xref: /freebsd/sys/kern/vfs_syscalls.c (revision 86c9d9918f1db7cdd968b60f8902466887bcd9e9)
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_ktrace.h"
43 
44 #include <sys/param.h>
45 #include <sys/systm.h>
46 #include <sys/bio.h>
47 #include <sys/buf.h>
48 #include <sys/capsicum.h>
49 #include <sys/disk.h>
50 #include <sys/sysent.h>
51 #include <sys/malloc.h>
52 #include <sys/mount.h>
53 #include <sys/mutex.h>
54 #include <sys/sysproto.h>
55 #include <sys/namei.h>
56 #include <sys/filedesc.h>
57 #include <sys/kernel.h>
58 #include <sys/fcntl.h>
59 #include <sys/file.h>
60 #include <sys/filio.h>
61 #include <sys/limits.h>
62 #include <sys/linker.h>
63 #include <sys/rwlock.h>
64 #include <sys/sdt.h>
65 #include <sys/stat.h>
66 #include <sys/sx.h>
67 #include <sys/unistd.h>
68 #include <sys/vnode.h>
69 #include <sys/priv.h>
70 #include <sys/proc.h>
71 #include <sys/dirent.h>
72 #include <sys/jail.h>
73 #include <sys/syscallsubr.h>
74 #include <sys/sysctl.h>
75 #ifdef KTRACE
76 #include <sys/ktrace.h>
77 #endif
78 
79 #include <machine/stdarg.h>
80 
81 #include <security/audit/audit.h>
82 #include <security/mac/mac_framework.h>
83 
84 #include <vm/vm.h>
85 #include <vm/vm_object.h>
86 #include <vm/vm_page.h>
87 #include <vm/uma.h>
88 
89 #include <ufs/ufs/quota.h>
90 
91 MALLOC_DEFINE(M_FADVISE, "fadvise", "posix_fadvise(2) information");
92 
93 SDT_PROVIDER_DEFINE(vfs);
94 SDT_PROBE_DEFINE2(vfs, , stat, mode, "char *", "int");
95 SDT_PROBE_DEFINE2(vfs, , stat, reg, "char *", "int");
96 
97 static int kern_chflagsat(struct thread *td, int fd, const char *path,
98     enum uio_seg pathseg, u_long flags, int atflag);
99 static int setfflags(struct thread *td, struct vnode *, u_long);
100 static int getutimes(const struct timeval *, enum uio_seg, struct timespec *);
101 static int getutimens(const struct timespec *, enum uio_seg,
102     struct timespec *, int *);
103 static int setutimes(struct thread *td, struct vnode *,
104     const struct timespec *, int, int);
105 static int vn_access(struct vnode *vp, int user_flags, struct ucred *cred,
106     struct thread *td);
107 
108 /*
109  * Sync each mounted filesystem.
110  */
111 #ifndef _SYS_SYSPROTO_H_
112 struct sync_args {
113 	int     dummy;
114 };
115 #endif
116 /* ARGSUSED */
117 int
118 sys_sync(td, uap)
119 	struct thread *td;
120 	struct sync_args *uap;
121 {
122 	struct mount *mp, *nmp;
123 	int save;
124 
125 	mtx_lock(&mountlist_mtx);
126 	for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
127 		if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK)) {
128 			nmp = TAILQ_NEXT(mp, mnt_list);
129 			continue;
130 		}
131 		if ((mp->mnt_flag & MNT_RDONLY) == 0 &&
132 		    vn_start_write(NULL, &mp, V_NOWAIT) == 0) {
133 			save = curthread_pflags_set(TDP_SYNCIO);
134 			vfs_msync(mp, MNT_NOWAIT);
135 			VFS_SYNC(mp, MNT_NOWAIT);
136 			curthread_pflags_restore(save);
137 			vn_finished_write(mp);
138 		}
139 		mtx_lock(&mountlist_mtx);
140 		nmp = TAILQ_NEXT(mp, mnt_list);
141 		vfs_unbusy(mp);
142 	}
143 	mtx_unlock(&mountlist_mtx);
144 	return (0);
145 }
146 
147 /*
148  * Change filesystem quotas.
149  */
150 #ifndef _SYS_SYSPROTO_H_
151 struct quotactl_args {
152 	char *path;
153 	int cmd;
154 	int uid;
155 	caddr_t arg;
156 };
157 #endif
158 int
159 sys_quotactl(td, uap)
160 	struct thread *td;
161 	register struct quotactl_args /* {
162 		char *path;
163 		int cmd;
164 		int uid;
165 		caddr_t arg;
166 	} */ *uap;
167 {
168 	struct mount *mp;
169 	struct nameidata nd;
170 	int error;
171 
172 	AUDIT_ARG_CMD(uap->cmd);
173 	AUDIT_ARG_UID(uap->uid);
174 	if (!prison_allow(td->td_ucred, PR_ALLOW_QUOTAS))
175 		return (EPERM);
176 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
177 	    uap->path, td);
178 	if ((error = namei(&nd)) != 0)
179 		return (error);
180 	NDFREE(&nd, NDF_ONLY_PNBUF);
181 	mp = nd.ni_vp->v_mount;
182 	vfs_ref(mp);
183 	vput(nd.ni_vp);
184 	error = vfs_busy(mp, 0);
185 	vfs_rel(mp);
186 	if (error != 0)
187 		return (error);
188 	error = VFS_QUOTACTL(mp, uap->cmd, uap->uid, uap->arg);
189 
190 	/*
191 	 * Since quota on operation typically needs to open quota
192 	 * file, the Q_QUOTAON handler needs to unbusy the mount point
193 	 * before calling into namei.  Otherwise, unmount might be
194 	 * started between two vfs_busy() invocations (first is our,
195 	 * second is from mount point cross-walk code in lookup()),
196 	 * causing deadlock.
197 	 *
198 	 * Require that Q_QUOTAON handles the vfs_busy() reference on
199 	 * its own, always returning with ubusied mount point.
200 	 */
201 	if ((uap->cmd >> SUBCMDSHIFT) != Q_QUOTAON)
202 		vfs_unbusy(mp);
203 	return (error);
204 }
205 
206 /*
207  * Used by statfs conversion routines to scale the block size up if
208  * necessary so that all of the block counts are <= 'max_size'.  Note
209  * that 'max_size' should be a bitmask, i.e. 2^n - 1 for some non-zero
210  * value of 'n'.
211  */
212 void
213 statfs_scale_blocks(struct statfs *sf, long max_size)
214 {
215 	uint64_t count;
216 	int shift;
217 
218 	KASSERT(powerof2(max_size + 1), ("%s: invalid max_size", __func__));
219 
220 	/*
221 	 * Attempt to scale the block counts to give a more accurate
222 	 * overview to userland of the ratio of free space to used
223 	 * space.  To do this, find the largest block count and compute
224 	 * a divisor that lets it fit into a signed integer <= max_size.
225 	 */
226 	if (sf->f_bavail < 0)
227 		count = -sf->f_bavail;
228 	else
229 		count = sf->f_bavail;
230 	count = MAX(sf->f_blocks, MAX(sf->f_bfree, count));
231 	if (count <= max_size)
232 		return;
233 
234 	count >>= flsl(max_size);
235 	shift = 0;
236 	while (count > 0) {
237 		shift++;
238 		count >>=1;
239 	}
240 
241 	sf->f_bsize <<= shift;
242 	sf->f_blocks >>= shift;
243 	sf->f_bfree >>= shift;
244 	sf->f_bavail >>= shift;
245 }
246 
247 /*
248  * Get filesystem statistics.
249  */
250 #ifndef _SYS_SYSPROTO_H_
251 struct statfs_args {
252 	char *path;
253 	struct statfs *buf;
254 };
255 #endif
256 int
257 sys_statfs(td, uap)
258 	struct thread *td;
259 	register struct statfs_args /* {
260 		char *path;
261 		struct statfs *buf;
262 	} */ *uap;
263 {
264 	struct statfs sf;
265 	int error;
266 
267 	error = kern_statfs(td, uap->path, UIO_USERSPACE, &sf);
268 	if (error == 0)
269 		error = copyout(&sf, uap->buf, sizeof(sf));
270 	return (error);
271 }
272 
273 int
274 kern_statfs(struct thread *td, char *path, enum uio_seg pathseg,
275     struct statfs *buf)
276 {
277 	struct mount *mp;
278 	struct statfs *sp, sb;
279 	struct nameidata nd;
280 	int error;
281 
282 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
283 	    pathseg, path, td);
284 	error = namei(&nd);
285 	if (error != 0)
286 		return (error);
287 	mp = nd.ni_vp->v_mount;
288 	vfs_ref(mp);
289 	NDFREE(&nd, NDF_ONLY_PNBUF);
290 	vput(nd.ni_vp);
291 	error = vfs_busy(mp, 0);
292 	vfs_rel(mp);
293 	if (error != 0)
294 		return (error);
295 #ifdef MAC
296 	error = mac_mount_check_stat(td->td_ucred, mp);
297 	if (error != 0)
298 		goto out;
299 #endif
300 	/*
301 	 * Set these in case the underlying filesystem fails to do so.
302 	 */
303 	sp = &mp->mnt_stat;
304 	sp->f_version = STATFS_VERSION;
305 	sp->f_namemax = NAME_MAX;
306 	sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
307 	error = VFS_STATFS(mp, sp);
308 	if (error != 0)
309 		goto out;
310 	if (priv_check(td, PRIV_VFS_GENERATION)) {
311 		bcopy(sp, &sb, sizeof(sb));
312 		sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
313 		prison_enforce_statfs(td->td_ucred, mp, &sb);
314 		sp = &sb;
315 	}
316 	*buf = *sp;
317 out:
318 	vfs_unbusy(mp);
319 	return (error);
320 }
321 
322 /*
323  * Get filesystem statistics.
324  */
325 #ifndef _SYS_SYSPROTO_H_
326 struct fstatfs_args {
327 	int fd;
328 	struct statfs *buf;
329 };
330 #endif
331 int
332 sys_fstatfs(td, uap)
333 	struct thread *td;
334 	register struct fstatfs_args /* {
335 		int fd;
336 		struct statfs *buf;
337 	} */ *uap;
338 {
339 	struct statfs sf;
340 	int error;
341 
342 	error = kern_fstatfs(td, uap->fd, &sf);
343 	if (error == 0)
344 		error = copyout(&sf, uap->buf, sizeof(sf));
345 	return (error);
346 }
347 
348 int
349 kern_fstatfs(struct thread *td, int fd, struct statfs *buf)
350 {
351 	struct file *fp;
352 	struct mount *mp;
353 	struct statfs *sp, sb;
354 	struct vnode *vp;
355 	cap_rights_t rights;
356 	int error;
357 
358 	AUDIT_ARG_FD(fd);
359 	error = getvnode(td, fd, cap_rights_init(&rights, CAP_FSTATFS), &fp);
360 	if (error != 0)
361 		return (error);
362 	vp = fp->f_vnode;
363 	vn_lock(vp, LK_SHARED | LK_RETRY);
364 #ifdef AUDIT
365 	AUDIT_ARG_VNODE1(vp);
366 #endif
367 	mp = vp->v_mount;
368 	if (mp)
369 		vfs_ref(mp);
370 	VOP_UNLOCK(vp, 0);
371 	fdrop(fp, td);
372 	if (mp == NULL) {
373 		error = EBADF;
374 		goto out;
375 	}
376 	error = vfs_busy(mp, 0);
377 	vfs_rel(mp);
378 	if (error != 0)
379 		return (error);
380 #ifdef MAC
381 	error = mac_mount_check_stat(td->td_ucred, mp);
382 	if (error != 0)
383 		goto out;
384 #endif
385 	/*
386 	 * Set these in case the underlying filesystem fails to do so.
387 	 */
388 	sp = &mp->mnt_stat;
389 	sp->f_version = STATFS_VERSION;
390 	sp->f_namemax = NAME_MAX;
391 	sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
392 	error = VFS_STATFS(mp, sp);
393 	if (error != 0)
394 		goto out;
395 	if (priv_check(td, PRIV_VFS_GENERATION)) {
396 		bcopy(sp, &sb, sizeof(sb));
397 		sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
398 		prison_enforce_statfs(td->td_ucred, mp, &sb);
399 		sp = &sb;
400 	}
401 	*buf = *sp;
402 out:
403 	if (mp)
404 		vfs_unbusy(mp);
405 	return (error);
406 }
407 
408 /*
409  * Get statistics on all filesystems.
410  */
411 #ifndef _SYS_SYSPROTO_H_
412 struct getfsstat_args {
413 	struct statfs *buf;
414 	long bufsize;
415 	int flags;
416 };
417 #endif
418 int
419 sys_getfsstat(td, uap)
420 	struct thread *td;
421 	register struct getfsstat_args /* {
422 		struct statfs *buf;
423 		long bufsize;
424 		int flags;
425 	} */ *uap;
426 {
427 	size_t count;
428 	int error;
429 
430 	if (uap->bufsize < 0 || uap->bufsize > SIZE_MAX)
431 		return (EINVAL);
432 	error = kern_getfsstat(td, &uap->buf, uap->bufsize, &count,
433 	    UIO_USERSPACE, uap->flags);
434 	if (error == 0)
435 		td->td_retval[0] = count;
436 	return (error);
437 }
438 
439 /*
440  * If (bufsize > 0 && bufseg == UIO_SYSSPACE)
441  *	The caller is responsible for freeing memory which will be allocated
442  *	in '*buf'.
443  */
444 int
445 kern_getfsstat(struct thread *td, struct statfs **buf, size_t bufsize,
446     size_t *countp, enum uio_seg bufseg, int flags)
447 {
448 	struct mount *mp, *nmp;
449 	struct statfs *sfsp, *sp, sb;
450 	size_t count, maxcount;
451 	int error;
452 
453 	maxcount = bufsize / sizeof(struct statfs);
454 	if (bufsize == 0)
455 		sfsp = NULL;
456 	else if (bufseg == UIO_USERSPACE)
457 		sfsp = *buf;
458 	else /* if (bufseg == UIO_SYSSPACE) */ {
459 		count = 0;
460 		mtx_lock(&mountlist_mtx);
461 		TAILQ_FOREACH(mp, &mountlist, mnt_list) {
462 			count++;
463 		}
464 		mtx_unlock(&mountlist_mtx);
465 		if (maxcount > count)
466 			maxcount = count;
467 		sfsp = *buf = malloc(maxcount * sizeof(struct statfs), M_TEMP,
468 		    M_WAITOK);
469 	}
470 	count = 0;
471 	mtx_lock(&mountlist_mtx);
472 	for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
473 		if (prison_canseemount(td->td_ucred, mp) != 0) {
474 			nmp = TAILQ_NEXT(mp, mnt_list);
475 			continue;
476 		}
477 #ifdef MAC
478 		if (mac_mount_check_stat(td->td_ucred, mp) != 0) {
479 			nmp = TAILQ_NEXT(mp, mnt_list);
480 			continue;
481 		}
482 #endif
483 		if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK)) {
484 			nmp = TAILQ_NEXT(mp, mnt_list);
485 			continue;
486 		}
487 		if (sfsp && count < maxcount) {
488 			sp = &mp->mnt_stat;
489 			/*
490 			 * Set these in case the underlying filesystem
491 			 * fails to do so.
492 			 */
493 			sp->f_version = STATFS_VERSION;
494 			sp->f_namemax = NAME_MAX;
495 			sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
496 			/*
497 			 * If MNT_NOWAIT or MNT_LAZY is specified, do not
498 			 * refresh the fsstat cache. MNT_NOWAIT or MNT_LAZY
499 			 * overrides MNT_WAIT.
500 			 */
501 			if (((flags & (MNT_LAZY|MNT_NOWAIT)) == 0 ||
502 			    (flags & MNT_WAIT)) &&
503 			    (error = VFS_STATFS(mp, sp))) {
504 				mtx_lock(&mountlist_mtx);
505 				nmp = TAILQ_NEXT(mp, mnt_list);
506 				vfs_unbusy(mp);
507 				continue;
508 			}
509 			if (priv_check(td, PRIV_VFS_GENERATION)) {
510 				bcopy(sp, &sb, sizeof(sb));
511 				sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
512 				prison_enforce_statfs(td->td_ucred, mp, &sb);
513 				sp = &sb;
514 			}
515 			if (bufseg == UIO_SYSSPACE)
516 				bcopy(sp, sfsp, sizeof(*sp));
517 			else /* if (bufseg == UIO_USERSPACE) */ {
518 				error = copyout(sp, sfsp, sizeof(*sp));
519 				if (error != 0) {
520 					vfs_unbusy(mp);
521 					return (error);
522 				}
523 			}
524 			sfsp++;
525 		}
526 		count++;
527 		mtx_lock(&mountlist_mtx);
528 		nmp = TAILQ_NEXT(mp, mnt_list);
529 		vfs_unbusy(mp);
530 	}
531 	mtx_unlock(&mountlist_mtx);
532 	if (sfsp && count > maxcount)
533 		*countp = maxcount;
534 	else
535 		*countp = count;
536 	return (0);
537 }
538 
539 #ifdef COMPAT_FREEBSD4
540 /*
541  * Get old format filesystem statistics.
542  */
543 static void cvtstatfs(struct statfs *, struct ostatfs *);
544 
545 #ifndef _SYS_SYSPROTO_H_
546 struct freebsd4_statfs_args {
547 	char *path;
548 	struct ostatfs *buf;
549 };
550 #endif
551 int
552 freebsd4_statfs(td, uap)
553 	struct thread *td;
554 	struct freebsd4_statfs_args /* {
555 		char *path;
556 		struct ostatfs *buf;
557 	} */ *uap;
558 {
559 	struct ostatfs osb;
560 	struct statfs sf;
561 	int error;
562 
563 	error = kern_statfs(td, uap->path, UIO_USERSPACE, &sf);
564 	if (error != 0)
565 		return (error);
566 	cvtstatfs(&sf, &osb);
567 	return (copyout(&osb, uap->buf, sizeof(osb)));
568 }
569 
570 /*
571  * Get filesystem statistics.
572  */
573 #ifndef _SYS_SYSPROTO_H_
574 struct freebsd4_fstatfs_args {
575 	int fd;
576 	struct ostatfs *buf;
577 };
578 #endif
579 int
580 freebsd4_fstatfs(td, uap)
581 	struct thread *td;
582 	struct freebsd4_fstatfs_args /* {
583 		int fd;
584 		struct ostatfs *buf;
585 	} */ *uap;
586 {
587 	struct ostatfs osb;
588 	struct statfs sf;
589 	int error;
590 
591 	error = kern_fstatfs(td, uap->fd, &sf);
592 	if (error != 0)
593 		return (error);
594 	cvtstatfs(&sf, &osb);
595 	return (copyout(&osb, uap->buf, sizeof(osb)));
596 }
597 
598 /*
599  * Get statistics on all filesystems.
