xref: /freebsd/sys/kern/vfs_syscalls.c (revision 5596f836e7e04a272113e57b3a80f1f67c0fec7f)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (c) 1989, 1993
5  *	The Regents of the University of California.  All rights reserved.
6  * (c) UNIX System Laboratories, Inc.
7  * All or some portions of this file are derived from material licensed
8  * to the University of California by American Telephone and Telegraph
9  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
10  * the permission of UNIX System Laboratories, Inc.
11  *
12  * Redistribution and use in source and binary forms, with or without
13  * modification, are permitted provided that the following conditions
14  * are met:
15  * 1. Redistributions of source code must retain the above copyright
16  *    notice, this list of conditions and the following disclaimer.
17  * 2. Redistributions in binary form must reproduce the above copyright
18  *    notice, this list of conditions and the following disclaimer in the
19  *    documentation and/or other materials provided with the distribution.
20  * 3. Neither the name of the University nor the names of its contributors
21  *    may be used to endorse or promote products derived from this software
22  *    without specific prior written permission.
23  *
24  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
25  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
28  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
29  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
30  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
31  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
32  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
33  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
34  * SUCH DAMAGE.
35  *
36  *	@(#)vfs_syscalls.c	8.13 (Berkeley) 4/15/94
37  */
38 
39 #include <sys/cdefs.h>
40 __FBSDID("$FreeBSD$");
41 
42 #include "opt_capsicum.h"
43 #include "opt_ktrace.h"
44 
45 #include <sys/param.h>
46 #include <sys/systm.h>
47 #include <sys/bio.h>
48 #include <sys/buf.h>
49 #include <sys/capsicum.h>
50 #include <sys/disk.h>
51 #include <sys/sysent.h>
52 #include <sys/malloc.h>
53 #include <sys/mount.h>
54 #include <sys/mutex.h>
55 #include <sys/sysproto.h>
56 #include <sys/namei.h>
57 #include <sys/filedesc.h>
58 #include <sys/kernel.h>
59 #include <sys/fcntl.h>
60 #include <sys/file.h>
61 #include <sys/filio.h>
62 #include <sys/limits.h>
63 #include <sys/linker.h>
64 #include <sys/rwlock.h>
65 #include <sys/sdt.h>
66 #include <sys/stat.h>
67 #include <sys/sx.h>
68 #include <sys/unistd.h>
69 #include <sys/vnode.h>
70 #include <sys/priv.h>
71 #include <sys/proc.h>
72 #include <sys/dirent.h>
73 #include <sys/jail.h>
74 #include <sys/syscallsubr.h>
75 #include <sys/sysctl.h>
76 #ifdef KTRACE
77 #include <sys/ktrace.h>
78 #endif
79 
80 #include <machine/stdarg.h>
81 
82 #include <security/audit/audit.h>
83 #include <security/mac/mac_framework.h>
84 
85 #include <vm/vm.h>
86 #include <vm/vm_object.h>
87 #include <vm/vm_page.h>
88 #include <vm/uma.h>
89 
90 #include <fs/devfs/devfs.h>
91 
92 #include <ufs/ufs/quota.h>
93 
94 MALLOC_DEFINE(M_FADVISE, "fadvise", "posix_fadvise(2) information");
95 
96 static int kern_chflagsat(struct thread *td, int fd, const char *path,
97     enum uio_seg pathseg, u_long flags, int atflag);
98 static int setfflags(struct thread *td, struct vnode *, u_long);
99 static int getutimes(const struct timeval *, enum uio_seg, struct timespec *);
100 static int getutimens(const struct timespec *, enum uio_seg,
101     struct timespec *, int *);
102 static int setutimes(struct thread *td, struct vnode *,
103     const struct timespec *, int, int);
104 static int vn_access(struct vnode *vp, int user_flags, struct ucred *cred,
105     struct thread *td);
106 static int kern_fhlinkat(struct thread *td, int fd, const char *path,
107     enum uio_seg pathseg, fhandle_t *fhp);
108 static int kern_getfhat(struct thread *td, int flags, int fd,
109     const char *path, enum uio_seg pathseg, fhandle_t *fhp);
110 static int kern_readlink_vp(struct vnode *vp, char *buf, enum uio_seg bufseg,
111     size_t count, struct thread *td);
112 static int kern_linkat_vp(struct thread *td, struct vnode *vp, int fd,
113     const char *path, enum uio_seg segflag);
114 
115 static uint64_t
116 at2cnpflags(u_int at_flags, u_int mask)
117 {
118 	u_int64_t res;
119 
120 	MPASS((at_flags & (AT_SYMLINK_FOLLOW | AT_SYMLINK_NOFOLLOW)) !=
121 	    (AT_SYMLINK_FOLLOW | AT_SYMLINK_NOFOLLOW));
122 
123 	res = 0;
124 	at_flags &= mask;
125 	if ((at_flags & AT_BENEATH) != 0)
126 		res |= BENEATH;
127 	if ((at_flags & AT_RESOLVE_BENEATH) != 0)
128 		res |= RBENEATH;
129 	if ((at_flags & AT_SYMLINK_FOLLOW) != 0)
130 		res |= FOLLOW;
131 	/* NOFOLLOW is pseudo flag */
132 	if ((mask & AT_SYMLINK_NOFOLLOW) != 0) {
133 		res |= (at_flags & AT_SYMLINK_NOFOLLOW) != 0 ? NOFOLLOW :
134 		    FOLLOW;
135 	}
136 	return (res);
137 }
138 
139 int
140 kern_sync(struct thread *td)
141 {
142 	struct mount *mp, *nmp;
143 	int save;
144 
145 	mtx_lock(&mountlist_mtx);
146 	for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
147 		if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK)) {
148 			nmp = TAILQ_NEXT(mp, mnt_list);
149 			continue;
150 		}
151 		if ((mp->mnt_flag & MNT_RDONLY) == 0 &&
152 		    vn_start_write(NULL, &mp, V_NOWAIT) == 0) {
153 			save = curthread_pflags_set(TDP_SYNCIO);
154 			vfs_periodic(mp, MNT_NOWAIT);
155 			VFS_SYNC(mp, MNT_NOWAIT);
156 			curthread_pflags_restore(save);
157 			vn_finished_write(mp);
158 		}
159 		mtx_lock(&mountlist_mtx);
160 		nmp = TAILQ_NEXT(mp, mnt_list);
161 		vfs_unbusy(mp);
162 	}
163 	mtx_unlock(&mountlist_mtx);
164 	return (0);
165 }
166 
167 /*
168  * Sync each mounted filesystem.
169  */
170 #ifndef _SYS_SYSPROTO_H_
171 struct sync_args {
172 	int     dummy;
173 };
174 #endif
175 /* ARGSUSED */
176 int
177 sys_sync(struct thread *td, struct sync_args *uap)
178 {
179 
180 	return (kern_sync(td));
181 }
182 
183 /*
184  * Change filesystem quotas.
185  */
186 #ifndef _SYS_SYSPROTO_H_
187 struct quotactl_args {
188 	char *path;
189 	int cmd;
190 	int uid;
191 	caddr_t arg;
192 };
193 #endif
194 int
195 sys_quotactl(struct thread *td, struct quotactl_args *uap)
196 {
197 	struct mount *mp;
198 	struct nameidata nd;
199 	int error;
200 
201 	AUDIT_ARG_CMD(uap->cmd);
202 	AUDIT_ARG_UID(uap->uid);
203 	if (!prison_allow(td->td_ucred, PR_ALLOW_QUOTAS))
204 		return (EPERM);
205 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
206 	    uap->path, td);
207 	if ((error = namei(&nd)) != 0)
208 		return (error);
209 	NDFREE(&nd, NDF_ONLY_PNBUF);
210 	mp = nd.ni_vp->v_mount;
211 	vfs_ref(mp);
212 	vput(nd.ni_vp);
213 	error = vfs_busy(mp, 0);
214 	if (error != 0) {
215 		vfs_rel(mp);
216 		return (error);
217 	}
218 	error = VFS_QUOTACTL(mp, uap->cmd, uap->uid, uap->arg);
219 
220 	/*
221 	 * Since quota on operation typically needs to open quota
222 	 * file, the Q_QUOTAON handler needs to unbusy the mount point
223 	 * before calling into namei.  Otherwise, unmount might be
224 	 * started between two vfs_busy() invocations (first is our,
225 	 * second is from mount point cross-walk code in lookup()),
226 	 * causing deadlock.
227 	 *
228 	 * Require that Q_QUOTAON handles the vfs_busy() reference on
229 	 * its own, always returning with ubusied mount point.
230 	 */
231 	if ((uap->cmd >> SUBCMDSHIFT) != Q_QUOTAON &&
232 	    (uap->cmd >> SUBCMDSHIFT) != Q_QUOTAOFF)
233 		vfs_unbusy(mp);
234 	vfs_rel(mp);
235 	return (error);
236 }
237 
238 /*
239  * Used by statfs conversion routines to scale the block size up if
240  * necessary so that all of the block counts are <= 'max_size'.  Note
241  * that 'max_size' should be a bitmask, i.e. 2^n - 1 for some non-zero
242  * value of 'n'.
243  */
244 void
245 statfs_scale_blocks(struct statfs *sf, long max_size)
246 {
247 	uint64_t count;
248 	int shift;
249 
250 	KASSERT(powerof2(max_size + 1), ("%s: invalid max_size", __func__));
251 
252 	/*
253 	 * Attempt to scale the block counts to give a more accurate
254 	 * overview to userland of the ratio of free space to used
255 	 * space.  To do this, find the largest block count and compute
256 	 * a divisor that lets it fit into a signed integer <= max_size.
257 	 */
258 	if (sf->f_bavail < 0)
259 		count = -sf->f_bavail;
260 	else
261 		count = sf->f_bavail;
262 	count = MAX(sf->f_blocks, MAX(sf->f_bfree, count));
263 	if (count <= max_size)
264 		return;
265 
266 	count >>= flsl(max_size);
267 	shift = 0;
268 	while (count > 0) {
269 		shift++;
270 		count >>=1;
271 	}
272 
273 	sf->f_bsize <<= shift;
274 	sf->f_blocks >>= shift;
275 	sf->f_bfree >>= shift;
276 	sf->f_bavail >>= shift;
277 }
278 
279 static int
280 kern_do_statfs(struct thread *td, struct mount *mp, struct statfs *buf)
281 {
282 	int error;
283 
284 	if (mp == NULL)
285 		return (EBADF);
286 	error = vfs_busy(mp, 0);
287 	vfs_rel(mp);
288 	if (error != 0)
289 		return (error);
290 #ifdef MAC
291 	error = mac_mount_check_stat(td->td_ucred, mp);
292 	if (error != 0)
293 		goto out;
294 #endif
295 	error = VFS_STATFS(mp, buf);
296 	if (error != 0)
297 		goto out;
298 	if (priv_check_cred_vfs_generation(td->td_ucred)) {
299 		buf->f_fsid.val[0] = buf->f_fsid.val[1] = 0;
300 		prison_enforce_statfs(td->td_ucred, mp, buf);
301 	}
302 out:
303 	vfs_unbusy(mp);
304 	return (error);
305 }
306 
307 /*
308  * Get filesystem statistics.
309  */
310 #ifndef _SYS_SYSPROTO_H_
311 struct statfs_args {
312 	char *path;
313 	struct statfs *buf;
314 };
315 #endif
316 int
317 sys_statfs(struct thread *td, struct statfs_args *uap)
318 {
319 	struct statfs *sfp;
320 	int error;
321 
322 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
323 	error = kern_statfs(td, uap->path, UIO_USERSPACE, sfp);
324 	if (error == 0)
325 		error = copyout(sfp, uap->buf, sizeof(struct statfs));
326 	free(sfp, M_STATFS);
327 	return (error);
328 }
329 
330 int
331 kern_statfs(struct thread *td, const char *path, enum uio_seg pathseg,
332     struct statfs *buf)
333 {
334 	struct mount *mp;
335 	struct nameidata nd;
336 	int error;
337 
338 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
339 	    pathseg, path, td);
340 	error = namei(&nd);
341 	if (error != 0)
342 		return (error);
343 	mp = nd.ni_vp->v_mount;
344 	vfs_ref(mp);
345 	NDFREE(&nd, NDF_ONLY_PNBUF);
346 	vput(nd.ni_vp);
347 	return (kern_do_statfs(td, mp, buf));
348 }
349 
350 /*
351  * Get filesystem statistics.
352  */
353 #ifndef _SYS_SYSPROTO_H_
354 struct fstatfs_args {
355 	int fd;
356 	struct statfs *buf;
357 };
358 #endif
359 int
360 sys_fstatfs(struct thread *td, struct fstatfs_args *uap)
361 {
362 	struct statfs *sfp;
363 	int error;
364 
365 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
366 	error = kern_fstatfs(td, uap->fd, sfp);
367 	if (error == 0)
368 		error = copyout(sfp, uap->buf, sizeof(struct statfs));
369 	free(sfp, M_STATFS);
370 	return (error);
371 }
372 
373 int
374 kern_fstatfs(struct thread *td, int fd, struct statfs *buf)
375 {
376 	struct file *fp;
377 	struct mount *mp;
378 	struct vnode *vp;
379 	int error;
380 
381 	AUDIT_ARG_FD(fd);
382 	error = getvnode(td, fd, &cap_fstatfs_rights, &fp);
383 	if (error != 0)
384 		return (error);
385 	vp = fp->f_vnode;
386 	vn_lock(vp, LK_SHARED | LK_RETRY);
387 #ifdef AUDIT
388 	AUDIT_ARG_VNODE1(vp);
389 #endif
390 	mp = vp->v_mount;
391 	if (mp != NULL)
392 		vfs_ref(mp);
393 	VOP_UNLOCK(vp);
394 	fdrop(fp, td);
395 	return (kern_do_statfs(td, mp, buf));
396 }
397 
398 /*
399  * Get statistics on all filesystems.
400  */
401 #ifndef _SYS_SYSPROTO_H_
402 struct getfsstat_args {
403 	struct statfs *buf;
404 	long bufsize;
405 	int mode;
406 };
407 #endif
408 int
409 sys_getfsstat(struct thread *td, struct getfsstat_args *uap)
410 {
411 	size_t count;
412 	int error;
413 
414 	if (uap->bufsize < 0 || uap->bufsize > SIZE_MAX)
415 		return (EINVAL);
416 	error = kern_getfsstat(td, &uap->buf, uap->bufsize, &count,
417 	    UIO_USERSPACE, uap->mode);
418 	if (error == 0)
419 		td->td_retval[0] = count;
420 	return (error);
421 }
422 
423 /*
424  * If (bufsize > 0 && bufseg == UIO_SYSSPACE)
425  *	The caller is responsible for freeing memory which will be allocated
426  *	in '*buf'.
427  */
428 int
429 kern_getfsstat(struct thread *td, struct statfs **buf, size_t bufsize,
430     size_t *countp, enum uio_seg bufseg, int mode)
431 {
432 	struct mount *mp, *nmp;
433 	struct statfs *sfsp, *sp, *sptmp, *tofree;
434 	size_t count, maxcount;
435 	int error;
436 
437 	switch (mode) {
438 	case MNT_WAIT:
439 	case MNT_NOWAIT:
440 		break;
441 	default:
442 		if (bufseg == UIO_SYSSPACE)
443 			*buf = NULL;
444 		return (EINVAL);
445 	}
446 restart:
447 	maxcount = bufsize / sizeof(struct statfs);
448 	if (bufsize == 0) {
449 		sfsp = NULL;
450 		tofree = NULL;
451 	} else if (bufseg == UIO_USERSPACE) {
452 		sfsp = *buf;
453 		tofree = NULL;
454 	} else /* if (bufseg == UIO_SYSSPACE) */ {
455 		count = 0;
456 		mtx_lock(&mountlist_mtx);
457 		TAILQ_FOREACH(mp, &mountlist, mnt_list) {
458 			count++;
459 		}
460 		mtx_unlock(&mountlist_mtx);
461 		if (maxcount > count)
462 			maxcount = count;
463 		tofree = sfsp = *buf = malloc(maxcount * sizeof(struct statfs),
464 		    M_STATFS, M_WAITOK);
465 	}
466 
467 	count = 0;
468 
469 	/*
470 	 * If there is no target buffer they only want the count.
471 	 *
472 	 * This could be TAILQ_FOREACH but it is open-coded to match the original
473 	 * code below.
474 	 */
475 	if (sfsp == NULL) {
476 		mtx_lock(&mountlist_mtx);
477 		for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
478 			if (prison_canseemount(td->td_ucred, mp) != 0) {
479 				nmp = TAILQ_NEXT(mp, mnt_list);
480 				continue;
481 			}
482 #ifdef MAC
483 			if (mac_mount_check_stat(td->td_ucred, mp) != 0) {
484 				nmp = TAILQ_NEXT(mp, mnt_list);
485 				continue;
486 			}
487 #endif
488 			count++;
489 			nmp = TAILQ_NEXT(mp, mnt_list);
490 		}
491 		mtx_unlock(&mountlist_mtx);
492 		*countp = count;
493 		return (0);
494 	}
495 
496 	/*
497 	 * They want the entire thing.
498 	 *
499 	 * Short-circuit the corner case of no room for anything, avoids
500 	 * relocking below.
501 	 */
502 	if (maxcount < 1) {
503 		goto out;
504 	}
505 
506 	mtx_lock(&mountlist_mtx);
507 	for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
508 		if (prison_canseemount(td->td_ucred, mp) != 0) {
509 			nmp = TAILQ_NEXT(mp, mnt_list);
510 			continue;
511 		}
512 #ifdef MAC
513 		if (mac_mount_check_stat(td->td_ucred, mp) != 0) {
514 			nmp = TAILQ_NEXT(mp, mnt_list);
515 			continue;
516 		}
517 #endif
518 		if (mode == MNT_WAIT) {
519 			if (vfs_busy(mp, MBF_MNTLSTLOCK) != 0) {
520 				/*
521 				 * If vfs_busy() failed, and MBF_NOWAIT
522 				 * wasn't passed, then the mp is gone.
523 				 * Furthermore, because of MBF_MNTLSTLOCK,
524 				 * the mountlist_mtx was dropped.  We have
525 				 * no other choice than to start over.
526 				 */
527 				mtx_unlock(&mountlist_mtx);
528 				free(tofree, M_STATFS);
529 				goto restart;
530 			}
531 		} else {
532 			if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK) != 0) {
533 				nmp = TAILQ_NEXT(mp, mnt_list);
534 				continue;
535 			}
536 		}
537 		sp = &mp->mnt_stat;
538 		/*
539 		 * If MNT_NOWAIT is specified, do not refresh
540 		 * the fsstat cache.
541 		 */
542 		if (mode != MNT_NOWAIT) {
543 			error = VFS_STATFS(mp, sp);
544 			if (error != 0) {
545 				mtx_lock(&mountlist_mtx);
546 				nmp = TAILQ_NEXT(mp, mnt_list);
547 				vfs_unbusy(mp);
548 				continue;
549 			}
550 		}
551 		if (priv_check_cred_vfs_generation(td->td_ucred)) {
552 			sptmp = malloc(sizeof(struct statfs), M_STATFS,
553 			    M_WAITOK);
554 			*sptmp = *sp;
555 			sptmp->f_fsid.val[0] = sptmp->f_fsid.val[1] = 0;
556 			prison_enforce_statfs(td->td_ucred, mp, sptmp);
557 			sp = sptmp;
558 		} else
559 			sptmp = NULL;
560 		if (bufseg == UIO_SYSSPACE) {
561 			bcopy(sp, sfsp, sizeof(*sp));
562 			free(sptmp, M_STATFS);
563 		} else /* if (bufseg == UIO_USERSPACE) */ {
564 			error = copyout(sp, sfsp, sizeof(*sp));
565 			free(sptmp, M_STATFS);
566 			if (error != 0) {
567 				vfs_unbusy(mp);
568 				return (error);
569 			}
570 		}
571 		sfsp++;
572 		count++;
573 
574 		if (count == maxcount) {
575 			vfs_unbusy(mp);
576 			goto out;
577 		}
578 
579 		mtx_lock(&mountlist_mtx);
580 		nmp = TAILQ_NEXT(mp, mnt_list);
581 		vfs_unbusy(mp);
582 	}
583 	mtx_unlock(&mountlist_mtx);
584 out:
585 	*countp = count;
586 	return (0);
587 }
588 
589 #ifdef COMPAT_FREEBSD4
590 /*
591  * Get old format filesystem statistics.
592  */
593 static void freebsd4_cvtstatfs(struct statfs *, struct ostatfs *);
594 
595 #ifndef _SYS_SYSPROTO_H_
596 struct freebsd4_statfs_args {
597 	char *path;
598 	struct ostatfs *buf;
599 };
600 #endif
601 int
602 freebsd4_statfs(struct thread *td, struct freebsd4_statfs_args *uap)
603 {
604 	struct ostatfs osb;
605 	struct statfs *sfp;
606 	int error;
607 
608 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
609 	error = kern_statfs(td, uap->path, UIO_USERSPACE, sfp);
610 	if (error == 0) {
611 		freebsd4_cvtstatfs(sfp, &osb);
612 		error = copyout(&osb, uap->buf, sizeof(osb));
613 	}
614 	free(sfp, M_STATFS);
615 	return (error);
616 }
617 
618 /*
619  * Get filesystem statistics.
620  */
621 #ifndef _SYS_SYSPROTO_H_
622 struct freebsd4_fstatfs_args {
623 	int fd;
624 	struct ostatfs *buf;
625 };
626 #endif
627 int
628 freebsd4_fstatfs(struct thread *td, struct freebsd4_fstatfs_args *uap)
629 {
630 	struct ostatfs osb;
631 	struct statfs *sfp;
632 	int error;
633 
634 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
635 	error = kern_fstatfs(td, uap->fd, sfp);
636 	if (error == 0) {
637 		freebsd4_cvtstatfs(sfp, &osb);
638 		error = copyout(&osb, uap->buf, sizeof(osb));
639 	}
640 	free(sfp, M_STATFS);
641 	return (error);
642 }
643 
644 /*
645  * Get statistics on all filesystems.
