xref: /freebsd/sys/kern/vfs_syscalls.c (revision 17a0c1eb753c2f86a4a8ce5b0054b42ce9094957)
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  * 3. Neither the name of the University nor the names of its contributors
19  *    may be used to endorse or promote products derived from this software
20  *    without specific prior written permission.
21  *
22  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
23  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
24  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
25  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
26  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
27  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
28  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
29  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
30  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
31  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32  * SUCH DAMAGE.
33  *
34  *	@(#)vfs_syscalls.c	8.13 (Berkeley) 4/15/94
35  */
36 
37 #include <sys/cdefs.h>
38 __FBSDID("$FreeBSD$");
39 
40 #include "opt_capsicum.h"
41 #include "opt_compat.h"
42 #include "opt_ktrace.h"
43 
44 #include <sys/param.h>
45 #include <sys/systm.h>
46 #include <sys/bio.h>
47 #include <sys/buf.h>
48 #include <sys/capsicum.h>
49 #include <sys/disk.h>
50 #include <sys/sysent.h>
51 #include <sys/malloc.h>
52 #include <sys/mount.h>
53 #include <sys/mutex.h>
54 #include <sys/sysproto.h>
55 #include <sys/namei.h>
56 #include <sys/filedesc.h>
57 #include <sys/kernel.h>
58 #include <sys/fcntl.h>
59 #include <sys/file.h>
60 #include <sys/filio.h>
61 #include <sys/limits.h>
62 #include <sys/linker.h>
63 #include <sys/rwlock.h>
64 #include <sys/sdt.h>
65 #include <sys/stat.h>
66 #include <sys/sx.h>
67 #include <sys/unistd.h>
68 #include <sys/vnode.h>
69 #include <sys/priv.h>
70 #include <sys/proc.h>
71 #include <sys/dirent.h>
72 #include <sys/jail.h>
73 #include <sys/syscallsubr.h>
74 #include <sys/sysctl.h>
75 #ifdef KTRACE
76 #include <sys/ktrace.h>
77 #endif
78 
79 #include <machine/stdarg.h>
80 
81 #include <security/audit/audit.h>
82 #include <security/mac/mac_framework.h>
83 
84 #include <vm/vm.h>
85 #include <vm/vm_object.h>
86 #include <vm/vm_page.h>
87 #include <vm/uma.h>
88 
89 #include <ufs/ufs/quota.h>
90 
91 MALLOC_DEFINE(M_FADVISE, "fadvise", "posix_fadvise(2) information");
92 
93 SDT_PROVIDER_DEFINE(vfs);
94 SDT_PROBE_DEFINE2(vfs, , stat, mode, "char *", "int");
95 SDT_PROBE_DEFINE2(vfs, , stat, reg, "char *", "int");
96 
97 static int kern_chflagsat(struct thread *td, int fd, const char *path,
98     enum uio_seg pathseg, u_long flags, int atflag);
99 static int setfflags(struct thread *td, struct vnode *, u_long);
100 static int getutimes(const struct timeval *, enum uio_seg, struct timespec *);
101 static int getutimens(const struct timespec *, enum uio_seg,
102     struct timespec *, int *);
103 static int setutimes(struct thread *td, struct vnode *,
104     const struct timespec *, int, int);
105 static int vn_access(struct vnode *vp, int user_flags, struct ucred *cred,
106     struct thread *td);
107 
108 /*
109  * Sync each mounted filesystem.
110  */
111 #ifndef _SYS_SYSPROTO_H_
112 struct sync_args {
113 	int     dummy;
114 };
115 #endif
116 /* ARGSUSED */
117 int
118 sys_sync(td, uap)
119 	struct thread *td;
120 	struct sync_args *uap;
121 {
122 	struct mount *mp, *nmp;
123 	int save;
124 
125 	mtx_lock(&mountlist_mtx);
126 	for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
127 		if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK)) {
128 			nmp = TAILQ_NEXT(mp, mnt_list);
129 			continue;
130 		}
131 		if ((mp->mnt_flag & MNT_RDONLY) == 0 &&
132 		    vn_start_write(NULL, &mp, V_NOWAIT) == 0) {
133 			save = curthread_pflags_set(TDP_SYNCIO);
134 			vfs_msync(mp, MNT_NOWAIT);
135 			VFS_SYNC(mp, MNT_NOWAIT);
136 			curthread_pflags_restore(save);
137 			vn_finished_write(mp);
138 		}
139 		mtx_lock(&mountlist_mtx);
140 		nmp = TAILQ_NEXT(mp, mnt_list);
141 		vfs_unbusy(mp);
142 	}
143 	mtx_unlock(&mountlist_mtx);
144 	return (0);
145 }
146 
147 /*
148  * Change filesystem quotas.
149  */
150 #ifndef _SYS_SYSPROTO_H_
151 struct quotactl_args {
152 	char *path;
153 	int cmd;
154 	int uid;
155 	caddr_t arg;
156 };
157 #endif
158 int
159 sys_quotactl(td, uap)
160 	struct thread *td;
161 	register struct quotactl_args /* {
162 		char *path;
163 		int cmd;
164 		int uid;
165 		caddr_t arg;
166 	} */ *uap;
167 {
168 	struct mount *mp;
169 	struct nameidata nd;
170 	int error;
171 
172 	AUDIT_ARG_CMD(uap->cmd);
173 	AUDIT_ARG_UID(uap->uid);
174 	if (!prison_allow(td->td_ucred, PR_ALLOW_QUOTAS))
175 		return (EPERM);
176 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
177 	    uap->path, td);
178 	if ((error = namei(&nd)) != 0)
179 		return (error);
180 	NDFREE(&nd, NDF_ONLY_PNBUF);
181 	mp = nd.ni_vp->v_mount;
182 	vfs_ref(mp);
183 	vput(nd.ni_vp);
184 	error = vfs_busy(mp, 0);
185 	vfs_rel(mp);
186 	if (error != 0)
187 		return (error);
188 	error = VFS_QUOTACTL(mp, uap->cmd, uap->uid, uap->arg);
189 
190 	/*
191 	 * Since quota on operation typically needs to open quota
192 	 * file, the Q_QUOTAON handler needs to unbusy the mount point
193 	 * before calling into namei.  Otherwise, unmount might be
194 	 * started between two vfs_busy() invocations (first is our,
195 	 * second is from mount point cross-walk code in lookup()),
196 	 * causing deadlock.
197 	 *
198 	 * Require that Q_QUOTAON handles the vfs_busy() reference on
199 	 * its own, always returning with ubusied mount point.
200 	 */
201 	if ((uap->cmd >> SUBCMDSHIFT) != Q_QUOTAON)
202 		vfs_unbusy(mp);
203 	return (error);
204 }
205 
206 /*
207  * Used by statfs conversion routines to scale the block size up if
208  * necessary so that all of the block counts are <= 'max_size'.  Note
209  * that 'max_size' should be a bitmask, i.e. 2^n - 1 for some non-zero
210  * value of 'n'.
211  */
212 void
213 statfs_scale_blocks(struct statfs *sf, long max_size)
214 {
215 	uint64_t count;
216 	int shift;
217 
218 	KASSERT(powerof2(max_size + 1), ("%s: invalid max_size", __func__));
219 
220 	/*
221 	 * Attempt to scale the block counts to give a more accurate
222 	 * overview to userland of the ratio of free space to used
223 	 * space.  To do this, find the largest block count and compute
224 	 * a divisor that lets it fit into a signed integer <= max_size.
225 	 */
226 	if (sf->f_bavail < 0)
227 		count = -sf->f_bavail;
228 	else
229 		count = sf->f_bavail;
230 	count = MAX(sf->f_blocks, MAX(sf->f_bfree, count));
231 	if (count <= max_size)
232 		return;
233 
234 	count >>= flsl(max_size);
235 	shift = 0;
236 	while (count > 0) {
237 		shift++;
238 		count >>=1;
239 	}
240 
241 	sf->f_bsize <<= shift;
242 	sf->f_blocks >>= shift;
243 	sf->f_bfree >>= shift;
244 	sf->f_bavail >>= shift;
245 }
246 
247 static int
248 kern_do_statfs(struct thread *td, struct mount *mp, struct statfs *buf)
249 {
250 	struct statfs *sp;
251 	int error;
252 
253 	if (mp == NULL)
254 		return (EBADF);
255 	error = vfs_busy(mp, 0);
256 	vfs_rel(mp);
257 	if (error != 0)
258 		return (error);
259 #ifdef MAC
260 	error = mac_mount_check_stat(td->td_ucred, mp);
261 	if (error != 0)
262 		goto out;
263 #endif
264 	/*
265 	 * Set these in case the underlying filesystem fails to do so.
266 	 */
267 	sp = &mp->mnt_stat;
268 	sp->f_version = STATFS_VERSION;
269 	sp->f_namemax = NAME_MAX;
270 	sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
271 	error = VFS_STATFS(mp, sp);
272 	if (error != 0)
273 		goto out;
274 	*buf = *sp;
275 	if (priv_check(td, PRIV_VFS_GENERATION)) {
276 		buf->f_fsid.val[0] = buf->f_fsid.val[1] = 0;
277 		prison_enforce_statfs(td->td_ucred, mp, buf);
278 	}
279 out:
280 	vfs_unbusy(mp);
281 	return (error);
282 }
283 
284 /*
285  * Get filesystem statistics.
286  */
287 #ifndef _SYS_SYSPROTO_H_
288 struct statfs_args {
289 	char *path;
290 	struct statfs *buf;
291 };
292 #endif
293 int
294 sys_statfs(td, uap)
295 	struct thread *td;
296 	register struct statfs_args /* {
297 		char *path;
298 		struct statfs *buf;
299 	} */ *uap;
300 {
301 	struct statfs sf;
302 	int error;
303 
304 	error = kern_statfs(td, uap->path, UIO_USERSPACE, &sf);
305 	if (error == 0)
306 		error = copyout(&sf, uap->buf, sizeof(sf));
307 	return (error);
308 }
309 
310 int
311 kern_statfs(struct thread *td, char *path, enum uio_seg pathseg,
312     struct statfs *buf)
313 {
314 	struct mount *mp;
315 	struct nameidata nd;
316 	int error;
317 
318 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
319 	    pathseg, path, td);
320 	error = namei(&nd);
321 	if (error != 0)
322 		return (error);
323 	mp = nd.ni_vp->v_mount;
324 	vfs_ref(mp);
325 	NDFREE(&nd, NDF_ONLY_PNBUF);
326 	vput(nd.ni_vp);
327 	return (kern_do_statfs(td, mp, buf));
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 vnode *vp;
362 	cap_rights_t rights;
363 	int error;
364 
365 	AUDIT_ARG_FD(fd);
366 	error = getvnode(td, fd, cap_rights_init(&rights, CAP_FSTATFS), &fp);
367 	if (error != 0)
368 		return (error);
369 	vp = fp->f_vnode;
370 	vn_lock(vp, LK_SHARED | LK_RETRY);
371 #ifdef AUDIT
372 	AUDIT_ARG_VNODE1(vp);
373 #endif
374 	mp = vp->v_mount;
375 	if (mp != NULL)
376 		vfs_ref(mp);
377 	VOP_UNLOCK(vp, 0);
378 	fdrop(fp, td);
379 	return (kern_do_statfs(td, mp, buf));
380 }
381 
382 /*
383  * Get statistics on all filesystems.
384  */
385 #ifndef _SYS_SYSPROTO_H_
386 struct getfsstat_args {
387 	struct statfs *buf;
388 	long bufsize;
389 	int mode;
390 };
391 #endif
392 int
393 sys_getfsstat(td, uap)
394 	struct thread *td;
395 	register struct getfsstat_args /* {
396 		struct statfs *buf;
397 		long bufsize;
398 		int mode;
399 	} */ *uap;
400 {
401 	size_t count;
402 	int error;
403 
404 	if (uap->bufsize < 0 || uap->bufsize > SIZE_MAX)
405 		return (EINVAL);
406 	error = kern_getfsstat(td, &uap->buf, uap->bufsize, &count,
407 	    UIO_USERSPACE, uap->mode);
408 	if (error == 0)
409 		td->td_retval[0] = count;
410 	return (error);
411 }
412 
413 /*
414  * If (bufsize > 0 && bufseg == UIO_SYSSPACE)
415  *	The caller is responsible for freeing memory which will be allocated
416  *	in '*buf'.
417  */
418 int
419 kern_getfsstat(struct thread *td, struct statfs **buf, size_t bufsize,
420     size_t *countp, enum uio_seg bufseg, int mode)
421 {
422 	struct mount *mp, *nmp;
423 	struct statfs *sfsp, *sp, sb, *tofree;
424 	size_t count, maxcount;
425 	int error;
426 
427 	switch (mode) {
428 	case MNT_WAIT:
429 	case MNT_NOWAIT:
430 		break;
431 	default:
432 		return (EINVAL);
433 	}
434 restart:
435 	maxcount = bufsize / sizeof(struct statfs);
436 	if (bufsize == 0) {
437 		sfsp = NULL;
438 		tofree = NULL;
439 	} else if (bufseg == UIO_USERSPACE) {
440 		sfsp = *buf;
441 		tofree = NULL;
442 	} else /* if (bufseg == UIO_SYSSPACE) */ {
443 		count = 0;
444 		mtx_lock(&mountlist_mtx);
445 		TAILQ_FOREACH(mp, &mountlist, mnt_list) {
446 			count++;
447 		}
448 		mtx_unlock(&mountlist_mtx);
449 		if (maxcount > count)
450 			maxcount = count;
451 		tofree = sfsp = *buf = malloc(maxcount * sizeof(struct statfs),
452 		    M_TEMP, M_WAITOK);
453 	}
454 	count = 0;
455 	mtx_lock(&mountlist_mtx);
456 	for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
457 		if (prison_canseemount(td->td_ucred, mp) != 0) {
458 			nmp = TAILQ_NEXT(mp, mnt_list);
459 			continue;
460 		}
461 #ifdef MAC
462 		if (mac_mount_check_stat(td->td_ucred, mp) != 0) {
463 			nmp = TAILQ_NEXT(mp, mnt_list);
464 			continue;
465 		}
466 #endif
467 		if (mode == MNT_WAIT) {
468 			if (vfs_busy(mp, MBF_MNTLSTLOCK) != 0) {
469 				/*
470 				 * If vfs_busy() failed, and MBF_NOWAIT
471 				 * wasn't passed, then the mp is gone.
472 				 * Furthermore, because of MBF_MNTLSTLOCK,
473 				 * the mountlist_mtx was dropped.  We have
474 				 * no other choice than to start over.
475 				 */
476 				mtx_unlock(&mountlist_mtx);
477 				free(tofree, M_TEMP);
478 				goto restart;
479 			}
480 		} else {
481 			if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK) != 0) {
482 				nmp = TAILQ_NEXT(mp, mnt_list);
483 				continue;
484 			}
485 		}
486 		if (sfsp && count < maxcount) {
487 			sp = &mp->mnt_stat;
488 			/*
489 			 * Set these in case the underlying filesystem
490 			 * fails to do so.
491 			 */
492 			sp->f_version = STATFS_VERSION;
493 			sp->f_namemax = NAME_MAX;
494 			sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
495 			/*
496 			 * If MNT_NOWAIT is specified, do not refresh
497 			 * the fsstat cache.
498 			 */
499 			if (mode != MNT_NOWAIT) {
500 				error = VFS_STATFS(mp, sp);
501 				if (error != 0) {
502 					mtx_lock(&mountlist_mtx);
503 					nmp = TAILQ_NEXT(mp, mnt_list);
504 					vfs_unbusy(mp);
505 					continue;
506 				}
507 			}
508 			if (priv_check(td, PRIV_VFS_GENERATION)) {
509 				bcopy(sp, &sb, sizeof(sb));
510 				sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
511 				prison_enforce_statfs(td->td_ucred, mp, &sb);
512 				sp = &sb;
513 			}
514 			if (bufseg == UIO_SYSSPACE)
515 				bcopy(sp, sfsp, sizeof(*sp));
516 			else /* if (bufseg == UIO_USERSPACE) */ {
517 				error = copyout(sp, sfsp, sizeof(*sp));
518 				if (error != 0) {
519 					vfs_unbusy(mp);
520 					return (error);
521 				}
522 			}
523 			sfsp++;
524 		}
525 		count++;
526 		mtx_lock(&mountlist_mtx);
527 		nmp = TAILQ_NEXT(mp, mnt_list);
528 		vfs_unbusy(mp);
529 	}
530 	mtx_unlock(&mountlist_mtx);
531 	if (sfsp && count > maxcount)
532 		*countp = maxcount;
533 	else
534 		*countp = count;
535 	return (0);
536 }
537 
538 #ifdef COMPAT_FREEBSD4
539 /*
540  * Get old format filesystem statistics.
541  */
542 static void cvtstatfs(struct statfs *, struct ostatfs *);
543 
544 #ifndef _SYS_SYSPROTO_H_
545 struct freebsd4_statfs_args {
546 	char *path;
547 	struct ostatfs *buf;
548 };
549 #endif
550 int
551 freebsd4_statfs(td, uap)
552 	struct thread *td;
553 	struct freebsd4_statfs_args /* {
554 		char *path;
555 		struct ostatfs *buf;
556 	} */ *uap;
557 {
558 	struct ostatfs osb;
559 	struct statfs sf;
560 	int error;
561 
562 	error = kern_statfs(td, uap->path, UIO_USERSPACE, &sf);
563 	if (error != 0)
564 		return (error);
565 	cvtstatfs(&sf, &osb);
566 	return (copyout(&osb, uap->buf, sizeof(osb)));
567 }
568 
569 /*
570  * Get filesystem statistics.
571  */
572 #ifndef _SYS_SYSPROTO_H_
573 struct freebsd4_fstatfs_args {
574 	int fd;
575 	struct ostatfs *buf;
576 };
577 #endif
578 int
579 freebsd4_fstatfs(td, uap)
580 	struct thread *td;
581 	struct freebsd4_fstatfs_args /* {
582 		int fd;
583 		struct ostatfs *buf;
584 	} */ *uap;
585 {
586 	struct ostatfs osb;
587 	struct statfs sf;
588 	int error;
589 
590 	error = kern_fstatfs(td, uap->fd, &sf);
591 	if (error != 0)
592 		return (error);
593 	cvtstatfs(&sf, &osb);
594 	return (copyout(&osb, uap->buf, sizeof(osb)));
595 }
596 
597 /*
598  * Get statistics on all filesystems.
