xref: /freebsd/sys/kern/vfs_syscalls.c (revision 145992504973bd16cf3518af9ba5ce185fefa82a)
1 /*-
2  * Copyright (c) 1989, 1993
3  *	The Regents of the University of California.  All rights reserved.
4  * (c) UNIX System Laboratories, Inc.
5  * All or some portions of this file are derived from material licensed
6  * to the University of California by American Telephone and Telegraph
7  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
8  * the permission of UNIX System Laboratories, Inc.
9  *
10  * Redistribution and use in source and binary forms, with or without
11  * modification, are permitted provided that the following conditions
12  * are met:
13  * 1. Redistributions of source code must retain the above copyright
14  *    notice, this list of conditions and the following disclaimer.
15  * 2. Redistributions in binary form must reproduce the above copyright
16  *    notice, this list of conditions and the following disclaimer in the
17  *    documentation and/or other materials provided with the distribution.
18  * 4. Neither the name of the University nor the names of its contributors
19  *    may be used to endorse or promote products derived from this software
20  *    without specific prior written permission.
21  *
22  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
23  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
24  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
25  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
26  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
27  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
28  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
29  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
30  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
31  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32  * SUCH DAMAGE.
33  *
34  *	@(#)vfs_syscalls.c	8.13 (Berkeley) 4/15/94
35  */
36 
37 #include <sys/cdefs.h>
38 __FBSDID("$FreeBSD$");
39 
40 #include "opt_capsicum.h"
41 #include "opt_compat.h"
42 #include "opt_kdtrace.h"
43 #include "opt_ktrace.h"
44 
45 #include <sys/param.h>
46 #include <sys/systm.h>
47 #include <sys/bio.h>
48 #include <sys/buf.h>
49 #include <sys/capability.h>
50 #include <sys/disk.h>
51 #include <sys/sysent.h>
52 #include <sys/malloc.h>
53 #include <sys/mount.h>
54 #include <sys/mutex.h>
55 #include <sys/sysproto.h>
56 #include <sys/namei.h>
57 #include <sys/filedesc.h>
58 #include <sys/kernel.h>
59 #include <sys/fcntl.h>
60 #include <sys/file.h>
61 #include <sys/filio.h>
62 #include <sys/limits.h>
63 #include <sys/linker.h>
64 #include <sys/sdt.h>
65 #include <sys/stat.h>
66 #include <sys/sx.h>
67 #include <sys/unistd.h>
68 #include <sys/vnode.h>
69 #include <sys/priv.h>
70 #include <sys/proc.h>
71 #include <sys/dirent.h>
72 #include <sys/jail.h>
73 #include <sys/syscallsubr.h>
74 #include <sys/sysctl.h>
75 #ifdef KTRACE
76 #include <sys/ktrace.h>
77 #endif
78 
79 #include <machine/stdarg.h>
80 
81 #include <security/audit/audit.h>
82 #include <security/mac/mac_framework.h>
83 
84 #include <vm/vm.h>
85 #include <vm/vm_object.h>
86 #include <vm/vm_page.h>
87 #include <vm/uma.h>
88 
89 #include <ufs/ufs/quota.h>
90 
91 MALLOC_DEFINE(M_FADVISE, "fadvise", "posix_fadvise(2) information");
92 
93 SDT_PROVIDER_DEFINE(vfs);
94 SDT_PROBE_DEFINE(vfs, , stat, mode, mode);
95 SDT_PROBE_ARGTYPE(vfs, , stat, mode, 0, "char *");
96 SDT_PROBE_ARGTYPE(vfs, , stat, mode, 1, "int");
97 SDT_PROBE_DEFINE(vfs, , stat, reg, reg);
98 SDT_PROBE_ARGTYPE(vfs, , stat, reg, 0, "char *");
99 SDT_PROBE_ARGTYPE(vfs, , stat, reg, 1, "int");
100 
101 static int chroot_refuse_vdir_fds(struct filedesc *fdp);
102 static int getutimes(const struct timeval *, enum uio_seg, struct timespec *);
103 static int setfflags(struct thread *td, struct vnode *, int);
104 static int setutimes(struct thread *td, struct vnode *,
105     const struct timespec *, int, int);
106 static int vn_access(struct vnode *vp, int user_flags, struct ucred *cred,
107     struct thread *td);
108 
109 /*
110  * The module initialization routine for POSIX asynchronous I/O will
111  * set this to the version of AIO that it implements.  (Zero means
112  * that it is not implemented.)  This value is used here by pathconf()
113  * and in kern_descrip.c by fpathconf().
114  */
115 int async_io_version;
116 
117 #ifdef DEBUG
118 static int syncprt = 0;
119 SYSCTL_INT(_debug, OID_AUTO, syncprt, CTLFLAG_RW, &syncprt, 0, "");
120 #endif
121 
122 /*
123  * Sync each mounted filesystem.
124  */
125 #ifndef _SYS_SYSPROTO_H_
126 struct sync_args {
127 	int     dummy;
128 };
129 #endif
130 /* ARGSUSED */
131 int
132 sys_sync(td, uap)
133 	struct thread *td;
134 	struct sync_args *uap;
135 {
136 	struct mount *mp, *nmp;
137 	int save;
138 
139 	mtx_lock(&mountlist_mtx);
140 	for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
141 		if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK)) {
142 			nmp = TAILQ_NEXT(mp, mnt_list);
143 			continue;
144 		}
145 		if ((mp->mnt_flag & MNT_RDONLY) == 0 &&
146 		    vn_start_write(NULL, &mp, V_NOWAIT) == 0) {
147 			save = curthread_pflags_set(TDP_SYNCIO);
148 			vfs_msync(mp, MNT_NOWAIT);
149 			VFS_SYNC(mp, MNT_NOWAIT);
150 			curthread_pflags_restore(save);
151 			vn_finished_write(mp);
152 		}
153 		mtx_lock(&mountlist_mtx);
154 		nmp = TAILQ_NEXT(mp, mnt_list);
155 		vfs_unbusy(mp);
156 	}
157 	mtx_unlock(&mountlist_mtx);
158 	return (0);
159 }
160 
161 /*
162  * Change filesystem quotas.
163  */
164 #ifndef _SYS_SYSPROTO_H_
165 struct quotactl_args {
166 	char *path;
167 	int cmd;
168 	int uid;
169 	caddr_t arg;
170 };
171 #endif
172 int
173 sys_quotactl(td, uap)
174 	struct thread *td;
175 	register struct quotactl_args /* {
176 		char *path;
177 		int cmd;
178 		int uid;
179 		caddr_t arg;
180 	} */ *uap;
181 {
182 	struct mount *mp;
183 	int error;
184 	struct nameidata nd;
185 
186 	AUDIT_ARG_CMD(uap->cmd);
187 	AUDIT_ARG_UID(uap->uid);
188 	if (!prison_allow(td->td_ucred, PR_ALLOW_QUOTAS))
189 		return (EPERM);
190 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1,
191 	   UIO_USERSPACE, uap->path, td);
192 	if ((error = namei(&nd)) != 0)
193 		return (error);
194 	NDFREE(&nd, NDF_ONLY_PNBUF);
195 	mp = nd.ni_vp->v_mount;
196 	vfs_ref(mp);
197 	vput(nd.ni_vp);
198 	error = vfs_busy(mp, 0);
199 	vfs_rel(mp);
200 	if (error)
201 		return (error);
202 	error = VFS_QUOTACTL(mp, uap->cmd, uap->uid, uap->arg);
203 
204 	/*
205 	 * Since quota on operation typically needs to open quota
206 	 * file, the Q_QUOTAON handler needs to unbusy the mount point
207 	 * before calling into namei.  Otherwise, unmount might be
208 	 * started between two vfs_busy() invocations (first is our,
209 	 * second is from mount point cross-walk code in lookup()),
210 	 * causing deadlock.
211 	 *
212 	 * Require that Q_QUOTAON handles the vfs_busy() reference on
213 	 * its own, always returning with ubusied mount point.
214 	 */
215 	if ((uap->cmd >> SUBCMDSHIFT) != Q_QUOTAON)
216 		vfs_unbusy(mp);
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 error;
294 	struct nameidata nd;
295 
296 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF |
297 	    AUDITVNODE1, pathseg, path, td);
298 	error = namei(&nd);
299 	if (error)
300 		return (error);
301 	mp = nd.ni_vp->v_mount;
302 	vfs_ref(mp);
303 	NDFREE(&nd, NDF_ONLY_PNBUF);
304 	vput(nd.ni_vp);
305 	error = vfs_busy(mp, 0);
306 	vfs_rel(mp);
307 	if (error)
308 		return (error);
309 #ifdef MAC
310 	error = mac_mount_check_stat(td->td_ucred, mp);
311 	if (error)
312 		goto out;
313 #endif
314 	/*
315 	 * Set these in case the underlying filesystem fails to do so.
316 	 */
317 	sp = &mp->mnt_stat;
318 	sp->f_version = STATFS_VERSION;
319 	sp->f_namemax = NAME_MAX;
320 	sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
321 	error = VFS_STATFS(mp, sp);
322 	if (error)
323 		goto out;
324 	if (priv_check(td, PRIV_VFS_GENERATION)) {
325 		bcopy(sp, &sb, sizeof(sb));
326 		sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
327 		prison_enforce_statfs(td->td_ucred, mp, &sb);
328 		sp = &sb;
329 	}
330 	*buf = *sp;
331 out:
332 	vfs_unbusy(mp);
333 	return (error);
334 }
335 
336 /*
337  * Get filesystem statistics.
338  */
339 #ifndef _SYS_SYSPROTO_H_
340 struct fstatfs_args {
341 	int fd;
342 	struct statfs *buf;
343 };
344 #endif
345 int
346 sys_fstatfs(td, uap)
347 	struct thread *td;
348 	register struct fstatfs_args /* {
349 		int fd;
350 		struct statfs *buf;
351 	} */ *uap;
352 {
353 	struct statfs sf;
354 	int error;
355 
356 	error = kern_fstatfs(td, uap->fd, &sf);
357 	if (error == 0)
358 		error = copyout(&sf, uap->buf, sizeof(sf));
359 	return (error);
360 }
361 
362 int
363 kern_fstatfs(struct thread *td, int fd, struct statfs *buf)
364 {
365 	struct file *fp;
366 	struct mount *mp;
367 	struct statfs *sp, sb;
368 	struct vnode *vp;
369 	int error;
370 
371 	AUDIT_ARG_FD(fd);
372 	error = getvnode(td->td_proc->p_fd, fd, CAP_FSTATFS, &fp);
373 	if (error)
374 		return (error);
375 	vp = fp->f_vnode;
376 	vn_lock(vp, LK_SHARED | LK_RETRY);
377 #ifdef AUDIT
378 	AUDIT_ARG_VNODE1(vp);
379 #endif
380 	mp = vp->v_mount;
381 	if (mp)
382 		vfs_ref(mp);
383 	VOP_UNLOCK(vp, 0);
384 	fdrop(fp, td);
385 	if (mp == NULL) {
386 		error = EBADF;
387 		goto out;
388 	}
389 	error = vfs_busy(mp, 0);
390 	vfs_rel(mp);
391 	if (error)
392 		return (error);
393 #ifdef MAC
394 	error = mac_mount_check_stat(td->td_ucred, mp);
395 	if (error)
396 		goto out;
397 #endif
398 	/*
399 	 * Set these in case the underlying filesystem fails to do so.
400 	 */
401 	sp = &mp->mnt_stat;
402 	sp->f_version = STATFS_VERSION;
403 	sp->f_namemax = NAME_MAX;
404 	sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
405 	error = VFS_STATFS(mp, sp);
406 	if (error)
407 		goto out;
408 	if (priv_check(td, PRIV_VFS_GENERATION)) {
409 		bcopy(sp, &sb, sizeof(sb));
410 		sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
411 		prison_enforce_statfs(td->td_ucred, mp, &sb);
412 		sp = &sb;
413 	}
414 	*buf = *sp;
415 out:
416 	if (mp)
417 		vfs_unbusy(mp);
418 	return (error);
419 }
420 
421 /*
422  * Get statistics on all filesystems.
423  */
424 #ifndef _SYS_SYSPROTO_H_
425 struct getfsstat_args {
426 	struct statfs *buf;
427 	long bufsize;
428 	int flags;
429 };
430 #endif
431 int
432 sys_getfsstat(td, uap)
433 	struct thread *td;
434 	register struct getfsstat_args /* {
435 		struct statfs *buf;
436 		long bufsize;
437 		int flags;
438 	} */ *uap;
439 {
440 
441 	return (kern_getfsstat(td, &uap->buf, uap->bufsize, UIO_USERSPACE,
442 	    uap->flags));
443 }
444 
445 /*
446  * If (bufsize > 0 && bufseg == UIO_SYSSPACE)
447  * 	The caller is responsible for freeing memory which will be allocated
448  *	in '*buf'.
449  */
450 int
451 kern_getfsstat(struct thread *td, struct statfs **buf, size_t bufsize,
452     enum uio_seg bufseg, int flags)
453 {
454 	struct mount *mp, *nmp;
455 	struct statfs *sfsp, *sp, sb;
456 	size_t count, maxcount;
457 	int error;
458 
459 	maxcount = bufsize / sizeof(struct statfs);
460 	if (bufsize == 0)
461 		sfsp = NULL;
462 	else if (bufseg == UIO_USERSPACE)
463 		sfsp = *buf;
464 	else /* if (bufseg == UIO_SYSSPACE) */ {
465 		count = 0;
466 		mtx_lock(&mountlist_mtx);
467 		TAILQ_FOREACH(mp, &mountlist, mnt_list) {
468 			count++;
469 		}
470 		mtx_unlock(&mountlist_mtx);
471 		if (maxcount > count)
472 			maxcount = count;
473 		sfsp = *buf = malloc(maxcount * sizeof(struct statfs), M_TEMP,
474 		    M_WAITOK);
475 	}
476 	count = 0;
477 	mtx_lock(&mountlist_mtx);
478 	for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
479 		if (prison_canseemount(td->td_ucred, mp) != 0) {
480 			nmp = TAILQ_NEXT(mp, mnt_list);
481 			continue;
482 		}
483 #ifdef MAC
484 		if (mac_mount_check_stat(td->td_ucred, mp) != 0) {
485 			nmp = TAILQ_NEXT(mp, mnt_list);
486 			continue;
487 		}
488 #endif
489 		if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK)) {
490 			nmp = TAILQ_NEXT(mp, mnt_list);
491 			continue;
492 		}
493 		if (sfsp && count < maxcount) {
494 			sp = &mp->mnt_stat;
495 			/*
496 			 * Set these in case the underlying filesystem
497 			 * fails to do so.
498 			 */
499 			sp->f_version = STATFS_VERSION;
500 			sp->f_namemax = NAME_MAX;
501 			sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
502 			/*
503 			 * If MNT_NOWAIT or MNT_LAZY is specified, do not
504 			 * refresh the fsstat cache. MNT_NOWAIT or MNT_LAZY
505 			 * overrides MNT_WAIT.
506 			 */
507 			if (((flags & (MNT_LAZY|MNT_NOWAIT)) == 0 ||
508 			    (flags & MNT_WAIT)) &&
509 			    (error = VFS_STATFS(mp, sp))) {
510 				mtx_lock(&mountlist_mtx);
511 				nmp = TAILQ_NEXT(mp, mnt_list);
512 				vfs_unbusy(mp);
513 				continue;
514 			}
515 			if (priv_check(td, PRIV_VFS_GENERATION)) {
516 				bcopy(sp, &sb, sizeof(sb));
517 				sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
518 				prison_enforce_statfs(td->td_ucred, mp, &sb);
519 				sp = &sb;
520 			}
521 			if (bufseg == UIO_SYSSPACE)
522 				bcopy(sp, sfsp, sizeof(*sp));
523 			else /* if (bufseg == UIO_USERSPACE) */ {
524 				error = copyout(sp, sfsp, sizeof(*sp));
525 				if (error) {
526 					vfs_unbusy(mp);
527 					return (error);
528 				}
529 			}
530 			sfsp++;
531 		}
532 		count++;
533 		mtx_lock(&mountlist_mtx);
534 		nmp = TAILQ_NEXT(mp, mnt_list);
535 		vfs_unbusy(mp);
536 	}
537 	mtx_unlock(&mountlist_mtx);
538 	if (sfsp && count > maxcount)
539 		td->td_retval[0] = maxcount;
540 	else
541 		td->td_retval[0] = count;
542 	return (0);
543 }
544 
545 #ifdef COMPAT_FREEBSD4
546 /*
547  * Get old format filesystem statistics.
548  */
549 static void cvtstatfs(struct statfs *, struct ostatfs *);
550 
551 #ifndef _SYS_SYSPROTO_H_
552 struct freebsd4_statfs_args {
553 	char *path;
554 	struct ostatfs *buf;
555 };
556 #endif
557 int
558 freebsd4_statfs(td, uap)
559 	struct thread *td;
560 	struct freebsd4_statfs_args /* {
561 		char *path;
562 		struct ostatfs *buf;
563 	} */ *uap;
564 {
565 	struct ostatfs osb;
566 	struct statfs sf;
567 	int error;
568 
569 	error = kern_statfs(td, uap->path, UIO_USERSPACE, &sf);
570 	if (error)
571 		return (error);
572 	cvtstatfs(&sf, &osb);
573 	return (copyout(&osb, uap->buf, sizeof(osb)));
574 }
575 
576 /*
577  * Get filesystem statistics.
578  */
579 #ifndef _SYS_SYSPROTO_H_
580 struct freebsd4_fstatfs_args {
581 	int fd;
582 	struct ostatfs *buf;
583 };
584 #endif
585 int
586 freebsd4_fstatfs(td, uap)
587 	struct thread *td;
588 	struct freebsd4_fstatfs_args /* {
589 		int fd;
590 		struct ostatfs *buf;
591 	} */ *uap;
592 {
593 	struct ostatfs osb;
594 	struct statfs sf;
595 	int error;
596 
597 	error = kern_fstatfs(td, uap->fd, &sf);
598 	if (error)
599 		return (error);
600 	cvtstatfs(&sf, &osb);
601 	return (copyout(&osb, uap->buf, sizeof(osb)));
602 }
603 
604 /*
605  * Get statistics on all filesystems.
606  */
607 #ifndef _SYS_SYSPROTO_H_
608 struct freebsd4_getfsstat_args {
609 	struct ostatfs *buf;
610 	long bufsize;
611 	int flags;
612 };
613 #endif
614 int
615 freebsd4_getfsstat(td, uap)
616 	struct thread *td;
617 	register struct freebsd4_getfsstat_args /* {
618 		struct ostatfs *buf;
619 		long bufsize;
620 		int flags;
621 	} */ *uap;
622 {
623 	struct statfs *buf, *sp;
624 	struct ostatfs osb;
625 	size_t count, size;
626 	int error;
627 
628 	count = uap->bufsize / sizeof(struct ostatfs);
629 	size = count * sizeof(struct statfs);
630 	error = kern_getfsstat(td, &buf, size, UIO_SYSSPACE, uap->flags);
631 	if (size > 0) {
632 		count = td->td_retval[0];
633 		sp = buf;
634 		while (count > 0 && error == 0) {
635 			cvtstatfs(sp, &osb);
636 			error = copyout(&osb, uap->buf, sizeof(osb));
637 			sp++;
638 			uap->buf++;
639 			count--;
640 		}
641 		free(buf, M_TEMP);
642 	}
643 	return (error);
644 }
645 
646 /*
647  * Implement fstatfs() for (NFS) file handles.