646  */
647 #ifndef _SYS_SYSPROTO_H_
648 struct freebsd4_getfsstat_args {
649 	struct ostatfs *buf;
650 	long bufsize;
651 	int mode;
652 };
653 #endif
654 int
655 freebsd4_getfsstat(struct thread *td, struct freebsd4_getfsstat_args *uap)
656 {
657 	struct statfs *buf, *sp;
658 	struct ostatfs osb;
659 	size_t count, size;
660 	int error;
661 
662 	if (uap->bufsize < 0)
663 		return (EINVAL);
664 	count = uap->bufsize / sizeof(struct ostatfs);
665 	if (count > SIZE_MAX / sizeof(struct statfs))
666 		return (EINVAL);
667 	size = count * sizeof(struct statfs);
668 	error = kern_getfsstat(td, &buf, size, &count, UIO_SYSSPACE,
669 	    uap->mode);
670 	if (error == 0)
671 		td->td_retval[0] = count;
672 	if (size != 0) {
673 		sp = buf;
674 		while (count != 0 && error == 0) {
675 			freebsd4_cvtstatfs(sp, &osb);
676 			error = copyout(&osb, uap->buf, sizeof(osb));
677 			sp++;
678 			uap->buf++;
679 			count--;
680 		}
681 		free(buf, M_STATFS);
682 	}
683 	return (error);
684 }
685 
686 /*
687  * Implement fstatfs() for (NFS) file handles.
688  */
689 #ifndef _SYS_SYSPROTO_H_
690 struct freebsd4_fhstatfs_args {
691 	struct fhandle *u_fhp;
692 	struct ostatfs *buf;
693 };
694 #endif
695 int
696 freebsd4_fhstatfs(struct thread *td, struct freebsd4_fhstatfs_args *uap)
697 {
698 	struct ostatfs osb;
699 	struct statfs *sfp;
700 	fhandle_t fh;
701 	int error;
702 
703 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
704 	if (error != 0)
705 		return (error);
706 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
707 	error = kern_fhstatfs(td, fh, sfp);
708 	if (error == 0) {
709 		freebsd4_cvtstatfs(sfp, &osb);
710 		error = copyout(&osb, uap->buf, sizeof(osb));
711 	}
712 	free(sfp, M_STATFS);
713 	return (error);
714 }
715 
716 /*
717  * Convert a new format statfs structure to an old format statfs structure.
718  */
719 static void
720 freebsd4_cvtstatfs(struct statfs *nsp, struct ostatfs *osp)
721 {
722 
723 	statfs_scale_blocks(nsp, LONG_MAX);
724 	bzero(osp, sizeof(*osp));
725 	osp->f_bsize = nsp->f_bsize;
726 	osp->f_iosize = MIN(nsp->f_iosize, LONG_MAX);
727 	osp->f_blocks = nsp->f_blocks;
728 	osp->f_bfree = nsp->f_bfree;
729 	osp->f_bavail = nsp->f_bavail;
730 	osp->f_files = MIN(nsp->f_files, LONG_MAX);
731 	osp->f_ffree = MIN(nsp->f_ffree, LONG_MAX);
732 	osp->f_owner = nsp->f_owner;
733 	osp->f_type = nsp->f_type;
734 	osp->f_flags = nsp->f_flags;
735 	osp->f_syncwrites = MIN(nsp->f_syncwrites, LONG_MAX);
736 	osp->f_asyncwrites = MIN(nsp->f_asyncwrites, LONG_MAX);
737 	osp->f_syncreads = MIN(nsp->f_syncreads, LONG_MAX);
738 	osp->f_asyncreads = MIN(nsp->f_asyncreads, LONG_MAX);
739 	strlcpy(osp->f_fstypename, nsp->f_fstypename,
740 	    MIN(MFSNAMELEN, OMFSNAMELEN));
741 	strlcpy(osp->f_mntonname, nsp->f_mntonname,
742 	    MIN(MNAMELEN, OMNAMELEN));
743 	strlcpy(osp->f_mntfromname, nsp->f_mntfromname,
744 	    MIN(MNAMELEN, OMNAMELEN));
745 	osp->f_fsid = nsp->f_fsid;
746 }
747 #endif /* COMPAT_FREEBSD4 */
748 
749 #if defined(COMPAT_FREEBSD11)
750 /*
751  * Get old format filesystem statistics.
752  */
753 static void freebsd11_cvtstatfs(struct statfs *, struct freebsd11_statfs *);
754 
755 int
756 freebsd11_statfs(struct thread *td, struct freebsd11_statfs_args *uap)
757 {
758 	struct freebsd11_statfs osb;
759 	struct statfs *sfp;
760 	int error;
761 
762 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
763 	error = kern_statfs(td, uap->path, UIO_USERSPACE, sfp);
764 	if (error == 0) {
765 		freebsd11_cvtstatfs(sfp, &osb);
766 		error = copyout(&osb, uap->buf, sizeof(osb));
767 	}
768 	free(sfp, M_STATFS);
769 	return (error);
770 }
771 
772 /*
773  * Get filesystem statistics.
774  */
775 int
776 freebsd11_fstatfs(struct thread *td, struct freebsd11_fstatfs_args *uap)
777 {
778 	struct freebsd11_statfs osb;
779 	struct statfs *sfp;
780 	int error;
781 
782 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
783 	error = kern_fstatfs(td, uap->fd, sfp);
784 	if (error == 0) {
785 		freebsd11_cvtstatfs(sfp, &osb);
786 		error = copyout(&osb, uap->buf, sizeof(osb));
787 	}
788 	free(sfp, M_STATFS);
789 	return (error);
790 }
791 
792 /*
793  * Get statistics on all filesystems.
794  */
795 int
796 freebsd11_getfsstat(struct thread *td, struct freebsd11_getfsstat_args *uap)
797 {
798 	struct freebsd11_statfs osb;
799 	struct statfs *buf, *sp;
800 	size_t count, size;
801 	int error;
802 
803 	count = uap->bufsize / sizeof(struct ostatfs);
804 	size = count * sizeof(struct statfs);
805 	error = kern_getfsstat(td, &buf, size, &count, UIO_SYSSPACE,
806 	    uap->mode);
807 	if (error == 0)
808 		td->td_retval[0] = count;
809 	if (size > 0) {
810 		sp = buf;
811 		while (count > 0 && error == 0) {
812 			freebsd11_cvtstatfs(sp, &osb);
813 			error = copyout(&osb, uap->buf, sizeof(osb));
814 			sp++;
815 			uap->buf++;
816 			count--;
817 		}
818 		free(buf, M_STATFS);
819 	}
820 	return (error);
821 }
822 
823 /*
824  * Implement fstatfs() for (NFS) file handles.
825  */
826 int
827 freebsd11_fhstatfs(struct thread *td, struct freebsd11_fhstatfs_args *uap)
828 {
829 	struct freebsd11_statfs osb;
830 	struct statfs *sfp;
831 	fhandle_t fh;
832 	int error;
833 
834 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
835 	if (error)
836 		return (error);
837 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
838 	error = kern_fhstatfs(td, fh, sfp);
839 	if (error == 0) {
840 		freebsd11_cvtstatfs(sfp, &osb);
841 		error = copyout(&osb, uap->buf, sizeof(osb));
842 	}
843 	free(sfp, M_STATFS);
844 	return (error);
845 }
846 
847 /*
848  * Convert a new format statfs structure to an old format statfs structure.
849  */
850 static void
851 freebsd11_cvtstatfs(struct statfs *nsp, struct freebsd11_statfs *osp)
852 {
853 
854 	bzero(osp, sizeof(*osp));
855 	osp->f_version = FREEBSD11_STATFS_VERSION;
856 	osp->f_type = nsp->f_type;
857 	osp->f_flags = nsp->f_flags;
858 	osp->f_bsize = nsp->f_bsize;
859 	osp->f_iosize = nsp->f_iosize;
860 	osp->f_blocks = nsp->f_blocks;
861 	osp->f_bfree = nsp->f_bfree;
862 	osp->f_bavail = nsp->f_bavail;
863 	osp->f_files = nsp->f_files;
864 	osp->f_ffree = nsp->f_ffree;
865 	osp->f_syncwrites = nsp->f_syncwrites;
866 	osp->f_asyncwrites = nsp->f_asyncwrites;
867 	osp->f_syncreads = nsp->f_syncreads;
868 	osp->f_asyncreads = nsp->f_asyncreads;
869 	osp->f_namemax = nsp->f_namemax;
870 	osp->f_owner = nsp->f_owner;
871 	osp->f_fsid = nsp->f_fsid;
872 	strlcpy(osp->f_fstypename, nsp->f_fstypename,
873 	    MIN(MFSNAMELEN, sizeof(osp->f_fstypename)));
874 	strlcpy(osp->f_mntonname, nsp->f_mntonname,
875 	    MIN(MNAMELEN, sizeof(osp->f_mntonname)));
876 	strlcpy(osp->f_mntfromname, nsp->f_mntfromname,
877 	    MIN(MNAMELEN, sizeof(osp->f_mntfromname)));
878 }
879 #endif /* COMPAT_FREEBSD11 */
880 
881 /*
882  * Change current working directory to a given file descriptor.
883  */
884 #ifndef _SYS_SYSPROTO_H_
885 struct fchdir_args {
886 	int	fd;
887 };
888 #endif
889 int
890 sys_fchdir(struct thread *td, struct fchdir_args *uap)
891 {
892 	struct vnode *vp, *tdp;
893 	struct mount *mp;
894 	struct file *fp;
895 	int error;
896 
897 	AUDIT_ARG_FD(uap->fd);
898 	error = getvnode(td, uap->fd, &cap_fchdir_rights,
899 	    &fp);
900 	if (error != 0)
901 		return (error);
902 	vp = fp->f_vnode;
903 	vrefact(vp);
904 	fdrop(fp, td);
905 	vn_lock(vp, LK_SHARED | LK_RETRY);
906 	AUDIT_ARG_VNODE1(vp);
907 	error = change_dir(vp, td);
908 	while (!error && (mp = vp->v_mountedhere) != NULL) {
909 		if (vfs_busy(mp, 0))
910 			continue;
911 		error = VFS_ROOT(mp, LK_SHARED, &tdp);
912 		vfs_unbusy(mp);
913 		if (error != 0)
914 			break;
915 		vput(vp);
916 		vp = tdp;
917 	}
918 	if (error != 0) {
919 		vput(vp);
920 		return (error);
921 	}
922 	VOP_UNLOCK(vp);
923 	pwd_chdir(td, vp);
924 	return (0);
925 }
926 
927 /*
928  * Change current working directory (``.'').
929  */
930 #ifndef _SYS_SYSPROTO_H_
931 struct chdir_args {
932 	char	*path;
933 };
934 #endif
935 int
936 sys_chdir(struct thread *td, struct chdir_args *uap)
937 {
938 
939 	return (kern_chdir(td, uap->path, UIO_USERSPACE));
940 }
941 
942 int
943 kern_chdir(struct thread *td, const char *path, enum uio_seg pathseg)
944 {
945 	struct nameidata nd;
946 	int error;
947 
948 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
949 	    pathseg, path, td);
950 	if ((error = namei(&nd)) != 0)
951 		return (error);
952 	if ((error = change_dir(nd.ni_vp, td)) != 0) {
953 		vput(nd.ni_vp);
954 		NDFREE(&nd, NDF_ONLY_PNBUF);
955 		return (error);
956 	}
957 	VOP_UNLOCK(nd.ni_vp);
958 	NDFREE(&nd, NDF_ONLY_PNBUF);
959 	pwd_chdir(td, nd.ni_vp);
960 	return (0);
961 }
962 
963 /*
964  * Change notion of root (``/'') directory.
965  */
966 #ifndef _SYS_SYSPROTO_H_
967 struct chroot_args {
968 	char	*path;
969 };
970 #endif
971 int
972 sys_chroot(struct thread *td, struct chroot_args *uap)
973 {
974 	struct nameidata nd;
975 	int error;
976 
977 	error = priv_check(td, PRIV_VFS_CHROOT);
978 	if (error != 0)
979 		return (error);
980 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
981 	    UIO_USERSPACE, uap->path, td);
982 	error = namei(&nd);
983 	if (error != 0)
984 		goto error;
985 	error = change_dir(nd.ni_vp, td);
986 	if (error != 0)
987 		goto e_vunlock;
988 #ifdef MAC
989 	error = mac_vnode_check_chroot(td->td_ucred, nd.ni_vp);
990 	if (error != 0)
991 		goto e_vunlock;
992 #endif
993 	VOP_UNLOCK(nd.ni_vp);
994 	error = pwd_chroot(td, nd.ni_vp);
995 	vrele(nd.ni_vp);
996 	NDFREE(&nd, NDF_ONLY_PNBUF);
997 	return (error);
998 e_vunlock:
999 	vput(nd.ni_vp);
1000 error:
1001 	NDFREE(&nd, NDF_ONLY_PNBUF);
1002 	return (error);
1003 }
1004 
1005 /*
1006  * Common routine for chroot and chdir.  Callers must provide a locked vnode
1007  * instance.
1008  */
1009 int
1010 change_dir(struct vnode *vp, struct thread *td)
1011 {
1012 #ifdef MAC
1013 	int error;
1014 #endif
1015 
1016 	ASSERT_VOP_LOCKED(vp, "change_dir(): vp not locked");
1017 	if (vp->v_type != VDIR)
1018 		return (ENOTDIR);
1019 #ifdef MAC
1020 	error = mac_vnode_check_chdir(td->td_ucred, vp);
1021 	if (error != 0)
1022 		return (error);
1023 #endif
1024 	return (VOP_ACCESS(vp, VEXEC, td->td_ucred, td));
1025 }
1026 
1027 static __inline void
1028 flags_to_rights(int flags, cap_rights_t *rightsp)
1029 {
1030 
1031 	if (flags & O_EXEC) {
1032 		cap_rights_set_one(rightsp, CAP_FEXECVE);
1033 	} else {
1034 		switch ((flags & O_ACCMODE)) {
1035 		case O_RDONLY:
1036 			cap_rights_set_one(rightsp, CAP_READ);
1037 			break;
1038 		case O_RDWR:
1039 			cap_rights_set_one(rightsp, CAP_READ);
1040 			/* FALLTHROUGH */
1041 		case O_WRONLY:
1042 			cap_rights_set_one(rightsp, CAP_WRITE);
1043 			if (!(flags & (O_APPEND | O_TRUNC)))
1044 				cap_rights_set_one(rightsp, CAP_SEEK);
1045 			break;
1046 		}
1047 	}
1048 
1049 	if (flags & O_CREAT)
1050 		cap_rights_set_one(rightsp, CAP_CREATE);
1051 
1052 	if (flags & O_TRUNC)
1053 		cap_rights_set_one(rightsp, CAP_FTRUNCATE);
1054 
1055 	if (flags & (O_SYNC | O_FSYNC))
1056 		cap_rights_set_one(rightsp, CAP_FSYNC);
1057 
1058 	if (flags & (O_EXLOCK | O_SHLOCK))
1059 		cap_rights_set_one(rightsp, CAP_FLOCK);
1060 }
1061 
1062 /*
1063  * Check permissions, allocate an open file structure, and call the device
1064  * open routine if any.
1065  */
1066 #ifndef _SYS_SYSPROTO_H_
1067 struct open_args {
1068 	char	*path;
1069 	int	flags;
1070 	int	mode;
1071 };
1072 #endif
1073 int
1074 sys_open(struct thread *td, struct open_args *uap)
1075 {
1076 
1077 	return (kern_openat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1078 	    uap->flags, uap->mode));
1079 }
1080 
1081 #ifndef _SYS_SYSPROTO_H_
1082 struct openat_args {
1083 	int	fd;
1084 	char	*path;
1085 	int	flag;
1086 	int	mode;
1087 };
1088 #endif
1089 int
1090 sys_openat(struct thread *td, struct openat_args *uap)
1091 {
1092 
1093 	AUDIT_ARG_FD(uap->fd);
1094 	return (kern_openat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
1095 	    uap->mode));
1096 }
1097 
1098 int
1099 kern_openat(struct thread *td, int fd, const char *path, enum uio_seg pathseg,
1100     int flags, int mode)
1101 {
1102 	struct proc *p = td->td_proc;
1103 	struct filedesc *fdp = p->p_fd;
1104 	struct file *fp;
1105 	struct vnode *vp;
1106 	struct nameidata nd;
1107 	cap_rights_t rights;
1108 	int cmode, error, indx;
1109 
1110 	indx = -1;
1111 
1112 	AUDIT_ARG_FFLAGS(flags);
1113 	AUDIT_ARG_MODE(mode);
1114 	cap_rights_init_one(&rights, CAP_LOOKUP);
1115 	flags_to_rights(flags, &rights);
1116 	/*
1117 	 * Only one of the O_EXEC, O_RDONLY, O_WRONLY and O_RDWR flags
1118 	 * may be specified.
1119 	 */
1120 	if (flags & O_EXEC) {
1121 		if (flags & O_ACCMODE)
1122 			return (EINVAL);
1123 	} else if ((flags & O_ACCMODE) == O_ACCMODE) {
1124 		return (EINVAL);
1125 	} else {
1126 		flags = FFLAGS(flags);
1127 	}
1128 
1129 	/*
1130 	 * Allocate a file structure. The descriptor to reference it
1131 	 * is allocated and set by finstall() below.
1132 	 */
1133 	error = falloc_noinstall(td, &fp);
1134 	if (error != 0)
1135 		return (error);
1136 	/*
1137 	 * An extra reference on `fp' has been held for us by
1138 	 * falloc_noinstall().
1139 	 */
1140 	/* Set the flags early so the finit in devfs can pick them up. */
1141 	fp->f_flag = flags & FMASK;
1142 	cmode = ((mode & ~fdp->fd_cmask) & ALLPERMS) & ~S_ISTXT;
1143 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, fd,
1144 	    &rights, td);
1145 	td->td_dupfd = -1;		/* XXX check for fdopen */
1146 	error = vn_open(&nd, &flags, cmode, fp);
1147 	if (error != 0) {
1148 		/*
1149 		 * If the vn_open replaced the method vector, something
1150 		 * wonderous happened deep below and we just pass it up
1151 		 * pretending we know what we do.
1152 		 */
1153 		if (error == ENXIO && fp->f_ops != &badfileops)
1154 			goto success;
1155 
1156 		/*
1157 		 * Handle special fdopen() case. bleh.
1158 		 *
1159 		 * Don't do this for relative (capability) lookups; we don't
1160 		 * understand exactly what would happen, and we don't think
1161 		 * that it ever should.
1162 		 */
1163 		if ((nd.ni_resflags & NIRES_STRICTREL) == 0 &&
1164 		    (error == ENODEV || error == ENXIO) &&
1165 		    td->td_dupfd >= 0) {
1166 			error = dupfdopen(td, fdp, td->td_dupfd, flags, error,
1167 			    &indx);
1168 			if (error == 0)
1169 				goto success;
1170 		}
1171 
1172 		goto bad;
1173 	}
1174 	td->td_dupfd = 0;
1175 	NDFREE(&nd, NDF_ONLY_PNBUF);
1176 	vp = nd.ni_vp;
1177 
1178 	/*
1179 	 * Store the vnode, for any f_type. Typically, the vnode use
1180 	 * count is decremented by direct call to vn_closefile() for
1181 	 * files that switched type in the cdevsw fdopen() method.
1182 	 */
1183 	fp->f_vnode = vp;
1184 	/*
1185 	 * If the file wasn't claimed by devfs bind it to the normal
1186 	 * vnode operations here.
1187 	 */
1188 	if (fp->f_ops == &badfileops) {
1189 		KASSERT(vp->v_type != VFIFO, ("Unexpected fifo."));
1190 		finit_vnode(fp, flags, NULL, &vnops);
1191 	}
1192 
1193 	VOP_UNLOCK(vp);
1194 	if (flags & O_TRUNC) {
1195 		error = fo_truncate(fp, 0, td->td_ucred, td);
1196 		if (error != 0)
1197 			goto bad;
1198 	}
1199 success:
1200 	/*
1201 	 * If we haven't already installed the FD (for dupfdopen), do so now.
1202 	 */
1203 	if (indx == -1) {
1204 		struct filecaps *fcaps;
1205 
1206 #ifdef CAPABILITIES
1207 		if ((nd.ni_resflags & NIRES_STRICTREL) != 0)
1208 			fcaps = &nd.ni_filecaps;
1209 		else
1210 #endif
1211 			fcaps = NULL;
1212 		error = finstall(td, fp, &indx, flags, fcaps);
1213 		/* On success finstall() consumes fcaps. */
1214 		if (error != 0) {
1215 			filecaps_free(&nd.ni_filecaps);
1216 			goto bad;
1217 		}
1218 	} else {
1219 		filecaps_free(&nd.ni_filecaps);
1220 	}
1221 
1222 	/*
1223 	 * Release our private reference, leaving the one associated with
1224 	 * the descriptor table intact.
1225 	 */
1226 	fdrop(fp, td);
1227 	td->td_retval[0] = indx;
1228 	return (0);
1229 bad:
1230 	KASSERT(indx == -1, ("indx=%d, should be -1", indx));
1231 	fdrop(fp, td);
1232 	return (error);
1233 }
1234 
1235 #ifdef COMPAT_43
1236 /*
1237  * Create a file.
1238  */
1239 #ifndef _SYS_SYSPROTO_H_
1240 struct ocreat_args {
1241 	char	*path;
1242 	int	mode;
1243 };
1244 #endif
1245 int
1246 ocreat(struct thread *td, struct ocreat_args *uap)
1247 {
1248 
1249 	return (kern_openat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1250 	    O_WRONLY | O_CREAT | O_TRUNC, uap->mode));
1251 }
1252 #endif /* COMPAT_43 */
1253 
1254 /*
1255  * Create a special file.
