xref: /freebsd/sys/kern/vfs_syscalls.c (revision 2f1217877e155a8a7de97e74eb7e82096a5ef316)
1 /*-
2  * Copyright (c) 1989, 1993
3  *	The Regents of the University of California.  All rights reserved.
4  * (c) UNIX System Laboratories, Inc.
5  * All or some portions of this file are derived from material licensed
6  * to the University of California by American Telephone and Telegraph
7  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
8  * the permission of UNIX System Laboratories, Inc.
9  *
10  * Redistribution and use in source and binary forms, with or without
11  * modification, are permitted provided that the following conditions
12  * are met:
13  * 1. Redistributions of source code must retain the above copyright
14  *    notice, this list of conditions and the following disclaimer.
15  * 2. Redistributions in binary form must reproduce the above copyright
16  *    notice, this list of conditions and the following disclaimer in the
17  *    documentation and/or other materials provided with the distribution.
18  * 4. Neither the name of the University nor the names of its contributors
19  *    may be used to endorse or promote products derived from this software
20  *    without specific prior written permission.
21  *
22  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
23  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
24  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
25  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
26  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
27  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
28  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
29  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
30  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
31  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32  * SUCH DAMAGE.
33  *
34  *	@(#)vfs_syscalls.c	8.13 (Berkeley) 4/15/94
35  */
36 
37 #include <sys/cdefs.h>
38 __FBSDID("$FreeBSD$");
39 
40 #include "opt_capsicum.h"
41 #include "opt_compat.h"
42 #include "opt_ktrace.h"
43 
44 #include <sys/param.h>
45 #include <sys/systm.h>
46 #include <sys/bio.h>
47 #include <sys/buf.h>
48 #include <sys/capsicum.h>
49 #include <sys/disk.h>
50 #include <sys/sysent.h>
51 #include <sys/malloc.h>
52 #include <sys/mount.h>
53 #include <sys/mutex.h>
54 #include <sys/sysproto.h>
55 #include <sys/namei.h>
56 #include <sys/filedesc.h>
57 #include <sys/kernel.h>
58 #include <sys/fcntl.h>
59 #include <sys/file.h>
60 #include <sys/filio.h>
61 #include <sys/limits.h>
62 #include <sys/linker.h>
63 #include <sys/rwlock.h>
64 #include <sys/sdt.h>
65 #include <sys/stat.h>
66 #include <sys/sx.h>
67 #include <sys/unistd.h>
68 #include <sys/vnode.h>
69 #include <sys/priv.h>
70 #include <sys/proc.h>
71 #include <sys/dirent.h>
72 #include <sys/jail.h>
73 #include <sys/syscallsubr.h>
74 #include <sys/sysctl.h>
75 #ifdef KTRACE
76 #include <sys/ktrace.h>
77 #endif
78 
79 #include <machine/stdarg.h>
80 
81 #include <security/audit/audit.h>
82 #include <security/mac/mac_framework.h>
83 
84 #include <vm/vm.h>
85 #include <vm/vm_object.h>
86 #include <vm/vm_page.h>
87 #include <vm/uma.h>
88 
89 #include <ufs/ufs/quota.h>
90 
91 MALLOC_DEFINE(M_FADVISE, "fadvise", "posix_fadvise(2) information");
92 
93 SDT_PROVIDER_DEFINE(vfs);
94 SDT_PROBE_DEFINE2(vfs, , stat, mode, "char *", "int");
95 SDT_PROBE_DEFINE2(vfs, , stat, reg, "char *", "int");
96 
97 static int chroot_refuse_vdir_fds(struct filedesc *fdp);
98 static int getutimes(const struct timeval *, enum uio_seg, struct timespec *);
99 static int kern_chflags(struct thread *td, const char *path,
100     enum uio_seg pathseg, u_long flags);
101 static int kern_chflagsat(struct thread *td, int fd, const char *path,
102     enum uio_seg pathseg, u_long flags, int atflag);
103 static int setfflags(struct thread *td, struct vnode *, u_long);
104 static int setutimes(struct thread *td, struct vnode *,
105     const struct timespec *, int, int);
106 static int vn_access(struct vnode *vp, int user_flags, struct ucred *cred,
107     struct thread *td);
108 
109 /*
110  * The module initialization routine for POSIX asynchronous I/O will
111  * set this to the version of AIO that it implements.  (Zero means
112  * that it is not implemented.)  This value is used here by pathconf()
113  * and in kern_descrip.c by fpathconf().
114  */
115 int async_io_version;
116 
117 #ifdef DEBUG
118 static int syncprt = 0;
119 SYSCTL_INT(_debug, OID_AUTO, syncprt, CTLFLAG_RW, &syncprt, 0, "");
120 #endif
121 
122 /*
123  * Sync each mounted filesystem.
124  */
125 #ifndef _SYS_SYSPROTO_H_
126 struct sync_args {
127 	int     dummy;
128 };
129 #endif
130 /* ARGSUSED */
131 int
132 sys_sync(td, uap)
133 	struct thread *td;
134 	struct sync_args *uap;
135 {
136 	struct mount *mp, *nmp;
137 	int save;
138 
139 	mtx_lock(&mountlist_mtx);
140 	for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
141 		if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK)) {
142 			nmp = TAILQ_NEXT(mp, mnt_list);
143 			continue;
144 		}
145 		if ((mp->mnt_flag & MNT_RDONLY) == 0 &&
146 		    vn_start_write(NULL, &mp, V_NOWAIT) == 0) {
147 			save = curthread_pflags_set(TDP_SYNCIO);
148 			vfs_msync(mp, MNT_NOWAIT);
149 			VFS_SYNC(mp, MNT_NOWAIT);
150 			curthread_pflags_restore(save);
151 			vn_finished_write(mp);
152 		}
153 		mtx_lock(&mountlist_mtx);
154 		nmp = TAILQ_NEXT(mp, mnt_list);
155 		vfs_unbusy(mp);
156 	}
157 	mtx_unlock(&mountlist_mtx);
158 	return (0);
159 }
160 
161 /*
162  * Change filesystem quotas.
163  */
164 #ifndef _SYS_SYSPROTO_H_
165 struct quotactl_args {
166 	char *path;
167 	int cmd;
168 	int uid;
169 	caddr_t arg;
170 };
171 #endif
172 int
173 sys_quotactl(td, uap)
174 	struct thread *td;
175 	register struct quotactl_args /* {
176 		char *path;
177 		int cmd;
178 		int uid;
179 		caddr_t arg;
180 	} */ *uap;
181 {
182 	struct mount *mp;
183 	struct nameidata nd;
184 	int error;
185 
186 	AUDIT_ARG_CMD(uap->cmd);
187 	AUDIT_ARG_UID(uap->uid);
188 	if (!prison_allow(td->td_ucred, PR_ALLOW_QUOTAS))
189 		return (EPERM);
190 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
191 	    uap->path, td);
192 	if ((error = namei(&nd)) != 0)
193 		return (error);
194 	NDFREE(&nd, NDF_ONLY_PNBUF);
195 	mp = nd.ni_vp->v_mount;
196 	vfs_ref(mp);
197 	vput(nd.ni_vp);
198 	error = vfs_busy(mp, 0);
199 	vfs_rel(mp);
200 	if (error != 0)
201 		return (error);
202 	error = VFS_QUOTACTL(mp, uap->cmd, uap->uid, uap->arg);
203 
204 	/*
205 	 * Since quota on operation typically needs to open quota
206 	 * file, the Q_QUOTAON handler needs to unbusy the mount point
207 	 * before calling into namei.  Otherwise, unmount might be
208 	 * started between two vfs_busy() invocations (first is our,
209 	 * second is from mount point cross-walk code in lookup()),
210 	 * causing deadlock.
211 	 *
212 	 * Require that Q_QUOTAON handles the vfs_busy() reference on
213 	 * its own, always returning with ubusied mount point.
214 	 */
215 	if ((uap->cmd >> SUBCMDSHIFT) != Q_QUOTAON)
216 		vfs_unbusy(mp);
217 	return (error);
218 }
219 
220 /*
221  * Used by statfs conversion routines to scale the block size up if
222  * necessary so that all of the block counts are <= 'max_size'.  Note
223  * that 'max_size' should be a bitmask, i.e. 2^n - 1 for some non-zero
224  * value of 'n'.
225  */
226 void
227 statfs_scale_blocks(struct statfs *sf, long max_size)
228 {
229 	uint64_t count;
230 	int shift;
231 
232 	KASSERT(powerof2(max_size + 1), ("%s: invalid max_size", __func__));
233 
234 	/*
235 	 * Attempt to scale the block counts to give a more accurate
236 	 * overview to userland of the ratio of free space to used
237 	 * space.  To do this, find the largest block count and compute
238 	 * a divisor that lets it fit into a signed integer <= max_size.
239 	 */
240 	if (sf->f_bavail < 0)
241 		count = -sf->f_bavail;
242 	else
243 		count = sf->f_bavail;
244 	count = MAX(sf->f_blocks, MAX(sf->f_bfree, count));
245 	if (count <= max_size)
246 		return;
247 
248 	count >>= flsl(max_size);
249 	shift = 0;
250 	while (count > 0) {
251 		shift++;
252 		count >>=1;
253 	}
254 
255 	sf->f_bsize <<= shift;
256 	sf->f_blocks >>= shift;
257 	sf->f_bfree >>= shift;
258 	sf->f_bavail >>= shift;
259 }
260 
261 /*
262  * Get filesystem statistics.
263  */
264 #ifndef _SYS_SYSPROTO_H_
265 struct statfs_args {
266 	char *path;
267 	struct statfs *buf;
268 };
269 #endif
270 int
271 sys_statfs(td, uap)
272 	struct thread *td;
273 	register struct statfs_args /* {
274 		char *path;
275 		struct statfs *buf;
276 	} */ *uap;
277 {
278 	struct statfs sf;
279 	int error;
280 
281 	error = kern_statfs(td, uap->path, UIO_USERSPACE, &sf);
282 	if (error == 0)
283 		error = copyout(&sf, uap->buf, sizeof(sf));
284 	return (error);
285 }
286 
287 int
288 kern_statfs(struct thread *td, char *path, enum uio_seg pathseg,
289     struct statfs *buf)
290 {
291 	struct mount *mp;
292 	struct statfs *sp, sb;
293 	struct nameidata nd;
294 	int error;
295 
296 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
297 	    pathseg, path, td);
298 	error = namei(&nd);
299 	if (error != 0)
300 		return (error);
301 	mp = nd.ni_vp->v_mount;
302 	vfs_ref(mp);
303 	NDFREE(&nd, NDF_ONLY_PNBUF);
304 	vput(nd.ni_vp);
305 	error = vfs_busy(mp, 0);
306 	vfs_rel(mp);
307 	if (error != 0)
308 		return (error);
309 #ifdef MAC
310 	error = mac_mount_check_stat(td->td_ucred, mp);
311 	if (error != 0)
312 		goto out;
313 #endif
314 	/*
315 	 * Set these in case the underlying filesystem fails to do so.
316 	 */
317 	sp = &mp->mnt_stat;
318 	sp->f_version = STATFS_VERSION;
319 	sp->f_namemax = NAME_MAX;
320 	sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
321 	error = VFS_STATFS(mp, sp);
322 	if (error != 0)
323 		goto out;
324 	if (priv_check(td, PRIV_VFS_GENERATION)) {
325 		bcopy(sp, &sb, sizeof(sb));
326 		sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
327 		prison_enforce_statfs(td->td_ucred, mp, &sb);
328 		sp = &sb;
329 	}
330 	*buf = *sp;
331 out:
332 	vfs_unbusy(mp);
333 	return (error);
334 }
335 
336 /*
337  * Get filesystem statistics.
338  */
339 #ifndef _SYS_SYSPROTO_H_
340 struct fstatfs_args {
341 	int fd;
342 	struct statfs *buf;
343 };
344 #endif
345 int
346 sys_fstatfs(td, uap)
347 	struct thread *td;
348 	register struct fstatfs_args /* {
349 		int fd;
350 		struct statfs *buf;
351 	} */ *uap;
352 {
353 	struct statfs sf;
354 	int error;
355 
356 	error = kern_fstatfs(td, uap->fd, &sf);
357 	if (error == 0)
358 		error = copyout(&sf, uap->buf, sizeof(sf));
359 	return (error);
360 }
361 
362 int
363 kern_fstatfs(struct thread *td, int fd, struct statfs *buf)
364 {
365 	struct file *fp;
366 	struct mount *mp;
367 	struct statfs *sp, sb;
368 	struct vnode *vp;
369 	cap_rights_t rights;
370 	int error;
371 
372 	AUDIT_ARG_FD(fd);
373 	error = getvnode(td->td_proc->p_fd, fd,
374 	    cap_rights_init(&rights, CAP_FSTATFS), &fp);
375 	if (error != 0)
376 		return (error);
377 	vp = fp->f_vnode;
378 	vn_lock(vp, LK_SHARED | LK_RETRY);
379 #ifdef AUDIT
380 	AUDIT_ARG_VNODE1(vp);
381 #endif
382 	mp = vp->v_mount;
383 	if (mp)
384 		vfs_ref(mp);
385 	VOP_UNLOCK(vp, 0);
386 	fdrop(fp, td);
387 	if (mp == NULL) {
388 		error = EBADF;
389 		goto out;
390 	}
391 	error = vfs_busy(mp, 0);
392 	vfs_rel(mp);
393 	if (error != 0)
394 		return (error);
395 #ifdef MAC
396 	error = mac_mount_check_stat(td->td_ucred, mp);
397 	if (error != 0)
398 		goto out;
399 #endif
400 	/*
401 	 * Set these in case the underlying filesystem fails to do so.
402 	 */
403 	sp = &mp->mnt_stat;
404 	sp->f_version = STATFS_VERSION;
405 	sp->f_namemax = NAME_MAX;
406 	sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
407 	error = VFS_STATFS(mp, sp);
408 	if (error != 0)
409 		goto out;
410 	if (priv_check(td, PRIV_VFS_GENERATION)) {
411 		bcopy(sp, &sb, sizeof(sb));
412 		sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
413 		prison_enforce_statfs(td->td_ucred, mp, &sb);
414 		sp = &sb;
415 	}
416 	*buf = *sp;
417 out:
418 	if (mp)
419 		vfs_unbusy(mp);
420 	return (error);
421 }
422 
423 /*
424  * Get statistics on all filesystems.
425  */
426 #ifndef _SYS_SYSPROTO_H_
427 struct getfsstat_args {
428 	struct statfs *buf;
429 	long bufsize;
430 	int flags;
431 };
432 #endif
433 int
434 sys_getfsstat(td, uap)
435 	struct thread *td;
436 	register struct getfsstat_args /* {
437 		struct statfs *buf;
438 		long bufsize;
439 		int flags;
440 	} */ *uap;
441 {
442 
443 	return (kern_getfsstat(td, &uap->buf, uap->bufsize, UIO_USERSPACE,
444 	    uap->flags));
445 }
446 
447 /*
448  * If (bufsize > 0 && bufseg == UIO_SYSSPACE)
449  *	The caller is responsible for freeing memory which will be allocated
450  *	in '*buf'.
451  */
452 int
453 kern_getfsstat(struct thread *td, struct statfs **buf, size_t bufsize,
454     enum uio_seg bufseg, int flags)
455 {
456 	struct mount *mp, *nmp;
457 	struct statfs *sfsp, *sp, sb;
458 	size_t count, maxcount;
459 	int error;
460 
461 	maxcount = bufsize / sizeof(struct statfs);
462 	if (bufsize == 0)
463 		sfsp = NULL;
464 	else if (bufseg == UIO_USERSPACE)
465 		sfsp = *buf;
466 	else /* if (bufseg == UIO_SYSSPACE) */ {
467 		count = 0;
468 		mtx_lock(&mountlist_mtx);
469 		TAILQ_FOREACH(mp, &mountlist, mnt_list) {
470 			count++;
471 		}
472 		mtx_unlock(&mountlist_mtx);
473 		if (maxcount > count)
474 			maxcount = count;
475 		sfsp = *buf = malloc(maxcount * sizeof(struct statfs), M_TEMP,
476 		    M_WAITOK);
477 	}
478 	count = 0;
479 	mtx_lock(&mountlist_mtx);
480 	for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
481 		if (prison_canseemount(td->td_ucred, mp) != 0) {
482 			nmp = TAILQ_NEXT(mp, mnt_list);
483 			continue;
484 		}
485 #ifdef MAC
486 		if (mac_mount_check_stat(td->td_ucred, mp) != 0) {
487 			nmp = TAILQ_NEXT(mp, mnt_list);
488 			continue;
489 		}
490 #endif
491 		if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK)) {
492 			nmp = TAILQ_NEXT(mp, mnt_list);
493 			continue;
494 		}
495 		if (sfsp && count < maxcount) {
496 			sp = &mp->mnt_stat;
497 			/*
498 			 * Set these in case the underlying filesystem
499 			 * fails to do so.
500 			 */
501 			sp->f_version = STATFS_VERSION;
502 			sp->f_namemax = NAME_MAX;
503 			sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
504 			/*
505 			 * If MNT_NOWAIT or MNT_LAZY is specified, do not
506 			 * refresh the fsstat cache. MNT_NOWAIT or MNT_LAZY
507 			 * overrides MNT_WAIT.
508 			 */
509 			if (((flags & (MNT_LAZY|MNT_NOWAIT)) == 0 ||
510 			    (flags & MNT_WAIT)) &&
511 			    (error = VFS_STATFS(mp, sp))) {
512 				mtx_lock(&mountlist_mtx);
513 				nmp = TAILQ_NEXT(mp, mnt_list);
514 				vfs_unbusy(mp);
515 				continue;
516 			}
517 			if (priv_check(td, PRIV_VFS_GENERATION)) {
518 				bcopy(sp, &sb, sizeof(sb));
519 				sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
520 				prison_enforce_statfs(td->td_ucred, mp, &sb);
521 				sp = &sb;
522 			}
523 			if (bufseg == UIO_SYSSPACE)
524 				bcopy(sp, sfsp, sizeof(*sp));
525 			else /* if (bufseg == UIO_USERSPACE) */ {
526 				error = copyout(sp, sfsp, sizeof(*sp));
527 				if (error != 0) {
528 					vfs_unbusy(mp);
529 					return (error);
530 				}
531 			}
532 			sfsp++;
533 		}
534 		count++;
535 		mtx_lock(&mountlist_mtx);
536 		nmp = TAILQ_NEXT(mp, mnt_list);
537 		vfs_unbusy(mp);
538 	}
539 	mtx_unlock(&mountlist_mtx);
540 	if (sfsp && count > maxcount)
541 		td->td_retval[0] = maxcount;
542 	else
543 		td->td_retval[0] = count;
544 	return (0);
545 }
546 
547 #ifdef COMPAT_FREEBSD4
548 /*
549  * Get old format filesystem statistics.
550  */
551 static void cvtstatfs(struct statfs *, struct ostatfs *);
552 
553 #ifndef _SYS_SYSPROTO_H_
554 struct freebsd4_statfs_args {
555 	char *path;
556 	struct ostatfs *buf;
557 };
558 #endif
559 int
560 freebsd4_statfs(td, uap)
561 	struct thread *td;
562 	struct freebsd4_statfs_args /* {
563 		char *path;
564 		struct ostatfs *buf;
565 	} */ *uap;
566 {
567 	struct ostatfs osb;
568 	struct statfs sf;
569 	int error;
570 
571 	error = kern_statfs(td, uap->path, UIO_USERSPACE, &sf);
572 	if (error != 0)
573 		return (error);
574 	cvtstatfs(&sf, &osb);
575 	return (copyout(&osb, uap->buf, sizeof(osb)));
576 }
577 
578 /*
579  * Get filesystem statistics.
580  */
581 #ifndef _SYS_SYSPROTO_H_
582 struct freebsd4_fstatfs_args {
583 	int fd;
584 	struct ostatfs *buf;
585 };
586 #endif
587 int
588 freebsd4_fstatfs(td, uap)
589 	struct thread *td;
590 	struct freebsd4_fstatfs_args /* {
591 		int fd;
592 		struct ostatfs *buf;
593 	} */ *uap;
594 {
595 	struct ostatfs osb;
596 	struct statfs sf;
597 	int error;
598 
599 	error = kern_fstatfs(td, uap->fd, &sf);
600 	if (error != 0)
601 		return (error);
602 	cvtstatfs(&sf, &osb);
603 	return (copyout(&osb, uap->buf, sizeof(osb)));
604 }
605 
606 /*
607  * Get statistics on all filesystems.
