xref: /freebsd/sys/kern/vfs_syscalls.c (revision d940bfec8c329dd82d8d54efebd81c8aa420503b)
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/rwlock.h>
65 #include <sys/sdt.h>
66 #include <sys/stat.h>
67 #include <sys/sx.h>
68 #include <sys/unistd.h>
69 #include <sys/vnode.h>
70 #include <sys/priv.h>
71 #include <sys/proc.h>
72 #include <sys/dirent.h>
73 #include <sys/jail.h>
74 #include <sys/syscallsubr.h>
75 #include <sys/sysctl.h>
76 #ifdef KTRACE
77 #include <sys/ktrace.h>
78 #endif
79 
80 #include <machine/stdarg.h>
81 
82 #include <security/audit/audit.h>
83 #include <security/mac/mac_framework.h>
84 
85 #include <vm/vm.h>
86 #include <vm/vm_object.h>
87 #include <vm/vm_page.h>
88 #include <vm/uma.h>
89 
90 #include <ufs/ufs/quota.h>
91 
92 MALLOC_DEFINE(M_FADVISE, "fadvise", "posix_fadvise(2) information");
93 
94 SDT_PROVIDER_DEFINE(vfs);
95 SDT_PROBE_DEFINE2(vfs, , stat, mode, mode, "char *", "int");
96 SDT_PROBE_DEFINE2(vfs, , stat, reg, reg, "char *", "int");
97 
98 static int chroot_refuse_vdir_fds(struct filedesc *fdp);
99 static int getutimes(const struct timeval *, enum uio_seg, struct timespec *);
100 static int kern_chflags(struct thread *td, const char *path,
101     enum uio_seg pathseg, u_long flags);
102 static int kern_chflagsat(struct thread *td, int fd, const char *path,
103     enum uio_seg pathseg, u_long flags, int atflag);
104 static int setfflags(struct thread *td, struct vnode *, u_long);
105 static int setutimes(struct thread *td, struct vnode *,
106     const struct timespec *, int, int);
107 static int vn_access(struct vnode *vp, int user_flags, struct ucred *cred,
108     struct thread *td);
109 
110 /*
111  * The module initialization routine for POSIX asynchronous I/O will
112  * set this to the version of AIO that it implements.  (Zero means
113  * that it is not implemented.)  This value is used here by pathconf()
114  * and in kern_descrip.c by fpathconf().
115  */
116 int async_io_version;
117 
118 #ifdef DEBUG
119 static int syncprt = 0;
120 SYSCTL_INT(_debug, OID_AUTO, syncprt, CTLFLAG_RW, &syncprt, 0, "");
121 #endif
122 
123 /*
124  * Sync each mounted filesystem.
125  */
126 #ifndef _SYS_SYSPROTO_H_
127 struct sync_args {
128 	int     dummy;
129 };
130 #endif
131 /* ARGSUSED */
132 int
133 sys_sync(td, uap)
134 	struct thread *td;
135 	struct sync_args *uap;
136 {
137 	struct mount *mp, *nmp;
138 	int save;
139 
140 	mtx_lock(&mountlist_mtx);
141 	for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
142 		if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK)) {
143 			nmp = TAILQ_NEXT(mp, mnt_list);
144 			continue;
145 		}
146 		if ((mp->mnt_flag & MNT_RDONLY) == 0 &&
147 		    vn_start_write(NULL, &mp, V_NOWAIT) == 0) {
148 			save = curthread_pflags_set(TDP_SYNCIO);
149 			vfs_msync(mp, MNT_NOWAIT);
150 			VFS_SYNC(mp, MNT_NOWAIT);
151 			curthread_pflags_restore(save);
152 			vn_finished_write(mp);
153 		}
154 		mtx_lock(&mountlist_mtx);
155 		nmp = TAILQ_NEXT(mp, mnt_list);
156 		vfs_unbusy(mp);
157 	}
158 	mtx_unlock(&mountlist_mtx);
159 	return (0);
160 }
161 
162 /*
163  * Change filesystem quotas.
164  */
165 #ifndef _SYS_SYSPROTO_H_
166 struct quotactl_args {
167 	char *path;
168 	int cmd;
169 	int uid;
170 	caddr_t arg;
171 };
172 #endif
173 int
174 sys_quotactl(td, uap)
175 	struct thread *td;
176 	register struct quotactl_args /* {
177 		char *path;
178 		int cmd;
179 		int uid;
180 		caddr_t arg;
181 	} */ *uap;
182 {
183 	struct mount *mp;
184 	int error;
185 	struct nameidata nd;
186 
187 	AUDIT_ARG_CMD(uap->cmd);
188 	AUDIT_ARG_UID(uap->uid);
189 	if (!prison_allow(td->td_ucred, PR_ALLOW_QUOTAS))
190 		return (EPERM);
191 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
192 	    uap->path, td);
193 	if ((error = namei(&nd)) != 0)
194 		return (error);
195 	NDFREE(&nd, NDF_ONLY_PNBUF);
196 	mp = nd.ni_vp->v_mount;
197 	vfs_ref(mp);
198 	vput(nd.ni_vp);
199 	error = vfs_busy(mp, 0);
200 	vfs_rel(mp);
201 	if (error)
202 		return (error);
203 	error = VFS_QUOTACTL(mp, uap->cmd, uap->uid, uap->arg);
204 
205 	/*
206 	 * Since quota on operation typically needs to open quota
207 	 * file, the Q_QUOTAON handler needs to unbusy the mount point
208 	 * before calling into namei.  Otherwise, unmount might be
209 	 * started between two vfs_busy() invocations (first is our,
210 	 * second is from mount point cross-walk code in lookup()),
211 	 * causing deadlock.
212 	 *
213 	 * Require that Q_QUOTAON handles the vfs_busy() reference on
214 	 * its own, always returning with ubusied mount point.
215 	 */
216 	if ((uap->cmd >> SUBCMDSHIFT) != Q_QUOTAON)
217 		vfs_unbusy(mp);
218 	return (error);
219 }
220 
221 /*
222  * Used by statfs conversion routines to scale the block size up if
223  * necessary so that all of the block counts are <= 'max_size'.  Note
224  * that 'max_size' should be a bitmask, i.e. 2^n - 1 for some non-zero
225  * value of 'n'.
226  */
227 void
228 statfs_scale_blocks(struct statfs *sf, long max_size)
229 {
230 	uint64_t count;
231 	int shift;
232 
233 	KASSERT(powerof2(max_size + 1), ("%s: invalid max_size", __func__));
234 
235 	/*
236 	 * Attempt to scale the block counts to give a more accurate
237 	 * overview to userland of the ratio of free space to used
238 	 * space.  To do this, find the largest block count and compute
239 	 * a divisor that lets it fit into a signed integer <= max_size.