1256  */
1257 #ifndef _SYS_SYSPROTO_H_
1258 struct mknodat_args {
1259 	int	fd;
1260 	char	*path;
1261 	mode_t	mode;
1262 	dev_t	dev;
1263 };
1264 #endif
1265 int
1266 sys_mknodat(struct thread *td, struct mknodat_args *uap)
1267 {
1268 
1269 	return (kern_mknodat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode,
1270 	    uap->dev));
1271 }
1272 
1273 #if defined(COMPAT_FREEBSD11)
1274 int
1275 freebsd11_mknod(struct thread *td,
1276     struct freebsd11_mknod_args *uap)
1277 {
1278 
1279 	return (kern_mknodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1280 	    uap->mode, uap->dev));
1281 }
1282 
1283 int
1284 freebsd11_mknodat(struct thread *td,
1285     struct freebsd11_mknodat_args *uap)
1286 {
1287 
1288 	return (kern_mknodat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode,
1289 	    uap->dev));
1290 }
1291 #endif /* COMPAT_FREEBSD11 */
1292 
1293 int
1294 kern_mknodat(struct thread *td, int fd, const char *path, enum uio_seg pathseg,
1295     int mode, dev_t dev)
1296 {
1297 	struct vnode *vp;
1298 	struct mount *mp;
1299 	struct vattr vattr;
1300 	struct nameidata nd;
1301 	int error, whiteout = 0;
1302 
1303 	AUDIT_ARG_MODE(mode);
1304 	AUDIT_ARG_DEV(dev);
1305 	switch (mode & S_IFMT) {
1306 	case S_IFCHR:
1307 	case S_IFBLK:
1308 		error = priv_check(td, PRIV_VFS_MKNOD_DEV);
1309 		if (error == 0 && dev == VNOVAL)
1310 			error = EINVAL;
1311 		break;
1312 	case S_IFWHT:
1313 		error = priv_check(td, PRIV_VFS_MKNOD_WHT);
1314 		break;
1315 	case S_IFIFO:
1316 		if (dev == 0)
1317 			return (kern_mkfifoat(td, fd, path, pathseg, mode));
1318 		/* FALLTHROUGH */
1319 	default:
1320 		error = EINVAL;
1321 		break;
1322 	}
1323 	if (error != 0)
1324 		return (error);
1325 restart:
1326 	bwillwrite();
1327 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
1328 	    NOCACHE, pathseg, path, fd, &cap_mknodat_rights,
1329 	    td);
1330 	if ((error = namei(&nd)) != 0)
1331 		return (error);
1332 	vp = nd.ni_vp;
1333 	if (vp != NULL) {
1334 		NDFREE(&nd, NDF_ONLY_PNBUF);
1335 		if (vp == nd.ni_dvp)
1336 			vrele(nd.ni_dvp);
1337 		else
1338 			vput(nd.ni_dvp);
1339 		vrele(vp);
1340 		return (EEXIST);
1341 	} else {
1342 		VATTR_NULL(&vattr);
1343 		vattr.va_mode = (mode & ALLPERMS) &
1344 		    ~td->td_proc->p_fd->fd_cmask;
1345 		vattr.va_rdev = dev;
1346 		whiteout = 0;
1347 
1348 		switch (mode & S_IFMT) {
1349 		case S_IFCHR:
1350 			vattr.va_type = VCHR;
1351 			break;
1352 		case S_IFBLK:
1353 			vattr.va_type = VBLK;
1354 			break;
1355 		case S_IFWHT:
1356 			whiteout = 1;
1357 			break;
1358 		default:
1359 			panic("kern_mknod: invalid mode");
1360 		}
1361 	}
1362 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1363 		NDFREE(&nd, NDF_ONLY_PNBUF);
1364 		vput(nd.ni_dvp);
1365 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1366 			return (error);
1367 		goto restart;
1368 	}
1369 #ifdef MAC
1370 	if (error == 0 && !whiteout)
1371 		error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp,
1372 		    &nd.ni_cnd, &vattr);
1373 #endif
1374 	if (error == 0) {
1375 		if (whiteout)
1376 			error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE);
1377 		else {
1378 			error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp,
1379 						&nd.ni_cnd, &vattr);
1380 			if (error == 0)
1381 				vput(nd.ni_vp);
1382 		}
1383 	}
1384 	NDFREE(&nd, NDF_ONLY_PNBUF);
1385 	vput(nd.ni_dvp);
1386 	vn_finished_write(mp);
1387 	if (error == ERELOOKUP)
1388 		goto restart;
1389 	return (error);
1390 }
1391 
1392 /*
1393  * Create a named pipe.
1394  */
1395 #ifndef _SYS_SYSPROTO_H_
1396 struct mkfifo_args {
1397 	char	*path;
1398 	int	mode;
1399 };
1400 #endif
1401 int
1402 sys_mkfifo(struct thread *td, struct mkfifo_args *uap)
1403 {
1404 
1405 	return (kern_mkfifoat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1406 	    uap->mode));
1407 }
1408 
1409 #ifndef _SYS_SYSPROTO_H_
1410 struct mkfifoat_args {
1411 	int	fd;
1412 	char	*path;
1413 	mode_t	mode;
1414 };
1415 #endif
1416 int
1417 sys_mkfifoat(struct thread *td, struct mkfifoat_args *uap)
1418 {
1419 
1420 	return (kern_mkfifoat(td, uap->fd, uap->path, UIO_USERSPACE,
1421 	    uap->mode));
1422 }
1423 
1424 int
1425 kern_mkfifoat(struct thread *td, int fd, const char *path,
1426     enum uio_seg pathseg, int mode)
1427 {
1428 	struct mount *mp;
1429 	struct vattr vattr;
1430 	struct nameidata nd;
1431 	int error;
1432 
1433 	AUDIT_ARG_MODE(mode);
1434 restart:
1435 	bwillwrite();
1436 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
1437 	    NOCACHE, pathseg, path, fd, &cap_mkfifoat_rights,
1438 	    td);
1439 	if ((error = namei(&nd)) != 0)
1440 		return (error);
1441 	if (nd.ni_vp != NULL) {
1442 		NDFREE(&nd, NDF_ONLY_PNBUF);
1443 		if (nd.ni_vp == nd.ni_dvp)
1444 			vrele(nd.ni_dvp);
1445 		else
1446 			vput(nd.ni_dvp);
1447 		vrele(nd.ni_vp);
1448 		return (EEXIST);
1449 	}
1450 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1451 		NDFREE(&nd, NDF_ONLY_PNBUF);
1452 		vput(nd.ni_dvp);
1453 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1454 			return (error);
1455 		goto restart;
1456 	}
1457 	VATTR_NULL(&vattr);
1458 	vattr.va_type = VFIFO;
1459 	vattr.va_mode = (mode & ALLPERMS) & ~td->td_proc->p_fd->fd_cmask;
1460 #ifdef MAC
1461 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1462 	    &vattr);
1463 	if (error != 0)
1464 		goto out;
1465 #endif
1466 	error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
1467 	if (error == 0)
1468 		vput(nd.ni_vp);
1469 #ifdef MAC
1470 out:
1471 #endif
1472 	vput(nd.ni_dvp);
1473 	vn_finished_write(mp);
1474 	NDFREE(&nd, NDF_ONLY_PNBUF);
1475 	if (error == ERELOOKUP)
1476 		goto restart;
1477 	return (error);
1478 }
1479 
1480 /*
1481  * Make a hard file link.
1482  */
1483 #ifndef _SYS_SYSPROTO_H_
1484 struct link_args {
1485 	char	*path;
1486 	char	*link;
1487 };
1488 #endif
1489 int
1490 sys_link(struct thread *td, struct link_args *uap)
1491 {
1492 
1493 	return (kern_linkat(td, AT_FDCWD, AT_FDCWD, uap->path, uap->link,
1494 	    UIO_USERSPACE, FOLLOW));
1495 }
1496 
1497 #ifndef _SYS_SYSPROTO_H_
1498 struct linkat_args {
1499 	int	fd1;
1500 	char	*path1;
1501 	int	fd2;
1502 	char	*path2;
1503 	int	flag;
1504 };
1505 #endif
1506 int
1507 sys_linkat(struct thread *td, struct linkat_args *uap)
1508 {
1509 	int flag;
1510 
1511 	flag = uap->flag;
1512 	if ((flag & ~(AT_SYMLINK_FOLLOW | AT_BENEATH |
1513 	    AT_RESOLVE_BENEATH)) != 0)
1514 		return (EINVAL);
1515 
1516 	return (kern_linkat(td, uap->fd1, uap->fd2, uap->path1, uap->path2,
1517 	    UIO_USERSPACE, at2cnpflags(flag, AT_SYMLINK_FOLLOW | AT_BENEATH |
1518 	    AT_RESOLVE_BENEATH)));
1519 }
1520 
1521 int hardlink_check_uid = 0;
1522 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_uid, CTLFLAG_RW,
1523     &hardlink_check_uid, 0,
1524     "Unprivileged processes cannot create hard links to files owned by other "
1525     "users");
1526 static int hardlink_check_gid = 0;
1527 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_gid, CTLFLAG_RW,
1528     &hardlink_check_gid, 0,
1529     "Unprivileged processes cannot create hard links to files owned by other "
1530     "groups");
1531 
1532 static int
1533 can_hardlink(struct vnode *vp, struct ucred *cred)
1534 {
1535 	struct vattr va;
1536 	int error;
1537 
1538 	if (!hardlink_check_uid && !hardlink_check_gid)
1539 		return (0);
1540 
1541 	error = VOP_GETATTR(vp, &va, cred);
1542 	if (error != 0)
1543 		return (error);
1544 
1545 	if (hardlink_check_uid && cred->cr_uid != va.va_uid) {
1546 		error = priv_check_cred(cred, PRIV_VFS_LINK);
1547 		if (error != 0)
1548 			return (error);
1549 	}
1550 
1551 	if (hardlink_check_gid && !groupmember(va.va_gid, cred)) {
1552 		error = priv_check_cred(cred, PRIV_VFS_LINK);
1553 		if (error != 0)
1554 			return (error);
1555 	}
1556 
1557 	return (0);
1558 }
1559 
1560 int
1561 kern_linkat(struct thread *td, int fd1, int fd2, const char *path1,
1562     const char *path2, enum uio_seg segflag, int follow)
1563 {
1564 	struct nameidata nd;
1565 	int error;
1566 
1567 	do {
1568 		bwillwrite();
1569 		NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, segflag,
1570 		    path1, fd1, &cap_linkat_source_rights, td);
1571 		if ((error = namei(&nd)) != 0)
1572 			return (error);
1573 		NDFREE(&nd, NDF_ONLY_PNBUF);
1574 		error = kern_linkat_vp(td, nd.ni_vp, fd2, path2, segflag);
1575 	} while (error ==  EAGAIN || error == ERELOOKUP);
1576 	return (error);
1577 }
1578 
1579 static int
1580 kern_linkat_vp(struct thread *td, struct vnode *vp, int fd, const char *path,
1581     enum uio_seg segflag)
1582 {
1583 	struct nameidata nd;
1584 	struct mount *mp;
1585 	int error;
1586 
1587 	if (vp->v_type == VDIR) {
1588 		vrele(vp);
1589 		return (EPERM);		/* POSIX */
1590 	}
1591 	NDINIT_ATRIGHTS(&nd, CREATE,
1592 	    LOCKPARENT | SAVENAME | AUDITVNODE2 | NOCACHE, segflag, path, fd,
1593 	    &cap_linkat_target_rights, td);
1594 	if ((error = namei(&nd)) == 0) {
1595 		if (nd.ni_vp != NULL) {
1596 			NDFREE(&nd, NDF_ONLY_PNBUF);
1597 			if (nd.ni_dvp == nd.ni_vp)
1598 				vrele(nd.ni_dvp);
1599 			else
1600 				vput(nd.ni_dvp);
1601 			vrele(nd.ni_vp);
1602 			vrele(vp);
1603 			return (EEXIST);
1604 		} else if (nd.ni_dvp->v_mount != vp->v_mount) {
1605 			/*
1606 			 * Cross-device link.  No need to recheck
1607 			 * vp->v_type, since it cannot change, except
1608 			 * to VBAD.
1609 			 */
1610 			NDFREE(&nd, NDF_ONLY_PNBUF);
1611 			vput(nd.ni_dvp);
1612 			vrele(vp);
1613 			return (EXDEV);
1614 		} else if ((error = vn_lock(vp, LK_EXCLUSIVE)) == 0) {
1615 			error = can_hardlink(vp, td->td_ucred);
1616 #ifdef MAC
1617 			if (error == 0)
1618 				error = mac_vnode_check_link(td->td_ucred,
1619 				    nd.ni_dvp, vp, &nd.ni_cnd);
1620 #endif
1621 			if (error != 0) {
1622 				vput(vp);
1623 				vput(nd.ni_dvp);
1624 				NDFREE(&nd, NDF_ONLY_PNBUF);
1625 				return (error);
1626 			}
1627 			error = vn_start_write(vp, &mp, V_NOWAIT);
1628 			if (error != 0) {
1629 				vput(vp);
1630 				vput(nd.ni_dvp);
1631 				NDFREE(&nd, NDF_ONLY_PNBUF);
1632 				error = vn_start_write(NULL, &mp,
1633 				    V_XSLEEP | PCATCH);
1634 				if (error != 0)
1635 					return (error);
1636 				return (EAGAIN);
1637 			}
1638 			error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
1639 			VOP_UNLOCK(vp);
1640 			vput(nd.ni_dvp);
1641 			vn_finished_write(mp);
1642 			NDFREE(&nd, NDF_ONLY_PNBUF);
1643 		} else {
1644 			vput(nd.ni_dvp);
1645 			NDFREE(&nd, NDF_ONLY_PNBUF);
1646 			vrele(vp);
1647 			return (EAGAIN);
1648 		}
1649 	}
1650 	vrele(vp);
1651 	return (error);
1652 }
1653 
1654 /*
1655  * Make a symbolic link.
1656  */
1657 #ifndef _SYS_SYSPROTO_H_
1658 struct symlink_args {
1659 	char	*path;
1660 	char	*link;
1661 };
1662 #endif
1663 int
1664 sys_symlink(struct thread *td, struct symlink_args *uap)
1665 {
1666 
1667 	return (kern_symlinkat(td, uap->path, AT_FDCWD, uap->link,
1668 	    UIO_USERSPACE));
1669 }
1670 
1671 #ifndef _SYS_SYSPROTO_H_
1672 struct symlinkat_args {
1673 	char	*path;
1674 	int	fd;
1675 	char	*path2;
1676 };
1677 #endif
1678 int
1679 sys_symlinkat(struct thread *td, struct symlinkat_args *uap)
1680 {
1681 
1682 	return (kern_symlinkat(td, uap->path1, uap->fd, uap->path2,
1683 	    UIO_USERSPACE));
1684 }
1685 
1686 int
1687 kern_symlinkat(struct thread *td, const char *path1, int fd, const char *path2,
1688     enum uio_seg segflg)
1689 {
1690 	struct mount *mp;
1691 	struct vattr vattr;
1692 	const char *syspath;
1693 	char *tmppath;
1694 	struct nameidata nd;
1695 	int error;
1696 
1697 	if (segflg == UIO_SYSSPACE) {
1698 		syspath = path1;
1699 	} else {
1700 		tmppath = uma_zalloc(namei_zone, M_WAITOK);
1701 		if ((error = copyinstr(path1, tmppath, MAXPATHLEN, NULL)) != 0)
1702 			goto out;
1703 		syspath = tmppath;
1704 	}
1705 	AUDIT_ARG_TEXT(syspath);
1706 restart:
1707 	bwillwrite();
1708 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
1709 	    NOCACHE, segflg, path2, fd, &cap_symlinkat_rights,
1710 	    td);
1711 	if ((error = namei(&nd)) != 0)
1712 		goto out;
1713 	if (nd.ni_vp) {
1714 		NDFREE(&nd, NDF_ONLY_PNBUF);
1715 		if (nd.ni_vp == nd.ni_dvp)
1716 			vrele(nd.ni_dvp);
1717 		else
1718 			vput(nd.ni_dvp);
1719 		vrele(nd.ni_vp);
1720 		error = EEXIST;
1721 		goto out;
1722 	}
1723 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1724 		NDFREE(&nd, NDF_ONLY_PNBUF);
1725 		vput(nd.ni_dvp);
1726 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1727 			goto out;
1728 		goto restart;
1729 	}
1730 	VATTR_NULL(&vattr);
1731 	vattr.va_mode = ACCESSPERMS &~ td->td_proc->p_fd->fd_cmask;
1732 #ifdef MAC
1733 	vattr.va_type = VLNK;
1734 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1735 	    &vattr);
1736 	if (error != 0)
1737 		goto out2;
1738 #endif
1739 	error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, syspath);
1740 	if (error == 0)
1741 		vput(nd.ni_vp);
1742 #ifdef MAC
1743 out2:
1744 #endif
1745 	NDFREE(&nd, NDF_ONLY_PNBUF);
1746 	vput(nd.ni_dvp);
1747 	vn_finished_write(mp);
1748 	if (error == ERELOOKUP)
1749 		goto restart;
1750 out:
1751 	if (segflg != UIO_SYSSPACE)
1752 		uma_zfree(namei_zone, tmppath);
1753 	return (error);
1754 }
1755 
1756 /*
1757  * Delete a whiteout from the filesystem.
1758  */
1759 #ifndef _SYS_SYSPROTO_H_
1760 struct undelete_args {
1761 	char *path;
1762 };
1763 #endif
1764 int
1765 sys_undelete(struct thread *td, struct undelete_args *uap)
1766 {
1767 	struct mount *mp;
1768 	struct nameidata nd;
1769 	int error;
1770 
1771 restart:
1772 	bwillwrite();
1773 	NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | AUDITVNODE1,
1774 	    UIO_USERSPACE, uap->path, td);
1775 	error = namei(&nd);
1776 	if (error != 0)
1777 		return (error);
1778 
1779 	if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) {
1780 		NDFREE(&nd, NDF_ONLY_PNBUF);
1781 		if (nd.ni_vp == nd.ni_dvp)
1782 			vrele(nd.ni_dvp);
1783 		else
1784 			vput(nd.ni_dvp);
1785 		if (nd.ni_vp)
1786 			vrele(nd.ni_vp);
1787 		return (EEXIST);
1788 	}
1789 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1790 		NDFREE(&nd, NDF_ONLY_PNBUF);
1791 		vput(nd.ni_dvp);
1792 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1793 			return (error);
1794 		goto restart;
1795 	}
1796 	error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE);
1797 	NDFREE(&nd, NDF_ONLY_PNBUF);
1798 	vput(nd.ni_dvp);
1799 	vn_finished_write(mp);
1800 	if (error == ERELOOKUP)
1801 		goto restart;
1802 	return (error);
1803 }
1804 
1805 /*
1806  * Delete a name from the filesystem.
1807  */
1808 #ifndef _SYS_SYSPROTO_H_
1809 struct unlink_args {
1810 	char	*path;
1811 };
1812 #endif
1813 int
1814 sys_unlink(struct thread *td, struct unlink_args *uap)
1815 {
1816 
1817 	return (kern_funlinkat(td, AT_FDCWD, uap->path, FD_NONE, UIO_USERSPACE,
1818 	    0, 0));
1819 }
1820 
1821 static int
1822 kern_funlinkat_ex(struct thread *td, int dfd, const char *path, int fd,
1823     int flag, enum uio_seg pathseg, ino_t oldinum)
1824 {
1825 
1826 	if ((flag & ~AT_REMOVEDIR) != 0)
1827 		return (EINVAL);
1828 
1829 	if ((flag & AT_REMOVEDIR) != 0)
1830 		return (kern_frmdirat(td, dfd, path, fd, UIO_USERSPACE, 0));
1831 
1832 	return (kern_funlinkat(td, dfd, path, fd, UIO_USERSPACE, 0, 0));
1833 }
1834 
1835 #ifndef _SYS_SYSPROTO_H_
1836 struct unlinkat_args {
1837 	int	fd;
1838 	char	*path;
1839 	int	flag;
1840 };
1841 #endif
1842 int
1843 sys_unlinkat(struct thread *td, struct unlinkat_args *uap)
1844 {
1845 
1846 	return (kern_funlinkat_ex(td, uap->fd, uap->path, FD_NONE, uap->flag,
1847 	    UIO_USERSPACE, 0));
1848 }
1849 
1850 #ifndef _SYS_SYSPROTO_H_
1851 struct funlinkat_args {
1852 	int		dfd;
1853 	const char	*path;
1854 	int		fd;
1855 	int		flag;
1856 };
1857 #endif
1858 int
1859 sys_funlinkat(struct thread *td, struct funlinkat_args *uap)
1860 {
1861 
1862 	return (kern_funlinkat_ex(td, uap->dfd, uap->path, uap->fd, uap->flag,
1863 	    UIO_USERSPACE, 0));
1864 }
1865 
1866 int
1867 kern_funlinkat(struct thread *td, int dfd, const char *path, int fd,
1868     enum uio_seg pathseg, int flag, ino_t oldinum)
1869 {
1870 	struct mount *mp;
1871 	struct file *fp;
1872 	struct vnode *vp;
1873 	struct nameidata nd;
1874 	struct stat sb;
1875 	int error;
1876 
1877 	fp = NULL;
1878 	if (fd != FD_NONE) {
1879 		error = getvnode(td, fd, &cap_no_rights, &fp);
1880 		if (error != 0)
1881 			return (error);
1882 	}
1883 
1884 restart:
1885 	bwillwrite();
1886 	NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1 |
1887 	    at2cnpflags(flag, AT_BENEATH | AT_RESOLVE_BENEATH),
1888 	    pathseg, path, dfd, &cap_unlinkat_rights, td);
1889 	if ((error = namei(&nd)) != 0) {
1890 		if (error == EINVAL)
1891 			error = EPERM;
1892 		goto fdout;
1893 	}
1894 	vp = nd.ni_vp;
1895 	if (vp->v_type == VDIR && oldinum == 0) {
1896 		error = EPERM;		/* POSIX */
1897 	} else if (oldinum != 0 &&
1898 		  ((error = VOP_STAT(vp, &sb, td->td_ucred, NOCRED, td)) == 0) &&
1899 		  sb.st_ino != oldinum) {
1900 		error = EIDRM;	/* Identifier removed */
1901 	} else if (fp != NULL && fp->f_vnode != vp) {
1902 		if (VN_IS_DOOMED(fp->f_vnode))
1903 			error = EBADF;
1904 		else
1905 			error = EDEADLK;
1906 	} else {
1907 		/*
1908 		 * The root of a mounted filesystem cannot be deleted.
