xref: /freebsd/sys/kern/vfs_syscalls.c (revision f5f7c05209ca2c3748fd8b27c5e80ffad49120eb)
1 /*-
2  * Copyright (c) 1989, 1993
3  *	The Regents of the University of California.  All rights reserved.
4  * (c) UNIX System Laboratories, Inc.
5  * All or some portions of this file are derived from material licensed
6  * to the University of California by American Telephone and Telegraph
7  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
8  * the permission of UNIX System Laboratories, Inc.
9  *
10  * Redistribution and use in source and binary forms, with or without
11  * modification, are permitted provided that the following conditions
12  * are met:
13  * 1. Redistributions of source code must retain the above copyright
14  *    notice, this list of conditions and the following disclaimer.
15  * 2. Redistributions in binary form must reproduce the above copyright
16  *    notice, this list of conditions and the following disclaimer in the
17  *    documentation and/or other materials provided with the distribution.
18  * 4. Neither the name of the University nor the names of its contributors
19  *    may be used to endorse or promote products derived from this software
20  *    without specific prior written permission.
21  *
22  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
23  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
24  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
25  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
26  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
27  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
28  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
29  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
30  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
31  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32  * SUCH DAMAGE.
33  *
34  *	@(#)vfs_syscalls.c	8.13 (Berkeley) 4/15/94
35  */
36 
37 #include <sys/cdefs.h>
38 __FBSDID("$FreeBSD$");
39 
40 #include "opt_capsicum.h"
41 #include "opt_compat.h"
42 #include "opt_kdtrace.h"
43 #include "opt_ktrace.h"
44 
45 #include <sys/param.h>
46 #include <sys/systm.h>
47 #include <sys/bio.h>
48 #include <sys/buf.h>
49 #include <sys/capability.h>
50 #include <sys/disk.h>
51 #include <sys/sysent.h>
52 #include <sys/malloc.h>
53 #include <sys/mount.h>
54 #include <sys/mutex.h>
55 #include <sys/sysproto.h>
56 #include <sys/namei.h>
57 #include <sys/filedesc.h>
58 #include <sys/kernel.h>
59 #include <sys/fcntl.h>
60 #include <sys/file.h>
61 #include <sys/filio.h>
62 #include <sys/limits.h>
63 #include <sys/linker.h>
64 #include <sys/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_DEFINE(vfs, , stat, mode, mode);
95 SDT_PROBE_ARGTYPE(vfs, , stat, mode, 0, "char *");
96 SDT_PROBE_ARGTYPE(vfs, , stat, mode, 1, "int");
97 SDT_PROBE_DEFINE(vfs, , stat, reg, reg);
98 SDT_PROBE_ARGTYPE(vfs, , stat, reg, 0, "char *");
99 SDT_PROBE_ARGTYPE(vfs, , stat, reg, 1, "int");
100 
101 static int chroot_refuse_vdir_fds(struct filedesc *fdp);
102 static int getutimes(const struct timeval *, enum uio_seg, struct timespec *);
103 static int setfflags(struct thread *td, struct vnode *, int);
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 	int error;
184 	struct nameidata nd;
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)
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 	int error;
294 	struct nameidata nd;
295 
296 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
297 	    pathseg, path, td);
298 	error = namei(&nd);
299 	if (error)
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)
308 		return (error);
309 #ifdef MAC
310 	error = mac_mount_check_stat(td->td_ucred, mp);
311 	if (error)
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)
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 	int error;
370 
371 	AUDIT_ARG_FD(fd);
372 	error = getvnode(td->td_proc->p_fd, fd, CAP_FSTATFS, &fp);
373 	if (error)
374 		return (error);
375 	vp = fp->f_vnode;
376 	vn_lock(vp, LK_SHARED | LK_RETRY);
377 #ifdef AUDIT
378 	AUDIT_ARG_VNODE1(vp);
379 #endif
380 	mp = vp->v_mount;
381 	if (mp)
382 		vfs_ref(mp);
383 	VOP_UNLOCK(vp, 0);
384 	fdrop(fp, td);
385 	if (mp == NULL) {
386 		error = EBADF;
387 		goto out;
388 	}
389 	error = vfs_busy(mp, 0);
390 	vfs_rel(mp);
391 	if (error)
392 		return (error);
393 #ifdef MAC
394 	error = mac_mount_check_stat(td->td_ucred, mp);
395 	if (error)
396 		goto out;
397 #endif
398 	/*
399 	 * Set these in case the underlying filesystem fails to do so.
400 	 */
401 	sp = &mp->mnt_stat;
402 	sp->f_version = STATFS_VERSION;
403 	sp->f_namemax = NAME_MAX;
404 	sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
405 	error = VFS_STATFS(mp, sp);
406 	if (error)
407 		goto out;
408 	if (priv_check(td, PRIV_VFS_GENERATION)) {
409 		bcopy(sp, &sb, sizeof(sb));
410 		sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
411 		prison_enforce_statfs(td->td_ucred, mp, &sb);
412 		sp = &sb;
413 	}
414 	*buf = *sp;
415 out:
416 	if (mp)
417 		vfs_unbusy(mp);
418 	return (error);
419 }
420 
421 /*
422  * Get statistics on all filesystems.
423  */
424 #ifndef _SYS_SYSPROTO_H_
425 struct getfsstat_args {
426 	struct statfs *buf;
427 	long bufsize;
428 	int flags;
429 };
430 #endif
431 int
432 sys_getfsstat(td, uap)
433 	struct thread *td;
434 	register struct getfsstat_args /* {
435 		struct statfs *buf;
436 		long bufsize;
437 		int flags;
438 	} */ *uap;
439 {
440 
441 	return (kern_getfsstat(td, &uap->buf, uap->bufsize, UIO_USERSPACE,
442 	    uap->flags));
443 }
444 
445 /*
446  * If (bufsize > 0 && bufseg == UIO_SYSSPACE)
447  * 	The caller is responsible for freeing memory which will be allocated
448  *	in '*buf'.
449  */
450 int
451 kern_getfsstat(struct thread *td, struct statfs **buf, size_t bufsize,
452     enum uio_seg bufseg, int flags)
453 {
454 	struct mount *mp, *nmp;
455 	struct statfs *sfsp, *sp, sb;
456 	size_t count, maxcount;
457 	int error;
458 
459 	maxcount = bufsize / sizeof(struct statfs);
460 	if (bufsize == 0)
461 		sfsp = NULL;
462 	else if (bufseg == UIO_USERSPACE)
463 		sfsp = *buf;
464 	else /* if (bufseg == UIO_SYSSPACE) */ {
465 		count = 0;
466 		mtx_lock(&mountlist_mtx);
467 		TAILQ_FOREACH(mp, &mountlist, mnt_list) {
468 			count++;
469 		}
470 		mtx_unlock(&mountlist_mtx);
471 		if (maxcount > count)
472 			maxcount = count;
473 		sfsp = *buf = malloc(maxcount * sizeof(struct statfs), M_TEMP,
474 		    M_WAITOK);
475 	}
476 	count = 0;
477 	mtx_lock(&mountlist_mtx);
478 	for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
479 		if (prison_canseemount(td->td_ucred, mp) != 0) {
480 			nmp = TAILQ_NEXT(mp, mnt_list);
481 			continue;
482 		}
483 #ifdef MAC
484 		if (mac_mount_check_stat(td->td_ucred, mp) != 0) {
485 			nmp = TAILQ_NEXT(mp, mnt_list);
486 			continue;
487 		}
488 #endif
489 		if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK)) {
490 			nmp = TAILQ_NEXT(mp, mnt_list);
491 			continue;
492 		}
493 		if (sfsp && count < maxcount) {
494 			sp = &mp->mnt_stat;
495 			/*
496 			 * Set these in case the underlying filesystem
497 			 * fails to do so.
498 			 */
499 			sp->f_version = STATFS_VERSION;
500 			sp->f_namemax = NAME_MAX;
501 			sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
502 			/*
503 			 * If MNT_NOWAIT or MNT_LAZY is specified, do not
504 			 * refresh the fsstat cache. MNT_NOWAIT or MNT_LAZY
505 			 * overrides MNT_WAIT.
506 			 */
507 			if (((flags & (MNT_LAZY|MNT_NOWAIT)) == 0 ||
508 			    (flags & MNT_WAIT)) &&
509 			    (error = VFS_STATFS(mp, sp))) {
510 				mtx_lock(&mountlist_mtx);
511 				nmp = TAILQ_NEXT(mp, mnt_list);
512 				vfs_unbusy(mp);
513 				continue;
514 			}
515 			if (priv_check(td, PRIV_VFS_GENERATION)) {
516 				bcopy(sp, &sb, sizeof(sb));
517 				sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
518 				prison_enforce_statfs(td->td_ucred, mp, &sb);
519 				sp = &sb;
520 			}
521 			if (bufseg == UIO_SYSSPACE)
522 				bcopy(sp, sfsp, sizeof(*sp));
523 			else /* if (bufseg == UIO_USERSPACE) */ {
524 				error = copyout(sp, sfsp, sizeof(*sp));
525 				if (error) {
526 					vfs_unbusy(mp);
527 					return (error);
528 				}
529 			}
530 			sfsp++;
531 		}
532 		count++;
533 		mtx_lock(&mountlist_mtx);
534 		nmp = TAILQ_NEXT(mp, mnt_list);
535 		vfs_unbusy(mp);
536 	}
537 	mtx_unlock(&mountlist_mtx);
538 	if (sfsp && count > maxcount)
539 		td->td_retval[0] = maxcount;
540 	else
541 		td->td_retval[0] = count;
542 	return (0);
543 }
544 
545 #ifdef COMPAT_FREEBSD4
546 /*
547  * Get old format filesystem statistics.
548  */
549 static void cvtstatfs(struct statfs *, struct ostatfs *);
550 
551 #ifndef _SYS_SYSPROTO_H_
552 struct freebsd4_statfs_args {
553 	char *path;
554 	struct ostatfs *buf;
555 };
556 #endif
557 int
558 freebsd4_statfs(td, uap)
559 	struct thread *td;
560 	struct freebsd4_statfs_args /* {
561 		char *path;
562 		struct ostatfs *buf;
563 	} */ *uap;
564 {
565 	struct ostatfs osb;
566 	struct statfs sf;
567 	int error;
568 
569 	error = kern_statfs(td, uap->path, UIO_USERSPACE, &sf);
570 	if (error)
571 		return (error);
572 	cvtstatfs(&sf, &osb);
573 	return (copyout(&osb, uap->buf, sizeof(osb)));
574 }
575 
576 /*
577  * Get filesystem statistics.
578  */
579 #ifndef _SYS_SYSPROTO_H_
580 struct freebsd4_fstatfs_args {
581 	int fd;
582 	struct ostatfs *buf;
583 };
584 #endif
585 int
586 freebsd4_fstatfs(td, uap)
587 	struct thread *td;
588 	struct freebsd4_fstatfs_args /* {
589 		int fd;
590 		struct ostatfs *buf;
591 	} */ *uap;
592 {
593 	struct ostatfs osb;
594 	struct statfs sf;
595 	int error;
596 
597 	error = kern_fstatfs(td, uap->fd, &sf);
598 	if (error)
599 		return (error);
600 	cvtstatfs(&sf, &osb);
601 	return (copyout(&osb, uap->buf, sizeof(osb)));
602 }
603 
604 /*
605  * Get statistics on all filesystems.
606  */
607 #ifndef _SYS_SYSPROTO_H_
608 struct freebsd4_getfsstat_args {
609 	struct ostatfs *buf;
610 	long bufsize;
611 	int flags;
612 };
613 #endif
614 int
615 freebsd4_getfsstat(td, uap)
616 	struct thread *td;
617 	register struct freebsd4_getfsstat_args /* {
618 		struct ostatfs *buf;
619 		long bufsize;
620 		int flags;
621 	} */ *uap;
622 {
623 	struct statfs *buf, *sp;
624 	struct ostatfs osb;
625 	size_t count, size;
626 	int error;
627 
628 	count = uap->bufsize / sizeof(struct ostatfs);
629 	size = count * sizeof(struct statfs);
630 	error = kern_getfsstat(td, &buf, size, UIO_SYSSPACE, uap->flags);
631 	if (size > 0) {
632 		count = td->td_retval[0];
633 		sp = buf;
634 		while (count > 0 && error == 0) {
635 			cvtstatfs(sp, &osb);
636 			error = copyout(&osb, uap->buf, sizeof(osb));
637 			sp++;
638 			uap->buf++;
639 			count--;
640 		}
641 		free(buf, M_TEMP);
642 	}
643 	return (error);
644 }
645 
646 /*
647  * Implement fstatfs() for (NFS) file handles.
648  */
649 #ifndef _SYS_SYSPROTO_H_
650 struct freebsd4_fhstatfs_args {
651 	struct fhandle *u_fhp;
652 	struct ostatfs *buf;
653 };
654 #endif
655 int
656 freebsd4_fhstatfs(td, uap)
657 	struct thread *td;
658 	struct freebsd4_fhstatfs_args /* {
659 		struct fhandle *u_fhp;
660 		struct ostatfs *buf;
661 	} */ *uap;
662 {
663 	struct ostatfs osb;
664 	struct statfs sf;
665 	fhandle_t fh;
666 	int error;
667 
668 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
669 	if (error)
670 		return (error);
671 	error = kern_fhstatfs(td, fh, &sf);
672 	if (error)
673 		return (error);
674 	cvtstatfs(&sf, &osb);
675 	return (copyout(&osb, uap->buf, sizeof(osb)));
676 }
677 
678 /*
679  * Convert a new format statfs structure to an old format statfs structure.
680  */
681 static void
682 cvtstatfs(nsp, osp)
683 	struct statfs *nsp;
684 	struct ostatfs *osp;
685 {
686 
687 	statfs_scale_blocks(nsp, LONG_MAX);
688 	bzero(osp, sizeof(*osp));
689 	osp->f_bsize = nsp->f_bsize;
690 	osp->f_iosize = MIN(nsp->f_iosize, LONG_MAX);
691 	osp->f_blocks = nsp->f_blocks;
692 	osp->f_bfree = nsp->f_bfree;
693 	osp->f_bavail = nsp->f_bavail;
694 	osp->f_files = MIN(nsp->f_files, LONG_MAX);
695 	osp->f_ffree = MIN(nsp->f_ffree, LONG_MAX);
696 	osp->f_owner = nsp->f_owner;
697 	osp->f_type = nsp->f_type;
698 	osp->f_flags = nsp->f_flags;
699 	osp->f_syncwrites = MIN(nsp->f_syncwrites, LONG_MAX);
700 	osp->f_asyncwrites = MIN(nsp->f_asyncwrites, LONG_MAX);
701 	osp->f_syncreads = MIN(nsp->f_syncreads, LONG_MAX);
702 	osp->f_asyncreads = MIN(nsp->f_asyncreads, LONG_MAX);
703 	strlcpy(osp->f_fstypename, nsp->f_fstypename,
704 	    MIN(MFSNAMELEN, OMFSNAMELEN));
705 	strlcpy(osp->f_mntonname, nsp->f_mntonname,
706 	    MIN(MNAMELEN, OMNAMELEN));
707 	strlcpy(osp->f_mntfromname, nsp->f_mntfromname,
708 	    MIN(MNAMELEN, OMNAMELEN));
709 	osp->f_fsid = nsp->f_fsid;
710 }
711 #endif /* COMPAT_FREEBSD4 */
712 
713 /*
714  * Change current working directory to a given file descriptor.
715  */
716 #ifndef _SYS_SYSPROTO_H_
717 struct fchdir_args {
718 	int	fd;
719 };
720 #endif
721 int
722 sys_fchdir(td, uap)
723 	struct thread *td;
724 	struct fchdir_args /* {
725 		int fd;
726 	} */ *uap;
727 {
728 	register struct filedesc *fdp = td->td_proc->p_fd;
729 	struct vnode *vp, *tdp, *vpold;
730 	struct mount *mp;
731 	struct file *fp;
732 	int error;
733 
734 	AUDIT_ARG_FD(uap->fd);
735 	if ((error = getvnode(fdp, uap->fd, CAP_FCHDIR, &fp)) != 0)
736 		return (error);
737 	vp = fp->f_vnode;
738 	VREF(vp);
739 	fdrop(fp, td);
740 	vn_lock(vp, LK_SHARED | LK_RETRY);
741 	AUDIT_ARG_VNODE1(vp);
742 	error = change_dir(vp, td);
743 	while (!error && (mp = vp->v_mountedhere) != NULL) {
744 		if (vfs_busy(mp, 0))
745 			continue;
746 		error = VFS_ROOT(mp, LK_SHARED, &tdp);
747 		vfs_unbusy(mp);
748 		if (error)
749 			break;
750 		vput(vp);
751 		vp = tdp;
752 	}
753 	if (error) {
754 		vput(vp);
755 		return (error);
756 	}
757 	VOP_UNLOCK(vp, 0);
758 	FILEDESC_XLOCK(fdp);
759 	vpold = fdp->fd_cdir;
760 	fdp->fd_cdir = vp;
761 	FILEDESC_XUNLOCK(fdp);
762 	vrele(vpold);
763 	return (0);
764 }
765 
766 /*
767  * Change current working directory (``.'').
768  */
769 #ifndef _SYS_SYSPROTO_H_
770 struct chdir_args {
771 	char	*path;
772 };
773 #endif
774 int
775 sys_chdir(td, uap)
776 	struct thread *td;
777 	struct chdir_args /* {
778 		char *path;
779 	} */ *uap;
780 {
781 
782 	return (kern_chdir(td, uap->path, UIO_USERSPACE));
783 }
784 
785 int
786 kern_chdir(struct thread *td, char *path, enum uio_seg pathseg)
787 {
788 	register struct filedesc *fdp = td->td_proc->p_fd;
789 	int error;
790 	struct nameidata nd;
791 	struct vnode *vp;
792 
793 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
794 	    pathseg, path, td);
795 	if ((error = namei(&nd)) != 0)
796 		return (error);
797 	if ((error = change_dir(nd.ni_vp, td)) != 0) {
798 		vput(nd.ni_vp);
799 		NDFREE(&nd, NDF_ONLY_PNBUF);
800 		return (error);
801 	}
802 	VOP_UNLOCK(nd.ni_vp, 0);
803 	NDFREE(&nd, NDF_ONLY_PNBUF);
804 	FILEDESC_XLOCK(fdp);
805 	vp = fdp->fd_cdir;
806 	fdp->fd_cdir = nd.ni_vp;
807 	FILEDESC_XUNLOCK(fdp);
808 	vrele(vp);
809 	return (0);
810 }
811 
812 /*
813  * Helper function for raised chroot(2) security function:  Refuse if
814  * any filedescriptors are open directories.