608  */
609 #ifndef _SYS_SYSPROTO_H_
610 struct freebsd4_getfsstat_args {
611 	struct ostatfs *buf;
612 	long bufsize;
613 	int flags;
614 };
615 #endif
616 int
617 freebsd4_getfsstat(td, uap)
618 	struct thread *td;
619 	register struct freebsd4_getfsstat_args /* {
620 		struct ostatfs *buf;
621 		long bufsize;
622 		int flags;
623 	} */ *uap;
624 {
625 	struct statfs *buf, *sp;
626 	struct ostatfs osb;
627 	size_t count, size;
628 	int error;
629 
630 	count = uap->bufsize / sizeof(struct ostatfs);
631 	size = count * sizeof(struct statfs);
632 	error = kern_getfsstat(td, &buf, size, UIO_SYSSPACE, uap->flags);
633 	if (size > 0) {
634 		count = td->td_retval[0];
635 		sp = buf;
636 		while (count > 0 && error == 0) {
637 			cvtstatfs(sp, &osb);
638 			error = copyout(&osb, uap->buf, sizeof(osb));
639 			sp++;
640 			uap->buf++;
641 			count--;
642 		}
643 		free(buf, M_TEMP);
644 	}
645 	return (error);
646 }
647 
648 /*
649  * Implement fstatfs() for (NFS) file handles.
650  */
651 #ifndef _SYS_SYSPROTO_H_
652 struct freebsd4_fhstatfs_args {
653 	struct fhandle *u_fhp;
654 	struct ostatfs *buf;
655 };
656 #endif
657 int
658 freebsd4_fhstatfs(td, uap)
659 	struct thread *td;
660 	struct freebsd4_fhstatfs_args /* {
661 		struct fhandle *u_fhp;
662 		struct ostatfs *buf;
663 	} */ *uap;
664 {
665 	struct ostatfs osb;
666 	struct statfs sf;
667 	fhandle_t fh;
668 	int error;
669 
670 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
671 	if (error != 0)
672 		return (error);
673 	error = kern_fhstatfs(td, fh, &sf);
674 	if (error != 0)
675 		return (error);
676 	cvtstatfs(&sf, &osb);
677 	return (copyout(&osb, uap->buf, sizeof(osb)));
678 }
679 
680 /*
681  * Convert a new format statfs structure to an old format statfs structure.
682  */
683 static void
684 cvtstatfs(nsp, osp)
685 	struct statfs *nsp;
686 	struct ostatfs *osp;
687 {
688 
689 	statfs_scale_blocks(nsp, LONG_MAX);
690 	bzero(osp, sizeof(*osp));
691 	osp->f_bsize = nsp->f_bsize;
692 	osp->f_iosize = MIN(nsp->f_iosize, LONG_MAX);
693 	osp->f_blocks = nsp->f_blocks;
694 	osp->f_bfree = nsp->f_bfree;
695 	osp->f_bavail = nsp->f_bavail;
696 	osp->f_files = MIN(nsp->f_files, LONG_MAX);
697 	osp->f_ffree = MIN(nsp->f_ffree, LONG_MAX);
698 	osp->f_owner = nsp->f_owner;
699 	osp->f_type = nsp->f_type;
700 	osp->f_flags = nsp->f_flags;
701 	osp->f_syncwrites = MIN(nsp->f_syncwrites, LONG_MAX);
702 	osp->f_asyncwrites = MIN(nsp->f_asyncwrites, LONG_MAX);
703 	osp->f_syncreads = MIN(nsp->f_syncreads, LONG_MAX);
704 	osp->f_asyncreads = MIN(nsp->f_asyncreads, LONG_MAX);
705 	strlcpy(osp->f_fstypename, nsp->f_fstypename,
706 	    MIN(MFSNAMELEN, OMFSNAMELEN));
707 	strlcpy(osp->f_mntonname, nsp->f_mntonname,
708 	    MIN(MNAMELEN, OMNAMELEN));
709 	strlcpy(osp->f_mntfromname, nsp->f_mntfromname,
710 	    MIN(MNAMELEN, OMNAMELEN));
711 	osp->f_fsid = nsp->f_fsid;
712 }
713 #endif /* COMPAT_FREEBSD4 */
714 
715 /*
716  * Change current working directory to a given file descriptor.
717  */
718 #ifndef _SYS_SYSPROTO_H_
719 struct fchdir_args {
720 	int	fd;
721 };
722 #endif
723 int
724 sys_fchdir(td, uap)
725 	struct thread *td;
726 	struct fchdir_args /* {
727 		int fd;
728 	} */ *uap;
729 {
730 	register struct filedesc *fdp = td->td_proc->p_fd;
731 	struct vnode *vp, *tdp, *vpold;
732 	struct mount *mp;
733 	struct file *fp;
734 	cap_rights_t rights;
735 	int error;
736 
737 	AUDIT_ARG_FD(uap->fd);
738 	error = getvnode(fdp, uap->fd, cap_rights_init(&rights, CAP_FCHDIR),
739 	    &fp);
740 	if (error != 0)
741 		return (error);
742 	vp = fp->f_vnode;
743 	VREF(vp);
744 	fdrop(fp, td);
745 	vn_lock(vp, LK_SHARED | LK_RETRY);
746 	AUDIT_ARG_VNODE1(vp);
747 	error = change_dir(vp, td);
748 	while (!error && (mp = vp->v_mountedhere) != NULL) {
749 		if (vfs_busy(mp, 0))
750 			continue;
751 		error = VFS_ROOT(mp, LK_SHARED, &tdp);
752 		vfs_unbusy(mp);
753 		if (error != 0)
754 			break;
755 		vput(vp);
756 		vp = tdp;
757 	}
758 	if (error != 0) {
759 		vput(vp);
760 		return (error);
761 	}
762 	VOP_UNLOCK(vp, 0);
763 	FILEDESC_XLOCK(fdp);
764 	vpold = fdp->fd_cdir;
765 	fdp->fd_cdir = vp;
766 	FILEDESC_XUNLOCK(fdp);
767 	vrele(vpold);
768 	return (0);
769 }
770 
771 /*
772  * Change current working directory (``.'').
773  */
774 #ifndef _SYS_SYSPROTO_H_
775 struct chdir_args {
776 	char	*path;
777 };
778 #endif
779 int
780 sys_chdir(td, uap)
781 	struct thread *td;
782 	struct chdir_args /* {
783 		char *path;
784 	} */ *uap;
785 {
786 
787 	return (kern_chdir(td, uap->path, UIO_USERSPACE));
788 }
789 
790 int
791 kern_chdir(struct thread *td, char *path, enum uio_seg pathseg)
792 {
793 	register struct filedesc *fdp = td->td_proc->p_fd;
794 	struct nameidata nd;
795 	struct vnode *vp;
796 	int error;
797 
798 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
799 	    pathseg, path, td);
800 	if ((error = namei(&nd)) != 0)
801 		return (error);
802 	if ((error = change_dir(nd.ni_vp, td)) != 0) {
803 		vput(nd.ni_vp);
804 		NDFREE(&nd, NDF_ONLY_PNBUF);
805 		return (error);
806 	}
807 	VOP_UNLOCK(nd.ni_vp, 0);
808 	NDFREE(&nd, NDF_ONLY_PNBUF);
809 	FILEDESC_XLOCK(fdp);
810 	vp = fdp->fd_cdir;
811 	fdp->fd_cdir = nd.ni_vp;
812 	FILEDESC_XUNLOCK(fdp);
813 	vrele(vp);
814 	return (0);
815 }
816 
817 /*
818  * Helper function for raised chroot(2) security function:  Refuse if
819  * any filedescriptors are open directories.
820  */
821 static int
822 chroot_refuse_vdir_fds(fdp)
823 	struct filedesc *fdp;
824 {
825 	struct vnode *vp;
826 	struct file *fp;
827 	int fd;
828 
829 	FILEDESC_LOCK_ASSERT(fdp);
830 
831 	for (fd = 0; fd <= fdp->fd_lastfile; fd++) {
832 		fp = fget_locked(fdp, fd);
833 		if (fp == NULL)
834 			continue;
835 		if (fp->f_type == DTYPE_VNODE) {
836 			vp = fp->f_vnode;
837 			if (vp->v_type == VDIR)
838 				return (EPERM);
839 		}
840 	}
841 	return (0);
842 }
843 
844 /*
845  * This sysctl determines if we will allow a process to chroot(2) if it
846  * has a directory open:
847  *	0: disallowed for all processes.
848  *	1: allowed for processes that were not already chroot(2)'ed.
849  *	2: allowed for all processes.
850  */
851 
852 static int chroot_allow_open_directories = 1;
853 
854 SYSCTL_INT(_kern, OID_AUTO, chroot_allow_open_directories, CTLFLAG_RW,
855      &chroot_allow_open_directories, 0,
856      "Allow a process to chroot(2) if it has a directory open");
857 
858 /*
859  * Change notion of root (``/'') directory.
860  */
861 #ifndef _SYS_SYSPROTO_H_
862 struct chroot_args {
863 	char	*path;
864 };
865 #endif
866 int
867 sys_chroot(td, uap)
868 	struct thread *td;
869 	struct chroot_args /* {
870 		char *path;
871 	} */ *uap;
872 {
873 	struct nameidata nd;
874 	int error;
875 
876 	error = priv_check(td, PRIV_VFS_CHROOT);
877 	if (error != 0)
878 		return (error);
879 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
880 	    UIO_USERSPACE, uap->path, td);
881 	error = namei(&nd);
882 	if (error != 0)
883 		goto error;
884 	error = change_dir(nd.ni_vp, td);
885 	if (error != 0)
886 		goto e_vunlock;
887 #ifdef MAC
888 	error = mac_vnode_check_chroot(td->td_ucred, nd.ni_vp);
889 	if (error != 0)
890 		goto e_vunlock;
891 #endif
892 	VOP_UNLOCK(nd.ni_vp, 0);
893 	error = change_root(nd.ni_vp, td);
894 	vrele(nd.ni_vp);
895 	NDFREE(&nd, NDF_ONLY_PNBUF);
896 	return (error);
897 e_vunlock:
898 	vput(nd.ni_vp);
899 error:
900 	NDFREE(&nd, NDF_ONLY_PNBUF);
901 	return (error);
902 }
903 
904 /*
905  * Common routine for chroot and chdir.  Callers must provide a locked vnode
906  * instance.
907  */
908 int
909 change_dir(vp, td)
910 	struct vnode *vp;
911 	struct thread *td;
912 {
913 #ifdef MAC
914 	int error;
915 #endif
916 
917 	ASSERT_VOP_LOCKED(vp, "change_dir(): vp not locked");
918 	if (vp->v_type != VDIR)
919 		return (ENOTDIR);
920 #ifdef MAC
921 	error = mac_vnode_check_chdir(td->td_ucred, vp);
922 	if (error != 0)
923 		return (error);
924 #endif
925 	return (VOP_ACCESS(vp, VEXEC, td->td_ucred, td));
926 }
927 
928 /*
929  * Common routine for kern_chroot() and jail_attach().  The caller is
930  * responsible for invoking priv_check() and mac_vnode_check_chroot() to
931  * authorize this operation.
932  */
933 int
934 change_root(vp, td)
935 	struct vnode *vp;
936 	struct thread *td;
937 {
938 	struct filedesc *fdp;
939 	struct vnode *oldvp;
940 	int error;
941 
942 	fdp = td->td_proc->p_fd;
943 	FILEDESC_XLOCK(fdp);
944 	if (chroot_allow_open_directories == 0 ||
945 	    (chroot_allow_open_directories == 1 && fdp->fd_rdir != rootvnode)) {
946 		error = chroot_refuse_vdir_fds(fdp);
947 		if (error != 0) {
948 			FILEDESC_XUNLOCK(fdp);
949 			return (error);
950 		}
951 	}
952 	oldvp = fdp->fd_rdir;
953 	fdp->fd_rdir = vp;
954 	VREF(fdp->fd_rdir);
955 	if (!fdp->fd_jdir) {
956 		fdp->fd_jdir = vp;
957 		VREF(fdp->fd_jdir);
958 	}
959 	FILEDESC_XUNLOCK(fdp);
960 	vrele(oldvp);
961 	return (0);
962 }
963 
964 static __inline void
965 flags_to_rights(int flags, cap_rights_t *rightsp)
966 {
967 
968 	if (flags & O_EXEC) {
969 		cap_rights_set(rightsp, CAP_FEXECVE);
970 	} else {
971 		switch ((flags & O_ACCMODE)) {
972 		case O_RDONLY:
973 			cap_rights_set(rightsp, CAP_READ);
974 			break;
975 		case O_RDWR:
976 			cap_rights_set(rightsp, CAP_READ);
977 			/* FALLTHROUGH */
978 		case O_WRONLY:
979 			cap_rights_set(rightsp, CAP_WRITE);
980 			if (!(flags & (O_APPEND | O_TRUNC)))
981 				cap_rights_set(rightsp, CAP_SEEK);
982 			break;
983 		}
984 	}
985 
986 	if (flags & O_CREAT)
987 		cap_rights_set(rightsp, CAP_CREATE);
988 
989 	if (flags & O_TRUNC)
990 		cap_rights_set(rightsp, CAP_FTRUNCATE);
991 
992 	if (flags & (O_SYNC | O_FSYNC))
993 		cap_rights_set(rightsp, CAP_FSYNC);
994 
995 	if (flags & (O_EXLOCK | O_SHLOCK))
996 		cap_rights_set(rightsp, CAP_FLOCK);
997 }
998 
999 /*
1000  * Check permissions, allocate an open file structure, and call the device
1001  * open routine if any.
1002  */
1003 #ifndef _SYS_SYSPROTO_H_
1004 struct open_args {
1005 	char	*path;
1006 	int	flags;
1007 	int	mode;
1008 };
1009 #endif
1010 int
1011 sys_open(td, uap)
1012 	struct thread *td;
1013 	register struct open_args /* {
1014 		char *path;
1015 		int flags;
1016 		int mode;
1017 	} */ *uap;
1018 {
1019 
1020 	return (kern_open(td, uap->path, UIO_USERSPACE, uap->flags, uap->mode));
1021 }
1022 
1023 #ifndef _SYS_SYSPROTO_H_
1024 struct openat_args {
1025 	int	fd;
1026 	char	*path;
1027 	int	flag;
1028 	int	mode;
1029 };
1030 #endif
1031 int
1032 sys_openat(struct thread *td, struct openat_args *uap)
1033 {
1034 
1035 	return (kern_openat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
1036 	    uap->mode));
1037 }
1038 
1039 int
1040 kern_open(struct thread *td, char *path, enum uio_seg pathseg, int flags,
1041     int mode)
1042 {
1043 
1044 	return (kern_openat(td, AT_FDCWD, path, pathseg, flags, mode));
1045 }
1046 
1047 int
1048 kern_openat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1049     int flags, int mode)
1050 {
1051 	struct proc *p = td->td_proc;
1052 	struct filedesc *fdp = p->p_fd;
1053 	struct file *fp;
1054 	struct vnode *vp;
1055 	struct nameidata nd;
1056 	cap_rights_t rights;
1057 	int cmode, error, indx;
1058 
1059 	indx = -1;
1060 
1061 	AUDIT_ARG_FFLAGS(flags);
1062 	AUDIT_ARG_MODE(mode);
1063 	/* XXX: audit dirfd */
1064 	cap_rights_init(&rights, CAP_LOOKUP);
1065 	flags_to_rights(flags, &rights);
1066 	/*
1067 	 * Only one of the O_EXEC, O_RDONLY, O_WRONLY and O_RDWR flags
1068 	 * may be specified.
1069 	 */
1070 	if (flags & O_EXEC) {
1071 		if (flags & O_ACCMODE)
1072 			return (EINVAL);
1073 	} else if ((flags & O_ACCMODE) == O_ACCMODE) {
1074 		return (EINVAL);
1075 	} else {
1076 		flags = FFLAGS(flags);
1077 	}
1078 
1079 	/*
1080 	 * Allocate the file descriptor, but don't install a descriptor yet.
1081 	 */
1082 	error = falloc_noinstall(td, &fp);
1083 	if (error != 0)
1084 		return (error);
1085 	/*
1086 	 * An extra reference on `fp' has been held for us by
1087 	 * falloc_noinstall().
1088 	 */
1089 	/* Set the flags early so the finit in devfs can pick them up. */
1090 	fp->f_flag = flags & FMASK;
1091 	cmode = ((mode & ~fdp->fd_cmask) & ALLPERMS) & ~S_ISTXT;
1092 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, fd,
1093 	    &rights, td);
1094 	td->td_dupfd = -1;		/* XXX check for fdopen */
1095 	error = vn_open(&nd, &flags, cmode, fp);
1096 	if (error != 0) {
1097 		/*
1098 		 * If the vn_open replaced the method vector, something
1099 		 * wonderous happened deep below and we just pass it up
1100 		 * pretending we know what we do.
1101 		 */
1102 		if (error == ENXIO && fp->f_ops != &badfileops)
1103 			goto success;
1104 
1105 		/*
1106 		 * Handle special fdopen() case. bleh.
1107 		 *
1108 		 * Don't do this for relative (capability) lookups; we don't
1109 		 * understand exactly what would happen, and we don't think
1110 		 * that it ever should.
1111 		 */
1112 		if (nd.ni_strictrelative == 0 &&
1113 		    (error == ENODEV || error == ENXIO) &&
1114 		    td->td_dupfd >= 0) {
1115 			error = dupfdopen(td, fdp, td->td_dupfd, flags, error,
1116 			    &indx);
1117 			if (error == 0)
1118 				goto success;
1119 		}
1120 
1121 		goto bad;
1122 	}
1123 	td->td_dupfd = 0;
1124 	NDFREE(&nd, NDF_ONLY_PNBUF);
1125 	vp = nd.ni_vp;
1126 
1127 	/*
1128 	 * Store the vnode, for any f_type. Typically, the vnode use
1129 	 * count is decremented by direct call to vn_closefile() for
1130 	 * files that switched type in the cdevsw fdopen() method.
1131 	 */
1132 	fp->f_vnode = vp;
1133 	/*
1134 	 * If the file wasn't claimed by devfs bind it to the normal
1135 	 * vnode operations here.
1136 	 */
1137 	if (fp->f_ops == &badfileops) {
1138 		KASSERT(vp->v_type != VFIFO, ("Unexpected fifo."));
1139 		fp->f_seqcount = 1;
1140 		finit(fp, (flags & FMASK) | (fp->f_flag & FHASLOCK),
1141 		    DTYPE_VNODE, vp, &vnops);
1142 	}
1143 
1144 	VOP_UNLOCK(vp, 0);
1145 	if (flags & O_TRUNC) {
1146 		error = fo_truncate(fp, 0, td->td_ucred, td);
1147 		if (error != 0)
1148 			goto bad;
1149 	}
1150 success:
1151 	/*
1152 	 * If we haven't already installed the FD (for dupfdopen), do so now.
1153 	 */
1154 	if (indx == -1) {
1155 		struct filecaps *fcaps;
1156 
1157 #ifdef CAPABILITIES
1158 		if (nd.ni_strictrelative == 1)
1159 			fcaps = &nd.ni_filecaps;
1160 		else
1161 #endif
1162 			fcaps = NULL;
1163 		error = finstall(td, fp, &indx, flags, fcaps);
1164 		/* On success finstall() consumes fcaps. */
1165 		if (error != 0) {
1166 			filecaps_free(&nd.ni_filecaps);
1167 			goto bad;
1168 		}
1169 	} else {
1170 		filecaps_free(&nd.ni_filecaps);
1171 	}
1172 
1173 	/*
1174 	 * Release our private reference, leaving the one associated with
1175 	 * the descriptor table intact.
1176 	 */
1177 	fdrop(fp, td);
1178 	td->td_retval[0] = indx;
1179 	return (0);
1180 bad:
1181 	KASSERT(indx == -1, ("indx=%d, should be -1", indx));
1182 	fdrop(fp, td);
1183 	return (error);
1184 }
1185 
1186 #ifdef COMPAT_43
1187 /*
1188  * Create a file.
