xref: /freebsd/sys/kern/vfs_syscalls.c (revision 2d0d168606f477d58d1a9aa72f94915da2f1733a)
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/rwlock.h>
65 #include <sys/sdt.h>
66 #include <sys/stat.h>
67 #include <sys/sx.h>
68 #include <sys/unistd.h>
69 #include <sys/vnode.h>
70 #include <sys/priv.h>
71 #include <sys/proc.h>
72 #include <sys/dirent.h>
73 #include <sys/jail.h>
74 #include <sys/syscallsubr.h>
75 #include <sys/sysctl.h>
76 #ifdef KTRACE
77 #include <sys/ktrace.h>
78 #endif
79 
80 #include <machine/stdarg.h>
81 
82 #include <security/audit/audit.h>
83 #include <security/mac/mac_framework.h>
84 
85 #include <vm/vm.h>
86 #include <vm/vm_object.h>
87 #include <vm/vm_page.h>
88 #include <vm/uma.h>
89 
90 #include <ufs/ufs/quota.h>
91 
92 MALLOC_DEFINE(M_FADVISE, "fadvise", "posix_fadvise(2) information");
93 
94 SDT_PROVIDER_DEFINE(vfs);
95 SDT_PROBE_DEFINE2(vfs, , stat, mode, mode, "char *", "int");
96 SDT_PROBE_DEFINE2(vfs, , stat, reg, reg, "char *", "int");
97 
98 static int chroot_refuse_vdir_fds(struct filedesc *fdp);
99 static int getutimes(const struct timeval *, enum uio_seg, struct timespec *);
100 static int kern_chflags(struct thread *td, const char *path,
101     enum uio_seg pathseg, u_long flags);
102 static int kern_chflagsat(struct thread *td, int fd, const char *path,
103     enum uio_seg pathseg, u_long flags, int atflag);
104 static int setfflags(struct thread *td, struct vnode *, u_long);
105 static int setutimes(struct thread *td, struct vnode *,
106     const struct timespec *, int, int);
107 static int vn_access(struct vnode *vp, int user_flags, struct ucred *cred,
108     struct thread *td);
109 
110 /*
111  * The module initialization routine for POSIX asynchronous I/O will
112  * set this to the version of AIO that it implements.  (Zero means
113  * that it is not implemented.)  This value is used here by pathconf()
114  * and in kern_descrip.c by fpathconf().
115  */
116 int async_io_version;
117 
118 #ifdef DEBUG
119 static int syncprt = 0;
120 SYSCTL_INT(_debug, OID_AUTO, syncprt, CTLFLAG_RW, &syncprt, 0, "");
121 #endif
122 
123 /*
124  * Sync each mounted filesystem.
125  */
126 #ifndef _SYS_SYSPROTO_H_
127 struct sync_args {
128 	int     dummy;
129 };
130 #endif
131 /* ARGSUSED */
132 int
133 sys_sync(td, uap)
134 	struct thread *td;
135 	struct sync_args *uap;
136 {
137 	struct mount *mp, *nmp;
138 	int save;
139 
140 	mtx_lock(&mountlist_mtx);
141 	for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
142 		if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK)) {
143 			nmp = TAILQ_NEXT(mp, mnt_list);
144 			continue;
145 		}
146 		if ((mp->mnt_flag & MNT_RDONLY) == 0 &&
147 		    vn_start_write(NULL, &mp, V_NOWAIT) == 0) {
148 			save = curthread_pflags_set(TDP_SYNCIO);
149 			vfs_msync(mp, MNT_NOWAIT);
150 			VFS_SYNC(mp, MNT_NOWAIT);
151 			curthread_pflags_restore(save);
152 			vn_finished_write(mp);
153 		}
154 		mtx_lock(&mountlist_mtx);
155 		nmp = TAILQ_NEXT(mp, mnt_list);
156 		vfs_unbusy(mp);
157 	}
158 	mtx_unlock(&mountlist_mtx);
159 	return (0);
160 }
161 
162 /*
163  * Change filesystem quotas.
164  */
165 #ifndef _SYS_SYSPROTO_H_
166 struct quotactl_args {
167 	char *path;
168 	int cmd;
169 	int uid;
170 	caddr_t arg;
171 };
172 #endif
173 int
174 sys_quotactl(td, uap)
175 	struct thread *td;
176 	register struct quotactl_args /* {
177 		char *path;
178 		int cmd;
179 		int uid;
180 		caddr_t arg;
181 	} */ *uap;
182 {
183 	struct mount *mp;
184 	struct nameidata nd;
185 	int error;
186 
187 	AUDIT_ARG_CMD(uap->cmd);
188 	AUDIT_ARG_UID(uap->uid);
189 	if (!prison_allow(td->td_ucred, PR_ALLOW_QUOTAS))
190 		return (EPERM);
191 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
192 	    uap->path, td);
193 	if ((error = namei(&nd)) != 0)
194 		return (error);
195 	NDFREE(&nd, NDF_ONLY_PNBUF);
196 	mp = nd.ni_vp->v_mount;
197 	vfs_ref(mp);
198 	vput(nd.ni_vp);
199 	error = vfs_busy(mp, 0);
200 	vfs_rel(mp);
201 	if (error != 0)
202 		return (error);
203 	error = VFS_QUOTACTL(mp, uap->cmd, uap->uid, uap->arg);
204 
205 	/*
206 	 * Since quota on operation typically needs to open quota
207 	 * file, the Q_QUOTAON handler needs to unbusy the mount point
208 	 * before calling into namei.  Otherwise, unmount might be
209 	 * started between two vfs_busy() invocations (first is our,
210 	 * second is from mount point cross-walk code in lookup()),
211 	 * causing deadlock.
212 	 *
213 	 * Require that Q_QUOTAON handles the vfs_busy() reference on
214 	 * its own, always returning with ubusied mount point.
215 	 */
216 	if ((uap->cmd >> SUBCMDSHIFT) != Q_QUOTAON)
217 		vfs_unbusy(mp);
218 	return (error);
219 }
220 
221 /*
222  * Used by statfs conversion routines to scale the block size up if
223  * necessary so that all of the block counts are <= 'max_size'.  Note
224  * that 'max_size' should be a bitmask, i.e. 2^n - 1 for some non-zero
225  * value of 'n'.
226  */
227 void
228 statfs_scale_blocks(struct statfs *sf, long max_size)
229 {
230 	uint64_t count;
231 	int shift;
232 
233 	KASSERT(powerof2(max_size + 1), ("%s: invalid max_size", __func__));
234 
235 	/*
236 	 * Attempt to scale the block counts to give a more accurate
237 	 * overview to userland of the ratio of free space to used
238 	 * space.  To do this, find the largest block count and compute
239 	 * a divisor that lets it fit into a signed integer <= max_size.
240 	 */
241 	if (sf->f_bavail < 0)
242 		count = -sf->f_bavail;
243 	else
244 		count = sf->f_bavail;
245 	count = MAX(sf->f_blocks, MAX(sf->f_bfree, count));
246 	if (count <= max_size)
247 		return;
248 
249 	count >>= flsl(max_size);
250 	shift = 0;
251 	while (count > 0) {
252 		shift++;
253 		count >>=1;
254 	}
255 
256 	sf->f_bsize <<= shift;
257 	sf->f_blocks >>= shift;
258 	sf->f_bfree >>= shift;
259 	sf->f_bavail >>= shift;
260 }
261 
262 /*
263  * Get filesystem statistics.
264  */
265 #ifndef _SYS_SYSPROTO_H_
266 struct statfs_args {
267 	char *path;
268 	struct statfs *buf;
269 };
270 #endif
271 int
272 sys_statfs(td, uap)
273 	struct thread *td;
274 	register struct statfs_args /* {
275 		char *path;
276 		struct statfs *buf;
277 	} */ *uap;
278 {
279 	struct statfs sf;
280 	int error;
281 
282 	error = kern_statfs(td, uap->path, UIO_USERSPACE, &sf);
283 	if (error == 0)
284 		error = copyout(&sf, uap->buf, sizeof(sf));
285 	return (error);
286 }
287 
288 int
289 kern_statfs(struct thread *td, char *path, enum uio_seg pathseg,
290     struct statfs *buf)
291 {
292 	struct mount *mp;
293 	struct statfs *sp, sb;
294 	struct nameidata nd;
295 	int error;
296 
297 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
298 	    pathseg, path, td);
299 	error = namei(&nd);
300 	if (error != 0)
301 		return (error);
302 	mp = nd.ni_vp->v_mount;
303 	vfs_ref(mp);
304 	NDFREE(&nd, NDF_ONLY_PNBUF);
305 	vput(nd.ni_vp);
306 	error = vfs_busy(mp, 0);
307 	vfs_rel(mp);
308 	if (error != 0)
309 		return (error);
310 #ifdef MAC
311 	error = mac_mount_check_stat(td->td_ucred, mp);
312 	if (error != 0)
313 		goto out;
314 #endif
315 	/*
316 	 * Set these in case the underlying filesystem fails to do so.
317 	 */
318 	sp = &mp->mnt_stat;
319 	sp->f_version = STATFS_VERSION;
320 	sp->f_namemax = NAME_MAX;
321 	sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
322 	error = VFS_STATFS(mp, sp);
323 	if (error != 0)
324 		goto out;
325 	if (priv_check(td, PRIV_VFS_GENERATION)) {
326 		bcopy(sp, &sb, sizeof(sb));
327 		sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
328 		prison_enforce_statfs(td->td_ucred, mp, &sb);
329 		sp = &sb;
330 	}
331 	*buf = *sp;
332 out:
333 	vfs_unbusy(mp);
334 	return (error);
335 }
336 
337 /*
338  * Get filesystem statistics.
339  */
340 #ifndef _SYS_SYSPROTO_H_
341 struct fstatfs_args {
342 	int fd;
343 	struct statfs *buf;
344 };
345 #endif
346 int
347 sys_fstatfs(td, uap)
348 	struct thread *td;
349 	register struct fstatfs_args /* {
350 		int fd;
351 		struct statfs *buf;
352 	} */ *uap;
353 {
354 	struct statfs sf;
355 	int error;
356 
357 	error = kern_fstatfs(td, uap->fd, &sf);
358 	if (error == 0)
359 		error = copyout(&sf, uap->buf, sizeof(sf));
360 	return (error);
361 }
362 
363 int
364 kern_fstatfs(struct thread *td, int fd, struct statfs *buf)
365 {
366 	struct file *fp;
367 	struct mount *mp;
368 	struct statfs *sp, sb;
369 	struct vnode *vp;
370 	cap_rights_t rights;
371 	int error;
372 
373 	AUDIT_ARG_FD(fd);
374 	error = getvnode(td->td_proc->p_fd, fd,
375 	    cap_rights_init(&rights, CAP_FSTATFS), &fp);
376 	if (error != 0)
377 		return (error);
378 	vp = fp->f_vnode;
379 	vn_lock(vp, LK_SHARED | LK_RETRY);
380 #ifdef AUDIT
381 	AUDIT_ARG_VNODE1(vp);
382 #endif
383 	mp = vp->v_mount;
384 	if (mp)
385 		vfs_ref(mp);
386 	VOP_UNLOCK(vp, 0);
387 	fdrop(fp, td);
388 	if (mp == NULL) {
389 		error = EBADF;
390 		goto out;
391 	}
392 	error = vfs_busy(mp, 0);
393 	vfs_rel(mp);
394 	if (error != 0)
395 		return (error);
396 #ifdef MAC
397 	error = mac_mount_check_stat(td->td_ucred, mp);
398 	if (error != 0)
399 		goto out;
400 #endif
401 	/*
402 	 * Set these in case the underlying filesystem fails to do so.
403 	 */
404 	sp = &mp->mnt_stat;
405 	sp->f_version = STATFS_VERSION;
406 	sp->f_namemax = NAME_MAX;
407 	sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
408 	error = VFS_STATFS(mp, sp);
409 	if (error != 0)
410 		goto out;
411 	if (priv_check(td, PRIV_VFS_GENERATION)) {
412 		bcopy(sp, &sb, sizeof(sb));
413 		sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
414 		prison_enforce_statfs(td->td_ucred, mp, &sb);
415 		sp = &sb;
416 	}
417 	*buf = *sp;
418 out:
419 	if (mp)
420 		vfs_unbusy(mp);
421 	return (error);
422 }
423 
424 /*
425  * Get statistics on all filesystems.
426  */
427 #ifndef _SYS_SYSPROTO_H_
428 struct getfsstat_args {
429 	struct statfs *buf;
430 	long bufsize;
431 	int flags;
432 };
433 #endif
434 int
435 sys_getfsstat(td, uap)
436 	struct thread *td;
437 	register struct getfsstat_args /* {
438 		struct statfs *buf;
439 		long bufsize;
440 		int flags;
441 	} */ *uap;
442 {
443 
444 	return (kern_getfsstat(td, &uap->buf, uap->bufsize, UIO_USERSPACE,
445 	    uap->flags));
446 }
447 
448 /*
449  * If (bufsize > 0 && bufseg == UIO_SYSSPACE)
450  *	The caller is responsible for freeing memory which will be allocated
451  *	in '*buf'.
452  */
453 int
454 kern_getfsstat(struct thread *td, struct statfs **buf, size_t bufsize,
455     enum uio_seg bufseg, int flags)
456 {
457 	struct mount *mp, *nmp;
458 	struct statfs *sfsp, *sp, sb;
459 	size_t count, maxcount;
460 	int error;
461 
462 	maxcount = bufsize / sizeof(struct statfs);
463 	if (bufsize == 0)
464 		sfsp = NULL;
465 	else if (bufseg == UIO_USERSPACE)
466 		sfsp = *buf;
467 	else /* if (bufseg == UIO_SYSSPACE) */ {
468 		count = 0;
469 		mtx_lock(&mountlist_mtx);
470 		TAILQ_FOREACH(mp, &mountlist, mnt_list) {
471 			count++;
472 		}
473 		mtx_unlock(&mountlist_mtx);
474 		if (maxcount > count)
475 			maxcount = count;
476 		sfsp = *buf = malloc(maxcount * sizeof(struct statfs), M_TEMP,
477 		    M_WAITOK);
478 	}
479 	count = 0;
480 	mtx_lock(&mountlist_mtx);
481 	for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
482 		if (prison_canseemount(td->td_ucred, mp) != 0) {
483 			nmp = TAILQ_NEXT(mp, mnt_list);
484 			continue;
485 		}
486 #ifdef MAC
487 		if (mac_mount_check_stat(td->td_ucred, mp) != 0) {
488 			nmp = TAILQ_NEXT(mp, mnt_list);
489 			continue;
490 		}
491 #endif
492 		if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK)) {
493 			nmp = TAILQ_NEXT(mp, mnt_list);
494 			continue;
495 		}
496 		if (sfsp && count < maxcount) {
497 			sp = &mp->mnt_stat;
498 			/*
499 			 * Set these in case the underlying filesystem
500 			 * fails to do so.
501 			 */
502 			sp->f_version = STATFS_VERSION;
503 			sp->f_namemax = NAME_MAX;
504 			sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
505 			/*
506 			 * If MNT_NOWAIT or MNT_LAZY is specified, do not
507 			 * refresh the fsstat cache. MNT_NOWAIT or MNT_LAZY
508 			 * overrides MNT_WAIT.
509 			 */
510 			if (((flags & (MNT_LAZY|MNT_NOWAIT)) == 0 ||
511 			    (flags & MNT_WAIT)) &&
512 			    (error = VFS_STATFS(mp, sp))) {
513 				mtx_lock(&mountlist_mtx);
514 				nmp = TAILQ_NEXT(mp, mnt_list);
515 				vfs_unbusy(mp);
516 				continue;
517 			}
518 			if (priv_check(td, PRIV_VFS_GENERATION)) {
519 				bcopy(sp, &sb, sizeof(sb));
520 				sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
521 				prison_enforce_statfs(td->td_ucred, mp, &sb);
522 				sp = &sb;
523 			}
524 			if (bufseg == UIO_SYSSPACE)
525 				bcopy(sp, sfsp, sizeof(*sp));
526 			else /* if (bufseg == UIO_USERSPACE) */ {
527 				error = copyout(sp, sfsp, sizeof(*sp));
528 				if (error != 0) {
529 					vfs_unbusy(mp);
530 					return (error);
531 				}
532 			}
533 			sfsp++;
534 		}
535 		count++;
536 		mtx_lock(&mountlist_mtx);
537 		nmp = TAILQ_NEXT(mp, mnt_list);
538 		vfs_unbusy(mp);
539 	}
540 	mtx_unlock(&mountlist_mtx);
541 	if (sfsp && count > maxcount)
542 		td->td_retval[0] = maxcount;
543 	else
544 		td->td_retval[0] = count;
545 	return (0);
546 }
547 
548 #ifdef COMPAT_FREEBSD4
549 /*
550  * Get old format filesystem statistics.
551  */
552 static void cvtstatfs(struct statfs *, struct ostatfs *);
553 
554 #ifndef _SYS_SYSPROTO_H_
555 struct freebsd4_statfs_args {
556 	char *path;
557 	struct ostatfs *buf;
558 };
559 #endif
560 int
561 freebsd4_statfs(td, uap)
562 	struct thread *td;
563 	struct freebsd4_statfs_args /* {
564 		char *path;
565 		struct ostatfs *buf;
566 	} */ *uap;
567 {
568 	struct ostatfs osb;
569 	struct statfs sf;
570 	int error;
571 
572 	error = kern_statfs(td, uap->path, UIO_USERSPACE, &sf);
573 	if (error != 0)
574 		return (error);
575 	cvtstatfs(&sf, &osb);
576 	return (copyout(&osb, uap->buf, sizeof(osb)));
577 }
578 
579 /*
580  * Get filesystem statistics.
581  */
582 #ifndef _SYS_SYSPROTO_H_
583 struct freebsd4_fstatfs_args {
584 	int fd;
585 	struct ostatfs *buf;
586 };
587 #endif
588 int
589 freebsd4_fstatfs(td, uap)
590 	struct thread *td;
591 	struct freebsd4_fstatfs_args /* {
592 		int fd;
593 		struct ostatfs *buf;
594 	} */ *uap;
595 {
596 	struct ostatfs osb;
597 	struct statfs sf;
598 	int error;
599 
600 	error = kern_fstatfs(td, uap->fd, &sf);
601 	if (error != 0)
602 		return (error);
603 	cvtstatfs(&sf, &osb);
604 	return (copyout(&osb, uap->buf, sizeof(osb)));
605 }
606 
607 /*
608  * Get statistics on all filesystems.
609  */
610 #ifndef _SYS_SYSPROTO_H_
611 struct freebsd4_getfsstat_args {
612 	struct ostatfs *buf;
613 	long bufsize;
614 	int flags;
615 };
616 #endif
617 int
618 freebsd4_getfsstat(td, uap)
619 	struct thread *td;
620 	register struct freebsd4_getfsstat_args /* {
621 		struct ostatfs *buf;
622 		long bufsize;
623 		int flags;
624 	} */ *uap;
625 {
626 	struct statfs *buf, *sp;
627 	struct ostatfs osb;
628 	size_t count, size;
629 	int error;
630 
631 	count = uap->bufsize / sizeof(struct ostatfs);
632 	size = count * sizeof(struct statfs);
633 	error = kern_getfsstat(td, &buf, size, UIO_SYSSPACE, uap->flags);
634 	if (size > 0) {
635 		count = td->td_retval[0];
636 		sp = buf;
637 		while (count > 0 && error == 0) {
638 			cvtstatfs(sp, &osb);
639 			error = copyout(&osb, uap->buf, sizeof(osb));
640 			sp++;
641 			uap->buf++;
642 			count--;
643 		}
644 		free(buf, M_TEMP);
645 	}
646 	return (error);
647 }
648 
649 /*
650  * Implement fstatfs() for (NFS) file handles.
