xref: /freebsd/sys/kern/vfs_syscalls.c (revision d8462732ce74ccaf077dac625168cfa75f4a3b7e)
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_readlink_vp(struct vnode *vp, char *buf, enum uio_seg bufseg,
109     size_t count, struct thread *td);
110 static int kern_linkat_vp(struct thread *td, struct vnode *vp, int fd,
111     const char *path, enum uio_seg segflag);
112 
113 static uint64_t
114 at2cnpflags(u_int at_flags, u_int mask)
115 {
116 	u_int64_t res;
117 
118 	MPASS((at_flags & (AT_SYMLINK_FOLLOW | AT_SYMLINK_NOFOLLOW)) !=
119 	    (AT_SYMLINK_FOLLOW | AT_SYMLINK_NOFOLLOW));
120 
121 	res = 0;
122 	at_flags &= mask;
123 	if ((at_flags & AT_BENEATH) != 0)
124 		res |= BENEATH;
125 	if ((at_flags & AT_RESOLVE_BENEATH) != 0)
126 		res |= RBENEATH;
127 	if ((at_flags & AT_SYMLINK_FOLLOW) != 0)
128 		res |= FOLLOW;
129 	/* NOFOLLOW is pseudo flag */
130 	if ((mask & AT_SYMLINK_NOFOLLOW) != 0) {
131 		res |= (at_flags & AT_SYMLINK_NOFOLLOW) != 0 ? NOFOLLOW :
132 		    FOLLOW;
133 	}
134 	return (res);
135 }
136 
137 int
138 kern_sync(struct thread *td)
139 {
140 	struct mount *mp, *nmp;
141 	int save;
142 
143 	mtx_lock(&mountlist_mtx);
144 	for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
145 		if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK)) {
146 			nmp = TAILQ_NEXT(mp, mnt_list);
147 			continue;
148 		}
149 		if ((mp->mnt_flag & MNT_RDONLY) == 0 &&
150 		    vn_start_write(NULL, &mp, V_NOWAIT) == 0) {
151 			save = curthread_pflags_set(TDP_SYNCIO);
152 			vfs_periodic(mp, MNT_NOWAIT);
153 			VFS_SYNC(mp, MNT_NOWAIT);
154 			curthread_pflags_restore(save);
155 			vn_finished_write(mp);
156 		}
157 		mtx_lock(&mountlist_mtx);
158 		nmp = TAILQ_NEXT(mp, mnt_list);
159 		vfs_unbusy(mp);
160 	}
161 	mtx_unlock(&mountlist_mtx);
162 	return (0);
163 }
164 
165 /*
166  * Sync each mounted filesystem.
167  */
168 #ifndef _SYS_SYSPROTO_H_
169 struct sync_args {
170 	int     dummy;
171 };
172 #endif
173 /* ARGSUSED */
174 int
175 sys_sync(struct thread *td, struct sync_args *uap)
176 {
177 
178 	return (kern_sync(td));
179 }
180 
181 /*
182  * Change filesystem quotas.
183  */
184 #ifndef _SYS_SYSPROTO_H_
185 struct quotactl_args {
186 	char *path;
187 	int cmd;
188 	int uid;
189 	caddr_t arg;
190 };
191 #endif
192 int
193 sys_quotactl(struct thread *td, struct quotactl_args *uap)
194 {
195 	struct mount *mp;
196 	struct nameidata nd;
197 	int error;
198 
199 	AUDIT_ARG_CMD(uap->cmd);
200 	AUDIT_ARG_UID(uap->uid);
201 	if (!prison_allow(td->td_ucred, PR_ALLOW_QUOTAS))
202 		return (EPERM);
203 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
204 	    uap->path, td);
205 	if ((error = namei(&nd)) != 0)
206 		return (error);
207 	NDFREE(&nd, NDF_ONLY_PNBUF);
208 	mp = nd.ni_vp->v_mount;
209 	vfs_ref(mp);
210 	vput(nd.ni_vp);
211 	error = vfs_busy(mp, 0);
212 	if (error != 0) {
213 		vfs_rel(mp);
214 		return (error);
215 	}
216 	error = VFS_QUOTACTL(mp, uap->cmd, uap->uid, uap->arg);
217 
218 	/*
219 	 * Since quota on operation typically needs to open quota
220 	 * file, the Q_QUOTAON handler needs to unbusy the mount point
221 	 * before calling into namei.  Otherwise, unmount might be
222 	 * started between two vfs_busy() invocations (first is our,
223 	 * second is from mount point cross-walk code in lookup()),
224 	 * causing deadlock.
225 	 *
226 	 * Require that Q_QUOTAON handles the vfs_busy() reference on
227 	 * its own, always returning with ubusied mount point.
228 	 */
229 	if ((uap->cmd >> SUBCMDSHIFT) != Q_QUOTAON &&
230 	    (uap->cmd >> SUBCMDSHIFT) != Q_QUOTAOFF)
231 		vfs_unbusy(mp);
232 	vfs_rel(mp);
233 	return (error);
234 }
235 
236 /*
237  * Used by statfs conversion routines to scale the block size up if
238  * necessary so that all of the block counts are <= 'max_size'.  Note
239  * that 'max_size' should be a bitmask, i.e. 2^n - 1 for some non-zero
240  * value of 'n'.
241  */
242 void
243 statfs_scale_blocks(struct statfs *sf, long max_size)
244 {
245 	uint64_t count;
246 	int shift;
247 
248 	KASSERT(powerof2(max_size + 1), ("%s: invalid max_size", __func__));
249 
250 	/*
251 	 * Attempt to scale the block counts to give a more accurate
252 	 * overview to userland of the ratio of free space to used
253 	 * space.  To do this, find the largest block count and compute
254 	 * a divisor that lets it fit into a signed integer <= max_size.
255 	 */
256 	if (sf->f_bavail < 0)
257 		count = -sf->f_bavail;
258 	else
259 		count = sf->f_bavail;
260 	count = MAX(sf->f_blocks, MAX(sf->f_bfree, count));
261 	if (count <= max_size)
262 		return;
263 
264 	count >>= flsl(max_size);
265 	shift = 0;
266 	while (count > 0) {
267 		shift++;
268 		count >>=1;
269 	}
270 
271 	sf->f_bsize <<= shift;
272 	sf->f_blocks >>= shift;
273 	sf->f_bfree >>= shift;
274 	sf->f_bavail >>= shift;
275 }
276 
277 static int
278 kern_do_statfs(struct thread *td, struct mount *mp, struct statfs *buf)
279 {
280 	int error;
281 
282 	if (mp == NULL)
283 		return (EBADF);
284 	error = vfs_busy(mp, 0);
285 	vfs_rel(mp);
286 	if (error != 0)
287 		return (error);
288 #ifdef MAC
289 	error = mac_mount_check_stat(td->td_ucred, mp);
290 	if (error != 0)
291 		goto out;
292 #endif
293 	error = VFS_STATFS(mp, buf);
294 	if (error != 0)
295 		goto out;
296 	if (priv_check_cred_vfs_generation(td->td_ucred)) {
297 		buf->f_fsid.val[0] = buf->f_fsid.val[1] = 0;
298 		prison_enforce_statfs(td->td_ucred, mp, buf);
299 	}
300 out:
301 	vfs_unbusy(mp);
302 	return (error);
303 }
304 
305 /*
306  * Get filesystem statistics.
307  */
308 #ifndef _SYS_SYSPROTO_H_
309 struct statfs_args {
310 	char *path;
311 	struct statfs *buf;
312 };
313 #endif
314 int
315 sys_statfs(struct thread *td, struct statfs_args *uap)
316 {
317 	struct statfs *sfp;
318 	int error;
319 
320 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
321 	error = kern_statfs(td, uap->path, UIO_USERSPACE, sfp);
322 	if (error == 0)
323 		error = copyout(sfp, uap->buf, sizeof(struct statfs));
324 	free(sfp, M_STATFS);
325 	return (error);
326 }
327 
328 int
329 kern_statfs(struct thread *td, const char *path, enum uio_seg pathseg,
330     struct statfs *buf)
331 {
332 	struct mount *mp;
333 	struct nameidata nd;
334 	int error;
335 
336 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
337 	    pathseg, path, td);
338 	error = namei(&nd);
339 	if (error != 0)
340 		return (error);
341 	mp = nd.ni_vp->v_mount;
342 	vfs_ref(mp);
343 	NDFREE(&nd, NDF_ONLY_PNBUF);
344 	vput(nd.ni_vp);
345 	return (kern_do_statfs(td, mp, buf));
346 }
347 
348 /*
349  * Get filesystem statistics.
350  */
351 #ifndef _SYS_SYSPROTO_H_
352 struct fstatfs_args {
353 	int fd;
354 	struct statfs *buf;
355 };
356 #endif
357 int
358 sys_fstatfs(struct thread *td, struct fstatfs_args *uap)
359 {
360 	struct statfs *sfp;
361 	int error;
362 
363 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
364 	error = kern_fstatfs(td, uap->fd, sfp);
365 	if (error == 0)
366 		error = copyout(sfp, uap->buf, sizeof(struct statfs));
367 	free(sfp, M_STATFS);
368 	return (error);
369 }
370 
371 int
372 kern_fstatfs(struct thread *td, int fd, struct statfs *buf)
373 {
374 	struct file *fp;
375 	struct mount *mp;
376 	struct vnode *vp;
377 	int error;
378 
379 	AUDIT_ARG_FD(fd);
380 	error = getvnode(td, fd, &cap_fstatfs_rights, &fp);
381 	if (error != 0)
382 		return (error);
383 	vp = fp->f_vnode;
384 	vn_lock(vp, LK_SHARED | LK_RETRY);
385 #ifdef AUDIT
386 	AUDIT_ARG_VNODE1(vp);
387 #endif
388 	mp = vp->v_mount;
389 	if (mp != NULL)
390 		vfs_ref(mp);
391 	VOP_UNLOCK(vp);
392 	fdrop(fp, td);
393 	return (kern_do_statfs(td, mp, buf));
394 }
395 
396 /*
397  * Get statistics on all filesystems.
398  */
399 #ifndef _SYS_SYSPROTO_H_
400 struct getfsstat_args {
401 	struct statfs *buf;
402 	long bufsize;
403 	int mode;
404 };
405 #endif
406 int
407 sys_getfsstat(struct thread *td, struct getfsstat_args *uap)
408 {
409 	size_t count;
410 	int error;
411 
412 	if (uap->bufsize < 0 || uap->bufsize > SIZE_MAX)
413 		return (EINVAL);
414 	error = kern_getfsstat(td, &uap->buf, uap->bufsize, &count,
415 	    UIO_USERSPACE, uap->mode);
416 	if (error == 0)
417 		td->td_retval[0] = count;
418 	return (error);
419 }
420 
421 /*
422  * If (bufsize > 0 && bufseg == UIO_SYSSPACE)
423  *	The caller is responsible for freeing memory which will be allocated
424  *	in '*buf'.
425  */
426 int
427 kern_getfsstat(struct thread *td, struct statfs **buf, size_t bufsize,
428     size_t *countp, enum uio_seg bufseg, int mode)
429 {
430 	struct mount *mp, *nmp;
431 	struct statfs *sfsp, *sp, *sptmp, *tofree;
432 	size_t count, maxcount;
433 	int error;
434 
435 	switch (mode) {
436 	case MNT_WAIT:
437 	case MNT_NOWAIT:
438 		break;
439 	default:
440 		if (bufseg == UIO_SYSSPACE)
441 			*buf = NULL;
442 		return (EINVAL);
443 	}
444 restart:
445 	maxcount = bufsize / sizeof(struct statfs);
446 	if (bufsize == 0) {
447 		sfsp = NULL;
448 		tofree = NULL;
449 	} else if (bufseg == UIO_USERSPACE) {
450 		sfsp = *buf;
451 		tofree = NULL;
452 	} else /* if (bufseg == UIO_SYSSPACE) */ {
453 		count = 0;
454 		mtx_lock(&mountlist_mtx);
455 		TAILQ_FOREACH(mp, &mountlist, mnt_list) {
456 			count++;
457 		}
458 		mtx_unlock(&mountlist_mtx);
459 		if (maxcount > count)
460 			maxcount = count;
461 		tofree = sfsp = *buf = malloc(maxcount * sizeof(struct statfs),
462 		    M_STATFS, M_WAITOK);
463 	}
464 
465 	count = 0;
466 
467 	/*
468 	 * If there is no target buffer they only want the count.
469 	 *
470 	 * This could be TAILQ_FOREACH but it is open-coded to match the original
471 	 * code below.
472 	 */
473 	if (sfsp == NULL) {
474 		mtx_lock(&mountlist_mtx);
475 		for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
476 			if (prison_canseemount(td->td_ucred, mp) != 0) {
477 				nmp = TAILQ_NEXT(mp, mnt_list);
478 				continue;
479 			}
480 #ifdef MAC
481 			if (mac_mount_check_stat(td->td_ucred, mp) != 0) {
482 				nmp = TAILQ_NEXT(mp, mnt_list);
483 				continue;
484 			}
485 #endif
486 			count++;
487 			nmp = TAILQ_NEXT(mp, mnt_list);
488 		}
489 		mtx_unlock(&mountlist_mtx);
490 		*countp = count;
491 		return (0);
492 	}
493 
494 	/*
495 	 * They want the entire thing.
496 	 *
497 	 * Short-circuit the corner case of no room for anything, avoids
498 	 * relocking below.
499 	 */
500 	if (maxcount < 1) {
501 		goto out;
502 	}
503 
504 	mtx_lock(&mountlist_mtx);
505 	for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
506 		if (prison_canseemount(td->td_ucred, mp) != 0) {
507 			nmp = TAILQ_NEXT(mp, mnt_list);
508 			continue;
509 		}
510 #ifdef MAC
511 		if (mac_mount_check_stat(td->td_ucred, mp) != 0) {
512 			nmp = TAILQ_NEXT(mp, mnt_list);
513 			continue;
514 		}
515 #endif
516 		if (mode == MNT_WAIT) {
517 			if (vfs_busy(mp, MBF_MNTLSTLOCK) != 0) {
518 				/*
519 				 * If vfs_busy() failed, and MBF_NOWAIT
520 				 * wasn't passed, then the mp is gone.
521 				 * Furthermore, because of MBF_MNTLSTLOCK,
522 				 * the mountlist_mtx was dropped.  We have
523 				 * no other choice than to start over.
524 				 */
525 				mtx_unlock(&mountlist_mtx);
526 				free(tofree, M_STATFS);
527 				goto restart;
528 			}
529 		} else {
530 			if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK) != 0) {
531 				nmp = TAILQ_NEXT(mp, mnt_list);
532 				continue;
533 			}
534 		}
535 		sp = &mp->mnt_stat;
536 		/*
537 		 * If MNT_NOWAIT is specified, do not refresh
538 		 * the fsstat cache.
539 		 */
540 		if (mode != MNT_NOWAIT) {
541 			error = VFS_STATFS(mp, sp);
542 			if (error != 0) {
543 				mtx_lock(&mountlist_mtx);
544 				nmp = TAILQ_NEXT(mp, mnt_list);
545 				vfs_unbusy(mp);
546 				continue;
547 			}
548 		}
549 		if (priv_check_cred_vfs_generation(td->td_ucred)) {
550 			sptmp = malloc(sizeof(struct statfs), M_STATFS,
551 			    M_WAITOK);
552 			*sptmp = *sp;
553 			sptmp->f_fsid.val[0] = sptmp->f_fsid.val[1] = 0;
554 			prison_enforce_statfs(td->td_ucred, mp, sptmp);
555 			sp = sptmp;
556 		} else
557 			sptmp = NULL;
558 		if (bufseg == UIO_SYSSPACE) {
559 			bcopy(sp, sfsp, sizeof(*sp));
560 			free(sptmp, M_STATFS);
561 		} else /* if (bufseg == UIO_USERSPACE) */ {
562 			error = copyout(sp, sfsp, sizeof(*sp));
563 			free(sptmp, M_STATFS);
564 			if (error != 0) {
565 				vfs_unbusy(mp);
566 				return (error);
567 			}
568 		}
569 		sfsp++;
570 		count++;
571 
572 		if (count == maxcount) {
573 			vfs_unbusy(mp);
574 			goto out;
575 		}
576 
577 		mtx_lock(&mountlist_mtx);
578 		nmp = TAILQ_NEXT(mp, mnt_list);
579 		vfs_unbusy(mp);
580 	}
581 	mtx_unlock(&mountlist_mtx);
582 out:
583 	*countp = count;
584 	return (0);
585 }
586 
587 #ifdef COMPAT_FREEBSD4
588 /*
589  * Get old format filesystem statistics.
590  */
591 static void freebsd4_cvtstatfs(struct statfs *, struct ostatfs *);
592 
593 #ifndef _SYS_SYSPROTO_H_
594 struct freebsd4_statfs_args {
595 	char *path;
596 	struct ostatfs *buf;
597 };
598 #endif
599 int
600 freebsd4_statfs(struct thread *td, struct freebsd4_statfs_args *uap)
601 {
602 	struct ostatfs osb;
603 	struct statfs *sfp;
604 	int error;
605 
606 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
607 	error = kern_statfs(td, uap->path, UIO_USERSPACE, sfp);
608 	if (error == 0) {
609 		freebsd4_cvtstatfs(sfp, &osb);
610 		error = copyout(&osb, uap->buf, sizeof(osb));
611 	}
612 	free(sfp, M_STATFS);
613 	return (error);
614 }
615 
616 /*
617  * Get filesystem statistics.
618  */
619 #ifndef _SYS_SYSPROTO_H_
620 struct freebsd4_fstatfs_args {
621 	int fd;
622 	struct ostatfs *buf;
623 };
624 #endif
625 int
626 freebsd4_fstatfs(struct thread *td, struct freebsd4_fstatfs_args *uap)
627 {
628 	struct ostatfs osb;
629 	struct statfs *sfp;
630 	int error;
631 
632 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
633 	error = kern_fstatfs(td, uap->fd, sfp);
634 	if (error == 0) {
635 		freebsd4_cvtstatfs(sfp, &osb);
636 		error = copyout(&osb, uap->buf, sizeof(osb));
637 	}
638 	free(sfp, M_STATFS);
639 	return (error);
640 }
641 
642 /*
643  * Get statistics on all filesystems.
644  */
645 #ifndef _SYS_SYSPROTO_H_
646 struct freebsd4_getfsstat_args {
647 	struct ostatfs *buf;
648 	long bufsize;
649 	int mode;
650 };
651 #endif
652 int
653 freebsd4_getfsstat(struct thread *td, struct freebsd4_getfsstat_args *uap)
654 {
655 	struct statfs *buf, *sp;
656 	struct ostatfs osb;
657 	size_t count, size;
658 	int error;
659 
660 	if (uap->bufsize < 0)
661 		return (EINVAL);
662 	count = uap->bufsize / sizeof(struct ostatfs);
663 	if (count > SIZE_MAX / sizeof(struct statfs))
664 		return (EINVAL);
665 	size = count * sizeof(struct statfs);
666 	error = kern_getfsstat(td, &buf, size, &count, UIO_SYSSPACE,
667 	    uap->mode);
668 	if (error == 0)
669 		td->td_retval[0] = count;
670 	if (size != 0) {
671 		sp = buf;
672 		while (count != 0 && error == 0) {
673 			freebsd4_cvtstatfs(sp, &osb);
674 			error = copyout(&osb, uap->buf, sizeof(osb));
675 			sp++;
676 			uap->buf++;
677 			count--;
678 		}
679 		free(buf, M_STATFS);
680 	}
681 	return (error);
682 }
683 
684 /*
685  * Implement fstatfs() for (NFS) file handles.
686  */
687 #ifndef _SYS_SYSPROTO_H_
688 struct freebsd4_fhstatfs_args {
689 	struct fhandle *u_fhp;
690 	struct ostatfs *buf;
691 };
692 #endif
693 int
694 freebsd4_fhstatfs(struct thread *td, struct freebsd4_fhstatfs_args *uap)
695 {
696 	struct ostatfs osb;
697 	struct statfs *sfp;
698 	fhandle_t fh;
699 	int error;
700 
701 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
702 	if (error != 0)
703 		return (error);
704 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
705 	error = kern_fhstatfs(td, fh, sfp);
706 	if (error == 0) {
707 		freebsd4_cvtstatfs(sfp, &osb);
708 		error = copyout(&osb, uap->buf, sizeof(osb));
709 	}
710 	free(sfp, M_STATFS);
711 	return (error);
712 }
713 
714 /*
715  * Convert a new format statfs structure to an old format statfs structure.