815  */
816 static int
817 chroot_refuse_vdir_fds(fdp)
818 	struct filedesc *fdp;
819 {
820 	struct vnode *vp;
821 	struct file *fp;
822 	int fd;
823 
824 	FILEDESC_LOCK_ASSERT(fdp);
825 
826 	for (fd = 0; fd < fdp->fd_nfiles ; fd++) {
827 		fp = fget_locked(fdp, fd);
828 		if (fp == NULL)
829 			continue;
830 		if (fp->f_type == DTYPE_VNODE) {
831 			vp = fp->f_vnode;
832 			if (vp->v_type == VDIR)
833 				return (EPERM);
834 		}
835 	}
836 	return (0);
837 }
838 
839 /*
840  * This sysctl determines if we will allow a process to chroot(2) if it
841  * has a directory open:
842  *	0: disallowed for all processes.
843  *	1: allowed for processes that were not already chroot(2)'ed.
844  *	2: allowed for all processes.
845  */
846 
847 static int chroot_allow_open_directories = 1;
848 
849 SYSCTL_INT(_kern, OID_AUTO, chroot_allow_open_directories, CTLFLAG_RW,
850      &chroot_allow_open_directories, 0,
851      "Allow a process to chroot(2) if it has a directory open");
852 
853 /*
854  * Change notion of root (``/'') directory.
855  */
856 #ifndef _SYS_SYSPROTO_H_
857 struct chroot_args {
858 	char	*path;
859 };
860 #endif
861 int
862 sys_chroot(td, uap)
863 	struct thread *td;
864 	struct chroot_args /* {
865 		char *path;
866 	} */ *uap;
867 {
868 	int error;
869 	struct nameidata nd;
870 
871 	error = priv_check(td, PRIV_VFS_CHROOT);
872 	if (error)
873 		return (error);
874 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
875 	    UIO_USERSPACE, uap->path, td);
876 	error = namei(&nd);
877 	if (error)
878 		goto error;
879 	if ((error = change_dir(nd.ni_vp, td)) != 0)
880 		goto e_vunlock;
881 #ifdef MAC
882 	if ((error = mac_vnode_check_chroot(td->td_ucred, nd.ni_vp)))
883 		goto e_vunlock;
884 #endif
885 	VOP_UNLOCK(nd.ni_vp, 0);
886 	error = change_root(nd.ni_vp, td);
887 	vrele(nd.ni_vp);
888 	NDFREE(&nd, NDF_ONLY_PNBUF);
889 	return (error);
890 e_vunlock:
891 	vput(nd.ni_vp);
892 error:
893 	NDFREE(&nd, NDF_ONLY_PNBUF);
894 	return (error);
895 }
896 
897 /*
898  * Common routine for chroot and chdir.  Callers must provide a locked vnode
899  * instance.
900  */
901 int
902 change_dir(vp, td)
903 	struct vnode *vp;
904 	struct thread *td;
905 {
906 	int error;
907 
908 	ASSERT_VOP_LOCKED(vp, "change_dir(): vp not locked");
909 	if (vp->v_type != VDIR)
910 		return (ENOTDIR);
911 #ifdef MAC
912 	error = mac_vnode_check_chdir(td->td_ucred, vp);
913 	if (error)
914 		return (error);
915 #endif
916 	error = VOP_ACCESS(vp, VEXEC, td->td_ucred, td);
917 	return (error);
918 }
919 
920 /*
921  * Common routine for kern_chroot() and jail_attach().  The caller is
922  * responsible for invoking priv_check() and mac_vnode_check_chroot() to
923  * authorize this operation.
924  */
925 int
926 change_root(vp, td)
927 	struct vnode *vp;
928 	struct thread *td;
929 {
930 	struct filedesc *fdp;
931 	struct vnode *oldvp;
932 	int error;
933 
934 	fdp = td->td_proc->p_fd;
935 	FILEDESC_XLOCK(fdp);
936 	if (chroot_allow_open_directories == 0 ||
937 	    (chroot_allow_open_directories == 1 && fdp->fd_rdir != rootvnode)) {
938 		error = chroot_refuse_vdir_fds(fdp);
939 		if (error) {
940 			FILEDESC_XUNLOCK(fdp);
941 			return (error);
942 		}
943 	}
944 	oldvp = fdp->fd_rdir;
945 	fdp->fd_rdir = vp;
946 	VREF(fdp->fd_rdir);
947 	if (!fdp->fd_jdir) {
948 		fdp->fd_jdir = vp;
949 		VREF(fdp->fd_jdir);
950 	}
951 	FILEDESC_XUNLOCK(fdp);
952 	vrele(oldvp);
953 	return (0);
954 }
955 
956 static __inline cap_rights_t
957 flags_to_rights(int flags)
958 {
959 	cap_rights_t rights = 0;
960 
961 	if (flags & O_EXEC) {
962 		rights |= CAP_FEXECVE;
963 	} else {
964 		switch ((flags & O_ACCMODE)) {
965 		case O_RDONLY:
966 			rights |= CAP_READ;
967 			break;
968 		case O_RDWR:
969 			rights |= CAP_READ;
970 			/* FALLTHROUGH */
971 		case O_WRONLY:
972 			rights |= CAP_WRITE;
973 			break;
974 		}
975 	}
976 
977 	if (flags & O_CREAT)
978 		rights |= CAP_CREATE;
979 
980 	if (flags & O_TRUNC)
981 		rights |= CAP_FTRUNCATE;
982 
983 	if ((flags & O_EXLOCK) || (flags & O_SHLOCK))
984 		rights |= CAP_FLOCK;
985 
986 	return (rights);
987 }
988 
989 /*
990  * Check permissions, allocate an open file structure, and call the device
991  * open routine if any.
992  */
993 #ifndef _SYS_SYSPROTO_H_
994 struct open_args {
995 	char	*path;
996 	int	flags;
997 	int	mode;
998 };
999 #endif
1000 int
1001 sys_open(td, uap)
1002 	struct thread *td;
1003 	register struct open_args /* {
1004 		char *path;
1005 		int flags;
1006 		int mode;
1007 	} */ *uap;
1008 {
1009 
1010 	return (kern_open(td, uap->path, UIO_USERSPACE, uap->flags, uap->mode));
1011 }
1012 
1013 #ifndef _SYS_SYSPROTO_H_
1014 struct openat_args {
1015 	int	fd;
1016 	char	*path;
1017 	int	flag;
1018 	int	mode;
1019 };
1020 #endif
1021 int
1022 sys_openat(struct thread *td, struct openat_args *uap)
1023 {
1024 
1025 	return (kern_openat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
1026 	    uap->mode));
1027 }
1028 
1029 int
1030 kern_open(struct thread *td, char *path, enum uio_seg pathseg, int flags,
1031     int mode)
1032 {
1033 
1034 	return (kern_openat(td, AT_FDCWD, path, pathseg, flags, mode));
1035 }
1036 
1037 int
1038 kern_openat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1039     int flags, int mode)
1040 {
1041 	struct proc *p = td->td_proc;
1042 	struct filedesc *fdp = p->p_fd;
1043 	struct file *fp;
1044 	struct vnode *vp;
1045 	int cmode;
1046 	int indx = -1, error;
1047 	struct nameidata nd;
1048 	cap_rights_t rights_needed = CAP_LOOKUP;
1049 
1050 	AUDIT_ARG_FFLAGS(flags);
1051 	AUDIT_ARG_MODE(mode);
1052 	/* XXX: audit dirfd */
1053 	rights_needed |= flags_to_rights(flags);
1054 	/*
1055 	 * Only one of the O_EXEC, O_RDONLY, O_WRONLY and O_RDWR flags
1056 	 * may be specified.
1057 	 */
1058 	if (flags & O_EXEC) {
1059 		if (flags & O_ACCMODE)
1060 			return (EINVAL);
1061 	} else if ((flags & O_ACCMODE) == O_ACCMODE) {
1062 		return (EINVAL);
1063 	} else {
1064 		flags = FFLAGS(flags);
1065 	}
1066 
1067 	/*
1068 	 * Allocate the file descriptor, but don't install a descriptor yet.
1069 	 */
1070 	error = falloc_noinstall(td, &fp);
1071 	if (error)
1072 		return (error);
1073 	/*
1074 	 * An extra reference on `fp' has been held for us by
1075 	 * falloc_noinstall().
1076 	 */
1077 	/* Set the flags early so the finit in devfs can pick them up. */
1078 	fp->f_flag = flags & FMASK;
1079 	cmode = ((mode &~ fdp->fd_cmask) & ALLPERMS) &~ S_ISTXT;
1080 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, fd,
1081 	    rights_needed, td);
1082 	td->td_dupfd = -1;		/* XXX check for fdopen */
1083 	error = vn_open(&nd, &flags, cmode, fp);
1084 	if (error) {
1085 		/*
1086 		 * If the vn_open replaced the method vector, something
1087 		 * wonderous happened deep below and we just pass it up
1088 		 * pretending we know what we do.
1089 		 */
1090 		if (error == ENXIO && fp->f_ops != &badfileops)
1091 			goto success;
1092 
1093 		/*
1094 		 * Handle special fdopen() case. bleh.
1095 		 *
1096 		 * Don't do this for relative (capability) lookups; we don't
1097 		 * understand exactly what would happen, and we don't think
1098 		 * that it ever should.
1099 		 */
1100 		if (nd.ni_strictrelative == 0 &&
1101 		    (error == ENODEV || error == ENXIO) &&
1102 		    td->td_dupfd >= 0) {
1103 			error = dupfdopen(td, fdp, td->td_dupfd, flags, error,
1104 			    &indx);
1105 			if (error == 0)
1106 				goto success;
1107 		}
1108 
1109 		goto bad;
1110 	}
1111 	td->td_dupfd = 0;
1112 	NDFREE(&nd, NDF_ONLY_PNBUF);
1113 	vp = nd.ni_vp;
1114 
1115 	/*
1116 	 * Store the vnode, for any f_type. Typically, the vnode use
1117 	 * count is decremented by direct call to vn_closefile() for
1118 	 * files that switched type in the cdevsw fdopen() method.
1119 	 */
1120 	fp->f_vnode = vp;
1121 	/*
1122 	 * If the file wasn't claimed by devfs bind it to the normal
1123 	 * vnode operations here.
1124 	 */
1125 	if (fp->f_ops == &badfileops) {
1126 		KASSERT(vp->v_type != VFIFO, ("Unexpected fifo."));
1127 		fp->f_seqcount = 1;
1128 		finit(fp, (flags & FMASK) | (fp->f_flag & FHASLOCK), DTYPE_VNODE,
1129 		    vp, &vnops);
1130 	}
1131 
1132 	VOP_UNLOCK(vp, 0);
1133 	if (flags & O_TRUNC) {
1134 		error = fo_truncate(fp, 0, td->td_ucred, td);
1135 		if (error)
1136 			goto bad;
1137 	}
1138 success:
1139 	/*
1140 	 * If we haven't already installed the FD (for dupfdopen), do so now.
1141 	 */
1142 	if (indx == -1) {
1143 #ifdef CAPABILITIES
1144 		if (nd.ni_strictrelative == 1) {
1145 			/*
1146 			 * We are doing a strict relative lookup; wrap the
1147 			 * result in a capability.
1148 			 */
1149 			if ((error = kern_capwrap(td, fp, nd.ni_baserights,
1150 			    &indx)) != 0)
1151 				goto bad;
1152 		} else
1153 #endif
1154 			if ((error = finstall(td, fp, &indx, flags)) != 0)
1155 				goto bad;
1156 	}
1157 
1158 	/*
1159 	 * Release our private reference, leaving the one associated with
1160 	 * the descriptor table intact.
1161 	 */
1162 	fdrop(fp, td);
1163 	td->td_retval[0] = indx;
1164 	return (0);
1165 bad:
1166 	KASSERT(indx == -1, ("indx=%d, should be -1", indx));
1167 	fdrop(fp, td);
1168 	return (error);
1169 }
1170 
1171 #ifdef COMPAT_43
1172 /*
1173  * Create a file.
1174  */
1175 #ifndef _SYS_SYSPROTO_H_
1176 struct ocreat_args {
1177 	char	*path;
1178 	int	mode;
1179 };
1180 #endif
1181 int
1182 ocreat(td, uap)
1183 	struct thread *td;
1184 	register struct ocreat_args /* {
1185 		char *path;
1186 		int mode;
1187 	} */ *uap;
1188 {
1189 
1190 	return (kern_open(td, uap->path, UIO_USERSPACE,
1191 	    O_WRONLY | O_CREAT | O_TRUNC, uap->mode));
1192 }
1193 #endif /* COMPAT_43 */
1194 
1195 /*
1196  * Create a special file.
1197  */
1198 #ifndef _SYS_SYSPROTO_H_
1199 struct mknod_args {
1200 	char	*path;
1201 	int	mode;
1202 	int	dev;
1203 };
1204 #endif
1205 int
1206 sys_mknod(td, uap)
1207 	struct thread *td;
1208 	register struct mknod_args /* {
1209 		char *path;
1210 		int mode;
1211 		int dev;
1212 	} */ *uap;
1213 {
1214 
1215 	return (kern_mknod(td, uap->path, UIO_USERSPACE, uap->mode, uap->dev));
1216 }
1217 
1218 #ifndef _SYS_SYSPROTO_H_
1219 struct mknodat_args {
1220 	int	fd;
1221 	char	*path;
1222 	mode_t	mode;
1223 	dev_t	dev;
1224 };
1225 #endif
1226 int
1227 sys_mknodat(struct thread *td, struct mknodat_args *uap)
1228 {
1229 
1230 	return (kern_mknodat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode,
1231 	    uap->dev));
1232 }
1233 
1234 int
1235 kern_mknod(struct thread *td, char *path, enum uio_seg pathseg, int mode,
1236     int dev)
1237 {
1238 
1239 	return (kern_mknodat(td, AT_FDCWD, path, pathseg, mode, dev));
1240 }
1241 
1242 int
1243 kern_mknodat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1244     int mode, int dev)
1245 {
1246 	struct vnode *vp;
1247 	struct mount *mp;
1248 	struct vattr vattr;
1249 	int error;
1250 	int whiteout = 0;
1251 	struct nameidata nd;
1252 
1253 	AUDIT_ARG_MODE(mode);
1254 	AUDIT_ARG_DEV(dev);
1255 	switch (mode & S_IFMT) {
1256 	case S_IFCHR:
1257 	case S_IFBLK:
1258 		error = priv_check(td, PRIV_VFS_MKNOD_DEV);
1259 		break;
1260 	case S_IFMT:
1261 		error = priv_check(td, PRIV_VFS_MKNOD_BAD);
1262 		break;
1263 	case S_IFWHT:
1264 		error = priv_check(td, PRIV_VFS_MKNOD_WHT);
1265 		break;
1266 	case S_IFIFO:
1267 		if (dev == 0)
1268 			return (kern_mkfifoat(td, fd, path, pathseg, mode));
1269 		/* FALLTHROUGH */
1270 	default:
1271 		error = EINVAL;
1272 		break;
1273 	}
1274 	if (error)
1275 		return (error);
1276 restart:
1277 	bwillwrite();
1278 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1,
1279 	    pathseg, path, fd, CAP_MKNOD, td);
1280 	if ((error = namei(&nd)) != 0)
1281 		return (error);
1282 	vp = nd.ni_vp;
1283 	if (vp != NULL) {
1284 		NDFREE(&nd, NDF_ONLY_PNBUF);
1285 		if (vp == nd.ni_dvp)
1286 			vrele(nd.ni_dvp);
1287 		else
1288 			vput(nd.ni_dvp);
1289 		vrele(vp);
1290 		return (EEXIST);
1291 	} else {
1292 		VATTR_NULL(&vattr);
1293 		vattr.va_mode = (mode & ALLPERMS) &
1294 		    ~td->td_proc->p_fd->fd_cmask;
1295 		vattr.va_rdev = dev;
1296 		whiteout = 0;
1297 
1298 		switch (mode & S_IFMT) {
1299 		case S_IFMT:	/* used by badsect to flag bad sectors */
1300 			vattr.va_type = VBAD;
1301 			break;
1302 		case S_IFCHR:
1303 			vattr.va_type = VCHR;
1304 			break;
1305 		case S_IFBLK:
1306 			vattr.va_type = VBLK;
1307 			break;
1308 		case S_IFWHT:
1309 			whiteout = 1;
1310 			break;
1311 		default:
1312 			panic("kern_mknod: invalid mode");
1313 		}
1314 	}
1315 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1316 		NDFREE(&nd, NDF_ONLY_PNBUF);
1317 		vput(nd.ni_dvp);
1318 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1319 			return (error);
1320 		goto restart;
1321 	}
1322 #ifdef MAC
1323 	if (error == 0 && !whiteout)
1324 		error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp,
1325 		    &nd.ni_cnd, &vattr);
1326 #endif
1327 	if (!error) {
1328 		if (whiteout)
1329 			error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE);
1330 		else {
1331 			error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp,
1332 						&nd.ni_cnd, &vattr);
1333 			if (error == 0)
1334 				vput(nd.ni_vp);
1335 		}
1336 	}
1337 	NDFREE(&nd, NDF_ONLY_PNBUF);
1338 	vput(nd.ni_dvp);
1339 	vn_finished_write(mp);
1340 	return (error);
1341 }
1342 
1343 /*
1344  * Create a named pipe.
