xref: /freebsd/sys/kern/vfs_syscalls.c (revision 65b71c48d226f3ef3eb1c4452dccc678a7d2ac96)
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  * The module initialization routine for POSIX asynchronous I/O will
110  * set this to the version of AIO that it implements.  (Zero means
111  * that it is not implemented.)  This value is used here by pathconf()
112  * and in kern_descrip.c by fpathconf().
113  */
114 int async_io_version;
115 
116 /*
117  * Sync each mounted filesystem.
118  */
119 #ifndef _SYS_SYSPROTO_H_
120 struct sync_args {
121 	int     dummy;
122 };
123 #endif
124 /* ARGSUSED */
125 int
126 sys_sync(td, uap)
127 	struct thread *td;
128 	struct sync_args *uap;
129 {
130 	struct mount *mp, *nmp;
131 	int save;
132 
133 	mtx_lock(&mountlist_mtx);
134 	for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
135 		if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK)) {
136 			nmp = TAILQ_NEXT(mp, mnt_list);
137 			continue;
138 		}
139 		if ((mp->mnt_flag & MNT_RDONLY) == 0 &&
140 		    vn_start_write(NULL, &mp, V_NOWAIT) == 0) {
141 			save = curthread_pflags_set(TDP_SYNCIO);
142 			vfs_msync(mp, MNT_NOWAIT);
143 			VFS_SYNC(mp, MNT_NOWAIT);
144 			curthread_pflags_restore(save);
145 			vn_finished_write(mp);
146 		}
147 		mtx_lock(&mountlist_mtx);
148 		nmp = TAILQ_NEXT(mp, mnt_list);
149 		vfs_unbusy(mp);
150 	}
151 	mtx_unlock(&mountlist_mtx);
152 	return (0);
153 }
154 
155 /*
156  * Change filesystem quotas.
157  */
158 #ifndef _SYS_SYSPROTO_H_
159 struct quotactl_args {
160 	char *path;
161 	int cmd;
162 	int uid;
163 	caddr_t arg;
164 };
165 #endif
166 int
167 sys_quotactl(td, uap)
168 	struct thread *td;
169 	register struct quotactl_args /* {
170 		char *path;
171 		int cmd;
172 		int uid;
173 		caddr_t arg;
174 	} */ *uap;
175 {
176 	struct mount *mp;
177 	struct nameidata nd;
178 	int error;
179 
180 	AUDIT_ARG_CMD(uap->cmd);
181 	AUDIT_ARG_UID(uap->uid);
182 	if (!prison_allow(td->td_ucred, PR_ALLOW_QUOTAS))
183 		return (EPERM);
184 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
185 	    uap->path, td);
186 	if ((error = namei(&nd)) != 0)
187 		return (error);
188 	NDFREE(&nd, NDF_ONLY_PNBUF);
189 	mp = nd.ni_vp->v_mount;
190 	vfs_ref(mp);
191 	vput(nd.ni_vp);
192 	error = vfs_busy(mp, 0);
193 	vfs_rel(mp);
194 	if (error != 0)
195 		return (error);
196 	error = VFS_QUOTACTL(mp, uap->cmd, uap->uid, uap->arg);
197 
198 	/*
199 	 * Since quota on operation typically needs to open quota
200 	 * file, the Q_QUOTAON handler needs to unbusy the mount point
201 	 * before calling into namei.  Otherwise, unmount might be
202 	 * started between two vfs_busy() invocations (first is our,
203 	 * second is from mount point cross-walk code in lookup()),
204 	 * causing deadlock.
205 	 *
206 	 * Require that Q_QUOTAON handles the vfs_busy() reference on
207 	 * its own, always returning with ubusied mount point.
208 	 */
209 	if ((uap->cmd >> SUBCMDSHIFT) != Q_QUOTAON)
210 		vfs_unbusy(mp);
211 	return (error);
212 }
213 
214 /*
215  * Used by statfs conversion routines to scale the block size up if
216  * necessary so that all of the block counts are <= 'max_size'.  Note
217  * that 'max_size' should be a bitmask, i.e. 2^n - 1 for some non-zero
218  * value of 'n'.
219  */
220 void
221 statfs_scale_blocks(struct statfs *sf, long max_size)
222 {
223 	uint64_t count;
224 	int shift;
225 
226 	KASSERT(powerof2(max_size + 1), ("%s: invalid max_size", __func__));
227 
228 	/*
229 	 * Attempt to scale the block counts to give a more accurate
230 	 * overview to userland of the ratio of free space to used
231 	 * space.  To do this, find the largest block count and compute
232 	 * a divisor that lets it fit into a signed integer <= max_size.
233 	 */
234 	if (sf->f_bavail < 0)
235 		count = -sf->f_bavail;
236 	else
237 		count = sf->f_bavail;
238 	count = MAX(sf->f_blocks, MAX(sf->f_bfree, count));
239 	if (count <= max_size)
240 		return;
241 
242 	count >>= flsl(max_size);
243 	shift = 0;
244 	while (count > 0) {
245 		shift++;
246 		count >>=1;
247 	}
248 
249 	sf->f_bsize <<= shift;
250 	sf->f_blocks >>= shift;
251 	sf->f_bfree >>= shift;
252 	sf->f_bavail >>= shift;
253 }
254 
255 /*
256  * Get filesystem statistics.
257  */
258 #ifndef _SYS_SYSPROTO_H_
259 struct statfs_args {
260 	char *path;
261 	struct statfs *buf;
262 };
263 #endif
264 int
265 sys_statfs(td, uap)
266 	struct thread *td;
267 	register struct statfs_args /* {
268 		char *path;
269 		struct statfs *buf;
270 	} */ *uap;
271 {
272 	struct statfs sf;
273 	int error;
274 
275 	error = kern_statfs(td, uap->path, UIO_USERSPACE, &sf);
276 	if (error == 0)
277 		error = copyout(&sf, uap->buf, sizeof(sf));
278 	return (error);
279 }
280 
281 int
282 kern_statfs(struct thread *td, char *path, enum uio_seg pathseg,
283     struct statfs *buf)
284 {
285 	struct mount *mp;
286 	struct statfs *sp, sb;
287 	struct nameidata nd;
288 	int error;
289 
290 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
291 	    pathseg, path, td);
292 	error = namei(&nd);
293 	if (error != 0)
294 		return (error);
295 	mp = nd.ni_vp->v_mount;
296 	vfs_ref(mp);
297 	NDFREE(&nd, NDF_ONLY_PNBUF);
298 	vput(nd.ni_vp);
299 	error = vfs_busy(mp, 0);
300 	vfs_rel(mp);
301 	if (error != 0)
302 		return (error);
303 #ifdef MAC
304 	error = mac_mount_check_stat(td->td_ucred, mp);
305 	if (error != 0)
306 		goto out;
307 #endif
308 	/*
309 	 * Set these in case the underlying filesystem fails to do so.
310 	 */
311 	sp = &mp->mnt_stat;
312 	sp->f_version = STATFS_VERSION;
313 	sp->f_namemax = NAME_MAX;
314 	sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
315 	error = VFS_STATFS(mp, sp);
316 	if (error != 0)
317 		goto out;
318 	if (priv_check(td, PRIV_VFS_GENERATION)) {
319 		bcopy(sp, &sb, sizeof(sb));
320 		sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
321 		prison_enforce_statfs(td->td_ucred, mp, &sb);
322 		sp = &sb;
323 	}
324 	*buf = *sp;
325 out:
326 	vfs_unbusy(mp);
327 	return (error);
328 }
329 
330 /*
331  * Get filesystem statistics.
332  */
333 #ifndef _SYS_SYSPROTO_H_
334 struct fstatfs_args {
335 	int fd;
336 	struct statfs *buf;
337 };
338 #endif
339 int
340 sys_fstatfs(td, uap)
341 	struct thread *td;
342 	register struct fstatfs_args /* {
343 		int fd;
344 		struct statfs *buf;
345 	} */ *uap;
346 {
347 	struct statfs sf;
348 	int error;
349 
350 	error = kern_fstatfs(td, uap->fd, &sf);
351 	if (error == 0)
352 		error = copyout(&sf, uap->buf, sizeof(sf));
353 	return (error);
354 }
355 
356 int
357 kern_fstatfs(struct thread *td, int fd, struct statfs *buf)
358 {
359 	struct file *fp;
360 	struct mount *mp;
361 	struct statfs *sp, sb;
362 	struct vnode *vp;
363 	cap_rights_t rights;
364 	int error;
365 
366 	AUDIT_ARG_FD(fd);
367 	error = getvnode(td, fd, cap_rights_init(&rights, CAP_FSTATFS), &fp);
368 	if (error != 0)
369 		return (error);
370 	vp = fp->f_vnode;
371 	vn_lock(vp, LK_SHARED | LK_RETRY);
372 #ifdef AUDIT
373 	AUDIT_ARG_VNODE1(vp);
374 #endif
375 	mp = vp->v_mount;
376 	if (mp)
377 		vfs_ref(mp);
378 	VOP_UNLOCK(vp, 0);
379 	fdrop(fp, td);
380 	if (mp == NULL) {
381 		error = EBADF;
382 		goto out;
383 	}
384 	error = vfs_busy(mp, 0);
385 	vfs_rel(mp);
386 	if (error != 0)
387 		return (error);
388 #ifdef MAC
389 	error = mac_mount_check_stat(td->td_ucred, mp);
390 	if (error != 0)
391 		goto out;
392 #endif
393 	/*
394 	 * Set these in case the underlying filesystem fails to do so.
395 	 */
396 	sp = &mp->mnt_stat;
397 	sp->f_version = STATFS_VERSION;
398 	sp->f_namemax = NAME_MAX;
399 	sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
400 	error = VFS_STATFS(mp, sp);
401 	if (error != 0)
402 		goto out;
403 	if (priv_check(td, PRIV_VFS_GENERATION)) {
404 		bcopy(sp, &sb, sizeof(sb));
405 		sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
406 		prison_enforce_statfs(td->td_ucred, mp, &sb);
407 		sp = &sb;
408 	}
409 	*buf = *sp;
410 out:
411 	if (mp)
412 		vfs_unbusy(mp);
413 	return (error);
414 }
415 
416 /*
417  * Get statistics on all filesystems.
418  */
419 #ifndef _SYS_SYSPROTO_H_
420 struct getfsstat_args {
421 	struct statfs *buf;
422 	long bufsize;
423 	int flags;
424 };
425 #endif
426 int
427 sys_getfsstat(td, uap)
428 	struct thread *td;
429 	register struct getfsstat_args /* {
430 		struct statfs *buf;
431 		long bufsize;
432 		int flags;
433 	} */ *uap;
434 {
435 	size_t count;
436 	int error;
437 
438 	error = kern_getfsstat(td, &uap->buf, uap->bufsize, &count,
439 	    UIO_USERSPACE, uap->flags);
440 	if (error == 0)
441 		td->td_retval[0] = count;
442 	return (error);
443 }
444 
445 /*
446  * If (bufsize > 0 && bufseg == UIO_SYSSPACE)
447  *	The caller is responsible for freeing memory which will be allocated
448  *	in '*buf'.
449  */
450 int
451 kern_getfsstat(struct thread *td, struct statfs **buf, size_t bufsize,
452     size_t *countp, enum uio_seg bufseg, int flags)
453 {
454 	struct mount *mp, *nmp;
455 	struct statfs *sfsp, *sp, sb;
456 	size_t count, maxcount;
457 	int error;
458 
459 	maxcount = bufsize / sizeof(struct statfs);
460 	if (bufsize == 0)
461 		sfsp = NULL;
462 	else if (bufseg == UIO_USERSPACE)
463 		sfsp = *buf;
464 	else /* if (bufseg == UIO_SYSSPACE) */ {
465 		count = 0;
466 		mtx_lock(&mountlist_mtx);
467 		TAILQ_FOREACH(mp, &mountlist, mnt_list) {
468 			count++;
469 		}
470 		mtx_unlock(&mountlist_mtx);
471 		if (maxcount > count)
472 			maxcount = count;
473 		sfsp = *buf = malloc(maxcount * sizeof(struct statfs), M_TEMP,
474 		    M_WAITOK);
475 	}
476 	count = 0;
477 	mtx_lock(&mountlist_mtx);
478 	for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
479 		if (prison_canseemount(td->td_ucred, mp) != 0) {
480 			nmp = TAILQ_NEXT(mp, mnt_list);
481 			continue;
482 		}
483 #ifdef MAC
484 		if (mac_mount_check_stat(td->td_ucred, mp) != 0) {
485 			nmp = TAILQ_NEXT(mp, mnt_list);
486 			continue;
487 		}
488 #endif
489 		if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK)) {
490 			nmp = TAILQ_NEXT(mp, mnt_list);
491 			continue;
492 		}
493 		if (sfsp && count < maxcount) {
494 			sp = &mp->mnt_stat;
495 			/*
496 			 * Set these in case the underlying filesystem
497 			 * fails to do so.
498 			 */
499 			sp->f_version = STATFS_VERSION;
500 			sp->f_namemax = NAME_MAX;
501 			sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
502 			/*
503 			 * If MNT_NOWAIT or MNT_LAZY is specified, do not
504 			 * refresh the fsstat cache. MNT_NOWAIT or MNT_LAZY
505 			 * overrides MNT_WAIT.
506 			 */
507 			if (((flags & (MNT_LAZY|MNT_NOWAIT)) == 0 ||
508 			    (flags & MNT_WAIT)) &&
509 			    (error = VFS_STATFS(mp, sp))) {
510 				mtx_lock(&mountlist_mtx);
511 				nmp = TAILQ_NEXT(mp, mnt_list);
512 				vfs_unbusy(mp);
513 				continue;
514 			}
515 			if (priv_check(td, PRIV_VFS_GENERATION)) {
516 				bcopy(sp, &sb, sizeof(sb));
517 				sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
518 				prison_enforce_statfs(td->td_ucred, mp, &sb);
519 				sp = &sb;
520 			}
521 			if (bufseg == UIO_SYSSPACE)
522 				bcopy(sp, sfsp, sizeof(*sp));
523 			else /* if (bufseg == UIO_USERSPACE) */ {
524 				error = copyout(sp, sfsp, sizeof(*sp));
525 				if (error != 0) {
526 					vfs_unbusy(mp);
527 					return (error);
528 				}
529 			}
530 			sfsp++;
531 		}
532 		count++;
533 		mtx_lock(&mountlist_mtx);
534 		nmp = TAILQ_NEXT(mp, mnt_list);
535 		vfs_unbusy(mp);
536 	}
537 	mtx_unlock(&mountlist_mtx);
538 	if (sfsp && count > maxcount)
539 		*countp = maxcount;
540 	else
541 		*countp = count;
542 	return (0);
543 }
544 
545 #ifdef COMPAT_FREEBSD4
546 /*
547  * Get old format filesystem statistics.
548  */
549 static void cvtstatfs(struct statfs *, struct ostatfs *);
550 
551 #ifndef _SYS_SYSPROTO_H_
552 struct freebsd4_statfs_args {
553 	char *path;
554 	struct ostatfs *buf;
555 };
556 #endif
557 int
558 freebsd4_statfs(td, uap)
559 	struct thread *td;
560 	struct freebsd4_statfs_args /* {
561 		char *path;
562 		struct ostatfs *buf;
563 	} */ *uap;
564 {
565 	struct ostatfs osb;
566 	struct statfs sf;
567 	int error;
568 
569 	error = kern_statfs(td, uap->path, UIO_USERSPACE, &sf);
570 	if (error != 0)
571 		return (error);
572 	cvtstatfs(&sf, &osb);
573 	return (copyout(&osb, uap->buf, sizeof(osb)));
574 }
575 
576 /*
577  * Get filesystem statistics.
578  */
579 #ifndef _SYS_SYSPROTO_H_
580 struct freebsd4_fstatfs_args {
581 	int fd;
582 	struct ostatfs *buf;
583 };
584 #endif
585 int
586 freebsd4_fstatfs(td, uap)
587 	struct thread *td;
588 	struct freebsd4_fstatfs_args /* {
589 		int fd;
590 		struct ostatfs *buf;
591 	} */ *uap;
592 {
593 	struct ostatfs osb;
594 	struct statfs sf;
595 	int error;
596 
597 	error = kern_fstatfs(td, uap->fd, &sf);
598 	if (error != 0)
599 		return (error);
600 	cvtstatfs(&sf, &osb);
601 	return (copyout(&osb, uap->buf, sizeof(osb)));
602 }
603 
604 /*
605  * Get statistics on all filesystems.
606  */
607 #ifndef _SYS_SYSPROTO_H_
608 struct freebsd4_getfsstat_args {
609 	struct ostatfs *buf;
610 	long bufsize;
611 	int flags;
612 };
613 #endif
614 int
615 freebsd4_getfsstat(td, uap)
616 	struct thread *td;
617 	register struct freebsd4_getfsstat_args /* {
618 		struct ostatfs *buf;
619 		long bufsize;
620 		int flags;
621 	} */ *uap;
622 {
623 	struct statfs *buf, *sp;
624 	struct ostatfs osb;
625 	size_t count, size;
626 	int error;
627 
628 	count = uap->bufsize / sizeof(struct ostatfs);
629 	size = count * sizeof(struct statfs);
630 	error = kern_getfsstat(td, &buf, size, &count, UIO_SYSSPACE,
631 	    uap->flags);
632 	if (size > 0) {
633 		sp = buf;
634 		while (count > 0 && error == 0) {
635 			cvtstatfs(sp, &osb);
636 			error = copyout(&osb, uap->buf, sizeof(osb));
637 			sp++;
638 			uap->buf++;
639 			count--;
640 		}
641 		free(buf, M_TEMP);
642 	}
643 	if (error == 0)
644 		td->td_retval[0] = count;
645 	return (error);
646 }
647 
648 /*
649  * Implement fstatfs() for (NFS) file handles.
650  */
651 #ifndef _SYS_SYSPROTO_H_
652 struct freebsd4_fhstatfs_args {
653 	struct fhandle *u_fhp;
654 	struct ostatfs *buf;
655 };
656 #endif
657 int
658 freebsd4_fhstatfs(td, uap)
659 	struct thread *td;
660 	struct freebsd4_fhstatfs_args /* {
661 		struct fhandle *u_fhp;
662 		struct ostatfs *buf;
663 	} */ *uap;
664 {
665 	struct ostatfs osb;
666 	struct statfs sf;
667 	fhandle_t fh;
668 	int error;
669 
670 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
671 	if (error != 0)
672 		return (error);
673 	error = kern_fhstatfs(td, fh, &sf);
674 	if (error != 0)
675 		return (error);
676 	cvtstatfs(&sf, &osb);
677 	return (copyout(&osb, uap->buf, sizeof(osb)));
678 }
679 
680 /*
681  * Convert a new format statfs structure to an old format statfs structure.