600  */
601 #ifndef _SYS_SYSPROTO_H_
602 struct freebsd4_getfsstat_args {
603 	struct ostatfs *buf;
604 	long bufsize;
605 	int flags;
606 };
607 #endif
608 int
609 freebsd4_getfsstat(td, uap)
610 	struct thread *td;
611 	register struct freebsd4_getfsstat_args /* {
612 		struct ostatfs *buf;
613 		long bufsize;
614 		int flags;
615 	} */ *uap;
616 {
617 	struct statfs *buf, *sp;
618 	struct ostatfs osb;
619 	size_t count, size;
620 	int error;
621 
622 	if (uap->bufsize < 0)
623 		return (EINVAL);
624 	count = uap->bufsize / sizeof(struct ostatfs);
625 	if (count > SIZE_MAX / sizeof(struct statfs))
626 		return (EINVAL);
627 	size = count * sizeof(struct statfs);
628 	error = kern_getfsstat(td, &buf, size, &count, UIO_SYSSPACE,
629 	    uap->flags);
630 	td->td_retval[0] = count;
631 	if (size != 0) {
632 		sp = buf;
633 		while (count != 0 && error == 0) {
634 			cvtstatfs(sp, &osb);
635 			error = copyout(&osb, uap->buf, sizeof(osb));
636 			sp++;
637 			uap->buf++;
638 			count--;
639 		}
640 		free(buf, M_TEMP);
641 	}
642 	return (error);
643 }
644 
645 /*
646  * Implement fstatfs() for (NFS) file handles.
647  */
648 #ifndef _SYS_SYSPROTO_H_
649 struct freebsd4_fhstatfs_args {
650 	struct fhandle *u_fhp;
651 	struct ostatfs *buf;
652 };
653 #endif
654 int
655 freebsd4_fhstatfs(td, uap)
656 	struct thread *td;
657 	struct freebsd4_fhstatfs_args /* {
658 		struct fhandle *u_fhp;
659 		struct ostatfs *buf;
660 	} */ *uap;
661 {
662 	struct ostatfs osb;
663 	struct statfs sf;
664 	fhandle_t fh;
665 	int error;
666 
667 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
668 	if (error != 0)
669 		return (error);
670 	error = kern_fhstatfs(td, fh, &sf);
671 	if (error != 0)
672 		return (error);
673 	cvtstatfs(&sf, &osb);
674 	return (copyout(&osb, uap->buf, sizeof(osb)));
675 }
676 
677 /*
678  * Convert a new format statfs structure to an old format statfs structure.
679  */
680 static void
681 cvtstatfs(nsp, osp)
682 	struct statfs *nsp;
683 	struct ostatfs *osp;
684 {
685 
686 	statfs_scale_blocks(nsp, LONG_MAX);
687 	bzero(osp, sizeof(*osp));
688 	osp->f_bsize = nsp->f_bsize;
689 	osp->f_iosize = MIN(nsp->f_iosize, LONG_MAX);
690 	osp->f_blocks = nsp->f_blocks;
691 	osp->f_bfree = nsp->f_bfree;
692 	osp->f_bavail = nsp->f_bavail;
693 	osp->f_files = MIN(nsp->f_files, LONG_MAX);
694 	osp->f_ffree = MIN(nsp->f_ffree, LONG_MAX);
695 	osp->f_owner = nsp->f_owner;
696 	osp->f_type = nsp->f_type;
697 	osp->f_flags = nsp->f_flags;
698 	osp->f_syncwrites = MIN(nsp->f_syncwrites, LONG_MAX);
699 	osp->f_asyncwrites = MIN(nsp->f_asyncwrites, LONG_MAX);
700 	osp->f_syncreads = MIN(nsp->f_syncreads, LONG_MAX);
701 	osp->f_asyncreads = MIN(nsp->f_asyncreads, LONG_MAX);
702 	strlcpy(osp->f_fstypename, nsp->f_fstypename,
703 	    MIN(MFSNAMELEN, OMFSNAMELEN));
704 	strlcpy(osp->f_mntonname, nsp->f_mntonname,
705 	    MIN(MNAMELEN, OMNAMELEN));
706 	strlcpy(osp->f_mntfromname, nsp->f_mntfromname,
707 	    MIN(MNAMELEN, OMNAMELEN));
708 	osp->f_fsid = nsp->f_fsid;
709 }
710 #endif /* COMPAT_FREEBSD4 */
711 
712 /*
713  * Change current working directory to a given file descriptor.
714  */
715 #ifndef _SYS_SYSPROTO_H_
716 struct fchdir_args {
717 	int	fd;
718 };
719 #endif
720 int
721 sys_fchdir(td, uap)
722 	struct thread *td;
723 	struct fchdir_args /* {
724 		int fd;
725 	} */ *uap;
726 {
727 	struct vnode *vp, *tdp;
728 	struct mount *mp;
729 	struct file *fp;
730 	cap_rights_t rights;
731 	int error;
732 
733 	AUDIT_ARG_FD(uap->fd);
734 	error = getvnode(td, uap->fd, cap_rights_init(&rights, CAP_FCHDIR),
735 	    &fp);
736 	if (error != 0)
737 		return (error);
738 	vp = fp->f_vnode;
739 	VREF(vp);
740 	fdrop(fp, td);
741 	vn_lock(vp, LK_SHARED | LK_RETRY);
742 	AUDIT_ARG_VNODE1(vp);
743 	error = change_dir(vp, td);
744 	while (!error && (mp = vp->v_mountedhere) != NULL) {
745 		if (vfs_busy(mp, 0))
746 			continue;
747 		error = VFS_ROOT(mp, LK_SHARED, &tdp);
748 		vfs_unbusy(mp);
749 		if (error != 0)
750 			break;
751 		vput(vp);
752 		vp = tdp;
753 	}
754 	if (error != 0) {
755 		vput(vp);
756 		return (error);
757 	}
758 	VOP_UNLOCK(vp, 0);
759 	pwd_chdir(td, vp);
760 	return (0);
761 }
762 
763 /*
764  * Change current working directory (``.'').
765  */
766 #ifndef _SYS_SYSPROTO_H_
767 struct chdir_args {
768 	char	*path;
769 };
770 #endif
771 int
772 sys_chdir(td, uap)
773 	struct thread *td;
774 	struct chdir_args /* {
775 		char *path;
776 	} */ *uap;
777 {
778 
779 	return (kern_chdir(td, uap->path, UIO_USERSPACE));
780 }
781 
782 int
783 kern_chdir(struct thread *td, char *path, enum uio_seg pathseg)
784 {
785 	struct nameidata nd;
786 	int error;
787 
788 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
789 	    pathseg, path, td);
790 	if ((error = namei(&nd)) != 0)
791 		return (error);
792 	if ((error = change_dir(nd.ni_vp, td)) != 0) {
793 		vput(nd.ni_vp);
794 		NDFREE(&nd, NDF_ONLY_PNBUF);
795 		return (error);
796 	}
797 	VOP_UNLOCK(nd.ni_vp, 0);
798 	NDFREE(&nd, NDF_ONLY_PNBUF);
799 	pwd_chdir(td, nd.ni_vp);
800 	return (0);
801 }
802 
803 /*
804  * Change notion of root (``/'') directory.
805  */
806 #ifndef _SYS_SYSPROTO_H_
807 struct chroot_args {
808 	char	*path;
809 };
810 #endif
811 int
812 sys_chroot(td, uap)
813 	struct thread *td;
814 	struct chroot_args /* {
815 		char *path;
816 	} */ *uap;
817 {
818 	struct nameidata nd;
819 	int error;
820 
821 	error = priv_check(td, PRIV_VFS_CHROOT);
822 	if (error != 0)
823 		return (error);
824 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
825 	    UIO_USERSPACE, uap->path, td);
826 	error = namei(&nd);
827 	if (error != 0)
828 		goto error;
829 	error = change_dir(nd.ni_vp, td);
830 	if (error != 0)
831 		goto e_vunlock;
832 #ifdef MAC
833 	error = mac_vnode_check_chroot(td->td_ucred, nd.ni_vp);
834 	if (error != 0)
835 		goto e_vunlock;
836 #endif
837 	VOP_UNLOCK(nd.ni_vp, 0);
838 	error = pwd_chroot(td, nd.ni_vp);
839 	vrele(nd.ni_vp);
840 	NDFREE(&nd, NDF_ONLY_PNBUF);
841 	return (error);
842 e_vunlock:
843 	vput(nd.ni_vp);
844 error:
845 	NDFREE(&nd, NDF_ONLY_PNBUF);
846 	return (error);
847 }
848 
849 /*
850  * Common routine for chroot and chdir.  Callers must provide a locked vnode
851  * instance.
852  */
853 int
854 change_dir(vp, td)
855 	struct vnode *vp;
856 	struct thread *td;
857 {
858 #ifdef MAC
859 	int error;
860 #endif
861 
862 	ASSERT_VOP_LOCKED(vp, "change_dir(): vp not locked");
863 	if (vp->v_type != VDIR)
864 		return (ENOTDIR);
865 #ifdef MAC
866 	error = mac_vnode_check_chdir(td->td_ucred, vp);
867 	if (error != 0)
868 		return (error);
869 #endif
870 	return (VOP_ACCESS(vp, VEXEC, td->td_ucred, td));
871 }
872 
873 static __inline void
874 flags_to_rights(int flags, cap_rights_t *rightsp)
875 {
876 
877 	if (flags & O_EXEC) {
878 		cap_rights_set(rightsp, CAP_FEXECVE);
879 	} else {
880 		switch ((flags & O_ACCMODE)) {
881 		case O_RDONLY:
882 			cap_rights_set(rightsp, CAP_READ);
883 			break;
884 		case O_RDWR:
885 			cap_rights_set(rightsp, CAP_READ);
886 			/* FALLTHROUGH */
887 		case O_WRONLY:
888 			cap_rights_set(rightsp, CAP_WRITE);
889 			if (!(flags & (O_APPEND | O_TRUNC)))
890 				cap_rights_set(rightsp, CAP_SEEK);
891 			break;
892 		}
893 	}
894 
895 	if (flags & O_CREAT)
896 		cap_rights_set(rightsp, CAP_CREATE);
897 
898 	if (flags & O_TRUNC)
899 		cap_rights_set(rightsp, CAP_FTRUNCATE);
900 
901 	if (flags & (O_SYNC | O_FSYNC))
902 		cap_rights_set(rightsp, CAP_FSYNC);
903 
904 	if (flags & (O_EXLOCK | O_SHLOCK))
905 		cap_rights_set(rightsp, CAP_FLOCK);
906 }
907 
908 /*
909  * Check permissions, allocate an open file structure, and call the device
910  * open routine if any.
911  */
912 #ifndef _SYS_SYSPROTO_H_
913 struct open_args {
914 	char	*path;
915 	int	flags;
916 	int	mode;
917 };
918 #endif
919 int
920 sys_open(td, uap)
921 	struct thread *td;
922 	register struct open_args /* {
923 		char *path;
924 		int flags;
925 		int mode;
926 	} */ *uap;
927 {
928 
929 	return (kern_openat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
930 	    uap->flags, uap->mode));
931 }
932 
933 #ifndef _SYS_SYSPROTO_H_
934 struct openat_args {
935 	int	fd;
936 	char	*path;
937 	int	flag;
938 	int	mode;
939 };
940 #endif
941 int
942 sys_openat(struct thread *td, struct openat_args *uap)
943 {
944 
945 	return (kern_openat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
946 	    uap->mode));
947 }
948 
949 int
950 kern_openat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
951     int flags, int mode)
952 {
953 	struct proc *p = td->td_proc;
954 	struct filedesc *fdp = p->p_fd;
955 	struct file *fp;
956 	struct vnode *vp;
957 	struct nameidata nd;
958 	cap_rights_t rights;
959 	int cmode, error, indx;
960 
961 	indx = -1;
962 
963 	AUDIT_ARG_FFLAGS(flags);
964 	AUDIT_ARG_MODE(mode);
965 	/* XXX: audit dirfd */
966 	cap_rights_init(&rights, CAP_LOOKUP);
967 	flags_to_rights(flags, &rights);
968 	/*
969 	 * Only one of the O_EXEC, O_RDONLY, O_WRONLY and O_RDWR flags
970 	 * may be specified.
971 	 */
972 	if (flags & O_EXEC) {
973 		if (flags & O_ACCMODE)
974 			return (EINVAL);
975 	} else if ((flags & O_ACCMODE) == O_ACCMODE) {
976 		return (EINVAL);
977 	} else {
978 		flags = FFLAGS(flags);
979 	}
980 
981 	/*
982 	 * Allocate a file structure. The descriptor to reference it
983 	 * is allocated and set by finstall() below.
984 	 */
985 	error = falloc_noinstall(td, &fp);
986 	if (error != 0)
987 		return (error);
988 	/*
989 	 * An extra reference on `fp' has been held for us by
990 	 * falloc_noinstall().
991 	 */
992 	/* Set the flags early so the finit in devfs can pick them up. */
993 	fp->f_flag = flags & FMASK;
994 	cmode = ((mode & ~fdp->fd_cmask) & ALLPERMS) & ~S_ISTXT;
995 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, fd,
996 	    &rights, td);
997 	td->td_dupfd = -1;		/* XXX check for fdopen */
998 	error = vn_open(&nd, &flags, cmode, fp);
999 	if (error != 0) {
1000 		/*
1001 		 * If the vn_open replaced the method vector, something
1002 		 * wonderous happened deep below and we just pass it up
1003 		 * pretending we know what we do.
1004 		 */
1005 		if (error == ENXIO && fp->f_ops != &badfileops)
1006 			goto success;
1007 
1008 		/*
1009 		 * Handle special fdopen() case. bleh.
1010 		 *
1011 		 * Don't do this for relative (capability) lookups; we don't
1012 		 * understand exactly what would happen, and we don't think
1013 		 * that it ever should.
1014 		 */
1015 		if (nd.ni_strictrelative == 0 &&
1016 		    (error == ENODEV || error == ENXIO) &&
1017 		    td->td_dupfd >= 0) {
1018 			error = dupfdopen(td, fdp, td->td_dupfd, flags, error,
1019 			    &indx);
1020 			if (error == 0)
1021 				goto success;
1022 		}
1023 
1024 		goto bad;
1025 	}
1026 	td->td_dupfd = 0;
1027 	NDFREE(&nd, NDF_ONLY_PNBUF);
1028 	vp = nd.ni_vp;
1029 
1030 	/*
1031 	 * Store the vnode, for any f_type. Typically, the vnode use
1032 	 * count is decremented by direct call to vn_closefile() for
1033 	 * files that switched type in the cdevsw fdopen() method.
1034 	 */
1035 	fp->f_vnode = vp;
1036 	/*
1037 	 * If the file wasn't claimed by devfs bind it to the normal
1038 	 * vnode operations here.
1039 	 */
1040 	if (fp->f_ops == &badfileops) {
1041 		KASSERT(vp->v_type != VFIFO, ("Unexpected fifo."));
1042 		fp->f_seqcount = 1;
1043 		finit(fp, (flags & FMASK) | (fp->f_flag & FHASLOCK),
1044 		    DTYPE_VNODE, vp, &vnops);
1045 	}
1046 
1047 	VOP_UNLOCK(vp, 0);
1048 	if (flags & O_TRUNC) {
1049 		error = fo_truncate(fp, 0, td->td_ucred, td);
1050 		if (error != 0)
1051 			goto bad;
1052 	}
1053 success:
1054 	/*
1055 	 * If we haven't already installed the FD (for dupfdopen), do so now.
1056 	 */
1057 	if (indx == -1) {
1058 		struct filecaps *fcaps;
1059 
1060 #ifdef CAPABILITIES
1061 		if (nd.ni_strictrelative == 1)
1062 			fcaps = &nd.ni_filecaps;
1063 		else
1064 #endif
1065 			fcaps = NULL;
1066 		error = finstall(td, fp, &indx, flags, fcaps);
1067 		/* On success finstall() consumes fcaps. */
1068 		if (error != 0) {
1069 			filecaps_free(&nd.ni_filecaps);
1070 			goto bad;
1071 		}
1072 	} else {
1073 		filecaps_free(&nd.ni_filecaps);
1074 	}
1075 
1076 	/*
1077 	 * Release our private reference, leaving the one associated with
1078 	 * the descriptor table intact.
1079 	 */
1080 	fdrop(fp, td);
1081 	td->td_retval[0] = indx;
1082 	return (0);
1083 bad:
1084 	KASSERT(indx == -1, ("indx=%d, should be -1", indx));
1085 	fdrop(fp, td);
1086 	return (error);
1087 }
1088 
1089 #ifdef COMPAT_43
1090 /*
1091  * Create a file.
1092  */
1093 #ifndef _SYS_SYSPROTO_H_
1094 struct ocreat_args {
1095 	char	*path;
1096 	int	mode;
1097 };
1098 #endif
1099 int
1100 ocreat(td, uap)
1101 	struct thread *td;
1102 	register struct ocreat_args /* {
1103 		char *path;
1104 		int mode;
1105 	} */ *uap;
1106 {
1107 
1108 	return (kern_openat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1109 	    O_WRONLY | O_CREAT | O_TRUNC, uap->mode));
1110 }
1111 #endif /* COMPAT_43 */
1112 
1113 /*
1114  * Create a special file.
1115  */
1116 #ifndef _SYS_SYSPROTO_H_
1117 struct mknod_args {
1118 	char	*path;
1119 	int	mode;
1120 	int	dev;
1121 };
1122 #endif
1123 int
1124 sys_mknod(td, uap)
1125 	struct thread *td;
1126 	register struct mknod_args /* {
1127 		char *path;
1128 		int mode;
1129 		int dev;
1130 	} */ *uap;
1131 {
1132 
1133 	return (kern_mknodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1134 	    uap->mode, uap->dev));
1135 }
1136 
1137 #ifndef _SYS_SYSPROTO_H_
1138 struct mknodat_args {
1139 	int	fd;
1140 	char	*path;
1141 	mode_t	mode;
1142 	dev_t	dev;
1143 };
1144 #endif
1145 int
1146 sys_mknodat(struct thread *td, struct mknodat_args *uap)
1147 {
1148 
1149 	return (kern_mknodat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode,
1150 	    uap->dev));
1151 }
1152 
1153 int
1154 kern_mknodat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1155     int mode, int dev)
1156 {
1157 	struct vnode *vp;
1158 	struct mount *mp;
1159 	struct vattr vattr;
1160 	struct nameidata nd;
1161 	cap_rights_t rights;
1162 	int error, whiteout = 0;
1163 
1164 	AUDIT_ARG_MODE(mode);
1165 	AUDIT_ARG_DEV(dev);
1166 	switch (mode & S_IFMT) {
1167 	case S_IFCHR:
1168 	case S_IFBLK:
1169 		error = priv_check(td, PRIV_VFS_MKNOD_DEV);
1170 		break;
1171 	case S_IFMT:
1172 		error = priv_check(td, PRIV_VFS_MKNOD_BAD);
1173 		break;
1174 	case S_IFWHT:
1175 		error = priv_check(td, PRIV_VFS_MKNOD_WHT);
1176 		break;
1177 	case S_IFIFO:
1178 		if (dev == 0)
1179 			return (kern_mkfifoat(td, fd, path, pathseg, mode));
1180 		/* FALLTHROUGH */
1181 	default:
1182 		error = EINVAL;
1183 		break;
1184 	}
1185 	if (error != 0)
1186 		return (error);
1187 restart:
1188 	bwillwrite();
1189 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
1190 	    NOCACHE, pathseg, path, fd, cap_rights_init(&rights, CAP_MKNODAT),
1191 	    td);
1192 	if ((error = namei(&nd)) != 0)
1193 		return (error);
1194 	vp = nd.ni_vp;
1195 	if (vp != NULL) {
1196 		NDFREE(&nd, NDF_ONLY_PNBUF);
1197 		if (vp == nd.ni_dvp)
1198 			vrele(nd.ni_dvp);
1199 		else
1200 			vput(nd.ni_dvp);
1201 		vrele(vp);
1202 		return (EEXIST);
1203 	} else {
1204 		VATTR_NULL(&vattr);
1205 		vattr.va_mode = (mode & ALLPERMS) &
1206 		    ~td->td_proc->p_fd->fd_cmask;
1207 		vattr.va_rdev = dev;
1208 		whiteout = 0;
1209 
1210 		switch (mode & S_IFMT) {
1211 		case S_IFMT:	/* used by badsect to flag bad sectors */
1212 			vattr.va_type = VBAD;
1213 			break;
1214 		case S_IFCHR:
1215 			vattr.va_type = VCHR;
1216 			break;
1217 		case S_IFBLK:
1218 			vattr.va_type = VBLK;
1219 			break;
1220 		case S_IFWHT:
1221 			whiteout = 1;
1222 			break;
1223 		default:
1224 			panic("kern_mknod: invalid mode");
1225 		}
1226 	}
1227 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1228 		NDFREE(&nd, NDF_ONLY_PNBUF);
1229 		vput(nd.ni_dvp);
1230 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1231 			return (error);
1232 		goto restart;
1233 	}
1234 #ifdef MAC
1235 	if (error == 0 && !whiteout)
1236 		error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp,
1237 		    &nd.ni_cnd, &vattr);
1238 #endif
1239 	if (error == 0) {
1240 		if (whiteout)
1241 			error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE);
1242 		else {
1243 			error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp,
1244 						&nd.ni_cnd, &vattr);
1245 			if (error == 0)
1246 				vput(nd.ni_vp);
1247 		}
1248 	}
1249 	NDFREE(&nd, NDF_ONLY_PNBUF);
1250 	vput(nd.ni_dvp);
1251 	vn_finished_write(mp);
1252 	return (error);
1253 }
1254 
1255 /*
1256  * Create a named pipe.