1345  */
1346 #ifndef _SYS_SYSPROTO_H_
1347 struct mkfifo_args {
1348 	char	*path;
1349 	int	mode;
1350 };
1351 #endif
1352 int
1353 sys_mkfifo(td, uap)
1354 	struct thread *td;
1355 	register struct mkfifo_args /* {
1356 		char *path;
1357 		int mode;
1358 	} */ *uap;
1359 {
1360 
1361 	return (kern_mkfifo(td, uap->path, UIO_USERSPACE, uap->mode));
1362 }
1363 
1364 #ifndef _SYS_SYSPROTO_H_
1365 struct mkfifoat_args {
1366 	int	fd;
1367 	char	*path;
1368 	mode_t	mode;
1369 };
1370 #endif
1371 int
1372 sys_mkfifoat(struct thread *td, struct mkfifoat_args *uap)
1373 {
1374 
1375 	return (kern_mkfifoat(td, uap->fd, uap->path, UIO_USERSPACE,
1376 	    uap->mode));
1377 }
1378 
1379 int
1380 kern_mkfifo(struct thread *td, char *path, enum uio_seg pathseg, int mode)
1381 {
1382 
1383 	return (kern_mkfifoat(td, AT_FDCWD, path, pathseg, mode));
1384 }
1385 
1386 int
1387 kern_mkfifoat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1388     int mode)
1389 {
1390 	struct mount *mp;
1391 	struct vattr vattr;
1392 	int error;
1393 	struct nameidata nd;
1394 
1395 	AUDIT_ARG_MODE(mode);
1396 restart:
1397 	bwillwrite();
1398 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1,
1399 	    pathseg, path, fd, CAP_MKFIFO, td);
1400 	if ((error = namei(&nd)) != 0)
1401 		return (error);
1402 	if (nd.ni_vp != NULL) {
1403 		NDFREE(&nd, NDF_ONLY_PNBUF);
1404 		if (nd.ni_vp == nd.ni_dvp)
1405 			vrele(nd.ni_dvp);
1406 		else
1407 			vput(nd.ni_dvp);
1408 		vrele(nd.ni_vp);
1409 		return (EEXIST);
1410 	}
1411 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1412 		NDFREE(&nd, NDF_ONLY_PNBUF);
1413 		vput(nd.ni_dvp);
1414 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1415 			return (error);
1416 		goto restart;
1417 	}
1418 	VATTR_NULL(&vattr);
1419 	vattr.va_type = VFIFO;
1420 	vattr.va_mode = (mode & ALLPERMS) & ~td->td_proc->p_fd->fd_cmask;
1421 #ifdef MAC
1422 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1423 	    &vattr);
1424 	if (error)
1425 		goto out;
1426 #endif
1427 	error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
1428 	if (error == 0)
1429 		vput(nd.ni_vp);
1430 #ifdef MAC
1431 out:
1432 #endif
1433 	vput(nd.ni_dvp);
1434 	vn_finished_write(mp);
1435 	NDFREE(&nd, NDF_ONLY_PNBUF);
1436 	return (error);
1437 }
1438 
1439 /*
1440  * Make a hard file link.
1441  */
1442 #ifndef _SYS_SYSPROTO_H_
1443 struct link_args {
1444 	char	*path;
1445 	char	*link;
1446 };
1447 #endif
1448 int
1449 sys_link(td, uap)
1450 	struct thread *td;
1451 	register struct link_args /* {
1452 		char *path;
1453 		char *link;
1454 	} */ *uap;
1455 {
1456 
1457 	return (kern_link(td, uap->path, uap->link, UIO_USERSPACE));
1458 }
1459 
1460 #ifndef _SYS_SYSPROTO_H_
1461 struct linkat_args {
1462 	int	fd1;
1463 	char	*path1;
1464 	int	fd2;
1465 	char	*path2;
1466 	int	flag;
1467 };
1468 #endif
1469 int
1470 sys_linkat(struct thread *td, struct linkat_args *uap)
1471 {
1472 	int flag;
1473 
1474 	flag = uap->flag;
1475 	if (flag & ~AT_SYMLINK_FOLLOW)
1476 		return (EINVAL);
1477 
1478 	return (kern_linkat(td, uap->fd1, uap->fd2, uap->path1, uap->path2,
1479 	    UIO_USERSPACE, (flag & AT_SYMLINK_FOLLOW) ? FOLLOW : NOFOLLOW));
1480 }
1481 
1482 int hardlink_check_uid = 0;
1483 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_uid, CTLFLAG_RW,
1484     &hardlink_check_uid, 0,
1485     "Unprivileged processes cannot create hard links to files owned by other "
1486     "users");
1487 static int hardlink_check_gid = 0;
1488 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_gid, CTLFLAG_RW,
1489     &hardlink_check_gid, 0,
1490     "Unprivileged processes cannot create hard links to files owned by other "
1491     "groups");
1492 
1493 static int
1494 can_hardlink(struct vnode *vp, struct ucred *cred)
1495 {
1496 	struct vattr va;
1497 	int error;
1498 
1499 	if (!hardlink_check_uid && !hardlink_check_gid)
1500 		return (0);
1501 
1502 	error = VOP_GETATTR(vp, &va, cred);
1503 	if (error != 0)
1504 		return (error);
1505 
1506 	if (hardlink_check_uid && cred->cr_uid != va.va_uid) {
1507 		error = priv_check_cred(cred, PRIV_VFS_LINK, 0);
1508 		if (error)
1509 			return (error);
1510 	}
1511 
1512 	if (hardlink_check_gid && !groupmember(va.va_gid, cred)) {
1513 		error = priv_check_cred(cred, PRIV_VFS_LINK, 0);
1514 		if (error)
1515 			return (error);
1516 	}
1517 
1518 	return (0);
1519 }
1520 
1521 int
1522 kern_link(struct thread *td, char *path, char *link, enum uio_seg segflg)
1523 {
1524 
1525 	return (kern_linkat(td, AT_FDCWD, AT_FDCWD, path,link, segflg, FOLLOW));
1526 }
1527 
1528 int
1529 kern_linkat(struct thread *td, int fd1, int fd2, char *path1, char *path2,
1530     enum uio_seg segflg, int follow)
1531 {
1532 	struct vnode *vp;
1533 	struct mount *mp;
1534 	struct nameidata nd;
1535 	int error;
1536 
1537 	bwillwrite();
1538 	NDINIT_AT(&nd, LOOKUP, follow | AUDITVNODE1, segflg, path1, fd1, td);
1539 
1540 	if ((error = namei(&nd)) != 0)
1541 		return (error);
1542 	NDFREE(&nd, NDF_ONLY_PNBUF);
1543 	vp = nd.ni_vp;
1544 	if (vp->v_type == VDIR) {
1545 		vrele(vp);
1546 		return (EPERM);		/* POSIX */
1547 	}
1548 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) {
1549 		vrele(vp);
1550 		return (error);
1551 	}
1552 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE2,
1553 	    segflg, path2, fd2, CAP_CREATE, td);
1554 	if ((error = namei(&nd)) == 0) {
1555 		if (nd.ni_vp != NULL) {
1556 			if (nd.ni_dvp == nd.ni_vp)
1557 				vrele(nd.ni_dvp);
1558 			else
1559 				vput(nd.ni_dvp);
1560 			vrele(nd.ni_vp);
1561 			error = EEXIST;
1562 		} else if ((error = vn_lock(vp, LK_EXCLUSIVE | LK_RETRY))
1563 		    == 0) {
1564 			error = can_hardlink(vp, td->td_ucred);
1565 			if (error == 0)
1566 #ifdef MAC
1567 				error = mac_vnode_check_link(td->td_ucred,
1568 				    nd.ni_dvp, vp, &nd.ni_cnd);
1569 			if (error == 0)
1570 #endif
1571 				error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
1572 			VOP_UNLOCK(vp, 0);
1573 			vput(nd.ni_dvp);
1574 		}
1575 		NDFREE(&nd, NDF_ONLY_PNBUF);
1576 	}
1577 	vrele(vp);
1578 	vn_finished_write(mp);
1579 	return (error);
1580 }
1581 
1582 /*
1583  * Make a symbolic link.
1584  */
1585 #ifndef _SYS_SYSPROTO_H_
1586 struct symlink_args {
1587 	char	*path;
1588 	char	*link;
1589 };
1590 #endif
1591 int
1592 sys_symlink(td, uap)
1593 	struct thread *td;
1594 	register struct symlink_args /* {
1595 		char *path;
1596 		char *link;
1597 	} */ *uap;
1598 {
1599 
1600 	return (kern_symlink(td, uap->path, uap->link, UIO_USERSPACE));
1601 }
1602 
1603 #ifndef _SYS_SYSPROTO_H_
1604 struct symlinkat_args {
1605 	char	*path;
1606 	int	fd;
1607 	char	*path2;
1608 };
1609 #endif
1610 int
1611 sys_symlinkat(struct thread *td, struct symlinkat_args *uap)
1612 {
1613 
1614 	return (kern_symlinkat(td, uap->path1, uap->fd, uap->path2,
1615 	    UIO_USERSPACE));
1616 }
1617 
1618 int
1619 kern_symlink(struct thread *td, char *path, char *link, enum uio_seg segflg)
1620 {
1621 
1622 	return (kern_symlinkat(td, path, AT_FDCWD, link, segflg));
1623 }
1624 
1625 int
1626 kern_symlinkat(struct thread *td, char *path1, int fd, char *path2,
1627     enum uio_seg segflg)
1628 {
1629 	struct mount *mp;
1630 	struct vattr vattr;
1631 	char *syspath;
1632 	int error;
1633 	struct nameidata nd;
1634 
1635 	if (segflg == UIO_SYSSPACE) {
1636 		syspath = path1;
1637 	} else {
1638 		syspath = uma_zalloc(namei_zone, M_WAITOK);
1639 		if ((error = copyinstr(path1, syspath, MAXPATHLEN, NULL)) != 0)
1640 			goto out;
1641 	}
1642 	AUDIT_ARG_TEXT(syspath);
1643 restart:
1644 	bwillwrite();
1645 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1,
1646 	    segflg, path2, fd, CAP_CREATE, td);
1647 	if ((error = namei(&nd)) != 0)
1648 		goto out;
1649 	if (nd.ni_vp) {
1650 		NDFREE(&nd, NDF_ONLY_PNBUF);
1651 		if (nd.ni_vp == nd.ni_dvp)
1652 			vrele(nd.ni_dvp);
1653 		else
1654 			vput(nd.ni_dvp);
1655 		vrele(nd.ni_vp);
1656 		error = EEXIST;
1657 		goto out;
1658 	}
1659 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1660 		NDFREE(&nd, NDF_ONLY_PNBUF);
1661 		vput(nd.ni_dvp);
1662 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1663 			goto out;
1664 		goto restart;
1665 	}
1666 	VATTR_NULL(&vattr);
1667 	vattr.va_mode = ACCESSPERMS &~ td->td_proc->p_fd->fd_cmask;
1668 #ifdef MAC
1669 	vattr.va_type = VLNK;
1670 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1671 	    &vattr);
1672 	if (error)
1673 		goto out2;
1674 #endif
1675 	error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, syspath);
1676 	if (error == 0)
1677 		vput(nd.ni_vp);
1678 #ifdef MAC
1679 out2:
1680 #endif
1681 	NDFREE(&nd, NDF_ONLY_PNBUF);
1682 	vput(nd.ni_dvp);
1683 	vn_finished_write(mp);
1684 out:
1685 	if (segflg != UIO_SYSSPACE)
1686 		uma_zfree(namei_zone, syspath);
1687 	return (error);
1688 }
1689 
1690 /*
1691  * Delete a whiteout from the filesystem.
1692  */
1693 int
1694 sys_undelete(td, uap)
1695 	struct thread *td;
1696 	register struct undelete_args /* {
1697 		char *path;
1698 	} */ *uap;
1699 {
1700 	int error;
1701 	struct mount *mp;
1702 	struct nameidata nd;
1703 
1704 restart:
1705 	bwillwrite();
1706 	NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | AUDITVNODE1,
1707 	    UIO_USERSPACE, uap->path, td);
1708 	error = namei(&nd);
1709 	if (error)
1710 		return (error);
1711 
1712 	if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) {
1713 		NDFREE(&nd, NDF_ONLY_PNBUF);
1714 		if (nd.ni_vp == nd.ni_dvp)
1715 			vrele(nd.ni_dvp);
1716 		else
1717 			vput(nd.ni_dvp);
1718 		if (nd.ni_vp)
1719 			vrele(nd.ni_vp);
1720 		return (EEXIST);
1721 	}
1722 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1723 		NDFREE(&nd, NDF_ONLY_PNBUF);
1724 		vput(nd.ni_dvp);
1725 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1726 			return (error);
1727 		goto restart;
1728 	}
1729 	error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE);
1730 	NDFREE(&nd, NDF_ONLY_PNBUF);
1731 	vput(nd.ni_dvp);
1732 	vn_finished_write(mp);
1733 	return (error);
1734 }
1735 
1736 /*
1737  * Delete a name from the filesystem.
1738  */
1739 #ifndef _SYS_SYSPROTO_H_
1740 struct unlink_args {
1741 	char	*path;
1742 };
1743 #endif
1744 int
1745 sys_unlink(td, uap)
1746 	struct thread *td;
1747 	struct unlink_args /* {
1748 		char *path;
1749 	} */ *uap;
1750 {
1751 
1752 	return (kern_unlink(td, uap->path, UIO_USERSPACE));
1753 }
1754 
1755 #ifndef _SYS_SYSPROTO_H_
1756 struct unlinkat_args {
1757 	int	fd;
1758 	char	*path;
1759 	int	flag;
1760 };
1761 #endif
1762 int
1763 sys_unlinkat(struct thread *td, struct unlinkat_args *uap)
1764 {
1765 	int flag = uap->flag;
1766 	int fd = uap->fd;
1767 	char *path = uap->path;
1768 
1769 	if (flag & ~AT_REMOVEDIR)
1770 		return (EINVAL);
1771 
1772 	if (flag & AT_REMOVEDIR)
1773 		return (kern_rmdirat(td, fd, path, UIO_USERSPACE));
1774 	else
1775 		return (kern_unlinkat(td, fd, path, UIO_USERSPACE, 0));
1776 }
1777 
1778 int
1779 kern_unlink(struct thread *td, char *path, enum uio_seg pathseg)
1780 {
1781 
1782 	return (kern_unlinkat(td, AT_FDCWD, path, pathseg, 0));
1783 }
1784 
1785 int
1786 kern_unlinkat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1787     ino_t oldinum)
1788 {
1789 	struct mount *mp;
1790 	struct vnode *vp;
1791 	int error;
1792 	struct nameidata nd;
1793 	struct stat sb;
1794 
1795 restart:
1796 	bwillwrite();
1797 	NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1,
1798 	    pathseg, path, fd, CAP_DELETE, td);
1799 	if ((error = namei(&nd)) != 0)
1800 		return (error == EINVAL ? EPERM : error);
1801 	vp = nd.ni_vp;
1802 	if (vp->v_type == VDIR && oldinum == 0) {
1803 		error = EPERM;		/* POSIX */
1804 	} else if (oldinum != 0 &&
1805 		  ((error = vn_stat(vp, &sb, td->td_ucred, NOCRED, td)) == 0) &&
1806 		  sb.st_ino != oldinum) {
1807 			error = EIDRM;	/* Identifier removed */
1808 	} else {
1809 		/*
1810 		 * The root of a mounted filesystem cannot be deleted.
1811 		 *
1812 		 * XXX: can this only be a VDIR case?
1813 		 */
1814 		if (vp->v_vflag & VV_ROOT)
1815 			error = EBUSY;
1816 	}
1817 	if (error == 0) {
1818 		if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1819 			NDFREE(&nd, NDF_ONLY_PNBUF);
1820 			vput(nd.ni_dvp);
1821 			if (vp == nd.ni_dvp)
1822 				vrele(vp);
1823 			else
1824 				vput(vp);
1825 			if ((error = vn_start_write(NULL, &mp,
1826 			    V_XSLEEP | PCATCH)) != 0)
1827 				return (error);
1828 			goto restart;
1829 		}
1830 #ifdef MAC
1831 		error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
1832 		    &nd.ni_cnd);
1833 		if (error)
1834 			goto out;
1835 #endif
1836 		error = VOP_REMOVE(nd.ni_dvp, vp, &nd.ni_cnd);
1837 #ifdef MAC
1838 out:
1839 #endif
1840 		vn_finished_write(mp);
1841 	}
1842 	NDFREE(&nd, NDF_ONLY_PNBUF);
1843 	vput(nd.ni_dvp);
1844 	if (vp == nd.ni_dvp)
1845 		vrele(vp);
1846 	else
1847 		vput(vp);
1848 	return (error);
1849 }
1850 
1851 /*
1852  * Reposition read/write file offset.
1853  */
1854 #ifndef _SYS_SYSPROTO_H_
1855 struct lseek_args {
1856 	int	fd;
1857 	int	pad;
1858 	off_t	offset;
1859 	int	whence;
1860 };
1861 #endif
1862 int
1863 sys_lseek(td, uap)
1864 	struct thread *td;
1865 	register struct lseek_args /* {
1866 		int fd;
1867 		int pad;
1868 		off_t offset;
1869 		int whence;
1870 	} */ *uap;
1871 {
1872 	struct ucred *cred = td->td_ucred;
1873 	struct file *fp;
1874 	struct vnode *vp;
1875 	struct vattr vattr;
1876 	off_t foffset, offset, size;
1877 	int error, noneg;
1878 
1879 	AUDIT_ARG_FD(uap->fd);
1880 	if ((error = fget(td, uap->fd, CAP_SEEK, &fp)) != 0)
1881 		return (error);
1882 	if (!(fp->f_ops->fo_flags & DFLAG_SEEKABLE)) {
1883 		fdrop(fp, td);
1884 		return (ESPIPE);
1885 	}
1886 	vp = fp->f_vnode;
1887 	foffset = foffset_lock(fp, 0);
1888 	noneg = (vp->v_type != VCHR);
1889 	offset = uap->offset;
1890 	switch (uap->whence) {
1891 	case L_INCR:
1892 		if (noneg &&
1893 		    (foffset < 0 ||
1894 		    (offset > 0 && foffset > OFF_MAX - offset))) {
1895 			error = EOVERFLOW;
1896 			break;
1897 		}
1898 		offset += foffset;
1899 		break;
1900 	case L_XTND:
1901 		vn_lock(vp, LK_SHARED | LK_RETRY);
1902 		error = VOP_GETATTR(vp, &vattr, cred);
1903 		VOP_UNLOCK(vp, 0);
1904 		if (error)
1905 			break;
1906 
1907 		/*
1908 		 * If the file references a disk device, then fetch
1909 		 * the media size and use that to determine the ending
1910 		 * offset.
1911 		 */
1912 		if (vattr.va_size == 0 && vp->v_type == VCHR &&
1913 		    fo_ioctl(fp, DIOCGMEDIASIZE, &size, cred, td) == 0)
1914 			vattr.va_size = size;
1915 		if (noneg &&
1916 		    (vattr.va_size > OFF_MAX ||
1917 		    (offset > 0 && vattr.va_size > OFF_MAX - offset))) {
1918 			error = EOVERFLOW;
1919 			break;
1920 		}
1921 		offset += vattr.va_size;
1922 		break;
1923 	case L_SET:
1924 		break;
1925 	case SEEK_DATA:
1926 		error = fo_ioctl(fp, FIOSEEKDATA, &offset, cred, td);
1927 		break;
1928 	case SEEK_HOLE:
1929 		error = fo_ioctl(fp, FIOSEEKHOLE, &offset, cred, td);
1930 		break;
1931 	default:
1932 		error = EINVAL;
1933 	}
1934 	if (error == 0 && noneg && offset < 0)
1935 		error = EINVAL;
1936 	if (error != 0)
1937 		goto drop;
1938 	VFS_KNOTE_UNLOCKED(vp, 0);
1939 	*(off_t *)(td->td_retval) = offset;
1940 drop:
1941 	fdrop(fp, td);
1942 	foffset_unlock(fp, offset, error != 0 ? FOF_NOUPDATE : 0);
1943 	return (error);
1944 }
1945 
1946 #if defined(COMPAT_43)
1947 /*
1948  * Reposition read/write file offset.
