xref: /freebsd/sys/kern/vfs_syscalls.c (revision 3e65b9c6e6b7b2081d54e1dc40983c3c00eaf738)
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 static MALLOC_DEFINE(M_FADVISE, "fadvise", "posix_fadvise(2) information");
90 
91 SDT_PROVIDER_DEFINE(vfs);
92 SDT_PROBE_DEFINE(vfs, , stat, mode, mode);
93 SDT_PROBE_ARGTYPE(vfs, , stat, mode, 0, "char *");
94 SDT_PROBE_ARGTYPE(vfs, , stat, mode, 1, "int");
95 SDT_PROBE_DEFINE(vfs, , stat, reg, reg);
96 SDT_PROBE_ARGTYPE(vfs, , stat, reg, 0, "char *");
97 SDT_PROBE_ARGTYPE(vfs, , stat, reg, 1, "int");
98 
99 static int chroot_refuse_vdir_fds(struct filedesc *fdp);
100 static int getutimes(const struct timeval *, enum uio_seg, struct timespec *);
101 static int setfflags(struct thread *td, struct vnode *, int);
102 static int setutimes(struct thread *td, struct vnode *,
103     const struct timespec *, int, int);
104 static int vn_access(struct vnode *vp, int user_flags, struct ucred *cred,
105     struct thread *td);
106 
107 /*
108  * The module initialization routine for POSIX asynchronous I/O will
109  * set this to the version of AIO that it implements.  (Zero means
110  * that it is not implemented.)  This value is used here by pathconf()
111  * and in kern_descrip.c by fpathconf().
112  */
113 int async_io_version;
114 
115 #ifdef DEBUG
116 static int syncprt = 0;
117 SYSCTL_INT(_debug, OID_AUTO, syncprt, CTLFLAG_RW, &syncprt, 0, "");
118 #endif
119 
120 /*
121  * Sync each mounted filesystem.
122  */
123 #ifndef _SYS_SYSPROTO_H_
124 struct sync_args {
125 	int     dummy;
126 };
127 #endif
128 /* ARGSUSED */
129 int
130 sys_sync(td, uap)
131 	struct thread *td;
132 	struct sync_args *uap;
133 {
134 	struct mount *mp, *nmp;
135 	int vfslocked;
136 
137 	mtx_lock(&mountlist_mtx);
138 	for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
139 		if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK)) {
140 			nmp = TAILQ_NEXT(mp, mnt_list);
141 			continue;
142 		}
143 		vfslocked = VFS_LOCK_GIANT(mp);
144 		if ((mp->mnt_flag & MNT_RDONLY) == 0 &&
145 		    vn_start_write(NULL, &mp, V_NOWAIT) == 0) {
146 			MNT_ILOCK(mp);
147 			mp->mnt_noasync++;
148 			mp->mnt_kern_flag &= ~MNTK_ASYNC;
149 			MNT_IUNLOCK(mp);
150 			vfs_msync(mp, MNT_NOWAIT);
151 			VFS_SYNC(mp, MNT_NOWAIT);
152 			MNT_ILOCK(mp);
153 			mp->mnt_noasync--;
154 			if ((mp->mnt_flag & MNT_ASYNC) != 0 &&
155 			    mp->mnt_noasync == 0)
156 				mp->mnt_kern_flag |= MNTK_ASYNC;
157 			MNT_IUNLOCK(mp);
158 			vn_finished_write(mp);
159 		}
160 		VFS_UNLOCK_GIANT(vfslocked);
161 		mtx_lock(&mountlist_mtx);
162 		nmp = TAILQ_NEXT(mp, mnt_list);
163 		vfs_unbusy(mp);
164 	}
165 	mtx_unlock(&mountlist_mtx);
166 	return (0);
167 }
168 
169 /*
170  * Change filesystem quotas.
171  */
172 #ifndef _SYS_SYSPROTO_H_
173 struct quotactl_args {
174 	char *path;
175 	int cmd;
176 	int uid;
177 	caddr_t arg;
178 };
179 #endif
180 int
181 sys_quotactl(td, uap)
182 	struct thread *td;
183 	register struct quotactl_args /* {
184 		char *path;
185 		int cmd;
186 		int uid;
187 		caddr_t arg;
188 	} */ *uap;
189 {
190 	struct mount *mp;
191 	int vfslocked;
192 	int error;
193 	struct nameidata nd;
194 
195 	AUDIT_ARG_CMD(uap->cmd);
196 	AUDIT_ARG_UID(uap->uid);
197 	if (!prison_allow(td->td_ucred, PR_ALLOW_QUOTAS))
198 		return (EPERM);
199 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | MPSAFE | AUDITVNODE1,
200 	   UIO_USERSPACE, uap->path, td);
201 	if ((error = namei(&nd)) != 0)
202 		return (error);
203 	vfslocked = NDHASGIANT(&nd);
204 	NDFREE(&nd, NDF_ONLY_PNBUF);
205 	mp = nd.ni_vp->v_mount;
206 	vfs_ref(mp);
207 	vput(nd.ni_vp);
208 	error = vfs_busy(mp, 0);
209 	vfs_rel(mp);
210 	if (error) {
211 		VFS_UNLOCK_GIANT(vfslocked);
212 		return (error);
213 	}
214 	error = VFS_QUOTACTL(mp, uap->cmd, uap->uid, uap->arg);
215 	vfs_unbusy(mp);
216 	VFS_UNLOCK_GIANT(vfslocked);
217 	return (error);
218 }
219 
220 /*
221  * Used by statfs conversion routines to scale the block size up if
222  * necessary so that all of the block counts are <= 'max_size'.  Note
223  * that 'max_size' should be a bitmask, i.e. 2^n - 1 for some non-zero
224  * value of 'n'.
225  */
226 void
227 statfs_scale_blocks(struct statfs *sf, long max_size)
228 {
229 	uint64_t count;
230 	int shift;
231 
232 	KASSERT(powerof2(max_size + 1), ("%s: invalid max_size", __func__));
233 
234 	/*
235 	 * Attempt to scale the block counts to give a more accurate
236 	 * overview to userland of the ratio of free space to used
237 	 * space.  To do this, find the largest block count and compute
238 	 * a divisor that lets it fit into a signed integer <= max_size.
239 	 */
240 	if (sf->f_bavail < 0)
241 		count = -sf->f_bavail;
242 	else
243 		count = sf->f_bavail;
244 	count = MAX(sf->f_blocks, MAX(sf->f_bfree, count));
245 	if (count <= max_size)
246 		return;
247 
248 	count >>= flsl(max_size);
249 	shift = 0;
250 	while (count > 0) {
251 		shift++;
252 		count >>=1;
253 	}
254 
255 	sf->f_bsize <<= shift;
256 	sf->f_blocks >>= shift;
257 	sf->f_bfree >>= shift;
258 	sf->f_bavail >>= shift;
259 }
260 
261 /*
262  * Get filesystem statistics.
263  */
264 #ifndef _SYS_SYSPROTO_H_
265 struct statfs_args {
266 	char *path;
267 	struct statfs *buf;
268 };
269 #endif
270 int
271 sys_statfs(td, uap)
272 	struct thread *td;
273 	register struct statfs_args /* {
274 		char *path;
275 		struct statfs *buf;
276 	} */ *uap;
277 {
278 	struct statfs sf;
279 	int error;
280 
281 	error = kern_statfs(td, uap->path, UIO_USERSPACE, &sf);
282 	if (error == 0)
283 		error = copyout(&sf, uap->buf, sizeof(sf));
284 	return (error);
285 }
286 
287 int
288 kern_statfs(struct thread *td, char *path, enum uio_seg pathseg,
289     struct statfs *buf)
290 {
291 	struct mount *mp;
292 	struct statfs *sp, sb;
293 	int vfslocked;
294 	int error;
295 	struct nameidata nd;
296 
297 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | MPSAFE |
298 	    AUDITVNODE1, pathseg, path, td);
299 	error = namei(&nd);
300 	if (error)
301 		return (error);
302 	vfslocked = NDHASGIANT(&nd);
303 	mp = nd.ni_vp->v_mount;
304 	vfs_ref(mp);
305 	NDFREE(&nd, NDF_ONLY_PNBUF);
306 	vput(nd.ni_vp);
307 	error = vfs_busy(mp, 0);
308 	vfs_rel(mp);
309 	if (error) {
310 		VFS_UNLOCK_GIANT(vfslocked);
311 		return (error);
312 	}
313 #ifdef MAC
314 	error = mac_mount_check_stat(td->td_ucred, mp);
315 	if (error)
316 		goto out;
317 #endif
318 	/*
319 	 * Set these in case the underlying filesystem fails to do so.
320 	 */
321 	sp = &mp->mnt_stat;
322 	sp->f_version = STATFS_VERSION;
323 	sp->f_namemax = NAME_MAX;
324 	sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
325 	error = VFS_STATFS(mp, sp);
326 	if (error)
327 		goto out;
328 	if (priv_check(td, PRIV_VFS_GENERATION)) {
329 		bcopy(sp, &sb, sizeof(sb));
330 		sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
331 		prison_enforce_statfs(td->td_ucred, mp, &sb);
332 		sp = &sb;
333 	}
334 	*buf = *sp;
335 out:
336 	vfs_unbusy(mp);
337 	VFS_UNLOCK_GIANT(vfslocked);
338 	return (error);
339 }
340 
341 /*
342  * Get filesystem statistics.
343  */
344 #ifndef _SYS_SYSPROTO_H_
345 struct fstatfs_args {
346 	int fd;
347 	struct statfs *buf;
348 };
349 #endif
350 int
351 sys_fstatfs(td, uap)
352 	struct thread *td;
353 	register struct fstatfs_args /* {
354 		int fd;
355 		struct statfs *buf;
356 	} */ *uap;
357 {
358 	struct statfs sf;
359 	int error;
360 
361 	error = kern_fstatfs(td, uap->fd, &sf);
362 	if (error == 0)
363 		error = copyout(&sf, uap->buf, sizeof(sf));
364 	return (error);
365 }
366 
367 int
368 kern_fstatfs(struct thread *td, int fd, struct statfs *buf)
369 {
370 	struct file *fp;
371 	struct mount *mp;
372 	struct statfs *sp, sb;
373 	int vfslocked;
374 	struct vnode *vp;
375 	int error;
376 
377 	AUDIT_ARG_FD(fd);
378 	error = getvnode(td->td_proc->p_fd, fd, CAP_FSTATFS, &fp);
379 	if (error)
380 		return (error);
381 	vp = fp->f_vnode;
382 	vfslocked = VFS_LOCK_GIANT(vp->v_mount);
383 	vn_lock(vp, LK_SHARED | LK_RETRY);
384 #ifdef AUDIT
385 	AUDIT_ARG_VNODE1(vp);
386 #endif
387 	mp = vp->v_mount;
388 	if (mp)
389 		vfs_ref(mp);
390 	VOP_UNLOCK(vp, 0);
391 	fdrop(fp, td);
392 	if (mp == NULL) {
393 		error = EBADF;
394 		goto out;
395 	}
396 	error = vfs_busy(mp, 0);
397 	vfs_rel(mp);
398 	if (error) {
399 		VFS_UNLOCK_GIANT(vfslocked);
400 		return (error);
401 	}
402 #ifdef MAC
403 	error = mac_mount_check_stat(td->td_ucred, mp);
404 	if (error)
405 		goto out;
406 #endif
407 	/*
408 	 * Set these in case the underlying filesystem fails to do so.
409 	 */
410 	sp = &mp->mnt_stat;
411 	sp->f_version = STATFS_VERSION;
412 	sp->f_namemax = NAME_MAX;
413 	sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
414 	error = VFS_STATFS(mp, sp);
415 	if (error)
416 		goto out;
417 	if (priv_check(td, PRIV_VFS_GENERATION)) {
418 		bcopy(sp, &sb, sizeof(sb));
419 		sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
420 		prison_enforce_statfs(td->td_ucred, mp, &sb);
421 		sp = &sb;
422 	}
423 	*buf = *sp;
424 out:
425 	if (mp)
426 		vfs_unbusy(mp);
427 	VFS_UNLOCK_GIANT(vfslocked);
428 	return (error);
429 }
430 
431 /*
432  * Get statistics on all filesystems.
433  */
434 #ifndef _SYS_SYSPROTO_H_
435 struct getfsstat_args {
436 	struct statfs *buf;
437 	long bufsize;
438 	int flags;
439 };
440 #endif
441 int
442 sys_getfsstat(td, uap)
443 	struct thread *td;
444 	register struct getfsstat_args /* {
445 		struct statfs *buf;
446 		long bufsize;
447 		int flags;
448 	} */ *uap;
449 {
450 
451 	return (kern_getfsstat(td, &uap->buf, uap->bufsize, UIO_USERSPACE,
452 	    uap->flags));
453 }
454 
455 /*
456  * If (bufsize > 0 && bufseg == UIO_SYSSPACE)
457  * 	The caller is responsible for freeing memory which will be allocated
458  *	in '*buf'.
459  */
460 int
461 kern_getfsstat(struct thread *td, struct statfs **buf, size_t bufsize,
462     enum uio_seg bufseg, int flags)
463 {
464 	struct mount *mp, *nmp;
465 	struct statfs *sfsp, *sp, sb;
466 	size_t count, maxcount;
467 	int vfslocked;
468 	int error;
469 
470 	maxcount = bufsize / sizeof(struct statfs);
471 	if (bufsize == 0)
472 		sfsp = NULL;
473 	else if (bufseg == UIO_USERSPACE)
474 		sfsp = *buf;
475 	else /* if (bufseg == UIO_SYSSPACE) */ {
476 		count = 0;
477 		mtx_lock(&mountlist_mtx);
478 		TAILQ_FOREACH(mp, &mountlist, mnt_list) {
479 			count++;
480 		}
481 		mtx_unlock(&mountlist_mtx);
482 		if (maxcount > count)
483 			maxcount = count;
484 		sfsp = *buf = malloc(maxcount * sizeof(struct statfs), M_TEMP,
485 		    M_WAITOK);
486 	}
487 	count = 0;
488 	mtx_lock(&mountlist_mtx);
489 	for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
490 		if (prison_canseemount(td->td_ucred, mp) != 0) {
491 			nmp = TAILQ_NEXT(mp, mnt_list);
492 			continue;
493 		}
494 #ifdef MAC
495 		if (mac_mount_check_stat(td->td_ucred, mp) != 0) {
496 			nmp = TAILQ_NEXT(mp, mnt_list);
497 			continue;
498 		}
499 #endif
500 		if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK)) {
501 			nmp = TAILQ_NEXT(mp, mnt_list);
502 			continue;
503 		}
504 		vfslocked = VFS_LOCK_GIANT(mp);
505 		if (sfsp && count < maxcount) {
506 			sp = &mp->mnt_stat;
507 			/*
508 			 * Set these in case the underlying filesystem
509 			 * fails to do so.
510 			 */
511 			sp->f_version = STATFS_VERSION;
512 			sp->f_namemax = NAME_MAX;
513 			sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
514 			/*
515 			 * If MNT_NOWAIT or MNT_LAZY is specified, do not
516 			 * refresh the fsstat cache. MNT_NOWAIT or MNT_LAZY
517 			 * overrides MNT_WAIT.
518 			 */
519 			if (((flags & (MNT_LAZY|MNT_NOWAIT)) == 0 ||
520 			    (flags & MNT_WAIT)) &&
521 			    (error = VFS_STATFS(mp, sp))) {
522 				VFS_UNLOCK_GIANT(vfslocked);
523 				mtx_lock(&mountlist_mtx);
524 				nmp = TAILQ_NEXT(mp, mnt_list);
525 				vfs_unbusy(mp);
526 				continue;
527 			}
528 			if (priv_check(td, PRIV_VFS_GENERATION)) {
529 				bcopy(sp, &sb, sizeof(sb));
530 				sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
531 				prison_enforce_statfs(td->td_ucred, mp, &sb);
532 				sp = &sb;
533 			}
534 			if (bufseg == UIO_SYSSPACE)
535 				bcopy(sp, sfsp, sizeof(*sp));
536 			else /* if (bufseg == UIO_USERSPACE) */ {
537 				error = copyout(sp, sfsp, sizeof(*sp));
538 				if (error) {
539 					vfs_unbusy(mp);
540 					VFS_UNLOCK_GIANT(vfslocked);
541 					return (error);
542 				}
543 			}
544 			sfsp++;
545 		}
546 		VFS_UNLOCK_GIANT(vfslocked);
547 		count++;
548 		mtx_lock(&mountlist_mtx);
549 		nmp = TAILQ_NEXT(mp, mnt_list);
550 		vfs_unbusy(mp);
551 	}
552 	mtx_unlock(&mountlist_mtx);
553 	if (sfsp && count > maxcount)
554 		td->td_retval[0] = maxcount;
555 	else
556 		td->td_retval[0] = count;
557 	return (0);
558 }
559 
560 #ifdef COMPAT_FREEBSD4
561 /*
562  * Get old format filesystem statistics.
563  */
564 static void cvtstatfs(struct statfs *, struct ostatfs *);
565 
566 #ifndef _SYS_SYSPROTO_H_
567 struct freebsd4_statfs_args {
568 	char *path;
569 	struct ostatfs *buf;
570 };
571 #endif
572 int
573 freebsd4_statfs(td, uap)
574 	struct thread *td;
575 	struct freebsd4_statfs_args /* {
576 		char *path;
577 		struct ostatfs *buf;
578 	} */ *uap;
579 {
580 	struct ostatfs osb;
581 	struct statfs sf;
582 	int error;
583 
584 	error = kern_statfs(td, uap->path, UIO_USERSPACE, &sf);
585 	if (error)
586 		return (error);
587 	cvtstatfs(&sf, &osb);
588 	return (copyout(&osb, uap->buf, sizeof(osb)));
589 }
590 
591 /*
592  * Get filesystem statistics.
593  */
594 #ifndef _SYS_SYSPROTO_H_
595 struct freebsd4_fstatfs_args {
596 	int fd;
597 	struct ostatfs *buf;
598 };
599 #endif
600 int
601 freebsd4_fstatfs(td, uap)
602 	struct thread *td;
603 	struct freebsd4_fstatfs_args /* {
604 		int fd;
605 		struct ostatfs *buf;
606 	} */ *uap;
607 {
608 	struct ostatfs osb;
609 	struct statfs sf;
610 	int error;
611 
612 	error = kern_fstatfs(td, uap->fd, &sf);
613 	if (error)
614 		return (error);
615 	cvtstatfs(&sf, &osb);
616 	return (copyout(&osb, uap->buf, sizeof(osb)));
617 }
618 
619 /*
620  * Get statistics on all filesystems.
621  */
622 #ifndef _SYS_SYSPROTO_H_
623 struct freebsd4_getfsstat_args {
624 	struct ostatfs *buf;
625 	long bufsize;
626 	int flags;
627 };
628 #endif
629 int
630 freebsd4_getfsstat(td, uap)
631 	struct thread *td;
632 	register struct freebsd4_getfsstat_args /* {
633 		struct ostatfs *buf;
634 		long bufsize;
635 		int flags;
636 	} */ *uap;
637 {
638 	struct statfs *buf, *sp;
639 	struct ostatfs osb;
640 	size_t count, size;
641 	int error;
642 
643 	count = uap->bufsize / sizeof(struct ostatfs);
644 	size = count * sizeof(struct statfs);
645 	error = kern_getfsstat(td, &buf, size, UIO_SYSSPACE, uap->flags);
646 	if (size > 0) {
647 		count = td->td_retval[0];
648 		sp = buf;
649 		while (count > 0 && error == 0) {
650 			cvtstatfs(sp, &osb);
651 			error = copyout(&osb, uap->buf, sizeof(osb));
652 			sp++;
653 			uap->buf++;
654 			count--;
655 		}
656 		free(buf, M_TEMP);
657 	}
658 	return (error);
659 }
660 
661 /*
662  * Implement fstatfs() for (NFS) file handles.
663  */
664 #ifndef _SYS_SYSPROTO_H_
665 struct freebsd4_fhstatfs_args {
666 	struct fhandle *u_fhp;
667 	struct ostatfs *buf;
668 };
669 #endif
670 int
671 freebsd4_fhstatfs(td, uap)
672 	struct thread *td;
673 	struct freebsd4_fhstatfs_args /* {
674 		struct fhandle *u_fhp;
675 		struct ostatfs *buf;
676 	} */ *uap;
677 {
678 	struct ostatfs osb;
679 	struct statfs sf;
680 	fhandle_t fh;
681 	int error;
682 
683 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
684 	if (error)
685 		return (error);
686 	error = kern_fhstatfs(td, fh, &sf);
687 	if (error)
688 		return (error);
689 	cvtstatfs(&sf, &osb);
690 	return (copyout(&osb, uap->buf, sizeof(osb)));
691 }
692 
693 /*
694  * Convert a new format statfs structure to an old format statfs structure.