1909 		 *
1910 		 * XXX: can this only be a VDIR case?
1911 		 */
1912 		if (vp->v_vflag & VV_ROOT)
1913 			error = EBUSY;
1914 	}
1915 	if (error == 0) {
1916 		if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1917 			NDFREE(&nd, NDF_ONLY_PNBUF);
1918 			vput(nd.ni_dvp);
1919 			if (vp == nd.ni_dvp)
1920 				vrele(vp);
1921 			else
1922 				vput(vp);
1923 			if ((error = vn_start_write(NULL, &mp,
1924 			    V_XSLEEP | PCATCH)) != 0) {
1925 				goto fdout;
1926 			}
1927 			goto restart;
1928 		}
1929 #ifdef MAC
1930 		error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
1931 		    &nd.ni_cnd);
1932 		if (error != 0)
1933 			goto out;
1934 #endif
1935 		vfs_notify_upper(vp, VFS_NOTIFY_UPPER_UNLINK);
1936 		error = VOP_REMOVE(nd.ni_dvp, vp, &nd.ni_cnd);
1937 #ifdef MAC
1938 out:
1939 #endif
1940 		vn_finished_write(mp);
1941 	}
1942 	NDFREE(&nd, NDF_ONLY_PNBUF);
1943 	vput(nd.ni_dvp);
1944 	if (vp == nd.ni_dvp)
1945 		vrele(vp);
1946 	else
1947 		vput(vp);
1948 	if (error == ERELOOKUP)
1949 		goto restart;
1950 fdout:
1951 	if (fp != NULL)
1952 		fdrop(fp, td);
1953 	return (error);
1954 }
1955 
1956 /*
1957  * Reposition read/write file offset.
1958  */
1959 #ifndef _SYS_SYSPROTO_H_
1960 struct lseek_args {
1961 	int	fd;
1962 	int	pad;
1963 	off_t	offset;
1964 	int	whence;
1965 };
1966 #endif
1967 int
1968 sys_lseek(struct thread *td, struct lseek_args *uap)
1969 {
1970 
1971 	return (kern_lseek(td, uap->fd, uap->offset, uap->whence));
1972 }
1973 
1974 int
1975 kern_lseek(struct thread *td, int fd, off_t offset, int whence)
1976 {
1977 	struct file *fp;
1978 	int error;
1979 
1980 	AUDIT_ARG_FD(fd);
1981 	error = fget(td, fd, &cap_seek_rights, &fp);
1982 	if (error != 0)
1983 		return (error);
1984 	error = (fp->f_ops->fo_flags & DFLAG_SEEKABLE) != 0 ?
1985 	    fo_seek(fp, offset, whence, td) : ESPIPE;
1986 	fdrop(fp, td);
1987 	return (error);
1988 }
1989 
1990 #if defined(COMPAT_43)
1991 /*
1992  * Reposition read/write file offset.
1993  */
1994 #ifndef _SYS_SYSPROTO_H_
1995 struct olseek_args {
1996 	int	fd;
1997 	long	offset;
1998 	int	whence;
1999 };
2000 #endif
2001 int
2002 olseek(struct thread *td, struct olseek_args *uap)
2003 {
2004 
2005 	return (kern_lseek(td, uap->fd, uap->offset, uap->whence));
2006 }
2007 #endif /* COMPAT_43 */
2008 
2009 #if defined(COMPAT_FREEBSD6)
2010 /* Version with the 'pad' argument */
2011 int
2012 freebsd6_lseek(struct thread *td, struct freebsd6_lseek_args *uap)
2013 {
2014 
2015 	return (kern_lseek(td, uap->fd, uap->offset, uap->whence));
2016 }
2017 #endif
2018 
2019 /*
2020  * Check access permissions using passed credentials.
2021  */
2022 static int
2023 vn_access(struct vnode *vp, int user_flags, struct ucred *cred,
2024      struct thread *td)
2025 {
2026 	accmode_t accmode;
2027 	int error;
2028 
2029 	/* Flags == 0 means only check for existence. */
2030 	if (user_flags == 0)
2031 		return (0);
2032 
2033 	accmode = 0;
2034 	if (user_flags & R_OK)
2035 		accmode |= VREAD;
2036 	if (user_flags & W_OK)
2037 		accmode |= VWRITE;
2038 	if (user_flags & X_OK)
2039 		accmode |= VEXEC;
2040 #ifdef MAC
2041 	error = mac_vnode_check_access(cred, vp, accmode);
2042 	if (error != 0)
2043 		return (error);
2044 #endif
2045 	if ((accmode & VWRITE) == 0 || (error = vn_writechk(vp)) == 0)
2046 		error = VOP_ACCESS(vp, accmode, cred, td);
2047 	return (error);
2048 }
2049 
2050 /*
2051  * Check access permissions using "real" credentials.
2052  */
2053 #ifndef _SYS_SYSPROTO_H_
2054 struct access_args {
2055 	char	*path;
2056 	int	amode;
2057 };
2058 #endif
2059 int
2060 sys_access(struct thread *td, struct access_args *uap)
2061 {
2062 
2063 	return (kern_accessat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2064 	    0, uap->amode));
2065 }
2066 
2067 #ifndef _SYS_SYSPROTO_H_
2068 struct faccessat_args {
2069 	int	dirfd;
2070 	char	*path;
2071 	int	amode;
2072 	int	flag;
2073 }
2074 #endif
2075 int
2076 sys_faccessat(struct thread *td, struct faccessat_args *uap)
2077 {
2078 
2079 	return (kern_accessat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
2080 	    uap->amode));
2081 }
2082 
2083 int
2084 kern_accessat(struct thread *td, int fd, const char *path,
2085     enum uio_seg pathseg, int flag, int amode)
2086 {
2087 	struct ucred *cred, *usecred;
2088 	struct vnode *vp;
2089 	struct nameidata nd;
2090 	int error;
2091 
2092 	if ((flag & ~(AT_EACCESS | AT_BENEATH | AT_RESOLVE_BENEATH)) != 0)
2093 		return (EINVAL);
2094 	if (amode != F_OK && (amode & ~(R_OK | W_OK | X_OK)) != 0)
2095 		return (EINVAL);
2096 
2097 	/*
2098 	 * Create and modify a temporary credential instead of one that
2099 	 * is potentially shared (if we need one).
2100 	 */
2101 	cred = td->td_ucred;
2102 	if ((flag & AT_EACCESS) == 0 &&
2103 	    ((cred->cr_uid != cred->cr_ruid ||
2104 	    cred->cr_rgid != cred->cr_groups[0]))) {
2105 		usecred = crdup(cred);
2106 		usecred->cr_uid = cred->cr_ruid;
2107 		usecred->cr_groups[0] = cred->cr_rgid;
2108 		td->td_ucred = usecred;
2109 	} else
2110 		usecred = cred;
2111 	AUDIT_ARG_VALUE(amode);
2112 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF |
2113 	    AUDITVNODE1 | at2cnpflags(flag, AT_BENEATH | AT_RESOLVE_BENEATH),
2114 	    pathseg, path, fd, &cap_fstat_rights, td);
2115 	if ((error = namei(&nd)) != 0)
2116 		goto out;
2117 	vp = nd.ni_vp;
2118 
2119 	error = vn_access(vp, amode, usecred, td);
2120 	NDFREE(&nd, NDF_ONLY_PNBUF);
2121 	vput(vp);
2122 out:
2123 	if (usecred != cred) {
2124 		td->td_ucred = cred;
2125 		crfree(usecred);
2126 	}
2127 	return (error);
2128 }
2129 
2130 /*
2131  * Check access permissions using "effective" credentials.
2132  */
2133 #ifndef _SYS_SYSPROTO_H_
2134 struct eaccess_args {
2135 	char	*path;
2136 	int	amode;
2137 };
2138 #endif
2139 int
2140 sys_eaccess(struct thread *td, struct eaccess_args *uap)
2141 {
2142 
2143 	return (kern_accessat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2144 	    AT_EACCESS, uap->amode));
2145 }
2146 
2147 #if defined(COMPAT_43)
2148 /*
2149  * Get file status; this version follows links.
2150  */
2151 #ifndef _SYS_SYSPROTO_H_
2152 struct ostat_args {
2153 	char	*path;
2154 	struct ostat *ub;
2155 };
2156 #endif
2157 int
2158 ostat(struct thread *td, struct ostat_args *uap)
2159 {
2160 	struct stat sb;
2161 	struct ostat osb;
2162 	int error;
2163 
2164 	error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE,
2165 	    &sb, NULL);
2166 	if (error != 0)
2167 		return (error);
2168 	cvtstat(&sb, &osb);
2169 	return (copyout(&osb, uap->ub, sizeof (osb)));
2170 }
2171 
2172 /*
2173  * Get file status; this version does not follow links.
2174  */
2175 #ifndef _SYS_SYSPROTO_H_
2176 struct olstat_args {
2177 	char	*path;
2178 	struct ostat *ub;
2179 };
2180 #endif
2181 int
2182 olstat(struct thread *td, struct olstat_args *uap)
2183 {
2184 	struct stat sb;
2185 	struct ostat osb;
2186 	int error;
2187 
2188 	error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2189 	    UIO_USERSPACE, &sb, NULL);
2190 	if (error != 0)
2191 		return (error);
2192 	cvtstat(&sb, &osb);
2193 	return (copyout(&osb, uap->ub, sizeof (osb)));
2194 }
2195 
2196 /*
2197  * Convert from an old to a new stat structure.
2198  * XXX: many values are blindly truncated.
2199  */
2200 void
2201 cvtstat(struct stat *st, struct ostat *ost)
2202 {
2203 
2204 	bzero(ost, sizeof(*ost));
2205 	ost->st_dev = st->st_dev;
2206 	ost->st_ino = st->st_ino;
2207 	ost->st_mode = st->st_mode;
2208 	ost->st_nlink = st->st_nlink;
2209 	ost->st_uid = st->st_uid;
2210 	ost->st_gid = st->st_gid;
2211 	ost->st_rdev = st->st_rdev;
2212 	ost->st_size = MIN(st->st_size, INT32_MAX);
2213 	ost->st_atim = st->st_atim;
2214 	ost->st_mtim = st->st_mtim;
2215 	ost->st_ctim = st->st_ctim;
2216 	ost->st_blksize = st->st_blksize;
2217 	ost->st_blocks = st->st_blocks;
2218 	ost->st_flags = st->st_flags;
2219 	ost->st_gen = st->st_gen;
2220 }
2221 #endif /* COMPAT_43 */
2222 
2223 #if defined(COMPAT_43) || defined(COMPAT_FREEBSD11)
2224 int ino64_trunc_error;
2225 SYSCTL_INT(_vfs, OID_AUTO, ino64_trunc_error, CTLFLAG_RW,
2226     &ino64_trunc_error, 0,
2227     "Error on truncation of device, file or inode number, or link count");
2228 
2229 int
2230 freebsd11_cvtstat(struct stat *st, struct freebsd11_stat *ost)
2231 {
2232 
2233 	ost->st_dev = st->st_dev;
2234 	if (ost->st_dev != st->st_dev) {
2235 		switch (ino64_trunc_error) {
2236 		default:
2237 			/*
2238 			 * Since dev_t is almost raw, don't clamp to the
2239 			 * maximum for case 2, but ignore the error.
2240 			 */
2241 			break;
2242 		case 1:
2243 			return (EOVERFLOW);
2244 		}
2245 	}
2246 	ost->st_ino = st->st_ino;
2247 	if (ost->st_ino != st->st_ino) {
2248 		switch (ino64_trunc_error) {
2249 		default:
2250 		case 0:
2251 			break;
2252 		case 1:
2253 			return (EOVERFLOW);
2254 		case 2:
2255 			ost->st_ino = UINT32_MAX;
2256 			break;
2257 		}
2258 	}
2259 	ost->st_mode = st->st_mode;
2260 	ost->st_nlink = st->st_nlink;
2261 	if (ost->st_nlink != st->st_nlink) {
2262 		switch (ino64_trunc_error) {
2263 		default:
2264 		case 0:
2265 			break;
2266 		case 1:
2267 			return (EOVERFLOW);
2268 		case 2:
2269 			ost->st_nlink = UINT16_MAX;
2270 			break;
2271 		}
2272 	}
2273 	ost->st_uid = st->st_uid;
2274 	ost->st_gid = st->st_gid;
2275 	ost->st_rdev = st->st_rdev;
2276 	if (ost->st_rdev != st->st_rdev) {
2277 		switch (ino64_trunc_error) {
2278 		default:
2279 			break;
2280 		case 1:
2281 			return (EOVERFLOW);
2282 		}
2283 	}
2284 	ost->st_atim = st->st_atim;
2285 	ost->st_mtim = st->st_mtim;
2286 	ost->st_ctim = st->st_ctim;
2287 	ost->st_size = st->st_size;
2288 	ost->st_blocks = st->st_blocks;
2289 	ost->st_blksize = st->st_blksize;
2290 	ost->st_flags = st->st_flags;
2291 	ost->st_gen = st->st_gen;
2292 	ost->st_lspare = 0;
2293 	ost->st_birthtim = st->st_birthtim;
2294 	bzero((char *)&ost->st_birthtim + sizeof(ost->st_birthtim),
2295 	    sizeof(*ost) - offsetof(struct freebsd11_stat,
2296 	    st_birthtim) - sizeof(ost->st_birthtim));
2297 	return (0);
2298 }
2299 
2300 int
2301 freebsd11_stat(struct thread *td, struct freebsd11_stat_args* uap)
2302 {
2303 	struct stat sb;
2304 	struct freebsd11_stat osb;
2305 	int error;
2306 
2307 	error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE,
2308 	    &sb, NULL);
2309 	if (error != 0)
2310 		return (error);
2311 	error = freebsd11_cvtstat(&sb, &osb);
2312 	if (error == 0)
2313 		error = copyout(&osb, uap->ub, sizeof(osb));
2314 	return (error);
2315 }
2316 
2317 int
2318 freebsd11_lstat(struct thread *td, struct freebsd11_lstat_args* uap)
2319 {
2320 	struct stat sb;
2321 	struct freebsd11_stat osb;
2322 	int error;
2323 
2324 	error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2325 	    UIO_USERSPACE, &sb, NULL);
2326 	if (error != 0)
2327 		return (error);
2328 	error = freebsd11_cvtstat(&sb, &osb);
2329 	if (error == 0)
2330 		error = copyout(&osb, uap->ub, sizeof(osb));
2331 	return (error);
2332 }
2333 
2334 int
2335 freebsd11_fhstat(struct thread *td, struct freebsd11_fhstat_args* uap)
2336 {
2337 	struct fhandle fh;
2338 	struct stat sb;
2339 	struct freebsd11_stat osb;
2340 	int error;
2341 
2342 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
2343 	if (error != 0)
2344 		return (error);
2345 	error = kern_fhstat(td, fh, &sb);
2346 	if (error != 0)
2347 		return (error);
2348 	error = freebsd11_cvtstat(&sb, &osb);
2349 	if (error == 0)
2350 		error = copyout(&osb, uap->sb, sizeof(osb));
2351 	return (error);
2352 }
2353 
2354 int
2355 freebsd11_fstatat(struct thread *td, struct freebsd11_fstatat_args* uap)
2356 {
2357 	struct stat sb;
2358 	struct freebsd11_stat osb;
2359 	int error;
2360 
2361 	error = kern_statat(td, uap->flag, uap->fd, uap->path,
2362 	    UIO_USERSPACE, &sb, NULL);
2363 	if (error != 0)
2364 		return (error);
2365 	error = freebsd11_cvtstat(&sb, &osb);
2366 	if (error == 0)
2367 		error = copyout(&osb, uap->buf, sizeof(osb));
2368 	return (error);
2369 }
2370 #endif	/* COMPAT_FREEBSD11 */
2371 
2372 /*
2373  * Get file status
2374  */
2375 #ifndef _SYS_SYSPROTO_H_
2376 struct fstatat_args {
2377 	int	fd;
2378 	char	*path;
2379 	struct stat	*buf;
2380 	int	flag;
2381 }
2382 #endif
2383 int
2384 sys_fstatat(struct thread *td, struct fstatat_args *uap)
2385 {
2386 	struct stat sb;
2387 	int error;
2388 
2389 	error = kern_statat(td, uap->flag, uap->fd, uap->path,
2390 	    UIO_USERSPACE, &sb, NULL);
2391 	if (error == 0)
2392 		error = copyout(&sb, uap->buf, sizeof (sb));
2393 	return (error);
2394 }
2395 
2396 int
2397 kern_statat(struct thread *td, int flag, int fd, const char *path,
2398     enum uio_seg pathseg, struct stat *sbp,
2399     void (*hook)(struct vnode *vp, struct stat *sbp))
2400 {
2401 	struct nameidata nd;
2402 	int error;
2403 
2404 	if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_BENEATH |
2405 	    AT_RESOLVE_BENEATH)) != 0)
2406 		return (EINVAL);
2407 
2408 	NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(flag, AT_BENEATH |
2409 	    AT_RESOLVE_BENEATH | AT_SYMLINK_NOFOLLOW) | LOCKSHARED | LOCKLEAF |
2410 	    AUDITVNODE1, pathseg, path, fd, &cap_fstat_rights, td);
2411 
2412 	if ((error = namei(&nd)) != 0)
2413 		return (error);
2414 	error = VOP_STAT(nd.ni_vp, sbp, td->td_ucred, NOCRED, td);
2415 	if (error == 0) {
2416 		if (__predict_false(hook != NULL))
2417 			hook(nd.ni_vp, sbp);
2418 	}
2419 	NDFREE(&nd, NDF_ONLY_PNBUF);
2420 	vput(nd.ni_vp);
2421 #ifdef __STAT_TIME_T_EXT
2422 	sbp->st_atim_ext = 0;
2423 	sbp->st_mtim_ext = 0;
2424 	sbp->st_ctim_ext = 0;
2425 	sbp->st_btim_ext = 0;
2426 #endif
2427 #ifdef KTRACE
2428 	if (KTRPOINT(td, KTR_STRUCT))
2429 		ktrstat_error(sbp, error);
2430 #endif
2431 	return (error);
2432 }
2433 
2434 #if defined(COMPAT_FREEBSD11)
2435 /*
2436  * Implementation of the NetBSD [l]stat() functions.
2437  */
2438 void
2439 freebsd11_cvtnstat(struct stat *sb, struct nstat *nsb)
2440 {
2441 
2442 	bzero(nsb, sizeof(*nsb));
2443 	nsb->st_dev = sb->st_dev;
2444 	nsb->st_ino = sb->st_ino;
2445 	nsb->st_mode = sb->st_mode;
2446 	nsb->st_nlink = sb->st_nlink;
2447 	nsb->st_uid = sb->st_uid;
2448 	nsb->st_gid = sb->st_gid;
2449 	nsb->st_rdev = sb->st_rdev;
2450 	nsb->st_atim = sb->st_atim;
2451 	nsb->st_mtim = sb->st_mtim;
2452 	nsb->st_ctim = sb->st_ctim;
2453 	nsb->st_size = sb->st_size;
2454 	nsb->st_blocks = sb->st_blocks;
2455 	nsb->st_blksize = sb->st_blksize;
2456 	nsb->st_flags = sb->st_flags;
2457 	nsb->st_gen = sb->st_gen;
2458 	nsb->st_birthtim = sb->st_birthtim;
2459 }
2460 
2461 #ifndef _SYS_SYSPROTO_H_
2462 struct freebsd11_nstat_args {
2463 	char	*path;
2464 	struct nstat *ub;
2465 };
2466 #endif
2467 int
2468 freebsd11_nstat(struct thread *td, struct freebsd11_nstat_args *uap)
2469 {
2470 	struct stat sb;
2471 	struct nstat nsb;
2472 	int error;
2473 
2474 	error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE,
2475 	    &sb, NULL);
2476 	if (error != 0)
2477 		return (error);
2478 	freebsd11_cvtnstat(&sb, &nsb);
2479 	return (copyout(&nsb, uap->ub, sizeof (nsb)));
2480 }
2481 
2482 /*
2483  * NetBSD lstat.  Get file status; this version does not follow links.
2484  */
2485 #ifndef _SYS_SYSPROTO_H_
2486 struct freebsd11_nlstat_args {
2487 	char	*path;
2488 	struct nstat *ub;
2489 };
2490 #endif
2491 int
2492 freebsd11_nlstat(struct thread *td, struct freebsd11_nlstat_args *uap)
2493 {
2494 	struct stat sb;
2495 	struct nstat nsb;
2496 	int error;
2497 
2498 	error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2499 	    UIO_USERSPACE, &sb, NULL);
2500 	if (error != 0)
2501 		return (error);
2502 	freebsd11_cvtnstat(&sb, &nsb);
2503 	return (copyout(&nsb, uap->ub, sizeof (nsb)));
2504 }
2505 #endif /* COMPAT_FREEBSD11 */
2506 
2507 /*
2508  * Get configurable pathname variables.