240 	 */
241 	if (sf->f_bavail < 0)
242 		count = -sf->f_bavail;
243 	else
244 		count = sf->f_bavail;
245 	count = MAX(sf->f_blocks, MAX(sf->f_bfree, count));
246 	if (count <= max_size)
247 		return;
248 
249 	count >>= flsl(max_size);
250 	shift = 0;
251 	while (count > 0) {
252 		shift++;
253 		count >>=1;
254 	}
255 
256 	sf->f_bsize <<= shift;
257 	sf->f_blocks >>= shift;
258 	sf->f_bfree >>= shift;
259 	sf->f_bavail >>= shift;
260 }
261 
262 /*
263  * Get filesystem statistics.
264  */
265 #ifndef _SYS_SYSPROTO_H_
266 struct statfs_args {
267 	char *path;
268 	struct statfs *buf;
269 };
270 #endif
271 int
272 sys_statfs(td, uap)
273 	struct thread *td;
274 	register struct statfs_args /* {
275 		char *path;
276 		struct statfs *buf;
277 	} */ *uap;
278 {
279 	struct statfs sf;
280 	int error;
281 
282 	error = kern_statfs(td, uap->path, UIO_USERSPACE, &sf);
283 	if (error == 0)
284 		error = copyout(&sf, uap->buf, sizeof(sf));
285 	return (error);
286 }
287 
288 int
289 kern_statfs(struct thread *td, char *path, enum uio_seg pathseg,
290     struct statfs *buf)
291 {
292 	struct mount *mp;
293 	struct statfs *sp, sb;
294 	int error;
295 	struct nameidata nd;
296 
297 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
298 	    pathseg, path, td);
299 	error = namei(&nd);
300 	if (error)
301 		return (error);
302 	mp = nd.ni_vp->v_mount;
303 	vfs_ref(mp);
304 	NDFREE(&nd, NDF_ONLY_PNBUF);
305 	vput(nd.ni_vp);
306 	error = vfs_busy(mp, 0);
307 	vfs_rel(mp);
308 	if (error)
309 		return (error);
310 #ifdef MAC
311 	error = mac_mount_check_stat(td->td_ucred, mp);
312 	if (error)
313 		goto out;
314 #endif
315 	/*
316 	 * Set these in case the underlying filesystem fails to do so.
317 	 */
318 	sp = &mp->mnt_stat;
319 	sp->f_version = STATFS_VERSION;
320 	sp->f_namemax = NAME_MAX;
321 	sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
322 	error = VFS_STATFS(mp, sp);
323 	if (error)
324 		goto out;
325 	if (priv_check(td, PRIV_VFS_GENERATION)) {
326 		bcopy(sp, &sb, sizeof(sb));
327 		sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
328 		prison_enforce_statfs(td->td_ucred, mp, &sb);
329 		sp = &sb;
330 	}
331 	*buf = *sp;
332 out:
333 	vfs_unbusy(mp);
334 	return (error);
335 }
336 
337 /*
338  * Get filesystem statistics.
339  */
340 #ifndef _SYS_SYSPROTO_H_
341 struct fstatfs_args {
342 	int fd;
343 	struct statfs *buf;
344 };
345 #endif
346 int
347 sys_fstatfs(td, uap)
348 	struct thread *td;
349 	register struct fstatfs_args /* {
350 		int fd;
351 		struct statfs *buf;
352 	} */ *uap;
353 {
354 	struct statfs sf;
355 	int error;
356 
357 	error = kern_fstatfs(td, uap->fd, &sf);
358 	if (error == 0)
359 		error = copyout(&sf, uap->buf, sizeof(sf));
360 	return (error);
361 }
362 
363 int
364 kern_fstatfs(struct thread *td, int fd, struct statfs *buf)
365 {
366 	struct file *fp;
367 	struct mount *mp;
368 	struct statfs *sp, sb;
369 	struct vnode *vp;
370 	int error;
371 
372 	AUDIT_ARG_FD(fd);
373 	error = getvnode(td->td_proc->p_fd, fd, CAP_FSTATFS, &fp);
374 	if (error)
375 		return (error);
376 	vp = fp->f_vnode;
377 	vn_lock(vp, LK_SHARED | LK_RETRY);
378 #ifdef AUDIT
379 	AUDIT_ARG_VNODE1(vp);
380 #endif
381 	mp = vp->v_mount;
382 	if (mp)
383 		vfs_ref(mp);
384 	VOP_UNLOCK(vp, 0);
385 	fdrop(fp, td);
386 	if (mp == NULL) {
387 		error = EBADF;
388 		goto out;
389 	}
390 	error = vfs_busy(mp, 0);
391 	vfs_rel(mp);
392 	if (error)
393 		return (error);
394 #ifdef MAC
395 	error = mac_mount_check_stat(td->td_ucred, mp);
396 	if (error)
397 		goto out;
398 #endif
399 	/*
400 	 * Set these in case the underlying filesystem fails to do so.
401 	 */
402 	sp = &mp->mnt_stat;
403 	sp->f_version = STATFS_VERSION;
404 	sp->f_namemax = NAME_MAX;
405 	sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
406 	error = VFS_STATFS(mp, sp);
407 	if (error)
408 		goto out;
409 	if (priv_check(td, PRIV_VFS_GENERATION)) {
410 		bcopy(sp, &sb, sizeof(sb));
411 		sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
412 		prison_enforce_statfs(td->td_ucred, mp, &sb);
413 		sp = &sb;
414 	}
415 	*buf = *sp;
416 out:
417 	if (mp)
418 		vfs_unbusy(mp);
419 	return (error);
420 }
421 
422 /*
423  * Get statistics on all filesystems.
424  */
425 #ifndef _SYS_SYSPROTO_H_
426 struct getfsstat_args {
427 	struct statfs *buf;
428 	long bufsize;
429 	int flags;
430 };
431 #endif
432 int
433 sys_getfsstat(td, uap)
434 	struct thread *td;
435 	register struct getfsstat_args /* {
436 		struct statfs *buf;
437 		long bufsize;
438 		int flags;
439 	} */ *uap;
440 {
441 
442 	return (kern_getfsstat(td, &uap->buf, uap->bufsize, UIO_USERSPACE,
443 	    uap->flags));
444 }
445 
446 /*
447  * If (bufsize > 0 && bufseg == UIO_SYSSPACE)
448  *	The caller is responsible for freeing memory which will be allocated
449  *	in '*buf'.