695  */
696 static void
697 cvtstatfs(nsp, osp)
698 	struct statfs *nsp;
699 	struct ostatfs *osp;
700 {
701 
702 	statfs_scale_blocks(nsp, LONG_MAX);
703 	bzero(osp, sizeof(*osp));
704 	osp->f_bsize = nsp->f_bsize;
705 	osp->f_iosize = MIN(nsp->f_iosize, LONG_MAX);
706 	osp->f_blocks = nsp->f_blocks;
707 	osp->f_bfree = nsp->f_bfree;
708 	osp->f_bavail = nsp->f_bavail;
709 	osp->f_files = MIN(nsp->f_files, LONG_MAX);
710 	osp->f_ffree = MIN(nsp->f_ffree, LONG_MAX);
711 	osp->f_owner = nsp->f_owner;
712 	osp->f_type = nsp->f_type;
713 	osp->f_flags = nsp->f_flags;
714 	osp->f_syncwrites = MIN(nsp->f_syncwrites, LONG_MAX);
715 	osp->f_asyncwrites = MIN(nsp->f_asyncwrites, LONG_MAX);
716 	osp->f_syncreads = MIN(nsp->f_syncreads, LONG_MAX);
717 	osp->f_asyncreads = MIN(nsp->f_asyncreads, LONG_MAX);
718 	strlcpy(osp->f_fstypename, nsp->f_fstypename,
719 	    MIN(MFSNAMELEN, OMFSNAMELEN));
720 	strlcpy(osp->f_mntonname, nsp->f_mntonname,
721 	    MIN(MNAMELEN, OMNAMELEN));
722 	strlcpy(osp->f_mntfromname, nsp->f_mntfromname,
723 	    MIN(MNAMELEN, OMNAMELEN));
724 	osp->f_fsid = nsp->f_fsid;
725 }
726 #endif /* COMPAT_FREEBSD4 */
727 
728 /*
729  * Change current working directory to a given file descriptor.
730  */
731 #ifndef _SYS_SYSPROTO_H_
732 struct fchdir_args {
733 	int	fd;
734 };
735 #endif
736 int
737 sys_fchdir(td, uap)
738 	struct thread *td;
739 	struct fchdir_args /* {
740 		int fd;
741 	} */ *uap;
742 {
743 	register struct filedesc *fdp = td->td_proc->p_fd;
744 	struct vnode *vp, *tdp, *vpold;
745 	struct mount *mp;
746 	struct file *fp;
747 	int vfslocked;
748 	int error;
749 
750 	AUDIT_ARG_FD(uap->fd);
751 	if ((error = getvnode(fdp, uap->fd, CAP_FCHDIR, &fp)) != 0)
752 		return (error);
753 	vp = fp->f_vnode;
754 	VREF(vp);
755 	fdrop(fp, td);
756 	vfslocked = VFS_LOCK_GIANT(vp->v_mount);
757 	vn_lock(vp, LK_SHARED | LK_RETRY);
758 	AUDIT_ARG_VNODE1(vp);
759 	error = change_dir(vp, td);
760 	while (!error && (mp = vp->v_mountedhere) != NULL) {
761 		int tvfslocked;
762 		if (vfs_busy(mp, 0))
763 			continue;
764 		tvfslocked = VFS_LOCK_GIANT(mp);
765 		error = VFS_ROOT(mp, LK_SHARED, &tdp);
766 		vfs_unbusy(mp);
767 		if (error) {
768 			VFS_UNLOCK_GIANT(tvfslocked);
769 			break;
770 		}
771 		vput(vp);
772 		VFS_UNLOCK_GIANT(vfslocked);
773 		vp = tdp;
774 		vfslocked = tvfslocked;
775 	}
776 	if (error) {
777 		vput(vp);
778 		VFS_UNLOCK_GIANT(vfslocked);
779 		return (error);
780 	}
781 	VOP_UNLOCK(vp, 0);
782 	VFS_UNLOCK_GIANT(vfslocked);
783 	FILEDESC_XLOCK(fdp);
784 	vpold = fdp->fd_cdir;
785 	fdp->fd_cdir = vp;
786 	FILEDESC_XUNLOCK(fdp);
787 	vfslocked = VFS_LOCK_GIANT(vpold->v_mount);
788 	vrele(vpold);
789 	VFS_UNLOCK_GIANT(vfslocked);
790 	return (0);
791 }
792 
793 /*
794  * Change current working directory (``.'').
795  */
796 #ifndef _SYS_SYSPROTO_H_
797 struct chdir_args {
798 	char	*path;
799 };
800 #endif
801 int
802 sys_chdir(td, uap)
803 	struct thread *td;
804 	struct chdir_args /* {
805 		char *path;
806 	} */ *uap;
807 {
808 
809 	return (kern_chdir(td, uap->path, UIO_USERSPACE));
810 }
811 
812 int
813 kern_chdir(struct thread *td, char *path, enum uio_seg pathseg)
814 {
815 	register struct filedesc *fdp = td->td_proc->p_fd;
816 	int error;
817 	struct nameidata nd;
818 	struct vnode *vp;
819 	int vfslocked;
820 
821 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1 |
822 	    MPSAFE, pathseg, path, td);
823 	if ((error = namei(&nd)) != 0)
824 		return (error);
825 	vfslocked = NDHASGIANT(&nd);
826 	if ((error = change_dir(nd.ni_vp, td)) != 0) {
827 		vput(nd.ni_vp);
828 		VFS_UNLOCK_GIANT(vfslocked);
829 		NDFREE(&nd, NDF_ONLY_PNBUF);
830 		return (error);
831 	}
832 	VOP_UNLOCK(nd.ni_vp, 0);
833 	VFS_UNLOCK_GIANT(vfslocked);
834 	NDFREE(&nd, NDF_ONLY_PNBUF);
835 	FILEDESC_XLOCK(fdp);
836 	vp = fdp->fd_cdir;
837 	fdp->fd_cdir = nd.ni_vp;
838 	FILEDESC_XUNLOCK(fdp);
839 	vfslocked = VFS_LOCK_GIANT(vp->v_mount);
840 	vrele(vp);
841 	VFS_UNLOCK_GIANT(vfslocked);
842 	return (0);
843 }
844 
845 /*
846  * Helper function for raised chroot(2) security function:  Refuse if
847  * any filedescriptors are open directories.
848  */
849 static int
850 chroot_refuse_vdir_fds(fdp)
851 	struct filedesc *fdp;
852 {
853 	struct vnode *vp;
854 	struct file *fp;
855 	int fd;
856 
857 	FILEDESC_LOCK_ASSERT(fdp);
858 
859 	for (fd = 0; fd < fdp->fd_nfiles ; fd++) {
860 		fp = fget_locked(fdp, fd);
861 		if (fp == NULL)
862 			continue;
863 		if (fp->f_type == DTYPE_VNODE) {
864 			vp = fp->f_vnode;
865 			if (vp->v_type == VDIR)
866 				return (EPERM);
867 		}
868 	}
869 	return (0);
870 }
871 
872 /*
873  * This sysctl determines if we will allow a process to chroot(2) if it
874  * has a directory open:
875  *	0: disallowed for all processes.
876  *	1: allowed for processes that were not already chroot(2)'ed.
877  *	2: allowed for all processes.
878  */
879 
880 static int chroot_allow_open_directories = 1;
881 
882 SYSCTL_INT(_kern, OID_AUTO, chroot_allow_open_directories, CTLFLAG_RW,
883      &chroot_allow_open_directories, 0,
884      "Allow a process to chroot(2) if it has a directory open");
885 
886 /*
887  * Change notion of root (``/'') directory.
888  */
889 #ifndef _SYS_SYSPROTO_H_
890 struct chroot_args {
891 	char	*path;
892 };
893 #endif
894 int
895 sys_chroot(td, uap)
896 	struct thread *td;
897 	struct chroot_args /* {
898 		char *path;
899 	} */ *uap;
900 {
901 	int error;
902 	struct nameidata nd;
903 	int vfslocked;
904 
905 	error = priv_check(td, PRIV_VFS_CHROOT);
906 	if (error)
907 		return (error);
908 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | MPSAFE |
909 	    AUDITVNODE1, UIO_USERSPACE, uap->path, td);
910 	error = namei(&nd);
911 	if (error)
912 		goto error;
913 	vfslocked = NDHASGIANT(&nd);
914 	if ((error = change_dir(nd.ni_vp, td)) != 0)
915 		goto e_vunlock;
916 #ifdef MAC
917 	if ((error = mac_vnode_check_chroot(td->td_ucred, nd.ni_vp)))
918 		goto e_vunlock;
919 #endif
920 	VOP_UNLOCK(nd.ni_vp, 0);
921 	error = change_root(nd.ni_vp, td);
922 	vrele(nd.ni_vp);
923 	VFS_UNLOCK_GIANT(vfslocked);
924 	NDFREE(&nd, NDF_ONLY_PNBUF);
925 	return (error);
926 e_vunlock:
927 	vput(nd.ni_vp);
928 	VFS_UNLOCK_GIANT(vfslocked);
929 error:
930 	NDFREE(&nd, NDF_ONLY_PNBUF);
931 	return (error);
932 }
933 
934 /*
935  * Common routine for chroot and chdir.  Callers must provide a locked vnode
936  * instance.
937  */
938 int
939 change_dir(vp, td)
940 	struct vnode *vp;
941 	struct thread *td;
942 {
943 	int error;
944 
945 	ASSERT_VOP_LOCKED(vp, "change_dir(): vp not locked");
946 	if (vp->v_type != VDIR)
947 		return (ENOTDIR);
948 #ifdef MAC
949 	error = mac_vnode_check_chdir(td->td_ucred, vp);
950 	if (error)
951 		return (error);
952 #endif
953 	error = VOP_ACCESS(vp, VEXEC, td->td_ucred, td);
954 	return (error);
955 }
956 
957 /*
958  * Common routine for kern_chroot() and jail_attach().  The caller is
959  * responsible for invoking priv_check() and mac_vnode_check_chroot() to
960  * authorize this operation.
961  */
962 int
963 change_root(vp, td)
964 	struct vnode *vp;
965 	struct thread *td;
966 {
967 	struct filedesc *fdp;
968 	struct vnode *oldvp;
969 	int vfslocked;
970 	int error;
971 
972 	VFS_ASSERT_GIANT(vp->v_mount);
973 	fdp = td->td_proc->p_fd;
974 	FILEDESC_XLOCK(fdp);
975 	if (chroot_allow_open_directories == 0 ||
976 	    (chroot_allow_open_directories == 1 && fdp->fd_rdir != rootvnode)) {
977 		error = chroot_refuse_vdir_fds(fdp);
978 		if (error) {
979 			FILEDESC_XUNLOCK(fdp);
980 			return (error);
981 		}
982 	}
983 	oldvp = fdp->fd_rdir;
984 	fdp->fd_rdir = vp;
985 	VREF(fdp->fd_rdir);
986 	if (!fdp->fd_jdir) {
987 		fdp->fd_jdir = vp;
988 		VREF(fdp->fd_jdir);
989 	}
990 	FILEDESC_XUNLOCK(fdp);
991 	vfslocked = VFS_LOCK_GIANT(oldvp->v_mount);
992 	vrele(oldvp);
993 	VFS_UNLOCK_GIANT(vfslocked);
994 	return (0);
995 }
996 
997 static __inline cap_rights_t
998 flags_to_rights(int flags)
999 {
1000 	cap_rights_t rights = 0;
1001 
1002 	switch ((flags & O_ACCMODE)) {
1003 	case O_RDONLY:
1004 		rights |= CAP_READ;
1005 		break;
1006 
1007 	case O_RDWR:
1008 		rights |= CAP_READ;
1009 		/* fall through */
1010 
1011 	case O_WRONLY:
1012 		rights |= CAP_WRITE;
1013 		break;
1014 
1015 	case O_EXEC:
1016 		rights |= CAP_FEXECVE;
1017 		break;
1018 	}
1019 
1020 	if (flags & O_CREAT)
1021 		rights |= CAP_CREATE;
1022 
1023 	if (flags & O_TRUNC)
1024 		rights |= CAP_FTRUNCATE;
1025 
1026 	if ((flags & O_EXLOCK) || (flags & O_SHLOCK))
1027 		rights |= CAP_FLOCK;
1028 
1029 	return (rights);
1030 }
1031 
1032 /*
1033  * Check permissions, allocate an open file structure, and call the device
1034  * open routine if any.
1035  */
1036 #ifndef _SYS_SYSPROTO_H_
1037 struct open_args {
1038 	char	*path;
1039 	int	flags;
1040 	int	mode;
1041 };
1042 #endif
1043 int
1044 sys_open(td, uap)
1045 	struct thread *td;
1046 	register struct open_args /* {
1047 		char *path;
1048 		int flags;
1049 		int mode;
1050 	} */ *uap;
1051 {
1052 
1053 	return (kern_open(td, uap->path, UIO_USERSPACE, uap->flags, uap->mode));
1054 }
1055 
1056 #ifndef _SYS_SYSPROTO_H_
1057 struct openat_args {
1058 	int	fd;
1059 	char	*path;
1060 	int	flag;
1061 	int	mode;
1062 };
1063 #endif
1064 int
1065 sys_openat(struct thread *td, struct openat_args *uap)
1066 {
1067 
1068 	return (kern_openat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
1069 	    uap->mode));
1070 }
1071 
1072 int
1073 kern_open(struct thread *td, char *path, enum uio_seg pathseg, int flags,
1074     int mode)
1075 {
1076 
1077 	return (kern_openat(td, AT_FDCWD, path, pathseg, flags, mode));
1078 }
1079 
1080 int
1081 kern_openat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1082     int flags, int mode)
1083 {
1084 	struct proc *p = td->td_proc;
1085 	struct filedesc *fdp = p->p_fd;
1086 	struct file *fp;
1087 	struct vnode *vp;
1088 	int cmode;
1089 	struct file *nfp;
1090 	int type, indx = -1, error, error_open;
1091 	struct flock lf;
1092 	struct nameidata nd;
1093 	int vfslocked;
1094 	cap_rights_t rights_needed = CAP_LOOKUP;
1095 
1096 	AUDIT_ARG_FFLAGS(flags);
1097 	AUDIT_ARG_MODE(mode);
1098 	/* XXX: audit dirfd */
1099 	rights_needed |= flags_to_rights(flags);
1100 	/*
1101 	 * Only one of the O_EXEC, O_RDONLY, O_WRONLY and O_RDWR flags
1102 	 * may be specified.
1103 	 */
1104 	if (flags & O_EXEC) {
1105 		if (flags & O_ACCMODE)
1106 			return (EINVAL);
1107 	} else if ((flags & O_ACCMODE) == O_ACCMODE)
1108 		return (EINVAL);
1109 	else
1110 		flags = FFLAGS(flags);
1111 
1112 	/*
1113 	 * allocate the file descriptor, but don't install a descriptor yet
1114 	 */
1115 	error = falloc_noinstall(td, &nfp);
1116 	if (error)
1117 		return (error);
1118 	/* An extra reference on `nfp' has been held for us by falloc_noinstall(). */
1119 	fp = nfp;
1120 	/* Set the flags early so the finit in devfs can pick them up. */
1121 	fp->f_flag = flags & FMASK;
1122 	cmode = ((mode &~ fdp->fd_cmask) & ALLPERMS) &~ S_ISTXT;
1123 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1 | MPSAFE, pathseg,
1124 	    path, fd, rights_needed, td);
1125 	td->td_dupfd = -1;		/* XXX check for fdopen */
1126 	error = vn_open(&nd, &flags, cmode, fp);
1127 	if (error) {
1128 		/*
1129 		 * If the vn_open replaced the method vector, something
1130 		 * wonderous happened deep below and we just pass it up
1131 		 * pretending we know what we do.
1132 		 */
1133 		if (error == ENXIO && fp->f_ops != &badfileops)
1134 			goto success;
1135 
1136 		/*
1137 		 * handle special fdopen() case.  bleh.  dupfdopen() is
1138 		 * responsible for dropping the old contents of ofiles[indx]
1139 		 * if it succeeds.
1140 		 *
1141 		 * Don't do this for relative (capability) lookups; we don't
1142 		 * understand exactly what would happen, and we don't think
1143 		 * that it ever should.
1144 		 */
1145 		if ((nd.ni_strictrelative == 0) &&
1146 		    (error == ENODEV || error == ENXIO) &&
1147 		    (td->td_dupfd >= 0)) {
1148 			/* XXX from fdopen */
1149 			error_open = error;
1150 			if ((error = finstall(td, fp, &indx, flags)) != 0)
1151 				goto bad_unlocked;
1152 			if ((error = dupfdopen(td, fdp, indx, td->td_dupfd,
1153 			    flags, error_open)) == 0)
1154 				goto success;
1155 		}
1156 		/*
1157 		 * Clean up the descriptor, but only if another thread hadn't
1158 		 * replaced or closed it.
1159 		 */
1160 		if (indx != -1)
1161 			fdclose(fdp, fp, indx, td);
1162 		fdrop(fp, td);
1163 
1164 		if (error == ERESTART)
1165 			error = EINTR;
1166 		return (error);
1167 	}
1168 	td->td_dupfd = 0;
1169 	vfslocked = NDHASGIANT(&nd);
1170 	NDFREE(&nd, NDF_ONLY_PNBUF);
1171 	vp = nd.ni_vp;
1172 
1173 	/*
1174 	 * Store the vnode, for any f_type. Typically, the vnode use
1175 	 * count is decremented by direct call to vn_closefile() for
1176 	 * files that switched type in the cdevsw fdopen() method.
1177 	 */
1178 	fp->f_vnode = vp;
1179 	/*
1180 	 * If the file wasn't claimed by devfs bind it to the normal
1181 	 * vnode operations here.
1182 	 */
1183 	if (fp->f_ops == &badfileops) {
1184 		KASSERT(vp->v_type != VFIFO, ("Unexpected fifo."));
1185 		fp->f_seqcount = 1;
1186 		finit(fp, flags & FMASK, DTYPE_VNODE, vp, &vnops);
1187 	}
1188 
1189 	VOP_UNLOCK(vp, 0);
1190 	if (fp->f_type == DTYPE_VNODE && (flags & (O_EXLOCK | O_SHLOCK)) != 0) {
1191 		lf.l_whence = SEEK_SET;
1192 		lf.l_start = 0;
1193 		lf.l_len = 0;
1194 		if (flags & O_EXLOCK)
1195 			lf.l_type = F_WRLCK;
1196 		else
1197 			lf.l_type = F_RDLCK;
1198 		type = F_FLOCK;
1199 		if ((flags & FNONBLOCK) == 0)
1200 			type |= F_WAIT;
1201 		if ((error = VOP_ADVLOCK(vp, (caddr_t)fp, F_SETLK, &lf,
1202 			    type)) != 0)
1203 			goto bad;
1204 		atomic_set_int(&fp->f_flag, FHASLOCK);
1205 	}
1206 	if (flags & O_TRUNC) {
1207 		error = fo_truncate(fp, 0, td->td_ucred, td);
1208 		if (error)
1209 			goto bad;
1210 	}
1211 	VFS_UNLOCK_GIANT(vfslocked);
1212 success:
1213 	/*
1214 	 * If we haven't already installed the FD (for dupfdopen), do so now.
1215 	 */
1216 	if (indx == -1) {
1217 #ifdef CAPABILITIES
1218 		if (nd.ni_strictrelative == 1) {
1219 			/*
1220 			 * We are doing a strict relative lookup; wrap the
1221 			 * result in a capability.
1222 			 */
1223 			if ((error = kern_capwrap(td, fp, nd.ni_baserights,
1224 			    &indx)) != 0)
1225 				goto bad_unlocked;
1226 		} else
1227 #endif
1228 			if ((error = finstall(td, fp, &indx, flags)) != 0)
1229 				goto bad_unlocked;
1230 
1231 	}
1232 
1233 	/*
1234 	 * Release our private reference, leaving the one associated with
1235 	 * the descriptor table intact.
1236 	 */
1237 	fdrop(fp, td);
1238 	td->td_retval[0] = indx;
1239 	return (0);
1240 bad:
1241 	VFS_UNLOCK_GIANT(vfslocked);
1242 bad_unlocked:
1243 	if (indx != -1)
1244 		fdclose(fdp, fp, indx, td);
1245 	fdrop(fp, td);
1246 	td->td_retval[0] = -1;
1247 	return (error);
1248 }
1249 
1250 #ifdef COMPAT_43
1251 /*
1252  * Create a file.
1253  */
1254 #ifndef _SYS_SYSPROTO_H_
1255 struct ocreat_args {
1256 	char	*path;
1257 	int	mode;
1258 };
1259 #endif
1260 int
1261 ocreat(td, uap)
1262 	struct thread *td;
1263 	register struct ocreat_args /* {
1264 		char *path;
1265 		int mode;
1266 	} */ *uap;
1267 {
1268 
1269 	return (kern_open(td, uap->path, UIO_USERSPACE,
1270 	    O_WRONLY | O_CREAT | O_TRUNC, uap->mode));
1271 }
1272 #endif /* COMPAT_43 */
1273 
1274 /*
1275  * Create a special file.