2509  */
2510 #ifndef _SYS_SYSPROTO_H_
2511 struct pathconf_args {
2512 	char	*path;
2513 	int	name;
2514 };
2515 #endif
2516 int
2517 sys_pathconf(struct thread *td, struct pathconf_args *uap)
2518 {
2519 	long value;
2520 	int error;
2521 
2522 	error = kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name, FOLLOW,
2523 	    &value);
2524 	if (error == 0)
2525 		td->td_retval[0] = value;
2526 	return (error);
2527 }
2528 
2529 #ifndef _SYS_SYSPROTO_H_
2530 struct lpathconf_args {
2531 	char	*path;
2532 	int	name;
2533 };
2534 #endif
2535 int
2536 sys_lpathconf(struct thread *td, struct lpathconf_args *uap)
2537 {
2538 	long value;
2539 	int error;
2540 
2541 	error = kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name,
2542 	    NOFOLLOW, &value);
2543 	if (error == 0)
2544 		td->td_retval[0] = value;
2545 	return (error);
2546 }
2547 
2548 int
2549 kern_pathconf(struct thread *td, const char *path, enum uio_seg pathseg,
2550     int name, u_long flags, long *valuep)
2551 {
2552 	struct nameidata nd;
2553 	int error;
2554 
2555 	NDINIT(&nd, LOOKUP, LOCKSHARED | LOCKLEAF | AUDITVNODE1 | flags,
2556 	    pathseg, path, td);
2557 	if ((error = namei(&nd)) != 0)
2558 		return (error);
2559 	NDFREE(&nd, NDF_ONLY_PNBUF);
2560 
2561 	error = VOP_PATHCONF(nd.ni_vp, name, valuep);
2562 	vput(nd.ni_vp);
2563 	return (error);
2564 }
2565 
2566 /*
2567  * Return target name of a symbolic link.
2568  */
2569 #ifndef _SYS_SYSPROTO_H_
2570 struct readlink_args {
2571 	char	*path;
2572 	char	*buf;
2573 	size_t	count;
2574 };
2575 #endif
2576 int
2577 sys_readlink(struct thread *td, struct readlink_args *uap)
2578 {
2579 
2580 	return (kern_readlinkat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2581 	    uap->buf, UIO_USERSPACE, uap->count));
2582 }
2583 #ifndef _SYS_SYSPROTO_H_
2584 struct readlinkat_args {
2585 	int	fd;
2586 	char	*path;
2587 	char	*buf;
2588 	size_t	bufsize;
2589 };
2590 #endif
2591 int
2592 sys_readlinkat(struct thread *td, struct readlinkat_args *uap)
2593 {
2594 
2595 	return (kern_readlinkat(td, uap->fd, uap->path, UIO_USERSPACE,
2596 	    uap->buf, UIO_USERSPACE, uap->bufsize));
2597 }
2598 
2599 int
2600 kern_readlinkat(struct thread *td, int fd, const char *path,
2601     enum uio_seg pathseg, char *buf, enum uio_seg bufseg, size_t count)
2602 {
2603 	struct vnode *vp;
2604 	struct nameidata nd;
2605 	int error;
2606 
2607 	if (count > IOSIZE_MAX)
2608 		return (EINVAL);
2609 
2610 	NDINIT_AT(&nd, LOOKUP, NOFOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
2611 	    pathseg, path, fd, td);
2612 
2613 	if ((error = namei(&nd)) != 0)
2614 		return (error);
2615 	NDFREE(&nd, NDF_ONLY_PNBUF);
2616 	vp = nd.ni_vp;
2617 
2618 	error = kern_readlink_vp(vp, buf, bufseg, count, td);
2619 	vput(vp);
2620 
2621 	return (error);
2622 }
2623 
2624 /*
2625  * Helper function to readlink from a vnode
2626  */
2627 static int
2628 kern_readlink_vp(struct vnode *vp, char *buf, enum uio_seg bufseg, size_t count,
2629     struct thread *td)
2630 {
2631 	struct iovec aiov;
2632 	struct uio auio;
2633 	int error;
2634 
2635 	ASSERT_VOP_LOCKED(vp, "kern_readlink_vp(): vp not locked");
2636 #ifdef MAC
2637 	error = mac_vnode_check_readlink(td->td_ucred, vp);
2638 	if (error != 0)
2639 		return (error);
2640 #endif
2641 	if (vp->v_type != VLNK && (vp->v_vflag & VV_READLINK) == 0)
2642 		return (EINVAL);
2643 
2644 	aiov.iov_base = buf;
2645 	aiov.iov_len = count;
2646 	auio.uio_iov = &aiov;
2647 	auio.uio_iovcnt = 1;
2648 	auio.uio_offset = 0;
2649 	auio.uio_rw = UIO_READ;
2650 	auio.uio_segflg = bufseg;
2651 	auio.uio_td = td;
2652 	auio.uio_resid = count;
2653 	error = VOP_READLINK(vp, &auio, td->td_ucred);
2654 	td->td_retval[0] = count - auio.uio_resid;
2655 	return (error);
2656 }
2657 
2658 /*
2659  * Common implementation code for chflags() and fchflags().
2660  */
2661 static int
2662 setfflags(struct thread *td, struct vnode *vp, u_long flags)
2663 {
2664 	struct mount *mp;
2665 	struct vattr vattr;
2666 	int error;
2667 
2668 	/* We can't support the value matching VNOVAL. */
2669 	if (flags == VNOVAL)
2670 		return (EOPNOTSUPP);
2671 
2672 	/*
2673 	 * Prevent non-root users from setting flags on devices.  When
2674 	 * a device is reused, users can retain ownership of the device
2675 	 * if they are allowed to set flags and programs assume that
2676 	 * chown can't fail when done as root.
2677 	 */
2678 	if (vp->v_type == VCHR || vp->v_type == VBLK) {
2679 		error = priv_check(td, PRIV_VFS_CHFLAGS_DEV);
2680 		if (error != 0)
2681 			return (error);
2682 	}
2683 
2684 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2685 		return (error);
2686 	VATTR_NULL(&vattr);
2687 	vattr.va_flags = flags;
2688 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2689 #ifdef MAC
2690 	error = mac_vnode_check_setflags(td->td_ucred, vp, vattr.va_flags);
2691 	if (error == 0)
2692 #endif
2693 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
2694 	VOP_UNLOCK(vp);
2695 	vn_finished_write(mp);
2696 	return (error);
2697 }
2698 
2699 /*
2700  * Change flags of a file given a path name.
2701  */
2702 #ifndef _SYS_SYSPROTO_H_
2703 struct chflags_args {
2704 	const char *path;
2705 	u_long	flags;
2706 };
2707 #endif
2708 int
2709 sys_chflags(struct thread *td, struct chflags_args *uap)
2710 {
2711 
2712 	return (kern_chflagsat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2713 	    uap->flags, 0));
2714 }
2715 
2716 #ifndef _SYS_SYSPROTO_H_
2717 struct chflagsat_args {
2718 	int	fd;
2719 	const char *path;
2720 	u_long	flags;
2721 	int	atflag;
2722 }
2723 #endif
2724 int
2725 sys_chflagsat(struct thread *td, struct chflagsat_args *uap)
2726 {
2727 
2728 	if ((uap->atflag & ~(AT_SYMLINK_NOFOLLOW | AT_BENEATH |
2729 	    AT_RESOLVE_BENEATH)) != 0)
2730 		return (EINVAL);
2731 
2732 	return (kern_chflagsat(td, uap->fd, uap->path, UIO_USERSPACE,
2733 	    uap->flags, uap->atflag));
2734 }
2735 
2736 /*
2737  * Same as chflags() but doesn't follow symlinks.
2738  */
2739 #ifndef _SYS_SYSPROTO_H_
2740 struct lchflags_args {
2741 	const char *path;
2742 	u_long flags;
2743 };
2744 #endif
2745 int
2746 sys_lchflags(struct thread *td, struct lchflags_args *uap)
2747 {
2748 
2749 	return (kern_chflagsat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2750 	    uap->flags, AT_SYMLINK_NOFOLLOW));
2751 }
2752 
2753 static int
2754 kern_chflagsat(struct thread *td, int fd, const char *path,
2755     enum uio_seg pathseg, u_long flags, int atflag)
2756 {
2757 	struct nameidata nd;
2758 	int error;
2759 
2760 	AUDIT_ARG_FFLAGS(flags);
2761 	NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(atflag, AT_SYMLINK_NOFOLLOW |
2762 	    AT_BENEATH | AT_RESOLVE_BENEATH) | AUDITVNODE1, pathseg, path, fd,
2763 	    &cap_fchflags_rights, td);
2764 	if ((error = namei(&nd)) != 0)
2765 		return (error);
2766 	NDFREE(&nd, NDF_ONLY_PNBUF);
2767 	error = setfflags(td, nd.ni_vp, flags);
2768 	vrele(nd.ni_vp);
2769 	return (error);
2770 }
2771 
2772 /*
2773  * Change flags of a file given a file descriptor.
2774  */
2775 #ifndef _SYS_SYSPROTO_H_
2776 struct fchflags_args {
2777 	int	fd;
2778 	u_long	flags;
2779 };
2780 #endif
2781 int
2782 sys_fchflags(struct thread *td, struct fchflags_args *uap)
2783 {
2784 	struct file *fp;
2785 	int error;
2786 
2787 	AUDIT_ARG_FD(uap->fd);
2788 	AUDIT_ARG_FFLAGS(uap->flags);
2789 	error = getvnode(td, uap->fd, &cap_fchflags_rights,
2790 	    &fp);
2791 	if (error != 0)
2792 		return (error);
2793 #ifdef AUDIT
2794 	vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
2795 	AUDIT_ARG_VNODE1(fp->f_vnode);
2796 	VOP_UNLOCK(fp->f_vnode);
2797 #endif
2798 	error = setfflags(td, fp->f_vnode, uap->flags);
2799 	fdrop(fp, td);
2800 	return (error);
2801 }
2802 
2803 /*
2804  * Common implementation code for chmod(), lchmod() and fchmod().
2805  */
2806 int
2807 setfmode(struct thread *td, struct ucred *cred, struct vnode *vp, int mode)
2808 {
2809 	struct mount *mp;
2810 	struct vattr vattr;
2811 	int error;
2812 
2813 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2814 		return (error);
2815 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2816 	VATTR_NULL(&vattr);
2817 	vattr.va_mode = mode & ALLPERMS;
2818 #ifdef MAC
2819 	error = mac_vnode_check_setmode(cred, vp, vattr.va_mode);
2820 	if (error == 0)
2821 #endif
2822 		error = VOP_SETATTR(vp, &vattr, cred);
2823 	VOP_UNLOCK(vp);
2824 	vn_finished_write(mp);
2825 	return (error);
2826 }
2827 
2828 /*
2829  * Change mode of a file given path name.
2830  */
2831 #ifndef _SYS_SYSPROTO_H_
2832 struct chmod_args {
2833 	char	*path;
2834 	int	mode;
2835 };
2836 #endif
2837 int
2838 sys_chmod(struct thread *td, struct chmod_args *uap)
2839 {
2840 
2841 	return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2842 	    uap->mode, 0));
2843 }
2844 
2845 #ifndef _SYS_SYSPROTO_H_
2846 struct fchmodat_args {
2847 	int	dirfd;
2848 	char	*path;
2849 	mode_t	mode;
2850 	int	flag;
2851 }
2852 #endif
2853 int
2854 sys_fchmodat(struct thread *td, struct fchmodat_args *uap)
2855 {
2856 
2857 	if ((uap->flag & ~(AT_SYMLINK_NOFOLLOW | AT_BENEATH |
2858 	    AT_RESOLVE_BENEATH)) != 0)
2859 		return (EINVAL);
2860 
2861 	return (kern_fchmodat(td, uap->fd, uap->path, UIO_USERSPACE,
2862 	    uap->mode, uap->flag));
2863 }
2864 
2865 /*
2866  * Change mode of a file given path name (don't follow links.)
2867  */
2868 #ifndef _SYS_SYSPROTO_H_
2869 struct lchmod_args {
2870 	char	*path;
2871 	int	mode;
2872 };
2873 #endif
2874 int
2875 sys_lchmod(struct thread *td, struct lchmod_args *uap)
2876 {
2877 
2878 	return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2879 	    uap->mode, AT_SYMLINK_NOFOLLOW));
2880 }
2881 
2882 int
2883 kern_fchmodat(struct thread *td, int fd, const char *path,
2884     enum uio_seg pathseg, mode_t mode, int flag)
2885 {
2886 	struct nameidata nd;
2887 	int error;
2888 
2889 	AUDIT_ARG_MODE(mode);
2890 	NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(flag, AT_SYMLINK_NOFOLLOW |
2891 	    AT_BENEATH | AT_RESOLVE_BENEATH) | AUDITVNODE1, pathseg, path, fd,
2892 	    &cap_fchmod_rights, td);
2893 	if ((error = namei(&nd)) != 0)
2894 		return (error);
2895 	NDFREE(&nd, NDF_ONLY_PNBUF);
2896 	error = setfmode(td, td->td_ucred, nd.ni_vp, mode);
2897 	vrele(nd.ni_vp);
2898 	return (error);
2899 }
2900 
2901 /*
2902  * Change mode of a file given a file descriptor.
2903  */
2904 #ifndef _SYS_SYSPROTO_H_
2905 struct fchmod_args {
2906 	int	fd;
2907 	int	mode;
2908 };
2909 #endif
2910 int
2911 sys_fchmod(struct thread *td, struct fchmod_args *uap)
2912 {
2913 	struct file *fp;
2914 	int error;
2915 
2916 	AUDIT_ARG_FD(uap->fd);
2917 	AUDIT_ARG_MODE(uap->mode);
2918 
2919 	error = fget(td, uap->fd, &cap_fchmod_rights, &fp);
2920 	if (error != 0)
2921 		return (error);
2922 	error = fo_chmod(fp, uap->mode, td->td_ucred, td);
2923 	fdrop(fp, td);
2924 	return (error);
2925 }
2926 
2927 /*
2928  * Common implementation for chown(), lchown(), and fchown()
2929  */
2930 int
2931 setfown(struct thread *td, struct ucred *cred, struct vnode *vp, uid_t uid,
2932     gid_t gid)
2933 {
2934 	struct mount *mp;
2935 	struct vattr vattr;
2936 	int error;
2937 
2938 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2939 		return (error);
2940 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2941 	VATTR_NULL(&vattr);
2942 	vattr.va_uid = uid;
2943 	vattr.va_gid = gid;
2944 #ifdef MAC
2945 	error = mac_vnode_check_setowner(cred, vp, vattr.va_uid,
2946 	    vattr.va_gid);
2947 	if (error == 0)
2948 #endif
2949 		error = VOP_SETATTR(vp, &vattr, cred);
2950 	VOP_UNLOCK(vp);
2951 	vn_finished_write(mp);
2952 	return (error);
2953 }
2954 
2955 /*
2956  * Set ownership given a path name.
2957  */
2958 #ifndef _SYS_SYSPROTO_H_
2959 struct chown_args {
2960 	char	*path;
2961 	int	uid;
2962 	int	gid;
2963 };
2964 #endif
2965 int
2966 sys_chown(struct thread *td, struct chown_args *uap)
2967 {
2968 
2969 	return (kern_fchownat(td, AT_FDCWD, uap->path, UIO_USERSPACE, uap->uid,
2970 	    uap->gid, 0));
2971 }
2972 
2973 #ifndef _SYS_SYSPROTO_H_
2974 struct fchownat_args {
2975 	int fd;
2976 	const char * path;
2977 	uid_t uid;
2978 	gid_t gid;
2979 	int flag;
2980 };
2981 #endif
2982 int
2983 sys_fchownat(struct thread *td, struct fchownat_args *uap)
2984 {
2985 
2986 	if ((uap->flag & ~(AT_SYMLINK_NOFOLLOW | AT_BENEATH |
2987 	    AT_RESOLVE_BENEATH)) != 0)
2988 		return (EINVAL);
2989 
2990 	return (kern_fchownat(td, uap->fd, uap->path, UIO_USERSPACE, uap->uid,
2991 	    uap->gid, uap->flag));
2992 }
2993 
2994 int
2995 kern_fchownat(struct thread *td, int fd, const char *path,
2996     enum uio_seg pathseg, int uid, int gid, int flag)
2997 {
2998 	struct nameidata nd;
2999 	int error;
3000 
3001 	AUDIT_ARG_OWNER(uid, gid);
3002 	NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(flag, AT_SYMLINK_NOFOLLOW |
3003 	    AT_BENEATH | AT_RESOLVE_BENEATH) | AUDITVNODE1, pathseg, path, fd,
3004 	    &cap_fchown_rights, td);
3005 
3006 	if ((error = namei(&nd)) != 0)
3007 		return (error);
3008 	NDFREE(&nd, NDF_ONLY_PNBUF);
3009 	error = setfown(td, td->td_ucred, nd.ni_vp, uid, gid);
3010 	vrele(nd.ni_vp);
3011 	return (error);
3012 }
3013 
3014 /*
3015  * Set ownership given a path name, do not cross symlinks.
3016  */
3017 #ifndef _SYS_SYSPROTO_H_
3018 struct lchown_args {
3019 	char	*path;
3020 	int	uid;
3021 	int	gid;
3022 };
3023 #endif
3024 int
3025 sys_lchown(struct thread *td, struct lchown_args *uap)
3026 {
3027 
3028 	return (kern_fchownat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
3029 	    uap->uid, uap->gid, AT_SYMLINK_NOFOLLOW));
3030 }
3031 
3032 /*
3033  * Set ownership given a file descriptor.
3034  */
3035 #ifndef _SYS_SYSPROTO_H_
3036 struct fchown_args {
3037 	int	fd;
3038 	int	uid;
3039 	int	gid;
3040 };
3041 #endif
3042 int
3043 sys_fchown(struct thread *td, struct fchown_args *uap)
3044 {
3045 	struct file *fp;
3046 	int error;
3047 
3048 	AUDIT_ARG_FD(uap->fd);
3049 	AUDIT_ARG_OWNER(uap->uid, uap->gid);
3050 	error = fget(td, uap->fd, &cap_fchown_rights, &fp);
3051 	if (error != 0)
3052 		return (error);
3053 	error = fo_chown(fp, uap->uid, uap->gid, td->td_ucred, td);
3054 	fdrop(fp, td);
3055 	return (error);
3056 }
3057 
3058 /*
3059  * Common implementation code for utimes(), lutimes(), and futimes().
3060  */
3061 static int
3062 getutimes(const struct timeval *usrtvp, enum uio_seg tvpseg,
3063     struct timespec *tsp)
3064 {
3065 	struct timeval tv[2];
3066 	const struct timeval *tvp;
3067 	int error;
3068 
3069 	if (usrtvp == NULL) {
3070 		vfs_timestamp(&tsp[0]);
3071 		tsp[1] = tsp[0];
3072 	} else {
3073 		if (tvpseg == UIO_SYSSPACE) {
3074 			tvp = usrtvp;
3075 		} else {
3076 			if ((error = copyin(usrtvp, tv, sizeof(tv))) != 0)
3077 				return (error);
3078 			tvp = tv;
3079 		}
3080 
3081 		if (tvp[0].tv_usec < 0 || tvp[0].tv_usec >= 1000000 ||
3082 		    tvp[1].tv_usec < 0 || tvp[1].tv_usec >= 1000000)
3083 			return (EINVAL);
3084 		TIMEVAL_TO_TIMESPEC(&tvp[0], &tsp[0]);
3085 		TIMEVAL_TO_TIMESPEC(&tvp[1], &tsp[1]);
3086 	}
3087 	return (0);
3088 }
3089 
3090 /*
3091  * Common implementation code for futimens(), utimensat().
3092  */
3093 #define	UTIMENS_NULL	0x1
3094 #define	UTIMENS_EXIT	0x2
3095 static int
3096 getutimens(const struct timespec *usrtsp, enum uio_seg tspseg,
3097     struct timespec *tsp, int *retflags)
3098 {
3099 	struct timespec tsnow;
3100 	int error;
3101 
3102 	vfs_timestamp(&tsnow);
3103 	*retflags = 0;
3104 	if (usrtsp == NULL) {
3105 		tsp[0] = tsnow;
3106 		tsp[1] = tsnow;
3107 		*retflags |= UTIMENS_NULL;
3108 		return (0);
3109 	}
3110 	if (tspseg == UIO_SYSSPACE) {
3111 		tsp[0] = usrtsp[0];
3112 		tsp[1] = usrtsp[1];
3113 	} else if ((error = copyin(usrtsp, tsp, sizeof(*tsp) * 2)) != 0)
3114 		return (error);
3115 	if (tsp[0].tv_nsec == UTIME_OMIT && tsp[1].tv_nsec == UTIME_OMIT)
3116 		*retflags |= UTIMENS_EXIT;
3117 	if (tsp[0].tv_nsec == UTIME_NOW && tsp[1].tv_nsec == UTIME_NOW)
3118 		*retflags |= UTIMENS_NULL;
3119 	if (tsp[0].tv_nsec == UTIME_OMIT)
3120 		tsp[0].tv_sec = VNOVAL;
3121 	else if (tsp[0].tv_nsec == UTIME_NOW)
3122 		tsp[0] = tsnow;
3123 	else if (tsp[0].tv_nsec < 0 || tsp[0].tv_nsec >= 1000000000L)
3124 		return (EINVAL);
3125 	if (tsp[1].tv_nsec == UTIME_OMIT)
3126 		tsp[1].tv_sec = VNOVAL;
3127 	else if (tsp[1].tv_nsec == UTIME_NOW)
3128 		tsp[1] = tsnow;
3129 	else if (tsp[1].tv_nsec < 0 || tsp[1].tv_nsec >= 1000000000L)
3130 		return (EINVAL);
3131 
3132 	return (0);
3133 }
3134 
3135 /*
3136  * Common implementation code for utimes(), lutimes(), futimes(), futimens(),
3137  * and utimensat().