599  */
600 #ifndef _SYS_SYSPROTO_H_
601 struct freebsd4_getfsstat_args {
602 	struct ostatfs *buf;
603 	long bufsize;
604 	int mode;
605 };
606 #endif
607 int
608 freebsd4_getfsstat(td, uap)
609 	struct thread *td;
610 	register struct freebsd4_getfsstat_args /* {
611 		struct ostatfs *buf;
612 		long bufsize;
613 		int mode;
614 	} */ *uap;
615 {
616 	struct statfs *buf, *sp;
617 	struct ostatfs osb;
618 	size_t count, size;
619 	int error;
620 
621 	if (uap->bufsize < 0)
622 		return (EINVAL);
623 	count = uap->bufsize / sizeof(struct ostatfs);
624 	if (count > SIZE_MAX / sizeof(struct statfs))
625 		return (EINVAL);
626 	size = count * sizeof(struct statfs);
627 	error = kern_getfsstat(td, &buf, size, &count, UIO_SYSSPACE,
628 	    uap->mode);
629 	td->td_retval[0] = count;
630 	if (size != 0) {
631 		sp = buf;
632 		while (count != 0 && error == 0) {
633 			cvtstatfs(sp, &osb);
634 			error = copyout(&osb, uap->buf, sizeof(osb));
635 			sp++;
636 			uap->buf++;
637 			count--;
638 		}
639 		free(buf, M_TEMP);
640 	}
641 	return (error);
642 }
643 
644 /*
645  * Implement fstatfs() for (NFS) file handles.
646  */
647 #ifndef _SYS_SYSPROTO_H_
648 struct freebsd4_fhstatfs_args {
649 	struct fhandle *u_fhp;
650 	struct ostatfs *buf;
651 };
652 #endif
653 int
654 freebsd4_fhstatfs(td, uap)
655 	struct thread *td;
656 	struct freebsd4_fhstatfs_args /* {
657 		struct fhandle *u_fhp;
658 		struct ostatfs *buf;
659 	} */ *uap;
660 {
661 	struct ostatfs osb;
662 	struct statfs sf;
663 	fhandle_t fh;
664 	int error;
665 
666 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
667 	if (error != 0)
668 		return (error);
669 	error = kern_fhstatfs(td, fh, &sf);
670 	if (error != 0)
671 		return (error);
672 	cvtstatfs(&sf, &osb);
673 	return (copyout(&osb, uap->buf, sizeof(osb)));
674 }
675 
676 /*
677  * Convert a new format statfs structure to an old format statfs structure.
678  */
679 static void
680 cvtstatfs(nsp, osp)
681 	struct statfs *nsp;
682 	struct ostatfs *osp;
683 {
684 
685 	statfs_scale_blocks(nsp, LONG_MAX);
686 	bzero(osp, sizeof(*osp));
687 	osp->f_bsize = nsp->f_bsize;
688 	osp->f_iosize = MIN(nsp->f_iosize, LONG_MAX);
689 	osp->f_blocks = nsp->f_blocks;
690 	osp->f_bfree = nsp->f_bfree;
691 	osp->f_bavail = nsp->f_bavail;
692 	osp->f_files = MIN(nsp->f_files, LONG_MAX);
693 	osp->f_ffree = MIN(nsp->f_ffree, LONG_MAX);
694 	osp->f_owner = nsp->f_owner;
695 	osp->f_type = nsp->f_type;
696 	osp->f_flags = nsp->f_flags;
697 	osp->f_syncwrites = MIN(nsp->f_syncwrites, LONG_MAX);
698 	osp->f_asyncwrites = MIN(nsp->f_asyncwrites, LONG_MAX);
699 	osp->f_syncreads = MIN(nsp->f_syncreads, LONG_MAX);
700 	osp->f_asyncreads = MIN(nsp->f_asyncreads, LONG_MAX);
701 	strlcpy(osp->f_fstypename, nsp->f_fstypename,
702 	    MIN(MFSNAMELEN, OMFSNAMELEN));
703 	strlcpy(osp->f_mntonname, nsp->f_mntonname,
704 	    MIN(MNAMELEN, OMNAMELEN));
705 	strlcpy(osp->f_mntfromname, nsp->f_mntfromname,
706 	    MIN(MNAMELEN, OMNAMELEN));
707 	osp->f_fsid = nsp->f_fsid;
708 }
709 #endif /* COMPAT_FREEBSD4 */
710 
711 /*
712  * Change current working directory to a given file descriptor.
713  */
714 #ifndef _SYS_SYSPROTO_H_
715 struct fchdir_args {
716 	int	fd;
717 };
718 #endif
719 int
720 sys_fchdir(td, uap)
721 	struct thread *td;
722 	struct fchdir_args /* {
723 		int fd;
724 	} */ *uap;
725 {
726 	struct vnode *vp, *tdp;
727 	struct mount *mp;
728 	struct file *fp;
729 	cap_rights_t rights;
730 	int error;
731 
732 	AUDIT_ARG_FD(uap->fd);
733 	error = getvnode(td, uap->fd, cap_rights_init(&rights, CAP_FCHDIR),
734 	    &fp);
735 	if (error != 0)
736 		return (error);
737 	vp = fp->f_vnode;
738 	vrefact(vp);
739 	fdrop(fp, td);
740 	vn_lock(vp, LK_SHARED | LK_RETRY);
741 	AUDIT_ARG_VNODE1(vp);
742 	error = change_dir(vp, td);
743 	while (!error && (mp = vp->v_mountedhere) != NULL) {
744 		if (vfs_busy(mp, 0))
745 			continue;
746 		error = VFS_ROOT(mp, LK_SHARED, &tdp);
747 		vfs_unbusy(mp);
748 		if (error != 0)
749 			break;
750 		vput(vp);
751 		vp = tdp;
752 	}
753 	if (error != 0) {
754 		vput(vp);
755 		return (error);
756 	}
757 	VOP_UNLOCK(vp, 0);
758 	pwd_chdir(td, vp);
759 	return (0);
760 }
761 
762 /*
763  * Change current working directory (``.'').
764  */
765 #ifndef _SYS_SYSPROTO_H_
766 struct chdir_args {
767 	char	*path;
768 };
769 #endif
770 int
771 sys_chdir(td, uap)
772 	struct thread *td;
773 	struct chdir_args /* {
774 		char *path;
775 	} */ *uap;
776 {
777 
778 	return (kern_chdir(td, uap->path, UIO_USERSPACE));
779 }
780 
781 int
782 kern_chdir(struct thread *td, char *path, enum uio_seg pathseg)
783 {
784 	struct nameidata nd;
785 	int error;
786 
787 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
788 	    pathseg, path, td);
789 	if ((error = namei(&nd)) != 0)
790 		return (error);
791 	if ((error = change_dir(nd.ni_vp, td)) != 0) {
792 		vput(nd.ni_vp);
793 		NDFREE(&nd, NDF_ONLY_PNBUF);
794 		return (error);
795 	}
796 	VOP_UNLOCK(nd.ni_vp, 0);
797 	NDFREE(&nd, NDF_ONLY_PNBUF);
798 	pwd_chdir(td, nd.ni_vp);
799 	return (0);
800 }
801 
802 /*
803  * Change notion of root (``/'') directory.
804  */
805 #ifndef _SYS_SYSPROTO_H_
806 struct chroot_args {
807 	char	*path;
808 };
809 #endif
810 int
811 sys_chroot(td, uap)
812 	struct thread *td;
813 	struct chroot_args /* {
814 		char *path;
815 	} */ *uap;
816 {
817 	struct nameidata nd;
818 	int error;
819 
820 	error = priv_check(td, PRIV_VFS_CHROOT);
821 	if (error != 0)
822 		return (error);
823 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
824 	    UIO_USERSPACE, uap->path, td);
825 	error = namei(&nd);
826 	if (error != 0)
827 		goto error;
828 	error = change_dir(nd.ni_vp, td);
829 	if (error != 0)
830 		goto e_vunlock;
831 #ifdef MAC
832 	error = mac_vnode_check_chroot(td->td_ucred, nd.ni_vp);
833 	if (error != 0)
834 		goto e_vunlock;
835 #endif
836 	VOP_UNLOCK(nd.ni_vp, 0);
837 	error = pwd_chroot(td, nd.ni_vp);
838 	vrele(nd.ni_vp);
839 	NDFREE(&nd, NDF_ONLY_PNBUF);
840 	return (error);
841 e_vunlock:
842 	vput(nd.ni_vp);
843 error:
844 	NDFREE(&nd, NDF_ONLY_PNBUF);
845 	return (error);
846 }
847 
848 /*
849  * Common routine for chroot and chdir.  Callers must provide a locked vnode
850  * instance.
851  */
852 int
853 change_dir(vp, td)
854 	struct vnode *vp;
855 	struct thread *td;
856 {
857 #ifdef MAC
858 	int error;
859 #endif
860 
861 	ASSERT_VOP_LOCKED(vp, "change_dir(): vp not locked");
862 	if (vp->v_type != VDIR)
863 		return (ENOTDIR);
864 #ifdef MAC
865 	error = mac_vnode_check_chdir(td->td_ucred, vp);
866 	if (error != 0)
867 		return (error);
868 #endif
869 	return (VOP_ACCESS(vp, VEXEC, td->td_ucred, td));
870 }
871 
872 static __inline void
873 flags_to_rights(int flags, cap_rights_t *rightsp)
874 {
875 
876 	if (flags & O_EXEC) {
877 		cap_rights_set(rightsp, CAP_FEXECVE);
878 	} else {
879 		switch ((flags & O_ACCMODE)) {
880 		case O_RDONLY:
881 			cap_rights_set(rightsp, CAP_READ);
882 			break;
883 		case O_RDWR:
884 			cap_rights_set(rightsp, CAP_READ);
885 			/* FALLTHROUGH */
886 		case O_WRONLY:
887 			cap_rights_set(rightsp, CAP_WRITE);
888 			if (!(flags & (O_APPEND | O_TRUNC)))
889 				cap_rights_set(rightsp, CAP_SEEK);
890 			break;
891 		}
892 	}
893 
894 	if (flags & O_CREAT)
895 		cap_rights_set(rightsp, CAP_CREATE);
896 
897 	if (flags & O_TRUNC)
898 		cap_rights_set(rightsp, CAP_FTRUNCATE);
899 
900 	if (flags & (O_SYNC | O_FSYNC))
901 		cap_rights_set(rightsp, CAP_FSYNC);
902 
903 	if (flags & (O_EXLOCK | O_SHLOCK))
904 		cap_rights_set(rightsp, CAP_FLOCK);
905 }
906 
907 /*
908  * Check permissions, allocate an open file structure, and call the device
909  * open routine if any.
910  */
911 #ifndef _SYS_SYSPROTO_H_
912 struct open_args {
913 	char	*path;
914 	int	flags;
915 	int	mode;
916 };
917 #endif
918 int
919 sys_open(td, uap)
920 	struct thread *td;
921 	register struct open_args /* {
922 		char *path;
923 		int flags;
924 		int mode;
925 	} */ *uap;
926 {
927 
928 	return (kern_openat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
929 	    uap->flags, uap->mode));
930 }
931 
932 #ifndef _SYS_SYSPROTO_H_
933 struct openat_args {
934 	int	fd;
935 	char	*path;
936 	int	flag;
937 	int	mode;
938 };
939 #endif
940 int
941 sys_openat(struct thread *td, struct openat_args *uap)
942 {
943 
944 	AUDIT_ARG_FD(uap->fd);
945 	return (kern_openat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
946 	    uap->mode));
947 }
948 
949 int
950 kern_openat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
951     int flags, int mode)
952 {
953 	struct proc *p = td->td_proc;
954 	struct filedesc *fdp = p->p_fd;
955 	struct file *fp;
956 	struct vnode *vp;
957 	struct nameidata nd;
958 	cap_rights_t rights;
959 	int cmode, error, indx;
960 
961 	indx = -1;
962 
963 	AUDIT_ARG_FFLAGS(flags);
964 	AUDIT_ARG_MODE(mode);
965 	cap_rights_init(&rights, CAP_LOOKUP);
966 	flags_to_rights(flags, &rights);
967 	/*
968 	 * Only one of the O_EXEC, O_RDONLY, O_WRONLY and O_RDWR flags
969 	 * may be specified.
970 	 */
971 	if (flags & O_EXEC) {
972 		if (flags & O_ACCMODE)
973 			return (EINVAL);
974 	} else if ((flags & O_ACCMODE) == O_ACCMODE) {
975 		return (EINVAL);
976 	} else {
977 		flags = FFLAGS(flags);
978 	}
979 
980 	/*
981 	 * Allocate a file structure. The descriptor to reference it
982 	 * is allocated and set by finstall() below.
983 	 */
984 	error = falloc_noinstall(td, &fp);
985 	if (error != 0)
986 		return (error);
987 	/*
988 	 * An extra reference on `fp' has been held for us by
989 	 * falloc_noinstall().
990 	 */
991 	/* Set the flags early so the finit in devfs can pick them up. */
992 	fp->f_flag = flags & FMASK;
993 	cmode = ((mode & ~fdp->fd_cmask) & ALLPERMS) & ~S_ISTXT;
994 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, fd,
995 	    &rights, td);
996 	td->td_dupfd = -1;		/* XXX check for fdopen */
997 	error = vn_open(&nd, &flags, cmode, fp);
998 	if (error != 0) {
999 		/*
1000 		 * If the vn_open replaced the method vector, something
1001 		 * wonderous happened deep below and we just pass it up
1002 		 * pretending we know what we do.
1003 		 */
1004 		if (error == ENXIO && fp->f_ops != &badfileops)
1005 			goto success;
1006 
1007 		/*
1008 		 * Handle special fdopen() case. bleh.
1009 		 *
1010 		 * Don't do this for relative (capability) lookups; we don't
1011 		 * understand exactly what would happen, and we don't think
1012 		 * that it ever should.
1013 		 */
1014 		if ((nd.ni_lcf & NI_LCF_STRICTRELATIVE) == 0 &&
1015 		    (error == ENODEV || error == ENXIO) &&
1016 		    td->td_dupfd >= 0) {
1017 			error = dupfdopen(td, fdp, td->td_dupfd, flags, error,
1018 			    &indx);
1019 			if (error == 0)
1020 				goto success;
1021 		}
1022 
1023 		goto bad;
1024 	}
1025 	td->td_dupfd = 0;
1026 	NDFREE(&nd, NDF_ONLY_PNBUF);
1027 	vp = nd.ni_vp;
1028 
1029 	/*
1030 	 * Store the vnode, for any f_type. Typically, the vnode use
1031 	 * count is decremented by direct call to vn_closefile() for
1032 	 * files that switched type in the cdevsw fdopen() method.
1033 	 */
1034 	fp->f_vnode = vp;
1035 	/*
1036 	 * If the file wasn't claimed by devfs bind it to the normal
1037 	 * vnode operations here.
1038 	 */
1039 	if (fp->f_ops == &badfileops) {
1040 		KASSERT(vp->v_type != VFIFO, ("Unexpected fifo."));
1041 		fp->f_seqcount = 1;
1042 		finit(fp, (flags & FMASK) | (fp->f_flag & FHASLOCK),
1043 		    DTYPE_VNODE, vp, &vnops);
1044 	}
1045 
1046 	VOP_UNLOCK(vp, 0);
1047 	if (flags & O_TRUNC) {
1048 		error = fo_truncate(fp, 0, td->td_ucred, td);
1049 		if (error != 0)
1050 			goto bad;
1051 	}
1052 success:
1053 	/*
1054 	 * If we haven't already installed the FD (for dupfdopen), do so now.
1055 	 */
1056 	if (indx == -1) {
1057 		struct filecaps *fcaps;
1058 
1059 #ifdef CAPABILITIES
1060 		if ((nd.ni_lcf & NI_LCF_STRICTRELATIVE) != 0)
1061 			fcaps = &nd.ni_filecaps;
1062 		else
1063 #endif
1064 			fcaps = NULL;
1065 		error = finstall(td, fp, &indx, flags, fcaps);
1066 		/* On success finstall() consumes fcaps. */
1067 		if (error != 0) {
1068 			filecaps_free(&nd.ni_filecaps);
1069 			goto bad;
1070 		}
1071 	} else {
1072 		filecaps_free(&nd.ni_filecaps);
1073 	}
1074 
1075 	/*
1076 	 * Release our private reference, leaving the one associated with
1077 	 * the descriptor table intact.
1078 	 */
1079 	fdrop(fp, td);
1080 	td->td_retval[0] = indx;
1081 	return (0);
1082 bad:
1083 	KASSERT(indx == -1, ("indx=%d, should be -1", indx));
1084 	fdrop(fp, td);
1085 	return (error);
1086 }
1087 
1088 #ifdef COMPAT_43
1089 /*
1090  * Create a file.
1091  */
1092 #ifndef _SYS_SYSPROTO_H_
1093 struct ocreat_args {
1094 	char	*path;
1095 	int	mode;
1096 };
1097 #endif
1098 int
1099 ocreat(td, uap)
1100 	struct thread *td;
1101 	register struct ocreat_args /* {
1102 		char *path;
1103 		int mode;
1104 	} */ *uap;
1105 {
1106 
1107 	return (kern_openat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1108 	    O_WRONLY | O_CREAT | O_TRUNC, uap->mode));
1109 }
1110 #endif /* COMPAT_43 */
1111 
1112 /*
1113  * Create a special file.