651  */
652 #ifndef _SYS_SYSPROTO_H_
653 struct freebsd4_fhstatfs_args {
654 	struct fhandle *u_fhp;
655 	struct ostatfs *buf;
656 };
657 #endif
658 int
659 freebsd4_fhstatfs(td, uap)
660 	struct thread *td;
661 	struct freebsd4_fhstatfs_args /* {
662 		struct fhandle *u_fhp;
663 		struct ostatfs *buf;
664 	} */ *uap;
665 {
666 	struct ostatfs osb;
667 	struct statfs sf;
668 	fhandle_t fh;
669 	int error;
670 
671 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
672 	if (error != 0)
673 		return (error);
674 	error = kern_fhstatfs(td, fh, &sf);
675 	if (error != 0)
676 		return (error);
677 	cvtstatfs(&sf, &osb);
678 	return (copyout(&osb, uap->buf, sizeof(osb)));
679 }
680 
681 /*
682  * Convert a new format statfs structure to an old format statfs structure.
683  */
684 static void
685 cvtstatfs(nsp, osp)
686 	struct statfs *nsp;
687 	struct ostatfs *osp;
688 {
689 
690 	statfs_scale_blocks(nsp, LONG_MAX);
691 	bzero(osp, sizeof(*osp));
692 	osp->f_bsize = nsp->f_bsize;
693 	osp->f_iosize = MIN(nsp->f_iosize, LONG_MAX);
694 	osp->f_blocks = nsp->f_blocks;
695 	osp->f_bfree = nsp->f_bfree;
696 	osp->f_bavail = nsp->f_bavail;
697 	osp->f_files = MIN(nsp->f_files, LONG_MAX);
698 	osp->f_ffree = MIN(nsp->f_ffree, LONG_MAX);
699 	osp->f_owner = nsp->f_owner;
700 	osp->f_type = nsp->f_type;
701 	osp->f_flags = nsp->f_flags;
702 	osp->f_syncwrites = MIN(nsp->f_syncwrites, LONG_MAX);
703 	osp->f_asyncwrites = MIN(nsp->f_asyncwrites, LONG_MAX);
704 	osp->f_syncreads = MIN(nsp->f_syncreads, LONG_MAX);
705 	osp->f_asyncreads = MIN(nsp->f_asyncreads, LONG_MAX);
706 	strlcpy(osp->f_fstypename, nsp->f_fstypename,
707 	    MIN(MFSNAMELEN, OMFSNAMELEN));
708 	strlcpy(osp->f_mntonname, nsp->f_mntonname,
709 	    MIN(MNAMELEN, OMNAMELEN));
710 	strlcpy(osp->f_mntfromname, nsp->f_mntfromname,
711 	    MIN(MNAMELEN, OMNAMELEN));
712 	osp->f_fsid = nsp->f_fsid;
713 }
714 #endif /* COMPAT_FREEBSD4 */
715 
716 /*
717  * Change current working directory to a given file descriptor.
718  */
719 #ifndef _SYS_SYSPROTO_H_
720 struct fchdir_args {
721 	int	fd;
722 };
723 #endif
724 int
725 sys_fchdir(td, uap)
726 	struct thread *td;
727 	struct fchdir_args /* {
728 		int fd;
729 	} */ *uap;
730 {
731 	register struct filedesc *fdp = td->td_proc->p_fd;
732 	struct vnode *vp, *tdp, *vpold;
733 	struct mount *mp;
734 	struct file *fp;
735 	cap_rights_t rights;
736 	int error;
737 
738 	AUDIT_ARG_FD(uap->fd);
739 	error = getvnode(fdp, uap->fd, cap_rights_init(&rights, CAP_FCHDIR),
740 	    &fp);
741 	if (error != 0)
742 		return (error);
743 	vp = fp->f_vnode;
744 	VREF(vp);
745 	fdrop(fp, td);
746 	vn_lock(vp, LK_SHARED | LK_RETRY);
747 	AUDIT_ARG_VNODE1(vp);
748 	error = change_dir(vp, td);
749 	while (!error && (mp = vp->v_mountedhere) != NULL) {
750 		if (vfs_busy(mp, 0))
751 			continue;
752 		error = VFS_ROOT(mp, LK_SHARED, &tdp);
753 		vfs_unbusy(mp);
754 		if (error != 0)
755 			break;
756 		vput(vp);
757 		vp = tdp;
758 	}
759 	if (error != 0) {
760 		vput(vp);
761 		return (error);
762 	}
763 	VOP_UNLOCK(vp, 0);
764 	FILEDESC_XLOCK(fdp);
765 	vpold = fdp->fd_cdir;
766 	fdp->fd_cdir = vp;
767 	FILEDESC_XUNLOCK(fdp);
768 	vrele(vpold);
769 	return (0);
770 }
771 
772 /*
773  * Change current working directory (``.'').
774  */
775 #ifndef _SYS_SYSPROTO_H_
776 struct chdir_args {
777 	char	*path;
778 };
779 #endif
780 int
781 sys_chdir(td, uap)
782 	struct thread *td;
783 	struct chdir_args /* {
784 		char *path;
785 	} */ *uap;
786 {
787 
788 	return (kern_chdir(td, uap->path, UIO_USERSPACE));
789 }
790 
791 int
792 kern_chdir(struct thread *td, char *path, enum uio_seg pathseg)
793 {
794 	register struct filedesc *fdp = td->td_proc->p_fd;
795 	struct nameidata nd;
796 	struct vnode *vp;
797 	int error;
798 
799 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
800 	    pathseg, path, td);
801 	if ((error = namei(&nd)) != 0)
802 		return (error);
803 	if ((error = change_dir(nd.ni_vp, td)) != 0) {
804 		vput(nd.ni_vp);
805 		NDFREE(&nd, NDF_ONLY_PNBUF);
806 		return (error);
807 	}
808 	VOP_UNLOCK(nd.ni_vp, 0);
809 	NDFREE(&nd, NDF_ONLY_PNBUF);
810 	FILEDESC_XLOCK(fdp);
811 	vp = fdp->fd_cdir;
812 	fdp->fd_cdir = nd.ni_vp;
813 	FILEDESC_XUNLOCK(fdp);
814 	vrele(vp);
815 	return (0);
816 }
817 
818 /*
819  * Helper function for raised chroot(2) security function:  Refuse if
820  * any filedescriptors are open directories.
821  */
822 static int
823 chroot_refuse_vdir_fds(fdp)
824 	struct filedesc *fdp;
825 {
826 	struct vnode *vp;
827 	struct file *fp;
828 	int fd;
829 
830 	FILEDESC_LOCK_ASSERT(fdp);
831 
832 	for (fd = 0; fd < fdp->fd_nfiles ; fd++) {
833 		fp = fget_locked(fdp, fd);
834 		if (fp == NULL)
835 			continue;
836 		if (fp->f_type == DTYPE_VNODE) {
837 			vp = fp->f_vnode;
838 			if (vp->v_type == VDIR)
839 				return (EPERM);
840 		}
841 	}
842 	return (0);
843 }
844 
845 /*
846  * This sysctl determines if we will allow a process to chroot(2) if it
847  * has a directory open:
848  *	0: disallowed for all processes.
849  *	1: allowed for processes that were not already chroot(2)'ed.
850  *	2: allowed for all processes.
851  */
852 
853 static int chroot_allow_open_directories = 1;
854 
855 SYSCTL_INT(_kern, OID_AUTO, chroot_allow_open_directories, CTLFLAG_RW,
856      &chroot_allow_open_directories, 0,
857      "Allow a process to chroot(2) if it has a directory open");
858 
859 /*
860  * Change notion of root (``/'') directory.
861  */
862 #ifndef _SYS_SYSPROTO_H_
863 struct chroot_args {
864 	char	*path;
865 };
866 #endif
867 int
868 sys_chroot(td, uap)
869 	struct thread *td;
870 	struct chroot_args /* {
871 		char *path;
872 	} */ *uap;
873 {
874 	struct nameidata nd;
875 	int error;
876 
877 	error = priv_check(td, PRIV_VFS_CHROOT);
878 	if (error != 0)
879 		return (error);
880 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
881 	    UIO_USERSPACE, uap->path, td);
882 	error = namei(&nd);
883 	if (error != 0)
884 		goto error;
885 	error = change_dir(nd.ni_vp, td);
886 	if (error != 0)
887 		goto e_vunlock;
888 #ifdef MAC
889 	error = mac_vnode_check_chroot(td->td_ucred, nd.ni_vp);
890 	if (error != 0)
891 		goto e_vunlock;
892 #endif
893 	VOP_UNLOCK(nd.ni_vp, 0);
894 	error = change_root(nd.ni_vp, td);
895 	vrele(nd.ni_vp);
896 	NDFREE(&nd, NDF_ONLY_PNBUF);
897 	return (error);
898 e_vunlock:
899 	vput(nd.ni_vp);
900 error:
901 	NDFREE(&nd, NDF_ONLY_PNBUF);
902 	return (error);
903 }
904 
905 /*
906  * Common routine for chroot and chdir.  Callers must provide a locked vnode
907  * instance.
908  */
909 int
910 change_dir(vp, td)
911 	struct vnode *vp;
912 	struct thread *td;
913 {
914 #ifdef MAC
915 	int error;
916 #endif
917 
918 	ASSERT_VOP_LOCKED(vp, "change_dir(): vp not locked");
919 	if (vp->v_type != VDIR)
920 		return (ENOTDIR);
921 #ifdef MAC
922 	error = mac_vnode_check_chdir(td->td_ucred, vp);
923 	if (error != 0)
924 		return (error);
925 #endif
926 	return (VOP_ACCESS(vp, VEXEC, td->td_ucred, td));
927 }
928 
929 /*
930  * Common routine for kern_chroot() and jail_attach().  The caller is
931  * responsible for invoking priv_check() and mac_vnode_check_chroot() to
932  * authorize this operation.
933  */
934 int
935 change_root(vp, td)
936 	struct vnode *vp;
937 	struct thread *td;
938 {
939 	struct filedesc *fdp;
940 	struct vnode *oldvp;
941 	int error;
942 
943 	fdp = td->td_proc->p_fd;
944 	FILEDESC_XLOCK(fdp);
945 	if (chroot_allow_open_directories == 0 ||
946 	    (chroot_allow_open_directories == 1 && fdp->fd_rdir != rootvnode)) {
947 		error = chroot_refuse_vdir_fds(fdp);
948 		if (error != 0) {
949 			FILEDESC_XUNLOCK(fdp);
950 			return (error);
951 		}
952 	}
953 	oldvp = fdp->fd_rdir;
954 	fdp->fd_rdir = vp;
955 	VREF(fdp->fd_rdir);
956 	if (!fdp->fd_jdir) {
957 		fdp->fd_jdir = vp;
958 		VREF(fdp->fd_jdir);
959 	}
960 	FILEDESC_XUNLOCK(fdp);
961 	vrele(oldvp);
962 	return (0);
963 }
964 
965 static __inline void
966 flags_to_rights(int flags, cap_rights_t *rightsp)
967 {
968 
969 	if (flags & O_EXEC) {
970 		cap_rights_set(rightsp, CAP_FEXECVE);
971 	} else {
972 		switch ((flags & O_ACCMODE)) {
973 		case O_RDONLY:
974 			cap_rights_set(rightsp, CAP_READ);
975 			break;
976 		case O_RDWR:
977 			cap_rights_set(rightsp, CAP_READ);
978 			/* FALLTHROUGH */
979 		case O_WRONLY:
980 			cap_rights_set(rightsp, CAP_WRITE);
981 			if (!(flags & (O_APPEND | O_TRUNC)))
982 				cap_rights_set(rightsp, CAP_SEEK);
983 			break;
984 		}
985 	}
986 
987 	if (flags & O_CREAT)
988 		cap_rights_set(rightsp, CAP_CREATE);
989 
990 	if (flags & O_TRUNC)
991 		cap_rights_set(rightsp, CAP_FTRUNCATE);
992 
993 	if (flags & (O_SYNC | O_FSYNC))
994 		cap_rights_set(rightsp, CAP_FSYNC);
995 
996 	if (flags & (O_EXLOCK | O_SHLOCK))
997 		cap_rights_set(rightsp, CAP_FLOCK);
998 }
999 
1000 /*
1001  * Check permissions, allocate an open file structure, and call the device
1002  * open routine if any.
1003  */
1004 #ifndef _SYS_SYSPROTO_H_
1005 struct open_args {
1006 	char	*path;
1007 	int	flags;
1008 	int	mode;
1009 };
1010 #endif
1011 int
1012 sys_open(td, uap)
1013 	struct thread *td;
1014 	register struct open_args /* {
1015 		char *path;
1016 		int flags;
1017 		int mode;
1018 	} */ *uap;
1019 {
1020 
1021 	return (kern_open(td, uap->path, UIO_USERSPACE, uap->flags, uap->mode));
1022 }
1023 
1024 #ifndef _SYS_SYSPROTO_H_
1025 struct openat_args {
1026 	int	fd;
1027 	char	*path;
1028 	int	flag;
1029 	int	mode;
1030 };
1031 #endif
1032 int
1033 sys_openat(struct thread *td, struct openat_args *uap)
1034 {
1035 
1036 	return (kern_openat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
1037 	    uap->mode));
1038 }
1039 
1040 int
1041 kern_open(struct thread *td, char *path, enum uio_seg pathseg, int flags,
1042     int mode)
1043 {
1044 
1045 	return (kern_openat(td, AT_FDCWD, path, pathseg, flags, mode));
1046 }
1047 
1048 int
1049 kern_openat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1050     int flags, int mode)
1051 {
1052 	struct proc *p = td->td_proc;
1053 	struct filedesc *fdp = p->p_fd;
1054 	struct file *fp;
1055 	struct vnode *vp;
1056 	struct nameidata nd;
1057 	cap_rights_t rights;
1058 	int cmode, error, indx;
1059 
1060 	indx = -1;
1061 
1062 	AUDIT_ARG_FFLAGS(flags);
1063 	AUDIT_ARG_MODE(mode);
1064 	/* XXX: audit dirfd */
1065 	cap_rights_init(&rights, CAP_LOOKUP);
1066 	flags_to_rights(flags, &rights);
1067 	/*
1068 	 * Only one of the O_EXEC, O_RDONLY, O_WRONLY and O_RDWR flags
1069 	 * may be specified.
1070 	 */
1071 	if (flags & O_EXEC) {
1072 		if (flags & O_ACCMODE)
1073 			return (EINVAL);
1074 	} else if ((flags & O_ACCMODE) == O_ACCMODE) {
1075 		return (EINVAL);
1076 	} else {
1077 		flags = FFLAGS(flags);
1078 	}
1079 
1080 	/*
1081 	 * Allocate the file descriptor, but don't install a descriptor yet.
1082 	 */
1083 	error = falloc_noinstall(td, &fp);
1084 	if (error != 0)
1085 		return (error);
1086 	/*
1087 	 * An extra reference on `fp' has been held for us by
1088 	 * falloc_noinstall().
1089 	 */
1090 	/* Set the flags early so the finit in devfs can pick them up. */
1091 	fp->f_flag = flags & FMASK;
1092 	cmode = ((mode & ~fdp->fd_cmask) & ALLPERMS) & ~S_ISTXT;
1093 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, fd,
1094 	    &rights, td);
1095 	td->td_dupfd = -1;		/* XXX check for fdopen */
1096 	error = vn_open(&nd, &flags, cmode, fp);
1097 	if (error != 0) {
1098 		/*
1099 		 * If the vn_open replaced the method vector, something
1100 		 * wonderous happened deep below and we just pass it up
1101 		 * pretending we know what we do.
1102 		 */
1103 		if (error == ENXIO && fp->f_ops != &badfileops)
1104 			goto success;
1105 
1106 		/*
1107 		 * Handle special fdopen() case. bleh.
1108 		 *
1109 		 * Don't do this for relative (capability) lookups; we don't
1110 		 * understand exactly what would happen, and we don't think
1111 		 * that it ever should.
1112 		 */
1113 		if (nd.ni_strictrelative == 0 &&
1114 		    (error == ENODEV || error == ENXIO) &&
1115 		    td->td_dupfd >= 0) {
1116 			error = dupfdopen(td, fdp, td->td_dupfd, flags, error,
1117 			    &indx);
1118 			if (error == 0)
1119 				goto success;
1120 		}
1121 
1122 		goto bad;
1123 	}
1124 	td->td_dupfd = 0;
1125 	NDFREE(&nd, NDF_ONLY_PNBUF);
1126 	vp = nd.ni_vp;
1127 
1128 	/*
1129 	 * Store the vnode, for any f_type. Typically, the vnode use
1130 	 * count is decremented by direct call to vn_closefile() for
1131 	 * files that switched type in the cdevsw fdopen() method.
1132 	 */
1133 	fp->f_vnode = vp;
1134 	/*
1135 	 * If the file wasn't claimed by devfs bind it to the normal
1136 	 * vnode operations here.
1137 	 */
1138 	if (fp->f_ops == &badfileops) {
1139 		KASSERT(vp->v_type != VFIFO, ("Unexpected fifo."));
1140 		fp->f_seqcount = 1;
1141 		finit(fp, (flags & FMASK) | (fp->f_flag & FHASLOCK),
1142 		    DTYPE_VNODE, vp, &vnops);
1143 	}
1144 
1145 	VOP_UNLOCK(vp, 0);
1146 	if (flags & O_TRUNC) {
1147 		error = fo_truncate(fp, 0, td->td_ucred, td);
1148 		if (error != 0)
1149 			goto bad;
1150 	}
1151 success:
1152 	/*
1153 	 * If we haven't already installed the FD (for dupfdopen), do so now.
1154 	 */
1155 	if (indx == -1) {
1156 		struct filecaps *fcaps;
1157 
1158 #ifdef CAPABILITIES
1159 		if (nd.ni_strictrelative == 1)
1160 			fcaps = &nd.ni_filecaps;
1161 		else
1162 #endif
1163 			fcaps = NULL;
1164 		error = finstall(td, fp, &indx, flags, fcaps);
1165 		/* On success finstall() consumes fcaps. */
1166 		if (error != 0) {
1167 			filecaps_free(&nd.ni_filecaps);
1168 			goto bad;
1169 		}
1170 	} else {
1171 		filecaps_free(&nd.ni_filecaps);
1172 	}
1173 
1174 	/*
1175 	 * Release our private reference, leaving the one associated with
1176 	 * the descriptor table intact.
1177 	 */
1178 	fdrop(fp, td);
1179 	td->td_retval[0] = indx;
1180 	return (0);
1181 bad:
1182 	KASSERT(indx == -1, ("indx=%d, should be -1", indx));
1183 	fdrop(fp, td);
1184 	return (error);
1185 }
1186 
1187 #ifdef COMPAT_43
1188 /*
1189  * Create a file.
1190  */
1191 #ifndef _SYS_SYSPROTO_H_
1192 struct ocreat_args {
1193 	char	*path;
1194 	int	mode;
1195 };
1196 #endif
1197 int
1198 ocreat(td, uap)
1199 	struct thread *td;
1200 	register struct ocreat_args /* {
1201 		char *path;
1202 		int mode;
1203 	} */ *uap;
1204 {
1205 
1206 	return (kern_open(td, uap->path, UIO_USERSPACE,
1207 	    O_WRONLY | O_CREAT | O_TRUNC, uap->mode));
1208 }
1209 #endif /* COMPAT_43 */
1210 
1211 /*
1212  * Create a special file.