1257  */
1258 #ifndef _SYS_SYSPROTO_H_
1259 struct mkfifo_args {
1260 	char	*path;
1261 	int	mode;
1262 };
1263 #endif
1264 int
1265 sys_mkfifo(td, uap)
1266 	struct thread *td;
1267 	register struct mkfifo_args /* {
1268 		char *path;
1269 		int mode;
1270 	} */ *uap;
1271 {
1272 
1273 	return (kern_mkfifoat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1274 	    uap->mode));
1275 }
1276 
1277 #ifndef _SYS_SYSPROTO_H_
1278 struct mkfifoat_args {
1279 	int	fd;
1280 	char	*path;
1281 	mode_t	mode;
1282 };
1283 #endif
1284 int
1285 sys_mkfifoat(struct thread *td, struct mkfifoat_args *uap)
1286 {
1287 
1288 	return (kern_mkfifoat(td, uap->fd, uap->path, UIO_USERSPACE,
1289 	    uap->mode));
1290 }
1291 
1292 int
1293 kern_mkfifoat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1294     int mode)
1295 {
1296 	struct mount *mp;
1297 	struct vattr vattr;
1298 	struct nameidata nd;
1299 	cap_rights_t rights;
1300 	int error;
1301 
1302 	AUDIT_ARG_MODE(mode);
1303 restart:
1304 	bwillwrite();
1305 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
1306 	    NOCACHE, pathseg, path, fd, cap_rights_init(&rights, CAP_MKFIFOAT),
1307 	    td);
1308 	if ((error = namei(&nd)) != 0)
1309 		return (error);
1310 	if (nd.ni_vp != NULL) {
1311 		NDFREE(&nd, NDF_ONLY_PNBUF);
1312 		if (nd.ni_vp == nd.ni_dvp)
1313 			vrele(nd.ni_dvp);
1314 		else
1315 			vput(nd.ni_dvp);
1316 		vrele(nd.ni_vp);
1317 		return (EEXIST);
1318 	}
1319 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1320 		NDFREE(&nd, NDF_ONLY_PNBUF);
1321 		vput(nd.ni_dvp);
1322 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1323 			return (error);
1324 		goto restart;
1325 	}
1326 	VATTR_NULL(&vattr);
1327 	vattr.va_type = VFIFO;
1328 	vattr.va_mode = (mode & ALLPERMS) & ~td->td_proc->p_fd->fd_cmask;
1329 #ifdef MAC
1330 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1331 	    &vattr);
1332 	if (error != 0)
1333 		goto out;
1334 #endif
1335 	error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
1336 	if (error == 0)
1337 		vput(nd.ni_vp);
1338 #ifdef MAC
1339 out:
1340 #endif
1341 	vput(nd.ni_dvp);
1342 	vn_finished_write(mp);
1343 	NDFREE(&nd, NDF_ONLY_PNBUF);
1344 	return (error);
1345 }
1346 
1347 /*
1348  * Make a hard file link.
1349  */
1350 #ifndef _SYS_SYSPROTO_H_
1351 struct link_args {
1352 	char	*path;
1353 	char	*link;
1354 };
1355 #endif
1356 int
1357 sys_link(td, uap)
1358 	struct thread *td;
1359 	register struct link_args /* {
1360 		char *path;
1361 		char *link;
1362 	} */ *uap;
1363 {
1364 
1365 	return (kern_linkat(td, AT_FDCWD, AT_FDCWD, uap->path, uap->link,
1366 	    UIO_USERSPACE, FOLLOW));
1367 }
1368 
1369 #ifndef _SYS_SYSPROTO_H_
1370 struct linkat_args {
1371 	int	fd1;
1372 	char	*path1;
1373 	int	fd2;
1374 	char	*path2;
1375 	int	flag;
1376 };
1377 #endif
1378 int
1379 sys_linkat(struct thread *td, struct linkat_args *uap)
1380 {
1381 	int flag;
1382 
1383 	flag = uap->flag;
1384 	if (flag & ~AT_SYMLINK_FOLLOW)
1385 		return (EINVAL);
1386 
1387 	return (kern_linkat(td, uap->fd1, uap->fd2, uap->path1, uap->path2,
1388 	    UIO_USERSPACE, (flag & AT_SYMLINK_FOLLOW) ? FOLLOW : NOFOLLOW));
1389 }
1390 
1391 int hardlink_check_uid = 0;
1392 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_uid, CTLFLAG_RW,
1393     &hardlink_check_uid, 0,
1394     "Unprivileged processes cannot create hard links to files owned by other "
1395     "users");
1396 static int hardlink_check_gid = 0;
1397 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_gid, CTLFLAG_RW,
1398     &hardlink_check_gid, 0,
1399     "Unprivileged processes cannot create hard links to files owned by other "
1400     "groups");
1401 
1402 static int
1403 can_hardlink(struct vnode *vp, struct ucred *cred)
1404 {
1405 	struct vattr va;
1406 	int error;
1407 
1408 	if (!hardlink_check_uid && !hardlink_check_gid)
1409 		return (0);
1410 
1411 	error = VOP_GETATTR(vp, &va, cred);
1412 	if (error != 0)
1413 		return (error);
1414 
1415 	if (hardlink_check_uid && cred->cr_uid != va.va_uid) {
1416 		error = priv_check_cred(cred, PRIV_VFS_LINK, 0);
1417 		if (error != 0)
1418 			return (error);
1419 	}
1420 
1421 	if (hardlink_check_gid && !groupmember(va.va_gid, cred)) {
1422 		error = priv_check_cred(cred, PRIV_VFS_LINK, 0);
1423 		if (error != 0)
1424 			return (error);
1425 	}
1426 
1427 	return (0);
1428 }
1429 
1430 int
1431 kern_linkat(struct thread *td, int fd1, int fd2, char *path1, char *path2,
1432     enum uio_seg segflg, int follow)
1433 {
1434 	struct vnode *vp;
1435 	struct mount *mp;
1436 	struct nameidata nd;
1437 	cap_rights_t rights;
1438 	int error;
1439 
1440 again:
1441 	bwillwrite();
1442 	NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, segflg, path1, fd1,
1443 	    cap_rights_init(&rights, CAP_LINKAT_SOURCE), td);
1444 
1445 	if ((error = namei(&nd)) != 0)
1446 		return (error);
1447 	NDFREE(&nd, NDF_ONLY_PNBUF);
1448 	vp = nd.ni_vp;
1449 	if (vp->v_type == VDIR) {
1450 		vrele(vp);
1451 		return (EPERM);		/* POSIX */
1452 	}
1453 	NDINIT_ATRIGHTS(&nd, CREATE,
1454 	    LOCKPARENT | SAVENAME | AUDITVNODE2 | NOCACHE, segflg, path2, fd2,
1455 	    cap_rights_init(&rights, CAP_LINKAT_TARGET), td);
1456 	if ((error = namei(&nd)) == 0) {
1457 		if (nd.ni_vp != NULL) {
1458 			NDFREE(&nd, NDF_ONLY_PNBUF);
1459 			if (nd.ni_dvp == nd.ni_vp)
1460 				vrele(nd.ni_dvp);
1461 			else
1462 				vput(nd.ni_dvp);
1463 			vrele(nd.ni_vp);
1464 			vrele(vp);
1465 			return (EEXIST);
1466 		} else if (nd.ni_dvp->v_mount != vp->v_mount) {
1467 			/*
1468 			 * Cross-device link.  No need to recheck
1469 			 * vp->v_type, since it cannot change, except
1470 			 * to VBAD.
1471 			 */
1472 			NDFREE(&nd, NDF_ONLY_PNBUF);
1473 			vput(nd.ni_dvp);
1474 			vrele(vp);
1475 			return (EXDEV);
1476 		} else if ((error = vn_lock(vp, LK_EXCLUSIVE)) == 0) {
1477 			error = can_hardlink(vp, td->td_ucred);
1478 #ifdef MAC
1479 			if (error == 0)
1480 				error = mac_vnode_check_link(td->td_ucred,
1481 				    nd.ni_dvp, vp, &nd.ni_cnd);
1482 #endif
1483 			if (error != 0) {
1484 				vput(vp);
1485 				vput(nd.ni_dvp);
1486 				NDFREE(&nd, NDF_ONLY_PNBUF);
1487 				return (error);
1488 			}
1489 			error = vn_start_write(vp, &mp, V_NOWAIT);
1490 			if (error != 0) {
1491 				vput(vp);
1492 				vput(nd.ni_dvp);
1493 				NDFREE(&nd, NDF_ONLY_PNBUF);
1494 				error = vn_start_write(NULL, &mp,
1495 				    V_XSLEEP | PCATCH);
1496 				if (error != 0)
1497 					return (error);
1498 				goto again;
1499 			}
1500 			error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
1501 			VOP_UNLOCK(vp, 0);
1502 			vput(nd.ni_dvp);
1503 			vn_finished_write(mp);
1504 			NDFREE(&nd, NDF_ONLY_PNBUF);
1505 		} else {
1506 			vput(nd.ni_dvp);
1507 			NDFREE(&nd, NDF_ONLY_PNBUF);
1508 			vrele(vp);
1509 			goto again;
1510 		}
1511 	}
1512 	vrele(vp);
1513 	return (error);
1514 }
1515 
1516 /*
1517  * Make a symbolic link.
1518  */
1519 #ifndef _SYS_SYSPROTO_H_
1520 struct symlink_args {
1521 	char	*path;
1522 	char	*link;
1523 };
1524 #endif
1525 int
1526 sys_symlink(td, uap)
1527 	struct thread *td;
1528 	register struct symlink_args /* {
1529 		char *path;
1530 		char *link;
1531 	} */ *uap;
1532 {
1533 
1534 	return (kern_symlinkat(td, uap->path, AT_FDCWD, uap->link,
1535 	    UIO_USERSPACE));
1536 }
1537 
1538 #ifndef _SYS_SYSPROTO_H_
1539 struct symlinkat_args {
1540 	char	*path;
1541 	int	fd;
1542 	char	*path2;
1543 };
1544 #endif
1545 int
1546 sys_symlinkat(struct thread *td, struct symlinkat_args *uap)
1547 {
1548 
1549 	return (kern_symlinkat(td, uap->path1, uap->fd, uap->path2,
1550 	    UIO_USERSPACE));
1551 }
1552 
1553 int
1554 kern_symlinkat(struct thread *td, char *path1, int fd, char *path2,
1555     enum uio_seg segflg)
1556 {
1557 	struct mount *mp;
1558 	struct vattr vattr;
1559 	char *syspath;
1560 	struct nameidata nd;
1561 	int error;
1562 	cap_rights_t rights;
1563 
1564 	if (segflg == UIO_SYSSPACE) {
1565 		syspath = path1;
1566 	} else {
1567 		syspath = uma_zalloc(namei_zone, M_WAITOK);
1568 		if ((error = copyinstr(path1, syspath, MAXPATHLEN, NULL)) != 0)
1569 			goto out;
1570 	}
1571 	AUDIT_ARG_TEXT(syspath);
1572 restart:
1573 	bwillwrite();
1574 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
1575 	    NOCACHE, segflg, path2, fd, cap_rights_init(&rights, CAP_SYMLINKAT),
1576 	    td);
1577 	if ((error = namei(&nd)) != 0)
1578 		goto out;
1579 	if (nd.ni_vp) {
1580 		NDFREE(&nd, NDF_ONLY_PNBUF);
1581 		if (nd.ni_vp == nd.ni_dvp)
1582 			vrele(nd.ni_dvp);
1583 		else
1584 			vput(nd.ni_dvp);
1585 		vrele(nd.ni_vp);
1586 		error = EEXIST;
1587 		goto out;
1588 	}
1589 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1590 		NDFREE(&nd, NDF_ONLY_PNBUF);
1591 		vput(nd.ni_dvp);
1592 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1593 			goto out;
1594 		goto restart;
1595 	}
1596 	VATTR_NULL(&vattr);
1597 	vattr.va_mode = ACCESSPERMS &~ td->td_proc->p_fd->fd_cmask;
1598 #ifdef MAC
1599 	vattr.va_type = VLNK;
1600 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1601 	    &vattr);
1602 	if (error != 0)
1603 		goto out2;
1604 #endif
1605 	error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, syspath);
1606 	if (error == 0)
1607 		vput(nd.ni_vp);
1608 #ifdef MAC
1609 out2:
1610 #endif
1611 	NDFREE(&nd, NDF_ONLY_PNBUF);
1612 	vput(nd.ni_dvp);
1613 	vn_finished_write(mp);
1614 out:
1615 	if (segflg != UIO_SYSSPACE)
1616 		uma_zfree(namei_zone, syspath);
1617 	return (error);
1618 }
1619 
1620 /*
1621  * Delete a whiteout from the filesystem.
1622  */
1623 int
1624 sys_undelete(td, uap)
1625 	struct thread *td;
1626 	register struct undelete_args /* {
1627 		char *path;
1628 	} */ *uap;
1629 {
1630 	struct mount *mp;
1631 	struct nameidata nd;
1632 	int error;
1633 
1634 restart:
1635 	bwillwrite();
1636 	NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | AUDITVNODE1,
1637 	    UIO_USERSPACE, uap->path, td);
1638 	error = namei(&nd);
1639 	if (error != 0)
1640 		return (error);
1641 
1642 	if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) {
1643 		NDFREE(&nd, NDF_ONLY_PNBUF);
1644 		if (nd.ni_vp == nd.ni_dvp)
1645 			vrele(nd.ni_dvp);
1646 		else
1647 			vput(nd.ni_dvp);
1648 		if (nd.ni_vp)
1649 			vrele(nd.ni_vp);
1650 		return (EEXIST);
1651 	}
1652 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1653 		NDFREE(&nd, NDF_ONLY_PNBUF);
1654 		vput(nd.ni_dvp);
1655 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1656 			return (error);
1657 		goto restart;
1658 	}
1659 	error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE);
1660 	NDFREE(&nd, NDF_ONLY_PNBUF);
1661 	vput(nd.ni_dvp);
1662 	vn_finished_write(mp);
1663 	return (error);
1664 }
1665 
1666 /*
1667  * Delete a name from the filesystem.
1668  */
1669 #ifndef _SYS_SYSPROTO_H_
1670 struct unlink_args {
1671 	char	*path;
1672 };
1673 #endif
1674 int
1675 sys_unlink(td, uap)
1676 	struct thread *td;
1677 	struct unlink_args /* {
1678 		char *path;
1679 	} */ *uap;
1680 {
1681 
1682 	return (kern_unlinkat(td, AT_FDCWD, uap->path, UIO_USERSPACE, 0));
1683 }
1684 
1685 #ifndef _SYS_SYSPROTO_H_
1686 struct unlinkat_args {
1687 	int	fd;
1688 	char	*path;
1689 	int	flag;
1690 };
1691 #endif
1692 int
1693 sys_unlinkat(struct thread *td, struct unlinkat_args *uap)
1694 {
1695 	int flag = uap->flag;
1696 	int fd = uap->fd;
1697 	char *path = uap->path;
1698 
1699 	if (flag & ~AT_REMOVEDIR)
1700 		return (EINVAL);
1701 
1702 	if (flag & AT_REMOVEDIR)
1703 		return (kern_rmdirat(td, fd, path, UIO_USERSPACE));
1704 	else
1705 		return (kern_unlinkat(td, fd, path, UIO_USERSPACE, 0));
1706 }
1707 
1708 int
1709 kern_unlinkat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1710     ino_t oldinum)
1711 {
1712 	struct mount *mp;
1713 	struct vnode *vp;
1714 	struct nameidata nd;
1715 	struct stat sb;
1716 	cap_rights_t rights;
1717 	int error;
1718 
1719 restart:
1720 	bwillwrite();
1721 	NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1,
1722 	    pathseg, path, fd, cap_rights_init(&rights, CAP_UNLINKAT), td);
1723 	if ((error = namei(&nd)) != 0)
1724 		return (error == EINVAL ? EPERM : error);
1725 	vp = nd.ni_vp;
1726 	if (vp->v_type == VDIR && oldinum == 0) {
1727 		error = EPERM;		/* POSIX */
1728 	} else if (oldinum != 0 &&
1729 		  ((error = vn_stat(vp, &sb, td->td_ucred, NOCRED, td)) == 0) &&
1730 		  sb.st_ino != oldinum) {
1731 			error = EIDRM;	/* Identifier removed */
1732 	} else {
1733 		/*
1734 		 * The root of a mounted filesystem cannot be deleted.
1735 		 *
1736 		 * XXX: can this only be a VDIR case?
1737 		 */
1738 		if (vp->v_vflag & VV_ROOT)
1739 			error = EBUSY;
1740 	}
1741 	if (error == 0) {
1742 		if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1743 			NDFREE(&nd, NDF_ONLY_PNBUF);
1744 			vput(nd.ni_dvp);
1745 			if (vp == nd.ni_dvp)
1746 				vrele(vp);
1747 			else
1748 				vput(vp);
1749 			if ((error = vn_start_write(NULL, &mp,
1750 			    V_XSLEEP | PCATCH)) != 0)
1751 				return (error);
1752 			goto restart;
1753 		}
1754 #ifdef MAC
1755 		error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
1756 		    &nd.ni_cnd);
1757 		if (error != 0)
1758 			goto out;
1759 #endif
1760 		vfs_notify_upper(vp, VFS_NOTIFY_UPPER_UNLINK);
1761 		error = VOP_REMOVE(nd.ni_dvp, vp, &nd.ni_cnd);
1762 #ifdef MAC
1763 out:
1764 #endif
1765 		vn_finished_write(mp);
1766 	}
1767 	NDFREE(&nd, NDF_ONLY_PNBUF);
1768 	vput(nd.ni_dvp);
1769 	if (vp == nd.ni_dvp)
1770 		vrele(vp);
1771 	else
1772 		vput(vp);
1773 	return (error);
1774 }
1775 
1776 /*
1777  * Reposition read/write file offset.