1114  */
1115 #ifndef _SYS_SYSPROTO_H_
1116 struct mknod_args {
1117 	char	*path;
1118 	int	mode;
1119 	int	dev;
1120 };
1121 #endif
1122 int
1123 sys_mknod(td, uap)
1124 	struct thread *td;
1125 	register struct mknod_args /* {
1126 		char *path;
1127 		int mode;
1128 		int dev;
1129 	} */ *uap;
1130 {
1131 
1132 	return (kern_mknodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1133 	    uap->mode, uap->dev));
1134 }
1135 
1136 #ifndef _SYS_SYSPROTO_H_
1137 struct mknodat_args {
1138 	int	fd;
1139 	char	*path;
1140 	mode_t	mode;
1141 	dev_t	dev;
1142 };
1143 #endif
1144 int
1145 sys_mknodat(struct thread *td, struct mknodat_args *uap)
1146 {
1147 
1148 	return (kern_mknodat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode,
1149 	    uap->dev));
1150 }
1151 
1152 int
1153 kern_mknodat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1154     int mode, int dev)
1155 {
1156 	struct vnode *vp;
1157 	struct mount *mp;
1158 	struct vattr vattr;
1159 	struct nameidata nd;
1160 	cap_rights_t rights;
1161 	int error, whiteout = 0;
1162 
1163 	AUDIT_ARG_MODE(mode);
1164 	AUDIT_ARG_DEV(dev);
1165 	switch (mode & S_IFMT) {
1166 	case S_IFCHR:
1167 	case S_IFBLK:
1168 		error = priv_check(td, PRIV_VFS_MKNOD_DEV);
1169 		if (error == 0 && dev == VNOVAL)
1170 			error = EINVAL;
1171 		break;
1172 	case S_IFMT:
1173 		error = priv_check(td, PRIV_VFS_MKNOD_BAD);
1174 		break;
1175 	case S_IFWHT:
1176 		error = priv_check(td, PRIV_VFS_MKNOD_WHT);
1177 		break;
1178 	case S_IFIFO:
1179 		if (dev == 0)
1180 			return (kern_mkfifoat(td, fd, path, pathseg, mode));
1181 		/* FALLTHROUGH */
1182 	default:
1183 		error = EINVAL;
1184 		break;
1185 	}
1186 	if (error != 0)
1187 		return (error);
1188 restart:
1189 	bwillwrite();
1190 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
1191 	    NOCACHE, pathseg, path, fd, cap_rights_init(&rights, CAP_MKNODAT),
1192 	    td);
1193 	if ((error = namei(&nd)) != 0)
1194 		return (error);
1195 	vp = nd.ni_vp;
1196 	if (vp != NULL) {
1197 		NDFREE(&nd, NDF_ONLY_PNBUF);
1198 		if (vp == nd.ni_dvp)
1199 			vrele(nd.ni_dvp);
1200 		else
1201 			vput(nd.ni_dvp);
1202 		vrele(vp);
1203 		return (EEXIST);
1204 	} else {
1205 		VATTR_NULL(&vattr);
1206 		vattr.va_mode = (mode & ALLPERMS) &
1207 		    ~td->td_proc->p_fd->fd_cmask;
1208 		vattr.va_rdev = dev;
1209 		whiteout = 0;
1210 
1211 		switch (mode & S_IFMT) {
1212 		case S_IFMT:	/* used by badsect to flag bad sectors */
1213 			vattr.va_type = VBAD;
1214 			break;
1215 		case S_IFCHR:
1216 			vattr.va_type = VCHR;
1217 			break;
1218 		case S_IFBLK:
1219 			vattr.va_type = VBLK;
1220 			break;
1221 		case S_IFWHT:
1222 			whiteout = 1;
1223 			break;
1224 		default:
1225 			panic("kern_mknod: invalid mode");
1226 		}
1227 	}
1228 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1229 		NDFREE(&nd, NDF_ONLY_PNBUF);
1230 		vput(nd.ni_dvp);
1231 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1232 			return (error);
1233 		goto restart;
1234 	}
1235 #ifdef MAC
1236 	if (error == 0 && !whiteout)
1237 		error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp,
1238 		    &nd.ni_cnd, &vattr);
1239 #endif
1240 	if (error == 0) {
1241 		if (whiteout)
1242 			error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE);
1243 		else {
1244 			error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp,
1245 						&nd.ni_cnd, &vattr);
1246 			if (error == 0)
1247 				vput(nd.ni_vp);
1248 		}
1249 	}
1250 	NDFREE(&nd, NDF_ONLY_PNBUF);
1251 	vput(nd.ni_dvp);
1252 	vn_finished_write(mp);
1253 	return (error);
1254 }
1255 
1256 /*
1257  * Create a named pipe.
1258  */
1259 #ifndef _SYS_SYSPROTO_H_
1260 struct mkfifo_args {
1261 	char	*path;
1262 	int	mode;
1263 };
1264 #endif
1265 int
1266 sys_mkfifo(td, uap)
1267 	struct thread *td;
1268 	register struct mkfifo_args /* {
1269 		char *path;
1270 		int mode;
1271 	} */ *uap;
1272 {
1273 
1274 	return (kern_mkfifoat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1275 	    uap->mode));
1276 }
1277 
1278 #ifndef _SYS_SYSPROTO_H_
1279 struct mkfifoat_args {
1280 	int	fd;
1281 	char	*path;
1282 	mode_t	mode;
1283 };
1284 #endif
1285 int
1286 sys_mkfifoat(struct thread *td, struct mkfifoat_args *uap)
1287 {
1288 
1289 	return (kern_mkfifoat(td, uap->fd, uap->path, UIO_USERSPACE,
1290 	    uap->mode));
1291 }
1292 
1293 int
1294 kern_mkfifoat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1295     int mode)
1296 {
1297 	struct mount *mp;
1298 	struct vattr vattr;
1299 	struct nameidata nd;
1300 	cap_rights_t rights;
1301 	int error;
1302 
1303 	AUDIT_ARG_MODE(mode);
1304 restart:
1305 	bwillwrite();
1306 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
1307 	    NOCACHE, pathseg, path, fd, cap_rights_init(&rights, CAP_MKFIFOAT),
1308 	    td);
1309 	if ((error = namei(&nd)) != 0)
1310 		return (error);
1311 	if (nd.ni_vp != NULL) {
1312 		NDFREE(&nd, NDF_ONLY_PNBUF);
1313 		if (nd.ni_vp == nd.ni_dvp)
1314 			vrele(nd.ni_dvp);
1315 		else
1316 			vput(nd.ni_dvp);
1317 		vrele(nd.ni_vp);
1318 		return (EEXIST);
1319 	}
1320 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1321 		NDFREE(&nd, NDF_ONLY_PNBUF);
1322 		vput(nd.ni_dvp);
1323 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1324 			return (error);
1325 		goto restart;
1326 	}
1327 	VATTR_NULL(&vattr);
1328 	vattr.va_type = VFIFO;
1329 	vattr.va_mode = (mode & ALLPERMS) & ~td->td_proc->p_fd->fd_cmask;
1330 #ifdef MAC
1331 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1332 	    &vattr);
1333 	if (error != 0)
1334 		goto out;
1335 #endif
1336 	error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
1337 	if (error == 0)
1338 		vput(nd.ni_vp);
1339 #ifdef MAC
1340 out:
1341 #endif
1342 	vput(nd.ni_dvp);
1343 	vn_finished_write(mp);
1344 	NDFREE(&nd, NDF_ONLY_PNBUF);
1345 	return (error);
1346 }
1347 
1348 /*
1349  * Make a hard file link.
1350  */
1351 #ifndef _SYS_SYSPROTO_H_
1352 struct link_args {
1353 	char	*path;
1354 	char	*link;
1355 };
1356 #endif
1357 int
1358 sys_link(td, uap)
1359 	struct thread *td;
1360 	register struct link_args /* {
1361 		char *path;
1362 		char *link;
1363 	} */ *uap;
1364 {
1365 
1366 	return (kern_linkat(td, AT_FDCWD, AT_FDCWD, uap->path, uap->link,
1367 	    UIO_USERSPACE, FOLLOW));
1368 }
1369 
1370 #ifndef _SYS_SYSPROTO_H_
1371 struct linkat_args {
1372 	int	fd1;
1373 	char	*path1;
1374 	int	fd2;
1375 	char	*path2;
1376 	int	flag;
1377 };
1378 #endif
1379 int
1380 sys_linkat(struct thread *td, struct linkat_args *uap)
1381 {
1382 	int flag;
1383 
1384 	flag = uap->flag;
1385 	if (flag & ~AT_SYMLINK_FOLLOW)
1386 		return (EINVAL);
1387 
1388 	return (kern_linkat(td, uap->fd1, uap->fd2, uap->path1, uap->path2,
1389 	    UIO_USERSPACE, (flag & AT_SYMLINK_FOLLOW) ? FOLLOW : NOFOLLOW));
1390 }
1391 
1392 int hardlink_check_uid = 0;
1393 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_uid, CTLFLAG_RW,
1394     &hardlink_check_uid, 0,
1395     "Unprivileged processes cannot create hard links to files owned by other "
1396     "users");
1397 static int hardlink_check_gid = 0;
1398 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_gid, CTLFLAG_RW,
1399     &hardlink_check_gid, 0,
1400     "Unprivileged processes cannot create hard links to files owned by other "
1401     "groups");
1402 
1403 static int
1404 can_hardlink(struct vnode *vp, struct ucred *cred)
1405 {
1406 	struct vattr va;
1407 	int error;
1408 
1409 	if (!hardlink_check_uid && !hardlink_check_gid)
1410 		return (0);
1411 
1412 	error = VOP_GETATTR(vp, &va, cred);
1413 	if (error != 0)
1414 		return (error);
1415 
1416 	if (hardlink_check_uid && cred->cr_uid != va.va_uid) {
1417 		error = priv_check_cred(cred, PRIV_VFS_LINK, 0);
1418 		if (error != 0)
1419 			return (error);
1420 	}
1421 
1422 	if (hardlink_check_gid && !groupmember(va.va_gid, cred)) {
1423 		error = priv_check_cred(cred, PRIV_VFS_LINK, 0);
1424 		if (error != 0)
1425 			return (error);
1426 	}
1427 
1428 	return (0);
1429 }
1430 
1431 int
1432 kern_linkat(struct thread *td, int fd1, int fd2, char *path1, char *path2,
1433     enum uio_seg segflg, int follow)
1434 {
1435 	struct vnode *vp;
1436 	struct mount *mp;
1437 	struct nameidata nd;
1438 	cap_rights_t rights;
1439 	int error;
1440 
1441 again:
1442 	bwillwrite();
1443 	NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, segflg, path1, fd1,
1444 	    cap_rights_init(&rights, CAP_LINKAT_SOURCE), td);
1445 
1446 	if ((error = namei(&nd)) != 0)
1447 		return (error);
1448 	NDFREE(&nd, NDF_ONLY_PNBUF);
1449 	vp = nd.ni_vp;
1450 	if (vp->v_type == VDIR) {
1451 		vrele(vp);
1452 		return (EPERM);		/* POSIX */
1453 	}
1454 	NDINIT_ATRIGHTS(&nd, CREATE,
1455 	    LOCKPARENT | SAVENAME | AUDITVNODE2 | NOCACHE, segflg, path2, fd2,
1456 	    cap_rights_init(&rights, CAP_LINKAT_TARGET), td);
1457 	if ((error = namei(&nd)) == 0) {
1458 		if (nd.ni_vp != NULL) {
1459 			NDFREE(&nd, NDF_ONLY_PNBUF);
1460 			if (nd.ni_dvp == nd.ni_vp)
1461 				vrele(nd.ni_dvp);
1462 			else
1463 				vput(nd.ni_dvp);
1464 			vrele(nd.ni_vp);
1465 			vrele(vp);
1466 			return (EEXIST);
1467 		} else if (nd.ni_dvp->v_mount != vp->v_mount) {
1468 			/*
1469 			 * Cross-device link.  No need to recheck
1470 			 * vp->v_type, since it cannot change, except
1471 			 * to VBAD.
1472 			 */
1473 			NDFREE(&nd, NDF_ONLY_PNBUF);
1474 			vput(nd.ni_dvp);
1475 			vrele(vp);
1476 			return (EXDEV);
1477 		} else if ((error = vn_lock(vp, LK_EXCLUSIVE)) == 0) {
1478 			error = can_hardlink(vp, td->td_ucred);
1479 #ifdef MAC
1480 			if (error == 0)
1481 				error = mac_vnode_check_link(td->td_ucred,
1482 				    nd.ni_dvp, vp, &nd.ni_cnd);
1483 #endif
1484 			if (error != 0) {
1485 				vput(vp);
1486 				vput(nd.ni_dvp);
1487 				NDFREE(&nd, NDF_ONLY_PNBUF);
1488 				return (error);
1489 			}
1490 			error = vn_start_write(vp, &mp, V_NOWAIT);
1491 			if (error != 0) {
1492 				vput(vp);
1493 				vput(nd.ni_dvp);
1494 				NDFREE(&nd, NDF_ONLY_PNBUF);
1495 				error = vn_start_write(NULL, &mp,
1496 				    V_XSLEEP | PCATCH);
1497 				if (error != 0)
1498 					return (error);
1499 				goto again;
1500 			}
1501 			error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
1502 			VOP_UNLOCK(vp, 0);
1503 			vput(nd.ni_dvp);
1504 			vn_finished_write(mp);
1505 			NDFREE(&nd, NDF_ONLY_PNBUF);
1506 		} else {
1507 			vput(nd.ni_dvp);
1508 			NDFREE(&nd, NDF_ONLY_PNBUF);
1509 			vrele(vp);
1510 			goto again;
1511 		}
1512 	}
1513 	vrele(vp);
1514 	return (error);
1515 }
1516 
1517 /*
1518  * Make a symbolic link.
1519  */
1520 #ifndef _SYS_SYSPROTO_H_
1521 struct symlink_args {
1522 	char	*path;
1523 	char	*link;
1524 };
1525 #endif
1526 int
1527 sys_symlink(td, uap)
1528 	struct thread *td;
1529 	register struct symlink_args /* {
1530 		char *path;
1531 		char *link;
1532 	} */ *uap;
1533 {
1534 
1535 	return (kern_symlinkat(td, uap->path, AT_FDCWD, uap->link,
1536 	    UIO_USERSPACE));
1537 }
1538 
1539 #ifndef _SYS_SYSPROTO_H_
1540 struct symlinkat_args {
1541 	char	*path;
1542 	int	fd;
1543 	char	*path2;
1544 };
1545 #endif
1546 int
1547 sys_symlinkat(struct thread *td, struct symlinkat_args *uap)
1548 {
1549 
1550 	return (kern_symlinkat(td, uap->path1, uap->fd, uap->path2,
1551 	    UIO_USERSPACE));
1552 }
1553 
1554 int
1555 kern_symlinkat(struct thread *td, char *path1, int fd, char *path2,
1556     enum uio_seg segflg)
1557 {
1558 	struct mount *mp;
1559 	struct vattr vattr;
1560 	char *syspath;
1561 	struct nameidata nd;
1562 	int error;
1563 	cap_rights_t rights;
1564 
1565 	if (segflg == UIO_SYSSPACE) {
1566 		syspath = path1;
1567 	} else {
1568 		syspath = uma_zalloc(namei_zone, M_WAITOK);
1569 		if ((error = copyinstr(path1, syspath, MAXPATHLEN, NULL)) != 0)
1570 			goto out;
1571 	}
1572 	AUDIT_ARG_TEXT(syspath);
1573 restart:
1574 	bwillwrite();
1575 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
1576 	    NOCACHE, segflg, path2, fd, cap_rights_init(&rights, CAP_SYMLINKAT),
1577 	    td);
1578 	if ((error = namei(&nd)) != 0)
1579 		goto out;
1580 	if (nd.ni_vp) {
1581 		NDFREE(&nd, NDF_ONLY_PNBUF);
1582 		if (nd.ni_vp == nd.ni_dvp)
1583 			vrele(nd.ni_dvp);
1584 		else
1585 			vput(nd.ni_dvp);
1586 		vrele(nd.ni_vp);
1587 		error = EEXIST;
1588 		goto out;
1589 	}
1590 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1591 		NDFREE(&nd, NDF_ONLY_PNBUF);
1592 		vput(nd.ni_dvp);
1593 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1594 			goto out;
1595 		goto restart;
1596 	}
1597 	VATTR_NULL(&vattr);
1598 	vattr.va_mode = ACCESSPERMS &~ td->td_proc->p_fd->fd_cmask;
1599 #ifdef MAC
1600 	vattr.va_type = VLNK;
1601 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1602 	    &vattr);
1603 	if (error != 0)
1604 		goto out2;
1605 #endif
1606 	error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, syspath);
1607 	if (error == 0)
1608 		vput(nd.ni_vp);
1609 #ifdef MAC
1610 out2:
1611 #endif
1612 	NDFREE(&nd, NDF_ONLY_PNBUF);
1613 	vput(nd.ni_dvp);
1614 	vn_finished_write(mp);
1615 out:
1616 	if (segflg != UIO_SYSSPACE)
1617 		uma_zfree(namei_zone, syspath);
1618 	return (error);
1619 }
1620 
1621 /*
1622  * Delete a whiteout from the filesystem.
1623  */
1624 int
1625 sys_undelete(td, uap)
1626 	struct thread *td;
1627 	register struct undelete_args /* {
1628 		char *path;
1629 	} */ *uap;
1630 {
1631 	struct mount *mp;
1632 	struct nameidata nd;
1633 	int error;
1634 
1635 restart:
1636 	bwillwrite();
1637 	NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | AUDITVNODE1,
1638 	    UIO_USERSPACE, uap->path, td);
1639 	error = namei(&nd);
1640 	if (error != 0)
1641 		return (error);
1642 
1643 	if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) {
1644 		NDFREE(&nd, NDF_ONLY_PNBUF);
1645 		if (nd.ni_vp == nd.ni_dvp)
1646 			vrele(nd.ni_dvp);
1647 		else
1648 			vput(nd.ni_dvp);
1649 		if (nd.ni_vp)
1650 			vrele(nd.ni_vp);
1651 		return (EEXIST);
1652 	}
1653 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1654 		NDFREE(&nd, NDF_ONLY_PNBUF);
1655 		vput(nd.ni_dvp);
1656 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1657 			return (error);
1658 		goto restart;
1659 	}
1660 	error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE);
1661 	NDFREE(&nd, NDF_ONLY_PNBUF);
1662 	vput(nd.ni_dvp);
1663 	vn_finished_write(mp);
1664 	return (error);
1665 }
1666 
1667 /*
1668  * Delete a name from the filesystem.