1213  */
1214 #ifndef _SYS_SYSPROTO_H_
1215 struct mknod_args {
1216 	char	*path;
1217 	int	mode;
1218 	int	dev;
1219 };
1220 #endif
1221 int
1222 sys_mknod(td, uap)
1223 	struct thread *td;
1224 	register struct mknod_args /* {
1225 		char *path;
1226 		int mode;
1227 		int dev;
1228 	} */ *uap;
1229 {
1230 
1231 	return (kern_mknod(td, uap->path, UIO_USERSPACE, uap->mode, uap->dev));
1232 }
1233 
1234 #ifndef _SYS_SYSPROTO_H_
1235 struct mknodat_args {
1236 	int	fd;
1237 	char	*path;
1238 	mode_t	mode;
1239 	dev_t	dev;
1240 };
1241 #endif
1242 int
1243 sys_mknodat(struct thread *td, struct mknodat_args *uap)
1244 {
1245 
1246 	return (kern_mknodat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode,
1247 	    uap->dev));
1248 }
1249 
1250 int
1251 kern_mknod(struct thread *td, char *path, enum uio_seg pathseg, int mode,
1252     int dev)
1253 {
1254 
1255 	return (kern_mknodat(td, AT_FDCWD, path, pathseg, mode, dev));
1256 }
1257 
1258 int
1259 kern_mknodat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1260     int mode, int dev)
1261 {
1262 	struct vnode *vp;
1263 	struct mount *mp;
1264 	struct vattr vattr;
1265 	struct nameidata nd;
1266 	cap_rights_t rights;
1267 	int error, whiteout = 0;
1268 
1269 	AUDIT_ARG_MODE(mode);
1270 	AUDIT_ARG_DEV(dev);
1271 	switch (mode & S_IFMT) {
1272 	case S_IFCHR:
1273 	case S_IFBLK:
1274 		error = priv_check(td, PRIV_VFS_MKNOD_DEV);
1275 		break;
1276 	case S_IFMT:
1277 		error = priv_check(td, PRIV_VFS_MKNOD_BAD);
1278 		break;
1279 	case S_IFWHT:
1280 		error = priv_check(td, PRIV_VFS_MKNOD_WHT);
1281 		break;
1282 	case S_IFIFO:
1283 		if (dev == 0)
1284 			return (kern_mkfifoat(td, fd, path, pathseg, mode));
1285 		/* FALLTHROUGH */
1286 	default:
1287 		error = EINVAL;
1288 		break;
1289 	}
1290 	if (error != 0)
1291 		return (error);
1292 restart:
1293 	bwillwrite();
1294 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1,
1295 	    pathseg, path, fd, cap_rights_init(&rights, CAP_MKNODAT), td);
1296 	if ((error = namei(&nd)) != 0)
1297 		return (error);
1298 	vp = nd.ni_vp;
1299 	if (vp != NULL) {
1300 		NDFREE(&nd, NDF_ONLY_PNBUF);
1301 		if (vp == nd.ni_dvp)
1302 			vrele(nd.ni_dvp);
1303 		else
1304 			vput(nd.ni_dvp);
1305 		vrele(vp);
1306 		return (EEXIST);
1307 	} else {
1308 		VATTR_NULL(&vattr);
1309 		vattr.va_mode = (mode & ALLPERMS) &
1310 		    ~td->td_proc->p_fd->fd_cmask;
1311 		vattr.va_rdev = dev;
1312 		whiteout = 0;
1313 
1314 		switch (mode & S_IFMT) {
1315 		case S_IFMT:	/* used by badsect to flag bad sectors */
1316 			vattr.va_type = VBAD;
1317 			break;
1318 		case S_IFCHR:
1319 			vattr.va_type = VCHR;
1320 			break;
1321 		case S_IFBLK:
1322 			vattr.va_type = VBLK;
1323 			break;
1324 		case S_IFWHT:
1325 			whiteout = 1;
1326 			break;
1327 		default:
1328 			panic("kern_mknod: invalid mode");
1329 		}
1330 	}
1331 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1332 		NDFREE(&nd, NDF_ONLY_PNBUF);
1333 		vput(nd.ni_dvp);
1334 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1335 			return (error);
1336 		goto restart;
1337 	}
1338 #ifdef MAC
1339 	if (error == 0 && !whiteout)
1340 		error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp,
1341 		    &nd.ni_cnd, &vattr);
1342 #endif
1343 	if (error == 0) {
1344 		if (whiteout)
1345 			error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE);
1346 		else {
1347 			error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp,
1348 						&nd.ni_cnd, &vattr);
1349 			if (error == 0)
1350 				vput(nd.ni_vp);
1351 		}
1352 	}
1353 	NDFREE(&nd, NDF_ONLY_PNBUF);
1354 	vput(nd.ni_dvp);
1355 	vn_finished_write(mp);
1356 	return (error);
1357 }
1358 
1359 /*
1360  * Create a named pipe.
1361  */
1362 #ifndef _SYS_SYSPROTO_H_
1363 struct mkfifo_args {
1364 	char	*path;
1365 	int	mode;
1366 };
1367 #endif
1368 int
1369 sys_mkfifo(td, uap)
1370 	struct thread *td;
1371 	register struct mkfifo_args /* {
1372 		char *path;
1373 		int mode;
1374 	} */ *uap;
1375 {
1376 
1377 	return (kern_mkfifo(td, uap->path, UIO_USERSPACE, uap->mode));
1378 }
1379 
1380 #ifndef _SYS_SYSPROTO_H_
1381 struct mkfifoat_args {
1382 	int	fd;
1383 	char	*path;
1384 	mode_t	mode;
1385 };
1386 #endif
1387 int
1388 sys_mkfifoat(struct thread *td, struct mkfifoat_args *uap)
1389 {
1390 
1391 	return (kern_mkfifoat(td, uap->fd, uap->path, UIO_USERSPACE,
1392 	    uap->mode));
1393 }
1394 
1395 int
1396 kern_mkfifo(struct thread *td, char *path, enum uio_seg pathseg, int mode)
1397 {
1398 
1399 	return (kern_mkfifoat(td, AT_FDCWD, path, pathseg, mode));
1400 }
1401 
1402 int
1403 kern_mkfifoat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1404     int mode)
1405 {
1406 	struct mount *mp;
1407 	struct vattr vattr;
1408 	struct nameidata nd;
1409 	cap_rights_t rights;
1410 	int error;
1411 
1412 	AUDIT_ARG_MODE(mode);
1413 restart:
1414 	bwillwrite();
1415 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1,
1416 	    pathseg, path, fd, cap_rights_init(&rights, CAP_MKFIFOAT), td);
1417 	if ((error = namei(&nd)) != 0)
1418 		return (error);
1419 	if (nd.ni_vp != NULL) {
1420 		NDFREE(&nd, NDF_ONLY_PNBUF);
1421 		if (nd.ni_vp == nd.ni_dvp)
1422 			vrele(nd.ni_dvp);
1423 		else
1424 			vput(nd.ni_dvp);
1425 		vrele(nd.ni_vp);
1426 		return (EEXIST);
1427 	}
1428 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1429 		NDFREE(&nd, NDF_ONLY_PNBUF);
1430 		vput(nd.ni_dvp);
1431 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1432 			return (error);
1433 		goto restart;
1434 	}
1435 	VATTR_NULL(&vattr);
1436 	vattr.va_type = VFIFO;
1437 	vattr.va_mode = (mode & ALLPERMS) & ~td->td_proc->p_fd->fd_cmask;
1438 #ifdef MAC
1439 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1440 	    &vattr);
1441 	if (error != 0)
1442 		goto out;
1443 #endif
1444 	error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
1445 	if (error == 0)
1446 		vput(nd.ni_vp);
1447 #ifdef MAC
1448 out:
1449 #endif
1450 	vput(nd.ni_dvp);
1451 	vn_finished_write(mp);
1452 	NDFREE(&nd, NDF_ONLY_PNBUF);
1453 	return (error);
1454 }
1455 
1456 /*
1457  * Make a hard file link.
1458  */
1459 #ifndef _SYS_SYSPROTO_H_
1460 struct link_args {
1461 	char	*path;
1462 	char	*link;
1463 };
1464 #endif
1465 int
1466 sys_link(td, uap)
1467 	struct thread *td;
1468 	register struct link_args /* {
1469 		char *path;
1470 		char *link;
1471 	} */ *uap;
1472 {
1473 
1474 	return (kern_link(td, uap->path, uap->link, UIO_USERSPACE));
1475 }
1476 
1477 #ifndef _SYS_SYSPROTO_H_
1478 struct linkat_args {
1479 	int	fd1;
1480 	char	*path1;
1481 	int	fd2;
1482 	char	*path2;
1483 	int	flag;
1484 };
1485 #endif
1486 int
1487 sys_linkat(struct thread *td, struct linkat_args *uap)
1488 {
1489 	int flag;
1490 
1491 	flag = uap->flag;
1492 	if (flag & ~AT_SYMLINK_FOLLOW)
1493 		return (EINVAL);
1494 
1495 	return (kern_linkat(td, uap->fd1, uap->fd2, uap->path1, uap->path2,
1496 	    UIO_USERSPACE, (flag & AT_SYMLINK_FOLLOW) ? FOLLOW : NOFOLLOW));
1497 }
1498 
1499 int hardlink_check_uid = 0;
1500 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_uid, CTLFLAG_RW,
1501     &hardlink_check_uid, 0,
1502     "Unprivileged processes cannot create hard links to files owned by other "
1503     "users");
1504 static int hardlink_check_gid = 0;
1505 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_gid, CTLFLAG_RW,
1506     &hardlink_check_gid, 0,
1507     "Unprivileged processes cannot create hard links to files owned by other "
1508     "groups");
1509 
1510 static int
1511 can_hardlink(struct vnode *vp, struct ucred *cred)
1512 {
1513 	struct vattr va;
1514 	int error;
1515 
1516 	if (!hardlink_check_uid && !hardlink_check_gid)
1517 		return (0);
1518 
1519 	error = VOP_GETATTR(vp, &va, cred);
1520 	if (error != 0)
1521 		return (error);
1522 
1523 	if (hardlink_check_uid && cred->cr_uid != va.va_uid) {
1524 		error = priv_check_cred(cred, PRIV_VFS_LINK, 0);
1525 		if (error != 0)
1526 			return (error);
1527 	}
1528 
1529 	if (hardlink_check_gid && !groupmember(va.va_gid, cred)) {
1530 		error = priv_check_cred(cred, PRIV_VFS_LINK, 0);
1531 		if (error != 0)
1532 			return (error);
1533 	}
1534 
1535 	return (0);
1536 }
1537 
1538 int
1539 kern_link(struct thread *td, char *path, char *link, enum uio_seg segflg)
1540 {
1541 
1542 	return (kern_linkat(td, AT_FDCWD, AT_FDCWD, path,link, segflg, FOLLOW));
1543 }
1544 
1545 int
1546 kern_linkat(struct thread *td, int fd1, int fd2, char *path1, char *path2,
1547     enum uio_seg segflg, int follow)
1548 {
1549 	struct vnode *vp;
1550 	struct mount *mp;
1551 	struct nameidata nd;
1552 	cap_rights_t rights;
1553 	int error;
1554 
1555 	bwillwrite();
1556 	NDINIT_AT(&nd, LOOKUP, follow | AUDITVNODE1, segflg, path1, fd1, td);
1557 
1558 	if ((error = namei(&nd)) != 0)
1559 		return (error);
1560 	NDFREE(&nd, NDF_ONLY_PNBUF);
1561 	vp = nd.ni_vp;
1562 	if (vp->v_type == VDIR) {
1563 		vrele(vp);
1564 		return (EPERM);		/* POSIX */
1565 	}
1566 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) {
1567 		vrele(vp);
1568 		return (error);
1569 	}
1570 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE2,
1571 	    segflg, path2, fd2, cap_rights_init(&rights, CAP_LINKAT), td);
1572 	if ((error = namei(&nd)) == 0) {
1573 		if (nd.ni_vp != NULL) {
1574 			if (nd.ni_dvp == nd.ni_vp)
1575 				vrele(nd.ni_dvp);
1576 			else
1577 				vput(nd.ni_dvp);
1578 			vrele(nd.ni_vp);
1579 			error = EEXIST;
1580 		} else if ((error = vn_lock(vp, LK_EXCLUSIVE | LK_RETRY))
1581 		    == 0) {
1582 			error = can_hardlink(vp, td->td_ucred);
1583 			if (error == 0)
1584 #ifdef MAC
1585 				error = mac_vnode_check_link(td->td_ucred,
1586 				    nd.ni_dvp, vp, &nd.ni_cnd);
1587 			if (error == 0)
1588 #endif
1589 				error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
1590 			VOP_UNLOCK(vp, 0);
1591 			vput(nd.ni_dvp);
1592 		}
1593 		NDFREE(&nd, NDF_ONLY_PNBUF);
1594 	}
1595 	vrele(vp);
1596 	vn_finished_write(mp);
1597 	return (error);
1598 }
1599 
1600 /*
1601  * Make a symbolic link.
1602  */
1603 #ifndef _SYS_SYSPROTO_H_
1604 struct symlink_args {
1605 	char	*path;
1606 	char	*link;
1607 };
1608 #endif
1609 int
1610 sys_symlink(td, uap)
1611 	struct thread *td;
1612 	register struct symlink_args /* {
1613 		char *path;
1614 		char *link;
1615 	} */ *uap;
1616 {
1617 
1618 	return (kern_symlink(td, uap->path, uap->link, UIO_USERSPACE));
1619 }
1620 
1621 #ifndef _SYS_SYSPROTO_H_
1622 struct symlinkat_args {
1623 	char	*path;
1624 	int	fd;
1625 	char	*path2;
1626 };
1627 #endif
1628 int
1629 sys_symlinkat(struct thread *td, struct symlinkat_args *uap)
1630 {
1631 
1632 	return (kern_symlinkat(td, uap->path1, uap->fd, uap->path2,
1633 	    UIO_USERSPACE));
1634 }
1635 
1636 int
1637 kern_symlink(struct thread *td, char *path, char *link, enum uio_seg segflg)
1638 {
1639 
1640 	return (kern_symlinkat(td, path, AT_FDCWD, link, segflg));
1641 }
1642 
1643 int
1644 kern_symlinkat(struct thread *td, char *path1, int fd, char *path2,
1645     enum uio_seg segflg)
1646 {
1647 	struct mount *mp;
1648 	struct vattr vattr;
1649 	char *syspath;
1650 	struct nameidata nd;
1651 	int error;
1652 	cap_rights_t rights;
1653 
1654 	if (segflg == UIO_SYSSPACE) {
1655 		syspath = path1;
1656 	} else {
1657 		syspath = uma_zalloc(namei_zone, M_WAITOK);
1658 		if ((error = copyinstr(path1, syspath, MAXPATHLEN, NULL)) != 0)
1659 			goto out;
1660 	}
1661 	AUDIT_ARG_TEXT(syspath);
1662 restart:
1663 	bwillwrite();
1664 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1,
1665 	    segflg, path2, fd, cap_rights_init(&rights, CAP_SYMLINKAT), td);
1666 	if ((error = namei(&nd)) != 0)
1667 		goto out;
1668 	if (nd.ni_vp) {
1669 		NDFREE(&nd, NDF_ONLY_PNBUF);
1670 		if (nd.ni_vp == nd.ni_dvp)
1671 			vrele(nd.ni_dvp);
1672 		else
1673 			vput(nd.ni_dvp);
1674 		vrele(nd.ni_vp);
1675 		error = EEXIST;
1676 		goto out;
1677 	}
1678 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1679 		NDFREE(&nd, NDF_ONLY_PNBUF);
1680 		vput(nd.ni_dvp);
1681 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1682 			goto out;
1683 		goto restart;
1684 	}
1685 	VATTR_NULL(&vattr);
1686 	vattr.va_mode = ACCESSPERMS &~ td->td_proc->p_fd->fd_cmask;
1687 #ifdef MAC
1688 	vattr.va_type = VLNK;
1689 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1690 	    &vattr);
1691 	if (error != 0)
1692 		goto out2;
1693 #endif
1694 	error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, syspath);
1695 	if (error == 0)
1696 		vput(nd.ni_vp);
1697 #ifdef MAC
1698 out2:
1699 #endif
1700 	NDFREE(&nd, NDF_ONLY_PNBUF);
1701 	vput(nd.ni_dvp);
1702 	vn_finished_write(mp);
1703 out:
1704 	if (segflg != UIO_SYSSPACE)
1705 		uma_zfree(namei_zone, syspath);
1706 	return (error);
1707 }
1708 
1709 /*
1710  * Delete a whiteout from the filesystem.
1711  */
1712 int
1713 sys_undelete(td, uap)
1714 	struct thread *td;
1715 	register struct undelete_args /* {
1716 		char *path;
1717 	} */ *uap;
1718 {
1719 	struct mount *mp;
1720 	struct nameidata nd;
1721 	int error;
1722 
1723 restart:
1724 	bwillwrite();
1725 	NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | AUDITVNODE1,
1726 	    UIO_USERSPACE, uap->path, td);
1727 	error = namei(&nd);
1728 	if (error != 0)
1729 		return (error);
1730 
1731 	if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) {
1732 		NDFREE(&nd, NDF_ONLY_PNBUF);
1733 		if (nd.ni_vp == nd.ni_dvp)
1734 			vrele(nd.ni_dvp);
1735 		else
1736 			vput(nd.ni_dvp);
1737 		if (nd.ni_vp)
1738 			vrele(nd.ni_vp);
1739 		return (EEXIST);
1740 	}
1741 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1742 		NDFREE(&nd, NDF_ONLY_PNBUF);
1743 		vput(nd.ni_dvp);
1744 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1745 			return (error);
1746 		goto restart;
1747 	}
1748 	error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE);
1749 	NDFREE(&nd, NDF_ONLY_PNBUF);
1750 	vput(nd.ni_dvp);
1751 	vn_finished_write(mp);
1752 	return (error);
1753 }
1754 
1755 /*
1756  * Delete a name from the filesystem.
1757  */
1758 #ifndef _SYS_SYSPROTO_H_
1759 struct unlink_args {
1760 	char	*path;
1761 };
1762 #endif
1763 int
1764 sys_unlink(td, uap)
1765 	struct thread *td;
1766 	struct unlink_args /* {
1767 		char *path;
1768 	} */ *uap;
1769 {
1770 
1771 	return (kern_unlink(td, uap->path, UIO_USERSPACE));
1772 }
1773 
1774 #ifndef _SYS_SYSPROTO_H_
1775 struct unlinkat_args {
1776 	int	fd;
1777 	char	*path;
1778 	int	flag;
1779 };
1780 #endif
1781 int
1782 sys_unlinkat(struct thread *td, struct unlinkat_args *uap)
1783 {
1784 	int flag = uap->flag;
1785 	int fd = uap->fd;
1786 	char *path = uap->path;
1787 
1788 	if (flag & ~AT_REMOVEDIR)
1789 		return (EINVAL);
1790 
1791 	if (flag & AT_REMOVEDIR)
1792 		return (kern_rmdirat(td, fd, path, UIO_USERSPACE));
1793 	else
1794 		return (kern_unlinkat(td, fd, path, UIO_USERSPACE, 0));
1795 }
1796 
1797 int
1798 kern_unlink(struct thread *td, char *path, enum uio_seg pathseg)
1799 {
1800 
1801 	return (kern_unlinkat(td, AT_FDCWD, path, pathseg, 0));
1802 }
1803 
1804 int
1805 kern_unlinkat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1806     ino_t oldinum)
1807 {
1808 	struct mount *mp;
1809 	struct vnode *vp;
1810 	struct nameidata nd;
1811 	struct stat sb;
1812 	cap_rights_t rights;
1813 	int error;
1814 
1815 restart:
1816 	bwillwrite();
1817 	NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1,
1818 	    pathseg, path, fd, cap_rights_init(&rights, CAP_UNLINKAT), td);
1819 	if ((error = namei(&nd)) != 0)
1820 		return (error == EINVAL ? EPERM : error);
1821 	vp = nd.ni_vp;
1822 	if (vp->v_type == VDIR && oldinum == 0) {
1823 		error = EPERM;		/* POSIX */
1824 	} else if (oldinum != 0 &&
1825 		  ((error = vn_stat(vp, &sb, td->td_ucred, NOCRED, td)) == 0) &&
1826 		  sb.st_ino != oldinum) {
1827 			error = EIDRM;	/* Identifier removed */
1828 	} else {
1829 		/*
1830 		 * The root of a mounted filesystem cannot be deleted.