1189  */
1190 #ifndef _SYS_SYSPROTO_H_
1191 struct ocreat_args {
1192 	char	*path;
1193 	int	mode;
1194 };
1195 #endif
1196 int
1197 ocreat(td, uap)
1198 	struct thread *td;
1199 	register struct ocreat_args /* {
1200 		char *path;
1201 		int mode;
1202 	} */ *uap;
1203 {
1204 
1205 	return (kern_open(td, uap->path, UIO_USERSPACE,
1206 	    O_WRONLY | O_CREAT | O_TRUNC, uap->mode));
1207 }
1208 #endif /* COMPAT_43 */
1209 
1210 /*
1211  * Create a special file.
1212  */
1213 #ifndef _SYS_SYSPROTO_H_
1214 struct mknod_args {
1215 	char	*path;
1216 	int	mode;
1217 	int	dev;
1218 };
1219 #endif
1220 int
1221 sys_mknod(td, uap)
1222 	struct thread *td;
1223 	register struct mknod_args /* {
1224 		char *path;
1225 		int mode;
1226 		int dev;
1227 	} */ *uap;
1228 {
1229 
1230 	return (kern_mknod(td, uap->path, UIO_USERSPACE, uap->mode, uap->dev));
1231 }
1232 
1233 #ifndef _SYS_SYSPROTO_H_
1234 struct mknodat_args {
1235 	int	fd;
1236 	char	*path;
1237 	mode_t	mode;
1238 	dev_t	dev;
1239 };
1240 #endif
1241 int
1242 sys_mknodat(struct thread *td, struct mknodat_args *uap)
1243 {
1244 
1245 	return (kern_mknodat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode,
1246 	    uap->dev));
1247 }
1248 
1249 int
1250 kern_mknod(struct thread *td, char *path, enum uio_seg pathseg, int mode,
1251     int dev)
1252 {
1253 
1254 	return (kern_mknodat(td, AT_FDCWD, path, pathseg, mode, dev));
1255 }
1256 
1257 int
1258 kern_mknodat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1259     int mode, int dev)
1260 {
1261 	struct vnode *vp;
1262 	struct mount *mp;
1263 	struct vattr vattr;
1264 	struct nameidata nd;
1265 	cap_rights_t rights;
1266 	int error, whiteout = 0;
1267 
1268 	AUDIT_ARG_MODE(mode);
1269 	AUDIT_ARG_DEV(dev);
1270 	switch (mode & S_IFMT) {
1271 	case S_IFCHR:
1272 	case S_IFBLK:
1273 		error = priv_check(td, PRIV_VFS_MKNOD_DEV);
1274 		break;
1275 	case S_IFMT:
1276 		error = priv_check(td, PRIV_VFS_MKNOD_BAD);
1277 		break;
1278 	case S_IFWHT:
1279 		error = priv_check(td, PRIV_VFS_MKNOD_WHT);
1280 		break;
1281 	case S_IFIFO:
1282 		if (dev == 0)
1283 			return (kern_mkfifoat(td, fd, path, pathseg, mode));
1284 		/* FALLTHROUGH */
1285 	default:
1286 		error = EINVAL;
1287 		break;
1288 	}
1289 	if (error != 0)
1290 		return (error);
1291 restart:
1292 	bwillwrite();
1293 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1,
1294 	    pathseg, path, fd, cap_rights_init(&rights, CAP_MKNODAT), td);
1295 	if ((error = namei(&nd)) != 0)
1296 		return (error);
1297 	vp = nd.ni_vp;
1298 	if (vp != NULL) {
1299 		NDFREE(&nd, NDF_ONLY_PNBUF);
1300 		if (vp == nd.ni_dvp)
1301 			vrele(nd.ni_dvp);
1302 		else
1303 			vput(nd.ni_dvp);
1304 		vrele(vp);
1305 		return (EEXIST);
1306 	} else {
1307 		VATTR_NULL(&vattr);
1308 		vattr.va_mode = (mode & ALLPERMS) &
1309 		    ~td->td_proc->p_fd->fd_cmask;
1310 		vattr.va_rdev = dev;
1311 		whiteout = 0;
1312 
1313 		switch (mode & S_IFMT) {
1314 		case S_IFMT:	/* used by badsect to flag bad sectors */
1315 			vattr.va_type = VBAD;
1316 			break;
1317 		case S_IFCHR:
1318 			vattr.va_type = VCHR;
1319 			break;
1320 		case S_IFBLK:
1321 			vattr.va_type = VBLK;
1322 			break;
1323 		case S_IFWHT:
1324 			whiteout = 1;
1325 			break;
1326 		default:
1327 			panic("kern_mknod: invalid mode");
1328 		}
1329 	}
1330 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1331 		NDFREE(&nd, NDF_ONLY_PNBUF);
1332 		vput(nd.ni_dvp);
1333 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1334 			return (error);
1335 		goto restart;
1336 	}
1337 #ifdef MAC
1338 	if (error == 0 && !whiteout)
1339 		error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp,
1340 		    &nd.ni_cnd, &vattr);
1341 #endif
1342 	if (error == 0) {
1343 		if (whiteout)
1344 			error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE);
1345 		else {
1346 			error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp,
1347 						&nd.ni_cnd, &vattr);
1348 			if (error == 0)
1349 				vput(nd.ni_vp);
1350 		}
1351 	}
1352 	NDFREE(&nd, NDF_ONLY_PNBUF);
1353 	vput(nd.ni_dvp);
1354 	vn_finished_write(mp);
1355 	return (error);
1356 }
1357 
1358 /*
1359  * Create a named pipe.
1360  */
1361 #ifndef _SYS_SYSPROTO_H_
1362 struct mkfifo_args {
1363 	char	*path;
1364 	int	mode;
1365 };
1366 #endif
1367 int
1368 sys_mkfifo(td, uap)
1369 	struct thread *td;
1370 	register struct mkfifo_args /* {
1371 		char *path;
1372 		int mode;
1373 	} */ *uap;
1374 {
1375 
1376 	return (kern_mkfifo(td, uap->path, UIO_USERSPACE, uap->mode));
1377 }
1378 
1379 #ifndef _SYS_SYSPROTO_H_
1380 struct mkfifoat_args {
1381 	int	fd;
1382 	char	*path;
1383 	mode_t	mode;
1384 };
1385 #endif
1386 int
1387 sys_mkfifoat(struct thread *td, struct mkfifoat_args *uap)
1388 {
1389 
1390 	return (kern_mkfifoat(td, uap->fd, uap->path, UIO_USERSPACE,
1391 	    uap->mode));
1392 }
1393 
1394 int
1395 kern_mkfifo(struct thread *td, char *path, enum uio_seg pathseg, int mode)
1396 {
1397 
1398 	return (kern_mkfifoat(td, AT_FDCWD, path, pathseg, mode));
1399 }
1400 
1401 int
1402 kern_mkfifoat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1403     int mode)
1404 {
1405 	struct mount *mp;
1406 	struct vattr vattr;
1407 	struct nameidata nd;
1408 	cap_rights_t rights;
1409 	int error;
1410 
1411 	AUDIT_ARG_MODE(mode);
1412 restart:
1413 	bwillwrite();
1414 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1,
1415 	    pathseg, path, fd, cap_rights_init(&rights, CAP_MKFIFOAT), td);
1416 	if ((error = namei(&nd)) != 0)
1417 		return (error);
1418 	if (nd.ni_vp != NULL) {
1419 		NDFREE(&nd, NDF_ONLY_PNBUF);
1420 		if (nd.ni_vp == nd.ni_dvp)
1421 			vrele(nd.ni_dvp);
1422 		else
1423 			vput(nd.ni_dvp);
1424 		vrele(nd.ni_vp);
1425 		return (EEXIST);
1426 	}
1427 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1428 		NDFREE(&nd, NDF_ONLY_PNBUF);
1429 		vput(nd.ni_dvp);
1430 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1431 			return (error);
1432 		goto restart;
1433 	}
1434 	VATTR_NULL(&vattr);
1435 	vattr.va_type = VFIFO;
1436 	vattr.va_mode = (mode & ALLPERMS) & ~td->td_proc->p_fd->fd_cmask;
1437 #ifdef MAC
1438 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1439 	    &vattr);
1440 	if (error != 0)
1441 		goto out;
1442 #endif
1443 	error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
1444 	if (error == 0)
1445 		vput(nd.ni_vp);
1446 #ifdef MAC
1447 out:
1448 #endif
1449 	vput(nd.ni_dvp);
1450 	vn_finished_write(mp);
1451 	NDFREE(&nd, NDF_ONLY_PNBUF);
1452 	return (error);
1453 }
1454 
1455 /*
1456  * Make a hard file link.
1457  */
1458 #ifndef _SYS_SYSPROTO_H_
1459 struct link_args {
1460 	char	*path;
1461 	char	*link;
1462 };
1463 #endif
1464 int
1465 sys_link(td, uap)
1466 	struct thread *td;
1467 	register struct link_args /* {
1468 		char *path;
1469 		char *link;
1470 	} */ *uap;
1471 {
1472 
1473 	return (kern_link(td, uap->path, uap->link, UIO_USERSPACE));
1474 }
1475 
1476 #ifndef _SYS_SYSPROTO_H_
1477 struct linkat_args {
1478 	int	fd1;
1479 	char	*path1;
1480 	int	fd2;
1481 	char	*path2;
1482 	int	flag;
1483 };
1484 #endif
1485 int
1486 sys_linkat(struct thread *td, struct linkat_args *uap)
1487 {
1488 	int flag;
1489 
1490 	flag = uap->flag;
1491 	if (flag & ~AT_SYMLINK_FOLLOW)
1492 		return (EINVAL);
1493 
1494 	return (kern_linkat(td, uap->fd1, uap->fd2, uap->path1, uap->path2,
1495 	    UIO_USERSPACE, (flag & AT_SYMLINK_FOLLOW) ? FOLLOW : NOFOLLOW));
1496 }
1497 
1498 int hardlink_check_uid = 0;
1499 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_uid, CTLFLAG_RW,
1500     &hardlink_check_uid, 0,
1501     "Unprivileged processes cannot create hard links to files owned by other "
1502     "users");
1503 static int hardlink_check_gid = 0;
1504 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_gid, CTLFLAG_RW,
1505     &hardlink_check_gid, 0,
1506     "Unprivileged processes cannot create hard links to files owned by other "
1507     "groups");
1508 
1509 static int
1510 can_hardlink(struct vnode *vp, struct ucred *cred)
1511 {
1512 	struct vattr va;
1513 	int error;
1514 
1515 	if (!hardlink_check_uid && !hardlink_check_gid)
1516 		return (0);
1517 
1518 	error = VOP_GETATTR(vp, &va, cred);
1519 	if (error != 0)
1520 		return (error);
1521 
1522 	if (hardlink_check_uid && cred->cr_uid != va.va_uid) {
1523 		error = priv_check_cred(cred, PRIV_VFS_LINK, 0);
1524 		if (error != 0)
1525 			return (error);
1526 	}
1527 
1528 	if (hardlink_check_gid && !groupmember(va.va_gid, cred)) {
1529 		error = priv_check_cred(cred, PRIV_VFS_LINK, 0);
1530 		if (error != 0)
1531 			return (error);
1532 	}
1533 
1534 	return (0);
1535 }
1536 
1537 int
1538 kern_link(struct thread *td, char *path, char *link, enum uio_seg segflg)
1539 {
1540 
1541 	return (kern_linkat(td, AT_FDCWD, AT_FDCWD, path,link, segflg, FOLLOW));
1542 }
1543 
1544 int
1545 kern_linkat(struct thread *td, int fd1, int fd2, char *path1, char *path2,
1546     enum uio_seg segflg, int follow)
1547 {
1548 	struct vnode *vp;
1549 	struct mount *mp;
1550 	struct nameidata nd;
1551 	cap_rights_t rights;
1552 	int error;
1553 
1554 again:
1555 	bwillwrite();
1556 	NDINIT_AT(&nd, LOOKUP, follow | AUDITVNODE1, segflg, path1, fd1, td);
1557 
1558 	if ((error = namei(&nd)) != 0)
1559 		return (error);
1560 	NDFREE(&nd, NDF_ONLY_PNBUF);
1561 	vp = nd.ni_vp;
1562 	if (vp->v_type == VDIR) {
1563 		vrele(vp);
1564 		return (EPERM);		/* POSIX */
1565 	}
1566 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE2,
1567 	    segflg, path2, fd2, cap_rights_init(&rights, CAP_LINKAT), td);
1568 	if ((error = namei(&nd)) == 0) {
1569 		if (nd.ni_vp != NULL) {
1570 			NDFREE(&nd, NDF_ONLY_PNBUF);
1571 			if (nd.ni_dvp == nd.ni_vp)
1572 				vrele(nd.ni_dvp);
1573 			else
1574 				vput(nd.ni_dvp);
1575 			vrele(nd.ni_vp);
1576 			vrele(vp);
1577 			return (EEXIST);
1578 		} else if (nd.ni_dvp->v_mount != vp->v_mount) {
1579 			/*
1580 			 * Cross-device link.  No need to recheck
1581 			 * vp->v_type, since it cannot change, except
1582 			 * to VBAD.
1583 			 */
1584 			NDFREE(&nd, NDF_ONLY_PNBUF);
1585 			vput(nd.ni_dvp);
1586 			vrele(vp);
1587 			return (EXDEV);
1588 		} else if ((error = vn_lock(vp, LK_EXCLUSIVE)) == 0) {
1589 			error = can_hardlink(vp, td->td_ucred);
1590 #ifdef MAC
1591 			if (error == 0)
1592 				error = mac_vnode_check_link(td->td_ucred,
1593 				    nd.ni_dvp, vp, &nd.ni_cnd);
1594 #endif
1595 			if (error != 0) {
1596 				vput(vp);
1597 				vput(nd.ni_dvp);
1598 				NDFREE(&nd, NDF_ONLY_PNBUF);
1599 				return (error);
1600 			}
1601 			error = vn_start_write(vp, &mp, V_NOWAIT);
1602 			if (error != 0) {
1603 				vput(vp);
1604 				vput(nd.ni_dvp);
1605 				NDFREE(&nd, NDF_ONLY_PNBUF);
1606 				error = vn_start_write(NULL, &mp,
1607 				    V_XSLEEP | PCATCH);
1608 				if (error != 0)
1609 					return (error);
1610 				goto again;
1611 			}
1612 			error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
1613 			VOP_UNLOCK(vp, 0);
1614 			vput(nd.ni_dvp);
1615 			vn_finished_write(mp);
1616 			NDFREE(&nd, NDF_ONLY_PNBUF);
1617 		} else {
1618 			vput(nd.ni_dvp);
1619 			NDFREE(&nd, NDF_ONLY_PNBUF);
1620 			vrele(vp);
1621 			goto again;
1622 		}
1623 	}
1624 	vrele(vp);
1625 	return (error);
1626 }
1627 
1628 /*
1629  * Make a symbolic link.
1630  */
1631 #ifndef _SYS_SYSPROTO_H_
1632 struct symlink_args {
1633 	char	*path;
1634 	char	*link;
1635 };
1636 #endif
1637 int
1638 sys_symlink(td, uap)
1639 	struct thread *td;
1640 	register struct symlink_args /* {
1641 		char *path;
1642 		char *link;
1643 	} */ *uap;
1644 {
1645 
1646 	return (kern_symlink(td, uap->path, uap->link, UIO_USERSPACE));
1647 }
1648 
1649 #ifndef _SYS_SYSPROTO_H_
1650 struct symlinkat_args {
1651 	char	*path;
1652 	int	fd;
1653 	char	*path2;
1654 };
1655 #endif
1656 int
1657 sys_symlinkat(struct thread *td, struct symlinkat_args *uap)
1658 {
1659 
1660 	return (kern_symlinkat(td, uap->path1, uap->fd, uap->path2,
1661 	    UIO_USERSPACE));
1662 }
1663 
1664 int
1665 kern_symlink(struct thread *td, char *path, char *link, enum uio_seg segflg)
1666 {
1667 
1668 	return (kern_symlinkat(td, path, AT_FDCWD, link, segflg));
1669 }
1670 
1671 int
1672 kern_symlinkat(struct thread *td, char *path1, int fd, char *path2,
1673     enum uio_seg segflg)
1674 {
1675 	struct mount *mp;
1676 	struct vattr vattr;
1677 	char *syspath;
1678 	struct nameidata nd;
1679 	int error;
1680 	cap_rights_t rights;
1681 
1682 	if (segflg == UIO_SYSSPACE) {
1683 		syspath = path1;
1684 	} else {
1685 		syspath = uma_zalloc(namei_zone, M_WAITOK);
1686 		if ((error = copyinstr(path1, syspath, MAXPATHLEN, NULL)) != 0)
1687 			goto out;
1688 	}
1689 	AUDIT_ARG_TEXT(syspath);
1690 restart:
1691 	bwillwrite();
1692 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1,
1693 	    segflg, path2, fd, cap_rights_init(&rights, CAP_SYMLINKAT), td);
1694 	if ((error = namei(&nd)) != 0)
1695 		goto out;
1696 	if (nd.ni_vp) {
1697 		NDFREE(&nd, NDF_ONLY_PNBUF);
1698 		if (nd.ni_vp == nd.ni_dvp)
1699 			vrele(nd.ni_dvp);
1700 		else
1701 			vput(nd.ni_dvp);
1702 		vrele(nd.ni_vp);
1703 		error = EEXIST;
1704 		goto out;
1705 	}
1706 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1707 		NDFREE(&nd, NDF_ONLY_PNBUF);
1708 		vput(nd.ni_dvp);
1709 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1710 			goto out;
1711 		goto restart;
1712 	}
1713 	VATTR_NULL(&vattr);
1714 	vattr.va_mode = ACCESSPERMS &~ td->td_proc->p_fd->fd_cmask;
1715 #ifdef MAC
1716 	vattr.va_type = VLNK;
1717 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1718 	    &vattr);
1719 	if (error != 0)
1720 		goto out2;
1721 #endif
1722 	error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, syspath);
1723 	if (error == 0)
1724 		vput(nd.ni_vp);
1725 #ifdef MAC
1726 out2:
1727 #endif
1728 	NDFREE(&nd, NDF_ONLY_PNBUF);
1729 	vput(nd.ni_dvp);
1730 	vn_finished_write(mp);
1731 out:
1732 	if (segflg != UIO_SYSSPACE)
1733 		uma_zfree(namei_zone, syspath);
1734 	return (error);
1735 }
1736 
1737 /*
1738  * Delete a whiteout from the filesystem.
1739  */
1740 int
1741 sys_undelete(td, uap)
1742 	struct thread *td;
1743 	register struct undelete_args /* {
1744 		char *path;
1745 	} */ *uap;
1746 {
1747 	struct mount *mp;
1748 	struct nameidata nd;
1749 	int error;
1750 
1751 restart:
1752 	bwillwrite();
1753 	NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | AUDITVNODE1,
1754 	    UIO_USERSPACE, uap->path, td);
1755 	error = namei(&nd);
1756 	if (error != 0)
1757 		return (error);
1758 
1759 	if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) {
1760 		NDFREE(&nd, NDF_ONLY_PNBUF);
1761 		if (nd.ni_vp == nd.ni_dvp)
1762 			vrele(nd.ni_dvp);
1763 		else
1764 			vput(nd.ni_dvp);
1765 		if (nd.ni_vp)
1766 			vrele(nd.ni_vp);
1767 		return (EEXIST);
1768 	}
1769 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1770 		NDFREE(&nd, NDF_ONLY_PNBUF);
1771 		vput(nd.ni_dvp);
1772 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1773 			return (error);
1774 		goto restart;
1775 	}
1776 	error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE);
1777 	NDFREE(&nd, NDF_ONLY_PNBUF);
1778 	vput(nd.ni_dvp);
1779 	vn_finished_write(mp);
1780 	return (error);
1781 }
1782 
1783 /*
1784  * Delete a name from the filesystem.