716  */
717 static void
718 freebsd4_cvtstatfs(struct statfs *nsp, struct ostatfs *osp)
719 {
720 
721 	statfs_scale_blocks(nsp, LONG_MAX);
722 	bzero(osp, sizeof(*osp));
723 	osp->f_bsize = nsp->f_bsize;
724 	osp->f_iosize = MIN(nsp->f_iosize, LONG_MAX);
725 	osp->f_blocks = nsp->f_blocks;
726 	osp->f_bfree = nsp->f_bfree;
727 	osp->f_bavail = nsp->f_bavail;
728 	osp->f_files = MIN(nsp->f_files, LONG_MAX);
729 	osp->f_ffree = MIN(nsp->f_ffree, LONG_MAX);
730 	osp->f_owner = nsp->f_owner;
731 	osp->f_type = nsp->f_type;
732 	osp->f_flags = nsp->f_flags;
733 	osp->f_syncwrites = MIN(nsp->f_syncwrites, LONG_MAX);
734 	osp->f_asyncwrites = MIN(nsp->f_asyncwrites, LONG_MAX);
735 	osp->f_syncreads = MIN(nsp->f_syncreads, LONG_MAX);
736 	osp->f_asyncreads = MIN(nsp->f_asyncreads, LONG_MAX);
737 	strlcpy(osp->f_fstypename, nsp->f_fstypename,
738 	    MIN(MFSNAMELEN, OMFSNAMELEN));
739 	strlcpy(osp->f_mntonname, nsp->f_mntonname,
740 	    MIN(MNAMELEN, OMNAMELEN));
741 	strlcpy(osp->f_mntfromname, nsp->f_mntfromname,
742 	    MIN(MNAMELEN, OMNAMELEN));
743 	osp->f_fsid = nsp->f_fsid;
744 }
745 #endif /* COMPAT_FREEBSD4 */
746 
747 #if defined(COMPAT_FREEBSD11)
748 /*
749  * Get old format filesystem statistics.
750  */
751 static void freebsd11_cvtstatfs(struct statfs *, struct freebsd11_statfs *);
752 
753 int
754 freebsd11_statfs(struct thread *td, struct freebsd11_statfs_args *uap)
755 {
756 	struct freebsd11_statfs osb;
757 	struct statfs *sfp;
758 	int error;
759 
760 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
761 	error = kern_statfs(td, uap->path, UIO_USERSPACE, sfp);
762 	if (error == 0) {
763 		freebsd11_cvtstatfs(sfp, &osb);
764 		error = copyout(&osb, uap->buf, sizeof(osb));
765 	}
766 	free(sfp, M_STATFS);
767 	return (error);
768 }
769 
770 /*
771  * Get filesystem statistics.
772  */
773 int
774 freebsd11_fstatfs(struct thread *td, struct freebsd11_fstatfs_args *uap)
775 {
776 	struct freebsd11_statfs osb;
777 	struct statfs *sfp;
778 	int error;
779 
780 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
781 	error = kern_fstatfs(td, uap->fd, sfp);
782 	if (error == 0) {
783 		freebsd11_cvtstatfs(sfp, &osb);
784 		error = copyout(&osb, uap->buf, sizeof(osb));
785 	}
786 	free(sfp, M_STATFS);
787 	return (error);
788 }
789 
790 /*
791  * Get statistics on all filesystems.
792  */
793 int
794 freebsd11_getfsstat(struct thread *td, struct freebsd11_getfsstat_args *uap)
795 {
796 	struct freebsd11_statfs osb;
797 	struct statfs *buf, *sp;
798 	size_t count, size;
799 	int error;
800 
801 	count = uap->bufsize / sizeof(struct ostatfs);
802 	size = count * sizeof(struct statfs);
803 	error = kern_getfsstat(td, &buf, size, &count, UIO_SYSSPACE,
804 	    uap->mode);
805 	if (error == 0)
806 		td->td_retval[0] = count;
807 	if (size > 0) {
808 		sp = buf;
809 		while (count > 0 && error == 0) {
810 			freebsd11_cvtstatfs(sp, &osb);
811 			error = copyout(&osb, uap->buf, sizeof(osb));
812 			sp++;
813 			uap->buf++;
814 			count--;
815 		}
816 		free(buf, M_STATFS);
817 	}
818 	return (error);
819 }
820 
821 /*
822  * Implement fstatfs() for (NFS) file handles.
823  */
824 int
825 freebsd11_fhstatfs(struct thread *td, struct freebsd11_fhstatfs_args *uap)
826 {
827 	struct freebsd11_statfs osb;
828 	struct statfs *sfp;
829 	fhandle_t fh;
830 	int error;
831 
832 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
833 	if (error)
834 		return (error);
835 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
836 	error = kern_fhstatfs(td, fh, sfp);
837 	if (error == 0) {
838 		freebsd11_cvtstatfs(sfp, &osb);
839 		error = copyout(&osb, uap->buf, sizeof(osb));
840 	}
841 	free(sfp, M_STATFS);
842 	return (error);
843 }
844 
845 /*
846  * Convert a new format statfs structure to an old format statfs structure.
847  */
848 static void
849 freebsd11_cvtstatfs(struct statfs *nsp, struct freebsd11_statfs *osp)
850 {
851 
852 	bzero(osp, sizeof(*osp));
853 	osp->f_version = FREEBSD11_STATFS_VERSION;
854 	osp->f_type = nsp->f_type;
855 	osp->f_flags = nsp->f_flags;
856 	osp->f_bsize = nsp->f_bsize;
857 	osp->f_iosize = nsp->f_iosize;
858 	osp->f_blocks = nsp->f_blocks;
859 	osp->f_bfree = nsp->f_bfree;
860 	osp->f_bavail = nsp->f_bavail;
861 	osp->f_files = nsp->f_files;
862 	osp->f_ffree = nsp->f_ffree;
863 	osp->f_syncwrites = nsp->f_syncwrites;
864 	osp->f_asyncwrites = nsp->f_asyncwrites;
865 	osp->f_syncreads = nsp->f_syncreads;
866 	osp->f_asyncreads = nsp->f_asyncreads;
867 	osp->f_namemax = nsp->f_namemax;
868 	osp->f_owner = nsp->f_owner;
869 	osp->f_fsid = nsp->f_fsid;
870 	strlcpy(osp->f_fstypename, nsp->f_fstypename,
871 	    MIN(MFSNAMELEN, sizeof(osp->f_fstypename)));
872 	strlcpy(osp->f_mntonname, nsp->f_mntonname,
873 	    MIN(MNAMELEN, sizeof(osp->f_mntonname)));
874 	strlcpy(osp->f_mntfromname, nsp->f_mntfromname,
875 	    MIN(MNAMELEN, sizeof(osp->f_mntfromname)));
876 }
877 #endif /* COMPAT_FREEBSD11 */
878 
879 /*
880  * Change current working directory to a given file descriptor.
881  */
882 #ifndef _SYS_SYSPROTO_H_
883 struct fchdir_args {
884 	int	fd;
885 };
886 #endif
887 int
888 sys_fchdir(struct thread *td, struct fchdir_args *uap)
889 {
890 	struct vnode *vp, *tdp;
891 	struct mount *mp;
892 	struct file *fp;
893 	int error;
894 
895 	AUDIT_ARG_FD(uap->fd);
896 	error = getvnode(td, uap->fd, &cap_fchdir_rights,
897 	    &fp);
898 	if (error != 0)
899 		return (error);
900 	vp = fp->f_vnode;
901 	vrefact(vp);
902 	fdrop(fp, td);
903 	vn_lock(vp, LK_SHARED | LK_RETRY);
904 	AUDIT_ARG_VNODE1(vp);
905 	error = change_dir(vp, td);
906 	while (!error && (mp = vp->v_mountedhere) != NULL) {
907 		if (vfs_busy(mp, 0))
908 			continue;
909 		error = VFS_ROOT(mp, LK_SHARED, &tdp);
910 		vfs_unbusy(mp);
911 		if (error != 0)
912 			break;
913 		vput(vp);
914 		vp = tdp;
915 	}
916 	if (error != 0) {
917 		vput(vp);
918 		return (error);
919 	}
920 	VOP_UNLOCK(vp);
921 	pwd_chdir(td, vp);
922 	return (0);
923 }
924 
925 /*
926  * Change current working directory (``.'').
927  */
928 #ifndef _SYS_SYSPROTO_H_
929 struct chdir_args {
930 	char	*path;
931 };
932 #endif
933 int
934 sys_chdir(struct thread *td, struct chdir_args *uap)
935 {
936 
937 	return (kern_chdir(td, uap->path, UIO_USERSPACE));
938 }
939 
940 int
941 kern_chdir(struct thread *td, const char *path, enum uio_seg pathseg)
942 {
943 	struct nameidata nd;
944 	int error;
945 
946 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
947 	    pathseg, path, td);
948 	if ((error = namei(&nd)) != 0)
949 		return (error);
950 	if ((error = change_dir(nd.ni_vp, td)) != 0) {
951 		vput(nd.ni_vp);
952 		NDFREE(&nd, NDF_ONLY_PNBUF);
953 		return (error);
954 	}
955 	VOP_UNLOCK(nd.ni_vp);
956 	NDFREE(&nd, NDF_ONLY_PNBUF);
957 	pwd_chdir(td, nd.ni_vp);
958 	return (0);
959 }
960 
961 /*
962  * Change notion of root (``/'') directory.
963  */
964 #ifndef _SYS_SYSPROTO_H_
965 struct chroot_args {
966 	char	*path;
967 };
968 #endif
969 int
970 sys_chroot(struct thread *td, struct chroot_args *uap)
971 {
972 	struct nameidata nd;
973 	int error;
974 
975 	error = priv_check(td, PRIV_VFS_CHROOT);
976 	if (error != 0)
977 		return (error);
978 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
979 	    UIO_USERSPACE, uap->path, td);
980 	error = namei(&nd);
981 	if (error != 0)
982 		goto error;
983 	error = change_dir(nd.ni_vp, td);
984 	if (error != 0)
985 		goto e_vunlock;
986 #ifdef MAC
987 	error = mac_vnode_check_chroot(td->td_ucred, nd.ni_vp);
988 	if (error != 0)
989 		goto e_vunlock;
990 #endif
991 	VOP_UNLOCK(nd.ni_vp);
992 	error = pwd_chroot(td, nd.ni_vp);
993 	vrele(nd.ni_vp);
994 	NDFREE(&nd, NDF_ONLY_PNBUF);
995 	return (error);
996 e_vunlock:
997 	vput(nd.ni_vp);
998 error:
999 	NDFREE(&nd, NDF_ONLY_PNBUF);
1000 	return (error);
1001 }
1002 
1003 /*
1004  * Common routine for chroot and chdir.  Callers must provide a locked vnode
1005  * instance.
1006  */
1007 int
1008 change_dir(struct vnode *vp, struct thread *td)
1009 {
1010 #ifdef MAC
1011 	int error;
1012 #endif
1013 
1014 	ASSERT_VOP_LOCKED(vp, "change_dir(): vp not locked");
1015 	if (vp->v_type != VDIR)
1016 		return (ENOTDIR);
1017 #ifdef MAC
1018 	error = mac_vnode_check_chdir(td->td_ucred, vp);
1019 	if (error != 0)
1020 		return (error);
1021 #endif
1022 	return (VOP_ACCESS(vp, VEXEC, td->td_ucred, td));
1023 }
1024 
1025 static __inline void
1026 flags_to_rights(int flags, cap_rights_t *rightsp)
1027 {
1028 
1029 	if (flags & O_EXEC) {
1030 		cap_rights_set_one(rightsp, CAP_FEXECVE);
1031 	} else {
1032 		switch ((flags & O_ACCMODE)) {
1033 		case O_RDONLY:
1034 			cap_rights_set_one(rightsp, CAP_READ);
1035 			break;
1036 		case O_RDWR:
1037 			cap_rights_set_one(rightsp, CAP_READ);
1038 			/* FALLTHROUGH */
1039 		case O_WRONLY:
1040 			cap_rights_set_one(rightsp, CAP_WRITE);
1041 			if (!(flags & (O_APPEND | O_TRUNC)))
1042 				cap_rights_set_one(rightsp, CAP_SEEK);
1043 			break;
1044 		}
1045 	}
1046 
1047 	if (flags & O_CREAT)
1048 		cap_rights_set_one(rightsp, CAP_CREATE);
1049 
1050 	if (flags & O_TRUNC)
1051 		cap_rights_set_one(rightsp, CAP_FTRUNCATE);
1052 
1053 	if (flags & (O_SYNC | O_FSYNC))
1054 		cap_rights_set_one(rightsp, CAP_FSYNC);
1055 
1056 	if (flags & (O_EXLOCK | O_SHLOCK))
1057 		cap_rights_set_one(rightsp, CAP_FLOCK);
1058 }
1059 
1060 /*
1061  * Check permissions, allocate an open file structure, and call the device
1062  * open routine if any.
1063  */
1064 #ifndef _SYS_SYSPROTO_H_
1065 struct open_args {
1066 	char	*path;
1067 	int	flags;
1068 	int	mode;
1069 };
1070 #endif
1071 int
1072 sys_open(struct thread *td, struct open_args *uap)
1073 {
1074 
1075 	return (kern_openat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1076 	    uap->flags, uap->mode));
1077 }
1078 
1079 #ifndef _SYS_SYSPROTO_H_
1080 struct openat_args {
1081 	int	fd;
1082 	char	*path;
1083 	int	flag;
1084 	int	mode;
1085 };
1086 #endif
1087 int
1088 sys_openat(struct thread *td, struct openat_args *uap)
1089 {
1090 
1091 	AUDIT_ARG_FD(uap->fd);
1092 	return (kern_openat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
1093 	    uap->mode));
1094 }
1095 
1096 int
1097 kern_openat(struct thread *td, int fd, const char *path, enum uio_seg pathseg,
1098     int flags, int mode)
1099 {
1100 	struct proc *p = td->td_proc;
1101 	struct filedesc *fdp;
1102 	struct pwddesc *pdp;
1103 	struct file *fp;
1104 	struct vnode *vp;
1105 	struct nameidata nd;
1106 	cap_rights_t rights;
1107 	int cmode, error, indx;
1108 
1109 	indx = -1;
1110 	fdp = p->p_fd;
1111 	pdp = p->p_pd;
1112 
1113 	AUDIT_ARG_FFLAGS(flags);
1114 	AUDIT_ARG_MODE(mode);
1115 	cap_rights_init_one(&rights, CAP_LOOKUP);
1116 	flags_to_rights(flags, &rights);
1117 	/*
1118 	 * Only one of the O_EXEC, O_RDONLY, O_WRONLY and O_RDWR flags
1119 	 * may be specified.
1120 	 */
1121 	if (flags & O_EXEC) {
1122 		if (flags & O_ACCMODE)
1123 			return (EINVAL);
1124 	} else if ((flags & O_ACCMODE) == O_ACCMODE) {
1125 		return (EINVAL);
1126 	} else {
1127 		flags = FFLAGS(flags);
1128 	}
1129 
1130 	/*
1131 	 * Allocate a file structure. The descriptor to reference it
1132 	 * is allocated and set by finstall() below.
1133 	 */
1134 	error = falloc_noinstall(td, &fp);
1135 	if (error != 0)
1136 		return (error);
1137 	/*
1138 	 * An extra reference on `fp' has been held for us by
1139 	 * falloc_noinstall().
1140 	 */
1141 	/* Set the flags early so the finit in devfs can pick them up. */
1142 	fp->f_flag = flags & FMASK;
1143 	cmode = ((mode & ~pdp->pd_cmask) & ALLPERMS) & ~S_ISTXT;
1144 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, fd,
1145 	    &rights, td);
1146 	td->td_dupfd = -1;		/* XXX check for fdopen */
1147 	error = vn_open(&nd, &flags, cmode, fp);
1148 	if (error != 0) {
1149 		/*
1150 		 * If the vn_open replaced the method vector, something
1151 		 * wonderous happened deep below and we just pass it up
1152 		 * pretending we know what we do.
1153 		 */
1154 		if (error == ENXIO && fp->f_ops != &badfileops)
1155 			goto success;
1156 
1157 		/*
1158 		 * Handle special fdopen() case. bleh.
1159 		 *
1160 		 * Don't do this for relative (capability) lookups; we don't
1161 		 * understand exactly what would happen, and we don't think
1162 		 * that it ever should.
1163 		 */
1164 		if ((nd.ni_resflags & NIRES_STRICTREL) == 0 &&
1165 		    (error == ENODEV || error == ENXIO) &&
1166 		    td->td_dupfd >= 0) {
1167 			error = dupfdopen(td, fdp, td->td_dupfd, flags, error,
1168 			    &indx);
1169 			if (error == 0)
1170 				goto success;
1171 		}
1172 
1173 		goto bad;
1174 	}
1175 	td->td_dupfd = 0;
1176 	NDFREE(&nd, NDF_ONLY_PNBUF);
1177 	vp = nd.ni_vp;
1178 
1179 	/*
1180 	 * Store the vnode, for any f_type. Typically, the vnode use
1181 	 * count is decremented by direct call to vn_closefile() for
1182 	 * files that switched type in the cdevsw fdopen() method.
1183 	 */
1184 	fp->f_vnode = vp;
1185 	/*
1186 	 * If the file wasn't claimed by devfs bind it to the normal
1187 	 * vnode operations here.
1188 	 */
1189 	if (fp->f_ops == &badfileops) {
1190 		KASSERT(vp->v_type != VFIFO, ("Unexpected fifo."));
1191 		finit_vnode(fp, flags, NULL, &vnops);
1192 	}
1193 
1194 	VOP_UNLOCK(vp);
1195 	if (flags & O_TRUNC) {
1196 		error = fo_truncate(fp, 0, td->td_ucred, td);
1197 		if (error != 0)
1198 			goto bad;
1199 	}
1200 success:
1201 	/*
1202 	 * If we haven't already installed the FD (for dupfdopen), do so now.
1203 	 */
1204 	if (indx == -1) {
1205 		struct filecaps *fcaps;
1206 
1207 #ifdef CAPABILITIES
1208 		if ((nd.ni_resflags & NIRES_STRICTREL) != 0)
1209 			fcaps = &nd.ni_filecaps;
1210 		else
1211 #endif
1212 			fcaps = NULL;
1213 		error = finstall(td, fp, &indx, flags, fcaps);
1214 		/* On success finstall() consumes fcaps. */
1215 		if (error != 0) {
1216 			filecaps_free(&nd.ni_filecaps);
1217 			goto bad;
1218 		}
1219 	} else {
1220 		filecaps_free(&nd.ni_filecaps);
1221 	}
1222 
1223 	/*
1224 	 * Release our private reference, leaving the one associated with
1225 	 * the descriptor table intact.
1226 	 */
1227 	fdrop(fp, td);
1228 	td->td_retval[0] = indx;
1229 	return (0);
1230 bad:
1231 	KASSERT(indx == -1, ("indx=%d, should be -1", indx));
1232 	fdrop_close(fp, td);
1233 	return (error);
1234 }
1235 
1236 #ifdef COMPAT_43
1237 /*
1238  * Create a file.
1239  */
1240 #ifndef _SYS_SYSPROTO_H_
1241 struct ocreat_args {
1242 	char	*path;
1243 	int	mode;
1244 };
1245 #endif
1246 int
1247 ocreat(struct thread *td, struct ocreat_args *uap)
1248 {
1249 
1250 	return (kern_openat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1251 	    O_WRONLY | O_CREAT | O_TRUNC, uap->mode));
1252 }
1253 #endif /* COMPAT_43 */
1254 
1255 /*
1256  * Create a special file.
1257  */
1258 #ifndef _SYS_SYSPROTO_H_
1259 struct mknodat_args {
1260 	int	fd;
1261 	char	*path;
1262 	mode_t	mode;
1263 	dev_t	dev;
1264 };
1265 #endif
1266 int
1267 sys_mknodat(struct thread *td, struct mknodat_args *uap)
1268 {
1269 
1270 	return (kern_mknodat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode,
1271 	    uap->dev));
1272 }
1273 
1274 #if defined(COMPAT_FREEBSD11)
1275 int
1276 freebsd11_mknod(struct thread *td,
1277     struct freebsd11_mknod_args *uap)
1278 {
1279 
1280 	return (kern_mknodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1281 	    uap->mode, uap->dev));
1282 }
1283 
1284 int
1285 freebsd11_mknodat(struct thread *td,
1286     struct freebsd11_mknodat_args *uap)
1287 {
1288 
1289 	return (kern_mknodat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode,
1290 	    uap->dev));
1291 }
1292 #endif /* COMPAT_FREEBSD11 */
1293 
1294 int
1295 kern_mknodat(struct thread *td, int fd, const char *path, enum uio_seg pathseg,
1296     int mode, dev_t dev)
1297 {
1298 	struct vnode *vp;
1299 	struct mount *mp;
1300 	struct vattr vattr;
1301 	struct nameidata nd;
1302 	int error, whiteout = 0;
1303 
1304 	AUDIT_ARG_MODE(mode);
1305 	AUDIT_ARG_DEV(dev);
1306 	switch (mode & S_IFMT) {
1307 	case S_IFCHR:
1308 	case S_IFBLK:
1309 		error = priv_check(td, PRIV_VFS_MKNOD_DEV);
1310 		if (error == 0 && dev == VNOVAL)
1311 			error = EINVAL;
1312 		break;
1313 	case S_IFWHT:
1314 		error = priv_check(td, PRIV_VFS_MKNOD_WHT);
1315 		break;
1316 	case S_IFIFO:
1317 		if (dev == 0)
1318 			return (kern_mkfifoat(td, fd, path, pathseg, mode));
1319 		/* FALLTHROUGH */
1320 	default:
1321 		error = EINVAL;
1322 		break;
1323 	}
1324 	if (error != 0)
1325 		return (error);
1326 restart:
1327 	bwillwrite();
1328 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
1329 	    NOCACHE, pathseg, path, fd, &cap_mknodat_rights,
1330 	    td);
1331 	if ((error = namei(&nd)) != 0)
1332 		return (error);
1333 	vp = nd.ni_vp;
1334 	if (vp != NULL) {
1335 		NDFREE(&nd, NDF_ONLY_PNBUF);
1336 		if (vp == nd.ni_dvp)
1337 			vrele(nd.ni_dvp);
1338 		else
1339 			vput(nd.ni_dvp);
1340 		vrele(vp);
1341 		return (EEXIST);
1342 	} else {
1343 		VATTR_NULL(&vattr);
1344 		vattr.va_mode = (mode & ALLPERMS) &
1345 		    ~td->td_proc->p_pd->pd_cmask;
1346 		vattr.va_rdev = dev;
1347 		whiteout = 0;
1348 
1349 		switch (mode & S_IFMT) {
1350 		case S_IFCHR:
1351 			vattr.va_type = VCHR;
1352 			break;
1353 		case S_IFBLK:
1354 			vattr.va_type = VBLK;
1355 			break;
1356 		case S_IFWHT:
1357 			whiteout = 1;
1358 			break;
1359 		default:
1360 			panic("kern_mknod: invalid mode");
1361 		}
1362 	}
1363 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1364 		NDFREE(&nd, NDF_ONLY_PNBUF);
1365 		vput(nd.ni_dvp);
1366 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1367 			return (error);
1368 		goto restart;
1369 	}
1370 #ifdef MAC
1371 	if (error == 0 && !whiteout)
1372 		error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp,
1373 		    &nd.ni_cnd, &vattr);
1374 #endif
1375 	if (error == 0) {
1376 		if (whiteout)
1377 			error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE);
1378 		else {
1379 			error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp,
1380 						&nd.ni_cnd, &vattr);
1381 			if (error == 0)
1382 				vput(nd.ni_vp);
1383 		}
1384 	}
1385 	NDFREE(&nd, NDF_ONLY_PNBUF);
1386 	vput(nd.ni_dvp);
1387 	vn_finished_write(mp);
1388 	if (error == ERELOOKUP)
1389 		goto restart;
1390 	return (error);
1391 }
1392 
1393 /*
1394  * Create a named pipe.