1778  */
1779 #ifndef _SYS_SYSPROTO_H_
1780 struct lseek_args {
1781 	int	fd;
1782 	int	pad;
1783 	off_t	offset;
1784 	int	whence;
1785 };
1786 #endif
1787 int
1788 sys_lseek(td, uap)
1789 	struct thread *td;
1790 	register struct lseek_args /* {
1791 		int fd;
1792 		int pad;
1793 		off_t offset;
1794 		int whence;
1795 	} */ *uap;
1796 {
1797 	struct file *fp;
1798 	cap_rights_t rights;
1799 	int error;
1800 
1801 	AUDIT_ARG_FD(uap->fd);
1802 	error = fget(td, uap->fd, cap_rights_init(&rights, CAP_SEEK), &fp);
1803 	if (error != 0)
1804 		return (error);
1805 	error = (fp->f_ops->fo_flags & DFLAG_SEEKABLE) != 0 ?
1806 	    fo_seek(fp, uap->offset, uap->whence, td) : ESPIPE;
1807 	fdrop(fp, td);
1808 	return (error);
1809 }
1810 
1811 #if defined(COMPAT_43)
1812 /*
1813  * Reposition read/write file offset.
1814  */
1815 #ifndef _SYS_SYSPROTO_H_
1816 struct olseek_args {
1817 	int	fd;
1818 	long	offset;
1819 	int	whence;
1820 };
1821 #endif
1822 int
1823 olseek(td, uap)
1824 	struct thread *td;
1825 	register struct olseek_args /* {
1826 		int fd;
1827 		long offset;
1828 		int whence;
1829 	} */ *uap;
1830 {
1831 	struct lseek_args /* {
1832 		int fd;
1833 		int pad;
1834 		off_t offset;
1835 		int whence;
1836 	} */ nuap;
1837 
1838 	nuap.fd = uap->fd;
1839 	nuap.offset = uap->offset;
1840 	nuap.whence = uap->whence;
1841 	return (sys_lseek(td, &nuap));
1842 }
1843 #endif /* COMPAT_43 */
1844 
1845 #if defined(COMPAT_FREEBSD6)
1846 /* Version with the 'pad' argument */
1847 int
1848 freebsd6_lseek(td, uap)
1849 	struct thread *td;
1850 	register struct freebsd6_lseek_args *uap;
1851 {
1852 	struct lseek_args ouap;
1853 
1854 	ouap.fd = uap->fd;
1855 	ouap.offset = uap->offset;
1856 	ouap.whence = uap->whence;
1857 	return (sys_lseek(td, &ouap));
1858 }
1859 #endif
1860 
1861 /*
1862  * Check access permissions using passed credentials.
1863  */
1864 static int
1865 vn_access(vp, user_flags, cred, td)
1866 	struct vnode	*vp;
1867 	int		user_flags;
1868 	struct ucred	*cred;
1869 	struct thread	*td;
1870 {
1871 	accmode_t accmode;
1872 	int error;
1873 
1874 	/* Flags == 0 means only check for existence. */
1875 	if (user_flags == 0)
1876 		return (0);
1877 
1878 	accmode = 0;
1879 	if (user_flags & R_OK)
1880 		accmode |= VREAD;
1881 	if (user_flags & W_OK)
1882 		accmode |= VWRITE;
1883 	if (user_flags & X_OK)
1884 		accmode |= VEXEC;
1885 #ifdef MAC
1886 	error = mac_vnode_check_access(cred, vp, accmode);
1887 	if (error != 0)
1888 		return (error);
1889 #endif
1890 	if ((accmode & VWRITE) == 0 || (error = vn_writechk(vp)) == 0)
1891 		error = VOP_ACCESS(vp, accmode, cred, td);
1892 	return (error);
1893 }
1894 
1895 /*
1896  * Check access permissions using "real" credentials.
1897  */
1898 #ifndef _SYS_SYSPROTO_H_
1899 struct access_args {
1900 	char	*path;
1901 	int	amode;
1902 };
1903 #endif
1904 int
1905 sys_access(td, uap)
1906 	struct thread *td;
1907 	register struct access_args /* {
1908 		char *path;
1909 		int amode;
1910 	} */ *uap;
1911 {
1912 
1913 	return (kern_accessat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1914 	    0, uap->amode));
1915 }
1916 
1917 #ifndef _SYS_SYSPROTO_H_
1918 struct faccessat_args {
1919 	int	dirfd;
1920 	char	*path;
1921 	int	amode;
1922 	int	flag;
1923 }
1924 #endif
1925 int
1926 sys_faccessat(struct thread *td, struct faccessat_args *uap)
1927 {
1928 
1929 	return (kern_accessat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
1930 	    uap->amode));
1931 }
1932 
1933 int
1934 kern_accessat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1935     int flag, int amode)
1936 {
1937 	struct ucred *cred, *usecred;
1938 	struct vnode *vp;
1939 	struct nameidata nd;
1940 	cap_rights_t rights;
1941 	int error;
1942 
1943 	if (flag & ~AT_EACCESS)
1944 		return (EINVAL);
1945 	if (amode != F_OK && (amode & ~(R_OK | W_OK | X_OK)) != 0)
1946 		return (EINVAL);
1947 
1948 	/*
1949 	 * Create and modify a temporary credential instead of one that
1950 	 * is potentially shared (if we need one).
1951 	 */
1952 	cred = td->td_ucred;
1953 	if ((flag & AT_EACCESS) == 0 &&
1954 	    ((cred->cr_uid != cred->cr_ruid ||
1955 	    cred->cr_rgid != cred->cr_groups[0]))) {
1956 		usecred = crdup(cred);
1957 		usecred->cr_uid = cred->cr_ruid;
1958 		usecred->cr_groups[0] = cred->cr_rgid;
1959 		td->td_ucred = usecred;
1960 	} else
1961 		usecred = cred;
1962 	AUDIT_ARG_VALUE(amode);
1963 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF |
1964 	    AUDITVNODE1, pathseg, path, fd, cap_rights_init(&rights, CAP_FSTAT),
1965 	    td);
1966 	if ((error = namei(&nd)) != 0)
1967 		goto out;
1968 	vp = nd.ni_vp;
1969 
1970 	error = vn_access(vp, amode, usecred, td);
1971 	NDFREE(&nd, NDF_ONLY_PNBUF);
1972 	vput(vp);
1973 out:
1974 	if (usecred != cred) {
1975 		td->td_ucred = cred;
1976 		crfree(usecred);
1977 	}
1978 	return (error);
1979 }
1980 
1981 /*
1982  * Check access permissions using "effective" credentials.
1983  */
1984 #ifndef _SYS_SYSPROTO_H_
1985 struct eaccess_args {
1986 	char	*path;
1987 	int	amode;
1988 };
1989 #endif
1990 int
1991 sys_eaccess(td, uap)
1992 	struct thread *td;
1993 	register struct eaccess_args /* {
1994 		char *path;
1995 		int amode;
1996 	} */ *uap;
1997 {
1998 
1999 	return (kern_accessat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2000 	    AT_EACCESS, uap->amode));
2001 }
2002 
2003 #if defined(COMPAT_43)
2004 /*
2005  * Get file status; this version follows links.
2006  */
2007 #ifndef _SYS_SYSPROTO_H_
2008 struct ostat_args {
2009 	char	*path;
2010 	struct ostat *ub;
2011 };
2012 #endif
2013 int
2014 ostat(td, uap)
2015 	struct thread *td;
2016 	register struct ostat_args /* {
2017 		char *path;
2018 		struct ostat *ub;
2019 	} */ *uap;
2020 {
2021 	struct stat sb;
2022 	struct ostat osb;
2023 	int error;
2024 
2025 	error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE,
2026 	    &sb, NULL);
2027 	if (error != 0)
2028 		return (error);
2029 	cvtstat(&sb, &osb);
2030 	return (copyout(&osb, uap->ub, sizeof (osb)));
2031 }
2032 
2033 /*
2034  * Get file status; this version does not follow links.
2035  */
2036 #ifndef _SYS_SYSPROTO_H_
2037 struct olstat_args {
2038 	char	*path;
2039 	struct ostat *ub;
2040 };
2041 #endif
2042 int
2043 olstat(td, uap)
2044 	struct thread *td;
2045 	register struct olstat_args /* {
2046 		char *path;
2047 		struct ostat *ub;
2048 	} */ *uap;
2049 {
2050 	struct stat sb;
2051 	struct ostat osb;
2052 	int error;
2053 
2054 	error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2055 	    UIO_USERSPACE, &sb, NULL);
2056 	if (error != 0)
2057 		return (error);
2058 	cvtstat(&sb, &osb);
2059 	return (copyout(&osb, uap->ub, sizeof (osb)));
2060 }
2061 
2062 /*
2063  * Convert from an old to a new stat structure.
2064  */
2065 void
2066 cvtstat(st, ost)
2067 	struct stat *st;
2068 	struct ostat *ost;
2069 {
2070 
2071 	bzero(ost, sizeof(*ost));
2072 	ost->st_dev = st->st_dev;
2073 	ost->st_ino = st->st_ino;
2074 	ost->st_mode = st->st_mode;
2075 	ost->st_nlink = st->st_nlink;
2076 	ost->st_uid = st->st_uid;
2077 	ost->st_gid = st->st_gid;
2078 	ost->st_rdev = st->st_rdev;
2079 	if (st->st_size < (quad_t)1 << 32)
2080 		ost->st_size = st->st_size;
2081 	else
2082 		ost->st_size = -2;
2083 	ost->st_atim = st->st_atim;
2084 	ost->st_mtim = st->st_mtim;
2085 	ost->st_ctim = st->st_ctim;
2086 	ost->st_blksize = st->st_blksize;
2087 	ost->st_blocks = st->st_blocks;
2088 	ost->st_flags = st->st_flags;
2089 	ost->st_gen = st->st_gen;
2090 }
2091 #endif /* COMPAT_43 */
2092 
2093 /*
2094  * Get file status; this version follows links.
2095  */
2096 #ifndef _SYS_SYSPROTO_H_
2097 struct stat_args {
2098 	char	*path;
2099 	struct stat *ub;
2100 };
2101 #endif
2102 int
2103 sys_stat(td, uap)
2104 	struct thread *td;
2105 	register struct stat_args /* {
2106 		char *path;
2107 		struct stat *ub;
2108 	} */ *uap;
2109 {
2110 	struct stat sb;
2111 	int error;
2112 
2113 	error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE,
2114 	    &sb, NULL);
2115 	if (error == 0)
2116 		error = copyout(&sb, uap->ub, sizeof (sb));
2117 	return (error);
2118 }
2119 
2120 #ifndef _SYS_SYSPROTO_H_
2121 struct fstatat_args {
2122 	int	fd;
2123 	char	*path;
2124 	struct stat	*buf;
2125 	int	flag;
2126 }
2127 #endif
2128 int
2129 sys_fstatat(struct thread *td, struct fstatat_args *uap)
2130 {
2131 	struct stat sb;
2132 	int error;
2133 
2134 	error = kern_statat(td, uap->flag, uap->fd, uap->path,
2135 	    UIO_USERSPACE, &sb, NULL);
2136 	if (error == 0)
2137 		error = copyout(&sb, uap->buf, sizeof (sb));
2138 	return (error);
2139 }
2140 
2141 int
2142 kern_statat(struct thread *td, int flag, int fd, char *path,
2143     enum uio_seg pathseg, struct stat *sbp,
2144     void (*hook)(struct vnode *vp, struct stat *sbp))
2145 {
2146 	struct nameidata nd;
2147 	struct stat sb;
2148 	cap_rights_t rights;
2149 	int error;
2150 
2151 	if (flag & ~AT_SYMLINK_NOFOLLOW)
2152 		return (EINVAL);
2153 
2154 	NDINIT_ATRIGHTS(&nd, LOOKUP, ((flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW :
2155 	    FOLLOW) | LOCKSHARED | LOCKLEAF | AUDITVNODE1, pathseg, path, fd,
2156 	    cap_rights_init(&rights, CAP_FSTAT), td);
2157 
2158 	if ((error = namei(&nd)) != 0)
2159 		return (error);
2160 	error = vn_stat(nd.ni_vp, &sb, td->td_ucred, NOCRED, td);
2161 	if (error == 0) {
2162 		SDT_PROBE2(vfs, , stat, mode, path, sb.st_mode);
2163 		if (S_ISREG(sb.st_mode))
2164 			SDT_PROBE2(vfs, , stat, reg, path, pathseg);
2165 		if (__predict_false(hook != NULL))
2166 			hook(nd.ni_vp, &sb);
2167 	}
2168 	NDFREE(&nd, NDF_ONLY_PNBUF);
2169 	vput(nd.ni_vp);
2170 	if (error != 0)
2171 		return (error);
2172 	*sbp = sb;
2173 #ifdef KTRACE
2174 	if (KTRPOINT(td, KTR_STRUCT))
2175 		ktrstat(&sb);
2176 #endif
2177 	return (0);
2178 }
2179 
2180 /*
2181  * Get file status; this version does not follow links.
2182  */
2183 #ifndef _SYS_SYSPROTO_H_
2184 struct lstat_args {
2185 	char	*path;
2186 	struct stat *ub;
2187 };
2188 #endif
2189 int
2190 sys_lstat(td, uap)
2191 	struct thread *td;
2192 	register struct lstat_args /* {
2193 		char *path;
2194 		struct stat *ub;
2195 	} */ *uap;
2196 {
2197 	struct stat sb;
2198 	int error;
2199 
2200 	error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2201 	    UIO_USERSPACE, &sb, NULL);
2202 	if (error == 0)
2203 		error = copyout(&sb, uap->ub, sizeof (sb));
2204 	return (error);
2205 }
2206 
2207 /*
2208  * Implementation of the NetBSD [l]stat() functions.
2209  */
2210 void
2211 cvtnstat(sb, nsb)
2212 	struct stat *sb;
2213 	struct nstat *nsb;
2214 {
2215 
2216 	bzero(nsb, sizeof *nsb);
2217 	nsb->st_dev = sb->st_dev;
2218 	nsb->st_ino = sb->st_ino;
2219 	nsb->st_mode = sb->st_mode;
2220 	nsb->st_nlink = sb->st_nlink;
2221 	nsb->st_uid = sb->st_uid;
2222 	nsb->st_gid = sb->st_gid;
2223 	nsb->st_rdev = sb->st_rdev;
2224 	nsb->st_atim = sb->st_atim;
2225 	nsb->st_mtim = sb->st_mtim;
2226 	nsb->st_ctim = sb->st_ctim;
2227 	nsb->st_size = sb->st_size;
2228 	nsb->st_blocks = sb->st_blocks;
2229 	nsb->st_blksize = sb->st_blksize;
2230 	nsb->st_flags = sb->st_flags;
2231 	nsb->st_gen = sb->st_gen;
2232 	nsb->st_birthtim = sb->st_birthtim;
2233 }
2234 
2235 #ifndef _SYS_SYSPROTO_H_
2236 struct nstat_args {
2237 	char	*path;
2238 	struct nstat *ub;
2239 };
2240 #endif
2241 int
2242 sys_nstat(td, uap)
2243 	struct thread *td;
2244 	register struct nstat_args /* {
2245 		char *path;
2246 		struct nstat *ub;
2247 	} */ *uap;
2248 {
2249 	struct stat sb;
2250 	struct nstat nsb;
2251 	int error;
2252 
2253 	error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE,
2254 	    &sb, NULL);
2255 	if (error != 0)
2256 		return (error);
2257 	cvtnstat(&sb, &nsb);
2258 	return (copyout(&nsb, uap->ub, sizeof (nsb)));
2259 }
2260 
2261 /*
2262  * NetBSD lstat.  Get file status; this version does not follow links.
2263  */
2264 #ifndef _SYS_SYSPROTO_H_
2265 struct lstat_args {
2266 	char	*path;
2267 	struct stat *ub;
2268 };
2269 #endif
2270 int
2271 sys_nlstat(td, uap)
2272 	struct thread *td;
2273 	register struct nlstat_args /* {
2274 		char *path;
2275 		struct nstat *ub;
2276 	} */ *uap;
2277 {
2278 	struct stat sb;
2279 	struct nstat nsb;
2280 	int error;
2281 
2282 	error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2283 	    UIO_USERSPACE, &sb, NULL);
2284 	if (error != 0)
2285 		return (error);
2286 	cvtnstat(&sb, &nsb);
2287 	return (copyout(&nsb, uap->ub, sizeof (nsb)));
2288 }
2289 
2290 /*
2291  * Get configurable pathname variables.
2292  */
2293 #ifndef _SYS_SYSPROTO_H_
2294 struct pathconf_args {
2295 	char	*path;
2296 	int	name;
2297 };
2298 #endif
2299 int
2300 sys_pathconf(td, uap)
2301 	struct thread *td;
2302 	register struct pathconf_args /* {
2303 		char *path;
2304 		int name;
2305 	} */ *uap;
2306 {
2307 
2308 	return (kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name, FOLLOW));
2309 }
2310 
2311 #ifndef _SYS_SYSPROTO_H_
2312 struct lpathconf_args {
2313 	char	*path;
2314 	int	name;
2315 };
2316 #endif
2317 int
2318 sys_lpathconf(td, uap)
2319 	struct thread *td;
2320 	register struct lpathconf_args /* {
2321 		char *path;
2322 		int name;
2323 	} */ *uap;
2324 {
2325 
2326 	return (kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name,
2327 	    NOFOLLOW));
2328 }
2329 
2330 int
2331 kern_pathconf(struct thread *td, char *path, enum uio_seg pathseg, int name,
2332     u_long flags)
2333 {
2334 	struct nameidata nd;
2335 	int error;
2336 
2337 	NDINIT(&nd, LOOKUP, LOCKSHARED | LOCKLEAF | AUDITVNODE1 | flags,
2338 	    pathseg, path, td);
2339 	if ((error = namei(&nd)) != 0)
2340 		return (error);
2341 	NDFREE(&nd, NDF_ONLY_PNBUF);
2342 
2343 	error = VOP_PATHCONF(nd.ni_vp, name, td->td_retval);
2344 	vput(nd.ni_vp);
2345 	return (error);
2346 }
2347 
2348 /*
2349  * Return target name of a symbolic link.
2350  */
2351 #ifndef _SYS_SYSPROTO_H_
2352 struct readlink_args {
2353 	char	*path;
2354 	char	*buf;
2355 	size_t	count;
2356 };
2357 #endif
2358 int
2359 sys_readlink(td, uap)
2360 	struct thread *td;
2361 	register struct readlink_args /* {
2362 		char *path;
2363 		char *buf;
2364 		size_t count;
2365 	} */ *uap;
2366 {
2367 
2368 	return (kern_readlinkat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2369 	    uap->buf, UIO_USERSPACE, uap->count));
2370 }
2371 #ifndef _SYS_SYSPROTO_H_
2372 struct readlinkat_args {
2373 	int	fd;
2374 	char	*path;
2375 	char	*buf;
2376 	size_t	bufsize;
2377 };
2378 #endif
2379 int
2380 sys_readlinkat(struct thread *td, struct readlinkat_args *uap)
2381 {
2382 
2383 	return (kern_readlinkat(td, uap->fd, uap->path, UIO_USERSPACE,
2384 	    uap->buf, UIO_USERSPACE, uap->bufsize));
2385 }
2386 
2387 int
2388 kern_readlinkat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2389     char *buf, enum uio_seg bufseg, size_t count)
2390 {
2391 	struct vnode *vp;
2392 	struct iovec aiov;
2393 	struct uio auio;
2394 	struct nameidata nd;
2395 	int error;
2396 
2397 	if (count > IOSIZE_MAX)
2398 		return (EINVAL);
2399 
2400 	NDINIT_AT(&nd, LOOKUP, NOFOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
2401 	    pathseg, path, fd, td);
2402 
2403 	if ((error = namei(&nd)) != 0)
2404 		return (error);
2405 	NDFREE(&nd, NDF_ONLY_PNBUF);
2406 	vp = nd.ni_vp;
2407 #ifdef MAC
2408 	error = mac_vnode_check_readlink(td->td_ucred, vp);
2409 	if (error != 0) {
2410 		vput(vp);
2411 		return (error);
2412 	}
2413 #endif
2414 	if (vp->v_type != VLNK)
2415 		error = EINVAL;
2416 	else {
2417 		aiov.iov_base = buf;
2418 		aiov.iov_len = count;
2419 		auio.uio_iov = &aiov;
2420 		auio.uio_iovcnt = 1;
2421 		auio.uio_offset = 0;
2422 		auio.uio_rw = UIO_READ;
2423 		auio.uio_segflg = bufseg;
2424 		auio.uio_td = td;
2425 		auio.uio_resid = count;
2426 		error = VOP_READLINK(vp, &auio, td->td_ucred);
2427 		td->td_retval[0] = count - auio.uio_resid;
2428 	}
2429 	vput(vp);
2430 	return (error);
2431 }
2432 
2433 /*
2434  * Common implementation code for chflags() and fchflags().