450  */
451 int
452 kern_getfsstat(struct thread *td, struct statfs **buf, size_t bufsize,
453     enum uio_seg bufseg, int flags)
454 {
455 	struct mount *mp, *nmp;
456 	struct statfs *sfsp, *sp, sb;
457 	size_t count, maxcount;
458 	int error;
459 
460 	maxcount = bufsize / sizeof(struct statfs);
461 	if (bufsize == 0)
462 		sfsp = NULL;
463 	else if (bufseg == UIO_USERSPACE)
464 		sfsp = *buf;
465 	else /* if (bufseg == UIO_SYSSPACE) */ {
466 		count = 0;
467 		mtx_lock(&mountlist_mtx);
468 		TAILQ_FOREACH(mp, &mountlist, mnt_list) {
469 			count++;
470 		}
471 		mtx_unlock(&mountlist_mtx);
472 		if (maxcount > count)
473 			maxcount = count;
474 		sfsp = *buf = malloc(maxcount * sizeof(struct statfs), M_TEMP,
475 		    M_WAITOK);
476 	}
477 	count = 0;
478 	mtx_lock(&mountlist_mtx);
479 	for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
480 		if (prison_canseemount(td->td_ucred, mp) != 0) {
481 			nmp = TAILQ_NEXT(mp, mnt_list);
482 			continue;
483 		}
484 #ifdef MAC
485 		if (mac_mount_check_stat(td->td_ucred, mp) != 0) {
486 			nmp = TAILQ_NEXT(mp, mnt_list);
487 			continue;
488 		}
489 #endif
490 		if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK)) {
491 			nmp = TAILQ_NEXT(mp, mnt_list);
492 			continue;
493 		}
494 		if (sfsp && count < maxcount) {
495 			sp = &mp->mnt_stat;
496 			/*
497 			 * Set these in case the underlying filesystem
498 			 * fails to do so.
499 			 */
500 			sp->f_version = STATFS_VERSION;
501 			sp->f_namemax = NAME_MAX;
502 			sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
503 			/*
504 			 * If MNT_NOWAIT or MNT_LAZY is specified, do not
505 			 * refresh the fsstat cache. MNT_NOWAIT or MNT_LAZY
506 			 * overrides MNT_WAIT.
507 			 */
508 			if (((flags & (MNT_LAZY|MNT_NOWAIT)) == 0 ||
509 			    (flags & MNT_WAIT)) &&
510 			    (error = VFS_STATFS(mp, sp))) {
511 				mtx_lock(&mountlist_mtx);
512 				nmp = TAILQ_NEXT(mp, mnt_list);
513 				vfs_unbusy(mp);
514 				continue;
515 			}
516 			if (priv_check(td, PRIV_VFS_GENERATION)) {
517 				bcopy(sp, &sb, sizeof(sb));
518 				sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
519 				prison_enforce_statfs(td->td_ucred, mp, &sb);
520 				sp = &sb;
521 			}
522 			if (bufseg == UIO_SYSSPACE)
523 				bcopy(sp, sfsp, sizeof(*sp));
524 			else /* if (bufseg == UIO_USERSPACE) */ {
525 				error = copyout(sp, sfsp, sizeof(*sp));
526 				if (error) {
527 					vfs_unbusy(mp);
528 					return (error);
529 				}
530 			}
531 			sfsp++;
532 		}
533 		count++;
534 		mtx_lock(&mountlist_mtx);
535 		nmp = TAILQ_NEXT(mp, mnt_list);
536 		vfs_unbusy(mp);
537 	}
538 	mtx_unlock(&mountlist_mtx);
539 	if (sfsp && count > maxcount)
540 		td->td_retval[0] = maxcount;
541 	else
542 		td->td_retval[0] = count;
543 	return (0);
544 }
545 
546 #ifdef COMPAT_FREEBSD4
547 /*
548  * Get old format filesystem statistics.
549  */
550 static void cvtstatfs(struct statfs *, struct ostatfs *);
551 
552 #ifndef _SYS_SYSPROTO_H_
553 struct freebsd4_statfs_args {
554 	char *path;
555 	struct ostatfs *buf;
556 };
557 #endif
558 int
559 freebsd4_statfs(td, uap)
560 	struct thread *td;
561 	struct freebsd4_statfs_args /* {
562 		char *path;
563 		struct ostatfs *buf;
564 	} */ *uap;
565 {
566 	struct ostatfs osb;
567 	struct statfs sf;
568 	int error;
569 
570 	error = kern_statfs(td, uap->path, UIO_USERSPACE, &sf);
571 	if (error)
572 		return (error);
573 	cvtstatfs(&sf, &osb);
574 	return (copyout(&osb, uap->buf, sizeof(osb)));
575 }
576 
577 /*
578  * Get filesystem statistics.
579  */
580 #ifndef _SYS_SYSPROTO_H_
581 struct freebsd4_fstatfs_args {
582 	int fd;
583 	struct ostatfs *buf;
584 };
585 #endif
586 int
587 freebsd4_fstatfs(td, uap)
588 	struct thread *td;
589 	struct freebsd4_fstatfs_args /* {
590 		int fd;
591 		struct ostatfs *buf;
592 	} */ *uap;
593 {
594 	struct ostatfs osb;
595 	struct statfs sf;
596 	int error;
597 
598 	error = kern_fstatfs(td, uap->fd, &sf);
599 	if (error)
600 		return (error);
601 	cvtstatfs(&sf, &osb);
602 	return (copyout(&osb, uap->buf, sizeof(osb)));
603 }
604 
605 /*
606  * Get statistics on all filesystems.
607  */
608 #ifndef _SYS_SYSPROTO_H_
609 struct freebsd4_getfsstat_args {
610 	struct ostatfs *buf;
611 	long bufsize;
612 	int flags;
613 };
614 #endif
615 int
616 freebsd4_getfsstat(td, uap)
617 	struct thread *td;
618 	register struct freebsd4_getfsstat_args /* {
619 		struct ostatfs *buf;
620 		long bufsize;
621 		int flags;
622 	} */ *uap;
623 {
624 	struct statfs *buf, *sp;
625 	struct ostatfs osb;
626 	size_t count, size;
627 	int error;
628 
629 	count = uap->bufsize / sizeof(struct ostatfs);
630 	size = count * sizeof(struct statfs);
631 	error = kern_getfsstat(td, &buf, size, UIO_SYSSPACE, uap->flags);
632 	if (size > 0) {
633 		count = td->td_retval[0];
634 		sp = buf;
635 		while (count > 0 && error == 0) {
636 			cvtstatfs(sp, &osb);
637 			error = copyout(&osb, uap->buf, sizeof(osb));
638 			sp++;
639 			uap->buf++;
640 			count--;
641 		}
642 		free(buf, M_TEMP);
643 	}
644 	return (error);
645 }
646 
647 /*
648  * Implement fstatfs() for (NFS) file handles.