1395  */
1396 #ifndef _SYS_SYSPROTO_H_
1397 struct mkfifo_args {
1398 	char	*path;
1399 	int	mode;
1400 };
1401 #endif
1402 int
1403 sys_mkfifo(struct thread *td, struct mkfifo_args *uap)
1404 {
1405 
1406 	return (kern_mkfifoat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1407 	    uap->mode));
1408 }
1409 
1410 #ifndef _SYS_SYSPROTO_H_
1411 struct mkfifoat_args {
1412 	int	fd;
1413 	char	*path;
1414 	mode_t	mode;
1415 };
1416 #endif
1417 int
1418 sys_mkfifoat(struct thread *td, struct mkfifoat_args *uap)
1419 {
1420 
1421 	return (kern_mkfifoat(td, uap->fd, uap->path, UIO_USERSPACE,
1422 	    uap->mode));
1423 }
1424 
1425 int
1426 kern_mkfifoat(struct thread *td, int fd, const char *path,
1427     enum uio_seg pathseg, int mode)
1428 {
1429 	struct mount *mp;
1430 	struct vattr vattr;
1431 	struct nameidata nd;
1432 	int error;
1433 
1434 	AUDIT_ARG_MODE(mode);
1435 restart:
1436 	bwillwrite();
1437 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
1438 	    NOCACHE, pathseg, path, fd, &cap_mkfifoat_rights,
1439 	    td);
1440 	if ((error = namei(&nd)) != 0)
1441 		return (error);
1442 	if (nd.ni_vp != NULL) {
1443 		NDFREE(&nd, NDF_ONLY_PNBUF);
1444 		if (nd.ni_vp == nd.ni_dvp)
1445 			vrele(nd.ni_dvp);
1446 		else
1447 			vput(nd.ni_dvp);
1448 		vrele(nd.ni_vp);
1449 		return (EEXIST);
1450 	}
1451 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1452 		NDFREE(&nd, NDF_ONLY_PNBUF);
1453 		vput(nd.ni_dvp);
1454 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1455 			return (error);
1456 		goto restart;
1457 	}
1458 	VATTR_NULL(&vattr);
1459 	vattr.va_type = VFIFO;
1460 	vattr.va_mode = (mode & ALLPERMS) & ~td->td_proc->p_pd->pd_cmask;
1461 #ifdef MAC
1462 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1463 	    &vattr);
1464 	if (error != 0)
1465 		goto out;
1466 #endif
1467 	error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
1468 	if (error == 0)
1469 		vput(nd.ni_vp);
1470 #ifdef MAC
1471 out:
1472 #endif
1473 	vput(nd.ni_dvp);
1474 	vn_finished_write(mp);
1475 	NDFREE(&nd, NDF_ONLY_PNBUF);
1476 	if (error == ERELOOKUP)
1477 		goto restart;
1478 	return (error);
1479 }
1480 
1481 /*
1482  * Make a hard file link.
1483  */
1484 #ifndef _SYS_SYSPROTO_H_
1485 struct link_args {
1486 	char	*path;
1487 	char	*link;
1488 };
1489 #endif
1490 int
1491 sys_link(struct thread *td, struct link_args *uap)
1492 {
1493 
1494 	return (kern_linkat(td, AT_FDCWD, AT_FDCWD, uap->path, uap->link,
1495 	    UIO_USERSPACE, FOLLOW));
1496 }
1497 
1498 #ifndef _SYS_SYSPROTO_H_
1499 struct linkat_args {
1500 	int	fd1;
1501 	char	*path1;
1502 	int	fd2;
1503 	char	*path2;
1504 	int	flag;
1505 };
1506 #endif
1507 int
1508 sys_linkat(struct thread *td, struct linkat_args *uap)
1509 {
1510 	int flag;
1511 
1512 	flag = uap->flag;
1513 	if ((flag & ~(AT_SYMLINK_FOLLOW | AT_BENEATH |
1514 	    AT_RESOLVE_BENEATH)) != 0)
1515 		return (EINVAL);
1516 
1517 	return (kern_linkat(td, uap->fd1, uap->fd2, uap->path1, uap->path2,
1518 	    UIO_USERSPACE, at2cnpflags(flag, AT_SYMLINK_FOLLOW | AT_BENEATH |
1519 	    AT_RESOLVE_BENEATH)));
1520 }
1521 
1522 int hardlink_check_uid = 0;
1523 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_uid, CTLFLAG_RW,
1524     &hardlink_check_uid, 0,
1525     "Unprivileged processes cannot create hard links to files owned by other "
1526     "users");
1527 static int hardlink_check_gid = 0;
1528 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_gid, CTLFLAG_RW,
1529     &hardlink_check_gid, 0,
1530     "Unprivileged processes cannot create hard links to files owned by other "
1531     "groups");
1532 
1533 static int
1534 can_hardlink(struct vnode *vp, struct ucred *cred)
1535 {
1536 	struct vattr va;
1537 	int error;
1538 
1539 	if (!hardlink_check_uid && !hardlink_check_gid)
1540 		return (0);
1541 
1542 	error = VOP_GETATTR(vp, &va, cred);
1543 	if (error != 0)
1544 		return (error);
1545 
1546 	if (hardlink_check_uid && cred->cr_uid != va.va_uid) {
1547 		error = priv_check_cred(cred, PRIV_VFS_LINK);
1548 		if (error != 0)
1549 			return (error);
1550 	}
1551 
1552 	if (hardlink_check_gid && !groupmember(va.va_gid, cred)) {
1553 		error = priv_check_cred(cred, PRIV_VFS_LINK);
1554 		if (error != 0)
1555 			return (error);
1556 	}
1557 
1558 	return (0);
1559 }
1560 
1561 int
1562 kern_linkat(struct thread *td, int fd1, int fd2, const char *path1,
1563     const char *path2, enum uio_seg segflag, int follow)
1564 {
1565 	struct nameidata nd;
1566 	int error;
1567 
1568 	do {
1569 		bwillwrite();
1570 		NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, segflag,
1571 		    path1, fd1, &cap_linkat_source_rights, td);
1572 		if ((error = namei(&nd)) != 0)
1573 			return (error);
1574 		NDFREE(&nd, NDF_ONLY_PNBUF);
1575 		error = kern_linkat_vp(td, nd.ni_vp, fd2, path2, segflag);
1576 	} while (error ==  EAGAIN || error == ERELOOKUP);
1577 	return (error);
1578 }
1579 
1580 static int
1581 kern_linkat_vp(struct thread *td, struct vnode *vp, int fd, const char *path,
1582     enum uio_seg segflag)
1583 {
1584 	struct nameidata nd;
1585 	struct mount *mp;
1586 	int error;
1587 
1588 	if (vp->v_type == VDIR) {
1589 		vrele(vp);
1590 		return (EPERM);		/* POSIX */
1591 	}
1592 	NDINIT_ATRIGHTS(&nd, CREATE,
1593 	    LOCKPARENT | SAVENAME | AUDITVNODE2 | NOCACHE, segflag, path, fd,
1594 	    &cap_linkat_target_rights, td);
1595 	if ((error = namei(&nd)) == 0) {
1596 		if (nd.ni_vp != NULL) {
1597 			NDFREE(&nd, NDF_ONLY_PNBUF);
1598 			if (nd.ni_dvp == nd.ni_vp)
1599 				vrele(nd.ni_dvp);
1600 			else
1601 				vput(nd.ni_dvp);
1602 			vrele(nd.ni_vp);
1603 			vrele(vp);
1604 			return (EEXIST);
1605 		} else if (nd.ni_dvp->v_mount != vp->v_mount) {
1606 			/*
1607 			 * Cross-device link.  No need to recheck
1608 			 * vp->v_type, since it cannot change, except
1609 			 * to VBAD.
1610 			 */
1611 			NDFREE(&nd, NDF_ONLY_PNBUF);
1612 			vput(nd.ni_dvp);
1613 			vrele(vp);
1614 			return (EXDEV);
1615 		} else if ((error = vn_lock(vp, LK_EXCLUSIVE)) == 0) {
1616 			error = can_hardlink(vp, td->td_ucred);
1617 #ifdef MAC
1618 			if (error == 0)
1619 				error = mac_vnode_check_link(td->td_ucred,
1620 				    nd.ni_dvp, vp, &nd.ni_cnd);
1621 #endif
1622 			if (error != 0) {
1623 				vput(vp);
1624 				vput(nd.ni_dvp);
1625 				NDFREE(&nd, NDF_ONLY_PNBUF);
1626 				return (error);
1627 			}
1628 			error = vn_start_write(vp, &mp, V_NOWAIT);
1629 			if (error != 0) {
1630 				vput(vp);
1631 				vput(nd.ni_dvp);
1632 				NDFREE(&nd, NDF_ONLY_PNBUF);
1633 				error = vn_start_write(NULL, &mp,
1634 				    V_XSLEEP | PCATCH);
1635 				if (error != 0)
1636 					return (error);
1637 				return (EAGAIN);
1638 			}
1639 			error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
1640 			VOP_UNLOCK(vp);
1641 			vput(nd.ni_dvp);
1642 			vn_finished_write(mp);
1643 			NDFREE(&nd, NDF_ONLY_PNBUF);
1644 		} else {
1645 			vput(nd.ni_dvp);
1646 			NDFREE(&nd, NDF_ONLY_PNBUF);
1647 			vrele(vp);
1648 			return (EAGAIN);
1649 		}
1650 	}
1651 	vrele(vp);
1652 	return (error);
1653 }
1654 
1655 /*
1656  * Make a symbolic link.
1657  */
1658 #ifndef _SYS_SYSPROTO_H_
1659 struct symlink_args {
1660 	char	*path;
1661 	char	*link;
1662 };
1663 #endif
1664 int
1665 sys_symlink(struct thread *td, struct symlink_args *uap)
1666 {
1667 
1668 	return (kern_symlinkat(td, uap->path, AT_FDCWD, uap->link,
1669 	    UIO_USERSPACE));
1670 }
1671 
1672 #ifndef _SYS_SYSPROTO_H_
1673 struct symlinkat_args {
1674 	char	*path;
1675 	int	fd;
1676 	char	*path2;
1677 };
1678 #endif
1679 int
1680 sys_symlinkat(struct thread *td, struct symlinkat_args *uap)
1681 {
1682 
1683 	return (kern_symlinkat(td, uap->path1, uap->fd, uap->path2,
1684 	    UIO_USERSPACE));
1685 }
1686 
1687 int
1688 kern_symlinkat(struct thread *td, const char *path1, int fd, const char *path2,
1689     enum uio_seg segflg)
1690 {
1691 	struct mount *mp;
1692 	struct vattr vattr;
1693 	const char *syspath;
1694 	char *tmppath;
1695 	struct nameidata nd;
1696 	int error;
1697 
1698 	if (segflg == UIO_SYSSPACE) {
1699 		syspath = path1;
1700 	} else {
1701 		tmppath = uma_zalloc(namei_zone, M_WAITOK);
1702 		if ((error = copyinstr(path1, tmppath, MAXPATHLEN, NULL)) != 0)
1703 			goto out;
1704 		syspath = tmppath;
1705 	}
1706 	AUDIT_ARG_TEXT(syspath);
1707 restart:
1708 	bwillwrite();
1709 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
1710 	    NOCACHE, segflg, path2, fd, &cap_symlinkat_rights,
1711 	    td);
1712 	if ((error = namei(&nd)) != 0)
1713 		goto out;
1714 	if (nd.ni_vp) {
1715 		NDFREE(&nd, NDF_ONLY_PNBUF);
1716 		if (nd.ni_vp == nd.ni_dvp)
1717 			vrele(nd.ni_dvp);
1718 		else
1719 			vput(nd.ni_dvp);
1720 		vrele(nd.ni_vp);
1721 		error = EEXIST;
1722 		goto out;
1723 	}
1724 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1725 		NDFREE(&nd, NDF_ONLY_PNBUF);
1726 		vput(nd.ni_dvp);
1727 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1728 			goto out;
1729 		goto restart;
1730 	}
1731 	VATTR_NULL(&vattr);
1732 	vattr.va_mode = ACCESSPERMS &~ td->td_proc->p_pd->pd_cmask;
1733 #ifdef MAC
1734 	vattr.va_type = VLNK;
1735 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1736 	    &vattr);
1737 	if (error != 0)
1738 		goto out2;
1739 #endif
1740 	error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, syspath);
1741 	if (error == 0)
1742 		vput(nd.ni_vp);
1743 #ifdef MAC
1744 out2:
1745 #endif
1746 	NDFREE(&nd, NDF_ONLY_PNBUF);
1747 	vput(nd.ni_dvp);
1748 	vn_finished_write(mp);
1749 	if (error == ERELOOKUP)
1750 		goto restart;
1751 out:
1752 	if (segflg != UIO_SYSSPACE)
1753 		uma_zfree(namei_zone, tmppath);
1754 	return (error);
1755 }
1756 
1757 /*
1758  * Delete a whiteout from the filesystem.
1759  */
1760 #ifndef _SYS_SYSPROTO_H_
1761 struct undelete_args {
1762 	char *path;
1763 };
1764 #endif
1765 int
1766 sys_undelete(struct thread *td, struct undelete_args *uap)
1767 {
1768 	struct mount *mp;
1769 	struct nameidata nd;
1770 	int error;
1771 
1772 restart:
1773 	bwillwrite();
1774 	NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | AUDITVNODE1,
1775 	    UIO_USERSPACE, uap->path, td);
1776 	error = namei(&nd);
1777 	if (error != 0)
1778 		return (error);
1779 
1780 	if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) {
1781 		NDFREE(&nd, NDF_ONLY_PNBUF);
1782 		if (nd.ni_vp == nd.ni_dvp)
1783 			vrele(nd.ni_dvp);
1784 		else
1785 			vput(nd.ni_dvp);
1786 		if (nd.ni_vp)
1787 			vrele(nd.ni_vp);
1788 		return (EEXIST);
1789 	}
1790 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1791 		NDFREE(&nd, NDF_ONLY_PNBUF);
1792 		vput(nd.ni_dvp);
1793 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1794 			return (error);
1795 		goto restart;
1796 	}
1797 	error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE);
1798 	NDFREE(&nd, NDF_ONLY_PNBUF);
1799 	vput(nd.ni_dvp);
1800 	vn_finished_write(mp);
1801 	if (error == ERELOOKUP)
1802 		goto restart;
1803 	return (error);
1804 }
1805 
1806 /*
1807  * Delete a name from the filesystem.
1808  */
1809 #ifndef _SYS_SYSPROTO_H_
1810 struct unlink_args {
1811 	char	*path;
1812 };
1813 #endif
1814 int
1815 sys_unlink(struct thread *td, struct unlink_args *uap)
1816 {
1817 
1818 	return (kern_funlinkat(td, AT_FDCWD, uap->path, FD_NONE, UIO_USERSPACE,
1819 	    0, 0));
1820 }
1821 
1822 static int
1823 kern_funlinkat_ex(struct thread *td, int dfd, const char *path, int fd,
1824     int flag, enum uio_seg pathseg, ino_t oldinum)
1825 {
1826 
1827 	if ((flag & ~AT_REMOVEDIR) != 0)
1828 		return (EINVAL);
1829 
1830 	if ((flag & AT_REMOVEDIR) != 0)
1831 		return (kern_frmdirat(td, dfd, path, fd, UIO_USERSPACE, 0));
1832 
1833 	return (kern_funlinkat(td, dfd, path, fd, UIO_USERSPACE, 0, 0));
1834 }
1835 
1836 #ifndef _SYS_SYSPROTO_H_
1837 struct unlinkat_args {
1838 	int	fd;
1839 	char	*path;
1840 	int	flag;
1841 };
1842 #endif
1843 int
1844 sys_unlinkat(struct thread *td, struct unlinkat_args *uap)
1845 {
1846 
1847 	return (kern_funlinkat_ex(td, uap->fd, uap->path, FD_NONE, uap->flag,
1848 	    UIO_USERSPACE, 0));
1849 }
1850 
1851 #ifndef _SYS_SYSPROTO_H_
1852 struct funlinkat_args {
1853 	int		dfd;
1854 	const char	*path;
1855 	int		fd;
1856 	int		flag;
1857 };
1858 #endif
1859 int
1860 sys_funlinkat(struct thread *td, struct funlinkat_args *uap)
1861 {
1862 
1863 	return (kern_funlinkat_ex(td, uap->dfd, uap->path, uap->fd, uap->flag,
1864 	    UIO_USERSPACE, 0));
1865 }
1866 
1867 int
1868 kern_funlinkat(struct thread *td, int dfd, const char *path, int fd,
1869     enum uio_seg pathseg, int flag, ino_t oldinum)
1870 {
1871 	struct mount *mp;
1872 	struct file *fp;
1873 	struct vnode *vp;
1874 	struct nameidata nd;
1875 	struct stat sb;
1876 	int error;
1877 
1878 	fp = NULL;
1879 	if (fd != FD_NONE) {
1880 		error = getvnode(td, fd, &cap_no_rights, &fp);
1881 		if (error != 0)
1882 			return (error);
1883 	}
1884 
1885 restart:
1886 	bwillwrite();
1887 	NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1 |
1888 	    at2cnpflags(flag, AT_BENEATH | AT_RESOLVE_BENEATH),
1889 	    pathseg, path, dfd, &cap_unlinkat_rights, td);
1890 	if ((error = namei(&nd)) != 0) {
1891 		if (error == EINVAL)
1892 			error = EPERM;
1893 		goto fdout;
1894 	}
1895 	vp = nd.ni_vp;
1896 	if (vp->v_type == VDIR && oldinum == 0) {
1897 		error = EPERM;		/* POSIX */
1898 	} else if (oldinum != 0 &&
1899 		  ((error = VOP_STAT(vp, &sb, td->td_ucred, NOCRED, td)) == 0) &&
1900 		  sb.st_ino != oldinum) {
1901 		error = EIDRM;	/* Identifier removed */
1902 	} else if (fp != NULL && fp->f_vnode != vp) {
1903 		if (VN_IS_DOOMED(fp->f_vnode))
1904 			error = EBADF;
1905 		else
1906 			error = EDEADLK;
1907 	} else {
1908 		/*
1909 		 * The root of a mounted filesystem cannot be deleted.
1910 		 *
1911 		 * XXX: can this only be a VDIR case?
1912 		 */
1913 		if (vp->v_vflag & VV_ROOT)
1914 			error = EBUSY;
1915 	}
1916 	if (error == 0) {
1917 		if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1918 			NDFREE(&nd, NDF_ONLY_PNBUF);
1919 			vput(nd.ni_dvp);
1920 			if (vp == nd.ni_dvp)
1921 				vrele(vp);
1922 			else
1923 				vput(vp);
1924 			if ((error = vn_start_write(NULL, &mp,
1925 			    V_XSLEEP | PCATCH)) != 0) {
1926 				goto fdout;
1927 			}
1928 			goto restart;
1929 		}
1930 #ifdef MAC
1931 		error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
1932 		    &nd.ni_cnd);
1933 		if (error != 0)
1934 			goto out;
1935 #endif
1936 		vfs_notify_upper(vp, VFS_NOTIFY_UPPER_UNLINK);
1937 		error = VOP_REMOVE(nd.ni_dvp, vp, &nd.ni_cnd);
1938 #ifdef MAC
1939 out:
1940 #endif
1941 		vn_finished_write(mp);
1942 	}
1943 	NDFREE(&nd, NDF_ONLY_PNBUF);
1944 	vput(nd.ni_dvp);
1945 	if (vp == nd.ni_dvp)
1946 		vrele(vp);
1947 	else
1948 		vput(vp);
1949 	if (error == ERELOOKUP)
1950 		goto restart;
1951 fdout:
1952 	if (fp != NULL)
1953 		fdrop(fp, td);
1954 	return (error);
1955 }
1956 
1957 /*
1958  * Reposition read/write file offset.
