *outoffp +
4847 		    len);
4848 		rl_rcookie = vn_rangelock_tryrlock(invp, *inoffp, *inoffp +
4849 		    len);
4850 		if (rl_rcookie != NULL)
4851 			break;
4852 		vn_rangelock_unlock(outvp, rl_wcookie);
4853 		rl_rcookie = vn_rangelock_rlock(invp, *inoffp, *inoffp + len);
4854 		vn_rangelock_unlock(invp, rl_rcookie);
4855 	}
4856 
4857 	retlen = len;
4858 	error = vn_copy_file_range(invp, inoffp, outvp, outoffp, &retlen,
4859 	    flags, infp->f_cred, outfp->f_cred, td);
4860 out:
4861 	if (rl_rcookie != NULL)
4862 		vn_rangelock_unlock(invp, rl_rcookie);
4863 	if (rl_wcookie != NULL)
4864 		vn_rangelock_unlock(outvp, rl_wcookie);
4865 	if (savinoff != -1 && (error == EINTR || error == ERESTART)) {
4866 		*inoffp = savinoff;
4867 		*outoffp = savoutoff;
4868 	}
4869 	if (outfp != NULL)
4870 		fdrop(outfp, td);
4871 	if (infp != NULL)
4872 		fdrop(infp, td);
4873 	td->td_retval[0] = retlen;
4874 	return (error);
4875 }
4876 
4877 int
4878 sys_copy_file_range(struct thread *td, struct copy_file_range_args *uap)
4879 {
4880 	off_t inoff, outoff, *inoffp, *outoffp;
4881 	int error;
4882 
4883 	inoffp = outoffp = NULL;
4884 	if (uap->inoffp != NULL) {
4885 		error = copyin(uap->inoffp, &inoff, sizeof(off_t));
4886 		if (error != 0)
4887 			return (error);
4888 		inoffp = &inoff;
4889 	}
4890 	if (uap->outoffp != NULL) {
4891 		error = copyin(uap->outoffp, &outoff, sizeof(off_t));
4892 		if (error != 0)
4893 			return (error);
4894 		outoffp = &outoff;
4895 	}
4896 	error = kern_copy_file_range(td, uap->infd, inoffp, uap->outfd,
4897 	    outoffp, uap->len, uap->flags);
4898 	if (error == 0 && uap->inoffp != NULL)
4899 		error = copyout(inoffp, uap->inoffp, sizeof(off_t));
4900 	if (error == 0 && uap->outoffp != NULL)
4901 		error = copyout(outoffp, uap->outoffp, sizeof(off_t));
4902 	return (error);
4903 }
4904