1276  */
1277 #ifndef _SYS_SYSPROTO_H_
1278 struct mknod_args {
1279 	char	*path;
1280 	int	mode;
1281 	int	dev;
1282 };
1283 #endif
1284 int
1285 sys_mknod(td, uap)
1286 	struct thread *td;
1287 	register struct mknod_args /* {
1288 		char *path;
1289 		int mode;
1290 		int dev;
1291 	} */ *uap;
1292 {
1293 
1294 	return (kern_mknod(td, uap->path, UIO_USERSPACE, uap->mode, uap->dev));
1295 }
1296 
1297 #ifndef _SYS_SYSPROTO_H_
1298 struct mknodat_args {
1299 	int	fd;
1300 	char	*path;
1301 	mode_t	mode;
1302 	dev_t	dev;
1303 };
1304 #endif
1305 int
1306 sys_mknodat(struct thread *td, struct mknodat_args *uap)
1307 {
1308 
1309 	return (kern_mknodat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode,
1310 	    uap->dev));
1311 }
1312 
1313 int
1314 kern_mknod(struct thread *td, char *path, enum uio_seg pathseg, int mode,
1315     int dev)
1316 {
1317 
1318 	return (kern_mknodat(td, AT_FDCWD, path, pathseg, mode, dev));
1319 }
1320 
1321 int
1322 kern_mknodat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1323     int mode, int dev)
1324 {
1325 	struct vnode *vp;
1326 	struct mount *mp;
1327 	struct vattr vattr;
1328 	int error;
1329 	int whiteout = 0;
1330 	struct nameidata nd;
1331 	int vfslocked;
1332 
1333 	AUDIT_ARG_MODE(mode);
1334 	AUDIT_ARG_DEV(dev);
1335 	switch (mode & S_IFMT) {
1336 	case S_IFCHR:
1337 	case S_IFBLK:
1338 		error = priv_check(td, PRIV_VFS_MKNOD_DEV);
1339 		break;
1340 	case S_IFMT:
1341 		error = priv_check(td, PRIV_VFS_MKNOD_BAD);
1342 		break;
1343 	case S_IFWHT:
1344 		error = priv_check(td, PRIV_VFS_MKNOD_WHT);
1345 		break;
1346 	case S_IFIFO:
1347 		if (dev == 0)
1348 			return (kern_mkfifoat(td, fd, path, pathseg, mode));
1349 		/* FALLTHROUGH */
1350 	default:
1351 		error = EINVAL;
1352 		break;
1353 	}
1354 	if (error)
1355 		return (error);
1356 restart:
1357 	bwillwrite();
1358 	NDINIT_ATRIGHTS(&nd, CREATE,
1359 	    LOCKPARENT | SAVENAME | MPSAFE | AUDITVNODE1, pathseg, path, fd,
1360 	    CAP_MKFIFO, td);
1361 	if ((error = namei(&nd)) != 0)
1362 		return (error);
1363 	vfslocked = NDHASGIANT(&nd);
1364 	vp = nd.ni_vp;
1365 	if (vp != NULL) {
1366 		NDFREE(&nd, NDF_ONLY_PNBUF);
1367 		if (vp == nd.ni_dvp)
1368 			vrele(nd.ni_dvp);
1369 		else
1370 			vput(nd.ni_dvp);
1371 		vrele(vp);
1372 		VFS_UNLOCK_GIANT(vfslocked);
1373 		return (EEXIST);
1374 	} else {
1375 		VATTR_NULL(&vattr);
1376 		vattr.va_mode = (mode & ALLPERMS) &
1377 		    ~td->td_proc->p_fd->fd_cmask;
1378 		vattr.va_rdev = dev;
1379 		whiteout = 0;
1380 
1381 		switch (mode & S_IFMT) {
1382 		case S_IFMT:	/* used by badsect to flag bad sectors */
1383 			vattr.va_type = VBAD;
1384 			break;
1385 		case S_IFCHR:
1386 			vattr.va_type = VCHR;
1387 			break;
1388 		case S_IFBLK:
1389 			vattr.va_type = VBLK;
1390 			break;
1391 		case S_IFWHT:
1392 			whiteout = 1;
1393 			break;
1394 		default:
1395 			panic("kern_mknod: invalid mode");
1396 		}
1397 	}
1398 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1399 		NDFREE(&nd, NDF_ONLY_PNBUF);
1400 		vput(nd.ni_dvp);
1401 		VFS_UNLOCK_GIANT(vfslocked);
1402 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1403 			return (error);
1404 		goto restart;
1405 	}
1406 #ifdef MAC
1407 	if (error == 0 && !whiteout)
1408 		error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp,
1409 		    &nd.ni_cnd, &vattr);
1410 #endif
1411 	if (!error) {
1412 		if (whiteout)
1413 			error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE);
1414 		else {
1415 			error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp,
1416 						&nd.ni_cnd, &vattr);
1417 			if (error == 0)
1418 				vput(nd.ni_vp);
1419 		}
1420 	}
1421 	NDFREE(&nd, NDF_ONLY_PNBUF);
1422 	vput(nd.ni_dvp);
1423 	vn_finished_write(mp);
1424 	VFS_UNLOCK_GIANT(vfslocked);
1425 	return (error);
1426 }
1427 
1428 /*
1429  * Create a named pipe.
1430  */
1431 #ifndef _SYS_SYSPROTO_H_
1432 struct mkfifo_args {
1433 	char	*path;
1434 	int	mode;
1435 };
1436 #endif
1437 int
1438 sys_mkfifo(td, uap)
1439 	struct thread *td;
1440 	register struct mkfifo_args /* {
1441 		char *path;
1442 		int mode;
1443 	} */ *uap;
1444 {
1445 
1446 	return (kern_mkfifo(td, uap->path, UIO_USERSPACE, uap->mode));
1447 }
1448 
1449 #ifndef _SYS_SYSPROTO_H_
1450 struct mkfifoat_args {
1451 	int	fd;
1452 	char	*path;
1453 	mode_t	mode;
1454 };
1455 #endif
1456 int
1457 sys_mkfifoat(struct thread *td, struct mkfifoat_args *uap)
1458 {
1459 
1460 	return (kern_mkfifoat(td, uap->fd, uap->path, UIO_USERSPACE,
1461 	    uap->mode));
1462 }
1463 
1464 int
1465 kern_mkfifo(struct thread *td, char *path, enum uio_seg pathseg, int mode)
1466 {
1467 
1468 	return (kern_mkfifoat(td, AT_FDCWD, path, pathseg, mode));
1469 }
1470 
1471 int
1472 kern_mkfifoat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1473     int mode)
1474 {
1475 	struct mount *mp;
1476 	struct vattr vattr;
1477 	int error;
1478 	struct nameidata nd;
1479 	int vfslocked;
1480 
1481 	AUDIT_ARG_MODE(mode);
1482 restart:
1483 	bwillwrite();
1484 	NDINIT_AT(&nd, CREATE, LOCKPARENT | SAVENAME | MPSAFE | AUDITVNODE1,
1485 	    pathseg, path, fd, td);
1486 	if ((error = namei(&nd)) != 0)
1487 		return (error);
1488 	vfslocked = NDHASGIANT(&nd);
1489 	if (nd.ni_vp != NULL) {
1490 		NDFREE(&nd, NDF_ONLY_PNBUF);
1491 		if (nd.ni_vp == nd.ni_dvp)
1492 			vrele(nd.ni_dvp);
1493 		else
1494 			vput(nd.ni_dvp);
1495 		vrele(nd.ni_vp);
1496 		VFS_UNLOCK_GIANT(vfslocked);
1497 		return (EEXIST);
1498 	}
1499 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1500 		NDFREE(&nd, NDF_ONLY_PNBUF);
1501 		vput(nd.ni_dvp);
1502 		VFS_UNLOCK_GIANT(vfslocked);
1503 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1504 			return (error);
1505 		goto restart;
1506 	}
1507 	VATTR_NULL(&vattr);
1508 	vattr.va_type = VFIFO;
1509 	vattr.va_mode = (mode & ALLPERMS) & ~td->td_proc->p_fd->fd_cmask;
1510 #ifdef MAC
1511 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1512 	    &vattr);
1513 	if (error)
1514 		goto out;
1515 #endif
1516 	error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
1517 	if (error == 0)
1518 		vput(nd.ni_vp);
1519 #ifdef MAC
1520 out:
1521 #endif
1522 	vput(nd.ni_dvp);
1523 	vn_finished_write(mp);
1524 	VFS_UNLOCK_GIANT(vfslocked);
1525 	NDFREE(&nd, NDF_ONLY_PNBUF);
1526 	return (error);
1527 }
1528 
1529 /*
1530  * Make a hard file link.
1531  */
1532 #ifndef _SYS_SYSPROTO_H_
1533 struct link_args {
1534 	char	*path;
1535 	char	*link;
1536 };
1537 #endif
1538 int
1539 sys_link(td, uap)
1540 	struct thread *td;
1541 	register struct link_args /* {
1542 		char *path;
1543 		char *link;
1544 	} */ *uap;
1545 {
1546 
1547 	return (kern_link(td, uap->path, uap->link, UIO_USERSPACE));
1548 }
1549 
1550 #ifndef _SYS_SYSPROTO_H_
1551 struct linkat_args {
1552 	int	fd1;
1553 	char	*path1;
1554 	int	fd2;
1555 	char	*path2;
1556 	int	flag;
1557 };
1558 #endif
1559 int
1560 sys_linkat(struct thread *td, struct linkat_args *uap)
1561 {
1562 	int flag;
1563 
1564 	flag = uap->flag;
1565 	if (flag & ~AT_SYMLINK_FOLLOW)
1566 		return (EINVAL);
1567 
1568 	return (kern_linkat(td, uap->fd1, uap->fd2, uap->path1, uap->path2,
1569 	    UIO_USERSPACE, (flag & AT_SYMLINK_FOLLOW) ? FOLLOW : NOFOLLOW));
1570 }
1571 
1572 int hardlink_check_uid = 0;
1573 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_uid, CTLFLAG_RW,
1574     &hardlink_check_uid, 0,
1575     "Unprivileged processes cannot create hard links to files owned by other "
1576     "users");
1577 static int hardlink_check_gid = 0;
1578 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_gid, CTLFLAG_RW,
1579     &hardlink_check_gid, 0,
1580     "Unprivileged processes cannot create hard links to files owned by other "
1581     "groups");
1582 
1583 static int
1584 can_hardlink(struct vnode *vp, struct ucred *cred)
1585 {
1586 	struct vattr va;
1587 	int error;
1588 
1589 	if (!hardlink_check_uid && !hardlink_check_gid)
1590 		return (0);
1591 
1592 	error = VOP_GETATTR(vp, &va, cred);
1593 	if (error != 0)
1594 		return (error);
1595 
1596 	if (hardlink_check_uid && cred->cr_uid != va.va_uid) {
1597 		error = priv_check_cred(cred, PRIV_VFS_LINK, 0);
1598 		if (error)
1599 			return (error);
1600 	}
1601 
1602 	if (hardlink_check_gid && !groupmember(va.va_gid, cred)) {
1603 		error = priv_check_cred(cred, PRIV_VFS_LINK, 0);
1604 		if (error)
1605 			return (error);
1606 	}
1607 
1608 	return (0);
1609 }
1610 
1611 int
1612 kern_link(struct thread *td, char *path, char *link, enum uio_seg segflg)
1613 {
1614 
1615 	return (kern_linkat(td, AT_FDCWD, AT_FDCWD, path,link, segflg, FOLLOW));
1616 }
1617 
1618 int
1619 kern_linkat(struct thread *td, int fd1, int fd2, char *path1, char *path2,
1620     enum uio_seg segflg, int follow)
1621 {
1622 	struct vnode *vp;
1623 	struct mount *mp;
1624 	struct nameidata nd;
1625 	int vfslocked;
1626 	int lvfslocked;
1627 	int error;
1628 
1629 	bwillwrite();
1630 	NDINIT_AT(&nd, LOOKUP, follow | MPSAFE | AUDITVNODE1, segflg, path1,
1631 	    fd1, td);
1632 
1633 	if ((error = namei(&nd)) != 0)
1634 		return (error);
1635 	vfslocked = NDHASGIANT(&nd);
1636 	NDFREE(&nd, NDF_ONLY_PNBUF);
1637 	vp = nd.ni_vp;
1638 	if (vp->v_type == VDIR) {
1639 		vrele(vp);
1640 		VFS_UNLOCK_GIANT(vfslocked);
1641 		return (EPERM);		/* POSIX */
1642 	}
1643 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) {
1644 		vrele(vp);
1645 		VFS_UNLOCK_GIANT(vfslocked);
1646 		return (error);
1647 	}
1648 	NDINIT_AT(&nd, CREATE, LOCKPARENT | SAVENAME | MPSAFE | AUDITVNODE2,
1649 	    segflg, path2, fd2, td);
1650 	if ((error = namei(&nd)) == 0) {
1651 		lvfslocked = NDHASGIANT(&nd);
1652 		if (nd.ni_vp != NULL) {
1653 			if (nd.ni_dvp == nd.ni_vp)
1654 				vrele(nd.ni_dvp);
1655 			else
1656 				vput(nd.ni_dvp);
1657 			vrele(nd.ni_vp);
1658 			error = EEXIST;
1659 		} else if ((error = vn_lock(vp, LK_EXCLUSIVE | LK_RETRY))
1660 		    == 0) {
1661 			error = can_hardlink(vp, td->td_ucred);
1662 			if (error == 0)
1663 #ifdef MAC
1664 				error = mac_vnode_check_link(td->td_ucred,
1665 				    nd.ni_dvp, vp, &nd.ni_cnd);
1666 			if (error == 0)
1667 #endif
1668 				error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
1669 			VOP_UNLOCK(vp, 0);
1670 			vput(nd.ni_dvp);
1671 		}
1672 		NDFREE(&nd, NDF_ONLY_PNBUF);
1673 		VFS_UNLOCK_GIANT(lvfslocked);
1674 	}
1675 	vrele(vp);
1676 	vn_finished_write(mp);
1677 	VFS_UNLOCK_GIANT(vfslocked);
1678 	return (error);
1679 }
1680 
1681 /*
1682  * Make a symbolic link.
1683  */
1684 #ifndef _SYS_SYSPROTO_H_
1685 struct symlink_args {
1686 	char	*path;
1687 	char	*link;
1688 };
1689 #endif
1690 int
1691 sys_symlink(td, uap)
1692 	struct thread *td;
1693 	register struct symlink_args /* {
1694 		char *path;
1695 		char *link;
1696 	} */ *uap;
1697 {
1698 
1699 	return (kern_symlink(td, uap->path, uap->link, UIO_USERSPACE));
1700 }
1701 
1702 #ifndef _SYS_SYSPROTO_H_
1703 struct symlinkat_args {
1704 	char	*path;
1705 	int	fd;
1706 	char	*path2;
1707 };
1708 #endif
1709 int
1710 sys_symlinkat(struct thread *td, struct symlinkat_args *uap)
1711 {
1712 
1713 	return (kern_symlinkat(td, uap->path1, uap->fd, uap->path2,
1714 	    UIO_USERSPACE));
1715 }
1716 
1717 int
1718 kern_symlink(struct thread *td, char *path, char *link, enum uio_seg segflg)
1719 {
1720 
1721 	return (kern_symlinkat(td, path, AT_FDCWD, link, segflg));
1722 }
1723 
1724 int
1725 kern_symlinkat(struct thread *td, char *path1, int fd, char *path2,
1726     enum uio_seg segflg)
1727 {
1728 	struct mount *mp;
1729 	struct vattr vattr;
1730 	char *syspath;
1731 	int error;
1732 	struct nameidata nd;
1733 	int vfslocked;
1734 
1735 	if (segflg == UIO_SYSSPACE) {
1736 		syspath = path1;
1737 	} else {
1738 		syspath = uma_zalloc(namei_zone, M_WAITOK);
1739 		if ((error = copyinstr(path1, syspath, MAXPATHLEN, NULL)) != 0)
1740 			goto out;
1741 	}
1742 	AUDIT_ARG_TEXT(syspath);
1743 restart:
1744 	bwillwrite();
1745 	NDINIT_AT(&nd, CREATE, LOCKPARENT | SAVENAME | MPSAFE | AUDITVNODE1,
1746 	    segflg, path2, fd, td);
1747 	if ((error = namei(&nd)) != 0)
1748 		goto out;
1749 	vfslocked = NDHASGIANT(&nd);
1750 	if (nd.ni_vp) {
1751 		NDFREE(&nd, NDF_ONLY_PNBUF);
1752 		if (nd.ni_vp == nd.ni_dvp)
1753 			vrele(nd.ni_dvp);
1754 		else
1755 			vput(nd.ni_dvp);
1756 		vrele(nd.ni_vp);
1757 		VFS_UNLOCK_GIANT(vfslocked);
1758 		error = EEXIST;
1759 		goto out;
1760 	}
1761 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1762 		NDFREE(&nd, NDF_ONLY_PNBUF);
1763 		vput(nd.ni_dvp);
1764 		VFS_UNLOCK_GIANT(vfslocked);
1765 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1766 			goto out;
1767 		goto restart;
1768 	}
1769 	VATTR_NULL(&vattr);
1770 	vattr.va_mode = ACCESSPERMS &~ td->td_proc->p_fd->fd_cmask;
1771 #ifdef MAC
1772 	vattr.va_type = VLNK;
1773 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1774 	    &vattr);
1775 	if (error)
1776 		goto out2;
1777 #endif
1778 	error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, syspath);
1779 	if (error == 0)
1780 		vput(nd.ni_vp);
1781 #ifdef MAC
1782 out2:
1783 #endif
1784 	NDFREE(&nd, NDF_ONLY_PNBUF);
1785 	vput(nd.ni_dvp);
1786 	vn_finished_write(mp);
1787 	VFS_UNLOCK_GIANT(vfslocked);
1788 out:
1789 	if (segflg != UIO_SYSSPACE)
1790 		uma_zfree(namei_zone, syspath);
1791 	return (error);
1792 }
1793 
1794 /*
1795  * Delete a whiteout from the filesystem.
1796  */
1797 int
1798 sys_undelete(td, uap)
1799 	struct thread *td;
1800 	register struct undelete_args /* {
1801 		char *path;
1802 	} */ *uap;
1803 {
1804 	int error;
1805 	struct mount *mp;
1806 	struct nameidata nd;
1807 	int vfslocked;
1808 
1809 restart:
1810 	bwillwrite();
1811 	NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | MPSAFE | AUDITVNODE1,
1812 	    UIO_USERSPACE, uap->path, td);
1813 	error = namei(&nd);
1814 	if (error)
1815 		return (error);
1816 	vfslocked = NDHASGIANT(&nd);
1817 
1818 	if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) {
1819 		NDFREE(&nd, NDF_ONLY_PNBUF);
1820 		if (nd.ni_vp == nd.ni_dvp)
1821 			vrele(nd.ni_dvp);
1822 		else
1823 			vput(nd.ni_dvp);
1824 		if (nd.ni_vp)
1825 			vrele(nd.ni_vp);
1826 		VFS_UNLOCK_GIANT(vfslocked);
1827 		return (EEXIST);
1828 	}
1829 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1830 		NDFREE(&nd, NDF_ONLY_PNBUF);
1831 		vput(nd.ni_dvp);
1832 		VFS_UNLOCK_GIANT(vfslocked);
1833 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1834 			return (error);
1835 		goto restart;
1836 	}
1837 	error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE);
1838 	NDFREE(&nd, NDF_ONLY_PNBUF);
1839 	vput(nd.ni_dvp);
1840 	vn_finished_write(mp);
1841 	VFS_UNLOCK_GIANT(vfslocked);
1842 	return (error);
1843 }
1844 
1845 /*
1846  * Delete a name from the filesystem.
1847  */
1848 #ifndef _SYS_SYSPROTO_H_
1849 struct unlink_args {
1850 	char	*path;
1851 };
1852 #endif
1853 int
1854 sys_unlink(td, uap)
1855 	struct thread *td;
1856 	struct unlink_args /* {
1857 		char *path;
1858 	} */ *uap;
1859 {
1860 
1861 	return (kern_unlink(td, uap->path, UIO_USERSPACE));
1862 }
1863 
1864 #ifndef _SYS_SYSPROTO_H_
1865 struct unlinkat_args {
1866 	int	fd;
1867 	char	*path;
1868 	int	flag;
1869 };
1870 #endif
1871 int
1872 sys_unlinkat(struct thread *td, struct unlinkat_args *uap)
1873 {
1874 	int flag = uap->flag;
1875 	int fd = uap->fd;
1876 	char *path = uap->path;
1877 
1878 	if (flag & ~AT_REMOVEDIR)
1879 		return (EINVAL);
1880 
1881 	if (flag & AT_REMOVEDIR)
1882 		return (kern_rmdirat(td, fd, path, UIO_USERSPACE));
1883 	else
1884 		return (kern_unlinkat(td, fd, path, UIO_USERSPACE, 0));
1885 }
1886 
1887 int
1888 kern_unlink(struct thread *td, char *path, enum uio_seg pathseg)
1889 {
1890 
1891 	return (kern_unlinkat(td, AT_FDCWD, path, pathseg, 0));
1892 }
1893 
1894 int
1895 kern_unlinkat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1896     ino_t oldinum)
1897 {
1898 	struct mount *mp;
1899 	struct vnode *vp;
1900 	int error;
1901 	struct nameidata nd;
1902 	struct stat sb;
1903 	int vfslocked;
1904 
1905 restart:
1906 	bwillwrite();
1907 	NDINIT_AT(&nd, DELETE, LOCKPARENT | LOCKLEAF | MPSAFE | AUDITVNODE1,
1908 	    pathseg, path, fd, td);
1909 	if ((error = namei(&nd)) != 0)
1910 		return (error == EINVAL ? EPERM : error);
1911 	vfslocked = NDHASGIANT(&nd);
1912 	vp = nd.ni_vp;
1913 	if (vp->v_type == VDIR && oldinum == 0) {
1914 		error = EPERM;		/* POSIX */
1915 	} else if (oldinum != 0 &&
1916 		  ((error = vn_stat(vp, &sb, td->td_ucred, NOCRED, td)) == 0) &&
1917 		  sb.st_ino != oldinum) {
1918 			error = EIDRM;	/* Identifier removed */
1919 	} else {
1920 		/*
1921 		 * The root of a mounted filesystem cannot be deleted.