1959  */
1960 #ifndef _SYS_SYSPROTO_H_
1961 struct lseek_args {
1962 	int	fd;
1963 	int	pad;
1964 	off_t	offset;
1965 	int	whence;
1966 };
1967 #endif
1968 int
1969 sys_lseek(struct thread *td, struct lseek_args *uap)
1970 {
1971 
1972 	return (kern_lseek(td, uap->fd, uap->offset, uap->whence));
1973 }
1974 
1975 int
1976 kern_lseek(struct thread *td, int fd, off_t offset, int whence)
1977 {
1978 	struct file *fp;
1979 	int error;
1980 
1981 	AUDIT_ARG_FD(fd);
1982 	error = fget(td, fd, &cap_seek_rights, &fp);
1983 	if (error != 0)
1984 		return (error);
1985 	error = (fp->f_ops->fo_flags & DFLAG_SEEKABLE) != 0 ?
1986 	    fo_seek(fp, offset, whence, td) : ESPIPE;
1987 	fdrop(fp, td);
1988 	return (error);
1989 }
1990 
1991 #if defined(COMPAT_43)
1992 /*
1993  * Reposition read/write file offset.
1994  */
1995 #ifndef _SYS_SYSPROTO_H_
1996 struct olseek_args {
1997 	int	fd;
1998 	long	offset;
1999 	int	whence;
2000 };
2001 #endif
2002 int
2003 olseek(struct thread *td, struct olseek_args *uap)
2004 {
2005 
2006 	return (kern_lseek(td, uap->fd, uap->offset, uap->whence));
2007 }
2008 #endif /* COMPAT_43 */
2009 
2010 #if defined(COMPAT_FREEBSD6)
2011 /* Version with the 'pad' argument */
2012 int
2013 freebsd6_lseek(struct thread *td, struct freebsd6_lseek_args *uap)
2014 {
2015 
2016 	return (kern_lseek(td, uap->fd, uap->offset, uap->whence));
2017 }
2018 #endif
2019 
2020 /*
2021  * Check access permissions using passed credentials.
2022  */
2023 static int
2024 vn_access(struct vnode *vp, int user_flags, struct ucred *cred,
2025      struct thread *td)
2026 {
2027 	accmode_t accmode;
2028 	int error;
2029 
2030 	/* Flags == 0 means only check for existence. */
2031 	if (user_flags == 0)
2032 		return (0);
2033 
2034 	accmode = 0;
2035 	if (user_flags & R_OK)
2036 		accmode |= VREAD;
2037 	if (user_flags & W_OK)
2038 		accmode |= VWRITE;
2039 	if (user_flags & X_OK)
2040 		accmode |= VEXEC;
2041 #ifdef MAC
2042 	error = mac_vnode_check_access(cred, vp, accmode);
2043 	if (error != 0)
2044 		return (error);
2045 #endif
2046 	if ((accmode & VWRITE) == 0 || (error = vn_writechk(vp)) == 0)
2047 		error = VOP_ACCESS(vp, accmode, cred, td);
2048 	return (error);
2049 }
2050 
2051 /*
2052  * Check access permissions using "real" credentials.
2053  */
2054 #ifndef _SYS_SYSPROTO_H_
2055 struct access_args {
2056 	char	*path;
2057 	int	amode;
2058 };
2059 #endif
2060 int
2061 sys_access(struct thread *td, struct access_args *uap)
2062 {
2063 
2064 	return (kern_accessat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2065 	    0, uap->amode));
2066 }
2067 
2068 #ifndef _SYS_SYSPROTO_H_
2069 struct faccessat_args {
2070 	int	dirfd;
2071 	char	*path;
2072 	int	amode;
2073 	int	flag;
2074 }
2075 #endif
2076 int
2077 sys_faccessat(struct thread *td, struct faccessat_args *uap)
2078 {
2079 
2080 	return (kern_accessat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
2081 	    uap->amode));
2082 }
2083 
2084 int
2085 kern_accessat(struct thread *td, int fd, const char *path,
2086     enum uio_seg pathseg, int flag, int amode)
2087 {
2088 	struct ucred *cred, *usecred;
2089 	struct vnode *vp;
2090 	struct nameidata nd;
2091 	int error;
2092 
2093 	if ((flag & ~(AT_EACCESS | AT_BENEATH | AT_RESOLVE_BENEATH)) != 0)
2094 		return (EINVAL);
2095 	if (amode != F_OK && (amode & ~(R_OK | W_OK | X_OK)) != 0)
2096 		return (EINVAL);
2097 
2098 	/*
2099 	 * Create and modify a temporary credential instead of one that
2100 	 * is potentially shared (if we need one).
2101 	 */
2102 	cred = td->td_ucred;
2103 	if ((flag & AT_EACCESS) == 0 &&
2104 	    ((cred->cr_uid != cred->cr_ruid ||
2105 	    cred->cr_rgid != cred->cr_groups[0]))) {
2106 		usecred = crdup(cred);
2107 		usecred->cr_uid = cred->cr_ruid;
2108 		usecred->cr_groups[0] = cred->cr_rgid;
2109 		td->td_ucred = usecred;
2110 	} else
2111 		usecred = cred;
2112 	AUDIT_ARG_VALUE(amode);
2113 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF |
2114 	    AUDITVNODE1 | at2cnpflags(flag, AT_BENEATH | AT_RESOLVE_BENEATH),
2115 	    pathseg, path, fd, &cap_fstat_rights, td);
2116 	if ((error = namei(&nd)) != 0)
2117 		goto out;
2118 	vp = nd.ni_vp;
2119 
2120 	error = vn_access(vp, amode, usecred, td);
2121 	NDFREE(&nd, NDF_ONLY_PNBUF);
2122 	vput(vp);
2123 out:
2124 	if (usecred != cred) {
2125 		td->td_ucred = cred;
2126 		crfree(usecred);
2127 	}
2128 	return (error);
2129 }
2130 
2131 /*
2132  * Check access permissions using "effective" credentials.
2133  */
2134 #ifndef _SYS_SYSPROTO_H_
2135 struct eaccess_args {
2136 	char	*path;
2137 	int	amode;
2138 };
2139 #endif
2140 int
2141 sys_eaccess(struct thread *td, struct eaccess_args *uap)
2142 {
2143 
2144 	return (kern_accessat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2145 	    AT_EACCESS, uap->amode));
2146 }
2147 
2148 #if defined(COMPAT_43)
2149 /*
2150  * Get file status; this version follows links.
2151  */
2152 #ifndef _SYS_SYSPROTO_H_
2153 struct ostat_args {
2154 	char	*path;
2155 	struct ostat *ub;
2156 };
2157 #endif
2158 int
2159 ostat(struct thread *td, struct ostat_args *uap)
2160 {
2161 	struct stat sb;
2162 	struct ostat osb;
2163 	int error;
2164 
2165 	error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE,
2166 	    &sb, NULL);
2167 	if (error != 0)
2168 		return (error);
2169 	cvtstat(&sb, &osb);
2170 	return (copyout(&osb, uap->ub, sizeof (osb)));
2171 }
2172 
2173 /*
2174  * Get file status; this version does not follow links.
2175  */
2176 #ifndef _SYS_SYSPROTO_H_
2177 struct olstat_args {
2178 	char	*path;
2179 	struct ostat *ub;
2180 };
2181 #endif
2182 int
2183 olstat(struct thread *td, struct olstat_args *uap)
2184 {
2185 	struct stat sb;
2186 	struct ostat osb;
2187 	int error;
2188 
2189 	error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2190 	    UIO_USERSPACE, &sb, NULL);
2191 	if (error != 0)
2192 		return (error);
2193 	cvtstat(&sb, &osb);
2194 	return (copyout(&osb, uap->ub, sizeof (osb)));
2195 }
2196 
2197 /*
2198  * Convert from an old to a new stat structure.
2199  * XXX: many values are blindly truncated.
2200  */
2201 void
2202 cvtstat(struct stat *st, struct ostat *ost)
2203 {
2204 
2205 	bzero(ost, sizeof(*ost));
2206 	ost->st_dev = st->st_dev;
2207 	ost->st_ino = st->st_ino;
2208 	ost->st_mode = st->st_mode;
2209 	ost->st_nlink = st->st_nlink;
2210 	ost->st_uid = st->st_uid;
2211 	ost->st_gid = st->st_gid;
2212 	ost->st_rdev = st->st_rdev;
2213 	ost->st_size = MIN(st->st_size, INT32_MAX);
2214 	ost->st_atim = st->st_atim;
2215 	ost->st_mtim = st->st_mtim;
2216 	ost->st_ctim = st->st_ctim;
2217 	ost->st_blksize = st->st_blksize;
2218 	ost->st_blocks = st->st_blocks;
2219 	ost->st_flags = st->st_flags;
2220 	ost->st_gen = st->st_gen;
2221 }
2222 #endif /* COMPAT_43 */
2223 
2224 #if defined(COMPAT_43) || defined(COMPAT_FREEBSD11)
2225 int ino64_trunc_error;
2226 SYSCTL_INT(_vfs, OID_AUTO, ino64_trunc_error, CTLFLAG_RW,
2227     &ino64_trunc_error, 0,
2228     "Error on truncation of device, file or inode number, or link count");
2229 
2230 int
2231 freebsd11_cvtstat(struct stat *st, struct freebsd11_stat *ost)
2232 {
2233 
2234 	ost->st_dev = st->st_dev;
2235 	if (ost->st_dev != st->st_dev) {
2236 		switch (ino64_trunc_error) {
2237 		default:
2238 			/*
2239 			 * Since dev_t is almost raw, don't clamp to the
2240 			 * maximum for case 2, but ignore the error.
2241 			 */
2242 			break;
2243 		case 1:
2244 			return (EOVERFLOW);
2245 		}
2246 	}
2247 	ost->st_ino = st->st_ino;
2248 	if (ost->st_ino != st->st_ino) {
2249 		switch (ino64_trunc_error) {
2250 		default:
2251 		case 0:
2252 			break;
2253 		case 1:
2254 			return (EOVERFLOW);
2255 		case 2:
2256 			ost->st_ino = UINT32_MAX;
2257 			break;
2258 		}
2259 	}
2260 	ost->st_mode = st->st_mode;
2261 	ost->st_nlink = st->st_nlink;
2262 	if (ost->st_nlink != st->st_nlink) {
2263 		switch (ino64_trunc_error) {
2264 		default:
2265 		case 0:
2266 			break;
2267 		case 1:
2268 			return (EOVERFLOW);
2269 		case 2:
2270 			ost->st_nlink = UINT16_MAX;
2271 			break;
2272 		}
2273 	}
2274 	ost->st_uid = st->st_uid;
2275 	ost->st_gid = st->st_gid;
2276 	ost->st_rdev = st->st_rdev;
2277 	if (ost->st_rdev != st->st_rdev) {
2278 		switch (ino64_trunc_error) {
2279 		default:
2280 			break;
2281 		case 1:
2282 			return (EOVERFLOW);
2283 		}
2284 	}
2285 	ost->st_atim = st->st_atim;
2286 	ost->st_mtim = st->st_mtim;
2287 	ost->st_ctim = st->st_ctim;
2288 	ost->st_size = st->st_size;
2289 	ost->st_blocks = st->st_blocks;
2290 	ost->st_blksize = st->st_blksize;
2291 	ost->st_flags = st->st_flags;
2292 	ost->st_gen = st->st_gen;
2293 	ost->st_lspare = 0;
2294 	ost->st_birthtim = st->st_birthtim;
2295 	bzero((char *)&ost->st_birthtim + sizeof(ost->st_birthtim),
2296 	    sizeof(*ost) - offsetof(struct freebsd11_stat,
2297 	    st_birthtim) - sizeof(ost->st_birthtim));
2298 	return (0);
2299 }
2300 
2301 int
2302 freebsd11_stat(struct thread *td, struct freebsd11_stat_args* uap)
2303 {
2304 	struct stat sb;
2305 	struct freebsd11_stat osb;
2306 	int error;
2307 
2308 	error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE,
2309 	    &sb, NULL);
2310 	if (error != 0)
2311 		return (error);
2312 	error = freebsd11_cvtstat(&sb, &osb);
2313 	if (error == 0)
2314 		error = copyout(&osb, uap->ub, sizeof(osb));
2315 	return (error);
2316 }
2317 
2318 int
2319 freebsd11_lstat(struct thread *td, struct freebsd11_lstat_args* uap)
2320 {
2321 	struct stat sb;
2322 	struct freebsd11_stat osb;
2323 	int error;
2324 
2325 	error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2326 	    UIO_USERSPACE, &sb, NULL);
2327 	if (error != 0)
2328 		return (error);
2329 	error = freebsd11_cvtstat(&sb, &osb);
2330 	if (error == 0)
2331 		error = copyout(&osb, uap->ub, sizeof(osb));
2332 	return (error);
2333 }
2334 
2335 int
2336 freebsd11_fhstat(struct thread *td, struct freebsd11_fhstat_args* uap)
2337 {
2338 	struct fhandle fh;
2339 	struct stat sb;
2340 	struct freebsd11_stat osb;
2341 	int error;
2342 
2343 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
2344 	if (error != 0)
2345 		return (error);
2346 	error = kern_fhstat(td, fh, &sb);
2347 	if (error != 0)
2348 		return (error);
2349 	error = freebsd11_cvtstat(&sb, &osb);
2350 	if (error == 0)
2351 		error = copyout(&osb, uap->sb, sizeof(osb));
2352 	return (error);
2353 }
2354 
2355 int
2356 freebsd11_fstatat(struct thread *td, struct freebsd11_fstatat_args* uap)
2357 {
2358 	struct stat sb;
2359 	struct freebsd11_stat osb;
2360 	int error;
2361 
2362 	error = kern_statat(td, uap->flag, uap->fd, uap->path,
2363 	    UIO_USERSPACE, &sb, NULL);
2364 	if (error != 0)
2365 		return (error);
2366 	error = freebsd11_cvtstat(&sb, &osb);
2367 	if (error == 0)
2368 		error = copyout(&osb, uap->buf, sizeof(osb));
2369 	return (error);
2370 }
2371 #endif	/* COMPAT_FREEBSD11 */
2372 
2373 /*
2374  * Get file status
2375  */
2376 #ifndef _SYS_SYSPROTO_H_
2377 struct fstatat_args {
2378 	int	fd;
2379 	char	*path;
2380 	struct stat	*buf;
2381 	int	flag;
2382 }
2383 #endif
2384 int
2385 sys_fstatat(struct thread *td, struct fstatat_args *uap)
2386 {
2387 	struct stat sb;
2388 	int error;
2389 
2390 	error = kern_statat(td, uap->flag, uap->fd, uap->path,
2391 	    UIO_USERSPACE, &sb, NULL);
2392 	if (error == 0)
2393 		error = copyout(&sb, uap->buf, sizeof (sb));
2394 	return (error);
2395 }
2396 
2397 int
2398 kern_statat(struct thread *td, int flag, int fd, const char *path,
2399     enum uio_seg pathseg, struct stat *sbp,
2400     void (*hook)(struct vnode *vp, struct stat *sbp))
2401 {
2402 	struct nameidata nd;
2403 	int error;
2404 
2405 	if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_BENEATH |
2406 	    AT_RESOLVE_BENEATH)) != 0)
2407 		return (EINVAL);
2408 
2409 	NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(flag, AT_BENEATH |
2410 	    AT_RESOLVE_BENEATH | AT_SYMLINK_NOFOLLOW) | LOCKSHARED | LOCKLEAF |
2411 	    AUDITVNODE1, pathseg, path, fd, &cap_fstat_rights, td);
2412 
2413 	if ((error = namei(&nd)) != 0)
2414 		return (error);
2415 	error = VOP_STAT(nd.ni_vp, sbp, td->td_ucred, NOCRED, td);
2416 	if (error == 0) {
2417 		if (__predict_false(hook != NULL))
2418 			hook(nd.ni_vp, sbp);
2419 	}
2420 	NDFREE(&nd, NDF_ONLY_PNBUF);
2421 	vput(nd.ni_vp);
2422 #ifdef __STAT_TIME_T_EXT
2423 	sbp->st_atim_ext = 0;
2424 	sbp->st_mtim_ext = 0;
2425 	sbp->st_ctim_ext = 0;
2426 	sbp->st_btim_ext = 0;
2427 #endif
2428 #ifdef KTRACE
2429 	if (KTRPOINT(td, KTR_STRUCT))
2430 		ktrstat_error(sbp, error);
2431 #endif
2432 	return (error);
2433 }
2434 
2435 #if defined(COMPAT_FREEBSD11)
2436 /*
2437  * Implementation of the NetBSD [l]stat() functions.
2438  */
2439 void
2440 freebsd11_cvtnstat(struct stat *sb, struct nstat *nsb)
2441 {
2442 
2443 	bzero(nsb, sizeof(*nsb));
2444 	nsb->st_dev = sb->st_dev;
2445 	nsb->st_ino = sb->st_ino;
2446 	nsb->st_mode = sb->st_mode;
2447 	nsb->st_nlink = sb->st_nlink;
2448 	nsb->st_uid = sb->st_uid;
2449 	nsb->st_gid = sb->st_gid;
2450 	nsb->st_rdev = sb->st_rdev;
2451 	nsb->st_atim = sb->st_atim;
2452 	nsb->st_mtim = sb->st_mtim;
2453 	nsb->st_ctim = sb->st_ctim;
2454 	nsb->st_size = sb->st_size;
2455 	nsb->st_blocks = sb->st_blocks;
2456 	nsb->st_blksize = sb->st_blksize;
2457 	nsb->st_flags = sb->st_flags;
2458 	nsb->st_gen = sb->st_gen;
2459 	nsb->st_birthtim = sb->st_birthtim;
2460 }
2461 
2462 #ifndef _SYS_SYSPROTO_H_
2463 struct freebsd11_nstat_args {
2464 	char	*path;
2465 	struct nstat *ub;
2466 };
2467 #endif
2468 int
2469 freebsd11_nstat(struct thread *td, struct freebsd11_nstat_args *uap)
2470 {
2471 	struct stat sb;
2472 	struct nstat nsb;
2473 	int error;
2474 
2475 	error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE,
2476 	    &sb, NULL);
2477 	if (error != 0)
2478 		return (error);
2479 	freebsd11_cvtnstat(&sb, &nsb);
2480 	return (copyout(&nsb, uap->ub, sizeof (nsb)));
2481 }
2482 
2483 /*
2484  * NetBSD lstat.  Get file status; this version does not follow links.
2485  */
2486 #ifndef _SYS_SYSPROTO_H_
2487 struct freebsd11_nlstat_args {
2488 	char	*path;
2489 	struct nstat *ub;
2490 };
2491 #endif
2492 int
2493 freebsd11_nlstat(struct thread *td, struct freebsd11_nlstat_args *uap)
2494 {
2495 	struct stat sb;
2496 	struct nstat nsb;
2497 	int error;
2498 
2499 	error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2500 	    UIO_USERSPACE, &sb, NULL);
2501 	if (error != 0)
2502 		return (error);
2503 	freebsd11_cvtnstat(&sb, &nsb);
2504 	return (copyout(&nsb, uap->ub, sizeof (nsb)));
2505 }
2506 #endif /* COMPAT_FREEBSD11 */
2507 
2508 /*
2509  * Get configurable pathname variables.
2510  */
2511 #ifndef _SYS_SYSPROTO_H_
2512 struct pathconf_args {
2513 	char	*path;
2514 	int	name;
2515 };
2516 #endif
2517 int
2518 sys_pathconf(struct thread *td, struct pathconf_args *uap)
2519 {
2520 	long value;
2521 	int error;
2522 
2523 	error = kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name, FOLLOW,
2524 	    &value);
2525 	if (error == 0)
2526 		td->td_retval[0] = value;
2527 	return (error);
2528 }
2529 
2530 #ifndef _SYS_SYSPROTO_H_
2531 struct lpathconf_args {
2532 	char	*path;
2533 	int	name;
2534 };
2535 #endif
2536 int
2537 sys_lpathconf(struct thread *td, struct lpathconf_args *uap)
2538 {
2539 	long value;
2540 	int error;
2541 
2542 	error = kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name,
2543 	    NOFOLLOW, &value);
2544 	if (error == 0)
2545 		td->td_retval[0] = value;
2546 	return (error);
2547 }
2548 
2549 int
2550 kern_pathconf(struct thread *td, const char *path, enum uio_seg pathseg,
2551     int name, u_long flags, long *valuep)
2552 {
2553 	struct nameidata nd;
2554 	int error;
2555 
2556 	NDINIT(&nd, LOOKUP, LOCKSHARED | LOCKLEAF | AUDITVNODE1 | flags,
2557 	    pathseg, path, td);
2558 	if ((error = namei(&nd)) != 0)
2559 		return (error);
2560 	NDFREE(&nd, NDF_ONLY_PNBUF);
2561 
2562 	error = VOP_PATHCONF(nd.ni_vp, name, valuep);
2563 	vput(nd.ni_vp);
2564 	return (error);
2565 }
2566 
2567 /*
2568  * Return target name of a symbolic link.