3138  */
3139 static int
3140 setutimes(struct thread *td, struct vnode *vp, const struct timespec *ts,
3141     int numtimes, int nullflag)
3142 {
3143 	struct mount *mp;
3144 	struct vattr vattr;
3145 	int error, setbirthtime;
3146 
3147 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
3148 		return (error);
3149 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3150 	setbirthtime = 0;
3151 	if (numtimes < 3 && !VOP_GETATTR(vp, &vattr, td->td_ucred) &&
3152 	    timespeccmp(&ts[1], &vattr.va_birthtime, < ))
3153 		setbirthtime = 1;
3154 	VATTR_NULL(&vattr);
3155 	vattr.va_atime = ts[0];
3156 	vattr.va_mtime = ts[1];
3157 	if (setbirthtime)
3158 		vattr.va_birthtime = ts[1];
3159 	if (numtimes > 2)
3160 		vattr.va_birthtime = ts[2];
3161 	if (nullflag)
3162 		vattr.va_vaflags |= VA_UTIMES_NULL;
3163 #ifdef MAC
3164 	error = mac_vnode_check_setutimes(td->td_ucred, vp, vattr.va_atime,
3165 	    vattr.va_mtime);
3166 #endif
3167 	if (error == 0)
3168 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3169 	VOP_UNLOCK(vp);
3170 	vn_finished_write(mp);
3171 	return (error);
3172 }
3173 
3174 /*
3175  * Set the access and modification times of a file.
3176  */
3177 #ifndef _SYS_SYSPROTO_H_
3178 struct utimes_args {
3179 	char	*path;
3180 	struct	timeval *tptr;
3181 };
3182 #endif
3183 int
3184 sys_utimes(struct thread *td, struct utimes_args *uap)
3185 {
3186 
3187 	return (kern_utimesat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
3188 	    uap->tptr, UIO_USERSPACE));
3189 }
3190 
3191 #ifndef _SYS_SYSPROTO_H_
3192 struct futimesat_args {
3193 	int fd;
3194 	const char * path;
3195 	const struct timeval * times;
3196 };
3197 #endif
3198 int
3199 sys_futimesat(struct thread *td, struct futimesat_args *uap)
3200 {
3201 
3202 	return (kern_utimesat(td, uap->fd, uap->path, UIO_USERSPACE,
3203 	    uap->times, UIO_USERSPACE));
3204 }
3205 
3206 int
3207 kern_utimesat(struct thread *td, int fd, const char *path,
3208     enum uio_seg pathseg, struct timeval *tptr, enum uio_seg tptrseg)
3209 {
3210 	struct nameidata nd;
3211 	struct timespec ts[2];
3212 	int error;
3213 
3214 	if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3215 		return (error);
3216 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, fd,
3217 	    &cap_futimes_rights, td);
3218 
3219 	if ((error = namei(&nd)) != 0)
3220 		return (error);
3221 	NDFREE(&nd, NDF_ONLY_PNBUF);
3222 	error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3223 	vrele(nd.ni_vp);
3224 	return (error);
3225 }
3226 
3227 /*
3228  * Set the access and modification times of a file.
3229  */
3230 #ifndef _SYS_SYSPROTO_H_
3231 struct lutimes_args {
3232 	char	*path;
3233 	struct	timeval *tptr;
3234 };
3235 #endif
3236 int
3237 sys_lutimes(struct thread *td, struct lutimes_args *uap)
3238 {
3239 
3240 	return (kern_lutimes(td, uap->path, UIO_USERSPACE, uap->tptr,
3241 	    UIO_USERSPACE));
3242 }
3243 
3244 int
3245 kern_lutimes(struct thread *td, const char *path, enum uio_seg pathseg,
3246     struct timeval *tptr, enum uio_seg tptrseg)
3247 {
3248 	struct timespec ts[2];
3249 	struct nameidata nd;
3250 	int error;
3251 
3252 	if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3253 		return (error);
3254 	NDINIT(&nd, LOOKUP, NOFOLLOW | AUDITVNODE1, pathseg, path, td);
3255 	if ((error = namei(&nd)) != 0)
3256 		return (error);
3257 	NDFREE(&nd, NDF_ONLY_PNBUF);
3258 	error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3259 	vrele(nd.ni_vp);
3260 	return (error);
3261 }
3262 
3263 /*
3264  * Set the access and modification times of a file.
3265  */
3266 #ifndef _SYS_SYSPROTO_H_
3267 struct futimes_args {
3268 	int	fd;
3269 	struct	timeval *tptr;
3270 };
3271 #endif
3272 int
3273 sys_futimes(struct thread *td, struct futimes_args *uap)
3274 {
3275 
3276 	return (kern_futimes(td, uap->fd, uap->tptr, UIO_USERSPACE));
3277 }
3278 
3279 int
3280 kern_futimes(struct thread *td, int fd, struct timeval *tptr,
3281     enum uio_seg tptrseg)
3282 {
3283 	struct timespec ts[2];
3284 	struct file *fp;
3285 	int error;
3286 
3287 	AUDIT_ARG_FD(fd);
3288 	error = getutimes(tptr, tptrseg, ts);
3289 	if (error != 0)
3290 		return (error);
3291 	error = getvnode(td, fd, &cap_futimes_rights, &fp);
3292 	if (error != 0)
3293 		return (error);
3294 #ifdef AUDIT
3295 	vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
3296 	AUDIT_ARG_VNODE1(fp->f_vnode);
3297 	VOP_UNLOCK(fp->f_vnode);
3298 #endif
3299 	error = setutimes(td, fp->f_vnode, ts, 2, tptr == NULL);
3300 	fdrop(fp, td);
3301 	return (error);
3302 }
3303 
3304 int
3305 sys_futimens(struct thread *td, struct futimens_args *uap)
3306 {
3307 
3308 	return (kern_futimens(td, uap->fd, uap->times, UIO_USERSPACE));
3309 }
3310 
3311 int
3312 kern_futimens(struct thread *td, int fd, struct timespec *tptr,
3313     enum uio_seg tptrseg)
3314 {
3315 	struct timespec ts[2];
3316 	struct file *fp;
3317 	int error, flags;
3318 
3319 	AUDIT_ARG_FD(fd);
3320 	error = getutimens(tptr, tptrseg, ts, &flags);
3321 	if (error != 0)
3322 		return (error);
3323 	if (flags & UTIMENS_EXIT)
3324 		return (0);
3325 	error = getvnode(td, fd, &cap_futimes_rights, &fp);
3326 	if (error != 0)
3327 		return (error);
3328 #ifdef AUDIT
3329 	vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
3330 	AUDIT_ARG_VNODE1(fp->f_vnode);
3331 	VOP_UNLOCK(fp->f_vnode);
3332 #endif
3333 	error = setutimes(td, fp->f_vnode, ts, 2, flags & UTIMENS_NULL);
3334 	fdrop(fp, td);
3335 	return (error);
3336 }
3337 
3338 int
3339 sys_utimensat(struct thread *td, struct utimensat_args *uap)
3340 {
3341 
3342 	return (kern_utimensat(td, uap->fd, uap->path, UIO_USERSPACE,
3343 	    uap->times, UIO_USERSPACE, uap->flag));
3344 }
3345 
3346 int
3347 kern_utimensat(struct thread *td, int fd, const char *path,
3348     enum uio_seg pathseg, struct timespec *tptr, enum uio_seg tptrseg,
3349     int flag)
3350 {
3351 	struct nameidata nd;
3352 	struct timespec ts[2];
3353 	int error, flags;
3354 
3355 	if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_BENEATH |
3356 	    AT_RESOLVE_BENEATH)) != 0)
3357 		return (EINVAL);
3358 
3359 	if ((error = getutimens(tptr, tptrseg, ts, &flags)) != 0)
3360 		return (error);
3361 	NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(flag, AT_SYMLINK_NOFOLLOW |
3362 	    AT_BENEATH | AT_RESOLVE_BENEATH) | AUDITVNODE1,
3363 	    pathseg, path, fd, &cap_futimes_rights, td);
3364 	if ((error = namei(&nd)) != 0)
3365 		return (error);
3366 	/*
3367 	 * We are allowed to call namei() regardless of 2xUTIME_OMIT.
3368 	 * POSIX states:
3369 	 * "If both tv_nsec fields are UTIME_OMIT... EACCESS may be detected."
3370 	 * "Search permission is denied by a component of the path prefix."
3371 	 */
3372 	NDFREE(&nd, NDF_ONLY_PNBUF);
3373 	if ((flags & UTIMENS_EXIT) == 0)
3374 		error = setutimes(td, nd.ni_vp, ts, 2, flags & UTIMENS_NULL);
3375 	vrele(nd.ni_vp);
3376 	return (error);
3377 }
3378 
3379 /*
3380  * Truncate a file given its path name.
3381  */
3382 #ifndef _SYS_SYSPROTO_H_
3383 struct truncate_args {
3384 	char	*path;
3385 	int	pad;
3386 	off_t	length;
3387 };
3388 #endif
3389 int
3390 sys_truncate(struct thread *td, struct truncate_args *uap)
3391 {
3392 
3393 	return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3394 }
3395 
3396 int
3397 kern_truncate(struct thread *td, const char *path, enum uio_seg pathseg,
3398     off_t length)
3399 {
3400 	struct mount *mp;
3401 	struct vnode *vp;
3402 	void *rl_cookie;
3403 	struct vattr vattr;
3404 	struct nameidata nd;
3405 	int error;
3406 
3407 	if (length < 0)
3408 		return (EINVAL);
3409 retry:
3410 	NDINIT(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, td);
3411 	if ((error = namei(&nd)) != 0)
3412 		return (error);
3413 	vp = nd.ni_vp;
3414 	rl_cookie = vn_rangelock_wlock(vp, 0, OFF_MAX);
3415 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) {
3416 		vn_rangelock_unlock(vp, rl_cookie);
3417 		vrele(vp);
3418 		return (error);
3419 	}
3420 	NDFREE(&nd, NDF_ONLY_PNBUF);
3421 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3422 	if (vp->v_type == VDIR)
3423 		error = EISDIR;
3424 #ifdef MAC
3425 	else if ((error = mac_vnode_check_write(td->td_ucred, NOCRED, vp))) {
3426 	}
3427 #endif
3428 	else if ((error = vn_writechk(vp)) == 0 &&
3429 	    (error = VOP_ACCESS(vp, VWRITE, td->td_ucred, td)) == 0) {
3430 		VATTR_NULL(&vattr);
3431 		vattr.va_size = length;
3432 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3433 	}
3434 	VOP_UNLOCK(vp);
3435 	vn_finished_write(mp);
3436 	vn_rangelock_unlock(vp, rl_cookie);
3437 	vrele(vp);
3438 	if (error == ERELOOKUP)
3439 		goto retry;
3440 	return (error);
3441 }
3442 
3443 #if defined(COMPAT_43)
3444 /*
3445  * Truncate a file given its path name.
3446  */
3447 #ifndef _SYS_SYSPROTO_H_
3448 struct otruncate_args {
3449 	char	*path;
3450 	long	length;
3451 };
3452 #endif
3453 int
3454 otruncate(struct thread *td, struct otruncate_args *uap)
3455 {
3456 
3457 	return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3458 }
3459 #endif /* COMPAT_43 */
3460 
3461 #if defined(COMPAT_FREEBSD6)
3462 /* Versions with the pad argument */
3463 int
3464 freebsd6_truncate(struct thread *td, struct freebsd6_truncate_args *uap)
3465 {
3466 
3467 	return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3468 }
3469 
3470 int
3471 freebsd6_ftruncate(struct thread *td, struct freebsd6_ftruncate_args *uap)
3472 {
3473 
3474 	return (kern_ftruncate(td, uap->fd, uap->length));
3475 }
3476 #endif
3477 
3478 int
3479 kern_fsync(struct thread *td, int fd, bool fullsync)
3480 {
3481 	struct vnode *vp;
3482 	struct mount *mp;
3483 	struct file *fp;
3484 	int error, lock_flags;
3485 
3486 	AUDIT_ARG_FD(fd);
3487 	error = getvnode(td, fd, &cap_fsync_rights, &fp);
3488 	if (error != 0)
3489 		return (error);
3490 	vp = fp->f_vnode;
3491 #if 0
3492 	if (!fullsync)
3493 		/* XXXKIB: compete outstanding aio writes */;
3494 #endif
3495 retry:
3496 	error = vn_start_write(vp, &mp, V_WAIT | PCATCH);
3497 	if (error != 0)
3498 		goto drop;
3499 	if (MNT_SHARED_WRITES(mp) ||
3500 	    ((mp == NULL) && MNT_SHARED_WRITES(vp->v_mount))) {
3501 		lock_flags = LK_SHARED;
3502 	} else {
3503 		lock_flags = LK_EXCLUSIVE;
3504 	}
3505 	vn_lock(vp, lock_flags | LK_RETRY);
3506 	AUDIT_ARG_VNODE1(vp);
3507 	if (vp->v_object != NULL) {
3508 		VM_OBJECT_WLOCK(vp->v_object);
3509 		vm_object_page_clean(vp->v_object, 0, 0, 0);
3510 		VM_OBJECT_WUNLOCK(vp->v_object);
3511 	}
3512 	error = fullsync ? VOP_FSYNC(vp, MNT_WAIT, td) : VOP_FDATASYNC(vp, td);
3513 	VOP_UNLOCK(vp);
3514 	vn_finished_write(mp);
3515 	if (error == ERELOOKUP)
3516 		goto retry;
3517 drop:
3518 	fdrop(fp, td);
3519 	return (error);
3520 }
3521 
3522 /*
3523  * Sync an open file.
3524  */
3525 #ifndef _SYS_SYSPROTO_H_
3526 struct fsync_args {
3527 	int	fd;
3528 };
3529 #endif
3530 int
3531 sys_fsync(struct thread *td, struct fsync_args *uap)
3532 {
3533 
3534 	return (kern_fsync(td, uap->fd, true));
3535 }
3536 
3537 int
3538 sys_fdatasync(struct thread *td, struct fdatasync_args *uap)
3539 {
3540 
3541 	return (kern_fsync(td, uap->fd, false));
3542 }
3543 
3544 /*
3545  * Rename files.  Source and destination must either both be directories, or
3546  * both not be directories.  If target is a directory, it must be empty.
3547  */
3548 #ifndef _SYS_SYSPROTO_H_
3549 struct rename_args {
3550 	char	*from;
3551 	char	*to;
3552 };
3553 #endif
3554 int
3555 sys_rename(struct thread *td, struct rename_args *uap)
3556 {
3557 
3558 	return (kern_renameat(td, AT_FDCWD, uap->from, AT_FDCWD,
3559 	    uap->to, UIO_USERSPACE));
3560 }
3561 
3562 #ifndef _SYS_SYSPROTO_H_
3563 struct renameat_args {
3564 	int	oldfd;
3565 	char	*old;
3566 	int	newfd;
3567 	char	*new;
3568 };
3569 #endif
3570 int
3571 sys_renameat(struct thread *td, struct renameat_args *uap)
3572 {
3573 
3574 	return (kern_renameat(td, uap->oldfd, uap->old, uap->newfd, uap->new,
3575 	    UIO_USERSPACE));
3576 }
3577 
3578 #ifdef MAC
3579 static int
3580 kern_renameat_mac(struct thread *td, int oldfd, const char *old, int newfd,
3581     const char *new, enum uio_seg pathseg, struct nameidata *fromnd)
3582 {
3583 	int error;
3584 
3585 	NDINIT_ATRIGHTS(fromnd, DELETE, LOCKPARENT | LOCKLEAF | SAVESTART |
3586 	    AUDITVNODE1, pathseg, old, oldfd, &cap_renameat_source_rights, td);
3587 	if ((error = namei(fromnd)) != 0)
3588 		return (error);
3589 	error = mac_vnode_check_rename_from(td->td_ucred, fromnd->ni_dvp,
3590 	    fromnd->ni_vp, &fromnd->ni_cnd);
3591 	VOP_UNLOCK(fromnd->ni_dvp);
3592 	if (fromnd->ni_dvp != fromnd->ni_vp)
3593 		VOP_UNLOCK(fromnd->ni_vp);
3594 	if (error != 0) {
3595 		NDFREE(fromnd, NDF_ONLY_PNBUF);
3596 		vrele(fromnd->ni_dvp);
3597 		vrele(fromnd->ni_vp);
3598 		if (fromnd->ni_startdir)
3599 			vrele(fromnd->ni_startdir);
3600 	}
3601 	return (error);
3602 }
3603 #endif
3604 
3605 int
3606 kern_renameat(struct thread *td, int oldfd, const char *old, int newfd,
3607     const char *new, enum uio_seg pathseg)
3608 {
3609 	struct mount *mp = NULL;
3610 	struct vnode *tvp, *fvp, *tdvp;
3611 	struct nameidata fromnd, tond;
3612 	int error;
3613 
3614 again:
3615 	bwillwrite();
3616 #ifdef MAC
3617 	if (mac_vnode_check_rename_from_enabled()) {
3618 		error = kern_renameat_mac(td, oldfd, old, newfd, new, pathseg,
3619 		    &fromnd);
3620 		if (error != 0)
3621 			return (error);
3622 	} else {
3623 #endif
3624 	NDINIT_ATRIGHTS(&fromnd, DELETE, WANTPARENT | SAVESTART | AUDITVNODE1,
3625 	    pathseg, old, oldfd, &cap_renameat_source_rights, td);
3626 	if ((error = namei(&fromnd)) != 0)
3627 		return (error);
3628 #ifdef MAC
3629 	}
3630 #endif
3631 	fvp = fromnd.ni_vp;
3632 	NDINIT_ATRIGHTS(&tond, RENAME, LOCKPARENT | LOCKLEAF | NOCACHE |
3633 	    SAVESTART | AUDITVNODE2, pathseg, new, newfd,
3634 	    &cap_renameat_target_rights, td);
3635 	if (fromnd.ni_vp->v_type == VDIR)
3636 		tond.ni_cnd.cn_flags |= WILLBEDIR;
3637 	if ((error = namei(&tond)) != 0) {
3638 		/* Translate error code for rename("dir1", "dir2/."). */
3639 		if (error == EISDIR && fvp->v_type == VDIR)
3640 			error = EINVAL;
3641 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3642 		vrele(fromnd.ni_dvp);
3643 		vrele(fvp);
3644 		goto out1;
3645 	}
3646 	tdvp = tond.ni_dvp;
3647 	tvp = tond.ni_vp;
3648 	error = vn_start_write(fvp, &mp, V_NOWAIT);
3649 	if (error != 0) {
3650 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3651 		NDFREE(&tond, NDF_ONLY_PNBUF);
3652 		if (tvp != NULL)
3653 			vput(tvp);
3654 		if (tdvp == tvp)
3655 			vrele(tdvp);
3656 		else
3657 			vput(tdvp);
3658 		vrele(fromnd.ni_dvp);
3659 		vrele(fvp);
3660 		vrele(tond.ni_startdir);
3661 		if (fromnd.ni_startdir != NULL)
3662 			vrele(fromnd.ni_startdir);
3663 		error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH);
3664 		if (error != 0)
3665 			return (error);
3666 		goto again;
3667 	}
3668 	if (tvp != NULL) {
3669 		if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
3670 			error = ENOTDIR;
3671 			goto out;
3672 		} else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
3673 			error = EISDIR;
3674 			goto out;
3675 		}
3676 #ifdef CAPABILITIES
3677 		if (newfd != AT_FDCWD && (tond.ni_resflags & NIRES_ABS) == 0) {
3678 			/*
3679 			 * If the target already exists we require CAP_UNLINKAT
3680 			 * from 'newfd', when newfd was used for the lookup.
3681 			 */
3682 			error = cap_check(&tond.ni_filecaps.fc_rights,
3683 			    &cap_unlinkat_rights);
3684 			if (error != 0)
3685 				goto out;
3686 		}
3687 #endif
3688 	}
3689 	if (fvp == tdvp) {
3690 		error = EINVAL;
3691 		goto out;
3692 	}
3693 	/*
3694 	 * If the source is the same as the destination (that is, if they
3695 	 * are links to the same vnode), then there is nothing to do.
3696 	 */
3697 	if (fvp == tvp)
3698 		error = ERESTART;
3699 #ifdef MAC
3700 	else
3701 		error = mac_vnode_check_rename_to(td->td_ucred, tdvp,
3702 		    tond.ni_vp, fromnd.ni_dvp == tdvp, &tond.ni_cnd);
3703 #endif
3704 out:
3705 	if (error == 0) {
3706 		error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
3707 		    tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
3708 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3709 		NDFREE(&tond, NDF_ONLY_PNBUF);
3710 	} else {
3711 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3712 		NDFREE(&tond, NDF_ONLY_PNBUF);
3713 		if (tvp != NULL)
3714 			vput(tvp);
3715 		if (tdvp == tvp)
3716 			vrele(tdvp);
3717 		else
3718 			vput(tdvp);
3719 		vrele(fromnd.ni_dvp);
3720 		vrele(fvp);
3721 	}
3722 	vrele(tond.ni_startdir);
3723 	vn_finished_write(mp);
3724 out1:
3725 	if (fromnd.ni_startdir)
3726 		vrele(fromnd.ni_startdir);
3727 	if (error == ERESTART)
3728 		return (0);
3729 	if (error == ERELOOKUP)
3730 		goto again;
3731 	return (error);
3732 }
3733 
3734 /*
3735  * Make a directory file.