1922 		 *
1923 		 * XXX: can this only be a VDIR case?
1924 		 */
1925 		if (vp->v_vflag & VV_ROOT)
1926 			error = EBUSY;
1927 	}
1928 	if (error == 0) {
1929 		if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1930 			NDFREE(&nd, NDF_ONLY_PNBUF);
1931 			vput(nd.ni_dvp);
1932 			if (vp == nd.ni_dvp)
1933 				vrele(vp);
1934 			else
1935 				vput(vp);
1936 			VFS_UNLOCK_GIANT(vfslocked);
1937 			if ((error = vn_start_write(NULL, &mp,
1938 			    V_XSLEEP | PCATCH)) != 0)
1939 				return (error);
1940 			goto restart;
1941 		}
1942 #ifdef MAC
1943 		error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
1944 		    &nd.ni_cnd);
1945 		if (error)
1946 			goto out;
1947 #endif
1948 		error = VOP_REMOVE(nd.ni_dvp, vp, &nd.ni_cnd);
1949 #ifdef MAC
1950 out:
1951 #endif
1952 		vn_finished_write(mp);
1953 	}
1954 	NDFREE(&nd, NDF_ONLY_PNBUF);
1955 	vput(nd.ni_dvp);
1956 	if (vp == nd.ni_dvp)
1957 		vrele(vp);
1958 	else
1959 		vput(vp);
1960 	VFS_UNLOCK_GIANT(vfslocked);
1961 	return (error);
1962 }
1963 
1964 /*
1965  * Reposition read/write file offset.
1966  */
1967 #ifndef _SYS_SYSPROTO_H_
1968 struct lseek_args {
1969 	int	fd;
1970 	int	pad;
1971 	off_t	offset;
1972 	int	whence;
1973 };
1974 #endif
1975 int
1976 sys_lseek(td, uap)
1977 	struct thread *td;
1978 	register struct lseek_args /* {
1979 		int fd;
1980 		int pad;
1981 		off_t offset;
1982 		int whence;
1983 	} */ *uap;
1984 {
1985 	struct ucred *cred = td->td_ucred;
1986 	struct file *fp;
1987 	struct vnode *vp;
1988 	struct vattr vattr;
1989 	off_t offset, size;
1990 	int error, noneg;
1991 	int vfslocked;
1992 
1993 	AUDIT_ARG_FD(uap->fd);
1994 	if ((error = fget(td, uap->fd, CAP_SEEK, &fp)) != 0)
1995 		return (error);
1996 	if (!(fp->f_ops->fo_flags & DFLAG_SEEKABLE)) {
1997 		fdrop(fp, td);
1998 		return (ESPIPE);
1999 	}
2000 	vp = fp->f_vnode;
2001 	vfslocked = VFS_LOCK_GIANT(vp->v_mount);
2002 	noneg = (vp->v_type != VCHR);
2003 	offset = uap->offset;
2004 	switch (uap->whence) {
2005 	case L_INCR:
2006 		if (noneg &&
2007 		    (fp->f_offset < 0 ||
2008 		    (offset > 0 && fp->f_offset > OFF_MAX - offset))) {
2009 			error = EOVERFLOW;
2010 			break;
2011 		}
2012 		offset += fp->f_offset;
2013 		break;
2014 	case L_XTND:
2015 		vn_lock(vp, LK_SHARED | LK_RETRY);
2016 		error = VOP_GETATTR(vp, &vattr, cred);
2017 		VOP_UNLOCK(vp, 0);
2018 		if (error)
2019 			break;
2020 
2021 		/*
2022 		 * If the file references a disk device, then fetch
2023 		 * the media size and use that to determine the ending
2024 		 * offset.
2025 		 */
2026 		if (vattr.va_size == 0 && vp->v_type == VCHR &&
2027 		    fo_ioctl(fp, DIOCGMEDIASIZE, &size, cred, td) == 0)
2028 			vattr.va_size = size;
2029 		if (noneg &&
2030 		    (vattr.va_size > OFF_MAX ||
2031 		    (offset > 0 && vattr.va_size > OFF_MAX - offset))) {
2032 			error = EOVERFLOW;
2033 			break;
2034 		}
2035 		offset += vattr.va_size;
2036 		break;
2037 	case L_SET:
2038 		break;
2039 	case SEEK_DATA:
2040 		error = fo_ioctl(fp, FIOSEEKDATA, &offset, cred, td);
2041 		break;
2042 	case SEEK_HOLE:
2043 		error = fo_ioctl(fp, FIOSEEKHOLE, &offset, cred, td);
2044 		break;
2045 	default:
2046 		error = EINVAL;
2047 	}
2048 	if (error == 0 && noneg && offset < 0)
2049 		error = EINVAL;
2050 	if (error != 0)
2051 		goto drop;
2052 	fp->f_offset = offset;
2053 	VFS_KNOTE_UNLOCKED(vp, 0);
2054 	*(off_t *)(td->td_retval) = fp->f_offset;
2055 drop:
2056 	fdrop(fp, td);
2057 	VFS_UNLOCK_GIANT(vfslocked);
2058 	return (error);
2059 }
2060 
2061 #if defined(COMPAT_43)
2062 /*
2063  * Reposition read/write file offset.
2064  */
2065 #ifndef _SYS_SYSPROTO_H_
2066 struct olseek_args {
2067 	int	fd;
2068 	long	offset;
2069 	int	whence;
2070 };
2071 #endif
2072 int
2073 olseek(td, uap)
2074 	struct thread *td;
2075 	register struct olseek_args /* {
2076 		int fd;
2077 		long offset;
2078 		int whence;
2079 	} */ *uap;
2080 {
2081 	struct lseek_args /* {
2082 		int fd;
2083 		int pad;
2084 		off_t offset;
2085 		int whence;
2086 	} */ nuap;
2087 
2088 	nuap.fd = uap->fd;
2089 	nuap.offset = uap->offset;
2090 	nuap.whence = uap->whence;
2091 	return (sys_lseek(td, &nuap));
2092 }
2093 #endif /* COMPAT_43 */
2094 
2095 /* Version with the 'pad' argument */
2096 int
2097 freebsd6_lseek(td, uap)
2098 	struct thread *td;
2099 	register struct freebsd6_lseek_args *uap;
2100 {
2101 	struct lseek_args ouap;
2102 
2103 	ouap.fd = uap->fd;
2104 	ouap.offset = uap->offset;
2105 	ouap.whence = uap->whence;
2106 	return (sys_lseek(td, &ouap));
2107 }
2108 
2109 /*
2110  * Check access permissions using passed credentials.
2111  */
2112 static int
2113 vn_access(vp, user_flags, cred, td)
2114 	struct vnode	*vp;
2115 	int		user_flags;
2116 	struct ucred	*cred;
2117 	struct thread	*td;
2118 {
2119 	int error;
2120 	accmode_t accmode;
2121 
2122 	/* Flags == 0 means only check for existence. */
2123 	error = 0;
2124 	if (user_flags) {
2125 		accmode = 0;
2126 		if (user_flags & R_OK)
2127 			accmode |= VREAD;
2128 		if (user_flags & W_OK)
2129 			accmode |= VWRITE;
2130 		if (user_flags & X_OK)
2131 			accmode |= VEXEC;
2132 #ifdef MAC
2133 		error = mac_vnode_check_access(cred, vp, accmode);
2134 		if (error)
2135 			return (error);
2136 #endif
2137 		if ((accmode & VWRITE) == 0 || (error = vn_writechk(vp)) == 0)
2138 			error = VOP_ACCESS(vp, accmode, cred, td);
2139 	}
2140 	return (error);
2141 }
2142 
2143 /*
2144  * Check access permissions using "real" credentials.
2145  */
2146 #ifndef _SYS_SYSPROTO_H_
2147 struct access_args {
2148 	char	*path;
2149 	int	amode;
2150 };
2151 #endif
2152 int
2153 sys_access(td, uap)
2154 	struct thread *td;
2155 	register struct access_args /* {
2156 		char *path;
2157 		int amode;
2158 	} */ *uap;
2159 {
2160 
2161 	return (kern_access(td, uap->path, UIO_USERSPACE, uap->amode));
2162 }
2163 
2164 #ifndef _SYS_SYSPROTO_H_
2165 struct faccessat_args {
2166 	int	dirfd;
2167 	char	*path;
2168 	int	amode;
2169 	int	flag;
2170 }
2171 #endif
2172 int
2173 sys_faccessat(struct thread *td, struct faccessat_args *uap)
2174 {
2175 
2176 	if (uap->flag & ~AT_EACCESS)
2177 		return (EINVAL);
2178 	return (kern_accessat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
2179 	    uap->amode));
2180 }
2181 
2182 int
2183 kern_access(struct thread *td, char *path, enum uio_seg pathseg, int amode)
2184 {
2185 
2186 	return (kern_accessat(td, AT_FDCWD, path, pathseg, 0, amode));
2187 }
2188 
2189 int
2190 kern_accessat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2191     int flag, int amode)
2192 {
2193 	struct ucred *cred, *tmpcred;
2194 	struct vnode *vp;
2195 	struct nameidata nd;
2196 	int vfslocked;
2197 	int error;
2198 
2199 	/*
2200 	 * Create and modify a temporary credential instead of one that
2201 	 * is potentially shared.
2202 	 */
2203 	if (!(flag & AT_EACCESS)) {
2204 		cred = td->td_ucred;
2205 		tmpcred = crdup(cred);
2206 		tmpcred->cr_uid = cred->cr_ruid;
2207 		tmpcred->cr_groups[0] = cred->cr_rgid;
2208 		td->td_ucred = tmpcred;
2209 	} else
2210 		cred = tmpcred = td->td_ucred;
2211 	AUDIT_ARG_VALUE(amode);
2212 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | MPSAFE |
2213 	    AUDITVNODE1, pathseg, path, fd, CAP_FSTAT, td);
2214 	if ((error = namei(&nd)) != 0)
2215 		goto out1;
2216 	vfslocked = NDHASGIANT(&nd);
2217 	vp = nd.ni_vp;
2218 
2219 	error = vn_access(vp, amode, tmpcred, td);
2220 	NDFREE(&nd, NDF_ONLY_PNBUF);
2221 	vput(vp);
2222 	VFS_UNLOCK_GIANT(vfslocked);
2223 out1:
2224 	if (!(flag & AT_EACCESS)) {
2225 		td->td_ucred = cred;
2226 		crfree(tmpcred);
2227 	}
2228 	return (error);
2229 }
2230 
2231 /*
2232  * Check access permissions using "effective" credentials.
2233  */
2234 #ifndef _SYS_SYSPROTO_H_
2235 struct eaccess_args {
2236 	char	*path;
2237 	int	amode;
2238 };
2239 #endif
2240 int
2241 sys_eaccess(td, uap)
2242 	struct thread *td;
2243 	register struct eaccess_args /* {
2244 		char *path;
2245 		int amode;
2246 	} */ *uap;
2247 {
2248 
2249 	return (kern_eaccess(td, uap->path, UIO_USERSPACE, uap->amode));
2250 }
2251 
2252 int
2253 kern_eaccess(struct thread *td, char *path, enum uio_seg pathseg, int amode)
2254 {
2255 
2256 	return (kern_accessat(td, AT_FDCWD, path, pathseg, AT_EACCESS, amode));
2257 }
2258 
2259 #if defined(COMPAT_43)
2260 /*
2261  * Get file status; this version follows links.
2262  */
2263 #ifndef _SYS_SYSPROTO_H_
2264 struct ostat_args {
2265 	char	*path;
2266 	struct ostat *ub;
2267 };
2268 #endif
2269 int
2270 ostat(td, uap)
2271 	struct thread *td;
2272 	register struct ostat_args /* {
2273 		char *path;
2274 		struct ostat *ub;
2275 	} */ *uap;
2276 {
2277 	struct stat sb;
2278 	struct ostat osb;
2279 	int error;
2280 
2281 	error = kern_stat(td, uap->path, UIO_USERSPACE, &sb);
2282 	if (error)
2283 		return (error);
2284 	cvtstat(&sb, &osb);
2285 	error = copyout(&osb, uap->ub, sizeof (osb));
2286 	return (error);
2287 }
2288 
2289 /*
2290  * Get file status; this version does not follow links.
2291  */
2292 #ifndef _SYS_SYSPROTO_H_
2293 struct olstat_args {
2294 	char	*path;
2295 	struct ostat *ub;
2296 };
2297 #endif
2298 int
2299 olstat(td, uap)
2300 	struct thread *td;
2301 	register struct olstat_args /* {
2302 		char *path;
2303 		struct ostat *ub;
2304 	} */ *uap;
2305 {
2306 	struct stat sb;
2307 	struct ostat osb;
2308 	int error;
2309 
2310 	error = kern_lstat(td, uap->path, UIO_USERSPACE, &sb);
2311 	if (error)
2312 		return (error);
2313 	cvtstat(&sb, &osb);
2314 	error = copyout(&osb, uap->ub, sizeof (osb));
2315 	return (error);
2316 }
2317 
2318 /*
2319  * Convert from an old to a new stat structure.
2320  */
2321 void
2322 cvtstat(st, ost)
2323 	struct stat *st;
2324 	struct ostat *ost;
2325 {
2326 
2327 	ost->st_dev = st->st_dev;
2328 	ost->st_ino = st->st_ino;
2329 	ost->st_mode = st->st_mode;
2330 	ost->st_nlink = st->st_nlink;
2331 	ost->st_uid = st->st_uid;
2332 	ost->st_gid = st->st_gid;
2333 	ost->st_rdev = st->st_rdev;
2334 	if (st->st_size < (quad_t)1 << 32)
2335 		ost->st_size = st->st_size;
2336 	else
2337 		ost->st_size = -2;
2338 	ost->st_atim = st->st_atim;
2339 	ost->st_mtim = st->st_mtim;
2340 	ost->st_ctim = st->st_ctim;
2341 	ost->st_blksize = st->st_blksize;
2342 	ost->st_blocks = st->st_blocks;
2343 	ost->st_flags = st->st_flags;
2344 	ost->st_gen = st->st_gen;
2345 }
2346 #endif /* COMPAT_43 */
2347 
2348 /*
2349  * Get file status; this version follows links.
2350  */
2351 #ifndef _SYS_SYSPROTO_H_
2352 struct stat_args {
2353 	char	*path;
2354 	struct stat *ub;
2355 };
2356 #endif
2357 int
2358 sys_stat(td, uap)
2359 	struct thread *td;
2360 	register struct stat_args /* {
2361 		char *path;
2362 		struct stat *ub;
2363 	} */ *uap;
2364 {
2365 	struct stat sb;
2366 	int error;
2367 
2368 	error = kern_stat(td, uap->path, UIO_USERSPACE, &sb);
2369 	if (error == 0)
2370 		error = copyout(&sb, uap->ub, sizeof (sb));
2371 	return (error);
2372 }
2373 
2374 #ifndef _SYS_SYSPROTO_H_
2375 struct fstatat_args {
2376 	int	fd;
2377 	char	*path;
2378 	struct stat	*buf;
2379 	int	flag;
2380 }
2381 #endif
2382 int
2383 sys_fstatat(struct thread *td, struct fstatat_args *uap)
2384 {
2385 	struct stat sb;
2386 	int error;
2387 
2388 	error = kern_statat(td, uap->flag, uap->fd, uap->path,
2389 	    UIO_USERSPACE, &sb);
2390 	if (error == 0)
2391 		error = copyout(&sb, uap->buf, sizeof (sb));
2392 	return (error);
2393 }
2394 
2395 int
2396 kern_stat(struct thread *td, char *path, enum uio_seg pathseg, struct stat *sbp)
2397 {
2398 
2399 	return (kern_statat(td, 0, AT_FDCWD, path, pathseg, sbp));
2400 }
2401 
2402 int
2403 kern_statat(struct thread *td, int flag, int fd, char *path,
2404     enum uio_seg pathseg, struct stat *sbp)
2405 {
2406 
2407 	return (kern_statat_vnhook(td, flag, fd, path, pathseg, sbp, NULL));
2408 }
2409 
2410 int
2411 kern_statat_vnhook(struct thread *td, int flag, int fd, char *path,
2412     enum uio_seg pathseg, struct stat *sbp,
2413     void (*hook)(struct vnode *vp, struct stat *sbp))
2414 {
2415 	struct nameidata nd;
2416 	struct stat sb;
2417 	int error, vfslocked;
2418 
2419 	if (flag & ~AT_SYMLINK_NOFOLLOW)
2420 		return (EINVAL);
2421 
2422 	NDINIT_ATRIGHTS(&nd, LOOKUP, ((flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW :
2423 	    FOLLOW) | LOCKSHARED | LOCKLEAF | AUDITVNODE1 | MPSAFE, pathseg,
2424 	    path, fd, CAP_FSTAT, td);
2425 
2426 	if ((error = namei(&nd)) != 0)
2427 		return (error);
2428 	vfslocked = NDHASGIANT(&nd);
2429 	error = vn_stat(nd.ni_vp, &sb, td->td_ucred, NOCRED, td);
2430 	if (!error) {
2431 		SDT_PROBE(vfs, , stat, mode, path, sb.st_mode, 0, 0, 0);
2432 		if (S_ISREG(sb.st_mode))
2433 			SDT_PROBE(vfs, , stat, reg, path, pathseg, 0, 0, 0);
2434 		if (__predict_false(hook != NULL))
2435 			hook(nd.ni_vp, &sb);
2436 	}
2437 	NDFREE(&nd, NDF_ONLY_PNBUF);
2438 	vput(nd.ni_vp);
2439 	VFS_UNLOCK_GIANT(vfslocked);
2440 	if (error)
2441 		return (error);
2442 	*sbp = sb;
2443 #ifdef KTRACE
2444 	if (KTRPOINT(td, KTR_STRUCT))
2445 		ktrstat(&sb);
2446 #endif
2447 	return (0);
2448 }
2449 
2450 /*
2451  * Get file status; this version does not follow links.
2452  */
2453 #ifndef _SYS_SYSPROTO_H_
2454 struct lstat_args {
2455 	char	*path;
2456 	struct stat *ub;
2457 };
2458 #endif
2459 int
2460 sys_lstat(td, uap)
2461 	struct thread *td;
2462 	register struct lstat_args /* {
2463 		char *path;
2464 		struct stat *ub;
2465 	} */ *uap;
2466 {
2467 	struct stat sb;
2468 	int error;
2469 
2470 	error = kern_lstat(td, uap->path, UIO_USERSPACE, &sb);
2471 	if (error == 0)
2472 		error = copyout(&sb, uap->ub, sizeof (sb));
2473 	return (error);
2474 }
2475 
2476 int
2477 kern_lstat(struct thread *td, char *path, enum uio_seg pathseg, struct stat *sbp)
2478 {
2479 
2480 	return (kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, path, pathseg,
2481 	    sbp));
2482 }
2483 
2484 /*
2485  * Implementation of the NetBSD [l]stat() functions.
2486  */
2487 void
2488 cvtnstat(sb, nsb)
2489 	struct stat *sb;
2490 	struct nstat *nsb;
2491 {
2492 	bzero(nsb, sizeof *nsb);
2493 	nsb->st_dev = sb->st_dev;
2494 	nsb->st_ino = sb->st_ino;
2495 	nsb->st_mode = sb->st_mode;
2496 	nsb->st_nlink = sb->st_nlink;
2497 	nsb->st_uid = sb->st_uid;
2498 	nsb->st_gid = sb->st_gid;
2499 	nsb->st_rdev = sb->st_rdev;
2500 	nsb->st_atim = sb->st_atim;
2501 	nsb->st_mtim = sb->st_mtim;
2502 	nsb->st_ctim = sb->st_ctim;
2503 	nsb->st_size = sb->st_size;
2504 	nsb->st_blocks = sb->st_blocks;
2505 	nsb->st_blksize = sb->st_blksize;
2506 	nsb->st_flags = sb->st_flags;
2507 	nsb->st_gen = sb->st_gen;
2508 	nsb->st_birthtim = sb->st_birthtim;
2509 }
2510 
2511 #ifndef _SYS_SYSPROTO_H_
2512 struct nstat_args {
2513 	char	*path;
2514 	struct nstat *ub;
2515 };
2516 #endif
2517 int
2518 sys_nstat(td, uap)
2519 	struct thread *td;
2520 	register struct nstat_args /* {
2521 		char *path;
2522 		struct nstat *ub;
2523 	} */ *uap;
2524 {
2525 	struct stat sb;
2526 	struct nstat nsb;
2527 	int error;
2528 
2529 	error = kern_stat(td, uap->path, UIO_USERSPACE, &sb);
2530 	if (error)
2531 		return (error);
2532 	cvtnstat(&sb, &nsb);
2533 	error = copyout(&nsb, uap->ub, sizeof (nsb));
2534 	return (error);
2535 }
2536 
2537 /*
2538  * NetBSD lstat.  Get file status; this version does not follow links.
