xref: /freebsd/sys/kern/vfs_syscalls.c (revision f5f47d5068fb97df18eb114a66ae8ef51a0b3c8c)
1 /*-
2  * Copyright (c) 1989, 1993
3  *	The Regents of the University of California.  All rights reserved.
4  * (c) UNIX System Laboratories, Inc.
5  * All or some portions of this file are derived from material licensed
6  * to the University of California by American Telephone and Telegraph
7  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
8  * the permission of UNIX System Laboratories, Inc.
9  *
10  * Redistribution and use in source and binary forms, with or without
11  * modification, are permitted provided that the following conditions
12  * are met:
13  * 1. Redistributions of source code must retain the above copyright
14  *    notice, this list of conditions and the following disclaimer.
15  * 2. Redistributions in binary form must reproduce the above copyright
16  *    notice, this list of conditions and the following disclaimer in the
17  *    documentation and/or other materials provided with the distribution.
18  * 4. Neither the name of the University nor the names of its contributors
19  *    may be used to endorse or promote products derived from this software
20  *    without specific prior written permission.
21  *
22  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
23  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
24  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
25  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
26  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
27  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
28  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
29  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
30  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
31  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32  * SUCH DAMAGE.
33  *
34  *	@(#)vfs_syscalls.c	8.13 (Berkeley) 4/15/94
35  */
36 
37 #include <sys/cdefs.h>
38 __FBSDID("$FreeBSD$");
39 
40 #include "opt_capsicum.h"
41 #include "opt_compat.h"
42 #include "opt_kdtrace.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/capability.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/sdt.h>
65 #include <sys/stat.h>
66 #include <sys/sx.h>
67 #include <sys/unistd.h>
68 #include <sys/vnode.h>
69 #include <sys/priv.h>
70 #include <sys/proc.h>
71 #include <sys/dirent.h>
72 #include <sys/jail.h>
73 #include <sys/syscallsubr.h>
74 #include <sys/sysctl.h>
75 #ifdef KTRACE
76 #include <sys/ktrace.h>
77 #endif
78 
79 #include <machine/stdarg.h>
80 
81 #include <security/audit/audit.h>
82 #include <security/mac/mac_framework.h>
83 
84 #include <vm/vm.h>
85 #include <vm/vm_object.h>
86 #include <vm/vm_page.h>
87 #include <vm/uma.h>
88 
89 #include <ufs/ufs/quota.h>
90 
91 MALLOC_DEFINE(M_FADVISE, "fadvise", "posix_fadvise(2) information");
92 
93 SDT_PROVIDER_DEFINE(vfs);
94 SDT_PROBE_DEFINE(vfs, , stat, mode, mode);
95 SDT_PROBE_ARGTYPE(vfs, , stat, mode, 0, "char *");
96 SDT_PROBE_ARGTYPE(vfs, , stat, mode, 1, "int");
97 SDT_PROBE_DEFINE(vfs, , stat, reg, reg);
98 SDT_PROBE_ARGTYPE(vfs, , stat, reg, 0, "char *");
99 SDT_PROBE_ARGTYPE(vfs, , stat, reg, 1, "int");
100 
101 static int chroot_refuse_vdir_fds(struct filedesc *fdp);
102 static int getutimes(const struct timeval *, enum uio_seg, struct timespec *);
103 static int setfflags(struct thread *td, struct vnode *, int);
104 static int setutimes(struct thread *td, struct vnode *,
105     const struct timespec *, int, int);
106 static int vn_access(struct vnode *vp, int user_flags, struct ucred *cred,
107     struct thread *td);
108 
109 /*
110  * The module initialization routine for POSIX asynchronous I/O will
111  * set this to the version of AIO that it implements.  (Zero means
112  * that it is not implemented.)  This value is used here by pathconf()
113  * and in kern_descrip.c by fpathconf().
114  */
115 int async_io_version;
116 
117 #ifdef DEBUG
118 static int syncprt = 0;
119 SYSCTL_INT(_debug, OID_AUTO, syncprt, CTLFLAG_RW, &syncprt, 0, "");
120 #endif
121 
122 /*
123  * Sync each mounted filesystem.
124  */
125 #ifndef _SYS_SYSPROTO_H_
126 struct sync_args {
127 	int     dummy;
128 };
129 #endif
130 /* ARGSUSED */
131 int
132 sys_sync(td, uap)
133 	struct thread *td;
134 	struct sync_args *uap;
135 {
136 	struct mount *mp, *nmp;
137 	int save, vfslocked;
138 
139 	mtx_lock(&mountlist_mtx);
140 	for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
141 		if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK)) {
142 			nmp = TAILQ_NEXT(mp, mnt_list);
143 			continue;
144 		}
145 		vfslocked = VFS_LOCK_GIANT(mp);
146 		if ((mp->mnt_flag & MNT_RDONLY) == 0 &&
147 		    vn_start_write(NULL, &mp, V_NOWAIT) == 0) {
148 			save = curthread_pflags_set(TDP_SYNCIO);
149 			vfs_msync(mp, MNT_NOWAIT);
150 			VFS_SYNC(mp, MNT_NOWAIT);
151 			curthread_pflags_restore(save);
152 			vn_finished_write(mp);
153 		}
154 		VFS_UNLOCK_GIANT(vfslocked);
155 		mtx_lock(&mountlist_mtx);
156 		nmp = TAILQ_NEXT(mp, mnt_list);
157 		vfs_unbusy(mp);
158 	}
159 	mtx_unlock(&mountlist_mtx);
160 	return (0);
161 }
162 
163 /*
164  * Change filesystem quotas.
165  */
166 #ifndef _SYS_SYSPROTO_H_
167 struct quotactl_args {
168 	char *path;
169 	int cmd;
170 	int uid;
171 	caddr_t arg;
172 };
173 #endif
174 int
175 sys_quotactl(td, uap)
176 	struct thread *td;
177 	register struct quotactl_args /* {
178 		char *path;
179 		int cmd;
180 		int uid;
181 		caddr_t arg;
182 	} */ *uap;
183 {
184 	struct mount *mp;
185 	int vfslocked;
186 	int error;
187 	struct nameidata nd;
188 
189 	AUDIT_ARG_CMD(uap->cmd);
190 	AUDIT_ARG_UID(uap->uid);
191 	if (!prison_allow(td->td_ucred, PR_ALLOW_QUOTAS))
192 		return (EPERM);
193 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | MPSAFE | AUDITVNODE1,
194 	   UIO_USERSPACE, uap->path, td);
195 	if ((error = namei(&nd)) != 0)
196 		return (error);
197 	vfslocked = NDHASGIANT(&nd);
198 	NDFREE(&nd, NDF_ONLY_PNBUF);
199 	mp = nd.ni_vp->v_mount;
200 	vfs_ref(mp);
201 	vput(nd.ni_vp);
202 	error = vfs_busy(mp, 0);
203 	vfs_rel(mp);
204 	if (error) {
205 		VFS_UNLOCK_GIANT(vfslocked);
206 		return (error);
207 	}
208 	error = VFS_QUOTACTL(mp, uap->cmd, uap->uid, uap->arg);
209 
210 	/*
211 	 * Since quota on operation typically needs to open quota
212 	 * file, the Q_QUOTAON handler needs to unbusy the mount point
213 	 * before calling into namei.  Otherwise, unmount might be
214 	 * started between two vfs_busy() invocations (first is our,
215 	 * second is from mount point cross-walk code in lookup()),
216 	 * causing deadlock.
217 	 *
218 	 * Require that Q_QUOTAON handles the vfs_busy() reference on
219 	 * its own, always returning with ubusied mount point.
220 	 */
221 	if ((uap->cmd >> SUBCMDSHIFT) != Q_QUOTAON)
222 		vfs_unbusy(mp);
223 	VFS_UNLOCK_GIANT(vfslocked);
224 	return (error);
225 }
226 
227 /*
228  * Used by statfs conversion routines to scale the block size up if
229  * necessary so that all of the block counts are <= 'max_size'.  Note
230  * that 'max_size' should be a bitmask, i.e. 2^n - 1 for some non-zero
231  * value of 'n'.
232  */
233 void
234 statfs_scale_blocks(struct statfs *sf, long max_size)
235 {
236 	uint64_t count;
237 	int shift;
238 
239 	KASSERT(powerof2(max_size + 1), ("%s: invalid max_size", __func__));
240 
241 	/*
242 	 * Attempt to scale the block counts to give a more accurate
243 	 * overview to userland of the ratio of free space to used
244 	 * space.  To do this, find the largest block count and compute
245 	 * a divisor that lets it fit into a signed integer <= max_size.
246 	 */
247 	if (sf->f_bavail < 0)
248 		count = -sf->f_bavail;
249 	else
250 		count = sf->f_bavail;
251 	count = MAX(sf->f_blocks, MAX(sf->f_bfree, count));
252 	if (count <= max_size)
253 		return;
254 
255 	count >>= flsl(max_size);
256 	shift = 0;
257 	while (count > 0) {
258 		shift++;
259 		count >>=1;
260 	}
261 
262 	sf->f_bsize <<= shift;
263 	sf->f_blocks >>= shift;
264 	sf->f_bfree >>= shift;
265 	sf->f_bavail >>= shift;
266 }
267 
268 /*
269  * Get filesystem statistics.
270  */
271 #ifndef _SYS_SYSPROTO_H_
272 struct statfs_args {
273 	char *path;
274 	struct statfs *buf;
275 };
276 #endif
277 int
278 sys_statfs(td, uap)
279 	struct thread *td;
280 	register struct statfs_args /* {
281 		char *path;
282 		struct statfs *buf;
283 	} */ *uap;
284 {
285 	struct statfs sf;
286 	int error;
287 
288 	error = kern_statfs(td, uap->path, UIO_USERSPACE, &sf);
289 	if (error == 0)
290 		error = copyout(&sf, uap->buf, sizeof(sf));
291 	return (error);
292 }
293 
294 int
295 kern_statfs(struct thread *td, char *path, enum uio_seg pathseg,
296     struct statfs *buf)
297 {
298 	struct mount *mp;
299 	struct statfs *sp, sb;
300 	int vfslocked;
301 	int error;
302 	struct nameidata nd;
303 
304 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | MPSAFE |
305 	    AUDITVNODE1, pathseg, path, td);
306 	error = namei(&nd);
307 	if (error)
308 		return (error);
309 	vfslocked = NDHASGIANT(&nd);
310 	mp = nd.ni_vp->v_mount;
311 	vfs_ref(mp);
312 	NDFREE(&nd, NDF_ONLY_PNBUF);
313 	vput(nd.ni_vp);
314 	error = vfs_busy(mp, 0);
315 	vfs_rel(mp);
316 	if (error) {
317 		VFS_UNLOCK_GIANT(vfslocked);
318 		return (error);
319 	}
320 #ifdef MAC
321 	error = mac_mount_check_stat(td->td_ucred, mp);
322 	if (error)
323 		goto out;
324 #endif
325 	/*
326 	 * Set these in case the underlying filesystem fails to do so.
327 	 */
328 	sp = &mp->mnt_stat;
329 	sp->f_version = STATFS_VERSION;
330 	sp->f_namemax = NAME_MAX;
331 	sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
332 	error = VFS_STATFS(mp, sp);
333 	if (error)
334 		goto out;
335 	if (priv_check(td, PRIV_VFS_GENERATION)) {
336 		bcopy(sp, &sb, sizeof(sb));
337 		sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
338 		prison_enforce_statfs(td->td_ucred, mp, &sb);
339 		sp = &sb;
340 	}
341 	*buf = *sp;
342 out:
343 	vfs_unbusy(mp);
344 	VFS_UNLOCK_GIANT(vfslocked);
345 	return (error);
346 }
347 
348 /*
349  * Get filesystem statistics.
350  */
351 #ifndef _SYS_SYSPROTO_H_
352 struct fstatfs_args {
353 	int fd;
354 	struct statfs *buf;
355 };
356 #endif
357 int
358 sys_fstatfs(td, uap)
359 	struct thread *td;
360 	register struct fstatfs_args /* {
361 		int fd;
362 		struct statfs *buf;
363 	} */ *uap;
364 {
365 	struct statfs sf;
366 	int error;
367 
368 	error = kern_fstatfs(td, uap->fd, &sf);
369 	if (error == 0)
370 		error = copyout(&sf, uap->buf, sizeof(sf));
371 	return (error);
372 }
373 
374 int
375 kern_fstatfs(struct thread *td, int fd, struct statfs *buf)
376 {
377 	struct file *fp;
378 	struct mount *mp;
379 	struct statfs *sp, sb;
380 	int vfslocked;
381 	struct vnode *vp;
382 	int error;
383 
384 	AUDIT_ARG_FD(fd);
385 	error = getvnode(td->td_proc->p_fd, fd, CAP_FSTATFS, &fp);
386 	if (error)
387 		return (error);
388 	vp = fp->f_vnode;
389 	vfslocked = VFS_LOCK_GIANT(vp->v_mount);
390 	vn_lock(vp, LK_SHARED | LK_RETRY);
391 #ifdef AUDIT
392 	AUDIT_ARG_VNODE1(vp);
393 #endif
394 	mp = vp->v_mount;
395 	if (mp)
396 		vfs_ref(mp);
397 	VOP_UNLOCK(vp, 0);
398 	fdrop(fp, td);
399 	if (mp == NULL) {
400 		error = EBADF;
401 		goto out;
402 	}
403 	error = vfs_busy(mp, 0);
404 	vfs_rel(mp);
405 	if (error) {
406 		VFS_UNLOCK_GIANT(vfslocked);
407 		return (error);
408 	}
409 #ifdef MAC
410 	error = mac_mount_check_stat(td->td_ucred, mp);
411 	if (error)
412 		goto out;
413 #endif
414 	/*
415 	 * Set these in case the underlying filesystem fails to do so.
416 	 */
417 	sp = &mp->mnt_stat;
418 	sp->f_version = STATFS_VERSION;
419 	sp->f_namemax = NAME_MAX;
420 	sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
421 	error = VFS_STATFS(mp, sp);
422 	if (error)
423 		goto out;
424 	if (priv_check(td, PRIV_VFS_GENERATION)) {
425 		bcopy(sp, &sb, sizeof(sb));
426 		sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
427 		prison_enforce_statfs(td->td_ucred, mp, &sb);
428 		sp = &sb;
429 	}
430 	*buf = *sp;
431 out:
432 	if (mp)
433 		vfs_unbusy(mp);
434 	VFS_UNLOCK_GIANT(vfslocked);
435 	return (error);
436 }
437 
438 /*
439  * Get statistics on all filesystems.
440  */
441 #ifndef _SYS_SYSPROTO_H_
442 struct getfsstat_args {
443 	struct statfs *buf;
444 	long bufsize;
445 	int flags;
446 };
447 #endif
448 int
449 sys_getfsstat(td, uap)
450 	struct thread *td;
451 	register struct getfsstat_args /* {
452 		struct statfs *buf;
453 		long bufsize;
454 		int flags;
455 	} */ *uap;
456 {
457 
458 	return (kern_getfsstat(td, &uap->buf, uap->bufsize, UIO_USERSPACE,
459 	    uap->flags));
460 }
461 
462 /*
463  * If (bufsize > 0 && bufseg == UIO_SYSSPACE)
464  * 	The caller is responsible for freeing memory which will be allocated
465  *	in '*buf'.
466  */
467 int
468 kern_getfsstat(struct thread *td, struct statfs **buf, size_t bufsize,
469     enum uio_seg bufseg, int flags)
470 {
471 	struct mount *mp, *nmp;
472 	struct statfs *sfsp, *sp, sb;
473 	size_t count, maxcount;
474 	int vfslocked;
475 	int error;
476 
477 	maxcount = bufsize / sizeof(struct statfs);
478 	if (bufsize == 0)
479 		sfsp = NULL;
480 	else if (bufseg == UIO_USERSPACE)
481 		sfsp = *buf;
482 	else /* if (bufseg == UIO_SYSSPACE) */ {
483 		count = 0;
484 		mtx_lock(&mountlist_mtx);
485 		TAILQ_FOREACH(mp, &mountlist, mnt_list) {
486 			count++;
487 		}
488 		mtx_unlock(&mountlist_mtx);
489 		if (maxcount > count)
490 			maxcount = count;
491 		sfsp = *buf = malloc(maxcount * sizeof(struct statfs), M_TEMP,
492 		    M_WAITOK);
493 	}
494 	count = 0;
495 	mtx_lock(&mountlist_mtx);
496 	for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
497 		if (prison_canseemount(td->td_ucred, mp) != 0) {
498 			nmp = TAILQ_NEXT(mp, mnt_list);
499 			continue;
500 		}
501 #ifdef MAC
502 		if (mac_mount_check_stat(td->td_ucred, mp) != 0) {
503 			nmp = TAILQ_NEXT(mp, mnt_list);
504 			continue;
505 		}
506 #endif
507 		if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK)) {
508 			nmp = TAILQ_NEXT(mp, mnt_list);
509 			continue;
510 		}
511 		vfslocked = VFS_LOCK_GIANT(mp);
512 		if (sfsp && count < maxcount) {
513 			sp = &mp->mnt_stat;
514 			/*
515 			 * Set these in case the underlying filesystem
516 			 * fails to do so.
517 			 */
518 			sp->f_version = STATFS_VERSION;
519 			sp->f_namemax = NAME_MAX;
520 			sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
521 			/*
522 			 * If MNT_NOWAIT or MNT_LAZY is specified, do not
523 			 * refresh the fsstat cache. MNT_NOWAIT or MNT_LAZY
524 			 * overrides MNT_WAIT.
525 			 */
526 			if (((flags & (MNT_LAZY|MNT_NOWAIT)) == 0 ||
527 			    (flags & MNT_WAIT)) &&
528 			    (error = VFS_STATFS(mp, sp))) {
529 				VFS_UNLOCK_GIANT(vfslocked);
530 				mtx_lock(&mountlist_mtx);
531 				nmp = TAILQ_NEXT(mp, mnt_list);
532 				vfs_unbusy(mp);
533 				continue;
534 			}
535 			if (priv_check(td, PRIV_VFS_GENERATION)) {
536 				bcopy(sp, &sb, sizeof(sb));
537 				sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
538 				prison_enforce_statfs(td->td_ucred, mp, &sb);
539 				sp = &sb;
540 			}
541 			if (bufseg == UIO_SYSSPACE)
542 				bcopy(sp, sfsp, sizeof(*sp));
543 			else /* if (bufseg == UIO_USERSPACE) */ {
544 				error = copyout(sp, sfsp, sizeof(*sp));
545 				if (error) {
546 					vfs_unbusy(mp);
547 					VFS_UNLOCK_GIANT(vfslocked);
548 					return (error);
549 				}
550 			}
551 			sfsp++;
552 		}
553 		VFS_UNLOCK_GIANT(vfslocked);
554 		count++;
555 		mtx_lock(&mountlist_mtx);
556 		nmp = TAILQ_NEXT(mp, mnt_list);
557 		vfs_unbusy(mp);
558 	}
559 	mtx_unlock(&mountlist_mtx);
560 	if (sfsp && count > maxcount)
561 		td->td_retval[0] = maxcount;
562 	else
563 		td->td_retval[0] = count;
564 	return (0);
565 }
566 
567 #ifdef COMPAT_FREEBSD4
568 /*
569  * Get old format filesystem statistics.
570  */
571 static void cvtstatfs(struct statfs *, struct ostatfs *);
572 
573 #ifndef _SYS_SYSPROTO_H_
574 struct freebsd4_statfs_args {
575 	char *path;
576 	struct ostatfs *buf;
577 };
578 #endif
579 int
580 freebsd4_statfs(td, uap)
581 	struct thread *td;
582 	struct freebsd4_statfs_args /* {
583 		char *path;
584 		struct ostatfs *buf;
585 	} */ *uap;
586 {
587 	struct ostatfs osb;
588 	struct statfs sf;
589 	int error;
590 
591 	error = kern_statfs(td, uap->path, UIO_USERSPACE, &sf);
592 	if (error)
593 		return (error);
594 	cvtstatfs(&sf, &osb);
595 	return (copyout(&osb, uap->buf, sizeof(osb)));
596 }
597 
598 /*
599  * Get filesystem statistics.
600  */
601 #ifndef _SYS_SYSPROTO_H_
602 struct freebsd4_fstatfs_args {
603 	int fd;
604 	struct ostatfs *buf;
605 };
606 #endif
607 int
608 freebsd4_fstatfs(td, uap)
609 	struct thread *td;
610 	struct freebsd4_fstatfs_args /* {
611 		int fd;
612 		struct ostatfs *buf;
613 	} */ *uap;
614 {
615 	struct ostatfs osb;
616 	struct statfs sf;
617 	int error;
618 
619 	error = kern_fstatfs(td, uap->fd, &sf);
620 	if (error)
621 		return (error);
622 	cvtstatfs(&sf, &osb);
623 	return (copyout(&osb, uap->buf, sizeof(osb)));
624 }
625 
626 /*
627  * Get statistics on all filesystems.