1669  */
1670 #ifndef _SYS_SYSPROTO_H_
1671 struct unlink_args {
1672 	char	*path;
1673 };
1674 #endif
1675 int
1676 sys_unlink(td, uap)
1677 	struct thread *td;
1678 	struct unlink_args /* {
1679 		char *path;
1680 	} */ *uap;
1681 {
1682 
1683 	return (kern_unlinkat(td, AT_FDCWD, uap->path, UIO_USERSPACE, 0));
1684 }
1685 
1686 #ifndef _SYS_SYSPROTO_H_
1687 struct unlinkat_args {
1688 	int	fd;
1689 	char	*path;
1690 	int	flag;
1691 };
1692 #endif
1693 int
1694 sys_unlinkat(struct thread *td, struct unlinkat_args *uap)
1695 {
1696 	int flag = uap->flag;
1697 	int fd = uap->fd;
1698 	char *path = uap->path;
1699 
1700 	if (flag & ~AT_REMOVEDIR)
1701 		return (EINVAL);
1702 
1703 	if (flag & AT_REMOVEDIR)
1704 		return (kern_rmdirat(td, fd, path, UIO_USERSPACE));
1705 	else
1706 		return (kern_unlinkat(td, fd, path, UIO_USERSPACE, 0));
1707 }
1708 
1709 int
1710 kern_unlinkat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1711     ino_t oldinum)
1712 {
1713 	struct mount *mp;
1714 	struct vnode *vp;
1715 	struct nameidata nd;
1716 	struct stat sb;
1717 	cap_rights_t rights;
1718 	int error;
1719 
1720 restart:
1721 	bwillwrite();
1722 	NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1,
1723 	    pathseg, path, fd, cap_rights_init(&rights, CAP_UNLINKAT), td);
1724 	if ((error = namei(&nd)) != 0)
1725 		return (error == EINVAL ? EPERM : error);
1726 	vp = nd.ni_vp;
1727 	if (vp->v_type == VDIR && oldinum == 0) {
1728 		error = EPERM;		/* POSIX */
1729 	} else if (oldinum != 0 &&
1730 		  ((error = vn_stat(vp, &sb, td->td_ucred, NOCRED, td)) == 0) &&
1731 		  sb.st_ino != oldinum) {
1732 			error = EIDRM;	/* Identifier removed */
1733 	} else {
1734 		/*
1735 		 * The root of a mounted filesystem cannot be deleted.
1736 		 *
1737 		 * XXX: can this only be a VDIR case?
1738 		 */
1739 		if (vp->v_vflag & VV_ROOT)
1740 			error = EBUSY;
1741 	}
1742 	if (error == 0) {
1743 		if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1744 			NDFREE(&nd, NDF_ONLY_PNBUF);
1745 			vput(nd.ni_dvp);
1746 			if (vp == nd.ni_dvp)
1747 				vrele(vp);
1748 			else
1749 				vput(vp);
1750 			if ((error = vn_start_write(NULL, &mp,
1751 			    V_XSLEEP | PCATCH)) != 0)
1752 				return (error);
1753 			goto restart;
1754 		}
1755 #ifdef MAC
1756 		error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
1757 		    &nd.ni_cnd);
1758 		if (error != 0)
1759 			goto out;
1760 #endif
1761 		vfs_notify_upper(vp, VFS_NOTIFY_UPPER_UNLINK);
1762 		error = VOP_REMOVE(nd.ni_dvp, vp, &nd.ni_cnd);
1763 #ifdef MAC
1764 out:
1765 #endif
1766 		vn_finished_write(mp);
1767 	}
1768 	NDFREE(&nd, NDF_ONLY_PNBUF);
1769 	vput(nd.ni_dvp);
1770 	if (vp == nd.ni_dvp)
1771 		vrele(vp);
1772 	else
1773 		vput(vp);
1774 	return (error);
1775 }
1776 
1777 /*
1778  * Reposition read/write file offset.
1779  */
1780 #ifndef _SYS_SYSPROTO_H_
1781 struct lseek_args {
1782 	int	fd;
1783 	int	pad;
1784 	off_t	offset;
1785 	int	whence;
1786 };
1787 #endif
1788 int
1789 sys_lseek(td, uap)
1790 	struct thread *td;
1791 	register struct lseek_args /* {
1792 		int fd;
1793 		int pad;
1794 		off_t offset;
1795 		int whence;
1796 	} */ *uap;
1797 {
1798 	struct file *fp;
1799 	cap_rights_t rights;
1800 	int error;
1801 
1802 	AUDIT_ARG_FD(uap->fd);
1803 	error = fget(td, uap->fd, cap_rights_init(&rights, CAP_SEEK), &fp);
1804 	if (error != 0)
1805 		return (error);
1806 	error = (fp->f_ops->fo_flags & DFLAG_SEEKABLE) != 0 ?
1807 	    fo_seek(fp, uap->offset, uap->whence, td) : ESPIPE;
1808 	fdrop(fp, td);
1809 	return (error);
1810 }
1811 
1812 #if defined(COMPAT_43)
1813 /*
1814  * Reposition read/write file offset.
1815  */
1816 #ifndef _SYS_SYSPROTO_H_
1817 struct olseek_args {
1818 	int	fd;
1819 	long	offset;
1820 	int	whence;
1821 };
1822 #endif
1823 int
1824 olseek(td, uap)
1825 	struct thread *td;
1826 	register struct olseek_args /* {
1827 		int fd;
1828 		long offset;
1829 		int whence;
1830 	} */ *uap;
1831 {
1832 	struct lseek_args /* {
1833 		int fd;
1834 		int pad;
1835 		off_t offset;
1836 		int whence;
1837 	} */ nuap;
1838 
1839 	nuap.fd = uap->fd;
1840 	nuap.offset = uap->offset;
1841 	nuap.whence = uap->whence;
1842 	return (sys_lseek(td, &nuap));
1843 }
1844 #endif /* COMPAT_43 */
1845 
1846 #if defined(COMPAT_FREEBSD6)
1847 /* Version with the 'pad' argument */
1848 int
1849 freebsd6_lseek(td, uap)
1850 	struct thread *td;
1851 	register struct freebsd6_lseek_args *uap;
1852 {
1853 	struct lseek_args ouap;
1854 
1855 	ouap.fd = uap->fd;
1856 	ouap.offset = uap->offset;
1857 	ouap.whence = uap->whence;
1858 	return (sys_lseek(td, &ouap));
1859 }
1860 #endif
1861 
1862 /*
1863  * Check access permissions using passed credentials.
1864  */
1865 static int
1866 vn_access(vp, user_flags, cred, td)
1867 	struct vnode	*vp;
1868 	int		user_flags;
1869 	struct ucred	*cred;
1870 	struct thread	*td;
1871 {
1872 	accmode_t accmode;
1873 	int error;
1874 
1875 	/* Flags == 0 means only check for existence. */
1876 	if (user_flags == 0)
1877 		return (0);
1878 
1879 	accmode = 0;
1880 	if (user_flags & R_OK)
1881 		accmode |= VREAD;
1882 	if (user_flags & W_OK)
1883 		accmode |= VWRITE;
1884 	if (user_flags & X_OK)
1885 		accmode |= VEXEC;
1886 #ifdef MAC
1887 	error = mac_vnode_check_access(cred, vp, accmode);
1888 	if (error != 0)
1889 		return (error);
1890 #endif
1891 	if ((accmode & VWRITE) == 0 || (error = vn_writechk(vp)) == 0)
1892 		error = VOP_ACCESS(vp, accmode, cred, td);
1893 	return (error);
1894 }
1895 
1896 /*
1897  * Check access permissions using "real" credentials.
1898  */
1899 #ifndef _SYS_SYSPROTO_H_
1900 struct access_args {
1901 	char	*path;
1902 	int	amode;
1903 };
1904 #endif
1905 int
1906 sys_access(td, uap)
1907 	struct thread *td;
1908 	register struct access_args /* {
1909 		char *path;
1910 		int amode;
1911 	} */ *uap;
1912 {
1913 
1914 	return (kern_accessat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1915 	    0, uap->amode));
1916 }
1917 
1918 #ifndef _SYS_SYSPROTO_H_
1919 struct faccessat_args {
1920 	int	dirfd;
1921 	char	*path;
1922 	int	amode;
1923 	int	flag;
1924 }
1925 #endif
1926 int
1927 sys_faccessat(struct thread *td, struct faccessat_args *uap)
1928 {
1929 
1930 	return (kern_accessat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
1931 	    uap->amode));
1932 }
1933 
1934 int
1935 kern_accessat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1936     int flag, int amode)
1937 {
1938 	struct ucred *cred, *usecred;
1939 	struct vnode *vp;
1940 	struct nameidata nd;
1941 	cap_rights_t rights;
1942 	int error;
1943 
1944 	if (flag & ~AT_EACCESS)
1945 		return (EINVAL);
1946 	if (amode != F_OK && (amode & ~(R_OK | W_OK | X_OK)) != 0)
1947 		return (EINVAL);
1948 
1949 	/*
1950 	 * Create and modify a temporary credential instead of one that
1951 	 * is potentially shared (if we need one).
1952 	 */
1953 	cred = td->td_ucred;
1954 	if ((flag & AT_EACCESS) == 0 &&
1955 	    ((cred->cr_uid != cred->cr_ruid ||
1956 	    cred->cr_rgid != cred->cr_groups[0]))) {
1957 		usecred = crdup(cred);
1958 		usecred->cr_uid = cred->cr_ruid;
1959 		usecred->cr_groups[0] = cred->cr_rgid;
1960 		td->td_ucred = usecred;
1961 	} else
1962 		usecred = cred;
1963 	AUDIT_ARG_VALUE(amode);
1964 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF |
1965 	    AUDITVNODE1, pathseg, path, fd, cap_rights_init(&rights, CAP_FSTAT),
1966 	    td);
1967 	if ((error = namei(&nd)) != 0)
1968 		goto out;
1969 	vp = nd.ni_vp;
1970 
1971 	error = vn_access(vp, amode, usecred, td);
1972 	NDFREE(&nd, NDF_ONLY_PNBUF);
1973 	vput(vp);
1974 out:
1975 	if (usecred != cred) {
1976 		td->td_ucred = cred;
1977 		crfree(usecred);
1978 	}
1979 	return (error);
1980 }
1981 
1982 /*
1983  * Check access permissions using "effective" credentials.
1984  */
1985 #ifndef _SYS_SYSPROTO_H_
1986 struct eaccess_args {
1987 	char	*path;
1988 	int	amode;
1989 };
1990 #endif
1991 int
1992 sys_eaccess(td, uap)
1993 	struct thread *td;
1994 	register struct eaccess_args /* {
1995 		char *path;
1996 		int amode;
1997 	} */ *uap;
1998 {
1999 
2000 	return (kern_accessat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2001 	    AT_EACCESS, uap->amode));
2002 }
2003 
2004 #if defined(COMPAT_43)
2005 /*
2006  * Get file status; this version follows links.
2007  */
2008 #ifndef _SYS_SYSPROTO_H_
2009 struct ostat_args {
2010 	char	*path;
2011 	struct ostat *ub;
2012 };
2013 #endif
2014 int
2015 ostat(td, uap)
2016 	struct thread *td;
2017 	register struct ostat_args /* {
2018 		char *path;
2019 		struct ostat *ub;
2020 	} */ *uap;
2021 {
2022 	struct stat sb;
2023 	struct ostat osb;
2024 	int error;
2025 
2026 	error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE,
2027 	    &sb, NULL);
2028 	if (error != 0)
2029 		return (error);
2030 	cvtstat(&sb, &osb);
2031 	return (copyout(&osb, uap->ub, sizeof (osb)));
2032 }
2033 
2034 /*
2035  * Get file status; this version does not follow links.
2036  */
2037 #ifndef _SYS_SYSPROTO_H_
2038 struct olstat_args {
2039 	char	*path;
2040 	struct ostat *ub;
2041 };
2042 #endif
2043 int
2044 olstat(td, uap)
2045 	struct thread *td;
2046 	register struct olstat_args /* {
2047 		char *path;
2048 		struct ostat *ub;
2049 	} */ *uap;
2050 {
2051 	struct stat sb;
2052 	struct ostat osb;
2053 	int error;
2054 
2055 	error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2056 	    UIO_USERSPACE, &sb, NULL);
2057 	if (error != 0)
2058 		return (error);
2059 	cvtstat(&sb, &osb);
2060 	return (copyout(&osb, uap->ub, sizeof (osb)));
2061 }
2062 
2063 /*
2064  * Convert from an old to a new stat structure.
2065  */
2066 void
2067 cvtstat(st, ost)
2068 	struct stat *st;
2069 	struct ostat *ost;
2070 {
2071 
2072 	bzero(ost, sizeof(*ost));
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 	error = VOP_PATHCONF(nd.ni_vp, name, td->td_retval);
2345 	vput(nd.ni_vp);
2346 	return (error);
2347 }
2348 
2349 /*
2350  * Return target name of a symbolic link.
2351  */
2352 #ifndef _SYS_SYSPROTO_H_
2353 struct readlink_args {
2354 	char	*path;
2355 	char	*buf;
2356 	size_t	count;
2357 };
2358 #endif
2359 int
2360 sys_readlink(td, uap)
2361 	struct thread *td;
2362 	register struct readlink_args /* {
2363 		char *path;
2364 		char *buf;
2365 		size_t count;
2366 	} */ *uap;
2367 {
2368 
2369 	return (kern_readlinkat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2370 	    uap->buf, UIO_USERSPACE, uap->count));
2371 }
2372 #ifndef _SYS_SYSPROTO_H_
2373 struct readlinkat_args {
2374 	int	fd;
2375 	char	*path;
2376 	char	*buf;
2377 	size_t	bufsize;
2378 };
2379 #endif
2380 int
2381 sys_readlinkat(struct thread *td, struct readlinkat_args *uap)
2382 {
2383 
2384 	return (kern_readlinkat(td, uap->fd, uap->path, UIO_USERSPACE,
2385 	    uap->buf, UIO_USERSPACE, uap->bufsize));
2386 }
2387 
2388 int
2389 kern_readlinkat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2390     char *buf, enum uio_seg bufseg, size_t count)
2391 {
2392 	struct vnode *vp;
2393 	struct iovec aiov;
2394 	struct uio auio;
2395 	struct nameidata nd;
2396 	int error;
2397 
2398 	if (count > IOSIZE_MAX)
2399 		return (EINVAL);
2400 
2401 	NDINIT_AT(&nd, LOOKUP, NOFOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
2402 	    pathseg, path, fd, td);
2403 
2404 	if ((error = namei(&nd)) != 0)
2405 		return (error);
2406 	NDFREE(&nd, NDF_ONLY_PNBUF);
2407 	vp = nd.ni_vp;
2408 #ifdef MAC
2409 	error = mac_vnode_check_readlink(td->td_ucred, vp);
2410 	if (error != 0) {
2411 		vput(vp);
2412 		return (error);
2413 	}
2414 #endif
2415 	if (vp->v_type != VLNK)
2416 		error = EINVAL;
2417 	else {
2418 		aiov.iov_base = buf;
2419 		aiov.iov_len = count;
2420 		auio.uio_iov = &aiov;
2421 		auio.uio_iovcnt = 1;
2422 		auio.uio_offset = 0;
2423 		auio.uio_rw = UIO_READ;
2424 		auio.uio_segflg = bufseg;
2425 		auio.uio_td = td;
2426 		auio.uio_resid = count;
2427 		error = VOP_READLINK(vp, &auio, td->td_ucred);
2428 		td->td_retval[0] = count - auio.uio_resid;
2429 	}
2430 	vput(vp);
2431 	return (error);
2432 }
2433 
2434 /*
2435  * Common implementation code for chflags() and fchflags().
2436  */
2437 static int
2438 setfflags(td, vp, flags)
2439 	struct thread *td;
2440 	struct vnode *vp;
2441 	u_long flags;
2442 {
2443 	struct mount *mp;
2444 	struct vattr vattr;
2445 	int error;
2446 
2447 	/* We can't support the value matching VNOVAL. */
2448 	if (flags == VNOVAL)
2449 		return (EOPNOTSUPP);
2450 
2451 	/*
2452 	 * Prevent non-root users from setting flags on devices.  When
2453 	 * a device is reused, users can retain ownership of the device
2454 	 * if they are allowed to set flags and programs assume that
2455 	 * chown can't fail when done as root.
2456 	 */
2457 	if (vp->v_type == VCHR || vp->v_type == VBLK) {
2458 		error = priv_check(td, PRIV_VFS_CHFLAGS_DEV);
2459 		if (error != 0)
2460 			return (error);
2461 	}
2462 
2463 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2464 		return (error);
2465 	VATTR_NULL(&vattr);
2466 	vattr.va_flags = flags;
2467 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2468 #ifdef MAC
2469 	error = mac_vnode_check_setflags(td->td_ucred, vp, vattr.va_flags);
2470 	if (error == 0)
2471 #endif
2472 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
2473 	VOP_UNLOCK(vp, 0);
2474 	vn_finished_write(mp);
2475 	return (error);
2476 }
2477 
2478 /*
2479  * Change flags of a file given a path name.
2480  */
2481 #ifndef _SYS_SYSPROTO_H_
2482 struct chflags_args {
2483 	const char *path;
2484 	u_long	flags;
2485 };
2486 #endif
2487 int
2488 sys_chflags(td, uap)
2489 	struct thread *td;
2490 	register struct chflags_args /* {
2491 		const char *path;
2492 		u_long flags;
2493 	} */ *uap;
2494 {
2495 
2496 	return (kern_chflagsat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2497 	    uap->flags, 0));
2498 }
2499 
2500 #ifndef _SYS_SYSPROTO_H_
2501 struct chflagsat_args {
2502 	int	fd;
2503 	const char *path;
2504 	u_long	flags;
2505 	int	atflag;
2506 }
2507 #endif
2508 int
2509 sys_chflagsat(struct thread *td, struct chflagsat_args *uap)
2510 {
2511 	int fd = uap->fd;
2512 	const char *path = uap->path;
2513 	u_long flags = uap->flags;
2514 	int atflag = uap->atflag;
2515 
2516 	if (atflag & ~AT_SYMLINK_NOFOLLOW)
2517 		return (EINVAL);
2518 
2519 	return (kern_chflagsat(td, fd, path, UIO_USERSPACE, flags, atflag));
2520 }
2521 
2522 /*
2523  * Same as chflags() but doesn't follow symlinks.
2524  */
2525 int
2526 sys_lchflags(td, uap)
2527 	struct thread *td;
2528 	register struct lchflags_args /* {
2529 		const char *path;
2530 		u_long flags;
2531 	} */ *uap;
2532 {
2533 
2534 	return (kern_chflagsat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2535 	    uap->flags, AT_SYMLINK_NOFOLLOW));
2536 }
2537 
2538 static int
2539 kern_chflagsat(struct thread *td, int fd, const char *path,
2540     enum uio_seg pathseg, u_long flags, int atflag)
2541 {
2542 	struct nameidata nd;
2543 	cap_rights_t rights;
2544 	int error, follow;
2545 
2546 	AUDIT_ARG_FFLAGS(flags);
2547 	follow = (atflag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
2548 	NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, pathseg, path, fd,
2549 	    cap_rights_init(&rights, CAP_FCHFLAGS), td);
2550 	if ((error = namei(&nd)) != 0)
2551 		return (error);
2552 	NDFREE(&nd, NDF_ONLY_PNBUF);
2553 	error = setfflags(td, nd.ni_vp, flags);
2554 	vrele(nd.ni_vp);
2555 	return (error);
2556 }
2557 
2558 /*
2559  * Change flags of a file given a file descriptor.