2569  */
2570 #ifndef _SYS_SYSPROTO_H_
2571 struct readlink_args {
2572 	char	*path;
2573 	char	*buf;
2574 	size_t	count;
2575 };
2576 #endif
2577 int
2578 sys_readlink(struct thread *td, struct readlink_args *uap)
2579 {
2580 
2581 	return (kern_readlinkat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2582 	    uap->buf, UIO_USERSPACE, uap->count));
2583 }
2584 #ifndef _SYS_SYSPROTO_H_
2585 struct readlinkat_args {
2586 	int	fd;
2587 	char	*path;
2588 	char	*buf;
2589 	size_t	bufsize;
2590 };
2591 #endif
2592 int
2593 sys_readlinkat(struct thread *td, struct readlinkat_args *uap)
2594 {
2595 
2596 	return (kern_readlinkat(td, uap->fd, uap->path, UIO_USERSPACE,
2597 	    uap->buf, UIO_USERSPACE, uap->bufsize));
2598 }
2599 
2600 int
2601 kern_readlinkat(struct thread *td, int fd, const char *path,
2602     enum uio_seg pathseg, char *buf, enum uio_seg bufseg, size_t count)
2603 {
2604 	struct vnode *vp;
2605 	struct nameidata nd;
2606 	int error;
2607 
2608 	if (count > IOSIZE_MAX)
2609 		return (EINVAL);
2610 
2611 	NDINIT_AT(&nd, LOOKUP, NOFOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
2612 	    pathseg, path, fd, td);
2613 
2614 	if ((error = namei(&nd)) != 0)
2615 		return (error);
2616 	NDFREE(&nd, NDF_ONLY_PNBUF);
2617 	vp = nd.ni_vp;
2618 
2619 	error = kern_readlink_vp(vp, buf, bufseg, count, td);
2620 	vput(vp);
2621 
2622 	return (error);
2623 }
2624 
2625 /*
2626  * Helper function to readlink from a vnode
2627  */
2628 static int
2629 kern_readlink_vp(struct vnode *vp, char *buf, enum uio_seg bufseg, size_t count,
2630     struct thread *td)
2631 {
2632 	struct iovec aiov;
2633 	struct uio auio;
2634 	int error;
2635 
2636 	ASSERT_VOP_LOCKED(vp, "kern_readlink_vp(): vp not locked");
2637 #ifdef MAC
2638 	error = mac_vnode_check_readlink(td->td_ucred, vp);
2639 	if (error != 0)
2640 		return (error);
2641 #endif
2642 	if (vp->v_type != VLNK && (vp->v_vflag & VV_READLINK) == 0)
2643 		return (EINVAL);
2644 
2645 	aiov.iov_base = buf;
2646 	aiov.iov_len = count;
2647 	auio.uio_iov = &aiov;
2648 	auio.uio_iovcnt = 1;
2649 	auio.uio_offset = 0;
2650 	auio.uio_rw = UIO_READ;
2651 	auio.uio_segflg = bufseg;
2652 	auio.uio_td = td;
2653 	auio.uio_resid = count;
2654 	error = VOP_READLINK(vp, &auio, td->td_ucred);
2655 	td->td_retval[0] = count - auio.uio_resid;
2656 	return (error);
2657 }
2658 
2659 /*
2660  * Common implementation code for chflags() and fchflags().
2661  */
2662 static int
2663 setfflags(struct thread *td, struct vnode *vp, u_long flags)
2664 {
2665 	struct mount *mp;
2666 	struct vattr vattr;
2667 	int error;
2668 
2669 	/* We can't support the value matching VNOVAL. */
2670 	if (flags == VNOVAL)
2671 		return (EOPNOTSUPP);
2672 
2673 	/*
2674 	 * Prevent non-root users from setting flags on devices.  When
2675 	 * a device is reused, users can retain ownership of the device
2676 	 * if they are allowed to set flags and programs assume that
2677 	 * chown can't fail when done as root.
2678 	 */
2679 	if (vp->v_type == VCHR || vp->v_type == VBLK) {
2680 		error = priv_check(td, PRIV_VFS_CHFLAGS_DEV);
2681 		if (error != 0)
2682 			return (error);
2683 	}
2684 
2685 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2686 		return (error);
2687 	VATTR_NULL(&vattr);
2688 	vattr.va_flags = flags;
2689 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2690 #ifdef MAC
2691 	error = mac_vnode_check_setflags(td->td_ucred, vp, vattr.va_flags);
2692 	if (error == 0)
2693 #endif
2694 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
2695 	VOP_UNLOCK(vp);
2696 	vn_finished_write(mp);
2697 	return (error);
2698 }
2699 
2700 /*
2701  * Change flags of a file given a path name.
2702  */
2703 #ifndef _SYS_SYSPROTO_H_
2704 struct chflags_args {
2705 	const char *path;
2706 	u_long	flags;
2707 };
2708 #endif
2709 int
2710 sys_chflags(struct thread *td, struct chflags_args *uap)
2711 {
2712 
2713 	return (kern_chflagsat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2714 	    uap->flags, 0));
2715 }
2716 
2717 #ifndef _SYS_SYSPROTO_H_
2718 struct chflagsat_args {
2719 	int	fd;
2720 	const char *path;
2721 	u_long	flags;
2722 	int	atflag;
2723 }
2724 #endif
2725 int
2726 sys_chflagsat(struct thread *td, struct chflagsat_args *uap)
2727 {
2728 
2729 	if ((uap->atflag & ~(AT_SYMLINK_NOFOLLOW | AT_BENEATH |
2730 	    AT_RESOLVE_BENEATH)) != 0)
2731 		return (EINVAL);
2732 
2733 	return (kern_chflagsat(td, uap->fd, uap->path, UIO_USERSPACE,
2734 	    uap->flags, uap->atflag));
2735 }
2736 
2737 /*
2738  * Same as chflags() but doesn't follow symlinks.
2739  */
2740 #ifndef _SYS_SYSPROTO_H_
2741 struct lchflags_args {
2742 	const char *path;
2743 	u_long flags;
2744 };
2745 #endif
2746 int
2747 sys_lchflags(struct thread *td, struct lchflags_args *uap)
2748 {
2749 
2750 	return (kern_chflagsat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2751 	    uap->flags, AT_SYMLINK_NOFOLLOW));
2752 }
2753 
2754 static int
2755 kern_chflagsat(struct thread *td, int fd, const char *path,
2756     enum uio_seg pathseg, u_long flags, int atflag)
2757 {
2758 	struct nameidata nd;
2759 	int error;
2760 
2761 	AUDIT_ARG_FFLAGS(flags);
2762 	NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(atflag, AT_SYMLINK_NOFOLLOW |
2763 	    AT_BENEATH | AT_RESOLVE_BENEATH) | AUDITVNODE1, pathseg, path, fd,
2764 	    &cap_fchflags_rights, td);
2765 	if ((error = namei(&nd)) != 0)
2766 		return (error);
2767 	NDFREE(&nd, NDF_ONLY_PNBUF);
2768 	error = setfflags(td, nd.ni_vp, flags);
2769 	vrele(nd.ni_vp);
2770 	return (error);
2771 }
2772 
2773 /*
2774  * Change flags of a file given a file descriptor.
2775  */
2776 #ifndef _SYS_SYSPROTO_H_
2777 struct fchflags_args {
2778 	int	fd;
2779 	u_long	flags;
2780 };
2781 #endif
2782 int
2783 sys_fchflags(struct thread *td, struct fchflags_args *uap)
2784 {
2785 	struct file *fp;
2786 	int error;
2787 
2788 	AUDIT_ARG_FD(uap->fd);
2789 	AUDIT_ARG_FFLAGS(uap->flags);
2790 	error = getvnode(td, uap->fd, &cap_fchflags_rights,
2791 	    &fp);
2792 	if (error != 0)
2793 		return (error);
2794 #ifdef AUDIT
2795 	vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
2796 	AUDIT_ARG_VNODE1(fp->f_vnode);
2797 	VOP_UNLOCK(fp->f_vnode);
2798 #endif
2799 	error = setfflags(td, fp->f_vnode, uap->flags);
2800 	fdrop(fp, td);
2801 	return (error);
2802 }
2803 
2804 /*
2805  * Common implementation code for chmod(), lchmod() and fchmod().
2806  */
2807 int
2808 setfmode(struct thread *td, struct ucred *cred, struct vnode *vp, int mode)
2809 {
2810 	struct mount *mp;
2811 	struct vattr vattr;
2812 	int error;
2813 
2814 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2815 		return (error);
2816 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2817 	VATTR_NULL(&vattr);
2818 	vattr.va_mode = mode & ALLPERMS;
2819 #ifdef MAC
2820 	error = mac_vnode_check_setmode(cred, vp, vattr.va_mode);
2821 	if (error == 0)
2822 #endif
2823 		error = VOP_SETATTR(vp, &vattr, cred);
2824 	VOP_UNLOCK(vp);
2825 	vn_finished_write(mp);
2826 	return (error);
2827 }
2828 
2829 /*
2830  * Change mode of a file given path name.
2831  */
2832 #ifndef _SYS_SYSPROTO_H_
2833 struct chmod_args {
2834 	char	*path;
2835 	int	mode;
2836 };
2837 #endif
2838 int
2839 sys_chmod(struct thread *td, struct chmod_args *uap)
2840 {
2841 
2842 	return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2843 	    uap->mode, 0));
2844 }
2845 
2846 #ifndef _SYS_SYSPROTO_H_
2847 struct fchmodat_args {
2848 	int	dirfd;
2849 	char	*path;
2850 	mode_t	mode;
2851 	int	flag;
2852 }
2853 #endif
2854 int
2855 sys_fchmodat(struct thread *td, struct fchmodat_args *uap)
2856 {
2857 
2858 	if ((uap->flag & ~(AT_SYMLINK_NOFOLLOW | AT_BENEATH |
2859 	    AT_RESOLVE_BENEATH)) != 0)
2860 		return (EINVAL);
2861 
2862 	return (kern_fchmodat(td, uap->fd, uap->path, UIO_USERSPACE,
2863 	    uap->mode, uap->flag));
2864 }
2865 
2866 /*
2867  * Change mode of a file given path name (don't follow links.)
2868  */
2869 #ifndef _SYS_SYSPROTO_H_
2870 struct lchmod_args {
2871 	char	*path;
2872 	int	mode;
2873 };
2874 #endif
2875 int
2876 sys_lchmod(struct thread *td, struct lchmod_args *uap)
2877 {
2878 
2879 	return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2880 	    uap->mode, AT_SYMLINK_NOFOLLOW));
2881 }
2882 
2883 int
2884 kern_fchmodat(struct thread *td, int fd, const char *path,
2885     enum uio_seg pathseg, mode_t mode, int flag)
2886 {
2887 	struct nameidata nd;
2888 	int error;
2889 
2890 	AUDIT_ARG_MODE(mode);
2891 	NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(flag, AT_SYMLINK_NOFOLLOW |
2892 	    AT_BENEATH | AT_RESOLVE_BENEATH) | AUDITVNODE1, pathseg, path, fd,
2893 	    &cap_fchmod_rights, td);
2894 	if ((error = namei(&nd)) != 0)
2895 		return (error);
2896 	NDFREE(&nd, NDF_ONLY_PNBUF);
2897 	error = setfmode(td, td->td_ucred, nd.ni_vp, mode);
2898 	vrele(nd.ni_vp);
2899 	return (error);
2900 }
2901 
2902 /*
2903  * Change mode of a file given a file descriptor.
2904  */
2905 #ifndef _SYS_SYSPROTO_H_
2906 struct fchmod_args {
2907 	int	fd;
2908 	int	mode;
2909 };
2910 #endif
2911 int
2912 sys_fchmod(struct thread *td, struct fchmod_args *uap)
2913 {
2914 	struct file *fp;
2915 	int error;
2916 
2917 	AUDIT_ARG_FD(uap->fd);
2918 	AUDIT_ARG_MODE(uap->mode);
2919 
2920 	error = fget(td, uap->fd, &cap_fchmod_rights, &fp);
2921 	if (error != 0)
2922 		return (error);
2923 	error = fo_chmod(fp, uap->mode, td->td_ucred, td);
2924 	fdrop(fp, td);
2925 	return (error);
2926 }
2927 
2928 /*
2929  * Common implementation for chown(), lchown(), and fchown()
2930  */
2931 int
2932 setfown(struct thread *td, struct ucred *cred, struct vnode *vp, uid_t uid,
2933     gid_t gid)
2934 {
2935 	struct mount *mp;
2936 	struct vattr vattr;
2937 	int error;
2938 
2939 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2940 		return (error);
2941 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2942 	VATTR_NULL(&vattr);
2943 	vattr.va_uid = uid;
2944 	vattr.va_gid = gid;
2945 #ifdef MAC
2946 	error = mac_vnode_check_setowner(cred, vp, vattr.va_uid,
2947 	    vattr.va_gid);
2948 	if (error == 0)
2949 #endif
2950 		error = VOP_SETATTR(vp, &vattr, cred);
2951 	VOP_UNLOCK(vp);
2952 	vn_finished_write(mp);
2953 	return (error);
2954 }
2955 
2956 /*
2957  * Set ownership given a path name.
2958  */
2959 #ifndef _SYS_SYSPROTO_H_
2960 struct chown_args {
2961 	char	*path;
2962 	int	uid;
2963 	int	gid;
2964 };
2965 #endif
2966 int
2967 sys_chown(struct thread *td, struct chown_args *uap)
2968 {
2969 
2970 	return (kern_fchownat(td, AT_FDCWD, uap->path, UIO_USERSPACE, uap->uid,
2971 	    uap->gid, 0));
2972 }
2973 
2974 #ifndef _SYS_SYSPROTO_H_
2975 struct fchownat_args {
2976 	int fd;
2977 	const char * path;
2978 	uid_t uid;
2979 	gid_t gid;
2980 	int flag;
2981 };
2982 #endif
2983 int
2984 sys_fchownat(struct thread *td, struct fchownat_args *uap)
2985 {
2986 
2987 	if ((uap->flag & ~(AT_SYMLINK_NOFOLLOW | AT_BENEATH |
2988 	    AT_RESOLVE_BENEATH)) != 0)
2989 		return (EINVAL);
2990 
2991 	return (kern_fchownat(td, uap->fd, uap->path, UIO_USERSPACE, uap->uid,
2992 	    uap->gid, uap->flag));
2993 }
2994 
2995 int
2996 kern_fchownat(struct thread *td, int fd, const char *path,
2997     enum uio_seg pathseg, int uid, int gid, int flag)
2998 {
2999 	struct nameidata nd;
3000 	int error;
3001 
3002 	AUDIT_ARG_OWNER(uid, gid);
3003 	NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(flag, AT_SYMLINK_NOFOLLOW |
3004 	    AT_BENEATH | AT_RESOLVE_BENEATH) | AUDITVNODE1, pathseg, path, fd,
3005 	    &cap_fchown_rights, td);
3006 
3007 	if ((error = namei(&nd)) != 0)
3008 		return (error);
3009 	NDFREE(&nd, NDF_ONLY_PNBUF);
3010 	error = setfown(td, td->td_ucred, nd.ni_vp, uid, gid);
3011 	vrele(nd.ni_vp);
3012 	return (error);
3013 }
3014 
3015 /*
3016  * Set ownership given a path name, do not cross symlinks.
3017  */
3018 #ifndef _SYS_SYSPROTO_H_
3019 struct lchown_args {
3020 	char	*path;
3021 	int	uid;
3022 	int	gid;
3023 };
3024 #endif
3025 int
3026 sys_lchown(struct thread *td, struct lchown_args *uap)
3027 {
3028 
3029 	return (kern_fchownat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
3030 	    uap->uid, uap->gid, AT_SYMLINK_NOFOLLOW));
3031 }
3032 
3033 /*
3034  * Set ownership given a file descriptor.
3035  */
3036 #ifndef _SYS_SYSPROTO_H_
3037 struct fchown_args {
3038 	int	fd;
3039 	int	uid;
3040 	int	gid;
3041 };
3042 #endif
3043 int
3044 sys_fchown(struct thread *td, struct fchown_args *uap)
3045 {
3046 	struct file *fp;
3047 	int error;
3048 
3049 	AUDIT_ARG_FD(uap->fd);
3050 	AUDIT_ARG_OWNER(uap->uid, uap->gid);
3051 	error = fget(td, uap->fd, &cap_fchown_rights, &fp);
3052 	if (error != 0)
3053 		return (error);
3054 	error = fo_chown(fp, uap->uid, uap->gid, td->td_ucred, td);
3055 	fdrop(fp, td);
3056 	return (error);
3057 }
3058 
3059 /*
3060  * Common implementation code for utimes(), lutimes(), and futimes().
3061  */
3062 static int
3063 getutimes(const struct timeval *usrtvp, enum uio_seg tvpseg,
3064     struct timespec *tsp)
3065 {
3066 	struct timeval tv[2];
3067 	const struct timeval *tvp;
3068 	int error;
3069 
3070 	if (usrtvp == NULL) {
3071 		vfs_timestamp(&tsp[0]);
3072 		tsp[1] = tsp[0];
3073 	} else {
3074 		if (tvpseg == UIO_SYSSPACE) {
3075 			tvp = usrtvp;
3076 		} else {
3077 			if ((error = copyin(usrtvp, tv, sizeof(tv))) != 0)
3078 				return (error);
3079 			tvp = tv;
3080 		}
3081 
3082 		if (tvp[0].tv_usec < 0 || tvp[0].tv_usec >= 1000000 ||
3083 		    tvp[1].tv_usec < 0 || tvp[1].tv_usec >= 1000000)
3084 			return (EINVAL);
3085 		TIMEVAL_TO_TIMESPEC(&tvp[0], &tsp[0]);
3086 		TIMEVAL_TO_TIMESPEC(&tvp[1], &tsp[1]);
3087 	}
3088 	return (0);
3089 }
3090 
3091 /*
3092  * Common implementation code for futimens(), utimensat().
3093  */
3094 #define	UTIMENS_NULL	0x1
3095 #define	UTIMENS_EXIT	0x2
3096 static int
3097 getutimens(const struct timespec *usrtsp, enum uio_seg tspseg,
3098     struct timespec *tsp, int *retflags)
3099 {
3100 	struct timespec tsnow;
3101 	int error;
3102 
3103 	vfs_timestamp(&tsnow);
3104 	*retflags = 0;
3105 	if (usrtsp == NULL) {
3106 		tsp[0] = tsnow;
3107 		tsp[1] = tsnow;
3108 		*retflags |= UTIMENS_NULL;
3109 		return (0);
3110 	}
3111 	if (tspseg == UIO_SYSSPACE) {
3112 		tsp[0] = usrtsp[0];
3113 		tsp[1] = usrtsp[1];
3114 	} else if ((error = copyin(usrtsp, tsp, sizeof(*tsp) * 2)) != 0)
3115 		return (error);
3116 	if (tsp[0].tv_nsec == UTIME_OMIT && tsp[1].tv_nsec == UTIME_OMIT)
3117 		*retflags |= UTIMENS_EXIT;
3118 	if (tsp[0].tv_nsec == UTIME_NOW && tsp[1].tv_nsec == UTIME_NOW)
3119 		*retflags |= UTIMENS_NULL;
3120 	if (tsp[0].tv_nsec == UTIME_OMIT)
3121 		tsp[0].tv_sec = VNOVAL;
3122 	else if (tsp[0].tv_nsec == UTIME_NOW)
3123 		tsp[0] = tsnow;
3124 	else if (tsp[0].tv_nsec < 0 || tsp[0].tv_nsec >= 1000000000L)
3125 		return (EINVAL);
3126 	if (tsp[1].tv_nsec == UTIME_OMIT)
3127 		tsp[1].tv_sec = VNOVAL;
3128 	else if (tsp[1].tv_nsec == UTIME_NOW)
3129 		tsp[1] = tsnow;
3130 	else if (tsp[1].tv_nsec < 0 || tsp[1].tv_nsec >= 1000000000L)
3131 		return (EINVAL);
3132 
3133 	return (0);
3134 }
3135 
3136 /*
3137  * Common implementation code for utimes(), lutimes(), futimes(), futimens(),
3138  * and utimensat().
3139  */
3140 static int
3141 setutimes(struct thread *td, struct vnode *vp, const struct timespec *ts,
3142     int numtimes, int nullflag)
3143 {
3144 	struct mount *mp;
3145 	struct vattr vattr;
3146 	int error, setbirthtime;
3147 
3148 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
3149 		return (error);
3150 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3151 	setbirthtime = 0;
3152 	if (numtimes < 3 && !VOP_GETATTR(vp, &vattr, td->td_ucred) &&
3153 	    timespeccmp(&ts[1], &vattr.va_birthtime, < ))
3154 		setbirthtime = 1;
3155 	VATTR_NULL(&vattr);
3156 	vattr.va_atime = ts[0];
3157 	vattr.va_mtime = ts[1];
3158 	if (setbirthtime)
3159 		vattr.va_birthtime = ts[1];
3160 	if (numtimes > 2)
3161 		vattr.va_birthtime = ts[2];
3162 	if (nullflag)
3163 		vattr.va_vaflags |= VA_UTIMES_NULL;
3164 #ifdef MAC
3165 	error = mac_vnode_check_setutimes(td->td_ucred, vp, vattr.va_atime,
3166 	    vattr.va_mtime);
3167 #endif
3168 	if (error == 0)
3169 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3170 	VOP_UNLOCK(vp);
3171 	vn_finished_write(mp);
3172 	return (error);
3173 }
3174 
3175 /*
3176  * Set the access and modification times of a file.