682  */
683 static void
684 cvtstatfs(nsp, osp)
685 	struct statfs *nsp;
686 	struct ostatfs *osp;
687 {
688 
689 	statfs_scale_blocks(nsp, LONG_MAX);
690 	bzero(osp, sizeof(*osp));
691 	osp->f_bsize = nsp->f_bsize;
692 	osp->f_iosize = MIN(nsp->f_iosize, LONG_MAX);
693 	osp->f_blocks = nsp->f_blocks;
694 	osp->f_bfree = nsp->f_bfree;
695 	osp->f_bavail = nsp->f_bavail;
696 	osp->f_files = MIN(nsp->f_files, LONG_MAX);
697 	osp->f_ffree = MIN(nsp->f_ffree, LONG_MAX);
698 	osp->f_owner = nsp->f_owner;
699 	osp->f_type = nsp->f_type;
700 	osp->f_flags = nsp->f_flags;
701 	osp->f_syncwrites = MIN(nsp->f_syncwrites, LONG_MAX);
702 	osp->f_asyncwrites = MIN(nsp->f_asyncwrites, LONG_MAX);
703 	osp->f_syncreads = MIN(nsp->f_syncreads, LONG_MAX);
704 	osp->f_asyncreads = MIN(nsp->f_asyncreads, LONG_MAX);
705 	strlcpy(osp->f_fstypename, nsp->f_fstypename,
706 	    MIN(MFSNAMELEN, OMFSNAMELEN));
707 	strlcpy(osp->f_mntonname, nsp->f_mntonname,
708 	    MIN(MNAMELEN, OMNAMELEN));
709 	strlcpy(osp->f_mntfromname, nsp->f_mntfromname,
710 	    MIN(MNAMELEN, OMNAMELEN));
711 	osp->f_fsid = nsp->f_fsid;
712 }
713 #endif /* COMPAT_FREEBSD4 */
714 
715 /*
716  * Change current working directory to a given file descriptor.
717  */
718 #ifndef _SYS_SYSPROTO_H_
719 struct fchdir_args {
720 	int	fd;
721 };
722 #endif
723 int
724 sys_fchdir(td, uap)
725 	struct thread *td;
726 	struct fchdir_args /* {
727 		int fd;
728 	} */ *uap;
729 {
730 	struct vnode *vp, *tdp;
731 	struct mount *mp;
732 	struct file *fp;
733 	cap_rights_t rights;
734 	int error;
735 
736 	AUDIT_ARG_FD(uap->fd);
737 	error = getvnode(td, uap->fd, cap_rights_init(&rights, CAP_FCHDIR),
738 	    &fp);
739 	if (error != 0)
740 		return (error);
741 	vp = fp->f_vnode;
742 	VREF(vp);
743 	fdrop(fp, td);
744 	vn_lock(vp, LK_SHARED | LK_RETRY);
745 	AUDIT_ARG_VNODE1(vp);
746 	error = change_dir(vp, td);
747 	while (!error && (mp = vp->v_mountedhere) != NULL) {
748 		if (vfs_busy(mp, 0))
749 			continue;
750 		error = VFS_ROOT(mp, LK_SHARED, &tdp);
751 		vfs_unbusy(mp);
752 		if (error != 0)
753 			break;
754 		vput(vp);
755 		vp = tdp;
756 	}
757 	if (error != 0) {
758 		vput(vp);
759 		return (error);
760 	}
761 	VOP_UNLOCK(vp, 0);
762 	pwd_chdir(td, vp);
763 	return (0);
764 }
765 
766 /*
767  * Change current working directory (``.'').
768  */
769 #ifndef _SYS_SYSPROTO_H_
770 struct chdir_args {
771 	char	*path;
772 };
773 #endif
774 int
775 sys_chdir(td, uap)
776 	struct thread *td;
777 	struct chdir_args /* {
778 		char *path;
779 	} */ *uap;
780 {
781 
782 	return (kern_chdir(td, uap->path, UIO_USERSPACE));
783 }
784 
785 int
786 kern_chdir(struct thread *td, char *path, enum uio_seg pathseg)
787 {
788 	struct nameidata nd;
789 	int error;
790 
791 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
792 	    pathseg, path, td);
793 	if ((error = namei(&nd)) != 0)
794 		return (error);
795 	if ((error = change_dir(nd.ni_vp, td)) != 0) {
796 		vput(nd.ni_vp);
797 		NDFREE(&nd, NDF_ONLY_PNBUF);
798 		return (error);
799 	}
800 	VOP_UNLOCK(nd.ni_vp, 0);
801 	NDFREE(&nd, NDF_ONLY_PNBUF);
802 	pwd_chdir(td, nd.ni_vp);
803 	return (0);
804 }
805 
806 /*
807  * Change notion of root (``/'') directory.
808  */
809 #ifndef _SYS_SYSPROTO_H_
810 struct chroot_args {
811 	char	*path;
812 };
813 #endif
814 int
815 sys_chroot(td, uap)
816 	struct thread *td;
817 	struct chroot_args /* {
818 		char *path;
819 	} */ *uap;
820 {
821 	struct nameidata nd;
822 	int error;
823 
824 	error = priv_check(td, PRIV_VFS_CHROOT);
825 	if (error != 0)
826 		return (error);
827 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
828 	    UIO_USERSPACE, uap->path, td);
829 	error = namei(&nd);
830 	if (error != 0)
831 		goto error;
832 	error = change_dir(nd.ni_vp, td);
833 	if (error != 0)
834 		goto e_vunlock;
835 #ifdef MAC
836 	error = mac_vnode_check_chroot(td->td_ucred, nd.ni_vp);
837 	if (error != 0)
838 		goto e_vunlock;
839 #endif
840 	VOP_UNLOCK(nd.ni_vp, 0);
841 	error = pwd_chroot(td, nd.ni_vp);
842 	vrele(nd.ni_vp);
843 	NDFREE(&nd, NDF_ONLY_PNBUF);
844 	return (error);
845 e_vunlock:
846 	vput(nd.ni_vp);
847 error:
848 	NDFREE(&nd, NDF_ONLY_PNBUF);
849 	return (error);
850 }
851 
852 /*
853  * Common routine for chroot and chdir.  Callers must provide a locked vnode
854  * instance.
855  */
856 int
857 change_dir(vp, td)
858 	struct vnode *vp;
859 	struct thread *td;
860 {
861 #ifdef MAC
862 	int error;
863 #endif
864 
865 	ASSERT_VOP_LOCKED(vp, "change_dir(): vp not locked");
866 	if (vp->v_type != VDIR)
867 		return (ENOTDIR);
868 #ifdef MAC
869 	error = mac_vnode_check_chdir(td->td_ucred, vp);
870 	if (error != 0)
871 		return (error);
872 #endif
873 	return (VOP_ACCESS(vp, VEXEC, td->td_ucred, td));
874 }
875 
876 static __inline void
877 flags_to_rights(int flags, cap_rights_t *rightsp)
878 {
879 
880 	if (flags & O_EXEC) {
881 		cap_rights_set(rightsp, CAP_FEXECVE);
882 	} else {
883 		switch ((flags & O_ACCMODE)) {
884 		case O_RDONLY:
885 			cap_rights_set(rightsp, CAP_READ);
886 			break;
887 		case O_RDWR:
888 			cap_rights_set(rightsp, CAP_READ);
889 			/* FALLTHROUGH */
890 		case O_WRONLY:
891 			cap_rights_set(rightsp, CAP_WRITE);
892 			if (!(flags & (O_APPEND | O_TRUNC)))
893 				cap_rights_set(rightsp, CAP_SEEK);
894 			break;
895 		}
896 	}
897 
898 	if (flags & O_CREAT)
899 		cap_rights_set(rightsp, CAP_CREATE);
900 
901 	if (flags & O_TRUNC)
902 		cap_rights_set(rightsp, CAP_FTRUNCATE);
903 
904 	if (flags & (O_SYNC | O_FSYNC))
905 		cap_rights_set(rightsp, CAP_FSYNC);
906 
907 	if (flags & (O_EXLOCK | O_SHLOCK))
908 		cap_rights_set(rightsp, CAP_FLOCK);
909 }
910 
911 /*
912  * Check permissions, allocate an open file structure, and call the device
913  * open routine if any.
914  */
915 #ifndef _SYS_SYSPROTO_H_
916 struct open_args {
917 	char	*path;
918 	int	flags;
919 	int	mode;
920 };
921 #endif
922 int
923 sys_open(td, uap)
924 	struct thread *td;
925 	register struct open_args /* {
926 		char *path;
927 		int flags;
928 		int mode;
929 	} */ *uap;
930 {
931 
932 	return (kern_openat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
933 	    uap->flags, uap->mode));
934 }
935 
936 #ifndef _SYS_SYSPROTO_H_
937 struct openat_args {
938 	int	fd;
939 	char	*path;
940 	int	flag;
941 	int	mode;
942 };
943 #endif
944 int
945 sys_openat(struct thread *td, struct openat_args *uap)
946 {
947 
948 	return (kern_openat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
949 	    uap->mode));
950 }
951 
952 int
953 kern_openat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
954     int flags, int mode)
955 {
956 	struct proc *p = td->td_proc;
957 	struct filedesc *fdp = p->p_fd;
958 	struct file *fp;
959 	struct vnode *vp;
960 	struct nameidata nd;
961 	cap_rights_t rights;
962 	int cmode, error, indx;
963 
964 	indx = -1;
965 
966 	AUDIT_ARG_FFLAGS(flags);
967 	AUDIT_ARG_MODE(mode);
968 	/* XXX: audit dirfd */
969 	cap_rights_init(&rights, CAP_LOOKUP);
970 	flags_to_rights(flags, &rights);
971 	/*
972 	 * Only one of the O_EXEC, O_RDONLY, O_WRONLY and O_RDWR flags
973 	 * may be specified.
974 	 */
975 	if (flags & O_EXEC) {
976 		if (flags & O_ACCMODE)
977 			return (EINVAL);
978 	} else if ((flags & O_ACCMODE) == O_ACCMODE) {
979 		return (EINVAL);
980 	} else {
981 		flags = FFLAGS(flags);
982 	}
983 
984 	/*
985 	 * Allocate the file descriptor, but don't install a descriptor yet.
986 	 */
987 	error = falloc_noinstall(td, &fp);
988 	if (error != 0)
989 		return (error);
990 	/*
991 	 * An extra reference on `fp' has been held for us by
992 	 * falloc_noinstall().
993 	 */
994 	/* Set the flags early so the finit in devfs can pick them up. */
995 	fp->f_flag = flags & FMASK;
996 	cmode = ((mode & ~fdp->fd_cmask) & ALLPERMS) & ~S_ISTXT;
997 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, fd,
998 	    &rights, td);
999 	td->td_dupfd = -1;		/* XXX check for fdopen */
1000 	error = vn_open(&nd, &flags, cmode, fp);
1001 	if (error != 0) {
1002 		/*
1003 		 * If the vn_open replaced the method vector, something
1004 		 * wonderous happened deep below and we just pass it up
1005 		 * pretending we know what we do.
1006 		 */
1007 		if (error == ENXIO && fp->f_ops != &badfileops)
1008 			goto success;
1009 
1010 		/*
1011 		 * Handle special fdopen() case. bleh.
1012 		 *
1013 		 * Don't do this for relative (capability) lookups; we don't
1014 		 * understand exactly what would happen, and we don't think
1015 		 * that it ever should.
1016 		 */
1017 		if (nd.ni_strictrelative == 0 &&
1018 		    (error == ENODEV || error == ENXIO) &&
1019 		    td->td_dupfd >= 0) {
1020 			error = dupfdopen(td, fdp, td->td_dupfd, flags, error,
1021 			    &indx);
1022 			if (error == 0)
1023 				goto success;
1024 		}
1025 
1026 		goto bad;
1027 	}
1028 	td->td_dupfd = 0;
1029 	NDFREE(&nd, NDF_ONLY_PNBUF);
1030 	vp = nd.ni_vp;
1031 
1032 	/*
1033 	 * Store the vnode, for any f_type. Typically, the vnode use
1034 	 * count is decremented by direct call to vn_closefile() for
1035 	 * files that switched type in the cdevsw fdopen() method.
1036 	 */
1037 	fp->f_vnode = vp;
1038 	/*
1039 	 * If the file wasn't claimed by devfs bind it to the normal
1040 	 * vnode operations here.
1041 	 */
1042 	if (fp->f_ops == &badfileops) {
1043 		KASSERT(vp->v_type != VFIFO, ("Unexpected fifo."));
1044 		fp->f_seqcount = 1;
1045 		finit(fp, (flags & FMASK) | (fp->f_flag & FHASLOCK),
1046 		    DTYPE_VNODE, vp, &vnops);
1047 	}
1048 
1049 	VOP_UNLOCK(vp, 0);
1050 	if (flags & O_TRUNC) {
1051 		error = fo_truncate(fp, 0, td->td_ucred, td);
1052 		if (error != 0)
1053 			goto bad;
1054 	}
1055 success:
1056 	/*
1057 	 * If we haven't already installed the FD (for dupfdopen), do so now.
1058 	 */
1059 	if (indx == -1) {
1060 		struct filecaps *fcaps;
1061 
1062 #ifdef CAPABILITIES
1063 		if (nd.ni_strictrelative == 1)
1064 			fcaps = &nd.ni_filecaps;
1065 		else
1066 #endif
1067 			fcaps = NULL;
1068 		error = finstall(td, fp, &indx, flags, fcaps);
1069 		/* On success finstall() consumes fcaps. */
1070 		if (error != 0) {
1071 			filecaps_free(&nd.ni_filecaps);
1072 			goto bad;
1073 		}
1074 	} else {
1075 		filecaps_free(&nd.ni_filecaps);
1076 	}
1077 
1078 	/*
1079 	 * Release our private reference, leaving the one associated with
1080 	 * the descriptor table intact.
1081 	 */
1082 	fdrop(fp, td);
1083 	td->td_retval[0] = indx;
1084 	return (0);
1085 bad:
1086 	KASSERT(indx == -1, ("indx=%d, should be -1", indx));
1087 	fdrop(fp, td);
1088 	return (error);
1089 }
1090 
1091 #ifdef COMPAT_43
1092 /*
1093  * Create a file.
1094  */
1095 #ifndef _SYS_SYSPROTO_H_
1096 struct ocreat_args {
1097 	char	*path;
1098 	int	mode;
1099 };
1100 #endif
1101 int
1102 ocreat(td, uap)
1103 	struct thread *td;
1104 	register struct ocreat_args /* {
1105 		char *path;
1106 		int mode;
1107 	} */ *uap;
1108 {
1109 
1110 	return (kern_openat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1111 	    O_WRONLY | O_CREAT | O_TRUNC, uap->mode));
1112 }
1113 #endif /* COMPAT_43 */
1114 
1115 /*
1116  * Create a special file.
1117  */
1118 #ifndef _SYS_SYSPROTO_H_
1119 struct mknod_args {
1120 	char	*path;
1121 	int	mode;
1122 	int	dev;
1123 };
1124 #endif
1125 int
1126 sys_mknod(td, uap)
1127 	struct thread *td;
1128 	register struct mknod_args /* {
1129 		char *path;
1130 		int mode;
1131 		int dev;
1132 	} */ *uap;
1133 {
1134 
1135 	return (kern_mknodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1136 	    uap->mode, uap->dev));
1137 }
1138 
1139 #ifndef _SYS_SYSPROTO_H_
1140 struct mknodat_args {
1141 	int	fd;
1142 	char	*path;
1143 	mode_t	mode;
1144 	dev_t	dev;
1145 };
1146 #endif
1147 int
1148 sys_mknodat(struct thread *td, struct mknodat_args *uap)
1149 {
1150 
1151 	return (kern_mknodat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode,
1152 	    uap->dev));
1153 }
1154 
1155 int
1156 kern_mknodat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1157     int mode, int dev)
1158 {
1159 	struct vnode *vp;
1160 	struct mount *mp;
1161 	struct vattr vattr;
1162 	struct nameidata nd;
1163 	cap_rights_t rights;
1164 	int error, whiteout = 0;
1165 
1166 	AUDIT_ARG_MODE(mode);
1167 	AUDIT_ARG_DEV(dev);
1168 	switch (mode & S_IFMT) {
1169 	case S_IFCHR:
1170 	case S_IFBLK:
1171 		error = priv_check(td, PRIV_VFS_MKNOD_DEV);
1172 		break;
1173 	case S_IFMT:
1174 		error = priv_check(td, PRIV_VFS_MKNOD_BAD);
1175 		break;
1176 	case S_IFWHT:
1177 		error = priv_check(td, PRIV_VFS_MKNOD_WHT);
1178 		break;
1179 	case S_IFIFO:
1180 		if (dev == 0)
1181 			return (kern_mkfifoat(td, fd, path, pathseg, mode));
1182 		/* FALLTHROUGH */
1183 	default:
1184 		error = EINVAL;
1185 		break;
1186 	}
1187 	if (error != 0)
1188 		return (error);
1189 restart:
1190 	bwillwrite();
1191 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
1192 	    NOCACHE, pathseg, path, fd, cap_rights_init(&rights, CAP_MKNODAT),
1193 	    td);
1194 	if ((error = namei(&nd)) != 0)
1195 		return (error);
1196 	vp = nd.ni_vp;
1197 	if (vp != NULL) {
1198 		NDFREE(&nd, NDF_ONLY_PNBUF);
1199 		if (vp == nd.ni_dvp)
1200 			vrele(nd.ni_dvp);
1201 		else
1202 			vput(nd.ni_dvp);
1203 		vrele(vp);
1204 		return (EEXIST);
1205 	} else {
1206 		VATTR_NULL(&vattr);
1207 		vattr.va_mode = (mode & ALLPERMS) &
1208 		    ~td->td_proc->p_fd->fd_cmask;
1209 		vattr.va_rdev = dev;
1210 		whiteout = 0;
1211 
1212 		switch (mode & S_IFMT) {
1213 		case S_IFMT:	/* used by badsect to flag bad sectors */
1214 			vattr.va_type = VBAD;
1215 			break;
1216 		case S_IFCHR:
1217 			vattr.va_type = VCHR;
1218 			break;
1219 		case S_IFBLK:
1220 			vattr.va_type = VBLK;
1221 			break;
1222 		case S_IFWHT:
1223 			whiteout = 1;
1224 			break;
1225 		default:
1226 			panic("kern_mknod: invalid mode");
1227 		}
1228 	}
1229 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1230 		NDFREE(&nd, NDF_ONLY_PNBUF);
1231 		vput(nd.ni_dvp);
1232 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1233 			return (error);
1234 		goto restart;
1235 	}
1236 #ifdef MAC
1237 	if (error == 0 && !whiteout)
1238 		error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp,
1239 		    &nd.ni_cnd, &vattr);
1240 #endif
1241 	if (error == 0) {
1242 		if (whiteout)
1243 			error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE);
1244 		else {
1245 			error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp,
1246 						&nd.ni_cnd, &vattr);
1247 			if (error == 0)
1248 				vput(nd.ni_vp);
1249 		}
1250 	}
1251 	NDFREE(&nd, NDF_ONLY_PNBUF);
1252 	vput(nd.ni_dvp);
1253 	vn_finished_write(mp);
1254 	return (error);
1255 }
1256 
1257 /*
1258  * Create a named pipe.