1785  */
1786 #ifndef _SYS_SYSPROTO_H_
1787 struct unlink_args {
1788 	char	*path;
1789 };
1790 #endif
1791 int
1792 sys_unlink(td, uap)
1793 	struct thread *td;
1794 	struct unlink_args /* {
1795 		char *path;
1796 	} */ *uap;
1797 {
1798 
1799 	return (kern_unlink(td, uap->path, UIO_USERSPACE));
1800 }
1801 
1802 #ifndef _SYS_SYSPROTO_H_
1803 struct unlinkat_args {
1804 	int	fd;
1805 	char	*path;
1806 	int	flag;
1807 };
1808 #endif
1809 int
1810 sys_unlinkat(struct thread *td, struct unlinkat_args *uap)
1811 {
1812 	int flag = uap->flag;
1813 	int fd = uap->fd;
1814 	char *path = uap->path;
1815 
1816 	if (flag & ~AT_REMOVEDIR)
1817 		return (EINVAL);
1818 
1819 	if (flag & AT_REMOVEDIR)
1820 		return (kern_rmdirat(td, fd, path, UIO_USERSPACE));
1821 	else
1822 		return (kern_unlinkat(td, fd, path, UIO_USERSPACE, 0));
1823 }
1824 
1825 int
1826 kern_unlink(struct thread *td, char *path, enum uio_seg pathseg)
1827 {
1828 
1829 	return (kern_unlinkat(td, AT_FDCWD, path, pathseg, 0));
1830 }
1831 
1832 int
1833 kern_unlinkat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1834     ino_t oldinum)
1835 {
1836 	struct mount *mp;
1837 	struct vnode *vp;
1838 	struct nameidata nd;
1839 	struct stat sb;
1840 	cap_rights_t rights;
1841 	int error;
1842 
1843 restart:
1844 	bwillwrite();
1845 	NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1,
1846 	    pathseg, path, fd, cap_rights_init(&rights, CAP_UNLINKAT), td);
1847 	if ((error = namei(&nd)) != 0)
1848 		return (error == EINVAL ? EPERM : error);
1849 	vp = nd.ni_vp;
1850 	if (vp->v_type == VDIR && oldinum == 0) {
1851 		error = EPERM;		/* POSIX */
1852 	} else if (oldinum != 0 &&
1853 		  ((error = vn_stat(vp, &sb, td->td_ucred, NOCRED, td)) == 0) &&
1854 		  sb.st_ino != oldinum) {
1855 			error = EIDRM;	/* Identifier removed */
1856 	} else {
1857 		/*
1858 		 * The root of a mounted filesystem cannot be deleted.
1859 		 *
1860 		 * XXX: can this only be a VDIR case?
1861 		 */
1862 		if (vp->v_vflag & VV_ROOT)
1863 			error = EBUSY;
1864 	}
1865 	if (error == 0) {
1866 		if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1867 			NDFREE(&nd, NDF_ONLY_PNBUF);
1868 			vput(nd.ni_dvp);
1869 			if (vp == nd.ni_dvp)
1870 				vrele(vp);
1871 			else
1872 				vput(vp);
1873 			if ((error = vn_start_write(NULL, &mp,
1874 			    V_XSLEEP | PCATCH)) != 0)
1875 				return (error);
1876 			goto restart;
1877 		}
1878 #ifdef MAC
1879 		error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
1880 		    &nd.ni_cnd);
1881 		if (error != 0)
1882 			goto out;
1883 #endif
1884 		vfs_notify_upper(vp, VFS_NOTIFY_UPPER_UNLINK);
1885 		error = VOP_REMOVE(nd.ni_dvp, vp, &nd.ni_cnd);
1886 #ifdef MAC
1887 out:
1888 #endif
1889 		vn_finished_write(mp);
1890 	}
1891 	NDFREE(&nd, NDF_ONLY_PNBUF);
1892 	vput(nd.ni_dvp);
1893 	if (vp == nd.ni_dvp)
1894 		vrele(vp);
1895 	else
1896 		vput(vp);
1897 	return (error);
1898 }
1899 
1900 /*
1901  * Reposition read/write file offset.
1902  */
1903 #ifndef _SYS_SYSPROTO_H_
1904 struct lseek_args {
1905 	int	fd;
1906 	int	pad;
1907 	off_t	offset;
1908 	int	whence;
1909 };
1910 #endif
1911 int
1912 sys_lseek(td, uap)
1913 	struct thread *td;
1914 	register struct lseek_args /* {
1915 		int fd;
1916 		int pad;
1917 		off_t offset;
1918 		int whence;
1919 	} */ *uap;
1920 {
1921 	struct file *fp;
1922 	cap_rights_t rights;
1923 	int error;
1924 
1925 	AUDIT_ARG_FD(uap->fd);
1926 	error = fget(td, uap->fd, cap_rights_init(&rights, CAP_SEEK), &fp);
1927 	if (error != 0)
1928 		return (error);
1929 	error = (fp->f_ops->fo_flags & DFLAG_SEEKABLE) != 0 ?
1930 	    fo_seek(fp, uap->offset, uap->whence, td) : ESPIPE;
1931 	fdrop(fp, td);
1932 	return (error);
1933 }
1934 
1935 #if defined(COMPAT_43)
1936 /*
1937  * Reposition read/write file offset.
1938  */
1939 #ifndef _SYS_SYSPROTO_H_
1940 struct olseek_args {
1941 	int	fd;
1942 	long	offset;
1943 	int	whence;
1944 };
1945 #endif
1946 int
1947 olseek(td, uap)
1948 	struct thread *td;
1949 	register struct olseek_args /* {
1950 		int fd;
1951 		long offset;
1952 		int whence;
1953 	} */ *uap;
1954 {
1955 	struct lseek_args /* {
1956 		int fd;
1957 		int pad;
1958 		off_t offset;
1959 		int whence;
1960 	} */ nuap;
1961 
1962 	nuap.fd = uap->fd;
1963 	nuap.offset = uap->offset;
1964 	nuap.whence = uap->whence;
1965 	return (sys_lseek(td, &nuap));
1966 }
1967 #endif /* COMPAT_43 */
1968 
1969 /* Version with the 'pad' argument */
1970 int
1971 freebsd6_lseek(td, uap)
1972 	struct thread *td;
1973 	register struct freebsd6_lseek_args *uap;
1974 {
1975 	struct lseek_args ouap;
1976 
1977 	ouap.fd = uap->fd;
1978 	ouap.offset = uap->offset;
1979 	ouap.whence = uap->whence;
1980 	return (sys_lseek(td, &ouap));
1981 }
1982 
1983 /*
1984  * Check access permissions using passed credentials.
1985  */
1986 static int
1987 vn_access(vp, user_flags, cred, td)
1988 	struct vnode	*vp;
1989 	int		user_flags;
1990 	struct ucred	*cred;
1991 	struct thread	*td;
1992 {
1993 	accmode_t accmode;
1994 	int error;
1995 
1996 	/* Flags == 0 means only check for existence. */
1997 	if (user_flags == 0)
1998 		return (0);
1999 
2000 	accmode = 0;
2001 	if (user_flags & R_OK)
2002 		accmode |= VREAD;
2003 	if (user_flags & W_OK)
2004 		accmode |= VWRITE;
2005 	if (user_flags & X_OK)
2006 		accmode |= VEXEC;
2007 #ifdef MAC
2008 	error = mac_vnode_check_access(cred, vp, accmode);
2009 	if (error != 0)
2010 		return (error);
2011 #endif
2012 	if ((accmode & VWRITE) == 0 || (error = vn_writechk(vp)) == 0)
2013 		error = VOP_ACCESS(vp, accmode, cred, td);
2014 	return (error);
2015 }
2016 
2017 /*
2018  * Check access permissions using "real" credentials.
2019  */
2020 #ifndef _SYS_SYSPROTO_H_
2021 struct access_args {
2022 	char	*path;
2023 	int	amode;
2024 };
2025 #endif
2026 int
2027 sys_access(td, uap)
2028 	struct thread *td;
2029 	register struct access_args /* {
2030 		char *path;
2031 		int amode;
2032 	} */ *uap;
2033 {
2034 
2035 	return (kern_access(td, uap->path, UIO_USERSPACE, uap->amode));
2036 }
2037 
2038 #ifndef _SYS_SYSPROTO_H_
2039 struct faccessat_args {
2040 	int	dirfd;
2041 	char	*path;
2042 	int	amode;
2043 	int	flag;
2044 }
2045 #endif
2046 int
2047 sys_faccessat(struct thread *td, struct faccessat_args *uap)
2048 {
2049 
2050 	if (uap->flag & ~AT_EACCESS)
2051 		return (EINVAL);
2052 	return (kern_accessat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
2053 	    uap->amode));
2054 }
2055 
2056 int
2057 kern_access(struct thread *td, char *path, enum uio_seg pathseg, int amode)
2058 {
2059 
2060 	return (kern_accessat(td, AT_FDCWD, path, pathseg, 0, amode));
2061 }
2062 
2063 int
2064 kern_accessat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2065     int flag, int amode)
2066 {
2067 	struct ucred *cred, *usecred;
2068 	struct vnode *vp;
2069 	struct nameidata nd;
2070 	cap_rights_t rights;
2071 	int error;
2072 
2073 	if (amode != F_OK && (amode & ~(R_OK | W_OK | X_OK)) != 0)
2074 		return (EINVAL);
2075 
2076 	/*
2077 	 * Create and modify a temporary credential instead of one that
2078 	 * is potentially shared (if we need one).
2079 	 */
2080 	cred = td->td_ucred;
2081 	if ((flag & AT_EACCESS) == 0 &&
2082 	    ((cred->cr_uid != cred->cr_ruid ||
2083 	    cred->cr_rgid != cred->cr_groups[0]))) {
2084 		usecred = crdup(cred);
2085 		usecred->cr_uid = cred->cr_ruid;
2086 		usecred->cr_groups[0] = cred->cr_rgid;
2087 		td->td_ucred = usecred;
2088 	} else
2089 		usecred = cred;
2090 	AUDIT_ARG_VALUE(amode);
2091 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF |
2092 	    AUDITVNODE1, pathseg, path, fd, cap_rights_init(&rights, CAP_FSTAT),
2093 	    td);
2094 	if ((error = namei(&nd)) != 0)
2095 		goto out;
2096 	vp = nd.ni_vp;
2097 
2098 	error = vn_access(vp, amode, usecred, td);
2099 	NDFREE(&nd, NDF_ONLY_PNBUF);
2100 	vput(vp);
2101 out:
2102 	if (usecred != cred) {
2103 		td->td_ucred = cred;
2104 		crfree(usecred);
2105 	}
2106 	return (error);
2107 }
2108 
2109 /*
2110  * Check access permissions using "effective" credentials.
2111  */
2112 #ifndef _SYS_SYSPROTO_H_
2113 struct eaccess_args {
2114 	char	*path;
2115 	int	amode;
2116 };
2117 #endif
2118 int
2119 sys_eaccess(td, uap)
2120 	struct thread *td;
2121 	register struct eaccess_args /* {
2122 		char *path;
2123 		int amode;
2124 	} */ *uap;
2125 {
2126 
2127 	return (kern_eaccess(td, uap->path, UIO_USERSPACE, uap->amode));
2128 }
2129 
2130 int
2131 kern_eaccess(struct thread *td, char *path, enum uio_seg pathseg, int amode)
2132 {
2133 
2134 	return (kern_accessat(td, AT_FDCWD, path, pathseg, AT_EACCESS, amode));
2135 }
2136 
2137 #if defined(COMPAT_43)
2138 /*
2139  * Get file status; this version follows links.
2140  */
2141 #ifndef _SYS_SYSPROTO_H_
2142 struct ostat_args {
2143 	char	*path;
2144 	struct ostat *ub;
2145 };
2146 #endif
2147 int
2148 ostat(td, uap)
2149 	struct thread *td;
2150 	register struct ostat_args /* {
2151 		char *path;
2152 		struct ostat *ub;
2153 	} */ *uap;
2154 {
2155 	struct stat sb;
2156 	struct ostat osb;
2157 	int error;
2158 
2159 	error = kern_stat(td, uap->path, UIO_USERSPACE, &sb);
2160 	if (error != 0)
2161 		return (error);
2162 	cvtstat(&sb, &osb);
2163 	return (copyout(&osb, uap->ub, sizeof (osb)));
2164 }
2165 
2166 /*
2167  * Get file status; this version does not follow links.
2168  */
2169 #ifndef _SYS_SYSPROTO_H_
2170 struct olstat_args {
2171 	char	*path;
2172 	struct ostat *ub;
2173 };
2174 #endif
2175 int
2176 olstat(td, uap)
2177 	struct thread *td;
2178 	register struct olstat_args /* {
2179 		char *path;
2180 		struct ostat *ub;
2181 	} */ *uap;
2182 {
2183 	struct stat sb;
2184 	struct ostat osb;
2185 	int error;
2186 
2187 	error = kern_lstat(td, uap->path, UIO_USERSPACE, &sb);
2188 	if (error != 0)
2189 		return (error);
2190 	cvtstat(&sb, &osb);
2191 	return (copyout(&osb, uap->ub, sizeof (osb)));
2192 }
2193 
2194 /*
2195  * Convert from an old to a new stat structure.
2196  */
2197 void
2198 cvtstat(st, ost)
2199 	struct stat *st;
2200 	struct ostat *ost;
2201 {
2202 
2203 	ost->st_dev = st->st_dev;
2204 	ost->st_ino = st->st_ino;
2205 	ost->st_mode = st->st_mode;
2206 	ost->st_nlink = st->st_nlink;
2207 	ost->st_uid = st->st_uid;
2208 	ost->st_gid = st->st_gid;
2209 	ost->st_rdev = st->st_rdev;
2210 	if (st->st_size < (quad_t)1 << 32)
2211 		ost->st_size = st->st_size;
2212 	else
2213 		ost->st_size = -2;
2214 	ost->st_atim = st->st_atim;
2215 	ost->st_mtim = st->st_mtim;
2216 	ost->st_ctim = st->st_ctim;
2217 	ost->st_blksize = st->st_blksize;
2218 	ost->st_blocks = st->st_blocks;
2219 	ost->st_flags = st->st_flags;
2220 	ost->st_gen = st->st_gen;
2221 }
2222 #endif /* COMPAT_43 */
2223 
2224 /*
2225  * Get file status; this version follows links.
2226  */
2227 #ifndef _SYS_SYSPROTO_H_
2228 struct stat_args {
2229 	char	*path;
2230 	struct stat *ub;
2231 };
2232 #endif
2233 int
2234 sys_stat(td, uap)
2235 	struct thread *td;
2236 	register struct stat_args /* {
2237 		char *path;
2238 		struct stat *ub;
2239 	} */ *uap;
2240 {
2241 	struct stat sb;
2242 	int error;
2243 
2244 	error = kern_stat(td, uap->path, UIO_USERSPACE, &sb);
2245 	if (error == 0)
2246 		error = copyout(&sb, uap->ub, sizeof (sb));
2247 	return (error);
2248 }
2249 
2250 #ifndef _SYS_SYSPROTO_H_
2251 struct fstatat_args {
2252 	int	fd;
2253 	char	*path;
2254 	struct stat	*buf;
2255 	int	flag;
2256 }
2257 #endif
2258 int
2259 sys_fstatat(struct thread *td, struct fstatat_args *uap)
2260 {
2261 	struct stat sb;
2262 	int error;
2263 
2264 	error = kern_statat(td, uap->flag, uap->fd, uap->path,
2265 	    UIO_USERSPACE, &sb);
2266 	if (error == 0)
2267 		error = copyout(&sb, uap->buf, sizeof (sb));
2268 	return (error);
2269 }
2270 
2271 int
2272 kern_stat(struct thread *td, char *path, enum uio_seg pathseg, struct stat *sbp)
2273 {
2274 
2275 	return (kern_statat(td, 0, AT_FDCWD, path, pathseg, sbp));
2276 }
2277 
2278 int
2279 kern_statat(struct thread *td, int flag, int fd, char *path,
2280     enum uio_seg pathseg, struct stat *sbp)
2281 {
2282 
2283 	return (kern_statat_vnhook(td, flag, fd, path, pathseg, sbp, NULL));
2284 }
2285 
2286 int
2287 kern_statat_vnhook(struct thread *td, int flag, int fd, char *path,
2288     enum uio_seg pathseg, struct stat *sbp,
2289     void (*hook)(struct vnode *vp, struct stat *sbp))
2290 {
2291 	struct nameidata nd;
2292 	struct stat sb;
2293 	cap_rights_t rights;
2294 	int error;
2295 
2296 	if (flag & ~AT_SYMLINK_NOFOLLOW)
2297 		return (EINVAL);
2298 
2299 	NDINIT_ATRIGHTS(&nd, LOOKUP, ((flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW :
2300 	    FOLLOW) | LOCKSHARED | LOCKLEAF | AUDITVNODE1, pathseg, path, fd,
2301 	    cap_rights_init(&rights, CAP_FSTAT), td);
2302 
2303 	if ((error = namei(&nd)) != 0)
2304 		return (error);
2305 	error = vn_stat(nd.ni_vp, &sb, td->td_ucred, NOCRED, td);
2306 	if (error == 0) {
2307 		SDT_PROBE(vfs, , stat, mode, path, sb.st_mode, 0, 0, 0);
2308 		if (S_ISREG(sb.st_mode))
2309 			SDT_PROBE(vfs, , stat, reg, path, pathseg, 0, 0, 0);
2310 		if (__predict_false(hook != NULL))
2311 			hook(nd.ni_vp, &sb);
2312 	}
2313 	NDFREE(&nd, NDF_ONLY_PNBUF);
2314 	vput(nd.ni_vp);
2315 	if (error != 0)
2316 		return (error);
2317 	*sbp = sb;
2318 #ifdef KTRACE
2319 	if (KTRPOINT(td, KTR_STRUCT))
2320 		ktrstat(&sb);
2321 #endif
2322 	return (0);
2323 }
2324 
2325 /*
2326  * Get file status; this version does not follow links.
2327  */
2328 #ifndef _SYS_SYSPROTO_H_
2329 struct lstat_args {
2330 	char	*path;
2331 	struct stat *ub;
2332 };
2333 #endif
2334 int
2335 sys_lstat(td, uap)
2336 	struct thread *td;
2337 	register struct lstat_args /* {
2338 		char *path;
2339 		struct stat *ub;
2340 	} */ *uap;
2341 {
2342 	struct stat sb;
2343 	int error;
2344 
2345 	error = kern_lstat(td, uap->path, UIO_USERSPACE, &sb);
2346 	if (error == 0)
2347 		error = copyout(&sb, uap->ub, sizeof (sb));
2348 	return (error);
2349 }
2350 
2351 int
2352 kern_lstat(struct thread *td, char *path, enum uio_seg pathseg, struct stat *sbp)
2353 {
2354 
2355 	return (kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, path, pathseg,
2356 	    sbp));
2357 }
2358 
2359 /*
2360  * Implementation of the NetBSD [l]stat() functions.
2361  */
2362 void
2363 cvtnstat(sb, nsb)
2364 	struct stat *sb;
2365 	struct nstat *nsb;
2366 {
2367 
2368 	bzero(nsb, sizeof *nsb);
2369 	nsb->st_dev = sb->st_dev;
2370 	nsb->st_ino = sb->st_ino;
2371 	nsb->st_mode = sb->st_mode;
2372 	nsb->st_nlink = sb->st_nlink;
2373 	nsb->st_uid = sb->st_uid;
2374 	nsb->st_gid = sb->st_gid;
2375 	nsb->st_rdev = sb->st_rdev;
2376 	nsb->st_atim = sb->st_atim;
2377 	nsb->st_mtim = sb->st_mtim;
2378 	nsb->st_ctim = sb->st_ctim;
2379 	nsb->st_size = sb->st_size;
2380 	nsb->st_blocks = sb->st_blocks;
2381 	nsb->st_blksize = sb->st_blksize;
2382 	nsb->st_flags = sb->st_flags;
2383 	nsb->st_gen = sb->st_gen;
2384 	nsb->st_birthtim = sb->st_birthtim;
2385 }
2386 
2387 #ifndef _SYS_SYSPROTO_H_
2388 struct nstat_args {
2389 	char	*path;
2390 	struct nstat *ub;
2391 };
2392 #endif
2393 int
2394 sys_nstat(td, uap)
2395 	struct thread *td;
2396 	register struct nstat_args /* {
2397 		char *path;
2398 		struct nstat *ub;
2399 	} */ *uap;
2400 {
2401 	struct stat sb;
2402 	struct nstat nsb;
2403 	int error;
2404 
2405 	error = kern_stat(td, uap->path, UIO_USERSPACE, &sb);
2406 	if (error != 0)
2407 		return (error);
2408 	cvtnstat(&sb, &nsb);
2409 	return (copyout(&nsb, uap->ub, sizeof (nsb)));
2410 }
2411 
2412 /*
2413  * NetBSD lstat.  Get file status; this version does not follow links.