1831 		 *
1832 		 * XXX: can this only be a VDIR case?
1833 		 */
1834 		if (vp->v_vflag & VV_ROOT)
1835 			error = EBUSY;
1836 	}
1837 	if (error == 0) {
1838 		if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1839 			NDFREE(&nd, NDF_ONLY_PNBUF);
1840 			vput(nd.ni_dvp);
1841 			if (vp == nd.ni_dvp)
1842 				vrele(vp);
1843 			else
1844 				vput(vp);
1845 			if ((error = vn_start_write(NULL, &mp,
1846 			    V_XSLEEP | PCATCH)) != 0)
1847 				return (error);
1848 			goto restart;
1849 		}
1850 #ifdef MAC
1851 		error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
1852 		    &nd.ni_cnd);
1853 		if (error != 0)
1854 			goto out;
1855 #endif
1856 		vfs_notify_upper(vp, VFS_NOTIFY_UPPER_UNLINK);
1857 		error = VOP_REMOVE(nd.ni_dvp, vp, &nd.ni_cnd);
1858 #ifdef MAC
1859 out:
1860 #endif
1861 		vn_finished_write(mp);
1862 	}
1863 	NDFREE(&nd, NDF_ONLY_PNBUF);
1864 	vput(nd.ni_dvp);
1865 	if (vp == nd.ni_dvp)
1866 		vrele(vp);
1867 	else
1868 		vput(vp);
1869 	return (error);
1870 }
1871 
1872 /*
1873  * Reposition read/write file offset.
1874  */
1875 #ifndef _SYS_SYSPROTO_H_
1876 struct lseek_args {
1877 	int	fd;
1878 	int	pad;
1879 	off_t	offset;
1880 	int	whence;
1881 };
1882 #endif
1883 int
1884 sys_lseek(td, uap)
1885 	struct thread *td;
1886 	register struct lseek_args /* {
1887 		int fd;
1888 		int pad;
1889 		off_t offset;
1890 		int whence;
1891 	} */ *uap;
1892 {
1893 	struct file *fp;
1894 	cap_rights_t rights;
1895 	int error;
1896 
1897 	AUDIT_ARG_FD(uap->fd);
1898 	error = fget(td, uap->fd, cap_rights_init(&rights, CAP_SEEK), &fp);
1899 	if (error != 0)
1900 		return (error);
1901 	error = (fp->f_ops->fo_flags & DFLAG_SEEKABLE) != 0 ?
1902 	    fo_seek(fp, uap->offset, uap->whence, td) : ESPIPE;
1903 	fdrop(fp, td);
1904 	return (error);
1905 }
1906 
1907 #if defined(COMPAT_43)
1908 /*
1909  * Reposition read/write file offset.
1910  */
1911 #ifndef _SYS_SYSPROTO_H_
1912 struct olseek_args {
1913 	int	fd;
1914 	long	offset;
1915 	int	whence;
1916 };
1917 #endif
1918 int
1919 olseek(td, uap)
1920 	struct thread *td;
1921 	register struct olseek_args /* {
1922 		int fd;
1923 		long offset;
1924 		int whence;
1925 	} */ *uap;
1926 {
1927 	struct lseek_args /* {
1928 		int fd;
1929 		int pad;
1930 		off_t offset;
1931 		int whence;
1932 	} */ nuap;
1933 
1934 	nuap.fd = uap->fd;
1935 	nuap.offset = uap->offset;
1936 	nuap.whence = uap->whence;
1937 	return (sys_lseek(td, &nuap));
1938 }
1939 #endif /* COMPAT_43 */
1940 
1941 /* Version with the 'pad' argument */
1942 int
1943 freebsd6_lseek(td, uap)
1944 	struct thread *td;
1945 	register struct freebsd6_lseek_args *uap;
1946 {
1947 	struct lseek_args ouap;
1948 
1949 	ouap.fd = uap->fd;
1950 	ouap.offset = uap->offset;
1951 	ouap.whence = uap->whence;
1952 	return (sys_lseek(td, &ouap));
1953 }
1954 
1955 /*
1956  * Check access permissions using passed credentials.
1957  */
1958 static int
1959 vn_access(vp, user_flags, cred, td)
1960 	struct vnode	*vp;
1961 	int		user_flags;
1962 	struct ucred	*cred;
1963 	struct thread	*td;
1964 {
1965 	accmode_t accmode;
1966 	int error;
1967 
1968 	/* Flags == 0 means only check for existence. */
1969 	error = 0;
1970 	if (user_flags) {
1971 		accmode = 0;
1972 		if (user_flags & R_OK)
1973 			accmode |= VREAD;
1974 		if (user_flags & W_OK)
1975 			accmode |= VWRITE;
1976 		if (user_flags & X_OK)
1977 			accmode |= VEXEC;
1978 #ifdef MAC
1979 		error = mac_vnode_check_access(cred, vp, accmode);
1980 		if (error != 0)
1981 			return (error);
1982 #endif
1983 		if ((accmode & VWRITE) == 0 || (error = vn_writechk(vp)) == 0)
1984 			error = VOP_ACCESS(vp, accmode, cred, td);
1985 	}
1986 	return (error);
1987 }
1988 
1989 /*
1990  * Check access permissions using "real" credentials.
1991  */
1992 #ifndef _SYS_SYSPROTO_H_
1993 struct access_args {
1994 	char	*path;
1995 	int	amode;
1996 };
1997 #endif
1998 int
1999 sys_access(td, uap)
2000 	struct thread *td;
2001 	register struct access_args /* {
2002 		char *path;
2003 		int amode;
2004 	} */ *uap;
2005 {
2006 
2007 	return (kern_access(td, uap->path, UIO_USERSPACE, uap->amode));
2008 }
2009 
2010 #ifndef _SYS_SYSPROTO_H_
2011 struct faccessat_args {
2012 	int	dirfd;
2013 	char	*path;
2014 	int	amode;
2015 	int	flag;
2016 }
2017 #endif
2018 int
2019 sys_faccessat(struct thread *td, struct faccessat_args *uap)
2020 {
2021 
2022 	if (uap->flag & ~AT_EACCESS)
2023 		return (EINVAL);
2024 	return (kern_accessat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
2025 	    uap->amode));
2026 }
2027 
2028 int
2029 kern_access(struct thread *td, char *path, enum uio_seg pathseg, int amode)
2030 {
2031 
2032 	return (kern_accessat(td, AT_FDCWD, path, pathseg, 0, amode));
2033 }
2034 
2035 int
2036 kern_accessat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2037     int flag, int amode)
2038 {
2039 	struct ucred *cred, *tmpcred;
2040 	struct vnode *vp;
2041 	struct nameidata nd;
2042 	cap_rights_t rights;
2043 	int error;
2044 
2045 	/*
2046 	 * Create and modify a temporary credential instead of one that
2047 	 * is potentially shared.
2048 	 */
2049 	if (!(flag & AT_EACCESS)) {
2050 		cred = td->td_ucred;
2051 		tmpcred = crdup(cred);
2052 		tmpcred->cr_uid = cred->cr_ruid;
2053 		tmpcred->cr_groups[0] = cred->cr_rgid;
2054 		td->td_ucred = tmpcred;
2055 	} else
2056 		cred = tmpcred = td->td_ucred;
2057 	AUDIT_ARG_VALUE(amode);
2058 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF |
2059 	    AUDITVNODE1, pathseg, path, fd, cap_rights_init(&rights, CAP_FSTAT),
2060 	    td);
2061 	if ((error = namei(&nd)) != 0)
2062 		goto out1;
2063 	vp = nd.ni_vp;
2064 
2065 	error = vn_access(vp, amode, tmpcred, td);
2066 	NDFREE(&nd, NDF_ONLY_PNBUF);
2067 	vput(vp);
2068 out1:
2069 	if (!(flag & AT_EACCESS)) {
2070 		td->td_ucred = cred;
2071 		crfree(tmpcred);
2072 	}
2073 	return (error);
2074 }
2075 
2076 /*
2077  * Check access permissions using "effective" credentials.
2078  */
2079 #ifndef _SYS_SYSPROTO_H_
2080 struct eaccess_args {
2081 	char	*path;
2082 	int	amode;
2083 };
2084 #endif
2085 int
2086 sys_eaccess(td, uap)
2087 	struct thread *td;
2088 	register struct eaccess_args /* {
2089 		char *path;
2090 		int amode;
2091 	} */ *uap;
2092 {
2093 
2094 	return (kern_eaccess(td, uap->path, UIO_USERSPACE, uap->amode));
2095 }
2096 
2097 int
2098 kern_eaccess(struct thread *td, char *path, enum uio_seg pathseg, int amode)
2099 {
2100 
2101 	return (kern_accessat(td, AT_FDCWD, path, pathseg, AT_EACCESS, amode));
2102 }
2103 
2104 #if defined(COMPAT_43)
2105 /*
2106  * Get file status; this version follows links.
2107  */
2108 #ifndef _SYS_SYSPROTO_H_
2109 struct ostat_args {
2110 	char	*path;
2111 	struct ostat *ub;
2112 };
2113 #endif
2114 int
2115 ostat(td, uap)
2116 	struct thread *td;
2117 	register struct ostat_args /* {
2118 		char *path;
2119 		struct ostat *ub;
2120 	} */ *uap;
2121 {
2122 	struct stat sb;
2123 	struct ostat osb;
2124 	int error;
2125 
2126 	error = kern_stat(td, uap->path, UIO_USERSPACE, &sb);
2127 	if (error != 0)
2128 		return (error);
2129 	cvtstat(&sb, &osb);
2130 	return (copyout(&osb, uap->ub, sizeof (osb)));
2131 }
2132 
2133 /*
2134  * Get file status; this version does not follow links.
2135  */
2136 #ifndef _SYS_SYSPROTO_H_
2137 struct olstat_args {
2138 	char	*path;
2139 	struct ostat *ub;
2140 };
2141 #endif
2142 int
2143 olstat(td, uap)
2144 	struct thread *td;
2145 	register struct olstat_args /* {
2146 		char *path;
2147 		struct ostat *ub;
2148 	} */ *uap;
2149 {
2150 	struct stat sb;
2151 	struct ostat osb;
2152 	int error;
2153 
2154 	error = kern_lstat(td, uap->path, UIO_USERSPACE, &sb);
2155 	if (error != 0)
2156 		return (error);
2157 	cvtstat(&sb, &osb);
2158 	return (copyout(&osb, uap->ub, sizeof (osb)));
2159 }
2160 
2161 /*
2162  * Convert from an old to a new stat structure.
2163  */
2164 void
2165 cvtstat(st, ost)
2166 	struct stat *st;
2167 	struct ostat *ost;
2168 {
2169 
2170 	ost->st_dev = st->st_dev;
2171 	ost->st_ino = st->st_ino;
2172 	ost->st_mode = st->st_mode;
2173 	ost->st_nlink = st->st_nlink;
2174 	ost->st_uid = st->st_uid;
2175 	ost->st_gid = st->st_gid;
2176 	ost->st_rdev = st->st_rdev;
2177 	if (st->st_size < (quad_t)1 << 32)
2178 		ost->st_size = st->st_size;
2179 	else
2180 		ost->st_size = -2;
2181 	ost->st_atim = st->st_atim;
2182 	ost->st_mtim = st->st_mtim;
2183 	ost->st_ctim = st->st_ctim;
2184 	ost->st_blksize = st->st_blksize;
2185 	ost->st_blocks = st->st_blocks;
2186 	ost->st_flags = st->st_flags;
2187 	ost->st_gen = st->st_gen;
2188 }
2189 #endif /* COMPAT_43 */
2190 
2191 /*
2192  * Get file status; this version follows links.
2193  */
2194 #ifndef _SYS_SYSPROTO_H_
2195 struct stat_args {
2196 	char	*path;
2197 	struct stat *ub;
2198 };
2199 #endif
2200 int
2201 sys_stat(td, uap)
2202 	struct thread *td;
2203 	register struct stat_args /* {
2204 		char *path;
2205 		struct stat *ub;
2206 	} */ *uap;
2207 {
2208 	struct stat sb;
2209 	int error;
2210 
2211 	error = kern_stat(td, uap->path, UIO_USERSPACE, &sb);
2212 	if (error == 0)
2213 		error = copyout(&sb, uap->ub, sizeof (sb));
2214 	return (error);
2215 }
2216 
2217 #ifndef _SYS_SYSPROTO_H_
2218 struct fstatat_args {
2219 	int	fd;
2220 	char	*path;
2221 	struct stat	*buf;
2222 	int	flag;
2223 }
2224 #endif
2225 int
2226 sys_fstatat(struct thread *td, struct fstatat_args *uap)
2227 {
2228 	struct stat sb;
2229 	int error;
2230 
2231 	error = kern_statat(td, uap->flag, uap->fd, uap->path,
2232 	    UIO_USERSPACE, &sb);
2233 	if (error == 0)
2234 		error = copyout(&sb, uap->buf, sizeof (sb));
2235 	return (error);
2236 }
2237 
2238 int
2239 kern_stat(struct thread *td, char *path, enum uio_seg pathseg, struct stat *sbp)
2240 {
2241 
2242 	return (kern_statat(td, 0, AT_FDCWD, path, pathseg, sbp));
2243 }
2244 
2245 int
2246 kern_statat(struct thread *td, int flag, int fd, char *path,
2247     enum uio_seg pathseg, struct stat *sbp)
2248 {
2249 
2250 	return (kern_statat_vnhook(td, flag, fd, path, pathseg, sbp, NULL));
2251 }
2252 
2253 int
2254 kern_statat_vnhook(struct thread *td, int flag, int fd, char *path,
2255     enum uio_seg pathseg, struct stat *sbp,
2256     void (*hook)(struct vnode *vp, struct stat *sbp))
2257 {
2258 	struct nameidata nd;
2259 	struct stat sb;
2260 	cap_rights_t rights;
2261 	int error;
2262 
2263 	if (flag & ~AT_SYMLINK_NOFOLLOW)
2264 		return (EINVAL);
2265 
2266 	NDINIT_ATRIGHTS(&nd, LOOKUP, ((flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW :
2267 	    FOLLOW) | LOCKSHARED | LOCKLEAF | AUDITVNODE1, pathseg, path, fd,
2268 	    cap_rights_init(&rights, CAP_FSTAT), td);
2269 
2270 	if ((error = namei(&nd)) != 0)
2271 		return (error);
2272 	error = vn_stat(nd.ni_vp, &sb, td->td_ucred, NOCRED, td);
2273 	if (error == 0) {
2274 		SDT_PROBE(vfs, , stat, mode, path, sb.st_mode, 0, 0, 0);
2275 		if (S_ISREG(sb.st_mode))
2276 			SDT_PROBE(vfs, , stat, reg, path, pathseg, 0, 0, 0);
2277 		if (__predict_false(hook != NULL))
2278 			hook(nd.ni_vp, &sb);
2279 	}
2280 	NDFREE(&nd, NDF_ONLY_PNBUF);
2281 	vput(nd.ni_vp);
2282 	if (error != 0)
2283 		return (error);
2284 	*sbp = sb;
2285 #ifdef KTRACE
2286 	if (KTRPOINT(td, KTR_STRUCT))
2287 		ktrstat(&sb);
2288 #endif
2289 	return (0);
2290 }
2291 
2292 /*
2293  * Get file status; this version does not follow links.
2294  */
2295 #ifndef _SYS_SYSPROTO_H_
2296 struct lstat_args {
2297 	char	*path;
2298 	struct stat *ub;
2299 };
2300 #endif
2301 int
2302 sys_lstat(td, uap)
2303 	struct thread *td;
2304 	register struct lstat_args /* {
2305 		char *path;
2306 		struct stat *ub;
2307 	} */ *uap;
2308 {
2309 	struct stat sb;
2310 	int error;
2311 
2312 	error = kern_lstat(td, uap->path, UIO_USERSPACE, &sb);
2313 	if (error == 0)
2314 		error = copyout(&sb, uap->ub, sizeof (sb));
2315 	return (error);
2316 }
2317 
2318 int
2319 kern_lstat(struct thread *td, char *path, enum uio_seg pathseg, struct stat *sbp)
2320 {
2321 
2322 	return (kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, path, pathseg,
2323 	    sbp));
2324 }
2325 
2326 /*
2327  * Implementation of the NetBSD [l]stat() functions.
2328  */
2329 void
2330 cvtnstat(sb, nsb)
2331 	struct stat *sb;
2332 	struct nstat *nsb;
2333 {
2334 
2335 	bzero(nsb, sizeof *nsb);
2336 	nsb->st_dev = sb->st_dev;
2337 	nsb->st_ino = sb->st_ino;
2338 	nsb->st_mode = sb->st_mode;
2339 	nsb->st_nlink = sb->st_nlink;
2340 	nsb->st_uid = sb->st_uid;
2341 	nsb->st_gid = sb->st_gid;
2342 	nsb->st_rdev = sb->st_rdev;
2343 	nsb->st_atim = sb->st_atim;
2344 	nsb->st_mtim = sb->st_mtim;
2345 	nsb->st_ctim = sb->st_ctim;
2346 	nsb->st_size = sb->st_size;
2347 	nsb->st_blocks = sb->st_blocks;
2348 	nsb->st_blksize = sb->st_blksize;
2349 	nsb->st_flags = sb->st_flags;
2350 	nsb->st_gen = sb->st_gen;
2351 	nsb->st_birthtim = sb->st_birthtim;
2352 }
2353 
2354 #ifndef _SYS_SYSPROTO_H_
2355 struct nstat_args {
2356 	char	*path;
2357 	struct nstat *ub;
2358 };
2359 #endif
2360 int
2361 sys_nstat(td, uap)
2362 	struct thread *td;
2363 	register struct nstat_args /* {
2364 		char *path;
2365 		struct nstat *ub;
2366 	} */ *uap;
2367 {
2368 	struct stat sb;
2369 	struct nstat nsb;
2370 	int error;
2371 
2372 	error = kern_stat(td, uap->path, UIO_USERSPACE, &sb);
2373 	if (error != 0)
2374 		return (error);
2375 	cvtnstat(&sb, &nsb);
2376 	return (copyout(&nsb, uap->ub, sizeof (nsb)));
2377 }
2378 
2379 /*
2380  * NetBSD lstat.  Get file status; this version does not follow links.
