xref: /freebsd/sys/kern/vfs_syscalls.c (revision ac099daf6742ead81ea7ea86351a8ef4e783041b)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (c) 1989, 1993
5  *	The Regents of the University of California.  All rights reserved.
6  * (c) UNIX System Laboratories, Inc.
7  * All or some portions of this file are derived from material licensed
8  * to the University of California by American Telephone and Telegraph
9  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
10  * the permission of UNIX System Laboratories, Inc.
11  *
12  * Redistribution and use in source and binary forms, with or without
13  * modification, are permitted provided that the following conditions
14  * are met:
15  * 1. Redistributions of source code must retain the above copyright
16  *    notice, this list of conditions and the following disclaimer.
17  * 2. Redistributions in binary form must reproduce the above copyright
18  *    notice, this list of conditions and the following disclaimer in the
19  *    documentation and/or other materials provided with the distribution.
20  * 3. Neither the name of the University nor the names of its contributors
21  *    may be used to endorse or promote products derived from this software
22  *    without specific prior written permission.
23  *
24  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
25  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
28  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
29  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
30  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
31  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
32  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
33  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
34  * SUCH DAMAGE.
35  *
36  *	@(#)vfs_syscalls.c	8.13 (Berkeley) 4/15/94
37  */
38 
39 #include <sys/cdefs.h>
40 __FBSDID("$FreeBSD$");
41 
42 #include "opt_capsicum.h"
43 #include "opt_ktrace.h"
44 
45 #include <sys/param.h>
46 #include <sys/systm.h>
47 #include <sys/bio.h>
48 #include <sys/buf.h>
49 #include <sys/capsicum.h>
50 #include <sys/disk.h>
51 #include <sys/sysent.h>
52 #include <sys/malloc.h>
53 #include <sys/mount.h>
54 #include <sys/mutex.h>
55 #include <sys/sysproto.h>
56 #include <sys/namei.h>
57 #include <sys/filedesc.h>
58 #include <sys/kernel.h>
59 #include <sys/fcntl.h>
60 #include <sys/file.h>
61 #include <sys/filio.h>
62 #include <sys/limits.h>
63 #include <sys/linker.h>
64 #include <sys/rwlock.h>
65 #include <sys/sdt.h>
66 #include <sys/stat.h>
67 #include <sys/sx.h>
68 #include <sys/unistd.h>
69 #include <sys/vnode.h>
70 #include <sys/priv.h>
71 #include <sys/proc.h>
72 #include <sys/dirent.h>
73 #include <sys/jail.h>
74 #include <sys/syscallsubr.h>
75 #include <sys/sysctl.h>
76 #ifdef KTRACE
77 #include <sys/ktrace.h>
78 #endif
79 
80 #include <machine/stdarg.h>
81 
82 #include <security/audit/audit.h>
83 #include <security/mac/mac_framework.h>
84 
85 #include <vm/vm.h>
86 #include <vm/vm_object.h>
87 #include <vm/vm_page.h>
88 #include <vm/uma.h>
89 
90 #include <fs/devfs/devfs.h>
91 
92 #include <ufs/ufs/quota.h>
93 
94 MALLOC_DEFINE(M_FADVISE, "fadvise", "posix_fadvise(2) information");
95 
96 static int kern_chflagsat(struct thread *td, int fd, const char *path,
97     enum uio_seg pathseg, u_long flags, int atflag);
98 static int setfflags(struct thread *td, struct vnode *, u_long);
99 static int getutimes(const struct timeval *, enum uio_seg, struct timespec *);
100 static int getutimens(const struct timespec *, enum uio_seg,
101     struct timespec *, int *);
102 static int setutimes(struct thread *td, struct vnode *,
103     const struct timespec *, int, int);
104 static int vn_access(struct vnode *vp, int user_flags, struct ucred *cred,
105     struct thread *td);
106 static int kern_fhlinkat(struct thread *td, int fd, const char *path,
107     enum uio_seg pathseg, fhandle_t *fhp);
108 static int kern_readlink_vp(struct vnode *vp, char *buf, enum uio_seg bufseg,
109     size_t count, struct thread *td);
110 static int kern_linkat_vp(struct thread *td, struct vnode *vp, int fd,
111     const char *path, enum uio_seg segflag);
112 
113 static uint64_t
114 at2cnpflags(u_int at_flags, u_int mask)
115 {
116 	u_int64_t res;
117 
118 	MPASS((at_flags & (AT_SYMLINK_FOLLOW | AT_SYMLINK_NOFOLLOW)) !=
119 	    (AT_SYMLINK_FOLLOW | AT_SYMLINK_NOFOLLOW));
120 
121 	res = 0;
122 	at_flags &= mask;
123 	if ((at_flags & AT_RESOLVE_BENEATH) != 0)
124 		res |= RBENEATH;
125 	if ((at_flags & AT_SYMLINK_FOLLOW) != 0)
126 		res |= FOLLOW;
127 	/* NOFOLLOW is pseudo flag */
128 	if ((mask & AT_SYMLINK_NOFOLLOW) != 0) {
129 		res |= (at_flags & AT_SYMLINK_NOFOLLOW) != 0 ? NOFOLLOW :
130 		    FOLLOW;
131 	}
132 	if ((mask & AT_EMPTY_PATH) != 0 && (at_flags & AT_EMPTY_PATH) != 0)
133 		res |= EMPTYPATH;
134 	return (res);
135 }
136 
137 int
138 kern_sync(struct thread *td)
139 {
140 	struct mount *mp, *nmp;
141 	int save;
142 
143 	mtx_lock(&mountlist_mtx);
144 	for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
145 		if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK)) {
146 			nmp = TAILQ_NEXT(mp, mnt_list);
147 			continue;
148 		}
149 		if ((mp->mnt_flag & MNT_RDONLY) == 0 &&
150 		    vn_start_write(NULL, &mp, V_NOWAIT) == 0) {
151 			save = curthread_pflags_set(TDP_SYNCIO);
152 			vfs_periodic(mp, MNT_NOWAIT);
153 			VFS_SYNC(mp, MNT_NOWAIT);
154 			curthread_pflags_restore(save);
155 			vn_finished_write(mp);
156 		}
157 		mtx_lock(&mountlist_mtx);
158 		nmp = TAILQ_NEXT(mp, mnt_list);
159 		vfs_unbusy(mp);
160 	}
161 	mtx_unlock(&mountlist_mtx);
162 	return (0);
163 }
164 
165 /*
166  * Sync each mounted filesystem.
167  */
168 #ifndef _SYS_SYSPROTO_H_
169 struct sync_args {
170 	int     dummy;
171 };
172 #endif
173 /* ARGSUSED */
174 int
175 sys_sync(struct thread *td, struct sync_args *uap)
176 {
177 
178 	return (kern_sync(td));
179 }
180 
181 /*
182  * Change filesystem quotas.
183  */
184 #ifndef _SYS_SYSPROTO_H_
185 struct quotactl_args {
186 	char *path;
187 	int cmd;
188 	int uid;
189 	caddr_t arg;
190 };
191 #endif
192 int
193 sys_quotactl(struct thread *td, struct quotactl_args *uap)
194 {
195 	struct mount *mp;
196 	struct nameidata nd;
197 	int error;
198 
199 	AUDIT_ARG_CMD(uap->cmd);
200 	AUDIT_ARG_UID(uap->uid);
201 	if (!prison_allow(td->td_ucred, PR_ALLOW_QUOTAS))
202 		return (EPERM);
203 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
204 	    uap->path, td);
205 	if ((error = namei(&nd)) != 0)
206 		return (error);
207 	NDFREE(&nd, NDF_ONLY_PNBUF);
208 	mp = nd.ni_vp->v_mount;
209 	vfs_ref(mp);
210 	vput(nd.ni_vp);
211 	error = vfs_busy(mp, 0);
212 	if (error != 0) {
213 		vfs_rel(mp);
214 		return (error);
215 	}
216 	error = VFS_QUOTACTL(mp, uap->cmd, uap->uid, uap->arg);
217 
218 	/*
219 	 * Since quota on operation typically needs to open quota
220 	 * file, the Q_QUOTAON handler needs to unbusy the mount point
221 	 * before calling into namei.  Otherwise, unmount might be
222 	 * started between two vfs_busy() invocations (first is our,
223 	 * second is from mount point cross-walk code in lookup()),
224 	 * causing deadlock.
225 	 *
226 	 * Require that Q_QUOTAON handles the vfs_busy() reference on
227 	 * its own, always returning with ubusied mount point.
228 	 */
229 	if ((uap->cmd >> SUBCMDSHIFT) != Q_QUOTAON &&
230 	    (uap->cmd >> SUBCMDSHIFT) != Q_QUOTAOFF)
231 		vfs_unbusy(mp);
232 	vfs_rel(mp);
233 	return (error);
234 }
235 
236 /*
237  * Used by statfs conversion routines to scale the block size up if
238  * necessary so that all of the block counts are <= 'max_size'.  Note
239  * that 'max_size' should be a bitmask, i.e. 2^n - 1 for some non-zero
240  * value of 'n'.
241  */
242 void
243 statfs_scale_blocks(struct statfs *sf, long max_size)
244 {
245 	uint64_t count;
246 	int shift;
247 
248 	KASSERT(powerof2(max_size + 1), ("%s: invalid max_size", __func__));
249 
250 	/*
251 	 * Attempt to scale the block counts to give a more accurate
252 	 * overview to userland of the ratio of free space to used
253 	 * space.  To do this, find the largest block count and compute
254 	 * a divisor that lets it fit into a signed integer <= max_size.
255 	 */
256 	if (sf->f_bavail < 0)
257 		count = -sf->f_bavail;
258 	else
259 		count = sf->f_bavail;
260 	count = MAX(sf->f_blocks, MAX(sf->f_bfree, count));
261 	if (count <= max_size)
262 		return;
263 
264 	count >>= flsl(max_size);
265 	shift = 0;
266 	while (count > 0) {
267 		shift++;
268 		count >>=1;
269 	}
270 
271 	sf->f_bsize <<= shift;
272 	sf->f_blocks >>= shift;
273 	sf->f_bfree >>= shift;
274 	sf->f_bavail >>= shift;
275 }
276 
277 static int
278 kern_do_statfs(struct thread *td, struct mount *mp, struct statfs *buf)
279 {
280 	int error;
281 
282 	if (mp == NULL)
283 		return (EBADF);
284 	error = vfs_busy(mp, 0);
285 	vfs_rel(mp);
286 	if (error != 0)
287 		return (error);
288 #ifdef MAC
289 	error = mac_mount_check_stat(td->td_ucred, mp);
290 	if (error != 0)
291 		goto out;
292 #endif
293 	error = VFS_STATFS(mp, buf);
294 	if (error != 0)
295 		goto out;
296 	if (priv_check_cred_vfs_generation(td->td_ucred)) {
297 		buf->f_fsid.val[0] = buf->f_fsid.val[1] = 0;
298 		prison_enforce_statfs(td->td_ucred, mp, buf);
299 	}
300 out:
301 	vfs_unbusy(mp);
302 	return (error);
303 }
304 
305 /*
306  * Get filesystem statistics.
307  */
308 #ifndef _SYS_SYSPROTO_H_
309 struct statfs_args {
310 	char *path;
311 	struct statfs *buf;
312 };
313 #endif
314 int
315 sys_statfs(struct thread *td, struct statfs_args *uap)
316 {
317 	struct statfs *sfp;
318 	int error;
319 
320 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
321 	error = kern_statfs(td, uap->path, UIO_USERSPACE, sfp);
322 	if (error == 0)
323 		error = copyout(sfp, uap->buf, sizeof(struct statfs));
324 	free(sfp, M_STATFS);
325 	return (error);
326 }
327 
328 int
329 kern_statfs(struct thread *td, const char *path, enum uio_seg pathseg,
330     struct statfs *buf)
331 {
332 	struct mount *mp;
333 	struct nameidata nd;
334 	int error;
335 
336 	NDINIT(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, td);
337 	error = namei(&nd);
338 	if (error != 0)
339 		return (error);
340 	mp = vfs_ref_from_vp(nd.ni_vp);
341 	NDFREE_NOTHING(&nd);
342 	vrele(nd.ni_vp);
343 	return (kern_do_statfs(td, mp, buf));
344 }
345 
346 /*
347  * Get filesystem statistics.
348  */
349 #ifndef _SYS_SYSPROTO_H_
350 struct fstatfs_args {
351 	int fd;
352 	struct statfs *buf;
353 };
354 #endif
355 int
356 sys_fstatfs(struct thread *td, struct fstatfs_args *uap)
357 {
358 	struct statfs *sfp;
359 	int error;
360 
361 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
362 	error = kern_fstatfs(td, uap->fd, sfp);
363 	if (error == 0)
364 		error = copyout(sfp, uap->buf, sizeof(struct statfs));
365 	free(sfp, M_STATFS);
366 	return (error);
367 }
368 
369 int
370 kern_fstatfs(struct thread *td, int fd, struct statfs *buf)
371 {
372 	struct file *fp;
373 	struct mount *mp;
374 	struct vnode *vp;
375 	int error;
376 
377 	AUDIT_ARG_FD(fd);
378 	error = getvnode_path(td, fd, &cap_fstatfs_rights, &fp);
379 	if (error != 0)
380 		return (error);
381 	vp = fp->f_vnode;
382 #ifdef AUDIT
383 	if (AUDITING_TD(td)) {
384 		vn_lock(vp, LK_SHARED | LK_RETRY);
385 		AUDIT_ARG_VNODE1(vp);
386 		VOP_UNLOCK(vp);
387 	}
388 #endif
389 	mp = vfs_ref_from_vp(vp);
390 	fdrop(fp, td);
391 	return (kern_do_statfs(td, mp, buf));
392 }
393 
394 /*
395  * Get statistics on all filesystems.
396  */
397 #ifndef _SYS_SYSPROTO_H_
398 struct getfsstat_args {
399 	struct statfs *buf;
400 	long bufsize;
401 	int mode;
402 };
403 #endif
404 int
405 sys_getfsstat(struct thread *td, struct getfsstat_args *uap)
406 {
407 	size_t count;
408 	int error;
409 
410 	if (uap->bufsize < 0 || uap->bufsize > SIZE_MAX)
411 		return (EINVAL);
412 	error = kern_getfsstat(td, &uap->buf, uap->bufsize, &count,
413 	    UIO_USERSPACE, uap->mode);
414 	if (error == 0)
415 		td->td_retval[0] = count;
416 	return (error);
417 }
418 
419 /*
420  * If (bufsize > 0 && bufseg == UIO_SYSSPACE)
421  *	The caller is responsible for freeing memory which will be allocated
422  *	in '*buf'.
423  */
424 int
425 kern_getfsstat(struct thread *td, struct statfs **buf, size_t bufsize,
426     size_t *countp, enum uio_seg bufseg, int mode)
427 {
428 	struct mount *mp, *nmp;
429 	struct statfs *sfsp, *sp, *sptmp, *tofree;
430 	size_t count, maxcount;
431 	int error;
432 
433 	switch (mode) {
434 	case MNT_WAIT:
435 	case MNT_NOWAIT:
436 		break;
437 	default:
438 		if (bufseg == UIO_SYSSPACE)
439 			*buf = NULL;
440 		return (EINVAL);
441 	}
442 restart:
443 	maxcount = bufsize / sizeof(struct statfs);
444 	if (bufsize == 0) {
445 		sfsp = NULL;
446 		tofree = NULL;
447 	} else if (bufseg == UIO_USERSPACE) {
448 		sfsp = *buf;
449 		tofree = NULL;
450 	} else /* if (bufseg == UIO_SYSSPACE) */ {
451 		count = 0;
452 		mtx_lock(&mountlist_mtx);
453 		TAILQ_FOREACH(mp, &mountlist, mnt_list) {
454 			count++;
455 		}
456 		mtx_unlock(&mountlist_mtx);
457 		if (maxcount > count)
458 			maxcount = count;
459 		tofree = sfsp = *buf = malloc(maxcount * sizeof(struct statfs),
460 		    M_STATFS, M_WAITOK);
461 	}
462 
463 	count = 0;
464 
465 	/*
466 	 * If there is no target buffer they only want the count.
467 	 *
468 	 * This could be TAILQ_FOREACH but it is open-coded to match the original
469 	 * code below.
470 	 */
471 	if (sfsp == NULL) {
472 		mtx_lock(&mountlist_mtx);
473 		for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
474 			if (prison_canseemount(td->td_ucred, mp) != 0) {
475 				nmp = TAILQ_NEXT(mp, mnt_list);
476 				continue;
477 			}
478 #ifdef MAC
479 			if (mac_mount_check_stat(td->td_ucred, mp) != 0) {
480 				nmp = TAILQ_NEXT(mp, mnt_list);
481 				continue;
482 			}
483 #endif
484 			count++;
485 			nmp = TAILQ_NEXT(mp, mnt_list);
486 		}
487 		mtx_unlock(&mountlist_mtx);
488 		*countp = count;
489 		return (0);
490 	}
491 
492 	/*
493 	 * They want the entire thing.
494 	 *
495 	 * Short-circuit the corner case of no room for anything, avoids
496 	 * relocking below.
497 	 */
498 	if (maxcount < 1) {
499 		goto out;
500 	}
501 
502 	mtx_lock(&mountlist_mtx);
503 	for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
504 		if (prison_canseemount(td->td_ucred, mp) != 0) {
505 			nmp = TAILQ_NEXT(mp, mnt_list);
506 			continue;
507 		}
508 #ifdef MAC
509 		if (mac_mount_check_stat(td->td_ucred, mp) != 0) {
510 			nmp = TAILQ_NEXT(mp, mnt_list);
511 			continue;
512 		}
513 #endif
514 		if (mode == MNT_WAIT) {
515 			if (vfs_busy(mp, MBF_MNTLSTLOCK) != 0) {
516 				/*
517 				 * If vfs_busy() failed, and MBF_NOWAIT
518 				 * wasn't passed, then the mp is gone.
519 				 * Furthermore, because of MBF_MNTLSTLOCK,
520 				 * the mountlist_mtx was dropped.  We have
521 				 * no other choice than to start over.
522 				 */
523 				mtx_unlock(&mountlist_mtx);
524 				free(tofree, M_STATFS);
525 				goto restart;
526 			}
527 		} else {
528 			if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK) != 0) {
529 				nmp = TAILQ_NEXT(mp, mnt_list);
530 				continue;
531 			}
532 		}
533 		sp = &mp->mnt_stat;
534 		/*
535 		 * If MNT_NOWAIT is specified, do not refresh
536 		 * the fsstat cache.
537 		 */
538 		if (mode != MNT_NOWAIT) {
539 			error = VFS_STATFS(mp, sp);
540 			if (error != 0) {
541 				mtx_lock(&mountlist_mtx);
542 				nmp = TAILQ_NEXT(mp, mnt_list);
543 				vfs_unbusy(mp);
544 				continue;
545 			}
546 		}
547 		if (priv_check_cred_vfs_generation(td->td_ucred)) {
548 			sptmp = malloc(sizeof(struct statfs), M_STATFS,
549 			    M_WAITOK);
550 			*sptmp = *sp;
551 			sptmp->f_fsid.val[0] = sptmp->f_fsid.val[1] = 0;
552 			prison_enforce_statfs(td->td_ucred, mp, sptmp);
553 			sp = sptmp;
554 		} else
555 			sptmp = NULL;
556 		if (bufseg == UIO_SYSSPACE) {
557 			bcopy(sp, sfsp, sizeof(*sp));
558 			free(sptmp, M_STATFS);
559 		} else /* if (bufseg == UIO_USERSPACE) */ {
560 			error = copyout(sp, sfsp, sizeof(*sp));
561 			free(sptmp, M_STATFS);
562 			if (error != 0) {
563 				vfs_unbusy(mp);
564 				return (error);
565 			}
566 		}
567 		sfsp++;
568 		count++;
569 
570 		if (count == maxcount) {
571 			vfs_unbusy(mp);
572 			goto out;
573 		}
574 
575 		mtx_lock(&mountlist_mtx);
576 		nmp = TAILQ_NEXT(mp, mnt_list);
577 		vfs_unbusy(mp);
578 	}
579 	mtx_unlock(&mountlist_mtx);
580 out:
581 	*countp = count;
582 	return (0);
583 }
584 
585 #ifdef COMPAT_FREEBSD4
586 /*
587  * Get old format filesystem statistics.
588  */
589 static void freebsd4_cvtstatfs(struct statfs *, struct ostatfs *);
590 
591 #ifndef _SYS_SYSPROTO_H_
592 struct freebsd4_statfs_args {
593 	char *path;
594 	struct ostatfs *buf;
595 };
596 #endif
597 int
598 freebsd4_statfs(struct thread *td, struct freebsd4_statfs_args *uap)
599 {
600 	struct ostatfs osb;
601 	struct statfs *sfp;
602 	int error;
603 
604 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
605 	error = kern_statfs(td, uap->path, UIO_USERSPACE, sfp);
606 	if (error == 0) {
607 		freebsd4_cvtstatfs(sfp, &osb);
608 		error = copyout(&osb, uap->buf, sizeof(osb));
609 	}
610 	free(sfp, M_STATFS);
611 	return (error);
612 }
613 
614 /*
615  * Get filesystem statistics.
616  */
617 #ifndef _SYS_SYSPROTO_H_
618 struct freebsd4_fstatfs_args {
619 	int fd;
620 	struct ostatfs *buf;
621 };
622 #endif
623 int
624 freebsd4_fstatfs(struct thread *td, struct freebsd4_fstatfs_args *uap)
625 {
626 	struct ostatfs osb;
627 	struct statfs *sfp;
628 	int error;
629 
630 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
631 	error = kern_fstatfs(td, uap->fd, sfp);
632 	if (error == 0) {
633 		freebsd4_cvtstatfs(sfp, &osb);
634 		error = copyout(&osb, uap->buf, sizeof(osb));
635 	}
636 	free(sfp, M_STATFS);
637 	return (error);
638 }
639 
640 /*
641  * Get statistics on all filesystems.