2435  */
2436 static int
2437 setfflags(td, vp, flags)
2438 	struct thread *td;
2439 	struct vnode *vp;
2440 	u_long flags;
2441 {
2442 	struct mount *mp;
2443 	struct vattr vattr;
2444 	int error;
2445 
2446 	/* We can't support the value matching VNOVAL. */
2447 	if (flags == VNOVAL)
2448 		return (EOPNOTSUPP);
2449 
2450 	/*
2451 	 * Prevent non-root users from setting flags on devices.  When
2452 	 * a device is reused, users can retain ownership of the device
2453 	 * if they are allowed to set flags and programs assume that
2454 	 * chown can't fail when done as root.
2455 	 */
2456 	if (vp->v_type == VCHR || vp->v_type == VBLK) {
2457 		error = priv_check(td, PRIV_VFS_CHFLAGS_DEV);
2458 		if (error != 0)
2459 			return (error);
2460 	}
2461 
2462 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2463 		return (error);
2464 	VATTR_NULL(&vattr);
2465 	vattr.va_flags = flags;
2466 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2467 #ifdef MAC
2468 	error = mac_vnode_check_setflags(td->td_ucred, vp, vattr.va_flags);
2469 	if (error == 0)
2470 #endif
2471 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
2472 	VOP_UNLOCK(vp, 0);
2473 	vn_finished_write(mp);
2474 	return (error);
2475 }
2476 
2477 /*
2478  * Change flags of a file given a path name.
2479  */
2480 #ifndef _SYS_SYSPROTO_H_
2481 struct chflags_args {
2482 	const char *path;
2483 	u_long	flags;
2484 };
2485 #endif
2486 int
2487 sys_chflags(td, uap)
2488 	struct thread *td;
2489 	register struct chflags_args /* {
2490 		const char *path;
2491 		u_long flags;
2492 	} */ *uap;
2493 {
2494 
2495 	return (kern_chflagsat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2496 	    uap->flags, 0));
2497 }
2498 
2499 #ifndef _SYS_SYSPROTO_H_
2500 struct chflagsat_args {
2501 	int	fd;
2502 	const char *path;
2503 	u_long	flags;
2504 	int	atflag;
2505 }
2506 #endif
2507 int
2508 sys_chflagsat(struct thread *td, struct chflagsat_args *uap)
2509 {
2510 	int fd = uap->fd;
2511 	const char *path = uap->path;
2512 	u_long flags = uap->flags;
2513 	int atflag = uap->atflag;
2514 
2515 	if (atflag & ~AT_SYMLINK_NOFOLLOW)
2516 		return (EINVAL);
2517 
2518 	return (kern_chflagsat(td, fd, path, UIO_USERSPACE, flags, atflag));
2519 }
2520 
2521 /*
2522  * Same as chflags() but doesn't follow symlinks.
2523  */
2524 int
2525 sys_lchflags(td, uap)
2526 	struct thread *td;
2527 	register struct lchflags_args /* {
2528 		const char *path;
2529 		u_long flags;
2530 	} */ *uap;
2531 {
2532 
2533 	return (kern_chflagsat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2534 	    uap->flags, AT_SYMLINK_NOFOLLOW));
2535 }
2536 
2537 static int
2538 kern_chflagsat(struct thread *td, int fd, const char *path,
2539     enum uio_seg pathseg, u_long flags, int atflag)
2540 {
2541 	struct nameidata nd;
2542 	cap_rights_t rights;
2543 	int error, follow;
2544 
2545 	AUDIT_ARG_FFLAGS(flags);
2546 	follow = (atflag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
2547 	NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, pathseg, path, fd,
2548 	    cap_rights_init(&rights, CAP_FCHFLAGS), td);
2549 	if ((error = namei(&nd)) != 0)
2550 		return (error);
2551 	NDFREE(&nd, NDF_ONLY_PNBUF);
2552 	error = setfflags(td, nd.ni_vp, flags);
2553 	vrele(nd.ni_vp);
2554 	return (error);
2555 }
2556 
2557 /*
2558  * Change flags of a file given a file descriptor.
2559  */
2560 #ifndef _SYS_SYSPROTO_H_
2561 struct fchflags_args {
2562 	int	fd;
2563 	u_long	flags;
2564 };
2565 #endif
2566 int
2567 sys_fchflags(td, uap)
2568 	struct thread *td;
2569 	register struct fchflags_args /* {
2570 		int fd;
2571 		u_long flags;
2572 	} */ *uap;
2573 {
2574 	struct file *fp;
2575 	cap_rights_t rights;
2576 	int error;
2577 
2578 	AUDIT_ARG_FD(uap->fd);
2579 	AUDIT_ARG_FFLAGS(uap->flags);
2580 	error = getvnode(td, uap->fd, cap_rights_init(&rights, CAP_FCHFLAGS),
2581 	    &fp);
2582 	if (error != 0)
2583 		return (error);
2584 #ifdef AUDIT
2585 	vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
2586 	AUDIT_ARG_VNODE1(fp->f_vnode);
2587 	VOP_UNLOCK(fp->f_vnode, 0);
2588 #endif
2589 	error = setfflags(td, fp->f_vnode, uap->flags);
2590 	fdrop(fp, td);
2591 	return (error);
2592 }
2593 
2594 /*
2595  * Common implementation code for chmod(), lchmod() and fchmod().
2596  */
2597 int
2598 setfmode(td, cred, vp, mode)
2599 	struct thread *td;
2600 	struct ucred *cred;
2601 	struct vnode *vp;
2602 	int mode;
2603 {
2604 	struct mount *mp;
2605 	struct vattr vattr;
2606 	int error;
2607 
2608 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2609 		return (error);
2610 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2611 	VATTR_NULL(&vattr);
2612 	vattr.va_mode = mode & ALLPERMS;
2613 #ifdef MAC
2614 	error = mac_vnode_check_setmode(cred, vp, vattr.va_mode);
2615 	if (error == 0)
2616 #endif
2617 		error = VOP_SETATTR(vp, &vattr, cred);
2618 	VOP_UNLOCK(vp, 0);
2619 	vn_finished_write(mp);
2620 	return (error);
2621 }
2622 
2623 /*
2624  * Change mode of a file given path name.
2625  */
2626 #ifndef _SYS_SYSPROTO_H_
2627 struct chmod_args {
2628 	char	*path;
2629 	int	mode;
2630 };
2631 #endif
2632 int
2633 sys_chmod(td, uap)
2634 	struct thread *td;
2635 	register struct chmod_args /* {
2636 		char *path;
2637 		int mode;
2638 	} */ *uap;
2639 {
2640 
2641 	return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2642 	    uap->mode, 0));
2643 }
2644 
2645 #ifndef _SYS_SYSPROTO_H_
2646 struct fchmodat_args {
2647 	int	dirfd;
2648 	char	*path;
2649 	mode_t	mode;
2650 	int	flag;
2651 }
2652 #endif
2653 int
2654 sys_fchmodat(struct thread *td, struct fchmodat_args *uap)
2655 {
2656 	int flag = uap->flag;
2657 	int fd = uap->fd;
2658 	char *path = uap->path;
2659 	mode_t mode = uap->mode;
2660 
2661 	if (flag & ~AT_SYMLINK_NOFOLLOW)
2662 		return (EINVAL);
2663 
2664 	return (kern_fchmodat(td, fd, path, UIO_USERSPACE, mode, flag));
2665 }
2666 
2667 /*
2668  * Change mode of a file given path name (don't follow links.)
2669  */
2670 #ifndef _SYS_SYSPROTO_H_
2671 struct lchmod_args {
2672 	char	*path;
2673 	int	mode;
2674 };
2675 #endif
2676 int
2677 sys_lchmod(td, uap)
2678 	struct thread *td;
2679 	register struct lchmod_args /* {
2680 		char *path;
2681 		int mode;
2682 	} */ *uap;
2683 {
2684 
2685 	return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2686 	    uap->mode, AT_SYMLINK_NOFOLLOW));
2687 }
2688 
2689 int
2690 kern_fchmodat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2691     mode_t mode, int flag)
2692 {
2693 	struct nameidata nd;
2694 	cap_rights_t rights;
2695 	int error, follow;
2696 
2697 	AUDIT_ARG_MODE(mode);
2698 	follow = (flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
2699 	NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, pathseg, path, fd,
2700 	    cap_rights_init(&rights, CAP_FCHMOD), td);
2701 	if ((error = namei(&nd)) != 0)
2702 		return (error);
2703 	NDFREE(&nd, NDF_ONLY_PNBUF);
2704 	error = setfmode(td, td->td_ucred, nd.ni_vp, mode);
2705 	vrele(nd.ni_vp);
2706 	return (error);
2707 }
2708 
2709 /*
2710  * Change mode of a file given a file descriptor.
2711  */
2712 #ifndef _SYS_SYSPROTO_H_
2713 struct fchmod_args {
2714 	int	fd;
2715 	int	mode;
2716 };
2717 #endif
2718 int
2719 sys_fchmod(struct thread *td, struct fchmod_args *uap)
2720 {
2721 	struct file *fp;
2722 	cap_rights_t rights;
2723 	int error;
2724 
2725 	AUDIT_ARG_FD(uap->fd);
2726 	AUDIT_ARG_MODE(uap->mode);
2727 
2728 	error = fget(td, uap->fd, cap_rights_init(&rights, CAP_FCHMOD), &fp);
2729 	if (error != 0)
2730 		return (error);
2731 	error = fo_chmod(fp, uap->mode, td->td_ucred, td);
2732 	fdrop(fp, td);
2733 	return (error);
2734 }
2735 
2736 /*
2737  * Common implementation for chown(), lchown(), and fchown()
2738  */
2739 int
2740 setfown(td, cred, vp, uid, gid)
2741 	struct thread *td;
2742 	struct ucred *cred;
2743 	struct vnode *vp;
2744 	uid_t uid;
2745 	gid_t gid;
2746 {
2747 	struct mount *mp;
2748 	struct vattr vattr;
2749 	int error;
2750 
2751 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2752 		return (error);
2753 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2754 	VATTR_NULL(&vattr);
2755 	vattr.va_uid = uid;
2756 	vattr.va_gid = gid;
2757 #ifdef MAC
2758 	error = mac_vnode_check_setowner(cred, vp, vattr.va_uid,
2759 	    vattr.va_gid);
2760 	if (error == 0)
2761 #endif
2762 		error = VOP_SETATTR(vp, &vattr, cred);
2763 	VOP_UNLOCK(vp, 0);
2764 	vn_finished_write(mp);
2765 	return (error);
2766 }
2767 
2768 /*
2769  * Set ownership given a path name.
2770  */
2771 #ifndef _SYS_SYSPROTO_H_
2772 struct chown_args {
2773 	char	*path;
2774 	int	uid;
2775 	int	gid;
2776 };
2777 #endif
2778 int
2779 sys_chown(td, uap)
2780 	struct thread *td;
2781 	register struct chown_args /* {
2782 		char *path;
2783 		int uid;
2784 		int gid;
2785 	} */ *uap;
2786 {
2787 
2788 	return (kern_fchownat(td, AT_FDCWD, uap->path, UIO_USERSPACE, uap->uid,
2789 	    uap->gid, 0));
2790 }
2791 
2792 #ifndef _SYS_SYSPROTO_H_
2793 struct fchownat_args {
2794 	int fd;
2795 	const char * path;
2796 	uid_t uid;
2797 	gid_t gid;
2798 	int flag;
2799 };
2800 #endif
2801 int
2802 sys_fchownat(struct thread *td, struct fchownat_args *uap)
2803 {
2804 	int flag;
2805 
2806 	flag = uap->flag;
2807 	if (flag & ~AT_SYMLINK_NOFOLLOW)
2808 		return (EINVAL);
2809 
2810 	return (kern_fchownat(td, uap->fd, uap->path, UIO_USERSPACE, uap->uid,
2811 	    uap->gid, uap->flag));
2812 }
2813 
2814 int
2815 kern_fchownat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2816     int uid, int gid, int flag)
2817 {
2818 	struct nameidata nd;
2819 	cap_rights_t rights;
2820 	int error, follow;
2821 
2822 	AUDIT_ARG_OWNER(uid, gid);
2823 	follow = (flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
2824 	NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, pathseg, path, fd,
2825 	    cap_rights_init(&rights, CAP_FCHOWN), td);
2826 
2827 	if ((error = namei(&nd)) != 0)
2828 		return (error);
2829 	NDFREE(&nd, NDF_ONLY_PNBUF);
2830 	error = setfown(td, td->td_ucred, nd.ni_vp, uid, gid);
2831 	vrele(nd.ni_vp);
2832 	return (error);
2833 }
2834 
2835 /*
2836  * Set ownership given a path name, do not cross symlinks.
2837  */
2838 #ifndef _SYS_SYSPROTO_H_
2839 struct lchown_args {
2840 	char	*path;
2841 	int	uid;
2842 	int	gid;
2843 };
2844 #endif
2845 int
2846 sys_lchown(td, uap)
2847 	struct thread *td;
2848 	register struct lchown_args /* {
2849 		char *path;
2850 		int uid;
2851 		int gid;
2852 	} */ *uap;
2853 {
2854 
2855 	return (kern_fchownat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2856 	    uap->uid, uap->gid, AT_SYMLINK_NOFOLLOW));
2857 }
2858 
2859 /*
2860  * Set ownership given a file descriptor.
2861  */
2862 #ifndef _SYS_SYSPROTO_H_
2863 struct fchown_args {
2864 	int	fd;
2865 	int	uid;
2866 	int	gid;
2867 };
2868 #endif
2869 int
2870 sys_fchown(td, uap)
2871 	struct thread *td;
2872 	register struct fchown_args /* {
2873 		int fd;
2874 		int uid;
2875 		int gid;
2876 	} */ *uap;
2877 {
2878 	struct file *fp;
2879 	cap_rights_t rights;
2880 	int error;
2881 
2882 	AUDIT_ARG_FD(uap->fd);
2883 	AUDIT_ARG_OWNER(uap->uid, uap->gid);
2884 	error = fget(td, uap->fd, cap_rights_init(&rights, CAP_FCHOWN), &fp);
2885 	if (error != 0)
2886 		return (error);
2887 	error = fo_chown(fp, uap->uid, uap->gid, td->td_ucred, td);
2888 	fdrop(fp, td);
2889 	return (error);
2890 }
2891 
2892 /*
2893  * Common implementation code for utimes(), lutimes(), and futimes().
2894  */
2895 static int
2896 getutimes(usrtvp, tvpseg, tsp)
2897 	const struct timeval *usrtvp;
2898 	enum uio_seg tvpseg;
2899 	struct timespec *tsp;
2900 {
2901 	struct timeval tv[2];
2902 	const struct timeval *tvp;
2903 	int error;
2904 
2905 	if (usrtvp == NULL) {
2906 		vfs_timestamp(&tsp[0]);
2907 		tsp[1] = tsp[0];
2908 	} else {
2909 		if (tvpseg == UIO_SYSSPACE) {
2910 			tvp = usrtvp;
2911 		} else {
2912 			if ((error = copyin(usrtvp, tv, sizeof(tv))) != 0)
2913 				return (error);
2914 			tvp = tv;
2915 		}
2916 
2917 		if (tvp[0].tv_usec < 0 || tvp[0].tv_usec >= 1000000 ||
2918 		    tvp[1].tv_usec < 0 || tvp[1].tv_usec >= 1000000)
2919 			return (EINVAL);
2920 		TIMEVAL_TO_TIMESPEC(&tvp[0], &tsp[0]);
2921 		TIMEVAL_TO_TIMESPEC(&tvp[1], &tsp[1]);
2922 	}
2923 	return (0);
2924 }
2925 
2926 /*
2927  * Common implementation code for futimens(), utimensat().
2928  */
2929 #define	UTIMENS_NULL	0x1
2930 #define	UTIMENS_EXIT	0x2
2931 static int
2932 getutimens(const struct timespec *usrtsp, enum uio_seg tspseg,
2933     struct timespec *tsp, int *retflags)
2934 {
2935 	struct timespec tsnow;
2936 	int error;
2937 
2938 	vfs_timestamp(&tsnow);
2939 	*retflags = 0;
2940 	if (usrtsp == NULL) {
2941 		tsp[0] = tsnow;
2942 		tsp[1] = tsnow;
2943 		*retflags |= UTIMENS_NULL;
2944 		return (0);
2945 	}
2946 	if (tspseg == UIO_SYSSPACE) {
2947 		tsp[0] = usrtsp[0];
2948 		tsp[1] = usrtsp[1];
2949 	} else if ((error = copyin(usrtsp, tsp, sizeof(*tsp) * 2)) != 0)
2950 		return (error);
2951 	if (tsp[0].tv_nsec == UTIME_OMIT && tsp[1].tv_nsec == UTIME_OMIT)
2952 		*retflags |= UTIMENS_EXIT;
2953 	if (tsp[0].tv_nsec == UTIME_NOW && tsp[1].tv_nsec == UTIME_NOW)
2954 		*retflags |= UTIMENS_NULL;
2955 	if (tsp[0].tv_nsec == UTIME_OMIT)
2956 		tsp[0].tv_sec = VNOVAL;
2957 	else if (tsp[0].tv_nsec == UTIME_NOW)
2958 		tsp[0] = tsnow;
2959 	else if (tsp[0].tv_nsec < 0 || tsp[0].tv_nsec >= 1000000000L)
2960 		return (EINVAL);
2961 	if (tsp[1].tv_nsec == UTIME_OMIT)
2962 		tsp[1].tv_sec = VNOVAL;
2963 	else if (tsp[1].tv_nsec == UTIME_NOW)
2964 		tsp[1] = tsnow;
2965 	else if (tsp[1].tv_nsec < 0 || tsp[1].tv_nsec >= 1000000000L)
2966 		return (EINVAL);
2967 
2968 	return (0);
2969 }
2970 
2971 /*
2972  * Common implementation code for utimes(), lutimes(), futimes(), futimens(),
2973  * and utimensat().