1259  */
1260 #ifndef _SYS_SYSPROTO_H_
1261 struct mkfifo_args {
1262 	char	*path;
1263 	int	mode;
1264 };
1265 #endif
1266 int
1267 sys_mkfifo(td, uap)
1268 	struct thread *td;
1269 	register struct mkfifo_args /* {
1270 		char *path;
1271 		int mode;
1272 	} */ *uap;
1273 {
1274 
1275 	return (kern_mkfifoat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1276 	    uap->mode));
1277 }
1278 
1279 #ifndef _SYS_SYSPROTO_H_
1280 struct mkfifoat_args {
1281 	int	fd;
1282 	char	*path;
1283 	mode_t	mode;
1284 };
1285 #endif
1286 int
1287 sys_mkfifoat(struct thread *td, struct mkfifoat_args *uap)
1288 {
1289 
1290 	return (kern_mkfifoat(td, uap->fd, uap->path, UIO_USERSPACE,
1291 	    uap->mode));
1292 }
1293 
1294 int
1295 kern_mkfifoat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1296     int mode)
1297 {
1298 	struct mount *mp;
1299 	struct vattr vattr;
1300 	struct nameidata nd;
1301 	cap_rights_t rights;
1302 	int error;
1303 
1304 	AUDIT_ARG_MODE(mode);
1305 restart:
1306 	bwillwrite();
1307 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
1308 	    NOCACHE, pathseg, path, fd, cap_rights_init(&rights, CAP_MKFIFOAT),
1309 	    td);
1310 	if ((error = namei(&nd)) != 0)
1311 		return (error);
1312 	if (nd.ni_vp != NULL) {
1313 		NDFREE(&nd, NDF_ONLY_PNBUF);
1314 		if (nd.ni_vp == nd.ni_dvp)
1315 			vrele(nd.ni_dvp);
1316 		else
1317 			vput(nd.ni_dvp);
1318 		vrele(nd.ni_vp);
1319 		return (EEXIST);
1320 	}
1321 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1322 		NDFREE(&nd, NDF_ONLY_PNBUF);
1323 		vput(nd.ni_dvp);
1324 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1325 			return (error);
1326 		goto restart;
1327 	}
1328 	VATTR_NULL(&vattr);
1329 	vattr.va_type = VFIFO;
1330 	vattr.va_mode = (mode & ALLPERMS) & ~td->td_proc->p_fd->fd_cmask;
1331 #ifdef MAC
1332 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1333 	    &vattr);
1334 	if (error != 0)
1335 		goto out;
1336 #endif
1337 	error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
1338 	if (error == 0)
1339 		vput(nd.ni_vp);
1340 #ifdef MAC
1341 out:
1342 #endif
1343 	vput(nd.ni_dvp);
1344 	vn_finished_write(mp);
1345 	NDFREE(&nd, NDF_ONLY_PNBUF);
1346 	return (error);
1347 }
1348 
1349 /*
1350  * Make a hard file link.
1351  */
1352 #ifndef _SYS_SYSPROTO_H_
1353 struct link_args {
1354 	char	*path;
1355 	char	*link;
1356 };
1357 #endif
1358 int
1359 sys_link(td, uap)
1360 	struct thread *td;
1361 	register struct link_args /* {
1362 		char *path;
1363 		char *link;
1364 	} */ *uap;
1365 {
1366 
1367 	return (kern_linkat(td, AT_FDCWD, AT_FDCWD, uap->path, uap->link,
1368 	    UIO_USERSPACE, FOLLOW));
1369 }
1370 
1371 #ifndef _SYS_SYSPROTO_H_
1372 struct linkat_args {
1373 	int	fd1;
1374 	char	*path1;
1375 	int	fd2;
1376 	char	*path2;
1377 	int	flag;
1378 };
1379 #endif
1380 int
1381 sys_linkat(struct thread *td, struct linkat_args *uap)
1382 {
1383 	int flag;
1384 
1385 	flag = uap->flag;
1386 	if (flag & ~AT_SYMLINK_FOLLOW)
1387 		return (EINVAL);
1388 
1389 	return (kern_linkat(td, uap->fd1, uap->fd2, uap->path1, uap->path2,
1390 	    UIO_USERSPACE, (flag & AT_SYMLINK_FOLLOW) ? FOLLOW : NOFOLLOW));
1391 }
1392 
1393 int hardlink_check_uid = 0;
1394 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_uid, CTLFLAG_RW,
1395     &hardlink_check_uid, 0,
1396     "Unprivileged processes cannot create hard links to files owned by other "
1397     "users");
1398 static int hardlink_check_gid = 0;
1399 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_gid, CTLFLAG_RW,
1400     &hardlink_check_gid, 0,
1401     "Unprivileged processes cannot create hard links to files owned by other "
1402     "groups");
1403 
1404 static int
1405 can_hardlink(struct vnode *vp, struct ucred *cred)
1406 {
1407 	struct vattr va;
1408 	int error;
1409 
1410 	if (!hardlink_check_uid && !hardlink_check_gid)
1411 		return (0);
1412 
1413 	error = VOP_GETATTR(vp, &va, cred);
1414 	if (error != 0)
1415 		return (error);
1416 
1417 	if (hardlink_check_uid && cred->cr_uid != va.va_uid) {
1418 		error = priv_check_cred(cred, PRIV_VFS_LINK, 0);
1419 		if (error != 0)
1420 			return (error);
1421 	}
1422 
1423 	if (hardlink_check_gid && !groupmember(va.va_gid, cred)) {
1424 		error = priv_check_cred(cred, PRIV_VFS_LINK, 0);
1425 		if (error != 0)
1426 			return (error);
1427 	}
1428 
1429 	return (0);
1430 }
1431 
1432 int
1433 kern_linkat(struct thread *td, int fd1, int fd2, char *path1, char *path2,
1434     enum uio_seg segflg, int follow)
1435 {
1436 	struct vnode *vp;
1437 	struct mount *mp;
1438 	struct nameidata nd;
1439 	cap_rights_t rights;
1440 	int error;
1441 
1442 again:
1443 	bwillwrite();
1444 	NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, segflg, path1, fd1,
1445 	    cap_rights_init(&rights, CAP_LINKAT_SOURCE), td);
1446 
1447 	if ((error = namei(&nd)) != 0)
1448 		return (error);
1449 	NDFREE(&nd, NDF_ONLY_PNBUF);
1450 	vp = nd.ni_vp;
1451 	if (vp->v_type == VDIR) {
1452 		vrele(vp);
1453 		return (EPERM);		/* POSIX */
1454 	}
1455 	NDINIT_ATRIGHTS(&nd, CREATE,
1456 	    LOCKPARENT | SAVENAME | AUDITVNODE2 | NOCACHE, segflg, path2, fd2,
1457 	    cap_rights_init(&rights, CAP_LINKAT_TARGET), td);
1458 	if ((error = namei(&nd)) == 0) {
1459 		if (nd.ni_vp != NULL) {
1460 			NDFREE(&nd, NDF_ONLY_PNBUF);
1461 			if (nd.ni_dvp == nd.ni_vp)
1462 				vrele(nd.ni_dvp);
1463 			else
1464 				vput(nd.ni_dvp);
1465 			vrele(nd.ni_vp);
1466 			vrele(vp);
1467 			return (EEXIST);
1468 		} else if (nd.ni_dvp->v_mount != vp->v_mount) {
1469 			/*
1470 			 * Cross-device link.  No need to recheck
1471 			 * vp->v_type, since it cannot change, except
1472 			 * to VBAD.
1473 			 */
1474 			NDFREE(&nd, NDF_ONLY_PNBUF);
1475 			vput(nd.ni_dvp);
1476 			vrele(vp);
1477 			return (EXDEV);
1478 		} else if ((error = vn_lock(vp, LK_EXCLUSIVE)) == 0) {
1479 			error = can_hardlink(vp, td->td_ucred);
1480 #ifdef MAC
1481 			if (error == 0)
1482 				error = mac_vnode_check_link(td->td_ucred,
1483 				    nd.ni_dvp, vp, &nd.ni_cnd);
1484 #endif
1485 			if (error != 0) {
1486 				vput(vp);
1487 				vput(nd.ni_dvp);
1488 				NDFREE(&nd, NDF_ONLY_PNBUF);
1489 				return (error);
1490 			}
1491 			error = vn_start_write(vp, &mp, V_NOWAIT);
1492 			if (error != 0) {
1493 				vput(vp);
1494 				vput(nd.ni_dvp);
1495 				NDFREE(&nd, NDF_ONLY_PNBUF);
1496 				error = vn_start_write(NULL, &mp,
1497 				    V_XSLEEP | PCATCH);
1498 				if (error != 0)
1499 					return (error);
1500 				goto again;
1501 			}
1502 			error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
1503 			VOP_UNLOCK(vp, 0);
1504 			vput(nd.ni_dvp);
1505 			vn_finished_write(mp);
1506 			NDFREE(&nd, NDF_ONLY_PNBUF);
1507 		} else {
1508 			vput(nd.ni_dvp);
1509 			NDFREE(&nd, NDF_ONLY_PNBUF);
1510 			vrele(vp);
1511 			goto again;
1512 		}
1513 	}
1514 	vrele(vp);
1515 	return (error);
1516 }
1517 
1518 /*
1519  * Make a symbolic link.
1520  */
1521 #ifndef _SYS_SYSPROTO_H_
1522 struct symlink_args {
1523 	char	*path;
1524 	char	*link;
1525 };
1526 #endif
1527 int
1528 sys_symlink(td, uap)
1529 	struct thread *td;
1530 	register struct symlink_args /* {
1531 		char *path;
1532 		char *link;
1533 	} */ *uap;
1534 {
1535 
1536 	return (kern_symlinkat(td, uap->path, AT_FDCWD, uap->link,
1537 	    UIO_USERSPACE));
1538 }
1539 
1540 #ifndef _SYS_SYSPROTO_H_
1541 struct symlinkat_args {
1542 	char	*path;
1543 	int	fd;
1544 	char	*path2;
1545 };
1546 #endif
1547 int
1548 sys_symlinkat(struct thread *td, struct symlinkat_args *uap)
1549 {
1550 
1551 	return (kern_symlinkat(td, uap->path1, uap->fd, uap->path2,
1552 	    UIO_USERSPACE));
1553 }
1554 
1555 int
1556 kern_symlinkat(struct thread *td, char *path1, int fd, char *path2,
1557     enum uio_seg segflg)
1558 {
1559 	struct mount *mp;
1560 	struct vattr vattr;
1561 	char *syspath;
1562 	struct nameidata nd;
1563 	int error;
1564 	cap_rights_t rights;
1565 
1566 	if (segflg == UIO_SYSSPACE) {
1567 		syspath = path1;
1568 	} else {
1569 		syspath = uma_zalloc(namei_zone, M_WAITOK);
1570 		if ((error = copyinstr(path1, syspath, MAXPATHLEN, NULL)) != 0)
1571 			goto out;
1572 	}
1573 	AUDIT_ARG_TEXT(syspath);
1574 restart:
1575 	bwillwrite();
1576 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
1577 	    NOCACHE, segflg, path2, fd, cap_rights_init(&rights, CAP_SYMLINKAT),
1578 	    td);
1579 	if ((error = namei(&nd)) != 0)
1580 		goto out;
1581 	if (nd.ni_vp) {
1582 		NDFREE(&nd, NDF_ONLY_PNBUF);
1583 		if (nd.ni_vp == nd.ni_dvp)
1584 			vrele(nd.ni_dvp);
1585 		else
1586 			vput(nd.ni_dvp);
1587 		vrele(nd.ni_vp);
1588 		error = EEXIST;
1589 		goto out;
1590 	}
1591 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1592 		NDFREE(&nd, NDF_ONLY_PNBUF);
1593 		vput(nd.ni_dvp);
1594 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1595 			goto out;
1596 		goto restart;
1597 	}
1598 	VATTR_NULL(&vattr);
1599 	vattr.va_mode = ACCESSPERMS &~ td->td_proc->p_fd->fd_cmask;
1600 #ifdef MAC
1601 	vattr.va_type = VLNK;
1602 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1603 	    &vattr);
1604 	if (error != 0)
1605 		goto out2;
1606 #endif
1607 	error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, syspath);
1608 	if (error == 0)
1609 		vput(nd.ni_vp);
1610 #ifdef MAC
1611 out2:
1612 #endif
1613 	NDFREE(&nd, NDF_ONLY_PNBUF);
1614 	vput(nd.ni_dvp);
1615 	vn_finished_write(mp);
1616 out:
1617 	if (segflg != UIO_SYSSPACE)
1618 		uma_zfree(namei_zone, syspath);
1619 	return (error);
1620 }
1621 
1622 /*
1623  * Delete a whiteout from the filesystem.
1624  */
1625 int
1626 sys_undelete(td, uap)
1627 	struct thread *td;
1628 	register struct undelete_args /* {
1629 		char *path;
1630 	} */ *uap;
1631 {
1632 	struct mount *mp;
1633 	struct nameidata nd;
1634 	int error;
1635 
1636 restart:
1637 	bwillwrite();
1638 	NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | AUDITVNODE1,
1639 	    UIO_USERSPACE, uap->path, td);
1640 	error = namei(&nd);
1641 	if (error != 0)
1642 		return (error);
1643 
1644 	if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) {
1645 		NDFREE(&nd, NDF_ONLY_PNBUF);
1646 		if (nd.ni_vp == nd.ni_dvp)
1647 			vrele(nd.ni_dvp);
1648 		else
1649 			vput(nd.ni_dvp);
1650 		if (nd.ni_vp)
1651 			vrele(nd.ni_vp);
1652 		return (EEXIST);
1653 	}
1654 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1655 		NDFREE(&nd, NDF_ONLY_PNBUF);
1656 		vput(nd.ni_dvp);
1657 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1658 			return (error);
1659 		goto restart;
1660 	}
1661 	error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE);
1662 	NDFREE(&nd, NDF_ONLY_PNBUF);
1663 	vput(nd.ni_dvp);
1664 	vn_finished_write(mp);
1665 	return (error);
1666 }
1667 
1668 /*
1669  * Delete a name from the filesystem.
1670  */
1671 #ifndef _SYS_SYSPROTO_H_
1672 struct unlink_args {
1673 	char	*path;
1674 };
1675 #endif
1676 int
1677 sys_unlink(td, uap)
1678 	struct thread *td;
1679 	struct unlink_args /* {
1680 		char *path;
1681 	} */ *uap;
1682 {
1683 
1684 	return (kern_unlinkat(td, AT_FDCWD, uap->path, UIO_USERSPACE, 0));
1685 }
1686 
1687 #ifndef _SYS_SYSPROTO_H_
1688 struct unlinkat_args {
1689 	int	fd;
1690 	char	*path;
1691 	int	flag;
1692 };
1693 #endif
1694 int
1695 sys_unlinkat(struct thread *td, struct unlinkat_args *uap)
1696 {
1697 	int flag = uap->flag;
1698 	int fd = uap->fd;
1699 	char *path = uap->path;
1700 
1701 	if (flag & ~AT_REMOVEDIR)
1702 		return (EINVAL);
1703 
1704 	if (flag & AT_REMOVEDIR)
1705 		return (kern_rmdirat(td, fd, path, UIO_USERSPACE));
1706 	else
1707 		return (kern_unlinkat(td, fd, path, UIO_USERSPACE, 0));
1708 }
1709 
1710 int
1711 kern_unlinkat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1712     ino_t oldinum)
1713 {
1714 	struct mount *mp;
1715 	struct vnode *vp;
1716 	struct nameidata nd;
1717 	struct stat sb;
1718 	cap_rights_t rights;
1719 	int error;
1720 
1721 restart:
1722 	bwillwrite();
1723 	NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1,
1724 	    pathseg, path, fd, cap_rights_init(&rights, CAP_UNLINKAT), td);
1725 	if ((error = namei(&nd)) != 0)
1726 		return (error == EINVAL ? EPERM : error);
1727 	vp = nd.ni_vp;
1728 	if (vp->v_type == VDIR && oldinum == 0) {
1729 		error = EPERM;		/* POSIX */
1730 	} else if (oldinum != 0 &&
1731 		  ((error = vn_stat(vp, &sb, td->td_ucred, NOCRED, td)) == 0) &&
1732 		  sb.st_ino != oldinum) {
1733 			error = EIDRM;	/* Identifier removed */
1734 	} else {
1735 		/*
1736 		 * The root of a mounted filesystem cannot be deleted.
1737 		 *
1738 		 * XXX: can this only be a VDIR case?
1739 		 */
1740 		if (vp->v_vflag & VV_ROOT)
1741 			error = EBUSY;
1742 	}
1743 	if (error == 0) {
1744 		if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1745 			NDFREE(&nd, NDF_ONLY_PNBUF);
1746 			vput(nd.ni_dvp);
1747 			if (vp == nd.ni_dvp)
1748 				vrele(vp);
1749 			else
1750 				vput(vp);
1751 			if ((error = vn_start_write(NULL, &mp,
1752 			    V_XSLEEP | PCATCH)) != 0)
1753 				return (error);
1754 			goto restart;
1755 		}
1756 #ifdef MAC
1757 		error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
1758 		    &nd.ni_cnd);
1759 		if (error != 0)
1760 			goto out;
1761 #endif
1762 		vfs_notify_upper(vp, VFS_NOTIFY_UPPER_UNLINK);
1763 		error = VOP_REMOVE(nd.ni_dvp, vp, &nd.ni_cnd);
1764 #ifdef MAC
1765 out:
1766 #endif
1767 		vn_finished_write(mp);
1768 	}
1769 	NDFREE(&nd, NDF_ONLY_PNBUF);
1770 	vput(nd.ni_dvp);
1771 	if (vp == nd.ni_dvp)
1772 		vrele(vp);
1773 	else
1774 		vput(vp);
1775 	return (error);
1776 }
1777 
1778 /*
1779  * Reposition read/write file offset.
1780  */
1781 #ifndef _SYS_SYSPROTO_H_
1782 struct lseek_args {
1783 	int	fd;
1784 	int	pad;
1785 	off_t	offset;
1786 	int	whence;
1787 };
1788 #endif
1789 int
1790 sys_lseek(td, uap)
1791 	struct thread *td;
1792 	register struct lseek_args /* {
1793 		int fd;
1794 		int pad;
1795 		off_t offset;
1796 		int whence;
1797 	} */ *uap;
1798 {
1799 	struct file *fp;
1800 	cap_rights_t rights;
1801 	int error;
1802 
1803 	AUDIT_ARG_FD(uap->fd);
1804 	error = fget(td, uap->fd, cap_rights_init(&rights, CAP_SEEK), &fp);
1805 	if (error != 0)
1806 		return (error);
1807 	error = (fp->f_ops->fo_flags & DFLAG_SEEKABLE) != 0 ?
1808 	    fo_seek(fp, uap->offset, uap->whence, td) : ESPIPE;
1809 	fdrop(fp, td);
1810 	return (error);
1811 }
1812 
1813 #if defined(COMPAT_43)
1814 /*
1815  * Reposition read/write file offset.
1816  */
1817 #ifndef _SYS_SYSPROTO_H_
1818 struct olseek_args {
1819 	int	fd;
1820 	long	offset;
1821 	int	whence;
1822 };
1823 #endif
1824 int
1825 olseek(td, uap)
1826 	struct thread *td;
1827 	register struct olseek_args /* {
1828 		int fd;
1829 		long offset;
1830 		int whence;
1831 	} */ *uap;
1832 {
1833 	struct lseek_args /* {
1834 		int fd;
1835 		int pad;
1836 		off_t offset;
1837 		int whence;
1838 	} */ nuap;
1839 
1840 	nuap.fd = uap->fd;
1841 	nuap.offset = uap->offset;
1842 	nuap.whence = uap->whence;
1843 	return (sys_lseek(td, &nuap));
1844 }
1845 #endif /* COMPAT_43 */
1846 
1847 #if defined(COMPAT_FREEBSD6)
1848 /* Version with the 'pad' argument */
1849 int
1850 freebsd6_lseek(td, uap)
1851 	struct thread *td;
1852 	register struct freebsd6_lseek_args *uap;
1853 {
1854 	struct lseek_args ouap;
1855 
1856 	ouap.fd = uap->fd;
1857 	ouap.offset = uap->offset;
1858 	ouap.whence = uap->whence;
1859 	return (sys_lseek(td, &ouap));
1860 }
1861 #endif
1862 
1863 /*
1864  * Check access permissions using passed credentials.