628  */
629 #ifndef _SYS_SYSPROTO_H_
630 struct freebsd4_getfsstat_args {
631 	struct ostatfs *buf;
632 	long bufsize;
633 	int flags;
634 };
635 #endif
636 int
637 freebsd4_getfsstat(td, uap)
638 	struct thread *td;
639 	register struct freebsd4_getfsstat_args /* {
640 		struct ostatfs *buf;
641 		long bufsize;
642 		int flags;
643 	} */ *uap;
644 {
645 	struct statfs *buf, *sp;
646 	struct ostatfs osb;
647 	size_t count, size;
648 	int error;
649 
650 	count = uap->bufsize / sizeof(struct ostatfs);
651 	size = count * sizeof(struct statfs);
652 	error = kern_getfsstat(td, &buf, size, UIO_SYSSPACE, uap->flags);
653 	if (size > 0) {
654 		count = td->td_retval[0];
655 		sp = buf;
656 		while (count > 0 && error == 0) {
657 			cvtstatfs(sp, &osb);
658 			error = copyout(&osb, uap->buf, sizeof(osb));
659 			sp++;
660 			uap->buf++;
661 			count--;
662 		}
663 		free(buf, M_TEMP);
664 	}
665 	return (error);
666 }
667 
668 /*
669  * Implement fstatfs() for (NFS) file handles.
670  */
671 #ifndef _SYS_SYSPROTO_H_
672 struct freebsd4_fhstatfs_args {
673 	struct fhandle *u_fhp;
674 	struct ostatfs *buf;
675 };
676 #endif
677 int
678 freebsd4_fhstatfs(td, uap)
679 	struct thread *td;
680 	struct freebsd4_fhstatfs_args /* {
681 		struct fhandle *u_fhp;
682 		struct ostatfs *buf;
683 	} */ *uap;
684 {
685 	struct ostatfs osb;
686 	struct statfs sf;
687 	fhandle_t fh;
688 	int error;
689 
690 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
691 	if (error)
692 		return (error);
693 	error = kern_fhstatfs(td, fh, &sf);
694 	if (error)
695 		return (error);
696 	cvtstatfs(&sf, &osb);
697 	return (copyout(&osb, uap->buf, sizeof(osb)));
698 }
699 
700 /*
701  * Convert a new format statfs structure to an old format statfs structure.
702  */
703 static void
704 cvtstatfs(nsp, osp)
705 	struct statfs *nsp;
706 	struct ostatfs *osp;
707 {
708 
709 	statfs_scale_blocks(nsp, LONG_MAX);
710 	bzero(osp, sizeof(*osp));
711 	osp->f_bsize = nsp->f_bsize;
712 	osp->f_iosize = MIN(nsp->f_iosize, LONG_MAX);
713 	osp->f_blocks = nsp->f_blocks;
714 	osp->f_bfree = nsp->f_bfree;
715 	osp->f_bavail = nsp->f_bavail;
716 	osp->f_files = MIN(nsp->f_files, LONG_MAX);
717 	osp->f_ffree = MIN(nsp->f_ffree, LONG_MAX);
718 	osp->f_owner = nsp->f_owner;
719 	osp->f_type = nsp->f_type;
720 	osp->f_flags = nsp->f_flags;
721 	osp->f_syncwrites = MIN(nsp->f_syncwrites, LONG_MAX);
722 	osp->f_asyncwrites = MIN(nsp->f_asyncwrites, LONG_MAX);
723 	osp->f_syncreads = MIN(nsp->f_syncreads, LONG_MAX);
724 	osp->f_asyncreads = MIN(nsp->f_asyncreads, LONG_MAX);
725 	strlcpy(osp->f_fstypename, nsp->f_fstypename,
726 	    MIN(MFSNAMELEN, OMFSNAMELEN));
727 	strlcpy(osp->f_mntonname, nsp->f_mntonname,
728 	    MIN(MNAMELEN, OMNAMELEN));
729 	strlcpy(osp->f_mntfromname, nsp->f_mntfromname,
730 	    MIN(MNAMELEN, OMNAMELEN));
731 	osp->f_fsid = nsp->f_fsid;
732 }
733 #endif /* COMPAT_FREEBSD4 */
734 
735 /*
736  * Change current working directory to a given file descriptor.
737  */
738 #ifndef _SYS_SYSPROTO_H_
739 struct fchdir_args {
740 	int	fd;
741 };
742 #endif
743 int
744 sys_fchdir(td, uap)
745 	struct thread *td;
746 	struct fchdir_args /* {
747 		int fd;
748 	} */ *uap;
749 {
750 	register struct filedesc *fdp = td->td_proc->p_fd;
751 	struct vnode *vp, *tdp, *vpold;
752 	struct mount *mp;
753 	struct file *fp;
754 	int vfslocked;
755 	int error;
756 
757 	AUDIT_ARG_FD(uap->fd);
758 	if ((error = getvnode(fdp, uap->fd, CAP_FCHDIR, &fp)) != 0)
759 		return (error);
760 	vp = fp->f_vnode;
761 	VREF(vp);
762 	fdrop(fp, td);
763 	vfslocked = VFS_LOCK_GIANT(vp->v_mount);
764 	vn_lock(vp, LK_SHARED | LK_RETRY);
765 	AUDIT_ARG_VNODE1(vp);
766 	error = change_dir(vp, td);
767 	while (!error && (mp = vp->v_mountedhere) != NULL) {
768 		int tvfslocked;
769 		if (vfs_busy(mp, 0))
770 			continue;
771 		tvfslocked = VFS_LOCK_GIANT(mp);
772 		error = VFS_ROOT(mp, LK_SHARED, &tdp);
773 		vfs_unbusy(mp);
774 		if (error) {
775 			VFS_UNLOCK_GIANT(tvfslocked);
776 			break;
777 		}
778 		vput(vp);
779 		VFS_UNLOCK_GIANT(vfslocked);
780 		vp = tdp;
781 		vfslocked = tvfslocked;
782 	}
783 	if (error) {
784 		vput(vp);
785 		VFS_UNLOCK_GIANT(vfslocked);
786 		return (error);
787 	}
788 	VOP_UNLOCK(vp, 0);
789 	VFS_UNLOCK_GIANT(vfslocked);
790 	FILEDESC_XLOCK(fdp);
791 	vpold = fdp->fd_cdir;
792 	fdp->fd_cdir = vp;
793 	FILEDESC_XUNLOCK(fdp);
794 	vfslocked = VFS_LOCK_GIANT(vpold->v_mount);
795 	vrele(vpold);
796 	VFS_UNLOCK_GIANT(vfslocked);
797 	return (0);
798 }
799 
800 /*
801  * Change current working directory (``.'').
802  */
803 #ifndef _SYS_SYSPROTO_H_
804 struct chdir_args {
805 	char	*path;
806 };
807 #endif
808 int
809 sys_chdir(td, uap)
810 	struct thread *td;
811 	struct chdir_args /* {
812 		char *path;
813 	} */ *uap;
814 {
815 
816 	return (kern_chdir(td, uap->path, UIO_USERSPACE));
817 }
818 
819 int
820 kern_chdir(struct thread *td, char *path, enum uio_seg pathseg)
821 {
822 	register struct filedesc *fdp = td->td_proc->p_fd;
823 	int error;
824 	struct nameidata nd;
825 	struct vnode *vp;
826 	int vfslocked;
827 
828 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1 |
829 	    MPSAFE, pathseg, path, td);
830 	if ((error = namei(&nd)) != 0)
831 		return (error);
832 	vfslocked = NDHASGIANT(&nd);
833 	if ((error = change_dir(nd.ni_vp, td)) != 0) {
834 		vput(nd.ni_vp);
835 		VFS_UNLOCK_GIANT(vfslocked);
836 		NDFREE(&nd, NDF_ONLY_PNBUF);
837 		return (error);
838 	}
839 	VOP_UNLOCK(nd.ni_vp, 0);
840 	VFS_UNLOCK_GIANT(vfslocked);
841 	NDFREE(&nd, NDF_ONLY_PNBUF);
842 	FILEDESC_XLOCK(fdp);
843 	vp = fdp->fd_cdir;
844 	fdp->fd_cdir = nd.ni_vp;
845 	FILEDESC_XUNLOCK(fdp);
846 	vfslocked = VFS_LOCK_GIANT(vp->v_mount);
847 	vrele(vp);
848 	VFS_UNLOCK_GIANT(vfslocked);
849 	return (0);
850 }
851 
852 /*
853  * Helper function for raised chroot(2) security function:  Refuse if
854  * any filedescriptors are open directories.
855  */
856 static int
857 chroot_refuse_vdir_fds(fdp)
858 	struct filedesc *fdp;
859 {
860 	struct vnode *vp;
861 	struct file *fp;
862 	int fd;
863 
864 	FILEDESC_LOCK_ASSERT(fdp);
865 
866 	for (fd = 0; fd < fdp->fd_nfiles ; fd++) {
867 		fp = fget_locked(fdp, fd);
868 		if (fp == NULL)
869 			continue;
870 		if (fp->f_type == DTYPE_VNODE) {
871 			vp = fp->f_vnode;
872 			if (vp->v_type == VDIR)
873 				return (EPERM);
874 		}
875 	}
876 	return (0);
877 }
878 
879 /*
880  * This sysctl determines if we will allow a process to chroot(2) if it
881  * has a directory open:
882  *	0: disallowed for all processes.
883  *	1: allowed for processes that were not already chroot(2)'ed.
884  *	2: allowed for all processes.
885  */
886 
887 static int chroot_allow_open_directories = 1;
888 
889 SYSCTL_INT(_kern, OID_AUTO, chroot_allow_open_directories, CTLFLAG_RW,
890      &chroot_allow_open_directories, 0,
891      "Allow a process to chroot(2) if it has a directory open");
892 
893 /*
894  * Change notion of root (``/'') directory.
895  */
896 #ifndef _SYS_SYSPROTO_H_
897 struct chroot_args {
898 	char	*path;
899 };
900 #endif
901 int
902 sys_chroot(td, uap)
903 	struct thread *td;
904 	struct chroot_args /* {
905 		char *path;
906 	} */ *uap;
907 {
908 	int error;
909 	struct nameidata nd;
910 	int vfslocked;
911 
912 	error = priv_check(td, PRIV_VFS_CHROOT);
913 	if (error)
914 		return (error);
915 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | MPSAFE |
916 	    AUDITVNODE1, UIO_USERSPACE, uap->path, td);
917 	error = namei(&nd);
918 	if (error)
919 		goto error;
920 	vfslocked = NDHASGIANT(&nd);
921 	if ((error = change_dir(nd.ni_vp, td)) != 0)
922 		goto e_vunlock;
923 #ifdef MAC
924 	if ((error = mac_vnode_check_chroot(td->td_ucred, nd.ni_vp)))
925 		goto e_vunlock;
926 #endif
927 	VOP_UNLOCK(nd.ni_vp, 0);
928 	error = change_root(nd.ni_vp, td);
929 	vrele(nd.ni_vp);
930 	VFS_UNLOCK_GIANT(vfslocked);
931 	NDFREE(&nd, NDF_ONLY_PNBUF);
932 	return (error);
933 e_vunlock:
934 	vput(nd.ni_vp);
935 	VFS_UNLOCK_GIANT(vfslocked);
936 error:
937 	NDFREE(&nd, NDF_ONLY_PNBUF);
938 	return (error);
939 }
940 
941 /*
942  * Common routine for chroot and chdir.  Callers must provide a locked vnode
943  * instance.
944  */
945 int
946 change_dir(vp, td)
947 	struct vnode *vp;
948 	struct thread *td;
949 {
950 	int error;
951 
952 	ASSERT_VOP_LOCKED(vp, "change_dir(): vp not locked");
953 	if (vp->v_type != VDIR)
954 		return (ENOTDIR);
955 #ifdef MAC
956 	error = mac_vnode_check_chdir(td->td_ucred, vp);
957 	if (error)
958 		return (error);
959 #endif
960 	error = VOP_ACCESS(vp, VEXEC, td->td_ucred, td);
961 	return (error);
962 }
963 
964 /*
965  * Common routine for kern_chroot() and jail_attach().  The caller is
966  * responsible for invoking priv_check() and mac_vnode_check_chroot() to
967  * authorize this operation.
968  */
969 int
970 change_root(vp, td)
971 	struct vnode *vp;
972 	struct thread *td;
973 {
974 	struct filedesc *fdp;
975 	struct vnode *oldvp;
976 	int vfslocked;
977 	int error;
978 
979 	VFS_ASSERT_GIANT(vp->v_mount);
980 	fdp = td->td_proc->p_fd;
981 	FILEDESC_XLOCK(fdp);
982 	if (chroot_allow_open_directories == 0 ||
983 	    (chroot_allow_open_directories == 1 && fdp->fd_rdir != rootvnode)) {
984 		error = chroot_refuse_vdir_fds(fdp);
985 		if (error) {
986 			FILEDESC_XUNLOCK(fdp);
987 			return (error);
988 		}
989 	}
990 	oldvp = fdp->fd_rdir;
991 	fdp->fd_rdir = vp;
992 	VREF(fdp->fd_rdir);
993 	if (!fdp->fd_jdir) {
994 		fdp->fd_jdir = vp;
995 		VREF(fdp->fd_jdir);
996 	}
997 	FILEDESC_XUNLOCK(fdp);
998 	vfslocked = VFS_LOCK_GIANT(oldvp->v_mount);
999 	vrele(oldvp);
1000 	VFS_UNLOCK_GIANT(vfslocked);
1001 	return (0);
1002 }
1003 
1004 static __inline cap_rights_t
1005 flags_to_rights(int flags)
1006 {
1007 	cap_rights_t rights = 0;
1008 
1009 	switch ((flags & O_ACCMODE)) {
1010 	case O_RDONLY:
1011 		rights |= CAP_READ;
1012 		break;
1013 
1014 	case O_RDWR:
1015 		rights |= CAP_READ;
1016 		/* fall through */
1017 
1018 	case O_WRONLY:
1019 		rights |= CAP_WRITE;
1020 		break;
1021 
1022 	case O_EXEC:
1023 		rights |= CAP_FEXECVE;
1024 		break;
1025 	}
1026 
1027 	if (flags & O_CREAT)
1028 		rights |= CAP_CREATE;
1029 
1030 	if (flags & O_TRUNC)
1031 		rights |= CAP_FTRUNCATE;
1032 
1033 	if ((flags & O_EXLOCK) || (flags & O_SHLOCK))
1034 		rights |= CAP_FLOCK;
1035 
1036 	return (rights);
1037 }
1038 
1039 /*
1040  * Check permissions, allocate an open file structure, and call the device
1041  * open routine if any.
1042  */
1043 #ifndef _SYS_SYSPROTO_H_
1044 struct open_args {
1045 	char	*path;
1046 	int	flags;
1047 	int	mode;
1048 };
1049 #endif
1050 int
1051 sys_open(td, uap)
1052 	struct thread *td;
1053 	register struct open_args /* {
1054 		char *path;
1055 		int flags;
1056 		int mode;
1057 	} */ *uap;
1058 {
1059 
1060 	return (kern_open(td, uap->path, UIO_USERSPACE, uap->flags, uap->mode));
1061 }
1062 
1063 #ifndef _SYS_SYSPROTO_H_
1064 struct openat_args {
1065 	int	fd;
1066 	char	*path;
1067 	int	flag;
1068 	int	mode;
1069 };
1070 #endif
1071 int
1072 sys_openat(struct thread *td, struct openat_args *uap)
1073 {
1074 
1075 	return (kern_openat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
1076 	    uap->mode));
1077 }
1078 
1079 int
1080 kern_open(struct thread *td, char *path, enum uio_seg pathseg, int flags,
1081     int mode)
1082 {
1083 
1084 	return (kern_openat(td, AT_FDCWD, path, pathseg, flags, mode));
1085 }
1086 
1087 int
1088 kern_openat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1089     int flags, int mode)
1090 {
1091 	struct proc *p = td->td_proc;
1092 	struct filedesc *fdp = p->p_fd;
1093 	struct file *fp;
1094 	struct vnode *vp;
1095 	int cmode;
1096 	int indx = -1, error;
1097 	struct nameidata nd;
1098 	int vfslocked;
1099 	cap_rights_t rights_needed = CAP_LOOKUP;
1100 
1101 	AUDIT_ARG_FFLAGS(flags);
1102 	AUDIT_ARG_MODE(mode);
1103 	/* XXX: audit dirfd */
1104 	rights_needed |= flags_to_rights(flags);
1105 	/*
1106 	 * Only one of the O_EXEC, O_RDONLY, O_WRONLY and O_RDWR flags
1107 	 * may be specified.
1108 	 */
1109 	if (flags & O_EXEC) {
1110 		if (flags & O_ACCMODE)
1111 			return (EINVAL);
1112 	} else if ((flags & O_ACCMODE) == O_ACCMODE) {
1113 		return (EINVAL);
1114 	} else {
1115 		flags = FFLAGS(flags);
1116 	}
1117 
1118 	/*
1119 	 * Allocate the file descriptor, but don't install a descriptor yet.
1120 	 */
1121 	error = falloc_noinstall(td, &fp);
1122 	if (error)
1123 		return (error);
1124 	/*
1125 	 * An extra reference on `fp' has been held for us by
1126 	 * falloc_noinstall().
1127 	 */
1128 	/* Set the flags early so the finit in devfs can pick them up. */
1129 	fp->f_flag = flags & FMASK;
1130 	cmode = ((mode &~ fdp->fd_cmask) & ALLPERMS) &~ S_ISTXT;
1131 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1 | MPSAFE, pathseg,
1132 	    path, fd, rights_needed, td);
1133 	td->td_dupfd = -1;		/* XXX check for fdopen */
1134 	error = vn_open(&nd, &flags, cmode, fp);
1135 	if (error) {
1136 		/*
1137 		 * If the vn_open replaced the method vector, something
1138 		 * wonderous happened deep below and we just pass it up
1139 		 * pretending we know what we do.
1140 		 */
1141 		if (error == ENXIO && fp->f_ops != &badfileops)
1142 			goto success;
1143 
1144 		/*
1145 		 * Handle special fdopen() case. bleh.
1146 		 *
1147 		 * Don't do this for relative (capability) lookups; we don't
1148 		 * understand exactly what would happen, and we don't think
1149 		 * that it ever should.
1150 		 */
1151 		if (nd.ni_strictrelative == 0 &&
1152 		    (error == ENODEV || error == ENXIO) &&
1153 		    td->td_dupfd >= 0) {
1154 			error = dupfdopen(td, fdp, td->td_dupfd, flags, error,
1155 			    &indx);
1156 			if (error == 0)
1157 				goto success;
1158 		}
1159 
1160 		if (error == ERESTART)
1161 			error = EINTR;
1162 		goto bad_unlocked;
1163 	}
1164 	td->td_dupfd = 0;
1165 	vfslocked = NDHASGIANT(&nd);
1166 	NDFREE(&nd, NDF_ONLY_PNBUF);
1167 	vp = nd.ni_vp;
1168 
1169 	/*
1170 	 * Store the vnode, for any f_type. Typically, the vnode use
1171 	 * count is decremented by direct call to vn_closefile() for
1172 	 * files that switched type in the cdevsw fdopen() method.
1173 	 */
1174 	fp->f_vnode = vp;
1175 	/*
1176 	 * If the file wasn't claimed by devfs bind it to the normal
1177 	 * vnode operations here.
1178 	 */
1179 	if (fp->f_ops == &badfileops) {
1180 		KASSERT(vp->v_type != VFIFO, ("Unexpected fifo."));
1181 		fp->f_seqcount = 1;
1182 		finit(fp, (flags & FMASK) | (fp->f_flag & FHASLOCK), DTYPE_VNODE,
1183 		    vp, &vnops);
1184 	}
1185 
1186 	VOP_UNLOCK(vp, 0);
1187 	if (flags & O_TRUNC) {
1188 		error = fo_truncate(fp, 0, td->td_ucred, td);
1189 		if (error)
1190 			goto bad;
1191 	}
1192 	VFS_UNLOCK_GIANT(vfslocked);
1193 success:
1194 	/*
1195 	 * If we haven't already installed the FD (for dupfdopen), do so now.
1196 	 */
1197 	if (indx == -1) {
1198 #ifdef CAPABILITIES
1199 		if (nd.ni_strictrelative == 1) {
1200 			/*
1201 			 * We are doing a strict relative lookup; wrap the
1202 			 * result in a capability.
1203 			 */
1204 			if ((error = kern_capwrap(td, fp, nd.ni_baserights,
1205 			    &indx)) != 0)
1206 				goto bad_unlocked;
1207 		} else
1208 #endif
1209 			if ((error = finstall(td, fp, &indx, flags)) != 0)
1210 				goto bad_unlocked;
1211 
1212 	}
1213 
1214 	/*
1215 	 * Release our private reference, leaving the one associated with
1216 	 * the descriptor table intact.
1217 	 */
1218 	fdrop(fp, td);
1219 	td->td_retval[0] = indx;
1220 	return (0);
1221 bad:
1222 	VFS_UNLOCK_GIANT(vfslocked);
1223 bad_unlocked:
1224 	KASSERT(indx == -1, ("indx=%d, should be -1", indx));
1225 	fdrop(fp, td);
1226 	return (error);
1227 }
1228 
1229 #ifdef COMPAT_43
1230 /*
1231  * Create a file.