2560  */
2561 #ifndef _SYS_SYSPROTO_H_
2562 struct fchflags_args {
2563 	int	fd;
2564 	u_long	flags;
2565 };
2566 #endif
2567 int
2568 sys_fchflags(td, uap)
2569 	struct thread *td;
2570 	register struct fchflags_args /* {
2571 		int fd;
2572 		u_long flags;
2573 	} */ *uap;
2574 {
2575 	struct file *fp;
2576 	cap_rights_t rights;
2577 	int error;
2578 
2579 	AUDIT_ARG_FD(uap->fd);
2580 	AUDIT_ARG_FFLAGS(uap->flags);
2581 	error = getvnode(td, uap->fd, cap_rights_init(&rights, CAP_FCHFLAGS),
2582 	    &fp);
2583 	if (error != 0)
2584 		return (error);
2585 #ifdef AUDIT
2586 	vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
2587 	AUDIT_ARG_VNODE1(fp->f_vnode);
2588 	VOP_UNLOCK(fp->f_vnode, 0);
2589 #endif
2590 	error = setfflags(td, fp->f_vnode, uap->flags);
2591 	fdrop(fp, td);
2592 	return (error);
2593 }
2594 
2595 /*
2596  * Common implementation code for chmod(), lchmod() and fchmod().
2597  */
2598 int
2599 setfmode(td, cred, vp, mode)
2600 	struct thread *td;
2601 	struct ucred *cred;
2602 	struct vnode *vp;
2603 	int mode;
2604 {
2605 	struct mount *mp;
2606 	struct vattr vattr;
2607 	int error;
2608 
2609 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2610 		return (error);
2611 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2612 	VATTR_NULL(&vattr);
2613 	vattr.va_mode = mode & ALLPERMS;
2614 #ifdef MAC
2615 	error = mac_vnode_check_setmode(cred, vp, vattr.va_mode);
2616 	if (error == 0)
2617 #endif
2618 		error = VOP_SETATTR(vp, &vattr, cred);
2619 	VOP_UNLOCK(vp, 0);
2620 	vn_finished_write(mp);
2621 	return (error);
2622 }
2623 
2624 /*
2625  * Change mode of a file given path name.
2626  */
2627 #ifndef _SYS_SYSPROTO_H_
2628 struct chmod_args {
2629 	char	*path;
2630 	int	mode;
2631 };
2632 #endif
2633 int
2634 sys_chmod(td, uap)
2635 	struct thread *td;
2636 	register struct chmod_args /* {
2637 		char *path;
2638 		int mode;
2639 	} */ *uap;
2640 {
2641 
2642 	return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2643 	    uap->mode, 0));
2644 }
2645 
2646 #ifndef _SYS_SYSPROTO_H_
2647 struct fchmodat_args {
2648 	int	dirfd;
2649 	char	*path;
2650 	mode_t	mode;
2651 	int	flag;
2652 }
2653 #endif
2654 int
2655 sys_fchmodat(struct thread *td, struct fchmodat_args *uap)
2656 {
2657 	int flag = uap->flag;
2658 	int fd = uap->fd;
2659 	char *path = uap->path;
2660 	mode_t mode = uap->mode;
2661 
2662 	if (flag & ~AT_SYMLINK_NOFOLLOW)
2663 		return (EINVAL);
2664 
2665 	return (kern_fchmodat(td, fd, path, UIO_USERSPACE, mode, flag));
2666 }
2667 
2668 /*
2669  * Change mode of a file given path name (don't follow links.)
2670  */
2671 #ifndef _SYS_SYSPROTO_H_
2672 struct lchmod_args {
2673 	char	*path;
2674 	int	mode;
2675 };
2676 #endif
2677 int
2678 sys_lchmod(td, uap)
2679 	struct thread *td;
2680 	register struct lchmod_args /* {
2681 		char *path;
2682 		int mode;
2683 	} */ *uap;
2684 {
2685 
2686 	return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2687 	    uap->mode, AT_SYMLINK_NOFOLLOW));
2688 }
2689 
2690 int
2691 kern_fchmodat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2692     mode_t mode, int flag)
2693 {
2694 	struct nameidata nd;
2695 	cap_rights_t rights;
2696 	int error, follow;
2697 
2698 	AUDIT_ARG_MODE(mode);
2699 	follow = (flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
2700 	NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, pathseg, path, fd,
2701 	    cap_rights_init(&rights, CAP_FCHMOD), td);
2702 	if ((error = namei(&nd)) != 0)
2703 		return (error);
2704 	NDFREE(&nd, NDF_ONLY_PNBUF);
2705 	error = setfmode(td, td->td_ucred, nd.ni_vp, mode);
2706 	vrele(nd.ni_vp);
2707 	return (error);
2708 }
2709 
2710 /*
2711  * Change mode of a file given a file descriptor.
2712  */
2713 #ifndef _SYS_SYSPROTO_H_
2714 struct fchmod_args {
2715 	int	fd;
2716 	int	mode;
2717 };
2718 #endif
2719 int
2720 sys_fchmod(struct thread *td, struct fchmod_args *uap)
2721 {
2722 	struct file *fp;
2723 	cap_rights_t rights;
2724 	int error;
2725 
2726 	AUDIT_ARG_FD(uap->fd);
2727 	AUDIT_ARG_MODE(uap->mode);
2728 
2729 	error = fget(td, uap->fd, cap_rights_init(&rights, CAP_FCHMOD), &fp);
2730 	if (error != 0)
2731 		return (error);
2732 	error = fo_chmod(fp, uap->mode, td->td_ucred, td);
2733 	fdrop(fp, td);
2734 	return (error);
2735 }
2736 
2737 /*
2738  * Common implementation for chown(), lchown(), and fchown()
2739  */
2740 int
2741 setfown(td, cred, vp, uid, gid)
2742 	struct thread *td;
2743 	struct ucred *cred;
2744 	struct vnode *vp;
2745 	uid_t uid;
2746 	gid_t gid;
2747 {
2748 	struct mount *mp;
2749 	struct vattr vattr;
2750 	int error;
2751 
2752 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2753 		return (error);
2754 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2755 	VATTR_NULL(&vattr);
2756 	vattr.va_uid = uid;
2757 	vattr.va_gid = gid;
2758 #ifdef MAC
2759 	error = mac_vnode_check_setowner(cred, vp, vattr.va_uid,
2760 	    vattr.va_gid);
2761 	if (error == 0)
2762 #endif
2763 		error = VOP_SETATTR(vp, &vattr, cred);
2764 	VOP_UNLOCK(vp, 0);
2765 	vn_finished_write(mp);
2766 	return (error);
2767 }
2768 
2769 /*
2770  * Set ownership given a path name.
2771  */
2772 #ifndef _SYS_SYSPROTO_H_
2773 struct chown_args {
2774 	char	*path;
2775 	int	uid;
2776 	int	gid;
2777 };
2778 #endif
2779 int
2780 sys_chown(td, uap)
2781 	struct thread *td;
2782 	register struct chown_args /* {
2783 		char *path;
2784 		int uid;
2785 		int gid;
2786 	} */ *uap;
2787 {
2788 
2789 	return (kern_fchownat(td, AT_FDCWD, uap->path, UIO_USERSPACE, uap->uid,
2790 	    uap->gid, 0));
2791 }
2792 
2793 #ifndef _SYS_SYSPROTO_H_
2794 struct fchownat_args {
2795 	int fd;
2796 	const char * path;
2797 	uid_t uid;
2798 	gid_t gid;
2799 	int flag;
2800 };
2801 #endif
2802 int
2803 sys_fchownat(struct thread *td, struct fchownat_args *uap)
2804 {
2805 	int flag;
2806 
2807 	flag = uap->flag;
2808 	if (flag & ~AT_SYMLINK_NOFOLLOW)
2809 		return (EINVAL);
2810 
2811 	return (kern_fchownat(td, uap->fd, uap->path, UIO_USERSPACE, uap->uid,
2812 	    uap->gid, uap->flag));
2813 }
2814 
2815 int
2816 kern_fchownat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2817     int uid, int gid, int flag)
2818 {
2819 	struct nameidata nd;
2820 	cap_rights_t rights;
2821 	int error, follow;
2822 
2823 	AUDIT_ARG_OWNER(uid, gid);
2824 	follow = (flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
2825 	NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, pathseg, path, fd,
2826 	    cap_rights_init(&rights, CAP_FCHOWN), td);
2827 
2828 	if ((error = namei(&nd)) != 0)
2829 		return (error);
2830 	NDFREE(&nd, NDF_ONLY_PNBUF);
2831 	error = setfown(td, td->td_ucred, nd.ni_vp, uid, gid);
2832 	vrele(nd.ni_vp);
2833 	return (error);
2834 }
2835 
2836 /*
2837  * Set ownership given a path name, do not cross symlinks.
2838  */
2839 #ifndef _SYS_SYSPROTO_H_
2840 struct lchown_args {
2841 	char	*path;
2842 	int	uid;
2843 	int	gid;
2844 };
2845 #endif
2846 int
2847 sys_lchown(td, uap)
2848 	struct thread *td;
2849 	register struct lchown_args /* {
2850 		char *path;
2851 		int uid;
2852 		int gid;
2853 	} */ *uap;
2854 {
2855 
2856 	return (kern_fchownat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2857 	    uap->uid, uap->gid, AT_SYMLINK_NOFOLLOW));
2858 }
2859 
2860 /*
2861  * Set ownership given a file descriptor.
2862  */
2863 #ifndef _SYS_SYSPROTO_H_
2864 struct fchown_args {
2865 	int	fd;
2866 	int	uid;
2867 	int	gid;
2868 };
2869 #endif
2870 int
2871 sys_fchown(td, uap)
2872 	struct thread *td;
2873 	register struct fchown_args /* {
2874 		int fd;
2875 		int uid;
2876 		int gid;
2877 	} */ *uap;
2878 {
2879 	struct file *fp;
2880 	cap_rights_t rights;
2881 	int error;
2882 
2883 	AUDIT_ARG_FD(uap->fd);
2884 	AUDIT_ARG_OWNER(uap->uid, uap->gid);
2885 	error = fget(td, uap->fd, cap_rights_init(&rights, CAP_FCHOWN), &fp);
2886 	if (error != 0)
2887 		return (error);
2888 	error = fo_chown(fp, uap->uid, uap->gid, td->td_ucred, td);
2889 	fdrop(fp, td);
2890 	return (error);
2891 }
2892 
2893 /*
2894  * Common implementation code for utimes(), lutimes(), and futimes().
2895  */
2896 static int
2897 getutimes(usrtvp, tvpseg, tsp)
2898 	const struct timeval *usrtvp;
2899 	enum uio_seg tvpseg;
2900 	struct timespec *tsp;
2901 {
2902 	struct timeval tv[2];
2903 	const struct timeval *tvp;
2904 	int error;
2905 
2906 	if (usrtvp == NULL) {
2907 		vfs_timestamp(&tsp[0]);
2908 		tsp[1] = tsp[0];
2909 	} else {
2910 		if (tvpseg == UIO_SYSSPACE) {
2911 			tvp = usrtvp;
2912 		} else {
2913 			if ((error = copyin(usrtvp, tv, sizeof(tv))) != 0)
2914 				return (error);
2915 			tvp = tv;
2916 		}
2917 
2918 		if (tvp[0].tv_usec < 0 || tvp[0].tv_usec >= 1000000 ||
2919 		    tvp[1].tv_usec < 0 || tvp[1].tv_usec >= 1000000)
2920 			return (EINVAL);
2921 		TIMEVAL_TO_TIMESPEC(&tvp[0], &tsp[0]);
2922 		TIMEVAL_TO_TIMESPEC(&tvp[1], &tsp[1]);
2923 	}
2924 	return (0);
2925 }
2926 
2927 /*
2928  * Common implementation code for futimens(), utimensat().
2929  */
2930 #define	UTIMENS_NULL	0x1
2931 #define	UTIMENS_EXIT	0x2
2932 static int
2933 getutimens(const struct timespec *usrtsp, enum uio_seg tspseg,
2934     struct timespec *tsp, int *retflags)
2935 {
2936 	struct timespec tsnow;
2937 	int error;
2938 
2939 	vfs_timestamp(&tsnow);
2940 	*retflags = 0;
2941 	if (usrtsp == NULL) {
2942 		tsp[0] = tsnow;
2943 		tsp[1] = tsnow;
2944 		*retflags |= UTIMENS_NULL;
2945 		return (0);
2946 	}
2947 	if (tspseg == UIO_SYSSPACE) {
2948 		tsp[0] = usrtsp[0];
2949 		tsp[1] = usrtsp[1];
2950 	} else if ((error = copyin(usrtsp, tsp, sizeof(*tsp) * 2)) != 0)
2951 		return (error);
2952 	if (tsp[0].tv_nsec == UTIME_OMIT && tsp[1].tv_nsec == UTIME_OMIT)
2953 		*retflags |= UTIMENS_EXIT;
2954 	if (tsp[0].tv_nsec == UTIME_NOW && tsp[1].tv_nsec == UTIME_NOW)
2955 		*retflags |= UTIMENS_NULL;
2956 	if (tsp[0].tv_nsec == UTIME_OMIT)
2957 		tsp[0].tv_sec = VNOVAL;
2958 	else if (tsp[0].tv_nsec == UTIME_NOW)
2959 		tsp[0] = tsnow;
2960 	else if (tsp[0].tv_nsec < 0 || tsp[0].tv_nsec >= 1000000000L)
2961 		return (EINVAL);
2962 	if (tsp[1].tv_nsec == UTIME_OMIT)
2963 		tsp[1].tv_sec = VNOVAL;
2964 	else if (tsp[1].tv_nsec == UTIME_NOW)
2965 		tsp[1] = tsnow;
2966 	else if (tsp[1].tv_nsec < 0 || tsp[1].tv_nsec >= 1000000000L)
2967 		return (EINVAL);
2968 
2969 	return (0);
2970 }
2971 
2972 /*
2973  * Common implementation code for utimes(), lutimes(), futimes(), futimens(),
2974  * and utimensat().
2975  */
2976 static int
2977 setutimes(td, vp, ts, numtimes, nullflag)
2978 	struct thread *td;
2979 	struct vnode *vp;
2980 	const struct timespec *ts;
2981 	int numtimes;
2982 	int nullflag;
2983 {
2984 	struct mount *mp;
2985 	struct vattr vattr;
2986 	int error, setbirthtime;
2987 
2988 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2989 		return (error);
2990 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2991 	setbirthtime = 0;
2992 	if (numtimes < 3 && !VOP_GETATTR(vp, &vattr, td->td_ucred) &&
2993 	    timespeccmp(&ts[1], &vattr.va_birthtime, < ))
2994 		setbirthtime = 1;
2995 	VATTR_NULL(&vattr);
2996 	vattr.va_atime = ts[0];
2997 	vattr.va_mtime = ts[1];
2998 	if (setbirthtime)
2999 		vattr.va_birthtime = ts[1];
3000 	if (numtimes > 2)
3001 		vattr.va_birthtime = ts[2];
3002 	if (nullflag)
3003 		vattr.va_vaflags |= VA_UTIMES_NULL;
3004 #ifdef MAC
3005 	error = mac_vnode_check_setutimes(td->td_ucred, vp, vattr.va_atime,
3006 	    vattr.va_mtime);
3007 #endif
3008 	if (error == 0)
3009 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3010 	VOP_UNLOCK(vp, 0);
3011 	vn_finished_write(mp);
3012 	return (error);
3013 }
3014 
3015 /*
3016  * Set the access and modification times of a file.
3017  */
3018 #ifndef _SYS_SYSPROTO_H_
3019 struct utimes_args {
3020 	char	*path;
3021 	struct	timeval *tptr;
3022 };
3023 #endif
3024 int
3025 sys_utimes(td, uap)
3026 	struct thread *td;
3027 	register struct utimes_args /* {
3028 		char *path;
3029 		struct timeval *tptr;
3030 	} */ *uap;
3031 {
3032 
3033 	return (kern_utimesat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
3034 	    uap->tptr, UIO_USERSPACE));
3035 }
3036 
3037 #ifndef _SYS_SYSPROTO_H_
3038 struct futimesat_args {
3039 	int fd;
3040 	const char * path;
3041 	const struct timeval * times;
3042 };
3043 #endif
3044 int
3045 sys_futimesat(struct thread *td, struct futimesat_args *uap)
3046 {
3047 
3048 	return (kern_utimesat(td, uap->fd, uap->path, UIO_USERSPACE,
3049 	    uap->times, UIO_USERSPACE));
3050 }
3051 
3052 int
3053 kern_utimesat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
3054     struct timeval *tptr, enum uio_seg tptrseg)
3055 {
3056 	struct nameidata nd;
3057 	struct timespec ts[2];
3058 	cap_rights_t rights;
3059 	int error;
3060 
3061 	if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3062 		return (error);
3063 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, fd,
3064 	    cap_rights_init(&rights, CAP_FUTIMES), td);
3065 
3066 	if ((error = namei(&nd)) != 0)
3067 		return (error);
3068 	NDFREE(&nd, NDF_ONLY_PNBUF);
3069 	error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3070 	vrele(nd.ni_vp);
3071 	return (error);
3072 }
3073 
3074 /*
3075  * Set the access and modification times of a file.
3076  */
3077 #ifndef _SYS_SYSPROTO_H_
3078 struct lutimes_args {
3079 	char	*path;
3080 	struct	timeval *tptr;
3081 };
3082 #endif
3083 int
3084 sys_lutimes(td, uap)
3085 	struct thread *td;
3086 	register struct lutimes_args /* {
3087 		char *path;
3088 		struct timeval *tptr;
3089 	} */ *uap;
3090 {
3091 
3092 	return (kern_lutimes(td, uap->path, UIO_USERSPACE, uap->tptr,
3093 	    UIO_USERSPACE));
3094 }
3095 
3096 int
3097 kern_lutimes(struct thread *td, char *path, enum uio_seg pathseg,
3098     struct timeval *tptr, enum uio_seg tptrseg)
3099 {
3100 	struct timespec ts[2];
3101 	struct nameidata nd;
3102 	int error;
3103 
3104 	if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3105 		return (error);
3106 	NDINIT(&nd, LOOKUP, NOFOLLOW | AUDITVNODE1, pathseg, path, td);
3107 	if ((error = namei(&nd)) != 0)
3108 		return (error);
3109 	NDFREE(&nd, NDF_ONLY_PNBUF);
3110 	error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3111 	vrele(nd.ni_vp);
3112 	return (error);
3113 }
3114 
3115 /*
3116  * Set the access and modification times of a file.