2974  */
2975 static int
2976 setutimes(td, vp, ts, numtimes, nullflag)
2977 	struct thread *td;
2978 	struct vnode *vp;
2979 	const struct timespec *ts;
2980 	int numtimes;
2981 	int nullflag;
2982 {
2983 	struct mount *mp;
2984 	struct vattr vattr;
2985 	int error, setbirthtime;
2986 
2987 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2988 		return (error);
2989 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2990 	setbirthtime = 0;
2991 	if (numtimes < 3 && !VOP_GETATTR(vp, &vattr, td->td_ucred) &&
2992 	    timespeccmp(&ts[1], &vattr.va_birthtime, < ))
2993 		setbirthtime = 1;
2994 	VATTR_NULL(&vattr);
2995 	vattr.va_atime = ts[0];
2996 	vattr.va_mtime = ts[1];
2997 	if (setbirthtime)
2998 		vattr.va_birthtime = ts[1];
2999 	if (numtimes > 2)
3000 		vattr.va_birthtime = ts[2];
3001 	if (nullflag)
3002 		vattr.va_vaflags |= VA_UTIMES_NULL;
3003 #ifdef MAC
3004 	error = mac_vnode_check_setutimes(td->td_ucred, vp, vattr.va_atime,
3005 	    vattr.va_mtime);
3006 #endif
3007 	if (error == 0)
3008 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3009 	VOP_UNLOCK(vp, 0);
3010 	vn_finished_write(mp);
3011 	return (error);
3012 }
3013 
3014 /*
3015  * Set the access and modification times of a file.
3016  */
3017 #ifndef _SYS_SYSPROTO_H_
3018 struct utimes_args {
3019 	char	*path;
3020 	struct	timeval *tptr;
3021 };
3022 #endif
3023 int
3024 sys_utimes(td, uap)
3025 	struct thread *td;
3026 	register struct utimes_args /* {
3027 		char *path;
3028 		struct timeval *tptr;
3029 	} */ *uap;
3030 {
3031 
3032 	return (kern_utimesat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
3033 	    uap->tptr, UIO_USERSPACE));
3034 }
3035 
3036 #ifndef _SYS_SYSPROTO_H_
3037 struct futimesat_args {
3038 	int fd;
3039 	const char * path;
3040 	const struct timeval * times;
3041 };
3042 #endif
3043 int
3044 sys_futimesat(struct thread *td, struct futimesat_args *uap)
3045 {
3046 
3047 	return (kern_utimesat(td, uap->fd, uap->path, UIO_USERSPACE,
3048 	    uap->times, UIO_USERSPACE));
3049 }
3050 
3051 int
3052 kern_utimesat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
3053     struct timeval *tptr, enum uio_seg tptrseg)
3054 {
3055 	struct nameidata nd;
3056 	struct timespec ts[2];
3057 	cap_rights_t rights;
3058 	int error;
3059 
3060 	if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3061 		return (error);
3062 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, fd,
3063 	    cap_rights_init(&rights, CAP_FUTIMES), td);
3064 
3065 	if ((error = namei(&nd)) != 0)
3066 		return (error);
3067 	NDFREE(&nd, NDF_ONLY_PNBUF);
3068 	error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3069 	vrele(nd.ni_vp);
3070 	return (error);
3071 }
3072 
3073 /*
3074  * Set the access and modification times of a file.
3075  */
3076 #ifndef _SYS_SYSPROTO_H_
3077 struct lutimes_args {
3078 	char	*path;
3079 	struct	timeval *tptr;
3080 };
3081 #endif
3082 int
3083 sys_lutimes(td, uap)
3084 	struct thread *td;
3085 	register struct lutimes_args /* {
3086 		char *path;
3087 		struct timeval *tptr;
3088 	} */ *uap;
3089 {
3090 
3091 	return (kern_lutimes(td, uap->path, UIO_USERSPACE, uap->tptr,
3092 	    UIO_USERSPACE));
3093 }
3094 
3095 int
3096 kern_lutimes(struct thread *td, char *path, enum uio_seg pathseg,
3097     struct timeval *tptr, enum uio_seg tptrseg)
3098 {
3099 	struct timespec ts[2];
3100 	struct nameidata nd;
3101 	int error;
3102 
3103 	if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3104 		return (error);
3105 	NDINIT(&nd, LOOKUP, NOFOLLOW | AUDITVNODE1, pathseg, path, td);
3106 	if ((error = namei(&nd)) != 0)
3107 		return (error);
3108 	NDFREE(&nd, NDF_ONLY_PNBUF);
3109 	error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3110 	vrele(nd.ni_vp);
3111 	return (error);
3112 }
3113 
3114 /*
3115  * Set the access and modification times of a file.
3116  */
3117 #ifndef _SYS_SYSPROTO_H_
3118 struct futimes_args {
3119 	int	fd;
3120 	struct	timeval *tptr;
3121 };
3122 #endif
3123 int
3124 sys_futimes(td, uap)
3125 	struct thread *td;
3126 	register struct futimes_args /* {
3127 		int  fd;
3128 		struct timeval *tptr;
3129 	} */ *uap;
3130 {
3131 
3132 	return (kern_futimes(td, uap->fd, uap->tptr, UIO_USERSPACE));
3133 }
3134 
3135 int
3136 kern_futimes(struct thread *td, int fd, struct timeval *tptr,
3137     enum uio_seg tptrseg)
3138 {
3139 	struct timespec ts[2];
3140 	struct file *fp;
3141 	cap_rights_t rights;
3142 	int error;
3143 
3144 	AUDIT_ARG_FD(fd);
3145 	error = getutimes(tptr, tptrseg, ts);
3146 	if (error != 0)
3147 		return (error);
3148 	error = getvnode(td, fd, cap_rights_init(&rights, CAP_FUTIMES), &fp);
3149 	if (error != 0)
3150 		return (error);
3151 #ifdef AUDIT
3152 	vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
3153 	AUDIT_ARG_VNODE1(fp->f_vnode);
3154 	VOP_UNLOCK(fp->f_vnode, 0);
3155 #endif
3156 	error = setutimes(td, fp->f_vnode, ts, 2, tptr == NULL);
3157 	fdrop(fp, td);
3158 	return (error);
3159 }
3160 
3161 int
3162 sys_futimens(struct thread *td, struct futimens_args *uap)
3163 {
3164 
3165 	return (kern_futimens(td, uap->fd, uap->times, UIO_USERSPACE));
3166 }
3167 
3168 int
3169 kern_futimens(struct thread *td, int fd, struct timespec *tptr,
3170     enum uio_seg tptrseg)
3171 {
3172 	struct timespec ts[2];
3173 	struct file *fp;
3174 	cap_rights_t rights;
3175 	int error, flags;
3176 
3177 	AUDIT_ARG_FD(fd);
3178 	error = getutimens(tptr, tptrseg, ts, &flags);
3179 	if (error != 0)
3180 		return (error);
3181 	if (flags & UTIMENS_EXIT)
3182 		return (0);
3183 	error = getvnode(td, fd, cap_rights_init(&rights, CAP_FUTIMES), &fp);
3184 	if (error != 0)
3185 		return (error);
3186 #ifdef AUDIT
3187 	vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
3188 	AUDIT_ARG_VNODE1(fp->f_vnode);
3189 	VOP_UNLOCK(fp->f_vnode, 0);
3190 #endif
3191 	error = setutimes(td, fp->f_vnode, ts, 2, flags & UTIMENS_NULL);
3192 	fdrop(fp, td);
3193 	return (error);
3194 }
3195 
3196 int
3197 sys_utimensat(struct thread *td, struct utimensat_args *uap)
3198 {
3199 
3200 	return (kern_utimensat(td, uap->fd, uap->path, UIO_USERSPACE,
3201 	    uap->times, UIO_USERSPACE, uap->flag));
3202 }
3203 
3204 int
3205 kern_utimensat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
3206     struct timespec *tptr, enum uio_seg tptrseg, int flag)
3207 {
3208 	struct nameidata nd;
3209 	struct timespec ts[2];
3210 	cap_rights_t rights;
3211 	int error, flags;
3212 
3213 	if (flag & ~AT_SYMLINK_NOFOLLOW)
3214 		return (EINVAL);
3215 
3216 	if ((error = getutimens(tptr, tptrseg, ts, &flags)) != 0)
3217 		return (error);
3218 	NDINIT_ATRIGHTS(&nd, LOOKUP, ((flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW :
3219 	    FOLLOW) | AUDITVNODE1, pathseg, path, fd,
3220 	    cap_rights_init(&rights, CAP_FUTIMES), td);
3221 	if ((error = namei(&nd)) != 0)
3222 		return (error);
3223 	/*
3224 	 * We are allowed to call namei() regardless of 2xUTIME_OMIT.
3225 	 * POSIX states:
3226 	 * "If both tv_nsec fields are UTIME_OMIT... EACCESS may be detected."
3227 	 * "Search permission is denied by a component of the path prefix."
3228 	 */
3229 	NDFREE(&nd, NDF_ONLY_PNBUF);
3230 	if ((flags & UTIMENS_EXIT) == 0)
3231 		error = setutimes(td, nd.ni_vp, ts, 2, flags & UTIMENS_NULL);
3232 	vrele(nd.ni_vp);
3233 	return (error);
3234 }
3235 
3236 /*
3237  * Truncate a file given its path name.
3238  */
3239 #ifndef _SYS_SYSPROTO_H_
3240 struct truncate_args {
3241 	char	*path;
3242 	int	pad;
3243 	off_t	length;
3244 };
3245 #endif
3246 int
3247 sys_truncate(td, uap)
3248 	struct thread *td;
3249 	register struct truncate_args /* {
3250 		char *path;
3251 		int pad;
3252 		off_t length;
3253 	} */ *uap;
3254 {
3255 
3256 	return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3257 }
3258 
3259 int
3260 kern_truncate(struct thread *td, char *path, enum uio_seg pathseg, off_t length)
3261 {
3262 	struct mount *mp;
3263 	struct vnode *vp;
3264 	void *rl_cookie;
3265 	struct vattr vattr;
3266 	struct nameidata nd;
3267 	int error;
3268 
3269 	if (length < 0)
3270 		return(EINVAL);
3271 	NDINIT(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, td);
3272 	if ((error = namei(&nd)) != 0)
3273 		return (error);
3274 	vp = nd.ni_vp;
3275 	rl_cookie = vn_rangelock_wlock(vp, 0, OFF_MAX);
3276 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) {
3277 		vn_rangelock_unlock(vp, rl_cookie);
3278 		vrele(vp);
3279 		return (error);
3280 	}
3281 	NDFREE(&nd, NDF_ONLY_PNBUF);
3282 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3283 	if (vp->v_type == VDIR)
3284 		error = EISDIR;
3285 #ifdef MAC
3286 	else if ((error = mac_vnode_check_write(td->td_ucred, NOCRED, vp))) {
3287 	}
3288 #endif
3289 	else if ((error = vn_writechk(vp)) == 0 &&
3290 	    (error = VOP_ACCESS(vp, VWRITE, td->td_ucred, td)) == 0) {
3291 		VATTR_NULL(&vattr);
3292 		vattr.va_size = length;
3293 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3294 	}
3295 	VOP_UNLOCK(vp, 0);
3296 	vn_finished_write(mp);
3297 	vn_rangelock_unlock(vp, rl_cookie);
3298 	vrele(vp);
3299 	return (error);
3300 }
3301 
3302 #if defined(COMPAT_43)
3303 /*
3304  * Truncate a file given its path name.
3305  */
3306 #ifndef _SYS_SYSPROTO_H_
3307 struct otruncate_args {
3308 	char	*path;
3309 	long	length;
3310 };
3311 #endif
3312 int
3313 otruncate(td, uap)
3314 	struct thread *td;
3315 	register struct otruncate_args /* {
3316 		char *path;
3317 		long length;
3318 	} */ *uap;
3319 {
3320 	struct truncate_args /* {
3321 		char *path;
3322 		int pad;
3323 		off_t length;
3324 	} */ nuap;
3325 
3326 	nuap.path = uap->path;
3327 	nuap.length = uap->length;
3328 	return (sys_truncate(td, &nuap));
3329 }
3330 #endif /* COMPAT_43 */
3331 
3332 #if defined(COMPAT_FREEBSD6)
3333 /* Versions with the pad argument */
3334 int
3335 freebsd6_truncate(struct thread *td, struct freebsd6_truncate_args *uap)
3336 {
3337 	struct truncate_args ouap;
3338 
3339 	ouap.path = uap->path;
3340 	ouap.length = uap->length;
3341 	return (sys_truncate(td, &ouap));
3342 }
3343 
3344 int
3345 freebsd6_ftruncate(struct thread *td, struct freebsd6_ftruncate_args *uap)
3346 {
3347 	struct ftruncate_args ouap;
3348 
3349 	ouap.fd = uap->fd;
3350 	ouap.length = uap->length;
3351 	return (sys_ftruncate(td, &ouap));
3352 }
3353 #endif
3354 
3355 /*
3356  * Sync an open file.
3357  */
3358 #ifndef _SYS_SYSPROTO_H_
3359 struct fsync_args {
3360 	int	fd;
3361 };
3362 #endif
3363 int
3364 sys_fsync(td, uap)
3365 	struct thread *td;
3366 	struct fsync_args /* {
3367 		int fd;
3368 	} */ *uap;
3369 {
3370 	struct vnode *vp;
3371 	struct mount *mp;
3372 	struct file *fp;
3373 	cap_rights_t rights;
3374 	int error, lock_flags;
3375 
3376 	AUDIT_ARG_FD(uap->fd);
3377 	error = getvnode(td, uap->fd, cap_rights_init(&rights, CAP_FSYNC), &fp);
3378 	if (error != 0)
3379 		return (error);
3380 	vp = fp->f_vnode;
3381 	error = vn_start_write(vp, &mp, V_WAIT | PCATCH);
3382 	if (error != 0)
3383 		goto drop;
3384 	if (MNT_SHARED_WRITES(mp) ||
3385 	    ((mp == NULL) && MNT_SHARED_WRITES(vp->v_mount))) {
3386 		lock_flags = LK_SHARED;
3387 	} else {
3388 		lock_flags = LK_EXCLUSIVE;
3389 	}
3390 	vn_lock(vp, lock_flags | LK_RETRY);
3391 	AUDIT_ARG_VNODE1(vp);
3392 	if (vp->v_object != NULL) {
3393 		VM_OBJECT_WLOCK(vp->v_object);
3394 		vm_object_page_clean(vp->v_object, 0, 0, 0);
3395 		VM_OBJECT_WUNLOCK(vp->v_object);
3396 	}
3397 	error = VOP_FSYNC(vp, MNT_WAIT, td);
3398 
3399 	VOP_UNLOCK(vp, 0);
3400 	vn_finished_write(mp);
3401 drop:
3402 	fdrop(fp, td);
3403 	return (error);
3404 }
3405 
3406 /*
3407  * Rename files.  Source and destination must either both be directories, or
3408  * both not be directories.  If target is a directory, it must be empty.
3409  */
3410 #ifndef _SYS_SYSPROTO_H_
3411 struct rename_args {
3412 	char	*from;
3413 	char	*to;
3414 };
3415 #endif
3416 int
3417 sys_rename(td, uap)
3418 	struct thread *td;
3419 	register struct rename_args /* {
3420 		char *from;
3421 		char *to;
3422 	} */ *uap;
3423 {
3424 
3425 	return (kern_renameat(td, AT_FDCWD, uap->from, AT_FDCWD,
3426 	    uap->to, UIO_USERSPACE));
3427 }
3428 
3429 #ifndef _SYS_SYSPROTO_H_
3430 struct renameat_args {
3431 	int	oldfd;
3432 	char	*old;
3433 	int	newfd;
3434 	char	*new;
3435 };
3436 #endif
3437 int
3438 sys_renameat(struct thread *td, struct renameat_args *uap)
3439 {
3440 
3441 	return (kern_renameat(td, uap->oldfd, uap->old, uap->newfd, uap->new,
3442 	    UIO_USERSPACE));
3443 }
3444 
3445 int
3446 kern_renameat(struct thread *td, int oldfd, char *old, int newfd, char *new,
3447     enum uio_seg pathseg)
3448 {
3449 	struct mount *mp = NULL;
3450 	struct vnode *tvp, *fvp, *tdvp;
3451 	struct nameidata fromnd, tond;
3452 	cap_rights_t rights;
3453 	int error;
3454 
3455 again:
3456 	bwillwrite();
3457 #ifdef MAC
3458 	NDINIT_ATRIGHTS(&fromnd, DELETE, LOCKPARENT | LOCKLEAF | SAVESTART |
3459 	    AUDITVNODE1, pathseg, old, oldfd,
3460 	    cap_rights_init(&rights, CAP_RENAMEAT_SOURCE), td);
3461 #else
3462 	NDINIT_ATRIGHTS(&fromnd, DELETE, WANTPARENT | SAVESTART | AUDITVNODE1,
3463 	    pathseg, old, oldfd,
3464 	    cap_rights_init(&rights, CAP_RENAMEAT_SOURCE), td);
3465 #endif
3466 
3467 	if ((error = namei(&fromnd)) != 0)
3468 		return (error);
3469 #ifdef MAC
3470 	error = mac_vnode_check_rename_from(td->td_ucred, fromnd.ni_dvp,
3471 	    fromnd.ni_vp, &fromnd.ni_cnd);
3472 	VOP_UNLOCK(fromnd.ni_dvp, 0);
3473 	if (fromnd.ni_dvp != fromnd.ni_vp)
3474 		VOP_UNLOCK(fromnd.ni_vp, 0);
3475 #endif
3476 	fvp = fromnd.ni_vp;
3477 	NDINIT_ATRIGHTS(&tond, RENAME, LOCKPARENT | LOCKLEAF | NOCACHE |
3478 	    SAVESTART | AUDITVNODE2, pathseg, new, newfd,
3479 	    cap_rights_init(&rights, CAP_RENAMEAT_TARGET), td);
3480 	if (fromnd.ni_vp->v_type == VDIR)
3481 		tond.ni_cnd.cn_flags |= WILLBEDIR;
3482 	if ((error = namei(&tond)) != 0) {
3483 		/* Translate error code for rename("dir1", "dir2/."). */
3484 		if (error == EISDIR && fvp->v_type == VDIR)
3485 			error = EINVAL;
3486 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3487 		vrele(fromnd.ni_dvp);
3488 		vrele(fvp);
3489 		goto out1;
3490 	}
3491 	tdvp = tond.ni_dvp;
3492 	tvp = tond.ni_vp;
3493 	error = vn_start_write(fvp, &mp, V_NOWAIT);
3494 	if (error != 0) {
3495 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3496 		NDFREE(&tond, NDF_ONLY_PNBUF);
3497 		if (tvp != NULL)
3498 			vput(tvp);
3499 		if (tdvp == tvp)
3500 			vrele(tdvp);
3501 		else
3502 			vput(tdvp);
3503 		vrele(fromnd.ni_dvp);
3504 		vrele(fvp);
3505 		vrele(tond.ni_startdir);
3506 		if (fromnd.ni_startdir != NULL)
3507 			vrele(fromnd.ni_startdir);
3508 		error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH);
3509 		if (error != 0)
3510 			return (error);
3511 		goto again;
3512 	}
3513 	if (tvp != NULL) {
3514 		if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
3515 			error = ENOTDIR;
3516 			goto out;
3517 		} else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
3518 			error = EISDIR;
3519 			goto out;
3520 		}
3521 #ifdef CAPABILITIES
3522 		if (newfd != AT_FDCWD) {
3523 			/*
3524 			 * If the target already exists we require CAP_UNLINKAT
3525 			 * from 'newfd'.
3526 			 */
3527 			error = cap_check(&tond.ni_filecaps.fc_rights,
3528 			    cap_rights_init(&rights, CAP_UNLINKAT));
3529 			if (error != 0)
3530 				goto out;
3531 		}
3532 #endif
3533 	}
3534 	if (fvp == tdvp) {
3535 		error = EINVAL;
3536 		goto out;
3537 	}
3538 	/*
3539 	 * If the source is the same as the destination (that is, if they
3540 	 * are links to the same vnode), then there is nothing to do.