3736  */
3737 #ifndef _SYS_SYSPROTO_H_
3738 struct mkdir_args {
3739 	char	*path;
3740 	int	mode;
3741 };
3742 #endif
3743 int
3744 sys_mkdir(struct thread *td, struct mkdir_args *uap)
3745 {
3746 
3747 	return (kern_mkdirat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
3748 	    uap->mode));
3749 }
3750 
3751 #ifndef _SYS_SYSPROTO_H_
3752 struct mkdirat_args {
3753 	int	fd;
3754 	char	*path;
3755 	mode_t	mode;
3756 };
3757 #endif
3758 int
3759 sys_mkdirat(struct thread *td, struct mkdirat_args *uap)
3760 {
3761 
3762 	return (kern_mkdirat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode));
3763 }
3764 
3765 int
3766 kern_mkdirat(struct thread *td, int fd, const char *path, enum uio_seg segflg,
3767     int mode)
3768 {
3769 	struct mount *mp;
3770 	struct vnode *vp;
3771 	struct vattr vattr;
3772 	struct nameidata nd;
3773 	int error;
3774 
3775 	AUDIT_ARG_MODE(mode);
3776 restart:
3777 	bwillwrite();
3778 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
3779 	    NC_NOMAKEENTRY | NC_KEEPPOSENTRY, segflg, path, fd,
3780 	    &cap_mkdirat_rights, td);
3781 	nd.ni_cnd.cn_flags |= WILLBEDIR;
3782 	if ((error = namei(&nd)) != 0)
3783 		return (error);
3784 	vp = nd.ni_vp;
3785 	if (vp != NULL) {
3786 		NDFREE(&nd, NDF_ONLY_PNBUF);
3787 		/*
3788 		 * XXX namei called with LOCKPARENT but not LOCKLEAF has
3789 		 * the strange behaviour of leaving the vnode unlocked
3790 		 * if the target is the same vnode as the parent.
3791 		 */
3792 		if (vp == nd.ni_dvp)
3793 			vrele(nd.ni_dvp);
3794 		else
3795 			vput(nd.ni_dvp);
3796 		vrele(vp);
3797 		return (EEXIST);
3798 	}
3799 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3800 		NDFREE(&nd, NDF_ONLY_PNBUF);
3801 		vput(nd.ni_dvp);
3802 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3803 			return (error);
3804 		goto restart;
3805 	}
3806 	VATTR_NULL(&vattr);
3807 	vattr.va_type = VDIR;
3808 	vattr.va_mode = (mode & ACCESSPERMS) &~ td->td_proc->p_fd->fd_cmask;
3809 #ifdef MAC
3810 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
3811 	    &vattr);
3812 	if (error != 0)
3813 		goto out;
3814 #endif
3815 	error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
3816 #ifdef MAC
3817 out:
3818 #endif
3819 	NDFREE(&nd, NDF_ONLY_PNBUF);
3820 	vput(nd.ni_dvp);
3821 	if (error == 0)
3822 		vput(nd.ni_vp);
3823 	vn_finished_write(mp);
3824 	if (error == ERELOOKUP)
3825 		goto restart;
3826 	return (error);
3827 }
3828 
3829 /*
3830  * Remove a directory file.
3831  */
3832 #ifndef _SYS_SYSPROTO_H_
3833 struct rmdir_args {
3834 	char	*path;
3835 };
3836 #endif
3837 int
3838 sys_rmdir(struct thread *td, struct rmdir_args *uap)
3839 {
3840 
3841 	return (kern_frmdirat(td, AT_FDCWD, uap->path, FD_NONE, UIO_USERSPACE,
3842 	    0));
3843 }
3844 
3845 int
3846 kern_frmdirat(struct thread *td, int dfd, const char *path, int fd,
3847     enum uio_seg pathseg, int flag)
3848 {
3849 	struct mount *mp;
3850 	struct vnode *vp;
3851 	struct file *fp;
3852 	struct nameidata nd;
3853 	cap_rights_t rights;
3854 	int error;
3855 
3856 	fp = NULL;
3857 	if (fd != FD_NONE) {
3858 		error = getvnode(td, fd, cap_rights_init_one(&rights, CAP_LOOKUP),
3859 		    &fp);
3860 		if (error != 0)
3861 			return (error);
3862 	}
3863 
3864 restart:
3865 	bwillwrite();
3866 	NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1 |
3867 	    at2cnpflags(flag, AT_BENEATH | AT_RESOLVE_BENEATH),
3868 	    pathseg, path, dfd, &cap_unlinkat_rights, td);
3869 	if ((error = namei(&nd)) != 0)
3870 		goto fdout;
3871 	vp = nd.ni_vp;
3872 	if (vp->v_type != VDIR) {
3873 		error = ENOTDIR;
3874 		goto out;
3875 	}
3876 	/*
3877 	 * No rmdir "." please.
3878 	 */
3879 	if (nd.ni_dvp == vp) {
3880 		error = EINVAL;
3881 		goto out;
3882 	}
3883 	/*
3884 	 * The root of a mounted filesystem cannot be deleted.
3885 	 */
3886 	if (vp->v_vflag & VV_ROOT) {
3887 		error = EBUSY;
3888 		goto out;
3889 	}
3890 
3891 	if (fp != NULL && fp->f_vnode != vp) {
3892 		if (VN_IS_DOOMED(fp->f_vnode))
3893 			error = EBADF;
3894 		else
3895 			error = EDEADLK;
3896 		goto out;
3897 	}
3898 
3899 #ifdef MAC
3900 	error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
3901 	    &nd.ni_cnd);
3902 	if (error != 0)
3903 		goto out;
3904 #endif
3905 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3906 		NDFREE(&nd, NDF_ONLY_PNBUF);
3907 		vput(vp);
3908 		if (nd.ni_dvp == vp)
3909 			vrele(nd.ni_dvp);
3910 		else
3911 			vput(nd.ni_dvp);
3912 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3913 			goto fdout;
3914 		goto restart;
3915 	}
3916 	vfs_notify_upper(vp, VFS_NOTIFY_UPPER_UNLINK);
3917 	error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
3918 	vn_finished_write(mp);
3919 out:
3920 	NDFREE(&nd, NDF_ONLY_PNBUF);
3921 	vput(vp);
3922 	if (nd.ni_dvp == vp)
3923 		vrele(nd.ni_dvp);
3924 	else
3925 		vput(nd.ni_dvp);
3926 	if (error == ERELOOKUP)
3927 		goto restart;
3928 fdout:
3929 	if (fp != NULL)
3930 		fdrop(fp, td);
3931 	return (error);
3932 }
3933 
3934 #if defined(COMPAT_43) || defined(COMPAT_FREEBSD11)
3935 int
3936 freebsd11_kern_getdirentries(struct thread *td, int fd, char *ubuf, u_int count,
3937     long *basep, void (*func)(struct freebsd11_dirent *))
3938 {
3939 	struct freebsd11_dirent dstdp;
3940 	struct dirent *dp, *edp;
3941 	char *dirbuf;
3942 	off_t base;
3943 	ssize_t resid, ucount;
3944 	int error;
3945 
3946 	/* XXX arbitrary sanity limit on `count'. */
3947 	count = min(count, 64 * 1024);
3948 
3949 	dirbuf = malloc(count, M_TEMP, M_WAITOK);
3950 
3951 	error = kern_getdirentries(td, fd, dirbuf, count, &base, &resid,
3952 	    UIO_SYSSPACE);
3953 	if (error != 0)
3954 		goto done;
3955 	if (basep != NULL)
3956 		*basep = base;
3957 
3958 	ucount = 0;
3959 	for (dp = (struct dirent *)dirbuf,
3960 	    edp = (struct dirent *)&dirbuf[count - resid];
3961 	    ucount < count && dp < edp; ) {
3962 		if (dp->d_reclen == 0)
3963 			break;
3964 		MPASS(dp->d_reclen >= _GENERIC_DIRLEN(0));
3965 		if (dp->d_namlen >= sizeof(dstdp.d_name))
3966 			continue;
3967 		dstdp.d_type = dp->d_type;
3968 		dstdp.d_namlen = dp->d_namlen;
3969 		dstdp.d_fileno = dp->d_fileno;		/* truncate */
3970 		if (dstdp.d_fileno != dp->d_fileno) {
3971 			switch (ino64_trunc_error) {
3972 			default:
3973 			case 0:
3974 				break;
3975 			case 1:
3976 				error = EOVERFLOW;
3977 				goto done;
3978 			case 2:
3979 				dstdp.d_fileno = UINT32_MAX;
3980 				break;
3981 			}
3982 		}
3983 		dstdp.d_reclen = sizeof(dstdp) - sizeof(dstdp.d_name) +
3984 		    ((dp->d_namlen + 1 + 3) &~ 3);
3985 		bcopy(dp->d_name, dstdp.d_name, dstdp.d_namlen);
3986 		bzero(dstdp.d_name + dstdp.d_namlen,
3987 		    dstdp.d_reclen - offsetof(struct freebsd11_dirent, d_name) -
3988 		    dstdp.d_namlen);
3989 		MPASS(dstdp.d_reclen <= dp->d_reclen);
3990 		MPASS(ucount + dstdp.d_reclen <= count);
3991 		if (func != NULL)
3992 			func(&dstdp);
3993 		error = copyout(&dstdp, ubuf + ucount, dstdp.d_reclen);
3994 		if (error != 0)
3995 			break;
3996 		dp = (struct dirent *)((char *)dp + dp->d_reclen);
3997 		ucount += dstdp.d_reclen;
3998 	}
3999 
4000 done:
4001 	free(dirbuf, M_TEMP);
4002 	if (error == 0)
4003 		td->td_retval[0] = ucount;
4004 	return (error);
4005 }
4006 #endif /* COMPAT */
4007 
4008 #ifdef COMPAT_43
4009 static void
4010 ogetdirentries_cvt(struct freebsd11_dirent *dp)
4011 {
4012 #if (BYTE_ORDER == LITTLE_ENDIAN)
4013 	/*
4014 	 * The expected low byte of dp->d_namlen is our dp->d_type.
4015 	 * The high MBZ byte of dp->d_namlen is our dp->d_namlen.
4016 	 */
4017 	dp->d_type = dp->d_namlen;
4018 	dp->d_namlen = 0;
4019 #else
4020 	/*
4021 	 * The dp->d_type is the high byte of the expected dp->d_namlen,
4022 	 * so must be zero'ed.
4023 	 */
4024 	dp->d_type = 0;
4025 #endif
4026 }
4027 
4028 /*
4029  * Read a block of directory entries in a filesystem independent format.
4030  */
4031 #ifndef _SYS_SYSPROTO_H_
4032 struct ogetdirentries_args {
4033 	int	fd;
4034 	char	*buf;
4035 	u_int	count;
4036 	long	*basep;
4037 };
4038 #endif
4039 int
4040 ogetdirentries(struct thread *td, struct ogetdirentries_args *uap)
4041 {
4042 	long loff;
4043 	int error;
4044 
4045 	error = kern_ogetdirentries(td, uap, &loff);
4046 	if (error == 0)
4047 		error = copyout(&loff, uap->basep, sizeof(long));
4048 	return (error);
4049 }
4050 
4051 int
4052 kern_ogetdirentries(struct thread *td, struct ogetdirentries_args *uap,
4053     long *ploff)
4054 {
4055 	long base;
4056 	int error;
4057 
4058 	/* XXX arbitrary sanity limit on `count'. */
4059 	if (uap->count > 64 * 1024)
4060 		return (EINVAL);
4061 
4062 	error = freebsd11_kern_getdirentries(td, uap->fd, uap->buf, uap->count,
4063 	    &base, ogetdirentries_cvt);
4064 
4065 	if (error == 0 && uap->basep != NULL)
4066 		error = copyout(&base, uap->basep, sizeof(long));
4067 
4068 	return (error);
4069 }
4070 #endif /* COMPAT_43 */
4071 
4072 #if defined(COMPAT_FREEBSD11)
4073 #ifndef _SYS_SYSPROTO_H_
4074 struct freebsd11_getdirentries_args {
4075 	int	fd;
4076 	char	*buf;
4077 	u_int	count;
4078 	long	*basep;
4079 };
4080 #endif
4081 int
4082 freebsd11_getdirentries(struct thread *td,
4083     struct freebsd11_getdirentries_args *uap)
4084 {
4085 	long base;
4086 	int error;
4087 
4088 	error = freebsd11_kern_getdirentries(td, uap->fd, uap->buf, uap->count,
4089 	    &base, NULL);
4090 
4091 	if (error == 0 && uap->basep != NULL)
4092 		error = copyout(&base, uap->basep, sizeof(long));
4093 	return (error);
4094 }
4095 
4096 int
4097 freebsd11_getdents(struct thread *td, struct freebsd11_getdents_args *uap)
4098 {
4099 	struct freebsd11_getdirentries_args ap;
4100 
4101 	ap.fd = uap->fd;
4102 	ap.buf = uap->buf;
4103 	ap.count = uap->count;
4104 	ap.basep = NULL;
4105 	return (freebsd11_getdirentries(td, &ap));
4106 }
4107 #endif /* COMPAT_FREEBSD11 */
4108 
4109 /*
4110  * Read a block of directory entries in a filesystem independent format.
4111  */
4112 int
4113 sys_getdirentries(struct thread *td, struct getdirentries_args *uap)
4114 {
4115 	off_t base;
4116 	int error;
4117 
4118 	error = kern_getdirentries(td, uap->fd, uap->buf, uap->count, &base,
4119 	    NULL, UIO_USERSPACE);
4120 	if (error != 0)
4121 		return (error);
4122 	if (uap->basep != NULL)
4123 		error = copyout(&base, uap->basep, sizeof(off_t));
4124 	return (error);
4125 }
4126 
4127 int
4128 kern_getdirentries(struct thread *td, int fd, char *buf, size_t count,
4129     off_t *basep, ssize_t *residp, enum uio_seg bufseg)
4130 {
4131 	struct vnode *vp;
4132 	struct file *fp;
4133 	struct uio auio;
4134 	struct iovec aiov;
4135 	off_t loff;
4136 	int error, eofflag;
4137 	off_t foffset;
4138 
4139 	AUDIT_ARG_FD(fd);
4140 	if (count > IOSIZE_MAX)
4141 		return (EINVAL);
4142 	auio.uio_resid = count;
4143 	error = getvnode(td, fd, &cap_read_rights, &fp);
4144 	if (error != 0)
4145 		return (error);
4146 	if ((fp->f_flag & FREAD) == 0) {
4147 		fdrop(fp, td);
4148 		return (EBADF);
4149 	}
4150 	vp = fp->f_vnode;
4151 	foffset = foffset_lock(fp, 0);
4152 unionread:
4153 	if (vp->v_type != VDIR) {
4154 		error = EINVAL;
4155 		goto fail;
4156 	}
4157 	aiov.iov_base = buf;
4158 	aiov.iov_len = count;
4159 	auio.uio_iov = &aiov;
4160 	auio.uio_iovcnt = 1;
4161 	auio.uio_rw = UIO_READ;
4162 	auio.uio_segflg = bufseg;
4163 	auio.uio_td = td;
4164 	vn_lock(vp, LK_SHARED | LK_RETRY);
4165 	AUDIT_ARG_VNODE1(vp);
4166 	loff = auio.uio_offset = foffset;
4167 #ifdef MAC
4168 	error = mac_vnode_check_readdir(td->td_ucred, vp);
4169 	if (error == 0)
4170 #endif
4171 		error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, NULL,
4172 		    NULL);
4173 	foffset = auio.uio_offset;
4174 	if (error != 0) {
4175 		VOP_UNLOCK(vp);
4176 		goto fail;
4177 	}
4178 	if (count == auio.uio_resid &&
4179 	    (vp->v_vflag & VV_ROOT) &&
4180 	    (vp->v_mount->mnt_flag & MNT_UNION)) {
4181 		struct vnode *tvp = vp;
4182 
4183 		vp = vp->v_mount->mnt_vnodecovered;
4184 		VREF(vp);
4185 		fp->f_vnode = vp;
4186 		foffset = 0;
4187 		vput(tvp);
4188 		goto unionread;
4189 	}
4190 	VOP_UNLOCK(vp);
4191 	*basep = loff;
4192 	if (residp != NULL)
4193 		*residp = auio.uio_resid;
4194 	td->td_retval[0] = count - auio.uio_resid;
4195 fail:
4196 	foffset_unlock(fp, foffset, 0);
4197 	fdrop(fp, td);
4198 	return (error);
4199 }
4200 
4201 /*
4202  * Set the mode mask for creation of filesystem nodes.
4203  */
4204 #ifndef _SYS_SYSPROTO_H_
4205 struct umask_args {
4206 	int	newmask;
4207 };
4208 #endif
4209 int
4210 sys_umask(struct thread *td, struct umask_args *uap)
4211 {
4212 	struct filedesc *fdp;
4213 
4214 	fdp = td->td_proc->p_fd;
4215 	FILEDESC_XLOCK(fdp);
4216 	td->td_retval[0] = fdp->fd_cmask;
4217 	fdp->fd_cmask = uap->newmask & ALLPERMS;
4218 	FILEDESC_XUNLOCK(fdp);
4219 	return (0);
4220 }
4221 
4222 /*
4223  * Void all references to file by ripping underlying filesystem away from
4224  * vnode.
4225  */
4226 #ifndef _SYS_SYSPROTO_H_
4227 struct revoke_args {
4228 	char	*path;
4229 };
4230 #endif
4231 int
4232 sys_revoke(struct thread *td, struct revoke_args *uap)
4233 {
4234 	struct vnode *vp;
4235 	struct vattr vattr;
4236 	struct nameidata nd;
4237 	int error;
4238 
4239 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
4240 	    uap->path, td);
4241 	if ((error = namei(&nd)) != 0)
4242 		return (error);
4243 	vp = nd.ni_vp;
4244 	NDFREE(&nd, NDF_ONLY_PNBUF);
4245 	if (vp->v_type != VCHR || vp->v_rdev == NULL) {
4246 		error = EINVAL;
4247 		goto out;
4248 	}
4249 #ifdef MAC
4250 	error = mac_vnode_check_revoke(td->td_ucred, vp);
4251 	if (error != 0)
4252 		goto out;
4253 #endif
4254 	error = VOP_GETATTR(vp, &vattr, td->td_ucred);
4255 	if (error != 0)
4256 		goto out;
4257 	if (td->td_ucred->cr_uid != vattr.va_uid) {
4258 		error = priv_check(td, PRIV_VFS_ADMIN);
4259 		if (error != 0)
4260 			goto out;
4261 	}
4262 	if (devfs_usecount(vp) > 0)
4263 		VOP_REVOKE(vp, REVOKEALL);
4264 out:
4265 	vput(vp);
4266 	return (error);
4267 }
4268 
4269 /*
4270  * Convert a user file descriptor to a kernel file entry and check that, if it
4271  * is a capability, the correct rights are present. A reference on the file
4272  * entry is held upon returning.
4273  */
4274 int
4275 getvnode(struct thread *td, int fd, cap_rights_t *rightsp, struct file **fpp)
4276 {
4277 	struct file *fp;
4278 	int error;
4279 
4280 	error = fget_unlocked(td->td_proc->p_fd, fd, rightsp, &fp);
4281 	if (error != 0)
4282 		return (error);
4283 
4284 	/*
4285 	 * The file could be not of the vnode type, or it may be not
4286 	 * yet fully initialized, in which case the f_vnode pointer
4287 	 * may be set, but f_ops is still badfileops.  E.g.,
4288 	 * devfs_open() transiently create such situation to
4289 	 * facilitate csw d_fdopen().
4290 	 *
4291 	 * Dupfdopen() handling in kern_openat() installs the
4292 	 * half-baked file into the process descriptor table, allowing
4293 	 * other thread to dereference it. Guard against the race by
4294 	 * checking f_ops.
4295 	 */
4296 	if (fp->f_vnode == NULL || fp->f_ops == &badfileops) {
4297 		fdrop(fp, td);
4298 		return (EINVAL);
4299 	}
4300 	*fpp = fp;
4301 	return (0);
4302 }
4303 
4304 /*
4305  * Get an (NFS) file handle.
4306  */
4307 #ifndef _SYS_SYSPROTO_H_
4308 struct lgetfh_args {
4309 	char *fname;
4310 	fhandle_t *fhp;
4311 };
4312 #endif
4313 int
4314 sys_lgetfh(struct thread *td, struct lgetfh_args *uap)
4315 {
4316 
4317 	return (kern_getfhat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->fname,
4318 	    UIO_USERSPACE, uap->fhp));
4319 }
4320 
4321 #ifndef _SYS_SYSPROTO_H_
4322 struct getfh_args {
4323 	char *fname;
4324 	fhandle_t *fhp;
4325 };
4326 #endif
4327 int
4328 sys_getfh(struct thread *td, struct getfh_args *uap)
4329 {
4330 
4331 	return (kern_getfhat(td, 0, AT_FDCWD, uap->fname, UIO_USERSPACE,
4332 	    uap->fhp));
4333 }
4334 
4335 /*
4336  * syscall for the rpc.lockd to use to translate an open descriptor into
4337  * a NFS file handle.
4338  *
4339  * warning: do not remove the priv_check() call or this becomes one giant
4340  * security hole.