1232  */
1233 #ifndef _SYS_SYSPROTO_H_
1234 struct ocreat_args {
1235 	char	*path;
1236 	int	mode;
1237 };
1238 #endif
1239 int
1240 ocreat(td, uap)
1241 	struct thread *td;
1242 	register struct ocreat_args /* {
1243 		char *path;
1244 		int mode;
1245 	} */ *uap;
1246 {
1247 
1248 	return (kern_open(td, uap->path, UIO_USERSPACE,
1249 	    O_WRONLY | O_CREAT | O_TRUNC, uap->mode));
1250 }
1251 #endif /* COMPAT_43 */
1252 
1253 /*
1254  * Create a special file.
1255  */
1256 #ifndef _SYS_SYSPROTO_H_
1257 struct mknod_args {
1258 	char	*path;
1259 	int	mode;
1260 	int	dev;
1261 };
1262 #endif
1263 int
1264 sys_mknod(td, uap)
1265 	struct thread *td;
1266 	register struct mknod_args /* {
1267 		char *path;
1268 		int mode;
1269 		int dev;
1270 	} */ *uap;
1271 {
1272 
1273 	return (kern_mknod(td, uap->path, UIO_USERSPACE, uap->mode, uap->dev));
1274 }
1275 
1276 #ifndef _SYS_SYSPROTO_H_
1277 struct mknodat_args {
1278 	int	fd;
1279 	char	*path;
1280 	mode_t	mode;
1281 	dev_t	dev;
1282 };
1283 #endif
1284 int
1285 sys_mknodat(struct thread *td, struct mknodat_args *uap)
1286 {
1287 
1288 	return (kern_mknodat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode,
1289 	    uap->dev));
1290 }
1291 
1292 int
1293 kern_mknod(struct thread *td, char *path, enum uio_seg pathseg, int mode,
1294     int dev)
1295 {
1296 
1297 	return (kern_mknodat(td, AT_FDCWD, path, pathseg, mode, dev));
1298 }
1299 
1300 int
1301 kern_mknodat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1302     int mode, int dev)
1303 {
1304 	struct vnode *vp;
1305 	struct mount *mp;
1306 	struct vattr vattr;
1307 	int error;
1308 	int whiteout = 0;
1309 	struct nameidata nd;
1310 	int vfslocked;
1311 
1312 	AUDIT_ARG_MODE(mode);
1313 	AUDIT_ARG_DEV(dev);
1314 	switch (mode & S_IFMT) {
1315 	case S_IFCHR:
1316 	case S_IFBLK:
1317 		error = priv_check(td, PRIV_VFS_MKNOD_DEV);
1318 		break;
1319 	case S_IFMT:
1320 		error = priv_check(td, PRIV_VFS_MKNOD_BAD);
1321 		break;
1322 	case S_IFWHT:
1323 		error = priv_check(td, PRIV_VFS_MKNOD_WHT);
1324 		break;
1325 	case S_IFIFO:
1326 		if (dev == 0)
1327 			return (kern_mkfifoat(td, fd, path, pathseg, mode));
1328 		/* FALLTHROUGH */
1329 	default:
1330 		error = EINVAL;
1331 		break;
1332 	}
1333 	if (error)
1334 		return (error);
1335 restart:
1336 	bwillwrite();
1337 	NDINIT_ATRIGHTS(&nd, CREATE,
1338 	    LOCKPARENT | SAVENAME | MPSAFE | AUDITVNODE1, pathseg, path, fd,
1339 	    CAP_MKFIFO, td);
1340 	if ((error = namei(&nd)) != 0)
1341 		return (error);
1342 	vfslocked = NDHASGIANT(&nd);
1343 	vp = nd.ni_vp;
1344 	if (vp != NULL) {
1345 		NDFREE(&nd, NDF_ONLY_PNBUF);
1346 		if (vp == nd.ni_dvp)
1347 			vrele(nd.ni_dvp);
1348 		else
1349 			vput(nd.ni_dvp);
1350 		vrele(vp);
1351 		VFS_UNLOCK_GIANT(vfslocked);
1352 		return (EEXIST);
1353 	} else {
1354 		VATTR_NULL(&vattr);
1355 		vattr.va_mode = (mode & ALLPERMS) &
1356 		    ~td->td_proc->p_fd->fd_cmask;
1357 		vattr.va_rdev = dev;
1358 		whiteout = 0;
1359 
1360 		switch (mode & S_IFMT) {
1361 		case S_IFMT:	/* used by badsect to flag bad sectors */
1362 			vattr.va_type = VBAD;
1363 			break;
1364 		case S_IFCHR:
1365 			vattr.va_type = VCHR;
1366 			break;
1367 		case S_IFBLK:
1368 			vattr.va_type = VBLK;
1369 			break;
1370 		case S_IFWHT:
1371 			whiteout = 1;
1372 			break;
1373 		default:
1374 			panic("kern_mknod: invalid mode");
1375 		}
1376 	}
1377 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1378 		NDFREE(&nd, NDF_ONLY_PNBUF);
1379 		vput(nd.ni_dvp);
1380 		VFS_UNLOCK_GIANT(vfslocked);
1381 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1382 			return (error);
1383 		goto restart;
1384 	}
1385 #ifdef MAC
1386 	if (error == 0 && !whiteout)
1387 		error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp,
1388 		    &nd.ni_cnd, &vattr);
1389 #endif
1390 	if (!error) {
1391 		if (whiteout)
1392 			error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE);
1393 		else {
1394 			error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp,
1395 						&nd.ni_cnd, &vattr);
1396 			if (error == 0)
1397 				vput(nd.ni_vp);
1398 		}
1399 	}
1400 	NDFREE(&nd, NDF_ONLY_PNBUF);
1401 	vput(nd.ni_dvp);
1402 	vn_finished_write(mp);
1403 	VFS_UNLOCK_GIANT(vfslocked);
1404 	return (error);
1405 }
1406 
1407 /*
1408  * Create a named pipe.
1409  */
1410 #ifndef _SYS_SYSPROTO_H_
1411 struct mkfifo_args {
1412 	char	*path;
1413 	int	mode;
1414 };
1415 #endif
1416 int
1417 sys_mkfifo(td, uap)
1418 	struct thread *td;
1419 	register struct mkfifo_args /* {
1420 		char *path;
1421 		int mode;
1422 	} */ *uap;
1423 {
1424 
1425 	return (kern_mkfifo(td, uap->path, UIO_USERSPACE, uap->mode));
1426 }
1427 
1428 #ifndef _SYS_SYSPROTO_H_
1429 struct mkfifoat_args {
1430 	int	fd;
1431 	char	*path;
1432 	mode_t	mode;
1433 };
1434 #endif
1435 int
1436 sys_mkfifoat(struct thread *td, struct mkfifoat_args *uap)
1437 {
1438 
1439 	return (kern_mkfifoat(td, uap->fd, uap->path, UIO_USERSPACE,
1440 	    uap->mode));
1441 }
1442 
1443 int
1444 kern_mkfifo(struct thread *td, char *path, enum uio_seg pathseg, int mode)
1445 {
1446 
1447 	return (kern_mkfifoat(td, AT_FDCWD, path, pathseg, mode));
1448 }
1449 
1450 int
1451 kern_mkfifoat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1452     int mode)
1453 {
1454 	struct mount *mp;
1455 	struct vattr vattr;
1456 	int error;
1457 	struct nameidata nd;
1458 	int vfslocked;
1459 
1460 	AUDIT_ARG_MODE(mode);
1461 restart:
1462 	bwillwrite();
1463 	NDINIT_AT(&nd, CREATE, LOCKPARENT | SAVENAME | MPSAFE | AUDITVNODE1,
1464 	    pathseg, path, fd, td);
1465 	if ((error = namei(&nd)) != 0)
1466 		return (error);
1467 	vfslocked = NDHASGIANT(&nd);
1468 	if (nd.ni_vp != NULL) {
1469 		NDFREE(&nd, NDF_ONLY_PNBUF);
1470 		if (nd.ni_vp == nd.ni_dvp)
1471 			vrele(nd.ni_dvp);
1472 		else
1473 			vput(nd.ni_dvp);
1474 		vrele(nd.ni_vp);
1475 		VFS_UNLOCK_GIANT(vfslocked);
1476 		return (EEXIST);
1477 	}
1478 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1479 		NDFREE(&nd, NDF_ONLY_PNBUF);
1480 		vput(nd.ni_dvp);
1481 		VFS_UNLOCK_GIANT(vfslocked);
1482 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1483 			return (error);
1484 		goto restart;
1485 	}
1486 	VATTR_NULL(&vattr);
1487 	vattr.va_type = VFIFO;
1488 	vattr.va_mode = (mode & ALLPERMS) & ~td->td_proc->p_fd->fd_cmask;
1489 #ifdef MAC
1490 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1491 	    &vattr);
1492 	if (error)
1493 		goto out;
1494 #endif
1495 	error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
1496 	if (error == 0)
1497 		vput(nd.ni_vp);
1498 #ifdef MAC
1499 out:
1500 #endif
1501 	vput(nd.ni_dvp);
1502 	vn_finished_write(mp);
1503 	VFS_UNLOCK_GIANT(vfslocked);
1504 	NDFREE(&nd, NDF_ONLY_PNBUF);
1505 	return (error);
1506 }
1507 
1508 /*
1509  * Make a hard file link.
1510  */
1511 #ifndef _SYS_SYSPROTO_H_
1512 struct link_args {
1513 	char	*path;
1514 	char	*link;
1515 };
1516 #endif
1517 int
1518 sys_link(td, uap)
1519 	struct thread *td;
1520 	register struct link_args /* {
1521 		char *path;
1522 		char *link;
1523 	} */ *uap;
1524 {
1525 
1526 	return (kern_link(td, uap->path, uap->link, UIO_USERSPACE));
1527 }
1528 
1529 #ifndef _SYS_SYSPROTO_H_
1530 struct linkat_args {
1531 	int	fd1;
1532 	char	*path1;
1533 	int	fd2;
1534 	char	*path2;
1535 	int	flag;
1536 };
1537 #endif
1538 int
1539 sys_linkat(struct thread *td, struct linkat_args *uap)
1540 {
1541 	int flag;
1542 
1543 	flag = uap->flag;
1544 	if (flag & ~AT_SYMLINK_FOLLOW)
1545 		return (EINVAL);
1546 
1547 	return (kern_linkat(td, uap->fd1, uap->fd2, uap->path1, uap->path2,
1548 	    UIO_USERSPACE, (flag & AT_SYMLINK_FOLLOW) ? FOLLOW : NOFOLLOW));
1549 }
1550 
1551 int hardlink_check_uid = 0;
1552 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_uid, CTLFLAG_RW,
1553     &hardlink_check_uid, 0,
1554     "Unprivileged processes cannot create hard links to files owned by other "
1555     "users");
1556 static int hardlink_check_gid = 0;
1557 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_gid, CTLFLAG_RW,
1558     &hardlink_check_gid, 0,
1559     "Unprivileged processes cannot create hard links to files owned by other "
1560     "groups");
1561 
1562 static int
1563 can_hardlink(struct vnode *vp, struct ucred *cred)
1564 {
1565 	struct vattr va;
1566 	int error;
1567 
1568 	if (!hardlink_check_uid && !hardlink_check_gid)
1569 		return (0);
1570 
1571 	error = VOP_GETATTR(vp, &va, cred);
1572 	if (error != 0)
1573 		return (error);
1574 
1575 	if (hardlink_check_uid && cred->cr_uid != va.va_uid) {
1576 		error = priv_check_cred(cred, PRIV_VFS_LINK, 0);
1577 		if (error)
1578 			return (error);
1579 	}
1580 
1581 	if (hardlink_check_gid && !groupmember(va.va_gid, cred)) {
1582 		error = priv_check_cred(cred, PRIV_VFS_LINK, 0);
1583 		if (error)
1584 			return (error);
1585 	}
1586 
1587 	return (0);
1588 }
1589 
1590 int
1591 kern_link(struct thread *td, char *path, char *link, enum uio_seg segflg)
1592 {
1593 
1594 	return (kern_linkat(td, AT_FDCWD, AT_FDCWD, path,link, segflg, FOLLOW));
1595 }
1596 
1597 int
1598 kern_linkat(struct thread *td, int fd1, int fd2, char *path1, char *path2,
1599     enum uio_seg segflg, int follow)
1600 {
1601 	struct vnode *vp;
1602 	struct mount *mp;
1603 	struct nameidata nd;
1604 	int vfslocked;
1605 	int lvfslocked;
1606 	int error;
1607 
1608 	bwillwrite();
1609 	NDINIT_AT(&nd, LOOKUP, follow | MPSAFE | AUDITVNODE1, segflg, path1,
1610 	    fd1, td);
1611 
1612 	if ((error = namei(&nd)) != 0)
1613 		return (error);
1614 	vfslocked = NDHASGIANT(&nd);
1615 	NDFREE(&nd, NDF_ONLY_PNBUF);
1616 	vp = nd.ni_vp;
1617 	if (vp->v_type == VDIR) {
1618 		vrele(vp);
1619 		VFS_UNLOCK_GIANT(vfslocked);
1620 		return (EPERM);		/* POSIX */
1621 	}
1622 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) {
1623 		vrele(vp);
1624 		VFS_UNLOCK_GIANT(vfslocked);
1625 		return (error);
1626 	}
1627 	NDINIT_AT(&nd, CREATE, LOCKPARENT | SAVENAME | MPSAFE | AUDITVNODE2,
1628 	    segflg, path2, fd2, td);
1629 	if ((error = namei(&nd)) == 0) {
1630 		lvfslocked = NDHASGIANT(&nd);
1631 		if (nd.ni_vp != NULL) {
1632 			if (nd.ni_dvp == nd.ni_vp)
1633 				vrele(nd.ni_dvp);
1634 			else
1635 				vput(nd.ni_dvp);
1636 			vrele(nd.ni_vp);
1637 			error = EEXIST;
1638 		} else if ((error = vn_lock(vp, LK_EXCLUSIVE | LK_RETRY))
1639 		    == 0) {
1640 			error = can_hardlink(vp, td->td_ucred);
1641 			if (error == 0)
1642 #ifdef MAC
1643 				error = mac_vnode_check_link(td->td_ucred,
1644 				    nd.ni_dvp, vp, &nd.ni_cnd);
1645 			if (error == 0)
1646 #endif
1647 				error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
1648 			VOP_UNLOCK(vp, 0);
1649 			vput(nd.ni_dvp);
1650 		}
1651 		NDFREE(&nd, NDF_ONLY_PNBUF);
1652 		VFS_UNLOCK_GIANT(lvfslocked);
1653 	}
1654 	vrele(vp);
1655 	vn_finished_write(mp);
1656 	VFS_UNLOCK_GIANT(vfslocked);
1657 	return (error);
1658 }
1659 
1660 /*
1661  * Make a symbolic link.
1662  */
1663 #ifndef _SYS_SYSPROTO_H_
1664 struct symlink_args {
1665 	char	*path;
1666 	char	*link;
1667 };
1668 #endif
1669 int
1670 sys_symlink(td, uap)
1671 	struct thread *td;
1672 	register struct symlink_args /* {
1673 		char *path;
1674 		char *link;
1675 	} */ *uap;
1676 {
1677 
1678 	return (kern_symlink(td, uap->path, uap->link, UIO_USERSPACE));
1679 }
1680 
1681 #ifndef _SYS_SYSPROTO_H_
1682 struct symlinkat_args {
1683 	char	*path;
1684 	int	fd;
1685 	char	*path2;
1686 };
1687 #endif
1688 int
1689 sys_symlinkat(struct thread *td, struct symlinkat_args *uap)
1690 {
1691 
1692 	return (kern_symlinkat(td, uap->path1, uap->fd, uap->path2,
1693 	    UIO_USERSPACE));
1694 }
1695 
1696 int
1697 kern_symlink(struct thread *td, char *path, char *link, enum uio_seg segflg)
1698 {
1699 
1700 	return (kern_symlinkat(td, path, AT_FDCWD, link, segflg));
1701 }
1702 
1703 int
1704 kern_symlinkat(struct thread *td, char *path1, int fd, char *path2,
1705     enum uio_seg segflg)
1706 {
1707 	struct mount *mp;
1708 	struct vattr vattr;
1709 	char *syspath;
1710 	int error;
1711 	struct nameidata nd;
1712 	int vfslocked;
1713 
1714 	if (segflg == UIO_SYSSPACE) {
1715 		syspath = path1;
1716 	} else {
1717 		syspath = uma_zalloc(namei_zone, M_WAITOK);
1718 		if ((error = copyinstr(path1, syspath, MAXPATHLEN, NULL)) != 0)
1719 			goto out;
1720 	}
1721 	AUDIT_ARG_TEXT(syspath);
1722 restart:
1723 	bwillwrite();
1724 	NDINIT_AT(&nd, CREATE, LOCKPARENT | SAVENAME | MPSAFE | AUDITVNODE1,
1725 	    segflg, path2, fd, td);
1726 	if ((error = namei(&nd)) != 0)
1727 		goto out;
1728 	vfslocked = NDHASGIANT(&nd);
1729 	if (nd.ni_vp) {
1730 		NDFREE(&nd, NDF_ONLY_PNBUF);
1731 		if (nd.ni_vp == nd.ni_dvp)
1732 			vrele(nd.ni_dvp);
1733 		else
1734 			vput(nd.ni_dvp);
1735 		vrele(nd.ni_vp);
1736 		VFS_UNLOCK_GIANT(vfslocked);
1737 		error = EEXIST;
1738 		goto out;
1739 	}
1740 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1741 		NDFREE(&nd, NDF_ONLY_PNBUF);
1742 		vput(nd.ni_dvp);
1743 		VFS_UNLOCK_GIANT(vfslocked);
1744 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1745 			goto out;
1746 		goto restart;
1747 	}
1748 	VATTR_NULL(&vattr);
1749 	vattr.va_mode = ACCESSPERMS &~ td->td_proc->p_fd->fd_cmask;
1750 #ifdef MAC
1751 	vattr.va_type = VLNK;
1752 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1753 	    &vattr);
1754 	if (error)
1755 		goto out2;
1756 #endif
1757 	error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, syspath);
1758 	if (error == 0)
1759 		vput(nd.ni_vp);
1760 #ifdef MAC
1761 out2:
1762 #endif
1763 	NDFREE(&nd, NDF_ONLY_PNBUF);
1764 	vput(nd.ni_dvp);
1765 	vn_finished_write(mp);
1766 	VFS_UNLOCK_GIANT(vfslocked);
1767 out:
1768 	if (segflg != UIO_SYSSPACE)
1769 		uma_zfree(namei_zone, syspath);
1770 	return (error);
1771 }
1772 
1773 /*
1774  * Delete a whiteout from the filesystem.
1775  */
1776 int
1777 sys_undelete(td, uap)
1778 	struct thread *td;
1779 	register struct undelete_args /* {
1780 		char *path;
1781 	} */ *uap;
1782 {
1783 	int error;
1784 	struct mount *mp;
1785 	struct nameidata nd;
1786 	int vfslocked;
1787 
1788 restart:
1789 	bwillwrite();
1790 	NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | MPSAFE | AUDITVNODE1,
1791 	    UIO_USERSPACE, uap->path, td);
1792 	error = namei(&nd);
1793 	if (error)
1794 		return (error);
1795 	vfslocked = NDHASGIANT(&nd);
1796 
1797 	if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) {
1798 		NDFREE(&nd, NDF_ONLY_PNBUF);
1799 		if (nd.ni_vp == nd.ni_dvp)
1800 			vrele(nd.ni_dvp);
1801 		else
1802 			vput(nd.ni_dvp);
1803 		if (nd.ni_vp)
1804 			vrele(nd.ni_vp);
1805 		VFS_UNLOCK_GIANT(vfslocked);
1806 		return (EEXIST);
1807 	}
1808 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1809 		NDFREE(&nd, NDF_ONLY_PNBUF);
1810 		vput(nd.ni_dvp);
1811 		VFS_UNLOCK_GIANT(vfslocked);
1812 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1813 			return (error);
1814 		goto restart;
1815 	}
1816 	error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE);
1817 	NDFREE(&nd, NDF_ONLY_PNBUF);
1818 	vput(nd.ni_dvp);
1819 	vn_finished_write(mp);
1820 	VFS_UNLOCK_GIANT(vfslocked);
1821 	return (error);
1822 }
1823 
1824 /*
1825  * Delete a name from the filesystem.