642  */
643 #ifndef _SYS_SYSPROTO_H_
644 struct freebsd4_getfsstat_args {
645 	struct ostatfs *buf;
646 	long bufsize;
647 	int mode;
648 };
649 #endif
650 int
651 freebsd4_getfsstat(struct thread *td, struct freebsd4_getfsstat_args *uap)
652 {
653 	struct statfs *buf, *sp;
654 	struct ostatfs osb;
655 	size_t count, size;
656 	int error;
657 
658 	if (uap->bufsize < 0)
659 		return (EINVAL);
660 	count = uap->bufsize / sizeof(struct ostatfs);
661 	if (count > SIZE_MAX / sizeof(struct statfs))
662 		return (EINVAL);
663 	size = count * sizeof(struct statfs);
664 	error = kern_getfsstat(td, &buf, size, &count, UIO_SYSSPACE,
665 	    uap->mode);
666 	if (error == 0)
667 		td->td_retval[0] = count;
668 	if (size != 0) {
669 		sp = buf;
670 		while (count != 0 && error == 0) {
671 			freebsd4_cvtstatfs(sp, &osb);
672 			error = copyout(&osb, uap->buf, sizeof(osb));
673 			sp++;
674 			uap->buf++;
675 			count--;
676 		}
677 		free(buf, M_STATFS);
678 	}
679 	return (error);
680 }
681 
682 /*
683  * Implement fstatfs() for (NFS) file handles.
684  */
685 #ifndef _SYS_SYSPROTO_H_
686 struct freebsd4_fhstatfs_args {
687 	struct fhandle *u_fhp;
688 	struct ostatfs *buf;
689 };
690 #endif
691 int
692 freebsd4_fhstatfs(struct thread *td, struct freebsd4_fhstatfs_args *uap)
693 {
694 	struct ostatfs osb;
695 	struct statfs *sfp;
696 	fhandle_t fh;
697 	int error;
698 
699 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
700 	if (error != 0)
701 		return (error);
702 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
703 	error = kern_fhstatfs(td, fh, sfp);
704 	if (error == 0) {
705 		freebsd4_cvtstatfs(sfp, &osb);
706 		error = copyout(&osb, uap->buf, sizeof(osb));
707 	}
708 	free(sfp, M_STATFS);
709 	return (error);
710 }
711 
712 /*
713  * Convert a new format statfs structure to an old format statfs structure.
714  */
715 static void
716 freebsd4_cvtstatfs(struct statfs *nsp, struct ostatfs *osp)
717 {
718 
719 	statfs_scale_blocks(nsp, LONG_MAX);
720 	bzero(osp, sizeof(*osp));
721 	osp->f_bsize = nsp->f_bsize;
722 	osp->f_iosize = MIN(nsp->f_iosize, LONG_MAX);
723 	osp->f_blocks = nsp->f_blocks;
724 	osp->f_bfree = nsp->f_bfree;
725 	osp->f_bavail = nsp->f_bavail;
726 	osp->f_files = MIN(nsp->f_files, LONG_MAX);
727 	osp->f_ffree = MIN(nsp->f_ffree, LONG_MAX);
728 	osp->f_owner = nsp->f_owner;
729 	osp->f_type = nsp->f_type;
730 	osp->f_flags = nsp->f_flags;
731 	osp->f_syncwrites = MIN(nsp->f_syncwrites, LONG_MAX);
732 	osp->f_asyncwrites = MIN(nsp->f_asyncwrites, LONG_MAX);
733 	osp->f_syncreads = MIN(nsp->f_syncreads, LONG_MAX);
734 	osp->f_asyncreads = MIN(nsp->f_asyncreads, LONG_MAX);
735 	strlcpy(osp->f_fstypename, nsp->f_fstypename,
736 	    MIN(MFSNAMELEN, OMFSNAMELEN));
737 	strlcpy(osp->f_mntonname, nsp->f_mntonname,
738 	    MIN(MNAMELEN, OMNAMELEN));
739 	strlcpy(osp->f_mntfromname, nsp->f_mntfromname,
740 	    MIN(MNAMELEN, OMNAMELEN));
741 	osp->f_fsid = nsp->f_fsid;
742 }
743 #endif /* COMPAT_FREEBSD4 */
744 
745 #if defined(COMPAT_FREEBSD11)
746 /*
747  * Get old format filesystem statistics.
748  */
749 static void freebsd11_cvtstatfs(struct statfs *, struct freebsd11_statfs *);
750 
751 int
752 freebsd11_statfs(struct thread *td, struct freebsd11_statfs_args *uap)
753 {
754 	struct freebsd11_statfs osb;
755 	struct statfs *sfp;
756 	int error;
757 
758 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
759 	error = kern_statfs(td, uap->path, UIO_USERSPACE, sfp);
760 	if (error == 0) {
761 		freebsd11_cvtstatfs(sfp, &osb);
762 		error = copyout(&osb, uap->buf, sizeof(osb));
763 	}
764 	free(sfp, M_STATFS);
765 	return (error);
766 }
767 
768 /*
769  * Get filesystem statistics.
770  */
771 int
772 freebsd11_fstatfs(struct thread *td, struct freebsd11_fstatfs_args *uap)
773 {
774 	struct freebsd11_statfs osb;
775 	struct statfs *sfp;
776 	int error;
777 
778 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
779 	error = kern_fstatfs(td, uap->fd, sfp);
780 	if (error == 0) {
781 		freebsd11_cvtstatfs(sfp, &osb);
782 		error = copyout(&osb, uap->buf, sizeof(osb));
783 	}
784 	free(sfp, M_STATFS);
785 	return (error);
786 }
787 
788 /*
789  * Get statistics on all filesystems.
790  */
791 int
792 freebsd11_getfsstat(struct thread *td, struct freebsd11_getfsstat_args *uap)
793 {
794 	struct freebsd11_statfs osb;
795 	struct statfs *buf, *sp;
796 	size_t count, size;
797 	int error;
798 
799 	count = uap->bufsize / sizeof(struct ostatfs);
800 	size = count * sizeof(struct statfs);
801 	error = kern_getfsstat(td, &buf, size, &count, UIO_SYSSPACE,
802 	    uap->mode);
803 	if (error == 0)
804 		td->td_retval[0] = count;
805 	if (size > 0) {
806 		sp = buf;
807 		while (count > 0 && error == 0) {
808 			freebsd11_cvtstatfs(sp, &osb);
809 			error = copyout(&osb, uap->buf, sizeof(osb));
810 			sp++;
811 			uap->buf++;
812 			count--;
813 		}
814 		free(buf, M_STATFS);
815 	}
816 	return (error);
817 }
818 
819 /*
820  * Implement fstatfs() for (NFS) file handles.
821  */
822 int
823 freebsd11_fhstatfs(struct thread *td, struct freebsd11_fhstatfs_args *uap)
824 {
825 	struct freebsd11_statfs osb;
826 	struct statfs *sfp;
827 	fhandle_t fh;
828 	int error;
829 
830 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
831 	if (error)
832 		return (error);
833 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
834 	error = kern_fhstatfs(td, fh, sfp);
835 	if (error == 0) {
836 		freebsd11_cvtstatfs(sfp, &osb);
837 		error = copyout(&osb, uap->buf, sizeof(osb));
838 	}
839 	free(sfp, M_STATFS);
840 	return (error);
841 }
842 
843 /*
844  * Convert a new format statfs structure to an old format statfs structure.
845  */
846 static void
847 freebsd11_cvtstatfs(struct statfs *nsp, struct freebsd11_statfs *osp)
848 {
849 
850 	bzero(osp, sizeof(*osp));
851 	osp->f_version = FREEBSD11_STATFS_VERSION;
852 	osp->f_type = nsp->f_type;
853 	osp->f_flags = nsp->f_flags;
854 	osp->f_bsize = nsp->f_bsize;
855 	osp->f_iosize = nsp->f_iosize;
856 	osp->f_blocks = nsp->f_blocks;
857 	osp->f_bfree = nsp->f_bfree;
858 	osp->f_bavail = nsp->f_bavail;
859 	osp->f_files = nsp->f_files;
860 	osp->f_ffree = nsp->f_ffree;
861 	osp->f_syncwrites = nsp->f_syncwrites;
862 	osp->f_asyncwrites = nsp->f_asyncwrites;
863 	osp->f_syncreads = nsp->f_syncreads;
864 	osp->f_asyncreads = nsp->f_asyncreads;
865 	osp->f_namemax = nsp->f_namemax;
866 	osp->f_owner = nsp->f_owner;
867 	osp->f_fsid = nsp->f_fsid;
868 	strlcpy(osp->f_fstypename, nsp->f_fstypename,
869 	    MIN(MFSNAMELEN, sizeof(osp->f_fstypename)));
870 	strlcpy(osp->f_mntonname, nsp->f_mntonname,
871 	    MIN(MNAMELEN, sizeof(osp->f_mntonname)));
872 	strlcpy(osp->f_mntfromname, nsp->f_mntfromname,
873 	    MIN(MNAMELEN, sizeof(osp->f_mntfromname)));
874 }
875 #endif /* COMPAT_FREEBSD11 */
876 
877 /*
878  * Change current working directory to a given file descriptor.
879  */
880 #ifndef _SYS_SYSPROTO_H_
881 struct fchdir_args {
882 	int	fd;
883 };
884 #endif
885 int
886 sys_fchdir(struct thread *td, struct fchdir_args *uap)
887 {
888 	struct vnode *vp, *tdp;
889 	struct mount *mp;
890 	struct file *fp;
891 	int error;
892 
893 	AUDIT_ARG_FD(uap->fd);
894 	error = getvnode_path(td, uap->fd, &cap_fchdir_rights,
895 	    &fp);
896 	if (error != 0)
897 		return (error);
898 	vp = fp->f_vnode;
899 	vref(vp);
900 	fdrop(fp, td);
901 	vn_lock(vp, LK_SHARED | LK_RETRY);
902 	AUDIT_ARG_VNODE1(vp);
903 	error = change_dir(vp, td);
904 	while (!error && (mp = vp->v_mountedhere) != NULL) {
905 		if (vfs_busy(mp, 0))
906 			continue;
907 		error = VFS_ROOT(mp, LK_SHARED, &tdp);
908 		vfs_unbusy(mp);
909 		if (error != 0)
910 			break;
911 		vput(vp);
912 		vp = tdp;
913 	}
914 	if (error != 0) {
915 		vput(vp);
916 		return (error);
917 	}
918 	VOP_UNLOCK(vp);
919 	pwd_chdir(td, vp);
920 	return (0);
921 }
922 
923 /*
924  * Change current working directory (``.'').
925  */
926 #ifndef _SYS_SYSPROTO_H_
927 struct chdir_args {
928 	char	*path;
929 };
930 #endif
931 int
932 sys_chdir(struct thread *td, struct chdir_args *uap)
933 {
934 
935 	return (kern_chdir(td, uap->path, UIO_USERSPACE));
936 }
937 
938 int
939 kern_chdir(struct thread *td, const char *path, enum uio_seg pathseg)
940 {
941 	struct nameidata nd;
942 	int error;
943 
944 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
945 	    pathseg, path, td);
946 	if ((error = namei(&nd)) != 0)
947 		return (error);
948 	if ((error = change_dir(nd.ni_vp, td)) != 0) {
949 		vput(nd.ni_vp);
950 		NDFREE_NOTHING(&nd);
951 		return (error);
952 	}
953 	VOP_UNLOCK(nd.ni_vp);
954 	NDFREE_NOTHING(&nd);
955 	pwd_chdir(td, nd.ni_vp);
956 	return (0);
957 }
958 
959 /*
960  * Change notion of root (``/'') directory.
961  */
962 #ifndef _SYS_SYSPROTO_H_
963 struct chroot_args {
964 	char	*path;
965 };
966 #endif
967 int
968 sys_chroot(struct thread *td, struct chroot_args *uap)
969 {
970 	struct nameidata nd;
971 	int error;
972 
973 	error = priv_check(td, PRIV_VFS_CHROOT);
974 	if (error != 0)
975 		return (error);
976 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
977 	    UIO_USERSPACE, uap->path, td);
978 	error = namei(&nd);
979 	if (error != 0)
980 		goto error;
981 	error = change_dir(nd.ni_vp, td);
982 	if (error != 0)
983 		goto e_vunlock;
984 #ifdef MAC
985 	error = mac_vnode_check_chroot(td->td_ucred, nd.ni_vp);
986 	if (error != 0)
987 		goto e_vunlock;
988 #endif
989 	VOP_UNLOCK(nd.ni_vp);
990 	error = pwd_chroot(td, nd.ni_vp);
991 	vrele(nd.ni_vp);
992 	NDFREE_NOTHING(&nd);
993 	return (error);
994 e_vunlock:
995 	vput(nd.ni_vp);
996 error:
997 	NDFREE_NOTHING(&nd);
998 	return (error);
999 }
1000 
1001 /*
1002  * Common routine for chroot and chdir.  Callers must provide a locked vnode
1003  * instance.
1004  */
1005 int
1006 change_dir(struct vnode *vp, struct thread *td)
1007 {
1008 #ifdef MAC
1009 	int error;
1010 #endif
1011 
1012 	ASSERT_VOP_LOCKED(vp, "change_dir(): vp not locked");
1013 	if (vp->v_type != VDIR)
1014 		return (ENOTDIR);
1015 #ifdef MAC
1016 	error = mac_vnode_check_chdir(td->td_ucred, vp);
1017 	if (error != 0)
1018 		return (error);
1019 #endif
1020 	return (VOP_ACCESS(vp, VEXEC, td->td_ucred, td));
1021 }
1022 
1023 static __inline void
1024 flags_to_rights(int flags, cap_rights_t *rightsp)
1025 {
1026 	if (flags & O_EXEC) {
1027 		cap_rights_set_one(rightsp, CAP_FEXECVE);
1028 		if (flags & O_PATH)
1029 			return;
1030 	} else {
1031 		switch ((flags & O_ACCMODE)) {
1032 		case O_RDONLY:
1033 			cap_rights_set_one(rightsp, CAP_READ);
1034 			break;
1035 		case O_RDWR:
1036 			cap_rights_set_one(rightsp, CAP_READ);
1037 			/* FALLTHROUGH */
1038 		case O_WRONLY:
1039 			cap_rights_set_one(rightsp, CAP_WRITE);
1040 			if (!(flags & (O_APPEND | O_TRUNC)))
1041 				cap_rights_set_one(rightsp, CAP_SEEK);
1042 			break;
1043 		}
1044 	}
1045 
1046 	if (flags & O_CREAT)
1047 		cap_rights_set_one(rightsp, CAP_CREATE);
1048 
1049 	if (flags & O_TRUNC)
1050 		cap_rights_set_one(rightsp, CAP_FTRUNCATE);
1051 
1052 	if (flags & (O_SYNC | O_FSYNC))
1053 		cap_rights_set_one(rightsp, CAP_FSYNC);
1054 
1055 	if (flags & (O_EXLOCK | O_SHLOCK))
1056 		cap_rights_set_one(rightsp, CAP_FLOCK);
1057 }
1058 
1059 /*
1060  * Check permissions, allocate an open file structure, and call the device
1061  * open routine if any.
1062  */
1063 #ifndef _SYS_SYSPROTO_H_
1064 struct open_args {
1065 	char	*path;
1066 	int	flags;
1067 	int	mode;
1068 };
1069 #endif
1070 int
1071 sys_open(struct thread *td, struct open_args *uap)
1072 {
1073 
1074 	return (kern_openat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1075 	    uap->flags, uap->mode));
1076 }
1077 
1078 #ifndef _SYS_SYSPROTO_H_
1079 struct openat_args {
1080 	int	fd;
1081 	char	*path;
1082 	int	flag;
1083 	int	mode;
1084 };
1085 #endif
1086 int
1087 sys_openat(struct thread *td, struct openat_args *uap)
1088 {
1089 
1090 	AUDIT_ARG_FD(uap->fd);
1091 	return (kern_openat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
1092 	    uap->mode));
1093 }
1094 
1095 int
1096 kern_openat(struct thread *td, int fd, const char *path, enum uio_seg pathseg,
1097     int flags, int mode)
1098 {
1099 	struct proc *p = td->td_proc;
1100 	struct filedesc *fdp;
1101 	struct pwddesc *pdp;
1102 	struct file *fp;
1103 	struct vnode *vp;
1104 	struct nameidata nd;
1105 	cap_rights_t rights;
1106 	int cmode, error, indx;
1107 
1108 	indx = -1;
1109 	fdp = p->p_fd;
1110 	pdp = p->p_pd;
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 	/*
1118 	 * Only one of the O_EXEC, O_RDONLY, O_WRONLY and O_RDWR flags
1119 	 * may be specified.  On the other hand, for O_PATH any mode
1120 	 * except O_EXEC is ignored.
1121 	 */
1122 	if ((flags & O_PATH) != 0) {
1123 		flags &= ~(O_CREAT | O_ACCMODE);
1124 	} else if ((flags & O_EXEC) != 0) {
1125 		if (flags & O_ACCMODE)
1126 			return (EINVAL);
1127 	} else if ((flags & O_ACCMODE) == O_ACCMODE) {
1128 		return (EINVAL);
1129 	} else {
1130 		flags = FFLAGS(flags);
1131 	}
1132 
1133 	/*
1134 	 * Allocate a file structure. The descriptor to reference it
1135 	 * is allocated and used by finstall_refed() below.
1136 	 */
1137 	error = falloc_noinstall(td, &fp);
1138 	if (error != 0)
1139 		return (error);
1140 	/* Set the flags early so the finit in devfs can pick them up. */
1141 	fp->f_flag = flags & FMASK;
1142 	cmode = ((mode & ~pdp->pd_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 			MPASS((flags & O_PATH) == 0);
1155 			goto success;
1156 		}
1157 
1158 		/*
1159 		 * Handle special fdopen() case. bleh.
1160 		 *
1161 		 * Don't do this for relative (capability) lookups; we don't
1162 		 * understand exactly what would happen, and we don't think
1163 		 * that it ever should.
1164 		 */
1165 		if ((nd.ni_resflags & NIRES_STRICTREL) == 0 &&
1166 		    (error == ENODEV || error == ENXIO) &&
1167 		    td->td_dupfd >= 0) {
1168 			error = dupfdopen(td, fdp, td->td_dupfd, flags, error,
1169 			    &indx);
1170 			if (error == 0)
1171 				goto success;
1172 		}
1173 
1174 		goto bad;
1175 	}
1176 	td->td_dupfd = 0;
1177 	NDFREE(&nd, NDF_ONLY_PNBUF);
1178 	vp = nd.ni_vp;
1179 
1180 	/*
1181 	 * Store the vnode, for any f_type. Typically, the vnode use
1182 	 * count is decremented by direct call to vn_closefile() for
1183 	 * files that switched type in the cdevsw fdopen() method.
1184 	 */
1185 	fp->f_vnode = vp;
1186 
1187 	/*
1188 	 * If the file wasn't claimed by devfs bind it to the normal
1189 	 * vnode operations here.
1190 	 */
1191 	if (fp->f_ops == &badfileops) {
1192 		KASSERT(vp->v_type != VFIFO || (flags & O_PATH) != 0,
1193 		    ("Unexpected fifo fp %p vp %p", fp, vp));
1194 		if ((flags & O_PATH) != 0) {
1195 			finit(fp, (flags & FMASK) | (fp->f_flag & FKQALLOWED),
1196 			    DTYPE_VNODE, NULL, &path_fileops);
1197 			vhold(vp);
1198 			vunref(vp);
1199 		} else {
1200 			finit_vnode(fp, flags, NULL, &vnops);
1201 		}
1202 	}
1203 
1204 	VOP_UNLOCK(vp);
1205 	if (flags & O_TRUNC) {
1206 		error = fo_truncate(fp, 0, td->td_ucred, td);
1207 		if (error != 0)
1208 			goto bad;
1209 	}
1210 success:
1211 	/*
1212 	 * If we haven't already installed the FD (for dupfdopen), do so now.
1213 	 */
1214 	if (indx == -1) {
1215 		struct filecaps *fcaps;
1216 
1217 #ifdef CAPABILITIES
1218 		if ((nd.ni_resflags & NIRES_STRICTREL) != 0)
1219 			fcaps = &nd.ni_filecaps;
1220 		else
1221 #endif
1222 			fcaps = NULL;
1223 		error = finstall_refed(td, fp, &indx, flags, fcaps);
1224 		/* On success finstall_refed() consumes fcaps. */
1225 		if (error != 0) {
1226 			filecaps_free(&nd.ni_filecaps);
1227 			goto bad;
1228 		}
1229 	} else {
1230 		filecaps_free(&nd.ni_filecaps);
1231 		falloc_abort(td, fp);
1232 	}
1233 
1234 	td->td_retval[0] = indx;
1235 	return (0);
1236 bad:
1237 	KASSERT(indx == -1, ("indx=%d, should be -1", indx));
1238 	falloc_abort(td, fp);
1239 	return (error);
1240 }
1241 
1242 #ifdef COMPAT_43
1243 /*
1244  * Create a file.
1245  */
1246 #ifndef _SYS_SYSPROTO_H_
1247 struct ocreat_args {
1248 	char	*path;
1249 	int	mode;
1250 };
1251 #endif
1252 int
1253 ocreat(struct thread *td, struct ocreat_args *uap)
1254 {
1255 
1256 	return (kern_openat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1257 	    O_WRONLY | O_CREAT | O_TRUNC, uap->mode));
1258 }
1259 #endif /* COMPAT_43 */
1260 
1261 /*
1262  * Create a special file.