4341  */
4342 #ifndef _SYS_SYSPROTO_H_
4343 struct getfhat_args {
4344 	int fd;
4345 	char *path;
4346 	fhandle_t *fhp;
4347 	int flags;
4348 };
4349 #endif
4350 int
4351 sys_getfhat(struct thread *td, struct getfhat_args *uap)
4352 {
4353 
4354 	if ((uap->flags & ~(AT_SYMLINK_NOFOLLOW | AT_BENEATH |
4355 	    AT_RESOLVE_BENEATH)) != 0)
4356 		return (EINVAL);
4357 	return (kern_getfhat(td, uap->flags, uap->fd, uap->path, UIO_USERSPACE,
4358 	    uap->fhp));
4359 }
4360 
4361 static int
4362 kern_getfhat(struct thread *td, int flags, int fd, const char *path,
4363     enum uio_seg pathseg, fhandle_t *fhp)
4364 {
4365 	struct nameidata nd;
4366 	fhandle_t fh;
4367 	struct vnode *vp;
4368 	int error;
4369 
4370 	error = priv_check(td, PRIV_VFS_GETFH);
4371 	if (error != 0)
4372 		return (error);
4373 	NDINIT_AT(&nd, LOOKUP, at2cnpflags(flags, AT_SYMLINK_NOFOLLOW |
4374 	    AT_BENEATH | AT_RESOLVE_BENEATH) | LOCKLEAF | AUDITVNODE1,
4375 	    pathseg, path, fd, td);
4376 	error = namei(&nd);
4377 	if (error != 0)
4378 		return (error);
4379 	NDFREE(&nd, NDF_ONLY_PNBUF);
4380 	vp = nd.ni_vp;
4381 	bzero(&fh, sizeof(fh));
4382 	fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
4383 	error = VOP_VPTOFH(vp, &fh.fh_fid);
4384 	vput(vp);
4385 	if (error == 0)
4386 		error = copyout(&fh, fhp, sizeof (fh));
4387 	return (error);
4388 }
4389 
4390 #ifndef _SYS_SYSPROTO_H_
4391 struct fhlink_args {
4392 	fhandle_t *fhp;
4393 	const char *to;
4394 };
4395 #endif
4396 int
4397 sys_fhlink(struct thread *td, struct fhlink_args *uap)
4398 {
4399 
4400 	return (kern_fhlinkat(td, AT_FDCWD, uap->to, UIO_USERSPACE, uap->fhp));
4401 }
4402 
4403 #ifndef _SYS_SYSPROTO_H_
4404 struct fhlinkat_args {
4405 	fhandle_t *fhp;
4406 	int tofd;
4407 	const char *to;
4408 };
4409 #endif
4410 int
4411 sys_fhlinkat(struct thread *td, struct fhlinkat_args *uap)
4412 {
4413 
4414 	return (kern_fhlinkat(td, uap->tofd, uap->to, UIO_USERSPACE, uap->fhp));
4415 }
4416 
4417 static int
4418 kern_fhlinkat(struct thread *td, int fd, const char *path,
4419     enum uio_seg pathseg, fhandle_t *fhp)
4420 {
4421 	fhandle_t fh;
4422 	struct mount *mp;
4423 	struct vnode *vp;
4424 	int error;
4425 
4426 	error = priv_check(td, PRIV_VFS_GETFH);
4427 	if (error != 0)
4428 		return (error);
4429 	error = copyin(fhp, &fh, sizeof(fh));
4430 	if (error != 0)
4431 		return (error);
4432 	do {
4433 		bwillwrite();
4434 		if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4435 			return (ESTALE);
4436 		error = VFS_FHTOVP(mp, &fh.fh_fid, LK_SHARED, &vp);
4437 		vfs_unbusy(mp);
4438 		if (error != 0)
4439 			return (error);
4440 		VOP_UNLOCK(vp);
4441 		error = kern_linkat_vp(td, vp, fd, path, pathseg);
4442 	} while (error == EAGAIN || error == ERELOOKUP);
4443 	return (error);
4444 }
4445 
4446 #ifndef _SYS_SYSPROTO_H_
4447 struct fhreadlink_args {
4448 	fhandle_t *fhp;
4449 	char *buf;
4450 	size_t bufsize;
4451 };
4452 #endif
4453 int
4454 sys_fhreadlink(struct thread *td, struct fhreadlink_args *uap)
4455 {
4456 	fhandle_t fh;
4457 	struct mount *mp;
4458 	struct vnode *vp;
4459 	int error;
4460 
4461 	error = priv_check(td, PRIV_VFS_GETFH);
4462 	if (error != 0)
4463 		return (error);
4464 	if (uap->bufsize > IOSIZE_MAX)
4465 		return (EINVAL);
4466 	error = copyin(uap->fhp, &fh, sizeof(fh));
4467 	if (error != 0)
4468 		return (error);
4469 	if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4470 		return (ESTALE);
4471 	error = VFS_FHTOVP(mp, &fh.fh_fid, LK_SHARED, &vp);
4472 	vfs_unbusy(mp);
4473 	if (error != 0)
4474 		return (error);
4475 	error = kern_readlink_vp(vp, uap->buf, UIO_USERSPACE, uap->bufsize, td);
4476 	vput(vp);
4477 	return (error);
4478 }
4479 
4480 /*
4481  * syscall for the rpc.lockd to use to translate a NFS file handle into an
4482  * open descriptor.
4483  *
4484  * warning: do not remove the priv_check() call or this becomes one giant
4485  * security hole.
4486  */
4487 #ifndef _SYS_SYSPROTO_H_
4488 struct fhopen_args {
4489 	const struct fhandle *u_fhp;
4490 	int flags;
4491 };
4492 #endif
4493 int
4494 sys_fhopen(struct thread *td, struct fhopen_args *uap)
4495 {
4496 	struct mount *mp;
4497 	struct vnode *vp;
4498 	struct fhandle fhp;
4499 	struct file *fp;
4500 	int fmode, error;
4501 	int indx;
4502 
4503 	error = priv_check(td, PRIV_VFS_FHOPEN);
4504 	if (error != 0)
4505 		return (error);
4506 	indx = -1;
4507 	fmode = FFLAGS(uap->flags);
4508 	/* why not allow a non-read/write open for our lockd? */
4509 	if (((fmode & (FREAD | FWRITE)) == 0) || (fmode & O_CREAT))
4510 		return (EINVAL);
4511 	error = copyin(uap->u_fhp, &fhp, sizeof(fhp));
4512 	if (error != 0)
4513 		return(error);
4514 	/* find the mount point */
4515 	mp = vfs_busyfs(&fhp.fh_fsid);
4516 	if (mp == NULL)
4517 		return (ESTALE);
4518 	/* now give me my vnode, it gets returned to me locked */
4519 	error = VFS_FHTOVP(mp, &fhp.fh_fid, LK_EXCLUSIVE, &vp);
4520 	vfs_unbusy(mp);
4521 	if (error != 0)
4522 		return (error);
4523 
4524 	error = falloc_noinstall(td, &fp);
4525 	if (error != 0) {
4526 		vput(vp);
4527 		return (error);
4528 	}
4529 	/*
4530 	 * An extra reference on `fp' has been held for us by
4531 	 * falloc_noinstall().
4532 	 */
4533 
4534 #ifdef INVARIANTS
4535 	td->td_dupfd = -1;
4536 #endif
4537 	error = vn_open_vnode(vp, fmode, td->td_ucred, td, fp);
4538 	if (error != 0) {
4539 		KASSERT(fp->f_ops == &badfileops,
4540 		    ("VOP_OPEN in fhopen() set f_ops"));
4541 		KASSERT(td->td_dupfd < 0,
4542 		    ("fhopen() encountered fdopen()"));
4543 
4544 		vput(vp);
4545 		goto bad;
4546 	}
4547 #ifdef INVARIANTS
4548 	td->td_dupfd = 0;
4549 #endif
4550 	fp->f_vnode = vp;
4551 	finit_vnode(fp, fmode, NULL, &vnops);
4552 	VOP_UNLOCK(vp);
4553 	if ((fmode & O_TRUNC) != 0) {
4554 		error = fo_truncate(fp, 0, td->td_ucred, td);
4555 		if (error != 0)
4556 			goto bad;
4557 	}
4558 
4559 	error = finstall(td, fp, &indx, fmode, NULL);
4560 bad:
4561 	fdrop(fp, td);
4562 	td->td_retval[0] = indx;
4563 	return (error);
4564 }
4565 
4566 /*
4567  * Stat an (NFS) file handle.
4568  */
4569 #ifndef _SYS_SYSPROTO_H_
4570 struct fhstat_args {
4571 	struct fhandle *u_fhp;
4572 	struct stat *sb;
4573 };
4574 #endif
4575 int
4576 sys_fhstat(struct thread *td, struct fhstat_args *uap)
4577 {
4578 	struct stat sb;
4579 	struct fhandle fh;
4580 	int error;
4581 
4582 	error = copyin(uap->u_fhp, &fh, sizeof(fh));
4583 	if (error != 0)
4584 		return (error);
4585 	error = kern_fhstat(td, fh, &sb);
4586 	if (error == 0)
4587 		error = copyout(&sb, uap->sb, sizeof(sb));
4588 	return (error);
4589 }
4590 
4591 int
4592 kern_fhstat(struct thread *td, struct fhandle fh, struct stat *sb)
4593 {
4594 	struct mount *mp;
4595 	struct vnode *vp;
4596 	int error;
4597 
4598 	error = priv_check(td, PRIV_VFS_FHSTAT);
4599 	if (error != 0)
4600 		return (error);
4601 	if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4602 		return (ESTALE);
4603 	error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4604 	vfs_unbusy(mp);
4605 	if (error != 0)
4606 		return (error);
4607 	error = VOP_STAT(vp, sb, td->td_ucred, NOCRED, td);
4608 	vput(vp);
4609 	return (error);
4610 }
4611 
4612 /*
4613  * Implement fstatfs() for (NFS) file handles.
4614  */
4615 #ifndef _SYS_SYSPROTO_H_
4616 struct fhstatfs_args {
4617 	struct fhandle *u_fhp;
4618 	struct statfs *buf;
4619 };
4620 #endif
4621 int
4622 sys_fhstatfs(struct thread *td, struct fhstatfs_args *uap)
4623 {
4624 	struct statfs *sfp;
4625 	fhandle_t fh;
4626 	int error;
4627 
4628 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
4629 	if (error != 0)
4630 		return (error);
4631 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
4632 	error = kern_fhstatfs(td, fh, sfp);
4633 	if (error == 0)
4634 		error = copyout(sfp, uap->buf, sizeof(*sfp));
4635 	free(sfp, M_STATFS);
4636 	return (error);
4637 }
4638 
4639 int
4640 kern_fhstatfs(struct thread *td, fhandle_t fh, struct statfs *buf)
4641 {
4642 	struct mount *mp;
4643 	struct vnode *vp;
4644 	int error;
4645 
4646 	error = priv_check(td, PRIV_VFS_FHSTATFS);
4647 	if (error != 0)
4648 		return (error);
4649 	if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4650 		return (ESTALE);
4651 	error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4652 	if (error != 0) {
4653 		vfs_unbusy(mp);
4654 		return (error);
4655 	}
4656 	vput(vp);
4657 	error = prison_canseemount(td->td_ucred, mp);
4658 	if (error != 0)
4659 		goto out;
4660 #ifdef MAC
4661 	error = mac_mount_check_stat(td->td_ucred, mp);
4662 	if (error != 0)
4663 		goto out;
4664 #endif
4665 	error = VFS_STATFS(mp, buf);
4666 out:
4667 	vfs_unbusy(mp);
4668 	return (error);
4669 }
4670 
4671 /*
4672  * Unlike madvise(2), we do not make a best effort to remember every
4673  * possible caching hint.  Instead, we remember the last setting with
4674  * the exception that we will allow POSIX_FADV_NORMAL to adjust the
4675  * region of any current setting.
4676  */
4677 int
4678 kern_posix_fadvise(struct thread *td, int fd, off_t offset, off_t len,
4679     int advice)
4680 {
4681 	struct fadvise_info *fa, *new;
4682 	struct file *fp;
4683 	struct vnode *vp;
4684 	off_t end;
4685 	int error;
4686 
4687 	if (offset < 0 || len < 0 || offset > OFF_MAX - len)
4688 		return (EINVAL);
4689 	AUDIT_ARG_VALUE(advice);
4690 	switch (advice) {
4691 	case POSIX_FADV_SEQUENTIAL:
4692 	case POSIX_FADV_RANDOM:
4693 	case POSIX_FADV_NOREUSE:
4694 		new = malloc(sizeof(*fa), M_FADVISE, M_WAITOK);
4695 		break;
4696 	case POSIX_FADV_NORMAL:
4697 	case POSIX_FADV_WILLNEED:
4698 	case POSIX_FADV_DONTNEED:
4699 		new = NULL;
4700 		break;
4701 	default:
4702 		return (EINVAL);
4703 	}
4704 	/* XXX: CAP_POSIX_FADVISE? */
4705 	AUDIT_ARG_FD(fd);
4706 	error = fget(td, fd, &cap_no_rights, &fp);
4707 	if (error != 0)
4708 		goto out;
4709 	AUDIT_ARG_FILE(td->td_proc, fp);
4710 	if ((fp->f_ops->fo_flags & DFLAG_SEEKABLE) == 0) {
4711 		error = ESPIPE;
4712 		goto out;
4713 	}
4714 	if (fp->f_type != DTYPE_VNODE) {
4715 		error = ENODEV;
4716 		goto out;
4717 	}
4718 	vp = fp->f_vnode;
4719 	if (vp->v_type != VREG) {
4720 		error = ENODEV;
4721 		goto out;
4722 	}
4723 	if (len == 0)
4724 		end = OFF_MAX;
4725 	else
4726 		end = offset + len - 1;
4727 	switch (advice) {
4728 	case POSIX_FADV_SEQUENTIAL:
4729 	case POSIX_FADV_RANDOM:
4730 	case POSIX_FADV_NOREUSE:
4731 		/*
4732 		 * Try to merge any existing non-standard region with
4733 		 * this new region if possible, otherwise create a new
4734 		 * non-standard region for this request.
4735 		 */
4736 		mtx_pool_lock(mtxpool_sleep, fp);
4737 		fa = fp->f_advice;
4738 		if (fa != NULL && fa->fa_advice == advice &&
4739 		    ((fa->fa_start <= end && fa->fa_end >= offset) ||
4740 		    (end != OFF_MAX && fa->fa_start == end + 1) ||
4741 		    (fa->fa_end != OFF_MAX && fa->fa_end + 1 == offset))) {
4742 			if (offset < fa->fa_start)
4743 				fa->fa_start = offset;
4744 			if (end > fa->fa_end)
4745 				fa->fa_end = end;
4746 		} else {
4747 			new->fa_advice = advice;
4748 			new->fa_start = offset;
4749 			new->fa_end = end;
4750 			fp->f_advice = new;
4751 			new = fa;
4752 		}
4753 		mtx_pool_unlock(mtxpool_sleep, fp);
4754 		break;
4755 	case POSIX_FADV_NORMAL:
4756 		/*
4757 		 * If a the "normal" region overlaps with an existing
4758 		 * non-standard region, trim or remove the
4759 		 * non-standard region.
4760 		 */
4761 		mtx_pool_lock(mtxpool_sleep, fp);
4762 		fa = fp->f_advice;
4763 		if (fa != NULL) {
4764 			if (offset <= fa->fa_start && end >= fa->fa_end) {
4765 				new = fa;
4766 				fp->f_advice = NULL;
4767 			} else if (offset <= fa->fa_start &&
4768 			    end >= fa->fa_start)
4769 				fa->fa_start = end + 1;
4770 			else if (offset <= fa->fa_end && end >= fa->fa_end)
4771 				fa->fa_end = offset - 1;
4772 			else if (offset >= fa->fa_start && end <= fa->fa_end) {
4773 				/*
4774 				 * If the "normal" region is a middle
4775 				 * portion of the existing
4776 				 * non-standard region, just remove
4777 				 * the whole thing rather than picking
4778 				 * one side or the other to
4779 				 * preserve.
4780 				 */
4781 				new = fa;
4782 				fp->f_advice = NULL;
4783 			}
4784 		}
4785 		mtx_pool_unlock(mtxpool_sleep, fp);
4786 		break;
4787 	case POSIX_FADV_WILLNEED:
4788 	case POSIX_FADV_DONTNEED:
4789 		error = VOP_ADVISE(vp, offset, end, advice);
4790 		break;
4791 	}
4792 out:
4793 	if (fp != NULL)
4794 		fdrop(fp, td);
4795 	free(new, M_FADVISE);
4796 	return (error);
4797 }
4798 
4799 int
4800 sys_posix_fadvise(struct thread *td, struct posix_fadvise_args *uap)
4801 {
4802 	int error;
4803 
4804 	error = kern_posix_fadvise(td, uap->fd, uap->offset, uap->len,
4805 	    uap->advice);
4806 	return (kern_posix_error(td, error));
4807 }
4808 
4809 int
4810 kern_copy_file_range(struct thread *td, int infd, off_t *inoffp, int outfd,
4811     off_t *outoffp, size_t len, unsigned int flags)
4812 {
4813 	struct file *infp, *outfp;
4814 	struct vnode *invp, *outvp;
4815 	int error;
4816 	size_t retlen;
4817 	void *rl_rcookie, *rl_wcookie;
4818 	off_t savinoff, savoutoff;
4819 
4820 	infp = outfp = NULL;
4821 	rl_rcookie = rl_wcookie = NULL;
4822 	savinoff = -1;
4823 	error = 0;
4824 	retlen = 0;
4825 
4826 	if (flags != 0) {
4827 		error = EINVAL;
4828 		goto out;
4829 	}
4830 	if (len > SSIZE_MAX)
4831 		/*
4832 		 * Although the len argument is size_t, the return argument
4833 		 * is ssize_t (which is signed).  Therefore a size that won't
4834 		 * fit in ssize_t can't be returned.
4835 		 */
4836 		len = SSIZE_MAX;
4837 
4838 	/* Get the file structures for the file descriptors. */
4839 	error = fget_read(td, infd, &cap_read_rights, &infp);
4840 	if (error != 0)
4841 		goto out;
4842 	if (infp->f_ops == &badfileops) {
4843 		error = EBADF;
4844 		goto out;
4845 	}
4846 	if (infp->f_vnode == NULL) {
4847 		error = EINVAL;
4848 		goto out;
4849 	}
4850 	error = fget_write(td, outfd, &cap_write_rights, &outfp);
4851 	if (error != 0)
4852 		goto out;
4853 	if (outfp->f_ops == &badfileops) {
4854 		error = EBADF;
4855 		goto out;
4856 	}
4857 	if (outfp->f_vnode == NULL) {
4858 		error = EINVAL;
4859 		goto out;
4860 	}
4861 
4862 	/* Set the offset pointers to the correct place. */
4863 	if (inoffp == NULL)
4864 		inoffp = &infp->f_offset;
4865 	if (outoffp == NULL)
4866 		outoffp = &outfp->f_offset;
4867 	savinoff = *inoffp;
4868 	savoutoff = *outoffp;
4869 
4870 	invp = infp->f_vnode;
4871 	outvp = outfp->f_vnode;
4872 	/* Sanity check the f_flag bits. */
4873 	if ((outfp->f_flag & (FWRITE | FAPPEND)) != FWRITE ||
4874 	    (infp->f_flag & FREAD) == 0) {
4875 		error = EBADF;
4876 		goto out;
4877 	}
4878 
4879 	/* If len == 0, just return 0. */
4880 	if (len == 0)
4881 		goto out;
4882 
4883 	/*
4884 	 * If infp and outfp refer to the same file, the byte ranges cannot
4885 	 * overlap.
4886 	 */
4887 	if (invp == outvp && ((savinoff <= savoutoff && savinoff + len >
4888 	    savoutoff) || (savinoff > savoutoff && savoutoff + len >
4889 	    savinoff))) {
4890 		error = EINVAL;
4891 		goto out;
4892 	}
4893 
4894 	/* Range lock the byte ranges for both invp and outvp. */
4895 	for (;;) {
4896 		rl_wcookie = vn_rangelock_wlock(outvp, *outoffp, *outoffp +
4897 		    len);
4898 		rl_rcookie = vn_rangelock_tryrlock(invp, *inoffp, *inoffp +
4899 		    len);
4900 		if (rl_rcookie != NULL)
4901 			break;
4902 		vn_rangelock_unlock(outvp, rl_wcookie);
4903 		rl_rcookie = vn_rangelock_rlock(invp, *inoffp, *inoffp + len);
4904 		vn_rangelock_unlock(invp, rl_rcookie);
4905 	}
4906 
4907 	retlen = len;
4908 	error = vn_copy_file_range(invp, inoffp, outvp, outoffp, &retlen,
4909 	    flags, infp->f_cred, outfp->f_cred, td);
4910 out:
4911 	if (rl_rcookie != NULL)
4912 		vn_rangelock_unlock(invp, rl_rcookie);
4913 	if (rl_wcookie != NULL)
4914 		vn_rangelock_unlock(outvp, rl_wcookie);
4915 	if (savinoff != -1 && (error == EINTR || error == ERESTART)) {
4916 		*inoffp = savinoff;
4917 		*outoffp = savoutoff;
4918 	}
4919 	if (outfp != NULL)
4920 		fdrop(outfp, td);
4921 	if (infp != NULL)
4922 		fdrop(infp, td);
4923 	td->td_retval[0] = retlen;
4924 	return (error);
4925 }
4926 
4927 int
4928 sys_copy_file_range(struct thread *td, struct copy_file_range_args *uap)
4929 {
4930 	off_t inoff, outoff, *inoffp, *outoffp;
4931 	int error;
4932 
4933 	inoffp = outoffp = NULL;
4934 	if (uap->inoffp != NULL) {
4935 		error = copyin(uap->inoffp, &inoff, sizeof(off_t));
4936 		if (error != 0)
4937 			return (error);
4938 		inoffp = &inoff;
4939 	}
4940 	if (uap->outoffp != NULL) {
4941 		error = copyin(uap->outoffp, &outoff, sizeof(off_t));
4942 		if (error != 0)
4943 			return (error);
4944 		outoffp = &outoff;
4945 	}
4946 	error = kern_copy_file_range(td, uap->infd, inoffp, uap->outfd,
4947 	    outoffp, uap->len, uap->flags);
4948 	if (error == 0 && uap->inoffp != NULL)
4949 		error = copyout(inoffp, uap->inoffp, sizeof(off_t));
4950 	if (error == 0 && uap->outoffp != NULL)
4951 		error = copyout(outoffp, uap->outoffp, sizeof(off_t));
4952 	return (error);
4953 }
4954