648  */
649 #ifndef _SYS_SYSPROTO_H_
650 struct freebsd4_fhstatfs_args {
651 	struct fhandle *u_fhp;
652 	struct ostatfs *buf;
653 };
654 #endif
655 int
656 freebsd4_fhstatfs(td, uap)
657 	struct thread *td;
658 	struct freebsd4_fhstatfs_args /* {
659 		struct fhandle *u_fhp;
660 		struct ostatfs *buf;
661 	} */ *uap;
662 {
663 	struct ostatfs osb;
664 	struct statfs sf;
665 	fhandle_t fh;
666 	int error;
667 
668 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
669 	if (error)
670 		return (error);
671 	error = kern_fhstatfs(td, fh, &sf);
672 	if (error)
673 		return (error);
674 	cvtstatfs(&sf, &osb);
675 	return (copyout(&osb, uap->buf, sizeof(osb)));
676 }
677 
678 /*
679  * Convert a new format statfs structure to an old format statfs structure.
680  */
681 static void
682 cvtstatfs(nsp, osp)
683 	struct statfs *nsp;
684 	struct ostatfs *osp;
685 {
686 
687 	statfs_scale_blocks(nsp, LONG_MAX);
688 	bzero(osp, sizeof(*osp));
689 	osp->f_bsize = nsp->f_bsize;
690 	osp->f_iosize = MIN(nsp->f_iosize, LONG_MAX);
691 	osp->f_blocks = nsp->f_blocks;
692 	osp->f_bfree = nsp->f_bfree;
693 	osp->f_bavail = nsp->f_bavail;
694 	osp->f_files = MIN(nsp->f_files, LONG_MAX);
695 	osp->f_ffree = MIN(nsp->f_ffree, LONG_MAX);
696 	osp->f_owner = nsp->f_owner;
697 	osp->f_type = nsp->f_type;
698 	osp->f_flags = nsp->f_flags;
699 	osp->f_syncwrites = MIN(nsp->f_syncwrites, LONG_MAX);
700 	osp->f_asyncwrites = MIN(nsp->f_asyncwrites, LONG_MAX);
701 	osp->f_syncreads = MIN(nsp->f_syncreads, LONG_MAX);
702 	osp->f_asyncreads = MIN(nsp->f_asyncreads, LONG_MAX);
703 	strlcpy(osp->f_fstypename, nsp->f_fstypename,
704 	    MIN(MFSNAMELEN, OMFSNAMELEN));
705 	strlcpy(osp->f_mntonname, nsp->f_mntonname,
706 	    MIN(MNAMELEN, OMNAMELEN));
707 	strlcpy(osp->f_mntfromname, nsp->f_mntfromname,
708 	    MIN(MNAMELEN, OMNAMELEN));
709 	osp->f_fsid = nsp->f_fsid;
710 }
711 #endif /* COMPAT_FREEBSD4 */
712 
713 /*
714  * Change current working directory to a given file descriptor.
715  */
716 #ifndef _SYS_SYSPROTO_H_
717 struct fchdir_args {
718 	int	fd;
719 };
720 #endif
721 int
722 sys_fchdir(td, uap)
723 	struct thread *td;
724 	struct fchdir_args /* {
725 		int fd;
726 	} */ *uap;
727 {
728 	register struct filedesc *fdp = td->td_proc->p_fd;
729 	struct vnode *vp, *tdp, *vpold;
730 	struct mount *mp;
731 	struct file *fp;
732 	int error;
733 
734 	AUDIT_ARG_FD(uap->fd);
735 	if ((error = getvnode(fdp, uap->fd, CAP_FCHDIR, &fp)) != 0)
736 		return (error);
737 	vp = fp->f_vnode;
738 	VREF(vp);
739 	fdrop(fp, td);
740 	vn_lock(vp, LK_SHARED | LK_RETRY);
741 	AUDIT_ARG_VNODE1(vp);
742 	error = change_dir(vp, td);
743 	while (!error && (mp = vp->v_mountedhere) != NULL) {
744 		if (vfs_busy(mp, 0))
745 			continue;
746 		error = VFS_ROOT(mp, LK_SHARED, &tdp);
747 		vfs_unbusy(mp);
748 		if (error)
749 			break;
750 		vput(vp);
751 		vp = tdp;
752 	}
753 	if (error) {
754 		vput(vp);
755 		return (error);
756 	}
757 	VOP_UNLOCK(vp, 0);
758 	FILEDESC_XLOCK(fdp);
759 	vpold = fdp->fd_cdir;
760 	fdp->fd_cdir = vp;
761 	FILEDESC_XUNLOCK(fdp);
762 	vrele(vpold);
763 	return (0);
764 }
765 
766 /*
767  * Change current working directory (``.'').
768  */
769 #ifndef _SYS_SYSPROTO_H_
770 struct chdir_args {
771 	char	*path;
772 };
773 #endif
774 int
775 sys_chdir(td, uap)
776 	struct thread *td;
777 	struct chdir_args /* {
778 		char *path;
779 	} */ *uap;
780 {
781 
782 	return (kern_chdir(td, uap->path, UIO_USERSPACE));
783 }
784 
785 int
786 kern_chdir(struct thread *td, char *path, enum uio_seg pathseg)
787 {
788 	register struct filedesc *fdp = td->td_proc->p_fd;
789 	int error;
790 	struct nameidata nd;
791 	struct vnode *vp;
792 
793 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
794 	    pathseg, path, td);
795 	if ((error = namei(&nd)) != 0)
796 		return (error);
797 	if ((error = change_dir(nd.ni_vp, td)) != 0) {
798 		vput(nd.ni_vp);
799 		NDFREE(&nd, NDF_ONLY_PNBUF);
800 		return (error);
801 	}
802 	VOP_UNLOCK(nd.ni_vp, 0);
803 	NDFREE(&nd, NDF_ONLY_PNBUF);
804 	FILEDESC_XLOCK(fdp);
805 	vp = fdp->fd_cdir;
806 	fdp->fd_cdir = nd.ni_vp;
807 	FILEDESC_XUNLOCK(fdp);
808 	vrele(vp);
809 	return (0);
810 }
811 
812 /*
813  * Helper function for raised chroot(2) security function:  Refuse if
814  * any filedescriptors are open directories.
815  */
816 static int
817 chroot_refuse_vdir_fds(fdp)
818 	struct filedesc *fdp;
819 {
820 	struct vnode *vp;
821 	struct file *fp;
822 	int fd;
823 
824 	FILEDESC_LOCK_ASSERT(fdp);
825 
826 	for (fd = 0; fd < fdp->fd_nfiles ; fd++) {
827 		fp = fget_locked(fdp, fd);
828 		if (fp == NULL)
829 			continue;
830 		if (fp->f_type == DTYPE_VNODE) {
831 			vp = fp->f_vnode;
832 			if (vp->v_type == VDIR)
833 				return (EPERM);
834 		}
835 	}
836 	return (0);
837 }
838 
839 /*
840  * This sysctl determines if we will allow a process to chroot(2) if it
841  * has a directory open:
842  *	0: disallowed for all processes.
843  *	1: allowed for processes that were not already chroot(2)'ed.
844  *	2: allowed for all processes.
845  */
846 
847 static int chroot_allow_open_directories = 1;
848 
849 SYSCTL_INT(_kern, OID_AUTO, chroot_allow_open_directories, CTLFLAG_RW,
850      &chroot_allow_open_directories, 0,
851      "Allow a process to chroot(2) if it has a directory open");
852 
853 /*
854  * Change notion of root (``/'') directory.
855  */
856 #ifndef _SYS_SYSPROTO_H_
857 struct chroot_args {
858 	char	*path;
859 };
860 #endif
861 int
862 sys_chroot(td, uap)
863 	struct thread *td;
864 	struct chroot_args /* {
865 		char *path;
866 	} */ *uap;
867 {
868 	int error;
869 	struct nameidata nd;
870 
871 	error = priv_check(td, PRIV_VFS_CHROOT);
872 	if (error)
873 		return (error);
874 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF |
875 	    AUDITVNODE1, UIO_USERSPACE, uap->path, td);
876 	error = namei(&nd);
877 	if (error)
878 		goto error;
879 	if ((error = change_dir(nd.ni_vp, td)) != 0)
880 		goto e_vunlock;
881 #ifdef MAC
882 	if ((error = mac_vnode_check_chroot(td->td_ucred, nd.ni_vp)))
883 		goto e_vunlock;
884 #endif
885 	VOP_UNLOCK(nd.ni_vp, 0);
886 	error = change_root(nd.ni_vp, td);
887 	vrele(nd.ni_vp);
888 	NDFREE(&nd, NDF_ONLY_PNBUF);
889 	return (error);
890 e_vunlock:
891 	vput(nd.ni_vp);
892 error:
893 	NDFREE(&nd, NDF_ONLY_PNBUF);
894 	return (error);
895 }
896 
897 /*
898  * Common routine for chroot and chdir.  Callers must provide a locked vnode
899  * instance.
900  */
901 int
902 change_dir(vp, td)
903 	struct vnode *vp;
904 	struct thread *td;
905 {
906 	int error;
907 
908 	ASSERT_VOP_LOCKED(vp, "change_dir(): vp not locked");
909 	if (vp->v_type != VDIR)
910 		return (ENOTDIR);
911 #ifdef MAC
912 	error = mac_vnode_check_chdir(td->td_ucred, vp);
913 	if (error)
914 		return (error);
915 #endif
916 	error = VOP_ACCESS(vp, VEXEC, td->td_ucred, td);
917 	return (error);
918 }
919 
920 /*
921  * Common routine for kern_chroot() and jail_attach().  The caller is
922  * responsible for invoking priv_check() and mac_vnode_check_chroot() to
923  * authorize this operation.
924  */
925 int
926 change_root(vp, td)
927 	struct vnode *vp;
928 	struct thread *td;
929 {
930 	struct filedesc *fdp;
931 	struct vnode *oldvp;
932 	int error;
933 
934 	fdp = td->td_proc->p_fd;
935 	FILEDESC_XLOCK(fdp);
936 	if (chroot_allow_open_directories == 0 ||
937 	    (chroot_allow_open_directories == 1 && fdp->fd_rdir != rootvnode)) {
938 		error = chroot_refuse_vdir_fds(fdp);
939 		if (error) {
940 			FILEDESC_XUNLOCK(fdp);
941 			return (error);
942 		}
943 	}
944 	oldvp = fdp->fd_rdir;
945 	fdp->fd_rdir = vp;
946 	VREF(fdp->fd_rdir);
947 	if (!fdp->fd_jdir) {
948 		fdp->fd_jdir = vp;
949 		VREF(fdp->fd_jdir);
950 	}
951 	FILEDESC_XUNLOCK(fdp);
952 	vrele(oldvp);
953 	return (0);
954 }
955 
956 static __inline cap_rights_t
957 flags_to_rights(int flags)
958 {
959 	cap_rights_t rights = 0;
960 
961 	if (flags & O_EXEC) {
962 		rights |= CAP_FEXECVE;
963 	} else {
964 		switch ((flags & O_ACCMODE)) {
965 		case O_RDONLY:
966 			rights |= CAP_READ;
967 			break;
968 		case O_RDWR:
969 			rights |= CAP_READ;
970 			/* FALLTHROUGH */
971 		case O_WRONLY:
972 			rights |= CAP_WRITE;
973 			break;
974 		}
975 	}
976 
977 	if (flags & O_CREAT)
978 		rights |= CAP_CREATE;
979 
980 	if (flags & O_TRUNC)
981 		rights |= CAP_FTRUNCATE;
982 
983 	if ((flags & O_EXLOCK) || (flags & O_SHLOCK))
984 		rights |= CAP_FLOCK;
985 
986 	return (rights);
987 }
988 
989 /*
990  * Check permissions, allocate an open file structure, and call the device
991  * open routine if any.
992  */
993 #ifndef _SYS_SYSPROTO_H_
994 struct open_args {
995 	char	*path;
996 	int	flags;
997 	int	mode;
998 };
999 #endif
1000 int
1001 sys_open(td, uap)
1002 	struct thread *td;
1003 	register struct open_args /* {
1004 		char *path;
1005 		int flags;
1006 		int mode;
1007 	} */ *uap;
1008 {
1009 
1010 	return (kern_open(td, uap->path, UIO_USERSPACE, uap->flags, uap->mode));
1011 }
1012 
1013 #ifndef _SYS_SYSPROTO_H_
1014 struct openat_args {
1015 	int	fd;
1016 	char	*path;
1017 	int	flag;
1018 	int	mode;
1019 };
1020 #endif
1021 int
1022 sys_openat(struct thread *td, struct openat_args *uap)
1023 {
1024 
1025 	return (kern_openat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
1026 	    uap->mode));
1027 }
1028 
1029 int
1030 kern_open(struct thread *td, char *path, enum uio_seg pathseg, int flags,
1031     int mode)
1032 {
1033 
1034 	return (kern_openat(td, AT_FDCWD, path, pathseg, flags, mode));
1035 }
1036 
1037 int
1038 kern_openat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1039     int flags, int mode)
1040 {
1041 	struct proc *p = td->td_proc;
1042 	struct filedesc *fdp = p->p_fd;
1043 	struct file *fp;
1044 	struct vnode *vp;
1045 	int cmode;
1046 	int indx = -1, error;
1047 	struct nameidata nd;
1048 	cap_rights_t rights_needed = CAP_LOOKUP;
1049 
1050 	AUDIT_ARG_FFLAGS(flags);
1051 	AUDIT_ARG_MODE(mode);
1052 	/* XXX: audit dirfd */
1053 	rights_needed |= flags_to_rights(flags);
1054 	/*
1055 	 * Only one of the O_EXEC, O_RDONLY, O_WRONLY and O_RDWR flags
1056 	 * may be specified.
1057 	 */
1058 	if (flags & O_EXEC) {
1059 		if (flags & O_ACCMODE)
1060 			return (EINVAL);
1061 	} else if ((flags & O_ACCMODE) == O_ACCMODE) {
1062 		return (EINVAL);
1063 	} else {
1064 		flags = FFLAGS(flags);
1065 	}
1066 
1067 	/*
1068 	 * Allocate the file descriptor, but don't install a descriptor yet.
1069 	 */
1070 	error = falloc_noinstall(td, &fp);
1071 	if (error)
1072 		return (error);
1073 	/*
1074 	 * An extra reference on `fp' has been held for us by
1075 	 * falloc_noinstall().
1076 	 */
1077 	/* Set the flags early so the finit in devfs can pick them up. */
1078 	fp->f_flag = flags & FMASK;
1079 	cmode = ((mode &~ fdp->fd_cmask) & ALLPERMS) &~ S_ISTXT;
1080 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg,
1081 	    path, fd, rights_needed, td);
1082 	td->td_dupfd = -1;		/* XXX check for fdopen */
1083 	error = vn_open(&nd, &flags, cmode, fp);
1084 	if (error) {
1085 		/*
1086 		 * If the vn_open replaced the method vector, something
1087 		 * wonderous happened deep below and we just pass it up
1088 		 * pretending we know what we do.
1089 		 */
1090 		if (error == ENXIO && fp->f_ops != &badfileops)
1091 			goto success;
1092 
1093 		/*
1094 		 * Handle special fdopen() case. bleh.
1095 		 *
1096 		 * Don't do this for relative (capability) lookups; we don't
1097 		 * understand exactly what would happen, and we don't think
1098 		 * that it ever should.
1099 		 */
1100 		if (nd.ni_strictrelative == 0 &&
1101 		    (error == ENODEV || error == ENXIO) &&
1102 		    td->td_dupfd >= 0) {
1103 			error = dupfdopen(td, fdp, td->td_dupfd, flags, error,
1104 			    &indx);
1105 			if (error == 0)
1106 				goto success;
1107 		}
1108 
1109 		if (error == ERESTART)
1110 			error = EINTR;
1111 		goto bad_unlocked;
1112 	}
1113 	td->td_dupfd = 0;
1114 	NDFREE(&nd, NDF_ONLY_PNBUF);
1115 	vp = nd.ni_vp;
1116 
1117 	/*
1118 	 * Store the vnode, for any f_type. Typically, the vnode use
1119 	 * count is decremented by direct call to vn_closefile() for
1120 	 * files that switched type in the cdevsw fdopen() method.
1121 	 */
1122 	fp->f_vnode = vp;
1123 	/*
1124 	 * If the file wasn't claimed by devfs bind it to the normal
1125 	 * vnode operations here.
1126 	 */
1127 	if (fp->f_ops == &badfileops) {
1128 		KASSERT(vp->v_type != VFIFO, ("Unexpected fifo."));
1129 		fp->f_seqcount = 1;
1130 		finit(fp, (flags & FMASK) | (fp->f_flag & FHASLOCK), DTYPE_VNODE,
1131 		    vp, &vnops);
1132 	}
1133 
1134 	VOP_UNLOCK(vp, 0);
1135 	if (flags & O_TRUNC) {
1136 		error = fo_truncate(fp, 0, td->td_ucred, td);
1137 		if (error)
1138 			goto bad;
1139 	}
1140 success:
1141 	/*
1142 	 * If we haven't already installed the FD (for dupfdopen), do so now.
1143 	 */
1144 	if (indx == -1) {
1145 #ifdef CAPABILITIES
1146 		if (nd.ni_strictrelative == 1) {
1147 			/*
1148 			 * We are doing a strict relative lookup; wrap the
1149 			 * result in a capability.
1150 			 */
1151 			if ((error = kern_capwrap(td, fp, nd.ni_baserights,
1152 			    &indx)) != 0)
1153 				goto bad_unlocked;
1154 		} else
1155 #endif
1156 			if ((error = finstall(td, fp, &indx, flags)) != 0)
1157 				goto bad_unlocked;
1158 
1159 	}
1160 
1161 	/*
1162 	 * Release our private reference, leaving the one associated with
1163 	 * the descriptor table intact.
1164 	 */
1165 	fdrop(fp, td);
1166 	td->td_retval[0] = indx;
1167 	return (0);
1168 bad:
1169 bad_unlocked:
1170 	KASSERT(indx == -1, ("indx=%d, should be -1", indx));
1171 	fdrop(fp, td);
1172 	return (error);
1173 }
1174 
1175 #ifdef COMPAT_43
1176 /*
1177  * Create a file.
1178  */
1179 #ifndef _SYS_SYSPROTO_H_
1180 struct ocreat_args {
1181 	char	*path;
1182 	int	mode;
1183 };
1184 #endif
1185 int
1186 ocreat(td, uap)
1187 	struct thread *td;
1188 	register struct ocreat_args /* {
1189 		char *path;
1190 		int mode;
1191 	} */ *uap;
1192 {
1193 
1194 	return (kern_open(td, uap->path, UIO_USERSPACE,
1195 	    O_WRONLY | O_CREAT | O_TRUNC, uap->mode));
1196 }
1197 #endif /* COMPAT_43 */
1198 
1199 /*
1200  * Create a special file.