3177  */
3178 #ifndef _SYS_SYSPROTO_H_
3179 struct utimes_args {
3180 	char	*path;
3181 	struct	timeval *tptr;
3182 };
3183 #endif
3184 int
3185 sys_utimes(struct thread *td, struct utimes_args *uap)
3186 {
3187 
3188 	return (kern_utimesat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
3189 	    uap->tptr, UIO_USERSPACE));
3190 }
3191 
3192 #ifndef _SYS_SYSPROTO_H_
3193 struct futimesat_args {
3194 	int fd;
3195 	const char * path;
3196 	const struct timeval * times;
3197 };
3198 #endif
3199 int
3200 sys_futimesat(struct thread *td, struct futimesat_args *uap)
3201 {
3202 
3203 	return (kern_utimesat(td, uap->fd, uap->path, UIO_USERSPACE,
3204 	    uap->times, UIO_USERSPACE));
3205 }
3206 
3207 int
3208 kern_utimesat(struct thread *td, int fd, const char *path,
3209     enum uio_seg pathseg, struct timeval *tptr, enum uio_seg tptrseg)
3210 {
3211 	struct nameidata nd;
3212 	struct timespec ts[2];
3213 	int error;
3214 
3215 	if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3216 		return (error);
3217 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, fd,
3218 	    &cap_futimes_rights, td);
3219 
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 lutimes_args {
3233 	char	*path;
3234 	struct	timeval *tptr;
3235 };
3236 #endif
3237 int
3238 sys_lutimes(struct thread *td, struct lutimes_args *uap)
3239 {
3240 
3241 	return (kern_lutimes(td, uap->path, UIO_USERSPACE, uap->tptr,
3242 	    UIO_USERSPACE));
3243 }
3244 
3245 int
3246 kern_lutimes(struct thread *td, const char *path, enum uio_seg pathseg,
3247     struct timeval *tptr, enum uio_seg tptrseg)
3248 {
3249 	struct timespec ts[2];
3250 	struct nameidata nd;
3251 	int error;
3252 
3253 	if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3254 		return (error);
3255 	NDINIT(&nd, LOOKUP, NOFOLLOW | AUDITVNODE1, pathseg, path, td);
3256 	if ((error = namei(&nd)) != 0)
3257 		return (error);
3258 	NDFREE(&nd, NDF_ONLY_PNBUF);
3259 	error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3260 	vrele(nd.ni_vp);
3261 	return (error);
3262 }
3263 
3264 /*
3265  * Set the access and modification times of a file.
3266  */
3267 #ifndef _SYS_SYSPROTO_H_
3268 struct futimes_args {
3269 	int	fd;
3270 	struct	timeval *tptr;
3271 };
3272 #endif
3273 int
3274 sys_futimes(struct thread *td, struct futimes_args *uap)
3275 {
3276 
3277 	return (kern_futimes(td, uap->fd, uap->tptr, UIO_USERSPACE));
3278 }
3279 
3280 int
3281 kern_futimes(struct thread *td, int fd, struct timeval *tptr,
3282     enum uio_seg tptrseg)
3283 {
3284 	struct timespec ts[2];
3285 	struct file *fp;
3286 	int error;
3287 
3288 	AUDIT_ARG_FD(fd);
3289 	error = getutimes(tptr, tptrseg, ts);
3290 	if (error != 0)
3291 		return (error);
3292 	error = getvnode(td, fd, &cap_futimes_rights, &fp);
3293 	if (error != 0)
3294 		return (error);
3295 #ifdef AUDIT
3296 	vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
3297 	AUDIT_ARG_VNODE1(fp->f_vnode);
3298 	VOP_UNLOCK(fp->f_vnode);
3299 #endif
3300 	error = setutimes(td, fp->f_vnode, ts, 2, tptr == NULL);
3301 	fdrop(fp, td);
3302 	return (error);
3303 }
3304 
3305 int
3306 sys_futimens(struct thread *td, struct futimens_args *uap)
3307 {
3308 
3309 	return (kern_futimens(td, uap->fd, uap->times, UIO_USERSPACE));
3310 }
3311 
3312 int
3313 kern_futimens(struct thread *td, int fd, struct timespec *tptr,
3314     enum uio_seg tptrseg)
3315 {
3316 	struct timespec ts[2];
3317 	struct file *fp;
3318 	int error, flags;
3319 
3320 	AUDIT_ARG_FD(fd);
3321 	error = getutimens(tptr, tptrseg, ts, &flags);
3322 	if (error != 0)
3323 		return (error);
3324 	if (flags & UTIMENS_EXIT)
3325 		return (0);
3326 	error = getvnode(td, fd, &cap_futimes_rights, &fp);
3327 	if (error != 0)
3328 		return (error);
3329 #ifdef AUDIT
3330 	vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
3331 	AUDIT_ARG_VNODE1(fp->f_vnode);
3332 	VOP_UNLOCK(fp->f_vnode);
3333 #endif
3334 	error = setutimes(td, fp->f_vnode, ts, 2, flags & UTIMENS_NULL);
3335 	fdrop(fp, td);
3336 	return (error);
3337 }
3338 
3339 int
3340 sys_utimensat(struct thread *td, struct utimensat_args *uap)
3341 {
3342 
3343 	return (kern_utimensat(td, uap->fd, uap->path, UIO_USERSPACE,
3344 	    uap->times, UIO_USERSPACE, uap->flag));
3345 }
3346 
3347 int
3348 kern_utimensat(struct thread *td, int fd, const char *path,
3349     enum uio_seg pathseg, struct timespec *tptr, enum uio_seg tptrseg,
3350     int flag)
3351 {
3352 	struct nameidata nd;
3353 	struct timespec ts[2];
3354 	int error, flags;
3355 
3356 	if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_BENEATH |
3357 	    AT_RESOLVE_BENEATH)) != 0)
3358 		return (EINVAL);
3359 
3360 	if ((error = getutimens(tptr, tptrseg, ts, &flags)) != 0)
3361 		return (error);
3362 	NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(flag, AT_SYMLINK_NOFOLLOW |
3363 	    AT_BENEATH | AT_RESOLVE_BENEATH) | AUDITVNODE1,
3364 	    pathseg, path, fd, &cap_futimes_rights, td);
3365 	if ((error = namei(&nd)) != 0)
3366 		return (error);
3367 	/*
3368 	 * We are allowed to call namei() regardless of 2xUTIME_OMIT.
3369 	 * POSIX states:
3370 	 * "If both tv_nsec fields are UTIME_OMIT... EACCESS may be detected."
3371 	 * "Search permission is denied by a component of the path prefix."
3372 	 */
3373 	NDFREE(&nd, NDF_ONLY_PNBUF);
3374 	if ((flags & UTIMENS_EXIT) == 0)
3375 		error = setutimes(td, nd.ni_vp, ts, 2, flags & UTIMENS_NULL);
3376 	vrele(nd.ni_vp);
3377 	return (error);
3378 }
3379 
3380 /*
3381  * Truncate a file given its path name.
3382  */
3383 #ifndef _SYS_SYSPROTO_H_
3384 struct truncate_args {
3385 	char	*path;
3386 	int	pad;
3387 	off_t	length;
3388 };
3389 #endif
3390 int
3391 sys_truncate(struct thread *td, struct truncate_args *uap)
3392 {
3393 
3394 	return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3395 }
3396 
3397 int
3398 kern_truncate(struct thread *td, const char *path, enum uio_seg pathseg,
3399     off_t length)
3400 {
3401 	struct mount *mp;
3402 	struct vnode *vp;
3403 	void *rl_cookie;
3404 	struct vattr vattr;
3405 	struct nameidata nd;
3406 	int error;
3407 
3408 	if (length < 0)
3409 		return (EINVAL);
3410 retry:
3411 	NDINIT(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, td);
3412 	if ((error = namei(&nd)) != 0)
3413 		return (error);
3414 	vp = nd.ni_vp;
3415 	rl_cookie = vn_rangelock_wlock(vp, 0, OFF_MAX);
3416 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) {
3417 		vn_rangelock_unlock(vp, rl_cookie);
3418 		vrele(vp);
3419 		return (error);
3420 	}
3421 	NDFREE(&nd, NDF_ONLY_PNBUF);
3422 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3423 	if (vp->v_type == VDIR)
3424 		error = EISDIR;
3425 #ifdef MAC
3426 	else if ((error = mac_vnode_check_write(td->td_ucred, NOCRED, vp))) {
3427 	}
3428 #endif
3429 	else if ((error = vn_writechk(vp)) == 0 &&
3430 	    (error = VOP_ACCESS(vp, VWRITE, td->td_ucred, td)) == 0) {
3431 		VATTR_NULL(&vattr);
3432 		vattr.va_size = length;
3433 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3434 	}
3435 	VOP_UNLOCK(vp);
3436 	vn_finished_write(mp);
3437 	vn_rangelock_unlock(vp, rl_cookie);
3438 	vrele(vp);
3439 	if (error == ERELOOKUP)
3440 		goto retry;
3441 	return (error);
3442 }
3443 
3444 #if defined(COMPAT_43)
3445 /*
3446  * Truncate a file given its path name.
3447  */
3448 #ifndef _SYS_SYSPROTO_H_
3449 struct otruncate_args {
3450 	char	*path;
3451 	long	length;
3452 };
3453 #endif
3454 int
3455 otruncate(struct thread *td, struct otruncate_args *uap)
3456 {
3457 
3458 	return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3459 }
3460 #endif /* COMPAT_43 */
3461 
3462 #if defined(COMPAT_FREEBSD6)
3463 /* Versions with the pad argument */
3464 int
3465 freebsd6_truncate(struct thread *td, struct freebsd6_truncate_args *uap)
3466 {
3467 
3468 	return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3469 }
3470 
3471 int
3472 freebsd6_ftruncate(struct thread *td, struct freebsd6_ftruncate_args *uap)
3473 {
3474 
3475 	return (kern_ftruncate(td, uap->fd, uap->length));
3476 }
3477 #endif
3478 
3479 int
3480 kern_fsync(struct thread *td, int fd, bool fullsync)
3481 {
3482 	struct vnode *vp;
3483 	struct mount *mp;
3484 	struct file *fp;
3485 	int error, lock_flags;
3486 
3487 	AUDIT_ARG_FD(fd);
3488 	error = getvnode(td, fd, &cap_fsync_rights, &fp);
3489 	if (error != 0)
3490 		return (error);
3491 	vp = fp->f_vnode;
3492 #if 0
3493 	if (!fullsync)
3494 		/* XXXKIB: compete outstanding aio writes */;
3495 #endif
3496 retry:
3497 	error = vn_start_write(vp, &mp, V_WAIT | PCATCH);
3498 	if (error != 0)
3499 		goto drop;
3500 	if (MNT_SHARED_WRITES(mp) ||
3501 	    ((mp == NULL) && MNT_SHARED_WRITES(vp->v_mount))) {
3502 		lock_flags = LK_SHARED;
3503 	} else {
3504 		lock_flags = LK_EXCLUSIVE;
3505 	}
3506 	vn_lock(vp, lock_flags | LK_RETRY);
3507 	AUDIT_ARG_VNODE1(vp);
3508 	if (vp->v_object != NULL) {
3509 		VM_OBJECT_WLOCK(vp->v_object);
3510 		vm_object_page_clean(vp->v_object, 0, 0, 0);
3511 		VM_OBJECT_WUNLOCK(vp->v_object);
3512 	}
3513 	error = fullsync ? VOP_FSYNC(vp, MNT_WAIT, td) : VOP_FDATASYNC(vp, td);
3514 	VOP_UNLOCK(vp);
3515 	vn_finished_write(mp);
3516 	if (error == ERELOOKUP)
3517 		goto retry;
3518 drop:
3519 	fdrop(fp, td);
3520 	return (error);
3521 }
3522 
3523 /*
3524  * Sync an open file.
3525  */
3526 #ifndef _SYS_SYSPROTO_H_
3527 struct fsync_args {
3528 	int	fd;
3529 };
3530 #endif
3531 int
3532 sys_fsync(struct thread *td, struct fsync_args *uap)
3533 {
3534 
3535 	return (kern_fsync(td, uap->fd, true));
3536 }
3537 
3538 int
3539 sys_fdatasync(struct thread *td, struct fdatasync_args *uap)
3540 {
3541 
3542 	return (kern_fsync(td, uap->fd, false));
3543 }
3544 
3545 /*
3546  * Rename files.  Source and destination must either both be directories, or
3547  * both not be directories.  If target is a directory, it must be empty.
3548  */
3549 #ifndef _SYS_SYSPROTO_H_
3550 struct rename_args {
3551 	char	*from;
3552 	char	*to;
3553 };
3554 #endif
3555 int
3556 sys_rename(struct thread *td, struct rename_args *uap)
3557 {
3558 
3559 	return (kern_renameat(td, AT_FDCWD, uap->from, AT_FDCWD,
3560 	    uap->to, UIO_USERSPACE));
3561 }
3562 
3563 #ifndef _SYS_SYSPROTO_H_
3564 struct renameat_args {
3565 	int	oldfd;
3566 	char	*old;
3567 	int	newfd;
3568 	char	*new;
3569 };
3570 #endif
3571 int
3572 sys_renameat(struct thread *td, struct renameat_args *uap)
3573 {
3574 
3575 	return (kern_renameat(td, uap->oldfd, uap->old, uap->newfd, uap->new,
3576 	    UIO_USERSPACE));
3577 }
3578 
3579 #ifdef MAC
3580 static int
3581 kern_renameat_mac(struct thread *td, int oldfd, const char *old, int newfd,
3582     const char *new, enum uio_seg pathseg, struct nameidata *fromnd)
3583 {
3584 	int error;
3585 
3586 	NDINIT_ATRIGHTS(fromnd, DELETE, LOCKPARENT | LOCKLEAF | SAVESTART |
3587 	    AUDITVNODE1, pathseg, old, oldfd, &cap_renameat_source_rights, td);
3588 	if ((error = namei(fromnd)) != 0)
3589 		return (error);
3590 	error = mac_vnode_check_rename_from(td->td_ucred, fromnd->ni_dvp,
3591 	    fromnd->ni_vp, &fromnd->ni_cnd);
3592 	VOP_UNLOCK(fromnd->ni_dvp);
3593 	if (fromnd->ni_dvp != fromnd->ni_vp)
3594 		VOP_UNLOCK(fromnd->ni_vp);
3595 	if (error != 0) {
3596 		NDFREE(fromnd, NDF_ONLY_PNBUF);
3597 		vrele(fromnd->ni_dvp);
3598 		vrele(fromnd->ni_vp);
3599 		if (fromnd->ni_startdir)
3600 			vrele(fromnd->ni_startdir);
3601 	}
3602 	return (error);
3603 }
3604 #endif
3605 
3606 int
3607 kern_renameat(struct thread *td, int oldfd, const char *old, int newfd,
3608     const char *new, enum uio_seg pathseg)
3609 {
3610 	struct mount *mp = NULL;
3611 	struct vnode *tvp, *fvp, *tdvp;
3612 	struct nameidata fromnd, tond;
3613 	int error;
3614 
3615 again:
3616 	bwillwrite();
3617 #ifdef MAC
3618 	if (mac_vnode_check_rename_from_enabled()) {
3619 		error = kern_renameat_mac(td, oldfd, old, newfd, new, pathseg,
3620 		    &fromnd);
3621 		if (error != 0)
3622 			return (error);
3623 	} else {
3624 #endif
3625 	NDINIT_ATRIGHTS(&fromnd, DELETE, WANTPARENT | SAVESTART | AUDITVNODE1,
3626 	    pathseg, old, oldfd, &cap_renameat_source_rights, td);
3627 	if ((error = namei(&fromnd)) != 0)
3628 		return (error);
3629 #ifdef MAC
3630 	}
3631 #endif
3632 	fvp = fromnd.ni_vp;
3633 	NDINIT_ATRIGHTS(&tond, RENAME, LOCKPARENT | LOCKLEAF | NOCACHE |
3634 	    SAVESTART | AUDITVNODE2, pathseg, new, newfd,
3635 	    &cap_renameat_target_rights, td);
3636 	if (fromnd.ni_vp->v_type == VDIR)
3637 		tond.ni_cnd.cn_flags |= WILLBEDIR;
3638 	if ((error = namei(&tond)) != 0) {
3639 		/* Translate error code for rename("dir1", "dir2/."). */
3640 		if (error == EISDIR && fvp->v_type == VDIR)
3641 			error = EINVAL;
3642 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3643 		vrele(fromnd.ni_dvp);
3644 		vrele(fvp);
3645 		goto out1;
3646 	}
3647 	tdvp = tond.ni_dvp;
3648 	tvp = tond.ni_vp;
3649 	error = vn_start_write(fvp, &mp, V_NOWAIT);
3650 	if (error != 0) {
3651 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3652 		NDFREE(&tond, NDF_ONLY_PNBUF);
3653 		if (tvp != NULL)
3654 			vput(tvp);
3655 		if (tdvp == tvp)
3656 			vrele(tdvp);
3657 		else
3658 			vput(tdvp);
3659 		vrele(fromnd.ni_dvp);
3660 		vrele(fvp);
3661 		vrele(tond.ni_startdir);
3662 		if (fromnd.ni_startdir != NULL)
3663 			vrele(fromnd.ni_startdir);
3664 		error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH);
3665 		if (error != 0)
3666 			return (error);
3667 		goto again;
3668 	}
3669 	if (tvp != NULL) {
3670 		if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
3671 			error = ENOTDIR;
3672 			goto out;
3673 		} else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
3674 			error = EISDIR;
3675 			goto out;
3676 		}
3677 #ifdef CAPABILITIES
3678 		if (newfd != AT_FDCWD && (tond.ni_resflags & NIRES_ABS) == 0) {
3679 			/*
3680 			 * If the target already exists we require CAP_UNLINKAT
3681 			 * from 'newfd', when newfd was used for the lookup.
3682 			 */
3683 			error = cap_check(&tond.ni_filecaps.fc_rights,
3684 			    &cap_unlinkat_rights);
3685 			if (error != 0)
3686 				goto out;
3687 		}
3688 #endif
3689 	}
3690 	if (fvp == tdvp) {
3691 		error = EINVAL;
3692 		goto out;
3693 	}
3694 	/*
3695 	 * If the source is the same as the destination (that is, if they
3696 	 * are links to the same vnode), then there is nothing to do.
3697 	 */
3698 	if (fvp == tvp)
3699 		error = ERESTART;
3700 #ifdef MAC
3701 	else
3702 		error = mac_vnode_check_rename_to(td->td_ucred, tdvp,
3703 		    tond.ni_vp, fromnd.ni_dvp == tdvp, &tond.ni_cnd);
3704 #endif
3705 out:
3706 	if (error == 0) {
3707 		error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
3708 		    tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
3709 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3710 		NDFREE(&tond, NDF_ONLY_PNBUF);
3711 	} else {
3712 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3713 		NDFREE(&tond, NDF_ONLY_PNBUF);
3714 		if (tvp != NULL)
3715 			vput(tvp);
3716 		if (tdvp == tvp)
3717 			vrele(tdvp);
3718 		else
3719 			vput(tdvp);
3720 		vrele(fromnd.ni_dvp);
3721 		vrele(fvp);
3722 	}
3723 	vrele(tond.ni_startdir);
3724 	vn_finished_write(mp);
3725 out1:
3726 	if (fromnd.ni_startdir)
3727 		vrele(fromnd.ni_startdir);
3728 	if (error == ERESTART)
3729 		return (0);
3730 	if (error == ERELOOKUP)
3731 		goto again;
3732 	return (error);
3733 }
3734 
3735 /*
3736  * Make a directory file.
3737  */
3738 #ifndef _SYS_SYSPROTO_H_
3739 struct mkdir_args {
3740 	char	*path;
3741 	int	mode;
3742 };
3743 #endif
3744 int
3745 sys_mkdir(struct thread *td, struct mkdir_args *uap)
3746 {
3747 
3748 	return (kern_mkdirat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
3749 	    uap->mode));
3750 }
3751 
3752 #ifndef _SYS_SYSPROTO_H_
3753 struct mkdirat_args {
3754 	int	fd;
3755 	char	*path;
3756 	mode_t	mode;
3757 };
3758 #endif
3759 int
3760 sys_mkdirat(struct thread *td, struct mkdirat_args *uap)
3761 {
3762 
3763 	return (kern_mkdirat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode));
3764 }
3765 
3766 int
3767 kern_mkdirat(struct thread *td, int fd, const char *path, enum uio_seg segflg,
3768     int mode)
3769 {
3770 	struct mount *mp;
3771 	struct vnode *vp;
3772 	struct vattr vattr;
3773 	struct nameidata nd;
3774 	int error;
3775 
3776 	AUDIT_ARG_MODE(mode);
3777 restart:
3778 	bwillwrite();
3779 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
3780 	    NC_NOMAKEENTRY | NC_KEEPPOSENTRY, segflg, path, fd,
3781 	    &cap_mkdirat_rights, td);
3782 	nd.ni_cnd.cn_flags |= WILLBEDIR;
3783 	if ((error = namei(&nd)) != 0)
3784 		return (error);
3785 	vp = nd.ni_vp;
3786 	if (vp != NULL) {
3787 		NDFREE(&nd, NDF_ONLY_PNBUF);
3788 		/*
3789 		 * XXX namei called with LOCKPARENT but not LOCKLEAF has
3790 		 * the strange behaviour of leaving the vnode unlocked
3791 		 * if the target is the same vnode as the parent.