649  */
650 #ifndef _SYS_SYSPROTO_H_
651 struct freebsd4_fhstatfs_args {
652 	struct fhandle *u_fhp;
653 	struct ostatfs *buf;
654 };
655 #endif
656 int
657 freebsd4_fhstatfs(td, uap)
658 	struct thread *td;
659 	struct freebsd4_fhstatfs_args /* {
660 		struct fhandle *u_fhp;
661 		struct ostatfs *buf;
662 	} */ *uap;
663 {
664 	struct ostatfs osb;
665 	struct statfs sf;
666 	fhandle_t fh;
667 	int error;
668 
669 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
670 	if (error)
671 		return (error);
672 	error = kern_fhstatfs(td, fh, &sf);
673 	if (error)
674 		return (error);
675 	cvtstatfs(&sf, &osb);
676 	return (copyout(&osb, uap->buf, sizeof(osb)));
677 }
678 
679 /*
680  * Convert a new format statfs structure to an old format statfs structure.
681  */
682 static void
683 cvtstatfs(nsp, osp)
684 	struct statfs *nsp;
685 	struct ostatfs *osp;
686 {
687 
688 	statfs_scale_blocks(nsp, LONG_MAX);
689 	bzero(osp, sizeof(*osp));
690 	osp->f_bsize = nsp->f_bsize;
691 	osp->f_iosize = MIN(nsp->f_iosize, LONG_MAX);
692 	osp->f_blocks = nsp->f_blocks;
693 	osp->f_bfree = nsp->f_bfree;
694 	osp->f_bavail = nsp->f_bavail;
695 	osp->f_files = MIN(nsp->f_files, LONG_MAX);
696 	osp->f_ffree = MIN(nsp->f_ffree, LONG_MAX);
697 	osp->f_owner = nsp->f_owner;
698 	osp->f_type = nsp->f_type;
699 	osp->f_flags = nsp->f_flags;
700 	osp->f_syncwrites = MIN(nsp->f_syncwrites, LONG_MAX);
701 	osp->f_asyncwrites = MIN(nsp->f_asyncwrites, LONG_MAX);
702 	osp->f_syncreads = MIN(nsp->f_syncreads, LONG_MAX);
703 	osp->f_asyncreads = MIN(nsp->f_asyncreads, LONG_MAX);
704 	strlcpy(osp->f_fstypename, nsp->f_fstypename,
705 	    MIN(MFSNAMELEN, OMFSNAMELEN));
706 	strlcpy(osp->f_mntonname, nsp->f_mntonname,
707 	    MIN(MNAMELEN, OMNAMELEN));
708 	strlcpy(osp->f_mntfromname, nsp->f_mntfromname,
709 	    MIN(MNAMELEN, OMNAMELEN));
710 	osp->f_fsid = nsp->f_fsid;
711 }
712 #endif /* COMPAT_FREEBSD4 */
713 
714 /*
715  * Change current working directory to a given file descriptor.
716  */
717 #ifndef _SYS_SYSPROTO_H_
718 struct fchdir_args {
719 	int	fd;
720 };
721 #endif
722 int
723 sys_fchdir(td, uap)
724 	struct thread *td;
725 	struct fchdir_args /* {
726 		int fd;
727 	} */ *uap;
728 {
729 	register struct filedesc *fdp = td->td_proc->p_fd;
730 	struct vnode *vp, *tdp, *vpold;
731 	struct mount *mp;
732 	struct file *fp;
733 	int error;
734 
735 	AUDIT_ARG_FD(uap->fd);
736 	if ((error = getvnode(fdp, uap->fd, CAP_FCHDIR, &fp)) != 0)
737 		return (error);
738 	vp = fp->f_vnode;
739 	VREF(vp);
740 	fdrop(fp, td);
741 	vn_lock(vp, LK_SHARED | LK_RETRY);
742 	AUDIT_ARG_VNODE1(vp);
743 	error = change_dir(vp, td);
744 	while (!error && (mp = vp->v_mountedhere) != NULL) {
745 		if (vfs_busy(mp, 0))
746 			continue;
747 		error = VFS_ROOT(mp, LK_SHARED, &tdp);
748 		vfs_unbusy(mp);
749 		if (error)
750 			break;
751 		vput(vp);
752 		vp = tdp;
753 	}
754 	if (error) {
755 		vput(vp);
756 		return (error);
757 	}
758 	VOP_UNLOCK(vp, 0);
759 	FILEDESC_XLOCK(fdp);
760 	vpold = fdp->fd_cdir;
761 	fdp->fd_cdir = vp;
762 	FILEDESC_XUNLOCK(fdp);
763 	vrele(vpold);
764 	return (0);
765 }
766 
767 /*
768  * Change current working directory (``.'').
769  */
770 #ifndef _SYS_SYSPROTO_H_
771 struct chdir_args {
772 	char	*path;
773 };
774 #endif
775 int
776 sys_chdir(td, uap)
777 	struct thread *td;
778 	struct chdir_args /* {
779 		char *path;
780 	} */ *uap;
781 {
782 
783 	return (kern_chdir(td, uap->path, UIO_USERSPACE));
784 }
785 
786 int
787 kern_chdir(struct thread *td, char *path, enum uio_seg pathseg)
788 {
789 	register struct filedesc *fdp = td->td_proc->p_fd;
790 	int error;
791 	struct nameidata nd;
792 	struct vnode *vp;
793 
794 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
795 	    pathseg, path, td);
796 	if ((error = namei(&nd)) != 0)
797 		return (error);
798 	if ((error = change_dir(nd.ni_vp, td)) != 0) {
799 		vput(nd.ni_vp);
800 		NDFREE(&nd, NDF_ONLY_PNBUF);
801 		return (error);
802 	}
803 	VOP_UNLOCK(nd.ni_vp, 0);
804 	NDFREE(&nd, NDF_ONLY_PNBUF);
805 	FILEDESC_XLOCK(fdp);
806 	vp = fdp->fd_cdir;
807 	fdp->fd_cdir = nd.ni_vp;
808 	FILEDESC_XUNLOCK(fdp);
809 	vrele(vp);
810 	return (0);
811 }
812 
813 /*
814  * Helper function for raised chroot(2) security function:  Refuse if
815  * any filedescriptors are open directories.