1201  */
1202 #ifndef _SYS_SYSPROTO_H_
1203 struct mknod_args {
1204 	char	*path;
1205 	int	mode;
1206 	int	dev;
1207 };
1208 #endif
1209 int
1210 sys_mknod(td, uap)
1211 	struct thread *td;
1212 	register struct mknod_args /* {
1213 		char *path;
1214 		int mode;
1215 		int dev;
1216 	} */ *uap;
1217 {
1218 
1219 	return (kern_mknod(td, uap->path, UIO_USERSPACE, uap->mode, uap->dev));
1220 }
1221 
1222 #ifndef _SYS_SYSPROTO_H_
1223 struct mknodat_args {
1224 	int	fd;
1225 	char	*path;
1226 	mode_t	mode;
1227 	dev_t	dev;
1228 };
1229 #endif
1230 int
1231 sys_mknodat(struct thread *td, struct mknodat_args *uap)
1232 {
1233 
1234 	return (kern_mknodat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode,
1235 	    uap->dev));
1236 }
1237 
1238 int
1239 kern_mknod(struct thread *td, char *path, enum uio_seg pathseg, int mode,
1240     int dev)
1241 {
1242 
1243 	return (kern_mknodat(td, AT_FDCWD, path, pathseg, mode, dev));
1244 }
1245 
1246 int
1247 kern_mknodat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1248     int mode, int dev)
1249 {
1250 	struct vnode *vp;
1251 	struct mount *mp;
1252 	struct vattr vattr;
1253 	int error;
1254 	int whiteout = 0;
1255 	struct nameidata nd;
1256 
1257 	AUDIT_ARG_MODE(mode);
1258 	AUDIT_ARG_DEV(dev);
1259 	switch (mode & S_IFMT) {
1260 	case S_IFCHR:
1261 	case S_IFBLK:
1262 		error = priv_check(td, PRIV_VFS_MKNOD_DEV);
1263 		break;
1264 	case S_IFMT:
1265 		error = priv_check(td, PRIV_VFS_MKNOD_BAD);
1266 		break;
1267 	case S_IFWHT:
1268 		error = priv_check(td, PRIV_VFS_MKNOD_WHT);
1269 		break;
1270 	case S_IFIFO:
1271 		if (dev == 0)
1272 			return (kern_mkfifoat(td, fd, path, pathseg, mode));
1273 		/* FALLTHROUGH */
1274 	default:
1275 		error = EINVAL;
1276 		break;
1277 	}
1278 	if (error)
1279 		return (error);
1280 restart:
1281 	bwillwrite();
1282 	NDINIT_ATRIGHTS(&nd, CREATE,
1283 	    LOCKPARENT | SAVENAME | AUDITVNODE1, pathseg, path, fd,
1284 	    CAP_MKNOD, td);
1285 	if ((error = namei(&nd)) != 0)
1286 		return (error);
1287 	vp = nd.ni_vp;
1288 	if (vp != NULL) {
1289 		NDFREE(&nd, NDF_ONLY_PNBUF);
1290 		if (vp == nd.ni_dvp)
1291 			vrele(nd.ni_dvp);
1292 		else
1293 			vput(nd.ni_dvp);
1294 		vrele(vp);
1295 		return (EEXIST);
1296 	} else {
1297 		VATTR_NULL(&vattr);
1298 		vattr.va_mode = (mode & ALLPERMS) &
1299 		    ~td->td_proc->p_fd->fd_cmask;
1300 		vattr.va_rdev = dev;
1301 		whiteout = 0;
1302 
1303 		switch (mode & S_IFMT) {
1304 		case S_IFMT:	/* used by badsect to flag bad sectors */
1305 			vattr.va_type = VBAD;
1306 			break;
1307 		case S_IFCHR:
1308 			vattr.va_type = VCHR;
1309 			break;
1310 		case S_IFBLK:
1311 			vattr.va_type = VBLK;
1312 			break;
1313 		case S_IFWHT:
1314 			whiteout = 1;
1315 			break;
1316 		default:
1317 			panic("kern_mknod: invalid mode");
1318 		}
1319 	}
1320 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1321 		NDFREE(&nd, NDF_ONLY_PNBUF);
1322 		vput(nd.ni_dvp);
1323 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1324 			return (error);
1325 		goto restart;
1326 	}
1327 #ifdef MAC
1328 	if (error == 0 && !whiteout)
1329 		error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp,
1330 		    &nd.ni_cnd, &vattr);
1331 #endif
1332 	if (!error) {
1333 		if (whiteout)
1334 			error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE);
1335 		else {
1336 			error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp,
1337 						&nd.ni_cnd, &vattr);
1338 			if (error == 0)
1339 				vput(nd.ni_vp);
1340 		}
1341 	}
1342 	NDFREE(&nd, NDF_ONLY_PNBUF);
1343 	vput(nd.ni_dvp);
1344 	vn_finished_write(mp);
1345 	return (error);
1346 }
1347 
1348 /*
1349  * Create a named pipe.
1350  */
1351 #ifndef _SYS_SYSPROTO_H_
1352 struct mkfifo_args {
1353 	char	*path;
1354 	int	mode;
1355 };
1356 #endif
1357 int
1358 sys_mkfifo(td, uap)
1359 	struct thread *td;
1360 	register struct mkfifo_args /* {
1361 		char *path;
1362 		int mode;
1363 	} */ *uap;
1364 {
1365 
1366 	return (kern_mkfifo(td, uap->path, UIO_USERSPACE, uap->mode));
1367 }
1368 
1369 #ifndef _SYS_SYSPROTO_H_
1370 struct mkfifoat_args {
1371 	int	fd;
1372 	char	*path;
1373 	mode_t	mode;
1374 };
1375 #endif
1376 int
1377 sys_mkfifoat(struct thread *td, struct mkfifoat_args *uap)
1378 {
1379 
1380 	return (kern_mkfifoat(td, uap->fd, uap->path, UIO_USERSPACE,
1381 	    uap->mode));
1382 }
1383 
1384 int
1385 kern_mkfifo(struct thread *td, char *path, enum uio_seg pathseg, int mode)
1386 {
1387 
1388 	return (kern_mkfifoat(td, AT_FDCWD, path, pathseg, mode));
1389 }
1390 
1391 int
1392 kern_mkfifoat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1393     int mode)
1394 {
1395 	struct mount *mp;
1396 	struct vattr vattr;
1397 	int error;
1398 	struct nameidata nd;
1399 
1400 	AUDIT_ARG_MODE(mode);
1401 restart:
1402 	bwillwrite();
1403 	NDINIT_ATRIGHTS(&nd, CREATE,
1404 	    LOCKPARENT | SAVENAME | AUDITVNODE1, pathseg, path, fd,
1405 	    CAP_MKFIFO, td);
1406 	if ((error = namei(&nd)) != 0)
1407 		return (error);
1408 	if (nd.ni_vp != NULL) {
1409 		NDFREE(&nd, NDF_ONLY_PNBUF);
1410 		if (nd.ni_vp == nd.ni_dvp)
1411 			vrele(nd.ni_dvp);
1412 		else
1413 			vput(nd.ni_dvp);
1414 		vrele(nd.ni_vp);
1415 		return (EEXIST);
1416 	}
1417 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1418 		NDFREE(&nd, NDF_ONLY_PNBUF);
1419 		vput(nd.ni_dvp);
1420 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1421 			return (error);
1422 		goto restart;
1423 	}
1424 	VATTR_NULL(&vattr);
1425 	vattr.va_type = VFIFO;
1426 	vattr.va_mode = (mode & ALLPERMS) & ~td->td_proc->p_fd->fd_cmask;
1427 #ifdef MAC
1428 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1429 	    &vattr);
1430 	if (error)
1431 		goto out;
1432 #endif
1433 	error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
1434 	if (error == 0)
1435 		vput(nd.ni_vp);
1436 #ifdef MAC
1437 out:
1438 #endif
1439 	vput(nd.ni_dvp);
1440 	vn_finished_write(mp);
1441 	NDFREE(&nd, NDF_ONLY_PNBUF);
1442 	return (error);
1443 }
1444 
1445 /*
1446  * Make a hard file link.
1447  */
1448 #ifndef _SYS_SYSPROTO_H_
1449 struct link_args {
1450 	char	*path;
1451 	char	*link;
1452 };
1453 #endif
1454 int
1455 sys_link(td, uap)
1456 	struct thread *td;
1457 	register struct link_args /* {
1458 		char *path;
1459 		char *link;
1460 	} */ *uap;
1461 {
1462 
1463 	return (kern_link(td, uap->path, uap->link, UIO_USERSPACE));
1464 }
1465 
1466 #ifndef _SYS_SYSPROTO_H_
1467 struct linkat_args {
1468 	int	fd1;
1469 	char	*path1;
1470 	int	fd2;
1471 	char	*path2;
1472 	int	flag;
1473 };
1474 #endif
1475 int
1476 sys_linkat(struct thread *td, struct linkat_args *uap)
1477 {
1478 	int flag;
1479 
1480 	flag = uap->flag;
1481 	if (flag & ~AT_SYMLINK_FOLLOW)
1482 		return (EINVAL);
1483 
1484 	return (kern_linkat(td, uap->fd1, uap->fd2, uap->path1, uap->path2,
1485 	    UIO_USERSPACE, (flag & AT_SYMLINK_FOLLOW) ? FOLLOW : NOFOLLOW));
1486 }
1487 
1488 int hardlink_check_uid = 0;
1489 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_uid, CTLFLAG_RW,
1490     &hardlink_check_uid, 0,
1491     "Unprivileged processes cannot create hard links to files owned by other "
1492     "users");
1493 static int hardlink_check_gid = 0;
1494 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_gid, CTLFLAG_RW,
1495     &hardlink_check_gid, 0,
1496     "Unprivileged processes cannot create hard links to files owned by other "
1497     "groups");
1498 
1499 static int
1500 can_hardlink(struct vnode *vp, struct ucred *cred)
1501 {
1502 	struct vattr va;
1503 	int error;
1504 
1505 	if (!hardlink_check_uid && !hardlink_check_gid)
1506 		return (0);
1507 
1508 	error = VOP_GETATTR(vp, &va, cred);
1509 	if (error != 0)
1510 		return (error);
1511 
1512 	if (hardlink_check_uid && cred->cr_uid != va.va_uid) {
1513 		error = priv_check_cred(cred, PRIV_VFS_LINK, 0);
1514 		if (error)
1515 			return (error);
1516 	}
1517 
1518 	if (hardlink_check_gid && !groupmember(va.va_gid, cred)) {
1519 		error = priv_check_cred(cred, PRIV_VFS_LINK, 0);
1520 		if (error)
1521 			return (error);
1522 	}
1523 
1524 	return (0);
1525 }
1526 
1527 int
1528 kern_link(struct thread *td, char *path, char *link, enum uio_seg segflg)
1529 {
1530 
1531 	return (kern_linkat(td, AT_FDCWD, AT_FDCWD, path,link, segflg, FOLLOW));
1532 }
1533 
1534 int
1535 kern_linkat(struct thread *td, int fd1, int fd2, char *path1, char *path2,
1536     enum uio_seg segflg, int follow)
1537 {
1538 	struct vnode *vp;
1539 	struct mount *mp;
1540 	struct nameidata nd;
1541 	int error;
1542 
1543 	bwillwrite();
1544 	NDINIT_AT(&nd, LOOKUP, follow | AUDITVNODE1, segflg, path1,
1545 	    fd1, td);
1546 
1547 	if ((error = namei(&nd)) != 0)
1548 		return (error);
1549 	NDFREE(&nd, NDF_ONLY_PNBUF);
1550 	vp = nd.ni_vp;
1551 	if (vp->v_type == VDIR) {
1552 		vrele(vp);
1553 		return (EPERM);		/* POSIX */
1554 	}
1555 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) {
1556 		vrele(vp);
1557 		return (error);
1558 	}
1559 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME |
1560 	    AUDITVNODE2, segflg, path2, fd2, CAP_CREATE, td);
1561 	if ((error = namei(&nd)) == 0) {
1562 		if (nd.ni_vp != NULL) {
1563 			if (nd.ni_dvp == nd.ni_vp)
1564 				vrele(nd.ni_dvp);
1565 			else
1566 				vput(nd.ni_dvp);
1567 			vrele(nd.ni_vp);
1568 			error = EEXIST;
1569 		} else if ((error = vn_lock(vp, LK_EXCLUSIVE | LK_RETRY))
1570 		    == 0) {
1571 			error = can_hardlink(vp, td->td_ucred);
1572 			if (error == 0)
1573 #ifdef MAC
1574 				error = mac_vnode_check_link(td->td_ucred,
1575 				    nd.ni_dvp, vp, &nd.ni_cnd);
1576 			if (error == 0)
1577 #endif
1578 				error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
1579 			VOP_UNLOCK(vp, 0);
1580 			vput(nd.ni_dvp);
1581 		}
1582 		NDFREE(&nd, NDF_ONLY_PNBUF);
1583 	}
1584 	vrele(vp);
1585 	vn_finished_write(mp);
1586 	return (error);
1587 }
1588 
1589 /*
1590  * Make a symbolic link.
1591  */
1592 #ifndef _SYS_SYSPROTO_H_
1593 struct symlink_args {
1594 	char	*path;
1595 	char	*link;
1596 };
1597 #endif
1598 int
1599 sys_symlink(td, uap)
1600 	struct thread *td;
1601 	register struct symlink_args /* {
1602 		char *path;
1603 		char *link;
1604 	} */ *uap;
1605 {
1606 
1607 	return (kern_symlink(td, uap->path, uap->link, UIO_USERSPACE));
1608 }
1609 
1610 #ifndef _SYS_SYSPROTO_H_
1611 struct symlinkat_args {
1612 	char	*path;
1613 	int	fd;
1614 	char	*path2;
1615 };
1616 #endif
1617 int
1618 sys_symlinkat(struct thread *td, struct symlinkat_args *uap)
1619 {
1620 
1621 	return (kern_symlinkat(td, uap->path1, uap->fd, uap->path2,
1622 	    UIO_USERSPACE));
1623 }
1624 
1625 int
1626 kern_symlink(struct thread *td, char *path, char *link, enum uio_seg segflg)
1627 {
1628 
1629 	return (kern_symlinkat(td, path, AT_FDCWD, link, segflg));
1630 }
1631 
1632 int
1633 kern_symlinkat(struct thread *td, char *path1, int fd, char *path2,
1634     enum uio_seg segflg)
1635 {
1636 	struct mount *mp;
1637 	struct vattr vattr;
1638 	char *syspath;
1639 	int error;
1640 	struct nameidata nd;
1641 
1642 	if (segflg == UIO_SYSSPACE) {
1643 		syspath = path1;
1644 	} else {
1645 		syspath = uma_zalloc(namei_zone, M_WAITOK);
1646 		if ((error = copyinstr(path1, syspath, MAXPATHLEN, NULL)) != 0)
1647 			goto out;
1648 	}
1649 	AUDIT_ARG_TEXT(syspath);
1650 restart:
1651 	bwillwrite();
1652 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME |
1653 	    AUDITVNODE1, segflg, path2, fd, CAP_CREATE, td);
1654 	if ((error = namei(&nd)) != 0)
1655 		goto out;
1656 	if (nd.ni_vp) {
1657 		NDFREE(&nd, NDF_ONLY_PNBUF);
1658 		if (nd.ni_vp == nd.ni_dvp)
1659 			vrele(nd.ni_dvp);
1660 		else
1661 			vput(nd.ni_dvp);
1662 		vrele(nd.ni_vp);
1663 		error = EEXIST;
1664 		goto out;
1665 	}
1666 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1667 		NDFREE(&nd, NDF_ONLY_PNBUF);
1668 		vput(nd.ni_dvp);
1669 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1670 			goto out;
1671 		goto restart;
1672 	}
1673 	VATTR_NULL(&vattr);
1674 	vattr.va_mode = ACCESSPERMS &~ td->td_proc->p_fd->fd_cmask;
1675 #ifdef MAC
1676 	vattr.va_type = VLNK;
1677 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1678 	    &vattr);
1679 	if (error)
1680 		goto out2;
1681 #endif
1682 	error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, syspath);
1683 	if (error == 0)
1684 		vput(nd.ni_vp);
1685 #ifdef MAC
1686 out2:
1687 #endif
1688 	NDFREE(&nd, NDF_ONLY_PNBUF);
1689 	vput(nd.ni_dvp);
1690 	vn_finished_write(mp);
1691 out:
1692 	if (segflg != UIO_SYSSPACE)
1693 		uma_zfree(namei_zone, syspath);
1694 	return (error);
1695 }
1696 
1697 /*
1698  * Delete a whiteout from the filesystem.
1699  */
1700 int
1701 sys_undelete(td, uap)
1702 	struct thread *td;
1703 	register struct undelete_args /* {
1704 		char *path;
1705 	} */ *uap;
1706 {
1707 	int error;
1708 	struct mount *mp;
1709 	struct nameidata nd;
1710 
1711 restart:
1712 	bwillwrite();
1713 	NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | AUDITVNODE1,
1714 	    UIO_USERSPACE, uap->path, td);
1715 	error = namei(&nd);
1716 	if (error)
1717 		return (error);
1718 
1719 	if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) {
1720 		NDFREE(&nd, NDF_ONLY_PNBUF);
1721 		if (nd.ni_vp == nd.ni_dvp)
1722 			vrele(nd.ni_dvp);
1723 		else
1724 			vput(nd.ni_dvp);
1725 		if (nd.ni_vp)
1726 			vrele(nd.ni_vp);
1727 		return (EEXIST);
1728 	}
1729 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1730 		NDFREE(&nd, NDF_ONLY_PNBUF);
1731 		vput(nd.ni_dvp);
1732 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1733 			return (error);
1734 		goto restart;
1735 	}
1736 	error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE);
1737 	NDFREE(&nd, NDF_ONLY_PNBUF);
1738 	vput(nd.ni_dvp);
1739 	vn_finished_write(mp);
1740 	return (error);
1741 }
1742 
1743 /*
1744  * Delete a name from the filesystem.
1745  */
1746 #ifndef _SYS_SYSPROTO_H_
1747 struct unlink_args {
1748 	char	*path;
1749 };
1750 #endif
1751 int
1752 sys_unlink(td, uap)
1753 	struct thread *td;
1754 	struct unlink_args /* {
1755 		char *path;
1756 	} */ *uap;
1757 {
1758 
1759 	return (kern_unlink(td, uap->path, UIO_USERSPACE));
1760 }
1761 
1762 #ifndef _SYS_SYSPROTO_H_
1763 struct unlinkat_args {
1764 	int	fd;
1765 	char	*path;
1766 	int	flag;
1767 };
1768 #endif
1769 int
1770 sys_unlinkat(struct thread *td, struct unlinkat_args *uap)
1771 {
1772 	int flag = uap->flag;
1773 	int fd = uap->fd;
1774 	char *path = uap->path;
1775 
1776 	if (flag & ~AT_REMOVEDIR)
1777 		return (EINVAL);
1778 
1779 	if (flag & AT_REMOVEDIR)
1780 		return (kern_rmdirat(td, fd, path, UIO_USERSPACE));
1781 	else
1782 		return (kern_unlinkat(td, fd, path, UIO_USERSPACE, 0));
1783 }
1784 
1785 int
1786 kern_unlink(struct thread *td, char *path, enum uio_seg pathseg)
1787 {
1788 
1789 	return (kern_unlinkat(td, AT_FDCWD, path, pathseg, 0));
1790 }
1791 
1792 int
1793 kern_unlinkat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1794     ino_t oldinum)
1795 {
1796 	struct mount *mp;
1797 	struct vnode *vp;
1798 	int error;
1799 	struct nameidata nd;
1800 	struct stat sb;
1801 
1802 restart:
1803 	bwillwrite();
1804 	NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF |
1805 	    AUDITVNODE1, pathseg, path, fd, CAP_DELETE, td);
1806 	if ((error = namei(&nd)) != 0)
1807 		return (error == EINVAL ? EPERM : error);
1808 	vp = nd.ni_vp;
1809 	if (vp->v_type == VDIR && oldinum == 0) {
1810 		error = EPERM;		/* POSIX */
1811 	} else if (oldinum != 0 &&
1812 		  ((error = vn_stat(vp, &sb, td->td_ucred, NOCRED, td)) == 0) &&
1813 		  sb.st_ino != oldinum) {
1814 			error = EIDRM;	/* Identifier removed */
1815 	} else {
1816 		/*
1817 		 * The root of a mounted filesystem cannot be deleted.