1865  */
1866 static int
1867 vn_access(vp, user_flags, cred, td)
1868 	struct vnode	*vp;
1869 	int		user_flags;
1870 	struct ucred	*cred;
1871 	struct thread	*td;
1872 {
1873 	accmode_t accmode;
1874 	int error;
1875 
1876 	/* Flags == 0 means only check for existence. */
1877 	if (user_flags == 0)
1878 		return (0);
1879 
1880 	accmode = 0;
1881 	if (user_flags & R_OK)
1882 		accmode |= VREAD;
1883 	if (user_flags & W_OK)
1884 		accmode |= VWRITE;
1885 	if (user_flags & X_OK)
1886 		accmode |= VEXEC;
1887 #ifdef MAC
1888 	error = mac_vnode_check_access(cred, vp, accmode);
1889 	if (error != 0)
1890 		return (error);
1891 #endif
1892 	if ((accmode & VWRITE) == 0 || (error = vn_writechk(vp)) == 0)
1893 		error = VOP_ACCESS(vp, accmode, cred, td);
1894 	return (error);
1895 }
1896 
1897 /*
1898  * Check access permissions using "real" credentials.
1899  */
1900 #ifndef _SYS_SYSPROTO_H_
1901 struct access_args {
1902 	char	*path;
1903 	int	amode;
1904 };
1905 #endif
1906 int
1907 sys_access(td, uap)
1908 	struct thread *td;
1909 	register struct access_args /* {
1910 		char *path;
1911 		int amode;
1912 	} */ *uap;
1913 {
1914 
1915 	return (kern_accessat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1916 	    0, uap->amode));
1917 }
1918 
1919 #ifndef _SYS_SYSPROTO_H_
1920 struct faccessat_args {
1921 	int	dirfd;
1922 	char	*path;
1923 	int	amode;
1924 	int	flag;
1925 }
1926 #endif
1927 int
1928 sys_faccessat(struct thread *td, struct faccessat_args *uap)
1929 {
1930 
1931 	return (kern_accessat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
1932 	    uap->amode));
1933 }
1934 
1935 int
1936 kern_accessat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1937     int flag, int amode)
1938 {
1939 	struct ucred *cred, *usecred;
1940 	struct vnode *vp;
1941 	struct nameidata nd;
1942 	cap_rights_t rights;
1943 	int error;
1944 
1945 	if (flag & ~AT_EACCESS)
1946 		return (EINVAL);
1947 	if (amode != F_OK && (amode & ~(R_OK | W_OK | X_OK)) != 0)
1948 		return (EINVAL);
1949 
1950 	/*
1951 	 * Create and modify a temporary credential instead of one that
1952 	 * is potentially shared (if we need one).
1953 	 */
1954 	cred = td->td_ucred;
1955 	if ((flag & AT_EACCESS) == 0 &&
1956 	    ((cred->cr_uid != cred->cr_ruid ||
1957 	    cred->cr_rgid != cred->cr_groups[0]))) {
1958 		usecred = crdup(cred);
1959 		usecred->cr_uid = cred->cr_ruid;
1960 		usecred->cr_groups[0] = cred->cr_rgid;
1961 		td->td_ucred = usecred;
1962 	} else
1963 		usecred = cred;
1964 	AUDIT_ARG_VALUE(amode);
1965 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF |
1966 	    AUDITVNODE1, pathseg, path, fd, cap_rights_init(&rights, CAP_FSTAT),
1967 	    td);
1968 	if ((error = namei(&nd)) != 0)
1969 		goto out;
1970 	vp = nd.ni_vp;
1971 
1972 	error = vn_access(vp, amode, usecred, td);
1973 	NDFREE(&nd, NDF_ONLY_PNBUF);
1974 	vput(vp);
1975 out:
1976 	if (usecred != cred) {
1977 		td->td_ucred = cred;
1978 		crfree(usecred);
1979 	}
1980 	return (error);
1981 }
1982 
1983 /*
1984  * Check access permissions using "effective" credentials.
1985  */
1986 #ifndef _SYS_SYSPROTO_H_
1987 struct eaccess_args {
1988 	char	*path;
1989 	int	amode;
1990 };
1991 #endif
1992 int
1993 sys_eaccess(td, uap)
1994 	struct thread *td;
1995 	register struct eaccess_args /* {
1996 		char *path;
1997 		int amode;
1998 	} */ *uap;
1999 {
2000 
2001 	return (kern_accessat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2002 	    AT_EACCESS, uap->amode));
2003 }
2004 
2005 #if defined(COMPAT_43)
2006 /*
2007  * Get file status; this version follows links.
2008  */
2009 #ifndef _SYS_SYSPROTO_H_
2010 struct ostat_args {
2011 	char	*path;
2012 	struct ostat *ub;
2013 };
2014 #endif
2015 int
2016 ostat(td, uap)
2017 	struct thread *td;
2018 	register struct ostat_args /* {
2019 		char *path;
2020 		struct ostat *ub;
2021 	} */ *uap;
2022 {
2023 	struct stat sb;
2024 	struct ostat osb;
2025 	int error;
2026 
2027 	error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE,
2028 	    &sb, NULL);
2029 	if (error != 0)
2030 		return (error);
2031 	cvtstat(&sb, &osb);
2032 	return (copyout(&osb, uap->ub, sizeof (osb)));
2033 }
2034 
2035 /*
2036  * Get file status; this version does not follow links.
2037  */
2038 #ifndef _SYS_SYSPROTO_H_
2039 struct olstat_args {
2040 	char	*path;
2041 	struct ostat *ub;
2042 };
2043 #endif
2044 int
2045 olstat(td, uap)
2046 	struct thread *td;
2047 	register struct olstat_args /* {
2048 		char *path;
2049 		struct ostat *ub;
2050 	} */ *uap;
2051 {
2052 	struct stat sb;
2053 	struct ostat osb;
2054 	int error;
2055 
2056 	error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2057 	    UIO_USERSPACE, &sb, NULL);
2058 	if (error != 0)
2059 		return (error);
2060 	cvtstat(&sb, &osb);
2061 	return (copyout(&osb, uap->ub, sizeof (osb)));
2062 }
2063 
2064 /*
2065  * Convert from an old to a new stat structure.
2066  */
2067 void
2068 cvtstat(st, ost)
2069 	struct stat *st;
2070 	struct ostat *ost;
2071 {
2072 
2073 	ost->st_dev = st->st_dev;
2074 	ost->st_ino = st->st_ino;
2075 	ost->st_mode = st->st_mode;
2076 	ost->st_nlink = st->st_nlink;
2077 	ost->st_uid = st->st_uid;
2078 	ost->st_gid = st->st_gid;
2079 	ost->st_rdev = st->st_rdev;
2080 	if (st->st_size < (quad_t)1 << 32)
2081 		ost->st_size = st->st_size;
2082 	else
2083 		ost->st_size = -2;
2084 	ost->st_atim = st->st_atim;
2085 	ost->st_mtim = st->st_mtim;
2086 	ost->st_ctim = st->st_ctim;
2087 	ost->st_blksize = st->st_blksize;
2088 	ost->st_blocks = st->st_blocks;
2089 	ost->st_flags = st->st_flags;
2090 	ost->st_gen = st->st_gen;
2091 }
2092 #endif /* COMPAT_43 */
2093 
2094 /*
2095  * Get file status; this version follows links.
2096  */
2097 #ifndef _SYS_SYSPROTO_H_
2098 struct stat_args {
2099 	char	*path;
2100 	struct stat *ub;
2101 };
2102 #endif
2103 int
2104 sys_stat(td, uap)
2105 	struct thread *td;
2106 	register struct stat_args /* {
2107 		char *path;
2108 		struct stat *ub;
2109 	} */ *uap;
2110 {
2111 	struct stat sb;
2112 	int error;
2113 
2114 	error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE,
2115 	    &sb, NULL);
2116 	if (error == 0)
2117 		error = copyout(&sb, uap->ub, sizeof (sb));
2118 	return (error);
2119 }
2120 
2121 #ifndef _SYS_SYSPROTO_H_
2122 struct fstatat_args {
2123 	int	fd;
2124 	char	*path;
2125 	struct stat	*buf;
2126 	int	flag;
2127 }
2128 #endif
2129 int
2130 sys_fstatat(struct thread *td, struct fstatat_args *uap)
2131 {
2132 	struct stat sb;
2133 	int error;
2134 
2135 	error = kern_statat(td, uap->flag, uap->fd, uap->path,
2136 	    UIO_USERSPACE, &sb, NULL);
2137 	if (error == 0)
2138 		error = copyout(&sb, uap->buf, sizeof (sb));
2139 	return (error);
2140 }
2141 
2142 int
2143 kern_statat(struct thread *td, int flag, int fd, char *path,
2144     enum uio_seg pathseg, struct stat *sbp,
2145     void (*hook)(struct vnode *vp, struct stat *sbp))
2146 {
2147 	struct nameidata nd;
2148 	struct stat sb;
2149 	cap_rights_t rights;
2150 	int error;
2151 
2152 	if (flag & ~AT_SYMLINK_NOFOLLOW)
2153 		return (EINVAL);
2154 
2155 	NDINIT_ATRIGHTS(&nd, LOOKUP, ((flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW :
2156 	    FOLLOW) | LOCKSHARED | LOCKLEAF | AUDITVNODE1, pathseg, path, fd,
2157 	    cap_rights_init(&rights, CAP_FSTAT), td);
2158 
2159 	if ((error = namei(&nd)) != 0)
2160 		return (error);
2161 	error = vn_stat(nd.ni_vp, &sb, td->td_ucred, NOCRED, td);
2162 	if (error == 0) {
2163 		SDT_PROBE2(vfs, , stat, mode, path, sb.st_mode);
2164 		if (S_ISREG(sb.st_mode))
2165 			SDT_PROBE2(vfs, , stat, reg, path, pathseg);
2166 		if (__predict_false(hook != NULL))
2167 			hook(nd.ni_vp, &sb);
2168 	}
2169 	NDFREE(&nd, NDF_ONLY_PNBUF);
2170 	vput(nd.ni_vp);
2171 	if (error != 0)
2172 		return (error);
2173 	*sbp = sb;
2174 #ifdef KTRACE
2175 	if (KTRPOINT(td, KTR_STRUCT))
2176 		ktrstat(&sb);
2177 #endif
2178 	return (0);
2179 }
2180 
2181 /*
2182  * Get file status; this version does not follow links.
2183  */
2184 #ifndef _SYS_SYSPROTO_H_
2185 struct lstat_args {
2186 	char	*path;
2187 	struct stat *ub;
2188 };
2189 #endif
2190 int
2191 sys_lstat(td, uap)
2192 	struct thread *td;
2193 	register struct lstat_args /* {
2194 		char *path;
2195 		struct stat *ub;
2196 	} */ *uap;
2197 {
2198 	struct stat sb;
2199 	int error;
2200 
2201 	error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2202 	    UIO_USERSPACE, &sb, NULL);
2203 	if (error == 0)
2204 		error = copyout(&sb, uap->ub, sizeof (sb));
2205 	return (error);
2206 }
2207 
2208 /*
2209  * Implementation of the NetBSD [l]stat() functions.
2210  */
2211 void
2212 cvtnstat(sb, nsb)
2213 	struct stat *sb;
2214 	struct nstat *nsb;
2215 {
2216 
2217 	bzero(nsb, sizeof *nsb);
2218 	nsb->st_dev = sb->st_dev;
2219 	nsb->st_ino = sb->st_ino;
2220 	nsb->st_mode = sb->st_mode;
2221 	nsb->st_nlink = sb->st_nlink;
2222 	nsb->st_uid = sb->st_uid;
2223 	nsb->st_gid = sb->st_gid;
2224 	nsb->st_rdev = sb->st_rdev;
2225 	nsb->st_atim = sb->st_atim;
2226 	nsb->st_mtim = sb->st_mtim;
2227 	nsb->st_ctim = sb->st_ctim;
2228 	nsb->st_size = sb->st_size;
2229 	nsb->st_blocks = sb->st_blocks;
2230 	nsb->st_blksize = sb->st_blksize;
2231 	nsb->st_flags = sb->st_flags;
2232 	nsb->st_gen = sb->st_gen;
2233 	nsb->st_birthtim = sb->st_birthtim;
2234 }
2235 
2236 #ifndef _SYS_SYSPROTO_H_
2237 struct nstat_args {
2238 	char	*path;
2239 	struct nstat *ub;
2240 };
2241 #endif
2242 int
2243 sys_nstat(td, uap)
2244 	struct thread *td;
2245 	register struct nstat_args /* {
2246 		char *path;
2247 		struct nstat *ub;
2248 	} */ *uap;
2249 {
2250 	struct stat sb;
2251 	struct nstat nsb;
2252 	int error;
2253 
2254 	error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE,
2255 	    &sb, NULL);
2256 	if (error != 0)
2257 		return (error);
2258 	cvtnstat(&sb, &nsb);
2259 	return (copyout(&nsb, uap->ub, sizeof (nsb)));
2260 }
2261 
2262 /*
2263  * NetBSD lstat.  Get file status; this version does not follow links.
2264  */
2265 #ifndef _SYS_SYSPROTO_H_
2266 struct lstat_args {
2267 	char	*path;
2268 	struct stat *ub;
2269 };
2270 #endif
2271 int
2272 sys_nlstat(td, uap)
2273 	struct thread *td;
2274 	register struct nlstat_args /* {
2275 		char *path;
2276 		struct nstat *ub;
2277 	} */ *uap;
2278 {
2279 	struct stat sb;
2280 	struct nstat nsb;
2281 	int error;
2282 
2283 	error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2284 	    UIO_USERSPACE, &sb, NULL);
2285 	if (error != 0)
2286 		return (error);
2287 	cvtnstat(&sb, &nsb);
2288 	return (copyout(&nsb, uap->ub, sizeof (nsb)));
2289 }
2290 
2291 /*
2292  * Get configurable pathname variables.
2293  */
2294 #ifndef _SYS_SYSPROTO_H_
2295 struct pathconf_args {
2296 	char	*path;
2297 	int	name;
2298 };
2299 #endif
2300 int
2301 sys_pathconf(td, uap)
2302 	struct thread *td;
2303 	register struct pathconf_args /* {
2304 		char *path;
2305 		int name;
2306 	} */ *uap;
2307 {
2308 
2309 	return (kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name, FOLLOW));
2310 }
2311 
2312 #ifndef _SYS_SYSPROTO_H_
2313 struct lpathconf_args {
2314 	char	*path;
2315 	int	name;
2316 };
2317 #endif
2318 int
2319 sys_lpathconf(td, uap)
2320 	struct thread *td;
2321 	register struct lpathconf_args /* {
2322 		char *path;
2323 		int name;
2324 	} */ *uap;
2325 {
2326 
2327 	return (kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name,
2328 	    NOFOLLOW));
2329 }
2330 
2331 int
2332 kern_pathconf(struct thread *td, char *path, enum uio_seg pathseg, int name,
2333     u_long flags)
2334 {
2335 	struct nameidata nd;
2336 	int error;
2337 
2338 	NDINIT(&nd, LOOKUP, LOCKSHARED | LOCKLEAF | AUDITVNODE1 | flags,
2339 	    pathseg, path, td);
2340 	if ((error = namei(&nd)) != 0)
2341 		return (error);
2342 	NDFREE(&nd, NDF_ONLY_PNBUF);
2343 
2344 	/* If asynchronous I/O is available, it works for all files. */
2345 	if (name == _PC_ASYNC_IO)
2346 		td->td_retval[0] = async_io_version;
2347 	else
2348 		error = VOP_PATHCONF(nd.ni_vp, name, td->td_retval);
2349 	vput(nd.ni_vp);
2350 	return (error);
2351 }
2352 
2353 /*
2354  * Return target name of a symbolic link.
2355  */
2356 #ifndef _SYS_SYSPROTO_H_
2357 struct readlink_args {
2358 	char	*path;
2359 	char	*buf;
2360 	size_t	count;
2361 };
2362 #endif
2363 int
2364 sys_readlink(td, uap)
2365 	struct thread *td;
2366 	register struct readlink_args /* {
2367 		char *path;
2368 		char *buf;
2369 		size_t count;
2370 	} */ *uap;
2371 {
2372 
2373 	return (kern_readlinkat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2374 	    uap->buf, UIO_USERSPACE, uap->count));
2375 }
2376 #ifndef _SYS_SYSPROTO_H_
2377 struct readlinkat_args {
2378 	int	fd;
2379 	char	*path;
2380 	char	*buf;
2381 	size_t	bufsize;
2382 };
2383 #endif
2384 int
2385 sys_readlinkat(struct thread *td, struct readlinkat_args *uap)
2386 {
2387 
2388 	return (kern_readlinkat(td, uap->fd, uap->path, UIO_USERSPACE,
2389 	    uap->buf, UIO_USERSPACE, uap->bufsize));
2390 }
2391 
2392 int
2393 kern_readlinkat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2394     char *buf, enum uio_seg bufseg, size_t count)
2395 {
2396 	struct vnode *vp;
2397 	struct iovec aiov;
2398 	struct uio auio;
2399 	struct nameidata nd;
2400 	int error;
2401 
2402 	if (count > IOSIZE_MAX)
2403 		return (EINVAL);
2404 
2405 	NDINIT_AT(&nd, LOOKUP, NOFOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
2406 	    pathseg, path, fd, td);
2407 
2408 	if ((error = namei(&nd)) != 0)
2409 		return (error);
2410 	NDFREE(&nd, NDF_ONLY_PNBUF);
2411 	vp = nd.ni_vp;
2412 #ifdef MAC
2413 	error = mac_vnode_check_readlink(td->td_ucred, vp);
2414 	if (error != 0) {
2415 		vput(vp);
2416 		return (error);
2417 	}
2418 #endif
2419 	if (vp->v_type != VLNK)
2420 		error = EINVAL;
2421 	else {
2422 		aiov.iov_base = buf;
2423 		aiov.iov_len = count;
2424 		auio.uio_iov = &aiov;
2425 		auio.uio_iovcnt = 1;
2426 		auio.uio_offset = 0;
2427 		auio.uio_rw = UIO_READ;
2428 		auio.uio_segflg = bufseg;
2429 		auio.uio_td = td;
2430 		auio.uio_resid = count;
2431 		error = VOP_READLINK(vp, &auio, td->td_ucred);
2432 		td->td_retval[0] = count - auio.uio_resid;
2433 	}
2434 	vput(vp);
2435 	return (error);
2436 }
2437 
2438 /*
2439  * Common implementation code for chflags() and fchflags().