1263  */
1264 #ifndef _SYS_SYSPROTO_H_
1265 struct mknodat_args {
1266 	int	fd;
1267 	char	*path;
1268 	mode_t	mode;
1269 	dev_t	dev;
1270 };
1271 #endif
1272 int
1273 sys_mknodat(struct thread *td, struct mknodat_args *uap)
1274 {
1275 
1276 	return (kern_mknodat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode,
1277 	    uap->dev));
1278 }
1279 
1280 #if defined(COMPAT_FREEBSD11)
1281 int
1282 freebsd11_mknod(struct thread *td,
1283     struct freebsd11_mknod_args *uap)
1284 {
1285 
1286 	return (kern_mknodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1287 	    uap->mode, uap->dev));
1288 }
1289 
1290 int
1291 freebsd11_mknodat(struct thread *td,
1292     struct freebsd11_mknodat_args *uap)
1293 {
1294 
1295 	return (kern_mknodat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode,
1296 	    uap->dev));
1297 }
1298 #endif /* COMPAT_FREEBSD11 */
1299 
1300 int
1301 kern_mknodat(struct thread *td, int fd, const char *path, enum uio_seg pathseg,
1302     int mode, dev_t dev)
1303 {
1304 	struct vnode *vp;
1305 	struct mount *mp;
1306 	struct vattr vattr;
1307 	struct nameidata nd;
1308 	int error, whiteout = 0;
1309 
1310 	AUDIT_ARG_MODE(mode);
1311 	AUDIT_ARG_DEV(dev);
1312 	switch (mode & S_IFMT) {
1313 	case S_IFCHR:
1314 	case S_IFBLK:
1315 		error = priv_check(td, PRIV_VFS_MKNOD_DEV);
1316 		if (error == 0 && dev == VNOVAL)
1317 			error = EINVAL;
1318 		break;
1319 	case S_IFWHT:
1320 		error = priv_check(td, PRIV_VFS_MKNOD_WHT);
1321 		break;
1322 	case S_IFIFO:
1323 		if (dev == 0)
1324 			return (kern_mkfifoat(td, fd, path, pathseg, mode));
1325 		/* FALLTHROUGH */
1326 	default:
1327 		error = EINVAL;
1328 		break;
1329 	}
1330 	if (error != 0)
1331 		return (error);
1332 restart:
1333 	bwillwrite();
1334 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
1335 	    NOCACHE, pathseg, path, fd, &cap_mknodat_rights,
1336 	    td);
1337 	if ((error = namei(&nd)) != 0)
1338 		return (error);
1339 	vp = nd.ni_vp;
1340 	if (vp != NULL) {
1341 		NDFREE(&nd, NDF_ONLY_PNBUF);
1342 		if (vp == nd.ni_dvp)
1343 			vrele(nd.ni_dvp);
1344 		else
1345 			vput(nd.ni_dvp);
1346 		vrele(vp);
1347 		return (EEXIST);
1348 	} else {
1349 		VATTR_NULL(&vattr);
1350 		vattr.va_mode = (mode & ALLPERMS) &
1351 		    ~td->td_proc->p_pd->pd_cmask;
1352 		vattr.va_rdev = dev;
1353 		whiteout = 0;
1354 
1355 		switch (mode & S_IFMT) {
1356 		case S_IFCHR:
1357 			vattr.va_type = VCHR;
1358 			break;
1359 		case S_IFBLK:
1360 			vattr.va_type = VBLK;
1361 			break;
1362 		case S_IFWHT:
1363 			whiteout = 1;
1364 			break;
1365 		default:
1366 			panic("kern_mknod: invalid mode");
1367 		}
1368 	}
1369 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1370 		NDFREE(&nd, NDF_ONLY_PNBUF);
1371 		vput(nd.ni_dvp);
1372 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1373 			return (error);
1374 		goto restart;
1375 	}
1376 #ifdef MAC
1377 	if (error == 0 && !whiteout)
1378 		error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp,
1379 		    &nd.ni_cnd, &vattr);
1380 #endif
1381 	if (error == 0) {
1382 		if (whiteout)
1383 			error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE);
1384 		else {
1385 			error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp,
1386 						&nd.ni_cnd, &vattr);
1387 		}
1388 	}
1389 	VOP_VPUT_PAIR(nd.ni_dvp, error == 0 && !whiteout ? &nd.ni_vp : NULL,
1390 	    true);
1391 	vn_finished_write(mp);
1392 	NDFREE(&nd, NDF_ONLY_PNBUF);
1393 	if (error == ERELOOKUP)
1394 		goto restart;
1395 	return (error);
1396 }
1397 
1398 /*
1399  * Create a named pipe.
1400  */
1401 #ifndef _SYS_SYSPROTO_H_
1402 struct mkfifo_args {
1403 	char	*path;
1404 	int	mode;
1405 };
1406 #endif
1407 int
1408 sys_mkfifo(struct thread *td, struct mkfifo_args *uap)
1409 {
1410 
1411 	return (kern_mkfifoat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1412 	    uap->mode));
1413 }
1414 
1415 #ifndef _SYS_SYSPROTO_H_
1416 struct mkfifoat_args {
1417 	int	fd;
1418 	char	*path;
1419 	mode_t	mode;
1420 };
1421 #endif
1422 int
1423 sys_mkfifoat(struct thread *td, struct mkfifoat_args *uap)
1424 {
1425 
1426 	return (kern_mkfifoat(td, uap->fd, uap->path, UIO_USERSPACE,
1427 	    uap->mode));
1428 }
1429 
1430 int
1431 kern_mkfifoat(struct thread *td, int fd, const char *path,
1432     enum uio_seg pathseg, int mode)
1433 {
1434 	struct mount *mp;
1435 	struct vattr vattr;
1436 	struct nameidata nd;
1437 	int error;
1438 
1439 	AUDIT_ARG_MODE(mode);
1440 restart:
1441 	bwillwrite();
1442 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
1443 	    NOCACHE, pathseg, path, fd, &cap_mkfifoat_rights,
1444 	    td);
1445 	if ((error = namei(&nd)) != 0)
1446 		return (error);
1447 	if (nd.ni_vp != NULL) {
1448 		NDFREE(&nd, NDF_ONLY_PNBUF);
1449 		if (nd.ni_vp == nd.ni_dvp)
1450 			vrele(nd.ni_dvp);
1451 		else
1452 			vput(nd.ni_dvp);
1453 		vrele(nd.ni_vp);
1454 		return (EEXIST);
1455 	}
1456 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1457 		NDFREE(&nd, NDF_ONLY_PNBUF);
1458 		vput(nd.ni_dvp);
1459 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1460 			return (error);
1461 		goto restart;
1462 	}
1463 	VATTR_NULL(&vattr);
1464 	vattr.va_type = VFIFO;
1465 	vattr.va_mode = (mode & ALLPERMS) & ~td->td_proc->p_pd->pd_cmask;
1466 #ifdef MAC
1467 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1468 	    &vattr);
1469 	if (error != 0)
1470 		goto out;
1471 #endif
1472 	error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
1473 #ifdef MAC
1474 out:
1475 #endif
1476 	VOP_VPUT_PAIR(nd.ni_dvp, error == 0 ? &nd.ni_vp : NULL, true);
1477 	vn_finished_write(mp);
1478 	NDFREE(&nd, NDF_ONLY_PNBUF);
1479 	if (error == ERELOOKUP)
1480 		goto restart;
1481 	return (error);
1482 }
1483 
1484 /*
1485  * Make a hard file link.
1486  */
1487 #ifndef _SYS_SYSPROTO_H_
1488 struct link_args {
1489 	char	*path;
1490 	char	*link;
1491 };
1492 #endif
1493 int
1494 sys_link(struct thread *td, struct link_args *uap)
1495 {
1496 
1497 	return (kern_linkat(td, AT_FDCWD, AT_FDCWD, uap->path, uap->link,
1498 	    UIO_USERSPACE, AT_SYMLINK_FOLLOW));
1499 }
1500 
1501 #ifndef _SYS_SYSPROTO_H_
1502 struct linkat_args {
1503 	int	fd1;
1504 	char	*path1;
1505 	int	fd2;
1506 	char	*path2;
1507 	int	flag;
1508 };
1509 #endif
1510 int
1511 sys_linkat(struct thread *td, struct linkat_args *uap)
1512 {
1513 	int flag;
1514 
1515 	flag = uap->flag;
1516 	if ((flag & ~(AT_SYMLINK_FOLLOW | AT_RESOLVE_BENEATH |
1517 	    AT_EMPTY_PATH)) != 0)
1518 		return (EINVAL);
1519 
1520 	return (kern_linkat(td, uap->fd1, uap->fd2, uap->path1, uap->path2,
1521 	    UIO_USERSPACE, flag));
1522 }
1523 
1524 int hardlink_check_uid = 0;
1525 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_uid, CTLFLAG_RW,
1526     &hardlink_check_uid, 0,
1527     "Unprivileged processes cannot create hard links to files owned by other "
1528     "users");
1529 static int hardlink_check_gid = 0;
1530 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_gid, CTLFLAG_RW,
1531     &hardlink_check_gid, 0,
1532     "Unprivileged processes cannot create hard links to files owned by other "
1533     "groups");
1534 
1535 static int
1536 can_hardlink(struct vnode *vp, struct ucred *cred)
1537 {
1538 	struct vattr va;
1539 	int error;
1540 
1541 	if (!hardlink_check_uid && !hardlink_check_gid)
1542 		return (0);
1543 
1544 	error = VOP_GETATTR(vp, &va, cred);
1545 	if (error != 0)
1546 		return (error);
1547 
1548 	if (hardlink_check_uid && cred->cr_uid != va.va_uid) {
1549 		error = priv_check_cred(cred, PRIV_VFS_LINK);
1550 		if (error != 0)
1551 			return (error);
1552 	}
1553 
1554 	if (hardlink_check_gid && !groupmember(va.va_gid, cred)) {
1555 		error = priv_check_cred(cred, PRIV_VFS_LINK);
1556 		if (error != 0)
1557 			return (error);
1558 	}
1559 
1560 	return (0);
1561 }
1562 
1563 int
1564 kern_linkat(struct thread *td, int fd1, int fd2, const char *path1,
1565     const char *path2, enum uio_seg segflag, int flag)
1566 {
1567 	struct nameidata nd;
1568 	int error;
1569 
1570 	do {
1571 		bwillwrite();
1572 		NDINIT_ATRIGHTS(&nd, LOOKUP, AUDITVNODE1 | at2cnpflags(flag,
1573 		    AT_SYMLINK_FOLLOW | AT_RESOLVE_BENEATH | AT_EMPTY_PATH),
1574 		    segflag, path1, fd1, &cap_linkat_source_rights, td);
1575 		if ((error = namei(&nd)) != 0)
1576 			return (error);
1577 		NDFREE(&nd, NDF_ONLY_PNBUF);
1578 		if ((nd.ni_resflags & NIRES_EMPTYPATH) != 0) {
1579 			error = priv_check(td, PRIV_VFS_FHOPEN);
1580 			if (error != 0) {
1581 				vrele(nd.ni_vp);
1582 				return (error);
1583 			}
1584 		}
1585 		error = kern_linkat_vp(td, nd.ni_vp, fd2, path2, segflag);
1586 	} while (error ==  EAGAIN || error == ERELOOKUP);
1587 	return (error);
1588 }
1589 
1590 static int
1591 kern_linkat_vp(struct thread *td, struct vnode *vp, int fd, const char *path,
1592     enum uio_seg segflag)
1593 {
1594 	struct nameidata nd;
1595 	struct mount *mp;
1596 	int error;
1597 
1598 	if (vp->v_type == VDIR) {
1599 		vrele(vp);
1600 		return (EPERM);		/* POSIX */
1601 	}
1602 	NDINIT_ATRIGHTS(&nd, CREATE,
1603 	    LOCKPARENT | SAVENAME | AUDITVNODE2 | NOCACHE, segflag, path, fd,
1604 	    &cap_linkat_target_rights, td);
1605 	if ((error = namei(&nd)) == 0) {
1606 		if (nd.ni_vp != NULL) {
1607 			NDFREE(&nd, NDF_ONLY_PNBUF);
1608 			if (nd.ni_dvp == nd.ni_vp)
1609 				vrele(nd.ni_dvp);
1610 			else
1611 				vput(nd.ni_dvp);
1612 			vrele(nd.ni_vp);
1613 			vrele(vp);
1614 			return (EEXIST);
1615 		} else if (nd.ni_dvp->v_mount != vp->v_mount) {
1616 			/*
1617 			 * Cross-device link.  No need to recheck
1618 			 * vp->v_type, since it cannot change, except
1619 			 * to VBAD.
1620 			 */
1621 			NDFREE(&nd, NDF_ONLY_PNBUF);
1622 			vput(nd.ni_dvp);
1623 			vrele(vp);
1624 			return (EXDEV);
1625 		} else if ((error = vn_lock(vp, LK_EXCLUSIVE)) == 0) {
1626 			error = can_hardlink(vp, td->td_ucred);
1627 #ifdef MAC
1628 			if (error == 0)
1629 				error = mac_vnode_check_link(td->td_ucred,
1630 				    nd.ni_dvp, vp, &nd.ni_cnd);
1631 #endif
1632 			if (error != 0) {
1633 				vput(vp);
1634 				vput(nd.ni_dvp);
1635 				NDFREE(&nd, NDF_ONLY_PNBUF);
1636 				return (error);
1637 			}
1638 			error = vn_start_write(vp, &mp, V_NOWAIT);
1639 			if (error != 0) {
1640 				vput(vp);
1641 				vput(nd.ni_dvp);
1642 				NDFREE(&nd, NDF_ONLY_PNBUF);
1643 				error = vn_start_write(NULL, &mp,
1644 				    V_XSLEEP | PCATCH);
1645 				if (error != 0)
1646 					return (error);
1647 				return (EAGAIN);
1648 			}
1649 			error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
1650 			VOP_VPUT_PAIR(nd.ni_dvp, &vp, true);
1651 			vn_finished_write(mp);
1652 			NDFREE(&nd, NDF_ONLY_PNBUF);
1653 			vp = NULL;
1654 		} else {
1655 			vput(nd.ni_dvp);
1656 			NDFREE(&nd, NDF_ONLY_PNBUF);
1657 			vrele(vp);
1658 			return (EAGAIN);
1659 		}
1660 	}
1661 	if (vp != NULL)
1662 		vrele(vp);
1663 	return (error);
1664 }
1665 
1666 /*
1667  * Make a symbolic link.
1668  */
1669 #ifndef _SYS_SYSPROTO_H_
1670 struct symlink_args {
1671 	char	*path;
1672 	char	*link;
1673 };
1674 #endif
1675 int
1676 sys_symlink(struct thread *td, struct symlink_args *uap)
1677 {
1678 
1679 	return (kern_symlinkat(td, uap->path, AT_FDCWD, uap->link,
1680 	    UIO_USERSPACE));
1681 }
1682 
1683 #ifndef _SYS_SYSPROTO_H_
1684 struct symlinkat_args {
1685 	char	*path;
1686 	int	fd;
1687 	char	*path2;
1688 };
1689 #endif
1690 int
1691 sys_symlinkat(struct thread *td, struct symlinkat_args *uap)
1692 {
1693 
1694 	return (kern_symlinkat(td, uap->path1, uap->fd, uap->path2,
1695 	    UIO_USERSPACE));
1696 }
1697 
1698 int
1699 kern_symlinkat(struct thread *td, const char *path1, int fd, const char *path2,
1700     enum uio_seg segflg)
1701 {
1702 	struct mount *mp;
1703 	struct vattr vattr;
1704 	const char *syspath;
1705 	char *tmppath;
1706 	struct nameidata nd;
1707 	int error;
1708 
1709 	if (segflg == UIO_SYSSPACE) {
1710 		syspath = path1;
1711 	} else {
1712 		tmppath = uma_zalloc(namei_zone, M_WAITOK);
1713 		if ((error = copyinstr(path1, tmppath, MAXPATHLEN, NULL)) != 0)
1714 			goto out;
1715 		syspath = tmppath;
1716 	}
1717 	AUDIT_ARG_TEXT(syspath);
1718 restart:
1719 	bwillwrite();
1720 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
1721 	    NOCACHE, segflg, path2, fd, &cap_symlinkat_rights,
1722 	    td);
1723 	if ((error = namei(&nd)) != 0)
1724 		goto out;
1725 	if (nd.ni_vp) {
1726 		NDFREE(&nd, NDF_ONLY_PNBUF);
1727 		if (nd.ni_vp == nd.ni_dvp)
1728 			vrele(nd.ni_dvp);
1729 		else
1730 			vput(nd.ni_dvp);
1731 		vrele(nd.ni_vp);
1732 		nd.ni_vp = NULL;
1733 		error = EEXIST;
1734 		goto out;
1735 	}
1736 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1737 		NDFREE(&nd, NDF_ONLY_PNBUF);
1738 		vput(nd.ni_dvp);
1739 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1740 			goto out;
1741 		goto restart;
1742 	}
1743 	VATTR_NULL(&vattr);
1744 	vattr.va_mode = ACCESSPERMS &~ td->td_proc->p_pd->pd_cmask;
1745 #ifdef MAC
1746 	vattr.va_type = VLNK;
1747 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1748 	    &vattr);
1749 	if (error != 0)
1750 		goto out2;
1751 #endif
1752 	error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, syspath);
1753 #ifdef MAC
1754 out2:
1755 #endif
1756 	VOP_VPUT_PAIR(nd.ni_dvp, error == 0 ? &nd.ni_vp : NULL, true);
1757 	vn_finished_write(mp);
1758 	NDFREE(&nd, NDF_ONLY_PNBUF);
1759 	if (error == ERELOOKUP)
1760 		goto restart;
1761 out:
1762 	if (segflg != UIO_SYSSPACE)
1763 		uma_zfree(namei_zone, tmppath);
1764 	return (error);
1765 }
1766 
1767 /*
1768  * Delete a whiteout from the filesystem.
1769  */
1770 #ifndef _SYS_SYSPROTO_H_
1771 struct undelete_args {
1772 	char *path;
1773 };
1774 #endif
1775 int
1776 sys_undelete(struct thread *td, struct undelete_args *uap)
1777 {
1778 	struct mount *mp;
1779 	struct nameidata nd;
1780 	int error;
1781 
1782 restart:
1783 	bwillwrite();
1784 	NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | AUDITVNODE1,
1785 	    UIO_USERSPACE, uap->path, td);
1786 	error = namei(&nd);
1787 	if (error != 0)
1788 		return (error);
1789 
1790 	if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) {
1791 		NDFREE(&nd, NDF_ONLY_PNBUF);
1792 		if (nd.ni_vp == nd.ni_dvp)
1793 			vrele(nd.ni_dvp);
1794 		else
1795 			vput(nd.ni_dvp);
1796 		if (nd.ni_vp)
1797 			vrele(nd.ni_vp);
1798 		return (EEXIST);
1799 	}
1800 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1801 		NDFREE(&nd, NDF_ONLY_PNBUF);
1802 		vput(nd.ni_dvp);
1803 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1804 			return (error);
1805 		goto restart;
1806 	}
1807 	error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE);
1808 	NDFREE(&nd, NDF_ONLY_PNBUF);
1809 	vput(nd.ni_dvp);
1810 	vn_finished_write(mp);
1811 	if (error == ERELOOKUP)
1812 		goto restart;
1813 	return (error);
1814 }
1815 
1816 /*
1817  * Delete a name from the filesystem.
1818  */
1819 #ifndef _SYS_SYSPROTO_H_
1820 struct unlink_args {
1821 	char	*path;
1822 };
1823 #endif
1824 int
1825 sys_unlink(struct thread *td, struct unlink_args *uap)
1826 {
1827 
1828 	return (kern_funlinkat(td, AT_FDCWD, uap->path, FD_NONE, UIO_USERSPACE,
1829 	    0, 0));
1830 }
1831 
1832 static int
1833 kern_funlinkat_ex(struct thread *td, int dfd, const char *path, int fd,
1834     int flag, enum uio_seg pathseg, ino_t oldinum)
1835 {
1836 
1837 	if ((flag & ~(AT_REMOVEDIR | AT_RESOLVE_BENEATH)) != 0)
1838 		return (EINVAL);
1839 
1840 	if ((flag & AT_REMOVEDIR) != 0)
1841 		return (kern_frmdirat(td, dfd, path, fd, UIO_USERSPACE, 0));
1842 
1843 	return (kern_funlinkat(td, dfd, path, fd, UIO_USERSPACE, 0, 0));
1844 }
1845 
1846 #ifndef _SYS_SYSPROTO_H_
1847 struct unlinkat_args {
1848 	int	fd;
1849 	char	*path;
1850 	int	flag;
1851 };
1852 #endif
1853 int
1854 sys_unlinkat(struct thread *td, struct unlinkat_args *uap)
1855 {
1856 
1857 	return (kern_funlinkat_ex(td, uap->fd, uap->path, FD_NONE, uap->flag,
1858 	    UIO_USERSPACE, 0));
1859 }
1860 
1861 #ifndef _SYS_SYSPROTO_H_
1862 struct funlinkat_args {
1863 	int		dfd;
1864 	const char	*path;
1865 	int		fd;
1866 	int		flag;
1867 };
1868 #endif
1869 int
1870 sys_funlinkat(struct thread *td, struct funlinkat_args *uap)
1871 {
1872 
1873 	return (kern_funlinkat_ex(td, uap->dfd, uap->path, uap->fd, uap->flag,
1874 	    UIO_USERSPACE, 0));
1875 }
1876 
1877 int
1878 kern_funlinkat(struct thread *td, int dfd, const char *path, int fd,
1879     enum uio_seg pathseg, int flag, ino_t oldinum)
1880 {
1881 	struct mount *mp;
1882 	struct file *fp;
1883 	struct vnode *vp;
1884 	struct nameidata nd;
1885 	struct stat sb;
1886 	int error;
1887 
1888 	fp = NULL;
1889 	if (fd != FD_NONE) {
1890 		error = getvnode_path(td, fd, &cap_no_rights, &fp);
1891 		if (error != 0)
1892 			return (error);
1893 	}
1894 
1895 restart:
1896 	bwillwrite();
1897 	NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1 |
1898 	    at2cnpflags(flag, AT_RESOLVE_BENEATH),
1899 	    pathseg, path, dfd, &cap_unlinkat_rights, td);
1900 	if ((error = namei(&nd)) != 0) {
1901 		if (error == EINVAL)
1902 			error = EPERM;
1903 		goto fdout;
1904 	}
1905 	vp = nd.ni_vp;
1906 	if (vp->v_type == VDIR && oldinum == 0) {
1907 		error = EPERM;		/* POSIX */
1908 	} else if (oldinum != 0 &&
1909 	    ((error = VOP_STAT(vp, &sb, td->td_ucred, NOCRED, td)) == 0) &&
1910 	    sb.st_ino != oldinum) {
1911 		error = EIDRM;	/* Identifier removed */
1912 	} else if (fp != NULL && fp->f_vnode != vp) {
1913 		if (VN_IS_DOOMED(fp->f_vnode))
1914 			error = EBADF;
1915 		else
1916 			error = EDEADLK;
1917 	} else {
1918 		/*
1919 		 * The root of a mounted filesystem cannot be deleted.