3117  */
3118 #ifndef _SYS_SYSPROTO_H_
3119 struct futimes_args {
3120 	int	fd;
3121 	struct	timeval *tptr;
3122 };
3123 #endif
3124 int
3125 sys_futimes(td, uap)
3126 	struct thread *td;
3127 	register struct futimes_args /* {
3128 		int  fd;
3129 		struct timeval *tptr;
3130 	} */ *uap;
3131 {
3132 
3133 	return (kern_futimes(td, uap->fd, uap->tptr, UIO_USERSPACE));
3134 }
3135 
3136 int
3137 kern_futimes(struct thread *td, int fd, struct timeval *tptr,
3138     enum uio_seg tptrseg)
3139 {
3140 	struct timespec ts[2];
3141 	struct file *fp;
3142 	cap_rights_t rights;
3143 	int error;
3144 
3145 	AUDIT_ARG_FD(fd);
3146 	error = getutimes(tptr, tptrseg, ts);
3147 	if (error != 0)
3148 		return (error);
3149 	error = getvnode(td, fd, cap_rights_init(&rights, CAP_FUTIMES), &fp);
3150 	if (error != 0)
3151 		return (error);
3152 #ifdef AUDIT
3153 	vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
3154 	AUDIT_ARG_VNODE1(fp->f_vnode);
3155 	VOP_UNLOCK(fp->f_vnode, 0);
3156 #endif
3157 	error = setutimes(td, fp->f_vnode, ts, 2, tptr == NULL);
3158 	fdrop(fp, td);
3159 	return (error);
3160 }
3161 
3162 int
3163 sys_futimens(struct thread *td, struct futimens_args *uap)
3164 {
3165 
3166 	return (kern_futimens(td, uap->fd, uap->times, UIO_USERSPACE));
3167 }
3168 
3169 int
3170 kern_futimens(struct thread *td, int fd, struct timespec *tptr,
3171     enum uio_seg tptrseg)
3172 {
3173 	struct timespec ts[2];
3174 	struct file *fp;
3175 	cap_rights_t rights;
3176 	int error, flags;
3177 
3178 	AUDIT_ARG_FD(fd);
3179 	error = getutimens(tptr, tptrseg, ts, &flags);
3180 	if (error != 0)
3181 		return (error);
3182 	if (flags & UTIMENS_EXIT)
3183 		return (0);
3184 	error = getvnode(td, fd, cap_rights_init(&rights, CAP_FUTIMES), &fp);
3185 	if (error != 0)
3186 		return (error);
3187 #ifdef AUDIT
3188 	vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
3189 	AUDIT_ARG_VNODE1(fp->f_vnode);
3190 	VOP_UNLOCK(fp->f_vnode, 0);
3191 #endif
3192 	error = setutimes(td, fp->f_vnode, ts, 2, flags & UTIMENS_NULL);
3193 	fdrop(fp, td);
3194 	return (error);
3195 }
3196 
3197 int
3198 sys_utimensat(struct thread *td, struct utimensat_args *uap)
3199 {
3200 
3201 	return (kern_utimensat(td, uap->fd, uap->path, UIO_USERSPACE,
3202 	    uap->times, UIO_USERSPACE, uap->flag));
3203 }
3204 
3205 int
3206 kern_utimensat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
3207     struct timespec *tptr, enum uio_seg tptrseg, int flag)
3208 {
3209 	struct nameidata nd;
3210 	struct timespec ts[2];
3211 	cap_rights_t rights;
3212 	int error, flags;
3213 
3214 	if (flag & ~AT_SYMLINK_NOFOLLOW)
3215 		return (EINVAL);
3216 
3217 	if ((error = getutimens(tptr, tptrseg, ts, &flags)) != 0)
3218 		return (error);
3219 	NDINIT_ATRIGHTS(&nd, LOOKUP, ((flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW :
3220 	    FOLLOW) | AUDITVNODE1, pathseg, path, fd,
3221 	    cap_rights_init(&rights, CAP_FUTIMES), td);
3222 	if ((error = namei(&nd)) != 0)
3223 		return (error);
3224 	/*
3225 	 * We are allowed to call namei() regardless of 2xUTIME_OMIT.
3226 	 * POSIX states:
3227 	 * "If both tv_nsec fields are UTIME_OMIT... EACCESS may be detected."
3228 	 * "Search permission is denied by a component of the path prefix."
3229 	 */
3230 	NDFREE(&nd, NDF_ONLY_PNBUF);
3231 	if ((flags & UTIMENS_EXIT) == 0)
3232 		error = setutimes(td, nd.ni_vp, ts, 2, flags & UTIMENS_NULL);
3233 	vrele(nd.ni_vp);
3234 	return (error);
3235 }
3236 
3237 /*
3238  * Truncate a file given its path name.
3239  */
3240 #ifndef _SYS_SYSPROTO_H_
3241 struct truncate_args {
3242 	char	*path;
3243 	int	pad;
3244 	off_t	length;
3245 };
3246 #endif
3247 int
3248 sys_truncate(td, uap)
3249 	struct thread *td;
3250 	register struct truncate_args /* {
3251 		char *path;
3252 		int pad;
3253 		off_t length;
3254 	} */ *uap;
3255 {
3256 
3257 	return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3258 }
3259 
3260 int
3261 kern_truncate(struct thread *td, char *path, enum uio_seg pathseg, off_t length)
3262 {
3263 	struct mount *mp;
3264 	struct vnode *vp;
3265 	void *rl_cookie;
3266 	struct vattr vattr;
3267 	struct nameidata nd;
3268 	int error;
3269 
3270 	if (length < 0)
3271 		return(EINVAL);
3272 	NDINIT(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, td);
3273 	if ((error = namei(&nd)) != 0)
3274 		return (error);
3275 	vp = nd.ni_vp;
3276 	rl_cookie = vn_rangelock_wlock(vp, 0, OFF_MAX);
3277 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) {
3278 		vn_rangelock_unlock(vp, rl_cookie);
3279 		vrele(vp);
3280 		return (error);
3281 	}
3282 	NDFREE(&nd, NDF_ONLY_PNBUF);
3283 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3284 	if (vp->v_type == VDIR)
3285 		error = EISDIR;
3286 #ifdef MAC
3287 	else if ((error = mac_vnode_check_write(td->td_ucred, NOCRED, vp))) {
3288 	}
3289 #endif
3290 	else if ((error = vn_writechk(vp)) == 0 &&
3291 	    (error = VOP_ACCESS(vp, VWRITE, td->td_ucred, td)) == 0) {
3292 		VATTR_NULL(&vattr);
3293 		vattr.va_size = length;
3294 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3295 	}
3296 	VOP_UNLOCK(vp, 0);
3297 	vn_finished_write(mp);
3298 	vn_rangelock_unlock(vp, rl_cookie);
3299 	vrele(vp);
3300 	return (error);
3301 }
3302 
3303 #if defined(COMPAT_43)
3304 /*
3305  * Truncate a file given its path name.
3306  */
3307 #ifndef _SYS_SYSPROTO_H_
3308 struct otruncate_args {
3309 	char	*path;
3310 	long	length;
3311 };
3312 #endif
3313 int
3314 otruncate(td, uap)
3315 	struct thread *td;
3316 	register struct otruncate_args /* {
3317 		char *path;
3318 		long length;
3319 	} */ *uap;
3320 {
3321 	struct truncate_args /* {
3322 		char *path;
3323 		int pad;
3324 		off_t length;
3325 	} */ nuap;
3326 
3327 	nuap.path = uap->path;
3328 	nuap.length = uap->length;
3329 	return (sys_truncate(td, &nuap));
3330 }
3331 #endif /* COMPAT_43 */
3332 
3333 #if defined(COMPAT_FREEBSD6)
3334 /* Versions with the pad argument */
3335 int
3336 freebsd6_truncate(struct thread *td, struct freebsd6_truncate_args *uap)
3337 {
3338 	struct truncate_args ouap;
3339 
3340 	ouap.path = uap->path;
3341 	ouap.length = uap->length;
3342 	return (sys_truncate(td, &ouap));
3343 }
3344 
3345 int
3346 freebsd6_ftruncate(struct thread *td, struct freebsd6_ftruncate_args *uap)
3347 {
3348 	struct ftruncate_args ouap;
3349 
3350 	ouap.fd = uap->fd;
3351 	ouap.length = uap->length;
3352 	return (sys_ftruncate(td, &ouap));
3353 }
3354 #endif
3355 
3356 int
3357 kern_fsync(struct thread *td, int fd, bool fullsync)
3358 {
3359 	struct vnode *vp;
3360 	struct mount *mp;
3361 	struct file *fp;
3362 	cap_rights_t rights;
3363 	int error, lock_flags;
3364 
3365 	AUDIT_ARG_FD(fd);
3366 	error = getvnode(td, fd, cap_rights_init(&rights, CAP_FSYNC), &fp);
3367 	if (error != 0)
3368 		return (error);
3369 	vp = fp->f_vnode;
3370 #if 0
3371 	if (!fullsync)
3372 		/* XXXKIB: compete outstanding aio writes */;
3373 #endif
3374 	error = vn_start_write(vp, &mp, V_WAIT | PCATCH);
3375 	if (error != 0)
3376 		goto drop;
3377 	if (MNT_SHARED_WRITES(mp) ||
3378 	    ((mp == NULL) && MNT_SHARED_WRITES(vp->v_mount))) {
3379 		lock_flags = LK_SHARED;
3380 	} else {
3381 		lock_flags = LK_EXCLUSIVE;
3382 	}
3383 	vn_lock(vp, lock_flags | LK_RETRY);
3384 	AUDIT_ARG_VNODE1(vp);
3385 	if (vp->v_object != NULL) {
3386 		VM_OBJECT_WLOCK(vp->v_object);
3387 		vm_object_page_clean(vp->v_object, 0, 0, 0);
3388 		VM_OBJECT_WUNLOCK(vp->v_object);
3389 	}
3390 	error = fullsync ? VOP_FSYNC(vp, MNT_WAIT, td) : VOP_FDATASYNC(vp, td);
3391 	VOP_UNLOCK(vp, 0);
3392 	vn_finished_write(mp);
3393 drop:
3394 	fdrop(fp, td);
3395 	return (error);
3396 }
3397 
3398 /*
3399  * Sync an open file.
3400  */
3401 #ifndef _SYS_SYSPROTO_H_
3402 struct fsync_args {
3403 	int	fd;
3404 };
3405 #endif
3406 int
3407 sys_fsync(struct thread *td, struct fsync_args *uap)
3408 {
3409 
3410 	return (kern_fsync(td, uap->fd, true));
3411 }
3412 
3413 int
3414 sys_fdatasync(struct thread *td, struct fdatasync_args *uap)
3415 {
3416 
3417 	return (kern_fsync(td, uap->fd, false));
3418 }
3419 
3420 /*
3421  * Rename files.  Source and destination must either both be directories, or
3422  * both not be directories.  If target is a directory, it must be empty.
3423  */
3424 #ifndef _SYS_SYSPROTO_H_
3425 struct rename_args {
3426 	char	*from;
3427 	char	*to;
3428 };
3429 #endif
3430 int
3431 sys_rename(td, uap)
3432 	struct thread *td;
3433 	register struct rename_args /* {
3434 		char *from;
3435 		char *to;
3436 	} */ *uap;
3437 {
3438 
3439 	return (kern_renameat(td, AT_FDCWD, uap->from, AT_FDCWD,
3440 	    uap->to, UIO_USERSPACE));
3441 }
3442 
3443 #ifndef _SYS_SYSPROTO_H_
3444 struct renameat_args {
3445 	int	oldfd;
3446 	char	*old;
3447 	int	newfd;
3448 	char	*new;
3449 };
3450 #endif
3451 int
3452 sys_renameat(struct thread *td, struct renameat_args *uap)
3453 {
3454 
3455 	return (kern_renameat(td, uap->oldfd, uap->old, uap->newfd, uap->new,
3456 	    UIO_USERSPACE));
3457 }
3458 
3459 int
3460 kern_renameat(struct thread *td, int oldfd, char *old, int newfd, char *new,
3461     enum uio_seg pathseg)
3462 {
3463 	struct mount *mp = NULL;
3464 	struct vnode *tvp, *fvp, *tdvp;
3465 	struct nameidata fromnd, tond;
3466 	cap_rights_t rights;
3467 	int error;
3468 
3469 again:
3470 	bwillwrite();
3471 #ifdef MAC
3472 	NDINIT_ATRIGHTS(&fromnd, DELETE, LOCKPARENT | LOCKLEAF | SAVESTART |
3473 	    AUDITVNODE1, pathseg, old, oldfd,
3474 	    cap_rights_init(&rights, CAP_RENAMEAT_SOURCE), td);
3475 #else
3476 	NDINIT_ATRIGHTS(&fromnd, DELETE, WANTPARENT | SAVESTART | AUDITVNODE1,
3477 	    pathseg, old, oldfd,
3478 	    cap_rights_init(&rights, CAP_RENAMEAT_SOURCE), td);
3479 #endif
3480 
3481 	if ((error = namei(&fromnd)) != 0)
3482 		return (error);
3483 #ifdef MAC
3484 	error = mac_vnode_check_rename_from(td->td_ucred, fromnd.ni_dvp,
3485 	    fromnd.ni_vp, &fromnd.ni_cnd);
3486 	VOP_UNLOCK(fromnd.ni_dvp, 0);
3487 	if (fromnd.ni_dvp != fromnd.ni_vp)
3488 		VOP_UNLOCK(fromnd.ni_vp, 0);
3489 #endif
3490 	fvp = fromnd.ni_vp;
3491 	NDINIT_ATRIGHTS(&tond, RENAME, LOCKPARENT | LOCKLEAF | NOCACHE |
3492 	    SAVESTART | AUDITVNODE2, pathseg, new, newfd,
3493 	    cap_rights_init(&rights, CAP_RENAMEAT_TARGET), td);
3494 	if (fromnd.ni_vp->v_type == VDIR)
3495 		tond.ni_cnd.cn_flags |= WILLBEDIR;
3496 	if ((error = namei(&tond)) != 0) {
3497 		/* Translate error code for rename("dir1", "dir2/."). */
3498 		if (error == EISDIR && fvp->v_type == VDIR)
3499 			error = EINVAL;
3500 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3501 		vrele(fromnd.ni_dvp);
3502 		vrele(fvp);
3503 		goto out1;
3504 	}
3505 	tdvp = tond.ni_dvp;
3506 	tvp = tond.ni_vp;
3507 	error = vn_start_write(fvp, &mp, V_NOWAIT);
3508 	if (error != 0) {
3509 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3510 		NDFREE(&tond, NDF_ONLY_PNBUF);
3511 		if (tvp != NULL)
3512 			vput(tvp);
3513 		if (tdvp == tvp)
3514 			vrele(tdvp);
3515 		else
3516 			vput(tdvp);
3517 		vrele(fromnd.ni_dvp);
3518 		vrele(fvp);
3519 		vrele(tond.ni_startdir);
3520 		if (fromnd.ni_startdir != NULL)
3521 			vrele(fromnd.ni_startdir);
3522 		error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH);
3523 		if (error != 0)
3524 			return (error);
3525 		goto again;
3526 	}
3527 	if (tvp != NULL) {
3528 		if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
3529 			error = ENOTDIR;
3530 			goto out;
3531 		} else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
3532 			error = EISDIR;
3533 			goto out;
3534 		}
3535 #ifdef CAPABILITIES
3536 		if (newfd != AT_FDCWD) {
3537 			/*
3538 			 * If the target already exists we require CAP_UNLINKAT
3539 			 * from 'newfd'.
3540 			 */
3541 			error = cap_check(&tond.ni_filecaps.fc_rights,
3542 			    cap_rights_init(&rights, CAP_UNLINKAT));
3543 			if (error != 0)
3544 				goto out;
3545 		}
3546 #endif
3547 	}
3548 	if (fvp == tdvp) {
3549 		error = EINVAL;
3550 		goto out;
3551 	}
3552 	/*
3553 	 * If the source is the same as the destination (that is, if they
3554 	 * are links to the same vnode), then there is nothing to do.
3555 	 */
3556 	if (fvp == tvp)
3557 		error = -1;
3558 #ifdef MAC
3559 	else
3560 		error = mac_vnode_check_rename_to(td->td_ucred, tdvp,
3561 		    tond.ni_vp, fromnd.ni_dvp == tdvp, &tond.ni_cnd);
3562 #endif
3563 out:
3564 	if (error == 0) {
3565 		error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
3566 		    tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
3567 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3568 		NDFREE(&tond, NDF_ONLY_PNBUF);
3569 	} else {
3570 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3571 		NDFREE(&tond, NDF_ONLY_PNBUF);
3572 		if (tvp != NULL)
3573 			vput(tvp);
3574 		if (tdvp == tvp)
3575 			vrele(tdvp);
3576 		else
3577 			vput(tdvp);
3578 		vrele(fromnd.ni_dvp);
3579 		vrele(fvp);
3580 	}
3581 	vrele(tond.ni_startdir);
3582 	vn_finished_write(mp);
3583 out1:
3584 	if (fromnd.ni_startdir)
3585 		vrele(fromnd.ni_startdir);
3586 	if (error == -1)
3587 		return (0);
3588 	return (error);
3589 }
3590 
3591 /*
3592  * Make a directory file.