2539  */
2540 #ifndef _SYS_SYSPROTO_H_
2541 struct lstat_args {
2542 	char	*path;
2543 	struct stat *ub;
2544 };
2545 #endif
2546 int
2547 sys_nlstat(td, uap)
2548 	struct thread *td;
2549 	register struct nlstat_args /* {
2550 		char *path;
2551 		struct nstat *ub;
2552 	} */ *uap;
2553 {
2554 	struct stat sb;
2555 	struct nstat nsb;
2556 	int error;
2557 
2558 	error = kern_lstat(td, uap->path, UIO_USERSPACE, &sb);
2559 	if (error)
2560 		return (error);
2561 	cvtnstat(&sb, &nsb);
2562 	error = copyout(&nsb, uap->ub, sizeof (nsb));
2563 	return (error);
2564 }
2565 
2566 /*
2567  * Get configurable pathname variables.
2568  */
2569 #ifndef _SYS_SYSPROTO_H_
2570 struct pathconf_args {
2571 	char	*path;
2572 	int	name;
2573 };
2574 #endif
2575 int
2576 sys_pathconf(td, uap)
2577 	struct thread *td;
2578 	register struct pathconf_args /* {
2579 		char *path;
2580 		int name;
2581 	} */ *uap;
2582 {
2583 
2584 	return (kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name, FOLLOW));
2585 }
2586 
2587 #ifndef _SYS_SYSPROTO_H_
2588 struct lpathconf_args {
2589 	char	*path;
2590 	int	name;
2591 };
2592 #endif
2593 int
2594 sys_lpathconf(td, uap)
2595 	struct thread *td;
2596 	register struct lpathconf_args /* {
2597 		char *path;
2598 		int name;
2599 	} */ *uap;
2600 {
2601 
2602 	return (kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name, NOFOLLOW));
2603 }
2604 
2605 int
2606 kern_pathconf(struct thread *td, char *path, enum uio_seg pathseg, int name,
2607     u_long flags)
2608 {
2609 	struct nameidata nd;
2610 	int error, vfslocked;
2611 
2612 	NDINIT(&nd, LOOKUP, LOCKSHARED | LOCKLEAF | MPSAFE | AUDITVNODE1 |
2613 	    flags, pathseg, path, td);
2614 	if ((error = namei(&nd)) != 0)
2615 		return (error);
2616 	vfslocked = NDHASGIANT(&nd);
2617 	NDFREE(&nd, NDF_ONLY_PNBUF);
2618 
2619 	/* If asynchronous I/O is available, it works for all files. */
2620 	if (name == _PC_ASYNC_IO)
2621 		td->td_retval[0] = async_io_version;
2622 	else
2623 		error = VOP_PATHCONF(nd.ni_vp, name, td->td_retval);
2624 	vput(nd.ni_vp);
2625 	VFS_UNLOCK_GIANT(vfslocked);
2626 	return (error);
2627 }
2628 
2629 /*
2630  * Return target name of a symbolic link.
2631  */
2632 #ifndef _SYS_SYSPROTO_H_
2633 struct readlink_args {
2634 	char	*path;
2635 	char	*buf;
2636 	size_t	count;
2637 };
2638 #endif
2639 int
2640 sys_readlink(td, uap)
2641 	struct thread *td;
2642 	register struct readlink_args /* {
2643 		char *path;
2644 		char *buf;
2645 		size_t count;
2646 	} */ *uap;
2647 {
2648 
2649 	return (kern_readlink(td, uap->path, UIO_USERSPACE, uap->buf,
2650 	    UIO_USERSPACE, uap->count));
2651 }
2652 #ifndef _SYS_SYSPROTO_H_
2653 struct readlinkat_args {
2654 	int	fd;
2655 	char	*path;
2656 	char	*buf;
2657 	size_t	bufsize;
2658 };
2659 #endif
2660 int
2661 sys_readlinkat(struct thread *td, struct readlinkat_args *uap)
2662 {
2663 
2664 	return (kern_readlinkat(td, uap->fd, uap->path, UIO_USERSPACE,
2665 	    uap->buf, UIO_USERSPACE, uap->bufsize));
2666 }
2667 
2668 int
2669 kern_readlink(struct thread *td, char *path, enum uio_seg pathseg, char *buf,
2670     enum uio_seg bufseg, size_t count)
2671 {
2672 
2673 	return (kern_readlinkat(td, AT_FDCWD, path, pathseg, buf, bufseg,
2674 	    count));
2675 }
2676 
2677 int
2678 kern_readlinkat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2679     char *buf, enum uio_seg bufseg, size_t count)
2680 {
2681 	struct vnode *vp;
2682 	struct iovec aiov;
2683 	struct uio auio;
2684 	int error;
2685 	struct nameidata nd;
2686 	int vfslocked;
2687 
2688 	if (count > INT_MAX)
2689 		return (EINVAL);
2690 
2691 	NDINIT_AT(&nd, LOOKUP, NOFOLLOW | LOCKSHARED | LOCKLEAF | MPSAFE |
2692 	    AUDITVNODE1, pathseg, path, fd, td);
2693 
2694 	if ((error = namei(&nd)) != 0)
2695 		return (error);
2696 	NDFREE(&nd, NDF_ONLY_PNBUF);
2697 	vfslocked = NDHASGIANT(&nd);
2698 	vp = nd.ni_vp;
2699 #ifdef MAC
2700 	error = mac_vnode_check_readlink(td->td_ucred, vp);
2701 	if (error) {
2702 		vput(vp);
2703 		VFS_UNLOCK_GIANT(vfslocked);
2704 		return (error);
2705 	}
2706 #endif
2707 	if (vp->v_type != VLNK)
2708 		error = EINVAL;
2709 	else {
2710 		aiov.iov_base = buf;
2711 		aiov.iov_len = count;
2712 		auio.uio_iov = &aiov;
2713 		auio.uio_iovcnt = 1;
2714 		auio.uio_offset = 0;
2715 		auio.uio_rw = UIO_READ;
2716 		auio.uio_segflg = bufseg;
2717 		auio.uio_td = td;
2718 		auio.uio_resid = count;
2719 		error = VOP_READLINK(vp, &auio, td->td_ucred);
2720 	}
2721 	vput(vp);
2722 	VFS_UNLOCK_GIANT(vfslocked);
2723 	td->td_retval[0] = count - auio.uio_resid;
2724 	return (error);
2725 }
2726 
2727 /*
2728  * Common implementation code for chflags() and fchflags().
2729  */
2730 static int
2731 setfflags(td, vp, flags)
2732 	struct thread *td;
2733 	struct vnode *vp;
2734 	int flags;
2735 {
2736 	int error;
2737 	struct mount *mp;
2738 	struct vattr vattr;
2739 
2740 	/*
2741 	 * Prevent non-root users from setting flags on devices.  When
2742 	 * a device is reused, users can retain ownership of the device
2743 	 * if they are allowed to set flags and programs assume that
2744 	 * chown can't fail when done as root.
2745 	 */
2746 	if (vp->v_type == VCHR || vp->v_type == VBLK) {
2747 		error = priv_check(td, PRIV_VFS_CHFLAGS_DEV);
2748 		if (error)
2749 			return (error);
2750 	}
2751 
2752 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2753 		return (error);
2754 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2755 	VATTR_NULL(&vattr);
2756 	vattr.va_flags = flags;
2757 #ifdef MAC
2758 	error = mac_vnode_check_setflags(td->td_ucred, vp, vattr.va_flags);
2759 	if (error == 0)
2760 #endif
2761 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
2762 	VOP_UNLOCK(vp, 0);
2763 	vn_finished_write(mp);
2764 	return (error);
2765 }
2766 
2767 /*
2768  * Change flags of a file given a path name.
2769  */
2770 #ifndef _SYS_SYSPROTO_H_
2771 struct chflags_args {
2772 	char	*path;
2773 	int	flags;
2774 };
2775 #endif
2776 int
2777 sys_chflags(td, uap)
2778 	struct thread *td;
2779 	register struct chflags_args /* {
2780 		char *path;
2781 		int flags;
2782 	} */ *uap;
2783 {
2784 	int error;
2785 	struct nameidata nd;
2786 	int vfslocked;
2787 
2788 	AUDIT_ARG_FFLAGS(uap->flags);
2789 	NDINIT(&nd, LOOKUP, FOLLOW | MPSAFE | AUDITVNODE1, UIO_USERSPACE,
2790 	    uap->path, td);
2791 	if ((error = namei(&nd)) != 0)
2792 		return (error);
2793 	NDFREE(&nd, NDF_ONLY_PNBUF);
2794 	vfslocked = NDHASGIANT(&nd);
2795 	error = setfflags(td, nd.ni_vp, uap->flags);
2796 	vrele(nd.ni_vp);
2797 	VFS_UNLOCK_GIANT(vfslocked);
2798 	return (error);
2799 }
2800 
2801 /*
2802  * Same as chflags() but doesn't follow symlinks.
2803  */
2804 int
2805 sys_lchflags(td, uap)
2806 	struct thread *td;
2807 	register struct lchflags_args /* {
2808 		char *path;
2809 		int flags;
2810 	} */ *uap;
2811 {
2812 	int error;
2813 	struct nameidata nd;
2814 	int vfslocked;
2815 
2816 	AUDIT_ARG_FFLAGS(uap->flags);
2817 	NDINIT(&nd, LOOKUP, NOFOLLOW | MPSAFE | AUDITVNODE1, UIO_USERSPACE,
2818 	    uap->path, td);
2819 	if ((error = namei(&nd)) != 0)
2820 		return (error);
2821 	vfslocked = NDHASGIANT(&nd);
2822 	NDFREE(&nd, NDF_ONLY_PNBUF);
2823 	error = setfflags(td, nd.ni_vp, uap->flags);
2824 	vrele(nd.ni_vp);
2825 	VFS_UNLOCK_GIANT(vfslocked);
2826 	return (error);
2827 }
2828 
2829 /*
2830  * Change flags of a file given a file descriptor.
2831  */
2832 #ifndef _SYS_SYSPROTO_H_
2833 struct fchflags_args {
2834 	int	fd;
2835 	int	flags;
2836 };
2837 #endif
2838 int
2839 sys_fchflags(td, uap)
2840 	struct thread *td;
2841 	register struct fchflags_args /* {
2842 		int fd;
2843 		int flags;
2844 	} */ *uap;
2845 {
2846 	struct file *fp;
2847 	int vfslocked;
2848 	int error;
2849 
2850 	AUDIT_ARG_FD(uap->fd);
2851 	AUDIT_ARG_FFLAGS(uap->flags);
2852 	if ((error = getvnode(td->td_proc->p_fd, uap->fd, CAP_FCHFLAGS,
2853 	    &fp)) != 0)
2854 		return (error);
2855 	vfslocked = VFS_LOCK_GIANT(fp->f_vnode->v_mount);
2856 #ifdef AUDIT
2857 	vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
2858 	AUDIT_ARG_VNODE1(fp->f_vnode);
2859 	VOP_UNLOCK(fp->f_vnode, 0);
2860 #endif
2861 	error = setfflags(td, fp->f_vnode, uap->flags);
2862 	VFS_UNLOCK_GIANT(vfslocked);
2863 	fdrop(fp, td);
2864 	return (error);
2865 }
2866 
2867 /*
2868  * Common implementation code for chmod(), lchmod() and fchmod().
2869  */
2870 int
2871 setfmode(td, cred, vp, mode)
2872 	struct thread *td;
2873 	struct ucred *cred;
2874 	struct vnode *vp;
2875 	int mode;
2876 {
2877 	int error;
2878 	struct mount *mp;
2879 	struct vattr vattr;
2880 
2881 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2882 		return (error);
2883 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2884 	VATTR_NULL(&vattr);
2885 	vattr.va_mode = mode & ALLPERMS;
2886 #ifdef MAC
2887 	error = mac_vnode_check_setmode(cred, vp, vattr.va_mode);
2888 	if (error == 0)
2889 #endif
2890 		error = VOP_SETATTR(vp, &vattr, cred);
2891 	VOP_UNLOCK(vp, 0);
2892 	vn_finished_write(mp);
2893 	return (error);
2894 }
2895 
2896 /*
2897  * Change mode of a file given path name.
2898  */
2899 #ifndef _SYS_SYSPROTO_H_
2900 struct chmod_args {
2901 	char	*path;
2902 	int	mode;
2903 };
2904 #endif
2905 int
2906 sys_chmod(td, uap)
2907 	struct thread *td;
2908 	register struct chmod_args /* {
2909 		char *path;
2910 		int mode;
2911 	} */ *uap;
2912 {
2913 
2914 	return (kern_chmod(td, uap->path, UIO_USERSPACE, uap->mode));
2915 }
2916 
2917 #ifndef _SYS_SYSPROTO_H_
2918 struct fchmodat_args {
2919 	int	dirfd;
2920 	char	*path;
2921 	mode_t	mode;
2922 	int	flag;
2923 }
2924 #endif
2925 int
2926 sys_fchmodat(struct thread *td, struct fchmodat_args *uap)
2927 {
2928 	int flag = uap->flag;
2929 	int fd = uap->fd;
2930 	char *path = uap->path;
2931 	mode_t mode = uap->mode;
2932 
2933 	if (flag & ~AT_SYMLINK_NOFOLLOW)
2934 		return (EINVAL);
2935 
2936 	return (kern_fchmodat(td, fd, path, UIO_USERSPACE, mode, flag));
2937 }
2938 
2939 int
2940 kern_chmod(struct thread *td, char *path, enum uio_seg pathseg, int mode)
2941 {
2942 
2943 	return (kern_fchmodat(td, AT_FDCWD, path, pathseg, mode, 0));
2944 }
2945 
2946 /*
2947  * Change mode of a file given path name (don't follow links.)
2948  */
2949 #ifndef _SYS_SYSPROTO_H_
2950 struct lchmod_args {
2951 	char	*path;
2952 	int	mode;
2953 };
2954 #endif
2955 int
2956 sys_lchmod(td, uap)
2957 	struct thread *td;
2958 	register struct lchmod_args /* {
2959 		char *path;
2960 		int mode;
2961 	} */ *uap;
2962 {
2963 
2964 	return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2965 	    uap->mode, AT_SYMLINK_NOFOLLOW));
2966 }
2967 
2968 
2969 int
2970 kern_fchmodat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2971     mode_t mode, int flag)
2972 {
2973 	int error;
2974 	struct nameidata nd;
2975 	int vfslocked;
2976 	int follow;
2977 
2978 	AUDIT_ARG_MODE(mode);
2979 	follow = (flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
2980 	NDINIT_ATRIGHTS(&nd, LOOKUP,  follow | MPSAFE | AUDITVNODE1, pathseg,
2981 	    path, fd, CAP_FCHMOD, td);
2982 	if ((error = namei(&nd)) != 0)
2983 		return (error);
2984 	vfslocked = NDHASGIANT(&nd);
2985 	NDFREE(&nd, NDF_ONLY_PNBUF);
2986 	error = setfmode(td, td->td_ucred, nd.ni_vp, mode);
2987 	vrele(nd.ni_vp);
2988 	VFS_UNLOCK_GIANT(vfslocked);
2989 	return (error);
2990 }
2991 
2992 /*
2993  * Change mode of a file given a file descriptor.
2994  */
2995 #ifndef _SYS_SYSPROTO_H_
2996 struct fchmod_args {
2997 	int	fd;
2998 	int	mode;
2999 };
3000 #endif
3001 int
3002 sys_fchmod(struct thread *td, struct fchmod_args *uap)
3003 {
3004 	struct file *fp;
3005 	int error;
3006 
3007 	AUDIT_ARG_FD(uap->fd);
3008 	AUDIT_ARG_MODE(uap->mode);
3009 
3010 	error = fget(td, uap->fd, CAP_FCHMOD, &fp);
3011 	if (error != 0)
3012 		return (error);
3013 	error = fo_chmod(fp, uap->mode, td->td_ucred, td);
3014 	fdrop(fp, td);
3015 	return (error);
3016 }
3017 
3018 /*
3019  * Common implementation for chown(), lchown(), and fchown()
3020  */
3021 int
3022 setfown(td, cred, vp, uid, gid)
3023 	struct thread *td;
3024 	struct ucred *cred;
3025 	struct vnode *vp;
3026 	uid_t uid;
3027 	gid_t gid;
3028 {
3029 	int error;
3030 	struct mount *mp;
3031 	struct vattr vattr;
3032 
3033 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
3034 		return (error);
3035 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3036 	VATTR_NULL(&vattr);
3037 	vattr.va_uid = uid;
3038 	vattr.va_gid = gid;
3039 #ifdef MAC
3040 	error = mac_vnode_check_setowner(cred, vp, vattr.va_uid,
3041 	    vattr.va_gid);
3042 	if (error == 0)
3043 #endif
3044 		error = VOP_SETATTR(vp, &vattr, cred);
3045 	VOP_UNLOCK(vp, 0);
3046 	vn_finished_write(mp);
3047 	return (error);
3048 }
3049 
3050 /*
3051  * Set ownership given a path name.
3052  */
3053 #ifndef _SYS_SYSPROTO_H_
3054 struct chown_args {
3055 	char	*path;
3056 	int	uid;
3057 	int	gid;
3058 };
3059 #endif
3060 int
3061 sys_chown(td, uap)
3062 	struct thread *td;
3063 	register struct chown_args /* {
3064 		char *path;
3065 		int uid;
3066 		int gid;
3067 	} */ *uap;
3068 {
3069 
3070 	return (kern_chown(td, uap->path, UIO_USERSPACE, uap->uid, uap->gid));
3071 }
3072 
3073 #ifndef _SYS_SYSPROTO_H_
3074 struct fchownat_args {
3075 	int fd;
3076 	const char * path;
3077 	uid_t uid;
3078 	gid_t gid;
3079 	int flag;
3080 };
3081 #endif
3082 int
3083 sys_fchownat(struct thread *td, struct fchownat_args *uap)
3084 {
3085 	int flag;
3086 
3087 	flag = uap->flag;
3088 	if (flag & ~AT_SYMLINK_NOFOLLOW)
3089 		return (EINVAL);
3090 
3091 	return (kern_fchownat(td, uap->fd, uap->path, UIO_USERSPACE, uap->uid,
3092 	    uap->gid, uap->flag));
3093 }
3094 
3095 int
3096 kern_chown(struct thread *td, char *path, enum uio_seg pathseg, int uid,
3097     int gid)
3098 {
3099 
3100 	return (kern_fchownat(td, AT_FDCWD, path, pathseg, uid, gid, 0));
3101 }
3102 
3103 int
3104 kern_fchownat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
3105     int uid, int gid, int flag)
3106 {
3107 	struct nameidata nd;
3108 	int error, vfslocked, follow;
3109 
3110 	AUDIT_ARG_OWNER(uid, gid);
3111 	follow = (flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
3112 	NDINIT_ATRIGHTS(&nd, LOOKUP, follow | MPSAFE | AUDITVNODE1, pathseg,
3113 	    path, fd, CAP_FCHOWN, td);
3114 
3115 	if ((error = namei(&nd)) != 0)
3116 		return (error);
3117 	vfslocked = NDHASGIANT(&nd);
3118 	NDFREE(&nd, NDF_ONLY_PNBUF);
3119 	error = setfown(td, td->td_ucred, nd.ni_vp, uid, gid);
3120 	vrele(nd.ni_vp);
3121 	VFS_UNLOCK_GIANT(vfslocked);
3122 	return (error);
3123 }
3124 
3125 /*
3126  * Set ownership given a path name, do not cross symlinks.
3127  */
3128 #ifndef _SYS_SYSPROTO_H_
3129 struct lchown_args {
3130 	char	*path;
3131 	int	uid;
3132 	int	gid;
3133 };
3134 #endif
3135 int
3136 sys_lchown(td, uap)
3137 	struct thread *td;
3138 	register struct lchown_args /* {
3139 		char *path;
3140 		int uid;
3141 		int gid;
3142 	} */ *uap;
3143 {
3144 
3145 	return (kern_lchown(td, uap->path, UIO_USERSPACE, uap->uid, uap->gid));
3146 }
3147 
3148 int
3149 kern_lchown(struct thread *td, char *path, enum uio_seg pathseg, int uid,
3150     int gid)
3151 {
3152 
3153 	return (kern_fchownat(td, AT_FDCWD, path, pathseg, uid, gid,
3154 	    AT_SYMLINK_NOFOLLOW));
3155 }
3156 
3157 /*
3158  * Set ownership given a file descriptor.
3159  */
3160 #ifndef _SYS_SYSPROTO_H_
3161 struct fchown_args {
3162 	int	fd;
3163 	int	uid;
3164 	int	gid;
3165 };
3166 #endif
3167 int
3168 sys_fchown(td, uap)
3169 	struct thread *td;
3170 	register struct fchown_args /* {
3171 		int fd;
3172 		int uid;
3173 		int gid;
3174 	} */ *uap;
3175 {
3176 	struct file *fp;
3177 	int error;
3178 
3179 	AUDIT_ARG_FD(uap->fd);
3180 	AUDIT_ARG_OWNER(uap->uid, uap->gid);
3181 	error = fget(td, uap->fd, CAP_FCHOWN, &fp);
3182 	if (error != 0)
3183 		return (error);
3184 	error = fo_chown(fp, uap->uid, uap->gid, td->td_ucred, td);
3185 	fdrop(fp, td);
3186 	return (error);
3187 }
3188 
3189 /*
3190  * Common implementation code for utimes(), lutimes(), and futimes().