3792 		 */
3793 		if (vp == nd.ni_dvp)
3794 			vrele(nd.ni_dvp);
3795 		else
3796 			vput(nd.ni_dvp);
3797 		vrele(vp);
3798 		return (EEXIST);
3799 	}
3800 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3801 		NDFREE(&nd, NDF_ONLY_PNBUF);
3802 		vput(nd.ni_dvp);
3803 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3804 			return (error);
3805 		goto restart;
3806 	}
3807 	VATTR_NULL(&vattr);
3808 	vattr.va_type = VDIR;
3809 	vattr.va_mode = (mode & ACCESSPERMS) &~ td->td_proc->p_pd->pd_cmask;
3810 #ifdef MAC
3811 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
3812 	    &vattr);
3813 	if (error != 0)
3814 		goto out;
3815 #endif
3816 	error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
3817 #ifdef MAC
3818 out:
3819 #endif
3820 	NDFREE(&nd, NDF_ONLY_PNBUF);
3821 	vput(nd.ni_dvp);
3822 	if (error == 0)
3823 		vput(nd.ni_vp);
3824 	vn_finished_write(mp);
3825 	if (error == ERELOOKUP)
3826 		goto restart;
3827 	return (error);
3828 }
3829 
3830 /*
3831  * Remove a directory file.
3832  */
3833 #ifndef _SYS_SYSPROTO_H_
3834 struct rmdir_args {
3835 	char	*path;
3836 };
3837 #endif
3838 int
3839 sys_rmdir(struct thread *td, struct rmdir_args *uap)
3840 {
3841 
3842 	return (kern_frmdirat(td, AT_FDCWD, uap->path, FD_NONE, UIO_USERSPACE,
3843 	    0));
3844 }
3845 
3846 int
3847 kern_frmdirat(struct thread *td, int dfd, const char *path, int fd,
3848     enum uio_seg pathseg, int flag)
3849 {
3850 	struct mount *mp;
3851 	struct vnode *vp;
3852 	struct file *fp;
3853 	struct nameidata nd;
3854 	cap_rights_t rights;
3855 	int error;
3856 
3857 	fp = NULL;
3858 	if (fd != FD_NONE) {
3859 		error = getvnode(td, fd, cap_rights_init_one(&rights, CAP_LOOKUP),
3860 		    &fp);
3861 		if (error != 0)
3862 			return (error);
3863 	}
3864 
3865 restart:
3866 	bwillwrite();
3867 	NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1 |
3868 	    at2cnpflags(flag, AT_BENEATH | AT_RESOLVE_BENEATH),
3869 	    pathseg, path, dfd, &cap_unlinkat_rights, td);
3870 	if ((error = namei(&nd)) != 0)
3871 		goto fdout;
3872 	vp = nd.ni_vp;
3873 	if (vp->v_type != VDIR) {
3874 		error = ENOTDIR;
3875 		goto out;
3876 	}
3877 	/*
3878 	 * No rmdir "." please.
3879 	 */
3880 	if (nd.ni_dvp == vp) {
3881 		error = EINVAL;
3882 		goto out;
3883 	}
3884 	/*
3885 	 * The root of a mounted filesystem cannot be deleted.
3886 	 */
3887 	if (vp->v_vflag & VV_ROOT) {
3888 		error = EBUSY;
3889 		goto out;
3890 	}
3891 
3892 	if (fp != NULL && fp->f_vnode != vp) {
3893 		if (VN_IS_DOOMED(fp->f_vnode))
3894 			error = EBADF;
3895 		else
3896 			error = EDEADLK;
3897 		goto out;
3898 	}
3899 
3900 #ifdef MAC
3901 	error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
3902 	    &nd.ni_cnd);
3903 	if (error != 0)
3904 		goto out;
3905 #endif
3906 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3907 		NDFREE(&nd, NDF_ONLY_PNBUF);
3908 		vput(vp);
3909 		if (nd.ni_dvp == vp)
3910 			vrele(nd.ni_dvp);
3911 		else
3912 			vput(nd.ni_dvp);
3913 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3914 			goto fdout;
3915 		goto restart;
3916 	}
3917 	vfs_notify_upper(vp, VFS_NOTIFY_UPPER_UNLINK);
3918 	error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
3919 	vn_finished_write(mp);
3920 out:
3921 	NDFREE(&nd, NDF_ONLY_PNBUF);
3922 	vput(vp);
3923 	if (nd.ni_dvp == vp)
3924 		vrele(nd.ni_dvp);
3925 	else
3926 		vput(nd.ni_dvp);
3927 	if (error == ERELOOKUP)
3928 		goto restart;
3929 fdout:
3930 	if (fp != NULL)
3931 		fdrop(fp, td);
3932 	return (error);
3933 }
3934 
3935 #if defined(COMPAT_43) || defined(COMPAT_FREEBSD11)
3936 int
3937 freebsd11_kern_getdirentries(struct thread *td, int fd, char *ubuf, u_int count,
3938     long *basep, void (*func)(struct freebsd11_dirent *))
3939 {
3940 	struct freebsd11_dirent dstdp;
3941 	struct dirent *dp, *edp;
3942 	char *dirbuf;
3943 	off_t base;
3944 	ssize_t resid, ucount;
3945 	int error;
3946 
3947 	/* XXX arbitrary sanity limit on `count'. */
3948 	count = min(count, 64 * 1024);
3949 
3950 	dirbuf = malloc(count, M_TEMP, M_WAITOK);
3951 
3952 	error = kern_getdirentries(td, fd, dirbuf, count, &base, &resid,
3953 	    UIO_SYSSPACE);
3954 	if (error != 0)
3955 		goto done;
3956 	if (basep != NULL)
3957 		*basep = base;
3958 
3959 	ucount = 0;
3960 	for (dp = (struct dirent *)dirbuf,
3961 	    edp = (struct dirent *)&dirbuf[count - resid];
3962 	    ucount < count && dp < edp; ) {
3963 		if (dp->d_reclen == 0)
3964 			break;
3965 		MPASS(dp->d_reclen >= _GENERIC_DIRLEN(0));
3966 		if (dp->d_namlen >= sizeof(dstdp.d_name))
3967 			continue;
3968 		dstdp.d_type = dp->d_type;
3969 		dstdp.d_namlen = dp->d_namlen;
3970 		dstdp.d_fileno = dp->d_fileno;		/* truncate */
3971 		if (dstdp.d_fileno != dp->d_fileno) {
3972 			switch (ino64_trunc_error) {
3973 			default:
3974 			case 0:
3975 				break;
3976 			case 1:
3977 				error = EOVERFLOW;
3978 				goto done;
3979 			case 2:
3980 				dstdp.d_fileno = UINT32_MAX;
3981 				break;
3982 			}
3983 		}
3984 		dstdp.d_reclen = sizeof(dstdp) - sizeof(dstdp.d_name) +
3985 		    ((dp->d_namlen + 1 + 3) &~ 3);
3986 		bcopy(dp->d_name, dstdp.d_name, dstdp.d_namlen);
3987 		bzero(dstdp.d_name + dstdp.d_namlen,
3988 		    dstdp.d_reclen - offsetof(struct freebsd11_dirent, d_name) -
3989 		    dstdp.d_namlen);
3990 		MPASS(dstdp.d_reclen <= dp->d_reclen);
3991 		MPASS(ucount + dstdp.d_reclen <= count);
3992 		if (func != NULL)
3993 			func(&dstdp);
3994 		error = copyout(&dstdp, ubuf + ucount, dstdp.d_reclen);
3995 		if (error != 0)
3996 			break;
3997 		dp = (struct dirent *)((char *)dp + dp->d_reclen);
3998 		ucount += dstdp.d_reclen;
3999 	}
4000 
4001 done:
4002 	free(dirbuf, M_TEMP);
4003 	if (error == 0)
4004 		td->td_retval[0] = ucount;
4005 	return (error);
4006 }
4007 #endif /* COMPAT */
4008 
4009 #ifdef COMPAT_43
4010 static void
4011 ogetdirentries_cvt(struct freebsd11_dirent *dp)
4012 {
4013 #if (BYTE_ORDER == LITTLE_ENDIAN)
4014 	/*
4015 	 * The expected low byte of dp->d_namlen is our dp->d_type.
4016 	 * The high MBZ byte of dp->d_namlen is our dp->d_namlen.
4017 	 */
4018 	dp->d_type = dp->d_namlen;
4019 	dp->d_namlen = 0;
4020 #else
4021 	/*
4022 	 * The dp->d_type is the high byte of the expected dp->d_namlen,
4023 	 * so must be zero'ed.
4024 	 */
4025 	dp->d_type = 0;
4026 #endif
4027 }
4028 
4029 /*
4030  * Read a block of directory entries in a filesystem independent format.
4031  */
4032 #ifndef _SYS_SYSPROTO_H_
4033 struct ogetdirentries_args {
4034 	int	fd;
4035 	char	*buf;
4036 	u_int	count;
4037 	long	*basep;
4038 };
4039 #endif
4040 int
4041 ogetdirentries(struct thread *td, struct ogetdirentries_args *uap)
4042 {
4043 	long loff;
4044 	int error;
4045 
4046 	error = kern_ogetdirentries(td, uap, &loff);
4047 	if (error == 0)
4048 		error = copyout(&loff, uap->basep, sizeof(long));
4049 	return (error);
4050 }
4051 
4052 int
4053 kern_ogetdirentries(struct thread *td, struct ogetdirentries_args *uap,
4054     long *ploff)
4055 {
4056 	long base;
4057 	int error;
4058 
4059 	/* XXX arbitrary sanity limit on `count'. */
4060 	if (uap->count > 64 * 1024)
4061 		return (EINVAL);
4062 
4063 	error = freebsd11_kern_getdirentries(td, uap->fd, uap->buf, uap->count,
4064 	    &base, ogetdirentries_cvt);
4065 
4066 	if (error == 0 && uap->basep != NULL)
4067 		error = copyout(&base, uap->basep, sizeof(long));
4068 
4069 	return (error);
4070 }
4071 #endif /* COMPAT_43 */
4072 
4073 #if defined(COMPAT_FREEBSD11)
4074 #ifndef _SYS_SYSPROTO_H_
4075 struct freebsd11_getdirentries_args {
4076 	int	fd;
4077 	char	*buf;
4078 	u_int	count;
4079 	long	*basep;
4080 };
4081 #endif
4082 int
4083 freebsd11_getdirentries(struct thread *td,
4084     struct freebsd11_getdirentries_args *uap)
4085 {
4086 	long base;
4087 	int error;
4088 
4089 	error = freebsd11_kern_getdirentries(td, uap->fd, uap->buf, uap->count,
4090 	    &base, NULL);
4091 
4092 	if (error == 0 && uap->basep != NULL)
4093 		error = copyout(&base, uap->basep, sizeof(long));
4094 	return (error);
4095 }
4096 
4097 int
4098 freebsd11_getdents(struct thread *td, struct freebsd11_getdents_args *uap)
4099 {
4100 	struct freebsd11_getdirentries_args ap;
4101 
4102 	ap.fd = uap->fd;
4103 	ap.buf = uap->buf;
4104 	ap.count = uap->count;
4105 	ap.basep = NULL;
4106 	return (freebsd11_getdirentries(td, &ap));
4107 }
4108 #endif /* COMPAT_FREEBSD11 */
4109 
4110 /*
4111  * Read a block of directory entries in a filesystem independent format.
4112  */
4113 int
4114 sys_getdirentries(struct thread *td, struct getdirentries_args *uap)
4115 {
4116 	off_t base;
4117 	int error;
4118 
4119 	error = kern_getdirentries(td, uap->fd, uap->buf, uap->count, &base,
4120 	    NULL, UIO_USERSPACE);
4121 	if (error != 0)
4122 		return (error);
4123 	if (uap->basep != NULL)
4124 		error = copyout(&base, uap->basep, sizeof(off_t));
4125 	return (error);
4126 }
4127 
4128 int
4129 kern_getdirentries(struct thread *td, int fd, char *buf, size_t count,
4130     off_t *basep, ssize_t *residp, enum uio_seg bufseg)
4131 {
4132 	struct vnode *vp;
4133 	struct file *fp;
4134 	struct uio auio;
4135 	struct iovec aiov;
4136 	off_t loff;
4137 	int error, eofflag;
4138 	off_t foffset;
4139 
4140 	AUDIT_ARG_FD(fd);
4141 	if (count > IOSIZE_MAX)
4142 		return (EINVAL);
4143 	auio.uio_resid = count;
4144 	error = getvnode(td, fd, &cap_read_rights, &fp);
4145 	if (error != 0)
4146 		return (error);
4147 	if ((fp->f_flag & FREAD) == 0) {
4148 		fdrop(fp, td);
4149 		return (EBADF);
4150 	}
4151 	vp = fp->f_vnode;
4152 	foffset = foffset_lock(fp, 0);
4153 unionread:
4154 	if (vp->v_type != VDIR) {
4155 		error = EINVAL;
4156 		goto fail;
4157 	}
4158 	aiov.iov_base = buf;
4159 	aiov.iov_len = count;
4160 	auio.uio_iov = &aiov;
4161 	auio.uio_iovcnt = 1;
4162 	auio.uio_rw = UIO_READ;
4163 	auio.uio_segflg = bufseg;
4164 	auio.uio_td = td;
4165 	vn_lock(vp, LK_SHARED | LK_RETRY);
4166 	AUDIT_ARG_VNODE1(vp);
4167 	loff = auio.uio_offset = foffset;
4168 #ifdef MAC
4169 	error = mac_vnode_check_readdir(td->td_ucred, vp);
4170 	if (error == 0)
4171 #endif
4172 		error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, NULL,
4173 		    NULL);
4174 	foffset = auio.uio_offset;
4175 	if (error != 0) {
4176 		VOP_UNLOCK(vp);
4177 		goto fail;
4178 	}
4179 	if (count == auio.uio_resid &&
4180 	    (vp->v_vflag & VV_ROOT) &&
4181 	    (vp->v_mount->mnt_flag & MNT_UNION)) {
4182 		struct vnode *tvp = vp;
4183 
4184 		vp = vp->v_mount->mnt_vnodecovered;
4185 		VREF(vp);
4186 		fp->f_vnode = vp;
4187 		foffset = 0;
4188 		vput(tvp);
4189 		goto unionread;
4190 	}
4191 	VOP_UNLOCK(vp);
4192 	*basep = loff;
4193 	if (residp != NULL)
4194 		*residp = auio.uio_resid;
4195 	td->td_retval[0] = count - auio.uio_resid;
4196 fail:
4197 	foffset_unlock(fp, foffset, 0);
4198 	fdrop(fp, td);
4199 	return (error);
4200 }
4201 
4202 /*
4203  * Set the mode mask for creation of filesystem nodes.
4204  */
4205 #ifndef _SYS_SYSPROTO_H_
4206 struct umask_args {
4207 	int	newmask;
4208 };
4209 #endif
4210 int
4211 sys_umask(struct thread *td, struct umask_args *uap)
4212 {
4213 	struct pwddesc *pdp;
4214 
4215 	pdp = td->td_proc->p_pd;
4216 	PWDDESC_XLOCK(pdp);
4217 	td->td_retval[0] = pdp->pd_cmask;
4218 	pdp->pd_cmask = uap->newmask & ALLPERMS;
4219 	PWDDESC_XUNLOCK(pdp);
4220 	return (0);
4221 }
4222 
4223 /*
4224  * Void all references to file by ripping underlying filesystem away from
4225  * vnode.
4226  */
4227 #ifndef _SYS_SYSPROTO_H_
4228 struct revoke_args {
4229 	char	*path;
4230 };
4231 #endif
4232 int
4233 sys_revoke(struct thread *td, struct revoke_args *uap)
4234 {
4235 	struct vnode *vp;
4236 	struct vattr vattr;
4237 	struct nameidata nd;
4238 	int error;
4239 
4240 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
4241 	    uap->path, td);
4242 	if ((error = namei(&nd)) != 0)
4243 		return (error);
4244 	vp = nd.ni_vp;
4245 	NDFREE(&nd, NDF_ONLY_PNBUF);
4246 	if (vp->v_type != VCHR || vp->v_rdev == NULL) {
4247 		error = EINVAL;
4248 		goto out;
4249 	}
4250 #ifdef MAC
4251 	error = mac_vnode_check_revoke(td->td_ucred, vp);
4252 	if (error != 0)
4253 		goto out;
4254 #endif
4255 	error = VOP_GETATTR(vp, &vattr, td->td_ucred);
4256 	if (error != 0)
4257 		goto out;
4258 	if (td->td_ucred->cr_uid != vattr.va_uid) {
4259 		error = priv_check(td, PRIV_VFS_ADMIN);
4260 		if (error != 0)
4261 			goto out;
4262 	}
4263 	if (devfs_usecount(vp) > 0)
4264 		VOP_REVOKE(vp, REVOKEALL);
4265 out:
4266 	vput(vp);
4267 	return (error);
4268 }
4269 
4270 /*
4271  * Convert a user file descriptor to a kernel file entry and check that, if it
4272  * is a capability, the correct rights are present. A reference on the file
4273  * entry is held upon returning.
4274  */
4275 int
4276 getvnode(struct thread *td, int fd, cap_rights_t *rightsp, struct file **fpp)
4277 {
4278 	struct file *fp;
4279 	int error;
4280 
4281 	error = fget_unlocked(td->td_proc->p_fd, fd, rightsp, &fp);
4282 	if (error != 0)
4283 		return (error);
4284 
4285 	/*
4286 	 * The file could be not of the vnode type, or it may be not
4287 	 * yet fully initialized, in which case the f_vnode pointer
4288 	 * may be set, but f_ops is still badfileops.  E.g.,
4289 	 * devfs_open() transiently create such situation to
4290 	 * facilitate csw d_fdopen().
4291 	 *
4292 	 * Dupfdopen() handling in kern_openat() installs the
4293 	 * half-baked file into the process descriptor table, allowing
4294 	 * other thread to dereference it. Guard against the race by
4295 	 * checking f_ops.
4296 	 */
4297 	if (fp->f_vnode == NULL || fp->f_ops == &badfileops) {
4298 		fdrop(fp, td);
4299 		return (EINVAL);
4300 	}
4301 	*fpp = fp;
4302 	return (0);
4303 }
4304 
4305 /*
4306  * Get an (NFS) file handle.
4307  */
4308 #ifndef _SYS_SYSPROTO_H_
4309 struct lgetfh_args {
4310 	char *fname;
4311 	fhandle_t *fhp;
4312 };
4313 #endif
4314 int
4315 sys_lgetfh(struct thread *td, struct lgetfh_args *uap)
4316 {
4317 
4318 	return (kern_getfhat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->fname,
4319 	    UIO_USERSPACE, uap->fhp, UIO_USERSPACE));
4320 }
4321 
4322 #ifndef _SYS_SYSPROTO_H_
4323 struct getfh_args {
4324 	char *fname;
4325 	fhandle_t *fhp;
4326 };
4327 #endif
4328 int
4329 sys_getfh(struct thread *td, struct getfh_args *uap)
4330 {
4331 
4332 	return (kern_getfhat(td, 0, AT_FDCWD, uap->fname, UIO_USERSPACE,
4333 	    uap->fhp, UIO_USERSPACE));
4334 }
4335 
4336 /*
4337  * syscall for the rpc.lockd to use to translate an open descriptor into
4338  * a NFS file handle.
4339  *
4340  * warning: do not remove the priv_check() call or this becomes one giant
4341  * security hole.