1920 		 *
1921 		 * XXX: can this only be a VDIR case?
1922 		 */
1923 		if (vp->v_vflag & VV_ROOT)
1924 			error = EBUSY;
1925 	}
1926 	if (error == 0) {
1927 		if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1928 			NDFREE(&nd, NDF_ONLY_PNBUF);
1929 			vput(nd.ni_dvp);
1930 			if (vp == nd.ni_dvp)
1931 				vrele(vp);
1932 			else
1933 				vput(vp);
1934 			if ((error = vn_start_write(NULL, &mp,
1935 			    V_XSLEEP | PCATCH)) != 0) {
1936 				goto fdout;
1937 			}
1938 			goto restart;
1939 		}
1940 #ifdef MAC
1941 		error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
1942 		    &nd.ni_cnd);
1943 		if (error != 0)
1944 			goto out;
1945 #endif
1946 		vfs_notify_upper(vp, VFS_NOTIFY_UPPER_UNLINK);
1947 		error = VOP_REMOVE(nd.ni_dvp, vp, &nd.ni_cnd);
1948 #ifdef MAC
1949 out:
1950 #endif
1951 		vn_finished_write(mp);
1952 	}
1953 	NDFREE(&nd, NDF_ONLY_PNBUF);
1954 	vput(nd.ni_dvp);
1955 	if (vp == nd.ni_dvp)
1956 		vrele(vp);
1957 	else
1958 		vput(vp);
1959 	if (error == ERELOOKUP)
1960 		goto restart;
1961 fdout:
1962 	if (fp != NULL)
1963 		fdrop(fp, td);
1964 	return (error);
1965 }
1966 
1967 /*
1968  * Reposition read/write file offset.
1969  */
1970 #ifndef _SYS_SYSPROTO_H_
1971 struct lseek_args {
1972 	int	fd;
1973 	int	pad;
1974 	off_t	offset;
1975 	int	whence;
1976 };
1977 #endif
1978 int
1979 sys_lseek(struct thread *td, struct lseek_args *uap)
1980 {
1981 
1982 	return (kern_lseek(td, uap->fd, uap->offset, uap->whence));
1983 }
1984 
1985 int
1986 kern_lseek(struct thread *td, int fd, off_t offset, int whence)
1987 {
1988 	struct file *fp;
1989 	int error;
1990 
1991 	AUDIT_ARG_FD(fd);
1992 	error = fget(td, fd, &cap_seek_rights, &fp);
1993 	if (error != 0)
1994 		return (error);
1995 	error = (fp->f_ops->fo_flags & DFLAG_SEEKABLE) != 0 ?
1996 	    fo_seek(fp, offset, whence, td) : ESPIPE;
1997 	fdrop(fp, td);
1998 	return (error);
1999 }
2000 
2001 #if defined(COMPAT_43)
2002 /*
2003  * Reposition read/write file offset.
2004  */
2005 #ifndef _SYS_SYSPROTO_H_
2006 struct olseek_args {
2007 	int	fd;
2008 	long	offset;
2009 	int	whence;
2010 };
2011 #endif
2012 int
2013 olseek(struct thread *td, struct olseek_args *uap)
2014 {
2015 
2016 	return (kern_lseek(td, uap->fd, uap->offset, uap->whence));
2017 }
2018 #endif /* COMPAT_43 */
2019 
2020 #if defined(COMPAT_FREEBSD6)
2021 /* Version with the 'pad' argument */
2022 int
2023 freebsd6_lseek(struct thread *td, struct freebsd6_lseek_args *uap)
2024 {
2025 
2026 	return (kern_lseek(td, uap->fd, uap->offset, uap->whence));
2027 }
2028 #endif
2029 
2030 /*
2031  * Check access permissions using passed credentials.
2032  */
2033 static int
2034 vn_access(struct vnode *vp, int user_flags, struct ucred *cred,
2035      struct thread *td)
2036 {
2037 	accmode_t accmode;
2038 	int error;
2039 
2040 	/* Flags == 0 means only check for existence. */
2041 	if (user_flags == 0)
2042 		return (0);
2043 
2044 	accmode = 0;
2045 	if (user_flags & R_OK)
2046 		accmode |= VREAD;
2047 	if (user_flags & W_OK)
2048 		accmode |= VWRITE;
2049 	if (user_flags & X_OK)
2050 		accmode |= VEXEC;
2051 #ifdef MAC
2052 	error = mac_vnode_check_access(cred, vp, accmode);
2053 	if (error != 0)
2054 		return (error);
2055 #endif
2056 	if ((accmode & VWRITE) == 0 || (error = vn_writechk(vp)) == 0)
2057 		error = VOP_ACCESS(vp, accmode, cred, td);
2058 	return (error);
2059 }
2060 
2061 /*
2062  * Check access permissions using "real" credentials.
2063  */
2064 #ifndef _SYS_SYSPROTO_H_
2065 struct access_args {
2066 	char	*path;
2067 	int	amode;
2068 };
2069 #endif
2070 int
2071 sys_access(struct thread *td, struct access_args *uap)
2072 {
2073 
2074 	return (kern_accessat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2075 	    0, uap->amode));
2076 }
2077 
2078 #ifndef _SYS_SYSPROTO_H_
2079 struct faccessat_args {
2080 	int	dirfd;
2081 	char	*path;
2082 	int	amode;
2083 	int	flag;
2084 }
2085 #endif
2086 int
2087 sys_faccessat(struct thread *td, struct faccessat_args *uap)
2088 {
2089 
2090 	return (kern_accessat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
2091 	    uap->amode));
2092 }
2093 
2094 int
2095 kern_accessat(struct thread *td, int fd, const char *path,
2096     enum uio_seg pathseg, int flag, int amode)
2097 {
2098 	struct ucred *cred, *usecred;
2099 	struct vnode *vp;
2100 	struct nameidata nd;
2101 	int error;
2102 
2103 	if ((flag & ~(AT_EACCESS | AT_RESOLVE_BENEATH | AT_EMPTY_PATH)) != 0)
2104 		return (EINVAL);
2105 	if (amode != F_OK && (amode & ~(R_OK | W_OK | X_OK)) != 0)
2106 		return (EINVAL);
2107 
2108 	/*
2109 	 * Create and modify a temporary credential instead of one that
2110 	 * is potentially shared (if we need one).
2111 	 */
2112 	cred = td->td_ucred;
2113 	if ((flag & AT_EACCESS) == 0 &&
2114 	    ((cred->cr_uid != cred->cr_ruid ||
2115 	    cred->cr_rgid != cred->cr_groups[0]))) {
2116 		usecred = crdup(cred);
2117 		usecred->cr_uid = cred->cr_ruid;
2118 		usecred->cr_groups[0] = cred->cr_rgid;
2119 		td->td_ucred = usecred;
2120 	} else
2121 		usecred = cred;
2122 	AUDIT_ARG_VALUE(amode);
2123 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF |
2124 	    AUDITVNODE1 | at2cnpflags(flag, AT_RESOLVE_BENEATH |
2125 	    AT_EMPTY_PATH), pathseg, path, fd, &cap_fstat_rights, td);
2126 	if ((error = namei(&nd)) != 0)
2127 		goto out;
2128 	vp = nd.ni_vp;
2129 
2130 	error = vn_access(vp, amode, usecred, td);
2131 	NDFREE_NOTHING(&nd);
2132 	vput(vp);
2133 out:
2134 	if (usecred != cred) {
2135 		td->td_ucred = cred;
2136 		crfree(usecred);
2137 	}
2138 	return (error);
2139 }
2140 
2141 /*
2142  * Check access permissions using "effective" credentials.
2143  */
2144 #ifndef _SYS_SYSPROTO_H_
2145 struct eaccess_args {
2146 	char	*path;
2147 	int	amode;
2148 };
2149 #endif
2150 int
2151 sys_eaccess(struct thread *td, struct eaccess_args *uap)
2152 {
2153 
2154 	return (kern_accessat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2155 	    AT_EACCESS, uap->amode));
2156 }
2157 
2158 #if defined(COMPAT_43)
2159 /*
2160  * Get file status; this version follows links.
2161  */
2162 #ifndef _SYS_SYSPROTO_H_
2163 struct ostat_args {
2164 	char	*path;
2165 	struct ostat *ub;
2166 };
2167 #endif
2168 int
2169 ostat(struct thread *td, struct ostat_args *uap)
2170 {
2171 	struct stat sb;
2172 	struct ostat osb;
2173 	int error;
2174 
2175 	error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE,
2176 	    &sb, NULL);
2177 	if (error != 0)
2178 		return (error);
2179 	cvtstat(&sb, &osb);
2180 	return (copyout(&osb, uap->ub, sizeof (osb)));
2181 }
2182 
2183 /*
2184  * Get file status; this version does not follow links.
2185  */
2186 #ifndef _SYS_SYSPROTO_H_
2187 struct olstat_args {
2188 	char	*path;
2189 	struct ostat *ub;
2190 };
2191 #endif
2192 int
2193 olstat(struct thread *td, struct olstat_args *uap)
2194 {
2195 	struct stat sb;
2196 	struct ostat osb;
2197 	int error;
2198 
2199 	error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2200 	    UIO_USERSPACE, &sb, NULL);
2201 	if (error != 0)
2202 		return (error);
2203 	cvtstat(&sb, &osb);
2204 	return (copyout(&osb, uap->ub, sizeof (osb)));
2205 }
2206 
2207 /*
2208  * Convert from an old to a new stat structure.
2209  * XXX: many values are blindly truncated.
2210  */
2211 void
2212 cvtstat(struct stat *st, struct ostat *ost)
2213 {
2214 
2215 	bzero(ost, sizeof(*ost));
2216 	ost->st_dev = st->st_dev;
2217 	ost->st_ino = st->st_ino;
2218 	ost->st_mode = st->st_mode;
2219 	ost->st_nlink = st->st_nlink;
2220 	ost->st_uid = st->st_uid;
2221 	ost->st_gid = st->st_gid;
2222 	ost->st_rdev = st->st_rdev;
2223 	ost->st_size = MIN(st->st_size, INT32_MAX);
2224 	ost->st_atim = st->st_atim;
2225 	ost->st_mtim = st->st_mtim;
2226 	ost->st_ctim = st->st_ctim;
2227 	ost->st_blksize = st->st_blksize;
2228 	ost->st_blocks = st->st_blocks;
2229 	ost->st_flags = st->st_flags;
2230 	ost->st_gen = st->st_gen;
2231 }
2232 #endif /* COMPAT_43 */
2233 
2234 #if defined(COMPAT_43) || defined(COMPAT_FREEBSD11)
2235 int ino64_trunc_error;
2236 SYSCTL_INT(_vfs, OID_AUTO, ino64_trunc_error, CTLFLAG_RW,
2237     &ino64_trunc_error, 0,
2238     "Error on truncation of device, file or inode number, or link count");
2239 
2240 int
2241 freebsd11_cvtstat(struct stat *st, struct freebsd11_stat *ost)
2242 {
2243 
2244 	ost->st_dev = st->st_dev;
2245 	if (ost->st_dev != st->st_dev) {
2246 		switch (ino64_trunc_error) {
2247 		default:
2248 			/*
2249 			 * Since dev_t is almost raw, don't clamp to the
2250 			 * maximum for case 2, but ignore the error.
2251 			 */
2252 			break;
2253 		case 1:
2254 			return (EOVERFLOW);
2255 		}
2256 	}
2257 	ost->st_ino = st->st_ino;
2258 	if (ost->st_ino != st->st_ino) {
2259 		switch (ino64_trunc_error) {
2260 		default:
2261 		case 0:
2262 			break;
2263 		case 1:
2264 			return (EOVERFLOW);
2265 		case 2:
2266 			ost->st_ino = UINT32_MAX;
2267 			break;
2268 		}
2269 	}
2270 	ost->st_mode = st->st_mode;
2271 	ost->st_nlink = st->st_nlink;
2272 	if (ost->st_nlink != st->st_nlink) {
2273 		switch (ino64_trunc_error) {
2274 		default:
2275 		case 0:
2276 			break;
2277 		case 1:
2278 			return (EOVERFLOW);
2279 		case 2:
2280 			ost->st_nlink = UINT16_MAX;
2281 			break;
2282 		}
2283 	}
2284 	ost->st_uid = st->st_uid;
2285 	ost->st_gid = st->st_gid;
2286 	ost->st_rdev = st->st_rdev;
2287 	if (ost->st_rdev != st->st_rdev) {
2288 		switch (ino64_trunc_error) {
2289 		default:
2290 			break;
2291 		case 1:
2292 			return (EOVERFLOW);
2293 		}
2294 	}
2295 	ost->st_atim = st->st_atim;
2296 	ost->st_mtim = st->st_mtim;
2297 	ost->st_ctim = st->st_ctim;
2298 	ost->st_size = st->st_size;
2299 	ost->st_blocks = st->st_blocks;
2300 	ost->st_blksize = st->st_blksize;
2301 	ost->st_flags = st->st_flags;
2302 	ost->st_gen = st->st_gen;
2303 	ost->st_lspare = 0;
2304 	ost->st_birthtim = st->st_birthtim;
2305 	bzero((char *)&ost->st_birthtim + sizeof(ost->st_birthtim),
2306 	    sizeof(*ost) - offsetof(struct freebsd11_stat,
2307 	    st_birthtim) - sizeof(ost->st_birthtim));
2308 	return (0);
2309 }
2310 
2311 int
2312 freebsd11_stat(struct thread *td, struct freebsd11_stat_args* uap)
2313 {
2314 	struct stat sb;
2315 	struct freebsd11_stat osb;
2316 	int error;
2317 
2318 	error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE,
2319 	    &sb, NULL);
2320 	if (error != 0)
2321 		return (error);
2322 	error = freebsd11_cvtstat(&sb, &osb);
2323 	if (error == 0)
2324 		error = copyout(&osb, uap->ub, sizeof(osb));
2325 	return (error);
2326 }
2327 
2328 int
2329 freebsd11_lstat(struct thread *td, struct freebsd11_lstat_args* uap)
2330 {
2331 	struct stat sb;
2332 	struct freebsd11_stat osb;
2333 	int error;
2334 
2335 	error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2336 	    UIO_USERSPACE, &sb, NULL);
2337 	if (error != 0)
2338 		return (error);
2339 	error = freebsd11_cvtstat(&sb, &osb);
2340 	if (error == 0)
2341 		error = copyout(&osb, uap->ub, sizeof(osb));
2342 	return (error);
2343 }
2344 
2345 int
2346 freebsd11_fhstat(struct thread *td, struct freebsd11_fhstat_args* uap)
2347 {
2348 	struct fhandle fh;
2349 	struct stat sb;
2350 	struct freebsd11_stat osb;
2351 	int error;
2352 
2353 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
2354 	if (error != 0)
2355 		return (error);
2356 	error = kern_fhstat(td, fh, &sb);
2357 	if (error != 0)
2358 		return (error);
2359 	error = freebsd11_cvtstat(&sb, &osb);
2360 	if (error == 0)
2361 		error = copyout(&osb, uap->sb, sizeof(osb));
2362 	return (error);
2363 }
2364 
2365 int
2366 freebsd11_fstatat(struct thread *td, struct freebsd11_fstatat_args* uap)
2367 {
2368 	struct stat sb;
2369 	struct freebsd11_stat osb;
2370 	int error;
2371 
2372 	error = kern_statat(td, uap->flag, uap->fd, uap->path,
2373 	    UIO_USERSPACE, &sb, NULL);
2374 	if (error != 0)
2375 		return (error);
2376 	error = freebsd11_cvtstat(&sb, &osb);
2377 	if (error == 0)
2378 		error = copyout(&osb, uap->buf, sizeof(osb));
2379 	return (error);
2380 }
2381 #endif	/* COMPAT_FREEBSD11 */
2382 
2383 /*
2384  * Get file status
2385  */
2386 #ifndef _SYS_SYSPROTO_H_
2387 struct fstatat_args {
2388 	int	fd;
2389 	char	*path;
2390 	struct stat	*buf;
2391 	int	flag;
2392 }
2393 #endif
2394 int
2395 sys_fstatat(struct thread *td, struct fstatat_args *uap)
2396 {
2397 	struct stat sb;
2398 	int error;
2399 
2400 	error = kern_statat(td, uap->flag, uap->fd, uap->path,
2401 	    UIO_USERSPACE, &sb, NULL);
2402 	if (error == 0)
2403 		error = copyout(&sb, uap->buf, sizeof (sb));
2404 	return (error);
2405 }
2406 
2407 int
2408 kern_statat(struct thread *td, int flag, int fd, const char *path,
2409     enum uio_seg pathseg, struct stat *sbp,
2410     void (*hook)(struct vnode *vp, struct stat *sbp))
2411 {
2412 	struct nameidata nd;
2413 	int error;
2414 
2415 	if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH |
2416 	    AT_EMPTY_PATH)) != 0)
2417 		return (EINVAL);
2418 
2419 	NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(flag, AT_RESOLVE_BENEATH |
2420 	    AT_SYMLINK_NOFOLLOW | AT_EMPTY_PATH) | LOCKSHARED | LOCKLEAF |
2421 	    AUDITVNODE1, pathseg, path, fd, &cap_fstat_rights, td);
2422 
2423 	if ((error = namei(&nd)) != 0)
2424 		return (error);
2425 	error = VOP_STAT(nd.ni_vp, sbp, td->td_ucred, NOCRED, td);
2426 	if (error == 0) {
2427 		if (__predict_false(hook != NULL))
2428 			hook(nd.ni_vp, sbp);
2429 	}
2430 	NDFREE_NOTHING(&nd);
2431 	vput(nd.ni_vp);
2432 #ifdef __STAT_TIME_T_EXT
2433 	sbp->st_atim_ext = 0;
2434 	sbp->st_mtim_ext = 0;
2435 	sbp->st_ctim_ext = 0;
2436 	sbp->st_btim_ext = 0;
2437 #endif
2438 #ifdef KTRACE
2439 	if (KTRPOINT(td, KTR_STRUCT))
2440 		ktrstat_error(sbp, error);
2441 #endif
2442 	return (error);
2443 }
2444 
2445 #if defined(COMPAT_FREEBSD11)
2446 /*
2447  * Implementation of the NetBSD [l]stat() functions.
2448  */
2449 void
2450 freebsd11_cvtnstat(struct stat *sb, struct nstat *nsb)
2451 {
2452 
2453 	bzero(nsb, sizeof(*nsb));
2454 	nsb->st_dev = sb->st_dev;
2455 	nsb->st_ino = sb->st_ino;
2456 	nsb->st_mode = sb->st_mode;
2457 	nsb->st_nlink = sb->st_nlink;
2458 	nsb->st_uid = sb->st_uid;
2459 	nsb->st_gid = sb->st_gid;
2460 	nsb->st_rdev = sb->st_rdev;
2461 	nsb->st_atim = sb->st_atim;
2462 	nsb->st_mtim = sb->st_mtim;
2463 	nsb->st_ctim = sb->st_ctim;
2464 	nsb->st_size = sb->st_size;
2465 	nsb->st_blocks = sb->st_blocks;
2466 	nsb->st_blksize = sb->st_blksize;
2467 	nsb->st_flags = sb->st_flags;
2468 	nsb->st_gen = sb->st_gen;
2469 	nsb->st_birthtim = sb->st_birthtim;
2470 }
2471 
2472 #ifndef _SYS_SYSPROTO_H_
2473 struct freebsd11_nstat_args {
2474 	char	*path;
2475 	struct nstat *ub;
2476 };
2477 #endif
2478 int
2479 freebsd11_nstat(struct thread *td, struct freebsd11_nstat_args *uap)
2480 {
2481 	struct stat sb;
2482 	struct nstat nsb;
2483 	int error;
2484 
2485 	error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE,
2486 	    &sb, NULL);
2487 	if (error != 0)
2488 		return (error);
2489 	freebsd11_cvtnstat(&sb, &nsb);
2490 	return (copyout(&nsb, uap->ub, sizeof (nsb)));
2491 }
2492 
2493 /*
2494  * NetBSD lstat.  Get file status; this version does not follow links.
2495  */
2496 #ifndef _SYS_SYSPROTO_H_
2497 struct freebsd11_nlstat_args {
2498 	char	*path;
2499 	struct nstat *ub;
2500 };
2501 #endif
2502 int
2503 freebsd11_nlstat(struct thread *td, struct freebsd11_nlstat_args *uap)
2504 {
2505 	struct stat sb;
2506 	struct nstat nsb;
2507 	int error;
2508 
2509 	error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2510 	    UIO_USERSPACE, &sb, NULL);
2511 	if (error != 0)
2512 		return (error);
2513 	freebsd11_cvtnstat(&sb, &nsb);
2514 	return (copyout(&nsb, uap->ub, sizeof (nsb)));
2515 }
2516 #endif /* COMPAT_FREEBSD11 */
2517 
2518 /*
2519  * Get configurable pathname variables.