2381  */
2382 #ifndef _SYS_SYSPROTO_H_
2383 struct lstat_args {
2384 	char	*path;
2385 	struct stat *ub;
2386 };
2387 #endif
2388 int
2389 sys_nlstat(td, uap)
2390 	struct thread *td;
2391 	register struct nlstat_args /* {
2392 		char *path;
2393 		struct nstat *ub;
2394 	} */ *uap;
2395 {
2396 	struct stat sb;
2397 	struct nstat nsb;
2398 	int error;
2399 
2400 	error = kern_lstat(td, uap->path, UIO_USERSPACE, &sb);
2401 	if (error != 0)
2402 		return (error);
2403 	cvtnstat(&sb, &nsb);
2404 	return (copyout(&nsb, uap->ub, sizeof (nsb)));
2405 }
2406 
2407 /*
2408  * Get configurable pathname variables.
2409  */
2410 #ifndef _SYS_SYSPROTO_H_
2411 struct pathconf_args {
2412 	char	*path;
2413 	int	name;
2414 };
2415 #endif
2416 int
2417 sys_pathconf(td, uap)
2418 	struct thread *td;
2419 	register struct pathconf_args /* {
2420 		char *path;
2421 		int name;
2422 	} */ *uap;
2423 {
2424 
2425 	return (kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name, FOLLOW));
2426 }
2427 
2428 #ifndef _SYS_SYSPROTO_H_
2429 struct lpathconf_args {
2430 	char	*path;
2431 	int	name;
2432 };
2433 #endif
2434 int
2435 sys_lpathconf(td, uap)
2436 	struct thread *td;
2437 	register struct lpathconf_args /* {
2438 		char *path;
2439 		int name;
2440 	} */ *uap;
2441 {
2442 
2443 	return (kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name,
2444 	    NOFOLLOW));
2445 }
2446 
2447 int
2448 kern_pathconf(struct thread *td, char *path, enum uio_seg pathseg, int name,
2449     u_long flags)
2450 {
2451 	struct nameidata nd;
2452 	int error;
2453 
2454 	NDINIT(&nd, LOOKUP, LOCKSHARED | LOCKLEAF | AUDITVNODE1 | flags,
2455 	    pathseg, path, td);
2456 	if ((error = namei(&nd)) != 0)
2457 		return (error);
2458 	NDFREE(&nd, NDF_ONLY_PNBUF);
2459 
2460 	/* If asynchronous I/O is available, it works for all files. */
2461 	if (name == _PC_ASYNC_IO)
2462 		td->td_retval[0] = async_io_version;
2463 	else
2464 		error = VOP_PATHCONF(nd.ni_vp, name, td->td_retval);
2465 	vput(nd.ni_vp);
2466 	return (error);
2467 }
2468 
2469 /*
2470  * Return target name of a symbolic link.
2471  */
2472 #ifndef _SYS_SYSPROTO_H_
2473 struct readlink_args {
2474 	char	*path;
2475 	char	*buf;
2476 	size_t	count;
2477 };
2478 #endif
2479 int
2480 sys_readlink(td, uap)
2481 	struct thread *td;
2482 	register struct readlink_args /* {
2483 		char *path;
2484 		char *buf;
2485 		size_t count;
2486 	} */ *uap;
2487 {
2488 
2489 	return (kern_readlink(td, uap->path, UIO_USERSPACE, uap->buf,
2490 	    UIO_USERSPACE, uap->count));
2491 }
2492 #ifndef _SYS_SYSPROTO_H_
2493 struct readlinkat_args {
2494 	int	fd;
2495 	char	*path;
2496 	char	*buf;
2497 	size_t	bufsize;
2498 };
2499 #endif
2500 int
2501 sys_readlinkat(struct thread *td, struct readlinkat_args *uap)
2502 {
2503 
2504 	return (kern_readlinkat(td, uap->fd, uap->path, UIO_USERSPACE,
2505 	    uap->buf, UIO_USERSPACE, uap->bufsize));
2506 }
2507 
2508 int
2509 kern_readlink(struct thread *td, char *path, enum uio_seg pathseg, char *buf,
2510     enum uio_seg bufseg, size_t count)
2511 {
2512 
2513 	return (kern_readlinkat(td, AT_FDCWD, path, pathseg, buf, bufseg,
2514 	    count));
2515 }
2516 
2517 int
2518 kern_readlinkat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2519     char *buf, enum uio_seg bufseg, size_t count)
2520 {
2521 	struct vnode *vp;
2522 	struct iovec aiov;
2523 	struct uio auio;
2524 	struct nameidata nd;
2525 	int error;
2526 
2527 	if (count > IOSIZE_MAX)
2528 		return (EINVAL);
2529 
2530 	NDINIT_AT(&nd, LOOKUP, NOFOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
2531 	    pathseg, path, fd, td);
2532 
2533 	if ((error = namei(&nd)) != 0)
2534 		return (error);
2535 	NDFREE(&nd, NDF_ONLY_PNBUF);
2536 	vp = nd.ni_vp;
2537 #ifdef MAC
2538 	error = mac_vnode_check_readlink(td->td_ucred, vp);
2539 	if (error != 0) {
2540 		vput(vp);
2541 		return (error);
2542 	}
2543 #endif
2544 	if (vp->v_type != VLNK)
2545 		error = EINVAL;
2546 	else {
2547 		aiov.iov_base = buf;
2548 		aiov.iov_len = count;
2549 		auio.uio_iov = &aiov;
2550 		auio.uio_iovcnt = 1;
2551 		auio.uio_offset = 0;
2552 		auio.uio_rw = UIO_READ;
2553 		auio.uio_segflg = bufseg;
2554 		auio.uio_td = td;
2555 		auio.uio_resid = count;
2556 		error = VOP_READLINK(vp, &auio, td->td_ucred);
2557 	}
2558 	vput(vp);
2559 	td->td_retval[0] = count - auio.uio_resid;
2560 	return (error);
2561 }
2562 
2563 /*
2564  * Common implementation code for chflags() and fchflags().
2565  */
2566 static int
2567 setfflags(td, vp, flags)
2568 	struct thread *td;
2569 	struct vnode *vp;
2570 	u_long flags;
2571 {
2572 	struct mount *mp;
2573 	struct vattr vattr;
2574 	int error;
2575 
2576 	/* We can't support the value matching VNOVAL. */
2577 	if (flags == VNOVAL)
2578 		return (EOPNOTSUPP);
2579 
2580 	/*
2581 	 * Prevent non-root users from setting flags on devices.  When
2582 	 * a device is reused, users can retain ownership of the device
2583 	 * if they are allowed to set flags and programs assume that
2584 	 * chown can't fail when done as root.
2585 	 */
2586 	if (vp->v_type == VCHR || vp->v_type == VBLK) {
2587 		error = priv_check(td, PRIV_VFS_CHFLAGS_DEV);
2588 		if (error != 0)
2589 			return (error);
2590 	}
2591 
2592 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2593 		return (error);
2594 	VATTR_NULL(&vattr);
2595 	vattr.va_flags = flags;
2596 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2597 #ifdef MAC
2598 	error = mac_vnode_check_setflags(td->td_ucred, vp, vattr.va_flags);
2599 	if (error == 0)
2600 #endif
2601 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
2602 	VOP_UNLOCK(vp, 0);
2603 	vn_finished_write(mp);
2604 	return (error);
2605 }
2606 
2607 /*
2608  * Change flags of a file given a path name.
2609  */
2610 #ifndef _SYS_SYSPROTO_H_
2611 struct chflags_args {
2612 	const char *path;
2613 	u_long	flags;
2614 };
2615 #endif
2616 int
2617 sys_chflags(td, uap)
2618 	struct thread *td;
2619 	register struct chflags_args /* {
2620 		const char *path;
2621 		u_long flags;
2622 	} */ *uap;
2623 {
2624 
2625 	return (kern_chflags(td, uap->path, UIO_USERSPACE, uap->flags));
2626 }
2627 
2628 #ifndef _SYS_SYSPROTO_H_
2629 struct chflagsat_args {
2630 	int	fd;
2631 	const char *path;
2632 	u_long	flags;
2633 	int	atflag;
2634 }
2635 #endif
2636 int
2637 sys_chflagsat(struct thread *td, struct chflagsat_args *uap)
2638 {
2639 	int fd = uap->fd;
2640 	const char *path = uap->path;
2641 	u_long flags = uap->flags;
2642 	int atflag = uap->atflag;
2643 
2644 	if (atflag & ~AT_SYMLINK_NOFOLLOW)
2645 		return (EINVAL);
2646 
2647 	return (kern_chflagsat(td, fd, path, UIO_USERSPACE, flags, atflag));
2648 }
2649 
2650 static int
2651 kern_chflags(struct thread *td, const char *path, enum uio_seg pathseg,
2652     u_long flags)
2653 {
2654 
2655 	return (kern_chflagsat(td, AT_FDCWD, path, pathseg, flags, 0));
2656 }
2657 
2658 /*
2659  * Same as chflags() but doesn't follow symlinks.
2660  */
2661 int
2662 sys_lchflags(td, uap)
2663 	struct thread *td;
2664 	register struct lchflags_args /* {
2665 		const char *path;
2666 		u_long flags;
2667 	} */ *uap;
2668 {
2669 
2670 	return (kern_chflagsat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2671 	    uap->flags, AT_SYMLINK_NOFOLLOW));
2672 }
2673 
2674 static int
2675 kern_chflagsat(struct thread *td, int fd, const char *path,
2676     enum uio_seg pathseg, u_long flags, int atflag)
2677 {
2678 	struct nameidata nd;
2679 	cap_rights_t rights;
2680 	int error, follow;
2681 
2682 	AUDIT_ARG_FFLAGS(flags);
2683 	follow = (atflag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
2684 	NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, pathseg, path, fd,
2685 	    cap_rights_init(&rights, CAP_FCHFLAGS), td);
2686 	if ((error = namei(&nd)) != 0)
2687 		return (error);
2688 	NDFREE(&nd, NDF_ONLY_PNBUF);
2689 	error = setfflags(td, nd.ni_vp, flags);
2690 	vrele(nd.ni_vp);
2691 	return (error);
2692 }
2693 
2694 /*
2695  * Change flags of a file given a file descriptor.
2696  */
2697 #ifndef _SYS_SYSPROTO_H_
2698 struct fchflags_args {
2699 	int	fd;
2700 	u_long	flags;
2701 };
2702 #endif
2703 int
2704 sys_fchflags(td, uap)
2705 	struct thread *td;
2706 	register struct fchflags_args /* {
2707 		int fd;
2708 		u_long flags;
2709 	} */ *uap;
2710 {
2711 	struct file *fp;
2712 	cap_rights_t rights;
2713 	int error;
2714 
2715 	AUDIT_ARG_FD(uap->fd);
2716 	AUDIT_ARG_FFLAGS(uap->flags);
2717 	error = getvnode(td->td_proc->p_fd, uap->fd,
2718 	    cap_rights_init(&rights, CAP_FCHFLAGS), &fp);
2719 	if (error != 0)
2720 		return (error);
2721 #ifdef AUDIT
2722 	vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
2723 	AUDIT_ARG_VNODE1(fp->f_vnode);
2724 	VOP_UNLOCK(fp->f_vnode, 0);
2725 #endif
2726 	error = setfflags(td, fp->f_vnode, uap->flags);
2727 	fdrop(fp, td);
2728 	return (error);
2729 }
2730 
2731 /*
2732  * Common implementation code for chmod(), lchmod() and fchmod().
2733  */
2734 int
2735 setfmode(td, cred, vp, mode)
2736 	struct thread *td;
2737 	struct ucred *cred;
2738 	struct vnode *vp;
2739 	int mode;
2740 {
2741 	struct mount *mp;
2742 	struct vattr vattr;
2743 	int error;
2744 
2745 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2746 		return (error);
2747 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2748 	VATTR_NULL(&vattr);
2749 	vattr.va_mode = mode & ALLPERMS;
2750 #ifdef MAC
2751 	error = mac_vnode_check_setmode(cred, vp, vattr.va_mode);
2752 	if (error == 0)
2753 #endif
2754 		error = VOP_SETATTR(vp, &vattr, cred);
2755 	VOP_UNLOCK(vp, 0);
2756 	vn_finished_write(mp);
2757 	return (error);
2758 }
2759 
2760 /*
2761  * Change mode of a file given path name.
2762  */
2763 #ifndef _SYS_SYSPROTO_H_
2764 struct chmod_args {
2765 	char	*path;
2766 	int	mode;
2767 };
2768 #endif
2769 int
2770 sys_chmod(td, uap)
2771 	struct thread *td;
2772 	register struct chmod_args /* {
2773 		char *path;
2774 		int mode;
2775 	} */ *uap;
2776 {
2777 
2778 	return (kern_chmod(td, uap->path, UIO_USERSPACE, uap->mode));
2779 }
2780 
2781 #ifndef _SYS_SYSPROTO_H_
2782 struct fchmodat_args {
2783 	int	dirfd;
2784 	char	*path;
2785 	mode_t	mode;
2786 	int	flag;
2787 }
2788 #endif
2789 int
2790 sys_fchmodat(struct thread *td, struct fchmodat_args *uap)
2791 {
2792 	int flag = uap->flag;
2793 	int fd = uap->fd;
2794 	char *path = uap->path;
2795 	mode_t mode = uap->mode;
2796 
2797 	if (flag & ~AT_SYMLINK_NOFOLLOW)
2798 		return (EINVAL);
2799 
2800 	return (kern_fchmodat(td, fd, path, UIO_USERSPACE, mode, flag));
2801 }
2802 
2803 int
2804 kern_chmod(struct thread *td, char *path, enum uio_seg pathseg, int mode)
2805 {
2806 
2807 	return (kern_fchmodat(td, AT_FDCWD, path, pathseg, mode, 0));
2808 }
2809 
2810 /*
2811  * Change mode of a file given path name (don't follow links.)
2812  */
2813 #ifndef _SYS_SYSPROTO_H_
2814 struct lchmod_args {
2815 	char	*path;
2816 	int	mode;
2817 };
2818 #endif
2819 int
2820 sys_lchmod(td, uap)
2821 	struct thread *td;
2822 	register struct lchmod_args /* {
2823 		char *path;
2824 		int mode;
2825 	} */ *uap;
2826 {
2827 
2828 	return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2829 	    uap->mode, AT_SYMLINK_NOFOLLOW));
2830 }
2831 
2832 int
2833 kern_fchmodat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2834     mode_t mode, int flag)
2835 {
2836 	struct nameidata nd;
2837 	cap_rights_t rights;
2838 	int error, follow;
2839 
2840 	AUDIT_ARG_MODE(mode);
2841 	follow = (flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
2842 	NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, pathseg, path, fd,
2843 	    cap_rights_init(&rights, CAP_FCHMOD), td);
2844 	if ((error = namei(&nd)) != 0)
2845 		return (error);
2846 	NDFREE(&nd, NDF_ONLY_PNBUF);
2847 	error = setfmode(td, td->td_ucred, nd.ni_vp, mode);
2848 	vrele(nd.ni_vp);
2849 	return (error);
2850 }
2851 
2852 /*
2853  * Change mode of a file given a file descriptor.
2854  */
2855 #ifndef _SYS_SYSPROTO_H_
2856 struct fchmod_args {
2857 	int	fd;
2858 	int	mode;
2859 };
2860 #endif
2861 int
2862 sys_fchmod(struct thread *td, struct fchmod_args *uap)
2863 {
2864 	struct file *fp;
2865 	cap_rights_t rights;
2866 	int error;
2867 
2868 	AUDIT_ARG_FD(uap->fd);
2869 	AUDIT_ARG_MODE(uap->mode);
2870 
2871 	error = fget(td, uap->fd, cap_rights_init(&rights, CAP_FCHMOD), &fp);
2872 	if (error != 0)
2873 		return (error);
2874 	error = fo_chmod(fp, uap->mode, td->td_ucred, td);
2875 	fdrop(fp, td);
2876 	return (error);
2877 }
2878 
2879 /*
2880  * Common implementation for chown(), lchown(), and fchown()
2881  */
2882 int
2883 setfown(td, cred, vp, uid, gid)
2884 	struct thread *td;
2885 	struct ucred *cred;
2886 	struct vnode *vp;
2887 	uid_t uid;
2888 	gid_t gid;
2889 {
2890 	struct mount *mp;
2891 	struct vattr vattr;
2892 	int error;
2893 
2894 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2895 		return (error);
2896 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2897 	VATTR_NULL(&vattr);
2898 	vattr.va_uid = uid;
2899 	vattr.va_gid = gid;
2900 #ifdef MAC
2901 	error = mac_vnode_check_setowner(cred, vp, vattr.va_uid,
2902 	    vattr.va_gid);
2903 	if (error == 0)
2904 #endif
2905 		error = VOP_SETATTR(vp, &vattr, cred);
2906 	VOP_UNLOCK(vp, 0);
2907 	vn_finished_write(mp);
2908 	return (error);
2909 }
2910 
2911 /*
2912  * Set ownership given a path name.
2913  */
2914 #ifndef _SYS_SYSPROTO_H_
2915 struct chown_args {
2916 	char	*path;
2917 	int	uid;
2918 	int	gid;
2919 };
2920 #endif
2921 int
2922 sys_chown(td, uap)
2923 	struct thread *td;
2924 	register struct chown_args /* {
2925 		char *path;
2926 		int uid;
2927 		int gid;
2928 	} */ *uap;
2929 {
2930 
2931 	return (kern_chown(td, uap->path, UIO_USERSPACE, uap->uid, uap->gid));
2932 }
2933 
2934 #ifndef _SYS_SYSPROTO_H_
2935 struct fchownat_args {
2936 	int fd;
2937 	const char * path;
2938 	uid_t uid;
2939 	gid_t gid;
2940 	int flag;
2941 };
2942 #endif
2943 int
2944 sys_fchownat(struct thread *td, struct fchownat_args *uap)
2945 {
2946 	int flag;
2947 
2948 	flag = uap->flag;
2949 	if (flag & ~AT_SYMLINK_NOFOLLOW)
2950 		return (EINVAL);
2951 
2952 	return (kern_fchownat(td, uap->fd, uap->path, UIO_USERSPACE, uap->uid,
2953 	    uap->gid, uap->flag));
2954 }
2955 
2956 int
2957 kern_chown(struct thread *td, char *path, enum uio_seg pathseg, int uid,
2958     int gid)
2959 {
2960 
2961 	return (kern_fchownat(td, AT_FDCWD, path, pathseg, uid, gid, 0));
2962 }
2963 
2964 int
2965 kern_fchownat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2966     int uid, int gid, int flag)
2967 {
2968 	struct nameidata nd;
2969 	cap_rights_t rights;
2970 	int error, follow;
2971 
2972 	AUDIT_ARG_OWNER(uid, gid);
2973 	follow = (flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
2974 	NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, pathseg, path, fd,
2975 	    cap_rights_init(&rights, CAP_FCHOWN), td);
2976 
2977 	if ((error = namei(&nd)) != 0)
2978 		return (error);
2979 	NDFREE(&nd, NDF_ONLY_PNBUF);
2980 	error = setfown(td, td->td_ucred, nd.ni_vp, uid, gid);
2981 	vrele(nd.ni_vp);
2982 	return (error);
2983 }
2984 
2985 /*
2986  * Set ownership given a path name, do not cross symlinks.