1949  */
1950 #ifndef _SYS_SYSPROTO_H_
1951 struct olseek_args {
1952 	int	fd;
1953 	long	offset;
1954 	int	whence;
1955 };
1956 #endif
1957 int
1958 olseek(td, uap)
1959 	struct thread *td;
1960 	register struct olseek_args /* {
1961 		int fd;
1962 		long offset;
1963 		int whence;
1964 	} */ *uap;
1965 {
1966 	struct lseek_args /* {
1967 		int fd;
1968 		int pad;
1969 		off_t offset;
1970 		int whence;
1971 	} */ nuap;
1972 
1973 	nuap.fd = uap->fd;
1974 	nuap.offset = uap->offset;
1975 	nuap.whence = uap->whence;
1976 	return (sys_lseek(td, &nuap));
1977 }
1978 #endif /* COMPAT_43 */
1979 
1980 /* Version with the 'pad' argument */
1981 int
1982 freebsd6_lseek(td, uap)
1983 	struct thread *td;
1984 	register struct freebsd6_lseek_args *uap;
1985 {
1986 	struct lseek_args ouap;
1987 
1988 	ouap.fd = uap->fd;
1989 	ouap.offset = uap->offset;
1990 	ouap.whence = uap->whence;
1991 	return (sys_lseek(td, &ouap));
1992 }
1993 
1994 /*
1995  * Check access permissions using passed credentials.
1996  */
1997 static int
1998 vn_access(vp, user_flags, cred, td)
1999 	struct vnode	*vp;
2000 	int		user_flags;
2001 	struct ucred	*cred;
2002 	struct thread	*td;
2003 {
2004 	int error;
2005 	accmode_t accmode;
2006 
2007 	/* Flags == 0 means only check for existence. */
2008 	error = 0;
2009 	if (user_flags) {
2010 		accmode = 0;
2011 		if (user_flags & R_OK)
2012 			accmode |= VREAD;
2013 		if (user_flags & W_OK)
2014 			accmode |= VWRITE;
2015 		if (user_flags & X_OK)
2016 			accmode |= VEXEC;
2017 #ifdef MAC
2018 		error = mac_vnode_check_access(cred, vp, accmode);
2019 		if (error)
2020 			return (error);
2021 #endif
2022 		if ((accmode & VWRITE) == 0 || (error = vn_writechk(vp)) == 0)
2023 			error = VOP_ACCESS(vp, accmode, cred, td);
2024 	}
2025 	return (error);
2026 }
2027 
2028 /*
2029  * Check access permissions using "real" credentials.
2030  */
2031 #ifndef _SYS_SYSPROTO_H_
2032 struct access_args {
2033 	char	*path;
2034 	int	amode;
2035 };
2036 #endif
2037 int
2038 sys_access(td, uap)
2039 	struct thread *td;
2040 	register struct access_args /* {
2041 		char *path;
2042 		int amode;
2043 	} */ *uap;
2044 {
2045 
2046 	return (kern_access(td, uap->path, UIO_USERSPACE, uap->amode));
2047 }
2048 
2049 #ifndef _SYS_SYSPROTO_H_
2050 struct faccessat_args {
2051 	int	dirfd;
2052 	char	*path;
2053 	int	amode;
2054 	int	flag;
2055 }
2056 #endif
2057 int
2058 sys_faccessat(struct thread *td, struct faccessat_args *uap)
2059 {
2060 
2061 	if (uap->flag & ~AT_EACCESS)
2062 		return (EINVAL);
2063 	return (kern_accessat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
2064 	    uap->amode));
2065 }
2066 
2067 int
2068 kern_access(struct thread *td, char *path, enum uio_seg pathseg, int amode)
2069 {
2070 
2071 	return (kern_accessat(td, AT_FDCWD, path, pathseg, 0, amode));
2072 }
2073 
2074 int
2075 kern_accessat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2076     int flag, int amode)
2077 {
2078 	struct ucred *cred, *tmpcred;
2079 	struct vnode *vp;
2080 	struct nameidata nd;
2081 	int error;
2082 
2083 	/*
2084 	 * Create and modify a temporary credential instead of one that
2085 	 * is potentially shared.
2086 	 */
2087 	if (!(flag & AT_EACCESS)) {
2088 		cred = td->td_ucred;
2089 		tmpcred = crdup(cred);
2090 		tmpcred->cr_uid = cred->cr_ruid;
2091 		tmpcred->cr_groups[0] = cred->cr_rgid;
2092 		td->td_ucred = tmpcred;
2093 	} else
2094 		cred = tmpcred = td->td_ucred;
2095 	AUDIT_ARG_VALUE(amode);
2096 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF |
2097 	    AUDITVNODE1, pathseg, path, fd, CAP_FSTAT, td);
2098 	if ((error = namei(&nd)) != 0)
2099 		goto out1;
2100 	vp = nd.ni_vp;
2101 
2102 	error = vn_access(vp, amode, tmpcred, td);
2103 	NDFREE(&nd, NDF_ONLY_PNBUF);
2104 	vput(vp);
2105 out1:
2106 	if (!(flag & AT_EACCESS)) {
2107 		td->td_ucred = cred;
2108 		crfree(tmpcred);
2109 	}
2110 	return (error);
2111 }
2112 
2113 /*
2114  * Check access permissions using "effective" credentials.
2115  */
2116 #ifndef _SYS_SYSPROTO_H_
2117 struct eaccess_args {
2118 	char	*path;
2119 	int	amode;
2120 };
2121 #endif
2122 int
2123 sys_eaccess(td, uap)
2124 	struct thread *td;
2125 	register struct eaccess_args /* {
2126 		char *path;
2127 		int amode;
2128 	} */ *uap;
2129 {
2130 
2131 	return (kern_eaccess(td, uap->path, UIO_USERSPACE, uap->amode));
2132 }
2133 
2134 int
2135 kern_eaccess(struct thread *td, char *path, enum uio_seg pathseg, int amode)
2136 {
2137 
2138 	return (kern_accessat(td, AT_FDCWD, path, pathseg, AT_EACCESS, amode));
2139 }
2140 
2141 #if defined(COMPAT_43)
2142 /*
2143  * Get file status; this version follows links.
2144  */
2145 #ifndef _SYS_SYSPROTO_H_
2146 struct ostat_args {
2147 	char	*path;
2148 	struct ostat *ub;
2149 };
2150 #endif
2151 int
2152 ostat(td, uap)
2153 	struct thread *td;
2154 	register struct ostat_args /* {
2155 		char *path;
2156 		struct ostat *ub;
2157 	} */ *uap;
2158 {
2159 	struct stat sb;
2160 	struct ostat osb;
2161 	int error;
2162 
2163 	error = kern_stat(td, uap->path, UIO_USERSPACE, &sb);
2164 	if (error)
2165 		return (error);
2166 	cvtstat(&sb, &osb);
2167 	error = copyout(&osb, uap->ub, sizeof (osb));
2168 	return (error);
2169 }
2170 
2171 /*
2172  * Get file status; this version does not follow links.
2173  */
2174 #ifndef _SYS_SYSPROTO_H_
2175 struct olstat_args {
2176 	char	*path;
2177 	struct ostat *ub;
2178 };
2179 #endif
2180 int
2181 olstat(td, uap)
2182 	struct thread *td;
2183 	register struct olstat_args /* {
2184 		char *path;
2185 		struct ostat *ub;
2186 	} */ *uap;
2187 {
2188 	struct stat sb;
2189 	struct ostat osb;
2190 	int error;
2191 
2192 	error = kern_lstat(td, uap->path, UIO_USERSPACE, &sb);
2193 	if (error)
2194 		return (error);
2195 	cvtstat(&sb, &osb);
2196 	error = copyout(&osb, uap->ub, sizeof (osb));
2197 	return (error);
2198 }
2199 
2200 /*
2201  * Convert from an old to a new stat structure.
2202  */
2203 void
2204 cvtstat(st, ost)
2205 	struct stat *st;
2206 	struct ostat *ost;
2207 {
2208 
2209 	ost->st_dev = st->st_dev;
2210 	ost->st_ino = st->st_ino;
2211 	ost->st_mode = st->st_mode;
2212 	ost->st_nlink = st->st_nlink;
2213 	ost->st_uid = st->st_uid;
2214 	ost->st_gid = st->st_gid;
2215 	ost->st_rdev = st->st_rdev;
2216 	if (st->st_size < (quad_t)1 << 32)
2217 		ost->st_size = st->st_size;
2218 	else
2219 		ost->st_size = -2;
2220 	ost->st_atim = st->st_atim;
2221 	ost->st_mtim = st->st_mtim;
2222 	ost->st_ctim = st->st_ctim;
2223 	ost->st_blksize = st->st_blksize;
2224 	ost->st_blocks = st->st_blocks;
2225 	ost->st_flags = st->st_flags;
2226 	ost->st_gen = st->st_gen;
2227 }
2228 #endif /* COMPAT_43 */
2229 
2230 /*
2231  * Get file status; this version follows links.
2232  */
2233 #ifndef _SYS_SYSPROTO_H_
2234 struct stat_args {
2235 	char	*path;
2236 	struct stat *ub;
2237 };
2238 #endif
2239 int
2240 sys_stat(td, uap)
2241 	struct thread *td;
2242 	register struct stat_args /* {
2243 		char *path;
2244 		struct stat *ub;
2245 	} */ *uap;
2246 {
2247 	struct stat sb;
2248 	int error;
2249 
2250 	error = kern_stat(td, uap->path, UIO_USERSPACE, &sb);
2251 	if (error == 0)
2252 		error = copyout(&sb, uap->ub, sizeof (sb));
2253 	return (error);
2254 }
2255 
2256 #ifndef _SYS_SYSPROTO_H_
2257 struct fstatat_args {
2258 	int	fd;
2259 	char	*path;
2260 	struct stat	*buf;
2261 	int	flag;
2262 }
2263 #endif
2264 int
2265 sys_fstatat(struct thread *td, struct fstatat_args *uap)
2266 {
2267 	struct stat sb;
2268 	int error;
2269 
2270 	error = kern_statat(td, uap->flag, uap->fd, uap->path,
2271 	    UIO_USERSPACE, &sb);
2272 	if (error == 0)
2273 		error = copyout(&sb, uap->buf, sizeof (sb));
2274 	return (error);
2275 }
2276 
2277 int
2278 kern_stat(struct thread *td, char *path, enum uio_seg pathseg, struct stat *sbp)
2279 {
2280 
2281 	return (kern_statat(td, 0, AT_FDCWD, path, pathseg, sbp));
2282 }
2283 
2284 int
2285 kern_statat(struct thread *td, int flag, int fd, char *path,
2286     enum uio_seg pathseg, struct stat *sbp)
2287 {
2288 
2289 	return (kern_statat_vnhook(td, flag, fd, path, pathseg, sbp, NULL));
2290 }
2291 
2292 int
2293 kern_statat_vnhook(struct thread *td, int flag, int fd, char *path,
2294     enum uio_seg pathseg, struct stat *sbp,
2295     void (*hook)(struct vnode *vp, struct stat *sbp))
2296 {
2297 	struct nameidata nd;
2298 	struct stat sb;
2299 	int error;
2300 
2301 	if (flag & ~AT_SYMLINK_NOFOLLOW)
2302 		return (EINVAL);
2303 
2304 	NDINIT_ATRIGHTS(&nd, LOOKUP, ((flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW :
2305 	    FOLLOW) | LOCKSHARED | LOCKLEAF | AUDITVNODE1, pathseg, path, fd,
2306 	    CAP_FSTAT, td);
2307 
2308 	if ((error = namei(&nd)) != 0)
2309 		return (error);
2310 	error = vn_stat(nd.ni_vp, &sb, td->td_ucred, NOCRED, td);
2311 	if (!error) {
2312 		SDT_PROBE(vfs, , stat, mode, path, sb.st_mode, 0, 0, 0);
2313 		if (S_ISREG(sb.st_mode))
2314 			SDT_PROBE(vfs, , stat, reg, path, pathseg, 0, 0, 0);
2315 		if (__predict_false(hook != NULL))
2316 			hook(nd.ni_vp, &sb);
2317 	}
2318 	NDFREE(&nd, NDF_ONLY_PNBUF);
2319 	vput(nd.ni_vp);
2320 	if (error)
2321 		return (error);
2322 	*sbp = sb;
2323 #ifdef KTRACE
2324 	if (KTRPOINT(td, KTR_STRUCT))
2325 		ktrstat(&sb);
2326 #endif
2327 	return (0);
2328 }
2329 
2330 /*
2331  * Get file status; this version does not follow links.
2332  */
2333 #ifndef _SYS_SYSPROTO_H_
2334 struct lstat_args {
2335 	char	*path;
2336 	struct stat *ub;
2337 };
2338 #endif
2339 int
2340 sys_lstat(td, uap)
2341 	struct thread *td;
2342 	register struct lstat_args /* {
2343 		char *path;
2344 		struct stat *ub;
2345 	} */ *uap;
2346 {
2347 	struct stat sb;
2348 	int error;
2349 
2350 	error = kern_lstat(td, uap->path, UIO_USERSPACE, &sb);
2351 	if (error == 0)
2352 		error = copyout(&sb, uap->ub, sizeof (sb));
2353 	return (error);
2354 }
2355 
2356 int
2357 kern_lstat(struct thread *td, char *path, enum uio_seg pathseg, struct stat *sbp)
2358 {
2359 
2360 	return (kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, path, pathseg,
2361 	    sbp));
2362 }
2363 
2364 /*
2365  * Implementation of the NetBSD [l]stat() functions.
2366  */
2367 void
2368 cvtnstat(sb, nsb)
2369 	struct stat *sb;
2370 	struct nstat *nsb;
2371 {
2372 	bzero(nsb, sizeof *nsb);
2373 	nsb->st_dev = sb->st_dev;
2374 	nsb->st_ino = sb->st_ino;
2375 	nsb->st_mode = sb->st_mode;
2376 	nsb->st_nlink = sb->st_nlink;
2377 	nsb->st_uid = sb->st_uid;
2378 	nsb->st_gid = sb->st_gid;
2379 	nsb->st_rdev = sb->st_rdev;
2380 	nsb->st_atim = sb->st_atim;
2381 	nsb->st_mtim = sb->st_mtim;
2382 	nsb->st_ctim = sb->st_ctim;
2383 	nsb->st_size = sb->st_size;
2384 	nsb->st_blocks = sb->st_blocks;
2385 	nsb->st_blksize = sb->st_blksize;
2386 	nsb->st_flags = sb->st_flags;
2387 	nsb->st_gen = sb->st_gen;
2388 	nsb->st_birthtim = sb->st_birthtim;
2389 }
2390 
2391 #ifndef _SYS_SYSPROTO_H_
2392 struct nstat_args {
2393 	char	*path;
2394 	struct nstat *ub;
2395 };
2396 #endif
2397 int
2398 sys_nstat(td, uap)
2399 	struct thread *td;
2400 	register struct nstat_args /* {
2401 		char *path;
2402 		struct nstat *ub;
2403 	} */ *uap;
2404 {
2405 	struct stat sb;
2406 	struct nstat nsb;
2407 	int error;
2408 
2409 	error = kern_stat(td, uap->path, UIO_USERSPACE, &sb);
2410 	if (error)
2411 		return (error);
2412 	cvtnstat(&sb, &nsb);
2413 	error = copyout(&nsb, uap->ub, sizeof (nsb));
2414 	return (error);
2415 }
2416 
2417 /*
2418  * NetBSD lstat.  Get file status; this version does not follow links.
2419  */
2420 #ifndef _SYS_SYSPROTO_H_
2421 struct lstat_args {
2422 	char	*path;
2423 	struct stat *ub;
2424 };
2425 #endif
2426 int
2427 sys_nlstat(td, uap)
2428 	struct thread *td;
2429 	register struct nlstat_args /* {
2430 		char *path;
2431 		struct nstat *ub;
2432 	} */ *uap;
2433 {
2434 	struct stat sb;
2435 	struct nstat nsb;
2436 	int error;
2437 
2438 	error = kern_lstat(td, uap->path, UIO_USERSPACE, &sb);
2439 	if (error)
2440 		return (error);
2441 	cvtnstat(&sb, &nsb);
2442 	error = copyout(&nsb, uap->ub, sizeof (nsb));
2443 	return (error);
2444 }
2445 
2446 /*
2447  * Get configurable pathname variables.
2448  */
2449 #ifndef _SYS_SYSPROTO_H_
2450 struct pathconf_args {
2451 	char	*path;
2452 	int	name;
2453 };
2454 #endif
2455 int
2456 sys_pathconf(td, uap)
2457 	struct thread *td;
2458 	register struct pathconf_args /* {
2459 		char *path;
2460 		int name;
2461 	} */ *uap;
2462 {
2463 
2464 	return (kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name, FOLLOW));
2465 }
2466 
2467 #ifndef _SYS_SYSPROTO_H_
2468 struct lpathconf_args {
2469 	char	*path;
2470 	int	name;
2471 };
2472 #endif
2473 int
2474 sys_lpathconf(td, uap)
2475 	struct thread *td;
2476 	register struct lpathconf_args /* {
2477 		char *path;
2478 		int name;
2479 	} */ *uap;
2480 {
2481 
2482 	return (kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name,
2483 	    NOFOLLOW));
2484 }
2485 
2486 int
2487 kern_pathconf(struct thread *td, char *path, enum uio_seg pathseg, int name,
2488     u_long flags)
2489 {
2490 	struct nameidata nd;
2491 	int error;
2492 
2493 	NDINIT(&nd, LOOKUP, LOCKSHARED | LOCKLEAF | AUDITVNODE1 | flags,
2494 	    pathseg, path, td);
2495 	if ((error = namei(&nd)) != 0)
2496 		return (error);
2497 	NDFREE(&nd, NDF_ONLY_PNBUF);
2498 
2499 	/* If asynchronous I/O is available, it works for all files. */
2500 	if (name == _PC_ASYNC_IO)
2501 		td->td_retval[0] = async_io_version;
2502 	else
2503 		error = VOP_PATHCONF(nd.ni_vp, name, td->td_retval);
2504 	vput(nd.ni_vp);
2505 	return (error);
2506 }
2507 
2508 /*
2509  * Return target name of a symbolic link.