2520  */
2521 #ifndef _SYS_SYSPROTO_H_
2522 struct pathconf_args {
2523 	char	*path;
2524 	int	name;
2525 };
2526 #endif
2527 int
2528 sys_pathconf(struct thread *td, struct pathconf_args *uap)
2529 {
2530 	long value;
2531 	int error;
2532 
2533 	error = kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name, FOLLOW,
2534 	    &value);
2535 	if (error == 0)
2536 		td->td_retval[0] = value;
2537 	return (error);
2538 }
2539 
2540 #ifndef _SYS_SYSPROTO_H_
2541 struct lpathconf_args {
2542 	char	*path;
2543 	int	name;
2544 };
2545 #endif
2546 int
2547 sys_lpathconf(struct thread *td, struct lpathconf_args *uap)
2548 {
2549 	long value;
2550 	int error;
2551 
2552 	error = kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name,
2553 	    NOFOLLOW, &value);
2554 	if (error == 0)
2555 		td->td_retval[0] = value;
2556 	return (error);
2557 }
2558 
2559 int
2560 kern_pathconf(struct thread *td, const char *path, enum uio_seg pathseg,
2561     int name, u_long flags, long *valuep)
2562 {
2563 	struct nameidata nd;
2564 	int error;
2565 
2566 	NDINIT(&nd, LOOKUP, LOCKSHARED | LOCKLEAF | AUDITVNODE1 | flags,
2567 	    pathseg, path, td);
2568 	if ((error = namei(&nd)) != 0)
2569 		return (error);
2570 	NDFREE_NOTHING(&nd);
2571 
2572 	error = VOP_PATHCONF(nd.ni_vp, name, valuep);
2573 	vput(nd.ni_vp);
2574 	return (error);
2575 }
2576 
2577 /*
2578  * Return target name of a symbolic link.
2579  */
2580 #ifndef _SYS_SYSPROTO_H_
2581 struct readlink_args {
2582 	char	*path;
2583 	char	*buf;
2584 	size_t	count;
2585 };
2586 #endif
2587 int
2588 sys_readlink(struct thread *td, struct readlink_args *uap)
2589 {
2590 
2591 	return (kern_readlinkat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2592 	    uap->buf, UIO_USERSPACE, uap->count));
2593 }
2594 #ifndef _SYS_SYSPROTO_H_
2595 struct readlinkat_args {
2596 	int	fd;
2597 	char	*path;
2598 	char	*buf;
2599 	size_t	bufsize;
2600 };
2601 #endif
2602 int
2603 sys_readlinkat(struct thread *td, struct readlinkat_args *uap)
2604 {
2605 
2606 	return (kern_readlinkat(td, uap->fd, uap->path, UIO_USERSPACE,
2607 	    uap->buf, UIO_USERSPACE, uap->bufsize));
2608 }
2609 
2610 int
2611 kern_readlinkat(struct thread *td, int fd, const char *path,
2612     enum uio_seg pathseg, char *buf, enum uio_seg bufseg, size_t count)
2613 {
2614 	struct vnode *vp;
2615 	struct nameidata nd;
2616 	int error;
2617 
2618 	if (count > IOSIZE_MAX)
2619 		return (EINVAL);
2620 
2621 	NDINIT_AT(&nd, LOOKUP, NOFOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
2622 	    pathseg, path, fd, td);
2623 
2624 	if ((error = namei(&nd)) != 0)
2625 		return (error);
2626 	NDFREE_NOTHING(&nd);
2627 	vp = nd.ni_vp;
2628 
2629 	error = kern_readlink_vp(vp, buf, bufseg, count, td);
2630 	vput(vp);
2631 
2632 	return (error);
2633 }
2634 
2635 /*
2636  * Helper function to readlink from a vnode
2637  */
2638 static int
2639 kern_readlink_vp(struct vnode *vp, char *buf, enum uio_seg bufseg, size_t count,
2640     struct thread *td)
2641 {
2642 	struct iovec aiov;
2643 	struct uio auio;
2644 	int error;
2645 
2646 	ASSERT_VOP_LOCKED(vp, "kern_readlink_vp(): vp not locked");
2647 #ifdef MAC
2648 	error = mac_vnode_check_readlink(td->td_ucred, vp);
2649 	if (error != 0)
2650 		return (error);
2651 #endif
2652 	if (vp->v_type != VLNK && (vp->v_vflag & VV_READLINK) == 0)
2653 		return (EINVAL);
2654 
2655 	aiov.iov_base = buf;
2656 	aiov.iov_len = count;
2657 	auio.uio_iov = &aiov;
2658 	auio.uio_iovcnt = 1;
2659 	auio.uio_offset = 0;
2660 	auio.uio_rw = UIO_READ;
2661 	auio.uio_segflg = bufseg;
2662 	auio.uio_td = td;
2663 	auio.uio_resid = count;
2664 	error = VOP_READLINK(vp, &auio, td->td_ucred);
2665 	td->td_retval[0] = count - auio.uio_resid;
2666 	return (error);
2667 }
2668 
2669 /*
2670  * Common implementation code for chflags() and fchflags().
2671  */
2672 static int
2673 setfflags(struct thread *td, struct vnode *vp, u_long flags)
2674 {
2675 	struct mount *mp;
2676 	struct vattr vattr;
2677 	int error;
2678 
2679 	/* We can't support the value matching VNOVAL. */
2680 	if (flags == VNOVAL)
2681 		return (EOPNOTSUPP);
2682 
2683 	/*
2684 	 * Prevent non-root users from setting flags on devices.  When
2685 	 * a device is reused, users can retain ownership of the device
2686 	 * if they are allowed to set flags and programs assume that
2687 	 * chown can't fail when done as root.
2688 	 */
2689 	if (vp->v_type == VCHR || vp->v_type == VBLK) {
2690 		error = priv_check(td, PRIV_VFS_CHFLAGS_DEV);
2691 		if (error != 0)
2692 			return (error);
2693 	}
2694 
2695 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2696 		return (error);
2697 	VATTR_NULL(&vattr);
2698 	vattr.va_flags = flags;
2699 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2700 #ifdef MAC
2701 	error = mac_vnode_check_setflags(td->td_ucred, vp, vattr.va_flags);
2702 	if (error == 0)
2703 #endif
2704 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
2705 	VOP_UNLOCK(vp);
2706 	vn_finished_write(mp);
2707 	return (error);
2708 }
2709 
2710 /*
2711  * Change flags of a file given a path name.
2712  */
2713 #ifndef _SYS_SYSPROTO_H_
2714 struct chflags_args {
2715 	const char *path;
2716 	u_long	flags;
2717 };
2718 #endif
2719 int
2720 sys_chflags(struct thread *td, struct chflags_args *uap)
2721 {
2722 
2723 	return (kern_chflagsat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2724 	    uap->flags, 0));
2725 }
2726 
2727 #ifndef _SYS_SYSPROTO_H_
2728 struct chflagsat_args {
2729 	int	fd;
2730 	const char *path;
2731 	u_long	flags;
2732 	int	atflag;
2733 }
2734 #endif
2735 int
2736 sys_chflagsat(struct thread *td, struct chflagsat_args *uap)
2737 {
2738 
2739 	if ((uap->atflag & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH |
2740 	    AT_EMPTY_PATH)) != 0)
2741 		return (EINVAL);
2742 
2743 	return (kern_chflagsat(td, uap->fd, uap->path, UIO_USERSPACE,
2744 	    uap->flags, uap->atflag));
2745 }
2746 
2747 /*
2748  * Same as chflags() but doesn't follow symlinks.
2749  */
2750 #ifndef _SYS_SYSPROTO_H_
2751 struct lchflags_args {
2752 	const char *path;
2753 	u_long flags;
2754 };
2755 #endif
2756 int
2757 sys_lchflags(struct thread *td, struct lchflags_args *uap)
2758 {
2759 
2760 	return (kern_chflagsat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2761 	    uap->flags, AT_SYMLINK_NOFOLLOW));
2762 }
2763 
2764 static int
2765 kern_chflagsat(struct thread *td, int fd, const char *path,
2766     enum uio_seg pathseg, u_long flags, int atflag)
2767 {
2768 	struct nameidata nd;
2769 	int error;
2770 
2771 	AUDIT_ARG_FFLAGS(flags);
2772 	NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(atflag, AT_SYMLINK_NOFOLLOW |
2773 	    AT_RESOLVE_BENEATH | AT_EMPTY_PATH) | AUDITVNODE1, pathseg, path,
2774 	    fd, &cap_fchflags_rights, td);
2775 	if ((error = namei(&nd)) != 0)
2776 		return (error);
2777 	NDFREE_NOTHING(&nd);
2778 	error = setfflags(td, nd.ni_vp, flags);
2779 	vrele(nd.ni_vp);
2780 	return (error);
2781 }
2782 
2783 /*
2784  * Change flags of a file given a file descriptor.
2785  */
2786 #ifndef _SYS_SYSPROTO_H_
2787 struct fchflags_args {
2788 	int	fd;
2789 	u_long	flags;
2790 };
2791 #endif
2792 int
2793 sys_fchflags(struct thread *td, struct fchflags_args *uap)
2794 {
2795 	struct file *fp;
2796 	int error;
2797 
2798 	AUDIT_ARG_FD(uap->fd);
2799 	AUDIT_ARG_FFLAGS(uap->flags);
2800 	error = getvnode(td, uap->fd, &cap_fchflags_rights,
2801 	    &fp);
2802 	if (error != 0)
2803 		return (error);
2804 #ifdef AUDIT
2805 	vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
2806 	AUDIT_ARG_VNODE1(fp->f_vnode);
2807 	VOP_UNLOCK(fp->f_vnode);
2808 #endif
2809 	error = setfflags(td, fp->f_vnode, uap->flags);
2810 	fdrop(fp, td);
2811 	return (error);
2812 }
2813 
2814 /*
2815  * Common implementation code for chmod(), lchmod() and fchmod().
2816  */
2817 int
2818 setfmode(struct thread *td, struct ucred *cred, struct vnode *vp, int mode)
2819 {
2820 	struct mount *mp;
2821 	struct vattr vattr;
2822 	int error;
2823 
2824 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2825 		return (error);
2826 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2827 	VATTR_NULL(&vattr);
2828 	vattr.va_mode = mode & ALLPERMS;
2829 #ifdef MAC
2830 	error = mac_vnode_check_setmode(cred, vp, vattr.va_mode);
2831 	if (error == 0)
2832 #endif
2833 		error = VOP_SETATTR(vp, &vattr, cred);
2834 	VOP_UNLOCK(vp);
2835 	vn_finished_write(mp);
2836 	return (error);
2837 }
2838 
2839 /*
2840  * Change mode of a file given path name.
2841  */
2842 #ifndef _SYS_SYSPROTO_H_
2843 struct chmod_args {
2844 	char	*path;
2845 	int	mode;
2846 };
2847 #endif
2848 int
2849 sys_chmod(struct thread *td, struct chmod_args *uap)
2850 {
2851 
2852 	return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2853 	    uap->mode, 0));
2854 }
2855 
2856 #ifndef _SYS_SYSPROTO_H_
2857 struct fchmodat_args {
2858 	int	dirfd;
2859 	char	*path;
2860 	mode_t	mode;
2861 	int	flag;
2862 }
2863 #endif
2864 int
2865 sys_fchmodat(struct thread *td, struct fchmodat_args *uap)
2866 {
2867 
2868 	if ((uap->flag & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH |
2869 	    AT_EMPTY_PATH)) != 0)
2870 		return (EINVAL);
2871 
2872 	return (kern_fchmodat(td, uap->fd, uap->path, UIO_USERSPACE,
2873 	    uap->mode, uap->flag));
2874 }
2875 
2876 /*
2877  * Change mode of a file given path name (don't follow links.)
2878  */
2879 #ifndef _SYS_SYSPROTO_H_
2880 struct lchmod_args {
2881 	char	*path;
2882 	int	mode;
2883 };
2884 #endif
2885 int
2886 sys_lchmod(struct thread *td, struct lchmod_args *uap)
2887 {
2888 
2889 	return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2890 	    uap->mode, AT_SYMLINK_NOFOLLOW));
2891 }
2892 
2893 int
2894 kern_fchmodat(struct thread *td, int fd, const char *path,
2895     enum uio_seg pathseg, mode_t mode, int flag)
2896 {
2897 	struct nameidata nd;
2898 	int error;
2899 
2900 	AUDIT_ARG_MODE(mode);
2901 	NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(flag, AT_SYMLINK_NOFOLLOW |
2902 	    AT_RESOLVE_BENEATH | AT_EMPTY_PATH) | AUDITVNODE1, pathseg, path,
2903 	    fd, &cap_fchmod_rights, td);
2904 	if ((error = namei(&nd)) != 0)
2905 		return (error);
2906 	NDFREE_NOTHING(&nd);
2907 	error = setfmode(td, td->td_ucred, nd.ni_vp, mode);
2908 	vrele(nd.ni_vp);
2909 	return (error);
2910 }
2911 
2912 /*
2913  * Change mode of a file given a file descriptor.
2914  */
2915 #ifndef _SYS_SYSPROTO_H_
2916 struct fchmod_args {
2917 	int	fd;
2918 	int	mode;
2919 };
2920 #endif
2921 int
2922 sys_fchmod(struct thread *td, struct fchmod_args *uap)
2923 {
2924 	struct file *fp;
2925 	int error;
2926 
2927 	AUDIT_ARG_FD(uap->fd);
2928 	AUDIT_ARG_MODE(uap->mode);
2929 
2930 	error = fget(td, uap->fd, &cap_fchmod_rights, &fp);
2931 	if (error != 0)
2932 		return (error);
2933 	error = fo_chmod(fp, uap->mode, td->td_ucred, td);
2934 	fdrop(fp, td);
2935 	return (error);
2936 }
2937 
2938 /*
2939  * Common implementation for chown(), lchown(), and fchown()
2940  */
2941 int
2942 setfown(struct thread *td, struct ucred *cred, struct vnode *vp, uid_t uid,
2943     gid_t gid)
2944 {
2945 	struct mount *mp;
2946 	struct vattr vattr;
2947 	int error;
2948 
2949 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2950 		return (error);
2951 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2952 	VATTR_NULL(&vattr);
2953 	vattr.va_uid = uid;
2954 	vattr.va_gid = gid;
2955 #ifdef MAC
2956 	error = mac_vnode_check_setowner(cred, vp, vattr.va_uid,
2957 	    vattr.va_gid);
2958 	if (error == 0)
2959 #endif
2960 		error = VOP_SETATTR(vp, &vattr, cred);
2961 	VOP_UNLOCK(vp);
2962 	vn_finished_write(mp);
2963 	return (error);
2964 }
2965 
2966 /*
2967  * Set ownership given a path name.
2968  */
2969 #ifndef _SYS_SYSPROTO_H_
2970 struct chown_args {
2971 	char	*path;
2972 	int	uid;
2973 	int	gid;
2974 };
2975 #endif
2976 int
2977 sys_chown(struct thread *td, struct chown_args *uap)
2978 {
2979 
2980 	return (kern_fchownat(td, AT_FDCWD, uap->path, UIO_USERSPACE, uap->uid,
2981 	    uap->gid, 0));
2982 }
2983 
2984 #ifndef _SYS_SYSPROTO_H_
2985 struct fchownat_args {
2986 	int fd;
2987 	const char * path;
2988 	uid_t uid;
2989 	gid_t gid;
2990 	int flag;
2991 };
2992 #endif
2993 int
2994 sys_fchownat(struct thread *td, struct fchownat_args *uap)
2995 {
2996 
2997 	if ((uap->flag & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH |
2998 	    AT_EMPTY_PATH)) != 0)
2999 		return (EINVAL);
3000 
3001 	return (kern_fchownat(td, uap->fd, uap->path, UIO_USERSPACE, uap->uid,
3002 	    uap->gid, uap->flag));
3003 }
3004 
3005 int
3006 kern_fchownat(struct thread *td, int fd, const char *path,
3007     enum uio_seg pathseg, int uid, int gid, int flag)
3008 {
3009 	struct nameidata nd;
3010 	int error;
3011 
3012 	AUDIT_ARG_OWNER(uid, gid);
3013 	NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(flag, AT_SYMLINK_NOFOLLOW |
3014 	    AT_RESOLVE_BENEATH | AT_EMPTY_PATH) | AUDITVNODE1, pathseg, path,
3015 	    fd, &cap_fchown_rights, td);
3016 
3017 	if ((error = namei(&nd)) != 0)
3018 		return (error);
3019 	NDFREE_NOTHING(&nd);
3020 	error = setfown(td, td->td_ucred, nd.ni_vp, uid, gid);
3021 	vrele(nd.ni_vp);
3022 	return (error);
3023 }
3024 
3025 /*
3026  * Set ownership given a path name, do not cross symlinks.
3027  */
3028 #ifndef _SYS_SYSPROTO_H_
3029 struct lchown_args {
3030 	char	*path;
3031 	int	uid;
3032 	int	gid;
3033 };
3034 #endif
3035 int
3036 sys_lchown(struct thread *td, struct lchown_args *uap)
3037 {
3038 
3039 	return (kern_fchownat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
3040 	    uap->uid, uap->gid, AT_SYMLINK_NOFOLLOW));
3041 }
3042 
3043 /*
3044  * Set ownership given a file descriptor.
3045  */
3046 #ifndef _SYS_SYSPROTO_H_
3047 struct fchown_args {
3048 	int	fd;
3049 	int	uid;
3050 	int	gid;
3051 };
3052 #endif
3053 int
3054 sys_fchown(struct thread *td, struct fchown_args *uap)
3055 {
3056 	struct file *fp;
3057 	int error;
3058 
3059 	AUDIT_ARG_FD(uap->fd);
3060 	AUDIT_ARG_OWNER(uap->uid, uap->gid);
3061 	error = fget(td, uap->fd, &cap_fchown_rights, &fp);
3062 	if (error != 0)
3063 		return (error);
3064 	error = fo_chown(fp, uap->uid, uap->gid, td->td_ucred, td);
3065 	fdrop(fp, td);
3066 	return (error);
3067 }
3068 
3069 /*
3070  * Common implementation code for utimes(), lutimes(), and futimes().
3071  */
3072 static int
3073 getutimes(const struct timeval *usrtvp, enum uio_seg tvpseg,
3074     struct timespec *tsp)
3075 {
3076 	struct timeval tv[2];
3077 	const struct timeval *tvp;
3078 	int error;
3079 
3080 	if (usrtvp == NULL) {
3081 		vfs_timestamp(&tsp[0]);
3082 		tsp[1] = tsp[0];
3083 	} else {
3084 		if (tvpseg == UIO_SYSSPACE) {
3085 			tvp = usrtvp;
3086 		} else {
3087 			if ((error = copyin(usrtvp, tv, sizeof(tv))) != 0)
3088 				return (error);
3089 			tvp = tv;
3090 		}
3091 
3092 		if (tvp[0].tv_usec < 0 || tvp[0].tv_usec >= 1000000 ||
3093 		    tvp[1].tv_usec < 0 || tvp[1].tv_usec >= 1000000)
3094 			return (EINVAL);
3095 		TIMEVAL_TO_TIMESPEC(&tvp[0], &tsp[0]);
3096 		TIMEVAL_TO_TIMESPEC(&tvp[1], &tsp[1]);
3097 	}
3098 	return (0);
3099 }
3100 
3101 /*
3102  * Common implementation code for futimens(), utimensat().
3103  */
3104 #define	UTIMENS_NULL	0x1
3105 #define	UTIMENS_EXIT	0x2
3106 static int
3107 getutimens(const struct timespec *usrtsp, enum uio_seg tspseg,
3108     struct timespec *tsp, int *retflags)
3109 {
3110 	struct timespec tsnow;
3111 	int error;
3112 
3113 	vfs_timestamp(&tsnow);
3114 	*retflags = 0;
3115 	if (usrtsp == NULL) {
3116 		tsp[0] = tsnow;
3117 		tsp[1] = tsnow;
3118 		*retflags |= UTIMENS_NULL;
3119 		return (0);
3120 	}
3121 	if (tspseg == UIO_SYSSPACE) {
3122 		tsp[0] = usrtsp[0];
3123 		tsp[1] = usrtsp[1];
3124 	} else if ((error = copyin(usrtsp, tsp, sizeof(*tsp) * 2)) != 0)
3125 		return (error);
3126 	if (tsp[0].tv_nsec == UTIME_OMIT && tsp[1].tv_nsec == UTIME_OMIT)
3127 		*retflags |= UTIMENS_EXIT;
3128 	if (tsp[0].tv_nsec == UTIME_NOW && tsp[1].tv_nsec == UTIME_NOW)
3129 		*retflags |= UTIMENS_NULL;
3130 	if (tsp[0].tv_nsec == UTIME_OMIT)
3131 		tsp[0].tv_sec = VNOVAL;
3132 	else if (tsp[0].tv_nsec == UTIME_NOW)
3133 		tsp[0] = tsnow;
3134 	else if (tsp[0].tv_nsec < 0 || tsp[0].tv_nsec >= 1000000000L)
3135 		return (EINVAL);
3136 	if (tsp[1].tv_nsec == UTIME_OMIT)
3137 		tsp[1].tv_sec = VNOVAL;
3138 	else if (tsp[1].tv_nsec == UTIME_NOW)
3139 		tsp[1] = tsnow;
3140 	else if (tsp[1].tv_nsec < 0 || tsp[1].tv_nsec >= 1000000000L)
3141 		return (EINVAL);
3142 
3143 	return (0);
3144 }
3145 
3146 /*
3147  * Common implementation code for utimes(), lutimes(), futimes(), futimens(),
3148  * and utimensat().