3541 	 */
3542 	if (fvp == tvp)
3543 		error = -1;
3544 #ifdef MAC
3545 	else
3546 		error = mac_vnode_check_rename_to(td->td_ucred, tdvp,
3547 		    tond.ni_vp, fromnd.ni_dvp == tdvp, &tond.ni_cnd);
3548 #endif
3549 out:
3550 	if (error == 0) {
3551 		error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
3552 		    tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
3553 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3554 		NDFREE(&tond, NDF_ONLY_PNBUF);
3555 	} else {
3556 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3557 		NDFREE(&tond, NDF_ONLY_PNBUF);
3558 		if (tvp != NULL)
3559 			vput(tvp);
3560 		if (tdvp == tvp)
3561 			vrele(tdvp);
3562 		else
3563 			vput(tdvp);
3564 		vrele(fromnd.ni_dvp);
3565 		vrele(fvp);
3566 	}
3567 	vrele(tond.ni_startdir);
3568 	vn_finished_write(mp);
3569 out1:
3570 	if (fromnd.ni_startdir)
3571 		vrele(fromnd.ni_startdir);
3572 	if (error == -1)
3573 		return (0);
3574 	return (error);
3575 }
3576 
3577 /*
3578  * Make a directory file.
3579  */
3580 #ifndef _SYS_SYSPROTO_H_
3581 struct mkdir_args {
3582 	char	*path;
3583 	int	mode;
3584 };
3585 #endif
3586 int
3587 sys_mkdir(td, uap)
3588 	struct thread *td;
3589 	register struct mkdir_args /* {
3590 		char *path;
3591 		int mode;
3592 	} */ *uap;
3593 {
3594 
3595 	return (kern_mkdirat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
3596 	    uap->mode));
3597 }
3598 
3599 #ifndef _SYS_SYSPROTO_H_
3600 struct mkdirat_args {
3601 	int	fd;
3602 	char	*path;
3603 	mode_t	mode;
3604 };
3605 #endif
3606 int
3607 sys_mkdirat(struct thread *td, struct mkdirat_args *uap)
3608 {
3609 
3610 	return (kern_mkdirat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode));
3611 }
3612 
3613 int
3614 kern_mkdirat(struct thread *td, int fd, char *path, enum uio_seg segflg,
3615     int mode)
3616 {
3617 	struct mount *mp;
3618 	struct vnode *vp;
3619 	struct vattr vattr;
3620 	struct nameidata nd;
3621 	cap_rights_t rights;
3622 	int error;
3623 
3624 	AUDIT_ARG_MODE(mode);
3625 restart:
3626 	bwillwrite();
3627 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
3628 	    NOCACHE, segflg, path, fd, cap_rights_init(&rights, CAP_MKDIRAT),
3629 	    td);
3630 	nd.ni_cnd.cn_flags |= WILLBEDIR;
3631 	if ((error = namei(&nd)) != 0)
3632 		return (error);
3633 	vp = nd.ni_vp;
3634 	if (vp != NULL) {
3635 		NDFREE(&nd, NDF_ONLY_PNBUF);
3636 		/*
3637 		 * XXX namei called with LOCKPARENT but not LOCKLEAF has
3638 		 * the strange behaviour of leaving the vnode unlocked
3639 		 * if the target is the same vnode as the parent.
3640 		 */
3641 		if (vp == nd.ni_dvp)
3642 			vrele(nd.ni_dvp);
3643 		else
3644 			vput(nd.ni_dvp);
3645 		vrele(vp);
3646 		return (EEXIST);
3647 	}
3648 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3649 		NDFREE(&nd, NDF_ONLY_PNBUF);
3650 		vput(nd.ni_dvp);
3651 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3652 			return (error);
3653 		goto restart;
3654 	}
3655 	VATTR_NULL(&vattr);
3656 	vattr.va_type = VDIR;
3657 	vattr.va_mode = (mode & ACCESSPERMS) &~ td->td_proc->p_fd->fd_cmask;
3658 #ifdef MAC
3659 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
3660 	    &vattr);
3661 	if (error != 0)
3662 		goto out;
3663 #endif
3664 	error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
3665 #ifdef MAC
3666 out:
3667 #endif
3668 	NDFREE(&nd, NDF_ONLY_PNBUF);
3669 	vput(nd.ni_dvp);
3670 	if (error == 0)
3671 		vput(nd.ni_vp);
3672 	vn_finished_write(mp);
3673 	return (error);
3674 }
3675 
3676 /*
3677  * Remove a directory file.
3678  */
3679 #ifndef _SYS_SYSPROTO_H_
3680 struct rmdir_args {
3681 	char	*path;
3682 };
3683 #endif
3684 int
3685 sys_rmdir(td, uap)
3686 	struct thread *td;
3687 	struct rmdir_args /* {
3688 		char *path;
3689 	} */ *uap;
3690 {
3691 
3692 	return (kern_rmdirat(td, AT_FDCWD, uap->path, UIO_USERSPACE));
3693 }
3694 
3695 int
3696 kern_rmdirat(struct thread *td, int fd, char *path, enum uio_seg pathseg)
3697 {
3698 	struct mount *mp;
3699 	struct vnode *vp;
3700 	struct nameidata nd;
3701 	cap_rights_t rights;
3702 	int error;
3703 
3704 restart:
3705 	bwillwrite();
3706 	NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1,
3707 	    pathseg, path, fd, cap_rights_init(&rights, CAP_UNLINKAT), td);
3708 	if ((error = namei(&nd)) != 0)
3709 		return (error);
3710 	vp = nd.ni_vp;
3711 	if (vp->v_type != VDIR) {
3712 		error = ENOTDIR;
3713 		goto out;
3714 	}
3715 	/*
3716 	 * No rmdir "." please.
3717 	 */
3718 	if (nd.ni_dvp == vp) {
3719 		error = EINVAL;
3720 		goto out;
3721 	}
3722 	/*
3723 	 * The root of a mounted filesystem cannot be deleted.
3724 	 */
3725 	if (vp->v_vflag & VV_ROOT) {
3726 		error = EBUSY;
3727 		goto out;
3728 	}
3729 #ifdef MAC
3730 	error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
3731 	    &nd.ni_cnd);
3732 	if (error != 0)
3733 		goto out;
3734 #endif
3735 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3736 		NDFREE(&nd, NDF_ONLY_PNBUF);
3737 		vput(vp);
3738 		if (nd.ni_dvp == vp)
3739 			vrele(nd.ni_dvp);
3740 		else
3741 			vput(nd.ni_dvp);
3742 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3743 			return (error);
3744 		goto restart;
3745 	}
3746 	vfs_notify_upper(vp, VFS_NOTIFY_UPPER_UNLINK);
3747 	error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
3748 	vn_finished_write(mp);
3749 out:
3750 	NDFREE(&nd, NDF_ONLY_PNBUF);
3751 	vput(vp);
3752 	if (nd.ni_dvp == vp)
3753 		vrele(nd.ni_dvp);
3754 	else
3755 		vput(nd.ni_dvp);
3756 	return (error);
3757 }
3758 
3759 #ifdef COMPAT_43
3760 /*
3761  * Read a block of directory entries in a filesystem independent format.
3762  */
3763 #ifndef _SYS_SYSPROTO_H_
3764 struct ogetdirentries_args {
3765 	int	fd;
3766 	char	*buf;
3767 	u_int	count;
3768 	long	*basep;
3769 };
3770 #endif
3771 int
3772 ogetdirentries(struct thread *td, struct ogetdirentries_args *uap)
3773 {
3774 	long loff;
3775 	int error;
3776 
3777 	error = kern_ogetdirentries(td, uap, &loff);
3778 	if (error == 0)
3779 		error = copyout(&loff, uap->basep, sizeof(long));
3780 	return (error);
3781 }
3782 
3783 int
3784 kern_ogetdirentries(struct thread *td, struct ogetdirentries_args *uap,
3785     long *ploff)
3786 {
3787 	struct vnode *vp;
3788 	struct file *fp;
3789 	struct uio auio, kuio;
3790 	struct iovec aiov, kiov;
3791 	struct dirent *dp, *edp;
3792 	cap_rights_t rights;
3793 	caddr_t dirbuf;
3794 	int error, eofflag, readcnt;
3795 	long loff;
3796 	off_t foffset;
3797 
3798 	/* XXX arbitrary sanity limit on `count'. */
3799 	if (uap->count > 64 * 1024)
3800 		return (EINVAL);
3801 	error = getvnode(td, uap->fd, cap_rights_init(&rights, CAP_READ), &fp);
3802 	if (error != 0)
3803 		return (error);
3804 	if ((fp->f_flag & FREAD) == 0) {
3805 		fdrop(fp, td);
3806 		return (EBADF);
3807 	}
3808 	vp = fp->f_vnode;
3809 	foffset = foffset_lock(fp, 0);
3810 unionread:
3811 	if (vp->v_type != VDIR) {
3812 		foffset_unlock(fp, foffset, 0);
3813 		fdrop(fp, td);
3814 		return (EINVAL);
3815 	}
3816 	aiov.iov_base = uap->buf;
3817 	aiov.iov_len = uap->count;
3818 	auio.uio_iov = &aiov;
3819 	auio.uio_iovcnt = 1;
3820 	auio.uio_rw = UIO_READ;
3821 	auio.uio_segflg = UIO_USERSPACE;
3822 	auio.uio_td = td;
3823 	auio.uio_resid = uap->count;
3824 	vn_lock(vp, LK_SHARED | LK_RETRY);
3825 	loff = auio.uio_offset = foffset;
3826 #ifdef MAC
3827 	error = mac_vnode_check_readdir(td->td_ucred, vp);
3828 	if (error != 0) {
3829 		VOP_UNLOCK(vp, 0);
3830 		foffset_unlock(fp, foffset, FOF_NOUPDATE);
3831 		fdrop(fp, td);
3832 		return (error);
3833 	}
3834 #endif
3835 #	if (BYTE_ORDER != LITTLE_ENDIAN)
3836 		if (vp->v_mount->mnt_maxsymlinklen <= 0) {
3837 			error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag,
3838 			    NULL, NULL);
3839 			foffset = auio.uio_offset;
3840 		} else
3841 #	endif
3842 	{
3843 		kuio = auio;
3844 		kuio.uio_iov = &kiov;
3845 		kuio.uio_segflg = UIO_SYSSPACE;
3846 		kiov.iov_len = uap->count;
3847 		dirbuf = malloc(uap->count, M_TEMP, M_WAITOK);
3848 		kiov.iov_base = dirbuf;
3849 		error = VOP_READDIR(vp, &kuio, fp->f_cred, &eofflag,
3850 			    NULL, NULL);
3851 		foffset = kuio.uio_offset;
3852 		if (error == 0) {
3853 			readcnt = uap->count - kuio.uio_resid;
3854 			edp = (struct dirent *)&dirbuf[readcnt];
3855 			for (dp = (struct dirent *)dirbuf; dp < edp; ) {
3856 #				if (BYTE_ORDER == LITTLE_ENDIAN)
3857 					/*
3858 					 * The expected low byte of
3859 					 * dp->d_namlen is our dp->d_type.
3860 					 * The high MBZ byte of dp->d_namlen
3861 					 * is our dp->d_namlen.
3862 					 */
3863 					dp->d_type = dp->d_namlen;
3864 					dp->d_namlen = 0;
3865 #				else
3866 					/*
3867 					 * The dp->d_type is the high byte
3868 					 * of the expected dp->d_namlen,
3869 					 * so must be zero'ed.
3870 					 */
3871 					dp->d_type = 0;
3872 #				endif
3873 				if (dp->d_reclen > 0) {
3874 					dp = (struct dirent *)
3875 					    ((char *)dp + dp->d_reclen);
3876 				} else {
3877 					error = EIO;
3878 					break;
3879 				}
3880 			}
3881 			if (dp >= edp)
3882 				error = uiomove(dirbuf, readcnt, &auio);
3883 		}
3884 		free(dirbuf, M_TEMP);
3885 	}
3886 	if (error != 0) {
3887 		VOP_UNLOCK(vp, 0);
3888 		foffset_unlock(fp, foffset, 0);
3889 		fdrop(fp, td);
3890 		return (error);
3891 	}
3892 	if (uap->count == auio.uio_resid &&
3893 	    (vp->v_vflag & VV_ROOT) &&
3894 	    (vp->v_mount->mnt_flag & MNT_UNION)) {
3895 		struct vnode *tvp = vp;
3896 		vp = vp->v_mount->mnt_vnodecovered;
3897 		VREF(vp);
3898 		fp->f_vnode = vp;
3899 		fp->f_data = vp;
3900 		foffset = 0;
3901 		vput(tvp);
3902 		goto unionread;
3903 	}
3904 	VOP_UNLOCK(vp, 0);
3905 	foffset_unlock(fp, foffset, 0);
3906 	fdrop(fp, td);
3907 	td->td_retval[0] = uap->count - auio.uio_resid;
3908 	if (error == 0)
3909 		*ploff = loff;
3910 	return (error);
3911 }
3912 #endif /* COMPAT_43 */
3913 
3914 /*
3915  * Read a block of directory entries in a filesystem independent format.
3916  */
3917 #ifndef _SYS_SYSPROTO_H_
3918 struct getdirentries_args {
3919 	int	fd;
3920 	char	*buf;
3921 	u_int	count;
3922 	long	*basep;
3923 };
3924 #endif
3925 int
3926 sys_getdirentries(td, uap)
3927 	struct thread *td;
3928 	register struct getdirentries_args /* {
3929 		int fd;
3930 		char *buf;
3931 		u_int count;
3932 		long *basep;
3933 	} */ *uap;
3934 {
3935 	long base;
3936 	int error;
3937 
3938 	error = kern_getdirentries(td, uap->fd, uap->buf, uap->count, &base,
3939 	    NULL, UIO_USERSPACE);
3940 	if (error != 0)
3941 		return (error);
3942 	if (uap->basep != NULL)
3943 		error = copyout(&base, uap->basep, sizeof(long));
3944 	return (error);
3945 }
3946 
3947 int
3948 kern_getdirentries(struct thread *td, int fd, char *buf, u_int count,
3949     long *basep, ssize_t *residp, enum uio_seg bufseg)
3950 {
3951 	struct vnode *vp;
3952 	struct file *fp;
3953 	struct uio auio;
3954 	struct iovec aiov;
3955 	cap_rights_t rights;
3956 	long loff;
3957 	int error, eofflag;
3958 	off_t foffset;
3959 
3960 	AUDIT_ARG_FD(fd);
3961 	if (count > IOSIZE_MAX)
3962 		return (EINVAL);
3963 	auio.uio_resid = count;
3964 	error = getvnode(td, fd, cap_rights_init(&rights, CAP_READ), &fp);
3965 	if (error != 0)
3966 		return (error);
3967 	if ((fp->f_flag & FREAD) == 0) {
3968 		fdrop(fp, td);
3969 		return (EBADF);
3970 	}
3971 	vp = fp->f_vnode;
3972 	foffset = foffset_lock(fp, 0);
3973 unionread:
3974 	if (vp->v_type != VDIR) {
3975 		error = EINVAL;
3976 		goto fail;
3977 	}
3978 	aiov.iov_base = buf;
3979 	aiov.iov_len = count;
3980 	auio.uio_iov = &aiov;
3981 	auio.uio_iovcnt = 1;
3982 	auio.uio_rw = UIO_READ;
3983 	auio.uio_segflg = bufseg;
3984 	auio.uio_td = td;
3985 	vn_lock(vp, LK_SHARED | LK_RETRY);
3986 	AUDIT_ARG_VNODE1(vp);
3987 	loff = auio.uio_offset = foffset;
3988 #ifdef MAC
3989 	error = mac_vnode_check_readdir(td->td_ucred, vp);
3990 	if (error == 0)
3991 #endif
3992 		error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, NULL,
3993 		    NULL);
3994 	foffset = auio.uio_offset;
3995 	if (error != 0) {
3996 		VOP_UNLOCK(vp, 0);
3997 		goto fail;
3998 	}
3999 	if (count == auio.uio_resid &&
4000 	    (vp->v_vflag & VV_ROOT) &&
4001 	    (vp->v_mount->mnt_flag & MNT_UNION)) {
4002 		struct vnode *tvp = vp;
4003 
4004 		vp = vp->v_mount->mnt_vnodecovered;
4005 		VREF(vp);
4006 		fp->f_vnode = vp;
4007 		fp->f_data = vp;
4008 		foffset = 0;
4009 		vput(tvp);
4010 		goto unionread;
4011 	}
4012 	VOP_UNLOCK(vp, 0);
4013 	*basep = loff;
4014 	if (residp != NULL)
4015 		*residp = auio.uio_resid;
4016 	td->td_retval[0] = count - auio.uio_resid;
4017 fail:
4018 	foffset_unlock(fp, foffset, 0);
4019 	fdrop(fp, td);
4020 	return (error);
4021 }
4022 
4023 #ifndef _SYS_SYSPROTO_H_
4024 struct getdents_args {
4025 	int fd;
4026 	char *buf;
4027 	size_t count;
4028 };
4029 #endif
4030 int
4031 sys_getdents(td, uap)
4032 	struct thread *td;
4033 	register struct getdents_args /* {
4034 		int fd;
4035 		char *buf;
4036 		u_int count;
4037 	} */ *uap;
4038 {
4039 	struct getdirentries_args ap;
4040 
4041 	ap.fd = uap->fd;
4042 	ap.buf = uap->buf;
4043 	ap.count = uap->count;
4044 	ap.basep = NULL;
4045 	return (sys_getdirentries(td, &ap));
4046 }
4047 
4048 /*
4049  * Set the mode mask for creation of filesystem nodes.
4050  */
4051 #ifndef _SYS_SYSPROTO_H_
4052 struct umask_args {
4053 	int	newmask;
4054 };
4055 #endif
4056 int
4057 sys_umask(td, uap)
4058 	struct thread *td;
4059 	struct umask_args /* {
4060 		int newmask;
4061 	} */ *uap;
4062 {
4063 	struct filedesc *fdp;
4064 
4065 	fdp = td->td_proc->p_fd;
4066 	FILEDESC_XLOCK(fdp);
4067 	td->td_retval[0] = fdp->fd_cmask;
4068 	fdp->fd_cmask = uap->newmask & ALLPERMS;
4069 	FILEDESC_XUNLOCK(fdp);
4070 	return (0);
4071 }
4072 
4073 /*
4074  * Void all references to file by ripping underlying filesystem away from
4075  * vnode.
4076  */
4077 #ifndef _SYS_SYSPROTO_H_
4078 struct revoke_args {
4079 	char	*path;
4080 };
4081 #endif
4082 int
4083 sys_revoke(td, uap)
4084 	struct thread *td;
4085 	register struct revoke_args /* {
4086 		char *path;
4087 	} */ *uap;
4088 {
4089 	struct vnode *vp;
4090 	struct vattr vattr;
4091 	struct nameidata nd;
4092 	int error;
4093 
4094 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
4095 	    uap->path, td);
4096 	if ((error = namei(&nd)) != 0)
4097 		return (error);
4098 	vp = nd.ni_vp;
4099 	NDFREE(&nd, NDF_ONLY_PNBUF);
4100 	if (vp->v_type != VCHR || vp->v_rdev == NULL) {
4101 		error = EINVAL;
4102 		goto out;
4103 	}
4104 #ifdef MAC
4105 	error = mac_vnode_check_revoke(td->td_ucred, vp);
4106 	if (error != 0)
4107 		goto out;
4108 #endif
4109 	error = VOP_GETATTR(vp, &vattr, td->td_ucred);
4110 	if (error != 0)
4111 		goto out;
4112 	if (td->td_ucred->cr_uid != vattr.va_uid) {
4113 		error = priv_check(td, PRIV_VFS_ADMIN);
4114 		if (error != 0)
4115 			goto out;
4116 	}
4117 	if (vcount(vp) > 1)
4118 		VOP_REVOKE(vp, REVOKEALL);
4119 out:
4120 	vput(vp);
4121 	return (error);
4122 }
4123 
4124 /*
4125  * Convert a user file descriptor to a kernel file entry and check that, if it
4126  * is a capability, the correct rights are present. A reference on the file
4127  * entry is held upon returning.