2414  */
2415 #ifndef _SYS_SYSPROTO_H_
2416 struct lstat_args {
2417 	char	*path;
2418 	struct stat *ub;
2419 };
2420 #endif
2421 int
2422 sys_nlstat(td, uap)
2423 	struct thread *td;
2424 	register struct nlstat_args /* {
2425 		char *path;
2426 		struct nstat *ub;
2427 	} */ *uap;
2428 {
2429 	struct stat sb;
2430 	struct nstat nsb;
2431 	int error;
2432 
2433 	error = kern_lstat(td, uap->path, UIO_USERSPACE, &sb);
2434 	if (error != 0)
2435 		return (error);
2436 	cvtnstat(&sb, &nsb);
2437 	return (copyout(&nsb, uap->ub, sizeof (nsb)));
2438 }
2439 
2440 /*
2441  * Get configurable pathname variables.
2442  */
2443 #ifndef _SYS_SYSPROTO_H_
2444 struct pathconf_args {
2445 	char	*path;
2446 	int	name;
2447 };
2448 #endif
2449 int
2450 sys_pathconf(td, uap)
2451 	struct thread *td;
2452 	register struct pathconf_args /* {
2453 		char *path;
2454 		int name;
2455 	} */ *uap;
2456 {
2457 
2458 	return (kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name, FOLLOW));
2459 }
2460 
2461 #ifndef _SYS_SYSPROTO_H_
2462 struct lpathconf_args {
2463 	char	*path;
2464 	int	name;
2465 };
2466 #endif
2467 int
2468 sys_lpathconf(td, uap)
2469 	struct thread *td;
2470 	register struct lpathconf_args /* {
2471 		char *path;
2472 		int name;
2473 	} */ *uap;
2474 {
2475 
2476 	return (kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name,
2477 	    NOFOLLOW));
2478 }
2479 
2480 int
2481 kern_pathconf(struct thread *td, char *path, enum uio_seg pathseg, int name,
2482     u_long flags)
2483 {
2484 	struct nameidata nd;
2485 	int error;
2486 
2487 	NDINIT(&nd, LOOKUP, LOCKSHARED | LOCKLEAF | AUDITVNODE1 | flags,
2488 	    pathseg, path, td);
2489 	if ((error = namei(&nd)) != 0)
2490 		return (error);
2491 	NDFREE(&nd, NDF_ONLY_PNBUF);
2492 
2493 	/* If asynchronous I/O is available, it works for all files. */
2494 	if (name == _PC_ASYNC_IO)
2495 		td->td_retval[0] = async_io_version;
2496 	else
2497 		error = VOP_PATHCONF(nd.ni_vp, name, td->td_retval);
2498 	vput(nd.ni_vp);
2499 	return (error);
2500 }
2501 
2502 /*
2503  * Return target name of a symbolic link.
2504  */
2505 #ifndef _SYS_SYSPROTO_H_
2506 struct readlink_args {
2507 	char	*path;
2508 	char	*buf;
2509 	size_t	count;
2510 };
2511 #endif
2512 int
2513 sys_readlink(td, uap)
2514 	struct thread *td;
2515 	register struct readlink_args /* {
2516 		char *path;
2517 		char *buf;
2518 		size_t count;
2519 	} */ *uap;
2520 {
2521 
2522 	return (kern_readlink(td, uap->path, UIO_USERSPACE, uap->buf,
2523 	    UIO_USERSPACE, uap->count));
2524 }
2525 #ifndef _SYS_SYSPROTO_H_
2526 struct readlinkat_args {
2527 	int	fd;
2528 	char	*path;
2529 	char	*buf;
2530 	size_t	bufsize;
2531 };
2532 #endif
2533 int
2534 sys_readlinkat(struct thread *td, struct readlinkat_args *uap)
2535 {
2536 
2537 	return (kern_readlinkat(td, uap->fd, uap->path, UIO_USERSPACE,
2538 	    uap->buf, UIO_USERSPACE, uap->bufsize));
2539 }
2540 
2541 int
2542 kern_readlink(struct thread *td, char *path, enum uio_seg pathseg, char *buf,
2543     enum uio_seg bufseg, size_t count)
2544 {
2545 
2546 	return (kern_readlinkat(td, AT_FDCWD, path, pathseg, buf, bufseg,
2547 	    count));
2548 }
2549 
2550 int
2551 kern_readlinkat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2552     char *buf, enum uio_seg bufseg, size_t count)
2553 {
2554 	struct vnode *vp;
2555 	struct iovec aiov;
2556 	struct uio auio;
2557 	struct nameidata nd;
2558 	int error;
2559 
2560 	if (count > IOSIZE_MAX)
2561 		return (EINVAL);
2562 
2563 	NDINIT_AT(&nd, LOOKUP, NOFOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
2564 	    pathseg, path, fd, td);
2565 
2566 	if ((error = namei(&nd)) != 0)
2567 		return (error);
2568 	NDFREE(&nd, NDF_ONLY_PNBUF);
2569 	vp = nd.ni_vp;
2570 #ifdef MAC
2571 	error = mac_vnode_check_readlink(td->td_ucred, vp);
2572 	if (error != 0) {
2573 		vput(vp);
2574 		return (error);
2575 	}
2576 #endif
2577 	if (vp->v_type != VLNK)
2578 		error = EINVAL;
2579 	else {
2580 		aiov.iov_base = buf;
2581 		aiov.iov_len = count;
2582 		auio.uio_iov = &aiov;
2583 		auio.uio_iovcnt = 1;
2584 		auio.uio_offset = 0;
2585 		auio.uio_rw = UIO_READ;
2586 		auio.uio_segflg = bufseg;
2587 		auio.uio_td = td;
2588 		auio.uio_resid = count;
2589 		error = VOP_READLINK(vp, &auio, td->td_ucred);
2590 		td->td_retval[0] = count - auio.uio_resid;
2591 	}
2592 	vput(vp);
2593 	return (error);
2594 }
2595 
2596 /*
2597  * Common implementation code for chflags() and fchflags().
2598  */
2599 static int
2600 setfflags(td, vp, flags)
2601 	struct thread *td;
2602 	struct vnode *vp;
2603 	u_long flags;
2604 {
2605 	struct mount *mp;
2606 	struct vattr vattr;
2607 	int error;
2608 
2609 	/* We can't support the value matching VNOVAL. */
2610 	if (flags == VNOVAL)
2611 		return (EOPNOTSUPP);
2612 
2613 	/*
2614 	 * Prevent non-root users from setting flags on devices.  When
2615 	 * a device is reused, users can retain ownership of the device
2616 	 * if they are allowed to set flags and programs assume that
2617 	 * chown can't fail when done as root.
2618 	 */
2619 	if (vp->v_type == VCHR || vp->v_type == VBLK) {
2620 		error = priv_check(td, PRIV_VFS_CHFLAGS_DEV);
2621 		if (error != 0)
2622 			return (error);
2623 	}
2624 
2625 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2626 		return (error);
2627 	VATTR_NULL(&vattr);
2628 	vattr.va_flags = flags;
2629 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2630 #ifdef MAC
2631 	error = mac_vnode_check_setflags(td->td_ucred, vp, vattr.va_flags);
2632 	if (error == 0)
2633 #endif
2634 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
2635 	VOP_UNLOCK(vp, 0);
2636 	vn_finished_write(mp);
2637 	return (error);
2638 }
2639 
2640 /*
2641  * Change flags of a file given a path name.
2642  */
2643 #ifndef _SYS_SYSPROTO_H_
2644 struct chflags_args {
2645 	const char *path;
2646 	u_long	flags;
2647 };
2648 #endif
2649 int
2650 sys_chflags(td, uap)
2651 	struct thread *td;
2652 	register struct chflags_args /* {
2653 		const char *path;
2654 		u_long flags;
2655 	} */ *uap;
2656 {
2657 
2658 	return (kern_chflags(td, uap->path, UIO_USERSPACE, uap->flags));
2659 }
2660 
2661 #ifndef _SYS_SYSPROTO_H_
2662 struct chflagsat_args {
2663 	int	fd;
2664 	const char *path;
2665 	u_long	flags;
2666 	int	atflag;
2667 }
2668 #endif
2669 int
2670 sys_chflagsat(struct thread *td, struct chflagsat_args *uap)
2671 {
2672 	int fd = uap->fd;
2673 	const char *path = uap->path;
2674 	u_long flags = uap->flags;
2675 	int atflag = uap->atflag;
2676 
2677 	if (atflag & ~AT_SYMLINK_NOFOLLOW)
2678 		return (EINVAL);
2679 
2680 	return (kern_chflagsat(td, fd, path, UIO_USERSPACE, flags, atflag));
2681 }
2682 
2683 static int
2684 kern_chflags(struct thread *td, const char *path, enum uio_seg pathseg,
2685     u_long flags)
2686 {
2687 
2688 	return (kern_chflagsat(td, AT_FDCWD, path, pathseg, flags, 0));
2689 }
2690 
2691 /*
2692  * Same as chflags() but doesn't follow symlinks.
2693  */
2694 int
2695 sys_lchflags(td, uap)
2696 	struct thread *td;
2697 	register struct lchflags_args /* {
2698 		const char *path;
2699 		u_long flags;
2700 	} */ *uap;
2701 {
2702 
2703 	return (kern_chflagsat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2704 	    uap->flags, AT_SYMLINK_NOFOLLOW));
2705 }
2706 
2707 static int
2708 kern_chflagsat(struct thread *td, int fd, const char *path,
2709     enum uio_seg pathseg, u_long flags, int atflag)
2710 {
2711 	struct nameidata nd;
2712 	cap_rights_t rights;
2713 	int error, follow;
2714 
2715 	AUDIT_ARG_FFLAGS(flags);
2716 	follow = (atflag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
2717 	NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, pathseg, path, fd,
2718 	    cap_rights_init(&rights, CAP_FCHFLAGS), td);
2719 	if ((error = namei(&nd)) != 0)
2720 		return (error);
2721 	NDFREE(&nd, NDF_ONLY_PNBUF);
2722 	error = setfflags(td, nd.ni_vp, flags);
2723 	vrele(nd.ni_vp);
2724 	return (error);
2725 }
2726 
2727 /*
2728  * Change flags of a file given a file descriptor.
2729  */
2730 #ifndef _SYS_SYSPROTO_H_
2731 struct fchflags_args {
2732 	int	fd;
2733 	u_long	flags;
2734 };
2735 #endif
2736 int
2737 sys_fchflags(td, uap)
2738 	struct thread *td;
2739 	register struct fchflags_args /* {
2740 		int fd;
2741 		u_long flags;
2742 	} */ *uap;
2743 {
2744 	struct file *fp;
2745 	cap_rights_t rights;
2746 	int error;
2747 
2748 	AUDIT_ARG_FD(uap->fd);
2749 	AUDIT_ARG_FFLAGS(uap->flags);
2750 	error = getvnode(td->td_proc->p_fd, uap->fd,
2751 	    cap_rights_init(&rights, CAP_FCHFLAGS), &fp);
2752 	if (error != 0)
2753 		return (error);
2754 #ifdef AUDIT
2755 	vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
2756 	AUDIT_ARG_VNODE1(fp->f_vnode);
2757 	VOP_UNLOCK(fp->f_vnode, 0);
2758 #endif
2759 	error = setfflags(td, fp->f_vnode, uap->flags);
2760 	fdrop(fp, td);
2761 	return (error);
2762 }
2763 
2764 /*
2765  * Common implementation code for chmod(), lchmod() and fchmod().
2766  */
2767 int
2768 setfmode(td, cred, vp, mode)
2769 	struct thread *td;
2770 	struct ucred *cred;
2771 	struct vnode *vp;
2772 	int mode;
2773 {
2774 	struct mount *mp;
2775 	struct vattr vattr;
2776 	int error;
2777 
2778 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2779 		return (error);
2780 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2781 	VATTR_NULL(&vattr);
2782 	vattr.va_mode = mode & ALLPERMS;
2783 #ifdef MAC
2784 	error = mac_vnode_check_setmode(cred, vp, vattr.va_mode);
2785 	if (error == 0)
2786 #endif
2787 		error = VOP_SETATTR(vp, &vattr, cred);
2788 	VOP_UNLOCK(vp, 0);
2789 	vn_finished_write(mp);
2790 	return (error);
2791 }
2792 
2793 /*
2794  * Change mode of a file given path name.
2795  */
2796 #ifndef _SYS_SYSPROTO_H_
2797 struct chmod_args {
2798 	char	*path;
2799 	int	mode;
2800 };
2801 #endif
2802 int
2803 sys_chmod(td, uap)
2804 	struct thread *td;
2805 	register struct chmod_args /* {
2806 		char *path;
2807 		int mode;
2808 	} */ *uap;
2809 {
2810 
2811 	return (kern_chmod(td, uap->path, UIO_USERSPACE, uap->mode));
2812 }
2813 
2814 #ifndef _SYS_SYSPROTO_H_
2815 struct fchmodat_args {
2816 	int	dirfd;
2817 	char	*path;
2818 	mode_t	mode;
2819 	int	flag;
2820 }
2821 #endif
2822 int
2823 sys_fchmodat(struct thread *td, struct fchmodat_args *uap)
2824 {
2825 	int flag = uap->flag;
2826 	int fd = uap->fd;
2827 	char *path = uap->path;
2828 	mode_t mode = uap->mode;
2829 
2830 	if (flag & ~AT_SYMLINK_NOFOLLOW)
2831 		return (EINVAL);
2832 
2833 	return (kern_fchmodat(td, fd, path, UIO_USERSPACE, mode, flag));
2834 }
2835 
2836 int
2837 kern_chmod(struct thread *td, char *path, enum uio_seg pathseg, int mode)
2838 {
2839 
2840 	return (kern_fchmodat(td, AT_FDCWD, path, pathseg, mode, 0));
2841 }
2842 
2843 /*
2844  * Change mode of a file given path name (don't follow links.)
2845  */
2846 #ifndef _SYS_SYSPROTO_H_
2847 struct lchmod_args {
2848 	char	*path;
2849 	int	mode;
2850 };
2851 #endif
2852 int
2853 sys_lchmod(td, uap)
2854 	struct thread *td;
2855 	register struct lchmod_args /* {
2856 		char *path;
2857 		int mode;
2858 	} */ *uap;
2859 {
2860 
2861 	return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2862 	    uap->mode, AT_SYMLINK_NOFOLLOW));
2863 }
2864 
2865 int
2866 kern_fchmodat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2867     mode_t mode, int flag)
2868 {
2869 	struct nameidata nd;
2870 	cap_rights_t rights;
2871 	int error, follow;
2872 
2873 	AUDIT_ARG_MODE(mode);
2874 	follow = (flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
2875 	NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, pathseg, path, fd,
2876 	    cap_rights_init(&rights, CAP_FCHMOD), td);
2877 	if ((error = namei(&nd)) != 0)
2878 		return (error);
2879 	NDFREE(&nd, NDF_ONLY_PNBUF);
2880 	error = setfmode(td, td->td_ucred, nd.ni_vp, mode);
2881 	vrele(nd.ni_vp);
2882 	return (error);
2883 }
2884 
2885 /*
2886  * Change mode of a file given a file descriptor.
2887  */
2888 #ifndef _SYS_SYSPROTO_H_
2889 struct fchmod_args {
2890 	int	fd;
2891 	int	mode;
2892 };
2893 #endif
2894 int
2895 sys_fchmod(struct thread *td, struct fchmod_args *uap)
2896 {
2897 	struct file *fp;
2898 	cap_rights_t rights;
2899 	int error;
2900 
2901 	AUDIT_ARG_FD(uap->fd);
2902 	AUDIT_ARG_MODE(uap->mode);
2903 
2904 	error = fget(td, uap->fd, cap_rights_init(&rights, CAP_FCHMOD), &fp);
2905 	if (error != 0)
2906 		return (error);
2907 	error = fo_chmod(fp, uap->mode, td->td_ucred, td);
2908 	fdrop(fp, td);
2909 	return (error);
2910 }
2911 
2912 /*
2913  * Common implementation for chown(), lchown(), and fchown()
2914  */
2915 int
2916 setfown(td, cred, vp, uid, gid)
2917 	struct thread *td;
2918 	struct ucred *cred;
2919 	struct vnode *vp;
2920 	uid_t uid;
2921 	gid_t gid;
2922 {
2923 	struct mount *mp;
2924 	struct vattr vattr;
2925 	int error;
2926 
2927 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2928 		return (error);
2929 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2930 	VATTR_NULL(&vattr);
2931 	vattr.va_uid = uid;
2932 	vattr.va_gid = gid;
2933 #ifdef MAC
2934 	error = mac_vnode_check_setowner(cred, vp, vattr.va_uid,
2935 	    vattr.va_gid);
2936 	if (error == 0)
2937 #endif
2938 		error = VOP_SETATTR(vp, &vattr, cred);
2939 	VOP_UNLOCK(vp, 0);
2940 	vn_finished_write(mp);
2941 	return (error);
2942 }
2943 
2944 /*
2945  * Set ownership given a path name.
2946  */
2947 #ifndef _SYS_SYSPROTO_H_
2948 struct chown_args {
2949 	char	*path;
2950 	int	uid;
2951 	int	gid;
2952 };
2953 #endif
2954 int
2955 sys_chown(td, uap)
2956 	struct thread *td;
2957 	register struct chown_args /* {
2958 		char *path;
2959 		int uid;
2960 		int gid;
2961 	} */ *uap;
2962 {
2963 
2964 	return (kern_chown(td, uap->path, UIO_USERSPACE, uap->uid, uap->gid));
2965 }
2966 
2967 #ifndef _SYS_SYSPROTO_H_
2968 struct fchownat_args {
2969 	int fd;
2970 	const char * path;
2971 	uid_t uid;
2972 	gid_t gid;
2973 	int flag;
2974 };
2975 #endif
2976 int
2977 sys_fchownat(struct thread *td, struct fchownat_args *uap)
2978 {
2979 	int flag;
2980 
2981 	flag = uap->flag;
2982 	if (flag & ~AT_SYMLINK_NOFOLLOW)
2983 		return (EINVAL);
2984 
2985 	return (kern_fchownat(td, uap->fd, uap->path, UIO_USERSPACE, uap->uid,
2986 	    uap->gid, uap->flag));
2987 }
2988 
2989 int
2990 kern_chown(struct thread *td, char *path, enum uio_seg pathseg, int uid,
2991     int gid)
2992 {
2993 
2994 	return (kern_fchownat(td, AT_FDCWD, path, pathseg, uid, gid, 0));
2995 }
2996 
2997 int
2998 kern_fchownat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2999     int uid, int gid, int flag)
3000 {
3001 	struct nameidata nd;
3002 	cap_rights_t rights;
3003 	int error, follow;
3004 
3005 	AUDIT_ARG_OWNER(uid, gid);
3006 	follow = (flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
3007 	NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, pathseg, path, fd,
3008 	    cap_rights_init(&rights, CAP_FCHOWN), td);
3009 
3010 	if ((error = namei(&nd)) != 0)
3011 		return (error);
3012 	NDFREE(&nd, NDF_ONLY_PNBUF);
3013 	error = setfown(td, td->td_ucred, nd.ni_vp, uid, gid);
3014 	vrele(nd.ni_vp);
3015 	return (error);
3016 }
3017 
3018 /*
3019  * Set ownership given a path name, do not cross symlinks.