2987  */
2988 #ifndef _SYS_SYSPROTO_H_
2989 struct lchown_args {
2990 	char	*path;
2991 	int	uid;
2992 	int	gid;
2993 };
2994 #endif
2995 int
2996 sys_lchown(td, uap)
2997 	struct thread *td;
2998 	register struct lchown_args /* {
2999 		char *path;
3000 		int uid;
3001 		int gid;
3002 	} */ *uap;
3003 {
3004 
3005 	return (kern_lchown(td, uap->path, UIO_USERSPACE, uap->uid, uap->gid));
3006 }
3007 
3008 int
3009 kern_lchown(struct thread *td, char *path, enum uio_seg pathseg, int uid,
3010     int gid)
3011 {
3012 
3013 	return (kern_fchownat(td, AT_FDCWD, path, pathseg, uid, gid,
3014 	    AT_SYMLINK_NOFOLLOW));
3015 }
3016 
3017 /*
3018  * Set ownership given a file descriptor.
3019  */
3020 #ifndef _SYS_SYSPROTO_H_
3021 struct fchown_args {
3022 	int	fd;
3023 	int	uid;
3024 	int	gid;
3025 };
3026 #endif
3027 int
3028 sys_fchown(td, uap)
3029 	struct thread *td;
3030 	register struct fchown_args /* {
3031 		int fd;
3032 		int uid;
3033 		int gid;
3034 	} */ *uap;
3035 {
3036 	struct file *fp;
3037 	cap_rights_t rights;
3038 	int error;
3039 
3040 	AUDIT_ARG_FD(uap->fd);
3041 	AUDIT_ARG_OWNER(uap->uid, uap->gid);
3042 	error = fget(td, uap->fd, cap_rights_init(&rights, CAP_FCHOWN), &fp);
3043 	if (error != 0)
3044 		return (error);
3045 	error = fo_chown(fp, uap->uid, uap->gid, td->td_ucred, td);
3046 	fdrop(fp, td);
3047 	return (error);
3048 }
3049 
3050 /*
3051  * Common implementation code for utimes(), lutimes(), and futimes().
3052  */
3053 static int
3054 getutimes(usrtvp, tvpseg, tsp)
3055 	const struct timeval *usrtvp;
3056 	enum uio_seg tvpseg;
3057 	struct timespec *tsp;
3058 {
3059 	struct timeval tv[2];
3060 	const struct timeval *tvp;
3061 	int error;
3062 
3063 	if (usrtvp == NULL) {
3064 		vfs_timestamp(&tsp[0]);
3065 		tsp[1] = tsp[0];
3066 	} else {
3067 		if (tvpseg == UIO_SYSSPACE) {
3068 			tvp = usrtvp;
3069 		} else {
3070 			if ((error = copyin(usrtvp, tv, sizeof(tv))) != 0)
3071 				return (error);
3072 			tvp = tv;
3073 		}
3074 
3075 		if (tvp[0].tv_usec < 0 || tvp[0].tv_usec >= 1000000 ||
3076 		    tvp[1].tv_usec < 0 || tvp[1].tv_usec >= 1000000)
3077 			return (EINVAL);
3078 		TIMEVAL_TO_TIMESPEC(&tvp[0], &tsp[0]);
3079 		TIMEVAL_TO_TIMESPEC(&tvp[1], &tsp[1]);
3080 	}
3081 	return (0);
3082 }
3083 
3084 /*
3085  * Common implementation code for utimes(), lutimes(), and futimes().
3086  */
3087 static int
3088 setutimes(td, vp, ts, numtimes, nullflag)
3089 	struct thread *td;
3090 	struct vnode *vp;
3091 	const struct timespec *ts;
3092 	int numtimes;
3093 	int nullflag;
3094 {
3095 	struct mount *mp;
3096 	struct vattr vattr;
3097 	int error, setbirthtime;
3098 
3099 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
3100 		return (error);
3101 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3102 	setbirthtime = 0;
3103 	if (numtimes < 3 && !VOP_GETATTR(vp, &vattr, td->td_ucred) &&
3104 	    timespeccmp(&ts[1], &vattr.va_birthtime, < ))
3105 		setbirthtime = 1;
3106 	VATTR_NULL(&vattr);
3107 	vattr.va_atime = ts[0];
3108 	vattr.va_mtime = ts[1];
3109 	if (setbirthtime)
3110 		vattr.va_birthtime = ts[1];
3111 	if (numtimes > 2)
3112 		vattr.va_birthtime = ts[2];
3113 	if (nullflag)
3114 		vattr.va_vaflags |= VA_UTIMES_NULL;
3115 #ifdef MAC
3116 	error = mac_vnode_check_setutimes(td->td_ucred, vp, vattr.va_atime,
3117 	    vattr.va_mtime);
3118 #endif
3119 	if (error == 0)
3120 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3121 	VOP_UNLOCK(vp, 0);
3122 	vn_finished_write(mp);
3123 	return (error);
3124 }
3125 
3126 /*
3127  * Set the access and modification times of a file.
3128  */
3129 #ifndef _SYS_SYSPROTO_H_
3130 struct utimes_args {
3131 	char	*path;
3132 	struct	timeval *tptr;
3133 };
3134 #endif
3135 int
3136 sys_utimes(td, uap)
3137 	struct thread *td;
3138 	register struct utimes_args /* {
3139 		char *path;
3140 		struct timeval *tptr;
3141 	} */ *uap;
3142 {
3143 
3144 	return (kern_utimes(td, uap->path, UIO_USERSPACE, uap->tptr,
3145 	    UIO_USERSPACE));
3146 }
3147 
3148 #ifndef _SYS_SYSPROTO_H_
3149 struct futimesat_args {
3150 	int fd;
3151 	const char * path;
3152 	const struct timeval * times;
3153 };
3154 #endif
3155 int
3156 sys_futimesat(struct thread *td, struct futimesat_args *uap)
3157 {
3158 
3159 	return (kern_utimesat(td, uap->fd, uap->path, UIO_USERSPACE,
3160 	    uap->times, UIO_USERSPACE));
3161 }
3162 
3163 int
3164 kern_utimes(struct thread *td, char *path, enum uio_seg pathseg,
3165     struct timeval *tptr, enum uio_seg tptrseg)
3166 {
3167 
3168 	return (kern_utimesat(td, AT_FDCWD, path, pathseg, tptr, tptrseg));
3169 }
3170 
3171 int
3172 kern_utimesat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
3173     struct timeval *tptr, enum uio_seg tptrseg)
3174 {
3175 	struct nameidata nd;
3176 	struct timespec ts[2];
3177 	cap_rights_t rights;
3178 	int error;
3179 
3180 	if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3181 		return (error);
3182 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, fd,
3183 	    cap_rights_init(&rights, CAP_FUTIMES), td);
3184 
3185 	if ((error = namei(&nd)) != 0)
3186 		return (error);
3187 	NDFREE(&nd, NDF_ONLY_PNBUF);
3188 	error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3189 	vrele(nd.ni_vp);
3190 	return (error);
3191 }
3192 
3193 /*
3194  * Set the access and modification times of a file.
3195  */
3196 #ifndef _SYS_SYSPROTO_H_
3197 struct lutimes_args {
3198 	char	*path;
3199 	struct	timeval *tptr;
3200 };
3201 #endif
3202 int
3203 sys_lutimes(td, uap)
3204 	struct thread *td;
3205 	register struct lutimes_args /* {
3206 		char *path;
3207 		struct timeval *tptr;
3208 	} */ *uap;
3209 {
3210 
3211 	return (kern_lutimes(td, uap->path, UIO_USERSPACE, uap->tptr,
3212 	    UIO_USERSPACE));
3213 }
3214 
3215 int
3216 kern_lutimes(struct thread *td, char *path, enum uio_seg pathseg,
3217     struct timeval *tptr, enum uio_seg tptrseg)
3218 {
3219 	struct timespec ts[2];
3220 	struct nameidata nd;
3221 	int error;
3222 
3223 	if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3224 		return (error);
3225 	NDINIT(&nd, LOOKUP, NOFOLLOW | AUDITVNODE1, pathseg, path, td);
3226 	if ((error = namei(&nd)) != 0)
3227 		return (error);
3228 	NDFREE(&nd, NDF_ONLY_PNBUF);
3229 	error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3230 	vrele(nd.ni_vp);
3231 	return (error);
3232 }
3233 
3234 /*
3235  * Set the access and modification times of a file.
3236  */
3237 #ifndef _SYS_SYSPROTO_H_
3238 struct futimes_args {
3239 	int	fd;
3240 	struct	timeval *tptr;
3241 };
3242 #endif
3243 int
3244 sys_futimes(td, uap)
3245 	struct thread *td;
3246 	register struct futimes_args /* {
3247 		int  fd;
3248 		struct timeval *tptr;
3249 	} */ *uap;
3250 {
3251 
3252 	return (kern_futimes(td, uap->fd, uap->tptr, UIO_USERSPACE));
3253 }
3254 
3255 int
3256 kern_futimes(struct thread *td, int fd, struct timeval *tptr,
3257     enum uio_seg tptrseg)
3258 {
3259 	struct timespec ts[2];
3260 	struct file *fp;
3261 	cap_rights_t rights;
3262 	int error;
3263 
3264 	AUDIT_ARG_FD(fd);
3265 	error = getutimes(tptr, tptrseg, ts);
3266 	if (error != 0)
3267 		return (error);
3268 	error = getvnode(td->td_proc->p_fd, fd,
3269 	    cap_rights_init(&rights, CAP_FUTIMES), &fp);
3270 	if (error != 0)
3271 		return (error);
3272 #ifdef AUDIT
3273 	vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
3274 	AUDIT_ARG_VNODE1(fp->f_vnode);
3275 	VOP_UNLOCK(fp->f_vnode, 0);
3276 #endif
3277 	error = setutimes(td, fp->f_vnode, ts, 2, tptr == NULL);
3278 	fdrop(fp, td);
3279 	return (error);
3280 }
3281 
3282 /*
3283  * Truncate a file given its path name.
3284  */
3285 #ifndef _SYS_SYSPROTO_H_
3286 struct truncate_args {
3287 	char	*path;
3288 	int	pad;
3289 	off_t	length;
3290 };
3291 #endif
3292 int
3293 sys_truncate(td, uap)
3294 	struct thread *td;
3295 	register struct truncate_args /* {
3296 		char *path;
3297 		int pad;
3298 		off_t length;
3299 	} */ *uap;
3300 {
3301 
3302 	return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3303 }
3304 
3305 int
3306 kern_truncate(struct thread *td, char *path, enum uio_seg pathseg, off_t length)
3307 {
3308 	struct mount *mp;
3309 	struct vnode *vp;
3310 	void *rl_cookie;
3311 	struct vattr vattr;
3312 	struct nameidata nd;
3313 	int error;
3314 
3315 	if (length < 0)
3316 		return(EINVAL);
3317 	NDINIT(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, td);
3318 	if ((error = namei(&nd)) != 0)
3319 		return (error);
3320 	vp = nd.ni_vp;
3321 	rl_cookie = vn_rangelock_wlock(vp, 0, OFF_MAX);
3322 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) {
3323 		vn_rangelock_unlock(vp, rl_cookie);
3324 		vrele(vp);
3325 		return (error);
3326 	}
3327 	NDFREE(&nd, NDF_ONLY_PNBUF);
3328 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3329 	if (vp->v_type == VDIR)
3330 		error = EISDIR;
3331 #ifdef MAC
3332 	else if ((error = mac_vnode_check_write(td->td_ucred, NOCRED, vp))) {
3333 	}
3334 #endif
3335 	else if ((error = vn_writechk(vp)) == 0 &&
3336 	    (error = VOP_ACCESS(vp, VWRITE, td->td_ucred, td)) == 0) {
3337 		VATTR_NULL(&vattr);
3338 		vattr.va_size = length;
3339 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3340 	}
3341 	VOP_UNLOCK(vp, 0);
3342 	vn_finished_write(mp);
3343 	vn_rangelock_unlock(vp, rl_cookie);
3344 	vrele(vp);
3345 	return (error);
3346 }
3347 
3348 #if defined(COMPAT_43)
3349 /*
3350  * Truncate a file given its path name.
3351  */
3352 #ifndef _SYS_SYSPROTO_H_
3353 struct otruncate_args {
3354 	char	*path;
3355 	long	length;
3356 };
3357 #endif
3358 int
3359 otruncate(td, uap)
3360 	struct thread *td;
3361 	register struct otruncate_args /* {
3362 		char *path;
3363 		long length;
3364 	} */ *uap;
3365 {
3366 	struct truncate_args /* {
3367 		char *path;
3368 		int pad;
3369 		off_t length;
3370 	} */ nuap;
3371 
3372 	nuap.path = uap->path;
3373 	nuap.length = uap->length;
3374 	return (sys_truncate(td, &nuap));
3375 }
3376 #endif /* COMPAT_43 */
3377 
3378 /* Versions with the pad argument */
3379 int
3380 freebsd6_truncate(struct thread *td, struct freebsd6_truncate_args *uap)
3381 {
3382 	struct truncate_args ouap;
3383 
3384 	ouap.path = uap->path;
3385 	ouap.length = uap->length;
3386 	return (sys_truncate(td, &ouap));
3387 }
3388 
3389 int
3390 freebsd6_ftruncate(struct thread *td, struct freebsd6_ftruncate_args *uap)
3391 {
3392 	struct ftruncate_args ouap;
3393 
3394 	ouap.fd = uap->fd;
3395 	ouap.length = uap->length;
3396 	return (sys_ftruncate(td, &ouap));
3397 }
3398 
3399 /*
3400  * Sync an open file.
3401  */
3402 #ifndef _SYS_SYSPROTO_H_
3403 struct fsync_args {
3404 	int	fd;
3405 };
3406 #endif
3407 int
3408 sys_fsync(td, uap)
3409 	struct thread *td;
3410 	struct fsync_args /* {
3411 		int fd;
3412 	} */ *uap;
3413 {
3414 	struct vnode *vp;
3415 	struct mount *mp;
3416 	struct file *fp;
3417 	cap_rights_t rights;
3418 	int error, lock_flags;
3419 
3420 	AUDIT_ARG_FD(uap->fd);
3421 	error = getvnode(td->td_proc->p_fd, uap->fd,
3422 	    cap_rights_init(&rights, CAP_FSYNC), &fp);
3423 	if (error != 0)
3424 		return (error);
3425 	vp = fp->f_vnode;
3426 	error = vn_start_write(vp, &mp, V_WAIT | PCATCH);
3427 	if (error != 0)
3428 		goto drop;
3429 	if (MNT_SHARED_WRITES(mp) ||
3430 	    ((mp == NULL) && MNT_SHARED_WRITES(vp->v_mount))) {
3431 		lock_flags = LK_SHARED;
3432 	} else {
3433 		lock_flags = LK_EXCLUSIVE;
3434 	}
3435 	vn_lock(vp, lock_flags | LK_RETRY);
3436 	AUDIT_ARG_VNODE1(vp);
3437 	if (vp->v_object != NULL) {
3438 		VM_OBJECT_WLOCK(vp->v_object);
3439 		vm_object_page_clean(vp->v_object, 0, 0, 0);
3440 		VM_OBJECT_WUNLOCK(vp->v_object);
3441 	}
3442 	error = VOP_FSYNC(vp, MNT_WAIT, td);
3443 
3444 	VOP_UNLOCK(vp, 0);
3445 	vn_finished_write(mp);
3446 drop:
3447 	fdrop(fp, td);
3448 	return (error);
3449 }
3450 
3451 /*
3452  * Rename files.  Source and destination must either both be directories, or
3453  * both not be directories.  If target is a directory, it must be empty.
3454  */
3455 #ifndef _SYS_SYSPROTO_H_
3456 struct rename_args {
3457 	char	*from;
3458 	char	*to;
3459 };
3460 #endif
3461 int
3462 sys_rename(td, uap)
3463 	struct thread *td;
3464 	register struct rename_args /* {
3465 		char *from;
3466 		char *to;
3467 	} */ *uap;
3468 {
3469 
3470 	return (kern_rename(td, uap->from, uap->to, UIO_USERSPACE));
3471 }
3472 
3473 #ifndef _SYS_SYSPROTO_H_
3474 struct renameat_args {
3475 	int	oldfd;
3476 	char	*old;
3477 	int	newfd;
3478 	char	*new;
3479 };
3480 #endif
3481 int
3482 sys_renameat(struct thread *td, struct renameat_args *uap)
3483 {
3484 
3485 	return (kern_renameat(td, uap->oldfd, uap->old, uap->newfd, uap->new,
3486 	    UIO_USERSPACE));
3487 }
3488 
3489 int
3490 kern_rename(struct thread *td, char *from, char *to, enum uio_seg pathseg)
3491 {
3492 
3493 	return (kern_renameat(td, AT_FDCWD, from, AT_FDCWD, to, pathseg));
3494 }
3495 
3496 int
3497 kern_renameat(struct thread *td, int oldfd, char *old, int newfd, char *new,
3498     enum uio_seg pathseg)
3499 {
3500 	struct mount *mp = NULL;
3501 	struct vnode *tvp, *fvp, *tdvp;
3502 	struct nameidata fromnd, tond;
3503 	cap_rights_t rights;
3504 	int error;
3505 
3506 	bwillwrite();
3507 #ifdef MAC
3508 	NDINIT_ATRIGHTS(&fromnd, DELETE, LOCKPARENT | LOCKLEAF | SAVESTART |
3509 	    AUDITVNODE1, pathseg, old, oldfd,
3510 	    cap_rights_init(&rights, CAP_RENAMEAT), td);
3511 #else
3512 	NDINIT_ATRIGHTS(&fromnd, DELETE, WANTPARENT | SAVESTART | AUDITVNODE1,
3513 	    pathseg, old, oldfd, cap_rights_init(&rights, CAP_RENAMEAT), td);
3514 #endif
3515 
3516 	if ((error = namei(&fromnd)) != 0)
3517 		return (error);
3518 #ifdef MAC
3519 	error = mac_vnode_check_rename_from(td->td_ucred, fromnd.ni_dvp,
3520 	    fromnd.ni_vp, &fromnd.ni_cnd);
3521 	VOP_UNLOCK(fromnd.ni_dvp, 0);
3522 	if (fromnd.ni_dvp != fromnd.ni_vp)
3523 		VOP_UNLOCK(fromnd.ni_vp, 0);
3524 #endif
3525 	fvp = fromnd.ni_vp;
3526 	if (error == 0)
3527 		error = vn_start_write(fvp, &mp, V_WAIT | PCATCH);
3528 	if (error != 0) {
3529 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3530 		vrele(fromnd.ni_dvp);
3531 		vrele(fvp);
3532 		goto out1;
3533 	}
3534 	NDINIT_ATRIGHTS(&tond, RENAME, LOCKPARENT | LOCKLEAF | NOCACHE |
3535 	    SAVESTART | AUDITVNODE2, pathseg, new, newfd,
3536 	    cap_rights_init(&rights, CAP_LINKAT), td);
3537 	if (fromnd.ni_vp->v_type == VDIR)
3538 		tond.ni_cnd.cn_flags |= WILLBEDIR;
3539 	if ((error = namei(&tond)) != 0) {
3540 		/* Translate error code for rename("dir1", "dir2/."). */
3541 		if (error == EISDIR && fvp->v_type == VDIR)
3542 			error = EINVAL;
3543 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3544 		vrele(fromnd.ni_dvp);
3545 		vrele(fvp);
3546 		vn_finished_write(mp);
3547 		goto out1;
3548 	}
3549 	tdvp = tond.ni_dvp;
3550 	tvp = tond.ni_vp;
3551 	if (tvp != NULL) {
3552 		if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
3553 			error = ENOTDIR;
3554 			goto out;
3555 		} else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
3556 			error = EISDIR;
3557 			goto out;
3558 		}
3559 #ifdef CAPABILITIES
3560 		if (newfd != AT_FDCWD) {
3561 			/*
3562 			 * If the target already exists we require CAP_UNLINKAT
3563 			 * from 'newfd'.