2440  */
2441 static int
2442 setfflags(td, vp, flags)
2443 	struct thread *td;
2444 	struct vnode *vp;
2445 	u_long flags;
2446 {
2447 	struct mount *mp;
2448 	struct vattr vattr;
2449 	int error;
2450 
2451 	/* We can't support the value matching VNOVAL. */
2452 	if (flags == VNOVAL)
2453 		return (EOPNOTSUPP);
2454 
2455 	/*
2456 	 * Prevent non-root users from setting flags on devices.  When
2457 	 * a device is reused, users can retain ownership of the device
2458 	 * if they are allowed to set flags and programs assume that
2459 	 * chown can't fail when done as root.
2460 	 */
2461 	if (vp->v_type == VCHR || vp->v_type == VBLK) {
2462 		error = priv_check(td, PRIV_VFS_CHFLAGS_DEV);
2463 		if (error != 0)
2464 			return (error);
2465 	}
2466 
2467 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2468 		return (error);
2469 	VATTR_NULL(&vattr);
2470 	vattr.va_flags = flags;
2471 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2472 #ifdef MAC
2473 	error = mac_vnode_check_setflags(td->td_ucred, vp, vattr.va_flags);
2474 	if (error == 0)
2475 #endif
2476 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
2477 	VOP_UNLOCK(vp, 0);
2478 	vn_finished_write(mp);
2479 	return (error);
2480 }
2481 
2482 /*
2483  * Change flags of a file given a path name.
2484  */
2485 #ifndef _SYS_SYSPROTO_H_
2486 struct chflags_args {
2487 	const char *path;
2488 	u_long	flags;
2489 };
2490 #endif
2491 int
2492 sys_chflags(td, uap)
2493 	struct thread *td;
2494 	register struct chflags_args /* {
2495 		const char *path;
2496 		u_long flags;
2497 	} */ *uap;
2498 {
2499 
2500 	return (kern_chflagsat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2501 	    uap->flags, 0));
2502 }
2503 
2504 #ifndef _SYS_SYSPROTO_H_
2505 struct chflagsat_args {
2506 	int	fd;
2507 	const char *path;
2508 	u_long	flags;
2509 	int	atflag;
2510 }
2511 #endif
2512 int
2513 sys_chflagsat(struct thread *td, struct chflagsat_args *uap)
2514 {
2515 	int fd = uap->fd;
2516 	const char *path = uap->path;
2517 	u_long flags = uap->flags;
2518 	int atflag = uap->atflag;
2519 
2520 	if (atflag & ~AT_SYMLINK_NOFOLLOW)
2521 		return (EINVAL);
2522 
2523 	return (kern_chflagsat(td, fd, path, UIO_USERSPACE, flags, atflag));
2524 }
2525 
2526 /*
2527  * Same as chflags() but doesn't follow symlinks.
2528  */
2529 int
2530 sys_lchflags(td, uap)
2531 	struct thread *td;
2532 	register struct lchflags_args /* {
2533 		const char *path;
2534 		u_long flags;
2535 	} */ *uap;
2536 {
2537 
2538 	return (kern_chflagsat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2539 	    uap->flags, AT_SYMLINK_NOFOLLOW));
2540 }
2541 
2542 static int
2543 kern_chflagsat(struct thread *td, int fd, const char *path,
2544     enum uio_seg pathseg, u_long flags, int atflag)
2545 {
2546 	struct nameidata nd;
2547 	cap_rights_t rights;
2548 	int error, follow;
2549 
2550 	AUDIT_ARG_FFLAGS(flags);
2551 	follow = (atflag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
2552 	NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, pathseg, path, fd,
2553 	    cap_rights_init(&rights, CAP_FCHFLAGS), td);
2554 	if ((error = namei(&nd)) != 0)
2555 		return (error);
2556 	NDFREE(&nd, NDF_ONLY_PNBUF);
2557 	error = setfflags(td, nd.ni_vp, flags);
2558 	vrele(nd.ni_vp);
2559 	return (error);
2560 }
2561 
2562 /*
2563  * Change flags of a file given a file descriptor.
2564  */
2565 #ifndef _SYS_SYSPROTO_H_
2566 struct fchflags_args {
2567 	int	fd;
2568 	u_long	flags;
2569 };
2570 #endif
2571 int
2572 sys_fchflags(td, uap)
2573 	struct thread *td;
2574 	register struct fchflags_args /* {
2575 		int fd;
2576 		u_long flags;
2577 	} */ *uap;
2578 {
2579 	struct file *fp;
2580 	cap_rights_t rights;
2581 	int error;
2582 
2583 	AUDIT_ARG_FD(uap->fd);
2584 	AUDIT_ARG_FFLAGS(uap->flags);
2585 	error = getvnode(td, uap->fd, cap_rights_init(&rights, CAP_FCHFLAGS),
2586 	    &fp);
2587 	if (error != 0)
2588 		return (error);
2589 #ifdef AUDIT
2590 	vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
2591 	AUDIT_ARG_VNODE1(fp->f_vnode);
2592 	VOP_UNLOCK(fp->f_vnode, 0);
2593 #endif
2594 	error = setfflags(td, fp->f_vnode, uap->flags);
2595 	fdrop(fp, td);
2596 	return (error);
2597 }
2598 
2599 /*
2600  * Common implementation code for chmod(), lchmod() and fchmod().
2601  */
2602 int
2603 setfmode(td, cred, vp, mode)
2604 	struct thread *td;
2605 	struct ucred *cred;
2606 	struct vnode *vp;
2607 	int mode;
2608 {
2609 	struct mount *mp;
2610 	struct vattr vattr;
2611 	int error;
2612 
2613 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2614 		return (error);
2615 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2616 	VATTR_NULL(&vattr);
2617 	vattr.va_mode = mode & ALLPERMS;
2618 #ifdef MAC
2619 	error = mac_vnode_check_setmode(cred, vp, vattr.va_mode);
2620 	if (error == 0)
2621 #endif
2622 		error = VOP_SETATTR(vp, &vattr, cred);
2623 	VOP_UNLOCK(vp, 0);
2624 	vn_finished_write(mp);
2625 	return (error);
2626 }
2627 
2628 /*
2629  * Change mode of a file given path name.
2630  */
2631 #ifndef _SYS_SYSPROTO_H_
2632 struct chmod_args {
2633 	char	*path;
2634 	int	mode;
2635 };
2636 #endif
2637 int
2638 sys_chmod(td, uap)
2639 	struct thread *td;
2640 	register struct chmod_args /* {
2641 		char *path;
2642 		int mode;
2643 	} */ *uap;
2644 {
2645 
2646 	return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2647 	    uap->mode, 0));
2648 }
2649 
2650 #ifndef _SYS_SYSPROTO_H_
2651 struct fchmodat_args {
2652 	int	dirfd;
2653 	char	*path;
2654 	mode_t	mode;
2655 	int	flag;
2656 }
2657 #endif
2658 int
2659 sys_fchmodat(struct thread *td, struct fchmodat_args *uap)
2660 {
2661 	int flag = uap->flag;
2662 	int fd = uap->fd;
2663 	char *path = uap->path;
2664 	mode_t mode = uap->mode;
2665 
2666 	if (flag & ~AT_SYMLINK_NOFOLLOW)
2667 		return (EINVAL);
2668 
2669 	return (kern_fchmodat(td, fd, path, UIO_USERSPACE, mode, flag));
2670 }
2671 
2672 /*
2673  * Change mode of a file given path name (don't follow links.)
2674  */
2675 #ifndef _SYS_SYSPROTO_H_
2676 struct lchmod_args {
2677 	char	*path;
2678 	int	mode;
2679 };
2680 #endif
2681 int
2682 sys_lchmod(td, uap)
2683 	struct thread *td;
2684 	register struct lchmod_args /* {
2685 		char *path;
2686 		int mode;
2687 	} */ *uap;
2688 {
2689 
2690 	return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2691 	    uap->mode, AT_SYMLINK_NOFOLLOW));
2692 }
2693 
2694 int
2695 kern_fchmodat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2696     mode_t mode, int flag)
2697 {
2698 	struct nameidata nd;
2699 	cap_rights_t rights;
2700 	int error, follow;
2701 
2702 	AUDIT_ARG_MODE(mode);
2703 	follow = (flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
2704 	NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, pathseg, path, fd,
2705 	    cap_rights_init(&rights, CAP_FCHMOD), td);
2706 	if ((error = namei(&nd)) != 0)
2707 		return (error);
2708 	NDFREE(&nd, NDF_ONLY_PNBUF);
2709 	error = setfmode(td, td->td_ucred, nd.ni_vp, mode);
2710 	vrele(nd.ni_vp);
2711 	return (error);
2712 }
2713 
2714 /*
2715  * Change mode of a file given a file descriptor.
2716  */
2717 #ifndef _SYS_SYSPROTO_H_
2718 struct fchmod_args {
2719 	int	fd;
2720 	int	mode;
2721 };
2722 #endif
2723 int
2724 sys_fchmod(struct thread *td, struct fchmod_args *uap)
2725 {
2726 	struct file *fp;
2727 	cap_rights_t rights;
2728 	int error;
2729 
2730 	AUDIT_ARG_FD(uap->fd);
2731 	AUDIT_ARG_MODE(uap->mode);
2732 
2733 	error = fget(td, uap->fd, cap_rights_init(&rights, CAP_FCHMOD), &fp);
2734 	if (error != 0)
2735 		return (error);
2736 	error = fo_chmod(fp, uap->mode, td->td_ucred, td);
2737 	fdrop(fp, td);
2738 	return (error);
2739 }
2740 
2741 /*
2742  * Common implementation for chown(), lchown(), and fchown()
2743  */
2744 int
2745 setfown(td, cred, vp, uid, gid)
2746 	struct thread *td;
2747 	struct ucred *cred;
2748 	struct vnode *vp;
2749 	uid_t uid;
2750 	gid_t gid;
2751 {
2752 	struct mount *mp;
2753 	struct vattr vattr;
2754 	int error;
2755 
2756 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2757 		return (error);
2758 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2759 	VATTR_NULL(&vattr);
2760 	vattr.va_uid = uid;
2761 	vattr.va_gid = gid;
2762 #ifdef MAC
2763 	error = mac_vnode_check_setowner(cred, vp, vattr.va_uid,
2764 	    vattr.va_gid);
2765 	if (error == 0)
2766 #endif
2767 		error = VOP_SETATTR(vp, &vattr, cred);
2768 	VOP_UNLOCK(vp, 0);
2769 	vn_finished_write(mp);
2770 	return (error);
2771 }
2772 
2773 /*
2774  * Set ownership given a path name.
2775  */
2776 #ifndef _SYS_SYSPROTO_H_
2777 struct chown_args {
2778 	char	*path;
2779 	int	uid;
2780 	int	gid;
2781 };
2782 #endif
2783 int
2784 sys_chown(td, uap)
2785 	struct thread *td;
2786 	register struct chown_args /* {
2787 		char *path;
2788 		int uid;
2789 		int gid;
2790 	} */ *uap;
2791 {
2792 
2793 	return (kern_fchownat(td, AT_FDCWD, uap->path, UIO_USERSPACE, uap->uid,
2794 	    uap->gid, 0));
2795 }
2796 
2797 #ifndef _SYS_SYSPROTO_H_
2798 struct fchownat_args {
2799 	int fd;
2800 	const char * path;
2801 	uid_t uid;
2802 	gid_t gid;
2803 	int flag;
2804 };
2805 #endif
2806 int
2807 sys_fchownat(struct thread *td, struct fchownat_args *uap)
2808 {
2809 	int flag;
2810 
2811 	flag = uap->flag;
2812 	if (flag & ~AT_SYMLINK_NOFOLLOW)
2813 		return (EINVAL);
2814 
2815 	return (kern_fchownat(td, uap->fd, uap->path, UIO_USERSPACE, uap->uid,
2816 	    uap->gid, uap->flag));
2817 }
2818 
2819 int
2820 kern_fchownat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2821     int uid, int gid, int flag)
2822 {
2823 	struct nameidata nd;
2824 	cap_rights_t rights;
2825 	int error, follow;
2826 
2827 	AUDIT_ARG_OWNER(uid, gid);
2828 	follow = (flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
2829 	NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, pathseg, path, fd,
2830 	    cap_rights_init(&rights, CAP_FCHOWN), td);
2831 
2832 	if ((error = namei(&nd)) != 0)
2833 		return (error);
2834 	NDFREE(&nd, NDF_ONLY_PNBUF);
2835 	error = setfown(td, td->td_ucred, nd.ni_vp, uid, gid);
2836 	vrele(nd.ni_vp);
2837 	return (error);
2838 }
2839 
2840 /*
2841  * Set ownership given a path name, do not cross symlinks.
2842  */
2843 #ifndef _SYS_SYSPROTO_H_
2844 struct lchown_args {
2845 	char	*path;
2846 	int	uid;
2847 	int	gid;
2848 };
2849 #endif
2850 int
2851 sys_lchown(td, uap)
2852 	struct thread *td;
2853 	register struct lchown_args /* {
2854 		char *path;
2855 		int uid;
2856 		int gid;
2857 	} */ *uap;
2858 {
2859 
2860 	return (kern_fchownat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2861 	    uap->uid, uap->gid, AT_SYMLINK_NOFOLLOW));
2862 }
2863 
2864 /*
2865  * Set ownership given a file descriptor.
2866  */
2867 #ifndef _SYS_SYSPROTO_H_
2868 struct fchown_args {
2869 	int	fd;
2870 	int	uid;
2871 	int	gid;
2872 };
2873 #endif
2874 int
2875 sys_fchown(td, uap)
2876 	struct thread *td;
2877 	register struct fchown_args /* {
2878 		int fd;
2879 		int uid;
2880 		int gid;
2881 	} */ *uap;
2882 {
2883 	struct file *fp;
2884 	cap_rights_t rights;
2885 	int error;
2886 
2887 	AUDIT_ARG_FD(uap->fd);
2888 	AUDIT_ARG_OWNER(uap->uid, uap->gid);
2889 	error = fget(td, uap->fd, cap_rights_init(&rights, CAP_FCHOWN), &fp);
2890 	if (error != 0)
2891 		return (error);
2892 	error = fo_chown(fp, uap->uid, uap->gid, td->td_ucred, td);
2893 	fdrop(fp, td);
2894 	return (error);
2895 }
2896 
2897 /*
2898  * Common implementation code for utimes(), lutimes(), and futimes().
2899  */
2900 static int
2901 getutimes(usrtvp, tvpseg, tsp)
2902 	const struct timeval *usrtvp;
2903 	enum uio_seg tvpseg;
2904 	struct timespec *tsp;
2905 {
2906 	struct timeval tv[2];
2907 	const struct timeval *tvp;
2908 	int error;
2909 
2910 	if (usrtvp == NULL) {
2911 		vfs_timestamp(&tsp[0]);
2912 		tsp[1] = tsp[0];
2913 	} else {
2914 		if (tvpseg == UIO_SYSSPACE) {
2915 			tvp = usrtvp;
2916 		} else {
2917 			if ((error = copyin(usrtvp, tv, sizeof(tv))) != 0)
2918 				return (error);
2919 			tvp = tv;
2920 		}
2921 
2922 		if (tvp[0].tv_usec < 0 || tvp[0].tv_usec >= 1000000 ||
2923 		    tvp[1].tv_usec < 0 || tvp[1].tv_usec >= 1000000)
2924 			return (EINVAL);
2925 		TIMEVAL_TO_TIMESPEC(&tvp[0], &tsp[0]);
2926 		TIMEVAL_TO_TIMESPEC(&tvp[1], &tsp[1]);
2927 	}
2928 	return (0);
2929 }
2930 
2931 /*
2932  * Common implementation code for futimens(), utimensat().
2933  */
2934 #define	UTIMENS_NULL	0x1
2935 #define	UTIMENS_EXIT	0x2
2936 static int
2937 getutimens(const struct timespec *usrtsp, enum uio_seg tspseg,
2938     struct timespec *tsp, int *retflags)
2939 {
2940 	struct timespec tsnow;
2941 	int error;
2942 
2943 	vfs_timestamp(&tsnow);
2944 	*retflags = 0;
2945 	if (usrtsp == NULL) {
2946 		tsp[0] = tsnow;
2947 		tsp[1] = tsnow;
2948 		*retflags |= UTIMENS_NULL;
2949 		return (0);
2950 	}
2951 	if (tspseg == UIO_SYSSPACE) {
2952 		tsp[0] = usrtsp[0];
2953 		tsp[1] = usrtsp[1];
2954 	} else if ((error = copyin(usrtsp, tsp, sizeof(*tsp) * 2)) != 0)
2955 		return (error);
2956 	if (tsp[0].tv_nsec == UTIME_OMIT && tsp[1].tv_nsec == UTIME_OMIT)
2957 		*retflags |= UTIMENS_EXIT;
2958 	if (tsp[0].tv_nsec == UTIME_NOW && tsp[1].tv_nsec == UTIME_NOW)
2959 		*retflags |= UTIMENS_NULL;
2960 	if (tsp[0].tv_nsec == UTIME_OMIT)
2961 		tsp[0].tv_sec = VNOVAL;
2962 	else if (tsp[0].tv_nsec == UTIME_NOW)
2963 		tsp[0] = tsnow;
2964 	else if (tsp[0].tv_nsec < 0 || tsp[0].tv_nsec >= 1000000000L)
2965 		return (EINVAL);
2966 	if (tsp[1].tv_nsec == UTIME_OMIT)
2967 		tsp[1].tv_sec = VNOVAL;
2968 	else if (tsp[1].tv_nsec == UTIME_NOW)
2969 		tsp[1] = tsnow;
2970 	else if (tsp[1].tv_nsec < 0 || tsp[1].tv_nsec >= 1000000000L)
2971 		return (EINVAL);
2972 
2973 	return (0);
2974 }
2975 
2976 /*
2977  * Common implementation code for utimes(), lutimes(), futimes(), futimens(),
2978  * and utimensat().
2979  */
2980 static int
2981 setutimes(td, vp, ts, numtimes, nullflag)
2982 	struct thread *td;
2983 	struct vnode *vp;
2984 	const struct timespec *ts;
2985 	int numtimes;
2986 	int nullflag;
2987 {
2988 	struct mount *mp;
2989 	struct vattr vattr;
2990 	int error, setbirthtime;
2991 
2992 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2993 		return (error);
2994 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2995 	setbirthtime = 0;
2996 	if (numtimes < 3 && !VOP_GETATTR(vp, &vattr, td->td_ucred) &&
2997 	    timespeccmp(&ts[1], &vattr.va_birthtime, < ))
2998 		setbirthtime = 1;
2999 	VATTR_NULL(&vattr);
3000 	vattr.va_atime = ts[0];
3001 	vattr.va_mtime = ts[1];
3002 	if (setbirthtime)
3003 		vattr.va_birthtime = ts[1];
3004 	if (numtimes > 2)
3005 		vattr.va_birthtime = ts[2];
3006 	if (nullflag)
3007 		vattr.va_vaflags |= VA_UTIMES_NULL;
3008 #ifdef MAC
3009 	error = mac_vnode_check_setutimes(td->td_ucred, vp, vattr.va_atime,
3010 	    vattr.va_mtime);
3011 #endif
3012 	if (error == 0)
3013 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3014 	VOP_UNLOCK(vp, 0);
3015 	vn_finished_write(mp);
3016 	return (error);
3017 }
3018 
3019 /*
3020  * Set the access and modification times of a file.