4128  */
4129 int
4130 getvnode(struct thread *td, int fd, cap_rights_t *rightsp, struct file **fpp)
4131 {
4132 	struct file *fp;
4133 	int error;
4134 
4135 	error = fget_unlocked(td->td_proc->p_fd, fd, rightsp, &fp, NULL);
4136 	if (error != 0)
4137 		return (error);
4138 
4139 	/*
4140 	 * The file could be not of the vnode type, or it may be not
4141 	 * yet fully initialized, in which case the f_vnode pointer
4142 	 * may be set, but f_ops is still badfileops.  E.g.,
4143 	 * devfs_open() transiently create such situation to
4144 	 * facilitate csw d_fdopen().
4145 	 *
4146 	 * Dupfdopen() handling in kern_openat() installs the
4147 	 * half-baked file into the process descriptor table, allowing
4148 	 * other thread to dereference it. Guard against the race by
4149 	 * checking f_ops.
4150 	 */
4151 	if (fp->f_vnode == NULL || fp->f_ops == &badfileops) {
4152 		fdrop(fp, td);
4153 		return (EINVAL);
4154 	}
4155 	*fpp = fp;
4156 	return (0);
4157 }
4158 
4159 
4160 /*
4161  * Get an (NFS) file handle.
4162  */
4163 #ifndef _SYS_SYSPROTO_H_
4164 struct lgetfh_args {
4165 	char	*fname;
4166 	fhandle_t *fhp;
4167 };
4168 #endif
4169 int
4170 sys_lgetfh(td, uap)
4171 	struct thread *td;
4172 	register struct lgetfh_args *uap;
4173 {
4174 	struct nameidata nd;
4175 	fhandle_t fh;
4176 	register struct vnode *vp;
4177 	int error;
4178 
4179 	error = priv_check(td, PRIV_VFS_GETFH);
4180 	if (error != 0)
4181 		return (error);
4182 	NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
4183 	    uap->fname, td);
4184 	error = namei(&nd);
4185 	if (error != 0)
4186 		return (error);
4187 	NDFREE(&nd, NDF_ONLY_PNBUF);
4188 	vp = nd.ni_vp;
4189 	bzero(&fh, sizeof(fh));
4190 	fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
4191 	error = VOP_VPTOFH(vp, &fh.fh_fid);
4192 	vput(vp);
4193 	if (error == 0)
4194 		error = copyout(&fh, uap->fhp, sizeof (fh));
4195 	return (error);
4196 }
4197 
4198 #ifndef _SYS_SYSPROTO_H_
4199 struct getfh_args {
4200 	char	*fname;
4201 	fhandle_t *fhp;
4202 };
4203 #endif
4204 int
4205 sys_getfh(td, uap)
4206 	struct thread *td;
4207 	register struct getfh_args *uap;
4208 {
4209 	struct nameidata nd;
4210 	fhandle_t fh;
4211 	register struct vnode *vp;
4212 	int error;
4213 
4214 	error = priv_check(td, PRIV_VFS_GETFH);
4215 	if (error != 0)
4216 		return (error);
4217 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
4218 	    uap->fname, td);
4219 	error = namei(&nd);
4220 	if (error != 0)
4221 		return (error);
4222 	NDFREE(&nd, NDF_ONLY_PNBUF);
4223 	vp = nd.ni_vp;
4224 	bzero(&fh, sizeof(fh));
4225 	fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
4226 	error = VOP_VPTOFH(vp, &fh.fh_fid);
4227 	vput(vp);
4228 	if (error == 0)
4229 		error = copyout(&fh, uap->fhp, sizeof (fh));
4230 	return (error);
4231 }
4232 
4233 /*
4234  * syscall for the rpc.lockd to use to translate a NFS file handle into an
4235  * open descriptor.
4236  *
4237  * warning: do not remove the priv_check() call or this becomes one giant
4238  * security hole.
4239  */
4240 #ifndef _SYS_SYSPROTO_H_
4241 struct fhopen_args {
4242 	const struct fhandle *u_fhp;
4243 	int flags;
4244 };
4245 #endif
4246 int
4247 sys_fhopen(td, uap)
4248 	struct thread *td;
4249 	struct fhopen_args /* {
4250 		const struct fhandle *u_fhp;
4251 		int flags;
4252 	} */ *uap;
4253 {
4254 	struct mount *mp;
4255 	struct vnode *vp;
4256 	struct fhandle fhp;
4257 	struct file *fp;
4258 	int fmode, error;
4259 	int indx;
4260 
4261 	error = priv_check(td, PRIV_VFS_FHOPEN);
4262 	if (error != 0)
4263 		return (error);
4264 	indx = -1;
4265 	fmode = FFLAGS(uap->flags);
4266 	/* why not allow a non-read/write open for our lockd? */
4267 	if (((fmode & (FREAD | FWRITE)) == 0) || (fmode & O_CREAT))
4268 		return (EINVAL);
4269 	error = copyin(uap->u_fhp, &fhp, sizeof(fhp));
4270 	if (error != 0)
4271 		return(error);
4272 	/* find the mount point */
4273 	mp = vfs_busyfs(&fhp.fh_fsid);
4274 	if (mp == NULL)
4275 		return (ESTALE);
4276 	/* now give me my vnode, it gets returned to me locked */
4277 	error = VFS_FHTOVP(mp, &fhp.fh_fid, LK_EXCLUSIVE, &vp);
4278 	vfs_unbusy(mp);
4279 	if (error != 0)
4280 		return (error);
4281 
4282 	error = falloc_noinstall(td, &fp);
4283 	if (error != 0) {
4284 		vput(vp);
4285 		return (error);
4286 	}
4287 	/*
4288 	 * An extra reference on `fp' has been held for us by
4289 	 * falloc_noinstall().
4290 	 */
4291 
4292 #ifdef INVARIANTS
4293 	td->td_dupfd = -1;
4294 #endif
4295 	error = vn_open_vnode(vp, fmode, td->td_ucred, td, fp);
4296 	if (error != 0) {
4297 		KASSERT(fp->f_ops == &badfileops,
4298 		    ("VOP_OPEN in fhopen() set f_ops"));
4299 		KASSERT(td->td_dupfd < 0,
4300 		    ("fhopen() encountered fdopen()"));
4301 
4302 		vput(vp);
4303 		goto bad;
4304 	}
4305 #ifdef INVARIANTS
4306 	td->td_dupfd = 0;
4307 #endif
4308 	fp->f_vnode = vp;
4309 	fp->f_seqcount = 1;
4310 	finit(fp, (fmode & FMASK) | (fp->f_flag & FHASLOCK), DTYPE_VNODE, vp,
4311 	    &vnops);
4312 	VOP_UNLOCK(vp, 0);
4313 	if ((fmode & O_TRUNC) != 0) {
4314 		error = fo_truncate(fp, 0, td->td_ucred, td);
4315 		if (error != 0)
4316 			goto bad;
4317 	}
4318 
4319 	error = finstall(td, fp, &indx, fmode, NULL);
4320 bad:
4321 	fdrop(fp, td);
4322 	td->td_retval[0] = indx;
4323 	return (error);
4324 }
4325 
4326 /*
4327  * Stat an (NFS) file handle.
4328  */
4329 #ifndef _SYS_SYSPROTO_H_
4330 struct fhstat_args {
4331 	struct fhandle *u_fhp;
4332 	struct stat *sb;
4333 };
4334 #endif
4335 int
4336 sys_fhstat(td, uap)
4337 	struct thread *td;
4338 	register struct fhstat_args /* {
4339 		struct fhandle *u_fhp;
4340 		struct stat *sb;
4341 	} */ *uap;
4342 {
4343 	struct stat sb;
4344 	struct fhandle fh;
4345 	int error;
4346 
4347 	error = copyin(uap->u_fhp, &fh, sizeof(fh));
4348 	if (error != 0)
4349 		return (error);
4350 	error = kern_fhstat(td, fh, &sb);
4351 	if (error == 0)
4352 		error = copyout(&sb, uap->sb, sizeof(sb));
4353 	return (error);
4354 }
4355 
4356 int
4357 kern_fhstat(struct thread *td, struct fhandle fh, struct stat *sb)
4358 {
4359 	struct mount *mp;
4360 	struct vnode *vp;
4361 	int error;
4362 
4363 	error = priv_check(td, PRIV_VFS_FHSTAT);
4364 	if (error != 0)
4365 		return (error);
4366 	if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4367 		return (ESTALE);
4368 	error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4369 	vfs_unbusy(mp);
4370 	if (error != 0)
4371 		return (error);
4372 	error = vn_stat(vp, sb, td->td_ucred, NOCRED, td);
4373 	vput(vp);
4374 	return (error);
4375 }
4376 
4377 /*
4378  * Implement fstatfs() for (NFS) file handles.
4379  */
4380 #ifndef _SYS_SYSPROTO_H_
4381 struct fhstatfs_args {
4382 	struct fhandle *u_fhp;
4383 	struct statfs *buf;
4384 };
4385 #endif
4386 int
4387 sys_fhstatfs(td, uap)
4388 	struct thread *td;
4389 	struct fhstatfs_args /* {
4390 		struct fhandle *u_fhp;
4391 		struct statfs *buf;
4392 	} */ *uap;
4393 {
4394 	struct statfs sf;
4395 	fhandle_t fh;
4396 	int error;
4397 
4398 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
4399 	if (error != 0)
4400 		return (error);
4401 	error = kern_fhstatfs(td, fh, &sf);
4402 	if (error != 0)
4403 		return (error);
4404 	return (copyout(&sf, uap->buf, sizeof(sf)));
4405 }
4406 
4407 int
4408 kern_fhstatfs(struct thread *td, fhandle_t fh, struct statfs *buf)
4409 {
4410 	struct statfs *sp;
4411 	struct mount *mp;
4412 	struct vnode *vp;
4413 	int error;
4414 
4415 	error = priv_check(td, PRIV_VFS_FHSTATFS);
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 	if (error != 0) {
4422 		vfs_unbusy(mp);
4423 		return (error);
4424 	}
4425 	vput(vp);
4426 	error = prison_canseemount(td->td_ucred, mp);
4427 	if (error != 0)
4428 		goto out;
4429 #ifdef MAC
4430 	error = mac_mount_check_stat(td->td_ucred, mp);
4431 	if (error != 0)
4432 		goto out;
4433 #endif
4434 	/*
4435 	 * Set these in case the underlying filesystem fails to do so.
4436 	 */
4437 	sp = &mp->mnt_stat;
4438 	sp->f_version = STATFS_VERSION;
4439 	sp->f_namemax = NAME_MAX;
4440 	sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
4441 	error = VFS_STATFS(mp, sp);
4442 	if (error == 0)
4443 		*buf = *sp;
4444 out:
4445 	vfs_unbusy(mp);
4446 	return (error);
4447 }
4448 
4449 int
4450 kern_posix_fallocate(struct thread *td, int fd, off_t offset, off_t len)
4451 {
4452 	struct file *fp;
4453 	struct mount *mp;
4454 	struct vnode *vp;
4455 	cap_rights_t rights;
4456 	off_t olen, ooffset;
4457 	int error;
4458 
4459 	if (offset < 0 || len <= 0)
4460 		return (EINVAL);
4461 	/* Check for wrap. */
4462 	if (offset > OFF_MAX - len)
4463 		return (EFBIG);
4464 	error = fget(td, fd, cap_rights_init(&rights, CAP_WRITE), &fp);
4465 	if (error != 0)
4466 		return (error);
4467 	if ((fp->f_ops->fo_flags & DFLAG_SEEKABLE) == 0) {
4468 		error = ESPIPE;
4469 		goto out;
4470 	}
4471 	if ((fp->f_flag & FWRITE) == 0) {
4472 		error = EBADF;
4473 		goto out;
4474 	}
4475 	if (fp->f_type != DTYPE_VNODE) {
4476 		error = ENODEV;
4477 		goto out;
4478 	}
4479 	vp = fp->f_vnode;
4480 	if (vp->v_type != VREG) {
4481 		error = ENODEV;
4482 		goto out;
4483 	}
4484 
4485 	/* Allocating blocks may take a long time, so iterate. */
4486 	for (;;) {
4487 		olen = len;
4488 		ooffset = offset;
4489 
4490 		bwillwrite();
4491 		mp = NULL;
4492 		error = vn_start_write(vp, &mp, V_WAIT | PCATCH);
4493 		if (error != 0)
4494 			break;
4495 		error = vn_lock(vp, LK_EXCLUSIVE);
4496 		if (error != 0) {
4497 			vn_finished_write(mp);
4498 			break;
4499 		}
4500 #ifdef MAC
4501 		error = mac_vnode_check_write(td->td_ucred, fp->f_cred, vp);
4502 		if (error == 0)
4503 #endif
4504 			error = VOP_ALLOCATE(vp, &offset, &len);
4505 		VOP_UNLOCK(vp, 0);
4506 		vn_finished_write(mp);
4507 
4508 		if (olen + ooffset != offset + len) {
4509 			panic("offset + len changed from %jx/%jx to %jx/%jx",
4510 			    ooffset, olen, offset, len);
4511 		}
4512 		if (error != 0 || len == 0)
4513 			break;
4514 		KASSERT(olen > len, ("Iteration did not make progress?"));
4515 		maybe_yield();
4516 	}
4517  out:
4518 	fdrop(fp, td);
4519 	return (error);
4520 }
4521 
4522 int
4523 sys_posix_fallocate(struct thread *td, struct posix_fallocate_args *uap)
4524 {
4525 	int error;
4526 
4527 	error = kern_posix_fallocate(td, uap->fd, uap->offset, uap->len);
4528 	return (kern_posix_error(td, error));
4529 }
4530 
4531 /*
4532  * Unlike madvise(2), we do not make a best effort to remember every
4533  * possible caching hint.  Instead, we remember the last setting with
4534  * the exception that we will allow POSIX_FADV_NORMAL to adjust the
4535  * region of any current setting.
4536  */
4537 int
4538 kern_posix_fadvise(struct thread *td, int fd, off_t offset, off_t len,
4539     int advice)
4540 {
4541 	struct fadvise_info *fa, *new;
4542 	struct file *fp;
4543 	struct vnode *vp;
4544 	cap_rights_t rights;
4545 	off_t end;
4546 	int error;
4547 
4548 	if (offset < 0 || len < 0 || offset > OFF_MAX - len)
4549 		return (EINVAL);
4550 	switch (advice) {
4551 	case POSIX_FADV_SEQUENTIAL:
4552 	case POSIX_FADV_RANDOM:
4553 	case POSIX_FADV_NOREUSE:
4554 		new = malloc(sizeof(*fa), M_FADVISE, M_WAITOK);
4555 		break;
4556 	case POSIX_FADV_NORMAL:
4557 	case POSIX_FADV_WILLNEED:
4558 	case POSIX_FADV_DONTNEED:
4559 		new = NULL;
4560 		break;
4561 	default:
4562 		return (EINVAL);
4563 	}
4564 	/* XXX: CAP_POSIX_FADVISE? */
4565 	error = fget(td, fd, cap_rights_init(&rights), &fp);
4566 	if (error != 0)
4567 		goto out;
4568 	if ((fp->f_ops->fo_flags & DFLAG_SEEKABLE) == 0) {
4569 		error = ESPIPE;
4570 		goto out;
4571 	}
4572 	if (fp->f_type != DTYPE_VNODE) {
4573 		error = ENODEV;
4574 		goto out;
4575 	}
4576 	vp = fp->f_vnode;
4577 	if (vp->v_type != VREG) {
4578 		error = ENODEV;
4579 		goto out;
4580 	}
4581 	if (len == 0)
4582 		end = OFF_MAX;
4583 	else
4584 		end = offset + len - 1;
4585 	switch (advice) {
4586 	case POSIX_FADV_SEQUENTIAL:
4587 	case POSIX_FADV_RANDOM:
4588 	case POSIX_FADV_NOREUSE:
4589 		/*
4590 		 * Try to merge any existing non-standard region with
4591 		 * this new region if possible, otherwise create a new
4592 		 * non-standard region for this request.
4593 		 */
4594 		mtx_pool_lock(mtxpool_sleep, fp);
4595 		fa = fp->f_advice;
4596 		if (fa != NULL && fa->fa_advice == advice &&
4597 		    ((fa->fa_start <= end && fa->fa_end >= offset) ||
4598 		    (end != OFF_MAX && fa->fa_start == end + 1) ||
4599 		    (fa->fa_end != OFF_MAX && fa->fa_end + 1 == offset))) {
4600 			if (offset < fa->fa_start)
4601 				fa->fa_start = offset;
4602 			if (end > fa->fa_end)
4603 				fa->fa_end = end;
4604 		} else {
4605 			new->fa_advice = advice;
4606 			new->fa_start = offset;
4607 			new->fa_end = end;
4608 			fp->f_advice = new;
4609 			new = fa;
4610 		}
4611 		mtx_pool_unlock(mtxpool_sleep, fp);
4612 		break;
4613 	case POSIX_FADV_NORMAL:
4614 		/*
4615 		 * If a the "normal" region overlaps with an existing
4616 		 * non-standard region, trim or remove the
4617 		 * non-standard region.
4618 		 */
4619 		mtx_pool_lock(mtxpool_sleep, fp);
4620 		fa = fp->f_advice;
4621 		if (fa != NULL) {
4622 			if (offset <= fa->fa_start && end >= fa->fa_end) {
4623 				new = fa;
4624 				fp->f_advice = NULL;
4625 			} else if (offset <= fa->fa_start &&
4626 			    end >= fa->fa_start)
4627 				fa->fa_start = end + 1;
4628 			else if (offset <= fa->fa_end && end >= fa->fa_end)
4629 				fa->fa_end = offset - 1;
4630 			else if (offset >= fa->fa_start && end <= fa->fa_end) {
4631 				/*
4632 				 * If the "normal" region is a middle
4633 				 * portion of the existing
4634 				 * non-standard region, just remove
4635 				 * the whole thing rather than picking
4636 				 * one side or the other to
4637 				 * preserve.
4638 				 */
4639 				new = fa;
4640 				fp->f_advice = NULL;
4641 			}
4642 		}
4643 		mtx_pool_unlock(mtxpool_sleep, fp);
4644 		break;
4645 	case POSIX_FADV_WILLNEED:
4646 	case POSIX_FADV_DONTNEED:
4647 		error = VOP_ADVISE(vp, offset, end, advice);
4648 		break;
4649 	}
4650 out:
4651 	if (fp != NULL)
4652 		fdrop(fp, td);
4653 	free(new, M_FADVISE);
4654 	return (error);
4655 }
4656 
4657 int
4658 sys_posix_fadvise(struct thread *td, struct posix_fadvise_args *uap)
4659 {
4660 	int error;
4661 
4662 	error = kern_posix_fadvise(td, uap->fd, uap->offset, uap->len,
4663 	    uap->advice);
4664 	return (kern_posix_error(td, error));
4665 }
4666