1826  */
1827 #ifndef _SYS_SYSPROTO_H_
1828 struct unlink_args {
1829 	char	*path;
1830 };
1831 #endif
1832 int
1833 sys_unlink(td, uap)
1834 	struct thread *td;
1835 	struct unlink_args /* {
1836 		char *path;
1837 	} */ *uap;
1838 {
1839 
1840 	return (kern_unlink(td, uap->path, UIO_USERSPACE));
1841 }
1842 
1843 #ifndef _SYS_SYSPROTO_H_
1844 struct unlinkat_args {
1845 	int	fd;
1846 	char	*path;
1847 	int	flag;
1848 };
1849 #endif
1850 int
1851 sys_unlinkat(struct thread *td, struct unlinkat_args *uap)
1852 {
1853 	int flag = uap->flag;
1854 	int fd = uap->fd;
1855 	char *path = uap->path;
1856 
1857 	if (flag & ~AT_REMOVEDIR)
1858 		return (EINVAL);
1859 
1860 	if (flag & AT_REMOVEDIR)
1861 		return (kern_rmdirat(td, fd, path, UIO_USERSPACE));
1862 	else
1863 		return (kern_unlinkat(td, fd, path, UIO_USERSPACE, 0));
1864 }
1865 
1866 int
1867 kern_unlink(struct thread *td, char *path, enum uio_seg pathseg)
1868 {
1869 
1870 	return (kern_unlinkat(td, AT_FDCWD, path, pathseg, 0));
1871 }
1872 
1873 int
1874 kern_unlinkat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1875     ino_t oldinum)
1876 {
1877 	struct mount *mp;
1878 	struct vnode *vp;
1879 	int error;
1880 	struct nameidata nd;
1881 	struct stat sb;
1882 	int vfslocked;
1883 
1884 restart:
1885 	bwillwrite();
1886 	NDINIT_AT(&nd, DELETE, LOCKPARENT | LOCKLEAF | MPSAFE | AUDITVNODE1,
1887 	    pathseg, path, fd, td);
1888 	if ((error = namei(&nd)) != 0)
1889 		return (error == EINVAL ? EPERM : error);
1890 	vfslocked = NDHASGIANT(&nd);
1891 	vp = nd.ni_vp;
1892 	if (vp->v_type == VDIR && oldinum == 0) {
1893 		error = EPERM;		/* POSIX */
1894 	} else if (oldinum != 0 &&
1895 		  ((error = vn_stat(vp, &sb, td->td_ucred, NOCRED, td)) == 0) &&
1896 		  sb.st_ino != oldinum) {
1897 			error = EIDRM;	/* Identifier removed */
1898 	} else {
1899 		/*
1900 		 * The root of a mounted filesystem cannot be deleted.
1901 		 *
1902 		 * XXX: can this only be a VDIR case?
1903 		 */
1904 		if (vp->v_vflag & VV_ROOT)
1905 			error = EBUSY;
1906 	}
1907 	if (error == 0) {
1908 		if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1909 			NDFREE(&nd, NDF_ONLY_PNBUF);
1910 			vput(nd.ni_dvp);
1911 			if (vp == nd.ni_dvp)
1912 				vrele(vp);
1913 			else
1914 				vput(vp);
1915 			VFS_UNLOCK_GIANT(vfslocked);
1916 			if ((error = vn_start_write(NULL, &mp,
1917 			    V_XSLEEP | PCATCH)) != 0)
1918 				return (error);
1919 			goto restart;
1920 		}
1921 #ifdef MAC
1922 		error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
1923 		    &nd.ni_cnd);
1924 		if (error)
1925 			goto out;
1926 #endif
1927 		error = VOP_REMOVE(nd.ni_dvp, vp, &nd.ni_cnd);
1928 #ifdef MAC
1929 out:
1930 #endif
1931 		vn_finished_write(mp);
1932 	}
1933 	NDFREE(&nd, NDF_ONLY_PNBUF);
1934 	vput(nd.ni_dvp);
1935 	if (vp == nd.ni_dvp)
1936 		vrele(vp);
1937 	else
1938 		vput(vp);
1939 	VFS_UNLOCK_GIANT(vfslocked);
1940 	return (error);
1941 }
1942 
1943 /*
1944  * Reposition read/write file offset.
1945  */
1946 #ifndef _SYS_SYSPROTO_H_
1947 struct lseek_args {
1948 	int	fd;
1949 	int	pad;
1950 	off_t	offset;
1951 	int	whence;
1952 };
1953 #endif
1954 int
1955 sys_lseek(td, uap)
1956 	struct thread *td;
1957 	register struct lseek_args /* {
1958 		int fd;
1959 		int pad;
1960 		off_t offset;
1961 		int whence;
1962 	} */ *uap;
1963 {
1964 	struct ucred *cred = td->td_ucred;
1965 	struct file *fp;
1966 	struct vnode *vp;
1967 	struct vattr vattr;
1968 	off_t foffset, offset, size;
1969 	int error, noneg;
1970 	int vfslocked;
1971 
1972 	AUDIT_ARG_FD(uap->fd);
1973 	if ((error = fget(td, uap->fd, CAP_SEEK, &fp)) != 0)
1974 		return (error);
1975 	if (!(fp->f_ops->fo_flags & DFLAG_SEEKABLE)) {
1976 		fdrop(fp, td);
1977 		return (ESPIPE);
1978 	}
1979 	vp = fp->f_vnode;
1980 	foffset = foffset_lock(fp, 0);
1981 	vfslocked = VFS_LOCK_GIANT(vp->v_mount);
1982 	noneg = (vp->v_type != VCHR);
1983 	offset = uap->offset;
1984 	switch (uap->whence) {
1985 	case L_INCR:
1986 		if (noneg &&
1987 		    (foffset < 0 ||
1988 		    (offset > 0 && foffset > OFF_MAX - offset))) {
1989 			error = EOVERFLOW;
1990 			break;
1991 		}
1992 		offset += foffset;
1993 		break;
1994 	case L_XTND:
1995 		vn_lock(vp, LK_SHARED | LK_RETRY);
1996 		error = VOP_GETATTR(vp, &vattr, cred);
1997 		VOP_UNLOCK(vp, 0);
1998 		if (error)
1999 			break;
2000 
2001 		/*
2002 		 * If the file references a disk device, then fetch
2003 		 * the media size and use that to determine the ending
2004 		 * offset.
2005 		 */
2006 		if (vattr.va_size == 0 && vp->v_type == VCHR &&
2007 		    fo_ioctl(fp, DIOCGMEDIASIZE, &size, cred, td) == 0)
2008 			vattr.va_size = size;
2009 		if (noneg &&
2010 		    (vattr.va_size > OFF_MAX ||
2011 		    (offset > 0 && vattr.va_size > OFF_MAX - offset))) {
2012 			error = EOVERFLOW;
2013 			break;
2014 		}
2015 		offset += vattr.va_size;
2016 		break;
2017 	case L_SET:
2018 		break;
2019 	case SEEK_DATA:
2020 		error = fo_ioctl(fp, FIOSEEKDATA, &offset, cred, td);
2021 		break;
2022 	case SEEK_HOLE:
2023 		error = fo_ioctl(fp, FIOSEEKHOLE, &offset, cred, td);
2024 		break;
2025 	default:
2026 		error = EINVAL;
2027 	}
2028 	if (error == 0 && noneg && offset < 0)
2029 		error = EINVAL;
2030 	if (error != 0)
2031 		goto drop;
2032 	VFS_KNOTE_UNLOCKED(vp, 0);
2033 	*(off_t *)(td->td_retval) = offset;
2034 drop:
2035 	fdrop(fp, td);
2036 	VFS_UNLOCK_GIANT(vfslocked);
2037 	foffset_unlock(fp, offset, error != 0 ? FOF_NOUPDATE : 0);
2038 	return (error);
2039 }
2040 
2041 #if defined(COMPAT_43)
2042 /*
2043  * Reposition read/write file offset.
2044  */
2045 #ifndef _SYS_SYSPROTO_H_
2046 struct olseek_args {
2047 	int	fd;
2048 	long	offset;
2049 	int	whence;
2050 };
2051 #endif
2052 int
2053 olseek(td, uap)
2054 	struct thread *td;
2055 	register struct olseek_args /* {
2056 		int fd;
2057 		long offset;
2058 		int whence;
2059 	} */ *uap;
2060 {
2061 	struct lseek_args /* {
2062 		int fd;
2063 		int pad;
2064 		off_t offset;
2065 		int whence;
2066 	} */ nuap;
2067 
2068 	nuap.fd = uap->fd;
2069 	nuap.offset = uap->offset;
2070 	nuap.whence = uap->whence;
2071 	return (sys_lseek(td, &nuap));
2072 }
2073 #endif /* COMPAT_43 */
2074 
2075 /* Version with the 'pad' argument */
2076 int
2077 freebsd6_lseek(td, uap)
2078 	struct thread *td;
2079 	register struct freebsd6_lseek_args *uap;
2080 {
2081 	struct lseek_args ouap;
2082 
2083 	ouap.fd = uap->fd;
2084 	ouap.offset = uap->offset;
2085 	ouap.whence = uap->whence;
2086 	return (sys_lseek(td, &ouap));
2087 }
2088 
2089 /*
2090  * Check access permissions using passed credentials.
2091  */
2092 static int
2093 vn_access(vp, user_flags, cred, td)
2094 	struct vnode	*vp;
2095 	int		user_flags;
2096 	struct ucred	*cred;
2097 	struct thread	*td;
2098 {
2099 	int error;
2100 	accmode_t accmode;
2101 
2102 	/* Flags == 0 means only check for existence. */
2103 	error = 0;
2104 	if (user_flags) {
2105 		accmode = 0;
2106 		if (user_flags & R_OK)
2107 			accmode |= VREAD;
2108 		if (user_flags & W_OK)
2109 			accmode |= VWRITE;
2110 		if (user_flags & X_OK)
2111 			accmode |= VEXEC;
2112 #ifdef MAC
2113 		error = mac_vnode_check_access(cred, vp, accmode);
2114 		if (error)
2115 			return (error);
2116 #endif
2117 		if ((accmode & VWRITE) == 0 || (error = vn_writechk(vp)) == 0)
2118 			error = VOP_ACCESS(vp, accmode, cred, td);
2119 	}
2120 	return (error);
2121 }
2122 
2123 /*
2124  * Check access permissions using "real" credentials.
2125  */
2126 #ifndef _SYS_SYSPROTO_H_
2127 struct access_args {
2128 	char	*path;
2129 	int	amode;
2130 };
2131 #endif
2132 int
2133 sys_access(td, uap)
2134 	struct thread *td;
2135 	register struct access_args /* {
2136 		char *path;
2137 		int amode;
2138 	} */ *uap;
2139 {
2140 
2141 	return (kern_access(td, uap->path, UIO_USERSPACE, uap->amode));
2142 }
2143 
2144 #ifndef _SYS_SYSPROTO_H_
2145 struct faccessat_args {
2146 	int	dirfd;
2147 	char	*path;
2148 	int	amode;
2149 	int	flag;
2150 }
2151 #endif
2152 int
2153 sys_faccessat(struct thread *td, struct faccessat_args *uap)
2154 {
2155 
2156 	if (uap->flag & ~AT_EACCESS)
2157 		return (EINVAL);
2158 	return (kern_accessat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
2159 	    uap->amode));
2160 }
2161 
2162 int
2163 kern_access(struct thread *td, char *path, enum uio_seg pathseg, int amode)
2164 {
2165 
2166 	return (kern_accessat(td, AT_FDCWD, path, pathseg, 0, amode));
2167 }
2168 
2169 int
2170 kern_accessat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2171     int flag, int amode)
2172 {
2173 	struct ucred *cred, *tmpcred;
2174 	struct vnode *vp;
2175 	struct nameidata nd;
2176 	int vfslocked;
2177 	int error;
2178 
2179 	/*
2180 	 * Create and modify a temporary credential instead of one that
2181 	 * is potentially shared.
2182 	 */
2183 	if (!(flag & AT_EACCESS)) {
2184 		cred = td->td_ucred;
2185 		tmpcred = crdup(cred);
2186 		tmpcred->cr_uid = cred->cr_ruid;
2187 		tmpcred->cr_groups[0] = cred->cr_rgid;
2188 		td->td_ucred = tmpcred;
2189 	} else
2190 		cred = tmpcred = td->td_ucred;
2191 	AUDIT_ARG_VALUE(amode);
2192 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | MPSAFE |
2193 	    AUDITVNODE1, pathseg, path, fd, CAP_FSTAT, td);
2194 	if ((error = namei(&nd)) != 0)
2195 		goto out1;
2196 	vfslocked = NDHASGIANT(&nd);
2197 	vp = nd.ni_vp;
2198 
2199 	error = vn_access(vp, amode, tmpcred, td);
2200 	NDFREE(&nd, NDF_ONLY_PNBUF);
2201 	vput(vp);
2202 	VFS_UNLOCK_GIANT(vfslocked);
2203 out1:
2204 	if (!(flag & AT_EACCESS)) {
2205 		td->td_ucred = cred;
2206 		crfree(tmpcred);
2207 	}
2208 	return (error);
2209 }
2210 
2211 /*
2212  * Check access permissions using "effective" credentials.
2213  */
2214 #ifndef _SYS_SYSPROTO_H_
2215 struct eaccess_args {
2216 	char	*path;
2217 	int	amode;
2218 };
2219 #endif
2220 int
2221 sys_eaccess(td, uap)
2222 	struct thread *td;
2223 	register struct eaccess_args /* {
2224 		char *path;
2225 		int amode;
2226 	} */ *uap;
2227 {
2228 
2229 	return (kern_eaccess(td, uap->path, UIO_USERSPACE, uap->amode));
2230 }
2231 
2232 int
2233 kern_eaccess(struct thread *td, char *path, enum uio_seg pathseg, int amode)
2234 {
2235 
2236 	return (kern_accessat(td, AT_FDCWD, path, pathseg, AT_EACCESS, amode));
2237 }
2238 
2239 #if defined(COMPAT_43)
2240 /*
2241  * Get file status; this version follows links.
2242  */
2243 #ifndef _SYS_SYSPROTO_H_
2244 struct ostat_args {
2245 	char	*path;
2246 	struct ostat *ub;
2247 };
2248 #endif
2249 int
2250 ostat(td, uap)
2251 	struct thread *td;
2252 	register struct ostat_args /* {
2253 		char *path;
2254 		struct ostat *ub;
2255 	} */ *uap;
2256 {
2257 	struct stat sb;
2258 	struct ostat osb;
2259 	int error;
2260 
2261 	error = kern_stat(td, uap->path, UIO_USERSPACE, &sb);
2262 	if (error)
2263 		return (error);
2264 	cvtstat(&sb, &osb);
2265 	error = copyout(&osb, uap->ub, sizeof (osb));
2266 	return (error);
2267 }
2268 
2269 /*
2270  * Get file status; this version does not follow links.
2271  */
2272 #ifndef _SYS_SYSPROTO_H_
2273 struct olstat_args {
2274 	char	*path;
2275 	struct ostat *ub;
2276 };
2277 #endif
2278 int
2279 olstat(td, uap)
2280 	struct thread *td;
2281 	register struct olstat_args /* {
2282 		char *path;
2283 		struct ostat *ub;
2284 	} */ *uap;
2285 {
2286 	struct stat sb;
2287 	struct ostat osb;
2288 	int error;
2289 
2290 	error = kern_lstat(td, uap->path, UIO_USERSPACE, &sb);
2291 	if (error)
2292 		return (error);
2293 	cvtstat(&sb, &osb);
2294 	error = copyout(&osb, uap->ub, sizeof (osb));
2295 	return (error);
2296 }
2297 
2298 /*
2299  * Convert from an old to a new stat structure.
2300  */
2301 void
2302 cvtstat(st, ost)
2303 	struct stat *st;
2304 	struct ostat *ost;
2305 {
2306 
2307 	ost->st_dev = st->st_dev;
2308 	ost->st_ino = st->st_ino;
2309 	ost->st_mode = st->st_mode;
2310 	ost->st_nlink = st->st_nlink;
2311 	ost->st_uid = st->st_uid;
2312 	ost->st_gid = st->st_gid;
2313 	ost->st_rdev = st->st_rdev;
2314 	if (st->st_size < (quad_t)1 << 32)
2315 		ost->st_size = st->st_size;
2316 	else
2317 		ost->st_size = -2;
2318 	ost->st_atim = st->st_atim;
2319 	ost->st_mtim = st->st_mtim;
2320 	ost->st_ctim = st->st_ctim;
2321 	ost->st_blksize = st->st_blksize;
2322 	ost->st_blocks = st->st_blocks;
2323 	ost->st_flags = st->st_flags;
2324 	ost->st_gen = st->st_gen;
2325 }
2326 #endif /* COMPAT_43 */
2327 
2328 /*
2329  * Get file status; this version follows links.
2330  */
2331 #ifndef _SYS_SYSPROTO_H_
2332 struct stat_args {
2333 	char	*path;
2334 	struct stat *ub;
2335 };
2336 #endif
2337 int
2338 sys_stat(td, uap)
2339 	struct thread *td;
2340 	register struct stat_args /* {
2341 		char *path;
2342 		struct stat *ub;
2343 	} */ *uap;
2344 {
2345 	struct stat sb;
2346 	int error;
2347 
2348 	error = kern_stat(td, uap->path, UIO_USERSPACE, &sb);
2349 	if (error == 0)
2350 		error = copyout(&sb, uap->ub, sizeof (sb));
2351 	return (error);
2352 }
2353 
2354 #ifndef _SYS_SYSPROTO_H_
2355 struct fstatat_args {
2356 	int	fd;
2357 	char	*path;
2358 	struct stat	*buf;
2359 	int	flag;
2360 }
2361 #endif
2362 int
2363 sys_fstatat(struct thread *td, struct fstatat_args *uap)
2364 {
2365 	struct stat sb;
2366 	int error;
2367 
2368 	error = kern_statat(td, uap->flag, uap->fd, uap->path,
2369 	    UIO_USERSPACE, &sb);
2370 	if (error == 0)
2371 		error = copyout(&sb, uap->buf, sizeof (sb));
2372 	return (error);
2373 }
2374 
2375 int
2376 kern_stat(struct thread *td, char *path, enum uio_seg pathseg, struct stat *sbp)
2377 {
2378 
2379 	return (kern_statat(td, 0, AT_FDCWD, path, pathseg, sbp));
2380 }
2381 
2382 int
2383 kern_statat(struct thread *td, int flag, int fd, char *path,
2384     enum uio_seg pathseg, struct stat *sbp)
2385 {
2386 
2387 	return (kern_statat_vnhook(td, flag, fd, path, pathseg, sbp, NULL));
2388 }
2389 
2390 int
2391 kern_statat_vnhook(struct thread *td, int flag, int fd, char *path,
2392     enum uio_seg pathseg, struct stat *sbp,
2393     void (*hook)(struct vnode *vp, struct stat *sbp))
2394 {
2395 	struct nameidata nd;
2396 	struct stat sb;
2397 	int error, vfslocked;
2398 
2399 	if (flag & ~AT_SYMLINK_NOFOLLOW)
2400 		return (EINVAL);
2401 
2402 	NDINIT_ATRIGHTS(&nd, LOOKUP, ((flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW :
2403 	    FOLLOW) | LOCKSHARED | LOCKLEAF | AUDITVNODE1 | MPSAFE, pathseg,
2404 	    path, fd, CAP_FSTAT, td);
2405 
2406 	if ((error = namei(&nd)) != 0)
2407 		return (error);
2408 	vfslocked = NDHASGIANT(&nd);
2409 	error = vn_stat(nd.ni_vp, &sb, td->td_ucred, NOCRED, td);
2410 	if (!error) {
2411 		SDT_PROBE(vfs, , stat, mode, path, sb.st_mode, 0, 0, 0);
2412 		if (S_ISREG(sb.st_mode))
2413 			SDT_PROBE(vfs, , stat, reg, path, pathseg, 0, 0, 0);
2414 		if (__predict_false(hook != NULL))
2415 			hook(nd.ni_vp, &sb);
2416 	}
2417 	NDFREE(&nd, NDF_ONLY_PNBUF);
2418 	vput(nd.ni_vp);
2419 	VFS_UNLOCK_GIANT(vfslocked);
2420 	if (error)
2421 		return (error);
2422 	*sbp = sb;
2423 #ifdef KTRACE
2424 	if (KTRPOINT(td, KTR_STRUCT))
2425 		ktrstat(&sb);
2426 #endif
2427 	return (0);
2428 }
2429 
2430 /*
2431  * Get file status; this version does not follow links.
2432  */
2433 #ifndef _SYS_SYSPROTO_H_
2434 struct lstat_args {
2435 	char	*path;
2436 	struct stat *ub;
2437 };
2438 #endif
2439 int
2440 sys_lstat(td, uap)
2441 	struct thread *td;
2442 	register struct lstat_args /* {
2443 		char *path;
2444 		struct stat *ub;
2445 	} */ *uap;
2446 {
2447 	struct stat sb;
2448 	int error;
2449 
2450 	error = kern_lstat(td, uap->path, UIO_USERSPACE, &sb);
2451 	if (error == 0)
2452 		error = copyout(&sb, uap->ub, sizeof (sb));
2453 	return (error);
2454 }
2455 
2456 int
2457 kern_lstat(struct thread *td, char *path, enum uio_seg pathseg, struct stat *sbp)
2458 {
2459 
2460 	return (kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, path, pathseg,
2461 	    sbp));
2462 }
2463 
2464 /*
2465  * Implementation of the NetBSD [l]stat() functions.