2510  */
2511 #ifndef _SYS_SYSPROTO_H_
2512 struct readlink_args {
2513 	char	*path;
2514 	char	*buf;
2515 	size_t	count;
2516 };
2517 #endif
2518 int
2519 sys_readlink(td, uap)
2520 	struct thread *td;
2521 	register struct readlink_args /* {
2522 		char *path;
2523 		char *buf;
2524 		size_t count;
2525 	} */ *uap;
2526 {
2527 
2528 	return (kern_readlink(td, uap->path, UIO_USERSPACE, uap->buf,
2529 	    UIO_USERSPACE, uap->count));
2530 }
2531 #ifndef _SYS_SYSPROTO_H_
2532 struct readlinkat_args {
2533 	int	fd;
2534 	char	*path;
2535 	char	*buf;
2536 	size_t	bufsize;
2537 };
2538 #endif
2539 int
2540 sys_readlinkat(struct thread *td, struct readlinkat_args *uap)
2541 {
2542 
2543 	return (kern_readlinkat(td, uap->fd, uap->path, UIO_USERSPACE,
2544 	    uap->buf, UIO_USERSPACE, uap->bufsize));
2545 }
2546 
2547 int
2548 kern_readlink(struct thread *td, char *path, enum uio_seg pathseg, char *buf,
2549     enum uio_seg bufseg, size_t count)
2550 {
2551 
2552 	return (kern_readlinkat(td, AT_FDCWD, path, pathseg, buf, bufseg,
2553 	    count));
2554 }
2555 
2556 int
2557 kern_readlinkat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2558     char *buf, enum uio_seg bufseg, size_t count)
2559 {
2560 	struct vnode *vp;
2561 	struct iovec aiov;
2562 	struct uio auio;
2563 	int error;
2564 	struct nameidata nd;
2565 
2566 	if (count > IOSIZE_MAX)
2567 		return (EINVAL);
2568 
2569 	NDINIT_AT(&nd, LOOKUP, NOFOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
2570 	    pathseg, path, fd, td);
2571 
2572 	if ((error = namei(&nd)) != 0)
2573 		return (error);
2574 	NDFREE(&nd, NDF_ONLY_PNBUF);
2575 	vp = nd.ni_vp;
2576 #ifdef MAC
2577 	error = mac_vnode_check_readlink(td->td_ucred, vp);
2578 	if (error) {
2579 		vput(vp);
2580 		return (error);
2581 	}
2582 #endif
2583 	if (vp->v_type != VLNK)
2584 		error = EINVAL;
2585 	else {
2586 		aiov.iov_base = buf;
2587 		aiov.iov_len = count;
2588 		auio.uio_iov = &aiov;
2589 		auio.uio_iovcnt = 1;
2590 		auio.uio_offset = 0;
2591 		auio.uio_rw = UIO_READ;
2592 		auio.uio_segflg = bufseg;
2593 		auio.uio_td = td;
2594 		auio.uio_resid = count;
2595 		error = VOP_READLINK(vp, &auio, td->td_ucred);
2596 	}
2597 	vput(vp);
2598 	td->td_retval[0] = count - auio.uio_resid;
2599 	return (error);
2600 }
2601 
2602 /*
2603  * Common implementation code for chflags() and fchflags().
2604  */
2605 static int
2606 setfflags(td, vp, flags)
2607 	struct thread *td;
2608 	struct vnode *vp;
2609 	int flags;
2610 {
2611 	int error;
2612 	struct mount *mp;
2613 	struct vattr vattr;
2614 
2615 	/* We can't support the value matching VNOVAL. */
2616 	if (flags == VNOVAL)
2617 		return (EOPNOTSUPP);
2618 
2619 	/*
2620 	 * Prevent non-root users from setting flags on devices.  When
2621 	 * a device is reused, users can retain ownership of the device
2622 	 * if they are allowed to set flags and programs assume that
2623 	 * chown can't fail when done as root.
2624 	 */
2625 	if (vp->v_type == VCHR || vp->v_type == VBLK) {
2626 		error = priv_check(td, PRIV_VFS_CHFLAGS_DEV);
2627 		if (error)
2628 			return (error);
2629 	}
2630 
2631 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2632 		return (error);
2633 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2634 	VATTR_NULL(&vattr);
2635 	vattr.va_flags = flags;
2636 #ifdef MAC
2637 	error = mac_vnode_check_setflags(td->td_ucred, vp, vattr.va_flags);
2638 	if (error == 0)
2639 #endif
2640 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
2641 	VOP_UNLOCK(vp, 0);
2642 	vn_finished_write(mp);
2643 	return (error);
2644 }
2645 
2646 /*
2647  * Change flags of a file given a path name.
2648  */
2649 #ifndef _SYS_SYSPROTO_H_
2650 struct chflags_args {
2651 	char	*path;
2652 	int	flags;
2653 };
2654 #endif
2655 int
2656 sys_chflags(td, uap)
2657 	struct thread *td;
2658 	register struct chflags_args /* {
2659 		char *path;
2660 		int flags;
2661 	} */ *uap;
2662 {
2663 	int error;
2664 	struct nameidata nd;
2665 
2666 	AUDIT_ARG_FFLAGS(uap->flags);
2667 	NDINIT(&nd, LOOKUP, FOLLOW | AUDITVNODE1, UIO_USERSPACE, uap->path, td);
2668 	if ((error = namei(&nd)) != 0)
2669 		return (error);
2670 	NDFREE(&nd, NDF_ONLY_PNBUF);
2671 	error = setfflags(td, nd.ni_vp, uap->flags);
2672 	vrele(nd.ni_vp);
2673 	return (error);
2674 }
2675 
2676 /*
2677  * Same as chflags() but doesn't follow symlinks.
2678  */
2679 int
2680 sys_lchflags(td, uap)
2681 	struct thread *td;
2682 	register struct lchflags_args /* {
2683 		char *path;
2684 		int flags;
2685 	} */ *uap;
2686 {
2687 	int error;
2688 	struct nameidata nd;
2689 
2690 	AUDIT_ARG_FFLAGS(uap->flags);
2691 	NDINIT(&nd, LOOKUP, NOFOLLOW | AUDITVNODE1, UIO_USERSPACE, uap->path,
2692 	    td);
2693 	if ((error = namei(&nd)) != 0)
2694 		return (error);
2695 	NDFREE(&nd, NDF_ONLY_PNBUF);
2696 	error = setfflags(td, nd.ni_vp, uap->flags);
2697 	vrele(nd.ni_vp);
2698 	return (error);
2699 }
2700 
2701 /*
2702  * Change flags of a file given a file descriptor.
2703  */
2704 #ifndef _SYS_SYSPROTO_H_
2705 struct fchflags_args {
2706 	int	fd;
2707 	int	flags;
2708 };
2709 #endif
2710 int
2711 sys_fchflags(td, uap)
2712 	struct thread *td;
2713 	register struct fchflags_args /* {
2714 		int fd;
2715 		int flags;
2716 	} */ *uap;
2717 {
2718 	struct file *fp;
2719 	int error;
2720 
2721 	AUDIT_ARG_FD(uap->fd);
2722 	AUDIT_ARG_FFLAGS(uap->flags);
2723 	if ((error = getvnode(td->td_proc->p_fd, uap->fd, CAP_FCHFLAGS,
2724 	    &fp)) != 0)
2725 		return (error);
2726 #ifdef AUDIT
2727 	vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
2728 	AUDIT_ARG_VNODE1(fp->f_vnode);
2729 	VOP_UNLOCK(fp->f_vnode, 0);
2730 #endif
2731 	error = setfflags(td, fp->f_vnode, uap->flags);
2732 	fdrop(fp, td);
2733 	return (error);
2734 }
2735 
2736 /*
2737  * Common implementation code for chmod(), lchmod() and fchmod().
2738  */
2739 int
2740 setfmode(td, cred, vp, mode)
2741 	struct thread *td;
2742 	struct ucred *cred;
2743 	struct vnode *vp;
2744 	int mode;
2745 {
2746 	int error;
2747 	struct mount *mp;
2748 	struct vattr vattr;
2749 
2750 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2751 		return (error);
2752 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2753 	VATTR_NULL(&vattr);
2754 	vattr.va_mode = mode & ALLPERMS;
2755 #ifdef MAC
2756 	error = mac_vnode_check_setmode(cred, vp, vattr.va_mode);
2757 	if (error == 0)
2758 #endif
2759 		error = VOP_SETATTR(vp, &vattr, cred);
2760 	VOP_UNLOCK(vp, 0);
2761 	vn_finished_write(mp);
2762 	return (error);
2763 }
2764 
2765 /*
2766  * Change mode of a file given path name.
2767  */
2768 #ifndef _SYS_SYSPROTO_H_
2769 struct chmod_args {
2770 	char	*path;
2771 	int	mode;
2772 };
2773 #endif
2774 int
2775 sys_chmod(td, uap)
2776 	struct thread *td;
2777 	register struct chmod_args /* {
2778 		char *path;
2779 		int mode;
2780 	} */ *uap;
2781 {
2782 
2783 	return (kern_chmod(td, uap->path, UIO_USERSPACE, uap->mode));
2784 }
2785 
2786 #ifndef _SYS_SYSPROTO_H_
2787 struct fchmodat_args {
2788 	int	dirfd;
2789 	char	*path;
2790 	mode_t	mode;
2791 	int	flag;
2792 }
2793 #endif
2794 int
2795 sys_fchmodat(struct thread *td, struct fchmodat_args *uap)
2796 {
2797 	int flag = uap->flag;
2798 	int fd = uap->fd;
2799 	char *path = uap->path;
2800 	mode_t mode = uap->mode;
2801 
2802 	if (flag & ~AT_SYMLINK_NOFOLLOW)
2803 		return (EINVAL);
2804 
2805 	return (kern_fchmodat(td, fd, path, UIO_USERSPACE, mode, flag));
2806 }
2807 
2808 int
2809 kern_chmod(struct thread *td, char *path, enum uio_seg pathseg, int mode)
2810 {
2811 
2812 	return (kern_fchmodat(td, AT_FDCWD, path, pathseg, mode, 0));
2813 }
2814 
2815 /*
2816  * Change mode of a file given path name (don't follow links.)
2817  */
2818 #ifndef _SYS_SYSPROTO_H_
2819 struct lchmod_args {
2820 	char	*path;
2821 	int	mode;
2822 };
2823 #endif
2824 int
2825 sys_lchmod(td, uap)
2826 	struct thread *td;
2827 	register struct lchmod_args /* {
2828 		char *path;
2829 		int mode;
2830 	} */ *uap;
2831 {
2832 
2833 	return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2834 	    uap->mode, AT_SYMLINK_NOFOLLOW));
2835 }
2836 
2837 
2838 int
2839 kern_fchmodat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2840     mode_t mode, int flag)
2841 {
2842 	int error;
2843 	struct nameidata nd;
2844 	int follow;
2845 
2846 	AUDIT_ARG_MODE(mode);
2847 	follow = (flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
2848 	NDINIT_ATRIGHTS(&nd, LOOKUP,  follow | AUDITVNODE1, pathseg, path, fd,
2849 	    CAP_FCHMOD, td);
2850 	if ((error = namei(&nd)) != 0)
2851 		return (error);
2852 	NDFREE(&nd, NDF_ONLY_PNBUF);
2853 	error = setfmode(td, td->td_ucred, nd.ni_vp, mode);
2854 	vrele(nd.ni_vp);
2855 	return (error);
2856 }
2857 
2858 /*
2859  * Change mode of a file given a file descriptor.
2860  */
2861 #ifndef _SYS_SYSPROTO_H_
2862 struct fchmod_args {
2863 	int	fd;
2864 	int	mode;
2865 };
2866 #endif
2867 int
2868 sys_fchmod(struct thread *td, struct fchmod_args *uap)
2869 {
2870 	struct file *fp;
2871 	int error;
2872 
2873 	AUDIT_ARG_FD(uap->fd);
2874 	AUDIT_ARG_MODE(uap->mode);
2875 
2876 	error = fget(td, uap->fd, CAP_FCHMOD, &fp);
2877 	if (error != 0)
2878 		return (error);
2879 	error = fo_chmod(fp, uap->mode, td->td_ucred, td);
2880 	fdrop(fp, td);
2881 	return (error);
2882 }
2883 
2884 /*
2885  * Common implementation for chown(), lchown(), and fchown()
2886  */
2887 int
2888 setfown(td, cred, vp, uid, gid)
2889 	struct thread *td;
2890 	struct ucred *cred;
2891 	struct vnode *vp;
2892 	uid_t uid;
2893 	gid_t gid;
2894 {
2895 	int error;
2896 	struct mount *mp;
2897 	struct vattr vattr;
2898 
2899 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2900 		return (error);
2901 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2902 	VATTR_NULL(&vattr);
2903 	vattr.va_uid = uid;
2904 	vattr.va_gid = gid;
2905 #ifdef MAC
2906 	error = mac_vnode_check_setowner(cred, vp, vattr.va_uid,
2907 	    vattr.va_gid);
2908 	if (error == 0)
2909 #endif
2910 		error = VOP_SETATTR(vp, &vattr, cred);
2911 	VOP_UNLOCK(vp, 0);
2912 	vn_finished_write(mp);
2913 	return (error);
2914 }
2915 
2916 /*
2917  * Set ownership given a path name.
2918  */
2919 #ifndef _SYS_SYSPROTO_H_
2920 struct chown_args {
2921 	char	*path;
2922 	int	uid;
2923 	int	gid;
2924 };
2925 #endif
2926 int
2927 sys_chown(td, uap)
2928 	struct thread *td;
2929 	register struct chown_args /* {
2930 		char *path;
2931 		int uid;
2932 		int gid;
2933 	} */ *uap;
2934 {
2935 
2936 	return (kern_chown(td, uap->path, UIO_USERSPACE, uap->uid, uap->gid));
2937 }
2938 
2939 #ifndef _SYS_SYSPROTO_H_
2940 struct fchownat_args {
2941 	int fd;
2942 	const char * path;
2943 	uid_t uid;
2944 	gid_t gid;
2945 	int flag;
2946 };
2947 #endif
2948 int
2949 sys_fchownat(struct thread *td, struct fchownat_args *uap)
2950 {
2951 	int flag;
2952 
2953 	flag = uap->flag;
2954 	if (flag & ~AT_SYMLINK_NOFOLLOW)
2955 		return (EINVAL);
2956 
2957 	return (kern_fchownat(td, uap->fd, uap->path, UIO_USERSPACE, uap->uid,
2958 	    uap->gid, uap->flag));
2959 }
2960 
2961 int
2962 kern_chown(struct thread *td, char *path, enum uio_seg pathseg, int uid,
2963     int gid)
2964 {
2965 
2966 	return (kern_fchownat(td, AT_FDCWD, path, pathseg, uid, gid, 0));
2967 }
2968 
2969 int
2970 kern_fchownat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2971     int uid, int gid, int flag)
2972 {
2973 	struct nameidata nd;
2974 	int error, follow;
2975 
2976 	AUDIT_ARG_OWNER(uid, gid);
2977 	follow = (flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
2978 	NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, pathseg, path, fd,
2979 	    CAP_FCHOWN, td);
2980 
2981 	if ((error = namei(&nd)) != 0)
2982 		return (error);
2983 	NDFREE(&nd, NDF_ONLY_PNBUF);
2984 	error = setfown(td, td->td_ucred, nd.ni_vp, uid, gid);
2985 	vrele(nd.ni_vp);
2986 	return (error);
2987 }
2988 
2989 /*
2990  * Set ownership given a path name, do not cross symlinks.
2991  */
2992 #ifndef _SYS_SYSPROTO_H_
2993 struct lchown_args {
2994 	char	*path;
2995 	int	uid;
2996 	int	gid;
2997 };
2998 #endif
2999 int
3000 sys_lchown(td, uap)
3001 	struct thread *td;
3002 	register struct lchown_args /* {
3003 		char *path;
3004 		int uid;
3005 		int gid;
3006 	} */ *uap;
3007 {
3008 
3009 	return (kern_lchown(td, uap->path, UIO_USERSPACE, uap->uid, uap->gid));
3010 }
3011 
3012 int
3013 kern_lchown(struct thread *td, char *path, enum uio_seg pathseg, int uid,
3014     int gid)
3015 {
3016 
3017 	return (kern_fchownat(td, AT_FDCWD, path, pathseg, uid, gid,
3018 	    AT_SYMLINK_NOFOLLOW));
3019 }
3020 
3021 /*
3022  * Set ownership given a file descriptor.
3023  */
3024 #ifndef _SYS_SYSPROTO_H_
3025 struct fchown_args {
3026 	int	fd;
3027 	int	uid;
3028 	int	gid;
3029 };
3030 #endif
3031 int
3032 sys_fchown(td, uap)
3033 	struct thread *td;
3034 	register struct fchown_args /* {
3035 		int fd;
3036 		int uid;
3037 		int gid;
3038 	} */ *uap;
3039 {
3040 	struct file *fp;
3041 	int error;
3042 
3043 	AUDIT_ARG_FD(uap->fd);
3044 	AUDIT_ARG_OWNER(uap->uid, uap->gid);
3045 	error = fget(td, uap->fd, CAP_FCHOWN, &fp);
3046 	if (error != 0)
3047 		return (error);
3048 	error = fo_chown(fp, uap->uid, uap->gid, td->td_ucred, td);
3049 	fdrop(fp, td);
3050 	return (error);
3051 }
3052 
3053 /*
3054  * Common implementation code for utimes(), lutimes(), and futimes().
3055  */
3056 static int
3057 getutimes(usrtvp, tvpseg, tsp)
3058 	const struct timeval *usrtvp;
3059 	enum uio_seg tvpseg;
3060 	struct timespec *tsp;
3061 {
3062 	struct timeval tv[2];
3063 	const struct timeval *tvp;
3064 	int error;
3065 
3066 	if (usrtvp == NULL) {
3067 		vfs_timestamp(&tsp[0]);
3068 		tsp[1] = tsp[0];
3069 	} else {
3070 		if (tvpseg == UIO_SYSSPACE) {
3071 			tvp = usrtvp;
3072 		} else {
3073 			if ((error = copyin(usrtvp, tv, sizeof(tv))) != 0)
3074 				return (error);
3075 			tvp = tv;
3076 		}
3077 
3078 		if (tvp[0].tv_usec < 0 || tvp[0].tv_usec >= 1000000 ||
3079 		    tvp[1].tv_usec < 0 || tvp[1].tv_usec >= 1000000)
3080 			return (EINVAL);
3081 		TIMEVAL_TO_TIMESPEC(&tvp[0], &tsp[0]);
3082 		TIMEVAL_TO_TIMESPEC(&tvp[1], &tsp[1]);
3083 	}
3084 	return (0);
3085 }
3086 
3087 /*
3088  * Common implementation code for utimes(), lutimes(), and futimes().