3593  */
3594 #ifndef _SYS_SYSPROTO_H_
3595 struct mkdir_args {
3596 	char	*path;
3597 	int	mode;
3598 };
3599 #endif
3600 int
3601 sys_mkdir(td, uap)
3602 	struct thread *td;
3603 	register struct mkdir_args /* {
3604 		char *path;
3605 		int mode;
3606 	} */ *uap;
3607 {
3608 
3609 	return (kern_mkdirat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
3610 	    uap->mode));
3611 }
3612 
3613 #ifndef _SYS_SYSPROTO_H_
3614 struct mkdirat_args {
3615 	int	fd;
3616 	char	*path;
3617 	mode_t	mode;
3618 };
3619 #endif
3620 int
3621 sys_mkdirat(struct thread *td, struct mkdirat_args *uap)
3622 {
3623 
3624 	return (kern_mkdirat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode));
3625 }
3626 
3627 int
3628 kern_mkdirat(struct thread *td, int fd, char *path, enum uio_seg segflg,
3629     int mode)
3630 {
3631 	struct mount *mp;
3632 	struct vnode *vp;
3633 	struct vattr vattr;
3634 	struct nameidata nd;
3635 	cap_rights_t rights;
3636 	int error;
3637 
3638 	AUDIT_ARG_MODE(mode);
3639 restart:
3640 	bwillwrite();
3641 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
3642 	    NOCACHE, segflg, path, fd, cap_rights_init(&rights, CAP_MKDIRAT),
3643 	    td);
3644 	nd.ni_cnd.cn_flags |= WILLBEDIR;
3645 	if ((error = namei(&nd)) != 0)
3646 		return (error);
3647 	vp = nd.ni_vp;
3648 	if (vp != NULL) {
3649 		NDFREE(&nd, NDF_ONLY_PNBUF);
3650 		/*
3651 		 * XXX namei called with LOCKPARENT but not LOCKLEAF has
3652 		 * the strange behaviour of leaving the vnode unlocked
3653 		 * if the target is the same vnode as the parent.
3654 		 */
3655 		if (vp == nd.ni_dvp)
3656 			vrele(nd.ni_dvp);
3657 		else
3658 			vput(nd.ni_dvp);
3659 		vrele(vp);
3660 		return (EEXIST);
3661 	}
3662 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3663 		NDFREE(&nd, NDF_ONLY_PNBUF);
3664 		vput(nd.ni_dvp);
3665 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3666 			return (error);
3667 		goto restart;
3668 	}
3669 	VATTR_NULL(&vattr);
3670 	vattr.va_type = VDIR;
3671 	vattr.va_mode = (mode & ACCESSPERMS) &~ td->td_proc->p_fd->fd_cmask;
3672 #ifdef MAC
3673 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
3674 	    &vattr);
3675 	if (error != 0)
3676 		goto out;
3677 #endif
3678 	error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
3679 #ifdef MAC
3680 out:
3681 #endif
3682 	NDFREE(&nd, NDF_ONLY_PNBUF);
3683 	vput(nd.ni_dvp);
3684 	if (error == 0)
3685 		vput(nd.ni_vp);
3686 	vn_finished_write(mp);
3687 	return (error);
3688 }
3689 
3690 /*
3691  * Remove a directory file.
3692  */
3693 #ifndef _SYS_SYSPROTO_H_
3694 struct rmdir_args {
3695 	char	*path;
3696 };
3697 #endif
3698 int
3699 sys_rmdir(td, uap)
3700 	struct thread *td;
3701 	struct rmdir_args /* {
3702 		char *path;
3703 	} */ *uap;
3704 {
3705 
3706 	return (kern_rmdirat(td, AT_FDCWD, uap->path, UIO_USERSPACE));
3707 }
3708 
3709 int
3710 kern_rmdirat(struct thread *td, int fd, char *path, enum uio_seg pathseg)
3711 {
3712 	struct mount *mp;
3713 	struct vnode *vp;
3714 	struct nameidata nd;
3715 	cap_rights_t rights;
3716 	int error;
3717 
3718 restart:
3719 	bwillwrite();
3720 	NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1,
3721 	    pathseg, path, fd, cap_rights_init(&rights, CAP_UNLINKAT), td);
3722 	if ((error = namei(&nd)) != 0)
3723 		return (error);
3724 	vp = nd.ni_vp;
3725 	if (vp->v_type != VDIR) {
3726 		error = ENOTDIR;
3727 		goto out;
3728 	}
3729 	/*
3730 	 * No rmdir "." please.
3731 	 */
3732 	if (nd.ni_dvp == vp) {
3733 		error = EINVAL;
3734 		goto out;
3735 	}
3736 	/*
3737 	 * The root of a mounted filesystem cannot be deleted.
3738 	 */
3739 	if (vp->v_vflag & VV_ROOT) {
3740 		error = EBUSY;
3741 		goto out;
3742 	}
3743 #ifdef MAC
3744 	error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
3745 	    &nd.ni_cnd);
3746 	if (error != 0)
3747 		goto out;
3748 #endif
3749 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3750 		NDFREE(&nd, NDF_ONLY_PNBUF);
3751 		vput(vp);
3752 		if (nd.ni_dvp == vp)
3753 			vrele(nd.ni_dvp);
3754 		else
3755 			vput(nd.ni_dvp);
3756 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3757 			return (error);
3758 		goto restart;
3759 	}
3760 	vfs_notify_upper(vp, VFS_NOTIFY_UPPER_UNLINK);
3761 	error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
3762 	vn_finished_write(mp);
3763 out:
3764 	NDFREE(&nd, NDF_ONLY_PNBUF);
3765 	vput(vp);
3766 	if (nd.ni_dvp == vp)
3767 		vrele(nd.ni_dvp);
3768 	else
3769 		vput(nd.ni_dvp);
3770 	return (error);
3771 }
3772 
3773 #ifdef COMPAT_43
3774 /*
3775  * Read a block of directory entries in a filesystem independent format.
3776  */
3777 #ifndef _SYS_SYSPROTO_H_
3778 struct ogetdirentries_args {
3779 	int	fd;
3780 	char	*buf;
3781 	u_int	count;
3782 	long	*basep;
3783 };
3784 #endif
3785 int
3786 ogetdirentries(struct thread *td, struct ogetdirentries_args *uap)
3787 {
3788 	long loff;
3789 	int error;
3790 
3791 	error = kern_ogetdirentries(td, uap, &loff);
3792 	if (error == 0)
3793 		error = copyout(&loff, uap->basep, sizeof(long));
3794 	return (error);
3795 }
3796 
3797 int
3798 kern_ogetdirentries(struct thread *td, struct ogetdirentries_args *uap,
3799     long *ploff)
3800 {
3801 	struct vnode *vp;
3802 	struct file *fp;
3803 	struct uio auio, kuio;
3804 	struct iovec aiov, kiov;
3805 	struct dirent *dp, *edp;
3806 	cap_rights_t rights;
3807 	caddr_t dirbuf;
3808 	int error, eofflag, readcnt;
3809 	long loff;
3810 	off_t foffset;
3811 
3812 	/* XXX arbitrary sanity limit on `count'. */
3813 	if (uap->count > 64 * 1024)
3814 		return (EINVAL);
3815 	error = getvnode(td, uap->fd, cap_rights_init(&rights, CAP_READ), &fp);
3816 	if (error != 0)
3817 		return (error);
3818 	if ((fp->f_flag & FREAD) == 0) {
3819 		fdrop(fp, td);
3820 		return (EBADF);
3821 	}
3822 	vp = fp->f_vnode;
3823 	foffset = foffset_lock(fp, 0);
3824 unionread:
3825 	if (vp->v_type != VDIR) {
3826 		foffset_unlock(fp, foffset, 0);
3827 		fdrop(fp, td);
3828 		return (EINVAL);
3829 	}
3830 	aiov.iov_base = uap->buf;
3831 	aiov.iov_len = uap->count;
3832 	auio.uio_iov = &aiov;
3833 	auio.uio_iovcnt = 1;
3834 	auio.uio_rw = UIO_READ;
3835 	auio.uio_segflg = UIO_USERSPACE;
3836 	auio.uio_td = td;
3837 	auio.uio_resid = uap->count;
3838 	vn_lock(vp, LK_SHARED | LK_RETRY);
3839 	loff = auio.uio_offset = foffset;
3840 #ifdef MAC
3841 	error = mac_vnode_check_readdir(td->td_ucred, vp);
3842 	if (error != 0) {
3843 		VOP_UNLOCK(vp, 0);
3844 		foffset_unlock(fp, foffset, FOF_NOUPDATE);
3845 		fdrop(fp, td);
3846 		return (error);
3847 	}
3848 #endif
3849 #	if (BYTE_ORDER != LITTLE_ENDIAN)
3850 		if (vp->v_mount->mnt_maxsymlinklen <= 0) {
3851 			error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag,
3852 			    NULL, NULL);
3853 			foffset = auio.uio_offset;
3854 		} else
3855 #	endif
3856 	{
3857 		kuio = auio;
3858 		kuio.uio_iov = &kiov;
3859 		kuio.uio_segflg = UIO_SYSSPACE;
3860 		kiov.iov_len = uap->count;
3861 		dirbuf = malloc(uap->count, M_TEMP, M_WAITOK);
3862 		kiov.iov_base = dirbuf;
3863 		error = VOP_READDIR(vp, &kuio, fp->f_cred, &eofflag,
3864 			    NULL, NULL);
3865 		foffset = kuio.uio_offset;
3866 		if (error == 0) {
3867 			readcnt = uap->count - kuio.uio_resid;
3868 			edp = (struct dirent *)&dirbuf[readcnt];
3869 			for (dp = (struct dirent *)dirbuf; dp < edp; ) {
3870 #				if (BYTE_ORDER == LITTLE_ENDIAN)
3871 					/*
3872 					 * The expected low byte of
3873 					 * dp->d_namlen is our dp->d_type.
3874 					 * The high MBZ byte of dp->d_namlen
3875 					 * is our dp->d_namlen.
3876 					 */
3877 					dp->d_type = dp->d_namlen;
3878 					dp->d_namlen = 0;
3879 #				else
3880 					/*
3881 					 * The dp->d_type is the high byte
3882 					 * of the expected dp->d_namlen,
3883 					 * so must be zero'ed.
3884 					 */
3885 					dp->d_type = 0;
3886 #				endif
3887 				if (dp->d_reclen > 0) {
3888 					dp = (struct dirent *)
3889 					    ((char *)dp + dp->d_reclen);
3890 				} else {
3891 					error = EIO;
3892 					break;
3893 				}
3894 			}
3895 			if (dp >= edp)
3896 				error = uiomove(dirbuf, readcnt, &auio);
3897 		}
3898 		free(dirbuf, M_TEMP);
3899 	}
3900 	if (error != 0) {
3901 		VOP_UNLOCK(vp, 0);
3902 		foffset_unlock(fp, foffset, 0);
3903 		fdrop(fp, td);
3904 		return (error);
3905 	}
3906 	if (uap->count == auio.uio_resid &&
3907 	    (vp->v_vflag & VV_ROOT) &&
3908 	    (vp->v_mount->mnt_flag & MNT_UNION)) {
3909 		struct vnode *tvp = vp;
3910 		vp = vp->v_mount->mnt_vnodecovered;
3911 		VREF(vp);
3912 		fp->f_vnode = vp;
3913 		fp->f_data = vp;
3914 		foffset = 0;
3915 		vput(tvp);
3916 		goto unionread;
3917 	}
3918 	VOP_UNLOCK(vp, 0);
3919 	foffset_unlock(fp, foffset, 0);
3920 	fdrop(fp, td);
3921 	td->td_retval[0] = uap->count - auio.uio_resid;
3922 	if (error == 0)
3923 		*ploff = loff;
3924 	return (error);
3925 }
3926 #endif /* COMPAT_43 */
3927 
3928 /*
3929  * Read a block of directory entries in a filesystem independent format.
3930  */
3931 #ifndef _SYS_SYSPROTO_H_
3932 struct getdirentries_args {
3933 	int	fd;
3934 	char	*buf;
3935 	u_int	count;
3936 	long	*basep;
3937 };
3938 #endif
3939 int
3940 sys_getdirentries(td, uap)
3941 	struct thread *td;
3942 	register struct getdirentries_args /* {
3943 		int fd;
3944 		char *buf;
3945 		u_int count;
3946 		long *basep;
3947 	} */ *uap;
3948 {
3949 	long base;
3950 	int error;
3951 
3952 	error = kern_getdirentries(td, uap->fd, uap->buf, uap->count, &base,
3953 	    NULL, UIO_USERSPACE);
3954 	if (error != 0)
3955 		return (error);
3956 	if (uap->basep != NULL)
3957 		error = copyout(&base, uap->basep, sizeof(long));
3958 	return (error);
3959 }
3960 
3961 int
3962 kern_getdirentries(struct thread *td, int fd, char *buf, u_int count,
3963     long *basep, ssize_t *residp, enum uio_seg bufseg)
3964 {
3965 	struct vnode *vp;
3966 	struct file *fp;
3967 	struct uio auio;
3968 	struct iovec aiov;
3969 	cap_rights_t rights;
3970 	long loff;
3971 	int error, eofflag;
3972 	off_t foffset;
3973 
3974 	AUDIT_ARG_FD(fd);
3975 	if (count > IOSIZE_MAX)
3976 		return (EINVAL);
3977 	auio.uio_resid = count;
3978 	error = getvnode(td, fd, cap_rights_init(&rights, CAP_READ), &fp);
3979 	if (error != 0)
3980 		return (error);
3981 	if ((fp->f_flag & FREAD) == 0) {
3982 		fdrop(fp, td);
3983 		return (EBADF);
3984 	}
3985 	vp = fp->f_vnode;
3986 	foffset = foffset_lock(fp, 0);
3987 unionread:
3988 	if (vp->v_type != VDIR) {
3989 		error = EINVAL;
3990 		goto fail;
3991 	}
3992 	aiov.iov_base = buf;
3993 	aiov.iov_len = count;
3994 	auio.uio_iov = &aiov;
3995 	auio.uio_iovcnt = 1;
3996 	auio.uio_rw = UIO_READ;
3997 	auio.uio_segflg = bufseg;
3998 	auio.uio_td = td;
3999 	vn_lock(vp, LK_SHARED | LK_RETRY);
4000 	AUDIT_ARG_VNODE1(vp);
4001 	loff = auio.uio_offset = foffset;
4002 #ifdef MAC
4003 	error = mac_vnode_check_readdir(td->td_ucred, vp);
4004 	if (error == 0)
4005 #endif
4006 		error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, NULL,
4007 		    NULL);
4008 	foffset = auio.uio_offset;
4009 	if (error != 0) {
4010 		VOP_UNLOCK(vp, 0);
4011 		goto fail;
4012 	}
4013 	if (count == auio.uio_resid &&
4014 	    (vp->v_vflag & VV_ROOT) &&
4015 	    (vp->v_mount->mnt_flag & MNT_UNION)) {
4016 		struct vnode *tvp = vp;
4017 
4018 		vp = vp->v_mount->mnt_vnodecovered;
4019 		VREF(vp);
4020 		fp->f_vnode = vp;
4021 		fp->f_data = vp;
4022 		foffset = 0;
4023 		vput(tvp);
4024 		goto unionread;
4025 	}
4026 	VOP_UNLOCK(vp, 0);
4027 	*basep = loff;
4028 	if (residp != NULL)
4029 		*residp = auio.uio_resid;
4030 	td->td_retval[0] = count - auio.uio_resid;
4031 fail:
4032 	foffset_unlock(fp, foffset, 0);
4033 	fdrop(fp, td);
4034 	return (error);
4035 }
4036 
4037 #ifndef _SYS_SYSPROTO_H_
4038 struct getdents_args {
4039 	int fd;
4040 	char *buf;
4041 	size_t count;
4042 };
4043 #endif
4044 int
4045 sys_getdents(td, uap)
4046 	struct thread *td;
4047 	register struct getdents_args /* {
4048 		int fd;
4049 		char *buf;
4050 		u_int count;
4051 	} */ *uap;
4052 {
4053 	struct getdirentries_args ap;
4054 
4055 	ap.fd = uap->fd;
4056 	ap.buf = uap->buf;
4057 	ap.count = uap->count;
4058 	ap.basep = NULL;
4059 	return (sys_getdirentries(td, &ap));
4060 }
4061 
4062 /*
4063  * Set the mode mask for creation of filesystem nodes.
4064  */
4065 #ifndef _SYS_SYSPROTO_H_
4066 struct umask_args {
4067 	int	newmask;
4068 };
4069 #endif
4070 int
4071 sys_umask(td, uap)
4072 	struct thread *td;
4073 	struct umask_args /* {
4074 		int newmask;
4075 	} */ *uap;
4076 {
4077 	struct filedesc *fdp;
4078 
4079 	fdp = td->td_proc->p_fd;
4080 	FILEDESC_XLOCK(fdp);
4081 	td->td_retval[0] = fdp->fd_cmask;
4082 	fdp->fd_cmask = uap->newmask & ALLPERMS;
4083 	FILEDESC_XUNLOCK(fdp);
4084 	return (0);
4085 }
4086 
4087 /*
4088  * Void all references to file by ripping underlying filesystem away from
4089  * vnode.
4090  */
4091 #ifndef _SYS_SYSPROTO_H_
4092 struct revoke_args {
4093 	char	*path;
4094 };
4095 #endif
4096 int
4097 sys_revoke(td, uap)
4098 	struct thread *td;
4099 	register struct revoke_args /* {
4100 		char *path;
4101 	} */ *uap;
4102 {
4103 	struct vnode *vp;
4104 	struct vattr vattr;
4105 	struct nameidata nd;
4106 	int error;
4107 
4108 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
4109 	    uap->path, td);
4110 	if ((error = namei(&nd)) != 0)
4111 		return (error);
4112 	vp = nd.ni_vp;
4113 	NDFREE(&nd, NDF_ONLY_PNBUF);
4114 	if (vp->v_type != VCHR || vp->v_rdev == NULL) {
4115 		error = EINVAL;
4116 		goto out;
4117 	}
4118 #ifdef MAC
4119 	error = mac_vnode_check_revoke(td->td_ucred, vp);
4120 	if (error != 0)
4121 		goto out;
4122 #endif
4123 	error = VOP_GETATTR(vp, &vattr, td->td_ucred);
4124 	if (error != 0)
4125 		goto out;
4126 	if (td->td_ucred->cr_uid != vattr.va_uid) {
4127 		error = priv_check(td, PRIV_VFS_ADMIN);
4128 		if (error != 0)
4129 			goto out;
4130 	}
4131 	if (vcount(vp) > 1)
4132 		VOP_REVOKE(vp, REVOKEALL);
4133 out:
4134 	vput(vp);
4135 	return (error);
4136 }
4137 
4138 /*
4139  * Convert a user file descriptor to a kernel file entry and check that, if it
4140  * is a capability, the correct rights are present. A reference on the file
4141  * entry is held upon returning.
4142  */
4143 int
4144 getvnode(struct thread *td, int fd, cap_rights_t *rightsp, struct file **fpp)
4145 {
4146 	struct file *fp;
4147 	int error;
4148 
4149 	error = fget_unlocked(td->td_proc->p_fd, fd, rightsp, &fp, NULL);
4150 	if (error != 0)
4151 		return (error);
4152 
4153 	/*
4154 	 * The file could be not of the vnode type, or it may be not
4155 	 * yet fully initialized, in which case the f_vnode pointer
4156 	 * may be set, but f_ops is still badfileops.  E.g.,
4157 	 * devfs_open() transiently create such situation to
4158 	 * facilitate csw d_fdopen().