3149  */
3150 static int
3151 setutimes(struct thread *td, struct vnode *vp, const struct timespec *ts,
3152     int numtimes, int nullflag)
3153 {
3154 	struct mount *mp;
3155 	struct vattr vattr;
3156 	int error, setbirthtime;
3157 
3158 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
3159 		return (error);
3160 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3161 	setbirthtime = 0;
3162 	if (numtimes < 3 && !VOP_GETATTR(vp, &vattr, td->td_ucred) &&
3163 	    timespeccmp(&ts[1], &vattr.va_birthtime, < ))
3164 		setbirthtime = 1;
3165 	VATTR_NULL(&vattr);
3166 	vattr.va_atime = ts[0];
3167 	vattr.va_mtime = ts[1];
3168 	if (setbirthtime)
3169 		vattr.va_birthtime = ts[1];
3170 	if (numtimes > 2)
3171 		vattr.va_birthtime = ts[2];
3172 	if (nullflag)
3173 		vattr.va_vaflags |= VA_UTIMES_NULL;
3174 #ifdef MAC
3175 	error = mac_vnode_check_setutimes(td->td_ucred, vp, vattr.va_atime,
3176 	    vattr.va_mtime);
3177 #endif
3178 	if (error == 0)
3179 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3180 	VOP_UNLOCK(vp);
3181 	vn_finished_write(mp);
3182 	return (error);
3183 }
3184 
3185 /*
3186  * Set the access and modification times of a file.
3187  */
3188 #ifndef _SYS_SYSPROTO_H_
3189 struct utimes_args {
3190 	char	*path;
3191 	struct	timeval *tptr;
3192 };
3193 #endif
3194 int
3195 sys_utimes(struct thread *td, struct utimes_args *uap)
3196 {
3197 
3198 	return (kern_utimesat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
3199 	    uap->tptr, UIO_USERSPACE));
3200 }
3201 
3202 #ifndef _SYS_SYSPROTO_H_
3203 struct futimesat_args {
3204 	int fd;
3205 	const char * path;
3206 	const struct timeval * times;
3207 };
3208 #endif
3209 int
3210 sys_futimesat(struct thread *td, struct futimesat_args *uap)
3211 {
3212 
3213 	return (kern_utimesat(td, uap->fd, uap->path, UIO_USERSPACE,
3214 	    uap->times, UIO_USERSPACE));
3215 }
3216 
3217 int
3218 kern_utimesat(struct thread *td, int fd, const char *path,
3219     enum uio_seg pathseg, struct timeval *tptr, enum uio_seg tptrseg)
3220 {
3221 	struct nameidata nd;
3222 	struct timespec ts[2];
3223 	int error;
3224 
3225 	if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3226 		return (error);
3227 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, fd,
3228 	    &cap_futimes_rights, td);
3229 
3230 	if ((error = namei(&nd)) != 0)
3231 		return (error);
3232 	NDFREE_NOTHING(&nd);
3233 	error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3234 	vrele(nd.ni_vp);
3235 	return (error);
3236 }
3237 
3238 /*
3239  * Set the access and modification times of a file.
3240  */
3241 #ifndef _SYS_SYSPROTO_H_
3242 struct lutimes_args {
3243 	char	*path;
3244 	struct	timeval *tptr;
3245 };
3246 #endif
3247 int
3248 sys_lutimes(struct thread *td, struct lutimes_args *uap)
3249 {
3250 
3251 	return (kern_lutimes(td, uap->path, UIO_USERSPACE, uap->tptr,
3252 	    UIO_USERSPACE));
3253 }
3254 
3255 int
3256 kern_lutimes(struct thread *td, const char *path, enum uio_seg pathseg,
3257     struct timeval *tptr, enum uio_seg tptrseg)
3258 {
3259 	struct timespec ts[2];
3260 	struct nameidata nd;
3261 	int error;
3262 
3263 	if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3264 		return (error);
3265 	NDINIT(&nd, LOOKUP, NOFOLLOW | AUDITVNODE1, pathseg, path, td);
3266 	if ((error = namei(&nd)) != 0)
3267 		return (error);
3268 	NDFREE_NOTHING(&nd);
3269 	error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3270 	vrele(nd.ni_vp);
3271 	return (error);
3272 }
3273 
3274 /*
3275  * Set the access and modification times of a file.
3276  */
3277 #ifndef _SYS_SYSPROTO_H_
3278 struct futimes_args {
3279 	int	fd;
3280 	struct	timeval *tptr;
3281 };
3282 #endif
3283 int
3284 sys_futimes(struct thread *td, struct futimes_args *uap)
3285 {
3286 
3287 	return (kern_futimes(td, uap->fd, uap->tptr, UIO_USERSPACE));
3288 }
3289 
3290 int
3291 kern_futimes(struct thread *td, int fd, struct timeval *tptr,
3292     enum uio_seg tptrseg)
3293 {
3294 	struct timespec ts[2];
3295 	struct file *fp;
3296 	int error;
3297 
3298 	AUDIT_ARG_FD(fd);
3299 	error = getutimes(tptr, tptrseg, ts);
3300 	if (error != 0)
3301 		return (error);
3302 	error = getvnode(td, fd, &cap_futimes_rights, &fp);
3303 	if (error != 0)
3304 		return (error);
3305 #ifdef AUDIT
3306 	vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
3307 	AUDIT_ARG_VNODE1(fp->f_vnode);
3308 	VOP_UNLOCK(fp->f_vnode);
3309 #endif
3310 	error = setutimes(td, fp->f_vnode, ts, 2, tptr == NULL);
3311 	fdrop(fp, td);
3312 	return (error);
3313 }
3314 
3315 int
3316 sys_futimens(struct thread *td, struct futimens_args *uap)
3317 {
3318 
3319 	return (kern_futimens(td, uap->fd, uap->times, UIO_USERSPACE));
3320 }
3321 
3322 int
3323 kern_futimens(struct thread *td, int fd, struct timespec *tptr,
3324     enum uio_seg tptrseg)
3325 {
3326 	struct timespec ts[2];
3327 	struct file *fp;
3328 	int error, flags;
3329 
3330 	AUDIT_ARG_FD(fd);
3331 	error = getutimens(tptr, tptrseg, ts, &flags);
3332 	if (error != 0)
3333 		return (error);
3334 	if (flags & UTIMENS_EXIT)
3335 		return (0);
3336 	error = getvnode(td, fd, &cap_futimes_rights, &fp);
3337 	if (error != 0)
3338 		return (error);
3339 #ifdef AUDIT
3340 	vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
3341 	AUDIT_ARG_VNODE1(fp->f_vnode);
3342 	VOP_UNLOCK(fp->f_vnode);
3343 #endif
3344 	error = setutimes(td, fp->f_vnode, ts, 2, flags & UTIMENS_NULL);
3345 	fdrop(fp, td);
3346 	return (error);
3347 }
3348 
3349 int
3350 sys_utimensat(struct thread *td, struct utimensat_args *uap)
3351 {
3352 
3353 	return (kern_utimensat(td, uap->fd, uap->path, UIO_USERSPACE,
3354 	    uap->times, UIO_USERSPACE, uap->flag));
3355 }
3356 
3357 int
3358 kern_utimensat(struct thread *td, int fd, const char *path,
3359     enum uio_seg pathseg, struct timespec *tptr, enum uio_seg tptrseg,
3360     int flag)
3361 {
3362 	struct nameidata nd;
3363 	struct timespec ts[2];
3364 	int error, flags;
3365 
3366 	if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH |
3367 	    AT_EMPTY_PATH)) != 0)
3368 		return (EINVAL);
3369 
3370 	if ((error = getutimens(tptr, tptrseg, ts, &flags)) != 0)
3371 		return (error);
3372 	NDINIT_ATRIGHTS(&nd, LOOKUP, at2cnpflags(flag, AT_SYMLINK_NOFOLLOW |
3373 	    AT_RESOLVE_BENEATH | AT_EMPTY_PATH) | AUDITVNODE1,
3374 	    pathseg, path, fd, &cap_futimes_rights, td);
3375 	if ((error = namei(&nd)) != 0)
3376 		return (error);
3377 	/*
3378 	 * We are allowed to call namei() regardless of 2xUTIME_OMIT.
3379 	 * POSIX states:
3380 	 * "If both tv_nsec fields are UTIME_OMIT... EACCESS may be detected."
3381 	 * "Search permission is denied by a component of the path prefix."
3382 	 */
3383 	NDFREE_NOTHING(&nd);
3384 	if ((flags & UTIMENS_EXIT) == 0)
3385 		error = setutimes(td, nd.ni_vp, ts, 2, flags & UTIMENS_NULL);
3386 	vrele(nd.ni_vp);
3387 	return (error);
3388 }
3389 
3390 /*
3391  * Truncate a file given its path name.
3392  */
3393 #ifndef _SYS_SYSPROTO_H_
3394 struct truncate_args {
3395 	char	*path;
3396 	int	pad;
3397 	off_t	length;
3398 };
3399 #endif
3400 int
3401 sys_truncate(struct thread *td, struct truncate_args *uap)
3402 {
3403 
3404 	return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3405 }
3406 
3407 int
3408 kern_truncate(struct thread *td, const char *path, enum uio_seg pathseg,
3409     off_t length)
3410 {
3411 	struct mount *mp;
3412 	struct vnode *vp;
3413 	void *rl_cookie;
3414 	struct vattr vattr;
3415 	struct nameidata nd;
3416 	int error;
3417 
3418 	if (length < 0)
3419 		return (EINVAL);
3420 retry:
3421 	NDINIT(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, td);
3422 	if ((error = namei(&nd)) != 0)
3423 		return (error);
3424 	vp = nd.ni_vp;
3425 	rl_cookie = vn_rangelock_wlock(vp, 0, OFF_MAX);
3426 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) {
3427 		vn_rangelock_unlock(vp, rl_cookie);
3428 		vrele(vp);
3429 		return (error);
3430 	}
3431 	NDFREE(&nd, NDF_ONLY_PNBUF);
3432 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3433 	if (vp->v_type == VDIR)
3434 		error = EISDIR;
3435 #ifdef MAC
3436 	else if ((error = mac_vnode_check_write(td->td_ucred, NOCRED, vp))) {
3437 	}
3438 #endif
3439 	else if ((error = vn_writechk(vp)) == 0 &&
3440 	    (error = VOP_ACCESS(vp, VWRITE, td->td_ucred, td)) == 0) {
3441 		VATTR_NULL(&vattr);
3442 		vattr.va_size = length;
3443 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3444 	}
3445 	VOP_UNLOCK(vp);
3446 	vn_finished_write(mp);
3447 	vn_rangelock_unlock(vp, rl_cookie);
3448 	vrele(vp);
3449 	if (error == ERELOOKUP)
3450 		goto retry;
3451 	return (error);
3452 }
3453 
3454 #if defined(COMPAT_43)
3455 /*
3456  * Truncate a file given its path name.
3457  */
3458 #ifndef _SYS_SYSPROTO_H_
3459 struct otruncate_args {
3460 	char	*path;
3461 	long	length;
3462 };
3463 #endif
3464 int
3465 otruncate(struct thread *td, struct otruncate_args *uap)
3466 {
3467 
3468 	return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3469 }
3470 #endif /* COMPAT_43 */
3471 
3472 #if defined(COMPAT_FREEBSD6)
3473 /* Versions with the pad argument */
3474 int
3475 freebsd6_truncate(struct thread *td, struct freebsd6_truncate_args *uap)
3476 {
3477 
3478 	return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3479 }
3480 
3481 int
3482 freebsd6_ftruncate(struct thread *td, struct freebsd6_ftruncate_args *uap)
3483 {
3484 
3485 	return (kern_ftruncate(td, uap->fd, uap->length));
3486 }
3487 #endif
3488 
3489 int
3490 kern_fsync(struct thread *td, int fd, bool fullsync)
3491 {
3492 	struct vnode *vp;
3493 	struct mount *mp;
3494 	struct file *fp;
3495 	int error, lock_flags;
3496 
3497 	AUDIT_ARG_FD(fd);
3498 	error = getvnode(td, fd, &cap_fsync_rights, &fp);
3499 	if (error != 0)
3500 		return (error);
3501 	vp = fp->f_vnode;
3502 #if 0
3503 	if (!fullsync)
3504 		/* XXXKIB: compete outstanding aio writes */;
3505 #endif
3506 retry:
3507 	error = vn_start_write(vp, &mp, V_WAIT | PCATCH);
3508 	if (error != 0)
3509 		goto drop;
3510 	if (MNT_SHARED_WRITES(mp) ||
3511 	    ((mp == NULL) && MNT_SHARED_WRITES(vp->v_mount))) {
3512 		lock_flags = LK_SHARED;
3513 	} else {
3514 		lock_flags = LK_EXCLUSIVE;
3515 	}
3516 	vn_lock(vp, lock_flags | LK_RETRY);
3517 	AUDIT_ARG_VNODE1(vp);
3518 	if (vp->v_object != NULL) {
3519 		VM_OBJECT_WLOCK(vp->v_object);
3520 		vm_object_page_clean(vp->v_object, 0, 0, 0);
3521 		VM_OBJECT_WUNLOCK(vp->v_object);
3522 	}
3523 	error = fullsync ? VOP_FSYNC(vp, MNT_WAIT, td) : VOP_FDATASYNC(vp, td);
3524 	VOP_UNLOCK(vp);
3525 	vn_finished_write(mp);
3526 	if (error == ERELOOKUP)
3527 		goto retry;
3528 drop:
3529 	fdrop(fp, td);
3530 	return (error);
3531 }
3532 
3533 /*
3534  * Sync an open file.
3535  */
3536 #ifndef _SYS_SYSPROTO_H_
3537 struct fsync_args {
3538 	int	fd;
3539 };
3540 #endif
3541 int
3542 sys_fsync(struct thread *td, struct fsync_args *uap)
3543 {
3544 
3545 	return (kern_fsync(td, uap->fd, true));
3546 }
3547 
3548 int
3549 sys_fdatasync(struct thread *td, struct fdatasync_args *uap)
3550 {
3551 
3552 	return (kern_fsync(td, uap->fd, false));
3553 }
3554 
3555 /*
3556  * Rename files.  Source and destination must either both be directories, or
3557  * both not be directories.  If target is a directory, it must be empty.
3558  */
3559 #ifndef _SYS_SYSPROTO_H_
3560 struct rename_args {
3561 	char	*from;
3562 	char	*to;
3563 };
3564 #endif
3565 int
3566 sys_rename(struct thread *td, struct rename_args *uap)
3567 {
3568 
3569 	return (kern_renameat(td, AT_FDCWD, uap->from, AT_FDCWD,
3570 	    uap->to, UIO_USERSPACE));
3571 }
3572 
3573 #ifndef _SYS_SYSPROTO_H_
3574 struct renameat_args {
3575 	int	oldfd;
3576 	char	*old;
3577 	int	newfd;
3578 	char	*new;
3579 };
3580 #endif
3581 int
3582 sys_renameat(struct thread *td, struct renameat_args *uap)
3583 {
3584 
3585 	return (kern_renameat(td, uap->oldfd, uap->old, uap->newfd, uap->new,
3586 	    UIO_USERSPACE));
3587 }
3588 
3589 #ifdef MAC
3590 static int
3591 kern_renameat_mac(struct thread *td, int oldfd, const char *old, int newfd,
3592     const char *new, enum uio_seg pathseg, struct nameidata *fromnd)
3593 {
3594 	int error;
3595 
3596 	NDINIT_ATRIGHTS(fromnd, DELETE, LOCKPARENT | LOCKLEAF | SAVESTART |
3597 	    AUDITVNODE1, pathseg, old, oldfd, &cap_renameat_source_rights, td);
3598 	if ((error = namei(fromnd)) != 0)
3599 		return (error);
3600 	error = mac_vnode_check_rename_from(td->td_ucred, fromnd->ni_dvp,
3601 	    fromnd->ni_vp, &fromnd->ni_cnd);
3602 	VOP_UNLOCK(fromnd->ni_dvp);
3603 	if (fromnd->ni_dvp != fromnd->ni_vp)
3604 		VOP_UNLOCK(fromnd->ni_vp);
3605 	if (error != 0) {
3606 		NDFREE(fromnd, NDF_ONLY_PNBUF);
3607 		vrele(fromnd->ni_dvp);
3608 		vrele(fromnd->ni_vp);
3609 		if (fromnd->ni_startdir)
3610 			vrele(fromnd->ni_startdir);
3611 	}
3612 	return (error);
3613 }
3614 #endif
3615 
3616 int
3617 kern_renameat(struct thread *td, int oldfd, const char *old, int newfd,
3618     const char *new, enum uio_seg pathseg)
3619 {
3620 	struct mount *mp = NULL;
3621 	struct vnode *tvp, *fvp, *tdvp;
3622 	struct nameidata fromnd, tond;
3623 	u_int64_t tondflags;
3624 	int error;
3625 
3626 again:
3627 	bwillwrite();
3628 #ifdef MAC
3629 	if (mac_vnode_check_rename_from_enabled()) {
3630 		error = kern_renameat_mac(td, oldfd, old, newfd, new, pathseg,
3631 		    &fromnd);
3632 		if (error != 0)
3633 			return (error);
3634 	} else {
3635 #endif
3636 	NDINIT_ATRIGHTS(&fromnd, DELETE, WANTPARENT | SAVESTART | AUDITVNODE1,
3637 	    pathseg, old, oldfd, &cap_renameat_source_rights, td);
3638 	if ((error = namei(&fromnd)) != 0)
3639 		return (error);
3640 #ifdef MAC
3641 	}
3642 #endif
3643 	fvp = fromnd.ni_vp;
3644 	tondflags = LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART | AUDITVNODE2;
3645 	if (fromnd.ni_vp->v_type == VDIR)
3646 		tondflags |= WILLBEDIR;
3647 	NDINIT_ATRIGHTS(&tond, RENAME, tondflags, pathseg, new, newfd,
3648 	    &cap_renameat_target_rights, td);
3649 	if ((error = namei(&tond)) != 0) {
3650 		/* Translate error code for rename("dir1", "dir2/."). */
3651 		if (error == EISDIR && fvp->v_type == VDIR)
3652 			error = EINVAL;
3653 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3654 		vrele(fromnd.ni_dvp);
3655 		vrele(fvp);
3656 		goto out1;
3657 	}
3658 	tdvp = tond.ni_dvp;
3659 	tvp = tond.ni_vp;
3660 	error = vn_start_write(fvp, &mp, V_NOWAIT);
3661 	if (error != 0) {
3662 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3663 		NDFREE(&tond, NDF_ONLY_PNBUF);
3664 		if (tvp != NULL)
3665 			vput(tvp);
3666 		if (tdvp == tvp)
3667 			vrele(tdvp);
3668 		else
3669 			vput(tdvp);
3670 		vrele(fromnd.ni_dvp);
3671 		vrele(fvp);
3672 		vrele(tond.ni_startdir);
3673 		if (fromnd.ni_startdir != NULL)
3674 			vrele(fromnd.ni_startdir);
3675 		error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH);
3676 		if (error != 0)
3677 			return (error);
3678 		goto again;
3679 	}
3680 	if (tvp != NULL) {
3681 		if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
3682 			error = ENOTDIR;
3683 			goto out;
3684 		} else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
3685 			error = EISDIR;
3686 			goto out;
3687 		}
3688 #ifdef CAPABILITIES
3689 		if (newfd != AT_FDCWD && (tond.ni_resflags & NIRES_ABS) == 0) {
3690 			/*
3691 			 * If the target already exists we require CAP_UNLINKAT
3692 			 * from 'newfd', when newfd was used for the lookup.
3693 			 */
3694 			error = cap_check(&tond.ni_filecaps.fc_rights,
3695 			    &cap_unlinkat_rights);
3696 			if (error != 0)
3697 				goto out;
3698 		}
3699 #endif
3700 	}
3701 	if (fvp == tdvp) {
3702 		error = EINVAL;
3703 		goto out;
3704 	}
3705 	/*
3706 	 * If the source is the same as the destination (that is, if they
3707 	 * are links to the same vnode), then there is nothing to do.
3708 	 */
3709 	if (fvp == tvp)
3710 		error = ERESTART;
3711 #ifdef MAC
3712 	else
3713 		error = mac_vnode_check_rename_to(td->td_ucred, tdvp,
3714 		    tond.ni_vp, fromnd.ni_dvp == tdvp, &tond.ni_cnd);
3715 #endif
3716 out:
3717 	if (error == 0) {
3718 		error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
3719 		    tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
3720 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3721 		NDFREE(&tond, NDF_ONLY_PNBUF);
3722 	} else {
3723 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3724 		NDFREE(&tond, NDF_ONLY_PNBUF);
3725 		if (tvp != NULL)
3726 			vput(tvp);
3727 		if (tdvp == tvp)
3728 			vrele(tdvp);
3729 		else
3730 			vput(tdvp);
3731 		vrele(fromnd.ni_dvp);
3732 		vrele(fvp);
3733 	}
3734 	vrele(tond.ni_startdir);
3735 	vn_finished_write(mp);
3736 out1:
3737 	if (fromnd.ni_startdir)
3738 		vrele(fromnd.ni_startdir);
3739 	if (error == ERESTART)
3740 		return (0);
3741 	if (error == ERELOOKUP)
3742 		goto again;
3743 	return (error);
3744 }
3745 
3746 /*
3747  * Make a directory file.