2466  */
2467 void
2468 cvtnstat(sb, nsb)
2469 	struct stat *sb;
2470 	struct nstat *nsb;
2471 {
2472 	bzero(nsb, sizeof *nsb);
2473 	nsb->st_dev = sb->st_dev;
2474 	nsb->st_ino = sb->st_ino;
2475 	nsb->st_mode = sb->st_mode;
2476 	nsb->st_nlink = sb->st_nlink;
2477 	nsb->st_uid = sb->st_uid;
2478 	nsb->st_gid = sb->st_gid;
2479 	nsb->st_rdev = sb->st_rdev;
2480 	nsb->st_atim = sb->st_atim;
2481 	nsb->st_mtim = sb->st_mtim;
2482 	nsb->st_ctim = sb->st_ctim;
2483 	nsb->st_size = sb->st_size;
2484 	nsb->st_blocks = sb->st_blocks;
2485 	nsb->st_blksize = sb->st_blksize;
2486 	nsb->st_flags = sb->st_flags;
2487 	nsb->st_gen = sb->st_gen;
2488 	nsb->st_birthtim = sb->st_birthtim;
2489 }
2490 
2491 #ifndef _SYS_SYSPROTO_H_
2492 struct nstat_args {
2493 	char	*path;
2494 	struct nstat *ub;
2495 };
2496 #endif
2497 int
2498 sys_nstat(td, uap)
2499 	struct thread *td;
2500 	register struct nstat_args /* {
2501 		char *path;
2502 		struct nstat *ub;
2503 	} */ *uap;
2504 {
2505 	struct stat sb;
2506 	struct nstat nsb;
2507 	int error;
2508 
2509 	error = kern_stat(td, uap->path, UIO_USERSPACE, &sb);
2510 	if (error)
2511 		return (error);
2512 	cvtnstat(&sb, &nsb);
2513 	error = copyout(&nsb, uap->ub, sizeof (nsb));
2514 	return (error);
2515 }
2516 
2517 /*
2518  * NetBSD lstat.  Get file status; this version does not follow links.
2519  */
2520 #ifndef _SYS_SYSPROTO_H_
2521 struct lstat_args {
2522 	char	*path;
2523 	struct stat *ub;
2524 };
2525 #endif
2526 int
2527 sys_nlstat(td, uap)
2528 	struct thread *td;
2529 	register struct nlstat_args /* {
2530 		char *path;
2531 		struct nstat *ub;
2532 	} */ *uap;
2533 {
2534 	struct stat sb;
2535 	struct nstat nsb;
2536 	int error;
2537 
2538 	error = kern_lstat(td, uap->path, UIO_USERSPACE, &sb);
2539 	if (error)
2540 		return (error);
2541 	cvtnstat(&sb, &nsb);
2542 	error = copyout(&nsb, uap->ub, sizeof (nsb));
2543 	return (error);
2544 }
2545 
2546 /*
2547  * Get configurable pathname variables.
2548  */
2549 #ifndef _SYS_SYSPROTO_H_
2550 struct pathconf_args {
2551 	char	*path;
2552 	int	name;
2553 };
2554 #endif
2555 int
2556 sys_pathconf(td, uap)
2557 	struct thread *td;
2558 	register struct pathconf_args /* {
2559 		char *path;
2560 		int name;
2561 	} */ *uap;
2562 {
2563 
2564 	return (kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name, FOLLOW));
2565 }
2566 
2567 #ifndef _SYS_SYSPROTO_H_
2568 struct lpathconf_args {
2569 	char	*path;
2570 	int	name;
2571 };
2572 #endif
2573 int
2574 sys_lpathconf(td, uap)
2575 	struct thread *td;
2576 	register struct lpathconf_args /* {
2577 		char *path;
2578 		int name;
2579 	} */ *uap;
2580 {
2581 
2582 	return (kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name, NOFOLLOW));
2583 }
2584 
2585 int
2586 kern_pathconf(struct thread *td, char *path, enum uio_seg pathseg, int name,
2587     u_long flags)
2588 {
2589 	struct nameidata nd;
2590 	int error, vfslocked;
2591 
2592 	NDINIT(&nd, LOOKUP, LOCKSHARED | LOCKLEAF | MPSAFE | AUDITVNODE1 |
2593 	    flags, pathseg, path, td);
2594 	if ((error = namei(&nd)) != 0)
2595 		return (error);
2596 	vfslocked = NDHASGIANT(&nd);
2597 	NDFREE(&nd, NDF_ONLY_PNBUF);
2598 
2599 	/* If asynchronous I/O is available, it works for all files. */
2600 	if (name == _PC_ASYNC_IO)
2601 		td->td_retval[0] = async_io_version;
2602 	else
2603 		error = VOP_PATHCONF(nd.ni_vp, name, td->td_retval);
2604 	vput(nd.ni_vp);
2605 	VFS_UNLOCK_GIANT(vfslocked);
2606 	return (error);
2607 }
2608 
2609 /*
2610  * Return target name of a symbolic link.
2611  */
2612 #ifndef _SYS_SYSPROTO_H_
2613 struct readlink_args {
2614 	char	*path;
2615 	char	*buf;
2616 	size_t	count;
2617 };
2618 #endif
2619 int
2620 sys_readlink(td, uap)
2621 	struct thread *td;
2622 	register struct readlink_args /* {
2623 		char *path;
2624 		char *buf;
2625 		size_t count;
2626 	} */ *uap;
2627 {
2628 
2629 	return (kern_readlink(td, uap->path, UIO_USERSPACE, uap->buf,
2630 	    UIO_USERSPACE, uap->count));
2631 }
2632 #ifndef _SYS_SYSPROTO_H_
2633 struct readlinkat_args {
2634 	int	fd;
2635 	char	*path;
2636 	char	*buf;
2637 	size_t	bufsize;
2638 };
2639 #endif
2640 int
2641 sys_readlinkat(struct thread *td, struct readlinkat_args *uap)
2642 {
2643 
2644 	return (kern_readlinkat(td, uap->fd, uap->path, UIO_USERSPACE,
2645 	    uap->buf, UIO_USERSPACE, uap->bufsize));
2646 }
2647 
2648 int
2649 kern_readlink(struct thread *td, char *path, enum uio_seg pathseg, char *buf,
2650     enum uio_seg bufseg, size_t count)
2651 {
2652 
2653 	return (kern_readlinkat(td, AT_FDCWD, path, pathseg, buf, bufseg,
2654 	    count));
2655 }
2656 
2657 int
2658 kern_readlinkat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2659     char *buf, enum uio_seg bufseg, size_t count)
2660 {
2661 	struct vnode *vp;
2662 	struct iovec aiov;
2663 	struct uio auio;
2664 	int error;
2665 	struct nameidata nd;
2666 	int vfslocked;
2667 
2668 	if (count > IOSIZE_MAX)
2669 		return (EINVAL);
2670 
2671 	NDINIT_AT(&nd, LOOKUP, NOFOLLOW | LOCKSHARED | LOCKLEAF | MPSAFE |
2672 	    AUDITVNODE1, pathseg, path, fd, td);
2673 
2674 	if ((error = namei(&nd)) != 0)
2675 		return (error);
2676 	NDFREE(&nd, NDF_ONLY_PNBUF);
2677 	vfslocked = NDHASGIANT(&nd);
2678 	vp = nd.ni_vp;
2679 #ifdef MAC
2680 	error = mac_vnode_check_readlink(td->td_ucred, vp);
2681 	if (error) {
2682 		vput(vp);
2683 		VFS_UNLOCK_GIANT(vfslocked);
2684 		return (error);
2685 	}
2686 #endif
2687 	if (vp->v_type != VLNK)
2688 		error = EINVAL;
2689 	else {
2690 		aiov.iov_base = buf;
2691 		aiov.iov_len = count;
2692 		auio.uio_iov = &aiov;
2693 		auio.uio_iovcnt = 1;
2694 		auio.uio_offset = 0;
2695 		auio.uio_rw = UIO_READ;
2696 		auio.uio_segflg = bufseg;
2697 		auio.uio_td = td;
2698 		auio.uio_resid = count;
2699 		error = VOP_READLINK(vp, &auio, td->td_ucred);
2700 	}
2701 	vput(vp);
2702 	VFS_UNLOCK_GIANT(vfslocked);
2703 	td->td_retval[0] = count - auio.uio_resid;
2704 	return (error);
2705 }
2706 
2707 /*
2708  * Common implementation code for chflags() and fchflags().
2709  */
2710 static int
2711 setfflags(td, vp, flags)
2712 	struct thread *td;
2713 	struct vnode *vp;
2714 	int flags;
2715 {
2716 	int error;
2717 	struct mount *mp;
2718 	struct vattr vattr;
2719 
2720 	/* We can't support the value matching VNOVAL. */
2721 	if (flags == VNOVAL)
2722 		return (EOPNOTSUPP);
2723 
2724 	/*
2725 	 * Prevent non-root users from setting flags on devices.  When
2726 	 * a device is reused, users can retain ownership of the device
2727 	 * if they are allowed to set flags and programs assume that
2728 	 * chown can't fail when done as root.
2729 	 */
2730 	if (vp->v_type == VCHR || vp->v_type == VBLK) {
2731 		error = priv_check(td, PRIV_VFS_CHFLAGS_DEV);
2732 		if (error)
2733 			return (error);
2734 	}
2735 
2736 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2737 		return (error);
2738 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2739 	VATTR_NULL(&vattr);
2740 	vattr.va_flags = flags;
2741 #ifdef MAC
2742 	error = mac_vnode_check_setflags(td->td_ucred, vp, vattr.va_flags);
2743 	if (error == 0)
2744 #endif
2745 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
2746 	VOP_UNLOCK(vp, 0);
2747 	vn_finished_write(mp);
2748 	return (error);
2749 }
2750 
2751 /*
2752  * Change flags of a file given a path name.
2753  */
2754 #ifndef _SYS_SYSPROTO_H_
2755 struct chflags_args {
2756 	char	*path;
2757 	int	flags;
2758 };
2759 #endif
2760 int
2761 sys_chflags(td, uap)
2762 	struct thread *td;
2763 	register struct chflags_args /* {
2764 		char *path;
2765 		int flags;
2766 	} */ *uap;
2767 {
2768 	int error;
2769 	struct nameidata nd;
2770 	int vfslocked;
2771 
2772 	AUDIT_ARG_FFLAGS(uap->flags);
2773 	NDINIT(&nd, LOOKUP, FOLLOW | MPSAFE | AUDITVNODE1, UIO_USERSPACE,
2774 	    uap->path, td);
2775 	if ((error = namei(&nd)) != 0)
2776 		return (error);
2777 	NDFREE(&nd, NDF_ONLY_PNBUF);
2778 	vfslocked = NDHASGIANT(&nd);
2779 	error = setfflags(td, nd.ni_vp, uap->flags);
2780 	vrele(nd.ni_vp);
2781 	VFS_UNLOCK_GIANT(vfslocked);
2782 	return (error);
2783 }
2784 
2785 /*
2786  * Same as chflags() but doesn't follow symlinks.
2787  */
2788 int
2789 sys_lchflags(td, uap)
2790 	struct thread *td;
2791 	register struct lchflags_args /* {
2792 		char *path;
2793 		int flags;
2794 	} */ *uap;
2795 {
2796 	int error;
2797 	struct nameidata nd;
2798 	int vfslocked;
2799 
2800 	AUDIT_ARG_FFLAGS(uap->flags);
2801 	NDINIT(&nd, LOOKUP, NOFOLLOW | MPSAFE | AUDITVNODE1, UIO_USERSPACE,
2802 	    uap->path, td);
2803 	if ((error = namei(&nd)) != 0)
2804 		return (error);
2805 	vfslocked = NDHASGIANT(&nd);
2806 	NDFREE(&nd, NDF_ONLY_PNBUF);
2807 	error = setfflags(td, nd.ni_vp, uap->flags);
2808 	vrele(nd.ni_vp);
2809 	VFS_UNLOCK_GIANT(vfslocked);
2810 	return (error);
2811 }
2812 
2813 /*
2814  * Change flags of a file given a file descriptor.
2815  */
2816 #ifndef _SYS_SYSPROTO_H_
2817 struct fchflags_args {
2818 	int	fd;
2819 	int	flags;
2820 };
2821 #endif
2822 int
2823 sys_fchflags(td, uap)
2824 	struct thread *td;
2825 	register struct fchflags_args /* {
2826 		int fd;
2827 		int flags;
2828 	} */ *uap;
2829 {
2830 	struct file *fp;
2831 	int vfslocked;
2832 	int error;
2833 
2834 	AUDIT_ARG_FD(uap->fd);
2835 	AUDIT_ARG_FFLAGS(uap->flags);
2836 	if ((error = getvnode(td->td_proc->p_fd, uap->fd, CAP_FCHFLAGS,
2837 	    &fp)) != 0)
2838 		return (error);
2839 	vfslocked = VFS_LOCK_GIANT(fp->f_vnode->v_mount);
2840 #ifdef AUDIT
2841 	vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
2842 	AUDIT_ARG_VNODE1(fp->f_vnode);
2843 	VOP_UNLOCK(fp->f_vnode, 0);
2844 #endif
2845 	error = setfflags(td, fp->f_vnode, uap->flags);
2846 	VFS_UNLOCK_GIANT(vfslocked);
2847 	fdrop(fp, td);
2848 	return (error);
2849 }
2850 
2851 /*
2852  * Common implementation code for chmod(), lchmod() and fchmod().
2853  */
2854 int
2855 setfmode(td, cred, vp, mode)
2856 	struct thread *td;
2857 	struct ucred *cred;
2858 	struct vnode *vp;
2859 	int mode;
2860 {
2861 	int error;
2862 	struct mount *mp;
2863 	struct vattr vattr;
2864 
2865 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2866 		return (error);
2867 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2868 	VATTR_NULL(&vattr);
2869 	vattr.va_mode = mode & ALLPERMS;
2870 #ifdef MAC
2871 	error = mac_vnode_check_setmode(cred, vp, vattr.va_mode);
2872 	if (error == 0)
2873 #endif
2874 		error = VOP_SETATTR(vp, &vattr, cred);
2875 	VOP_UNLOCK(vp, 0);
2876 	vn_finished_write(mp);
2877 	return (error);
2878 }
2879 
2880 /*
2881  * Change mode of a file given path name.
2882  */
2883 #ifndef _SYS_SYSPROTO_H_
2884 struct chmod_args {
2885 	char	*path;
2886 	int	mode;
2887 };
2888 #endif
2889 int
2890 sys_chmod(td, uap)
2891 	struct thread *td;
2892 	register struct chmod_args /* {
2893 		char *path;
2894 		int mode;
2895 	} */ *uap;
2896 {
2897 
2898 	return (kern_chmod(td, uap->path, UIO_USERSPACE, uap->mode));
2899 }
2900 
2901 #ifndef _SYS_SYSPROTO_H_
2902 struct fchmodat_args {
2903 	int	dirfd;
2904 	char	*path;
2905 	mode_t	mode;
2906 	int	flag;
2907 }
2908 #endif
2909 int
2910 sys_fchmodat(struct thread *td, struct fchmodat_args *uap)
2911 {
2912 	int flag = uap->flag;
2913 	int fd = uap->fd;
2914 	char *path = uap->path;
2915 	mode_t mode = uap->mode;
2916 
2917 	if (flag & ~AT_SYMLINK_NOFOLLOW)
2918 		return (EINVAL);
2919 
2920 	return (kern_fchmodat(td, fd, path, UIO_USERSPACE, mode, flag));
2921 }
2922 
2923 int
2924 kern_chmod(struct thread *td, char *path, enum uio_seg pathseg, int mode)
2925 {
2926 
2927 	return (kern_fchmodat(td, AT_FDCWD, path, pathseg, mode, 0));
2928 }
2929 
2930 /*
2931  * Change mode of a file given path name (don't follow links.)
2932  */
2933 #ifndef _SYS_SYSPROTO_H_
2934 struct lchmod_args {
2935 	char	*path;
2936 	int	mode;
2937 };
2938 #endif
2939 int
2940 sys_lchmod(td, uap)
2941 	struct thread *td;
2942 	register struct lchmod_args /* {
2943 		char *path;
2944 		int mode;
2945 	} */ *uap;
2946 {
2947 
2948 	return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2949 	    uap->mode, AT_SYMLINK_NOFOLLOW));
2950 }
2951 
2952 
2953 int
2954 kern_fchmodat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2955     mode_t mode, int flag)
2956 {
2957 	int error;
2958 	struct nameidata nd;
2959 	int vfslocked;
2960 	int follow;
2961 
2962 	AUDIT_ARG_MODE(mode);
2963 	follow = (flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
2964 	NDINIT_ATRIGHTS(&nd, LOOKUP,  follow | MPSAFE | AUDITVNODE1, pathseg,
2965 	    path, fd, CAP_FCHMOD, td);
2966 	if ((error = namei(&nd)) != 0)
2967 		return (error);
2968 	vfslocked = NDHASGIANT(&nd);
2969 	NDFREE(&nd, NDF_ONLY_PNBUF);
2970 	error = setfmode(td, td->td_ucred, nd.ni_vp, mode);
2971 	vrele(nd.ni_vp);
2972 	VFS_UNLOCK_GIANT(vfslocked);
2973 	return (error);
2974 }
2975 
2976 /*
2977  * Change mode of a file given a file descriptor.
2978  */
2979 #ifndef _SYS_SYSPROTO_H_
2980 struct fchmod_args {
2981 	int	fd;
2982 	int	mode;
2983 };
2984 #endif
2985 int
2986 sys_fchmod(struct thread *td, struct fchmod_args *uap)
2987 {
2988 	struct file *fp;
2989 	int error;
2990 
2991 	AUDIT_ARG_FD(uap->fd);
2992 	AUDIT_ARG_MODE(uap->mode);
2993 
2994 	error = fget(td, uap->fd, CAP_FCHMOD, &fp);
2995 	if (error != 0)
2996 		return (error);
2997 	error = fo_chmod(fp, uap->mode, td->td_ucred, td);
2998 	fdrop(fp, td);
2999 	return (error);
3000 }
3001 
3002 /*
3003  * Common implementation for chown(), lchown(), and fchown()
3004  */
3005 int
3006 setfown(td, cred, vp, uid, gid)
3007 	struct thread *td;
3008 	struct ucred *cred;
3009 	struct vnode *vp;
3010 	uid_t uid;
3011 	gid_t gid;
3012 {
3013 	int error;
3014 	struct mount *mp;
3015 	struct vattr vattr;
3016 
3017 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
3018 		return (error);
3019 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3020 	VATTR_NULL(&vattr);
3021 	vattr.va_uid = uid;
3022 	vattr.va_gid = gid;
3023 #ifdef MAC
3024 	error = mac_vnode_check_setowner(cred, vp, vattr.va_uid,
3025 	    vattr.va_gid);
3026 	if (error == 0)
3027 #endif
3028 		error = VOP_SETATTR(vp, &vattr, cred);
3029 	VOP_UNLOCK(vp, 0);
3030 	vn_finished_write(mp);
3031 	return (error);
3032 }
3033 
3034 /*
3035  * Set ownership given a path name.
3036  */
3037 #ifndef _SYS_SYSPROTO_H_
3038 struct chown_args {
3039 	char	*path;
3040 	int	uid;
3041 	int	gid;
3042 };
3043 #endif
3044 int
3045 sys_chown(td, uap)
3046 	struct thread *td;
3047 	register struct chown_args /* {
3048 		char *path;
3049 		int uid;
3050 		int gid;
3051 	} */ *uap;
3052 {
3053 
3054 	return (kern_chown(td, uap->path, UIO_USERSPACE, uap->uid, uap->gid));
3055 }
3056 
3057 #ifndef _SYS_SYSPROTO_H_
3058 struct fchownat_args {
3059 	int fd;
3060 	const char * path;
3061 	uid_t uid;
3062 	gid_t gid;
3063 	int flag;
3064 };
3065 #endif
3066 int
3067 sys_fchownat(struct thread *td, struct fchownat_args *uap)
3068 {
3069 	int flag;
3070 
3071 	flag = uap->flag;
3072 	if (flag & ~AT_SYMLINK_NOFOLLOW)
3073 		return (EINVAL);
3074 
3075 	return (kern_fchownat(td, uap->fd, uap->path, UIO_USERSPACE, uap->uid,
3076 	    uap->gid, uap->flag));
3077 }
3078 
3079 int
3080 kern_chown(struct thread *td, char *path, enum uio_seg pathseg, int uid,
3081     int gid)
3082 {
3083 
3084 	return (kern_fchownat(td, AT_FDCWD, path, pathseg, uid, gid, 0));
3085 }
3086 
3087 int
3088 kern_fchownat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
3089     int uid, int gid, int flag)
3090 {
3091 	struct nameidata nd;
3092 	int error, vfslocked, follow;
3093 
3094 	AUDIT_ARG_OWNER(uid, gid);
3095 	follow = (flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
3096 	NDINIT_ATRIGHTS(&nd, LOOKUP, follow | MPSAFE | AUDITVNODE1, pathseg,
3097 	    path, fd, CAP_FCHOWN, td);
3098 
3099 	if ((error = namei(&nd)) != 0)
3100 		return (error);
3101 	vfslocked = NDHASGIANT(&nd);
3102 	NDFREE(&nd, NDF_ONLY_PNBUF);
3103 	error = setfown(td, td->td_ucred, nd.ni_vp, uid, gid);
3104 	vrele(nd.ni_vp);
3105 	VFS_UNLOCK_GIANT(vfslocked);
3106 	return (error);
3107 }
3108 
3109 /*
3110  * Set ownership given a path name, do not cross symlinks.