4159 	 *
4160 	 * Dupfdopen() handling in kern_openat() installs the
4161 	 * half-baked file into the process descriptor table, allowing
4162 	 * other thread to dereference it. Guard against the race by
4163 	 * checking f_ops.
4164 	 */
4165 	if (fp->f_vnode == NULL || fp->f_ops == &badfileops) {
4166 		fdrop(fp, td);
4167 		return (EINVAL);
4168 	}
4169 	*fpp = fp;
4170 	return (0);
4171 }
4172 
4173 
4174 /*
4175  * Get an (NFS) file handle.
4176  */
4177 #ifndef _SYS_SYSPROTO_H_
4178 struct lgetfh_args {
4179 	char	*fname;
4180 	fhandle_t *fhp;
4181 };
4182 #endif
4183 int
4184 sys_lgetfh(td, uap)
4185 	struct thread *td;
4186 	register struct lgetfh_args *uap;
4187 {
4188 	struct nameidata nd;
4189 	fhandle_t fh;
4190 	register struct vnode *vp;
4191 	int error;
4192 
4193 	error = priv_check(td, PRIV_VFS_GETFH);
4194 	if (error != 0)
4195 		return (error);
4196 	NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
4197 	    uap->fname, td);
4198 	error = namei(&nd);
4199 	if (error != 0)
4200 		return (error);
4201 	NDFREE(&nd, NDF_ONLY_PNBUF);
4202 	vp = nd.ni_vp;
4203 	bzero(&fh, sizeof(fh));
4204 	fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
4205 	error = VOP_VPTOFH(vp, &fh.fh_fid);
4206 	vput(vp);
4207 	if (error == 0)
4208 		error = copyout(&fh, uap->fhp, sizeof (fh));
4209 	return (error);
4210 }
4211 
4212 #ifndef _SYS_SYSPROTO_H_
4213 struct getfh_args {
4214 	char	*fname;
4215 	fhandle_t *fhp;
4216 };
4217 #endif
4218 int
4219 sys_getfh(td, uap)
4220 	struct thread *td;
4221 	register struct getfh_args *uap;
4222 {
4223 	struct nameidata nd;
4224 	fhandle_t fh;
4225 	register struct vnode *vp;
4226 	int error;
4227 
4228 	error = priv_check(td, PRIV_VFS_GETFH);
4229 	if (error != 0)
4230 		return (error);
4231 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
4232 	    uap->fname, td);
4233 	error = namei(&nd);
4234 	if (error != 0)
4235 		return (error);
4236 	NDFREE(&nd, NDF_ONLY_PNBUF);
4237 	vp = nd.ni_vp;
4238 	bzero(&fh, sizeof(fh));
4239 	fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
4240 	error = VOP_VPTOFH(vp, &fh.fh_fid);
4241 	vput(vp);
4242 	if (error == 0)
4243 		error = copyout(&fh, uap->fhp, sizeof (fh));
4244 	return (error);
4245 }
4246 
4247 /*
4248  * syscall for the rpc.lockd to use to translate a NFS file handle into an
4249  * open descriptor.
4250  *
4251  * warning: do not remove the priv_check() call or this becomes one giant
4252  * security hole.
4253  */
4254 #ifndef _SYS_SYSPROTO_H_
4255 struct fhopen_args {
4256 	const struct fhandle *u_fhp;
4257 	int flags;
4258 };
4259 #endif
4260 int
4261 sys_fhopen(td, uap)
4262 	struct thread *td;
4263 	struct fhopen_args /* {
4264 		const struct fhandle *u_fhp;
4265 		int flags;
4266 	} */ *uap;
4267 {
4268 	struct mount *mp;
4269 	struct vnode *vp;
4270 	struct fhandle fhp;
4271 	struct file *fp;
4272 	int fmode, error;
4273 	int indx;
4274 
4275 	error = priv_check(td, PRIV_VFS_FHOPEN);
4276 	if (error != 0)
4277 		return (error);
4278 	indx = -1;
4279 	fmode = FFLAGS(uap->flags);
4280 	/* why not allow a non-read/write open for our lockd? */
4281 	if (((fmode & (FREAD | FWRITE)) == 0) || (fmode & O_CREAT))
4282 		return (EINVAL);
4283 	error = copyin(uap->u_fhp, &fhp, sizeof(fhp));
4284 	if (error != 0)
4285 		return(error);
4286 	/* find the mount point */
4287 	mp = vfs_busyfs(&fhp.fh_fsid);
4288 	if (mp == NULL)
4289 		return (ESTALE);
4290 	/* now give me my vnode, it gets returned to me locked */
4291 	error = VFS_FHTOVP(mp, &fhp.fh_fid, LK_EXCLUSIVE, &vp);
4292 	vfs_unbusy(mp);
4293 	if (error != 0)
4294 		return (error);
4295 
4296 	error = falloc_noinstall(td, &fp);
4297 	if (error != 0) {
4298 		vput(vp);
4299 		return (error);
4300 	}
4301 	/*
4302 	 * An extra reference on `fp' has been held for us by
4303 	 * falloc_noinstall().
4304 	 */
4305 
4306 #ifdef INVARIANTS
4307 	td->td_dupfd = -1;
4308 #endif
4309 	error = vn_open_vnode(vp, fmode, td->td_ucred, td, fp);
4310 	if (error != 0) {
4311 		KASSERT(fp->f_ops == &badfileops,
4312 		    ("VOP_OPEN in fhopen() set f_ops"));
4313 		KASSERT(td->td_dupfd < 0,
4314 		    ("fhopen() encountered fdopen()"));
4315 
4316 		vput(vp);
4317 		goto bad;
4318 	}
4319 #ifdef INVARIANTS
4320 	td->td_dupfd = 0;
4321 #endif
4322 	fp->f_vnode = vp;
4323 	fp->f_seqcount = 1;
4324 	finit(fp, (fmode & FMASK) | (fp->f_flag & FHASLOCK), DTYPE_VNODE, vp,
4325 	    &vnops);
4326 	VOP_UNLOCK(vp, 0);
4327 	if ((fmode & O_TRUNC) != 0) {
4328 		error = fo_truncate(fp, 0, td->td_ucred, td);
4329 		if (error != 0)
4330 			goto bad;
4331 	}
4332 
4333 	error = finstall(td, fp, &indx, fmode, NULL);
4334 bad:
4335 	fdrop(fp, td);
4336 	td->td_retval[0] = indx;
4337 	return (error);
4338 }
4339 
4340 /*
4341  * Stat an (NFS) file handle.
4342  */
4343 #ifndef _SYS_SYSPROTO_H_
4344 struct fhstat_args {
4345 	struct fhandle *u_fhp;
4346 	struct stat *sb;
4347 };
4348 #endif
4349 int
4350 sys_fhstat(td, uap)
4351 	struct thread *td;
4352 	register struct fhstat_args /* {
4353 		struct fhandle *u_fhp;
4354 		struct stat *sb;
4355 	} */ *uap;
4356 {
4357 	struct stat sb;
4358 	struct fhandle fh;
4359 	int error;
4360 
4361 	error = copyin(uap->u_fhp, &fh, sizeof(fh));
4362 	if (error != 0)
4363 		return (error);
4364 	error = kern_fhstat(td, fh, &sb);
4365 	if (error == 0)
4366 		error = copyout(&sb, uap->sb, sizeof(sb));
4367 	return (error);
4368 }
4369 
4370 int
4371 kern_fhstat(struct thread *td, struct fhandle fh, struct stat *sb)
4372 {
4373 	struct mount *mp;
4374 	struct vnode *vp;
4375 	int error;
4376 
4377 	error = priv_check(td, PRIV_VFS_FHSTAT);
4378 	if (error != 0)
4379 		return (error);
4380 	if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4381 		return (ESTALE);
4382 	error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4383 	vfs_unbusy(mp);
4384 	if (error != 0)
4385 		return (error);
4386 	error = vn_stat(vp, sb, td->td_ucred, NOCRED, td);
4387 	vput(vp);
4388 	return (error);
4389 }
4390 
4391 /*
4392  * Implement fstatfs() for (NFS) file handles.
4393  */
4394 #ifndef _SYS_SYSPROTO_H_
4395 struct fhstatfs_args {
4396 	struct fhandle *u_fhp;
4397 	struct statfs *buf;
4398 };
4399 #endif
4400 int
4401 sys_fhstatfs(td, uap)
4402 	struct thread *td;
4403 	struct fhstatfs_args /* {
4404 		struct fhandle *u_fhp;
4405 		struct statfs *buf;
4406 	} */ *uap;
4407 {
4408 	struct statfs sf;
4409 	fhandle_t fh;
4410 	int error;
4411 
4412 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
4413 	if (error != 0)
4414 		return (error);
4415 	error = kern_fhstatfs(td, fh, &sf);
4416 	if (error != 0)
4417 		return (error);
4418 	return (copyout(&sf, uap->buf, sizeof(sf)));
4419 }
4420 
4421 int
4422 kern_fhstatfs(struct thread *td, fhandle_t fh, struct statfs *buf)
4423 {
4424 	struct statfs *sp;
4425 	struct mount *mp;
4426 	struct vnode *vp;
4427 	int error;
4428 
4429 	error = priv_check(td, PRIV_VFS_FHSTATFS);
4430 	if (error != 0)
4431 		return (error);
4432 	if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4433 		return (ESTALE);
4434 	error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4435 	if (error != 0) {
4436 		vfs_unbusy(mp);
4437 		return (error);
4438 	}
4439 	vput(vp);
4440 	error = prison_canseemount(td->td_ucred, mp);
4441 	if (error != 0)
4442 		goto out;
4443 #ifdef MAC
4444 	error = mac_mount_check_stat(td->td_ucred, mp);
4445 	if (error != 0)
4446 		goto out;
4447 #endif
4448 	/*
4449 	 * Set these in case the underlying filesystem fails to do so.
4450 	 */
4451 	sp = &mp->mnt_stat;
4452 	sp->f_version = STATFS_VERSION;
4453 	sp->f_namemax = NAME_MAX;
4454 	sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
4455 	error = VFS_STATFS(mp, sp);
4456 	if (error == 0)
4457 		*buf = *sp;
4458 out:
4459 	vfs_unbusy(mp);
4460 	return (error);
4461 }
4462 
4463 int
4464 kern_posix_fallocate(struct thread *td, int fd, off_t offset, off_t len)
4465 {
4466 	struct file *fp;
4467 	struct mount *mp;
4468 	struct vnode *vp;
4469 	cap_rights_t rights;
4470 	off_t olen, ooffset;
4471 	int error;
4472 
4473 	if (offset < 0 || len <= 0)
4474 		return (EINVAL);
4475 	/* Check for wrap. */
4476 	if (offset > OFF_MAX - len)
4477 		return (EFBIG);
4478 	error = fget(td, fd, cap_rights_init(&rights, CAP_WRITE), &fp);
4479 	if (error != 0)
4480 		return (error);
4481 	if ((fp->f_ops->fo_flags & DFLAG_SEEKABLE) == 0) {
4482 		error = ESPIPE;
4483 		goto out;
4484 	}
4485 	if ((fp->f_flag & FWRITE) == 0) {
4486 		error = EBADF;
4487 		goto out;
4488 	}
4489 	if (fp->f_type != DTYPE_VNODE) {
4490 		error = ENODEV;
4491 		goto out;
4492 	}
4493 	vp = fp->f_vnode;
4494 	if (vp->v_type != VREG) {
4495 		error = ENODEV;
4496 		goto out;
4497 	}
4498 
4499 	/* Allocating blocks may take a long time, so iterate. */
4500 	for (;;) {
4501 		olen = len;
4502 		ooffset = offset;
4503 
4504 		bwillwrite();
4505 		mp = NULL;
4506 		error = vn_start_write(vp, &mp, V_WAIT | PCATCH);
4507 		if (error != 0)
4508 			break;
4509 		error = vn_lock(vp, LK_EXCLUSIVE);
4510 		if (error != 0) {
4511 			vn_finished_write(mp);
4512 			break;
4513 		}
4514 #ifdef MAC
4515 		error = mac_vnode_check_write(td->td_ucred, fp->f_cred, vp);
4516 		if (error == 0)
4517 #endif
4518 			error = VOP_ALLOCATE(vp, &offset, &len);
4519 		VOP_UNLOCK(vp, 0);
4520 		vn_finished_write(mp);
4521 
4522 		if (olen + ooffset != offset + len) {
4523 			panic("offset + len changed from %jx/%jx to %jx/%jx",
4524 			    ooffset, olen, offset, len);
4525 		}
4526 		if (error != 0 || len == 0)
4527 			break;
4528 		KASSERT(olen > len, ("Iteration did not make progress?"));
4529 		maybe_yield();
4530 	}
4531  out:
4532 	fdrop(fp, td);
4533 	return (error);
4534 }
4535 
4536 int
4537 sys_posix_fallocate(struct thread *td, struct posix_fallocate_args *uap)
4538 {
4539 	int error;
4540 
4541 	error = kern_posix_fallocate(td, uap->fd, uap->offset, uap->len);
4542 	return (kern_posix_error(td, error));
4543 }
4544 
4545 /*
4546  * Unlike madvise(2), we do not make a best effort to remember every
4547  * possible caching hint.  Instead, we remember the last setting with
4548  * the exception that we will allow POSIX_FADV_NORMAL to adjust the
4549  * region of any current setting.
4550  */
4551 int
4552 kern_posix_fadvise(struct thread *td, int fd, off_t offset, off_t len,
4553     int advice)
4554 {
4555 	struct fadvise_info *fa, *new;
4556 	struct file *fp;
4557 	struct vnode *vp;
4558 	cap_rights_t rights;
4559 	off_t end;
4560 	int error;
4561 
4562 	if (offset < 0 || len < 0 || offset > OFF_MAX - len)
4563 		return (EINVAL);
4564 	switch (advice) {
4565 	case POSIX_FADV_SEQUENTIAL:
4566 	case POSIX_FADV_RANDOM:
4567 	case POSIX_FADV_NOREUSE:
4568 		new = malloc(sizeof(*fa), M_FADVISE, M_WAITOK);
4569 		break;
4570 	case POSIX_FADV_NORMAL:
4571 	case POSIX_FADV_WILLNEED:
4572 	case POSIX_FADV_DONTNEED:
4573 		new = NULL;
4574 		break;
4575 	default:
4576 		return (EINVAL);
4577 	}
4578 	/* XXX: CAP_POSIX_FADVISE? */
4579 	error = fget(td, fd, cap_rights_init(&rights), &fp);
4580 	if (error != 0)
4581 		goto out;
4582 	if ((fp->f_ops->fo_flags & DFLAG_SEEKABLE) == 0) {
4583 		error = ESPIPE;
4584 		goto out;
4585 	}
4586 	if (fp->f_type != DTYPE_VNODE) {
4587 		error = ENODEV;
4588 		goto out;
4589 	}
4590 	vp = fp->f_vnode;
4591 	if (vp->v_type != VREG) {
4592 		error = ENODEV;
4593 		goto out;
4594 	}
4595 	if (len == 0)
4596 		end = OFF_MAX;
4597 	else
4598 		end = offset + len - 1;
4599 	switch (advice) {
4600 	case POSIX_FADV_SEQUENTIAL:
4601 	case POSIX_FADV_RANDOM:
4602 	case POSIX_FADV_NOREUSE:
4603 		/*
4604 		 * Try to merge any existing non-standard region with
4605 		 * this new region if possible, otherwise create a new
4606 		 * non-standard region for this request.
4607 		 */
4608 		mtx_pool_lock(mtxpool_sleep, fp);
4609 		fa = fp->f_advice;
4610 		if (fa != NULL && fa->fa_advice == advice &&
4611 		    ((fa->fa_start <= end && fa->fa_end >= offset) ||
4612 		    (end != OFF_MAX && fa->fa_start == end + 1) ||
4613 		    (fa->fa_end != OFF_MAX && fa->fa_end + 1 == offset))) {
4614 			if (offset < fa->fa_start)
4615 				fa->fa_start = offset;
4616 			if (end > fa->fa_end)
4617 				fa->fa_end = end;
4618 		} else {
4619 			new->fa_advice = advice;
4620 			new->fa_start = offset;
4621 			new->fa_end = end;
4622 			fp->f_advice = new;
4623 			new = fa;
4624 		}
4625 		mtx_pool_unlock(mtxpool_sleep, fp);
4626 		break;
4627 	case POSIX_FADV_NORMAL:
4628 		/*
4629 		 * If a the "normal" region overlaps with an existing
4630 		 * non-standard region, trim or remove the
4631 		 * non-standard region.
4632 		 */
4633 		mtx_pool_lock(mtxpool_sleep, fp);
4634 		fa = fp->f_advice;
4635 		if (fa != NULL) {
4636 			if (offset <= fa->fa_start && end >= fa->fa_end) {
4637 				new = fa;
4638 				fp->f_advice = NULL;
4639 			} else if (offset <= fa->fa_start &&
4640 			    end >= fa->fa_start)
4641 				fa->fa_start = end + 1;
4642 			else if (offset <= fa->fa_end && end >= fa->fa_end)
4643 				fa->fa_end = offset - 1;
4644 			else if (offset >= fa->fa_start && end <= fa->fa_end) {
4645 				/*
4646 				 * If the "normal" region is a middle
4647 				 * portion of the existing
4648 				 * non-standard region, just remove
4649 				 * the whole thing rather than picking
4650 				 * one side or the other to
4651 				 * preserve.
4652 				 */
4653 				new = fa;
4654 				fp->f_advice = NULL;
4655 			}
4656 		}
4657 		mtx_pool_unlock(mtxpool_sleep, fp);
4658 		break;
4659 	case POSIX_FADV_WILLNEED:
4660 	case POSIX_FADV_DONTNEED:
4661 		error = VOP_ADVISE(vp, offset, end, advice);
4662 		break;
4663 	}
4664 out:
4665 	if (fp != NULL)
4666 		fdrop(fp, td);
4667 	free(new, M_FADVISE);
4668 	return (error);
4669 }
4670 
4671 int
4672 sys_posix_fadvise(struct thread *td, struct posix_fadvise_args *uap)
4673 {
4674 	int error;
4675 
4676 	error = kern_posix_fadvise(td, uap->fd, uap->offset, uap->len,
4677 	    uap->advice);
4678 	return (kern_posix_error(td, error));
4679 }
4680