3020  */
3021 #ifndef _SYS_SYSPROTO_H_
3022 struct lchown_args {
3023 	char	*path;
3024 	int	uid;
3025 	int	gid;
3026 };
3027 #endif
3028 int
3029 sys_lchown(td, uap)
3030 	struct thread *td;
3031 	register struct lchown_args /* {
3032 		char *path;
3033 		int uid;
3034 		int gid;
3035 	} */ *uap;
3036 {
3037 
3038 	return (kern_lchown(td, uap->path, UIO_USERSPACE, uap->uid, uap->gid));
3039 }
3040 
3041 int
3042 kern_lchown(struct thread *td, char *path, enum uio_seg pathseg, int uid,
3043     int gid)
3044 {
3045 
3046 	return (kern_fchownat(td, AT_FDCWD, path, pathseg, uid, gid,
3047 	    AT_SYMLINK_NOFOLLOW));
3048 }
3049 
3050 /*
3051  * Set ownership given a file descriptor.
3052  */
3053 #ifndef _SYS_SYSPROTO_H_
3054 struct fchown_args {
3055 	int	fd;
3056 	int	uid;
3057 	int	gid;
3058 };
3059 #endif
3060 int
3061 sys_fchown(td, uap)
3062 	struct thread *td;
3063 	register struct fchown_args /* {
3064 		int fd;
3065 		int uid;
3066 		int gid;
3067 	} */ *uap;
3068 {
3069 	struct file *fp;
3070 	cap_rights_t rights;
3071 	int error;
3072 
3073 	AUDIT_ARG_FD(uap->fd);
3074 	AUDIT_ARG_OWNER(uap->uid, uap->gid);
3075 	error = fget(td, uap->fd, cap_rights_init(&rights, CAP_FCHOWN), &fp);
3076 	if (error != 0)
3077 		return (error);
3078 	error = fo_chown(fp, uap->uid, uap->gid, td->td_ucred, td);
3079 	fdrop(fp, td);
3080 	return (error);
3081 }
3082 
3083 /*
3084  * Common implementation code for utimes(), lutimes(), and futimes().
3085  */
3086 static int
3087 getutimes(usrtvp, tvpseg, tsp)
3088 	const struct timeval *usrtvp;
3089 	enum uio_seg tvpseg;
3090 	struct timespec *tsp;
3091 {
3092 	struct timeval tv[2];
3093 	const struct timeval *tvp;
3094 	int error;
3095 
3096 	if (usrtvp == NULL) {
3097 		vfs_timestamp(&tsp[0]);
3098 		tsp[1] = tsp[0];
3099 	} else {
3100 		if (tvpseg == UIO_SYSSPACE) {
3101 			tvp = usrtvp;
3102 		} else {
3103 			if ((error = copyin(usrtvp, tv, sizeof(tv))) != 0)
3104 				return (error);
3105 			tvp = tv;
3106 		}
3107 
3108 		if (tvp[0].tv_usec < 0 || tvp[0].tv_usec >= 1000000 ||
3109 		    tvp[1].tv_usec < 0 || tvp[1].tv_usec >= 1000000)
3110 			return (EINVAL);
3111 		TIMEVAL_TO_TIMESPEC(&tvp[0], &tsp[0]);
3112 		TIMEVAL_TO_TIMESPEC(&tvp[1], &tsp[1]);
3113 	}
3114 	return (0);
3115 }
3116 
3117 /*
3118  * Common implementation code for utimes(), lutimes(), and futimes().
3119  */
3120 static int
3121 setutimes(td, vp, ts, numtimes, nullflag)
3122 	struct thread *td;
3123 	struct vnode *vp;
3124 	const struct timespec *ts;
3125 	int numtimes;
3126 	int nullflag;
3127 {
3128 	struct mount *mp;
3129 	struct vattr vattr;
3130 	int error, setbirthtime;
3131 
3132 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
3133 		return (error);
3134 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3135 	setbirthtime = 0;
3136 	if (numtimes < 3 && !VOP_GETATTR(vp, &vattr, td->td_ucred) &&
3137 	    timespeccmp(&ts[1], &vattr.va_birthtime, < ))
3138 		setbirthtime = 1;
3139 	VATTR_NULL(&vattr);
3140 	vattr.va_atime = ts[0];
3141 	vattr.va_mtime = ts[1];
3142 	if (setbirthtime)
3143 		vattr.va_birthtime = ts[1];
3144 	if (numtimes > 2)
3145 		vattr.va_birthtime = ts[2];
3146 	if (nullflag)
3147 		vattr.va_vaflags |= VA_UTIMES_NULL;
3148 #ifdef MAC
3149 	error = mac_vnode_check_setutimes(td->td_ucred, vp, vattr.va_atime,
3150 	    vattr.va_mtime);
3151 #endif
3152 	if (error == 0)
3153 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3154 	VOP_UNLOCK(vp, 0);
3155 	vn_finished_write(mp);
3156 	return (error);
3157 }
3158 
3159 /*
3160  * Set the access and modification times of a file.
3161  */
3162 #ifndef _SYS_SYSPROTO_H_
3163 struct utimes_args {
3164 	char	*path;
3165 	struct	timeval *tptr;
3166 };
3167 #endif
3168 int
3169 sys_utimes(td, uap)
3170 	struct thread *td;
3171 	register struct utimes_args /* {
3172 		char *path;
3173 		struct timeval *tptr;
3174 	} */ *uap;
3175 {
3176 
3177 	return (kern_utimes(td, uap->path, UIO_USERSPACE, uap->tptr,
3178 	    UIO_USERSPACE));
3179 }
3180 
3181 #ifndef _SYS_SYSPROTO_H_
3182 struct futimesat_args {
3183 	int fd;
3184 	const char * path;
3185 	const struct timeval * times;
3186 };
3187 #endif
3188 int
3189 sys_futimesat(struct thread *td, struct futimesat_args *uap)
3190 {
3191 
3192 	return (kern_utimesat(td, uap->fd, uap->path, UIO_USERSPACE,
3193 	    uap->times, UIO_USERSPACE));
3194 }
3195 
3196 int
3197 kern_utimes(struct thread *td, char *path, enum uio_seg pathseg,
3198     struct timeval *tptr, enum uio_seg tptrseg)
3199 {
3200 
3201 	return (kern_utimesat(td, AT_FDCWD, path, pathseg, tptr, tptrseg));
3202 }
3203 
3204 int
3205 kern_utimesat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
3206     struct timeval *tptr, enum uio_seg tptrseg)
3207 {
3208 	struct nameidata nd;
3209 	struct timespec ts[2];
3210 	cap_rights_t rights;
3211 	int error;
3212 
3213 	if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3214 		return (error);
3215 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, fd,
3216 	    cap_rights_init(&rights, CAP_FUTIMES), td);
3217 
3218 	if ((error = namei(&nd)) != 0)
3219 		return (error);
3220 	NDFREE(&nd, NDF_ONLY_PNBUF);
3221 	error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3222 	vrele(nd.ni_vp);
3223 	return (error);
3224 }
3225 
3226 /*
3227  * Set the access and modification times of a file.
3228  */
3229 #ifndef _SYS_SYSPROTO_H_
3230 struct lutimes_args {
3231 	char	*path;
3232 	struct	timeval *tptr;
3233 };
3234 #endif
3235 int
3236 sys_lutimes(td, uap)
3237 	struct thread *td;
3238 	register struct lutimes_args /* {
3239 		char *path;
3240 		struct timeval *tptr;
3241 	} */ *uap;
3242 {
3243 
3244 	return (kern_lutimes(td, uap->path, UIO_USERSPACE, uap->tptr,
3245 	    UIO_USERSPACE));
3246 }
3247 
3248 int
3249 kern_lutimes(struct thread *td, char *path, enum uio_seg pathseg,
3250     struct timeval *tptr, enum uio_seg tptrseg)
3251 {
3252 	struct timespec ts[2];
3253 	struct nameidata nd;
3254 	int error;
3255 
3256 	if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3257 		return (error);
3258 	NDINIT(&nd, LOOKUP, NOFOLLOW | AUDITVNODE1, pathseg, path, td);
3259 	if ((error = namei(&nd)) != 0)
3260 		return (error);
3261 	NDFREE(&nd, NDF_ONLY_PNBUF);
3262 	error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3263 	vrele(nd.ni_vp);
3264 	return (error);
3265 }
3266 
3267 /*
3268  * Set the access and modification times of a file.
3269  */
3270 #ifndef _SYS_SYSPROTO_H_
3271 struct futimes_args {
3272 	int	fd;
3273 	struct	timeval *tptr;
3274 };
3275 #endif
3276 int
3277 sys_futimes(td, uap)
3278 	struct thread *td;
3279 	register struct futimes_args /* {
3280 		int  fd;
3281 		struct timeval *tptr;
3282 	} */ *uap;
3283 {
3284 
3285 	return (kern_futimes(td, uap->fd, uap->tptr, UIO_USERSPACE));
3286 }
3287 
3288 int
3289 kern_futimes(struct thread *td, int fd, struct timeval *tptr,
3290     enum uio_seg tptrseg)
3291 {
3292 	struct timespec ts[2];
3293 	struct file *fp;
3294 	cap_rights_t rights;
3295 	int error;
3296 
3297 	AUDIT_ARG_FD(fd);
3298 	error = getutimes(tptr, tptrseg, ts);
3299 	if (error != 0)
3300 		return (error);
3301 	error = getvnode(td->td_proc->p_fd, fd,
3302 	    cap_rights_init(&rights, CAP_FUTIMES), &fp);
3303 	if (error != 0)
3304 		return (error);
3305 #ifdef AUDIT
3306 	vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
3307 	AUDIT_ARG_VNODE1(fp->f_vnode);
3308 	VOP_UNLOCK(fp->f_vnode, 0);
3309 #endif
3310 	error = setutimes(td, fp->f_vnode, ts, 2, tptr == NULL);
3311 	fdrop(fp, td);
3312 	return (error);
3313 }
3314 
3315 /*
3316  * Truncate a file given its path name.
3317  */
3318 #ifndef _SYS_SYSPROTO_H_
3319 struct truncate_args {
3320 	char	*path;
3321 	int	pad;
3322 	off_t	length;
3323 };
3324 #endif
3325 int
3326 sys_truncate(td, uap)
3327 	struct thread *td;
3328 	register struct truncate_args /* {
3329 		char *path;
3330 		int pad;
3331 		off_t length;
3332 	} */ *uap;
3333 {
3334 
3335 	return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3336 }
3337 
3338 int
3339 kern_truncate(struct thread *td, char *path, enum uio_seg pathseg, off_t length)
3340 {
3341 	struct mount *mp;
3342 	struct vnode *vp;
3343 	void *rl_cookie;
3344 	struct vattr vattr;
3345 	struct nameidata nd;
3346 	int error;
3347 
3348 	if (length < 0)
3349 		return(EINVAL);
3350 	NDINIT(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, td);
3351 	if ((error = namei(&nd)) != 0)
3352 		return (error);
3353 	vp = nd.ni_vp;
3354 	rl_cookie = vn_rangelock_wlock(vp, 0, OFF_MAX);
3355 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) {
3356 		vn_rangelock_unlock(vp, rl_cookie);
3357 		vrele(vp);
3358 		return (error);
3359 	}
3360 	NDFREE(&nd, NDF_ONLY_PNBUF);
3361 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3362 	if (vp->v_type == VDIR)
3363 		error = EISDIR;
3364 #ifdef MAC
3365 	else if ((error = mac_vnode_check_write(td->td_ucred, NOCRED, vp))) {
3366 	}
3367 #endif
3368 	else if ((error = vn_writechk(vp)) == 0 &&
3369 	    (error = VOP_ACCESS(vp, VWRITE, td->td_ucred, td)) == 0) {
3370 		VATTR_NULL(&vattr);
3371 		vattr.va_size = length;
3372 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3373 	}
3374 	VOP_UNLOCK(vp, 0);
3375 	vn_finished_write(mp);
3376 	vn_rangelock_unlock(vp, rl_cookie);
3377 	vrele(vp);
3378 	return (error);
3379 }
3380 
3381 #if defined(COMPAT_43)
3382 /*
3383  * Truncate a file given its path name.
3384  */
3385 #ifndef _SYS_SYSPROTO_H_
3386 struct otruncate_args {
3387 	char	*path;
3388 	long	length;
3389 };
3390 #endif
3391 int
3392 otruncate(td, uap)
3393 	struct thread *td;
3394 	register struct otruncate_args /* {
3395 		char *path;
3396 		long length;
3397 	} */ *uap;
3398 {
3399 	struct truncate_args /* {
3400 		char *path;
3401 		int pad;
3402 		off_t length;
3403 	} */ nuap;
3404 
3405 	nuap.path = uap->path;
3406 	nuap.length = uap->length;
3407 	return (sys_truncate(td, &nuap));
3408 }
3409 #endif /* COMPAT_43 */
3410 
3411 /* Versions with the pad argument */
3412 int
3413 freebsd6_truncate(struct thread *td, struct freebsd6_truncate_args *uap)
3414 {
3415 	struct truncate_args ouap;
3416 
3417 	ouap.path = uap->path;
3418 	ouap.length = uap->length;
3419 	return (sys_truncate(td, &ouap));
3420 }
3421 
3422 int
3423 freebsd6_ftruncate(struct thread *td, struct freebsd6_ftruncate_args *uap)
3424 {
3425 	struct ftruncate_args ouap;
3426 
3427 	ouap.fd = uap->fd;
3428 	ouap.length = uap->length;
3429 	return (sys_ftruncate(td, &ouap));
3430 }
3431 
3432 /*
3433  * Sync an open file.
3434  */
3435 #ifndef _SYS_SYSPROTO_H_
3436 struct fsync_args {
3437 	int	fd;
3438 };
3439 #endif
3440 int
3441 sys_fsync(td, uap)
3442 	struct thread *td;
3443 	struct fsync_args /* {
3444 		int fd;
3445 	} */ *uap;
3446 {
3447 	struct vnode *vp;
3448 	struct mount *mp;
3449 	struct file *fp;
3450 	cap_rights_t rights;
3451 	int error, lock_flags;
3452 
3453 	AUDIT_ARG_FD(uap->fd);
3454 	error = getvnode(td->td_proc->p_fd, uap->fd,
3455 	    cap_rights_init(&rights, CAP_FSYNC), &fp);
3456 	if (error != 0)
3457 		return (error);
3458 	vp = fp->f_vnode;
3459 	error = vn_start_write(vp, &mp, V_WAIT | PCATCH);
3460 	if (error != 0)
3461 		goto drop;
3462 	if (MNT_SHARED_WRITES(mp) ||
3463 	    ((mp == NULL) && MNT_SHARED_WRITES(vp->v_mount))) {
3464 		lock_flags = LK_SHARED;
3465 	} else {
3466 		lock_flags = LK_EXCLUSIVE;
3467 	}
3468 	vn_lock(vp, lock_flags | LK_RETRY);
3469 	AUDIT_ARG_VNODE1(vp);
3470 	if (vp->v_object != NULL) {
3471 		VM_OBJECT_WLOCK(vp->v_object);
3472 		vm_object_page_clean(vp->v_object, 0, 0, 0);
3473 		VM_OBJECT_WUNLOCK(vp->v_object);
3474 	}
3475 	error = VOP_FSYNC(vp, MNT_WAIT, td);
3476 
3477 	VOP_UNLOCK(vp, 0);
3478 	vn_finished_write(mp);
3479 drop:
3480 	fdrop(fp, td);
3481 	return (error);
3482 }
3483 
3484 /*
3485  * Rename files.  Source and destination must either both be directories, or
3486  * both not be directories.  If target is a directory, it must be empty.
3487  */
3488 #ifndef _SYS_SYSPROTO_H_
3489 struct rename_args {
3490 	char	*from;
3491 	char	*to;
3492 };
3493 #endif
3494 int
3495 sys_rename(td, uap)
3496 	struct thread *td;
3497 	register struct rename_args /* {
3498 		char *from;
3499 		char *to;
3500 	} */ *uap;
3501 {
3502 
3503 	return (kern_rename(td, uap->from, uap->to, UIO_USERSPACE));
3504 }
3505 
3506 #ifndef _SYS_SYSPROTO_H_
3507 struct renameat_args {
3508 	int	oldfd;
3509 	char	*old;
3510 	int	newfd;
3511 	char	*new;
3512 };
3513 #endif
3514 int
3515 sys_renameat(struct thread *td, struct renameat_args *uap)
3516 {
3517 
3518 	return (kern_renameat(td, uap->oldfd, uap->old, uap->newfd, uap->new,
3519 	    UIO_USERSPACE));
3520 }
3521 
3522 int
3523 kern_rename(struct thread *td, char *from, char *to, enum uio_seg pathseg)
3524 {
3525 
3526 	return (kern_renameat(td, AT_FDCWD, from, AT_FDCWD, to, pathseg));
3527 }
3528 
3529 int
3530 kern_renameat(struct thread *td, int oldfd, char *old, int newfd, char *new,
3531     enum uio_seg pathseg)
3532 {
3533 	struct mount *mp = NULL;
3534 	struct vnode *tvp, *fvp, *tdvp;
3535 	struct nameidata fromnd, tond;
3536 	cap_rights_t rights;
3537 	int error;
3538 
3539 again:
3540 	bwillwrite();
3541 #ifdef MAC
3542 	NDINIT_ATRIGHTS(&fromnd, DELETE, LOCKPARENT | LOCKLEAF | SAVESTART |
3543 	    AUDITVNODE1, pathseg, old, oldfd,
3544 	    cap_rights_init(&rights, CAP_RENAMEAT), td);
3545 #else
3546 	NDINIT_ATRIGHTS(&fromnd, DELETE, WANTPARENT | SAVESTART | AUDITVNODE1,
3547 	    pathseg, old, oldfd, cap_rights_init(&rights, CAP_RENAMEAT), td);
3548 #endif
3549 
3550 	if ((error = namei(&fromnd)) != 0)
3551 		return (error);
3552 #ifdef MAC
3553 	error = mac_vnode_check_rename_from(td->td_ucred, fromnd.ni_dvp,
3554 	    fromnd.ni_vp, &fromnd.ni_cnd);
3555 	VOP_UNLOCK(fromnd.ni_dvp, 0);
3556 	if (fromnd.ni_dvp != fromnd.ni_vp)
3557 		VOP_UNLOCK(fromnd.ni_vp, 0);
3558 #endif
3559 	fvp = fromnd.ni_vp;
3560 	NDINIT_ATRIGHTS(&tond, RENAME, LOCKPARENT | LOCKLEAF | NOCACHE |
3561 	    SAVESTART | AUDITVNODE2, pathseg, new, newfd,
3562 	    cap_rights_init(&rights, CAP_LINKAT), td);
3563 	if (fromnd.ni_vp->v_type == VDIR)
3564 		tond.ni_cnd.cn_flags |= WILLBEDIR;
3565 	if ((error = namei(&tond)) != 0) {
3566 		/* Translate error code for rename("dir1", "dir2/."). */
3567 		if (error == EISDIR && fvp->v_type == VDIR)
3568 			error = EINVAL;
3569 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3570 		vrele(fromnd.ni_dvp);
3571 		vrele(fvp);
3572 		goto out1;
3573 	}
3574 	tdvp = tond.ni_dvp;
3575 	tvp = tond.ni_vp;
3576 	error = vn_start_write(fvp, &mp, V_NOWAIT);
3577 	if (error != 0) {
3578 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3579 		NDFREE(&tond, NDF_ONLY_PNBUF);
3580 		if (tvp != NULL)
3581 			vput(tvp);
3582 		if (tdvp == tvp)
3583 			vrele(tdvp);
3584 		else
3585 			vput(tdvp);
3586 		vrele(fromnd.ni_dvp);
3587 		vrele(fvp);
3588 		vrele(tond.ni_startdir);
3589 		if (fromnd.ni_startdir != NULL)
3590 			vrele(fromnd.ni_startdir);
3591 		error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH);
3592 		if (error != 0)
3593 			return (error);
3594 		goto again;
3595 	}
3596 	if (tvp != NULL) {
3597 		if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
3598 			error = ENOTDIR;
3599 			goto out;
3600 		} else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
3601 			error = EISDIR;
3602 			goto out;
3603 		}
3604 #ifdef CAPABILITIES
3605 		if (newfd != AT_FDCWD) {
3606 			/*
3607 			 * If the target already exists we require CAP_UNLINKAT
3608 			 * from 'newfd'.