3021  */
3022 #ifndef _SYS_SYSPROTO_H_
3023 struct utimes_args {
3024 	char	*path;
3025 	struct	timeval *tptr;
3026 };
3027 #endif
3028 int
3029 sys_utimes(td, uap)
3030 	struct thread *td;
3031 	register struct utimes_args /* {
3032 		char *path;
3033 		struct timeval *tptr;
3034 	} */ *uap;
3035 {
3036 
3037 	return (kern_utimesat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
3038 	    uap->tptr, UIO_USERSPACE));
3039 }
3040 
3041 #ifndef _SYS_SYSPROTO_H_
3042 struct futimesat_args {
3043 	int fd;
3044 	const char * path;
3045 	const struct timeval * times;
3046 };
3047 #endif
3048 int
3049 sys_futimesat(struct thread *td, struct futimesat_args *uap)
3050 {
3051 
3052 	return (kern_utimesat(td, uap->fd, uap->path, UIO_USERSPACE,
3053 	    uap->times, UIO_USERSPACE));
3054 }
3055 
3056 int
3057 kern_utimesat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
3058     struct timeval *tptr, enum uio_seg tptrseg)
3059 {
3060 	struct nameidata nd;
3061 	struct timespec ts[2];
3062 	cap_rights_t rights;
3063 	int error;
3064 
3065 	if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3066 		return (error);
3067 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, fd,
3068 	    cap_rights_init(&rights, CAP_FUTIMES), td);
3069 
3070 	if ((error = namei(&nd)) != 0)
3071 		return (error);
3072 	NDFREE(&nd, NDF_ONLY_PNBUF);
3073 	error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3074 	vrele(nd.ni_vp);
3075 	return (error);
3076 }
3077 
3078 /*
3079  * Set the access and modification times of a file.
3080  */
3081 #ifndef _SYS_SYSPROTO_H_
3082 struct lutimes_args {
3083 	char	*path;
3084 	struct	timeval *tptr;
3085 };
3086 #endif
3087 int
3088 sys_lutimes(td, uap)
3089 	struct thread *td;
3090 	register struct lutimes_args /* {
3091 		char *path;
3092 		struct timeval *tptr;
3093 	} */ *uap;
3094 {
3095 
3096 	return (kern_lutimes(td, uap->path, UIO_USERSPACE, uap->tptr,
3097 	    UIO_USERSPACE));
3098 }
3099 
3100 int
3101 kern_lutimes(struct thread *td, char *path, enum uio_seg pathseg,
3102     struct timeval *tptr, enum uio_seg tptrseg)
3103 {
3104 	struct timespec ts[2];
3105 	struct nameidata nd;
3106 	int error;
3107 
3108 	if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3109 		return (error);
3110 	NDINIT(&nd, LOOKUP, NOFOLLOW | AUDITVNODE1, pathseg, path, td);
3111 	if ((error = namei(&nd)) != 0)
3112 		return (error);
3113 	NDFREE(&nd, NDF_ONLY_PNBUF);
3114 	error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3115 	vrele(nd.ni_vp);
3116 	return (error);
3117 }
3118 
3119 /*
3120  * Set the access and modification times of a file.
3121  */
3122 #ifndef _SYS_SYSPROTO_H_
3123 struct futimes_args {
3124 	int	fd;
3125 	struct	timeval *tptr;
3126 };
3127 #endif
3128 int
3129 sys_futimes(td, uap)
3130 	struct thread *td;
3131 	register struct futimes_args /* {
3132 		int  fd;
3133 		struct timeval *tptr;
3134 	} */ *uap;
3135 {
3136 
3137 	return (kern_futimes(td, uap->fd, uap->tptr, UIO_USERSPACE));
3138 }
3139 
3140 int
3141 kern_futimes(struct thread *td, int fd, struct timeval *tptr,
3142     enum uio_seg tptrseg)
3143 {
3144 	struct timespec ts[2];
3145 	struct file *fp;
3146 	cap_rights_t rights;
3147 	int error;
3148 
3149 	AUDIT_ARG_FD(fd);
3150 	error = getutimes(tptr, tptrseg, ts);
3151 	if (error != 0)
3152 		return (error);
3153 	error = getvnode(td, fd, cap_rights_init(&rights, CAP_FUTIMES), &fp);
3154 	if (error != 0)
3155 		return (error);
3156 #ifdef AUDIT
3157 	vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
3158 	AUDIT_ARG_VNODE1(fp->f_vnode);
3159 	VOP_UNLOCK(fp->f_vnode, 0);
3160 #endif
3161 	error = setutimes(td, fp->f_vnode, ts, 2, tptr == NULL);
3162 	fdrop(fp, td);
3163 	return (error);
3164 }
3165 
3166 int
3167 sys_futimens(struct thread *td, struct futimens_args *uap)
3168 {
3169 
3170 	return (kern_futimens(td, uap->fd, uap->times, UIO_USERSPACE));
3171 }
3172 
3173 int
3174 kern_futimens(struct thread *td, int fd, struct timespec *tptr,
3175     enum uio_seg tptrseg)
3176 {
3177 	struct timespec ts[2];
3178 	struct file *fp;
3179 	cap_rights_t rights;
3180 	int error, flags;
3181 
3182 	AUDIT_ARG_FD(fd);
3183 	error = getutimens(tptr, tptrseg, ts, &flags);
3184 	if (error != 0)
3185 		return (error);
3186 	if (flags & UTIMENS_EXIT)
3187 		return (0);
3188 	error = getvnode(td, fd, cap_rights_init(&rights, CAP_FUTIMES), &fp);
3189 	if (error != 0)
3190 		return (error);
3191 #ifdef AUDIT
3192 	vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
3193 	AUDIT_ARG_VNODE1(fp->f_vnode);
3194 	VOP_UNLOCK(fp->f_vnode, 0);
3195 #endif
3196 	error = setutimes(td, fp->f_vnode, ts, 2, flags & UTIMENS_NULL);
3197 	fdrop(fp, td);
3198 	return (error);
3199 }
3200 
3201 int
3202 sys_utimensat(struct thread *td, struct utimensat_args *uap)
3203 {
3204 
3205 	return (kern_utimensat(td, uap->fd, uap->path, UIO_USERSPACE,
3206 	    uap->times, UIO_USERSPACE, uap->flag));
3207 }
3208 
3209 int
3210 kern_utimensat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
3211     struct timespec *tptr, enum uio_seg tptrseg, int flag)
3212 {
3213 	struct nameidata nd;
3214 	struct timespec ts[2];
3215 	cap_rights_t rights;
3216 	int error, flags;
3217 
3218 	if (flag & ~AT_SYMLINK_NOFOLLOW)
3219 		return (EINVAL);
3220 
3221 	if ((error = getutimens(tptr, tptrseg, ts, &flags)) != 0)
3222 		return (error);
3223 	NDINIT_ATRIGHTS(&nd, LOOKUP, ((flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW :
3224 	    FOLLOW) | AUDITVNODE1, pathseg, path, fd,
3225 	    cap_rights_init(&rights, CAP_FUTIMES), td);
3226 	if ((error = namei(&nd)) != 0)
3227 		return (error);
3228 	/*
3229 	 * We are allowed to call namei() regardless of 2xUTIME_OMIT.
3230 	 * POSIX states:
3231 	 * "If both tv_nsec fields are UTIME_OMIT... EACCESS may be detected."
3232 	 * "Search permission is denied by a component of the path prefix."
3233 	 */
3234 	NDFREE(&nd, NDF_ONLY_PNBUF);
3235 	if ((flags & UTIMENS_EXIT) == 0)
3236 		error = setutimes(td, nd.ni_vp, ts, 2, flags & UTIMENS_NULL);
3237 	vrele(nd.ni_vp);
3238 	return (error);
3239 }
3240 
3241 /*
3242  * Truncate a file given its path name.
3243  */
3244 #ifndef _SYS_SYSPROTO_H_
3245 struct truncate_args {
3246 	char	*path;
3247 	int	pad;
3248 	off_t	length;
3249 };
3250 #endif
3251 int
3252 sys_truncate(td, uap)
3253 	struct thread *td;
3254 	register struct truncate_args /* {
3255 		char *path;
3256 		int pad;
3257 		off_t length;
3258 	} */ *uap;
3259 {
3260 
3261 	return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3262 }
3263 
3264 int
3265 kern_truncate(struct thread *td, char *path, enum uio_seg pathseg, off_t length)
3266 {
3267 	struct mount *mp;
3268 	struct vnode *vp;
3269 	void *rl_cookie;
3270 	struct vattr vattr;
3271 	struct nameidata nd;
3272 	int error;
3273 
3274 	if (length < 0)
3275 		return(EINVAL);
3276 	NDINIT(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, td);
3277 	if ((error = namei(&nd)) != 0)
3278 		return (error);
3279 	vp = nd.ni_vp;
3280 	rl_cookie = vn_rangelock_wlock(vp, 0, OFF_MAX);
3281 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) {
3282 		vn_rangelock_unlock(vp, rl_cookie);
3283 		vrele(vp);
3284 		return (error);
3285 	}
3286 	NDFREE(&nd, NDF_ONLY_PNBUF);
3287 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3288 	if (vp->v_type == VDIR)
3289 		error = EISDIR;
3290 #ifdef MAC
3291 	else if ((error = mac_vnode_check_write(td->td_ucred, NOCRED, vp))) {
3292 	}
3293 #endif
3294 	else if ((error = vn_writechk(vp)) == 0 &&
3295 	    (error = VOP_ACCESS(vp, VWRITE, td->td_ucred, td)) == 0) {
3296 		VATTR_NULL(&vattr);
3297 		vattr.va_size = length;
3298 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3299 	}
3300 	VOP_UNLOCK(vp, 0);
3301 	vn_finished_write(mp);
3302 	vn_rangelock_unlock(vp, rl_cookie);
3303 	vrele(vp);
3304 	return (error);
3305 }
3306 
3307 #if defined(COMPAT_43)
3308 /*
3309  * Truncate a file given its path name.
3310  */
3311 #ifndef _SYS_SYSPROTO_H_
3312 struct otruncate_args {
3313 	char	*path;
3314 	long	length;
3315 };
3316 #endif
3317 int
3318 otruncate(td, uap)
3319 	struct thread *td;
3320 	register struct otruncate_args /* {
3321 		char *path;
3322 		long length;
3323 	} */ *uap;
3324 {
3325 	struct truncate_args /* {
3326 		char *path;
3327 		int pad;
3328 		off_t length;
3329 	} */ nuap;
3330 
3331 	nuap.path = uap->path;
3332 	nuap.length = uap->length;
3333 	return (sys_truncate(td, &nuap));
3334 }
3335 #endif /* COMPAT_43 */
3336 
3337 #if defined(COMPAT_FREEBSD6)
3338 /* Versions with the pad argument */
3339 int
3340 freebsd6_truncate(struct thread *td, struct freebsd6_truncate_args *uap)
3341 {
3342 	struct truncate_args ouap;
3343 
3344 	ouap.path = uap->path;
3345 	ouap.length = uap->length;
3346 	return (sys_truncate(td, &ouap));
3347 }
3348 
3349 int
3350 freebsd6_ftruncate(struct thread *td, struct freebsd6_ftruncate_args *uap)
3351 {
3352 	struct ftruncate_args ouap;
3353 
3354 	ouap.fd = uap->fd;
3355 	ouap.length = uap->length;
3356 	return (sys_ftruncate(td, &ouap));
3357 }
3358 #endif
3359 
3360 /*
3361  * Sync an open file.
3362  */
3363 #ifndef _SYS_SYSPROTO_H_
3364 struct fsync_args {
3365 	int	fd;
3366 };
3367 #endif
3368 int
3369 sys_fsync(td, uap)
3370 	struct thread *td;
3371 	struct fsync_args /* {
3372 		int fd;
3373 	} */ *uap;
3374 {
3375 	struct vnode *vp;
3376 	struct mount *mp;
3377 	struct file *fp;
3378 	cap_rights_t rights;
3379 	int error, lock_flags;
3380 
3381 	AUDIT_ARG_FD(uap->fd);
3382 	error = getvnode(td, uap->fd, cap_rights_init(&rights, CAP_FSYNC), &fp);
3383 	if (error != 0)
3384 		return (error);
3385 	vp = fp->f_vnode;
3386 	error = vn_start_write(vp, &mp, V_WAIT | PCATCH);
3387 	if (error != 0)
3388 		goto drop;
3389 	if (MNT_SHARED_WRITES(mp) ||
3390 	    ((mp == NULL) && MNT_SHARED_WRITES(vp->v_mount))) {
3391 		lock_flags = LK_SHARED;
3392 	} else {
3393 		lock_flags = LK_EXCLUSIVE;
3394 	}
3395 	vn_lock(vp, lock_flags | LK_RETRY);
3396 	AUDIT_ARG_VNODE1(vp);
3397 	if (vp->v_object != NULL) {
3398 		VM_OBJECT_WLOCK(vp->v_object);
3399 		vm_object_page_clean(vp->v_object, 0, 0, 0);
3400 		VM_OBJECT_WUNLOCK(vp->v_object);
3401 	}
3402 	error = VOP_FSYNC(vp, MNT_WAIT, td);
3403 
3404 	VOP_UNLOCK(vp, 0);
3405 	vn_finished_write(mp);
3406 drop:
3407 	fdrop(fp, td);
3408 	return (error);
3409 }
3410 
3411 /*
3412  * Rename files.  Source and destination must either both be directories, or
3413  * both not be directories.  If target is a directory, it must be empty.
3414  */
3415 #ifndef _SYS_SYSPROTO_H_
3416 struct rename_args {
3417 	char	*from;
3418 	char	*to;
3419 };
3420 #endif
3421 int
3422 sys_rename(td, uap)
3423 	struct thread *td;
3424 	register struct rename_args /* {
3425 		char *from;
3426 		char *to;
3427 	} */ *uap;
3428 {
3429 
3430 	return (kern_renameat(td, AT_FDCWD, uap->from, AT_FDCWD,
3431 	    uap->to, UIO_USERSPACE));
3432 }
3433 
3434 #ifndef _SYS_SYSPROTO_H_
3435 struct renameat_args {
3436 	int	oldfd;
3437 	char	*old;
3438 	int	newfd;
3439 	char	*new;
3440 };
3441 #endif
3442 int
3443 sys_renameat(struct thread *td, struct renameat_args *uap)
3444 {
3445 
3446 	return (kern_renameat(td, uap->oldfd, uap->old, uap->newfd, uap->new,
3447 	    UIO_USERSPACE));
3448 }
3449 
3450 int
3451 kern_renameat(struct thread *td, int oldfd, char *old, int newfd, char *new,
3452     enum uio_seg pathseg)
3453 {
3454 	struct mount *mp = NULL;
3455 	struct vnode *tvp, *fvp, *tdvp;
3456 	struct nameidata fromnd, tond;
3457 	cap_rights_t rights;
3458 	int error;
3459 
3460 again:
3461 	bwillwrite();
3462 #ifdef MAC
3463 	NDINIT_ATRIGHTS(&fromnd, DELETE, LOCKPARENT | LOCKLEAF | SAVESTART |
3464 	    AUDITVNODE1, pathseg, old, oldfd,
3465 	    cap_rights_init(&rights, CAP_RENAMEAT_SOURCE), td);
3466 #else
3467 	NDINIT_ATRIGHTS(&fromnd, DELETE, WANTPARENT | SAVESTART | AUDITVNODE1,
3468 	    pathseg, old, oldfd,
3469 	    cap_rights_init(&rights, CAP_RENAMEAT_SOURCE), td);
3470 #endif
3471 
3472 	if ((error = namei(&fromnd)) != 0)
3473 		return (error);
3474 #ifdef MAC
3475 	error = mac_vnode_check_rename_from(td->td_ucred, fromnd.ni_dvp,
3476 	    fromnd.ni_vp, &fromnd.ni_cnd);
3477 	VOP_UNLOCK(fromnd.ni_dvp, 0);
3478 	if (fromnd.ni_dvp != fromnd.ni_vp)
3479 		VOP_UNLOCK(fromnd.ni_vp, 0);
3480 #endif
3481 	fvp = fromnd.ni_vp;
3482 	NDINIT_ATRIGHTS(&tond, RENAME, LOCKPARENT | LOCKLEAF | NOCACHE |
3483 	    SAVESTART | AUDITVNODE2, pathseg, new, newfd,
3484 	    cap_rights_init(&rights, CAP_RENAMEAT_TARGET), td);
3485 	if (fromnd.ni_vp->v_type == VDIR)
3486 		tond.ni_cnd.cn_flags |= WILLBEDIR;
3487 	if ((error = namei(&tond)) != 0) {
3488 		/* Translate error code for rename("dir1", "dir2/."). */
3489 		if (error == EISDIR && fvp->v_type == VDIR)
3490 			error = EINVAL;
3491 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3492 		vrele(fromnd.ni_dvp);
3493 		vrele(fvp);
3494 		goto out1;
3495 	}
3496 	tdvp = tond.ni_dvp;
3497 	tvp = tond.ni_vp;
3498 	error = vn_start_write(fvp, &mp, V_NOWAIT);
3499 	if (error != 0) {
3500 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3501 		NDFREE(&tond, NDF_ONLY_PNBUF);
3502 		if (tvp != NULL)
3503 			vput(tvp);
3504 		if (tdvp == tvp)
3505 			vrele(tdvp);
3506 		else
3507 			vput(tdvp);
3508 		vrele(fromnd.ni_dvp);
3509 		vrele(fvp);
3510 		vrele(tond.ni_startdir);
3511 		if (fromnd.ni_startdir != NULL)
3512 			vrele(fromnd.ni_startdir);
3513 		error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH);
3514 		if (error != 0)
3515 			return (error);
3516 		goto again;
3517 	}
3518 	if (tvp != NULL) {
3519 		if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
3520 			error = ENOTDIR;
3521 			goto out;
3522 		} else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
3523 			error = EISDIR;
3524 			goto out;
3525 		}
3526 #ifdef CAPABILITIES
3527 		if (newfd != AT_FDCWD) {
3528 			/*
3529 			 * If the target already exists we require CAP_UNLINKAT
3530 			 * from 'newfd'.
3531 			 */
3532 			error = cap_check(&tond.ni_filecaps.fc_rights,
3533 			    cap_rights_init(&rights, CAP_UNLINKAT));
3534 			if (error != 0)
3535 				goto out;
3536 		}
3537 #endif
3538 	}
3539 	if (fvp == tdvp) {
3540 		error = EINVAL;
3541 		goto out;
3542 	}
3543 	/*
3544 	 * If the source is the same as the destination (that is, if they
3545 	 * are links to the same vnode), then there is nothing to do.
3546 	 */
3547 	if (fvp == tvp)
3548 		error = -1;
3549 #ifdef MAC
3550 	else
3551 		error = mac_vnode_check_rename_to(td->td_ucred, tdvp,
3552 		    tond.ni_vp, fromnd.ni_dvp == tdvp, &tond.ni_cnd);
3553 #endif
3554 out:
3555 	if (error == 0) {
3556 		error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
3557 		    tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
3558 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3559 		NDFREE(&tond, NDF_ONLY_PNBUF);
3560 	} else {
3561 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3562 		NDFREE(&tond, NDF_ONLY_PNBUF);
3563 		if (tvp != NULL)
3564 			vput(tvp);
3565 		if (tdvp == tvp)
3566 			vrele(tdvp);
3567 		else
3568 			vput(tdvp);
3569 		vrele(fromnd.ni_dvp);
3570 		vrele(fvp);
3571 	}
3572 	vrele(tond.ni_startdir);
3573 	vn_finished_write(mp);
3574 out1:
3575 	if (fromnd.ni_startdir)
3576 		vrele(fromnd.ni_startdir);
3577 	if (error == -1)
3578 		return (0);
3579 	return (error);
3580 }
3581 
3582 /*
3583  * Make a directory file.