3089  */
3090 static int
3091 setutimes(td, vp, ts, numtimes, nullflag)
3092 	struct thread *td;
3093 	struct vnode *vp;
3094 	const struct timespec *ts;
3095 	int numtimes;
3096 	int nullflag;
3097 {
3098 	int error, setbirthtime;
3099 	struct mount *mp;
3100 	struct vattr vattr;
3101 
3102 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
3103 		return (error);
3104 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3105 	setbirthtime = 0;
3106 	if (numtimes < 3 && !VOP_GETATTR(vp, &vattr, td->td_ucred) &&
3107 	    timespeccmp(&ts[1], &vattr.va_birthtime, < ))
3108 		setbirthtime = 1;
3109 	VATTR_NULL(&vattr);
3110 	vattr.va_atime = ts[0];
3111 	vattr.va_mtime = ts[1];
3112 	if (setbirthtime)
3113 		vattr.va_birthtime = ts[1];
3114 	if (numtimes > 2)
3115 		vattr.va_birthtime = ts[2];
3116 	if (nullflag)
3117 		vattr.va_vaflags |= VA_UTIMES_NULL;
3118 #ifdef MAC
3119 	error = mac_vnode_check_setutimes(td->td_ucred, vp, vattr.va_atime,
3120 	    vattr.va_mtime);
3121 #endif
3122 	if (error == 0)
3123 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3124 	VOP_UNLOCK(vp, 0);
3125 	vn_finished_write(mp);
3126 	return (error);
3127 }
3128 
3129 /*
3130  * Set the access and modification times of a file.
3131  */
3132 #ifndef _SYS_SYSPROTO_H_
3133 struct utimes_args {
3134 	char	*path;
3135 	struct	timeval *tptr;
3136 };
3137 #endif
3138 int
3139 sys_utimes(td, uap)
3140 	struct thread *td;
3141 	register struct utimes_args /* {
3142 		char *path;
3143 		struct timeval *tptr;
3144 	} */ *uap;
3145 {
3146 
3147 	return (kern_utimes(td, uap->path, UIO_USERSPACE, uap->tptr,
3148 	    UIO_USERSPACE));
3149 }
3150 
3151 #ifndef _SYS_SYSPROTO_H_
3152 struct futimesat_args {
3153 	int fd;
3154 	const char * path;
3155 	const struct timeval * times;
3156 };
3157 #endif
3158 int
3159 sys_futimesat(struct thread *td, struct futimesat_args *uap)
3160 {
3161 
3162 	return (kern_utimesat(td, uap->fd, uap->path, UIO_USERSPACE,
3163 	    uap->times, UIO_USERSPACE));
3164 }
3165 
3166 int
3167 kern_utimes(struct thread *td, char *path, enum uio_seg pathseg,
3168     struct timeval *tptr, enum uio_seg tptrseg)
3169 {
3170 
3171 	return (kern_utimesat(td, AT_FDCWD, path, pathseg, tptr, tptrseg));
3172 }
3173 
3174 int
3175 kern_utimesat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
3176     struct timeval *tptr, enum uio_seg tptrseg)
3177 {
3178 	struct nameidata nd;
3179 	struct timespec ts[2];
3180 	int error;
3181 
3182 	if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3183 		return (error);
3184 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, fd,
3185 	    CAP_FUTIMES, td);
3186 
3187 	if ((error = namei(&nd)) != 0)
3188 		return (error);
3189 	NDFREE(&nd, NDF_ONLY_PNBUF);
3190 	error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3191 	vrele(nd.ni_vp);
3192 	return (error);
3193 }
3194 
3195 /*
3196  * Set the access and modification times of a file.
3197  */
3198 #ifndef _SYS_SYSPROTO_H_
3199 struct lutimes_args {
3200 	char	*path;
3201 	struct	timeval *tptr;
3202 };
3203 #endif
3204 int
3205 sys_lutimes(td, uap)
3206 	struct thread *td;
3207 	register struct lutimes_args /* {
3208 		char *path;
3209 		struct timeval *tptr;
3210 	} */ *uap;
3211 {
3212 
3213 	return (kern_lutimes(td, uap->path, UIO_USERSPACE, uap->tptr,
3214 	    UIO_USERSPACE));
3215 }
3216 
3217 int
3218 kern_lutimes(struct thread *td, char *path, enum uio_seg pathseg,
3219     struct timeval *tptr, enum uio_seg tptrseg)
3220 {
3221 	struct timespec ts[2];
3222 	int error;
3223 	struct nameidata nd;
3224 
3225 	if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3226 		return (error);
3227 	NDINIT(&nd, LOOKUP, NOFOLLOW | AUDITVNODE1, pathseg, path, td);
3228 	if ((error = namei(&nd)) != 0)
3229 		return (error);
3230 	NDFREE(&nd, NDF_ONLY_PNBUF);
3231 	error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3232 	vrele(nd.ni_vp);
3233 	return (error);
3234 }
3235 
3236 /*
3237  * Set the access and modification times of a file.
3238  */
3239 #ifndef _SYS_SYSPROTO_H_
3240 struct futimes_args {
3241 	int	fd;
3242 	struct	timeval *tptr;
3243 };
3244 #endif
3245 int
3246 sys_futimes(td, uap)
3247 	struct thread *td;
3248 	register struct futimes_args /* {
3249 		int  fd;
3250 		struct timeval *tptr;
3251 	} */ *uap;
3252 {
3253 
3254 	return (kern_futimes(td, uap->fd, uap->tptr, UIO_USERSPACE));
3255 }
3256 
3257 int
3258 kern_futimes(struct thread *td, int fd, struct timeval *tptr,
3259     enum uio_seg tptrseg)
3260 {
3261 	struct timespec ts[2];
3262 	struct file *fp;
3263 	int error;
3264 
3265 	AUDIT_ARG_FD(fd);
3266 	if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3267 		return (error);
3268 	if ((error = getvnode(td->td_proc->p_fd, fd, CAP_FUTIMES, &fp))
3269 	    != 0)
3270 		return (error);
3271 #ifdef AUDIT
3272 	vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
3273 	AUDIT_ARG_VNODE1(fp->f_vnode);
3274 	VOP_UNLOCK(fp->f_vnode, 0);
3275 #endif
3276 	error = setutimes(td, fp->f_vnode, ts, 2, tptr == NULL);
3277 	fdrop(fp, td);
3278 	return (error);
3279 }
3280 
3281 /*
3282  * Truncate a file given its path name.
3283  */
3284 #ifndef _SYS_SYSPROTO_H_
3285 struct truncate_args {
3286 	char	*path;
3287 	int	pad;
3288 	off_t	length;
3289 };
3290 #endif
3291 int
3292 sys_truncate(td, uap)
3293 	struct thread *td;
3294 	register struct truncate_args /* {
3295 		char *path;
3296 		int pad;
3297 		off_t length;
3298 	} */ *uap;
3299 {
3300 
3301 	return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3302 }
3303 
3304 int
3305 kern_truncate(struct thread *td, char *path, enum uio_seg pathseg, off_t length)
3306 {
3307 	struct mount *mp;
3308 	struct vnode *vp;
3309 	void *rl_cookie;
3310 	struct vattr vattr;
3311 	struct nameidata nd;
3312 	int error;
3313 
3314 	if (length < 0)
3315 		return(EINVAL);
3316 	NDINIT(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, td);
3317 	if ((error = namei(&nd)) != 0)
3318 		return (error);
3319 	vp = nd.ni_vp;
3320 	rl_cookie = vn_rangelock_wlock(vp, 0, OFF_MAX);
3321 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) {
3322 		vn_rangelock_unlock(vp, rl_cookie);
3323 		vrele(vp);
3324 		return (error);
3325 	}
3326 	NDFREE(&nd, NDF_ONLY_PNBUF);
3327 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3328 	if (vp->v_type == VDIR)
3329 		error = EISDIR;
3330 #ifdef MAC
3331 	else if ((error = mac_vnode_check_write(td->td_ucred, NOCRED, vp))) {
3332 	}
3333 #endif
3334 	else if ((error = vn_writechk(vp)) == 0 &&
3335 	    (error = VOP_ACCESS(vp, VWRITE, td->td_ucred, td)) == 0) {
3336 		VATTR_NULL(&vattr);
3337 		vattr.va_size = length;
3338 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3339 	}
3340 	VOP_UNLOCK(vp, 0);
3341 	vn_finished_write(mp);
3342 	vn_rangelock_unlock(vp, rl_cookie);
3343 	vrele(vp);
3344 	return (error);
3345 }
3346 
3347 #if defined(COMPAT_43)
3348 /*
3349  * Truncate a file given its path name.
3350  */
3351 #ifndef _SYS_SYSPROTO_H_
3352 struct otruncate_args {
3353 	char	*path;
3354 	long	length;
3355 };
3356 #endif
3357 int
3358 otruncate(td, uap)
3359 	struct thread *td;
3360 	register struct otruncate_args /* {
3361 		char *path;
3362 		long length;
3363 	} */ *uap;
3364 {
3365 	struct truncate_args /* {
3366 		char *path;
3367 		int pad;
3368 		off_t length;
3369 	} */ nuap;
3370 
3371 	nuap.path = uap->path;
3372 	nuap.length = uap->length;
3373 	return (sys_truncate(td, &nuap));
3374 }
3375 #endif /* COMPAT_43 */
3376 
3377 /* Versions with the pad argument */
3378 int
3379 freebsd6_truncate(struct thread *td, struct freebsd6_truncate_args *uap)
3380 {
3381 	struct truncate_args ouap;
3382 
3383 	ouap.path = uap->path;
3384 	ouap.length = uap->length;
3385 	return (sys_truncate(td, &ouap));
3386 }
3387 
3388 int
3389 freebsd6_ftruncate(struct thread *td, struct freebsd6_ftruncate_args *uap)
3390 {
3391 	struct ftruncate_args ouap;
3392 
3393 	ouap.fd = uap->fd;
3394 	ouap.length = uap->length;
3395 	return (sys_ftruncate(td, &ouap));
3396 }
3397 
3398 /*
3399  * Sync an open file.
3400  */
3401 #ifndef _SYS_SYSPROTO_H_
3402 struct fsync_args {
3403 	int	fd;
3404 };
3405 #endif
3406 int
3407 sys_fsync(td, uap)
3408 	struct thread *td;
3409 	struct fsync_args /* {
3410 		int fd;
3411 	} */ *uap;
3412 {
3413 	struct vnode *vp;
3414 	struct mount *mp;
3415 	struct file *fp;
3416 	int error, lock_flags;
3417 
3418 	AUDIT_ARG_FD(uap->fd);
3419 	if ((error = getvnode(td->td_proc->p_fd, uap->fd, CAP_FSYNC,
3420 	    &fp)) != 0)
3421 		return (error);
3422 	vp = fp->f_vnode;
3423 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
3424 		goto drop;
3425 	if (MNT_SHARED_WRITES(mp) ||
3426 	    ((mp == NULL) && MNT_SHARED_WRITES(vp->v_mount))) {
3427 		lock_flags = LK_SHARED;
3428 	} else {
3429 		lock_flags = LK_EXCLUSIVE;
3430 	}
3431 	vn_lock(vp, lock_flags | LK_RETRY);
3432 	AUDIT_ARG_VNODE1(vp);
3433 	if (vp->v_object != NULL) {
3434 		VM_OBJECT_LOCK(vp->v_object);
3435 		vm_object_page_clean(vp->v_object, 0, 0, 0);
3436 		VM_OBJECT_UNLOCK(vp->v_object);
3437 	}
3438 	error = VOP_FSYNC(vp, MNT_WAIT, td);
3439 
3440 	VOP_UNLOCK(vp, 0);
3441 	vn_finished_write(mp);
3442 drop:
3443 	fdrop(fp, td);
3444 	return (error);
3445 }
3446 
3447 /*
3448  * Rename files.  Source and destination must either both be directories, or
3449  * both not be directories.  If target is a directory, it must be empty.
3450  */
3451 #ifndef _SYS_SYSPROTO_H_
3452 struct rename_args {
3453 	char	*from;
3454 	char	*to;
3455 };
3456 #endif
3457 int
3458 sys_rename(td, uap)
3459 	struct thread *td;
3460 	register struct rename_args /* {
3461 		char *from;
3462 		char *to;
3463 	} */ *uap;
3464 {
3465 
3466 	return (kern_rename(td, uap->from, uap->to, UIO_USERSPACE));
3467 }
3468 
3469 #ifndef _SYS_SYSPROTO_H_
3470 struct renameat_args {
3471 	int	oldfd;
3472 	char	*old;
3473 	int	newfd;
3474 	char	*new;
3475 };
3476 #endif
3477 int
3478 sys_renameat(struct thread *td, struct renameat_args *uap)
3479 {
3480 
3481 	return (kern_renameat(td, uap->oldfd, uap->old, uap->newfd, uap->new,
3482 	    UIO_USERSPACE));
3483 }
3484 
3485 int
3486 kern_rename(struct thread *td, char *from, char *to, enum uio_seg pathseg)
3487 {
3488 
3489 	return (kern_renameat(td, AT_FDCWD, from, AT_FDCWD, to, pathseg));
3490 }
3491 
3492 int
3493 kern_renameat(struct thread *td, int oldfd, char *old, int newfd, char *new,
3494     enum uio_seg pathseg)
3495 {
3496 	struct mount *mp = NULL;
3497 	struct vnode *tvp, *fvp, *tdvp;
3498 	struct nameidata fromnd, tond;
3499 	int error;
3500 
3501 	bwillwrite();
3502 #ifdef MAC
3503 	NDINIT_ATRIGHTS(&fromnd, DELETE, LOCKPARENT | LOCKLEAF | SAVESTART |
3504 	    AUDITVNODE1, pathseg, old, oldfd, CAP_DELETE, td);
3505 #else
3506 	NDINIT_ATRIGHTS(&fromnd, DELETE, WANTPARENT | SAVESTART | AUDITVNODE1,
3507 	    pathseg, old, oldfd, CAP_DELETE, td);
3508 #endif
3509 
3510 	if ((error = namei(&fromnd)) != 0)
3511 		return (error);
3512 #ifdef MAC
3513 	error = mac_vnode_check_rename_from(td->td_ucred, fromnd.ni_dvp,
3514 	    fromnd.ni_vp, &fromnd.ni_cnd);
3515 	VOP_UNLOCK(fromnd.ni_dvp, 0);
3516 	if (fromnd.ni_dvp != fromnd.ni_vp)
3517 		VOP_UNLOCK(fromnd.ni_vp, 0);
3518 #endif
3519 	fvp = fromnd.ni_vp;
3520 	if (error == 0)
3521 		error = vn_start_write(fvp, &mp, V_WAIT | PCATCH);
3522 	if (error != 0) {
3523 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3524 		vrele(fromnd.ni_dvp);
3525 		vrele(fvp);
3526 		goto out1;
3527 	}
3528 	NDINIT_ATRIGHTS(&tond, RENAME, LOCKPARENT | LOCKLEAF | NOCACHE |
3529 	    SAVESTART | AUDITVNODE2, pathseg, new, newfd, CAP_CREATE,
3530 	    td);
3531 	if (fromnd.ni_vp->v_type == VDIR)
3532 		tond.ni_cnd.cn_flags |= WILLBEDIR;
3533 	if ((error = namei(&tond)) != 0) {
3534 		/* Translate error code for rename("dir1", "dir2/."). */
3535 		if (error == EISDIR && fvp->v_type == VDIR)
3536 			error = EINVAL;
3537 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3538 		vrele(fromnd.ni_dvp);
3539 		vrele(fvp);
3540 		vn_finished_write(mp);
3541 		goto out1;
3542 	}
3543 	tdvp = tond.ni_dvp;
3544 	tvp = tond.ni_vp;
3545 	if (tvp != NULL) {
3546 		if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
3547 			error = ENOTDIR;
3548 			goto out;
3549 		} else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
3550 			error = EISDIR;
3551 			goto out;
3552 		}
3553 	}
3554 	if (fvp == tdvp) {
3555 		error = EINVAL;
3556 		goto out;
3557 	}
3558 	/*
3559 	 * If the source is the same as the destination (that is, if they
3560 	 * are links to the same vnode), then there is nothing to do.
3561 	 */
3562 	if (fvp == tvp)
3563 		error = -1;
3564 #ifdef MAC
3565 	else
3566 		error = mac_vnode_check_rename_to(td->td_ucred, tdvp,
3567 		    tond.ni_vp, fromnd.ni_dvp == tdvp, &tond.ni_cnd);
3568 #endif
3569 out:
3570 	if (!error) {
3571 		error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
3572 				   tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
3573 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3574 		NDFREE(&tond, NDF_ONLY_PNBUF);
3575 	} else {
3576 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3577 		NDFREE(&tond, NDF_ONLY_PNBUF);
3578 		if (tvp)
3579 			vput(tvp);
3580 		if (tdvp == tvp)
3581 			vrele(tdvp);
3582 		else
3583 			vput(tdvp);
3584 		vrele(fromnd.ni_dvp);
3585 		vrele(fvp);
3586 	}
3587 	vrele(tond.ni_startdir);
3588 	vn_finished_write(mp);
3589 out1:
3590 	if (fromnd.ni_startdir)
3591 		vrele(fromnd.ni_startdir);
3592 	if (error == -1)
3593 		return (0);
3594 	return (error);
3595 }
3596 
3597 /*
3598  * Make a directory file.
3599  */
3600 #ifndef _SYS_SYSPROTO_H_
3601 struct mkdir_args {
3602 	char	*path;
3603 	int	mode;
3604 };
3605 #endif
3606 int
3607 sys_mkdir(td, uap)
3608 	struct thread *td;
3609 	register struct mkdir_args /* {
3610 		char *path;
3611 		int mode;
3612 	} */ *uap;
3613 {
3614 
3615 	return (kern_mkdir(td, uap->path, UIO_USERSPACE, uap->mode));
3616 }
3617 
3618 #ifndef _SYS_SYSPROTO_H_
3619 struct mkdirat_args {
3620 	int	fd;
3621 	char	*path;
3622 	mode_t	mode;
3623 };
3624 #endif
3625 int
3626 sys_mkdirat(struct thread *td, struct mkdirat_args *uap)
3627 {
3628 
3629 	return (kern_mkdirat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode));
3630 }
3631 
3632 int
3633 kern_mkdir(struct thread *td, char *path, enum uio_seg segflg, int mode)
3634 {
3635 
3636 	return (kern_mkdirat(td, AT_FDCWD, path, segflg, mode));
3637 }
3638 
3639 int
3640 kern_mkdirat(struct thread *td, int fd, char *path, enum uio_seg segflg,
3641     int mode)
3642 {
3643 	struct mount *mp;
3644 	struct vnode *vp;
3645 	struct vattr vattr;
3646 	int error;
3647 	struct nameidata nd;
3648 
3649 	AUDIT_ARG_MODE(mode);
3650 restart:
3651 	bwillwrite();
3652 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1,
3653 	    segflg, path, fd, CAP_MKDIR, td);
3654 	nd.ni_cnd.cn_flags |= WILLBEDIR;
3655 	if ((error = namei(&nd)) != 0)
3656 		return (error);
3657 	vp = nd.ni_vp;
3658 	if (vp != NULL) {
3659 		NDFREE(&nd, NDF_ONLY_PNBUF);
3660 		/*
3661 		 * XXX namei called with LOCKPARENT but not LOCKLEAF has
3662 		 * the strange behaviour of leaving the vnode unlocked
3663 		 * if the target is the same vnode as the parent.