816  */
817 static int
818 chroot_refuse_vdir_fds(fdp)
819 	struct filedesc *fdp;
820 {
821 	struct vnode *vp;
822 	struct file *fp;
823 	int fd;
824 
825 	FILEDESC_LOCK_ASSERT(fdp);
826 
827 	for (fd = 0; fd < fdp->fd_nfiles ; fd++) {
828 		fp = fget_locked(fdp, fd);
829 		if (fp == NULL)
830 			continue;
831 		if (fp->f_type == DTYPE_VNODE) {
832 			vp = fp->f_vnode;
833 			if (vp->v_type == VDIR)
834 				return (EPERM);
835 		}
836 	}
837 	return (0);
838 }
839 
840 /*
841  * This sysctl determines if we will allow a process to chroot(2) if it
842  * has a directory open:
843  *	0: disallowed for all processes.
844  *	1: allowed for processes that were not already chroot(2)'ed.
845  *	2: allowed for all processes.
846  */
847 
848 static int chroot_allow_open_directories = 1;
849 
850 SYSCTL_INT(_kern, OID_AUTO, chroot_allow_open_directories, CTLFLAG_RW,
851      &chroot_allow_open_directories, 0,
852      "Allow a process to chroot(2) if it has a directory open");
853 
854 /*
855  * Change notion of root (``/'') directory.
856  */
857 #ifndef _SYS_SYSPROTO_H_
858 struct chroot_args {
859 	char	*path;
860 };
861 #endif
862 int
863 sys_chroot(td, uap)
864 	struct thread *td;
865 	struct chroot_args /* {
866 		char *path;
867 	} */ *uap;
868 {
869 	int error;
870 	struct nameidata nd;
871 
872 	error = priv_check(td, PRIV_VFS_CHROOT);
873 	if (error)
874 		return (error);
875 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
876 	    UIO_USERSPACE, uap->path, td);
877 	error = namei(&nd);
878 	if (error)
879 		goto error;
880 	if ((error = change_dir(nd.ni_vp, td)) != 0)
881 		goto e_vunlock;
882 #ifdef MAC
883 	if ((error = mac_vnode_check_chroot(td->td_ucred, nd.ni_vp)))
884 		goto e_vunlock;
885 #endif
886 	VOP_UNLOCK(nd.ni_vp, 0);
887 	error = change_root(nd.ni_vp, td);
888 	vrele(nd.ni_vp);
889 	NDFREE(&nd, NDF_ONLY_PNBUF);
890 	return (error);
891 e_vunlock:
892 	vput(nd.ni_vp);
893 error:
894 	NDFREE(&nd, NDF_ONLY_PNBUF);
895 	return (error);
896 }
897 
898 /*
899  * Common routine for chroot and chdir.  Callers must provide a locked vnode
900  * instance.
901  */
902 int
903 change_dir(vp, td)
904 	struct vnode *vp;
905 	struct thread *td;
906 {
907 	int error;
908 
909 	ASSERT_VOP_LOCKED(vp, "change_dir(): vp not locked");
910 	if (vp->v_type != VDIR)
911 		return (ENOTDIR);
912 #ifdef MAC
913 	error = mac_vnode_check_chdir(td->td_ucred, vp);
914 	if (error)
915 		return (error);
916 #endif
917 	error = VOP_ACCESS(vp, VEXEC, td->td_ucred, td);
918 	return (error);
919 }
920 
921 /*
922  * Common routine for kern_chroot() and jail_attach().  The caller is
923  * responsible for invoking priv_check() and mac_vnode_check_chroot() to
924  * authorize this operation.
925  */
926 int
927 change_root(vp, td)
928 	struct vnode *vp;
929 	struct thread *td;
930 {
931 	struct filedesc *fdp;
932 	struct vnode *oldvp;
933 	int error;
934 
935 	fdp = td->td_proc->p_fd;
936 	FILEDESC_XLOCK(fdp);
937 	if (chroot_allow_open_directories == 0 ||
938 	    (chroot_allow_open_directories == 1 && fdp->fd_rdir != rootvnode)) {
939 		error = chroot_refuse_vdir_fds(fdp);
940 		if (error) {
941 			FILEDESC_XUNLOCK(fdp);
942 			return (error);
943 		}
944 	}
945 	oldvp = fdp->fd_rdir;
946 	fdp->fd_rdir = vp;
947 	VREF(fdp->fd_rdir);
948 	if (!fdp->fd_jdir) {
949 		fdp->fd_jdir = vp;
950 		VREF(fdp->fd_jdir);
951 	}
952 	FILEDESC_XUNLOCK(fdp);
953 	vrele(oldvp);
954 	return (0);
955 }
956 
957 static __inline cap_rights_t
958 flags_to_rights(int flags)
959 {
960 	cap_rights_t rights = 0;
961 
962 	if (flags & O_EXEC) {
963 		rights |= CAP_FEXECVE;
964 	} else {
965 		switch ((flags & O_ACCMODE)) {
966 		case O_RDONLY:
967 			rights |= CAP_READ;
968 			break;
969 		case O_RDWR:
970 			rights |= CAP_READ;
971 			/* FALLTHROUGH */
972 		case O_WRONLY:
973 			rights |= CAP_WRITE;
974 			if (!(flags & (O_APPEND | O_TRUNC)))
975 				rights |= CAP_SEEK;
976 			break;
977 		}
978 	}
979 
980 	if (flags & O_CREAT)
981 		rights |= CAP_CREATE;
982 
983 	if (flags & O_TRUNC)
984 		rights |= CAP_FTRUNCATE;
985 
986 	if (flags & (O_SYNC | O_FSYNC))
987 		rights |= CAP_FSYNC;
988 
989 	if (flags & (O_EXLOCK | O_SHLOCK))
990 		rights |= CAP_FLOCK;
991 
992 	return (rights);
993 }
994 
995 /*
996  * Check permissions, allocate an open file structure, and call the device
997  * open routine if any.