3191  */
3192 static int
3193 getutimes(usrtvp, tvpseg, tsp)
3194 	const struct timeval *usrtvp;
3195 	enum uio_seg tvpseg;
3196 	struct timespec *tsp;
3197 {
3198 	struct timeval tv[2];
3199 	const struct timeval *tvp;
3200 	int error;
3201 
3202 	if (usrtvp == NULL) {
3203 		vfs_timestamp(&tsp[0]);
3204 		tsp[1] = tsp[0];
3205 	} else {
3206 		if (tvpseg == UIO_SYSSPACE) {
3207 			tvp = usrtvp;
3208 		} else {
3209 			if ((error = copyin(usrtvp, tv, sizeof(tv))) != 0)
3210 				return (error);
3211 			tvp = tv;
3212 		}
3213 
3214 		if (tvp[0].tv_usec < 0 || tvp[0].tv_usec >= 1000000 ||
3215 		    tvp[1].tv_usec < 0 || tvp[1].tv_usec >= 1000000)
3216 			return (EINVAL);
3217 		TIMEVAL_TO_TIMESPEC(&tvp[0], &tsp[0]);
3218 		TIMEVAL_TO_TIMESPEC(&tvp[1], &tsp[1]);
3219 	}
3220 	return (0);
3221 }
3222 
3223 /*
3224  * Common implementation code for utimes(), lutimes(), and futimes().
3225  */
3226 static int
3227 setutimes(td, vp, ts, numtimes, nullflag)
3228 	struct thread *td;
3229 	struct vnode *vp;
3230 	const struct timespec *ts;
3231 	int numtimes;
3232 	int nullflag;
3233 {
3234 	int error, setbirthtime;
3235 	struct mount *mp;
3236 	struct vattr vattr;
3237 
3238 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
3239 		return (error);
3240 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3241 	setbirthtime = 0;
3242 	if (numtimes < 3 && !VOP_GETATTR(vp, &vattr, td->td_ucred) &&
3243 	    timespeccmp(&ts[1], &vattr.va_birthtime, < ))
3244 		setbirthtime = 1;
3245 	VATTR_NULL(&vattr);
3246 	vattr.va_atime = ts[0];
3247 	vattr.va_mtime = ts[1];
3248 	if (setbirthtime)
3249 		vattr.va_birthtime = ts[1];
3250 	if (numtimes > 2)
3251 		vattr.va_birthtime = ts[2];
3252 	if (nullflag)
3253 		vattr.va_vaflags |= VA_UTIMES_NULL;
3254 #ifdef MAC
3255 	error = mac_vnode_check_setutimes(td->td_ucred, vp, vattr.va_atime,
3256 	    vattr.va_mtime);
3257 #endif
3258 	if (error == 0)
3259 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3260 	VOP_UNLOCK(vp, 0);
3261 	vn_finished_write(mp);
3262 	return (error);
3263 }
3264 
3265 /*
3266  * Set the access and modification times of a file.
3267  */
3268 #ifndef _SYS_SYSPROTO_H_
3269 struct utimes_args {
3270 	char	*path;
3271 	struct	timeval *tptr;
3272 };
3273 #endif
3274 int
3275 sys_utimes(td, uap)
3276 	struct thread *td;
3277 	register struct utimes_args /* {
3278 		char *path;
3279 		struct timeval *tptr;
3280 	} */ *uap;
3281 {
3282 
3283 	return (kern_utimes(td, uap->path, UIO_USERSPACE, uap->tptr,
3284 	    UIO_USERSPACE));
3285 }
3286 
3287 #ifndef _SYS_SYSPROTO_H_
3288 struct futimesat_args {
3289 	int fd;
3290 	const char * path;
3291 	const struct timeval * times;
3292 };
3293 #endif
3294 int
3295 sys_futimesat(struct thread *td, struct futimesat_args *uap)
3296 {
3297 
3298 	return (kern_utimesat(td, uap->fd, uap->path, UIO_USERSPACE,
3299 	    uap->times, UIO_USERSPACE));
3300 }
3301 
3302 int
3303 kern_utimes(struct thread *td, char *path, enum uio_seg pathseg,
3304     struct timeval *tptr, enum uio_seg tptrseg)
3305 {
3306 
3307 	return (kern_utimesat(td, AT_FDCWD, path, pathseg, tptr, tptrseg));
3308 }
3309 
3310 int
3311 kern_utimesat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
3312     struct timeval *tptr, enum uio_seg tptrseg)
3313 {
3314 	struct nameidata nd;
3315 	struct timespec ts[2];
3316 	int error, vfslocked;
3317 
3318 	if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3319 		return (error);
3320 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | MPSAFE | AUDITVNODE1, pathseg,
3321 	    path, fd, CAP_FUTIMES, td);
3322 
3323 	if ((error = namei(&nd)) != 0)
3324 		return (error);
3325 	vfslocked = NDHASGIANT(&nd);
3326 	NDFREE(&nd, NDF_ONLY_PNBUF);
3327 	error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3328 	vrele(nd.ni_vp);
3329 	VFS_UNLOCK_GIANT(vfslocked);
3330 	return (error);
3331 }
3332 
3333 /*
3334  * Set the access and modification times of a file.
3335  */
3336 #ifndef _SYS_SYSPROTO_H_
3337 struct lutimes_args {
3338 	char	*path;
3339 	struct	timeval *tptr;
3340 };
3341 #endif
3342 int
3343 sys_lutimes(td, uap)
3344 	struct thread *td;
3345 	register struct lutimes_args /* {
3346 		char *path;
3347 		struct timeval *tptr;
3348 	} */ *uap;
3349 {
3350 
3351 	return (kern_lutimes(td, uap->path, UIO_USERSPACE, uap->tptr,
3352 	    UIO_USERSPACE));
3353 }
3354 
3355 int
3356 kern_lutimes(struct thread *td, char *path, enum uio_seg pathseg,
3357     struct timeval *tptr, enum uio_seg tptrseg)
3358 {
3359 	struct timespec ts[2];
3360 	int error;
3361 	struct nameidata nd;
3362 	int vfslocked;
3363 
3364 	if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3365 		return (error);
3366 	NDINIT(&nd, LOOKUP, NOFOLLOW | MPSAFE | AUDITVNODE1, pathseg, path, td);
3367 	if ((error = namei(&nd)) != 0)
3368 		return (error);
3369 	vfslocked = NDHASGIANT(&nd);
3370 	NDFREE(&nd, NDF_ONLY_PNBUF);
3371 	error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3372 	vrele(nd.ni_vp);
3373 	VFS_UNLOCK_GIANT(vfslocked);
3374 	return (error);
3375 }
3376 
3377 /*
3378  * Set the access and modification times of a file.
3379  */
3380 #ifndef _SYS_SYSPROTO_H_
3381 struct futimes_args {
3382 	int	fd;
3383 	struct	timeval *tptr;
3384 };
3385 #endif
3386 int
3387 sys_futimes(td, uap)
3388 	struct thread *td;
3389 	register struct futimes_args /* {
3390 		int  fd;
3391 		struct timeval *tptr;
3392 	} */ *uap;
3393 {
3394 
3395 	return (kern_futimes(td, uap->fd, uap->tptr, UIO_USERSPACE));
3396 }
3397 
3398 int
3399 kern_futimes(struct thread *td, int fd, struct timeval *tptr,
3400     enum uio_seg tptrseg)
3401 {
3402 	struct timespec ts[2];
3403 	struct file *fp;
3404 	int vfslocked;
3405 	int error;
3406 
3407 	AUDIT_ARG_FD(fd);
3408 	if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3409 		return (error);
3410 	if ((error = getvnode(td->td_proc->p_fd, fd, CAP_FUTIMES, &fp))
3411 	    != 0)
3412 		return (error);
3413 	vfslocked = VFS_LOCK_GIANT(fp->f_vnode->v_mount);
3414 #ifdef AUDIT
3415 	vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
3416 	AUDIT_ARG_VNODE1(fp->f_vnode);
3417 	VOP_UNLOCK(fp->f_vnode, 0);
3418 #endif
3419 	error = setutimes(td, fp->f_vnode, ts, 2, tptr == NULL);
3420 	VFS_UNLOCK_GIANT(vfslocked);
3421 	fdrop(fp, td);
3422 	return (error);
3423 }
3424 
3425 /*
3426  * Truncate a file given its path name.
3427  */
3428 #ifndef _SYS_SYSPROTO_H_
3429 struct truncate_args {
3430 	char	*path;
3431 	int	pad;
3432 	off_t	length;
3433 };
3434 #endif
3435 int
3436 sys_truncate(td, uap)
3437 	struct thread *td;
3438 	register struct truncate_args /* {
3439 		char *path;
3440 		int pad;
3441 		off_t length;
3442 	} */ *uap;
3443 {
3444 
3445 	return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3446 }
3447 
3448 int
3449 kern_truncate(struct thread *td, char *path, enum uio_seg pathseg, off_t length)
3450 {
3451 	struct mount *mp;
3452 	struct vnode *vp;
3453 	struct vattr vattr;
3454 	int error;
3455 	struct nameidata nd;
3456 	int vfslocked;
3457 
3458 	if (length < 0)
3459 		return(EINVAL);
3460 	NDINIT(&nd, LOOKUP, FOLLOW | MPSAFE | AUDITVNODE1, pathseg, path, td);
3461 	if ((error = namei(&nd)) != 0)
3462 		return (error);
3463 	vfslocked = NDHASGIANT(&nd);
3464 	vp = nd.ni_vp;
3465 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) {
3466 		vrele(vp);
3467 		VFS_UNLOCK_GIANT(vfslocked);
3468 		return (error);
3469 	}
3470 	NDFREE(&nd, NDF_ONLY_PNBUF);
3471 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3472 	if (vp->v_type == VDIR)
3473 		error = EISDIR;
3474 #ifdef MAC
3475 	else if ((error = mac_vnode_check_write(td->td_ucred, NOCRED, vp))) {
3476 	}
3477 #endif
3478 	else if ((error = vn_writechk(vp)) == 0 &&
3479 	    (error = VOP_ACCESS(vp, VWRITE, td->td_ucred, td)) == 0) {
3480 		VATTR_NULL(&vattr);
3481 		vattr.va_size = length;
3482 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3483 	}
3484 	vput(vp);
3485 	vn_finished_write(mp);
3486 	VFS_UNLOCK_GIANT(vfslocked);
3487 	return (error);
3488 }
3489 
3490 #if defined(COMPAT_43)
3491 /*
3492  * Truncate a file given its path name.
3493  */
3494 #ifndef _SYS_SYSPROTO_H_
3495 struct otruncate_args {
3496 	char	*path;
3497 	long	length;
3498 };
3499 #endif
3500 int
3501 otruncate(td, uap)
3502 	struct thread *td;
3503 	register struct otruncate_args /* {
3504 		char *path;
3505 		long length;
3506 	} */ *uap;
3507 {
3508 	struct truncate_args /* {
3509 		char *path;
3510 		int pad;
3511 		off_t length;
3512 	} */ nuap;
3513 
3514 	nuap.path = uap->path;
3515 	nuap.length = uap->length;
3516 	return (sys_truncate(td, &nuap));
3517 }
3518 #endif /* COMPAT_43 */
3519 
3520 /* Versions with the pad argument */
3521 int
3522 freebsd6_truncate(struct thread *td, struct freebsd6_truncate_args *uap)
3523 {
3524 	struct truncate_args ouap;
3525 
3526 	ouap.path = uap->path;
3527 	ouap.length = uap->length;
3528 	return (sys_truncate(td, &ouap));
3529 }
3530 
3531 int
3532 freebsd6_ftruncate(struct thread *td, struct freebsd6_ftruncate_args *uap)
3533 {
3534 	struct ftruncate_args ouap;
3535 
3536 	ouap.fd = uap->fd;
3537 	ouap.length = uap->length;
3538 	return (sys_ftruncate(td, &ouap));
3539 }
3540 
3541 /*
3542  * Sync an open file.
3543  */
3544 #ifndef _SYS_SYSPROTO_H_
3545 struct fsync_args {
3546 	int	fd;
3547 };
3548 #endif
3549 int
3550 sys_fsync(td, uap)
3551 	struct thread *td;
3552 	struct fsync_args /* {
3553 		int fd;
3554 	} */ *uap;
3555 {
3556 	struct vnode *vp;
3557 	struct mount *mp;
3558 	struct file *fp;
3559 	int vfslocked;
3560 	int error, lock_flags;
3561 
3562 	AUDIT_ARG_FD(uap->fd);
3563 	if ((error = getvnode(td->td_proc->p_fd, uap->fd, CAP_FSYNC,
3564 	    &fp)) != 0)
3565 		return (error);
3566 	vp = fp->f_vnode;
3567 	vfslocked = VFS_LOCK_GIANT(vp->v_mount);
3568 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
3569 		goto drop;
3570 	if (MNT_SHARED_WRITES(mp) ||
3571 	    ((mp == NULL) && MNT_SHARED_WRITES(vp->v_mount))) {
3572 		lock_flags = LK_SHARED;
3573 	} else {
3574 		lock_flags = LK_EXCLUSIVE;
3575 	}
3576 	vn_lock(vp, lock_flags | LK_RETRY);
3577 	AUDIT_ARG_VNODE1(vp);
3578 	if (vp->v_object != NULL) {
3579 		VM_OBJECT_LOCK(vp->v_object);
3580 		vm_object_page_clean(vp->v_object, 0, 0, 0);
3581 		VM_OBJECT_UNLOCK(vp->v_object);
3582 	}
3583 	error = VOP_FSYNC(vp, MNT_WAIT, td);
3584 
3585 	VOP_UNLOCK(vp, 0);
3586 	vn_finished_write(mp);
3587 drop:
3588 	VFS_UNLOCK_GIANT(vfslocked);
3589 	fdrop(fp, td);
3590 	return (error);
3591 }
3592 
3593 /*
3594  * Rename files.  Source and destination must either both be directories, or
3595  * both not be directories.  If target is a directory, it must be empty.
3596  */
3597 #ifndef _SYS_SYSPROTO_H_
3598 struct rename_args {
3599 	char	*from;
3600 	char	*to;
3601 };
3602 #endif
3603 int
3604 sys_rename(td, uap)
3605 	struct thread *td;
3606 	register struct rename_args /* {
3607 		char *from;
3608 		char *to;
3609 	} */ *uap;
3610 {
3611 
3612 	return (kern_rename(td, uap->from, uap->to, UIO_USERSPACE));
3613 }
3614 
3615 #ifndef _SYS_SYSPROTO_H_
3616 struct renameat_args {
3617 	int	oldfd;
3618 	char	*old;
3619 	int	newfd;
3620 	char	*new;
3621 };
3622 #endif
3623 int
3624 sys_renameat(struct thread *td, struct renameat_args *uap)
3625 {
3626 
3627 	return (kern_renameat(td, uap->oldfd, uap->old, uap->newfd, uap->new,
3628 	    UIO_USERSPACE));
3629 }
3630 
3631 int
3632 kern_rename(struct thread *td, char *from, char *to, enum uio_seg pathseg)
3633 {
3634 
3635 	return (kern_renameat(td, AT_FDCWD, from, AT_FDCWD, to, pathseg));
3636 }
3637 
3638 int
3639 kern_renameat(struct thread *td, int oldfd, char *old, int newfd, char *new,
3640     enum uio_seg pathseg)
3641 {
3642 	struct mount *mp = NULL;
3643 	struct vnode *tvp, *fvp, *tdvp;
3644 	struct nameidata fromnd, tond;
3645 	int tvfslocked;
3646 	int fvfslocked;
3647 	int error;
3648 
3649 	bwillwrite();
3650 #ifdef MAC
3651 	NDINIT_ATRIGHTS(&fromnd, DELETE, LOCKPARENT | LOCKLEAF | SAVESTART |
3652 	    MPSAFE | AUDITVNODE1, pathseg, old, oldfd, CAP_DELETE, td);
3653 #else
3654 	NDINIT_ATRIGHTS(&fromnd, DELETE, WANTPARENT | SAVESTART | MPSAFE |
3655 	    AUDITVNODE1, pathseg, old, oldfd, CAP_DELETE, td);
3656 #endif
3657 
3658 	if ((error = namei(&fromnd)) != 0)
3659 		return (error);
3660 	fvfslocked = NDHASGIANT(&fromnd);
3661 	tvfslocked = 0;
3662 #ifdef MAC
3663 	error = mac_vnode_check_rename_from(td->td_ucred, fromnd.ni_dvp,
3664 	    fromnd.ni_vp, &fromnd.ni_cnd);
3665 	VOP_UNLOCK(fromnd.ni_dvp, 0);
3666 	if (fromnd.ni_dvp != fromnd.ni_vp)
3667 		VOP_UNLOCK(fromnd.ni_vp, 0);
3668 #endif
3669 	fvp = fromnd.ni_vp;
3670 	if (error == 0)
3671 		error = vn_start_write(fvp, &mp, V_WAIT | PCATCH);
3672 	if (error != 0) {
3673 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3674 		vrele(fromnd.ni_dvp);
3675 		vrele(fvp);
3676 		goto out1;
3677 	}
3678 	NDINIT_ATRIGHTS(&tond, RENAME, LOCKPARENT | LOCKLEAF | NOCACHE |
3679 	    SAVESTART | MPSAFE | AUDITVNODE2, pathseg, new, newfd, CAP_CREATE,
3680 	    td);
3681 	if (fromnd.ni_vp->v_type == VDIR)
3682 		tond.ni_cnd.cn_flags |= WILLBEDIR;
3683 	if ((error = namei(&tond)) != 0) {
3684 		/* Translate error code for rename("dir1", "dir2/."). */
3685 		if (error == EISDIR && fvp->v_type == VDIR)
3686 			error = EINVAL;
3687 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3688 		vrele(fromnd.ni_dvp);
3689 		vrele(fvp);
3690 		vn_finished_write(mp);
3691 		goto out1;
3692 	}
3693 	tvfslocked = NDHASGIANT(&tond);
3694 	tdvp = tond.ni_dvp;
3695 	tvp = tond.ni_vp;
3696 	if (tvp != NULL) {
3697 		if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
3698 			error = ENOTDIR;
3699 			goto out;
3700 		} else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
3701 			error = EISDIR;
3702 			goto out;
3703 		}
3704 	}
3705 	if (fvp == tdvp) {
3706 		error = EINVAL;
3707 		goto out;
3708 	}
3709 	/*
3710 	 * If the source is the same as the destination (that is, if they
3711 	 * are links to the same vnode), then there is nothing to do.
3712 	 */
3713 	if (fvp == tvp)
3714 		error = -1;
3715 #ifdef MAC
3716 	else
3717 		error = mac_vnode_check_rename_to(td->td_ucred, tdvp,
3718 		    tond.ni_vp, fromnd.ni_dvp == tdvp, &tond.ni_cnd);
3719 #endif
3720 out:
3721 	if (!error) {
3722 		error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
3723 				   tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
3724 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3725 		NDFREE(&tond, NDF_ONLY_PNBUF);
3726 	} else {
3727 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3728 		NDFREE(&tond, NDF_ONLY_PNBUF);
3729 		if (tvp)
3730 			vput(tvp);
3731 		if (tdvp == tvp)
3732 			vrele(tdvp);
3733 		else
3734 			vput(tdvp);
3735 		vrele(fromnd.ni_dvp);
3736 		vrele(fvp);
3737 	}
3738 	vrele(tond.ni_startdir);
3739 	vn_finished_write(mp);
3740 out1:
3741 	if (fromnd.ni_startdir)
3742 		vrele(fromnd.ni_startdir);
3743 	VFS_UNLOCK_GIANT(fvfslocked);
3744 	VFS_UNLOCK_GIANT(tvfslocked);
3745 	if (error == -1)
3746 		return (0);
3747 	return (error);
3748 }
3749 
3750 /*
3751  * Make a directory file.
3752  */
3753 #ifndef _SYS_SYSPROTO_H_
3754 struct mkdir_args {
3755 	char	*path;
3756 	int	mode;
3757 };
3758 #endif
3759 int
3760 sys_mkdir(td, uap)
3761 	struct thread *td;
3762 	register struct mkdir_args /* {
3763 		char *path;
3764 		int mode;
3765 	} */ *uap;
3766 {
3767 
3768 	return (kern_mkdir(td, uap->path, UIO_USERSPACE, uap->mode));
3769 }
3770 
3771 #ifndef _SYS_SYSPROTO_H_
3772 struct mkdirat_args {
3773 	int	fd;
3774 	char	*path;
3775 	mode_t	mode;
3776 };
3777 #endif
3778 int
3779 sys_mkdirat(struct thread *td, struct mkdirat_args *uap)
3780 {
3781 
3782 	return (kern_mkdirat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode));
3783 }
3784 
3785 int
3786 kern_mkdir(struct thread *td, char *path, enum uio_seg segflg, int mode)
3787 {
3788 
3789 	return (kern_mkdirat(td, AT_FDCWD, path, segflg, mode));
3790 }
3791 
3792 int
3793 kern_mkdirat(struct thread *td, int fd, char *path, enum uio_seg segflg,
3794     int mode)
3795 {
3796 	struct mount *mp;
3797 	struct vnode *vp;
3798 	struct vattr vattr;
3799 	int error;
3800 	struct nameidata nd;
3801 	int vfslocked;
3802 
3803 	AUDIT_ARG_MODE(mode);
3804 restart:
3805 	bwillwrite();
3806 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | MPSAFE |
3807 	    AUDITVNODE1, segflg, path, fd, CAP_MKDIR, td);
3808 	nd.ni_cnd.cn_flags |= WILLBEDIR;
3809 	if ((error = namei(&nd)) != 0)
3810 		return (error);
3811 	vfslocked = NDHASGIANT(&nd);
3812 	vp = nd.ni_vp;
3813 	if (vp != NULL) {
3814 		NDFREE(&nd, NDF_ONLY_PNBUF);
3815 		/*
3816 		 * XXX namei called with LOCKPARENT but not LOCKLEAF has
3817 		 * the strange behaviour of leaving the vnode unlocked
3818 		 * if the target is the same vnode as the parent.