3584  */
3585 #ifndef _SYS_SYSPROTO_H_
3586 struct mkdir_args {
3587 	char	*path;
3588 	int	mode;
3589 };
3590 #endif
3591 int
3592 sys_mkdir(td, uap)
3593 	struct thread *td;
3594 	register struct mkdir_args /* {
3595 		char *path;
3596 		int mode;
3597 	} */ *uap;
3598 {
3599 
3600 	return (kern_mkdirat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
3601 	    uap->mode));
3602 }
3603 
3604 #ifndef _SYS_SYSPROTO_H_
3605 struct mkdirat_args {
3606 	int	fd;
3607 	char	*path;
3608 	mode_t	mode;
3609 };
3610 #endif
3611 int
3612 sys_mkdirat(struct thread *td, struct mkdirat_args *uap)
3613 {
3614 
3615 	return (kern_mkdirat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode));
3616 }
3617 
3618 int
3619 kern_mkdirat(struct thread *td, int fd, char *path, enum uio_seg segflg,
3620     int mode)
3621 {
3622 	struct mount *mp;
3623 	struct vnode *vp;
3624 	struct vattr vattr;
3625 	struct nameidata nd;
3626 	cap_rights_t rights;
3627 	int error;
3628 
3629 	AUDIT_ARG_MODE(mode);
3630 restart:
3631 	bwillwrite();
3632 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
3633 	    NOCACHE, segflg, path, fd, cap_rights_init(&rights, CAP_MKDIRAT),
3634 	    td);
3635 	nd.ni_cnd.cn_flags |= WILLBEDIR;
3636 	if ((error = namei(&nd)) != 0)
3637 		return (error);
3638 	vp = nd.ni_vp;
3639 	if (vp != NULL) {
3640 		NDFREE(&nd, NDF_ONLY_PNBUF);
3641 		/*
3642 		 * XXX namei called with LOCKPARENT but not LOCKLEAF has
3643 		 * the strange behaviour of leaving the vnode unlocked
3644 		 * if the target is the same vnode as the parent.
3645 		 */
3646 		if (vp == nd.ni_dvp)
3647 			vrele(nd.ni_dvp);
3648 		else
3649 			vput(nd.ni_dvp);
3650 		vrele(vp);
3651 		return (EEXIST);
3652 	}
3653 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3654 		NDFREE(&nd, NDF_ONLY_PNBUF);
3655 		vput(nd.ni_dvp);
3656 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3657 			return (error);
3658 		goto restart;
3659 	}
3660 	VATTR_NULL(&vattr);
3661 	vattr.va_type = VDIR;
3662 	vattr.va_mode = (mode & ACCESSPERMS) &~ td->td_proc->p_fd->fd_cmask;
3663 #ifdef MAC
3664 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
3665 	    &vattr);
3666 	if (error != 0)
3667 		goto out;
3668 #endif
3669 	error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
3670 #ifdef MAC
3671 out:
3672 #endif
3673 	NDFREE(&nd, NDF_ONLY_PNBUF);
3674 	vput(nd.ni_dvp);
3675 	if (error == 0)
3676 		vput(nd.ni_vp);
3677 	vn_finished_write(mp);
3678 	return (error);
3679 }
3680 
3681 /*
3682  * Remove a directory file.
3683  */
3684 #ifndef _SYS_SYSPROTO_H_
3685 struct rmdir_args {
3686 	char	*path;
3687 };
3688 #endif
3689 int
3690 sys_rmdir(td, uap)
3691 	struct thread *td;
3692 	struct rmdir_args /* {
3693 		char *path;
3694 	} */ *uap;
3695 {
3696 
3697 	return (kern_rmdirat(td, AT_FDCWD, uap->path, UIO_USERSPACE));
3698 }
3699 
3700 int
3701 kern_rmdirat(struct thread *td, int fd, char *path, enum uio_seg pathseg)
3702 {
3703 	struct mount *mp;
3704 	struct vnode *vp;
3705 	struct nameidata nd;
3706 	cap_rights_t rights;
3707 	int error;
3708 
3709 restart:
3710 	bwillwrite();
3711 	NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1,
3712 	    pathseg, path, fd, cap_rights_init(&rights, CAP_UNLINKAT), td);
3713 	if ((error = namei(&nd)) != 0)
3714 		return (error);
3715 	vp = nd.ni_vp;
3716 	if (vp->v_type != VDIR) {
3717 		error = ENOTDIR;
3718 		goto out;
3719 	}
3720 	/*
3721 	 * No rmdir "." please.
3722 	 */
3723 	if (nd.ni_dvp == vp) {
3724 		error = EINVAL;
3725 		goto out;
3726 	}
3727 	/*
3728 	 * The root of a mounted filesystem cannot be deleted.
3729 	 */
3730 	if (vp->v_vflag & VV_ROOT) {
3731 		error = EBUSY;
3732 		goto out;
3733 	}
3734 #ifdef MAC
3735 	error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
3736 	    &nd.ni_cnd);
3737 	if (error != 0)
3738 		goto out;
3739 #endif
3740 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3741 		NDFREE(&nd, NDF_ONLY_PNBUF);
3742 		vput(vp);
3743 		if (nd.ni_dvp == vp)
3744 			vrele(nd.ni_dvp);
3745 		else
3746 			vput(nd.ni_dvp);
3747 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3748 			return (error);
3749 		goto restart;
3750 	}
3751 	vfs_notify_upper(vp, VFS_NOTIFY_UPPER_UNLINK);
3752 	error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
3753 	vn_finished_write(mp);
3754 out:
3755 	NDFREE(&nd, NDF_ONLY_PNBUF);
3756 	vput(vp);
3757 	if (nd.ni_dvp == vp)
3758 		vrele(nd.ni_dvp);
3759 	else
3760 		vput(nd.ni_dvp);
3761 	return (error);
3762 }
3763 
3764 #ifdef COMPAT_43
3765 /*
3766  * Read a block of directory entries in a filesystem independent format.
3767  */
3768 #ifndef _SYS_SYSPROTO_H_
3769 struct ogetdirentries_args {
3770 	int	fd;
3771 	char	*buf;
3772 	u_int	count;
3773 	long	*basep;
3774 };
3775 #endif
3776 int
3777 ogetdirentries(struct thread *td, struct ogetdirentries_args *uap)
3778 {
3779 	long loff;
3780 	int error;
3781 
3782 	error = kern_ogetdirentries(td, uap, &loff);
3783 	if (error == 0)
3784 		error = copyout(&loff, uap->basep, sizeof(long));
3785 	return (error);
3786 }
3787 
3788 int
3789 kern_ogetdirentries(struct thread *td, struct ogetdirentries_args *uap,
3790     long *ploff)
3791 {
3792 	struct vnode *vp;
3793 	struct file *fp;
3794 	struct uio auio, kuio;
3795 	struct iovec aiov, kiov;
3796 	struct dirent *dp, *edp;
3797 	cap_rights_t rights;
3798 	caddr_t dirbuf;
3799 	int error, eofflag, readcnt;
3800 	long loff;
3801 	off_t foffset;
3802 
3803 	/* XXX arbitrary sanity limit on `count'. */
3804 	if (uap->count > 64 * 1024)
3805 		return (EINVAL);
3806 	error = getvnode(td, uap->fd, cap_rights_init(&rights, CAP_READ), &fp);
3807 	if (error != 0)
3808 		return (error);
3809 	if ((fp->f_flag & FREAD) == 0) {
3810 		fdrop(fp, td);
3811 		return (EBADF);
3812 	}
3813 	vp = fp->f_vnode;
3814 	foffset = foffset_lock(fp, 0);
3815 unionread:
3816 	if (vp->v_type != VDIR) {
3817 		foffset_unlock(fp, foffset, 0);
3818 		fdrop(fp, td);
3819 		return (EINVAL);
3820 	}
3821 	aiov.iov_base = uap->buf;
3822 	aiov.iov_len = uap->count;
3823 	auio.uio_iov = &aiov;
3824 	auio.uio_iovcnt = 1;
3825 	auio.uio_rw = UIO_READ;
3826 	auio.uio_segflg = UIO_USERSPACE;
3827 	auio.uio_td = td;
3828 	auio.uio_resid = uap->count;
3829 	vn_lock(vp, LK_SHARED | LK_RETRY);
3830 	loff = auio.uio_offset = foffset;
3831 #ifdef MAC
3832 	error = mac_vnode_check_readdir(td->td_ucred, vp);
3833 	if (error != 0) {
3834 		VOP_UNLOCK(vp, 0);
3835 		foffset_unlock(fp, foffset, FOF_NOUPDATE);
3836 		fdrop(fp, td);
3837 		return (error);
3838 	}
3839 #endif
3840 #	if (BYTE_ORDER != LITTLE_ENDIAN)
3841 		if (vp->v_mount->mnt_maxsymlinklen <= 0) {
3842 			error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag,
3843 			    NULL, NULL);
3844 			foffset = auio.uio_offset;
3845 		} else
3846 #	endif
3847 	{
3848 		kuio = auio;
3849 		kuio.uio_iov = &kiov;
3850 		kuio.uio_segflg = UIO_SYSSPACE;
3851 		kiov.iov_len = uap->count;
3852 		dirbuf = malloc(uap->count, M_TEMP, M_WAITOK);
3853 		kiov.iov_base = dirbuf;
3854 		error = VOP_READDIR(vp, &kuio, fp->f_cred, &eofflag,
3855 			    NULL, NULL);
3856 		foffset = kuio.uio_offset;
3857 		if (error == 0) {
3858 			readcnt = uap->count - kuio.uio_resid;
3859 			edp = (struct dirent *)&dirbuf[readcnt];
3860 			for (dp = (struct dirent *)dirbuf; dp < edp; ) {
3861 #				if (BYTE_ORDER == LITTLE_ENDIAN)
3862 					/*
3863 					 * The expected low byte of
3864 					 * dp->d_namlen is our dp->d_type.
3865 					 * The high MBZ byte of dp->d_namlen
3866 					 * is our dp->d_namlen.
3867 					 */
3868 					dp->d_type = dp->d_namlen;
3869 					dp->d_namlen = 0;
3870 #				else
3871 					/*
3872 					 * The dp->d_type is the high byte
3873 					 * of the expected dp->d_namlen,
3874 					 * so must be zero'ed.
3875 					 */
3876 					dp->d_type = 0;
3877 #				endif
3878 				if (dp->d_reclen > 0) {
3879 					dp = (struct dirent *)
3880 					    ((char *)dp + dp->d_reclen);
3881 				} else {
3882 					error = EIO;
3883 					break;
3884 				}
3885 			}
3886 			if (dp >= edp)
3887 				error = uiomove(dirbuf, readcnt, &auio);
3888 		}
3889 		free(dirbuf, M_TEMP);
3890 	}
3891 	if (error != 0) {
3892 		VOP_UNLOCK(vp, 0);
3893 		foffset_unlock(fp, foffset, 0);
3894 		fdrop(fp, td);
3895 		return (error);
3896 	}
3897 	if (uap->count == auio.uio_resid &&
3898 	    (vp->v_vflag & VV_ROOT) &&
3899 	    (vp->v_mount->mnt_flag & MNT_UNION)) {
3900 		struct vnode *tvp = vp;
3901 		vp = vp->v_mount->mnt_vnodecovered;
3902 		VREF(vp);
3903 		fp->f_vnode = vp;
3904 		fp->f_data = vp;
3905 		foffset = 0;
3906 		vput(tvp);
3907 		goto unionread;
3908 	}
3909 	VOP_UNLOCK(vp, 0);
3910 	foffset_unlock(fp, foffset, 0);
3911 	fdrop(fp, td);
3912 	td->td_retval[0] = uap->count - auio.uio_resid;
3913 	if (error == 0)
3914 		*ploff = loff;
3915 	return (error);
3916 }
3917 #endif /* COMPAT_43 */
3918 
3919 /*
3920  * Read a block of directory entries in a filesystem independent format.
3921  */
3922 #ifndef _SYS_SYSPROTO_H_
3923 struct getdirentries_args {
3924 	int	fd;
3925 	char	*buf;
3926 	u_int	count;
3927 	long	*basep;
3928 };
3929 #endif
3930 int
3931 sys_getdirentries(td, uap)
3932 	struct thread *td;
3933 	register struct getdirentries_args /* {
3934 		int fd;
3935 		char *buf;
3936 		u_int count;
3937 		long *basep;
3938 	} */ *uap;
3939 {
3940 	long base;
3941 	int error;
3942 
3943 	error = kern_getdirentries(td, uap->fd, uap->buf, uap->count, &base,
3944 	    NULL, UIO_USERSPACE);
3945 	if (error != 0)
3946 		return (error);
3947 	if (uap->basep != NULL)
3948 		error = copyout(&base, uap->basep, sizeof(long));
3949 	return (error);
3950 }
3951 
3952 int
3953 kern_getdirentries(struct thread *td, int fd, char *buf, u_int count,
3954     long *basep, ssize_t *residp, enum uio_seg bufseg)
3955 {
3956 	struct vnode *vp;
3957 	struct file *fp;
3958 	struct uio auio;
3959 	struct iovec aiov;
3960 	cap_rights_t rights;
3961 	long loff;
3962 	int error, eofflag;
3963 	off_t foffset;
3964 
3965 	AUDIT_ARG_FD(fd);
3966 	if (count > IOSIZE_MAX)
3967 		return (EINVAL);
3968 	auio.uio_resid = count;
3969 	error = getvnode(td, fd, cap_rights_init(&rights, CAP_READ), &fp);
3970 	if (error != 0)
3971 		return (error);
3972 	if ((fp->f_flag & FREAD) == 0) {
3973 		fdrop(fp, td);
3974 		return (EBADF);
3975 	}
3976 	vp = fp->f_vnode;
3977 	foffset = foffset_lock(fp, 0);
3978 unionread:
3979 	if (vp->v_type != VDIR) {
3980 		error = EINVAL;
3981 		goto fail;
3982 	}
3983 	aiov.iov_base = buf;
3984 	aiov.iov_len = count;
3985 	auio.uio_iov = &aiov;
3986 	auio.uio_iovcnt = 1;
3987 	auio.uio_rw = UIO_READ;
3988 	auio.uio_segflg = bufseg;
3989 	auio.uio_td = td;
3990 	vn_lock(vp, LK_SHARED | LK_RETRY);
3991 	AUDIT_ARG_VNODE1(vp);
3992 	loff = auio.uio_offset = foffset;
3993 #ifdef MAC
3994 	error = mac_vnode_check_readdir(td->td_ucred, vp);
3995 	if (error == 0)
3996 #endif
3997 		error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, NULL,
3998 		    NULL);
3999 	foffset = auio.uio_offset;
4000 	if (error != 0) {
4001 		VOP_UNLOCK(vp, 0);
4002 		goto fail;
4003 	}
4004 	if (count == auio.uio_resid &&
4005 	    (vp->v_vflag & VV_ROOT) &&
4006 	    (vp->v_mount->mnt_flag & MNT_UNION)) {
4007 		struct vnode *tvp = vp;
4008 
4009 		vp = vp->v_mount->mnt_vnodecovered;
4010 		VREF(vp);
4011 		fp->f_vnode = vp;
4012 		fp->f_data = vp;
4013 		foffset = 0;
4014 		vput(tvp);
4015 		goto unionread;
4016 	}
4017 	VOP_UNLOCK(vp, 0);
4018 	*basep = loff;
4019 	if (residp != NULL)
4020 		*residp = auio.uio_resid;
4021 	td->td_retval[0] = count - auio.uio_resid;
4022 fail:
4023 	foffset_unlock(fp, foffset, 0);
4024 	fdrop(fp, td);
4025 	return (error);
4026 }
4027 
4028 #ifndef _SYS_SYSPROTO_H_
4029 struct getdents_args {
4030 	int fd;
4031 	char *buf;
4032 	size_t count;
4033 };
4034 #endif
4035 int
4036 sys_getdents(td, uap)
4037 	struct thread *td;
4038 	register struct getdents_args /* {
4039 		int fd;
4040 		char *buf;
4041 		u_int count;
4042 	} */ *uap;
4043 {
4044 	struct getdirentries_args ap;
4045 
4046 	ap.fd = uap->fd;
4047 	ap.buf = uap->buf;
4048 	ap.count = uap->count;
4049 	ap.basep = NULL;
4050 	return (sys_getdirentries(td, &ap));
4051 }
4052 
4053 /*
4054  * Set the mode mask for creation of filesystem nodes.
4055  */
4056 #ifndef _SYS_SYSPROTO_H_
4057 struct umask_args {
4058 	int	newmask;
4059 };
4060 #endif
4061 int
4062 sys_umask(td, uap)
4063 	struct thread *td;
4064 	struct umask_args /* {
4065 		int newmask;
4066 	} */ *uap;
4067 {
4068 	struct filedesc *fdp;
4069 
4070 	fdp = td->td_proc->p_fd;
4071 	FILEDESC_XLOCK(fdp);
4072 	td->td_retval[0] = fdp->fd_cmask;
4073 	fdp->fd_cmask = uap->newmask & ALLPERMS;
4074 	FILEDESC_XUNLOCK(fdp);
4075 	return (0);
4076 }
4077 
4078 /*
4079  * Void all references to file by ripping underlying filesystem away from
4080  * vnode.
4081  */
4082 #ifndef _SYS_SYSPROTO_H_
4083 struct revoke_args {
4084 	char	*path;
4085 };
4086 #endif
4087 int
4088 sys_revoke(td, uap)
4089 	struct thread *td;
4090 	register struct revoke_args /* {
4091 		char *path;
4092 	} */ *uap;
4093 {
4094 	struct vnode *vp;
4095 	struct vattr vattr;
4096 	struct nameidata nd;
4097 	int error;
4098 
4099 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
4100 	    uap->path, td);
4101 	if ((error = namei(&nd)) != 0)
4102 		return (error);
4103 	vp = nd.ni_vp;
4104 	NDFREE(&nd, NDF_ONLY_PNBUF);
4105 	if (vp->v_type != VCHR || vp->v_rdev == NULL) {
4106 		error = EINVAL;
4107 		goto out;
4108 	}
4109 #ifdef MAC
4110 	error = mac_vnode_check_revoke(td->td_ucred, vp);
4111 	if (error != 0)
4112 		goto out;
4113 #endif
4114 	error = VOP_GETATTR(vp, &vattr, td->td_ucred);
4115 	if (error != 0)
4116 		goto out;
4117 	if (td->td_ucred->cr_uid != vattr.va_uid) {
4118 		error = priv_check(td, PRIV_VFS_ADMIN);
4119 		if (error != 0)
4120 			goto out;
4121 	}
4122 	if (vcount(vp) > 1)
4123 		VOP_REVOKE(vp, REVOKEALL);
4124 out:
4125 	vput(vp);
4126 	return (error);
4127 }
4128 
4129 /*
4130  * Convert a user file descriptor to a kernel file entry and check that, if it
4131  * is a capability, the correct rights are present. A reference on the file
4132  * entry is held upon returning.