998  */
999 #ifndef _SYS_SYSPROTO_H_
1000 struct open_args {
1001 	char	*path;
1002 	int	flags;
1003 	int	mode;
1004 };
1005 #endif
1006 int
1007 sys_open(td, uap)
1008 	struct thread *td;
1009 	register struct open_args /* {
1010 		char *path;
1011 		int flags;
1012 		int mode;
1013 	} */ *uap;
1014 {
1015 
1016 	return (kern_open(td, uap->path, UIO_USERSPACE, uap->flags, uap->mode));
1017 }
1018 
1019 #ifndef _SYS_SYSPROTO_H_
1020 struct openat_args {
1021 	int	fd;
1022 	char	*path;
1023 	int	flag;
1024 	int	mode;
1025 };
1026 #endif
1027 int
1028 sys_openat(struct thread *td, struct openat_args *uap)
1029 {
1030 
1031 	return (kern_openat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
1032 	    uap->mode));
1033 }
1034 
1035 int
1036 kern_open(struct thread *td, char *path, enum uio_seg pathseg, int flags,
1037     int mode)
1038 {
1039 
1040 	return (kern_openat(td, AT_FDCWD, path, pathseg, flags, mode));
1041 }
1042 
1043 int
1044 kern_openat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1045     int flags, int mode)
1046 {
1047 	struct proc *p = td->td_proc;
1048 	struct filedesc *fdp = p->p_fd;
1049 	struct file *fp;
1050 	struct vnode *vp;
1051 	int cmode;
1052 	int indx = -1, error;
1053 	struct nameidata nd;
1054 	cap_rights_t rights_needed = CAP_LOOKUP;
1055 
1056 	AUDIT_ARG_FFLAGS(flags);
1057 	AUDIT_ARG_MODE(mode);
1058 	/* XXX: audit dirfd */
1059 	rights_needed |= flags_to_rights(flags);
1060 	/*
1061 	 * Only one of the O_EXEC, O_RDONLY, O_WRONLY and O_RDWR flags
1062 	 * may be specified.
1063 	 */
1064 	if (flags & O_EXEC) {
1065 		if (flags & O_ACCMODE)
1066 			return (EINVAL);
1067 	} else if ((flags & O_ACCMODE) == O_ACCMODE) {
1068 		return (EINVAL);
1069 	} else {
1070 		flags = FFLAGS(flags);
1071 	}
1072 
1073 	/*
1074 	 * Allocate the file descriptor, but don't install a descriptor yet.
1075 	 */
1076 	error = falloc_noinstall(td, &fp);
1077 	if (error)
1078 		return (error);
1079 	/*
1080 	 * An extra reference on `fp' has been held for us by
1081 	 * falloc_noinstall().
1082 	 */
1083 	/* Set the flags early so the finit in devfs can pick them up. */
1084 	fp->f_flag = flags & FMASK;
1085 	cmode = ((mode &~ fdp->fd_cmask) & ALLPERMS) &~ S_ISTXT;
1086 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, fd,
1087 	    rights_needed, td);
1088 	td->td_dupfd = -1;		/* XXX check for fdopen */
1089 	error = vn_open(&nd, &flags, cmode, fp);
1090 	if (error) {
1091 		/*
1092 		 * If the vn_open replaced the method vector, something
1093 		 * wonderous happened deep below and we just pass it up
1094 		 * pretending we know what we do.
1095 		 */
1096 		if (error == ENXIO && fp->f_ops != &badfileops)
1097 			goto success;
1098 
1099 		/*
1100 		 * Handle special fdopen() case. bleh.
1101 		 *
1102 		 * Don't do this for relative (capability) lookups; we don't
1103 		 * understand exactly what would happen, and we don't think
1104 		 * that it ever should.
1105 		 */
1106 		if (nd.ni_strictrelative == 0 &&
1107 		    (error == ENODEV || error == ENXIO) &&
1108 		    td->td_dupfd >= 0) {
1109 			error = dupfdopen(td, fdp, td->td_dupfd, flags, error,
1110 			    &indx);
1111 			if (error == 0)
1112 				goto success;
1113 		}
1114 
1115 		goto bad;
1116 	}
1117 	td->td_dupfd = 0;
1118 	NDFREE(&nd, NDF_ONLY_PNBUF);
1119 	vp = nd.ni_vp;
1120 
1121 	/*
1122 	 * Store the vnode, for any f_type. Typically, the vnode use
1123 	 * count is decremented by direct call to vn_closefile() for
1124 	 * files that switched type in the cdevsw fdopen() method.
1125 	 */
1126 	fp->f_vnode = vp;
1127 	/*
1128 	 * If the file wasn't claimed by devfs bind it to the normal
1129 	 * vnode operations here.
1130 	 */
1131 	if (fp->f_ops == &badfileops) {
1132 		KASSERT(vp->v_type != VFIFO, ("Unexpected fifo."));
1133 		fp->f_seqcount = 1;
1134 		finit(fp, (flags & FMASK) | (fp->f_flag & FHASLOCK), DTYPE_VNODE,
1135 		    vp, &vnops);
1136 	}
1137 
1138 	VOP_UNLOCK(vp, 0);
1139 	if (flags & O_TRUNC) {
1140 		error = fo_truncate(fp, 0, td->td_ucred, td);
1141 		if (error)
1142 			goto bad;
1143 	}
1144 success:
1145 	/*
1146 	 * If we haven't already installed the FD (for dupfdopen), do so now.
1147 	 */
1148 	if (indx == -1) {
1149 		struct filecaps *fcaps;
1150 
1151 #ifdef CAPABILITIES
1152 		if (nd.ni_strictrelative == 1)
1153 			fcaps = &nd.ni_filecaps;
1154 		else
1155 #endif
1156 			fcaps = NULL;
1157 		error = finstall(td, fp, &indx, flags, fcaps);
1158 		/* On success finstall() consumes fcaps. */
1159 		if (error != 0) {
1160 			filecaps_free(&nd.ni_filecaps);
1161 			goto bad;
1162 		}
1163 	} else {
1164 		filecaps_free(&nd.ni_filecaps);
1165 	}
1166 
1167 	/*
1168 	 * Release our private reference, leaving the one associated with
1169 	 * the descriptor table intact.
1170 	 */
1171 	fdrop(fp, td);
1172 	td->td_retval[0] = indx;
1173 	return (0);
1174 bad:
1175 	KASSERT(indx == -1, ("indx=%d, should be -1", indx));
1176 	fdrop(fp, td);
1177 	return (error);
1178 }
1179 
1180 #ifdef COMPAT_43
1181 /*
1182  * Create a file.
1183  */
1184 #ifndef _SYS_SYSPROTO_H_
1185 struct ocreat_args {
1186 	char	*path;
1187 	int	mode;
1188 };
1189 #endif
1190 int
1191 ocreat(td, uap)
1192 	struct thread *td;
1193 	register struct ocreat_args /* {
1194 		char *path;
1195 		int mode;
1196 	} */ *uap;
1197 {
1198 
1199 	return (kern_open(td, uap->path, UIO_USERSPACE,
1200 	    O_WRONLY | O_CREAT | O_TRUNC, uap->mode));
1201 }
1202 #endif /* COMPAT_43 */
1203 
1204 /*
1205  * Create a special file.