3111  */
3112 #ifndef _SYS_SYSPROTO_H_
3113 struct lchown_args {
3114 	char	*path;
3115 	int	uid;
3116 	int	gid;
3117 };
3118 #endif
3119 int
3120 sys_lchown(td, uap)
3121 	struct thread *td;
3122 	register struct lchown_args /* {
3123 		char *path;
3124 		int uid;
3125 		int gid;
3126 	} */ *uap;
3127 {
3128 
3129 	return (kern_lchown(td, uap->path, UIO_USERSPACE, uap->uid, uap->gid));
3130 }
3131 
3132 int
3133 kern_lchown(struct thread *td, char *path, enum uio_seg pathseg, int uid,
3134     int gid)
3135 {
3136 
3137 	return (kern_fchownat(td, AT_FDCWD, path, pathseg, uid, gid,
3138 	    AT_SYMLINK_NOFOLLOW));
3139 }
3140 
3141 /*
3142  * Set ownership given a file descriptor.
3143  */
3144 #ifndef _SYS_SYSPROTO_H_
3145 struct fchown_args {
3146 	int	fd;
3147 	int	uid;
3148 	int	gid;
3149 };
3150 #endif
3151 int
3152 sys_fchown(td, uap)
3153 	struct thread *td;
3154 	register struct fchown_args /* {
3155 		int fd;
3156 		int uid;
3157 		int gid;
3158 	} */ *uap;
3159 {
3160 	struct file *fp;
3161 	int error;
3162 
3163 	AUDIT_ARG_FD(uap->fd);
3164 	AUDIT_ARG_OWNER(uap->uid, uap->gid);
3165 	error = fget(td, uap->fd, CAP_FCHOWN, &fp);
3166 	if (error != 0)
3167 		return (error);
3168 	error = fo_chown(fp, uap->uid, uap->gid, td->td_ucred, td);
3169 	fdrop(fp, td);
3170 	return (error);
3171 }
3172 
3173 /*
3174  * Common implementation code for utimes(), lutimes(), and futimes().
3175  */
3176 static int
3177 getutimes(usrtvp, tvpseg, tsp)
3178 	const struct timeval *usrtvp;
3179 	enum uio_seg tvpseg;
3180 	struct timespec *tsp;
3181 {
3182 	struct timeval tv[2];
3183 	const struct timeval *tvp;
3184 	int error;
3185 
3186 	if (usrtvp == NULL) {
3187 		vfs_timestamp(&tsp[0]);
3188 		tsp[1] = tsp[0];
3189 	} else {
3190 		if (tvpseg == UIO_SYSSPACE) {
3191 			tvp = usrtvp;
3192 		} else {
3193 			if ((error = copyin(usrtvp, tv, sizeof(tv))) != 0)
3194 				return (error);
3195 			tvp = tv;
3196 		}
3197 
3198 		if (tvp[0].tv_usec < 0 || tvp[0].tv_usec >= 1000000 ||
3199 		    tvp[1].tv_usec < 0 || tvp[1].tv_usec >= 1000000)
3200 			return (EINVAL);
3201 		TIMEVAL_TO_TIMESPEC(&tvp[0], &tsp[0]);
3202 		TIMEVAL_TO_TIMESPEC(&tvp[1], &tsp[1]);
3203 	}
3204 	return (0);
3205 }
3206 
3207 /*
3208  * Common implementation code for utimes(), lutimes(), and futimes().
3209  */
3210 static int
3211 setutimes(td, vp, ts, numtimes, nullflag)
3212 	struct thread *td;
3213 	struct vnode *vp;
3214 	const struct timespec *ts;
3215 	int numtimes;
3216 	int nullflag;
3217 {
3218 	int error, setbirthtime;
3219 	struct mount *mp;
3220 	struct vattr vattr;
3221 
3222 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
3223 		return (error);
3224 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3225 	setbirthtime = 0;
3226 	if (numtimes < 3 && !VOP_GETATTR(vp, &vattr, td->td_ucred) &&
3227 	    timespeccmp(&ts[1], &vattr.va_birthtime, < ))
3228 		setbirthtime = 1;
3229 	VATTR_NULL(&vattr);
3230 	vattr.va_atime = ts[0];
3231 	vattr.va_mtime = ts[1];
3232 	if (setbirthtime)
3233 		vattr.va_birthtime = ts[1];
3234 	if (numtimes > 2)
3235 		vattr.va_birthtime = ts[2];
3236 	if (nullflag)
3237 		vattr.va_vaflags |= VA_UTIMES_NULL;
3238 #ifdef MAC
3239 	error = mac_vnode_check_setutimes(td->td_ucred, vp, vattr.va_atime,
3240 	    vattr.va_mtime);
3241 #endif
3242 	if (error == 0)
3243 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3244 	VOP_UNLOCK(vp, 0);
3245 	vn_finished_write(mp);
3246 	return (error);
3247 }
3248 
3249 /*
3250  * Set the access and modification times of a file.
3251  */
3252 #ifndef _SYS_SYSPROTO_H_
3253 struct utimes_args {
3254 	char	*path;
3255 	struct	timeval *tptr;
3256 };
3257 #endif
3258 int
3259 sys_utimes(td, uap)
3260 	struct thread *td;
3261 	register struct utimes_args /* {
3262 		char *path;
3263 		struct timeval *tptr;
3264 	} */ *uap;
3265 {
3266 
3267 	return (kern_utimes(td, uap->path, UIO_USERSPACE, uap->tptr,
3268 	    UIO_USERSPACE));
3269 }
3270 
3271 #ifndef _SYS_SYSPROTO_H_
3272 struct futimesat_args {
3273 	int fd;
3274 	const char * path;
3275 	const struct timeval * times;
3276 };
3277 #endif
3278 int
3279 sys_futimesat(struct thread *td, struct futimesat_args *uap)
3280 {
3281 
3282 	return (kern_utimesat(td, uap->fd, uap->path, UIO_USERSPACE,
3283 	    uap->times, UIO_USERSPACE));
3284 }
3285 
3286 int
3287 kern_utimes(struct thread *td, char *path, enum uio_seg pathseg,
3288     struct timeval *tptr, enum uio_seg tptrseg)
3289 {
3290 
3291 	return (kern_utimesat(td, AT_FDCWD, path, pathseg, tptr, tptrseg));
3292 }
3293 
3294 int
3295 kern_utimesat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
3296     struct timeval *tptr, enum uio_seg tptrseg)
3297 {
3298 	struct nameidata nd;
3299 	struct timespec ts[2];
3300 	int error, vfslocked;
3301 
3302 	if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3303 		return (error);
3304 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | MPSAFE | AUDITVNODE1, pathseg,
3305 	    path, fd, CAP_FUTIMES, td);
3306 
3307 	if ((error = namei(&nd)) != 0)
3308 		return (error);
3309 	vfslocked = NDHASGIANT(&nd);
3310 	NDFREE(&nd, NDF_ONLY_PNBUF);
3311 	error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3312 	vrele(nd.ni_vp);
3313 	VFS_UNLOCK_GIANT(vfslocked);
3314 	return (error);
3315 }
3316 
3317 /*
3318  * Set the access and modification times of a file.
3319  */
3320 #ifndef _SYS_SYSPROTO_H_
3321 struct lutimes_args {
3322 	char	*path;
3323 	struct	timeval *tptr;
3324 };
3325 #endif
3326 int
3327 sys_lutimes(td, uap)
3328 	struct thread *td;
3329 	register struct lutimes_args /* {
3330 		char *path;
3331 		struct timeval *tptr;
3332 	} */ *uap;
3333 {
3334 
3335 	return (kern_lutimes(td, uap->path, UIO_USERSPACE, uap->tptr,
3336 	    UIO_USERSPACE));
3337 }
3338 
3339 int
3340 kern_lutimes(struct thread *td, char *path, enum uio_seg pathseg,
3341     struct timeval *tptr, enum uio_seg tptrseg)
3342 {
3343 	struct timespec ts[2];
3344 	int error;
3345 	struct nameidata nd;
3346 	int vfslocked;
3347 
3348 	if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3349 		return (error);
3350 	NDINIT(&nd, LOOKUP, NOFOLLOW | MPSAFE | AUDITVNODE1, pathseg, path, td);
3351 	if ((error = namei(&nd)) != 0)
3352 		return (error);
3353 	vfslocked = NDHASGIANT(&nd);
3354 	NDFREE(&nd, NDF_ONLY_PNBUF);
3355 	error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3356 	vrele(nd.ni_vp);
3357 	VFS_UNLOCK_GIANT(vfslocked);
3358 	return (error);
3359 }
3360 
3361 /*
3362  * Set the access and modification times of a file.
3363  */
3364 #ifndef _SYS_SYSPROTO_H_
3365 struct futimes_args {
3366 	int	fd;
3367 	struct	timeval *tptr;
3368 };
3369 #endif
3370 int
3371 sys_futimes(td, uap)
3372 	struct thread *td;
3373 	register struct futimes_args /* {
3374 		int  fd;
3375 		struct timeval *tptr;
3376 	} */ *uap;
3377 {
3378 
3379 	return (kern_futimes(td, uap->fd, uap->tptr, UIO_USERSPACE));
3380 }
3381 
3382 int
3383 kern_futimes(struct thread *td, int fd, struct timeval *tptr,
3384     enum uio_seg tptrseg)
3385 {
3386 	struct timespec ts[2];
3387 	struct file *fp;
3388 	int vfslocked;
3389 	int error;
3390 
3391 	AUDIT_ARG_FD(fd);
3392 	if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3393 		return (error);
3394 	if ((error = getvnode(td->td_proc->p_fd, fd, CAP_FUTIMES, &fp))
3395 	    != 0)
3396 		return (error);
3397 	vfslocked = VFS_LOCK_GIANT(fp->f_vnode->v_mount);
3398 #ifdef AUDIT
3399 	vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
3400 	AUDIT_ARG_VNODE1(fp->f_vnode);
3401 	VOP_UNLOCK(fp->f_vnode, 0);
3402 #endif
3403 	error = setutimes(td, fp->f_vnode, ts, 2, tptr == NULL);
3404 	VFS_UNLOCK_GIANT(vfslocked);
3405 	fdrop(fp, td);
3406 	return (error);
3407 }
3408 
3409 /*
3410  * Truncate a file given its path name.
3411  */
3412 #ifndef _SYS_SYSPROTO_H_
3413 struct truncate_args {
3414 	char	*path;
3415 	int	pad;
3416 	off_t	length;
3417 };
3418 #endif
3419 int
3420 sys_truncate(td, uap)
3421 	struct thread *td;
3422 	register struct truncate_args /* {
3423 		char *path;
3424 		int pad;
3425 		off_t length;
3426 	} */ *uap;
3427 {
3428 
3429 	return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3430 }
3431 
3432 int
3433 kern_truncate(struct thread *td, char *path, enum uio_seg pathseg, off_t length)
3434 {
3435 	struct mount *mp;
3436 	struct vnode *vp;
3437 	struct vattr vattr;
3438 	int error;
3439 	struct nameidata nd;
3440 	int vfslocked;
3441 
3442 	if (length < 0)
3443 		return(EINVAL);
3444 	NDINIT(&nd, LOOKUP, FOLLOW | MPSAFE | AUDITVNODE1, pathseg, path, td);
3445 	if ((error = namei(&nd)) != 0)
3446 		return (error);
3447 	vfslocked = NDHASGIANT(&nd);
3448 	vp = nd.ni_vp;
3449 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) {
3450 		vrele(vp);
3451 		VFS_UNLOCK_GIANT(vfslocked);
3452 		return (error);
3453 	}
3454 	NDFREE(&nd, NDF_ONLY_PNBUF);
3455 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3456 	if (vp->v_type == VDIR)
3457 		error = EISDIR;
3458 #ifdef MAC
3459 	else if ((error = mac_vnode_check_write(td->td_ucred, NOCRED, vp))) {
3460 	}
3461 #endif
3462 	else if ((error = vn_writechk(vp)) == 0 &&
3463 	    (error = VOP_ACCESS(vp, VWRITE, td->td_ucred, td)) == 0) {
3464 		VATTR_NULL(&vattr);
3465 		vattr.va_size = length;
3466 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3467 	}
3468 	vput(vp);
3469 	vn_finished_write(mp);
3470 	VFS_UNLOCK_GIANT(vfslocked);
3471 	return (error);
3472 }
3473 
3474 #if defined(COMPAT_43)
3475 /*
3476  * Truncate a file given its path name.
3477  */
3478 #ifndef _SYS_SYSPROTO_H_
3479 struct otruncate_args {
3480 	char	*path;
3481 	long	length;
3482 };
3483 #endif
3484 int
3485 otruncate(td, uap)
3486 	struct thread *td;
3487 	register struct otruncate_args /* {
3488 		char *path;
3489 		long length;
3490 	} */ *uap;
3491 {
3492 	struct truncate_args /* {
3493 		char *path;
3494 		int pad;
3495 		off_t length;
3496 	} */ nuap;
3497 
3498 	nuap.path = uap->path;
3499 	nuap.length = uap->length;
3500 	return (sys_truncate(td, &nuap));
3501 }
3502 #endif /* COMPAT_43 */
3503 
3504 /* Versions with the pad argument */
3505 int
3506 freebsd6_truncate(struct thread *td, struct freebsd6_truncate_args *uap)
3507 {
3508 	struct truncate_args ouap;
3509 
3510 	ouap.path = uap->path;
3511 	ouap.length = uap->length;
3512 	return (sys_truncate(td, &ouap));
3513 }
3514 
3515 int
3516 freebsd6_ftruncate(struct thread *td, struct freebsd6_ftruncate_args *uap)
3517 {
3518 	struct ftruncate_args ouap;
3519 
3520 	ouap.fd = uap->fd;
3521 	ouap.length = uap->length;
3522 	return (sys_ftruncate(td, &ouap));
3523 }
3524 
3525 /*
3526  * Sync an open file.
3527  */
3528 #ifndef _SYS_SYSPROTO_H_
3529 struct fsync_args {
3530 	int	fd;
3531 };
3532 #endif
3533 int
3534 sys_fsync(td, uap)
3535 	struct thread *td;
3536 	struct fsync_args /* {
3537 		int fd;
3538 	} */ *uap;
3539 {
3540 	struct vnode *vp;
3541 	struct mount *mp;
3542 	struct file *fp;
3543 	int vfslocked;
3544 	int error, lock_flags;
3545 
3546 	AUDIT_ARG_FD(uap->fd);
3547 	if ((error = getvnode(td->td_proc->p_fd, uap->fd, CAP_FSYNC,
3548 	    &fp)) != 0)
3549 		return (error);
3550 	vp = fp->f_vnode;
3551 	vfslocked = VFS_LOCK_GIANT(vp->v_mount);
3552 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
3553 		goto drop;
3554 	if (MNT_SHARED_WRITES(mp) ||
3555 	    ((mp == NULL) && MNT_SHARED_WRITES(vp->v_mount))) {
3556 		lock_flags = LK_SHARED;
3557 	} else {
3558 		lock_flags = LK_EXCLUSIVE;
3559 	}
3560 	vn_lock(vp, lock_flags | LK_RETRY);
3561 	AUDIT_ARG_VNODE1(vp);
3562 	if (vp->v_object != NULL) {
3563 		VM_OBJECT_LOCK(vp->v_object);
3564 		vm_object_page_clean(vp->v_object, 0, 0, 0);
3565 		VM_OBJECT_UNLOCK(vp->v_object);
3566 	}
3567 	error = VOP_FSYNC(vp, MNT_WAIT, td);
3568 
3569 	VOP_UNLOCK(vp, 0);
3570 	vn_finished_write(mp);
3571 drop:
3572 	VFS_UNLOCK_GIANT(vfslocked);
3573 	fdrop(fp, td);
3574 	return (error);
3575 }
3576 
3577 /*
3578  * Rename files.  Source and destination must either both be directories, or
3579  * both not be directories.  If target is a directory, it must be empty.
3580  */
3581 #ifndef _SYS_SYSPROTO_H_
3582 struct rename_args {
3583 	char	*from;
3584 	char	*to;
3585 };
3586 #endif
3587 int
3588 sys_rename(td, uap)
3589 	struct thread *td;
3590 	register struct rename_args /* {
3591 		char *from;
3592 		char *to;
3593 	} */ *uap;
3594 {
3595 
3596 	return (kern_rename(td, uap->from, uap->to, UIO_USERSPACE));
3597 }
3598 
3599 #ifndef _SYS_SYSPROTO_H_
3600 struct renameat_args {
3601 	int	oldfd;
3602 	char	*old;
3603 	int	newfd;
3604 	char	*new;
3605 };
3606 #endif
3607 int
3608 sys_renameat(struct thread *td, struct renameat_args *uap)
3609 {
3610 
3611 	return (kern_renameat(td, uap->oldfd, uap->old, uap->newfd, uap->new,
3612 	    UIO_USERSPACE));
3613 }
3614 
3615 int
3616 kern_rename(struct thread *td, char *from, char *to, enum uio_seg pathseg)
3617 {
3618 
3619 	return (kern_renameat(td, AT_FDCWD, from, AT_FDCWD, to, pathseg));
3620 }
3621 
3622 int
3623 kern_renameat(struct thread *td, int oldfd, char *old, int newfd, char *new,
3624     enum uio_seg pathseg)
3625 {
3626 	struct mount *mp = NULL;
3627 	struct vnode *tvp, *fvp, *tdvp;
3628 	struct nameidata fromnd, tond;
3629 	int tvfslocked;
3630 	int fvfslocked;
3631 	int error;
3632 
3633 	bwillwrite();
3634 #ifdef MAC
3635 	NDINIT_ATRIGHTS(&fromnd, DELETE, LOCKPARENT | LOCKLEAF | SAVESTART |
3636 	    MPSAFE | AUDITVNODE1, pathseg, old, oldfd, CAP_DELETE, td);
3637 #else
3638 	NDINIT_ATRIGHTS(&fromnd, DELETE, WANTPARENT | SAVESTART | MPSAFE |
3639 	    AUDITVNODE1, pathseg, old, oldfd, CAP_DELETE, td);
3640 #endif
3641 
3642 	if ((error = namei(&fromnd)) != 0)
3643 		return (error);
3644 	fvfslocked = NDHASGIANT(&fromnd);
3645 	tvfslocked = 0;
3646 #ifdef MAC
3647 	error = mac_vnode_check_rename_from(td->td_ucred, fromnd.ni_dvp,
3648 	    fromnd.ni_vp, &fromnd.ni_cnd);
3649 	VOP_UNLOCK(fromnd.ni_dvp, 0);
3650 	if (fromnd.ni_dvp != fromnd.ni_vp)
3651 		VOP_UNLOCK(fromnd.ni_vp, 0);
3652 #endif
3653 	fvp = fromnd.ni_vp;
3654 	if (error == 0)
3655 		error = vn_start_write(fvp, &mp, V_WAIT | PCATCH);
3656 	if (error != 0) {
3657 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3658 		vrele(fromnd.ni_dvp);
3659 		vrele(fvp);
3660 		goto out1;
3661 	}
3662 	NDINIT_ATRIGHTS(&tond, RENAME, LOCKPARENT | LOCKLEAF | NOCACHE |
3663 	    SAVESTART | MPSAFE | AUDITVNODE2, pathseg, new, newfd, CAP_CREATE,
3664 	    td);
3665 	if (fromnd.ni_vp->v_type == VDIR)
3666 		tond.ni_cnd.cn_flags |= WILLBEDIR;
3667 	if ((error = namei(&tond)) != 0) {
3668 		/* Translate error code for rename("dir1", "dir2/."). */
3669 		if (error == EISDIR && fvp->v_type == VDIR)
3670 			error = EINVAL;
3671 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3672 		vrele(fromnd.ni_dvp);
3673 		vrele(fvp);
3674 		vn_finished_write(mp);
3675 		goto out1;
3676 	}
3677 	tvfslocked = NDHASGIANT(&tond);
3678 	tdvp = tond.ni_dvp;
3679 	tvp = tond.ni_vp;
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 	}
3689 	if (fvp == tdvp) {
3690 		error = EINVAL;
3691 		goto out;
3692 	}
3693 	/*
3694 	 * If the source is the same as the destination (that is, if they
3695 	 * are links to the same vnode), then there is nothing to do.
3696 	 */
3697 	if (fvp == tvp)
3698 		error = -1;
3699 #ifdef MAC
3700 	else
3701 		error = mac_vnode_check_rename_to(td->td_ucred, tdvp,
3702 		    tond.ni_vp, fromnd.ni_dvp == tdvp, &tond.ni_cnd);
3703 #endif
3704 out:
3705 	if (!error) {
3706 		error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
3707 				   tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
3708 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3709 		NDFREE(&tond, NDF_ONLY_PNBUF);
3710 	} else {
3711 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3712 		NDFREE(&tond, NDF_ONLY_PNBUF);
3713 		if (tvp)
3714 			vput(tvp);
3715 		if (tdvp == tvp)
3716 			vrele(tdvp);
3717 		else
3718 			vput(tdvp);
3719 		vrele(fromnd.ni_dvp);
3720 		vrele(fvp);
3721 	}
3722 	vrele(tond.ni_startdir);
3723 	vn_finished_write(mp);
3724 out1:
3725 	if (fromnd.ni_startdir)
3726 		vrele(fromnd.ni_startdir);
3727 	VFS_UNLOCK_GIANT(fvfslocked);
3728 	VFS_UNLOCK_GIANT(tvfslocked);
3729 	if (error == -1)
3730 		return (0);
3731 	return (error);
3732 }
3733 
3734 /*
3735  * Make a directory file.