3664 		 */
3665 		if (vp == nd.ni_dvp)
3666 			vrele(nd.ni_dvp);
3667 		else
3668 			vput(nd.ni_dvp);
3669 		vrele(vp);
3670 		return (EEXIST);
3671 	}
3672 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3673 		NDFREE(&nd, NDF_ONLY_PNBUF);
3674 		vput(nd.ni_dvp);
3675 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3676 			return (error);
3677 		goto restart;
3678 	}
3679 	VATTR_NULL(&vattr);
3680 	vattr.va_type = VDIR;
3681 	vattr.va_mode = (mode & ACCESSPERMS) &~ td->td_proc->p_fd->fd_cmask;
3682 #ifdef MAC
3683 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
3684 	    &vattr);
3685 	if (error)
3686 		goto out;
3687 #endif
3688 	error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
3689 #ifdef MAC
3690 out:
3691 #endif
3692 	NDFREE(&nd, NDF_ONLY_PNBUF);
3693 	vput(nd.ni_dvp);
3694 	if (!error)
3695 		vput(nd.ni_vp);
3696 	vn_finished_write(mp);
3697 	return (error);
3698 }
3699 
3700 /*
3701  * Remove a directory file.
3702  */
3703 #ifndef _SYS_SYSPROTO_H_
3704 struct rmdir_args {
3705 	char	*path;
3706 };
3707 #endif
3708 int
3709 sys_rmdir(td, uap)
3710 	struct thread *td;
3711 	struct rmdir_args /* {
3712 		char *path;
3713 	} */ *uap;
3714 {
3715 
3716 	return (kern_rmdir(td, uap->path, UIO_USERSPACE));
3717 }
3718 
3719 int
3720 kern_rmdir(struct thread *td, char *path, enum uio_seg pathseg)
3721 {
3722 
3723 	return (kern_rmdirat(td, AT_FDCWD, path, pathseg));
3724 }
3725 
3726 int
3727 kern_rmdirat(struct thread *td, int fd, char *path, enum uio_seg pathseg)
3728 {
3729 	struct mount *mp;
3730 	struct vnode *vp;
3731 	int error;
3732 	struct nameidata nd;
3733 
3734 restart:
3735 	bwillwrite();
3736 	NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1,
3737 	    pathseg, path, fd, CAP_RMDIR, td);
3738 	if ((error = namei(&nd)) != 0)
3739 		return (error);
3740 	vp = nd.ni_vp;
3741 	if (vp->v_type != VDIR) {
3742 		error = ENOTDIR;
3743 		goto out;
3744 	}
3745 	/*
3746 	 * No rmdir "." please.
3747 	 */
3748 	if (nd.ni_dvp == vp) {
3749 		error = EINVAL;
3750 		goto out;
3751 	}
3752 	/*
3753 	 * The root of a mounted filesystem cannot be deleted.
3754 	 */
3755 	if (vp->v_vflag & VV_ROOT) {
3756 		error = EBUSY;
3757 		goto out;
3758 	}
3759 #ifdef MAC
3760 	error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
3761 	    &nd.ni_cnd);
3762 	if (error)
3763 		goto out;
3764 #endif
3765 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3766 		NDFREE(&nd, NDF_ONLY_PNBUF);
3767 		vput(vp);
3768 		if (nd.ni_dvp == vp)
3769 			vrele(nd.ni_dvp);
3770 		else
3771 			vput(nd.ni_dvp);
3772 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3773 			return (error);
3774 		goto restart;
3775 	}
3776 	error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
3777 	vn_finished_write(mp);
3778 out:
3779 	NDFREE(&nd, NDF_ONLY_PNBUF);
3780 	vput(vp);
3781 	if (nd.ni_dvp == vp)
3782 		vrele(nd.ni_dvp);
3783 	else
3784 		vput(nd.ni_dvp);
3785 	return (error);
3786 }
3787 
3788 #ifdef COMPAT_43
3789 /*
3790  * Read a block of directory entries in a filesystem independent format.
3791  */
3792 #ifndef _SYS_SYSPROTO_H_
3793 struct ogetdirentries_args {
3794 	int	fd;
3795 	char	*buf;
3796 	u_int	count;
3797 	long	*basep;
3798 };
3799 #endif
3800 int
3801 ogetdirentries(struct thread *td, struct ogetdirentries_args *uap)
3802 {
3803 	long loff;
3804 	int error;
3805 
3806 	error = kern_ogetdirentries(td, uap, &loff);
3807 	if (error == 0)
3808 		error = copyout(&loff, uap->basep, sizeof(long));
3809 	return (error);
3810 }
3811 
3812 int
3813 kern_ogetdirentries(struct thread *td, struct ogetdirentries_args *uap,
3814     long *ploff)
3815 {
3816 	struct vnode *vp;
3817 	struct file *fp;
3818 	struct uio auio, kuio;
3819 	struct iovec aiov, kiov;
3820 	struct dirent *dp, *edp;
3821 	caddr_t dirbuf;
3822 	int error, eofflag, readcnt;
3823 	long loff;
3824 	off_t foffset;
3825 
3826 	/* XXX arbitrary sanity limit on `count'. */
3827 	if (uap->count > 64 * 1024)
3828 		return (EINVAL);
3829 	if ((error = getvnode(td->td_proc->p_fd, uap->fd, CAP_READ,
3830 	    &fp)) != 0)
3831 		return (error);
3832 	if ((fp->f_flag & FREAD) == 0) {
3833 		fdrop(fp, td);
3834 		return (EBADF);
3835 	}
3836 	vp = fp->f_vnode;
3837 	foffset = foffset_lock(fp, 0);
3838 unionread:
3839 	if (vp->v_type != VDIR) {
3840 		foffset_unlock(fp, foffset, 0);
3841 		fdrop(fp, td);
3842 		return (EINVAL);
3843 	}
3844 	aiov.iov_base = uap->buf;
3845 	aiov.iov_len = uap->count;
3846 	auio.uio_iov = &aiov;
3847 	auio.uio_iovcnt = 1;
3848 	auio.uio_rw = UIO_READ;
3849 	auio.uio_segflg = UIO_USERSPACE;
3850 	auio.uio_td = td;
3851 	auio.uio_resid = uap->count;
3852 	vn_lock(vp, LK_SHARED | LK_RETRY);
3853 	loff = auio.uio_offset = foffset;
3854 #ifdef MAC
3855 	error = mac_vnode_check_readdir(td->td_ucred, vp);
3856 	if (error) {
3857 		VOP_UNLOCK(vp, 0);
3858 		foffset_unlock(fp, foffset, FOF_NOUPDATE);
3859 		fdrop(fp, td);
3860 		return (error);
3861 	}
3862 #endif
3863 #	if (BYTE_ORDER != LITTLE_ENDIAN)
3864 		if (vp->v_mount->mnt_maxsymlinklen <= 0) {
3865 			error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag,
3866 			    NULL, NULL);
3867 			foffset = auio.uio_offset;
3868 		} else
3869 #	endif
3870 	{
3871 		kuio = auio;
3872 		kuio.uio_iov = &kiov;
3873 		kuio.uio_segflg = UIO_SYSSPACE;
3874 		kiov.iov_len = uap->count;
3875 		dirbuf = malloc(uap->count, M_TEMP, M_WAITOK);
3876 		kiov.iov_base = dirbuf;
3877 		error = VOP_READDIR(vp, &kuio, fp->f_cred, &eofflag,
3878 			    NULL, NULL);
3879 		foffset = kuio.uio_offset;
3880 		if (error == 0) {
3881 			readcnt = uap->count - kuio.uio_resid;
3882 			edp = (struct dirent *)&dirbuf[readcnt];
3883 			for (dp = (struct dirent *)dirbuf; dp < edp; ) {
3884 #				if (BYTE_ORDER == LITTLE_ENDIAN)
3885 					/*
3886 					 * The expected low byte of
3887 					 * dp->d_namlen is our dp->d_type.
3888 					 * The high MBZ byte of dp->d_namlen
3889 					 * is our dp->d_namlen.
3890 					 */
3891 					dp->d_type = dp->d_namlen;
3892 					dp->d_namlen = 0;
3893 #				else
3894 					/*
3895 					 * The dp->d_type is the high byte
3896 					 * of the expected dp->d_namlen,
3897 					 * so must be zero'ed.
3898 					 */
3899 					dp->d_type = 0;
3900 #				endif
3901 				if (dp->d_reclen > 0) {
3902 					dp = (struct dirent *)
3903 					    ((char *)dp + dp->d_reclen);
3904 				} else {
3905 					error = EIO;
3906 					break;
3907 				}
3908 			}
3909 			if (dp >= edp)
3910 				error = uiomove(dirbuf, readcnt, &auio);
3911 		}
3912 		free(dirbuf, M_TEMP);
3913 	}
3914 	if (error) {
3915 		VOP_UNLOCK(vp, 0);
3916 		foffset_unlock(fp, foffset, 0);
3917 		fdrop(fp, td);
3918 		return (error);
3919 	}
3920 	if (uap->count == auio.uio_resid &&
3921 	    (vp->v_vflag & VV_ROOT) &&
3922 	    (vp->v_mount->mnt_flag & MNT_UNION)) {
3923 		struct vnode *tvp = vp;
3924 		vp = vp->v_mount->mnt_vnodecovered;
3925 		VREF(vp);
3926 		fp->f_vnode = vp;
3927 		fp->f_data = vp;
3928 		foffset = 0;
3929 		vput(tvp);
3930 		goto unionread;
3931 	}
3932 	VOP_UNLOCK(vp, 0);
3933 	foffset_unlock(fp, foffset, 0);
3934 	fdrop(fp, td);
3935 	td->td_retval[0] = uap->count - auio.uio_resid;
3936 	if (error == 0)
3937 		*ploff = loff;
3938 	return (error);
3939 }
3940 #endif /* COMPAT_43 */
3941 
3942 /*
3943  * Read a block of directory entries in a filesystem independent format.
3944  */
3945 #ifndef _SYS_SYSPROTO_H_
3946 struct getdirentries_args {
3947 	int	fd;
3948 	char	*buf;
3949 	u_int	count;
3950 	long	*basep;
3951 };
3952 #endif
3953 int
3954 sys_getdirentries(td, uap)
3955 	struct thread *td;
3956 	register struct getdirentries_args /* {
3957 		int fd;
3958 		char *buf;
3959 		u_int count;
3960 		long *basep;
3961 	} */ *uap;
3962 {
3963 	long base;
3964 	int error;
3965 
3966 	error = kern_getdirentries(td, uap->fd, uap->buf, uap->count, &base,
3967 	    NULL, UIO_USERSPACE);
3968 	if (error)
3969 		return (error);
3970 	if (uap->basep != NULL)
3971 		error = copyout(&base, uap->basep, sizeof(long));
3972 	return (error);
3973 }
3974 
3975 int
3976 kern_getdirentries(struct thread *td, int fd, char *buf, u_int count,
3977     long *basep, ssize_t *residp, enum uio_seg bufseg)
3978 {
3979 	struct vnode *vp;
3980 	struct file *fp;
3981 	struct uio auio;
3982 	struct iovec aiov;
3983 	long loff;
3984 	int error, eofflag;
3985 	off_t foffset;
3986 
3987 	AUDIT_ARG_FD(fd);
3988 	if (count > IOSIZE_MAX)
3989 		return (EINVAL);
3990 	auio.uio_resid = count;
3991 	if ((error = getvnode(td->td_proc->p_fd, fd, CAP_READ | CAP_SEEK,
3992 	    &fp)) != 0)
3993 		return (error);
3994 	if ((fp->f_flag & FREAD) == 0) {
3995 		fdrop(fp, td);
3996 		return (EBADF);
3997 	}
3998 	vp = fp->f_vnode;
3999 	foffset = foffset_lock(fp, 0);
4000 unionread:
4001 	if (vp->v_type != VDIR) {
4002 		error = EINVAL;
4003 		goto fail;
4004 	}
4005 	aiov.iov_base = buf;
4006 	aiov.iov_len = count;
4007 	auio.uio_iov = &aiov;
4008 	auio.uio_iovcnt = 1;
4009 	auio.uio_rw = UIO_READ;
4010 	auio.uio_segflg = bufseg;
4011 	auio.uio_td = td;
4012 	vn_lock(vp, LK_SHARED | LK_RETRY);
4013 	AUDIT_ARG_VNODE1(vp);
4014 	loff = auio.uio_offset = foffset;
4015 #ifdef MAC
4016 	error = mac_vnode_check_readdir(td->td_ucred, vp);
4017 	if (error == 0)
4018 #endif
4019 		error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, NULL,
4020 		    NULL);
4021 	foffset = auio.uio_offset;
4022 	if (error) {
4023 		VOP_UNLOCK(vp, 0);
4024 		goto fail;
4025 	}
4026 	if (count == auio.uio_resid &&
4027 	    (vp->v_vflag & VV_ROOT) &&
4028 	    (vp->v_mount->mnt_flag & MNT_UNION)) {
4029 		struct vnode *tvp = vp;
4030 		vp = vp->v_mount->mnt_vnodecovered;
4031 		VREF(vp);
4032 		fp->f_vnode = vp;
4033 		fp->f_data = vp;
4034 		foffset = 0;
4035 		vput(tvp);
4036 		goto unionread;
4037 	}
4038 	VOP_UNLOCK(vp, 0);
4039 	*basep = loff;
4040 	if (residp != NULL)
4041 		*residp = auio.uio_resid;
4042 	td->td_retval[0] = count - auio.uio_resid;
4043 fail:
4044 	foffset_unlock(fp, foffset, 0);
4045 	fdrop(fp, td);
4046 	return (error);
4047 }
4048 
4049 #ifndef _SYS_SYSPROTO_H_
4050 struct getdents_args {
4051 	int fd;
4052 	char *buf;
4053 	size_t count;
4054 };
4055 #endif
4056 int
4057 sys_getdents(td, uap)
4058 	struct thread *td;
4059 	register struct getdents_args /* {
4060 		int fd;
4061 		char *buf;
4062 		u_int count;
4063 	} */ *uap;
4064 {
4065 	struct getdirentries_args ap;
4066 	ap.fd = uap->fd;
4067 	ap.buf = uap->buf;
4068 	ap.count = uap->count;
4069 	ap.basep = NULL;
4070 	return (sys_getdirentries(td, &ap));
4071 }
4072 
4073 /*
4074  * Set the mode mask for creation of filesystem nodes.
4075  */
4076 #ifndef _SYS_SYSPROTO_H_
4077 struct umask_args {
4078 	int	newmask;
4079 };
4080 #endif
4081 int
4082 sys_umask(td, uap)
4083 	struct thread *td;
4084 	struct umask_args /* {
4085 		int newmask;
4086 	} */ *uap;
4087 {
4088 	register struct filedesc *fdp;
4089 
4090 	FILEDESC_XLOCK(td->td_proc->p_fd);
4091 	fdp = td->td_proc->p_fd;
4092 	td->td_retval[0] = fdp->fd_cmask;
4093 	fdp->fd_cmask = uap->newmask & ALLPERMS;
4094 	FILEDESC_XUNLOCK(td->td_proc->p_fd);
4095 	return (0);
4096 }
4097 
4098 /*
4099  * Void all references to file by ripping underlying filesystem away from
4100  * vnode.
4101  */
4102 #ifndef _SYS_SYSPROTO_H_
4103 struct revoke_args {
4104 	char	*path;
4105 };
4106 #endif
4107 int
4108 sys_revoke(td, uap)
4109 	struct thread *td;
4110 	register struct revoke_args /* {
4111 		char *path;
4112 	} */ *uap;
4113 {
4114 	struct vnode *vp;
4115 	struct vattr vattr;
4116 	int error;
4117 	struct nameidata nd;
4118 
4119 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
4120 	    uap->path, td);
4121 	if ((error = namei(&nd)) != 0)
4122 		return (error);
4123 	vp = nd.ni_vp;
4124 	NDFREE(&nd, NDF_ONLY_PNBUF);
4125 	if (vp->v_type != VCHR || vp->v_rdev == NULL) {
4126 		error = EINVAL;
4127 		goto out;
4128 	}
4129 #ifdef MAC
4130 	error = mac_vnode_check_revoke(td->td_ucred, vp);
4131 	if (error)
4132 		goto out;
4133 #endif
4134 	error = VOP_GETATTR(vp, &vattr, td->td_ucred);
4135 	if (error)
4136 		goto out;
4137 	if (td->td_ucred->cr_uid != vattr.va_uid) {
4138 		error = priv_check(td, PRIV_VFS_ADMIN);
4139 		if (error)
4140 			goto out;
4141 	}
4142 	if (vcount(vp) > 1)
4143 		VOP_REVOKE(vp, REVOKEALL);
4144 out:
4145 	vput(vp);
4146 	return (error);
4147 }
4148 
4149 /*
4150  * Convert a user file descriptor to a kernel file entry and check that, if it
4151  * is a capability, the correct rights are present. A reference on the file
4152  * entry is held upon returning.
4153  */
4154 int
4155 getvnode(struct filedesc *fdp, int fd, cap_rights_t rights,
4156     struct file **fpp)
4157 {
4158 	struct file *fp;
4159 #ifdef CAPABILITIES
4160 	struct file *fp_fromcap;
4161 	int error;
4162 #endif
4163 
4164 	if (fdp == NULL || (fp = fget_unlocked(fdp, fd)) == NULL)
4165 		return (EBADF);
4166 #ifdef CAPABILITIES
4167 	/*
4168 	 * If the file descriptor is for a capability, test rights and use the
4169 	 * file descriptor referenced by the capability.