3748  */
3749 #ifndef _SYS_SYSPROTO_H_
3750 struct mkdir_args {
3751 	char	*path;
3752 	int	mode;
3753 };
3754 #endif
3755 int
3756 sys_mkdir(struct thread *td, struct mkdir_args *uap)
3757 {
3758 
3759 	return (kern_mkdirat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
3760 	    uap->mode));
3761 }
3762 
3763 #ifndef _SYS_SYSPROTO_H_
3764 struct mkdirat_args {
3765 	int	fd;
3766 	char	*path;
3767 	mode_t	mode;
3768 };
3769 #endif
3770 int
3771 sys_mkdirat(struct thread *td, struct mkdirat_args *uap)
3772 {
3773 
3774 	return (kern_mkdirat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode));
3775 }
3776 
3777 int
3778 kern_mkdirat(struct thread *td, int fd, const char *path, enum uio_seg segflg,
3779     int mode)
3780 {
3781 	struct mount *mp;
3782 	struct vattr vattr;
3783 	struct nameidata nd;
3784 	int error;
3785 
3786 	AUDIT_ARG_MODE(mode);
3787 restart:
3788 	bwillwrite();
3789 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
3790 	    NC_NOMAKEENTRY | NC_KEEPPOSENTRY | FAILIFEXISTS | WILLBEDIR,
3791 	    segflg, path, fd, &cap_mkdirat_rights, td);
3792 	if ((error = namei(&nd)) != 0)
3793 		return (error);
3794 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3795 		NDFREE(&nd, NDF_ONLY_PNBUF);
3796 		vput(nd.ni_dvp);
3797 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3798 			return (error);
3799 		goto restart;
3800 	}
3801 	VATTR_NULL(&vattr);
3802 	vattr.va_type = VDIR;
3803 	vattr.va_mode = (mode & ACCESSPERMS) &~ td->td_proc->p_pd->pd_cmask;
3804 #ifdef MAC
3805 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
3806 	    &vattr);
3807 	if (error != 0)
3808 		goto out;
3809 #endif
3810 	error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
3811 #ifdef MAC
3812 out:
3813 #endif
3814 	NDFREE(&nd, NDF_ONLY_PNBUF);
3815 	VOP_VPUT_PAIR(nd.ni_dvp, error == 0 ? &nd.ni_vp : NULL, true);
3816 	vn_finished_write(mp);
3817 	if (error == ERELOOKUP)
3818 		goto restart;
3819 	return (error);
3820 }
3821 
3822 /*
3823  * Remove a directory file.
3824  */
3825 #ifndef _SYS_SYSPROTO_H_
3826 struct rmdir_args {
3827 	char	*path;
3828 };
3829 #endif
3830 int
3831 sys_rmdir(struct thread *td, struct rmdir_args *uap)
3832 {
3833 
3834 	return (kern_frmdirat(td, AT_FDCWD, uap->path, FD_NONE, UIO_USERSPACE,
3835 	    0));
3836 }
3837 
3838 int
3839 kern_frmdirat(struct thread *td, int dfd, const char *path, int fd,
3840     enum uio_seg pathseg, int flag)
3841 {
3842 	struct mount *mp;
3843 	struct vnode *vp;
3844 	struct file *fp;
3845 	struct nameidata nd;
3846 	cap_rights_t rights;
3847 	int error;
3848 
3849 	fp = NULL;
3850 	if (fd != FD_NONE) {
3851 		error = getvnode(td, fd, cap_rights_init_one(&rights,
3852 		    CAP_LOOKUP), &fp);
3853 		if (error != 0)
3854 			return (error);
3855 	}
3856 
3857 restart:
3858 	bwillwrite();
3859 	NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1 |
3860 	    at2cnpflags(flag, AT_RESOLVE_BENEATH),
3861 	    pathseg, path, dfd, &cap_unlinkat_rights, td);
3862 	if ((error = namei(&nd)) != 0)
3863 		goto fdout;
3864 	vp = nd.ni_vp;
3865 	if (vp->v_type != VDIR) {
3866 		error = ENOTDIR;
3867 		goto out;
3868 	}
3869 	/*
3870 	 * No rmdir "." please.
3871 	 */
3872 	if (nd.ni_dvp == vp) {
3873 		error = EINVAL;
3874 		goto out;
3875 	}
3876 	/*
3877 	 * The root of a mounted filesystem cannot be deleted.
3878 	 */
3879 	if (vp->v_vflag & VV_ROOT) {
3880 		error = EBUSY;
3881 		goto out;
3882 	}
3883 
3884 	if (fp != NULL && fp->f_vnode != vp) {
3885 		if (VN_IS_DOOMED(fp->f_vnode))
3886 			error = EBADF;
3887 		else
3888 			error = EDEADLK;
3889 		goto out;
3890 	}
3891 
3892 #ifdef MAC
3893 	error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
3894 	    &nd.ni_cnd);
3895 	if (error != 0)
3896 		goto out;
3897 #endif
3898 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3899 		NDFREE(&nd, NDF_ONLY_PNBUF);
3900 		vput(vp);
3901 		if (nd.ni_dvp == vp)
3902 			vrele(nd.ni_dvp);
3903 		else
3904 			vput(nd.ni_dvp);
3905 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3906 			goto fdout;
3907 		goto restart;
3908 	}
3909 	vfs_notify_upper(vp, VFS_NOTIFY_UPPER_UNLINK);
3910 	error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
3911 	vn_finished_write(mp);
3912 out:
3913 	NDFREE(&nd, NDF_ONLY_PNBUF);
3914 	vput(vp);
3915 	if (nd.ni_dvp == vp)
3916 		vrele(nd.ni_dvp);
3917 	else
3918 		vput(nd.ni_dvp);
3919 	if (error == ERELOOKUP)
3920 		goto restart;
3921 fdout:
3922 	if (fp != NULL)
3923 		fdrop(fp, td);
3924 	return (error);
3925 }
3926 
3927 #if defined(COMPAT_43) || defined(COMPAT_FREEBSD11)
3928 int
3929 freebsd11_kern_getdirentries(struct thread *td, int fd, char *ubuf, u_int count,
3930     long *basep, void (*func)(struct freebsd11_dirent *))
3931 {
3932 	struct freebsd11_dirent dstdp;
3933 	struct dirent *dp, *edp;
3934 	char *dirbuf;
3935 	off_t base;
3936 	ssize_t resid, ucount;
3937 	int error;
3938 
3939 	/* XXX arbitrary sanity limit on `count'. */
3940 	count = min(count, 64 * 1024);
3941 
3942 	dirbuf = malloc(count, M_TEMP, M_WAITOK);
3943 
3944 	error = kern_getdirentries(td, fd, dirbuf, count, &base, &resid,
3945 	    UIO_SYSSPACE);
3946 	if (error != 0)
3947 		goto done;
3948 	if (basep != NULL)
3949 		*basep = base;
3950 
3951 	ucount = 0;
3952 	for (dp = (struct dirent *)dirbuf,
3953 	    edp = (struct dirent *)&dirbuf[count - resid];
3954 	    ucount < count && dp < edp; ) {
3955 		if (dp->d_reclen == 0)
3956 			break;
3957 		MPASS(dp->d_reclen >= _GENERIC_DIRLEN(0));
3958 		if (dp->d_namlen >= sizeof(dstdp.d_name))
3959 			continue;
3960 		dstdp.d_type = dp->d_type;
3961 		dstdp.d_namlen = dp->d_namlen;
3962 		dstdp.d_fileno = dp->d_fileno;		/* truncate */
3963 		if (dstdp.d_fileno != dp->d_fileno) {
3964 			switch (ino64_trunc_error) {
3965 			default:
3966 			case 0:
3967 				break;
3968 			case 1:
3969 				error = EOVERFLOW;
3970 				goto done;
3971 			case 2:
3972 				dstdp.d_fileno = UINT32_MAX;
3973 				break;
3974 			}
3975 		}
3976 		dstdp.d_reclen = sizeof(dstdp) - sizeof(dstdp.d_name) +
3977 		    ((dp->d_namlen + 1 + 3) &~ 3);
3978 		bcopy(dp->d_name, dstdp.d_name, dstdp.d_namlen);
3979 		bzero(dstdp.d_name + dstdp.d_namlen,
3980 		    dstdp.d_reclen - offsetof(struct freebsd11_dirent, d_name) -
3981 		    dstdp.d_namlen);
3982 		MPASS(dstdp.d_reclen <= dp->d_reclen);
3983 		MPASS(ucount + dstdp.d_reclen <= count);
3984 		if (func != NULL)
3985 			func(&dstdp);
3986 		error = copyout(&dstdp, ubuf + ucount, dstdp.d_reclen);
3987 		if (error != 0)
3988 			break;
3989 		dp = (struct dirent *)((char *)dp + dp->d_reclen);
3990 		ucount += dstdp.d_reclen;
3991 	}
3992 
3993 done:
3994 	free(dirbuf, M_TEMP);
3995 	if (error == 0)
3996 		td->td_retval[0] = ucount;
3997 	return (error);
3998 }
3999 #endif /* COMPAT */
4000 
4001 #ifdef COMPAT_43
4002 static void
4003 ogetdirentries_cvt(struct freebsd11_dirent *dp)
4004 {
4005 #if (BYTE_ORDER == LITTLE_ENDIAN)
4006 	/*
4007 	 * The expected low byte of dp->d_namlen is our dp->d_type.
4008 	 * The high MBZ byte of dp->d_namlen is our dp->d_namlen.
4009 	 */
4010 	dp->d_type = dp->d_namlen;
4011 	dp->d_namlen = 0;
4012 #else
4013 	/*
4014 	 * The dp->d_type is the high byte of the expected dp->d_namlen,
4015 	 * so must be zero'ed.
4016 	 */
4017 	dp->d_type = 0;
4018 #endif
4019 }
4020 
4021 /*
4022  * Read a block of directory entries in a filesystem independent format.
4023  */
4024 #ifndef _SYS_SYSPROTO_H_
4025 struct ogetdirentries_args {
4026 	int	fd;
4027 	char	*buf;
4028 	u_int	count;
4029 	long	*basep;
4030 };
4031 #endif
4032 int
4033 ogetdirentries(struct thread *td, struct ogetdirentries_args *uap)
4034 {
4035 	long loff;
4036 	int error;
4037 
4038 	error = kern_ogetdirentries(td, uap, &loff);
4039 	if (error == 0)
4040 		error = copyout(&loff, uap->basep, sizeof(long));
4041 	return (error);
4042 }
4043 
4044 int
4045 kern_ogetdirentries(struct thread *td, struct ogetdirentries_args *uap,
4046     long *ploff)
4047 {
4048 	long base;
4049 	int error;
4050 
4051 	/* XXX arbitrary sanity limit on `count'. */
4052 	if (uap->count > 64 * 1024)
4053 		return (EINVAL);
4054 
4055 	error = freebsd11_kern_getdirentries(td, uap->fd, uap->buf, uap->count,
4056 	    &base, ogetdirentries_cvt);
4057 
4058 	if (error == 0 && uap->basep != NULL)
4059 		error = copyout(&base, uap->basep, sizeof(long));
4060 
4061 	return (error);
4062 }
4063 #endif /* COMPAT_43 */
4064 
4065 #if defined(COMPAT_FREEBSD11)
4066 #ifndef _SYS_SYSPROTO_H_
4067 struct freebsd11_getdirentries_args {
4068 	int	fd;
4069 	char	*buf;
4070 	u_int	count;
4071 	long	*basep;
4072 };
4073 #endif
4074 int
4075 freebsd11_getdirentries(struct thread *td,
4076     struct freebsd11_getdirentries_args *uap)
4077 {
4078 	long base;
4079 	int error;
4080 
4081 	error = freebsd11_kern_getdirentries(td, uap->fd, uap->buf, uap->count,
4082 	    &base, NULL);
4083 
4084 	if (error == 0 && uap->basep != NULL)
4085 		error = copyout(&base, uap->basep, sizeof(long));
4086 	return (error);
4087 }
4088 
4089 int
4090 freebsd11_getdents(struct thread *td, struct freebsd11_getdents_args *uap)
4091 {
4092 	struct freebsd11_getdirentries_args ap;
4093 
4094 	ap.fd = uap->fd;
4095 	ap.buf = uap->buf;
4096 	ap.count = uap->count;
4097 	ap.basep = NULL;
4098 	return (freebsd11_getdirentries(td, &ap));
4099 }
4100 #endif /* COMPAT_FREEBSD11 */
4101 
4102 /*
4103  * Read a block of directory entries in a filesystem independent format.
4104  */
4105 int
4106 sys_getdirentries(struct thread *td, struct getdirentries_args *uap)
4107 {
4108 	off_t base;
4109 	int error;
4110 
4111 	error = kern_getdirentries(td, uap->fd, uap->buf, uap->count, &base,
4112 	    NULL, UIO_USERSPACE);
4113 	if (error != 0)
4114 		return (error);
4115 	if (uap->basep != NULL)
4116 		error = copyout(&base, uap->basep, sizeof(off_t));
4117 	return (error);
4118 }
4119 
4120 int
4121 kern_getdirentries(struct thread *td, int fd, char *buf, size_t count,
4122     off_t *basep, ssize_t *residp, enum uio_seg bufseg)
4123 {
4124 	struct vnode *vp;
4125 	struct file *fp;
4126 	struct uio auio;
4127 	struct iovec aiov;
4128 	off_t loff;
4129 	int error, eofflag;
4130 	off_t foffset;
4131 
4132 	AUDIT_ARG_FD(fd);
4133 	if (count > IOSIZE_MAX)
4134 		return (EINVAL);
4135 	auio.uio_resid = count;
4136 	error = getvnode(td, fd, &cap_read_rights, &fp);
4137 	if (error != 0)
4138 		return (error);
4139 	if ((fp->f_flag & FREAD) == 0) {
4140 		fdrop(fp, td);
4141 		return (EBADF);
4142 	}
4143 	vp = fp->f_vnode;
4144 	foffset = foffset_lock(fp, 0);
4145 unionread:
4146 	if (vp->v_type != VDIR) {
4147 		error = EINVAL;
4148 		goto fail;
4149 	}
4150 	aiov.iov_base = buf;
4151 	aiov.iov_len = count;
4152 	auio.uio_iov = &aiov;
4153 	auio.uio_iovcnt = 1;
4154 	auio.uio_rw = UIO_READ;
4155 	auio.uio_segflg = bufseg;
4156 	auio.uio_td = td;
4157 	vn_lock(vp, LK_SHARED | LK_RETRY);
4158 	AUDIT_ARG_VNODE1(vp);
4159 	loff = auio.uio_offset = foffset;
4160 #ifdef MAC
4161 	error = mac_vnode_check_readdir(td->td_ucred, vp);
4162 	if (error == 0)
4163 #endif
4164 		error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, NULL,
4165 		    NULL);
4166 	foffset = auio.uio_offset;
4167 	if (error != 0) {
4168 		VOP_UNLOCK(vp);
4169 		goto fail;
4170 	}
4171 	if (count == auio.uio_resid &&
4172 	    (vp->v_vflag & VV_ROOT) &&
4173 	    (vp->v_mount->mnt_flag & MNT_UNION)) {
4174 		struct vnode *tvp = vp;
4175 
4176 		vp = vp->v_mount->mnt_vnodecovered;
4177 		VREF(vp);
4178 		fp->f_vnode = vp;
4179 		foffset = 0;
4180 		vput(tvp);
4181 		goto unionread;
4182 	}
4183 	VOP_UNLOCK(vp);
4184 	*basep = loff;
4185 	if (residp != NULL)
4186 		*residp = auio.uio_resid;
4187 	td->td_retval[0] = count - auio.uio_resid;
4188 fail:
4189 	foffset_unlock(fp, foffset, 0);
4190 	fdrop(fp, td);
4191 	return (error);
4192 }
4193 
4194 /*
4195  * Set the mode mask for creation of filesystem nodes.
4196  */
4197 #ifndef _SYS_SYSPROTO_H_
4198 struct umask_args {
4199 	int	newmask;
4200 };
4201 #endif
4202 int
4203 sys_umask(struct thread *td, struct umask_args *uap)
4204 {
4205 	struct pwddesc *pdp;
4206 
4207 	pdp = td->td_proc->p_pd;
4208 	PWDDESC_XLOCK(pdp);
4209 	td->td_retval[0] = pdp->pd_cmask;
4210 	pdp->pd_cmask = uap->newmask & ALLPERMS;
4211 	PWDDESC_XUNLOCK(pdp);
4212 	return (0);
4213 }
4214 
4215 /*
4216  * Void all references to file by ripping underlying filesystem away from
4217  * vnode.
4218  */
4219 #ifndef _SYS_SYSPROTO_H_
4220 struct revoke_args {
4221 	char	*path;
4222 };
4223 #endif
4224 int
4225 sys_revoke(struct thread *td, struct revoke_args *uap)
4226 {
4227 	struct vnode *vp;
4228 	struct vattr vattr;
4229 	struct nameidata nd;
4230 	int error;
4231 
4232 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
4233 	    uap->path, td);
4234 	if ((error = namei(&nd)) != 0)
4235 		return (error);
4236 	vp = nd.ni_vp;
4237 	NDFREE_NOTHING(&nd);
4238 	if (vp->v_type != VCHR || vp->v_rdev == NULL) {
4239 		error = EINVAL;
4240 		goto out;
4241 	}
4242 #ifdef MAC
4243 	error = mac_vnode_check_revoke(td->td_ucred, vp);
4244 	if (error != 0)
4245 		goto out;
4246 #endif
4247 	error = VOP_GETATTR(vp, &vattr, td->td_ucred);
4248 	if (error != 0)
4249 		goto out;
4250 	if (td->td_ucred->cr_uid != vattr.va_uid) {
4251 		error = priv_check(td, PRIV_VFS_ADMIN);
4252 		if (error != 0)
4253 			goto out;
4254 	}
4255 	if (devfs_usecount(vp) > 0)
4256 		VOP_REVOKE(vp, REVOKEALL);
4257 out:
4258 	vput(vp);
4259 	return (error);
4260 }
4261 
4262 /*
4263  * This variant of getvnode() allows O_PATH files.  Caller should
4264  * ensure that returned file and vnode are only used for compatible
4265  * semantics.
4266  */
4267 int
4268 getvnode_path(struct thread *td, int fd, cap_rights_t *rightsp,
4269     struct file **fpp)
4270 {
4271 	struct file *fp;
4272 	int error;
4273 
4274 	error = fget_unlocked(td->td_proc->p_fd, fd, rightsp, &fp);
4275 	if (error != 0)
4276 		return (error);
4277 
4278 	/*
4279 	 * The file could be not of the vnode type, or it may be not
4280 	 * yet fully initialized, in which case the f_vnode pointer
4281 	 * may be set, but f_ops is still badfileops.  E.g.,
4282 	 * devfs_open() transiently create such situation to
4283 	 * facilitate csw d_fdopen().
4284 	 *
4285 	 * Dupfdopen() handling in kern_openat() installs the
4286 	 * half-baked file into the process descriptor table, allowing
4287 	 * other thread to dereference it. Guard against the race by
4288 	 * checking f_ops.
4289 	 */
4290 	if (fp->f_vnode == NULL || fp->f_ops == &badfileops) {
4291 		fdrop(fp, td);
4292 		return (EINVAL);
4293 	}
4294 
4295 	*fpp = fp;
4296 	return (0);
4297 }
4298 
4299 /*
4300  * Convert a user file descriptor to a kernel file entry and check
4301  * that, if it is a capability, the correct rights are present.
4302  * A reference on the file entry is held upon returning.
4303  */
4304 int
4305 getvnode(struct thread *td, int fd, cap_rights_t *rightsp, struct file **fpp)
4306 {
4307 	int error;
4308 
4309 	error = getvnode_path(td, fd, rightsp, fpp);
4310 
4311 	/*
4312 	 * Filter out O_PATH file descriptors, most getvnode() callers
4313 	 * do not call fo_ methods.
4314 	 */
4315 	if (error == 0 && (*fpp)->f_ops == &path_fileops) {
4316 		fdrop(*fpp, td);
4317 		error = EBADF;
4318 	}
4319 
4320 	return (error);
4321 }
4322 
4323 /*
4324  * Get an (NFS) file handle.
4325  */
4326 #ifndef _SYS_SYSPROTO_H_
4327 struct lgetfh_args {
4328 	char *fname;
4329 	fhandle_t *fhp;
4330 };
4331 #endif
4332 int
4333 sys_lgetfh(struct thread *td, struct lgetfh_args *uap)
4334 {
4335 
4336 	return (kern_getfhat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->fname,
4337 	    UIO_USERSPACE, uap->fhp, UIO_USERSPACE));
4338 }
4339 
4340 #ifndef _SYS_SYSPROTO_H_
4341 struct getfh_args {
4342 	char *fname;
4343 	fhandle_t *fhp;
4344 };
4345 #endif
4346 int
4347 sys_getfh(struct thread *td, struct getfh_args *uap)
4348 {
4349 
4350 	return (kern_getfhat(td, 0, AT_FDCWD, uap->fname, UIO_USERSPACE,
4351 	    uap->fhp, UIO_USERSPACE));
4352 }
4353 
4354 /*
4355  * syscall for the rpc.lockd to use to translate an open descriptor into
4356  * a NFS file handle.
4357  *
4358  * warning: do not remove the priv_check() call or this becomes one giant
4359  * security hole.