3736  */
3737 #ifndef _SYS_SYSPROTO_H_
3738 struct mkdir_args {
3739 	char	*path;
3740 	int	mode;
3741 };
3742 #endif
3743 int
3744 sys_mkdir(td, uap)
3745 	struct thread *td;
3746 	register struct mkdir_args /* {
3747 		char *path;
3748 		int mode;
3749 	} */ *uap;
3750 {
3751 
3752 	return (kern_mkdir(td, uap->path, UIO_USERSPACE, uap->mode));
3753 }
3754 
3755 #ifndef _SYS_SYSPROTO_H_
3756 struct mkdirat_args {
3757 	int	fd;
3758 	char	*path;
3759 	mode_t	mode;
3760 };
3761 #endif
3762 int
3763 sys_mkdirat(struct thread *td, struct mkdirat_args *uap)
3764 {
3765 
3766 	return (kern_mkdirat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode));
3767 }
3768 
3769 int
3770 kern_mkdir(struct thread *td, char *path, enum uio_seg segflg, int mode)
3771 {
3772 
3773 	return (kern_mkdirat(td, AT_FDCWD, path, segflg, mode));
3774 }
3775 
3776 int
3777 kern_mkdirat(struct thread *td, int fd, char *path, enum uio_seg segflg,
3778     int mode)
3779 {
3780 	struct mount *mp;
3781 	struct vnode *vp;
3782 	struct vattr vattr;
3783 	int error;
3784 	struct nameidata nd;
3785 	int vfslocked;
3786 
3787 	AUDIT_ARG_MODE(mode);
3788 restart:
3789 	bwillwrite();
3790 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | MPSAFE |
3791 	    AUDITVNODE1, segflg, path, fd, CAP_MKDIR, td);
3792 	nd.ni_cnd.cn_flags |= WILLBEDIR;
3793 	if ((error = namei(&nd)) != 0)
3794 		return (error);
3795 	vfslocked = NDHASGIANT(&nd);
3796 	vp = nd.ni_vp;
3797 	if (vp != NULL) {
3798 		NDFREE(&nd, NDF_ONLY_PNBUF);
3799 		/*
3800 		 * XXX namei called with LOCKPARENT but not LOCKLEAF has
3801 		 * the strange behaviour of leaving the vnode unlocked
3802 		 * if the target is the same vnode as the parent.
3803 		 */
3804 		if (vp == nd.ni_dvp)
3805 			vrele(nd.ni_dvp);
3806 		else
3807 			vput(nd.ni_dvp);
3808 		vrele(vp);
3809 		VFS_UNLOCK_GIANT(vfslocked);
3810 		return (EEXIST);
3811 	}
3812 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3813 		NDFREE(&nd, NDF_ONLY_PNBUF);
3814 		vput(nd.ni_dvp);
3815 		VFS_UNLOCK_GIANT(vfslocked);
3816 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3817 			return (error);
3818 		goto restart;
3819 	}
3820 	VATTR_NULL(&vattr);
3821 	vattr.va_type = VDIR;
3822 	vattr.va_mode = (mode & ACCESSPERMS) &~ td->td_proc->p_fd->fd_cmask;
3823 #ifdef MAC
3824 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
3825 	    &vattr);
3826 	if (error)
3827 		goto out;
3828 #endif
3829 	error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
3830 #ifdef MAC
3831 out:
3832 #endif
3833 	NDFREE(&nd, NDF_ONLY_PNBUF);
3834 	vput(nd.ni_dvp);
3835 	if (!error)
3836 		vput(nd.ni_vp);
3837 	vn_finished_write(mp);
3838 	VFS_UNLOCK_GIANT(vfslocked);
3839 	return (error);
3840 }
3841 
3842 /*
3843  * Remove a directory file.
3844  */
3845 #ifndef _SYS_SYSPROTO_H_
3846 struct rmdir_args {
3847 	char	*path;
3848 };
3849 #endif
3850 int
3851 sys_rmdir(td, uap)
3852 	struct thread *td;
3853 	struct rmdir_args /* {
3854 		char *path;
3855 	} */ *uap;
3856 {
3857 
3858 	return (kern_rmdir(td, uap->path, UIO_USERSPACE));
3859 }
3860 
3861 int
3862 kern_rmdir(struct thread *td, char *path, enum uio_seg pathseg)
3863 {
3864 
3865 	return (kern_rmdirat(td, AT_FDCWD, path, pathseg));
3866 }
3867 
3868 int
3869 kern_rmdirat(struct thread *td, int fd, char *path, enum uio_seg pathseg)
3870 {
3871 	struct mount *mp;
3872 	struct vnode *vp;
3873 	int error;
3874 	struct nameidata nd;
3875 	int vfslocked;
3876 
3877 restart:
3878 	bwillwrite();
3879 	NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | MPSAFE |
3880 	    AUDITVNODE1, pathseg, path, fd, CAP_RMDIR, td);
3881 	if ((error = namei(&nd)) != 0)
3882 		return (error);
3883 	vfslocked = NDHASGIANT(&nd);
3884 	vp = nd.ni_vp;
3885 	if (vp->v_type != VDIR) {
3886 		error = ENOTDIR;
3887 		goto out;
3888 	}
3889 	/*
3890 	 * No rmdir "." please.
3891 	 */
3892 	if (nd.ni_dvp == vp) {
3893 		error = EINVAL;
3894 		goto out;
3895 	}
3896 	/*
3897 	 * The root of a mounted filesystem cannot be deleted.
3898 	 */
3899 	if (vp->v_vflag & VV_ROOT) {
3900 		error = EBUSY;
3901 		goto out;
3902 	}
3903 #ifdef MAC
3904 	error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
3905 	    &nd.ni_cnd);
3906 	if (error)
3907 		goto out;
3908 #endif
3909 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3910 		NDFREE(&nd, NDF_ONLY_PNBUF);
3911 		vput(vp);
3912 		if (nd.ni_dvp == vp)
3913 			vrele(nd.ni_dvp);
3914 		else
3915 			vput(nd.ni_dvp);
3916 		VFS_UNLOCK_GIANT(vfslocked);
3917 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3918 			return (error);
3919 		goto restart;
3920 	}
3921 	error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
3922 	vn_finished_write(mp);
3923 out:
3924 	NDFREE(&nd, NDF_ONLY_PNBUF);
3925 	vput(vp);
3926 	if (nd.ni_dvp == vp)
3927 		vrele(nd.ni_dvp);
3928 	else
3929 		vput(nd.ni_dvp);
3930 	VFS_UNLOCK_GIANT(vfslocked);
3931 	return (error);
3932 }
3933 
3934 #ifdef COMPAT_43
3935 /*
3936  * Read a block of directory entries in a filesystem independent format.
3937  */
3938 #ifndef _SYS_SYSPROTO_H_
3939 struct ogetdirentries_args {
3940 	int	fd;
3941 	char	*buf;
3942 	u_int	count;
3943 	long	*basep;
3944 };
3945 #endif
3946 int
3947 ogetdirentries(struct thread *td, struct ogetdirentries_args *uap)
3948 {
3949 	long loff;
3950 	int error;
3951 
3952 	error = kern_ogetdirentries(td, uap, &loff);
3953 	if (error == 0)
3954 		error = copyout(&loff, uap->basep, sizeof(long));
3955 	return (error);
3956 }
3957 
3958 int
3959 kern_ogetdirentries(struct thread *td, struct ogetdirentries_args *uap,
3960     long *ploff)
3961 {
3962 	struct vnode *vp;
3963 	struct file *fp;
3964 	struct uio auio, kuio;
3965 	struct iovec aiov, kiov;
3966 	struct dirent *dp, *edp;
3967 	caddr_t dirbuf;
3968 	int error, eofflag, readcnt, vfslocked;
3969 	long loff;
3970 	off_t foffset;
3971 
3972 	/* XXX arbitrary sanity limit on `count'. */
3973 	if (uap->count > 64 * 1024)
3974 		return (EINVAL);
3975 	if ((error = getvnode(td->td_proc->p_fd, uap->fd, CAP_READ,
3976 	    &fp)) != 0)
3977 		return (error);
3978 	if ((fp->f_flag & FREAD) == 0) {
3979 		fdrop(fp, td);
3980 		return (EBADF);
3981 	}
3982 	vp = fp->f_vnode;
3983 	foffset = foffset_lock(fp, 0);
3984 unionread:
3985 	vfslocked = VFS_LOCK_GIANT(vp->v_mount);
3986 	if (vp->v_type != VDIR) {
3987 		VFS_UNLOCK_GIANT(vfslocked);
3988 		foffset_unlock(fp, foffset, 0);
3989 		fdrop(fp, td);
3990 		return (EINVAL);
3991 	}
3992 	aiov.iov_base = uap->buf;
3993 	aiov.iov_len = uap->count;
3994 	auio.uio_iov = &aiov;
3995 	auio.uio_iovcnt = 1;
3996 	auio.uio_rw = UIO_READ;
3997 	auio.uio_segflg = UIO_USERSPACE;
3998 	auio.uio_td = td;
3999 	auio.uio_resid = uap->count;
4000 	vn_lock(vp, LK_SHARED | LK_RETRY);
4001 	loff = auio.uio_offset = foffset;
4002 #ifdef MAC
4003 	error = mac_vnode_check_readdir(td->td_ucred, vp);
4004 	if (error) {
4005 		VOP_UNLOCK(vp, 0);
4006 		VFS_UNLOCK_GIANT(vfslocked);
4007 		foffset_unlock(fp, foffset, FOF_NOUPDATE);
4008 		fdrop(fp, td);
4009 		return (error);
4010 	}
4011 #endif
4012 #	if (BYTE_ORDER != LITTLE_ENDIAN)
4013 		if (vp->v_mount->mnt_maxsymlinklen <= 0) {
4014 			error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag,
4015 			    NULL, NULL);
4016 			foffset = auio.uio_offset;
4017 		} else
4018 #	endif
4019 	{
4020 		kuio = auio;
4021 		kuio.uio_iov = &kiov;
4022 		kuio.uio_segflg = UIO_SYSSPACE;
4023 		kiov.iov_len = uap->count;
4024 		dirbuf = malloc(uap->count, M_TEMP, M_WAITOK);
4025 		kiov.iov_base = dirbuf;
4026 		error = VOP_READDIR(vp, &kuio, fp->f_cred, &eofflag,
4027 			    NULL, NULL);
4028 		foffset = kuio.uio_offset;
4029 		if (error == 0) {
4030 			readcnt = uap->count - kuio.uio_resid;
4031 			edp = (struct dirent *)&dirbuf[readcnt];
4032 			for (dp = (struct dirent *)dirbuf; dp < edp; ) {
4033 #				if (BYTE_ORDER == LITTLE_ENDIAN)
4034 					/*
4035 					 * The expected low byte of
4036 					 * dp->d_namlen is our dp->d_type.
4037 					 * The high MBZ byte of dp->d_namlen
4038 					 * is our dp->d_namlen.
4039 					 */
4040 					dp->d_type = dp->d_namlen;
4041 					dp->d_namlen = 0;
4042 #				else
4043 					/*
4044 					 * The dp->d_type is the high byte
4045 					 * of the expected dp->d_namlen,
4046 					 * so must be zero'ed.
4047 					 */
4048 					dp->d_type = 0;
4049 #				endif
4050 				if (dp->d_reclen > 0) {
4051 					dp = (struct dirent *)
4052 					    ((char *)dp + dp->d_reclen);
4053 				} else {
4054 					error = EIO;
4055 					break;
4056 				}
4057 			}
4058 			if (dp >= edp)
4059 				error = uiomove(dirbuf, readcnt, &auio);
4060 		}
4061 		free(dirbuf, M_TEMP);
4062 	}
4063 	if (error) {
4064 		VOP_UNLOCK(vp, 0);
4065 		VFS_UNLOCK_GIANT(vfslocked);
4066 		foffset_unlock(fp, foffset, 0);
4067 		fdrop(fp, td);
4068 		return (error);
4069 	}
4070 	if (uap->count == auio.uio_resid &&
4071 	    (vp->v_vflag & VV_ROOT) &&
4072 	    (vp->v_mount->mnt_flag & MNT_UNION)) {
4073 		struct vnode *tvp = vp;
4074 		vp = vp->v_mount->mnt_vnodecovered;
4075 		VREF(vp);
4076 		fp->f_vnode = vp;
4077 		fp->f_data = vp;
4078 		foffset = 0;
4079 		vput(tvp);
4080 		VFS_UNLOCK_GIANT(vfslocked);
4081 		goto unionread;
4082 	}
4083 	VOP_UNLOCK(vp, 0);
4084 	VFS_UNLOCK_GIANT(vfslocked);
4085 	foffset_unlock(fp, foffset, 0);
4086 	fdrop(fp, td);
4087 	td->td_retval[0] = uap->count - auio.uio_resid;
4088 	if (error == 0)
4089 		*ploff = loff;
4090 	return (error);
4091 }
4092 #endif /* COMPAT_43 */
4093 
4094 /*
4095  * Read a block of directory entries in a filesystem independent format.
4096  */
4097 #ifndef _SYS_SYSPROTO_H_
4098 struct getdirentries_args {
4099 	int	fd;
4100 	char	*buf;
4101 	u_int	count;
4102 	long	*basep;
4103 };
4104 #endif
4105 int
4106 sys_getdirentries(td, uap)
4107 	struct thread *td;
4108 	register struct getdirentries_args /* {
4109 		int fd;
4110 		char *buf;
4111 		u_int count;
4112 		long *basep;
4113 	} */ *uap;
4114 {
4115 	long base;
4116 	int error;
4117 
4118 	error = kern_getdirentries(td, uap->fd, uap->buf, uap->count, &base,
4119 	    NULL, UIO_USERSPACE);
4120 	if (error)
4121 		return (error);
4122 	if (uap->basep != NULL)
4123 		error = copyout(&base, uap->basep, sizeof(long));
4124 	return (error);
4125 }
4126 
4127 int
4128 kern_getdirentries(struct thread *td, int fd, char *buf, u_int count,
4129     long *basep, ssize_t *residp, enum uio_seg bufseg)
4130 {
4131 	struct vnode *vp;
4132 	struct file *fp;
4133 	struct uio auio;
4134 	struct iovec aiov;
4135 	int vfslocked;
4136 	long loff;
4137 	int error, eofflag;
4138 	off_t foffset;
4139 
4140 	AUDIT_ARG_FD(fd);
4141 	if (count > IOSIZE_MAX)
4142 		return (EINVAL);
4143 	auio.uio_resid = count;
4144 	if ((error = getvnode(td->td_proc->p_fd, fd, CAP_READ | CAP_SEEK,
4145 	    &fp)) != 0)
4146 		return (error);
4147 	if ((fp->f_flag & FREAD) == 0) {
4148 		fdrop(fp, td);
4149 		return (EBADF);
4150 	}
4151 	vp = fp->f_vnode;
4152 	foffset = foffset_lock(fp, 0);
4153 unionread:
4154 	vfslocked = VFS_LOCK_GIANT(vp->v_mount);
4155 	if (vp->v_type != VDIR) {
4156 		VFS_UNLOCK_GIANT(vfslocked);
4157 		error = EINVAL;
4158 		goto fail;
4159 	}
4160 	aiov.iov_base = buf;
4161 	aiov.iov_len = count;
4162 	auio.uio_iov = &aiov;
4163 	auio.uio_iovcnt = 1;
4164 	auio.uio_rw = UIO_READ;
4165 	auio.uio_segflg = bufseg;
4166 	auio.uio_td = td;
4167 	vn_lock(vp, LK_SHARED | LK_RETRY);
4168 	AUDIT_ARG_VNODE1(vp);
4169 	loff = auio.uio_offset = foffset;
4170 #ifdef MAC
4171 	error = mac_vnode_check_readdir(td->td_ucred, vp);
4172 	if (error == 0)
4173 #endif
4174 		error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, NULL,
4175 		    NULL);
4176 	foffset = auio.uio_offset;
4177 	if (error) {
4178 		VOP_UNLOCK(vp, 0);
4179 		VFS_UNLOCK_GIANT(vfslocked);
4180 		goto fail;
4181 	}
4182 	if (count == auio.uio_resid &&
4183 	    (vp->v_vflag & VV_ROOT) &&
4184 	    (vp->v_mount->mnt_flag & MNT_UNION)) {
4185 		struct vnode *tvp = vp;
4186 		vp = vp->v_mount->mnt_vnodecovered;
4187 		VREF(vp);
4188 		fp->f_vnode = vp;
4189 		fp->f_data = vp;
4190 		foffset = 0;
4191 		vput(tvp);
4192 		VFS_UNLOCK_GIANT(vfslocked);
4193 		goto unionread;
4194 	}
4195 	VOP_UNLOCK(vp, 0);
4196 	VFS_UNLOCK_GIANT(vfslocked);
4197 	*basep = loff;
4198 	if (residp != NULL)
4199 		*residp = auio.uio_resid;
4200 	td->td_retval[0] = count - auio.uio_resid;
4201 fail:
4202 	foffset_unlock(fp, foffset, 0);
4203 	fdrop(fp, td);
4204 	return (error);
4205 }
4206 
4207 #ifndef _SYS_SYSPROTO_H_
4208 struct getdents_args {
4209 	int fd;
4210 	char *buf;
4211 	size_t count;
4212 };
4213 #endif
4214 int
4215 sys_getdents(td, uap)
4216 	struct thread *td;
4217 	register struct getdents_args /* {
4218 		int fd;
4219 		char *buf;
4220 		u_int count;
4221 	} */ *uap;
4222 {
4223 	struct getdirentries_args ap;
4224 	ap.fd = uap->fd;
4225 	ap.buf = uap->buf;
4226 	ap.count = uap->count;
4227 	ap.basep = NULL;
4228 	return (sys_getdirentries(td, &ap));
4229 }
4230 
4231 /*
4232  * Set the mode mask for creation of filesystem nodes.
4233  */
4234 #ifndef _SYS_SYSPROTO_H_
4235 struct umask_args {
4236 	int	newmask;
4237 };
4238 #endif
4239 int
4240 sys_umask(td, uap)
4241 	struct thread *td;
4242 	struct umask_args /* {
4243 		int newmask;
4244 	} */ *uap;
4245 {
4246 	register struct filedesc *fdp;
4247 
4248 	FILEDESC_XLOCK(td->td_proc->p_fd);
4249 	fdp = td->td_proc->p_fd;
4250 	td->td_retval[0] = fdp->fd_cmask;
4251 	fdp->fd_cmask = uap->newmask & ALLPERMS;
4252 	FILEDESC_XUNLOCK(td->td_proc->p_fd);
4253 	return (0);
4254 }
4255 
4256 /*
4257  * Void all references to file by ripping underlying filesystem away from
4258  * vnode.
4259  */
4260 #ifndef _SYS_SYSPROTO_H_
4261 struct revoke_args {
4262 	char	*path;
4263 };
4264 #endif
4265 int
4266 sys_revoke(td, uap)
4267 	struct thread *td;
4268 	register struct revoke_args /* {
4269 		char *path;
4270 	} */ *uap;
4271 {
4272 	struct vnode *vp;
4273 	struct vattr vattr;
4274 	int error;
4275 	struct nameidata nd;
4276 	int vfslocked;
4277 
4278 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | MPSAFE | AUDITVNODE1,
4279 	    UIO_USERSPACE, uap->path, td);
4280 	if ((error = namei(&nd)) != 0)
4281 		return (error);
4282 	vfslocked = NDHASGIANT(&nd);
4283 	vp = nd.ni_vp;
4284 	NDFREE(&nd, NDF_ONLY_PNBUF);
4285 	if (vp->v_type != VCHR || vp->v_rdev == NULL) {
4286 		error = EINVAL;
4287 		goto out;
4288 	}
4289 #ifdef MAC
4290 	error = mac_vnode_check_revoke(td->td_ucred, vp);
4291 	if (error)
4292 		goto out;
4293 #endif
4294 	error = VOP_GETATTR(vp, &vattr, td->td_ucred);
4295 	if (error)
4296 		goto out;
4297 	if (td->td_ucred->cr_uid != vattr.va_uid) {
4298 		error = priv_check(td, PRIV_VFS_ADMIN);
4299 		if (error)
4300 			goto out;
4301 	}
4302 	if (vcount(vp) > 1)
4303 		VOP_REVOKE(vp, REVOKEALL);
4304 out:
4305 	vput(vp);
4306 	VFS_UNLOCK_GIANT(vfslocked);
4307 	return (error);
4308 }
4309 
4310 /*
4311  * Convert a user file descriptor to a kernel file entry and check that, if it
4312  * is a capability, the correct rights are present. A reference on the file
4313  * entry is held upon returning.