3819 		 */
3820 		if (vp == nd.ni_dvp)
3821 			vrele(nd.ni_dvp);
3822 		else
3823 			vput(nd.ni_dvp);
3824 		vrele(vp);
3825 		VFS_UNLOCK_GIANT(vfslocked);
3826 		return (EEXIST);
3827 	}
3828 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3829 		NDFREE(&nd, NDF_ONLY_PNBUF);
3830 		vput(nd.ni_dvp);
3831 		VFS_UNLOCK_GIANT(vfslocked);
3832 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3833 			return (error);
3834 		goto restart;
3835 	}
3836 	VATTR_NULL(&vattr);
3837 	vattr.va_type = VDIR;
3838 	vattr.va_mode = (mode & ACCESSPERMS) &~ td->td_proc->p_fd->fd_cmask;
3839 #ifdef MAC
3840 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
3841 	    &vattr);
3842 	if (error)
3843 		goto out;
3844 #endif
3845 	error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
3846 #ifdef MAC
3847 out:
3848 #endif
3849 	NDFREE(&nd, NDF_ONLY_PNBUF);
3850 	vput(nd.ni_dvp);
3851 	if (!error)
3852 		vput(nd.ni_vp);
3853 	vn_finished_write(mp);
3854 	VFS_UNLOCK_GIANT(vfslocked);
3855 	return (error);
3856 }
3857 
3858 /*
3859  * Remove a directory file.
3860  */
3861 #ifndef _SYS_SYSPROTO_H_
3862 struct rmdir_args {
3863 	char	*path;
3864 };
3865 #endif
3866 int
3867 sys_rmdir(td, uap)
3868 	struct thread *td;
3869 	struct rmdir_args /* {
3870 		char *path;
3871 	} */ *uap;
3872 {
3873 
3874 	return (kern_rmdir(td, uap->path, UIO_USERSPACE));
3875 }
3876 
3877 int
3878 kern_rmdir(struct thread *td, char *path, enum uio_seg pathseg)
3879 {
3880 
3881 	return (kern_rmdirat(td, AT_FDCWD, path, pathseg));
3882 }
3883 
3884 int
3885 kern_rmdirat(struct thread *td, int fd, char *path, enum uio_seg pathseg)
3886 {
3887 	struct mount *mp;
3888 	struct vnode *vp;
3889 	int error;
3890 	struct nameidata nd;
3891 	int vfslocked;
3892 
3893 restart:
3894 	bwillwrite();
3895 	NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | MPSAFE |
3896 	    AUDITVNODE1, pathseg, path, fd, CAP_RMDIR, td);
3897 	if ((error = namei(&nd)) != 0)
3898 		return (error);
3899 	vfslocked = NDHASGIANT(&nd);
3900 	vp = nd.ni_vp;
3901 	if (vp->v_type != VDIR) {
3902 		error = ENOTDIR;
3903 		goto out;
3904 	}
3905 	/*
3906 	 * No rmdir "." please.
3907 	 */
3908 	if (nd.ni_dvp == vp) {
3909 		error = EINVAL;
3910 		goto out;
3911 	}
3912 	/*
3913 	 * The root of a mounted filesystem cannot be deleted.
3914 	 */
3915 	if (vp->v_vflag & VV_ROOT) {
3916 		error = EBUSY;
3917 		goto out;
3918 	}
3919 #ifdef MAC
3920 	error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
3921 	    &nd.ni_cnd);
3922 	if (error)
3923 		goto out;
3924 #endif
3925 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3926 		NDFREE(&nd, NDF_ONLY_PNBUF);
3927 		vput(vp);
3928 		if (nd.ni_dvp == vp)
3929 			vrele(nd.ni_dvp);
3930 		else
3931 			vput(nd.ni_dvp);
3932 		VFS_UNLOCK_GIANT(vfslocked);
3933 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3934 			return (error);
3935 		goto restart;
3936 	}
3937 	error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
3938 	vn_finished_write(mp);
3939 out:
3940 	NDFREE(&nd, NDF_ONLY_PNBUF);
3941 	vput(vp);
3942 	if (nd.ni_dvp == vp)
3943 		vrele(nd.ni_dvp);
3944 	else
3945 		vput(nd.ni_dvp);
3946 	VFS_UNLOCK_GIANT(vfslocked);
3947 	return (error);
3948 }
3949 
3950 #ifdef COMPAT_43
3951 /*
3952  * Read a block of directory entries in a filesystem independent format.
3953  */
3954 #ifndef _SYS_SYSPROTO_H_
3955 struct ogetdirentries_args {
3956 	int	fd;
3957 	char	*buf;
3958 	u_int	count;
3959 	long	*basep;
3960 };
3961 #endif
3962 int
3963 ogetdirentries(struct thread *td, struct ogetdirentries_args *uap)
3964 {
3965 	long loff;
3966 	int error;
3967 
3968 	error = kern_ogetdirentries(td, uap, &loff);
3969 	if (error == 0)
3970 		error = copyout(&loff, uap->basep, sizeof(long));
3971 	return (error);
3972 }
3973 
3974 int
3975 kern_ogetdirentries(struct thread *td, struct ogetdirentries_args *uap,
3976     long *ploff)
3977 {
3978 	struct vnode *vp;
3979 	struct file *fp;
3980 	struct uio auio, kuio;
3981 	struct iovec aiov, kiov;
3982 	struct dirent *dp, *edp;
3983 	caddr_t dirbuf;
3984 	int error, eofflag, readcnt, vfslocked;
3985 	long loff;
3986 
3987 	/* XXX arbitrary sanity limit on `count'. */
3988 	if (uap->count > 64 * 1024)
3989 		return (EINVAL);
3990 	if ((error = getvnode(td->td_proc->p_fd, uap->fd, CAP_READ,
3991 	    &fp)) != 0)
3992 		return (error);
3993 	if ((fp->f_flag & FREAD) == 0) {
3994 		fdrop(fp, td);
3995 		return (EBADF);
3996 	}
3997 	vp = fp->f_vnode;
3998 unionread:
3999 	vfslocked = VFS_LOCK_GIANT(vp->v_mount);
4000 	if (vp->v_type != VDIR) {
4001 		VFS_UNLOCK_GIANT(vfslocked);
4002 		fdrop(fp, td);
4003 		return (EINVAL);
4004 	}
4005 	aiov.iov_base = uap->buf;
4006 	aiov.iov_len = uap->count;
4007 	auio.uio_iov = &aiov;
4008 	auio.uio_iovcnt = 1;
4009 	auio.uio_rw = UIO_READ;
4010 	auio.uio_segflg = UIO_USERSPACE;
4011 	auio.uio_td = td;
4012 	auio.uio_resid = uap->count;
4013 	vn_lock(vp, LK_SHARED | LK_RETRY);
4014 	loff = auio.uio_offset = fp->f_offset;
4015 #ifdef MAC
4016 	error = mac_vnode_check_readdir(td->td_ucred, vp);
4017 	if (error) {
4018 		VOP_UNLOCK(vp, 0);
4019 		VFS_UNLOCK_GIANT(vfslocked);
4020 		fdrop(fp, td);
4021 		return (error);
4022 	}
4023 #endif
4024 #	if (BYTE_ORDER != LITTLE_ENDIAN)
4025 		if (vp->v_mount->mnt_maxsymlinklen <= 0) {
4026 			error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag,
4027 			    NULL, NULL);
4028 			fp->f_offset = auio.uio_offset;
4029 		} else
4030 #	endif
4031 	{
4032 		kuio = auio;
4033 		kuio.uio_iov = &kiov;
4034 		kuio.uio_segflg = UIO_SYSSPACE;
4035 		kiov.iov_len = uap->count;
4036 		dirbuf = malloc(uap->count, M_TEMP, M_WAITOK);
4037 		kiov.iov_base = dirbuf;
4038 		error = VOP_READDIR(vp, &kuio, fp->f_cred, &eofflag,
4039 			    NULL, NULL);
4040 		fp->f_offset = kuio.uio_offset;
4041 		if (error == 0) {
4042 			readcnt = uap->count - kuio.uio_resid;
4043 			edp = (struct dirent *)&dirbuf[readcnt];
4044 			for (dp = (struct dirent *)dirbuf; dp < edp; ) {
4045 #				if (BYTE_ORDER == LITTLE_ENDIAN)
4046 					/*
4047 					 * The expected low byte of
4048 					 * dp->d_namlen is our dp->d_type.
4049 					 * The high MBZ byte of dp->d_namlen
4050 					 * is our dp->d_namlen.
4051 					 */
4052 					dp->d_type = dp->d_namlen;
4053 					dp->d_namlen = 0;
4054 #				else
4055 					/*
4056 					 * The dp->d_type is the high byte
4057 					 * of the expected dp->d_namlen,
4058 					 * so must be zero'ed.
4059 					 */
4060 					dp->d_type = 0;
4061 #				endif
4062 				if (dp->d_reclen > 0) {
4063 					dp = (struct dirent *)
4064 					    ((char *)dp + dp->d_reclen);
4065 				} else {
4066 					error = EIO;
4067 					break;
4068 				}
4069 			}
4070 			if (dp >= edp)
4071 				error = uiomove(dirbuf, readcnt, &auio);
4072 		}
4073 		free(dirbuf, M_TEMP);
4074 	}
4075 	if (error) {
4076 		VOP_UNLOCK(vp, 0);
4077 		VFS_UNLOCK_GIANT(vfslocked);
4078 		fdrop(fp, td);
4079 		return (error);
4080 	}
4081 	if (uap->count == auio.uio_resid &&
4082 	    (vp->v_vflag & VV_ROOT) &&
4083 	    (vp->v_mount->mnt_flag & MNT_UNION)) {
4084 		struct vnode *tvp = vp;
4085 		vp = vp->v_mount->mnt_vnodecovered;
4086 		VREF(vp);
4087 		fp->f_vnode = vp;
4088 		fp->f_data = vp;
4089 		fp->f_offset = 0;
4090 		vput(tvp);
4091 		VFS_UNLOCK_GIANT(vfslocked);
4092 		goto unionread;
4093 	}
4094 	VOP_UNLOCK(vp, 0);
4095 	VFS_UNLOCK_GIANT(vfslocked);
4096 	fdrop(fp, td);
4097 	td->td_retval[0] = uap->count - auio.uio_resid;
4098 	if (error == 0)
4099 		*ploff = loff;
4100 	return (error);
4101 }
4102 #endif /* COMPAT_43 */
4103 
4104 /*
4105  * Read a block of directory entries in a filesystem independent format.
4106  */
4107 #ifndef _SYS_SYSPROTO_H_
4108 struct getdirentries_args {
4109 	int	fd;
4110 	char	*buf;
4111 	u_int	count;
4112 	long	*basep;
4113 };
4114 #endif
4115 int
4116 sys_getdirentries(td, uap)
4117 	struct thread *td;
4118 	register struct getdirentries_args /* {
4119 		int fd;
4120 		char *buf;
4121 		u_int count;
4122 		long *basep;
4123 	} */ *uap;
4124 {
4125 	long base;
4126 	int error;
4127 
4128 	error = kern_getdirentries(td, uap->fd, uap->buf, uap->count, &base);
4129 	if (error)
4130 		return (error);
4131 	if (uap->basep != NULL)
4132 		error = copyout(&base, uap->basep, sizeof(long));
4133 	return (error);
4134 }
4135 
4136 int
4137 kern_getdirentries(struct thread *td, int fd, char *buf, u_int count,
4138     long *basep)
4139 {
4140 	struct vnode *vp;
4141 	struct file *fp;
4142 	struct uio auio;
4143 	struct iovec aiov;
4144 	int vfslocked;
4145 	long loff;
4146 	int error, eofflag;
4147 
4148 	AUDIT_ARG_FD(fd);
4149 	if (count > INT_MAX)
4150 		return (EINVAL);
4151 	if ((error = getvnode(td->td_proc->p_fd, fd, CAP_READ | CAP_SEEK,
4152 	    &fp)) != 0)
4153 		return (error);
4154 	if ((fp->f_flag & FREAD) == 0) {
4155 		fdrop(fp, td);
4156 		return (EBADF);
4157 	}
4158 	vp = fp->f_vnode;
4159 unionread:
4160 	vfslocked = VFS_LOCK_GIANT(vp->v_mount);
4161 	if (vp->v_type != VDIR) {
4162 		VFS_UNLOCK_GIANT(vfslocked);
4163 		error = EINVAL;
4164 		goto fail;
4165 	}
4166 	aiov.iov_base = buf;
4167 	aiov.iov_len = count;
4168 	auio.uio_iov = &aiov;
4169 	auio.uio_iovcnt = 1;
4170 	auio.uio_rw = UIO_READ;
4171 	auio.uio_segflg = UIO_USERSPACE;
4172 	auio.uio_td = td;
4173 	auio.uio_resid = count;
4174 	vn_lock(vp, LK_SHARED | LK_RETRY);
4175 	AUDIT_ARG_VNODE1(vp);
4176 	loff = auio.uio_offset = fp->f_offset;
4177 #ifdef MAC
4178 	error = mac_vnode_check_readdir(td->td_ucred, vp);
4179 	if (error == 0)
4180 #endif
4181 		error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, NULL,
4182 		    NULL);
4183 	fp->f_offset = auio.uio_offset;
4184 	if (error) {
4185 		VOP_UNLOCK(vp, 0);
4186 		VFS_UNLOCK_GIANT(vfslocked);
4187 		goto fail;
4188 	}
4189 	if (count == auio.uio_resid &&
4190 	    (vp->v_vflag & VV_ROOT) &&
4191 	    (vp->v_mount->mnt_flag & MNT_UNION)) {
4192 		struct vnode *tvp = vp;
4193 		vp = vp->v_mount->mnt_vnodecovered;
4194 		VREF(vp);
4195 		fp->f_vnode = vp;
4196 		fp->f_data = vp;
4197 		fp->f_offset = 0;
4198 		vput(tvp);
4199 		VFS_UNLOCK_GIANT(vfslocked);
4200 		goto unionread;
4201 	}
4202 	VOP_UNLOCK(vp, 0);
4203 	VFS_UNLOCK_GIANT(vfslocked);
4204 	*basep = loff;
4205 	td->td_retval[0] = count - auio.uio_resid;
4206 fail:
4207 	fdrop(fp, td);
4208 	return (error);
4209 }
4210 
4211 #ifndef _SYS_SYSPROTO_H_
4212 struct getdents_args {
4213 	int fd;
4214 	char *buf;
4215 	size_t count;
4216 };
4217 #endif
4218 int
4219 sys_getdents(td, uap)
4220 	struct thread *td;
4221 	register struct getdents_args /* {
4222 		int fd;
4223 		char *buf;
4224 		u_int count;
4225 	} */ *uap;
4226 {
4227 	struct getdirentries_args ap;
4228 	ap.fd = uap->fd;
4229 	ap.buf = uap->buf;
4230 	ap.count = uap->count;
4231 	ap.basep = NULL;
4232 	return (sys_getdirentries(td, &ap));
4233 }
4234 
4235 /*
4236  * Set the mode mask for creation of filesystem nodes.
4237  */
4238 #ifndef _SYS_SYSPROTO_H_
4239 struct umask_args {
4240 	int	newmask;
4241 };
4242 #endif
4243 int
4244 sys_umask(td, uap)
4245 	struct thread *td;
4246 	struct umask_args /* {
4247 		int newmask;
4248 	} */ *uap;
4249 {
4250 	register struct filedesc *fdp;
4251 
4252 	FILEDESC_XLOCK(td->td_proc->p_fd);
4253 	fdp = td->td_proc->p_fd;
4254 	td->td_retval[0] = fdp->fd_cmask;
4255 	fdp->fd_cmask = uap->newmask & ALLPERMS;
4256 	FILEDESC_XUNLOCK(td->td_proc->p_fd);
4257 	return (0);
4258 }
4259 
4260 /*
4261  * Void all references to file by ripping underlying filesystem away from
4262  * vnode.
4263  */
4264 #ifndef _SYS_SYSPROTO_H_
4265 struct revoke_args {
4266 	char	*path;
4267 };
4268 #endif
4269 int
4270 sys_revoke(td, uap)
4271 	struct thread *td;
4272 	register struct revoke_args /* {
4273 		char *path;
4274 	} */ *uap;
4275 {
4276 	struct vnode *vp;
4277 	struct vattr vattr;
4278 	int error;
4279 	struct nameidata nd;
4280 	int vfslocked;
4281 
4282 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | MPSAFE | AUDITVNODE1,
4283 	    UIO_USERSPACE, uap->path, td);
4284 	if ((error = namei(&nd)) != 0)
4285 		return (error);
4286 	vfslocked = NDHASGIANT(&nd);
4287 	vp = nd.ni_vp;
4288 	NDFREE(&nd, NDF_ONLY_PNBUF);
4289 	if (vp->v_type != VCHR || vp->v_rdev == NULL) {
4290 		error = EINVAL;
4291 		goto out;
4292 	}
4293 #ifdef MAC
4294 	error = mac_vnode_check_revoke(td->td_ucred, vp);
4295 	if (error)
4296 		goto out;
4297 #endif
4298 	error = VOP_GETATTR(vp, &vattr, td->td_ucred);
4299 	if (error)
4300 		goto out;
4301 	if (td->td_ucred->cr_uid != vattr.va_uid) {
4302 		error = priv_check(td, PRIV_VFS_ADMIN);
4303 		if (error)
4304 			goto out;
4305 	}
4306 	if (vcount(vp) > 1)
4307 		VOP_REVOKE(vp, REVOKEALL);
4308 out:
4309 	vput(vp);
4310 	VFS_UNLOCK_GIANT(vfslocked);
4311 	return (error);
4312 }
4313 
4314 /*
4315  * Convert a user file descriptor to a kernel file entry and check that, if it
4316  * is a capability, the correct rights are present. A reference on the file
4317  * entry is held upon returning.
4318  */
4319 int
4320 getvnode(struct filedesc *fdp, int fd, cap_rights_t rights,
4321     struct file **fpp)
4322 {
4323 	struct file *fp;
4324 #ifdef CAPABILITIES
4325 	struct file *fp_fromcap;
4326 #endif
4327 	int error;
4328 
4329 	error = 0;
4330 	fp = NULL;
4331 	if ((fdp == NULL) || (fp = fget_unlocked(fdp, fd)) == NULL)
4332 		return (EBADF);
4333 #ifdef CAPABILITIES
4334 	/*
4335 	 * If the file descriptor is for a capability, test rights and use the
4336 	 * file descriptor referenced by the capability.
4337 	 */
4338 	error = cap_funwrap(fp, rights, &fp_fromcap);
4339 	if (error) {
4340 		fdrop(fp, curthread);
4341 		return (error);
4342 	}
4343 	if (fp != fp_fromcap) {
4344 		fhold(fp_fromcap);
4345 		fdrop(fp, curthread);
4346 		fp = fp_fromcap;
4347 	}
4348 #endif /* CAPABILITIES */
4349 
4350 	/*
4351 	 * The file could be not of the vnode type, or it may be not
4352 	 * yet fully initialized, in which case the f_vnode pointer
4353 	 * may be set, but f_ops is still badfileops.  E.g.,
4354 	 * devfs_open() transiently create such situation to
4355 	 * facilitate csw d_fdopen().
4356 	 *
4357 	 * Dupfdopen() handling in kern_openat() installs the
4358 	 * half-baked file into the process descriptor table, allowing
4359 	 * other thread to dereference it. Guard against the race by
4360 	 * checking f_ops.
4361 	 */
4362 	if (fp->f_vnode == NULL || fp->f_ops == &badfileops) {
4363 		fdrop(fp, curthread);
4364 		return (EINVAL);
4365 	}
4366 	*fpp = fp;
4367 	return (0);
4368 }
4369 
4370 
4371 /*
4372  * Get an (NFS) file handle.