4342  */
4343 #ifndef _SYS_SYSPROTO_H_
4344 struct getfhat_args {
4345 	int fd;
4346 	char *path;
4347 	fhandle_t *fhp;
4348 	int flags;
4349 };
4350 #endif
4351 int
4352 sys_getfhat(struct thread *td, struct getfhat_args *uap)
4353 {
4354 
4355 	if ((uap->flags & ~(AT_SYMLINK_NOFOLLOW | AT_BENEATH |
4356 	    AT_RESOLVE_BENEATH)) != 0)
4357 		return (EINVAL);
4358 	return (kern_getfhat(td, uap->flags, uap->fd, uap->path, UIO_USERSPACE,
4359 	    uap->fhp, UIO_USERSPACE));
4360 }
4361 
4362 int
4363 kern_getfhat(struct thread *td, int flags, int fd, const char *path,
4364     enum uio_seg pathseg, fhandle_t *fhp, enum uio_seg fhseg)
4365 {
4366 	struct nameidata nd;
4367 	fhandle_t fh;
4368 	struct vnode *vp;
4369 	int error;
4370 
4371 	error = priv_check(td, PRIV_VFS_GETFH);
4372 	if (error != 0)
4373 		return (error);
4374 	NDINIT_AT(&nd, LOOKUP, at2cnpflags(flags, AT_SYMLINK_NOFOLLOW |
4375 	    AT_BENEATH | AT_RESOLVE_BENEATH) | LOCKLEAF | AUDITVNODE1,
4376 	    pathseg, path, fd, td);
4377 	error = namei(&nd);
4378 	if (error != 0)
4379 		return (error);
4380 	NDFREE(&nd, NDF_ONLY_PNBUF);
4381 	vp = nd.ni_vp;
4382 	bzero(&fh, sizeof(fh));
4383 	fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
4384 	error = VOP_VPTOFH(vp, &fh.fh_fid);
4385 	vput(vp);
4386 	if (error == 0) {
4387 		if (fhseg == UIO_USERSPACE)
4388 			error = copyout(&fh, fhp, sizeof (fh));
4389 		else
4390 			memcpy(fhp, &fh, sizeof(fh));
4391 	}
4392 	return (error);
4393 }
4394 
4395 #ifndef _SYS_SYSPROTO_H_
4396 struct fhlink_args {
4397 	fhandle_t *fhp;
4398 	const char *to;
4399 };
4400 #endif
4401 int
4402 sys_fhlink(struct thread *td, struct fhlink_args *uap)
4403 {
4404 
4405 	return (kern_fhlinkat(td, AT_FDCWD, uap->to, UIO_USERSPACE, uap->fhp));
4406 }
4407 
4408 #ifndef _SYS_SYSPROTO_H_
4409 struct fhlinkat_args {
4410 	fhandle_t *fhp;
4411 	int tofd;
4412 	const char *to;
4413 };
4414 #endif
4415 int
4416 sys_fhlinkat(struct thread *td, struct fhlinkat_args *uap)
4417 {
4418 
4419 	return (kern_fhlinkat(td, uap->tofd, uap->to, UIO_USERSPACE, uap->fhp));
4420 }
4421 
4422 static int
4423 kern_fhlinkat(struct thread *td, int fd, const char *path,
4424     enum uio_seg pathseg, fhandle_t *fhp)
4425 {
4426 	fhandle_t fh;
4427 	struct mount *mp;
4428 	struct vnode *vp;
4429 	int error;
4430 
4431 	error = priv_check(td, PRIV_VFS_GETFH);
4432 	if (error != 0)
4433 		return (error);
4434 	error = copyin(fhp, &fh, sizeof(fh));
4435 	if (error != 0)
4436 		return (error);
4437 	do {
4438 		bwillwrite();
4439 		if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4440 			return (ESTALE);
4441 		error = VFS_FHTOVP(mp, &fh.fh_fid, LK_SHARED, &vp);
4442 		vfs_unbusy(mp);
4443 		if (error != 0)
4444 			return (error);
4445 		VOP_UNLOCK(vp);
4446 		error = kern_linkat_vp(td, vp, fd, path, pathseg);
4447 	} while (error == EAGAIN || error == ERELOOKUP);
4448 	return (error);
4449 }
4450 
4451 #ifndef _SYS_SYSPROTO_H_
4452 struct fhreadlink_args {
4453 	fhandle_t *fhp;
4454 	char *buf;
4455 	size_t bufsize;
4456 };
4457 #endif
4458 int
4459 sys_fhreadlink(struct thread *td, struct fhreadlink_args *uap)
4460 {
4461 	fhandle_t fh;
4462 	struct mount *mp;
4463 	struct vnode *vp;
4464 	int error;
4465 
4466 	error = priv_check(td, PRIV_VFS_GETFH);
4467 	if (error != 0)
4468 		return (error);
4469 	if (uap->bufsize > IOSIZE_MAX)
4470 		return (EINVAL);
4471 	error = copyin(uap->fhp, &fh, sizeof(fh));
4472 	if (error != 0)
4473 		return (error);
4474 	if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4475 		return (ESTALE);
4476 	error = VFS_FHTOVP(mp, &fh.fh_fid, LK_SHARED, &vp);
4477 	vfs_unbusy(mp);
4478 	if (error != 0)
4479 		return (error);
4480 	error = kern_readlink_vp(vp, uap->buf, UIO_USERSPACE, uap->bufsize, td);
4481 	vput(vp);
4482 	return (error);
4483 }
4484 
4485 /*
4486  * syscall for the rpc.lockd to use to translate a NFS file handle into an
4487  * open descriptor.
4488  *
4489  * warning: do not remove the priv_check() call or this becomes one giant
4490  * security hole.
4491  */
4492 #ifndef _SYS_SYSPROTO_H_
4493 struct fhopen_args {
4494 	const struct fhandle *u_fhp;
4495 	int flags;
4496 };
4497 #endif
4498 int
4499 sys_fhopen(struct thread *td, struct fhopen_args *uap)
4500 {
4501 	return (kern_fhopen(td, uap->u_fhp, uap->flags));
4502 }
4503 
4504 int
4505 kern_fhopen(struct thread *td, const struct fhandle *u_fhp, int flags)
4506 {
4507 	struct mount *mp;
4508 	struct vnode *vp;
4509 	struct fhandle fhp;
4510 	struct file *fp;
4511 	int fmode, error;
4512 	int indx;
4513 
4514 	error = priv_check(td, PRIV_VFS_FHOPEN);
4515 	if (error != 0)
4516 		return (error);
4517 	indx = -1;
4518 	fmode = FFLAGS(flags);
4519 	/* why not allow a non-read/write open for our lockd? */
4520 	if (((fmode & (FREAD | FWRITE)) == 0) || (fmode & O_CREAT))
4521 		return (EINVAL);
4522 	error = copyin(u_fhp, &fhp, sizeof(fhp));
4523 	if (error != 0)
4524 		return(error);
4525 	/* find the mount point */
4526 	mp = vfs_busyfs(&fhp.fh_fsid);
4527 	if (mp == NULL)
4528 		return (ESTALE);
4529 	/* now give me my vnode, it gets returned to me locked */
4530 	error = VFS_FHTOVP(mp, &fhp.fh_fid, LK_EXCLUSIVE, &vp);
4531 	vfs_unbusy(mp);
4532 	if (error != 0)
4533 		return (error);
4534 
4535 	error = falloc_noinstall(td, &fp);
4536 	if (error != 0) {
4537 		vput(vp);
4538 		return (error);
4539 	}
4540 	/*
4541 	 * An extra reference on `fp' has been held for us by
4542 	 * falloc_noinstall().
4543 	 */
4544 
4545 #ifdef INVARIANTS
4546 	td->td_dupfd = -1;
4547 #endif
4548 	error = vn_open_vnode(vp, fmode, td->td_ucred, td, fp);
4549 	if (error != 0) {
4550 		KASSERT(fp->f_ops == &badfileops,
4551 		    ("VOP_OPEN in fhopen() set f_ops"));
4552 		KASSERT(td->td_dupfd < 0,
4553 		    ("fhopen() encountered fdopen()"));
4554 
4555 		vput(vp);
4556 		goto bad;
4557 	}
4558 #ifdef INVARIANTS
4559 	td->td_dupfd = 0;
4560 #endif
4561 	fp->f_vnode = vp;
4562 	finit_vnode(fp, fmode, NULL, &vnops);
4563 	VOP_UNLOCK(vp);
4564 	if ((fmode & O_TRUNC) != 0) {
4565 		error = fo_truncate(fp, 0, td->td_ucred, td);
4566 		if (error != 0)
4567 			goto bad;
4568 	}
4569 
4570 	error = finstall(td, fp, &indx, fmode, NULL);
4571 bad:
4572 	fdrop(fp, td);
4573 	td->td_retval[0] = indx;
4574 	return (error);
4575 }
4576 
4577 /*
4578  * Stat an (NFS) file handle.
4579  */
4580 #ifndef _SYS_SYSPROTO_H_
4581 struct fhstat_args {
4582 	struct fhandle *u_fhp;
4583 	struct stat *sb;
4584 };
4585 #endif
4586 int
4587 sys_fhstat(struct thread *td, struct fhstat_args *uap)
4588 {
4589 	struct stat sb;
4590 	struct fhandle fh;
4591 	int error;
4592 
4593 	error = copyin(uap->u_fhp, &fh, sizeof(fh));
4594 	if (error != 0)
4595 		return (error);
4596 	error = kern_fhstat(td, fh, &sb);
4597 	if (error == 0)
4598 		error = copyout(&sb, uap->sb, sizeof(sb));
4599 	return (error);
4600 }
4601 
4602 int
4603 kern_fhstat(struct thread *td, struct fhandle fh, struct stat *sb)
4604 {
4605 	struct mount *mp;
4606 	struct vnode *vp;
4607 	int error;
4608 
4609 	error = priv_check(td, PRIV_VFS_FHSTAT);
4610 	if (error != 0)
4611 		return (error);
4612 	if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4613 		return (ESTALE);
4614 	error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4615 	vfs_unbusy(mp);
4616 	if (error != 0)
4617 		return (error);
4618 	error = VOP_STAT(vp, sb, td->td_ucred, NOCRED, td);
4619 	vput(vp);
4620 	return (error);
4621 }
4622 
4623 /*
4624  * Implement fstatfs() for (NFS) file handles.
4625  */
4626 #ifndef _SYS_SYSPROTO_H_
4627 struct fhstatfs_args {
4628 	struct fhandle *u_fhp;
4629 	struct statfs *buf;
4630 };
4631 #endif
4632 int
4633 sys_fhstatfs(struct thread *td, struct fhstatfs_args *uap)
4634 {
4635 	struct statfs *sfp;
4636 	fhandle_t fh;
4637 	int error;
4638 
4639 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
4640 	if (error != 0)
4641 		return (error);
4642 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
4643 	error = kern_fhstatfs(td, fh, sfp);
4644 	if (error == 0)
4645 		error = copyout(sfp, uap->buf, sizeof(*sfp));
4646 	free(sfp, M_STATFS);
4647 	return (error);
4648 }
4649 
4650 int
4651 kern_fhstatfs(struct thread *td, fhandle_t fh, struct statfs *buf)
4652 {
4653 	struct mount *mp;
4654 	struct vnode *vp;
4655 	int error;
4656 
4657 	error = priv_check(td, PRIV_VFS_FHSTATFS);
4658 	if (error != 0)
4659 		return (error);
4660 	if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4661 		return (ESTALE);
4662 	error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4663 	if (error != 0) {
4664 		vfs_unbusy(mp);
4665 		return (error);
4666 	}
4667 	vput(vp);
4668 	error = prison_canseemount(td->td_ucred, mp);
4669 	if (error != 0)
4670 		goto out;
4671 #ifdef MAC
4672 	error = mac_mount_check_stat(td->td_ucred, mp);
4673 	if (error != 0)
4674 		goto out;
4675 #endif
4676 	error = VFS_STATFS(mp, buf);
4677 out:
4678 	vfs_unbusy(mp);
4679 	return (error);
4680 }
4681 
4682 /*
4683  * Unlike madvise(2), we do not make a best effort to remember every
4684  * possible caching hint.  Instead, we remember the last setting with
4685  * the exception that we will allow POSIX_FADV_NORMAL to adjust the
4686  * region of any current setting.
4687  */
4688 int
4689 kern_posix_fadvise(struct thread *td, int fd, off_t offset, off_t len,
4690     int advice)
4691 {
4692 	struct fadvise_info *fa, *new;
4693 	struct file *fp;
4694 	struct vnode *vp;
4695 	off_t end;
4696 	int error;
4697 
4698 	if (offset < 0 || len < 0 || offset > OFF_MAX - len)
4699 		return (EINVAL);
4700 	AUDIT_ARG_VALUE(advice);
4701 	switch (advice) {
4702 	case POSIX_FADV_SEQUENTIAL:
4703 	case POSIX_FADV_RANDOM:
4704 	case POSIX_FADV_NOREUSE:
4705 		new = malloc(sizeof(*fa), M_FADVISE, M_WAITOK);
4706 		break;
4707 	case POSIX_FADV_NORMAL:
4708 	case POSIX_FADV_WILLNEED:
4709 	case POSIX_FADV_DONTNEED:
4710 		new = NULL;
4711 		break;
4712 	default:
4713 		return (EINVAL);
4714 	}
4715 	/* XXX: CAP_POSIX_FADVISE? */
4716 	AUDIT_ARG_FD(fd);
4717 	error = fget(td, fd, &cap_no_rights, &fp);
4718 	if (error != 0)
4719 		goto out;
4720 	AUDIT_ARG_FILE(td->td_proc, fp);
4721 	if ((fp->f_ops->fo_flags & DFLAG_SEEKABLE) == 0) {
4722 		error = ESPIPE;
4723 		goto out;
4724 	}
4725 	if (fp->f_type != DTYPE_VNODE) {
4726 		error = ENODEV;
4727 		goto out;
4728 	}
4729 	vp = fp->f_vnode;
4730 	if (vp->v_type != VREG) {
4731 		error = ENODEV;
4732 		goto out;
4733 	}
4734 	if (len == 0)
4735 		end = OFF_MAX;
4736 	else
4737 		end = offset + len - 1;
4738 	switch (advice) {
4739 	case POSIX_FADV_SEQUENTIAL:
4740 	case POSIX_FADV_RANDOM:
4741 	case POSIX_FADV_NOREUSE:
4742 		/*
4743 		 * Try to merge any existing non-standard region with
4744 		 * this new region if possible, otherwise create a new
4745 		 * non-standard region for this request.
4746 		 */
4747 		mtx_pool_lock(mtxpool_sleep, fp);
4748 		fa = fp->f_advice;
4749 		if (fa != NULL && fa->fa_advice == advice &&
4750 		    ((fa->fa_start <= end && fa->fa_end >= offset) ||
4751 		    (end != OFF_MAX && fa->fa_start == end + 1) ||
4752 		    (fa->fa_end != OFF_MAX && fa->fa_end + 1 == offset))) {
4753 			if (offset < fa->fa_start)
4754 				fa->fa_start = offset;
4755 			if (end > fa->fa_end)
4756 				fa->fa_end = end;
4757 		} else {
4758 			new->fa_advice = advice;
4759 			new->fa_start = offset;
4760 			new->fa_end = end;
4761 			fp->f_advice = new;
4762 			new = fa;
4763 		}
4764 		mtx_pool_unlock(mtxpool_sleep, fp);
4765 		break;
4766 	case POSIX_FADV_NORMAL:
4767 		/*
4768 		 * If a the "normal" region overlaps with an existing
4769 		 * non-standard region, trim or remove the
4770 		 * non-standard region.
4771 		 */
4772 		mtx_pool_lock(mtxpool_sleep, fp);
4773 		fa = fp->f_advice;
4774 		if (fa != NULL) {
4775 			if (offset <= fa->fa_start && end >= fa->fa_end) {
4776 				new = fa;
4777 				fp->f_advice = NULL;
4778 			} else if (offset <= fa->fa_start &&
4779 			    end >= fa->fa_start)
4780 				fa->fa_start = end + 1;
4781 			else if (offset <= fa->fa_end && end >= fa->fa_end)
4782 				fa->fa_end = offset - 1;
4783 			else if (offset >= fa->fa_start && end <= fa->fa_end) {
4784 				/*
4785 				 * If the "normal" region is a middle
4786 				 * portion of the existing
4787 				 * non-standard region, just remove
4788 				 * the whole thing rather than picking
4789 				 * one side or the other to
4790 				 * preserve.
4791 				 */
4792 				new = fa;
4793 				fp->f_advice = NULL;
4794 			}
4795 		}
4796 		mtx_pool_unlock(mtxpool_sleep, fp);
4797 		break;
4798 	case POSIX_FADV_WILLNEED:
4799 	case POSIX_FADV_DONTNEED:
4800 		error = VOP_ADVISE(vp, offset, end, advice);
4801 		break;
4802 	}
4803 out:
4804 	if (fp != NULL)
4805 		fdrop(fp, td);
4806 	free(new, M_FADVISE);
4807 	return (error);
4808 }
4809 
4810 int
4811 sys_posix_fadvise(struct thread *td, struct posix_fadvise_args *uap)
4812 {
4813 	int error;
4814 
4815 	error = kern_posix_fadvise(td, uap->fd, uap->offset, uap->len,
4816 	    uap->advice);
4817 	return (kern_posix_error(td, error));
4818 }
4819 
4820 int
4821 kern_copy_file_range(struct thread *td, int infd, off_t *inoffp, int outfd,
4822     off_t *outoffp, size_t len, unsigned int flags)
4823 {
4824 	struct file *infp, *outfp;
4825 	struct vnode *invp, *outvp;
4826 	int error;
4827 	size_t retlen;
4828 	void *rl_rcookie, *rl_wcookie;
4829 	off_t savinoff, savoutoff;
4830 
4831 	infp = outfp = NULL;
4832 	rl_rcookie = rl_wcookie = NULL;
4833 	savinoff = -1;
4834 	error = 0;
4835 	retlen = 0;
4836 
4837 	if (flags != 0) {
4838 		error = EINVAL;
4839 		goto out;
4840 	}
4841 	if (len > SSIZE_MAX)
4842 		/*
4843 		 * Although the len argument is size_t, the return argument
4844 		 * is ssize_t (which is signed).  Therefore a size that won't
4845 		 * fit in ssize_t can't be returned.
4846 		 */
4847 		len = SSIZE_MAX;
4848 
4849 	/* Get the file structures for the file descriptors. */
4850 	error = fget_read(td, infd, &cap_read_rights, &infp);
4851 	if (error != 0)
4852 		goto out;
4853 	if (infp->f_ops == &badfileops) {
4854 		error = EBADF;
4855 		goto out;
4856 	}
4857 	if (infp->f_vnode == NULL) {
4858 		error = EINVAL;
4859 		goto out;
4860 	}
4861 	error = fget_write(td, outfd, &cap_write_rights, &outfp);
4862 	if (error != 0)
4863 		goto out;
4864 	if (outfp->f_ops == &badfileops) {
4865 		error = EBADF;
4866 		goto out;
4867 	}
4868 	if (outfp->f_vnode == NULL) {
4869 		error = EINVAL;
4870 		goto out;
4871 	}
4872 
4873 	/* Set the offset pointers to the correct place. */
4874 	if (inoffp == NULL)
4875 		inoffp = &infp->f_offset;
4876 	if (outoffp == NULL)
4877 		outoffp = &outfp->f_offset;
4878 	savinoff = *inoffp;
4879 	savoutoff = *outoffp;
4880 
4881 	invp = infp->f_vnode;
4882 	outvp = outfp->f_vnode;
4883 	/* Sanity check the f_flag bits. */
4884 	if ((outfp->f_flag & (FWRITE | FAPPEND)) != FWRITE ||
4885 	    (infp->f_flag & FREAD) == 0) {
4886 		error = EBADF;
4887 		goto out;
4888 	}
4889 
4890 	/* If len == 0, just return 0. */
4891 	if (len == 0)
4892 		goto out;
4893 
4894 	/*
4895 	 * If infp and outfp refer to the same file, the byte ranges cannot
4896 	 * overlap.
4897 	 */
4898 	if (invp == outvp && ((savinoff <= savoutoff && savinoff + len >
4899 	    savoutoff) || (savinoff > savoutoff && savoutoff + len >
4900 	    savinoff))) {
4901 		error = EINVAL;
4902 		goto out;
4903 	}
4904 
4905 	/* Range lock the byte ranges for both invp and outvp. */
4906 	for (;;) {
4907 		rl_wcookie = vn_rangelock_wlock(outvp, *outoffp, *outoffp +
4908 		    len);
4909 		rl_rcookie = vn_rangelock_tryrlock(invp, *inoffp, *inoffp +
4910 		    len);
4911 		if (rl_rcookie != NULL)
4912 			break;
4913 		vn_rangelock_unlock(outvp, rl_wcookie);
4914 		rl_rcookie = vn_rangelock_rlock(invp, *inoffp, *inoffp + len);
4915 		vn_rangelock_unlock(invp, rl_rcookie);
4916 	}
4917 
4918 	retlen = len;
4919 	error = vn_copy_file_range(invp, inoffp, outvp, outoffp, &retlen,
4920 	    flags, infp->f_cred, outfp->f_cred, td);
4921 out:
4922 	if (rl_rcookie != NULL)
4923 		vn_rangelock_unlock(invp, rl_rcookie);
4924 	if (rl_wcookie != NULL)
4925 		vn_rangelock_unlock(outvp, rl_wcookie);
4926 	if (savinoff != -1 && (error == EINTR || error == ERESTART)) {
4927 		*inoffp = savinoff;
4928 		*outoffp = savoutoff;
4929 	}
4930 	if (outfp != NULL)
4931 		fdrop(outfp, td);
4932 	if (infp != NULL)
4933 		fdrop(infp, td);
4934 	td->td_retval[0] = retlen;
4935 	return (error);
4936 }
4937 
4938 int
4939 sys_copy_file_range(struct thread *td, struct copy_file_range_args *uap)
4940 {
4941 	off_t inoff, outoff, *inoffp, *outoffp;
4942 	int error;
4943 
4944 	inoffp = outoffp = NULL;
4945 	if (uap->inoffp != NULL) {
4946 		error = copyin(uap->inoffp, &inoff, sizeof(off_t));
4947 		if (error != 0)
4948 			return (error);
4949 		inoffp = &inoff;
4950 	}
4951 	if (uap->outoffp != NULL) {
4952 		error = copyin(uap->outoffp, &outoff, sizeof(off_t));
4953 		if (error != 0)
4954 			return (error);
4955 		outoffp = &outoff;
4956 	}
4957 	error = kern_copy_file_range(td, uap->infd, inoffp, uap->outfd,
4958 	    outoffp, uap->len, uap->flags);
4959 	if (error == 0 && uap->inoffp != NULL)
4960 		error = copyout(inoffp, uap->inoffp, sizeof(off_t));
4961 	if (error == 0 && uap->outoffp != NULL)
4962 		error = copyout(outoffp, uap->outoffp, sizeof(off_t));
4963 	return (error);
4964 }
4965