4170 	 */
4171 	error = cap_funwrap(fp, rights, &fp_fromcap);
4172 	if (error) {
4173 		fdrop(fp, curthread);
4174 		return (error);
4175 	}
4176 	if (fp != fp_fromcap) {
4177 		fhold(fp_fromcap);
4178 		fdrop(fp, curthread);
4179 		fp = fp_fromcap;
4180 	}
4181 #endif /* CAPABILITIES */
4182 
4183 	/*
4184 	 * The file could be not of the vnode type, or it may be not
4185 	 * yet fully initialized, in which case the f_vnode pointer
4186 	 * may be set, but f_ops is still badfileops.  E.g.,
4187 	 * devfs_open() transiently create such situation to
4188 	 * facilitate csw d_fdopen().
4189 	 *
4190 	 * Dupfdopen() handling in kern_openat() installs the
4191 	 * half-baked file into the process descriptor table, allowing
4192 	 * other thread to dereference it. Guard against the race by
4193 	 * checking f_ops.
4194 	 */
4195 	if (fp->f_vnode == NULL || fp->f_ops == &badfileops) {
4196 		fdrop(fp, curthread);
4197 		return (EINVAL);
4198 	}
4199 	*fpp = fp;
4200 	return (0);
4201 }
4202 
4203 
4204 /*
4205  * Get an (NFS) file handle.
4206  */
4207 #ifndef _SYS_SYSPROTO_H_
4208 struct lgetfh_args {
4209 	char	*fname;
4210 	fhandle_t *fhp;
4211 };
4212 #endif
4213 int
4214 sys_lgetfh(td, uap)
4215 	struct thread *td;
4216 	register struct lgetfh_args *uap;
4217 {
4218 	struct nameidata nd;
4219 	fhandle_t fh;
4220 	register struct vnode *vp;
4221 	int error;
4222 
4223 	error = priv_check(td, PRIV_VFS_GETFH);
4224 	if (error)
4225 		return (error);
4226 	NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
4227 	    uap->fname, td);
4228 	error = namei(&nd);
4229 	if (error)
4230 		return (error);
4231 	NDFREE(&nd, NDF_ONLY_PNBUF);
4232 	vp = nd.ni_vp;
4233 	bzero(&fh, sizeof(fh));
4234 	fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
4235 	error = VOP_VPTOFH(vp, &fh.fh_fid);
4236 	vput(vp);
4237 	if (error)
4238 		return (error);
4239 	error = copyout(&fh, uap->fhp, sizeof (fh));
4240 	return (error);
4241 }
4242 
4243 #ifndef _SYS_SYSPROTO_H_
4244 struct getfh_args {
4245 	char	*fname;
4246 	fhandle_t *fhp;
4247 };
4248 #endif
4249 int
4250 sys_getfh(td, uap)
4251 	struct thread *td;
4252 	register struct getfh_args *uap;
4253 {
4254 	struct nameidata nd;
4255 	fhandle_t fh;
4256 	register struct vnode *vp;
4257 	int error;
4258 
4259 	error = priv_check(td, PRIV_VFS_GETFH);
4260 	if (error)
4261 		return (error);
4262 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
4263 	    uap->fname, td);
4264 	error = namei(&nd);
4265 	if (error)
4266 		return (error);
4267 	NDFREE(&nd, NDF_ONLY_PNBUF);
4268 	vp = nd.ni_vp;
4269 	bzero(&fh, sizeof(fh));
4270 	fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
4271 	error = VOP_VPTOFH(vp, &fh.fh_fid);
4272 	vput(vp);
4273 	if (error)
4274 		return (error);
4275 	error = copyout(&fh, uap->fhp, sizeof (fh));
4276 	return (error);
4277 }
4278 
4279 /*
4280  * syscall for the rpc.lockd to use to translate a NFS file handle into an
4281  * open descriptor.
4282  *
4283  * warning: do not remove the priv_check() call or this becomes one giant
4284  * security hole.
4285  */
4286 #ifndef _SYS_SYSPROTO_H_
4287 struct fhopen_args {
4288 	const struct fhandle *u_fhp;
4289 	int flags;
4290 };
4291 #endif
4292 int
4293 sys_fhopen(td, uap)
4294 	struct thread *td;
4295 	struct fhopen_args /* {
4296 		const struct fhandle *u_fhp;
4297 		int flags;
4298 	} */ *uap;
4299 {
4300 	struct mount *mp;
4301 	struct vnode *vp;
4302 	struct fhandle fhp;
4303 	struct file *fp;
4304 	int fmode, error;
4305 	int indx;
4306 
4307 	error = priv_check(td, PRIV_VFS_FHOPEN);
4308 	if (error)
4309 		return (error);
4310 	indx = -1;
4311 	fmode = FFLAGS(uap->flags);
4312 	/* why not allow a non-read/write open for our lockd? */
4313 	if (((fmode & (FREAD | FWRITE)) == 0) || (fmode & O_CREAT))
4314 		return (EINVAL);
4315 	error = copyin(uap->u_fhp, &fhp, sizeof(fhp));
4316 	if (error)
4317 		return(error);
4318 	/* find the mount point */
4319 	mp = vfs_busyfs(&fhp.fh_fsid);
4320 	if (mp == NULL)
4321 		return (ESTALE);
4322 	/* now give me my vnode, it gets returned to me locked */
4323 	error = VFS_FHTOVP(mp, &fhp.fh_fid, LK_EXCLUSIVE, &vp);
4324 	vfs_unbusy(mp);
4325 	if (error)
4326 		return (error);
4327 
4328 	error = falloc_noinstall(td, &fp);
4329 	if (error) {
4330 		vput(vp);
4331 		return (error);
4332 	}
4333 	/*
4334 	 * An extra reference on `fp' has been held for us by
4335 	 * falloc_noinstall().
4336 	 */
4337 
4338 #ifdef INVARIANTS
4339 	td->td_dupfd = -1;
4340 #endif
4341 	error = vn_open_vnode(vp, fmode, td->td_ucred, td, fp);
4342 	if (error) {
4343 		KASSERT(fp->f_ops == &badfileops,
4344 		    ("VOP_OPEN in fhopen() set f_ops"));
4345 		KASSERT(td->td_dupfd < 0,
4346 		    ("fhopen() encountered fdopen()"));
4347 
4348 		vput(vp);
4349 		goto bad;
4350 	}
4351 #ifdef INVARIANTS
4352 	td->td_dupfd = 0;
4353 #endif
4354 	fp->f_vnode = vp;
4355 	fp->f_seqcount = 1;
4356 	finit(fp, (fmode & FMASK) | (fp->f_flag & FHASLOCK), DTYPE_VNODE, vp,
4357 	    &vnops);
4358 	VOP_UNLOCK(vp, 0);
4359 	if (fmode & O_TRUNC) {
4360 		error = fo_truncate(fp, 0, td->td_ucred, td);
4361 		if (error)
4362 			goto bad;
4363 	}
4364 
4365 	error = finstall(td, fp, &indx, fmode);
4366 bad:
4367 	fdrop(fp, td);
4368 	td->td_retval[0] = indx;
4369 	return (error);
4370 }
4371 
4372 /*
4373  * Stat an (NFS) file handle.
4374  */
4375 #ifndef _SYS_SYSPROTO_H_
4376 struct fhstat_args {
4377 	struct fhandle *u_fhp;
4378 	struct stat *sb;
4379 };
4380 #endif
4381 int
4382 sys_fhstat(td, uap)
4383 	struct thread *td;
4384 	register struct fhstat_args /* {
4385 		struct fhandle *u_fhp;
4386 		struct stat *sb;
4387 	} */ *uap;
4388 {
4389 	struct stat sb;
4390 	struct fhandle fh;
4391 	int error;
4392 
4393 	error = copyin(uap->u_fhp, &fh, sizeof(fh));
4394 	if (error != 0)
4395 		return (error);
4396 	error = kern_fhstat(td, fh, &sb);
4397 	if (error != 0)
4398 		return (error);
4399 	error = copyout(&sb, uap->sb, sizeof(sb));
4400 	return (error);
4401 }
4402 
4403 int
4404 kern_fhstat(struct thread *td, struct fhandle fh, struct stat *sb)
4405 {
4406 	struct mount *mp;
4407 	struct vnode *vp;
4408 	int error;
4409 
4410 	error = priv_check(td, PRIV_VFS_FHSTAT);
4411 	if (error)
4412 		return (error);
4413 	if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4414 		return (ESTALE);
4415 	error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4416 	vfs_unbusy(mp);
4417 	if (error)
4418 		return (error);
4419 	error = vn_stat(vp, sb, td->td_ucred, NOCRED, td);
4420 	vput(vp);
4421 	return (error);
4422 }
4423 
4424 /*
4425  * Implement fstatfs() for (NFS) file handles.
4426  */
4427 #ifndef _SYS_SYSPROTO_H_
4428 struct fhstatfs_args {
4429 	struct fhandle *u_fhp;
4430 	struct statfs *buf;
4431 };
4432 #endif
4433 int
4434 sys_fhstatfs(td, uap)
4435 	struct thread *td;
4436 	struct fhstatfs_args /* {
4437 		struct fhandle *u_fhp;
4438 		struct statfs *buf;
4439 	} */ *uap;
4440 {
4441 	struct statfs sf;
4442 	fhandle_t fh;
4443 	int error;
4444 
4445 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
4446 	if (error)
4447 		return (error);
4448 	error = kern_fhstatfs(td, fh, &sf);
4449 	if (error)
4450 		return (error);
4451 	return (copyout(&sf, uap->buf, sizeof(sf)));
4452 }
4453 
4454 int
4455 kern_fhstatfs(struct thread *td, fhandle_t fh, struct statfs *buf)
4456 {
4457 	struct statfs *sp;
4458 	struct mount *mp;
4459 	struct vnode *vp;
4460 	int error;
4461 
4462 	error = priv_check(td, PRIV_VFS_FHSTATFS);
4463 	if (error)
4464 		return (error);
4465 	if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4466 		return (ESTALE);
4467 	error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4468 	if (error) {
4469 		vfs_unbusy(mp);
4470 		return (error);
4471 	}
4472 	vput(vp);
4473 	error = prison_canseemount(td->td_ucred, mp);
4474 	if (error)
4475 		goto out;
4476 #ifdef MAC
4477 	error = mac_mount_check_stat(td->td_ucred, mp);
4478 	if (error)
4479 		goto out;
4480 #endif
4481 	/*
4482 	 * Set these in case the underlying filesystem fails to do so.
4483 	 */
4484 	sp = &mp->mnt_stat;
4485 	sp->f_version = STATFS_VERSION;
4486 	sp->f_namemax = NAME_MAX;
4487 	sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
4488 	error = VFS_STATFS(mp, sp);
4489 	if (error == 0)
4490 		*buf = *sp;
4491 out:
4492 	vfs_unbusy(mp);
4493 	return (error);
4494 }
4495 
4496 int
4497 kern_posix_fallocate(struct thread *td, int fd, off_t offset, off_t len)
4498 {
4499 	struct file *fp;
4500 	struct mount *mp;
4501 	struct vnode *vp;
4502 	off_t olen, ooffset;
4503 	int error;
4504 
4505 	fp = NULL;
4506 	error = fget(td, fd, CAP_WRITE, &fp);
4507 	if (error != 0)
4508 		goto out;
4509 
4510 	switch (fp->f_type) {
4511 	case DTYPE_VNODE:
4512 		break;
4513 	case DTYPE_PIPE:
4514 	case DTYPE_FIFO:
4515 		error = ESPIPE;
4516 		goto out;
4517 	default:
4518 		error = ENODEV;
4519 		goto out;
4520 	}
4521 	if ((fp->f_flag & FWRITE) == 0) {
4522 		error = EBADF;
4523 		goto out;
4524 	}
4525 	vp = fp->f_vnode;
4526 	if (vp->v_type != VREG) {
4527 		error = ENODEV;
4528 		goto out;
4529 	}
4530 	if (offset < 0 || len <= 0) {
4531 		error = EINVAL;
4532 		goto out;
4533 	}
4534 	/* Check for wrap. */
4535 	if (offset > OFF_MAX - len) {
4536 		error = EFBIG;
4537 		goto out;
4538 	}
4539 
4540 	/* Allocating blocks may take a long time, so iterate. */
4541 	for (;;) {
4542 		olen = len;
4543 		ooffset = offset;
4544 
4545 		bwillwrite();
4546 		mp = NULL;
4547 		error = vn_start_write(vp, &mp, V_WAIT | PCATCH);
4548 		if (error != 0)
4549 			break;
4550 		error = vn_lock(vp, LK_EXCLUSIVE);
4551 		if (error != 0) {
4552 			vn_finished_write(mp);
4553 			break;
4554 		}
4555 #ifdef MAC
4556 		error = mac_vnode_check_write(td->td_ucred, fp->f_cred, vp);
4557 		if (error == 0)
4558 #endif
4559 			error = VOP_ALLOCATE(vp, &offset, &len);
4560 		VOP_UNLOCK(vp, 0);
4561 		vn_finished_write(mp);
4562 
4563 		if (olen + ooffset != offset + len) {
4564 			panic("offset + len changed from %jx/%jx to %jx/%jx",
4565 			    ooffset, olen, offset, len);
4566 		}
4567 		if (error != 0 || len == 0)
4568 			break;
4569 		KASSERT(olen > len, ("Iteration did not make progress?"));
4570 		maybe_yield();
4571 	}
4572  out:
4573 	if (fp != NULL)
4574 		fdrop(fp, td);
4575 	return (error);
4576 }
4577 
4578 int
4579 sys_posix_fallocate(struct thread *td, struct posix_fallocate_args *uap)
4580 {
4581 
4582 	return (kern_posix_fallocate(td, uap->fd, uap->offset, uap->len));
4583 }
4584 
4585 /*
4586  * Unlike madvise(2), we do not make a best effort to remember every
4587  * possible caching hint.  Instead, we remember the last setting with
4588  * the exception that we will allow POSIX_FADV_NORMAL to adjust the
4589  * region of any current setting.
4590  */
4591 int
4592 kern_posix_fadvise(struct thread *td, int fd, off_t offset, off_t len,
4593     int advice)
4594 {
4595 	struct fadvise_info *fa, *new;
4596 	struct file *fp;
4597 	struct vnode *vp;
4598 	off_t end;
4599 	int error;
4600 
4601 	if (offset < 0 || len < 0 || offset > OFF_MAX - len)
4602 		return (EINVAL);
4603 	switch (advice) {
4604 	case POSIX_FADV_SEQUENTIAL:
4605 	case POSIX_FADV_RANDOM:
4606 	case POSIX_FADV_NOREUSE:
4607 		new = malloc(sizeof(*fa), M_FADVISE, M_WAITOK);
4608 		break;
4609 	case POSIX_FADV_NORMAL:
4610 	case POSIX_FADV_WILLNEED:
4611 	case POSIX_FADV_DONTNEED:
4612 		new = NULL;
4613 		break;
4614 	default:
4615 		return (EINVAL);
4616 	}
4617 	/* XXX: CAP_POSIX_FADVISE? */
4618 	error = fget(td, fd, 0, &fp);
4619 	if (error != 0)
4620 		goto out;
4621 
4622 	switch (fp->f_type) {
4623 	case DTYPE_VNODE:
4624 		break;
4625 	case DTYPE_PIPE:
4626 	case DTYPE_FIFO:
4627 		error = ESPIPE;
4628 		goto out;
4629 	default:
4630 		error = ENODEV;
4631 		goto out;
4632 	}
4633 	vp = fp->f_vnode;
4634 	if (vp->v_type != VREG) {
4635 		error = ENODEV;
4636 		goto out;
4637 	}
4638 	if (len == 0)
4639 		end = OFF_MAX;
4640 	else
4641 		end = offset + len - 1;
4642 	switch (advice) {
4643 	case POSIX_FADV_SEQUENTIAL:
4644 	case POSIX_FADV_RANDOM:
4645 	case POSIX_FADV_NOREUSE:
4646 		/*
4647 		 * Try to merge any existing non-standard region with
4648 		 * this new region if possible, otherwise create a new
4649 		 * non-standard region for this request.
4650 		 */
4651 		mtx_pool_lock(mtxpool_sleep, fp);
4652 		fa = fp->f_advice;
4653 		if (fa != NULL && fa->fa_advice == advice &&
4654 		    ((fa->fa_start <= end && fa->fa_end >= offset) ||
4655 		    (end != OFF_MAX && fa->fa_start == end + 1) ||
4656 		    (fa->fa_end != OFF_MAX && fa->fa_end + 1 == offset))) {
4657 			if (offset < fa->fa_start)
4658 				fa->fa_start = offset;
4659 			if (end > fa->fa_end)
4660 				fa->fa_end = end;
4661 		} else {
4662 			new->fa_advice = advice;
4663 			new->fa_start = offset;
4664 			new->fa_end = end;
4665 			new->fa_prevstart = 0;
4666 			new->fa_prevend = 0;
4667 			fp->f_advice = new;
4668 			new = fa;
4669 		}
4670 		mtx_pool_unlock(mtxpool_sleep, fp);
4671 		break;
4672 	case POSIX_FADV_NORMAL:
4673 		/*
4674 		 * If a the "normal" region overlaps with an existing
4675 		 * non-standard region, trim or remove the
4676 		 * non-standard region.
4677 		 */
4678 		mtx_pool_lock(mtxpool_sleep, fp);
4679 		fa = fp->f_advice;
4680 		if (fa != NULL) {
4681 			if (offset <= fa->fa_start && end >= fa->fa_end) {
4682 				new = fa;
4683 				fp->f_advice = NULL;
4684 			} else if (offset <= fa->fa_start &&
4685  			    end >= fa->fa_start)
4686 				fa->fa_start = end + 1;
4687 			else if (offset <= fa->fa_end && end >= fa->fa_end)
4688 				fa->fa_end = offset - 1;
4689 			else if (offset >= fa->fa_start && end <= fa->fa_end) {
4690 				/*
4691 				 * If the "normal" region is a middle
4692 				 * portion of the existing
4693 				 * non-standard region, just remove
4694 				 * the whole thing rather than picking
4695 				 * one side or the other to
4696 				 * preserve.
4697 				 */
4698 				new = fa;
4699 				fp->f_advice = NULL;
4700 			}
4701 		}
4702 		mtx_pool_unlock(mtxpool_sleep, fp);
4703 		break;
4704 	case POSIX_FADV_WILLNEED:
4705 	case POSIX_FADV_DONTNEED:
4706 		error = VOP_ADVISE(vp, offset, end, advice);
4707 		break;
4708 	}
4709 out:
4710 	if (fp != NULL)
4711 		fdrop(fp, td);
4712 	free(new, M_FADVISE);
4713 	return (error);
4714 }
4715 
4716 int
4717 sys_posix_fadvise(struct thread *td, struct posix_fadvise_args *uap)
4718 {
4719 
4720 	return (kern_posix_fadvise(td, uap->fd, uap->offset, uap->len,
4721 	    uap->advice));
4722 }
4723