3609 			 */
3610 			error = cap_check(&tond.ni_filecaps.fc_rights,
3611 			    cap_rights_init(&rights, CAP_UNLINKAT));
3612 			if (error != 0)
3613 				goto out;
3614 		}
3615 #endif
3616 	}
3617 	if (fvp == tdvp) {
3618 		error = EINVAL;
3619 		goto out;
3620 	}
3621 	/*
3622 	 * If the source is the same as the destination (that is, if they
3623 	 * are links to the same vnode), then there is nothing to do.
3624 	 */
3625 	if (fvp == tvp)
3626 		error = -1;
3627 #ifdef MAC
3628 	else
3629 		error = mac_vnode_check_rename_to(td->td_ucred, tdvp,
3630 		    tond.ni_vp, fromnd.ni_dvp == tdvp, &tond.ni_cnd);
3631 #endif
3632 out:
3633 	if (error == 0) {
3634 		error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
3635 		    tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
3636 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3637 		NDFREE(&tond, NDF_ONLY_PNBUF);
3638 	} else {
3639 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3640 		NDFREE(&tond, NDF_ONLY_PNBUF);
3641 		if (tvp != NULL)
3642 			vput(tvp);
3643 		if (tdvp == tvp)
3644 			vrele(tdvp);
3645 		else
3646 			vput(tdvp);
3647 		vrele(fromnd.ni_dvp);
3648 		vrele(fvp);
3649 	}
3650 	vrele(tond.ni_startdir);
3651 	vn_finished_write(mp);
3652 out1:
3653 	if (fromnd.ni_startdir)
3654 		vrele(fromnd.ni_startdir);
3655 	if (error == -1)
3656 		return (0);
3657 	return (error);
3658 }
3659 
3660 /*
3661  * Make a directory file.
3662  */
3663 #ifndef _SYS_SYSPROTO_H_
3664 struct mkdir_args {
3665 	char	*path;
3666 	int	mode;
3667 };
3668 #endif
3669 int
3670 sys_mkdir(td, uap)
3671 	struct thread *td;
3672 	register struct mkdir_args /* {
3673 		char *path;
3674 		int mode;
3675 	} */ *uap;
3676 {
3677 
3678 	return (kern_mkdir(td, uap->path, UIO_USERSPACE, uap->mode));
3679 }
3680 
3681 #ifndef _SYS_SYSPROTO_H_
3682 struct mkdirat_args {
3683 	int	fd;
3684 	char	*path;
3685 	mode_t	mode;
3686 };
3687 #endif
3688 int
3689 sys_mkdirat(struct thread *td, struct mkdirat_args *uap)
3690 {
3691 
3692 	return (kern_mkdirat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode));
3693 }
3694 
3695 int
3696 kern_mkdir(struct thread *td, char *path, enum uio_seg segflg, int mode)
3697 {
3698 
3699 	return (kern_mkdirat(td, AT_FDCWD, path, segflg, mode));
3700 }
3701 
3702 int
3703 kern_mkdirat(struct thread *td, int fd, char *path, enum uio_seg segflg,
3704     int mode)
3705 {
3706 	struct mount *mp;
3707 	struct vnode *vp;
3708 	struct vattr vattr;
3709 	struct nameidata nd;
3710 	cap_rights_t rights;
3711 	int error;
3712 
3713 	AUDIT_ARG_MODE(mode);
3714 restart:
3715 	bwillwrite();
3716 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1,
3717 	    segflg, path, fd, cap_rights_init(&rights, CAP_MKDIRAT), td);
3718 	nd.ni_cnd.cn_flags |= WILLBEDIR;
3719 	if ((error = namei(&nd)) != 0)
3720 		return (error);
3721 	vp = nd.ni_vp;
3722 	if (vp != NULL) {
3723 		NDFREE(&nd, NDF_ONLY_PNBUF);
3724 		/*
3725 		 * XXX namei called with LOCKPARENT but not LOCKLEAF has
3726 		 * the strange behaviour of leaving the vnode unlocked
3727 		 * if the target is the same vnode as the parent.
3728 		 */
3729 		if (vp == nd.ni_dvp)
3730 			vrele(nd.ni_dvp);
3731 		else
3732 			vput(nd.ni_dvp);
3733 		vrele(vp);
3734 		return (EEXIST);
3735 	}
3736 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3737 		NDFREE(&nd, NDF_ONLY_PNBUF);
3738 		vput(nd.ni_dvp);
3739 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3740 			return (error);
3741 		goto restart;
3742 	}
3743 	VATTR_NULL(&vattr);
3744 	vattr.va_type = VDIR;
3745 	vattr.va_mode = (mode & ACCESSPERMS) &~ td->td_proc->p_fd->fd_cmask;
3746 #ifdef MAC
3747 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
3748 	    &vattr);
3749 	if (error != 0)
3750 		goto out;
3751 #endif
3752 	error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
3753 #ifdef MAC
3754 out:
3755 #endif
3756 	NDFREE(&nd, NDF_ONLY_PNBUF);
3757 	vput(nd.ni_dvp);
3758 	if (error == 0)
3759 		vput(nd.ni_vp);
3760 	vn_finished_write(mp);
3761 	return (error);
3762 }
3763 
3764 /*
3765  * Remove a directory file.
3766  */
3767 #ifndef _SYS_SYSPROTO_H_
3768 struct rmdir_args {
3769 	char	*path;
3770 };
3771 #endif
3772 int
3773 sys_rmdir(td, uap)
3774 	struct thread *td;
3775 	struct rmdir_args /* {
3776 		char *path;
3777 	} */ *uap;
3778 {
3779 
3780 	return (kern_rmdir(td, uap->path, UIO_USERSPACE));
3781 }
3782 
3783 int
3784 kern_rmdir(struct thread *td, char *path, enum uio_seg pathseg)
3785 {
3786 
3787 	return (kern_rmdirat(td, AT_FDCWD, path, pathseg));
3788 }
3789 
3790 int
3791 kern_rmdirat(struct thread *td, int fd, char *path, enum uio_seg pathseg)
3792 {
3793 	struct mount *mp;
3794 	struct vnode *vp;
3795 	struct nameidata nd;
3796 	cap_rights_t rights;
3797 	int error;
3798 
3799 restart:
3800 	bwillwrite();
3801 	NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1,
3802 	    pathseg, path, fd, cap_rights_init(&rights, CAP_UNLINKAT), td);
3803 	if ((error = namei(&nd)) != 0)
3804 		return (error);
3805 	vp = nd.ni_vp;
3806 	if (vp->v_type != VDIR) {
3807 		error = ENOTDIR;
3808 		goto out;
3809 	}
3810 	/*
3811 	 * No rmdir "." please.
3812 	 */
3813 	if (nd.ni_dvp == vp) {
3814 		error = EINVAL;
3815 		goto out;
3816 	}
3817 	/*
3818 	 * The root of a mounted filesystem cannot be deleted.
3819 	 */
3820 	if (vp->v_vflag & VV_ROOT) {
3821 		error = EBUSY;
3822 		goto out;
3823 	}
3824 #ifdef MAC
3825 	error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
3826 	    &nd.ni_cnd);
3827 	if (error != 0)
3828 		goto out;
3829 #endif
3830 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3831 		NDFREE(&nd, NDF_ONLY_PNBUF);
3832 		vput(vp);
3833 		if (nd.ni_dvp == vp)
3834 			vrele(nd.ni_dvp);
3835 		else
3836 			vput(nd.ni_dvp);
3837 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3838 			return (error);
3839 		goto restart;
3840 	}
3841 	vfs_notify_upper(vp, VFS_NOTIFY_UPPER_UNLINK);
3842 	error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
3843 	vn_finished_write(mp);
3844 out:
3845 	NDFREE(&nd, NDF_ONLY_PNBUF);
3846 	vput(vp);
3847 	if (nd.ni_dvp == vp)
3848 		vrele(nd.ni_dvp);
3849 	else
3850 		vput(nd.ni_dvp);
3851 	return (error);
3852 }
3853 
3854 #ifdef COMPAT_43
3855 /*
3856  * Read a block of directory entries in a filesystem independent format.
3857  */
3858 #ifndef _SYS_SYSPROTO_H_
3859 struct ogetdirentries_args {
3860 	int	fd;
3861 	char	*buf;
3862 	u_int	count;
3863 	long	*basep;
3864 };
3865 #endif
3866 int
3867 ogetdirentries(struct thread *td, struct ogetdirentries_args *uap)
3868 {
3869 	long loff;
3870 	int error;
3871 
3872 	error = kern_ogetdirentries(td, uap, &loff);
3873 	if (error == 0)
3874 		error = copyout(&loff, uap->basep, sizeof(long));
3875 	return (error);
3876 }
3877 
3878 int
3879 kern_ogetdirentries(struct thread *td, struct ogetdirentries_args *uap,
3880     long *ploff)
3881 {
3882 	struct vnode *vp;
3883 	struct file *fp;
3884 	struct uio auio, kuio;
3885 	struct iovec aiov, kiov;
3886 	struct dirent *dp, *edp;
3887 	cap_rights_t rights;
3888 	caddr_t dirbuf;
3889 	int error, eofflag, readcnt;
3890 	long loff;
3891 	off_t foffset;
3892 
3893 	/* XXX arbitrary sanity limit on `count'. */
3894 	if (uap->count > 64 * 1024)
3895 		return (EINVAL);
3896 	error = getvnode(td->td_proc->p_fd, uap->fd,
3897 	    cap_rights_init(&rights, CAP_READ), &fp);
3898 	if (error != 0)
3899 		return (error);
3900 	if ((fp->f_flag & FREAD) == 0) {
3901 		fdrop(fp, td);
3902 		return (EBADF);
3903 	}
3904 	vp = fp->f_vnode;
3905 	foffset = foffset_lock(fp, 0);
3906 unionread:
3907 	if (vp->v_type != VDIR) {
3908 		foffset_unlock(fp, foffset, 0);
3909 		fdrop(fp, td);
3910 		return (EINVAL);
3911 	}
3912 	aiov.iov_base = uap->buf;
3913 	aiov.iov_len = uap->count;
3914 	auio.uio_iov = &aiov;
3915 	auio.uio_iovcnt = 1;
3916 	auio.uio_rw = UIO_READ;
3917 	auio.uio_segflg = UIO_USERSPACE;
3918 	auio.uio_td = td;
3919 	auio.uio_resid = uap->count;
3920 	vn_lock(vp, LK_SHARED | LK_RETRY);
3921 	loff = auio.uio_offset = foffset;
3922 #ifdef MAC
3923 	error = mac_vnode_check_readdir(td->td_ucred, vp);
3924 	if (error != 0) {
3925 		VOP_UNLOCK(vp, 0);
3926 		foffset_unlock(fp, foffset, FOF_NOUPDATE);
3927 		fdrop(fp, td);
3928 		return (error);
3929 	}
3930 #endif
3931 #	if (BYTE_ORDER != LITTLE_ENDIAN)
3932 		if (vp->v_mount->mnt_maxsymlinklen <= 0) {
3933 			error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag,
3934 			    NULL, NULL);
3935 			foffset = auio.uio_offset;
3936 		} else
3937 #	endif
3938 	{
3939 		kuio = auio;
3940 		kuio.uio_iov = &kiov;
3941 		kuio.uio_segflg = UIO_SYSSPACE;
3942 		kiov.iov_len = uap->count;
3943 		dirbuf = malloc(uap->count, M_TEMP, M_WAITOK);
3944 		kiov.iov_base = dirbuf;
3945 		error = VOP_READDIR(vp, &kuio, fp->f_cred, &eofflag,
3946 			    NULL, NULL);
3947 		foffset = kuio.uio_offset;
3948 		if (error == 0) {
3949 			readcnt = uap->count - kuio.uio_resid;
3950 			edp = (struct dirent *)&dirbuf[readcnt];
3951 			for (dp = (struct dirent *)dirbuf; dp < edp; ) {
3952 #				if (BYTE_ORDER == LITTLE_ENDIAN)
3953 					/*
3954 					 * The expected low byte of
3955 					 * dp->d_namlen is our dp->d_type.
3956 					 * The high MBZ byte of dp->d_namlen
3957 					 * is our dp->d_namlen.
3958 					 */
3959 					dp->d_type = dp->d_namlen;
3960 					dp->d_namlen = 0;
3961 #				else
3962 					/*
3963 					 * The dp->d_type is the high byte
3964 					 * of the expected dp->d_namlen,
3965 					 * so must be zero'ed.
3966 					 */
3967 					dp->d_type = 0;
3968 #				endif
3969 				if (dp->d_reclen > 0) {
3970 					dp = (struct dirent *)
3971 					    ((char *)dp + dp->d_reclen);
3972 				} else {
3973 					error = EIO;
3974 					break;
3975 				}
3976 			}
3977 			if (dp >= edp)
3978 				error = uiomove(dirbuf, readcnt, &auio);
3979 		}
3980 		free(dirbuf, M_TEMP);
3981 	}
3982 	if (error != 0) {
3983 		VOP_UNLOCK(vp, 0);
3984 		foffset_unlock(fp, foffset, 0);
3985 		fdrop(fp, td);
3986 		return (error);
3987 	}
3988 	if (uap->count == auio.uio_resid &&
3989 	    (vp->v_vflag & VV_ROOT) &&
3990 	    (vp->v_mount->mnt_flag & MNT_UNION)) {
3991 		struct vnode *tvp = vp;
3992 		vp = vp->v_mount->mnt_vnodecovered;
3993 		VREF(vp);
3994 		fp->f_vnode = vp;
3995 		fp->f_data = vp;
3996 		foffset = 0;
3997 		vput(tvp);
3998 		goto unionread;
3999 	}
4000 	VOP_UNLOCK(vp, 0);
4001 	foffset_unlock(fp, foffset, 0);
4002 	fdrop(fp, td);
4003 	td->td_retval[0] = uap->count - auio.uio_resid;
4004 	if (error == 0)
4005 		*ploff = loff;
4006 	return (error);
4007 }
4008 #endif /* COMPAT_43 */
4009 
4010 /*
4011  * Read a block of directory entries in a filesystem independent format.
4012  */
4013 #ifndef _SYS_SYSPROTO_H_
4014 struct getdirentries_args {
4015 	int	fd;
4016 	char	*buf;
4017 	u_int	count;
4018 	long	*basep;
4019 };
4020 #endif
4021 int
4022 sys_getdirentries(td, uap)
4023 	struct thread *td;
4024 	register struct getdirentries_args /* {
4025 		int fd;
4026 		char *buf;
4027 		u_int count;
4028 		long *basep;
4029 	} */ *uap;
4030 {
4031 	long base;
4032 	int error;
4033 
4034 	error = kern_getdirentries(td, uap->fd, uap->buf, uap->count, &base,
4035 	    NULL, UIO_USERSPACE);
4036 	if (error != 0)
4037 		return (error);
4038 	if (uap->basep != NULL)
4039 		error = copyout(&base, uap->basep, sizeof(long));
4040 	return (error);
4041 }
4042 
4043 int
4044 kern_getdirentries(struct thread *td, int fd, char *buf, u_int count,
4045     long *basep, ssize_t *residp, enum uio_seg bufseg)
4046 {
4047 	struct vnode *vp;
4048 	struct file *fp;
4049 	struct uio auio;
4050 	struct iovec aiov;
4051 	cap_rights_t rights;
4052 	long loff;
4053 	int error, eofflag;
4054 	off_t foffset;
4055 
4056 	AUDIT_ARG_FD(fd);
4057 	if (count > IOSIZE_MAX)
4058 		return (EINVAL);
4059 	auio.uio_resid = count;
4060 	error = getvnode(td->td_proc->p_fd, fd,
4061 	    cap_rights_init(&rights, CAP_READ), &fp);
4062 	if (error != 0)
4063 		return (error);
4064 	if ((fp->f_flag & FREAD) == 0) {
4065 		fdrop(fp, td);
4066 		return (EBADF);
4067 	}
4068 	vp = fp->f_vnode;
4069 	foffset = foffset_lock(fp, 0);
4070 unionread:
4071 	if (vp->v_type != VDIR) {
4072 		error = EINVAL;
4073 		goto fail;
4074 	}
4075 	aiov.iov_base = buf;
4076 	aiov.iov_len = count;
4077 	auio.uio_iov = &aiov;
4078 	auio.uio_iovcnt = 1;
4079 	auio.uio_rw = UIO_READ;
4080 	auio.uio_segflg = bufseg;
4081 	auio.uio_td = td;
4082 	vn_lock(vp, LK_SHARED | LK_RETRY);
4083 	AUDIT_ARG_VNODE1(vp);
4084 	loff = auio.uio_offset = foffset;
4085 #ifdef MAC
4086 	error = mac_vnode_check_readdir(td->td_ucred, vp);
4087 	if (error == 0)
4088 #endif
4089 		error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, NULL,
4090 		    NULL);
4091 	foffset = auio.uio_offset;
4092 	if (error != 0) {
4093 		VOP_UNLOCK(vp, 0);
4094 		goto fail;
4095 	}
4096 	if (count == auio.uio_resid &&
4097 	    (vp->v_vflag & VV_ROOT) &&
4098 	    (vp->v_mount->mnt_flag & MNT_UNION)) {
4099 		struct vnode *tvp = vp;
4100 
4101 		vp = vp->v_mount->mnt_vnodecovered;
4102 		VREF(vp);
4103 		fp->f_vnode = vp;
4104 		fp->f_data = vp;
4105 		foffset = 0;
4106 		vput(tvp);
4107 		goto unionread;
4108 	}
4109 	VOP_UNLOCK(vp, 0);
4110 	*basep = loff;
4111 	if (residp != NULL)
4112 		*residp = auio.uio_resid;
4113 	td->td_retval[0] = count - auio.uio_resid;
4114 fail:
4115 	foffset_unlock(fp, foffset, 0);
4116 	fdrop(fp, td);
4117 	return (error);
4118 }
4119 
4120 #ifndef _SYS_SYSPROTO_H_
4121 struct getdents_args {
4122 	int fd;
4123 	char *buf;
4124 	size_t count;
4125 };
4126 #endif
4127 int
4128 sys_getdents(td, uap)
4129 	struct thread *td;
4130 	register struct getdents_args /* {
4131 		int fd;
4132 		char *buf;
4133 		u_int count;
4134 	} */ *uap;
4135 {
4136 	struct getdirentries_args ap;
4137 
4138 	ap.fd = uap->fd;
4139 	ap.buf = uap->buf;
4140 	ap.count = uap->count;
4141 	ap.basep = NULL;
4142 	return (sys_getdirentries(td, &ap));
4143 }
4144 
4145 /*
4146  * Set the mode mask for creation of filesystem nodes.
4147  */
4148 #ifndef _SYS_SYSPROTO_H_
4149 struct umask_args {
4150 	int	newmask;
4151 };
4152 #endif
4153 int
4154 sys_umask(td, uap)
4155 	struct thread *td;
4156 	struct umask_args /* {
4157 		int newmask;
4158 	} */ *uap;
4159 {
4160 	register struct filedesc *fdp;
4161 
4162 	FILEDESC_XLOCK(td->td_proc->p_fd);
4163 	fdp = td->td_proc->p_fd;
4164 	td->td_retval[0] = fdp->fd_cmask;
4165 	fdp->fd_cmask = uap->newmask & ALLPERMS;
4166 	FILEDESC_XUNLOCK(td->td_proc->p_fd);
4167 	return (0);
4168 }
4169 
4170 /*
4171  * Void all references to file by ripping underlying filesystem away from
4172  * vnode.