4360  */
4361 #ifndef _SYS_SYSPROTO_H_
4362 struct getfhat_args {
4363 	int fd;
4364 	char *path;
4365 	fhandle_t *fhp;
4366 	int flags;
4367 };
4368 #endif
4369 int
4370 sys_getfhat(struct thread *td, struct getfhat_args *uap)
4371 {
4372 
4373 	if ((uap->flags & ~(AT_SYMLINK_NOFOLLOW | AT_RESOLVE_BENEATH)) != 0)
4374 		return (EINVAL);
4375 	return (kern_getfhat(td, uap->flags, uap->fd, uap->path, UIO_USERSPACE,
4376 	    uap->fhp, UIO_USERSPACE));
4377 }
4378 
4379 int
4380 kern_getfhat(struct thread *td, int flags, int fd, const char *path,
4381     enum uio_seg pathseg, fhandle_t *fhp, enum uio_seg fhseg)
4382 {
4383 	struct nameidata nd;
4384 	fhandle_t fh;
4385 	struct vnode *vp;
4386 	int error;
4387 
4388 	error = priv_check(td, PRIV_VFS_GETFH);
4389 	if (error != 0)
4390 		return (error);
4391 	NDINIT_AT(&nd, LOOKUP, at2cnpflags(flags, AT_SYMLINK_NOFOLLOW |
4392 	    AT_RESOLVE_BENEATH) | LOCKLEAF | AUDITVNODE1, pathseg, path,
4393 	    fd, td);
4394 	error = namei(&nd);
4395 	if (error != 0)
4396 		return (error);
4397 	NDFREE_NOTHING(&nd);
4398 	vp = nd.ni_vp;
4399 	bzero(&fh, sizeof(fh));
4400 	fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
4401 	error = VOP_VPTOFH(vp, &fh.fh_fid);
4402 	vput(vp);
4403 	if (error == 0) {
4404 		if (fhseg == UIO_USERSPACE)
4405 			error = copyout(&fh, fhp, sizeof (fh));
4406 		else
4407 			memcpy(fhp, &fh, sizeof(fh));
4408 	}
4409 	return (error);
4410 }
4411 
4412 #ifndef _SYS_SYSPROTO_H_
4413 struct fhlink_args {
4414 	fhandle_t *fhp;
4415 	const char *to;
4416 };
4417 #endif
4418 int
4419 sys_fhlink(struct thread *td, struct fhlink_args *uap)
4420 {
4421 
4422 	return (kern_fhlinkat(td, AT_FDCWD, uap->to, UIO_USERSPACE, uap->fhp));
4423 }
4424 
4425 #ifndef _SYS_SYSPROTO_H_
4426 struct fhlinkat_args {
4427 	fhandle_t *fhp;
4428 	int tofd;
4429 	const char *to;
4430 };
4431 #endif
4432 int
4433 sys_fhlinkat(struct thread *td, struct fhlinkat_args *uap)
4434 {
4435 
4436 	return (kern_fhlinkat(td, uap->tofd, uap->to, UIO_USERSPACE, uap->fhp));
4437 }
4438 
4439 static int
4440 kern_fhlinkat(struct thread *td, int fd, const char *path,
4441     enum uio_seg pathseg, fhandle_t *fhp)
4442 {
4443 	fhandle_t fh;
4444 	struct mount *mp;
4445 	struct vnode *vp;
4446 	int error;
4447 
4448 	error = priv_check(td, PRIV_VFS_GETFH);
4449 	if (error != 0)
4450 		return (error);
4451 	error = copyin(fhp, &fh, sizeof(fh));
4452 	if (error != 0)
4453 		return (error);
4454 	do {
4455 		bwillwrite();
4456 		if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4457 			return (ESTALE);
4458 		error = VFS_FHTOVP(mp, &fh.fh_fid, LK_SHARED, &vp);
4459 		vfs_unbusy(mp);
4460 		if (error != 0)
4461 			return (error);
4462 		VOP_UNLOCK(vp);
4463 		error = kern_linkat_vp(td, vp, fd, path, pathseg);
4464 	} while (error == EAGAIN || error == ERELOOKUP);
4465 	return (error);
4466 }
4467 
4468 #ifndef _SYS_SYSPROTO_H_
4469 struct fhreadlink_args {
4470 	fhandle_t *fhp;
4471 	char *buf;
4472 	size_t bufsize;
4473 };
4474 #endif
4475 int
4476 sys_fhreadlink(struct thread *td, struct fhreadlink_args *uap)
4477 {
4478 	fhandle_t fh;
4479 	struct mount *mp;
4480 	struct vnode *vp;
4481 	int error;
4482 
4483 	error = priv_check(td, PRIV_VFS_GETFH);
4484 	if (error != 0)
4485 		return (error);
4486 	if (uap->bufsize > IOSIZE_MAX)
4487 		return (EINVAL);
4488 	error = copyin(uap->fhp, &fh, sizeof(fh));
4489 	if (error != 0)
4490 		return (error);
4491 	if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4492 		return (ESTALE);
4493 	error = VFS_FHTOVP(mp, &fh.fh_fid, LK_SHARED, &vp);
4494 	vfs_unbusy(mp);
4495 	if (error != 0)
4496 		return (error);
4497 	error = kern_readlink_vp(vp, uap->buf, UIO_USERSPACE, uap->bufsize, td);
4498 	vput(vp);
4499 	return (error);
4500 }
4501 
4502 /*
4503  * syscall for the rpc.lockd to use to translate a NFS file handle into an
4504  * open descriptor.
4505  *
4506  * warning: do not remove the priv_check() call or this becomes one giant
4507  * security hole.
4508  */
4509 #ifndef _SYS_SYSPROTO_H_
4510 struct fhopen_args {
4511 	const struct fhandle *u_fhp;
4512 	int flags;
4513 };
4514 #endif
4515 int
4516 sys_fhopen(struct thread *td, struct fhopen_args *uap)
4517 {
4518 	return (kern_fhopen(td, uap->u_fhp, uap->flags));
4519 }
4520 
4521 int
4522 kern_fhopen(struct thread *td, const struct fhandle *u_fhp, int flags)
4523 {
4524 	struct mount *mp;
4525 	struct vnode *vp;
4526 	struct fhandle fhp;
4527 	struct file *fp;
4528 	int fmode, error;
4529 	int indx;
4530 
4531 	error = priv_check(td, PRIV_VFS_FHOPEN);
4532 	if (error != 0)
4533 		return (error);
4534 	indx = -1;
4535 	fmode = FFLAGS(flags);
4536 	/* why not allow a non-read/write open for our lockd? */
4537 	if (((fmode & (FREAD | FWRITE)) == 0) || (fmode & O_CREAT))
4538 		return (EINVAL);
4539 	error = copyin(u_fhp, &fhp, sizeof(fhp));
4540 	if (error != 0)
4541 		return(error);
4542 	/* find the mount point */
4543 	mp = vfs_busyfs(&fhp.fh_fsid);
4544 	if (mp == NULL)
4545 		return (ESTALE);
4546 	/* now give me my vnode, it gets returned to me locked */
4547 	error = VFS_FHTOVP(mp, &fhp.fh_fid, LK_EXCLUSIVE, &vp);
4548 	vfs_unbusy(mp);
4549 	if (error != 0)
4550 		return (error);
4551 
4552 	error = falloc_noinstall(td, &fp);
4553 	if (error != 0) {
4554 		vput(vp);
4555 		return (error);
4556 	}
4557 	/*
4558 	 * An extra reference on `fp' has been held for us by
4559 	 * falloc_noinstall().
4560 	 */
4561 
4562 #ifdef INVARIANTS
4563 	td->td_dupfd = -1;
4564 #endif
4565 	error = vn_open_vnode(vp, fmode, td->td_ucred, td, fp);
4566 	if (error != 0) {
4567 		KASSERT(fp->f_ops == &badfileops,
4568 		    ("VOP_OPEN in fhopen() set f_ops"));
4569 		KASSERT(td->td_dupfd < 0,
4570 		    ("fhopen() encountered fdopen()"));
4571 
4572 		vput(vp);
4573 		goto bad;
4574 	}
4575 #ifdef INVARIANTS
4576 	td->td_dupfd = 0;
4577 #endif
4578 	fp->f_vnode = vp;
4579 	finit_vnode(fp, fmode, NULL, &vnops);
4580 	VOP_UNLOCK(vp);
4581 	if ((fmode & O_TRUNC) != 0) {
4582 		error = fo_truncate(fp, 0, td->td_ucred, td);
4583 		if (error != 0)
4584 			goto bad;
4585 	}
4586 
4587 	error = finstall(td, fp, &indx, fmode, NULL);
4588 bad:
4589 	fdrop(fp, td);
4590 	td->td_retval[0] = indx;
4591 	return (error);
4592 }
4593 
4594 /*
4595  * Stat an (NFS) file handle.
4596  */
4597 #ifndef _SYS_SYSPROTO_H_
4598 struct fhstat_args {
4599 	struct fhandle *u_fhp;
4600 	struct stat *sb;
4601 };
4602 #endif
4603 int
4604 sys_fhstat(struct thread *td, struct fhstat_args *uap)
4605 {
4606 	struct stat sb;
4607 	struct fhandle fh;
4608 	int error;
4609 
4610 	error = copyin(uap->u_fhp, &fh, sizeof(fh));
4611 	if (error != 0)
4612 		return (error);
4613 	error = kern_fhstat(td, fh, &sb);
4614 	if (error == 0)
4615 		error = copyout(&sb, uap->sb, sizeof(sb));
4616 	return (error);
4617 }
4618 
4619 int
4620 kern_fhstat(struct thread *td, struct fhandle fh, struct stat *sb)
4621 {
4622 	struct mount *mp;
4623 	struct vnode *vp;
4624 	int error;
4625 
4626 	error = priv_check(td, PRIV_VFS_FHSTAT);
4627 	if (error != 0)
4628 		return (error);
4629 	if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4630 		return (ESTALE);
4631 	error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4632 	vfs_unbusy(mp);
4633 	if (error != 0)
4634 		return (error);
4635 	error = VOP_STAT(vp, sb, td->td_ucred, NOCRED, td);
4636 	vput(vp);
4637 	return (error);
4638 }
4639 
4640 /*
4641  * Implement fstatfs() for (NFS) file handles.
4642  */
4643 #ifndef _SYS_SYSPROTO_H_
4644 struct fhstatfs_args {
4645 	struct fhandle *u_fhp;
4646 	struct statfs *buf;
4647 };
4648 #endif
4649 int
4650 sys_fhstatfs(struct thread *td, struct fhstatfs_args *uap)
4651 {
4652 	struct statfs *sfp;
4653 	fhandle_t fh;
4654 	int error;
4655 
4656 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
4657 	if (error != 0)
4658 		return (error);
4659 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
4660 	error = kern_fhstatfs(td, fh, sfp);
4661 	if (error == 0)
4662 		error = copyout(sfp, uap->buf, sizeof(*sfp));
4663 	free(sfp, M_STATFS);
4664 	return (error);
4665 }
4666 
4667 int
4668 kern_fhstatfs(struct thread *td, fhandle_t fh, struct statfs *buf)
4669 {
4670 	struct mount *mp;
4671 	struct vnode *vp;
4672 	int error;
4673 
4674 	error = priv_check(td, PRIV_VFS_FHSTATFS);
4675 	if (error != 0)
4676 		return (error);
4677 	if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4678 		return (ESTALE);
4679 	error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4680 	if (error != 0) {
4681 		vfs_unbusy(mp);
4682 		return (error);
4683 	}
4684 	vput(vp);
4685 	error = prison_canseemount(td->td_ucred, mp);
4686 	if (error != 0)
4687 		goto out;
4688 #ifdef MAC
4689 	error = mac_mount_check_stat(td->td_ucred, mp);
4690 	if (error != 0)
4691 		goto out;
4692 #endif
4693 	error = VFS_STATFS(mp, buf);
4694 out:
4695 	vfs_unbusy(mp);
4696 	return (error);
4697 }
4698 
4699 /*
4700  * Unlike madvise(2), we do not make a best effort to remember every
4701  * possible caching hint.  Instead, we remember the last setting with
4702  * the exception that we will allow POSIX_FADV_NORMAL to adjust the
4703  * region of any current setting.
4704  */
4705 int
4706 kern_posix_fadvise(struct thread *td, int fd, off_t offset, off_t len,
4707     int advice)
4708 {
4709 	struct fadvise_info *fa, *new;
4710 	struct file *fp;
4711 	struct vnode *vp;
4712 	off_t end;
4713 	int error;
4714 
4715 	if (offset < 0 || len < 0 || offset > OFF_MAX - len)
4716 		return (EINVAL);
4717 	AUDIT_ARG_VALUE(advice);
4718 	switch (advice) {
4719 	case POSIX_FADV_SEQUENTIAL:
4720 	case POSIX_FADV_RANDOM:
4721 	case POSIX_FADV_NOREUSE:
4722 		new = malloc(sizeof(*fa), M_FADVISE, M_WAITOK);
4723 		break;
4724 	case POSIX_FADV_NORMAL:
4725 	case POSIX_FADV_WILLNEED:
4726 	case POSIX_FADV_DONTNEED:
4727 		new = NULL;
4728 		break;
4729 	default:
4730 		return (EINVAL);
4731 	}
4732 	/* XXX: CAP_POSIX_FADVISE? */
4733 	AUDIT_ARG_FD(fd);
4734 	error = fget(td, fd, &cap_no_rights, &fp);
4735 	if (error != 0)
4736 		goto out;
4737 	AUDIT_ARG_FILE(td->td_proc, fp);
4738 	if ((fp->f_ops->fo_flags & DFLAG_SEEKABLE) == 0) {
4739 		error = ESPIPE;
4740 		goto out;
4741 	}
4742 	if (fp->f_type != DTYPE_VNODE) {
4743 		error = ENODEV;
4744 		goto out;
4745 	}
4746 	vp = fp->f_vnode;
4747 	if (vp->v_type != VREG) {
4748 		error = ENODEV;
4749 		goto out;
4750 	}
4751 	if (len == 0)
4752 		end = OFF_MAX;
4753 	else
4754 		end = offset + len - 1;
4755 	switch (advice) {
4756 	case POSIX_FADV_SEQUENTIAL:
4757 	case POSIX_FADV_RANDOM:
4758 	case POSIX_FADV_NOREUSE:
4759 		/*
4760 		 * Try to merge any existing non-standard region with
4761 		 * this new region if possible, otherwise create a new
4762 		 * non-standard region for this request.
4763 		 */
4764 		mtx_pool_lock(mtxpool_sleep, fp);
4765 		fa = fp->f_advice;
4766 		if (fa != NULL && fa->fa_advice == advice &&
4767 		    ((fa->fa_start <= end && fa->fa_end >= offset) ||
4768 		    (end != OFF_MAX && fa->fa_start == end + 1) ||
4769 		    (fa->fa_end != OFF_MAX && fa->fa_end + 1 == offset))) {
4770 			if (offset < fa->fa_start)
4771 				fa->fa_start = offset;
4772 			if (end > fa->fa_end)
4773 				fa->fa_end = end;
4774 		} else {
4775 			new->fa_advice = advice;
4776 			new->fa_start = offset;
4777 			new->fa_end = end;
4778 			fp->f_advice = new;
4779 			new = fa;
4780 		}
4781 		mtx_pool_unlock(mtxpool_sleep, fp);
4782 		break;
4783 	case POSIX_FADV_NORMAL:
4784 		/*
4785 		 * If a the "normal" region overlaps with an existing
4786 		 * non-standard region, trim or remove the
4787 		 * non-standard region.
4788 		 */
4789 		mtx_pool_lock(mtxpool_sleep, fp);
4790 		fa = fp->f_advice;
4791 		if (fa != NULL) {
4792 			if (offset <= fa->fa_start && end >= fa->fa_end) {
4793 				new = fa;
4794 				fp->f_advice = NULL;
4795 			} else if (offset <= fa->fa_start &&
4796 			    end >= fa->fa_start)
4797 				fa->fa_start = end + 1;
4798 			else if (offset <= fa->fa_end && end >= fa->fa_end)
4799 				fa->fa_end = offset - 1;
4800 			else if (offset >= fa->fa_start && end <= fa->fa_end) {
4801 				/*
4802 				 * If the "normal" region is a middle
4803 				 * portion of the existing
4804 				 * non-standard region, just remove
4805 				 * the whole thing rather than picking
4806 				 * one side or the other to
4807 				 * preserve.
4808 				 */
4809 				new = fa;
4810 				fp->f_advice = NULL;
4811 			}
4812 		}
4813 		mtx_pool_unlock(mtxpool_sleep, fp);
4814 		break;
4815 	case POSIX_FADV_WILLNEED:
4816 	case POSIX_FADV_DONTNEED:
4817 		error = VOP_ADVISE(vp, offset, end, advice);
4818 		break;
4819 	}
4820 out:
4821 	if (fp != NULL)
4822 		fdrop(fp, td);
4823 	free(new, M_FADVISE);
4824 	return (error);
4825 }
4826 
4827 int
4828 sys_posix_fadvise(struct thread *td, struct posix_fadvise_args *uap)
4829 {
4830 	int error;
4831 
4832 	error = kern_posix_fadvise(td, uap->fd, uap->offset, uap->len,
4833 	    uap->advice);
4834 	return (kern_posix_error(td, error));
4835 }
4836 
4837 int
4838 kern_copy_file_range(struct thread *td, int infd, off_t *inoffp, int outfd,
4839     off_t *outoffp, size_t len, unsigned int flags)
4840 {
4841 	struct file *infp, *outfp;
4842 	struct vnode *invp, *outvp;
4843 	int error;
4844 	size_t retlen;
4845 	void *rl_rcookie, *rl_wcookie;
4846 	off_t savinoff, savoutoff;
4847 
4848 	infp = outfp = NULL;
4849 	rl_rcookie = rl_wcookie = NULL;
4850 	savinoff = -1;
4851 	error = 0;
4852 	retlen = 0;
4853 
4854 	if (flags != 0) {
4855 		error = EINVAL;
4856 		goto out;
4857 	}
4858 	if (len > SSIZE_MAX)
4859 		/*
4860 		 * Although the len argument is size_t, the return argument
4861 		 * is ssize_t (which is signed).  Therefore a size that won't
4862 		 * fit in ssize_t can't be returned.
4863 		 */
4864 		len = SSIZE_MAX;
4865 
4866 	/* Get the file structures for the file descriptors. */
4867 	error = fget_read(td, infd, &cap_read_rights, &infp);
4868 	if (error != 0)
4869 		goto out;
4870 	if (infp->f_ops == &badfileops) {
4871 		error = EBADF;
4872 		goto out;
4873 	}
4874 	if (infp->f_vnode == NULL) {
4875 		error = EINVAL;
4876 		goto out;
4877 	}
4878 	error = fget_write(td, outfd, &cap_write_rights, &outfp);
4879 	if (error != 0)
4880 		goto out;
4881 	if (outfp->f_ops == &badfileops) {
4882 		error = EBADF;
4883 		goto out;
4884 	}
4885 	if (outfp->f_vnode == NULL) {
4886 		error = EINVAL;
4887 		goto out;
4888 	}
4889 
4890 	/* Set the offset pointers to the correct place. */
4891 	if (inoffp == NULL)
4892 		inoffp = &infp->f_offset;
4893 	if (outoffp == NULL)
4894 		outoffp = &outfp->f_offset;
4895 	savinoff = *inoffp;
4896 	savoutoff = *outoffp;
4897 
4898 	invp = infp->f_vnode;
4899 	outvp = outfp->f_vnode;
4900 	/* Sanity check the f_flag bits. */
4901 	if ((outfp->f_flag & (FWRITE | FAPPEND)) != FWRITE ||
4902 	    (infp->f_flag & FREAD) == 0) {
4903 		error = EBADF;
4904 		goto out;
4905 	}
4906 
4907 	/* If len == 0, just return 0. */
4908 	if (len == 0)
4909 		goto out;
4910 
4911 	/*
4912 	 * If infp and outfp refer to the same file, the byte ranges cannot
4913 	 * overlap.
4914 	 */
4915 	if (invp == outvp && ((savinoff <= savoutoff && savinoff + len >
4916 	    savoutoff) || (savinoff > savoutoff && savoutoff + len >
4917 	    savinoff))) {
4918 		error = EINVAL;
4919 		goto out;
4920 	}
4921 
4922 	/* Range lock the byte ranges for both invp and outvp. */
4923 	for (;;) {
4924 		rl_wcookie = vn_rangelock_wlock(outvp, *outoffp, *outoffp +
4925 		    len);
4926 		rl_rcookie = vn_rangelock_tryrlock(invp, *inoffp, *inoffp +
4927 		    len);
4928 		if (rl_rcookie != NULL)
4929 			break;
4930 		vn_rangelock_unlock(outvp, rl_wcookie);
4931 		rl_rcookie = vn_rangelock_rlock(invp, *inoffp, *inoffp + len);
4932 		vn_rangelock_unlock(invp, rl_rcookie);
4933 	}
4934 
4935 	retlen = len;
4936 	error = vn_copy_file_range(invp, inoffp, outvp, outoffp, &retlen,
4937 	    flags, infp->f_cred, outfp->f_cred, td);
4938 out:
4939 	if (rl_rcookie != NULL)
4940 		vn_rangelock_unlock(invp, rl_rcookie);
4941 	if (rl_wcookie != NULL)
4942 		vn_rangelock_unlock(outvp, rl_wcookie);
4943 	if (savinoff != -1 && (error == EINTR || error == ERESTART)) {
4944 		*inoffp = savinoff;
4945 		*outoffp = savoutoff;
4946 	}
4947 	if (outfp != NULL)
4948 		fdrop(outfp, td);
4949 	if (infp != NULL)
4950 		fdrop(infp, td);
4951 	td->td_retval[0] = retlen;
4952 	return (error);
4953 }
4954 
4955 int
4956 sys_copy_file_range(struct thread *td, struct copy_file_range_args *uap)
4957 {
4958 	off_t inoff, outoff, *inoffp, *outoffp;
4959 	int error;
4960 
4961 	inoffp = outoffp = NULL;
4962 	if (uap->inoffp != NULL) {
4963 		error = copyin(uap->inoffp, &inoff, sizeof(off_t));
4964 		if (error != 0)
4965 			return (error);
4966 		inoffp = &inoff;
4967 	}
4968 	if (uap->outoffp != NULL) {
4969 		error = copyin(uap->outoffp, &outoff, sizeof(off_t));
4970 		if (error != 0)
4971 			return (error);
4972 		outoffp = &outoff;
4973 	}
4974 	error = kern_copy_file_range(td, uap->infd, inoffp, uap->outfd,
4975 	    outoffp, uap->len, uap->flags);
4976 	if (error == 0 && uap->inoffp != NULL)
4977 		error = copyout(inoffp, uap->inoffp, sizeof(off_t));
4978 	if (error == 0 && uap->outoffp != NULL)
4979 		error = copyout(outoffp, uap->outoffp, sizeof(off_t));
4980 	return (error);
4981 }
4982