4373  */
4374 #ifndef _SYS_SYSPROTO_H_
4375 struct lgetfh_args {
4376 	char	*fname;
4377 	fhandle_t *fhp;
4378 };
4379 #endif
4380 int
4381 sys_lgetfh(td, uap)
4382 	struct thread *td;
4383 	register struct lgetfh_args *uap;
4384 {
4385 	struct nameidata nd;
4386 	fhandle_t fh;
4387 	register struct vnode *vp;
4388 	int vfslocked;
4389 	int error;
4390 
4391 	error = priv_check(td, PRIV_VFS_GETFH);
4392 	if (error)
4393 		return (error);
4394 	NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | MPSAFE | AUDITVNODE1,
4395 	    UIO_USERSPACE, uap->fname, td);
4396 	error = namei(&nd);
4397 	if (error)
4398 		return (error);
4399 	vfslocked = NDHASGIANT(&nd);
4400 	NDFREE(&nd, NDF_ONLY_PNBUF);
4401 	vp = nd.ni_vp;
4402 	bzero(&fh, sizeof(fh));
4403 	fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
4404 	error = VOP_VPTOFH(vp, &fh.fh_fid);
4405 	vput(vp);
4406 	VFS_UNLOCK_GIANT(vfslocked);
4407 	if (error)
4408 		return (error);
4409 	error = copyout(&fh, uap->fhp, sizeof (fh));
4410 	return (error);
4411 }
4412 
4413 #ifndef _SYS_SYSPROTO_H_
4414 struct getfh_args {
4415 	char	*fname;
4416 	fhandle_t *fhp;
4417 };
4418 #endif
4419 int
4420 sys_getfh(td, uap)
4421 	struct thread *td;
4422 	register struct getfh_args *uap;
4423 {
4424 	struct nameidata nd;
4425 	fhandle_t fh;
4426 	register struct vnode *vp;
4427 	int vfslocked;
4428 	int error;
4429 
4430 	error = priv_check(td, PRIV_VFS_GETFH);
4431 	if (error)
4432 		return (error);
4433 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | MPSAFE | AUDITVNODE1,
4434 	    UIO_USERSPACE, uap->fname, td);
4435 	error = namei(&nd);
4436 	if (error)
4437 		return (error);
4438 	vfslocked = NDHASGIANT(&nd);
4439 	NDFREE(&nd, NDF_ONLY_PNBUF);
4440 	vp = nd.ni_vp;
4441 	bzero(&fh, sizeof(fh));
4442 	fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
4443 	error = VOP_VPTOFH(vp, &fh.fh_fid);
4444 	vput(vp);
4445 	VFS_UNLOCK_GIANT(vfslocked);
4446 	if (error)
4447 		return (error);
4448 	error = copyout(&fh, uap->fhp, sizeof (fh));
4449 	return (error);
4450 }
4451 
4452 /*
4453  * syscall for the rpc.lockd to use to translate a NFS file handle into an
4454  * open descriptor.
4455  *
4456  * warning: do not remove the priv_check() call or this becomes one giant
4457  * security hole.
4458  */
4459 #ifndef _SYS_SYSPROTO_H_
4460 struct fhopen_args {
4461 	const struct fhandle *u_fhp;
4462 	int flags;
4463 };
4464 #endif
4465 int
4466 sys_fhopen(td, uap)
4467 	struct thread *td;
4468 	struct fhopen_args /* {
4469 		const struct fhandle *u_fhp;
4470 		int flags;
4471 	} */ *uap;
4472 {
4473 	struct proc *p = td->td_proc;
4474 	struct mount *mp;
4475 	struct vnode *vp;
4476 	struct fhandle fhp;
4477 	struct vattr vat;
4478 	struct vattr *vap = &vat;
4479 	struct flock lf;
4480 	struct file *fp;
4481 	register struct filedesc *fdp = p->p_fd;
4482 	int fmode, error, type;
4483 	accmode_t accmode;
4484 	struct file *nfp;
4485 	int vfslocked;
4486 	int indx;
4487 
4488 	error = priv_check(td, PRIV_VFS_FHOPEN);
4489 	if (error)
4490 		return (error);
4491 	fmode = FFLAGS(uap->flags);
4492 	/* why not allow a non-read/write open for our lockd? */
4493 	if (((fmode & (FREAD | FWRITE)) == 0) || (fmode & O_CREAT))
4494 		return (EINVAL);
4495 	error = copyin(uap->u_fhp, &fhp, sizeof(fhp));
4496 	if (error)
4497 		return(error);
4498 	/* find the mount point */
4499 	mp = vfs_busyfs(&fhp.fh_fsid);
4500 	if (mp == NULL)
4501 		return (ESTALE);
4502 	vfslocked = VFS_LOCK_GIANT(mp);
4503 	/* now give me my vnode, it gets returned to me locked */
4504 	error = VFS_FHTOVP(mp, &fhp.fh_fid, LK_EXCLUSIVE, &vp);
4505 	vfs_unbusy(mp);
4506 	if (error)
4507 		goto out;
4508 	/*
4509 	 * from now on we have to make sure not
4510 	 * to forget about the vnode
4511 	 * any error that causes an abort must vput(vp)
4512 	 * just set error = err and 'goto bad;'.
4513 	 */
4514 
4515 	/*
4516 	 * from vn_open
4517 	 */
4518 	if (vp->v_type == VLNK) {
4519 		error = EMLINK;
4520 		goto bad;
4521 	}
4522 	if (vp->v_type == VSOCK) {
4523 		error = EOPNOTSUPP;
4524 		goto bad;
4525 	}
4526 	if (vp->v_type != VDIR && fmode & O_DIRECTORY) {
4527 		error = ENOTDIR;
4528 		goto bad;
4529 	}
4530 	accmode = 0;
4531 	if (fmode & (FWRITE | O_TRUNC)) {
4532 		if (vp->v_type == VDIR) {
4533 			error = EISDIR;
4534 			goto bad;
4535 		}
4536 		error = vn_writechk(vp);
4537 		if (error)
4538 			goto bad;
4539 		accmode |= VWRITE;
4540 	}
4541 	if (fmode & FREAD)
4542 		accmode |= VREAD;
4543 	if ((fmode & O_APPEND) && (fmode & FWRITE))
4544 		accmode |= VAPPEND;
4545 #ifdef MAC
4546 	error = mac_vnode_check_open(td->td_ucred, vp, accmode);
4547 	if (error)
4548 		goto bad;
4549 #endif
4550 	if (accmode) {
4551 		error = VOP_ACCESS(vp, accmode, td->td_ucred, td);
4552 		if (error)
4553 			goto bad;
4554 	}
4555 	if (fmode & O_TRUNC) {
4556 		vfs_ref(mp);
4557 		VOP_UNLOCK(vp, 0);				/* XXX */
4558 		if ((error = vn_start_write(NULL, &mp, V_WAIT | PCATCH)) != 0) {
4559 			vrele(vp);
4560 			vfs_rel(mp);
4561 			goto out;
4562 		}
4563 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);	/* XXX */
4564 		vfs_rel(mp);
4565 #ifdef MAC
4566 		/*
4567 		 * We don't yet have fp->f_cred, so use td->td_ucred, which
4568 		 * should be right.
4569 		 */
4570 		error = mac_vnode_check_write(td->td_ucred, td->td_ucred, vp);
4571 		if (error == 0) {
4572 #endif
4573 			VATTR_NULL(vap);
4574 			vap->va_size = 0;
4575 			error = VOP_SETATTR(vp, vap, td->td_ucred);
4576 #ifdef MAC
4577 		}
4578 #endif
4579 		vn_finished_write(mp);
4580 		if (error)
4581 			goto bad;
4582 	}
4583 	error = VOP_OPEN(vp, fmode, td->td_ucred, td, NULL);
4584 	if (error)
4585 		goto bad;
4586 
4587 	if (fmode & FWRITE)
4588 		vp->v_writecount++;
4589 
4590 	/*
4591 	 * end of vn_open code
4592 	 */
4593 
4594 	if ((error = falloc(td, &nfp, &indx, fmode)) != 0) {
4595 		if (fmode & FWRITE)
4596 			vp->v_writecount--;
4597 		goto bad;
4598 	}
4599 	/* An extra reference on `nfp' has been held for us by falloc(). */
4600 	fp = nfp;
4601 	nfp->f_vnode = vp;
4602 	finit(nfp, fmode & FMASK, DTYPE_VNODE, vp, &vnops);
4603 	if (fmode & (O_EXLOCK | O_SHLOCK)) {
4604 		lf.l_whence = SEEK_SET;
4605 		lf.l_start = 0;
4606 		lf.l_len = 0;
4607 		if (fmode & O_EXLOCK)
4608 			lf.l_type = F_WRLCK;
4609 		else
4610 			lf.l_type = F_RDLCK;
4611 		type = F_FLOCK;
4612 		if ((fmode & FNONBLOCK) == 0)
4613 			type |= F_WAIT;
4614 		VOP_UNLOCK(vp, 0);
4615 		if ((error = VOP_ADVLOCK(vp, (caddr_t)fp, F_SETLK, &lf,
4616 			    type)) != 0) {
4617 			/*
4618 			 * The lock request failed.  Normally close the
4619 			 * descriptor but handle the case where someone might
4620 			 * have dup()d or close()d it when we weren't looking.
4621 			 */
4622 			fdclose(fdp, fp, indx, td);
4623 
4624 			/*
4625 			 * release our private reference
4626 			 */
4627 			fdrop(fp, td);
4628 			goto out;
4629 		}
4630 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
4631 		atomic_set_int(&fp->f_flag, FHASLOCK);
4632 	}
4633 
4634 	VOP_UNLOCK(vp, 0);
4635 	fdrop(fp, td);
4636 	VFS_UNLOCK_GIANT(vfslocked);
4637 	td->td_retval[0] = indx;
4638 	return (0);
4639 
4640 bad:
4641 	vput(vp);
4642 out:
4643 	VFS_UNLOCK_GIANT(vfslocked);
4644 	return (error);
4645 }
4646 
4647 /*
4648  * Stat an (NFS) file handle.
4649  */
4650 #ifndef _SYS_SYSPROTO_H_
4651 struct fhstat_args {
4652 	struct fhandle *u_fhp;
4653 	struct stat *sb;
4654 };
4655 #endif
4656 int
4657 sys_fhstat(td, uap)
4658 	struct thread *td;
4659 	register struct fhstat_args /* {
4660 		struct fhandle *u_fhp;
4661 		struct stat *sb;
4662 	} */ *uap;
4663 {
4664 	struct stat sb;
4665 	fhandle_t fh;
4666 	struct mount *mp;
4667 	struct vnode *vp;
4668 	int vfslocked;
4669 	int error;
4670 
4671 	error = priv_check(td, PRIV_VFS_FHSTAT);
4672 	if (error)
4673 		return (error);
4674 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
4675 	if (error)
4676 		return (error);
4677 	if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4678 		return (ESTALE);
4679 	vfslocked = VFS_LOCK_GIANT(mp);
4680 	error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4681 	vfs_unbusy(mp);
4682 	if (error) {
4683 		VFS_UNLOCK_GIANT(vfslocked);
4684 		return (error);
4685 	}
4686 	error = vn_stat(vp, &sb, td->td_ucred, NOCRED, td);
4687 	vput(vp);
4688 	VFS_UNLOCK_GIANT(vfslocked);
4689 	if (error)
4690 		return (error);
4691 	error = copyout(&sb, uap->sb, sizeof(sb));
4692 	return (error);
4693 }
4694 
4695 /*
4696  * Implement fstatfs() for (NFS) file handles.
4697  */
4698 #ifndef _SYS_SYSPROTO_H_
4699 struct fhstatfs_args {
4700 	struct fhandle *u_fhp;
4701 	struct statfs *buf;
4702 };
4703 #endif
4704 int
4705 sys_fhstatfs(td, uap)
4706 	struct thread *td;
4707 	struct fhstatfs_args /* {
4708 		struct fhandle *u_fhp;
4709 		struct statfs *buf;
4710 	} */ *uap;
4711 {
4712 	struct statfs sf;
4713 	fhandle_t fh;
4714 	int error;
4715 
4716 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
4717 	if (error)
4718 		return (error);
4719 	error = kern_fhstatfs(td, fh, &sf);
4720 	if (error)
4721 		return (error);
4722 	return (copyout(&sf, uap->buf, sizeof(sf)));
4723 }
4724 
4725 int
4726 kern_fhstatfs(struct thread *td, fhandle_t fh, struct statfs *buf)
4727 {
4728 	struct statfs *sp;
4729 	struct mount *mp;
4730 	struct vnode *vp;
4731 	int vfslocked;
4732 	int error;
4733 
4734 	error = priv_check(td, PRIV_VFS_FHSTATFS);
4735 	if (error)
4736 		return (error);
4737 	if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4738 		return (ESTALE);
4739 	vfslocked = VFS_LOCK_GIANT(mp);
4740 	error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4741 	if (error) {
4742 		vfs_unbusy(mp);
4743 		VFS_UNLOCK_GIANT(vfslocked);
4744 		return (error);
4745 	}
4746 	vput(vp);
4747 	error = prison_canseemount(td->td_ucred, mp);
4748 	if (error)
4749 		goto out;
4750 #ifdef MAC
4751 	error = mac_mount_check_stat(td->td_ucred, mp);
4752 	if (error)
4753 		goto out;
4754 #endif
4755 	/*
4756 	 * Set these in case the underlying filesystem fails to do so.
4757 	 */
4758 	sp = &mp->mnt_stat;
4759 	sp->f_version = STATFS_VERSION;
4760 	sp->f_namemax = NAME_MAX;
4761 	sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
4762 	error = VFS_STATFS(mp, sp);
4763 	if (error == 0)
4764 		*buf = *sp;
4765 out:
4766 	vfs_unbusy(mp);
4767 	VFS_UNLOCK_GIANT(vfslocked);
4768 	return (error);
4769 }
4770 
4771 int
4772 kern_posix_fallocate(struct thread *td, int fd, off_t offset, off_t len)
4773 {
4774 	struct file *fp;
4775 	struct mount *mp;
4776 	struct vnode *vp;
4777 	off_t olen, ooffset;
4778 	int error, vfslocked;
4779 
4780 	fp = NULL;
4781 	vfslocked = 0;
4782 	error = fget(td, fd, CAP_WRITE, &fp);
4783 	if (error != 0)
4784 		goto out;
4785 
4786 	switch (fp->f_type) {
4787 	case DTYPE_VNODE:
4788 		break;
4789 	case DTYPE_PIPE:
4790 	case DTYPE_FIFO:
4791 		error = ESPIPE;
4792 		goto out;
4793 	default:
4794 		error = ENODEV;
4795 		goto out;
4796 	}
4797 	if ((fp->f_flag & FWRITE) == 0) {
4798 		error = EBADF;
4799 		goto out;
4800 	}
4801 	vp = fp->f_vnode;
4802 	if (vp->v_type != VREG) {
4803 		error = ENODEV;
4804 		goto out;
4805 	}
4806 	if (offset < 0 || len <= 0) {
4807 		error = EINVAL;
4808 		goto out;
4809 	}
4810 	/* Check for wrap. */
4811 	if (offset > OFF_MAX - len) {
4812 		error = EFBIG;
4813 		goto out;
4814 	}
4815 
4816 	/* Allocating blocks may take a long time, so iterate. */
4817 	for (;;) {
4818 		olen = len;
4819 		ooffset = offset;
4820 
4821 		bwillwrite();
4822 		vfslocked = VFS_LOCK_GIANT(vp->v_mount);
4823 		mp = NULL;
4824 		error = vn_start_write(vp, &mp, V_WAIT | PCATCH);
4825 		if (error != 0) {
4826 			VFS_UNLOCK_GIANT(vfslocked);
4827 			break;
4828 		}
4829 		error = vn_lock(vp, LK_EXCLUSIVE);
4830 		if (error != 0) {
4831 			vn_finished_write(mp);
4832 			VFS_UNLOCK_GIANT(vfslocked);
4833 			break;
4834 		}
4835 #ifdef MAC
4836 		error = mac_vnode_check_write(td->td_ucred, fp->f_cred, vp);
4837 		if (error == 0)
4838 #endif
4839 			error = VOP_ALLOCATE(vp, &offset, &len);
4840 		VOP_UNLOCK(vp, 0);
4841 		vn_finished_write(mp);
4842 		VFS_UNLOCK_GIANT(vfslocked);
4843 
4844 		if (olen + ooffset != offset + len) {
4845 			panic("offset + len changed from %jx/%jx to %jx/%jx",
4846 			    ooffset, olen, offset, len);
4847 		}
4848 		if (error != 0 || len == 0)
4849 			break;
4850 		KASSERT(olen > len, ("Iteration did not make progress?"));
4851 		maybe_yield();
4852 	}
4853  out:
4854 	if (fp != NULL)
4855 		fdrop(fp, td);
4856 	return (error);
4857 }
4858 
4859 int
4860 sys_posix_fallocate(struct thread *td, struct posix_fallocate_args *uap)
4861 {
4862 
4863 	return (kern_posix_fallocate(td, uap->fd, uap->offset, uap->len));
4864 }
4865 
4866 /*
4867  * Unlike madvise(2), we do not make a best effort to remember every
4868  * possible caching hint.  Instead, we remember the last setting with
4869  * the exception that we will allow POSIX_FADV_NORMAL to adjust the
4870  * region of any current setting.
4871  */
4872 int
4873 kern_posix_fadvise(struct thread *td, int fd, off_t offset, off_t len,
4874     int advice)
4875 {
4876 	struct fadvise_info *fa, *new;
4877 	struct file *fp;
4878 	struct vnode *vp;
4879 	off_t end;
4880 	int error;
4881 
4882 	if (offset < 0 || len < 0 || offset > OFF_MAX - len)
4883 		return (EINVAL);
4884 	switch (advice) {
4885 	case POSIX_FADV_SEQUENTIAL:
4886 	case POSIX_FADV_RANDOM:
4887 	case POSIX_FADV_NOREUSE:
4888 		new = malloc(sizeof(*fa), M_FADVISE, M_WAITOK);
4889 		break;
4890 	case POSIX_FADV_NORMAL:
4891 	case POSIX_FADV_WILLNEED:
4892 	case POSIX_FADV_DONTNEED:
4893 		new = NULL;
4894 		break;
4895 	default:
4896 		return (EINVAL);
4897 	}
4898 	/* XXX: CAP_POSIX_FADVISE? */
4899 	error = fget(td, fd, 0, &fp);
4900 	if (error != 0)
4901 		goto out;
4902 
4903 	switch (fp->f_type) {
4904 	case DTYPE_VNODE:
4905 		break;
4906 	case DTYPE_PIPE:
4907 	case DTYPE_FIFO:
4908 		error = ESPIPE;
4909 		goto out;
4910 	default:
4911 		error = ENODEV;
4912 		goto out;
4913 	}
4914 	vp = fp->f_vnode;
4915 	if (vp->v_type != VREG) {
4916 		error = ENODEV;
4917 		goto out;
4918 	}
4919 	if (len == 0)
4920 		end = OFF_MAX;
4921 	else
4922 		end = offset + len - 1;
4923 	switch (advice) {
4924 	case POSIX_FADV_SEQUENTIAL:
4925 	case POSIX_FADV_RANDOM:
4926 	case POSIX_FADV_NOREUSE:
4927 		/*
4928 		 * Try to merge any existing non-standard region with
4929 		 * this new region if possible, otherwise create a new
4930 		 * non-standard region for this request.
4931 		 */
4932 		mtx_pool_lock(mtxpool_sleep, fp);
4933 		fa = fp->f_advice;
4934 		if (fa != NULL && fa->fa_advice == advice &&
4935 		    ((fa->fa_start <= end && fa->fa_end >= offset) ||
4936 		    (end != OFF_MAX && fa->fa_start == end + 1) ||
4937 		    (fa->fa_end != OFF_MAX && fa->fa_end + 1 == offset))) {
4938 			if (offset < fa->fa_start)
4939 				fa->fa_start = offset;
4940 			if (end > fa->fa_end)
4941 				fa->fa_end = end;
4942 		} else {
4943 			new->fa_advice = advice;
4944 			new->fa_start = offset;
4945 			new->fa_end = end;
4946 			fp->f_advice = new;
4947 			new = fa;
4948 		}
4949 		mtx_pool_unlock(mtxpool_sleep, fp);
4950 		break;
4951 	case POSIX_FADV_NORMAL:
4952 		/*
4953 		 * If a the "normal" region overlaps with an existing
4954 		 * non-standard region, trim or remove the
4955 		 * non-standard region.
4956 		 */
4957 		mtx_pool_lock(mtxpool_sleep, fp);
4958 		fa = fp->f_advice;
4959 		if (fa != NULL) {
4960 			if (offset <= fa->fa_start && end >= fa->fa_end) {
4961 				new = fa;
4962 				fp->f_advice = NULL;
4963 			} else if (offset <= fa->fa_start &&
4964  			    end >= fa->fa_start)
4965 				fa->fa_start = end + 1;
4966 			else if (offset <= fa->fa_end && end >= fa->fa_end)
4967 				fa->fa_end = offset - 1;
4968 			else if (offset >= fa->fa_start && end <= fa->fa_end) {
4969 				/*
4970 				 * If the "normal" region is a middle
4971 				 * portion of the existing
4972 				 * non-standard region, just remove
4973 				 * the whole thing rather than picking
4974 				 * one side or the other to
4975 				 * preserve.
4976 				 */
4977 				new = fa;
4978 				fp->f_advice = NULL;
4979 			}
4980 		}
4981 		mtx_pool_unlock(mtxpool_sleep, fp);
4982 		break;
4983 	case POSIX_FADV_WILLNEED:
4984 	case POSIX_FADV_DONTNEED:
4985 		error = VOP_ADVISE(vp, offset, end, advice);
4986 		break;
4987 	}
4988 out:
4989 	if (fp != NULL)
4990 		fdrop(fp, td);
4991 	free(new, M_FADVISE);
4992 	return (error);
4993 }
4994 
4995 int
4996 sys_posix_fadvise(struct thread *td, struct posix_fadvise_args *uap)
4997 {
4998 
4999 	return (kern_posix_fadvise(td, uap->fd, uap->offset, uap->len,
5000 	    uap->advice));
5001 }
5002