1818 		 *
1819 		 * XXX: can this only be a VDIR case?
1820 		 */
1821 		if (vp->v_vflag & VV_ROOT)
1822 			error = EBUSY;
1823 	}
1824 	if (error == 0) {
1825 		if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1826 			NDFREE(&nd, NDF_ONLY_PNBUF);
1827 			vput(nd.ni_dvp);
1828 			if (vp == nd.ni_dvp)
1829 				vrele(vp);
1830 			else
1831 				vput(vp);
1832 			if ((error = vn_start_write(NULL, &mp,
1833 			    V_XSLEEP | PCATCH)) != 0)
1834 				return (error);
1835 			goto restart;
1836 		}
1837 #ifdef MAC
1838 		error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
1839 		    &nd.ni_cnd);
1840 		if (error)
1841 			goto out;
1842 #endif
1843 		error = VOP_REMOVE(nd.ni_dvp, vp, &nd.ni_cnd);
1844 #ifdef MAC
1845 out:
1846 #endif
1847 		vn_finished_write(mp);
1848 	}
1849 	NDFREE(&nd, NDF_ONLY_PNBUF);
1850 	vput(nd.ni_dvp);
1851 	if (vp == nd.ni_dvp)
1852 		vrele(vp);
1853 	else
1854 		vput(vp);
1855 	return (error);
1856 }
1857 
1858 /*
1859  * Reposition read/write file offset.
1860  */
1861 #ifndef _SYS_SYSPROTO_H_
1862 struct lseek_args {
1863 	int	fd;
1864 	int	pad;
1865 	off_t	offset;
1866 	int	whence;
1867 };
1868 #endif
1869 int
1870 sys_lseek(td, uap)
1871 	struct thread *td;
1872 	register struct lseek_args /* {
1873 		int fd;
1874 		int pad;
1875 		off_t offset;
1876 		int whence;
1877 	} */ *uap;
1878 {
1879 	struct ucred *cred = td->td_ucred;
1880 	struct file *fp;
1881 	struct vnode *vp;
1882 	struct vattr vattr;
1883 	off_t foffset, offset, size;
1884 	int error, noneg;
1885 
1886 	AUDIT_ARG_FD(uap->fd);
1887 	if ((error = fget(td, uap->fd, CAP_SEEK, &fp)) != 0)
1888 		return (error);
1889 	if (!(fp->f_ops->fo_flags & DFLAG_SEEKABLE)) {
1890 		fdrop(fp, td);
1891 		return (ESPIPE);
1892 	}
1893 	vp = fp->f_vnode;
1894 	foffset = foffset_lock(fp, 0);
1895 	noneg = (vp->v_type != VCHR);
1896 	offset = uap->offset;
1897 	switch (uap->whence) {
1898 	case L_INCR:
1899 		if (noneg &&
1900 		    (foffset < 0 ||
1901 		    (offset > 0 && foffset > OFF_MAX - offset))) {
1902 			error = EOVERFLOW;
1903 			break;
1904 		}
1905 		offset += foffset;
1906 		break;
1907 	case L_XTND:
1908 		vn_lock(vp, LK_SHARED | LK_RETRY);
1909 		error = VOP_GETATTR(vp, &vattr, cred);
1910 		VOP_UNLOCK(vp, 0);
1911 		if (error)
1912 			break;
1913 
1914 		/*
1915 		 * If the file references a disk device, then fetch
1916 		 * the media size and use that to determine the ending
1917 		 * offset.
1918 		 */
1919 		if (vattr.va_size == 0 && vp->v_type == VCHR &&
1920 		    fo_ioctl(fp, DIOCGMEDIASIZE, &size, cred, td) == 0)
1921 			vattr.va_size = size;
1922 		if (noneg &&
1923 		    (vattr.va_size > OFF_MAX ||
1924 		    (offset > 0 && vattr.va_size > OFF_MAX - offset))) {
1925 			error = EOVERFLOW;
1926 			break;
1927 		}
1928 		offset += vattr.va_size;
1929 		break;
1930 	case L_SET:
1931 		break;
1932 	case SEEK_DATA:
1933 		error = fo_ioctl(fp, FIOSEEKDATA, &offset, cred, td);
1934 		break;
1935 	case SEEK_HOLE:
1936 		error = fo_ioctl(fp, FIOSEEKHOLE, &offset, cred, td);
1937 		break;
1938 	default:
1939 		error = EINVAL;
1940 	}
1941 	if (error == 0 && noneg && offset < 0)
1942 		error = EINVAL;
1943 	if (error != 0)
1944 		goto drop;
1945 	VFS_KNOTE_UNLOCKED(vp, 0);
1946 	*(off_t *)(td->td_retval) = offset;
1947 drop:
1948 	fdrop(fp, td);
1949 	foffset_unlock(fp, offset, error != 0 ? FOF_NOUPDATE : 0);
1950 	return (error);
1951 }
1952 
1953 #if defined(COMPAT_43)
1954 /*
1955  * Reposition read/write file offset.
1956  */
1957 #ifndef _SYS_SYSPROTO_H_
1958 struct olseek_args {
1959 	int	fd;
1960 	long	offset;
1961 	int	whence;
1962 };
1963 #endif
1964 int
1965 olseek(td, uap)
1966 	struct thread *td;
1967 	register struct olseek_args /* {
1968 		int fd;
1969 		long offset;
1970 		int whence;
1971 	} */ *uap;
1972 {
1973 	struct lseek_args /* {
1974 		int fd;
1975 		int pad;
1976 		off_t offset;
1977 		int whence;
1978 	} */ nuap;
1979 
1980 	nuap.fd = uap->fd;
1981 	nuap.offset = uap->offset;
1982 	nuap.whence = uap->whence;
1983 	return (sys_lseek(td, &nuap));
1984 }
1985 #endif /* COMPAT_43 */
1986 
1987 /* Version with the 'pad' argument */
1988 int
1989 freebsd6_lseek(td, uap)
1990 	struct thread *td;
1991 	register struct freebsd6_lseek_args *uap;
1992 {
1993 	struct lseek_args ouap;
1994 
1995 	ouap.fd = uap->fd;
1996 	ouap.offset = uap->offset;
1997 	ouap.whence = uap->whence;
1998 	return (sys_lseek(td, &ouap));
1999 }
2000 
2001 /*
2002  * Check access permissions using passed credentials.
2003  */
2004 static int
2005 vn_access(vp, user_flags, cred, td)
2006 	struct vnode	*vp;
2007 	int		user_flags;
2008 	struct ucred	*cred;
2009 	struct thread	*td;
2010 {
2011 	int error;
2012 	accmode_t accmode;
2013 
2014 	/* Flags == 0 means only check for existence. */
2015 	error = 0;
2016 	if (user_flags) {
2017 		accmode = 0;
2018 		if (user_flags & R_OK)
2019 			accmode |= VREAD;
2020 		if (user_flags & W_OK)
2021 			accmode |= VWRITE;
2022 		if (user_flags & X_OK)
2023 			accmode |= VEXEC;
2024 #ifdef MAC
2025 		error = mac_vnode_check_access(cred, vp, accmode);
2026 		if (error)
2027 			return (error);
2028 #endif
2029 		if ((accmode & VWRITE) == 0 || (error = vn_writechk(vp)) == 0)
2030 			error = VOP_ACCESS(vp, accmode, cred, td);
2031 	}
2032 	return (error);
2033 }
2034 
2035 /*
2036  * Check access permissions using "real" credentials.
2037  */
2038 #ifndef _SYS_SYSPROTO_H_
2039 struct access_args {
2040 	char	*path;
2041 	int	amode;
2042 };
2043 #endif
2044 int
2045 sys_access(td, uap)
2046 	struct thread *td;
2047 	register struct access_args /* {
2048 		char *path;
2049 		int amode;
2050 	} */ *uap;
2051 {
2052 
2053 	return (kern_access(td, uap->path, UIO_USERSPACE, uap->amode));
2054 }
2055 
2056 #ifndef _SYS_SYSPROTO_H_
2057 struct faccessat_args {
2058 	int	dirfd;
2059 	char	*path;
2060 	int	amode;
2061 	int	flag;
2062 }
2063 #endif
2064 int
2065 sys_faccessat(struct thread *td, struct faccessat_args *uap)
2066 {
2067 
2068 	if (uap->flag & ~AT_EACCESS)
2069 		return (EINVAL);
2070 	return (kern_accessat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
2071 	    uap->amode));
2072 }
2073 
2074 int
2075 kern_access(struct thread *td, char *path, enum uio_seg pathseg, int amode)
2076 {
2077 
2078 	return (kern_accessat(td, AT_FDCWD, path, pathseg, 0, amode));
2079 }
2080 
2081 int
2082 kern_accessat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2083     int flag, int amode)
2084 {
2085 	struct ucred *cred, *tmpcred;
2086 	struct vnode *vp;
2087 	struct nameidata nd;
2088 	int error;
2089 
2090 	/*
2091 	 * Create and modify a temporary credential instead of one that
2092 	 * is potentially shared.
2093 	 */
2094 	if (!(flag & AT_EACCESS)) {
2095 		cred = td->td_ucred;
2096 		tmpcred = crdup(cred);
2097 		tmpcred->cr_uid = cred->cr_ruid;
2098 		tmpcred->cr_groups[0] = cred->cr_rgid;
2099 		td->td_ucred = tmpcred;
2100 	} else
2101 		cred = tmpcred = td->td_ucred;
2102 	AUDIT_ARG_VALUE(amode);
2103 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF |
2104 	    AUDITVNODE1, pathseg, path, fd, CAP_FSTAT, td);
2105 	if ((error = namei(&nd)) != 0)
2106 		goto out1;
2107 	vp = nd.ni_vp;
2108 
2109 	error = vn_access(vp, amode, tmpcred, td);
2110 	NDFREE(&nd, NDF_ONLY_PNBUF);
2111 	vput(vp);
2112 out1:
2113 	if (!(flag & AT_EACCESS)) {
2114 		td->td_ucred = cred;
2115 		crfree(tmpcred);
2116 	}
2117 	return (error);
2118 }
2119 
2120 /*
2121  * Check access permissions using "effective" credentials.
2122  */
2123 #ifndef _SYS_SYSPROTO_H_
2124 struct eaccess_args {
2125 	char	*path;
2126 	int	amode;
2127 };
2128 #endif
2129 int
2130 sys_eaccess(td, uap)
2131 	struct thread *td;
2132 	register struct eaccess_args /* {
2133 		char *path;
2134 		int amode;
2135 	} */ *uap;
2136 {
2137 
2138 	return (kern_eaccess(td, uap->path, UIO_USERSPACE, uap->amode));
2139 }
2140 
2141 int
2142 kern_eaccess(struct thread *td, char *path, enum uio_seg pathseg, int amode)
2143 {
2144 
2145 	return (kern_accessat(td, AT_FDCWD, path, pathseg, AT_EACCESS, amode));
2146 }
2147 
2148 #if defined(COMPAT_43)
2149 /*
2150  * Get file status; this version follows links.
2151  */
2152 #ifndef _SYS_SYSPROTO_H_
2153 struct ostat_args {
2154 	char	*path;
2155 	struct ostat *ub;
2156 };
2157 #endif
2158 int
2159 ostat(td, uap)
2160 	struct thread *td;
2161 	register struct ostat_args /* {
2162 		char *path;
2163 		struct ostat *ub;
2164 	} */ *uap;
2165 {
2166 	struct stat sb;
2167 	struct ostat osb;
2168 	int error;
2169 
2170 	error = kern_stat(td, uap->path, UIO_USERSPACE, &sb);
2171 	if (error)
2172 		return (error);
2173 	cvtstat(&sb, &osb);
2174 	error = copyout(&osb, uap->ub, sizeof (osb));
2175 	return (error);
2176 }
2177 
2178 /*
2179  * Get file status; this version does not follow links.
2180  */
2181 #ifndef _SYS_SYSPROTO_H_
2182 struct olstat_args {
2183 	char	*path;
2184 	struct ostat *ub;
2185 };
2186 #endif
2187 int
2188 olstat(td, uap)
2189 	struct thread *td;
2190 	register struct olstat_args /* {
2191 		char *path;
2192 		struct ostat *ub;
2193 	} */ *uap;
2194 {
2195 	struct stat sb;
2196 	struct ostat osb;
2197 	int error;
2198 
2199 	error = kern_lstat(td, uap->path, UIO_USERSPACE, &sb);
2200 	if (error)
2201 		return (error);
2202 	cvtstat(&sb, &osb);
2203 	error = copyout(&osb, uap->ub, sizeof (osb));
2204 	return (error);
2205 }
2206 
2207 /*
2208  * Convert from an old to a new stat structure.
2209  */
2210 void
2211 cvtstat(st, ost)
2212 	struct stat *st;
2213 	struct ostat *ost;
2214 {
2215 
2216 	ost->st_dev = st->st_dev;
2217 	ost->st_ino = st->st_ino;
2218 	ost->st_mode = st->st_mode;
2219 	ost->st_nlink = st->st_nlink;
2220 	ost->st_uid = st->st_uid;
2221 	ost->st_gid = st->st_gid;
2222 	ost->st_rdev = st->st_rdev;
2223 	if (st->st_size < (quad_t)1 << 32)
2224 		ost->st_size = st->st_size;
2225 	else
2226 		ost->st_size = -2;
2227 	ost->st_atim = st->st_atim;
2228 	ost->st_mtim = st->st_mtim;
2229 	ost->st_ctim = st->st_ctim;
2230 	ost->st_blksize = st->st_blksize;
2231 	ost->st_blocks = st->st_blocks;
2232 	ost->st_flags = st->st_flags;
2233 	ost->st_gen = st->st_gen;
2234 }
2235 #endif /* COMPAT_43 */
2236 
2237 /*
2238  * Get file status; this version follows links.
2239  */
2240 #ifndef _SYS_SYSPROTO_H_
2241 struct stat_args {
2242 	char	*path;
2243 	struct stat *ub;
2244 };
2245 #endif
2246 int
2247 sys_stat(td, uap)
2248 	struct thread *td;
2249 	register struct stat_args /* {
2250 		char *path;
2251 		struct stat *ub;
2252 	} */ *uap;
2253 {
2254 	struct stat sb;
2255 	int error;
2256 
2257 	error = kern_stat(td, uap->path, UIO_USERSPACE, &sb);
2258 	if (error == 0)
2259 		error = copyout(&sb, uap->ub, sizeof (sb));
2260 	return (error);
2261 }
2262 
2263 #ifndef _SYS_SYSPROTO_H_
2264 struct fstatat_args {
2265 	int	fd;
2266 	char	*path;
2267 	struct stat	*buf;
2268 	int	flag;
2269 }
2270 #endif
2271 int
2272 sys_fstatat(struct thread *td, struct fstatat_args *uap)
2273 {
2274 	struct stat sb;
2275 	int error;
2276 
2277 	error = kern_statat(td, uap->flag, uap->fd, uap->path,
2278 	    UIO_USERSPACE, &sb);
2279 	if (error == 0)
2280 		error = copyout(&sb, uap->buf, sizeof (sb));
2281 	return (error);
2282 }
2283 
2284 int
2285 kern_stat(struct thread *td, char *path, enum uio_seg pathseg, struct stat *sbp)
2286 {
2287 
2288 	return (kern_statat(td, 0, AT_FDCWD, path, pathseg, sbp));
2289 }
2290 
2291 int
2292 kern_statat(struct thread *td, int flag, int fd, char *path,
2293     enum uio_seg pathseg, struct stat *sbp)
2294 {
2295 
2296 	return (kern_statat_vnhook(td, flag, fd, path, pathseg, sbp, NULL));
2297 }
2298 
2299 int
2300 kern_statat_vnhook(struct thread *td, int flag, int fd, char *path,
2301     enum uio_seg pathseg, struct stat *sbp,
2302     void (*hook)(struct vnode *vp, struct stat *sbp))
2303 {
2304 	struct nameidata nd;
2305 	struct stat sb;
2306 	int error;
2307 
2308 	if (flag & ~AT_SYMLINK_NOFOLLOW)
2309 		return (EINVAL);
2310 
2311 	NDINIT_ATRIGHTS(&nd, LOOKUP, ((flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW :
2312 	    FOLLOW) | LOCKSHARED | LOCKLEAF | AUDITVNODE1, pathseg,
2313 	    path, fd, CAP_FSTAT, td);
2314 
2315 	if ((error = namei(&nd)) != 0)
2316 		return (error);
2317 	error = vn_stat(nd.ni_vp, &sb, td->td_ucred, NOCRED, td);
2318 	if (!error) {
2319 		SDT_PROBE(vfs, , stat, mode, path, sb.st_mode, 0, 0, 0);
2320 		if (S_ISREG(sb.st_mode))
2321 			SDT_PROBE(vfs, , stat, reg, path, pathseg, 0, 0, 0);
2322 		if (__predict_false(hook != NULL))
2323 			hook(nd.ni_vp, &sb);
2324 	}
2325 	NDFREE(&nd, NDF_ONLY_PNBUF);
2326 	vput(nd.ni_vp);
2327 	if (error)
2328 		return (error);
2329 	*sbp = sb;
2330 #ifdef KTRACE
2331 	if (KTRPOINT(td, KTR_STRUCT))
2332 		ktrstat(&sb);
2333 #endif
2334 	return (0);
2335 }
2336 
2337 /*
2338  * Get file status; this version does not follow links.
2339  */
2340 #ifndef _SYS_SYSPROTO_H_
2341 struct lstat_args {
2342 	char	*path;
2343 	struct stat *ub;
2344 };
2345 #endif
2346 int
2347 sys_lstat(td, uap)
2348 	struct thread *td;
2349 	register struct lstat_args /* {
2350 		char *path;
2351 		struct stat *ub;
2352 	} */ *uap;
2353 {
2354 	struct stat sb;
2355 	int error;
2356 
2357 	error = kern_lstat(td, uap->path, UIO_USERSPACE, &sb);
2358 	if (error == 0)
2359 		error = copyout(&sb, uap->ub, sizeof (sb));
2360 	return (error);
2361 }
2362 
2363 int
2364 kern_lstat(struct thread *td, char *path, enum uio_seg pathseg, struct stat *sbp)
2365 {
2366 
2367 	return (kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, path, pathseg,
2368 	    sbp));
2369 }
2370 
2371 /*
2372  * Implementation of the NetBSD [l]stat() functions.
2373  */
2374 void
2375 cvtnstat(sb, nsb)
2376 	struct stat *sb;
2377 	struct nstat *nsb;
2378 {
2379 	bzero(nsb, sizeof *nsb);
2380 	nsb->st_dev = sb->st_dev;
2381 	nsb->st_ino = sb->st_ino;
2382 	nsb->st_mode = sb->st_mode;
2383 	nsb->st_nlink = sb->st_nlink;
2384 	nsb->st_uid = sb->st_uid;
2385 	nsb->st_gid = sb->st_gid;
2386 	nsb->st_rdev = sb->st_rdev;
2387 	nsb->st_atim = sb->st_atim;
2388 	nsb->st_mtim = sb->st_mtim;
2389 	nsb->st_ctim = sb->st_ctim;
2390 	nsb->st_size = sb->st_size;
2391 	nsb->st_blocks = sb->st_blocks;
2392 	nsb->st_blksize = sb->st_blksize;
2393 	nsb->st_flags = sb->st_flags;
2394 	nsb->st_gen = sb->st_gen;
2395 	nsb->st_birthtim = sb->st_birthtim;
2396 }
2397 
2398 #ifndef _SYS_SYSPROTO_H_
2399 struct nstat_args {
2400 	char	*path;
2401 	struct nstat *ub;
2402 };
2403 #endif
2404 int
2405 sys_nstat(td, uap)
2406 	struct thread *td;
2407 	register struct nstat_args /* {
2408 		char *path;
2409 		struct nstat *ub;
2410 	} */ *uap;
2411 {
2412 	struct stat sb;
2413 	struct nstat nsb;
2414 	int error;
2415 
2416 	error = kern_stat(td, uap->path, UIO_USERSPACE, &sb);
2417 	if (error)
2418 		return (error);
2419 	cvtnstat(&sb, &nsb);
2420 	error = copyout(&nsb, uap->ub, sizeof (nsb));
2421 	return (error);
2422 }
2423 
2424 /*
2425  * NetBSD lstat.  Get file status; this version does not follow links.