4133  */
4134 int
4135 getvnode(struct thread *td, int fd, cap_rights_t *rightsp, struct file **fpp)
4136 {
4137 	struct file *fp;
4138 	int error;
4139 
4140 	error = fget_unlocked(td->td_proc->p_fd, fd, rightsp, &fp, NULL);
4141 	if (error != 0)
4142 		return (error);
4143 
4144 	/*
4145 	 * The file could be not of the vnode type, or it may be not
4146 	 * yet fully initialized, in which case the f_vnode pointer
4147 	 * may be set, but f_ops is still badfileops.  E.g.,
4148 	 * devfs_open() transiently create such situation to
4149 	 * facilitate csw d_fdopen().
4150 	 *
4151 	 * Dupfdopen() handling in kern_openat() installs the
4152 	 * half-baked file into the process descriptor table, allowing
4153 	 * other thread to dereference it. Guard against the race by
4154 	 * checking f_ops.
4155 	 */
4156 	if (fp->f_vnode == NULL || fp->f_ops == &badfileops) {
4157 		fdrop(fp, td);
4158 		return (EINVAL);
4159 	}
4160 	*fpp = fp;
4161 	return (0);
4162 }
4163 
4164 
4165 /*
4166  * Get an (NFS) file handle.
4167  */
4168 #ifndef _SYS_SYSPROTO_H_
4169 struct lgetfh_args {
4170 	char	*fname;
4171 	fhandle_t *fhp;
4172 };
4173 #endif
4174 int
4175 sys_lgetfh(td, uap)
4176 	struct thread *td;
4177 	register struct lgetfh_args *uap;
4178 {
4179 	struct nameidata nd;
4180 	fhandle_t fh;
4181 	register struct vnode *vp;
4182 	int error;
4183 
4184 	error = priv_check(td, PRIV_VFS_GETFH);
4185 	if (error != 0)
4186 		return (error);
4187 	NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
4188 	    uap->fname, td);
4189 	error = namei(&nd);
4190 	if (error != 0)
4191 		return (error);
4192 	NDFREE(&nd, NDF_ONLY_PNBUF);
4193 	vp = nd.ni_vp;
4194 	bzero(&fh, sizeof(fh));
4195 	fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
4196 	error = VOP_VPTOFH(vp, &fh.fh_fid);
4197 	vput(vp);
4198 	if (error == 0)
4199 		error = copyout(&fh, uap->fhp, sizeof (fh));
4200 	return (error);
4201 }
4202 
4203 #ifndef _SYS_SYSPROTO_H_
4204 struct getfh_args {
4205 	char	*fname;
4206 	fhandle_t *fhp;
4207 };
4208 #endif
4209 int
4210 sys_getfh(td, uap)
4211 	struct thread *td;
4212 	register struct getfh_args *uap;
4213 {
4214 	struct nameidata nd;
4215 	fhandle_t fh;
4216 	register struct vnode *vp;
4217 	int error;
4218 
4219 	error = priv_check(td, PRIV_VFS_GETFH);
4220 	if (error != 0)
4221 		return (error);
4222 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
4223 	    uap->fname, td);
4224 	error = namei(&nd);
4225 	if (error != 0)
4226 		return (error);
4227 	NDFREE(&nd, NDF_ONLY_PNBUF);
4228 	vp = nd.ni_vp;
4229 	bzero(&fh, sizeof(fh));
4230 	fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
4231 	error = VOP_VPTOFH(vp, &fh.fh_fid);
4232 	vput(vp);
4233 	if (error == 0)
4234 		error = copyout(&fh, uap->fhp, sizeof (fh));
4235 	return (error);
4236 }
4237 
4238 /*
4239  * syscall for the rpc.lockd to use to translate a NFS file handle into an
4240  * open descriptor.
4241  *
4242  * warning: do not remove the priv_check() call or this becomes one giant
4243  * security hole.
4244  */
4245 #ifndef _SYS_SYSPROTO_H_
4246 struct fhopen_args {
4247 	const struct fhandle *u_fhp;
4248 	int flags;
4249 };
4250 #endif
4251 int
4252 sys_fhopen(td, uap)
4253 	struct thread *td;
4254 	struct fhopen_args /* {
4255 		const struct fhandle *u_fhp;
4256 		int flags;
4257 	} */ *uap;
4258 {
4259 	struct mount *mp;
4260 	struct vnode *vp;
4261 	struct fhandle fhp;
4262 	struct file *fp;
4263 	int fmode, error;
4264 	int indx;
4265 
4266 	error = priv_check(td, PRIV_VFS_FHOPEN);
4267 	if (error != 0)
4268 		return (error);
4269 	indx = -1;
4270 	fmode = FFLAGS(uap->flags);
4271 	/* why not allow a non-read/write open for our lockd? */
4272 	if (((fmode & (FREAD | FWRITE)) == 0) || (fmode & O_CREAT))
4273 		return (EINVAL);
4274 	error = copyin(uap->u_fhp, &fhp, sizeof(fhp));
4275 	if (error != 0)
4276 		return(error);
4277 	/* find the mount point */
4278 	mp = vfs_busyfs(&fhp.fh_fsid);
4279 	if (mp == NULL)
4280 		return (ESTALE);
4281 	/* now give me my vnode, it gets returned to me locked */
4282 	error = VFS_FHTOVP(mp, &fhp.fh_fid, LK_EXCLUSIVE, &vp);
4283 	vfs_unbusy(mp);
4284 	if (error != 0)
4285 		return (error);
4286 
4287 	error = falloc_noinstall(td, &fp);
4288 	if (error != 0) {
4289 		vput(vp);
4290 		return (error);
4291 	}
4292 	/*
4293 	 * An extra reference on `fp' has been held for us by
4294 	 * falloc_noinstall().
4295 	 */
4296 
4297 #ifdef INVARIANTS
4298 	td->td_dupfd = -1;
4299 #endif
4300 	error = vn_open_vnode(vp, fmode, td->td_ucred, td, fp);
4301 	if (error != 0) {
4302 		KASSERT(fp->f_ops == &badfileops,
4303 		    ("VOP_OPEN in fhopen() set f_ops"));
4304 		KASSERT(td->td_dupfd < 0,
4305 		    ("fhopen() encountered fdopen()"));
4306 
4307 		vput(vp);
4308 		goto bad;
4309 	}
4310 #ifdef INVARIANTS
4311 	td->td_dupfd = 0;
4312 #endif
4313 	fp->f_vnode = vp;
4314 	fp->f_seqcount = 1;
4315 	finit(fp, (fmode & FMASK) | (fp->f_flag & FHASLOCK), DTYPE_VNODE, vp,
4316 	    &vnops);
4317 	VOP_UNLOCK(vp, 0);
4318 	if ((fmode & O_TRUNC) != 0) {
4319 		error = fo_truncate(fp, 0, td->td_ucred, td);
4320 		if (error != 0)
4321 			goto bad;
4322 	}
4323 
4324 	error = finstall(td, fp, &indx, fmode, NULL);
4325 bad:
4326 	fdrop(fp, td);
4327 	td->td_retval[0] = indx;
4328 	return (error);
4329 }
4330 
4331 /*
4332  * Stat an (NFS) file handle.
4333  */
4334 #ifndef _SYS_SYSPROTO_H_
4335 struct fhstat_args {
4336 	struct fhandle *u_fhp;
4337 	struct stat *sb;
4338 };
4339 #endif
4340 int
4341 sys_fhstat(td, uap)
4342 	struct thread *td;
4343 	register struct fhstat_args /* {
4344 		struct fhandle *u_fhp;
4345 		struct stat *sb;
4346 	} */ *uap;
4347 {
4348 	struct stat sb;
4349 	struct fhandle fh;
4350 	int error;
4351 
4352 	error = copyin(uap->u_fhp, &fh, sizeof(fh));
4353 	if (error != 0)
4354 		return (error);
4355 	error = kern_fhstat(td, fh, &sb);
4356 	if (error == 0)
4357 		error = copyout(&sb, uap->sb, sizeof(sb));
4358 	return (error);
4359 }
4360 
4361 int
4362 kern_fhstat(struct thread *td, struct fhandle fh, struct stat *sb)
4363 {
4364 	struct mount *mp;
4365 	struct vnode *vp;
4366 	int error;
4367 
4368 	error = priv_check(td, PRIV_VFS_FHSTAT);
4369 	if (error != 0)
4370 		return (error);
4371 	if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4372 		return (ESTALE);
4373 	error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4374 	vfs_unbusy(mp);
4375 	if (error != 0)
4376 		return (error);
4377 	error = vn_stat(vp, sb, td->td_ucred, NOCRED, td);
4378 	vput(vp);
4379 	return (error);
4380 }
4381 
4382 /*
4383  * Implement fstatfs() for (NFS) file handles.
4384  */
4385 #ifndef _SYS_SYSPROTO_H_
4386 struct fhstatfs_args {
4387 	struct fhandle *u_fhp;
4388 	struct statfs *buf;
4389 };
4390 #endif
4391 int
4392 sys_fhstatfs(td, uap)
4393 	struct thread *td;
4394 	struct fhstatfs_args /* {
4395 		struct fhandle *u_fhp;
4396 		struct statfs *buf;
4397 	} */ *uap;
4398 {
4399 	struct statfs sf;
4400 	fhandle_t fh;
4401 	int error;
4402 
4403 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
4404 	if (error != 0)
4405 		return (error);
4406 	error = kern_fhstatfs(td, fh, &sf);
4407 	if (error != 0)
4408 		return (error);
4409 	return (copyout(&sf, uap->buf, sizeof(sf)));
4410 }
4411 
4412 int
4413 kern_fhstatfs(struct thread *td, fhandle_t fh, struct statfs *buf)
4414 {
4415 	struct statfs *sp;
4416 	struct mount *mp;
4417 	struct vnode *vp;
4418 	int error;
4419 
4420 	error = priv_check(td, PRIV_VFS_FHSTATFS);
4421 	if (error != 0)
4422 		return (error);
4423 	if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4424 		return (ESTALE);
4425 	error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4426 	if (error != 0) {
4427 		vfs_unbusy(mp);
4428 		return (error);
4429 	}
4430 	vput(vp);
4431 	error = prison_canseemount(td->td_ucred, mp);
4432 	if (error != 0)
4433 		goto out;
4434 #ifdef MAC
4435 	error = mac_mount_check_stat(td->td_ucred, mp);
4436 	if (error != 0)
4437 		goto out;
4438 #endif
4439 	/*
4440 	 * Set these in case the underlying filesystem fails to do so.
4441 	 */
4442 	sp = &mp->mnt_stat;
4443 	sp->f_version = STATFS_VERSION;
4444 	sp->f_namemax = NAME_MAX;
4445 	sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
4446 	error = VFS_STATFS(mp, sp);
4447 	if (error == 0)
4448 		*buf = *sp;
4449 out:
4450 	vfs_unbusy(mp);
4451 	return (error);
4452 }
4453 
4454 int
4455 kern_posix_fallocate(struct thread *td, int fd, off_t offset, off_t len)
4456 {
4457 	struct file *fp;
4458 	struct mount *mp;
4459 	struct vnode *vp;
4460 	cap_rights_t rights;
4461 	off_t olen, ooffset;
4462 	int error;
4463 
4464 	if (offset < 0 || len <= 0)
4465 		return (EINVAL);
4466 	/* Check for wrap. */
4467 	if (offset > OFF_MAX - len)
4468 		return (EFBIG);
4469 	error = fget(td, fd, cap_rights_init(&rights, CAP_WRITE), &fp);
4470 	if (error != 0)
4471 		return (error);
4472 	if ((fp->f_ops->fo_flags & DFLAG_SEEKABLE) == 0) {
4473 		error = ESPIPE;
4474 		goto out;
4475 	}
4476 	if ((fp->f_flag & FWRITE) == 0) {
4477 		error = EBADF;
4478 		goto out;
4479 	}
4480 	if (fp->f_type != DTYPE_VNODE) {
4481 		error = ENODEV;
4482 		goto out;
4483 	}
4484 	vp = fp->f_vnode;
4485 	if (vp->v_type != VREG) {
4486 		error = ENODEV;
4487 		goto out;
4488 	}
4489 
4490 	/* Allocating blocks may take a long time, so iterate. */
4491 	for (;;) {
4492 		olen = len;
4493 		ooffset = offset;
4494 
4495 		bwillwrite();
4496 		mp = NULL;
4497 		error = vn_start_write(vp, &mp, V_WAIT | PCATCH);
4498 		if (error != 0)
4499 			break;
4500 		error = vn_lock(vp, LK_EXCLUSIVE);
4501 		if (error != 0) {
4502 			vn_finished_write(mp);
4503 			break;
4504 		}
4505 #ifdef MAC
4506 		error = mac_vnode_check_write(td->td_ucred, fp->f_cred, vp);
4507 		if (error == 0)
4508 #endif
4509 			error = VOP_ALLOCATE(vp, &offset, &len);
4510 		VOP_UNLOCK(vp, 0);
4511 		vn_finished_write(mp);
4512 
4513 		if (olen + ooffset != offset + len) {
4514 			panic("offset + len changed from %jx/%jx to %jx/%jx",
4515 			    ooffset, olen, offset, len);
4516 		}
4517 		if (error != 0 || len == 0)
4518 			break;
4519 		KASSERT(olen > len, ("Iteration did not make progress?"));
4520 		maybe_yield();
4521 	}
4522  out:
4523 	fdrop(fp, td);
4524 	return (error);
4525 }
4526 
4527 int
4528 sys_posix_fallocate(struct thread *td, struct posix_fallocate_args *uap)
4529 {
4530 
4531 	td->td_retval[0] = kern_posix_fallocate(td, uap->fd, uap->offset,
4532 	    uap->len);
4533 	return (0);
4534 }
4535 
4536 /*
4537  * Unlike madvise(2), we do not make a best effort to remember every
4538  * possible caching hint.  Instead, we remember the last setting with
4539  * the exception that we will allow POSIX_FADV_NORMAL to adjust the
4540  * region of any current setting.
4541  */
4542 int
4543 kern_posix_fadvise(struct thread *td, int fd, off_t offset, off_t len,
4544     int advice)
4545 {
4546 	struct fadvise_info *fa, *new;
4547 	struct file *fp;
4548 	struct vnode *vp;
4549 	cap_rights_t rights;
4550 	off_t end;
4551 	int error;
4552 
4553 	if (offset < 0 || len < 0 || offset > OFF_MAX - len)
4554 		return (EINVAL);
4555 	switch (advice) {
4556 	case POSIX_FADV_SEQUENTIAL:
4557 	case POSIX_FADV_RANDOM:
4558 	case POSIX_FADV_NOREUSE:
4559 		new = malloc(sizeof(*fa), M_FADVISE, M_WAITOK);
4560 		break;
4561 	case POSIX_FADV_NORMAL:
4562 	case POSIX_FADV_WILLNEED:
4563 	case POSIX_FADV_DONTNEED:
4564 		new = NULL;
4565 		break;
4566 	default:
4567 		return (EINVAL);
4568 	}
4569 	/* XXX: CAP_POSIX_FADVISE? */
4570 	error = fget(td, fd, cap_rights_init(&rights), &fp);
4571 	if (error != 0)
4572 		goto out;
4573 	if ((fp->f_ops->fo_flags & DFLAG_SEEKABLE) == 0) {
4574 		error = ESPIPE;
4575 		goto out;
4576 	}
4577 	if (fp->f_type != DTYPE_VNODE) {
4578 		error = ENODEV;
4579 		goto out;
4580 	}
4581 	vp = fp->f_vnode;
4582 	if (vp->v_type != VREG) {
4583 		error = ENODEV;
4584 		goto out;
4585 	}
4586 	if (len == 0)
4587 		end = OFF_MAX;
4588 	else
4589 		end = offset + len - 1;
4590 	switch (advice) {
4591 	case POSIX_FADV_SEQUENTIAL:
4592 	case POSIX_FADV_RANDOM:
4593 	case POSIX_FADV_NOREUSE:
4594 		/*
4595 		 * Try to merge any existing non-standard region with
4596 		 * this new region if possible, otherwise create a new
4597 		 * non-standard region for this request.
4598 		 */
4599 		mtx_pool_lock(mtxpool_sleep, fp);
4600 		fa = fp->f_advice;
4601 		if (fa != NULL && fa->fa_advice == advice &&
4602 		    ((fa->fa_start <= end && fa->fa_end >= offset) ||
4603 		    (end != OFF_MAX && fa->fa_start == end + 1) ||
4604 		    (fa->fa_end != OFF_MAX && fa->fa_end + 1 == offset))) {
4605 			if (offset < fa->fa_start)
4606 				fa->fa_start = offset;
4607 			if (end > fa->fa_end)
4608 				fa->fa_end = end;
4609 		} else {
4610 			new->fa_advice = advice;
4611 			new->fa_start = offset;
4612 			new->fa_end = end;
4613 			new->fa_prevstart = 0;
4614 			new->fa_prevend = 0;
4615 			fp->f_advice = new;
4616 			new = fa;
4617 		}
4618 		mtx_pool_unlock(mtxpool_sleep, fp);
4619 		break;
4620 	case POSIX_FADV_NORMAL:
4621 		/*
4622 		 * If a the "normal" region overlaps with an existing
4623 		 * non-standard region, trim or remove the
4624 		 * non-standard region.
4625 		 */
4626 		mtx_pool_lock(mtxpool_sleep, fp);
4627 		fa = fp->f_advice;
4628 		if (fa != NULL) {
4629 			if (offset <= fa->fa_start && end >= fa->fa_end) {
4630 				new = fa;
4631 				fp->f_advice = NULL;
4632 			} else if (offset <= fa->fa_start &&
4633 			    end >= fa->fa_start)
4634 				fa->fa_start = end + 1;
4635 			else if (offset <= fa->fa_end && end >= fa->fa_end)
4636 				fa->fa_end = offset - 1;
4637 			else if (offset >= fa->fa_start && end <= fa->fa_end) {
4638 				/*
4639 				 * If the "normal" region is a middle
4640 				 * portion of the existing
4641 				 * non-standard region, just remove
4642 				 * the whole thing rather than picking
4643 				 * one side or the other to
4644 				 * preserve.
4645 				 */
4646 				new = fa;
4647 				fp->f_advice = NULL;
4648 			}
4649 		}
4650 		mtx_pool_unlock(mtxpool_sleep, fp);
4651 		break;
4652 	case POSIX_FADV_WILLNEED:
4653 	case POSIX_FADV_DONTNEED:
4654 		error = VOP_ADVISE(vp, offset, end, advice);
4655 		break;
4656 	}
4657 out:
4658 	if (fp != NULL)
4659 		fdrop(fp, td);
4660 	free(new, M_FADVISE);
4661 	return (error);
4662 }
4663 
4664 int
4665 sys_posix_fadvise(struct thread *td, struct posix_fadvise_args *uap)
4666 {
4667 
4668 	td->td_retval[0] = kern_posix_fadvise(td, uap->fd, uap->offset,
4669 	    uap->len, uap->advice);
4670 	return (0);
4671 }
4672