4314  */
4315 int
4316 getvnode(struct filedesc *fdp, int fd, cap_rights_t rights,
4317     struct file **fpp)
4318 {
4319 	struct file *fp;
4320 #ifdef CAPABILITIES
4321 	struct file *fp_fromcap;
4322 	int error;
4323 #endif
4324 
4325 	if (fdp == NULL || (fp = fget_unlocked(fdp, fd)) == NULL)
4326 		return (EBADF);
4327 #ifdef CAPABILITIES
4328 	/*
4329 	 * If the file descriptor is for a capability, test rights and use the
4330 	 * file descriptor referenced by the capability.
4331 	 */
4332 	error = cap_funwrap(fp, rights, &fp_fromcap);
4333 	if (error) {
4334 		fdrop(fp, curthread);
4335 		return (error);
4336 	}
4337 	if (fp != fp_fromcap) {
4338 		fhold(fp_fromcap);
4339 		fdrop(fp, curthread);
4340 		fp = fp_fromcap;
4341 	}
4342 #endif /* CAPABILITIES */
4343 
4344 	/*
4345 	 * The file could be not of the vnode type, or it may be not
4346 	 * yet fully initialized, in which case the f_vnode pointer
4347 	 * may be set, but f_ops is still badfileops.  E.g.,
4348 	 * devfs_open() transiently create such situation to
4349 	 * facilitate csw d_fdopen().
4350 	 *
4351 	 * Dupfdopen() handling in kern_openat() installs the
4352 	 * half-baked file into the process descriptor table, allowing
4353 	 * other thread to dereference it. Guard against the race by
4354 	 * checking f_ops.
4355 	 */
4356 	if (fp->f_vnode == NULL || fp->f_ops == &badfileops) {
4357 		fdrop(fp, curthread);
4358 		return (EINVAL);
4359 	}
4360 	*fpp = fp;
4361 	return (0);
4362 }
4363 
4364 
4365 /*
4366  * Get an (NFS) file handle.
4367  */
4368 #ifndef _SYS_SYSPROTO_H_
4369 struct lgetfh_args {
4370 	char	*fname;
4371 	fhandle_t *fhp;
4372 };
4373 #endif
4374 int
4375 sys_lgetfh(td, uap)
4376 	struct thread *td;
4377 	register struct lgetfh_args *uap;
4378 {
4379 	struct nameidata nd;
4380 	fhandle_t fh;
4381 	register struct vnode *vp;
4382 	int vfslocked;
4383 	int error;
4384 
4385 	error = priv_check(td, PRIV_VFS_GETFH);
4386 	if (error)
4387 		return (error);
4388 	NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | MPSAFE | AUDITVNODE1,
4389 	    UIO_USERSPACE, uap->fname, td);
4390 	error = namei(&nd);
4391 	if (error)
4392 		return (error);
4393 	vfslocked = NDHASGIANT(&nd);
4394 	NDFREE(&nd, NDF_ONLY_PNBUF);
4395 	vp = nd.ni_vp;
4396 	bzero(&fh, sizeof(fh));
4397 	fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
4398 	error = VOP_VPTOFH(vp, &fh.fh_fid);
4399 	vput(vp);
4400 	VFS_UNLOCK_GIANT(vfslocked);
4401 	if (error)
4402 		return (error);
4403 	error = copyout(&fh, uap->fhp, sizeof (fh));
4404 	return (error);
4405 }
4406 
4407 #ifndef _SYS_SYSPROTO_H_
4408 struct getfh_args {
4409 	char	*fname;
4410 	fhandle_t *fhp;
4411 };
4412 #endif
4413 int
4414 sys_getfh(td, uap)
4415 	struct thread *td;
4416 	register struct getfh_args *uap;
4417 {
4418 	struct nameidata nd;
4419 	fhandle_t fh;
4420 	register struct vnode *vp;
4421 	int vfslocked;
4422 	int error;
4423 
4424 	error = priv_check(td, PRIV_VFS_GETFH);
4425 	if (error)
4426 		return (error);
4427 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | MPSAFE | AUDITVNODE1,
4428 	    UIO_USERSPACE, uap->fname, td);
4429 	error = namei(&nd);
4430 	if (error)
4431 		return (error);
4432 	vfslocked = NDHASGIANT(&nd);
4433 	NDFREE(&nd, NDF_ONLY_PNBUF);
4434 	vp = nd.ni_vp;
4435 	bzero(&fh, sizeof(fh));
4436 	fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
4437 	error = VOP_VPTOFH(vp, &fh.fh_fid);
4438 	vput(vp);
4439 	VFS_UNLOCK_GIANT(vfslocked);
4440 	if (error)
4441 		return (error);
4442 	error = copyout(&fh, uap->fhp, sizeof (fh));
4443 	return (error);
4444 }
4445 
4446 /*
4447  * syscall for the rpc.lockd to use to translate a NFS file handle into an
4448  * open descriptor.
4449  *
4450  * warning: do not remove the priv_check() call or this becomes one giant
4451  * security hole.
4452  */
4453 #ifndef _SYS_SYSPROTO_H_
4454 struct fhopen_args {
4455 	const struct fhandle *u_fhp;
4456 	int flags;
4457 };
4458 #endif
4459 int
4460 sys_fhopen(td, uap)
4461 	struct thread *td;
4462 	struct fhopen_args /* {
4463 		const struct fhandle *u_fhp;
4464 		int flags;
4465 	} */ *uap;
4466 {
4467 	struct mount *mp;
4468 	struct vnode *vp;
4469 	struct fhandle fhp;
4470 	struct file *fp;
4471 	int fmode, error;
4472 	int vfslocked;
4473 	int indx;
4474 
4475 	error = priv_check(td, PRIV_VFS_FHOPEN);
4476 	if (error)
4477 		return (error);
4478 	indx = -1;
4479 	fmode = FFLAGS(uap->flags);
4480 	/* why not allow a non-read/write open for our lockd? */
4481 	if (((fmode & (FREAD | FWRITE)) == 0) || (fmode & O_CREAT))
4482 		return (EINVAL);
4483 	error = copyin(uap->u_fhp, &fhp, sizeof(fhp));
4484 	if (error)
4485 		return(error);
4486 	/* find the mount point */
4487 	mp = vfs_busyfs(&fhp.fh_fsid);
4488 	if (mp == NULL)
4489 		return (ESTALE);
4490 	vfslocked = VFS_LOCK_GIANT(mp);
4491 	/* now give me my vnode, it gets returned to me locked */
4492 	error = VFS_FHTOVP(mp, &fhp.fh_fid, LK_EXCLUSIVE, &vp);
4493 	vfs_unbusy(mp);
4494 	if (error) {
4495 		VFS_UNLOCK_GIANT(vfslocked);
4496 		return (error);
4497 	}
4498 
4499 	error = falloc_noinstall(td, &fp);
4500 	if (error) {
4501 		vput(vp);
4502 		VFS_UNLOCK_GIANT(vfslocked);
4503 		return (error);
4504 	}
4505 	/*
4506 	 * An extra reference on `fp' has been held for us by
4507 	 * falloc_noinstall().
4508 	 */
4509 
4510 #ifdef INVARIANTS
4511 	td->td_dupfd = -1;
4512 #endif
4513 	error = vn_open_vnode(vp, fmode, td->td_ucred, td, fp);
4514 	if (error) {
4515 		KASSERT(fp->f_ops == &badfileops,
4516 		    ("VOP_OPEN in fhopen() set f_ops"));
4517 		KASSERT(td->td_dupfd < 0,
4518 		    ("fhopen() encountered fdopen()"));
4519 
4520 		vput(vp);
4521 		goto bad;
4522 	}
4523 #ifdef INVARIANTS
4524 	td->td_dupfd = 0;
4525 #endif
4526 	fp->f_vnode = vp;
4527 	fp->f_seqcount = 1;
4528 	finit(fp, (fmode & FMASK) | (fp->f_flag & FHASLOCK), DTYPE_VNODE, vp,
4529 	    &vnops);
4530 	VOP_UNLOCK(vp, 0);
4531 	if (fmode & O_TRUNC) {
4532 		error = fo_truncate(fp, 0, td->td_ucred, td);
4533 		if (error)
4534 			goto bad;
4535 	}
4536 
4537 	error = finstall(td, fp, &indx, fmode);
4538 bad:
4539 	VFS_UNLOCK_GIANT(vfslocked);
4540 	fdrop(fp, td);
4541 	td->td_retval[0] = indx;
4542 	return (error);
4543 }
4544 
4545 /*
4546  * Stat an (NFS) file handle.
4547  */
4548 #ifndef _SYS_SYSPROTO_H_
4549 struct fhstat_args {
4550 	struct fhandle *u_fhp;
4551 	struct stat *sb;
4552 };
4553 #endif
4554 int
4555 sys_fhstat(td, uap)
4556 	struct thread *td;
4557 	register struct fhstat_args /* {
4558 		struct fhandle *u_fhp;
4559 		struct stat *sb;
4560 	} */ *uap;
4561 {
4562 	struct stat sb;
4563 	struct fhandle fh;
4564 	int error;
4565 
4566 	error = copyin(uap->u_fhp, &fh, sizeof(fh));
4567 	if (error != 0)
4568 		return (error);
4569 	error = kern_fhstat(td, fh, &sb);
4570 	if (error != 0)
4571 		return (error);
4572 	error = copyout(&sb, uap->sb, sizeof(sb));
4573 	return (error);
4574 }
4575 
4576 int
4577 kern_fhstat(struct thread *td, struct fhandle fh, struct stat *sb)
4578 {
4579 	struct mount *mp;
4580 	struct vnode *vp;
4581 	int vfslocked;
4582 	int error;
4583 
4584 	error = priv_check(td, PRIV_VFS_FHSTAT);
4585 	if (error)
4586 		return (error);
4587 	if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4588 		return (ESTALE);
4589 	vfslocked = VFS_LOCK_GIANT(mp);
4590 	error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4591 	vfs_unbusy(mp);
4592 	if (error) {
4593 		VFS_UNLOCK_GIANT(vfslocked);
4594 		return (error);
4595 	}
4596 	error = vn_stat(vp, sb, td->td_ucred, NOCRED, td);
4597 	vput(vp);
4598 	VFS_UNLOCK_GIANT(vfslocked);
4599 	return (error);
4600 }
4601 
4602 /*
4603  * Implement fstatfs() for (NFS) file handles.
4604  */
4605 #ifndef _SYS_SYSPROTO_H_
4606 struct fhstatfs_args {
4607 	struct fhandle *u_fhp;
4608 	struct statfs *buf;
4609 };
4610 #endif
4611 int
4612 sys_fhstatfs(td, uap)
4613 	struct thread *td;
4614 	struct fhstatfs_args /* {
4615 		struct fhandle *u_fhp;
4616 		struct statfs *buf;
4617 	} */ *uap;
4618 {
4619 	struct statfs sf;
4620 	fhandle_t fh;
4621 	int error;
4622 
4623 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
4624 	if (error)
4625 		return (error);
4626 	error = kern_fhstatfs(td, fh, &sf);
4627 	if (error)
4628 		return (error);
4629 	return (copyout(&sf, uap->buf, sizeof(sf)));
4630 }
4631 
4632 int
4633 kern_fhstatfs(struct thread *td, fhandle_t fh, struct statfs *buf)
4634 {
4635 	struct statfs *sp;
4636 	struct mount *mp;
4637 	struct vnode *vp;
4638 	int vfslocked;
4639 	int error;
4640 
4641 	error = priv_check(td, PRIV_VFS_FHSTATFS);
4642 	if (error)
4643 		return (error);
4644 	if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4645 		return (ESTALE);
4646 	vfslocked = VFS_LOCK_GIANT(mp);
4647 	error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4648 	if (error) {
4649 		vfs_unbusy(mp);
4650 		VFS_UNLOCK_GIANT(vfslocked);
4651 		return (error);
4652 	}
4653 	vput(vp);
4654 	error = prison_canseemount(td->td_ucred, mp);
4655 	if (error)
4656 		goto out;
4657 #ifdef MAC
4658 	error = mac_mount_check_stat(td->td_ucred, mp);
4659 	if (error)
4660 		goto out;
4661 #endif
4662 	/*
4663 	 * Set these in case the underlying filesystem fails to do so.
4664 	 */
4665 	sp = &mp->mnt_stat;
4666 	sp->f_version = STATFS_VERSION;
4667 	sp->f_namemax = NAME_MAX;
4668 	sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
4669 	error = VFS_STATFS(mp, sp);
4670 	if (error == 0)
4671 		*buf = *sp;
4672 out:
4673 	vfs_unbusy(mp);
4674 	VFS_UNLOCK_GIANT(vfslocked);
4675 	return (error);
4676 }
4677 
4678 int
4679 kern_posix_fallocate(struct thread *td, int fd, off_t offset, off_t len)
4680 {
4681 	struct file *fp;
4682 	struct mount *mp;
4683 	struct vnode *vp;
4684 	off_t olen, ooffset;
4685 	int error, vfslocked;
4686 
4687 	fp = NULL;
4688 	vfslocked = 0;
4689 	error = fget(td, fd, CAP_WRITE, &fp);
4690 	if (error != 0)
4691 		goto out;
4692 
4693 	switch (fp->f_type) {
4694 	case DTYPE_VNODE:
4695 		break;
4696 	case DTYPE_PIPE:
4697 	case DTYPE_FIFO:
4698 		error = ESPIPE;
4699 		goto out;
4700 	default:
4701 		error = ENODEV;
4702 		goto out;
4703 	}
4704 	if ((fp->f_flag & FWRITE) == 0) {
4705 		error = EBADF;
4706 		goto out;
4707 	}
4708 	vp = fp->f_vnode;
4709 	if (vp->v_type != VREG) {
4710 		error = ENODEV;
4711 		goto out;
4712 	}
4713 	if (offset < 0 || len <= 0) {
4714 		error = EINVAL;
4715 		goto out;
4716 	}
4717 	/* Check for wrap. */
4718 	if (offset > OFF_MAX - len) {
4719 		error = EFBIG;
4720 		goto out;
4721 	}
4722 
4723 	/* Allocating blocks may take a long time, so iterate. */
4724 	for (;;) {
4725 		olen = len;
4726 		ooffset = offset;
4727 
4728 		bwillwrite();
4729 		vfslocked = VFS_LOCK_GIANT(vp->v_mount);
4730 		mp = NULL;
4731 		error = vn_start_write(vp, &mp, V_WAIT | PCATCH);
4732 		if (error != 0) {
4733 			VFS_UNLOCK_GIANT(vfslocked);
4734 			break;
4735 		}
4736 		error = vn_lock(vp, LK_EXCLUSIVE);
4737 		if (error != 0) {
4738 			vn_finished_write(mp);
4739 			VFS_UNLOCK_GIANT(vfslocked);
4740 			break;
4741 		}
4742 #ifdef MAC
4743 		error = mac_vnode_check_write(td->td_ucred, fp->f_cred, vp);
4744 		if (error == 0)
4745 #endif
4746 			error = VOP_ALLOCATE(vp, &offset, &len);
4747 		VOP_UNLOCK(vp, 0);
4748 		vn_finished_write(mp);
4749 		VFS_UNLOCK_GIANT(vfslocked);
4750 
4751 		if (olen + ooffset != offset + len) {
4752 			panic("offset + len changed from %jx/%jx to %jx/%jx",
4753 			    ooffset, olen, offset, len);
4754 		}
4755 		if (error != 0 || len == 0)
4756 			break;
4757 		KASSERT(olen > len, ("Iteration did not make progress?"));
4758 		maybe_yield();
4759 	}
4760  out:
4761 	if (fp != NULL)
4762 		fdrop(fp, td);
4763 	return (error);
4764 }
4765 
4766 int
4767 sys_posix_fallocate(struct thread *td, struct posix_fallocate_args *uap)
4768 {
4769 
4770 	return (kern_posix_fallocate(td, uap->fd, uap->offset, uap->len));
4771 }
4772 
4773 /*
4774  * Unlike madvise(2), we do not make a best effort to remember every
4775  * possible caching hint.  Instead, we remember the last setting with
4776  * the exception that we will allow POSIX_FADV_NORMAL to adjust the
4777  * region of any current setting.
4778  */
4779 int
4780 kern_posix_fadvise(struct thread *td, int fd, off_t offset, off_t len,
4781     int advice)
4782 {
4783 	struct fadvise_info *fa, *new;
4784 	struct file *fp;
4785 	struct vnode *vp;
4786 	off_t end;
4787 	int error;
4788 
4789 	if (offset < 0 || len < 0 || offset > OFF_MAX - len)
4790 		return (EINVAL);
4791 	switch (advice) {
4792 	case POSIX_FADV_SEQUENTIAL:
4793 	case POSIX_FADV_RANDOM:
4794 	case POSIX_FADV_NOREUSE:
4795 		new = malloc(sizeof(*fa), M_FADVISE, M_WAITOK);
4796 		break;
4797 	case POSIX_FADV_NORMAL:
4798 	case POSIX_FADV_WILLNEED:
4799 	case POSIX_FADV_DONTNEED:
4800 		new = NULL;
4801 		break;
4802 	default:
4803 		return (EINVAL);
4804 	}
4805 	/* XXX: CAP_POSIX_FADVISE? */
4806 	error = fget(td, fd, 0, &fp);
4807 	if (error != 0)
4808 		goto out;
4809 
4810 	switch (fp->f_type) {
4811 	case DTYPE_VNODE:
4812 		break;
4813 	case DTYPE_PIPE:
4814 	case DTYPE_FIFO:
4815 		error = ESPIPE;
4816 		goto out;
4817 	default:
4818 		error = ENODEV;
4819 		goto out;
4820 	}
4821 	vp = fp->f_vnode;
4822 	if (vp->v_type != VREG) {
4823 		error = ENODEV;
4824 		goto out;
4825 	}
4826 	if (len == 0)
4827 		end = OFF_MAX;
4828 	else
4829 		end = offset + len - 1;
4830 	switch (advice) {
4831 	case POSIX_FADV_SEQUENTIAL:
4832 	case POSIX_FADV_RANDOM:
4833 	case POSIX_FADV_NOREUSE:
4834 		/*
4835 		 * Try to merge any existing non-standard region with
4836 		 * this new region if possible, otherwise create a new
4837 		 * non-standard region for this request.
4838 		 */
4839 		mtx_pool_lock(mtxpool_sleep, fp);
4840 		fa = fp->f_advice;
4841 		if (fa != NULL && fa->fa_advice == advice &&
4842 		    ((fa->fa_start <= end && fa->fa_end >= offset) ||
4843 		    (end != OFF_MAX && fa->fa_start == end + 1) ||
4844 		    (fa->fa_end != OFF_MAX && fa->fa_end + 1 == offset))) {
4845 			if (offset < fa->fa_start)
4846 				fa->fa_start = offset;
4847 			if (end > fa->fa_end)
4848 				fa->fa_end = end;
4849 		} else {
4850 			new->fa_advice = advice;
4851 			new->fa_start = offset;
4852 			new->fa_end = end;
4853 			new->fa_prevstart = 0;
4854 			new->fa_prevend = 0;
4855 			fp->f_advice = new;
4856 			new = fa;
4857 		}
4858 		mtx_pool_unlock(mtxpool_sleep, fp);
4859 		break;
4860 	case POSIX_FADV_NORMAL:
4861 		/*
4862 		 * If a the "normal" region overlaps with an existing
4863 		 * non-standard region, trim or remove the
4864 		 * non-standard region.
4865 		 */
4866 		mtx_pool_lock(mtxpool_sleep, fp);
4867 		fa = fp->f_advice;
4868 		if (fa != NULL) {
4869 			if (offset <= fa->fa_start && end >= fa->fa_end) {
4870 				new = fa;
4871 				fp->f_advice = NULL;
4872 			} else if (offset <= fa->fa_start &&
4873  			    end >= fa->fa_start)
4874 				fa->fa_start = end + 1;
4875 			else if (offset <= fa->fa_end && end >= fa->fa_end)
4876 				fa->fa_end = offset - 1;
4877 			else if (offset >= fa->fa_start && end <= fa->fa_end) {
4878 				/*
4879 				 * If the "normal" region is a middle
4880 				 * portion of the existing
4881 				 * non-standard region, just remove
4882 				 * the whole thing rather than picking
4883 				 * one side or the other to
4884 				 * preserve.
4885 				 */
4886 				new = fa;
4887 				fp->f_advice = NULL;
4888 			}
4889 		}
4890 		mtx_pool_unlock(mtxpool_sleep, fp);
4891 		break;
4892 	case POSIX_FADV_WILLNEED:
4893 	case POSIX_FADV_DONTNEED:
4894 		error = VOP_ADVISE(vp, offset, end, advice);
4895 		break;
4896 	}
4897 out:
4898 	if (fp != NULL)
4899 		fdrop(fp, td);
4900 	free(new, M_FADVISE);
4901 	return (error);
4902 }
4903 
4904 int
4905 sys_posix_fadvise(struct thread *td, struct posix_fadvise_args *uap)
4906 {
4907 
4908 	return (kern_posix_fadvise(td, uap->fd, uap->offset, uap->len,
4909 	    uap->advice));
4910 }
4911