2426  */
2427 #ifndef _SYS_SYSPROTO_H_
2428 struct lstat_args {
2429 	char	*path;
2430 	struct stat *ub;
2431 };
2432 #endif
2433 int
2434 sys_nlstat(td, uap)
2435 	struct thread *td;
2436 	register struct nlstat_args /* {
2437 		char *path;
2438 		struct nstat *ub;
2439 	} */ *uap;
2440 {
2441 	struct stat sb;
2442 	struct nstat nsb;
2443 	int error;
2444 
2445 	error = kern_lstat(td, uap->path, UIO_USERSPACE, &sb);
2446 	if (error)
2447 		return (error);
2448 	cvtnstat(&sb, &nsb);
2449 	error = copyout(&nsb, uap->ub, sizeof (nsb));
2450 	return (error);
2451 }
2452 
2453 /*
2454  * Get configurable pathname variables.
2455  */
2456 #ifndef _SYS_SYSPROTO_H_
2457 struct pathconf_args {
2458 	char	*path;
2459 	int	name;
2460 };
2461 #endif
2462 int
2463 sys_pathconf(td, uap)
2464 	struct thread *td;
2465 	register struct pathconf_args /* {
2466 		char *path;
2467 		int name;
2468 	} */ *uap;
2469 {
2470 
2471 	return (kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name, FOLLOW));
2472 }
2473 
2474 #ifndef _SYS_SYSPROTO_H_
2475 struct lpathconf_args {
2476 	char	*path;
2477 	int	name;
2478 };
2479 #endif
2480 int
2481 sys_lpathconf(td, uap)
2482 	struct thread *td;
2483 	register struct lpathconf_args /* {
2484 		char *path;
2485 		int name;
2486 	} */ *uap;
2487 {
2488 
2489 	return (kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name, NOFOLLOW));
2490 }
2491 
2492 int
2493 kern_pathconf(struct thread *td, char *path, enum uio_seg pathseg, int name,
2494     u_long flags)
2495 {
2496 	struct nameidata nd;
2497 	int error;
2498 
2499 	NDINIT(&nd, LOOKUP, LOCKSHARED | LOCKLEAF | AUDITVNODE1 | flags,
2500 	    pathseg, path, td);
2501 	if ((error = namei(&nd)) != 0)
2502 		return (error);
2503 	NDFREE(&nd, NDF_ONLY_PNBUF);
2504 
2505 	/* If asynchronous I/O is available, it works for all files. */
2506 	if (name == _PC_ASYNC_IO)
2507 		td->td_retval[0] = async_io_version;
2508 	else
2509 		error = VOP_PATHCONF(nd.ni_vp, name, td->td_retval);
2510 	vput(nd.ni_vp);
2511 	return (error);
2512 }
2513 
2514 /*
2515  * Return target name of a symbolic link.
2516  */
2517 #ifndef _SYS_SYSPROTO_H_
2518 struct readlink_args {
2519 	char	*path;
2520 	char	*buf;
2521 	size_t	count;
2522 };
2523 #endif
2524 int
2525 sys_readlink(td, uap)
2526 	struct thread *td;
2527 	register struct readlink_args /* {
2528 		char *path;
2529 		char *buf;
2530 		size_t count;
2531 	} */ *uap;
2532 {
2533 
2534 	return (kern_readlink(td, uap->path, UIO_USERSPACE, uap->buf,
2535 	    UIO_USERSPACE, uap->count));
2536 }
2537 #ifndef _SYS_SYSPROTO_H_
2538 struct readlinkat_args {
2539 	int	fd;
2540 	char	*path;
2541 	char	*buf;
2542 	size_t	bufsize;
2543 };
2544 #endif
2545 int
2546 sys_readlinkat(struct thread *td, struct readlinkat_args *uap)
2547 {
2548 
2549 	return (kern_readlinkat(td, uap->fd, uap->path, UIO_USERSPACE,
2550 	    uap->buf, UIO_USERSPACE, uap->bufsize));
2551 }
2552 
2553 int
2554 kern_readlink(struct thread *td, char *path, enum uio_seg pathseg, char *buf,
2555     enum uio_seg bufseg, size_t count)
2556 {
2557 
2558 	return (kern_readlinkat(td, AT_FDCWD, path, pathseg, buf, bufseg,
2559 	    count));
2560 }
2561 
2562 int
2563 kern_readlinkat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2564     char *buf, enum uio_seg bufseg, size_t count)
2565 {
2566 	struct vnode *vp;
2567 	struct iovec aiov;
2568 	struct uio auio;
2569 	int error;
2570 	struct nameidata nd;
2571 
2572 	if (count > IOSIZE_MAX)
2573 		return (EINVAL);
2574 
2575 	NDINIT_AT(&nd, LOOKUP, NOFOLLOW | LOCKSHARED | LOCKLEAF |
2576 	    AUDITVNODE1, pathseg, path, fd, td);
2577 
2578 	if ((error = namei(&nd)) != 0)
2579 		return (error);
2580 	NDFREE(&nd, NDF_ONLY_PNBUF);
2581 	vp = nd.ni_vp;
2582 #ifdef MAC
2583 	error = mac_vnode_check_readlink(td->td_ucred, vp);
2584 	if (error) {
2585 		vput(vp);
2586 		return (error);
2587 	}
2588 #endif
2589 	if (vp->v_type != VLNK)
2590 		error = EINVAL;
2591 	else {
2592 		aiov.iov_base = buf;
2593 		aiov.iov_len = count;
2594 		auio.uio_iov = &aiov;
2595 		auio.uio_iovcnt = 1;
2596 		auio.uio_offset = 0;
2597 		auio.uio_rw = UIO_READ;
2598 		auio.uio_segflg = bufseg;
2599 		auio.uio_td = td;
2600 		auio.uio_resid = count;
2601 		error = VOP_READLINK(vp, &auio, td->td_ucred);
2602 	}
2603 	vput(vp);
2604 	td->td_retval[0] = count - auio.uio_resid;
2605 	return (error);
2606 }
2607 
2608 /*
2609  * Common implementation code for chflags() and fchflags().
2610  */
2611 static int
2612 setfflags(td, vp, flags)
2613 	struct thread *td;
2614 	struct vnode *vp;
2615 	int flags;
2616 {
2617 	int error;
2618 	struct mount *mp;
2619 	struct vattr vattr;
2620 
2621 	/* We can't support the value matching VNOVAL. */
2622 	if (flags == VNOVAL)
2623 		return (EOPNOTSUPP);
2624 
2625 	/*
2626 	 * Prevent non-root users from setting flags on devices.  When
2627 	 * a device is reused, users can retain ownership of the device
2628 	 * if they are allowed to set flags and programs assume that
2629 	 * chown can't fail when done as root.
2630 	 */
2631 	if (vp->v_type == VCHR || vp->v_type == VBLK) {
2632 		error = priv_check(td, PRIV_VFS_CHFLAGS_DEV);
2633 		if (error)
2634 			return (error);
2635 	}
2636 
2637 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2638 		return (error);
2639 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2640 	VATTR_NULL(&vattr);
2641 	vattr.va_flags = flags;
2642 #ifdef MAC
2643 	error = mac_vnode_check_setflags(td->td_ucred, vp, vattr.va_flags);
2644 	if (error == 0)
2645 #endif
2646 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
2647 	VOP_UNLOCK(vp, 0);
2648 	vn_finished_write(mp);
2649 	return (error);
2650 }
2651 
2652 /*
2653  * Change flags of a file given a path name.
2654  */
2655 #ifndef _SYS_SYSPROTO_H_
2656 struct chflags_args {
2657 	char	*path;
2658 	int	flags;
2659 };
2660 #endif
2661 int
2662 sys_chflags(td, uap)
2663 	struct thread *td;
2664 	register struct chflags_args /* {
2665 		char *path;
2666 		int flags;
2667 	} */ *uap;
2668 {
2669 	int error;
2670 	struct nameidata nd;
2671 
2672 	AUDIT_ARG_FFLAGS(uap->flags);
2673 	NDINIT(&nd, LOOKUP, FOLLOW | AUDITVNODE1, UIO_USERSPACE,
2674 	    uap->path, td);
2675 	if ((error = namei(&nd)) != 0)
2676 		return (error);
2677 	NDFREE(&nd, NDF_ONLY_PNBUF);
2678 	error = setfflags(td, nd.ni_vp, uap->flags);
2679 	vrele(nd.ni_vp);
2680 	return (error);
2681 }
2682 
2683 /*
2684  * Same as chflags() but doesn't follow symlinks.
2685  */
2686 int
2687 sys_lchflags(td, uap)
2688 	struct thread *td;
2689 	register struct lchflags_args /* {
2690 		char *path;
2691 		int flags;
2692 	} */ *uap;
2693 {
2694 	int error;
2695 	struct nameidata nd;
2696 
2697 	AUDIT_ARG_FFLAGS(uap->flags);
2698 	NDINIT(&nd, LOOKUP, NOFOLLOW | AUDITVNODE1, UIO_USERSPACE,
2699 	    uap->path, td);
2700 	if ((error = namei(&nd)) != 0)
2701 		return (error);
2702 	NDFREE(&nd, NDF_ONLY_PNBUF);
2703 	error = setfflags(td, nd.ni_vp, uap->flags);
2704 	vrele(nd.ni_vp);
2705 	return (error);
2706 }
2707 
2708 /*
2709  * Change flags of a file given a file descriptor.
2710  */
2711 #ifndef _SYS_SYSPROTO_H_
2712 struct fchflags_args {
2713 	int	fd;
2714 	int	flags;
2715 };
2716 #endif
2717 int
2718 sys_fchflags(td, uap)
2719 	struct thread *td;
2720 	register struct fchflags_args /* {
2721 		int fd;
2722 		int flags;
2723 	} */ *uap;
2724 {
2725 	struct file *fp;
2726 	int error;
2727 
2728 	AUDIT_ARG_FD(uap->fd);
2729 	AUDIT_ARG_FFLAGS(uap->flags);
2730 	if ((error = getvnode(td->td_proc->p_fd, uap->fd, CAP_FCHFLAGS,
2731 	    &fp)) != 0)
2732 		return (error);
2733 #ifdef AUDIT
2734 	vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
2735 	AUDIT_ARG_VNODE1(fp->f_vnode);
2736 	VOP_UNLOCK(fp->f_vnode, 0);
2737 #endif
2738 	error = setfflags(td, fp->f_vnode, uap->flags);
2739 	fdrop(fp, td);
2740 	return (error);
2741 }
2742 
2743 /*
2744  * Common implementation code for chmod(), lchmod() and fchmod().
2745  */
2746 int
2747 setfmode(td, cred, vp, mode)
2748 	struct thread *td;
2749 	struct ucred *cred;
2750 	struct vnode *vp;
2751 	int mode;
2752 {
2753 	int error;
2754 	struct mount *mp;
2755 	struct vattr vattr;
2756 
2757 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2758 		return (error);
2759 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2760 	VATTR_NULL(&vattr);
2761 	vattr.va_mode = mode & ALLPERMS;
2762 #ifdef MAC
2763 	error = mac_vnode_check_setmode(cred, vp, vattr.va_mode);
2764 	if (error == 0)
2765 #endif
2766 		error = VOP_SETATTR(vp, &vattr, cred);
2767 	VOP_UNLOCK(vp, 0);
2768 	vn_finished_write(mp);
2769 	return (error);
2770 }
2771 
2772 /*
2773  * Change mode of a file given path name.
2774  */
2775 #ifndef _SYS_SYSPROTO_H_
2776 struct chmod_args {
2777 	char	*path;
2778 	int	mode;
2779 };
2780 #endif
2781 int
2782 sys_chmod(td, uap)
2783 	struct thread *td;
2784 	register struct chmod_args /* {
2785 		char *path;
2786 		int mode;
2787 	} */ *uap;
2788 {
2789 
2790 	return (kern_chmod(td, uap->path, UIO_USERSPACE, uap->mode));
2791 }
2792 
2793 #ifndef _SYS_SYSPROTO_H_
2794 struct fchmodat_args {
2795 	int	dirfd;
2796 	char	*path;
2797 	mode_t	mode;
2798 	int	flag;
2799 }
2800 #endif
2801 int
2802 sys_fchmodat(struct thread *td, struct fchmodat_args *uap)
2803 {
2804 	int flag = uap->flag;
2805 	int fd = uap->fd;
2806 	char *path = uap->path;
2807 	mode_t mode = uap->mode;
2808 
2809 	if (flag & ~AT_SYMLINK_NOFOLLOW)
2810 		return (EINVAL);
2811 
2812 	return (kern_fchmodat(td, fd, path, UIO_USERSPACE, mode, flag));
2813 }
2814 
2815 int
2816 kern_chmod(struct thread *td, char *path, enum uio_seg pathseg, int mode)
2817 {
2818 
2819 	return (kern_fchmodat(td, AT_FDCWD, path, pathseg, mode, 0));
2820 }
2821 
2822 /*
2823  * Change mode of a file given path name (don't follow links.)
2824  */
2825 #ifndef _SYS_SYSPROTO_H_
2826 struct lchmod_args {
2827 	char	*path;
2828 	int	mode;
2829 };
2830 #endif
2831 int
2832 sys_lchmod(td, uap)
2833 	struct thread *td;
2834 	register struct lchmod_args /* {
2835 		char *path;
2836 		int mode;
2837 	} */ *uap;
2838 {
2839 
2840 	return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2841 	    uap->mode, AT_SYMLINK_NOFOLLOW));
2842 }
2843 
2844 
2845 int
2846 kern_fchmodat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2847     mode_t mode, int flag)
2848 {
2849 	int error;
2850 	struct nameidata nd;
2851 	int follow;
2852 
2853 	AUDIT_ARG_MODE(mode);
2854 	follow = (flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
2855 	NDINIT_ATRIGHTS(&nd, LOOKUP,  follow | AUDITVNODE1, pathseg,
2856 	    path, fd, CAP_FCHMOD, td);
2857 	if ((error = namei(&nd)) != 0)
2858 		return (error);
2859 	NDFREE(&nd, NDF_ONLY_PNBUF);
2860 	error = setfmode(td, td->td_ucred, nd.ni_vp, mode);
2861 	vrele(nd.ni_vp);
2862 	return (error);
2863 }
2864 
2865 /*
2866  * Change mode of a file given a file descriptor.
2867  */
2868 #ifndef _SYS_SYSPROTO_H_
2869 struct fchmod_args {
2870 	int	fd;
2871 	int	mode;
2872 };
2873 #endif
2874 int
2875 sys_fchmod(struct thread *td, struct fchmod_args *uap)
2876 {
2877 	struct file *fp;
2878 	int error;
2879 
2880 	AUDIT_ARG_FD(uap->fd);
2881 	AUDIT_ARG_MODE(uap->mode);
2882 
2883 	error = fget(td, uap->fd, CAP_FCHMOD, &fp);
2884 	if (error != 0)
2885 		return (error);
2886 	error = fo_chmod(fp, uap->mode, td->td_ucred, td);
2887 	fdrop(fp, td);
2888 	return (error);
2889 }
2890 
2891 /*
2892  * Common implementation for chown(), lchown(), and fchown()
2893  */
2894 int
2895 setfown(td, cred, vp, uid, gid)
2896 	struct thread *td;
2897 	struct ucred *cred;
2898 	struct vnode *vp;
2899 	uid_t uid;
2900 	gid_t gid;
2901 {
2902 	int error;
2903 	struct mount *mp;
2904 	struct vattr vattr;
2905 
2906 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2907 		return (error);
2908 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2909 	VATTR_NULL(&vattr);
2910 	vattr.va_uid = uid;
2911 	vattr.va_gid = gid;
2912 #ifdef MAC
2913 	error = mac_vnode_check_setowner(cred, vp, vattr.va_uid,
2914 	    vattr.va_gid);
2915 	if (error == 0)
2916 #endif
2917 		error = VOP_SETATTR(vp, &vattr, cred);
2918 	VOP_UNLOCK(vp, 0);
2919 	vn_finished_write(mp);
2920 	return (error);
2921 }
2922 
2923 /*
2924  * Set ownership given a path name.
2925  */
2926 #ifndef _SYS_SYSPROTO_H_
2927 struct chown_args {
2928 	char	*path;
2929 	int	uid;
2930 	int	gid;
2931 };
2932 #endif
2933 int
2934 sys_chown(td, uap)
2935 	struct thread *td;
2936 	register struct chown_args /* {
2937 		char *path;
2938 		int uid;
2939 		int gid;
2940 	} */ *uap;
2941 {
2942 
2943 	return (kern_chown(td, uap->path, UIO_USERSPACE, uap->uid, uap->gid));
2944 }
2945 
2946 #ifndef _SYS_SYSPROTO_H_
2947 struct fchownat_args {
2948 	int fd;
2949 	const char * path;
2950 	uid_t uid;
2951 	gid_t gid;
2952 	int flag;
2953 };
2954 #endif
2955 int
2956 sys_fchownat(struct thread *td, struct fchownat_args *uap)
2957 {
2958 	int flag;
2959 
2960 	flag = uap->flag;
2961 	if (flag & ~AT_SYMLINK_NOFOLLOW)
2962 		return (EINVAL);
2963 
2964 	return (kern_fchownat(td, uap->fd, uap->path, UIO_USERSPACE, uap->uid,
2965 	    uap->gid, uap->flag));
2966 }
2967 
2968 int
2969 kern_chown(struct thread *td, char *path, enum uio_seg pathseg, int uid,
2970     int gid)
2971 {
2972 
2973 	return (kern_fchownat(td, AT_FDCWD, path, pathseg, uid, gid, 0));
2974 }
2975 
2976 int
2977 kern_fchownat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2978     int uid, int gid, int flag)
2979 {
2980 	struct nameidata nd;
2981 	int error, follow;
2982 
2983 	AUDIT_ARG_OWNER(uid, gid);
2984 	follow = (flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
2985 	NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, pathseg,
2986 	    path, fd, CAP_FCHOWN, td);
2987 
2988 	if ((error = namei(&nd)) != 0)
2989 		return (error);
2990 	NDFREE(&nd, NDF_ONLY_PNBUF);
2991 	error = setfown(td, td->td_ucred, nd.ni_vp, uid, gid);
2992 	vrele(nd.ni_vp);
2993 	return (error);
2994 }
2995 
2996 /*
2997  * Set ownership given a path name, do not cross symlinks.
2998  */
2999 #ifndef _SYS_SYSPROTO_H_
3000 struct lchown_args {
3001 	char	*path;
3002 	int	uid;
3003 	int	gid;
3004 };
3005 #endif
3006 int
3007 sys_lchown(td, uap)
3008 	struct thread *td;
3009 	register struct lchown_args /* {
3010 		char *path;
3011 		int uid;
3012 		int gid;
3013 	} */ *uap;
3014 {
3015 
3016 	return (kern_lchown(td, uap->path, UIO_USERSPACE, uap->uid, uap->gid));
3017 }
3018 
3019 int
3020 kern_lchown(struct thread *td, char *path, enum uio_seg pathseg, int uid,
3021     int gid)
3022 {
3023 
3024 	return (kern_fchownat(td, AT_FDCWD, path, pathseg, uid, gid,
3025 	    AT_SYMLINK_NOFOLLOW));
3026 }
3027 
3028 /*
3029  * Set ownership given a file descriptor.