1206  */
1207 #ifndef _SYS_SYSPROTO_H_
1208 struct mknod_args {
1209 	char	*path;
1210 	int	mode;
1211 	int	dev;
1212 };
1213 #endif
1214 int
1215 sys_mknod(td, uap)
1216 	struct thread *td;
1217 	register struct mknod_args /* {
1218 		char *path;
1219 		int mode;
1220 		int dev;
1221 	} */ *uap;
1222 {
1223 
1224 	return (kern_mknod(td, uap->path, UIO_USERSPACE, uap->mode, uap->dev));
1225 }
1226 
1227 #ifndef _SYS_SYSPROTO_H_
1228 struct mknodat_args {
1229 	int	fd;
1230 	char	*path;
1231 	mode_t	mode;
1232 	dev_t	dev;
1233 };
1234 #endif
1235 int
1236 sys_mknodat(struct thread *td, struct mknodat_args *uap)
1237 {
1238 
1239 	return (kern_mknodat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode,
1240 	    uap->dev));
1241 }
1242 
1243 int
1244 kern_mknod(struct thread *td, char *path, enum uio_seg pathseg, int mode,
1245     int dev)
1246 {
1247 
1248 	return (kern_mknodat(td, AT_FDCWD, path, pathseg, mode, dev));
1249 }
1250 
1251 int
1252 kern_mknodat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1253     int mode, int dev)
1254 {
1255 	struct vnode *vp;
1256 	struct mount *mp;
1257 	struct vattr vattr;
1258 	int error;
1259 	int whiteout = 0;
1260 	struct nameidata nd;
1261 
1262 	AUDIT_ARG_MODE(mode);
1263 	AUDIT_ARG_DEV(dev);
1264 	switch (mode & S_IFMT) {
1265 	case S_IFCHR:
1266 	case S_IFBLK:
1267 		error = priv_check(td, PRIV_VFS_MKNOD_DEV);
1268 		break;
1269 	case S_IFMT:
1270 		error = priv_check(td, PRIV_VFS_MKNOD_BAD);
1271 		break;
1272 	case S_IFWHT:
1273 		error = priv_check(td, PRIV_VFS_MKNOD_WHT);
1274 		break;
1275 	case S_IFIFO:
1276 		if (dev == 0)
1277 			return (kern_mkfifoat(td, fd, path, pathseg, mode));
1278 		/* FALLTHROUGH */
1279 	default:
1280 		error = EINVAL;
1281 		break;
1282 	}
1283 	if (error)
1284 		return (error);
1285 restart:
1286 	bwillwrite();
1287 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1,
1288 	    pathseg, path, fd, CAP_MKNODAT, td);
1289 	if ((error = namei(&nd)) != 0)
1290 		return (error);
1291 	vp = nd.ni_vp;
1292 	if (vp != NULL) {
1293 		NDFREE(&nd, NDF_ONLY_PNBUF);
1294 		if (vp == nd.ni_dvp)
1295 			vrele(nd.ni_dvp);
1296 		else
1297 			vput(nd.ni_dvp);
1298 		vrele(vp);
1299 		return (EEXIST);
1300 	} else {
1301 		VATTR_NULL(&vattr);
1302 		vattr.va_mode = (mode & ALLPERMS) &
1303 		    ~td->td_proc->p_fd->fd_cmask;
1304 		vattr.va_rdev = dev;
1305 		whiteout = 0;
1306 
1307 		switch (mode & S_IFMT) {
1308 		case S_IFMT:	/* used by badsect to flag bad sectors */
1309 			vattr.va_type = VBAD;
1310 			break;
1311 		case S_IFCHR:
1312 			vattr.va_type = VCHR;
1313 			break;
1314 		case S_IFBLK:
1315 			vattr.va_type = VBLK;
1316 			break;
1317 		case S_IFWHT:
1318 			whiteout = 1;
1319 			break;
1320 		default:
1321 			panic("kern_mknod: invalid mode");
1322 		}
1323 	}
1324 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1325 		NDFREE(&nd, NDF_ONLY_PNBUF);
1326 		vput(nd.ni_dvp);
1327 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1328 			return (error);
1329 		goto restart;
1330 	}
1331 #ifdef MAC
1332 	if (error == 0 && !whiteout)
1333 		error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp,
1334 		    &nd.ni_cnd, &vattr);
1335 #endif
1336 	if (!error) {
1337 		if (whiteout)
1338 			error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE);
1339 		else {
1340 			error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp,
1341 						&nd.ni_cnd, &vattr);
1342 			if (error == 0)
1343 				vput(nd.ni_vp);
1344 		}
1345 	}
1346 	NDFREE(&nd, NDF_ONLY_PNBUF);
1347 	vput(nd.ni_dvp);
1348 	vn_finished_write(mp);
1349 	return (error);
1350 }
1351 
1352 /*
1353  * Create a named pipe.