3564 			 */
3565 			error = cap_check(&tond.ni_filecaps.fc_rights,
3566 			    cap_rights_init(&rights, CAP_UNLINKAT));
3567 			if (error != 0)
3568 				goto out;
3569 		}
3570 #endif
3571 	}
3572 	if (fvp == tdvp) {
3573 		error = EINVAL;
3574 		goto out;
3575 	}
3576 	/*
3577 	 * If the source is the same as the destination (that is, if they
3578 	 * are links to the same vnode), then there is nothing to do.
3579 	 */
3580 	if (fvp == tvp)
3581 		error = -1;
3582 #ifdef MAC
3583 	else
3584 		error = mac_vnode_check_rename_to(td->td_ucred, tdvp,
3585 		    tond.ni_vp, fromnd.ni_dvp == tdvp, &tond.ni_cnd);
3586 #endif
3587 out:
3588 	if (error == 0) {
3589 		error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
3590 		    tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
3591 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3592 		NDFREE(&tond, NDF_ONLY_PNBUF);
3593 	} else {
3594 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3595 		NDFREE(&tond, NDF_ONLY_PNBUF);
3596 		if (tvp != NULL)
3597 			vput(tvp);
3598 		if (tdvp == tvp)
3599 			vrele(tdvp);
3600 		else
3601 			vput(tdvp);
3602 		vrele(fromnd.ni_dvp);
3603 		vrele(fvp);
3604 	}
3605 	vrele(tond.ni_startdir);
3606 	vn_finished_write(mp);
3607 out1:
3608 	if (fromnd.ni_startdir)
3609 		vrele(fromnd.ni_startdir);
3610 	if (error == -1)
3611 		return (0);
3612 	return (error);
3613 }
3614 
3615 /*
3616  * Make a directory file.
3617  */
3618 #ifndef _SYS_SYSPROTO_H_
3619 struct mkdir_args {
3620 	char	*path;
3621 	int	mode;
3622 };
3623 #endif
3624 int
3625 sys_mkdir(td, uap)
3626 	struct thread *td;
3627 	register struct mkdir_args /* {
3628 		char *path;
3629 		int mode;
3630 	} */ *uap;
3631 {
3632 
3633 	return (kern_mkdir(td, uap->path, UIO_USERSPACE, uap->mode));
3634 }
3635 
3636 #ifndef _SYS_SYSPROTO_H_
3637 struct mkdirat_args {
3638 	int	fd;
3639 	char	*path;
3640 	mode_t	mode;
3641 };
3642 #endif
3643 int
3644 sys_mkdirat(struct thread *td, struct mkdirat_args *uap)
3645 {
3646 
3647 	return (kern_mkdirat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode));
3648 }
3649 
3650 int
3651 kern_mkdir(struct thread *td, char *path, enum uio_seg segflg, int mode)
3652 {
3653 
3654 	return (kern_mkdirat(td, AT_FDCWD, path, segflg, mode));
3655 }
3656 
3657 int
3658 kern_mkdirat(struct thread *td, int fd, char *path, enum uio_seg segflg,
3659     int mode)
3660 {
3661 	struct mount *mp;
3662 	struct vnode *vp;
3663 	struct vattr vattr;
3664 	struct nameidata nd;
3665 	cap_rights_t rights;
3666 	int error;
3667 
3668 	AUDIT_ARG_MODE(mode);
3669 restart:
3670 	bwillwrite();
3671 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1,
3672 	    segflg, path, fd, cap_rights_init(&rights, CAP_MKDIRAT), td);
3673 	nd.ni_cnd.cn_flags |= WILLBEDIR;
3674 	if ((error = namei(&nd)) != 0)
3675 		return (error);
3676 	vp = nd.ni_vp;
3677 	if (vp != NULL) {
3678 		NDFREE(&nd, NDF_ONLY_PNBUF);
3679 		/*
3680 		 * XXX namei called with LOCKPARENT but not LOCKLEAF has
3681 		 * the strange behaviour of leaving the vnode unlocked
3682 		 * if the target is the same vnode as the parent.
3683 		 */
3684 		if (vp == nd.ni_dvp)
3685 			vrele(nd.ni_dvp);
3686 		else
3687 			vput(nd.ni_dvp);
3688 		vrele(vp);
3689 		return (EEXIST);
3690 	}
3691 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3692 		NDFREE(&nd, NDF_ONLY_PNBUF);
3693 		vput(nd.ni_dvp);
3694 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3695 			return (error);
3696 		goto restart;
3697 	}
3698 	VATTR_NULL(&vattr);
3699 	vattr.va_type = VDIR;
3700 	vattr.va_mode = (mode & ACCESSPERMS) &~ td->td_proc->p_fd->fd_cmask;
3701 #ifdef MAC
3702 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
3703 	    &vattr);
3704 	if (error != 0)
3705 		goto out;
3706 #endif
3707 	error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
3708 #ifdef MAC
3709 out:
3710 #endif
3711 	NDFREE(&nd, NDF_ONLY_PNBUF);
3712 	vput(nd.ni_dvp);
3713 	if (error == 0)
3714 		vput(nd.ni_vp);
3715 	vn_finished_write(mp);
3716 	return (error);
3717 }
3718 
3719 /*
3720  * Remove a directory file.
3721  */
3722 #ifndef _SYS_SYSPROTO_H_
3723 struct rmdir_args {
3724 	char	*path;
3725 };
3726 #endif
3727 int
3728 sys_rmdir(td, uap)
3729 	struct thread *td;
3730 	struct rmdir_args /* {
3731 		char *path;
3732 	} */ *uap;
3733 {
3734 
3735 	return (kern_rmdir(td, uap->path, UIO_USERSPACE));
3736 }
3737 
3738 int
3739 kern_rmdir(struct thread *td, char *path, enum uio_seg pathseg)
3740 {
3741 
3742 	return (kern_rmdirat(td, AT_FDCWD, path, pathseg));
3743 }
3744 
3745 int
3746 kern_rmdirat(struct thread *td, int fd, char *path, enum uio_seg pathseg)
3747 {
3748 	struct mount *mp;
3749 	struct vnode *vp;
3750 	struct nameidata nd;
3751 	cap_rights_t rights;
3752 	int error;
3753 
3754 restart:
3755 	bwillwrite();
3756 	NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1,
3757 	    pathseg, path, fd, cap_rights_init(&rights, CAP_UNLINKAT), td);
3758 	if ((error = namei(&nd)) != 0)
3759 		return (error);
3760 	vp = nd.ni_vp;
3761 	if (vp->v_type != VDIR) {
3762 		error = ENOTDIR;
3763 		goto out;
3764 	}
3765 	/*
3766 	 * No rmdir "." please.
3767 	 */
3768 	if (nd.ni_dvp == vp) {
3769 		error = EINVAL;
3770 		goto out;
3771 	}
3772 	/*
3773 	 * The root of a mounted filesystem cannot be deleted.
3774 	 */
3775 	if (vp->v_vflag & VV_ROOT) {
3776 		error = EBUSY;
3777 		goto out;
3778 	}
3779 #ifdef MAC
3780 	error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
3781 	    &nd.ni_cnd);
3782 	if (error != 0)
3783 		goto out;
3784 #endif
3785 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
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 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3793 			return (error);
3794 		goto restart;
3795 	}
3796 	vfs_notify_upper(vp, VFS_NOTIFY_UPPER_UNLINK);
3797 	error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
3798 	vn_finished_write(mp);
3799 out:
3800 	NDFREE(&nd, NDF_ONLY_PNBUF);
3801 	vput(vp);
3802 	if (nd.ni_dvp == vp)
3803 		vrele(nd.ni_dvp);
3804 	else
3805 		vput(nd.ni_dvp);
3806 	return (error);
3807 }
3808 
3809 #ifdef COMPAT_43
3810 /*
3811  * Read a block of directory entries in a filesystem independent format.
3812  */
3813 #ifndef _SYS_SYSPROTO_H_
3814 struct ogetdirentries_args {
3815 	int	fd;
3816 	char	*buf;
3817 	u_int	count;
3818 	long	*basep;
3819 };
3820 #endif
3821 int
3822 ogetdirentries(struct thread *td, struct ogetdirentries_args *uap)
3823 {
3824 	long loff;
3825 	int error;
3826 
3827 	error = kern_ogetdirentries(td, uap, &loff);
3828 	if (error == 0)
3829 		error = copyout(&loff, uap->basep, sizeof(long));
3830 	return (error);
3831 }
3832 
3833 int
3834 kern_ogetdirentries(struct thread *td, struct ogetdirentries_args *uap,
3835     long *ploff)
3836 {
3837 	struct vnode *vp;
3838 	struct file *fp;
3839 	struct uio auio, kuio;
3840 	struct iovec aiov, kiov;
3841 	struct dirent *dp, *edp;
3842 	cap_rights_t rights;
3843 	caddr_t dirbuf;
3844 	int error, eofflag, readcnt;
3845 	long loff;
3846 	off_t foffset;
3847 
3848 	/* XXX arbitrary sanity limit on `count'. */
3849 	if (uap->count > 64 * 1024)
3850 		return (EINVAL);
3851 	error = getvnode(td->td_proc->p_fd, uap->fd,
3852 	    cap_rights_init(&rights, CAP_READ), &fp);
3853 	if (error != 0)
3854 		return (error);
3855 	if ((fp->f_flag & FREAD) == 0) {
3856 		fdrop(fp, td);
3857 		return (EBADF);
3858 	}
3859 	vp = fp->f_vnode;
3860 	foffset = foffset_lock(fp, 0);
3861 unionread:
3862 	if (vp->v_type != VDIR) {
3863 		foffset_unlock(fp, foffset, 0);
3864 		fdrop(fp, td);
3865 		return (EINVAL);
3866 	}
3867 	aiov.iov_base = uap->buf;
3868 	aiov.iov_len = uap->count;
3869 	auio.uio_iov = &aiov;
3870 	auio.uio_iovcnt = 1;
3871 	auio.uio_rw = UIO_READ;
3872 	auio.uio_segflg = UIO_USERSPACE;
3873 	auio.uio_td = td;
3874 	auio.uio_resid = uap->count;
3875 	vn_lock(vp, LK_SHARED | LK_RETRY);
3876 	loff = auio.uio_offset = foffset;
3877 #ifdef MAC
3878 	error = mac_vnode_check_readdir(td->td_ucred, vp);
3879 	if (error != 0) {
3880 		VOP_UNLOCK(vp, 0);
3881 		foffset_unlock(fp, foffset, FOF_NOUPDATE);
3882 		fdrop(fp, td);
3883 		return (error);
3884 	}
3885 #endif
3886 #	if (BYTE_ORDER != LITTLE_ENDIAN)
3887 		if (vp->v_mount->mnt_maxsymlinklen <= 0) {
3888 			error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag,
3889 			    NULL, NULL);
3890 			foffset = auio.uio_offset;
3891 		} else
3892 #	endif
3893 	{
3894 		kuio = auio;
3895 		kuio.uio_iov = &kiov;
3896 		kuio.uio_segflg = UIO_SYSSPACE;
3897 		kiov.iov_len = uap->count;
3898 		dirbuf = malloc(uap->count, M_TEMP, M_WAITOK);
3899 		kiov.iov_base = dirbuf;
3900 		error = VOP_READDIR(vp, &kuio, fp->f_cred, &eofflag,
3901 			    NULL, NULL);
3902 		foffset = kuio.uio_offset;
3903 		if (error == 0) {
3904 			readcnt = uap->count - kuio.uio_resid;
3905 			edp = (struct dirent *)&dirbuf[readcnt];
3906 			for (dp = (struct dirent *)dirbuf; dp < edp; ) {
3907 #				if (BYTE_ORDER == LITTLE_ENDIAN)
3908 					/*
3909 					 * The expected low byte of
3910 					 * dp->d_namlen is our dp->d_type.
3911 					 * The high MBZ byte of dp->d_namlen
3912 					 * is our dp->d_namlen.
3913 					 */
3914 					dp->d_type = dp->d_namlen;
3915 					dp->d_namlen = 0;
3916 #				else
3917 					/*
3918 					 * The dp->d_type is the high byte
3919 					 * of the expected dp->d_namlen,
3920 					 * so must be zero'ed.
3921 					 */
3922 					dp->d_type = 0;
3923 #				endif
3924 				if (dp->d_reclen > 0) {
3925 					dp = (struct dirent *)
3926 					    ((char *)dp + dp->d_reclen);
3927 				} else {
3928 					error = EIO;
3929 					break;
3930 				}
3931 			}
3932 			if (dp >= edp)
3933 				error = uiomove(dirbuf, readcnt, &auio);
3934 		}
3935 		free(dirbuf, M_TEMP);
3936 	}
3937 	if (error != 0) {
3938 		VOP_UNLOCK(vp, 0);
3939 		foffset_unlock(fp, foffset, 0);
3940 		fdrop(fp, td);
3941 		return (error);
3942 	}
3943 	if (uap->count == auio.uio_resid &&
3944 	    (vp->v_vflag & VV_ROOT) &&
3945 	    (vp->v_mount->mnt_flag & MNT_UNION)) {
3946 		struct vnode *tvp = vp;
3947 		vp = vp->v_mount->mnt_vnodecovered;
3948 		VREF(vp);
3949 		fp->f_vnode = vp;
3950 		fp->f_data = vp;
3951 		foffset = 0;
3952 		vput(tvp);
3953 		goto unionread;
3954 	}
3955 	VOP_UNLOCK(vp, 0);
3956 	foffset_unlock(fp, foffset, 0);
3957 	fdrop(fp, td);
3958 	td->td_retval[0] = uap->count - auio.uio_resid;
3959 	if (error == 0)
3960 		*ploff = loff;
3961 	return (error);
3962 }
3963 #endif /* COMPAT_43 */
3964 
3965 /*
3966  * Read a block of directory entries in a filesystem independent format.
3967  */
3968 #ifndef _SYS_SYSPROTO_H_
3969 struct getdirentries_args {
3970 	int	fd;
3971 	char	*buf;
3972 	u_int	count;
3973 	long	*basep;
3974 };
3975 #endif
3976 int
3977 sys_getdirentries(td, uap)
3978 	struct thread *td;
3979 	register struct getdirentries_args /* {
3980 		int fd;
3981 		char *buf;
3982 		u_int count;
3983 		long *basep;
3984 	} */ *uap;
3985 {
3986 	long base;
3987 	int error;
3988 
3989 	error = kern_getdirentries(td, uap->fd, uap->buf, uap->count, &base,
3990 	    NULL, UIO_USERSPACE);
3991 	if (error != 0)
3992 		return (error);
3993 	if (uap->basep != NULL)
3994 		error = copyout(&base, uap->basep, sizeof(long));
3995 	return (error);
3996 }
3997 
3998 int
3999 kern_getdirentries(struct thread *td, int fd, char *buf, u_int count,
4000     long *basep, ssize_t *residp, enum uio_seg bufseg)
4001 {
4002 	struct vnode *vp;
4003 	struct file *fp;
4004 	struct uio auio;
4005 	struct iovec aiov;
4006 	cap_rights_t rights;
4007 	long loff;
4008 	int error, eofflag;
4009 	off_t foffset;
4010 
4011 	AUDIT_ARG_FD(fd);
4012 	if (count > IOSIZE_MAX)
4013 		return (EINVAL);
4014 	auio.uio_resid = count;
4015 	error = getvnode(td->td_proc->p_fd, fd,
4016 	    cap_rights_init(&rights, CAP_READ), &fp);
4017 	if (error != 0)
4018 		return (error);
4019 	if ((fp->f_flag & FREAD) == 0) {
4020 		fdrop(fp, td);
4021 		return (EBADF);
4022 	}
4023 	vp = fp->f_vnode;
4024 	foffset = foffset_lock(fp, 0);
4025 unionread:
4026 	if (vp->v_type != VDIR) {
4027 		error = EINVAL;
4028 		goto fail;
4029 	}
4030 	aiov.iov_base = buf;
4031 	aiov.iov_len = count;
4032 	auio.uio_iov = &aiov;
4033 	auio.uio_iovcnt = 1;
4034 	auio.uio_rw = UIO_READ;
4035 	auio.uio_segflg = bufseg;
4036 	auio.uio_td = td;
4037 	vn_lock(vp, LK_SHARED | LK_RETRY);
4038 	AUDIT_ARG_VNODE1(vp);
4039 	loff = auio.uio_offset = foffset;
4040 #ifdef MAC
4041 	error = mac_vnode_check_readdir(td->td_ucred, vp);
4042 	if (error == 0)
4043 #endif
4044 		error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, NULL,
4045 		    NULL);
4046 	foffset = auio.uio_offset;
4047 	if (error != 0) {
4048 		VOP_UNLOCK(vp, 0);
4049 		goto fail;
4050 	}
4051 	if (count == auio.uio_resid &&
4052 	    (vp->v_vflag & VV_ROOT) &&
4053 	    (vp->v_mount->mnt_flag & MNT_UNION)) {
4054 		struct vnode *tvp = vp;
4055 
4056 		vp = vp->v_mount->mnt_vnodecovered;
4057 		VREF(vp);
4058 		fp->f_vnode = vp;
4059 		fp->f_data = vp;
4060 		foffset = 0;
4061 		vput(tvp);
4062 		goto unionread;
4063 	}
4064 	VOP_UNLOCK(vp, 0);
4065 	*basep = loff;
4066 	if (residp != NULL)
4067 		*residp = auio.uio_resid;
4068 	td->td_retval[0] = count - auio.uio_resid;
4069 fail:
4070 	foffset_unlock(fp, foffset, 0);
4071 	fdrop(fp, td);
4072 	return (error);
4073 }
4074 
4075 #ifndef _SYS_SYSPROTO_H_
4076 struct getdents_args {
4077 	int fd;
4078 	char *buf;
4079 	size_t count;
4080 };
4081 #endif
4082 int
4083 sys_getdents(td, uap)
4084 	struct thread *td;
4085 	register struct getdents_args /* {
4086 		int fd;
4087 		char *buf;
4088 		u_int count;
4089 	} */ *uap;
4090 {
4091 	struct getdirentries_args ap;
4092 
4093 	ap.fd = uap->fd;
4094 	ap.buf = uap->buf;
4095 	ap.count = uap->count;
4096 	ap.basep = NULL;
4097 	return (sys_getdirentries(td, &ap));
4098 }
4099 
4100 /*
4101  * Set the mode mask for creation of filesystem nodes.
4102  */
4103 #ifndef _SYS_SYSPROTO_H_
4104 struct umask_args {
4105 	int	newmask;
4106 };
4107 #endif
4108 int
4109 sys_umask(td, uap)
4110 	struct thread *td;
4111 	struct umask_args /* {
4112 		int newmask;
4113 	} */ *uap;
4114 {
4115 	register struct filedesc *fdp;
4116 
4117 	FILEDESC_XLOCK(td->td_proc->p_fd);
4118 	fdp = td->td_proc->p_fd;
4119 	td->td_retval[0] = fdp->fd_cmask;
4120 	fdp->fd_cmask = uap->newmask & ALLPERMS;
4121 	FILEDESC_XUNLOCK(td->td_proc->p_fd);
4122 	return (0);
4123 }
4124 
4125 /*
4126  * Void all references to file by ripping underlying filesystem away from
4127  * vnode.