3030  */
3031 #ifndef _SYS_SYSPROTO_H_
3032 struct fchown_args {
3033 	int	fd;
3034 	int	uid;
3035 	int	gid;
3036 };
3037 #endif
3038 int
3039 sys_fchown(td, uap)
3040 	struct thread *td;
3041 	register struct fchown_args /* {
3042 		int fd;
3043 		int uid;
3044 		int gid;
3045 	} */ *uap;
3046 {
3047 	struct file *fp;
3048 	int error;
3049 
3050 	AUDIT_ARG_FD(uap->fd);
3051 	AUDIT_ARG_OWNER(uap->uid, uap->gid);
3052 	error = fget(td, uap->fd, CAP_FCHOWN, &fp);
3053 	if (error != 0)
3054 		return (error);
3055 	error = fo_chown(fp, uap->uid, uap->gid, td->td_ucred, td);
3056 	fdrop(fp, td);
3057 	return (error);
3058 }
3059 
3060 /*
3061  * Common implementation code for utimes(), lutimes(), and futimes().
3062  */
3063 static int
3064 getutimes(usrtvp, tvpseg, tsp)
3065 	const struct timeval *usrtvp;
3066 	enum uio_seg tvpseg;
3067 	struct timespec *tsp;
3068 {
3069 	struct timeval tv[2];
3070 	const struct timeval *tvp;
3071 	int error;
3072 
3073 	if (usrtvp == NULL) {
3074 		vfs_timestamp(&tsp[0]);
3075 		tsp[1] = tsp[0];
3076 	} else {
3077 		if (tvpseg == UIO_SYSSPACE) {
3078 			tvp = usrtvp;
3079 		} else {
3080 			if ((error = copyin(usrtvp, tv, sizeof(tv))) != 0)
3081 				return (error);
3082 			tvp = tv;
3083 		}
3084 
3085 		if (tvp[0].tv_usec < 0 || tvp[0].tv_usec >= 1000000 ||
3086 		    tvp[1].tv_usec < 0 || tvp[1].tv_usec >= 1000000)
3087 			return (EINVAL);
3088 		TIMEVAL_TO_TIMESPEC(&tvp[0], &tsp[0]);
3089 		TIMEVAL_TO_TIMESPEC(&tvp[1], &tsp[1]);
3090 	}
3091 	return (0);
3092 }
3093 
3094 /*
3095  * Common implementation code for utimes(), lutimes(), and futimes().
3096  */
3097 static int
3098 setutimes(td, vp, ts, numtimes, nullflag)
3099 	struct thread *td;
3100 	struct vnode *vp;
3101 	const struct timespec *ts;
3102 	int numtimes;
3103 	int nullflag;
3104 {
3105 	int error, setbirthtime;
3106 	struct mount *mp;
3107 	struct vattr vattr;
3108 
3109 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
3110 		return (error);
3111 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3112 	setbirthtime = 0;
3113 	if (numtimes < 3 && !VOP_GETATTR(vp, &vattr, td->td_ucred) &&
3114 	    timespeccmp(&ts[1], &vattr.va_birthtime, < ))
3115 		setbirthtime = 1;
3116 	VATTR_NULL(&vattr);
3117 	vattr.va_atime = ts[0];
3118 	vattr.va_mtime = ts[1];
3119 	if (setbirthtime)
3120 		vattr.va_birthtime = ts[1];
3121 	if (numtimes > 2)
3122 		vattr.va_birthtime = ts[2];
3123 	if (nullflag)
3124 		vattr.va_vaflags |= VA_UTIMES_NULL;
3125 #ifdef MAC
3126 	error = mac_vnode_check_setutimes(td->td_ucred, vp, vattr.va_atime,
3127 	    vattr.va_mtime);
3128 #endif
3129 	if (error == 0)
3130 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3131 	VOP_UNLOCK(vp, 0);
3132 	vn_finished_write(mp);
3133 	return (error);
3134 }
3135 
3136 /*
3137  * Set the access and modification times of a file.
3138  */
3139 #ifndef _SYS_SYSPROTO_H_
3140 struct utimes_args {
3141 	char	*path;
3142 	struct	timeval *tptr;
3143 };
3144 #endif
3145 int
3146 sys_utimes(td, uap)
3147 	struct thread *td;
3148 	register struct utimes_args /* {
3149 		char *path;
3150 		struct timeval *tptr;
3151 	} */ *uap;
3152 {
3153 
3154 	return (kern_utimes(td, uap->path, UIO_USERSPACE, uap->tptr,
3155 	    UIO_USERSPACE));
3156 }
3157 
3158 #ifndef _SYS_SYSPROTO_H_
3159 struct futimesat_args {
3160 	int fd;
3161 	const char * path;
3162 	const struct timeval * times;
3163 };
3164 #endif
3165 int
3166 sys_futimesat(struct thread *td, struct futimesat_args *uap)
3167 {
3168 
3169 	return (kern_utimesat(td, uap->fd, uap->path, UIO_USERSPACE,
3170 	    uap->times, UIO_USERSPACE));
3171 }
3172 
3173 int
3174 kern_utimes(struct thread *td, char *path, enum uio_seg pathseg,
3175     struct timeval *tptr, enum uio_seg tptrseg)
3176 {
3177 
3178 	return (kern_utimesat(td, AT_FDCWD, path, pathseg, tptr, tptrseg));
3179 }
3180 
3181 int
3182 kern_utimesat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
3183     struct timeval *tptr, enum uio_seg tptrseg)
3184 {
3185 	struct nameidata nd;
3186 	struct timespec ts[2];
3187 	int error;
3188 
3189 	if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3190 		return (error);
3191 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg,
3192 	    path, fd, CAP_FUTIMES, td);
3193 
3194 	if ((error = namei(&nd)) != 0)
3195 		return (error);
3196 	NDFREE(&nd, NDF_ONLY_PNBUF);
3197 	error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3198 	vrele(nd.ni_vp);
3199 	return (error);
3200 }
3201 
3202 /*
3203  * Set the access and modification times of a file.
3204  */
3205 #ifndef _SYS_SYSPROTO_H_
3206 struct lutimes_args {
3207 	char	*path;
3208 	struct	timeval *tptr;
3209 };
3210 #endif
3211 int
3212 sys_lutimes(td, uap)
3213 	struct thread *td;
3214 	register struct lutimes_args /* {
3215 		char *path;
3216 		struct timeval *tptr;
3217 	} */ *uap;
3218 {
3219 
3220 	return (kern_lutimes(td, uap->path, UIO_USERSPACE, uap->tptr,
3221 	    UIO_USERSPACE));
3222 }
3223 
3224 int
3225 kern_lutimes(struct thread *td, char *path, enum uio_seg pathseg,
3226     struct timeval *tptr, enum uio_seg tptrseg)
3227 {
3228 	struct timespec ts[2];
3229 	int error;
3230 	struct nameidata nd;
3231 
3232 	if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3233 		return (error);
3234 	NDINIT(&nd, LOOKUP, NOFOLLOW | AUDITVNODE1, pathseg, path, td);
3235 	if ((error = namei(&nd)) != 0)
3236 		return (error);
3237 	NDFREE(&nd, NDF_ONLY_PNBUF);
3238 	error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3239 	vrele(nd.ni_vp);
3240 	return (error);
3241 }
3242 
3243 /*
3244  * Set the access and modification times of a file.
3245  */
3246 #ifndef _SYS_SYSPROTO_H_
3247 struct futimes_args {
3248 	int	fd;
3249 	struct	timeval *tptr;
3250 };
3251 #endif
3252 int
3253 sys_futimes(td, uap)
3254 	struct thread *td;
3255 	register struct futimes_args /* {
3256 		int  fd;
3257 		struct timeval *tptr;
3258 	} */ *uap;
3259 {
3260 
3261 	return (kern_futimes(td, uap->fd, uap->tptr, UIO_USERSPACE));
3262 }
3263 
3264 int
3265 kern_futimes(struct thread *td, int fd, struct timeval *tptr,
3266     enum uio_seg tptrseg)
3267 {
3268 	struct timespec ts[2];
3269 	struct file *fp;
3270 	int error;
3271 
3272 	AUDIT_ARG_FD(fd);
3273 	if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3274 		return (error);
3275 	if ((error = getvnode(td->td_proc->p_fd, fd, CAP_FUTIMES, &fp))
3276 	    != 0)
3277 		return (error);
3278 #ifdef AUDIT
3279 	vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
3280 	AUDIT_ARG_VNODE1(fp->f_vnode);
3281 	VOP_UNLOCK(fp->f_vnode, 0);
3282 #endif
3283 	error = setutimes(td, fp->f_vnode, ts, 2, tptr == NULL);
3284 	fdrop(fp, td);
3285 	return (error);
3286 }
3287 
3288 /*
3289  * Truncate a file given its path name.
3290  */
3291 #ifndef _SYS_SYSPROTO_H_
3292 struct truncate_args {
3293 	char	*path;
3294 	int	pad;
3295 	off_t	length;
3296 };
3297 #endif
3298 int
3299 sys_truncate(td, uap)
3300 	struct thread *td;
3301 	register struct truncate_args /* {
3302 		char *path;
3303 		int pad;
3304 		off_t length;
3305 	} */ *uap;
3306 {
3307 
3308 	return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3309 }
3310 
3311 int
3312 kern_truncate(struct thread *td, char *path, enum uio_seg pathseg, off_t length)
3313 {
3314 	struct mount *mp;
3315 	struct vnode *vp;
3316 	void *rl_cookie;
3317 	struct vattr vattr;
3318 	struct nameidata nd;
3319 	int error;
3320 
3321 	if (length < 0)
3322 		return(EINVAL);
3323 	NDINIT(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, td);
3324 	if ((error = namei(&nd)) != 0)
3325 		return (error);
3326 	vp = nd.ni_vp;
3327 	rl_cookie = vn_rangelock_wlock(vp, 0, OFF_MAX);
3328 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) {
3329 		vn_rangelock_unlock(vp, rl_cookie);
3330 		vrele(vp);
3331 		return (error);
3332 	}
3333 	NDFREE(&nd, NDF_ONLY_PNBUF);
3334 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3335 	if (vp->v_type == VDIR)
3336 		error = EISDIR;
3337 #ifdef MAC
3338 	else if ((error = mac_vnode_check_write(td->td_ucred, NOCRED, vp))) {
3339 	}
3340 #endif
3341 	else if ((error = vn_writechk(vp)) == 0 &&
3342 	    (error = VOP_ACCESS(vp, VWRITE, td->td_ucred, td)) == 0) {
3343 		VATTR_NULL(&vattr);
3344 		vattr.va_size = length;
3345 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3346 	}
3347 	VOP_UNLOCK(vp, 0);
3348 	vn_finished_write(mp);
3349 	vn_rangelock_unlock(vp, rl_cookie);
3350 	vrele(vp);
3351 	return (error);
3352 }
3353 
3354 #if defined(COMPAT_43)
3355 /*
3356  * Truncate a file given its path name.
3357  */
3358 #ifndef _SYS_SYSPROTO_H_
3359 struct otruncate_args {
3360 	char	*path;
3361 	long	length;
3362 };
3363 #endif
3364 int
3365 otruncate(td, uap)
3366 	struct thread *td;
3367 	register struct otruncate_args /* {
3368 		char *path;
3369 		long length;
3370 	} */ *uap;
3371 {
3372 	struct truncate_args /* {
3373 		char *path;
3374 		int pad;
3375 		off_t length;
3376 	} */ nuap;
3377 
3378 	nuap.path = uap->path;
3379 	nuap.length = uap->length;
3380 	return (sys_truncate(td, &nuap));
3381 }
3382 #endif /* COMPAT_43 */
3383 
3384 /* Versions with the pad argument */
3385 int
3386 freebsd6_truncate(struct thread *td, struct freebsd6_truncate_args *uap)
3387 {
3388 	struct truncate_args ouap;
3389 
3390 	ouap.path = uap->path;
3391 	ouap.length = uap->length;
3392 	return (sys_truncate(td, &ouap));
3393 }
3394 
3395 int
3396 freebsd6_ftruncate(struct thread *td, struct freebsd6_ftruncate_args *uap)
3397 {
3398 	struct ftruncate_args ouap;
3399 
3400 	ouap.fd = uap->fd;
3401 	ouap.length = uap->length;
3402 	return (sys_ftruncate(td, &ouap));
3403 }
3404 
3405 /*
3406  * Sync an open file.
3407  */
3408 #ifndef _SYS_SYSPROTO_H_
3409 struct fsync_args {
3410 	int	fd;
3411 };
3412 #endif
3413 int
3414 sys_fsync(td, uap)
3415 	struct thread *td;
3416 	struct fsync_args /* {
3417 		int fd;
3418 	} */ *uap;
3419 {
3420 	struct vnode *vp;
3421 	struct mount *mp;
3422 	struct file *fp;
3423 	int error, lock_flags;
3424 
3425 	AUDIT_ARG_FD(uap->fd);
3426 	if ((error = getvnode(td->td_proc->p_fd, uap->fd, CAP_FSYNC,
3427 	    &fp)) != 0)
3428 		return (error);
3429 	vp = fp->f_vnode;
3430 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
3431 		goto drop;
3432 	if (MNT_SHARED_WRITES(mp) ||
3433 	    ((mp == NULL) && MNT_SHARED_WRITES(vp->v_mount))) {
3434 		lock_flags = LK_SHARED;
3435 	} else {
3436 		lock_flags = LK_EXCLUSIVE;
3437 	}
3438 	vn_lock(vp, lock_flags | LK_RETRY);
3439 	AUDIT_ARG_VNODE1(vp);
3440 	if (vp->v_object != NULL) {
3441 		VM_OBJECT_LOCK(vp->v_object);
3442 		vm_object_page_clean(vp->v_object, 0, 0, 0);
3443 		VM_OBJECT_UNLOCK(vp->v_object);
3444 	}
3445 	error = VOP_FSYNC(vp, MNT_WAIT, td);
3446 
3447 	VOP_UNLOCK(vp, 0);
3448 	vn_finished_write(mp);
3449 drop:
3450 	fdrop(fp, td);
3451 	return (error);
3452 }
3453 
3454 /*
3455  * Rename files.  Source and destination must either both be directories, or
3456  * both not be directories.  If target is a directory, it must be empty.
3457  */
3458 #ifndef _SYS_SYSPROTO_H_
3459 struct rename_args {
3460 	char	*from;
3461 	char	*to;
3462 };
3463 #endif
3464 int
3465 sys_rename(td, uap)
3466 	struct thread *td;
3467 	register struct rename_args /* {
3468 		char *from;
3469 		char *to;
3470 	} */ *uap;
3471 {
3472 
3473 	return (kern_rename(td, uap->from, uap->to, UIO_USERSPACE));
3474 }
3475 
3476 #ifndef _SYS_SYSPROTO_H_
3477 struct renameat_args {
3478 	int	oldfd;
3479 	char	*old;
3480 	int	newfd;
3481 	char	*new;
3482 };
3483 #endif
3484 int
3485 sys_renameat(struct thread *td, struct renameat_args *uap)
3486 {
3487 
3488 	return (kern_renameat(td, uap->oldfd, uap->old, uap->newfd, uap->new,
3489 	    UIO_USERSPACE));
3490 }
3491 
3492 int
3493 kern_rename(struct thread *td, char *from, char *to, enum uio_seg pathseg)
3494 {
3495 
3496 	return (kern_renameat(td, AT_FDCWD, from, AT_FDCWD, to, pathseg));
3497 }
3498 
3499 int
3500 kern_renameat(struct thread *td, int oldfd, char *old, int newfd, char *new,
3501     enum uio_seg pathseg)
3502 {
3503 	struct mount *mp = NULL;
3504 	struct vnode *tvp, *fvp, *tdvp;
3505 	struct nameidata fromnd, tond;
3506 	int error;
3507 
3508 	bwillwrite();
3509 #ifdef MAC
3510 	NDINIT_ATRIGHTS(&fromnd, DELETE, LOCKPARENT | LOCKLEAF | SAVESTART |
3511 	    AUDITVNODE1, pathseg, old, oldfd, CAP_DELETE, td);
3512 #else
3513 	NDINIT_ATRIGHTS(&fromnd, DELETE, WANTPARENT | SAVESTART |
3514 	    AUDITVNODE1, pathseg, old, oldfd, CAP_DELETE, td);
3515 #endif
3516 
3517 	if ((error = namei(&fromnd)) != 0)
3518 		return (error);
3519 #ifdef MAC
3520 	error = mac_vnode_check_rename_from(td->td_ucred, fromnd.ni_dvp,
3521 	    fromnd.ni_vp, &fromnd.ni_cnd);
3522 	VOP_UNLOCK(fromnd.ni_dvp, 0);
3523 	if (fromnd.ni_dvp != fromnd.ni_vp)
3524 		VOP_UNLOCK(fromnd.ni_vp, 0);
3525 #endif
3526 	fvp = fromnd.ni_vp;
3527 	if (error == 0)
3528 		error = vn_start_write(fvp, &mp, V_WAIT | PCATCH);
3529 	if (error != 0) {
3530 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3531 		vrele(fromnd.ni_dvp);
3532 		vrele(fvp);
3533 		goto out1;
3534 	}
3535 	NDINIT_ATRIGHTS(&tond, RENAME, LOCKPARENT | LOCKLEAF | NOCACHE |
3536 	    SAVESTART | AUDITVNODE2, pathseg, new, newfd, CAP_CREATE,
3537 	    td);
3538 	if (fromnd.ni_vp->v_type == VDIR)
3539 		tond.ni_cnd.cn_flags |= WILLBEDIR;
3540 	if ((error = namei(&tond)) != 0) {
3541 		/* Translate error code for rename("dir1", "dir2/."). */
3542 		if (error == EISDIR && fvp->v_type == VDIR)
3543 			error = EINVAL;
3544 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3545 		vrele(fromnd.ni_dvp);
3546 		vrele(fvp);
3547 		vn_finished_write(mp);
3548 		goto out1;
3549 	}
3550 	tdvp = tond.ni_dvp;
3551 	tvp = tond.ni_vp;
3552 	if (tvp != NULL) {
3553 		if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
3554 			error = ENOTDIR;
3555 			goto out;
3556 		} else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
3557 			error = EISDIR;
3558 			goto out;
3559 		}
3560 	}
3561 	if (fvp == tdvp) {
3562 		error = EINVAL;
3563 		goto out;
3564 	}
3565 	/*
3566 	 * If the source is the same as the destination (that is, if they
3567 	 * are links to the same vnode), then there is nothing to do.
3568 	 */
3569 	if (fvp == tvp)
3570 		error = -1;
3571 #ifdef MAC
3572 	else
3573 		error = mac_vnode_check_rename_to(td->td_ucred, tdvp,
3574 		    tond.ni_vp, fromnd.ni_dvp == tdvp, &tond.ni_cnd);
3575 #endif
3576 out:
3577 	if (!error) {
3578 		error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
3579 				   tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
3580 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3581 		NDFREE(&tond, NDF_ONLY_PNBUF);
3582 	} else {
3583 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3584 		NDFREE(&tond, NDF_ONLY_PNBUF);
3585 		if (tvp)
3586 			vput(tvp);
3587 		if (tdvp == tvp)
3588 			vrele(tdvp);
3589 		else
3590 			vput(tdvp);
3591 		vrele(fromnd.ni_dvp);
3592 		vrele(fvp);
3593 	}
3594 	vrele(tond.ni_startdir);
3595 	vn_finished_write(mp);
3596 out1:
3597 	if (fromnd.ni_startdir)
3598 		vrele(fromnd.ni_startdir);
3599 	if (error == -1)
3600 		return (0);
3601 	return (error);
3602 }
3603 
3604 /*
3605  * Make a directory file.