4173  */
4174 #ifndef _SYS_SYSPROTO_H_
4175 struct revoke_args {
4176 	char	*path;
4177 };
4178 #endif
4179 int
4180 sys_revoke(td, uap)
4181 	struct thread *td;
4182 	register struct revoke_args /* {
4183 		char *path;
4184 	} */ *uap;
4185 {
4186 	struct vnode *vp;
4187 	struct vattr vattr;
4188 	struct nameidata nd;
4189 	int error;
4190 
4191 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
4192 	    uap->path, td);
4193 	if ((error = namei(&nd)) != 0)
4194 		return (error);
4195 	vp = nd.ni_vp;
4196 	NDFREE(&nd, NDF_ONLY_PNBUF);
4197 	if (vp->v_type != VCHR || vp->v_rdev == NULL) {
4198 		error = EINVAL;
4199 		goto out;
4200 	}
4201 #ifdef MAC
4202 	error = mac_vnode_check_revoke(td->td_ucred, vp);
4203 	if (error != 0)
4204 		goto out;
4205 #endif
4206 	error = VOP_GETATTR(vp, &vattr, td->td_ucred);
4207 	if (error != 0)
4208 		goto out;
4209 	if (td->td_ucred->cr_uid != vattr.va_uid) {
4210 		error = priv_check(td, PRIV_VFS_ADMIN);
4211 		if (error != 0)
4212 			goto out;
4213 	}
4214 	if (vcount(vp) > 1)
4215 		VOP_REVOKE(vp, REVOKEALL);
4216 out:
4217 	vput(vp);
4218 	return (error);
4219 }
4220 
4221 /*
4222  * Convert a user file descriptor to a kernel file entry and check that, if it
4223  * is a capability, the correct rights are present. A reference on the file
4224  * entry is held upon returning.
4225  */
4226 int
4227 getvnode(struct filedesc *fdp, int fd, cap_rights_t *rightsp, struct file **fpp)
4228 {
4229 	struct file *fp;
4230 	int error;
4231 
4232 	error = fget_unlocked(fdp, fd, rightsp, 0, &fp, NULL);
4233 	if (error != 0)
4234 		return (error);
4235 
4236 	/*
4237 	 * The file could be not of the vnode type, or it may be not
4238 	 * yet fully initialized, in which case the f_vnode pointer
4239 	 * may be set, but f_ops is still badfileops.  E.g.,
4240 	 * devfs_open() transiently create such situation to
4241 	 * facilitate csw d_fdopen().
4242 	 *
4243 	 * Dupfdopen() handling in kern_openat() installs the
4244 	 * half-baked file into the process descriptor table, allowing
4245 	 * other thread to dereference it. Guard against the race by
4246 	 * checking f_ops.
4247 	 */
4248 	if (fp->f_vnode == NULL || fp->f_ops == &badfileops) {
4249 		fdrop(fp, curthread);
4250 		return (EINVAL);
4251 	}
4252 	*fpp = fp;
4253 	return (0);
4254 }
4255 
4256 
4257 /*
4258  * Get an (NFS) file handle.
4259  */
4260 #ifndef _SYS_SYSPROTO_H_
4261 struct lgetfh_args {
4262 	char	*fname;
4263 	fhandle_t *fhp;
4264 };
4265 #endif
4266 int
4267 sys_lgetfh(td, uap)
4268 	struct thread *td;
4269 	register struct lgetfh_args *uap;
4270 {
4271 	struct nameidata nd;
4272 	fhandle_t fh;
4273 	register struct vnode *vp;
4274 	int error;
4275 
4276 	error = priv_check(td, PRIV_VFS_GETFH);
4277 	if (error != 0)
4278 		return (error);
4279 	NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
4280 	    uap->fname, td);
4281 	error = namei(&nd);
4282 	if (error != 0)
4283 		return (error);
4284 	NDFREE(&nd, NDF_ONLY_PNBUF);
4285 	vp = nd.ni_vp;
4286 	bzero(&fh, sizeof(fh));
4287 	fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
4288 	error = VOP_VPTOFH(vp, &fh.fh_fid);
4289 	vput(vp);
4290 	if (error == 0)
4291 		error = copyout(&fh, uap->fhp, sizeof (fh));
4292 	return (error);
4293 }
4294 
4295 #ifndef _SYS_SYSPROTO_H_
4296 struct getfh_args {
4297 	char	*fname;
4298 	fhandle_t *fhp;
4299 };
4300 #endif
4301 int
4302 sys_getfh(td, uap)
4303 	struct thread *td;
4304 	register struct getfh_args *uap;
4305 {
4306 	struct nameidata nd;
4307 	fhandle_t fh;
4308 	register struct vnode *vp;
4309 	int error;
4310 
4311 	error = priv_check(td, PRIV_VFS_GETFH);
4312 	if (error != 0)
4313 		return (error);
4314 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
4315 	    uap->fname, td);
4316 	error = namei(&nd);
4317 	if (error != 0)
4318 		return (error);
4319 	NDFREE(&nd, NDF_ONLY_PNBUF);
4320 	vp = nd.ni_vp;
4321 	bzero(&fh, sizeof(fh));
4322 	fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
4323 	error = VOP_VPTOFH(vp, &fh.fh_fid);
4324 	vput(vp);
4325 	if (error == 0)
4326 		error = copyout(&fh, uap->fhp, sizeof (fh));
4327 	return (error);
4328 }
4329 
4330 /*
4331  * syscall for the rpc.lockd to use to translate a NFS file handle into an
4332  * open descriptor.
4333  *
4334  * warning: do not remove the priv_check() call or this becomes one giant
4335  * security hole.
4336  */
4337 #ifndef _SYS_SYSPROTO_H_
4338 struct fhopen_args {
4339 	const struct fhandle *u_fhp;
4340 	int flags;
4341 };
4342 #endif
4343 int
4344 sys_fhopen(td, uap)
4345 	struct thread *td;
4346 	struct fhopen_args /* {
4347 		const struct fhandle *u_fhp;
4348 		int flags;
4349 	} */ *uap;
4350 {
4351 	struct mount *mp;
4352 	struct vnode *vp;
4353 	struct fhandle fhp;
4354 	struct file *fp;
4355 	int fmode, error;
4356 	int indx;
4357 
4358 	error = priv_check(td, PRIV_VFS_FHOPEN);
4359 	if (error != 0)
4360 		return (error);
4361 	indx = -1;
4362 	fmode = FFLAGS(uap->flags);
4363 	/* why not allow a non-read/write open for our lockd? */
4364 	if (((fmode & (FREAD | FWRITE)) == 0) || (fmode & O_CREAT))
4365 		return (EINVAL);
4366 	error = copyin(uap->u_fhp, &fhp, sizeof(fhp));
4367 	if (error != 0)
4368 		return(error);
4369 	/* find the mount point */
4370 	mp = vfs_busyfs(&fhp.fh_fsid);
4371 	if (mp == NULL)
4372 		return (ESTALE);
4373 	/* now give me my vnode, it gets returned to me locked */
4374 	error = VFS_FHTOVP(mp, &fhp.fh_fid, LK_EXCLUSIVE, &vp);
4375 	vfs_unbusy(mp);
4376 	if (error != 0)
4377 		return (error);
4378 
4379 	error = falloc_noinstall(td, &fp);
4380 	if (error != 0) {
4381 		vput(vp);
4382 		return (error);
4383 	}
4384 	/*
4385 	 * An extra reference on `fp' has been held for us by
4386 	 * falloc_noinstall().
4387 	 */
4388 
4389 #ifdef INVARIANTS
4390 	td->td_dupfd = -1;
4391 #endif
4392 	error = vn_open_vnode(vp, fmode, td->td_ucred, td, fp);
4393 	if (error != 0) {
4394 		KASSERT(fp->f_ops == &badfileops,
4395 		    ("VOP_OPEN in fhopen() set f_ops"));
4396 		KASSERT(td->td_dupfd < 0,
4397 		    ("fhopen() encountered fdopen()"));
4398 
4399 		vput(vp);
4400 		goto bad;
4401 	}
4402 #ifdef INVARIANTS
4403 	td->td_dupfd = 0;
4404 #endif
4405 	fp->f_vnode = vp;
4406 	fp->f_seqcount = 1;
4407 	finit(fp, (fmode & FMASK) | (fp->f_flag & FHASLOCK), DTYPE_VNODE, vp,
4408 	    &vnops);
4409 	VOP_UNLOCK(vp, 0);
4410 	if ((fmode & O_TRUNC) != 0) {
4411 		error = fo_truncate(fp, 0, td->td_ucred, td);
4412 		if (error != 0)
4413 			goto bad;
4414 	}
4415 
4416 	error = finstall(td, fp, &indx, fmode, NULL);
4417 bad:
4418 	fdrop(fp, td);
4419 	td->td_retval[0] = indx;
4420 	return (error);
4421 }
4422 
4423 /*
4424  * Stat an (NFS) file handle.
4425  */
4426 #ifndef _SYS_SYSPROTO_H_
4427 struct fhstat_args {
4428 	struct fhandle *u_fhp;
4429 	struct stat *sb;
4430 };
4431 #endif
4432 int
4433 sys_fhstat(td, uap)
4434 	struct thread *td;
4435 	register struct fhstat_args /* {
4436 		struct fhandle *u_fhp;
4437 		struct stat *sb;
4438 	} */ *uap;
4439 {
4440 	struct stat sb;
4441 	struct fhandle fh;
4442 	int error;
4443 
4444 	error = copyin(uap->u_fhp, &fh, sizeof(fh));
4445 	if (error != 0)
4446 		return (error);
4447 	error = kern_fhstat(td, fh, &sb);
4448 	if (error == 0)
4449 		error = copyout(&sb, uap->sb, sizeof(sb));
4450 	return (error);
4451 }
4452 
4453 int
4454 kern_fhstat(struct thread *td, struct fhandle fh, struct stat *sb)
4455 {
4456 	struct mount *mp;
4457 	struct vnode *vp;
4458 	int error;
4459 
4460 	error = priv_check(td, PRIV_VFS_FHSTAT);
4461 	if (error != 0)
4462 		return (error);
4463 	if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4464 		return (ESTALE);
4465 	error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4466 	vfs_unbusy(mp);
4467 	if (error != 0)
4468 		return (error);
4469 	error = vn_stat(vp, sb, td->td_ucred, NOCRED, td);
4470 	vput(vp);
4471 	return (error);
4472 }
4473 
4474 /*
4475  * Implement fstatfs() for (NFS) file handles.
4476  */
4477 #ifndef _SYS_SYSPROTO_H_
4478 struct fhstatfs_args {
4479 	struct fhandle *u_fhp;
4480 	struct statfs *buf;
4481 };
4482 #endif
4483 int
4484 sys_fhstatfs(td, uap)
4485 	struct thread *td;
4486 	struct fhstatfs_args /* {
4487 		struct fhandle *u_fhp;
4488 		struct statfs *buf;
4489 	} */ *uap;
4490 {
4491 	struct statfs sf;
4492 	fhandle_t fh;
4493 	int error;
4494 
4495 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
4496 	if (error != 0)
4497 		return (error);
4498 	error = kern_fhstatfs(td, fh, &sf);
4499 	if (error != 0)
4500 		return (error);
4501 	return (copyout(&sf, uap->buf, sizeof(sf)));
4502 }
4503 
4504 int
4505 kern_fhstatfs(struct thread *td, fhandle_t fh, struct statfs *buf)
4506 {
4507 	struct statfs *sp;
4508 	struct mount *mp;
4509 	struct vnode *vp;
4510 	int error;
4511 
4512 	error = priv_check(td, PRIV_VFS_FHSTATFS);
4513 	if (error != 0)
4514 		return (error);
4515 	if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4516 		return (ESTALE);
4517 	error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4518 	if (error != 0) {
4519 		vfs_unbusy(mp);
4520 		return (error);
4521 	}
4522 	vput(vp);
4523 	error = prison_canseemount(td->td_ucred, mp);
4524 	if (error != 0)
4525 		goto out;
4526 #ifdef MAC
4527 	error = mac_mount_check_stat(td->td_ucred, mp);
4528 	if (error != 0)
4529 		goto out;
4530 #endif
4531 	/*
4532 	 * Set these in case the underlying filesystem fails to do so.
4533 	 */
4534 	sp = &mp->mnt_stat;
4535 	sp->f_version = STATFS_VERSION;
4536 	sp->f_namemax = NAME_MAX;
4537 	sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
4538 	error = VFS_STATFS(mp, sp);
4539 	if (error == 0)
4540 		*buf = *sp;
4541 out:
4542 	vfs_unbusy(mp);
4543 	return (error);
4544 }
4545 
4546 int
4547 kern_posix_fallocate(struct thread *td, int fd, off_t offset, off_t len)
4548 {
4549 	struct file *fp;
4550 	struct mount *mp;
4551 	struct vnode *vp;
4552 	cap_rights_t rights;
4553 	off_t olen, ooffset;
4554 	int error;
4555 
4556 	fp = NULL;
4557 	error = fget(td, fd, cap_rights_init(&rights, CAP_WRITE), &fp);
4558 	if (error != 0)
4559 		goto out;
4560 
4561 	switch (fp->f_type) {
4562 	case DTYPE_VNODE:
4563 		break;
4564 	case DTYPE_PIPE:
4565 	case DTYPE_FIFO:
4566 		error = ESPIPE;
4567 		goto out;
4568 	default:
4569 		error = ENODEV;
4570 		goto out;
4571 	}
4572 	if ((fp->f_flag & FWRITE) == 0) {
4573 		error = EBADF;
4574 		goto out;
4575 	}
4576 	vp = fp->f_vnode;
4577 	if (vp->v_type != VREG) {
4578 		error = ENODEV;
4579 		goto out;
4580 	}
4581 	if (offset < 0 || len <= 0) {
4582 		error = EINVAL;
4583 		goto out;
4584 	}
4585 	/* Check for wrap. */
4586 	if (offset > OFF_MAX - len) {
4587 		error = EFBIG;
4588 		goto out;
4589 	}
4590 
4591 	/* Allocating blocks may take a long time, so iterate. */
4592 	for (;;) {
4593 		olen = len;
4594 		ooffset = offset;
4595 
4596 		bwillwrite();
4597 		mp = NULL;
4598 		error = vn_start_write(vp, &mp, V_WAIT | PCATCH);
4599 		if (error != 0)
4600 			break;
4601 		error = vn_lock(vp, LK_EXCLUSIVE);
4602 		if (error != 0) {
4603 			vn_finished_write(mp);
4604 			break;
4605 		}
4606 #ifdef MAC
4607 		error = mac_vnode_check_write(td->td_ucred, fp->f_cred, vp);
4608 		if (error == 0)
4609 #endif
4610 			error = VOP_ALLOCATE(vp, &offset, &len);
4611 		VOP_UNLOCK(vp, 0);
4612 		vn_finished_write(mp);
4613 
4614 		if (olen + ooffset != offset + len) {
4615 			panic("offset + len changed from %jx/%jx to %jx/%jx",
4616 			    ooffset, olen, offset, len);
4617 		}
4618 		if (error != 0 || len == 0)
4619 			break;
4620 		KASSERT(olen > len, ("Iteration did not make progress?"));
4621 		maybe_yield();
4622 	}
4623  out:
4624 	if (fp != NULL)
4625 		fdrop(fp, td);
4626 	return (error);
4627 }
4628 
4629 int
4630 sys_posix_fallocate(struct thread *td, struct posix_fallocate_args *uap)
4631 {
4632 
4633 	td->td_retval[0] = kern_posix_fallocate(td, uap->fd, uap->offset,
4634 	    uap->len);
4635 	return (0);
4636 }
4637 
4638 /*
4639  * Unlike madvise(2), we do not make a best effort to remember every
4640  * possible caching hint.  Instead, we remember the last setting with
4641  * the exception that we will allow POSIX_FADV_NORMAL to adjust the
4642  * region of any current setting.
4643  */
4644 int
4645 kern_posix_fadvise(struct thread *td, int fd, off_t offset, off_t len,
4646     int advice)
4647 {
4648 	struct fadvise_info *fa, *new;
4649 	struct file *fp;
4650 	struct vnode *vp;
4651 	cap_rights_t rights;
4652 	off_t end;
4653 	int error;
4654 
4655 	if (offset < 0 || len < 0 || offset > OFF_MAX - len)
4656 		return (EINVAL);
4657 	switch (advice) {
4658 	case POSIX_FADV_SEQUENTIAL:
4659 	case POSIX_FADV_RANDOM:
4660 	case POSIX_FADV_NOREUSE:
4661 		new = malloc(sizeof(*fa), M_FADVISE, M_WAITOK);
4662 		break;
4663 	case POSIX_FADV_NORMAL:
4664 	case POSIX_FADV_WILLNEED:
4665 	case POSIX_FADV_DONTNEED:
4666 		new = NULL;
4667 		break;
4668 	default:
4669 		return (EINVAL);
4670 	}
4671 	/* XXX: CAP_POSIX_FADVISE? */
4672 	error = fget(td, fd, cap_rights_init(&rights), &fp);
4673 	if (error != 0)
4674 		goto out;
4675 
4676 	switch (fp->f_type) {
4677 	case DTYPE_VNODE:
4678 		break;
4679 	case DTYPE_PIPE:
4680 	case DTYPE_FIFO:
4681 		error = ESPIPE;
4682 		goto out;
4683 	default:
4684 		error = ENODEV;
4685 		goto out;
4686 	}
4687 	vp = fp->f_vnode;
4688 	if (vp->v_type != VREG) {
4689 		error = ENODEV;
4690 		goto out;
4691 	}
4692 	if (len == 0)
4693 		end = OFF_MAX;
4694 	else
4695 		end = offset + len - 1;
4696 	switch (advice) {
4697 	case POSIX_FADV_SEQUENTIAL:
4698 	case POSIX_FADV_RANDOM:
4699 	case POSIX_FADV_NOREUSE:
4700 		/*
4701 		 * Try to merge any existing non-standard region with
4702 		 * this new region if possible, otherwise create a new
4703 		 * non-standard region for this request.
4704 		 */
4705 		mtx_pool_lock(mtxpool_sleep, fp);
4706 		fa = fp->f_advice;
4707 		if (fa != NULL && fa->fa_advice == advice &&
4708 		    ((fa->fa_start <= end && fa->fa_end >= offset) ||
4709 		    (end != OFF_MAX && fa->fa_start == end + 1) ||
4710 		    (fa->fa_end != OFF_MAX && fa->fa_end + 1 == offset))) {
4711 			if (offset < fa->fa_start)
4712 				fa->fa_start = offset;
4713 			if (end > fa->fa_end)
4714 				fa->fa_end = end;
4715 		} else {
4716 			new->fa_advice = advice;
4717 			new->fa_start = offset;
4718 			new->fa_end = end;
4719 			new->fa_prevstart = 0;
4720 			new->fa_prevend = 0;
4721 			fp->f_advice = new;
4722 			new = fa;
4723 		}
4724 		mtx_pool_unlock(mtxpool_sleep, fp);
4725 		break;
4726 	case POSIX_FADV_NORMAL:
4727 		/*
4728 		 * If a the "normal" region overlaps with an existing
4729 		 * non-standard region, trim or remove the
4730 		 * non-standard region.
4731 		 */
4732 		mtx_pool_lock(mtxpool_sleep, fp);
4733 		fa = fp->f_advice;
4734 		if (fa != NULL) {
4735 			if (offset <= fa->fa_start && end >= fa->fa_end) {
4736 				new = fa;
4737 				fp->f_advice = NULL;
4738 			} else if (offset <= fa->fa_start &&
4739 			    end >= fa->fa_start)
4740 				fa->fa_start = end + 1;
4741 			else if (offset <= fa->fa_end && end >= fa->fa_end)
4742 				fa->fa_end = offset - 1;
4743 			else if (offset >= fa->fa_start && end <= fa->fa_end) {
4744 				/*
4745 				 * If the "normal" region is a middle
4746 				 * portion of the existing
4747 				 * non-standard region, just remove
4748 				 * the whole thing rather than picking
4749 				 * one side or the other to
4750 				 * preserve.
4751 				 */
4752 				new = fa;
4753 				fp->f_advice = NULL;
4754 			}
4755 		}
4756 		mtx_pool_unlock(mtxpool_sleep, fp);
4757 		break;
4758 	case POSIX_FADV_WILLNEED:
4759 	case POSIX_FADV_DONTNEED:
4760 		error = VOP_ADVISE(vp, offset, end, advice);
4761 		break;
4762 	}
4763 out:
4764 	if (fp != NULL)
4765 		fdrop(fp, td);
4766 	free(new, M_FADVISE);
4767 	return (error);
4768 }
4769 
4770 int
4771 sys_posix_fadvise(struct thread *td, struct posix_fadvise_args *uap)
4772 {
4773 
4774 	td->td_retval[0] = kern_posix_fadvise(td, uap->fd, uap->offset,
4775 	    uap->len, uap->advice);
4776 	return (0);
4777 }
4778