4128  */
4129 #ifndef _SYS_SYSPROTO_H_
4130 struct revoke_args {
4131 	char	*path;
4132 };
4133 #endif
4134 int
4135 sys_revoke(td, uap)
4136 	struct thread *td;
4137 	register struct revoke_args /* {
4138 		char *path;
4139 	} */ *uap;
4140 {
4141 	struct vnode *vp;
4142 	struct vattr vattr;
4143 	struct nameidata nd;
4144 	int error;
4145 
4146 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
4147 	    uap->path, td);
4148 	if ((error = namei(&nd)) != 0)
4149 		return (error);
4150 	vp = nd.ni_vp;
4151 	NDFREE(&nd, NDF_ONLY_PNBUF);
4152 	if (vp->v_type != VCHR || vp->v_rdev == NULL) {
4153 		error = EINVAL;
4154 		goto out;
4155 	}
4156 #ifdef MAC
4157 	error = mac_vnode_check_revoke(td->td_ucred, vp);
4158 	if (error != 0)
4159 		goto out;
4160 #endif
4161 	error = VOP_GETATTR(vp, &vattr, td->td_ucred);
4162 	if (error != 0)
4163 		goto out;
4164 	if (td->td_ucred->cr_uid != vattr.va_uid) {
4165 		error = priv_check(td, PRIV_VFS_ADMIN);
4166 		if (error != 0)
4167 			goto out;
4168 	}
4169 	if (vcount(vp) > 1)
4170 		VOP_REVOKE(vp, REVOKEALL);
4171 out:
4172 	vput(vp);
4173 	return (error);
4174 }
4175 
4176 /*
4177  * Convert a user file descriptor to a kernel file entry and check that, if it
4178  * is a capability, the correct rights are present. A reference on the file
4179  * entry is held upon returning.
4180  */
4181 int
4182 getvnode(struct filedesc *fdp, int fd, cap_rights_t *rightsp, struct file **fpp)
4183 {
4184 	struct file *fp;
4185 	int error;
4186 
4187 	error = fget_unlocked(fdp, fd, rightsp, 0, &fp, NULL);
4188 	if (error != 0)
4189 		return (error);
4190 
4191 	/*
4192 	 * The file could be not of the vnode type, or it may be not
4193 	 * yet fully initialized, in which case the f_vnode pointer
4194 	 * may be set, but f_ops is still badfileops.  E.g.,
4195 	 * devfs_open() transiently create such situation to
4196 	 * facilitate csw d_fdopen().
4197 	 *
4198 	 * Dupfdopen() handling in kern_openat() installs the
4199 	 * half-baked file into the process descriptor table, allowing
4200 	 * other thread to dereference it. Guard against the race by
4201 	 * checking f_ops.
4202 	 */
4203 	if (fp->f_vnode == NULL || fp->f_ops == &badfileops) {
4204 		fdrop(fp, curthread);
4205 		return (EINVAL);
4206 	}
4207 	*fpp = fp;
4208 	return (0);
4209 }
4210 
4211 
4212 /*
4213  * Get an (NFS) file handle.
4214  */
4215 #ifndef _SYS_SYSPROTO_H_
4216 struct lgetfh_args {
4217 	char	*fname;
4218 	fhandle_t *fhp;
4219 };
4220 #endif
4221 int
4222 sys_lgetfh(td, uap)
4223 	struct thread *td;
4224 	register struct lgetfh_args *uap;
4225 {
4226 	struct nameidata nd;
4227 	fhandle_t fh;
4228 	register struct vnode *vp;
4229 	int error;
4230 
4231 	error = priv_check(td, PRIV_VFS_GETFH);
4232 	if (error != 0)
4233 		return (error);
4234 	NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
4235 	    uap->fname, td);
4236 	error = namei(&nd);
4237 	if (error != 0)
4238 		return (error);
4239 	NDFREE(&nd, NDF_ONLY_PNBUF);
4240 	vp = nd.ni_vp;
4241 	bzero(&fh, sizeof(fh));
4242 	fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
4243 	error = VOP_VPTOFH(vp, &fh.fh_fid);
4244 	vput(vp);
4245 	if (error == 0)
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 != 0)
4268 		return (error);
4269 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
4270 	    uap->fname, td);
4271 	error = namei(&nd);
4272 	if (error != 0)
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 == 0)
4281 		error = copyout(&fh, uap->fhp, sizeof (fh));
4282 	return (error);
4283 }
4284 
4285 /*
4286  * syscall for the rpc.lockd to use to translate a NFS file handle into an
4287  * open descriptor.
4288  *
4289  * warning: do not remove the priv_check() call or this becomes one giant
4290  * security hole.
4291  */
4292 #ifndef _SYS_SYSPROTO_H_
4293 struct fhopen_args {
4294 	const struct fhandle *u_fhp;
4295 	int flags;
4296 };
4297 #endif
4298 int
4299 sys_fhopen(td, uap)
4300 	struct thread *td;
4301 	struct fhopen_args /* {
4302 		const struct fhandle *u_fhp;
4303 		int flags;
4304 	} */ *uap;
4305 {
4306 	struct mount *mp;
4307 	struct vnode *vp;
4308 	struct fhandle fhp;
4309 	struct file *fp;
4310 	int fmode, error;
4311 	int indx;
4312 
4313 	error = priv_check(td, PRIV_VFS_FHOPEN);
4314 	if (error != 0)
4315 		return (error);
4316 	indx = -1;
4317 	fmode = FFLAGS(uap->flags);
4318 	/* why not allow a non-read/write open for our lockd? */
4319 	if (((fmode & (FREAD | FWRITE)) == 0) || (fmode & O_CREAT))
4320 		return (EINVAL);
4321 	error = copyin(uap->u_fhp, &fhp, sizeof(fhp));
4322 	if (error != 0)
4323 		return(error);
4324 	/* find the mount point */
4325 	mp = vfs_busyfs(&fhp.fh_fsid);
4326 	if (mp == NULL)
4327 		return (ESTALE);
4328 	/* now give me my vnode, it gets returned to me locked */
4329 	error = VFS_FHTOVP(mp, &fhp.fh_fid, LK_EXCLUSIVE, &vp);
4330 	vfs_unbusy(mp);
4331 	if (error != 0)
4332 		return (error);
4333 
4334 	error = falloc_noinstall(td, &fp);
4335 	if (error != 0) {
4336 		vput(vp);
4337 		return (error);
4338 	}
4339 	/*
4340 	 * An extra reference on `fp' has been held for us by
4341 	 * falloc_noinstall().
4342 	 */
4343 
4344 #ifdef INVARIANTS
4345 	td->td_dupfd = -1;
4346 #endif
4347 	error = vn_open_vnode(vp, fmode, td->td_ucred, td, fp);
4348 	if (error != 0) {
4349 		KASSERT(fp->f_ops == &badfileops,
4350 		    ("VOP_OPEN in fhopen() set f_ops"));
4351 		KASSERT(td->td_dupfd < 0,
4352 		    ("fhopen() encountered fdopen()"));
4353 
4354 		vput(vp);
4355 		goto bad;
4356 	}
4357 #ifdef INVARIANTS
4358 	td->td_dupfd = 0;
4359 #endif
4360 	fp->f_vnode = vp;
4361 	fp->f_seqcount = 1;
4362 	finit(fp, (fmode & FMASK) | (fp->f_flag & FHASLOCK), DTYPE_VNODE, vp,
4363 	    &vnops);
4364 	VOP_UNLOCK(vp, 0);
4365 	if ((fmode & O_TRUNC) != 0) {
4366 		error = fo_truncate(fp, 0, td->td_ucred, td);
4367 		if (error != 0)
4368 			goto bad;
4369 	}
4370 
4371 	error = finstall(td, fp, &indx, fmode, NULL);
4372 bad:
4373 	fdrop(fp, td);
4374 	td->td_retval[0] = indx;
4375 	return (error);
4376 }
4377 
4378 /*
4379  * Stat an (NFS) file handle.
4380  */
4381 #ifndef _SYS_SYSPROTO_H_
4382 struct fhstat_args {
4383 	struct fhandle *u_fhp;
4384 	struct stat *sb;
4385 };
4386 #endif
4387 int
4388 sys_fhstat(td, uap)
4389 	struct thread *td;
4390 	register struct fhstat_args /* {
4391 		struct fhandle *u_fhp;
4392 		struct stat *sb;
4393 	} */ *uap;
4394 {
4395 	struct stat sb;
4396 	struct fhandle fh;
4397 	int error;
4398 
4399 	error = copyin(uap->u_fhp, &fh, sizeof(fh));
4400 	if (error != 0)
4401 		return (error);
4402 	error = kern_fhstat(td, fh, &sb);
4403 	if (error == 0)
4404 		error = copyout(&sb, uap->sb, sizeof(sb));
4405 	return (error);
4406 }
4407 
4408 int
4409 kern_fhstat(struct thread *td, struct fhandle fh, struct stat *sb)
4410 {
4411 	struct mount *mp;
4412 	struct vnode *vp;
4413 	int error;
4414 
4415 	error = priv_check(td, PRIV_VFS_FHSTAT);
4416 	if (error != 0)
4417 		return (error);
4418 	if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4419 		return (ESTALE);
4420 	error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4421 	vfs_unbusy(mp);
4422 	if (error != 0)
4423 		return (error);
4424 	error = vn_stat(vp, sb, td->td_ucred, NOCRED, td);
4425 	vput(vp);
4426 	return (error);
4427 }
4428 
4429 /*
4430  * Implement fstatfs() for (NFS) file handles.
4431  */
4432 #ifndef _SYS_SYSPROTO_H_
4433 struct fhstatfs_args {
4434 	struct fhandle *u_fhp;
4435 	struct statfs *buf;
4436 };
4437 #endif
4438 int
4439 sys_fhstatfs(td, uap)
4440 	struct thread *td;
4441 	struct fhstatfs_args /* {
4442 		struct fhandle *u_fhp;
4443 		struct statfs *buf;
4444 	} */ *uap;
4445 {
4446 	struct statfs sf;
4447 	fhandle_t fh;
4448 	int error;
4449 
4450 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
4451 	if (error != 0)
4452 		return (error);
4453 	error = kern_fhstatfs(td, fh, &sf);
4454 	if (error != 0)
4455 		return (error);
4456 	return (copyout(&sf, uap->buf, sizeof(sf)));
4457 }
4458 
4459 int
4460 kern_fhstatfs(struct thread *td, fhandle_t fh, struct statfs *buf)
4461 {
4462 	struct statfs *sp;
4463 	struct mount *mp;
4464 	struct vnode *vp;
4465 	int error;
4466 
4467 	error = priv_check(td, PRIV_VFS_FHSTATFS);
4468 	if (error != 0)
4469 		return (error);
4470 	if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4471 		return (ESTALE);
4472 	error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4473 	if (error != 0) {
4474 		vfs_unbusy(mp);
4475 		return (error);
4476 	}
4477 	vput(vp);
4478 	error = prison_canseemount(td->td_ucred, mp);
4479 	if (error != 0)
4480 		goto out;
4481 #ifdef MAC
4482 	error = mac_mount_check_stat(td->td_ucred, mp);
4483 	if (error != 0)
4484 		goto out;
4485 #endif
4486 	/*
4487 	 * Set these in case the underlying filesystem fails to do so.
4488 	 */
4489 	sp = &mp->mnt_stat;
4490 	sp->f_version = STATFS_VERSION;
4491 	sp->f_namemax = NAME_MAX;
4492 	sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
4493 	error = VFS_STATFS(mp, sp);
4494 	if (error == 0)
4495 		*buf = *sp;
4496 out:
4497 	vfs_unbusy(mp);
4498 	return (error);
4499 }
4500 
4501 int
4502 kern_posix_fallocate(struct thread *td, int fd, off_t offset, off_t len)
4503 {
4504 	struct file *fp;
4505 	struct mount *mp;
4506 	struct vnode *vp;
4507 	cap_rights_t rights;
4508 	off_t olen, ooffset;
4509 	int error;
4510 
4511 	fp = NULL;
4512 	error = fget(td, fd, cap_rights_init(&rights, CAP_WRITE), &fp);
4513 	if (error != 0)
4514 		goto out;
4515 
4516 	switch (fp->f_type) {
4517 	case DTYPE_VNODE:
4518 		break;
4519 	case DTYPE_PIPE:
4520 	case DTYPE_FIFO:
4521 		error = ESPIPE;
4522 		goto out;
4523 	default:
4524 		error = ENODEV;
4525 		goto out;
4526 	}
4527 	if ((fp->f_flag & FWRITE) == 0) {
4528 		error = EBADF;
4529 		goto out;
4530 	}
4531 	vp = fp->f_vnode;
4532 	if (vp->v_type != VREG) {
4533 		error = ENODEV;
4534 		goto out;
4535 	}
4536 	if (offset < 0 || len <= 0) {
4537 		error = EINVAL;
4538 		goto out;
4539 	}
4540 	/* Check for wrap. */
4541 	if (offset > OFF_MAX - len) {
4542 		error = EFBIG;
4543 		goto out;
4544 	}
4545 
4546 	/* Allocating blocks may take a long time, so iterate. */
4547 	for (;;) {
4548 		olen = len;
4549 		ooffset = offset;
4550 
4551 		bwillwrite();
4552 		mp = NULL;
4553 		error = vn_start_write(vp, &mp, V_WAIT | PCATCH);
4554 		if (error != 0)
4555 			break;
4556 		error = vn_lock(vp, LK_EXCLUSIVE);
4557 		if (error != 0) {
4558 			vn_finished_write(mp);
4559 			break;
4560 		}
4561 #ifdef MAC
4562 		error = mac_vnode_check_write(td->td_ucred, fp->f_cred, vp);
4563 		if (error == 0)
4564 #endif
4565 			error = VOP_ALLOCATE(vp, &offset, &len);
4566 		VOP_UNLOCK(vp, 0);
4567 		vn_finished_write(mp);
4568 
4569 		if (olen + ooffset != offset + len) {
4570 			panic("offset + len changed from %jx/%jx to %jx/%jx",
4571 			    ooffset, olen, offset, len);
4572 		}
4573 		if (error != 0 || len == 0)
4574 			break;
4575 		KASSERT(olen > len, ("Iteration did not make progress?"));
4576 		maybe_yield();
4577 	}
4578  out:
4579 	if (fp != NULL)
4580 		fdrop(fp, td);
4581 	return (error);
4582 }
4583 
4584 int
4585 sys_posix_fallocate(struct thread *td, struct posix_fallocate_args *uap)
4586 {
4587 
4588 	return (kern_posix_fallocate(td, uap->fd, uap->offset, uap->len));
4589 }
4590 
4591 /*
4592  * Unlike madvise(2), we do not make a best effort to remember every
4593  * possible caching hint.  Instead, we remember the last setting with
4594  * the exception that we will allow POSIX_FADV_NORMAL to adjust the
4595  * region of any current setting.
4596  */
4597 int
4598 kern_posix_fadvise(struct thread *td, int fd, off_t offset, off_t len,
4599     int advice)
4600 {
4601 	struct fadvise_info *fa, *new;
4602 	struct file *fp;
4603 	struct vnode *vp;
4604 	cap_rights_t rights;
4605 	off_t end;
4606 	int error;
4607 
4608 	if (offset < 0 || len < 0 || offset > OFF_MAX - len)
4609 		return (EINVAL);
4610 	switch (advice) {
4611 	case POSIX_FADV_SEQUENTIAL:
4612 	case POSIX_FADV_RANDOM:
4613 	case POSIX_FADV_NOREUSE:
4614 		new = malloc(sizeof(*fa), M_FADVISE, M_WAITOK);
4615 		break;
4616 	case POSIX_FADV_NORMAL:
4617 	case POSIX_FADV_WILLNEED:
4618 	case POSIX_FADV_DONTNEED:
4619 		new = NULL;
4620 		break;
4621 	default:
4622 		return (EINVAL);
4623 	}
4624 	/* XXX: CAP_POSIX_FADVISE? */
4625 	error = fget(td, fd, cap_rights_init(&rights), &fp);
4626 	if (error != 0)
4627 		goto out;
4628 
4629 	switch (fp->f_type) {
4630 	case DTYPE_VNODE:
4631 		break;
4632 	case DTYPE_PIPE:
4633 	case DTYPE_FIFO:
4634 		error = ESPIPE;
4635 		goto out;
4636 	default:
4637 		error = ENODEV;
4638 		goto out;
4639 	}
4640 	vp = fp->f_vnode;
4641 	if (vp->v_type != VREG) {
4642 		error = ENODEV;
4643 		goto out;
4644 	}
4645 	if (len == 0)
4646 		end = OFF_MAX;
4647 	else
4648 		end = offset + len - 1;
4649 	switch (advice) {
4650 	case POSIX_FADV_SEQUENTIAL:
4651 	case POSIX_FADV_RANDOM:
4652 	case POSIX_FADV_NOREUSE:
4653 		/*
4654 		 * Try to merge any existing non-standard region with
4655 		 * this new region if possible, otherwise create a new
4656 		 * non-standard region for this request.
4657 		 */
4658 		mtx_pool_lock(mtxpool_sleep, fp);
4659 		fa = fp->f_advice;
4660 		if (fa != NULL && fa->fa_advice == advice &&
4661 		    ((fa->fa_start <= end && fa->fa_end >= offset) ||
4662 		    (end != OFF_MAX && fa->fa_start == end + 1) ||
4663 		    (fa->fa_end != OFF_MAX && fa->fa_end + 1 == offset))) {
4664 			if (offset < fa->fa_start)
4665 				fa->fa_start = offset;
4666 			if (end > fa->fa_end)
4667 				fa->fa_end = end;
4668 		} else {
4669 			new->fa_advice = advice;
4670 			new->fa_start = offset;
4671 			new->fa_end = end;
4672 			new->fa_prevstart = 0;
4673 			new->fa_prevend = 0;
4674 			fp->f_advice = new;
4675 			new = fa;
4676 		}
4677 		mtx_pool_unlock(mtxpool_sleep, fp);
4678 		break;
4679 	case POSIX_FADV_NORMAL:
4680 		/*
4681 		 * If a the "normal" region overlaps with an existing
4682 		 * non-standard region, trim or remove the
4683 		 * non-standard region.
4684 		 */
4685 		mtx_pool_lock(mtxpool_sleep, fp);
4686 		fa = fp->f_advice;
4687 		if (fa != NULL) {
4688 			if (offset <= fa->fa_start && end >= fa->fa_end) {
4689 				new = fa;
4690 				fp->f_advice = NULL;
4691 			} else if (offset <= fa->fa_start &&
4692 			    end >= fa->fa_start)
4693 				fa->fa_start = end + 1;
4694 			else if (offset <= fa->fa_end && end >= fa->fa_end)
4695 				fa->fa_end = offset - 1;
4696 			else if (offset >= fa->fa_start && end <= fa->fa_end) {
4697 				/*
4698 				 * If the "normal" region is a middle
4699 				 * portion of the existing
4700 				 * non-standard region, just remove
4701 				 * the whole thing rather than picking
4702 				 * one side or the other to
4703 				 * preserve.
4704 				 */
4705 				new = fa;
4706 				fp->f_advice = NULL;
4707 			}
4708 		}
4709 		mtx_pool_unlock(mtxpool_sleep, fp);
4710 		break;
4711 	case POSIX_FADV_WILLNEED:
4712 	case POSIX_FADV_DONTNEED:
4713 		error = VOP_ADVISE(vp, offset, end, advice);
4714 		break;
4715 	}
4716 out:
4717 	if (fp != NULL)
4718 		fdrop(fp, td);
4719 	free(new, M_FADVISE);
4720 	return (error);
4721 }
4722 
4723 int
4724 sys_posix_fadvise(struct thread *td, struct posix_fadvise_args *uap)
4725 {
4726 
4727 	return (kern_posix_fadvise(td, uap->fd, uap->offset, uap->len,
4728 	    uap->advice));
4729 }
4730