3606  */
3607 #ifndef _SYS_SYSPROTO_H_
3608 struct mkdir_args {
3609 	char	*path;
3610 	int	mode;
3611 };
3612 #endif
3613 int
3614 sys_mkdir(td, uap)
3615 	struct thread *td;
3616 	register struct mkdir_args /* {
3617 		char *path;
3618 		int mode;
3619 	} */ *uap;
3620 {
3621 
3622 	return (kern_mkdir(td, uap->path, UIO_USERSPACE, uap->mode));
3623 }
3624 
3625 #ifndef _SYS_SYSPROTO_H_
3626 struct mkdirat_args {
3627 	int	fd;
3628 	char	*path;
3629 	mode_t	mode;
3630 };
3631 #endif
3632 int
3633 sys_mkdirat(struct thread *td, struct mkdirat_args *uap)
3634 {
3635 
3636 	return (kern_mkdirat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode));
3637 }
3638 
3639 int
3640 kern_mkdir(struct thread *td, char *path, enum uio_seg segflg, int mode)
3641 {
3642 
3643 	return (kern_mkdirat(td, AT_FDCWD, path, segflg, mode));
3644 }
3645 
3646 int
3647 kern_mkdirat(struct thread *td, int fd, char *path, enum uio_seg segflg,
3648     int mode)
3649 {
3650 	struct mount *mp;
3651 	struct vnode *vp;
3652 	struct vattr vattr;
3653 	int error;
3654 	struct nameidata nd;
3655 
3656 	AUDIT_ARG_MODE(mode);
3657 restart:
3658 	bwillwrite();
3659 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME |
3660 	    AUDITVNODE1, segflg, path, fd, CAP_MKDIR, td);
3661 	nd.ni_cnd.cn_flags |= WILLBEDIR;
3662 	if ((error = namei(&nd)) != 0)
3663 		return (error);
3664 	vp = nd.ni_vp;
3665 	if (vp != NULL) {
3666 		NDFREE(&nd, NDF_ONLY_PNBUF);
3667 		/*
3668 		 * XXX namei called with LOCKPARENT but not LOCKLEAF has
3669 		 * the strange behaviour of leaving the vnode unlocked
3670 		 * if the target is the same vnode as the parent.
3671 		 */
3672 		if (vp == nd.ni_dvp)
3673 			vrele(nd.ni_dvp);
3674 		else
3675 			vput(nd.ni_dvp);
3676 		vrele(vp);
3677 		return (EEXIST);
3678 	}
3679 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3680 		NDFREE(&nd, NDF_ONLY_PNBUF);
3681 		vput(nd.ni_dvp);
3682 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3683 			return (error);
3684 		goto restart;
3685 	}
3686 	VATTR_NULL(&vattr);
3687 	vattr.va_type = VDIR;
3688 	vattr.va_mode = (mode & ACCESSPERMS) &~ td->td_proc->p_fd->fd_cmask;
3689 #ifdef MAC
3690 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
3691 	    &vattr);
3692 	if (error)
3693 		goto out;
3694 #endif
3695 	error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
3696 #ifdef MAC
3697 out:
3698 #endif
3699 	NDFREE(&nd, NDF_ONLY_PNBUF);
3700 	vput(nd.ni_dvp);
3701 	if (!error)
3702 		vput(nd.ni_vp);
3703 	vn_finished_write(mp);
3704 	return (error);
3705 }
3706 
3707 /*
3708  * Remove a directory file.
3709  */
3710 #ifndef _SYS_SYSPROTO_H_
3711 struct rmdir_args {
3712 	char	*path;
3713 };
3714 #endif
3715 int
3716 sys_rmdir(td, uap)
3717 	struct thread *td;
3718 	struct rmdir_args /* {
3719 		char *path;
3720 	} */ *uap;
3721 {
3722 
3723 	return (kern_rmdir(td, uap->path, UIO_USERSPACE));
3724 }
3725 
3726 int
3727 kern_rmdir(struct thread *td, char *path, enum uio_seg pathseg)
3728 {
3729 
3730 	return (kern_rmdirat(td, AT_FDCWD, path, pathseg));
3731 }
3732 
3733 int
3734 kern_rmdirat(struct thread *td, int fd, char *path, enum uio_seg pathseg)
3735 {
3736 	struct mount *mp;
3737 	struct vnode *vp;
3738 	int error;
3739 	struct nameidata nd;
3740 
3741 restart:
3742 	bwillwrite();
3743 	NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF |
3744 	    AUDITVNODE1, pathseg, path, fd, CAP_RMDIR, td);
3745 	if ((error = namei(&nd)) != 0)
3746 		return (error);
3747 	vp = nd.ni_vp;
3748 	if (vp->v_type != VDIR) {
3749 		error = ENOTDIR;
3750 		goto out;
3751 	}
3752 	/*
3753 	 * No rmdir "." please.
3754 	 */
3755 	if (nd.ni_dvp == vp) {
3756 		error = EINVAL;
3757 		goto out;
3758 	}
3759 	/*
3760 	 * The root of a mounted filesystem cannot be deleted.
3761 	 */
3762 	if (vp->v_vflag & VV_ROOT) {
3763 		error = EBUSY;
3764 		goto out;
3765 	}
3766 #ifdef MAC
3767 	error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
3768 	    &nd.ni_cnd);
3769 	if (error)
3770 		goto out;
3771 #endif
3772 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3773 		NDFREE(&nd, NDF_ONLY_PNBUF);
3774 		vput(vp);
3775 		if (nd.ni_dvp == vp)
3776 			vrele(nd.ni_dvp);
3777 		else
3778 			vput(nd.ni_dvp);
3779 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3780 			return (error);
3781 		goto restart;
3782 	}
3783 	error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
3784 	vn_finished_write(mp);
3785 out:
3786 	NDFREE(&nd, NDF_ONLY_PNBUF);
3787 	vput(vp);
3788 	if (nd.ni_dvp == vp)
3789 		vrele(nd.ni_dvp);
3790 	else
3791 		vput(nd.ni_dvp);
3792 	return (error);
3793 }
3794 
3795 #ifdef COMPAT_43
3796 /*
3797  * Read a block of directory entries in a filesystem independent format.
3798  */
3799 #ifndef _SYS_SYSPROTO_H_
3800 struct ogetdirentries_args {
3801 	int	fd;
3802 	char	*buf;
3803 	u_int	count;
3804 	long	*basep;
3805 };
3806 #endif
3807 int
3808 ogetdirentries(struct thread *td, struct ogetdirentries_args *uap)
3809 {
3810 	long loff;
3811 	int error;
3812 
3813 	error = kern_ogetdirentries(td, uap, &loff);
3814 	if (error == 0)
3815 		error = copyout(&loff, uap->basep, sizeof(long));
3816 	return (error);
3817 }
3818 
3819 int
3820 kern_ogetdirentries(struct thread *td, struct ogetdirentries_args *uap,
3821     long *ploff)
3822 {
3823 	struct vnode *vp;
3824 	struct file *fp;
3825 	struct uio auio, kuio;
3826 	struct iovec aiov, kiov;
3827 	struct dirent *dp, *edp;
3828 	caddr_t dirbuf;
3829 	int error, eofflag, readcnt;
3830 	long loff;
3831 	off_t foffset;
3832 
3833 	/* XXX arbitrary sanity limit on `count'. */
3834 	if (uap->count > 64 * 1024)
3835 		return (EINVAL);
3836 	if ((error = getvnode(td->td_proc->p_fd, uap->fd, CAP_READ,
3837 	    &fp)) != 0)
3838 		return (error);
3839 	if ((fp->f_flag & FREAD) == 0) {
3840 		fdrop(fp, td);
3841 		return (EBADF);
3842 	}
3843 	vp = fp->f_vnode;
3844 	foffset = foffset_lock(fp, 0);
3845 unionread:
3846 	if (vp->v_type != VDIR) {
3847 		foffset_unlock(fp, foffset, 0);
3848 		fdrop(fp, td);
3849 		return (EINVAL);
3850 	}
3851 	aiov.iov_base = uap->buf;
3852 	aiov.iov_len = uap->count;
3853 	auio.uio_iov = &aiov;
3854 	auio.uio_iovcnt = 1;
3855 	auio.uio_rw = UIO_READ;
3856 	auio.uio_segflg = UIO_USERSPACE;
3857 	auio.uio_td = td;
3858 	auio.uio_resid = uap->count;
3859 	vn_lock(vp, LK_SHARED | LK_RETRY);
3860 	loff = auio.uio_offset = foffset;
3861 #ifdef MAC
3862 	error = mac_vnode_check_readdir(td->td_ucred, vp);
3863 	if (error) {
3864 		VOP_UNLOCK(vp, 0);
3865 		foffset_unlock(fp, foffset, FOF_NOUPDATE);
3866 		fdrop(fp, td);
3867 		return (error);
3868 	}
3869 #endif
3870 #	if (BYTE_ORDER != LITTLE_ENDIAN)
3871 		if (vp->v_mount->mnt_maxsymlinklen <= 0) {
3872 			error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag,
3873 			    NULL, NULL);
3874 			foffset = auio.uio_offset;
3875 		} else
3876 #	endif
3877 	{
3878 		kuio = auio;
3879 		kuio.uio_iov = &kiov;
3880 		kuio.uio_segflg = UIO_SYSSPACE;
3881 		kiov.iov_len = uap->count;
3882 		dirbuf = malloc(uap->count, M_TEMP, M_WAITOK);
3883 		kiov.iov_base = dirbuf;
3884 		error = VOP_READDIR(vp, &kuio, fp->f_cred, &eofflag,
3885 			    NULL, NULL);
3886 		foffset = kuio.uio_offset;
3887 		if (error == 0) {
3888 			readcnt = uap->count - kuio.uio_resid;
3889 			edp = (struct dirent *)&dirbuf[readcnt];
3890 			for (dp = (struct dirent *)dirbuf; dp < edp; ) {
3891 #				if (BYTE_ORDER == LITTLE_ENDIAN)
3892 					/*
3893 					 * The expected low byte of
3894 					 * dp->d_namlen is our dp->d_type.
3895 					 * The high MBZ byte of dp->d_namlen
3896 					 * is our dp->d_namlen.
3897 					 */
3898 					dp->d_type = dp->d_namlen;
3899 					dp->d_namlen = 0;
3900 #				else
3901 					/*
3902 					 * The dp->d_type is the high byte
3903 					 * of the expected dp->d_namlen,
3904 					 * so must be zero'ed.
3905 					 */
3906 					dp->d_type = 0;
3907 #				endif
3908 				if (dp->d_reclen > 0) {
3909 					dp = (struct dirent *)
3910 					    ((char *)dp + dp->d_reclen);
3911 				} else {
3912 					error = EIO;
3913 					break;
3914 				}
3915 			}
3916 			if (dp >= edp)
3917 				error = uiomove(dirbuf, readcnt, &auio);
3918 		}
3919 		free(dirbuf, M_TEMP);
3920 	}
3921 	if (error) {
3922 		VOP_UNLOCK(vp, 0);
3923 		foffset_unlock(fp, foffset, 0);
3924 		fdrop(fp, td);
3925 		return (error);
3926 	}
3927 	if (uap->count == auio.uio_resid &&
3928 	    (vp->v_vflag & VV_ROOT) &&
3929 	    (vp->v_mount->mnt_flag & MNT_UNION)) {
3930 		struct vnode *tvp = vp;
3931 		vp = vp->v_mount->mnt_vnodecovered;
3932 		VREF(vp);
3933 		fp->f_vnode = vp;
3934 		fp->f_data = vp;
3935 		foffset = 0;
3936 		vput(tvp);
3937 		goto unionread;
3938 	}
3939 	VOP_UNLOCK(vp, 0);
3940 	foffset_unlock(fp, foffset, 0);
3941 	fdrop(fp, td);
3942 	td->td_retval[0] = uap->count - auio.uio_resid;
3943 	if (error == 0)
3944 		*ploff = loff;
3945 	return (error);
3946 }
3947 #endif /* COMPAT_43 */
3948 
3949 /*
3950  * Read a block of directory entries in a filesystem independent format.
3951  */
3952 #ifndef _SYS_SYSPROTO_H_
3953 struct getdirentries_args {
3954 	int	fd;
3955 	char	*buf;
3956 	u_int	count;
3957 	long	*basep;
3958 };
3959 #endif
3960 int
3961 sys_getdirentries(td, uap)
3962 	struct thread *td;
3963 	register struct getdirentries_args /* {
3964 		int fd;
3965 		char *buf;
3966 		u_int count;
3967 		long *basep;
3968 	} */ *uap;
3969 {
3970 	long base;
3971 	int error;
3972 
3973 	error = kern_getdirentries(td, uap->fd, uap->buf, uap->count, &base,
3974 	    NULL, UIO_USERSPACE);
3975 	if (error)
3976 		return (error);
3977 	if (uap->basep != NULL)
3978 		error = copyout(&base, uap->basep, sizeof(long));
3979 	return (error);
3980 }
3981 
3982 int
3983 kern_getdirentries(struct thread *td, int fd, char *buf, u_int count,
3984     long *basep, ssize_t *residp, enum uio_seg bufseg)
3985 {
3986 	struct vnode *vp;
3987 	struct file *fp;
3988 	struct uio auio;
3989 	struct iovec aiov;
3990 	long loff;
3991 	int error, eofflag;
3992 	off_t foffset;
3993 
3994 	AUDIT_ARG_FD(fd);
3995 	if (count > IOSIZE_MAX)
3996 		return (EINVAL);
3997 	auio.uio_resid = count;
3998 	if ((error = getvnode(td->td_proc->p_fd, fd, CAP_READ | CAP_SEEK,
3999 	    &fp)) != 0)
4000 		return (error);
4001 	if ((fp->f_flag & FREAD) == 0) {
4002 		fdrop(fp, td);
4003 		return (EBADF);
4004 	}
4005 	vp = fp->f_vnode;
4006 	foffset = foffset_lock(fp, 0);
4007 unionread:
4008 	if (vp->v_type != VDIR) {
4009 		error = EINVAL;
4010 		goto fail;
4011 	}
4012 	aiov.iov_base = buf;
4013 	aiov.iov_len = count;
4014 	auio.uio_iov = &aiov;
4015 	auio.uio_iovcnt = 1;
4016 	auio.uio_rw = UIO_READ;
4017 	auio.uio_segflg = bufseg;
4018 	auio.uio_td = td;
4019 	vn_lock(vp, LK_SHARED | LK_RETRY);
4020 	AUDIT_ARG_VNODE1(vp);
4021 	loff = auio.uio_offset = foffset;
4022 #ifdef MAC
4023 	error = mac_vnode_check_readdir(td->td_ucred, vp);
4024 	if (error == 0)
4025 #endif
4026 		error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, NULL,
4027 		    NULL);
4028 	foffset = auio.uio_offset;
4029 	if (error) {
4030 		VOP_UNLOCK(vp, 0);
4031 		goto fail;
4032 	}
4033 	if (count == auio.uio_resid &&
4034 	    (vp->v_vflag & VV_ROOT) &&
4035 	    (vp->v_mount->mnt_flag & MNT_UNION)) {
4036 		struct vnode *tvp = vp;
4037 		vp = vp->v_mount->mnt_vnodecovered;
4038 		VREF(vp);
4039 		fp->f_vnode = vp;
4040 		fp->f_data = vp;
4041 		foffset = 0;
4042 		vput(tvp);
4043 		goto unionread;
4044 	}
4045 	VOP_UNLOCK(vp, 0);
4046 	*basep = loff;
4047 	if (residp != NULL)
4048 		*residp = auio.uio_resid;
4049 	td->td_retval[0] = count - auio.uio_resid;
4050 fail:
4051 	foffset_unlock(fp, foffset, 0);
4052 	fdrop(fp, td);
4053 	return (error);
4054 }
4055 
4056 #ifndef _SYS_SYSPROTO_H_
4057 struct getdents_args {
4058 	int fd;
4059 	char *buf;
4060 	size_t count;
4061 };
4062 #endif
4063 int
4064 sys_getdents(td, uap)
4065 	struct thread *td;
4066 	register struct getdents_args /* {
4067 		int fd;
4068 		char *buf;
4069 		u_int count;
4070 	} */ *uap;
4071 {
4072 	struct getdirentries_args ap;
4073 	ap.fd = uap->fd;
4074 	ap.buf = uap->buf;
4075 	ap.count = uap->count;
4076 	ap.basep = NULL;
4077 	return (sys_getdirentries(td, &ap));
4078 }
4079 
4080 /*
4081  * Set the mode mask for creation of filesystem nodes.
4082  */
4083 #ifndef _SYS_SYSPROTO_H_
4084 struct umask_args {
4085 	int	newmask;
4086 };
4087 #endif
4088 int
4089 sys_umask(td, uap)
4090 	struct thread *td;
4091 	struct umask_args /* {
4092 		int newmask;
4093 	} */ *uap;
4094 {
4095 	register struct filedesc *fdp;
4096 
4097 	FILEDESC_XLOCK(td->td_proc->p_fd);
4098 	fdp = td->td_proc->p_fd;
4099 	td->td_retval[0] = fdp->fd_cmask;
4100 	fdp->fd_cmask = uap->newmask & ALLPERMS;
4101 	FILEDESC_XUNLOCK(td->td_proc->p_fd);
4102 	return (0);
4103 }
4104 
4105 /*
4106  * Void all references to file by ripping underlying filesystem away from
4107  * vnode.
4108  */
4109 #ifndef _SYS_SYSPROTO_H_
4110 struct revoke_args {
4111 	char	*path;
4112 };
4113 #endif
4114 int
4115 sys_revoke(td, uap)
4116 	struct thread *td;
4117 	register struct revoke_args /* {
4118 		char *path;
4119 	} */ *uap;
4120 {
4121 	struct vnode *vp;
4122 	struct vattr vattr;
4123 	int error;
4124 	struct nameidata nd;
4125 
4126 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1,
4127 	    UIO_USERSPACE, uap->path, td);
4128 	if ((error = namei(&nd)) != 0)
4129 		return (error);
4130 	vp = nd.ni_vp;
4131 	NDFREE(&nd, NDF_ONLY_PNBUF);
4132 	if (vp->v_type != VCHR || vp->v_rdev == NULL) {
4133 		error = EINVAL;
4134 		goto out;
4135 	}
4136 #ifdef MAC
4137 	error = mac_vnode_check_revoke(td->td_ucred, vp);
4138 	if (error)
4139 		goto out;
4140 #endif
4141 	error = VOP_GETATTR(vp, &vattr, td->td_ucred);
4142 	if (error)
4143 		goto out;
4144 	if (td->td_ucred->cr_uid != vattr.va_uid) {
4145 		error = priv_check(td, PRIV_VFS_ADMIN);
4146 		if (error)
4147 			goto out;
4148 	}
4149 	if (vcount(vp) > 1)
4150 		VOP_REVOKE(vp, REVOKEALL);
4151 out:
4152 	vput(vp);
4153 	return (error);
4154 }
4155 
4156 /*
4157  * Convert a user file descriptor to a kernel file entry and check that, if it
4158  * is a capability, the correct rights are present. A reference on the file
4159  * entry is held upon returning.
4160  */
4161 int
4162 getvnode(struct filedesc *fdp, int fd, cap_rights_t rights,
4163     struct file **fpp)
4164 {
4165 	struct file *fp;
4166 #ifdef CAPABILITIES
4167 	struct file *fp_fromcap;
4168 	int error;
4169 #endif
4170 
4171 	if (fdp == NULL || (fp = fget_unlocked(fdp, fd)) == NULL)
4172 		return (EBADF);
4173 #ifdef CAPABILITIES
4174 	/*
4175 	 * If the file descriptor is for a capability, test rights and use the
4176 	 * file descriptor referenced by the capability.
4177 	 */
4178 	error = cap_funwrap(fp, rights, &fp_fromcap);
4179 	if (error) {
4180 		fdrop(fp, curthread);
4181 		return (error);
4182 	}
4183 	if (fp != fp_fromcap) {
4184 		fhold(fp_fromcap);
4185 		fdrop(fp, curthread);
4186 		fp = fp_fromcap;
4187 	}
4188 #endif /* CAPABILITIES */
4189 
4190 	/*
4191 	 * The file could be not of the vnode type, or it may be not
4192 	 * yet fully initialized, in which case the f_vnode pointer
4193 	 * may be set, but f_ops is still badfileops.  E.g.,
4194 	 * devfs_open() transiently create such situation to
4195 	 * facilitate csw d_fdopen().
4196 	 *
4197 	 * Dupfdopen() handling in kern_openat() installs the
4198 	 * half-baked file into the process descriptor table, allowing
4199 	 * other thread to dereference it. Guard against the race by
4200 	 * checking f_ops.
4201 	 */
4202 	if (fp->f_vnode == NULL || fp->f_ops == &badfileops) {
4203 		fdrop(fp, curthread);
4204 		return (EINVAL);
4205 	}
4206 	*fpp = fp;
4207 	return (0);
4208 }
4209 
4210 
4211 /*
4212  * Get an (NFS) file handle.
4213  */
4214 #ifndef _SYS_SYSPROTO_H_
4215 struct lgetfh_args {
4216 	char	*fname;
4217 	fhandle_t *fhp;
4218 };
4219 #endif
4220 int
4221 sys_lgetfh(td, uap)
4222 	struct thread *td;
4223 	register struct lgetfh_args *uap;
4224 {
4225 	struct nameidata nd;
4226 	fhandle_t fh;
4227 	register struct vnode *vp;
4228 	int error;
4229 
4230 	error = priv_check(td, PRIV_VFS_GETFH);
4231 	if (error)
4232 		return (error);
4233 	NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | AUDITVNODE1,
4234 	    UIO_USERSPACE, uap->fname, td);
4235 	error = namei(&nd);
4236 	if (error)
4237 		return (error);
4238 	NDFREE(&nd, NDF_ONLY_PNBUF);
4239 	vp = nd.ni_vp;
4240 	bzero(&fh, sizeof(fh));
4241 	fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
4242 	error = VOP_VPTOFH(vp, &fh.fh_fid);
4243 	vput(vp);
4244 	if (error)
4245 		return (error);
4246 	error = copyout(&fh, uap->fhp, sizeof (fh));
4247 	return (error);
4248 }
4249 
4250 #ifndef _SYS_SYSPROTO_H_
4251 struct getfh_args {
4252 	char	*fname;
4253 	fhandle_t *fhp;
4254 };
4255 #endif
4256 int
4257 sys_getfh(td, uap)
4258 	struct thread *td;
4259 	register struct getfh_args *uap;
4260 {
4261 	struct nameidata nd;
4262 	fhandle_t fh;
4263 	register struct vnode *vp;
4264 	int error;
4265 
4266 	error = priv_check(td, PRIV_VFS_GETFH);
4267 	if (error)
4268 		return (error);
4269 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1,
4270 	    UIO_USERSPACE, uap->fname, td);
4271 	error = namei(&nd);
4272 	if (error)
4273 		return (error);
4274 	NDFREE(&nd, NDF_ONLY_PNBUF);
4275 	vp = nd.ni_vp;
4276 	bzero(&fh, sizeof(fh));
4277 	fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
4278 	error = VOP_VPTOFH(vp, &fh.fh_fid);
4279 	vput(vp);
4280 	if (error)
4281 		return (error);
4282 	error = copyout(&fh, uap->fhp, sizeof (fh));
4283 	return (error);
4284 }
4285 
4286 /*
4287  * syscall for the rpc.lockd to use to translate a NFS file handle into an
4288  * open descriptor.
4289  *
4290  * warning: do not remove the priv_check() call or this becomes one giant
4291  * security hole.
4292  */
4293 #ifndef _SYS_SYSPROTO_H_
4294 struct fhopen_args {
4295 	const struct fhandle *u_fhp;
4296 	int flags;
4297 };
4298 #endif
4299 int
4300 sys_fhopen(td, uap)
4301 	struct thread *td;
4302 	struct fhopen_args /* {
4303 		const struct fhandle *u_fhp;
4304 		int flags;
4305 	} */ *uap;
4306 {
4307 	struct mount *mp;
4308 	struct vnode *vp;
4309 	struct fhandle fhp;
4310 	struct file *fp;
4311 	int fmode, error;
4312 	int indx;
4313 
4314 	error = priv_check(td, PRIV_VFS_FHOPEN);
4315 	if (error)
4316 		return (error);
4317 	indx = -1;
4318 	fmode = FFLAGS(uap->flags);
4319 	/* why not allow a non-read/write open for our lockd? */
4320 	if (((fmode & (FREAD | FWRITE)) == 0) || (fmode & O_CREAT))
4321 		return (EINVAL);
4322 	error = copyin(uap->u_fhp, &fhp, sizeof(fhp));
4323 	if (error)
4324 		return(error);
4325 	/* find the mount point */
4326 	mp = vfs_busyfs(&fhp.fh_fsid);
4327 	if (mp == NULL)
4328 		return (ESTALE);
4329 	/* now give me my vnode, it gets returned to me locked */
4330 	error = VFS_FHTOVP(mp, &fhp.fh_fid, LK_EXCLUSIVE, &vp);
4331 	vfs_unbusy(mp);
4332 	if (error)
4333 		return (error);
4334 
4335 	error = falloc_noinstall(td, &fp);
4336 	if (error) {
4337 		vput(vp);
4338 		return (error);
4339 	}
4340 	/*
4341 	 * An extra reference on `fp' has been held for us by
4342 	 * falloc_noinstall().
4343 	 */
4344 
4345 #ifdef INVARIANTS
4346 	td->td_dupfd = -1;
4347 #endif
4348 	error = vn_open_vnode(vp, fmode, td->td_ucred, td, fp);
4349 	if (error) {
4350 		KASSERT(fp->f_ops == &badfileops,
4351 		    ("VOP_OPEN in fhopen() set f_ops"));
4352 		KASSERT(td->td_dupfd < 0,
4353 		    ("fhopen() encountered fdopen()"));
4354 
4355 		vput(vp);
4356 		goto bad;
4357 	}
4358 #ifdef INVARIANTS
4359 	td->td_dupfd = 0;
4360 #endif
4361 	fp->f_vnode = vp;
4362 	fp->f_seqcount = 1;
4363 	finit(fp, (fmode & FMASK) | (fp->f_flag & FHASLOCK), DTYPE_VNODE, vp,
4364 	    &vnops);
4365 	VOP_UNLOCK(vp, 0);
4366 	if (fmode & O_TRUNC) {
4367 		error = fo_truncate(fp, 0, td->td_ucred, td);
4368 		if (error)
4369 			goto bad;
4370 	}
4371 
4372 	error = finstall(td, fp, &indx, fmode);
4373 bad:
4374 	fdrop(fp, td);
4375 	td->td_retval[0] = indx;
4376 	return (error);
4377 }
4378 
4379 /*
4380  * Stat an (NFS) file handle.
4381  */
4382 #ifndef _SYS_SYSPROTO_H_
4383 struct fhstat_args {
4384 	struct fhandle *u_fhp;
4385 	struct stat *sb;
4386 };
4387 #endif
4388 int
4389 sys_fhstat(td, uap)
4390 	struct thread *td;
4391 	register struct fhstat_args /* {
4392 		struct fhandle *u_fhp;
4393 		struct stat *sb;
4394 	} */ *uap;
4395 {
4396 	struct stat sb;
4397 	struct fhandle fh;
4398 	int error;
4399 
4400 	error = copyin(uap->u_fhp, &fh, sizeof(fh));
4401 	if (error != 0)
4402 		return (error);
4403 	error = kern_fhstat(td, fh, &sb);
4404 	if (error != 0)
4405 		return (error);
4406 	error = copyout(&sb, uap->sb, sizeof(sb));
4407 	return (error);
4408 }
4409 
4410 int
4411 kern_fhstat(struct thread *td, struct fhandle fh, struct stat *sb)
4412 {
4413 	struct mount *mp;
4414 	struct vnode *vp;
4415 	int error;
4416 
4417 	error = priv_check(td, PRIV_VFS_FHSTAT);
4418 	if (error)
4419 		return (error);
4420 	if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4421 		return (ESTALE);
4422 	error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4423 	vfs_unbusy(mp);
4424 	if (error)
4425 		return (error);
4426 	error = vn_stat(vp, sb, td->td_ucred, NOCRED, td);
4427 	vput(vp);
4428 	return (error);
4429 }
4430 
4431 /*
4432  * Implement fstatfs() for (NFS) file handles.
4433  */
4434 #ifndef _SYS_SYSPROTO_H_
4435 struct fhstatfs_args {
4436 	struct fhandle *u_fhp;
4437 	struct statfs *buf;
4438 };
4439 #endif
4440 int
4441 sys_fhstatfs(td, uap)
4442 	struct thread *td;
4443 	struct fhstatfs_args /* {
4444 		struct fhandle *u_fhp;
4445 		struct statfs *buf;
4446 	} */ *uap;
4447 {
4448 	struct statfs sf;
4449 	fhandle_t fh;
4450 	int error;
4451 
4452 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
4453 	if (error)
4454 		return (error);
4455 	error = kern_fhstatfs(td, fh, &sf);
4456 	if (error)
4457 		return (error);
4458 	return (copyout(&sf, uap->buf, sizeof(sf)));
4459 }
4460 
4461 int
4462 kern_fhstatfs(struct thread *td, fhandle_t fh, struct statfs *buf)
4463 {
4464 	struct statfs *sp;
4465 	struct mount *mp;
4466 	struct vnode *vp;
4467 	int error;
4468 
4469 	error = priv_check(td, PRIV_VFS_FHSTATFS);
4470 	if (error)
4471 		return (error);
4472 	if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4473 		return (ESTALE);
4474 	error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4475 	if (error) {
4476 		vfs_unbusy(mp);
4477 		return (error);
4478 	}
4479 	vput(vp);
4480 	error = prison_canseemount(td->td_ucred, mp);
4481 	if (error)
4482 		goto out;
4483 #ifdef MAC
4484 	error = mac_mount_check_stat(td->td_ucred, mp);
4485 	if (error)
4486 		goto out;
4487 #endif
4488 	/*
4489 	 * Set these in case the underlying filesystem fails to do so.
4490 	 */
4491 	sp = &mp->mnt_stat;
4492 	sp->f_version = STATFS_VERSION;
4493 	sp->f_namemax = NAME_MAX;
4494 	sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
4495 	error = VFS_STATFS(mp, sp);
4496 	if (error == 0)
4497 		*buf = *sp;
4498 out:
4499 	vfs_unbusy(mp);
4500 	return (error);
4501 }
4502 
4503 int
4504 kern_posix_fallocate(struct thread *td, int fd, off_t offset, off_t len)
4505 {
4506 	struct file *fp;
4507 	struct mount *mp;
4508 	struct vnode *vp;
4509 	off_t olen, ooffset;
4510 	int error, vfslocked;
4511 
4512 	fp = NULL;
4513 	vfslocked = 0;
4514 	error = fget(td, fd, CAP_WRITE, &fp);
4515 	if (error != 0)
4516 		goto out;
4517 
4518 	switch (fp->f_type) {
4519 	case DTYPE_VNODE:
4520 		break;
4521 	case DTYPE_PIPE:
4522 	case DTYPE_FIFO:
4523 		error = ESPIPE;
4524 		goto out;
4525 	default:
4526 		error = ENODEV;
4527 		goto out;
4528 	}
4529 	if ((fp->f_flag & FWRITE) == 0) {
4530 		error = EBADF;
4531 		goto out;
4532 	}
4533 	vp = fp->f_vnode;
4534 	if (vp->v_type != VREG) {
4535 		error = ENODEV;
4536 		goto out;
4537 	}
4538 	if (offset < 0 || len <= 0) {
4539 		error = EINVAL;
4540 		goto out;
4541 	}
4542 	/* Check for wrap. */
4543 	if (offset > OFF_MAX - len) {
4544 		error = EFBIG;
4545 		goto out;
4546 	}
4547 
4548 	/* Allocating blocks may take a long time, so iterate. */
4549 	for (;;) {
4550 		olen = len;
4551 		ooffset = offset;
4552 
4553 		bwillwrite();
4554 		mp = NULL;
4555 		error = vn_start_write(vp, &mp, V_WAIT | PCATCH);
4556 		if (error != 0)
4557 			break;
4558 		error = vn_lock(vp, LK_EXCLUSIVE);
4559 		if (error != 0) {
4560 			vn_finished_write(mp);
4561 			break;
4562 		}
4563 #ifdef MAC
4564 		error = mac_vnode_check_write(td->td_ucred, fp->f_cred, vp);
4565 		if (error == 0)
4566 #endif
4567 			error = VOP_ALLOCATE(vp, &offset, &len);
4568 		VOP_UNLOCK(vp, 0);
4569 		vn_finished_write(mp);
4570 
4571 		if (olen + ooffset != offset + len) {
4572 			panic("offset + len changed from %jx/%jx to %jx/%jx",
4573 			    ooffset, olen, offset, len);
4574 		}
4575 		if (error != 0 || len == 0)
4576 			break;
4577 		KASSERT(olen > len, ("Iteration did not make progress?"));
4578 		maybe_yield();
4579 	}
4580  out:
4581 	if (fp != NULL)
4582 		fdrop(fp, td);
4583 	return (error);
4584 }
4585 
4586 int
4587 sys_posix_fallocate(struct thread *td, struct posix_fallocate_args *uap)
4588 {
4589 
4590 	return (kern_posix_fallocate(td, uap->fd, uap->offset, uap->len));
4591 }
4592 
4593 /*
4594  * Unlike madvise(2), we do not make a best effort to remember every
4595  * possible caching hint.  Instead, we remember the last setting with
4596  * the exception that we will allow POSIX_FADV_NORMAL to adjust the
4597  * region of any current setting.
4598  */
4599 int
4600 kern_posix_fadvise(struct thread *td, int fd, off_t offset, off_t len,
4601     int advice)
4602 {
4603 	struct fadvise_info *fa, *new;
4604 	struct file *fp;
4605 	struct vnode *vp;
4606 	off_t end;
4607 	int error;
4608 
4609 	if (offset < 0 || len < 0 || offset > OFF_MAX - len)
4610 		return (EINVAL);
4611 	switch (advice) {
4612 	case POSIX_FADV_SEQUENTIAL:
4613 	case POSIX_FADV_RANDOM:
4614 	case POSIX_FADV_NOREUSE:
4615 		new = malloc(sizeof(*fa), M_FADVISE, M_WAITOK);
4616 		break;
4617 	case POSIX_FADV_NORMAL:
4618 	case POSIX_FADV_WILLNEED:
4619 	case POSIX_FADV_DONTNEED:
4620 		new = NULL;
4621 		break;
4622 	default:
4623 		return (EINVAL);
4624 	}
4625 	/* XXX: CAP_POSIX_FADVISE? */
4626 	error = fget(td, fd, 0, &fp);
4627 	if (error != 0)
4628 		goto out;
4629 
4630 	switch (fp->f_type) {
4631 	case DTYPE_VNODE:
4632 		break;
4633 	case DTYPE_PIPE:
4634 	case DTYPE_FIFO:
4635 		error = ESPIPE;
4636 		goto out;
4637 	default:
4638 		error = ENODEV;
4639 		goto out;
4640 	}
4641 	vp = fp->f_vnode;
4642 	if (vp->v_type != VREG) {
4643 		error = ENODEV;
4644 		goto out;
4645 	}
4646 	if (len == 0)
4647 		end = OFF_MAX;
4648 	else
4649 		end = offset + len - 1;
4650 	switch (advice) {
4651 	case POSIX_FADV_SEQUENTIAL:
4652 	case POSIX_FADV_RANDOM:
4653 	case POSIX_FADV_NOREUSE:
4654 		/*
4655 		 * Try to merge any existing non-standard region with
4656 		 * this new region if possible, otherwise create a new
4657 		 * non-standard region for this request.
4658 		 */
4659 		mtx_pool_lock(mtxpool_sleep, fp);
4660 		fa = fp->f_advice;
4661 		if (fa != NULL && fa->fa_advice == advice &&
4662 		    ((fa->fa_start <= end && fa->fa_end >= offset) ||
4663 		    (end != OFF_MAX && fa->fa_start == end + 1) ||
4664 		    (fa->fa_end != OFF_MAX && fa->fa_end + 1 == offset))) {
4665 			if (offset < fa->fa_start)
4666 				fa->fa_start = offset;
4667 			if (end > fa->fa_end)
4668 				fa->fa_end = end;
4669 		} else {
4670 			new->fa_advice = advice;
4671 			new->fa_start = offset;
4672 			new->fa_end = end;
4673 			new->fa_prevstart = 0;
4674 			new->fa_prevend = 0;
4675 			fp->f_advice = new;
4676 			new = fa;
4677 		}
4678 		mtx_pool_unlock(mtxpool_sleep, fp);
4679 		break;
4680 	case POSIX_FADV_NORMAL:
4681 		/*
4682 		 * If a the "normal" region overlaps with an existing
4683 		 * non-standard region, trim or remove the
4684 		 * non-standard region.
4685 		 */
4686 		mtx_pool_lock(mtxpool_sleep, fp);
4687 		fa = fp->f_advice;
4688 		if (fa != NULL) {
4689 			if (offset <= fa->fa_start && end >= fa->fa_end) {
4690 				new = fa;
4691 				fp->f_advice = NULL;
4692 			} else if (offset <= fa->fa_start &&
4693  			    end >= fa->fa_start)
4694 				fa->fa_start = end + 1;
4695 			else if (offset <= fa->fa_end && end >= fa->fa_end)
4696 				fa->fa_end = offset - 1;
4697 			else if (offset >= fa->fa_start && end <= fa->fa_end) {
4698 				/*
4699 				 * If the "normal" region is a middle
4700 				 * portion of the existing
4701 				 * non-standard region, just remove
4702 				 * the whole thing rather than picking
4703 				 * one side or the other to
4704 				 * preserve.
4705 				 */
4706 				new = fa;
4707 				fp->f_advice = NULL;
4708 			}
4709 		}
4710 		mtx_pool_unlock(mtxpool_sleep, fp);
4711 		break;
4712 	case POSIX_FADV_WILLNEED:
4713 	case POSIX_FADV_DONTNEED:
4714 		error = VOP_ADVISE(vp, offset, end, advice);
4715 		break;
4716 	}
4717 out:
4718 	if (fp != NULL)
4719 		fdrop(fp, td);
4720 	free(new, M_FADVISE);
4721 	return (error);
4722 }
4723 
4724 int
4725 sys_posix_fadvise(struct thread *td, struct posix_fadvise_args *uap)
4726 {
4727 
4728 	return (kern_posix_fadvise(td, uap->fd, uap->offset, uap->len,
4729 	    uap->advice));
4730 }
4731