1354  */
1355 #ifndef _SYS_SYSPROTO_H_
1356 struct mkfifo_args {
1357 	char	*path;
1358 	int	mode;
1359 };
1360 #endif
1361 int
1362 sys_mkfifo(td, uap)
1363 	struct thread *td;
1364 	register struct mkfifo_args /* {
1365 		char *path;
1366 		int mode;
1367 	} */ *uap;
1368 {
1369 
1370 	return (kern_mkfifo(td, uap->path, UIO_USERSPACE, uap->mode));
1371 }
1372 
1373 #ifndef _SYS_SYSPROTO_H_
1374 struct mkfifoat_args {
1375 	int	fd;
1376 	char	*path;
1377 	mode_t	mode;
1378 };
1379 #endif
1380 int
1381 sys_mkfifoat(struct thread *td, struct mkfifoat_args *uap)
1382 {
1383 
1384 	return (kern_mkfifoat(td, uap->fd, uap->path, UIO_USERSPACE,
1385 	    uap->mode));
1386 }
1387 
1388 int
1389 kern_mkfifo(struct thread *td, char *path, enum uio_seg pathseg, int mode)
1390 {
1391 
1392 	return (kern_mkfifoat(td, AT_FDCWD, path, pathseg, mode));
1393 }
1394 
1395 int
1396 kern_mkfifoat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1397     int mode)
1398 {
1399 	struct mount *mp;
1400 	struct vattr vattr;
1401 	int error;
1402 	struct nameidata nd;
1403 
1404 	AUDIT_ARG_MODE(mode);
1405 restart:
1406 	bwillwrite();
1407 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1,
1408 	    pathseg, path, fd, CAP_MKFIFOAT, td);
1409 	if ((error = namei(&nd)) != 0)
1410 		return (error);
1411 	if (nd.ni_vp != NULL) {
1412 		NDFREE(&nd, NDF_ONLY_PNBUF);
1413 		if (nd.ni_vp == nd.ni_dvp)
1414 			vrele(nd.ni_dvp);
1415 		else
1416 			vput(nd.ni_dvp);
1417 		vrele(nd.ni_vp);
1418 		return (EEXIST);
1419 	}
1420 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1421 		NDFREE(&nd, NDF_ONLY_PNBUF);
1422 		vput(nd.ni_dvp);
1423 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1424 			return (error);
1425 		goto restart;
1426 	}
1427 	VATTR_NULL(&vattr);
1428 	vattr.va_type = VFIFO;
1429 	vattr.va_mode = (mode & ALLPERMS) & ~td->td_proc->p_fd->fd_cmask;
1430 #ifdef MAC
1431 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1432 	    &vattr);
1433 	if (error)
1434 		goto out;
1435 #endif
1436 	error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
1437 	if (error == 0)
1438 		vput(nd.ni_vp);
1439 #ifdef MAC
1440 out:
1441 #endif
1442 	vput(nd.ni_dvp);
1443 	vn_finished_write(mp);
1444 	NDFREE(&nd, NDF_ONLY_PNBUF);
1445 	return (error);
1446 }
1447 
1448 /*
1449  * Make a hard file link.
1450  */
1451 #ifndef _SYS_SYSPROTO_H_
1452 struct link_args {
1453 	char	*path;
1454 	char	*link;
1455 };
1456 #endif
1457 int
1458 sys_link(td, uap)
1459 	struct thread *td;
1460 	register struct link_args /* {
1461 		char *path;
1462 		char *link;
1463 	} */ *uap;
1464 {
1465 
1466 	return (kern_link(td, uap->path, uap->link, UIO_USERSPACE));
1467 }
1468 
1469 #ifndef _SYS_SYSPROTO_H_
1470 struct linkat_args {
1471 	int	fd1;
1472 	char	*path1;
1473 	int	fd2;
1474 	char	*path2;
1475 	int	flag;
1476 };
1477 #endif
1478 int
1479 sys_linkat(struct thread *td, struct linkat_args *uap)
1480 {
1481 	int flag;
1482 
1483 	flag = uap->flag;
1484 	if (flag & ~AT_SYMLINK_FOLLOW)
1485 		return (EINVAL);
1486 
1487 	return (kern_linkat(td, uap->fd1, uap->fd2, uap->path1, uap->path2,
1488 	    UIO_USERSPACE, (flag & AT_SYMLINK_FOLLOW) ? FOLLOW : NOFOLLOW));
1489 }
1490 
1491 int hardlink_check_uid = 0;
1492 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_uid, CTLFLAG_RW,
1493     &hardlink_check_uid, 0,
1494     "Unprivileged processes cannot create hard links to files owned by other "
1495     "users");
1496 static int hardlink_check_gid = 0;
1497 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_gid, CTLFLAG_RW,
1498     &hardlink_check_gid, 0,
1499     "Unprivileged processes cannot create hard links to files owned by other "
1500     "groups");
1501 
1502 static int
1503 can_hardlink(struct vnode *vp, struct ucred *cred)
1504 {
1505 	struct vattr va;
1506 	int error;
1507 
1508 	if (!hardlink_check_uid && !hardlink_check_gid)
1509 		return (0);
1510 
1511 	error = VOP_GETATTR(vp, &va, cred);
1512 	if (error != 0)
1513 		return (error);
1514 
1515 	if (hardlink_check_uid && cred->cr_uid != va.va_uid) {
1516 		error = priv_check_cred(cred, PRIV_VFS_LINK, 0);
1517 		if (error)
1518 			return (error);
1519 	}
1520 
1521 	if (hardlink_check_gid && !groupmember(va.va_gid, cred)) {
1522 		error = priv_check_cred(cred, PRIV_VFS_LINK, 0);
1523 		if (error)
1524 			return (error);
1525 	}
1526 
1527 	return (0);
1528 }
1529 
1530 int
1531 kern_link(struct thread *td, char *path, char *link, enum uio_seg segflg)
1532 {
1533 
1534 	return (kern_linkat(td, AT_FDCWD, AT_FDCWD, path,link, segflg, FOLLOW));
1535 }
1536 
1537 int
1538 kern_linkat(struct thread *td, int fd1, int fd2, char *path1, char *path2,
1539     enum uio_seg segflg, int follow)
1540 {
1541 	struct vnode *vp;
1542 	struct mount *mp;
1543 	struct nameidata nd;
1544 	int error;
1545 
1546 	bwillwrite();
1547 	NDINIT_AT(&nd, LOOKUP, follow | AUDITVNODE1, segflg, path1, fd1, td);
1548 
1549 	if ((error = namei(&nd)) != 0)
1550 		return (error);
1551 	NDFREE(&nd, NDF_ONLY_PNBUF);
1552 	vp = nd.ni_vp;
1553 	if (vp->v_type == VDIR) {
1554 		vrele(vp);
1555 		return (EPERM);		/* POSIX */
1556 	}
1557 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) {
1558 		vrele(vp);
1559 		return (error);
1560 	}
1561 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE2,
1562 	    segflg, path2, fd2, CAP_LINKAT, td);
1563 	if ((error = namei(&nd)) == 0) {
1564 		if (nd.ni_vp != NULL) {
1565 			if (nd.ni_dvp == nd.ni_vp)
1566 				vrele(nd.ni_dvp);
1567 			else
1568 				vput(nd.ni_dvp);
1569 			vrele(nd.ni_vp);
1570 			error = EEXIST;
1571 		} else if ((error = vn_lock(vp, LK_EXCLUSIVE | LK_RETRY))
1572 		    == 0) {
1573 			error = can_hardlink(vp, td->td_ucred);
1574 			if (error == 0)
1575 #ifdef MAC
1576 				error = mac_vnode_check_link(td->td_ucred,
1577 				    nd.ni_dvp, vp, &nd.ni_cnd);
1578 			if (error == 0)
1579 #endif
1580 				error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
1581 			VOP_UNLOCK(vp, 0);
1582 			vput(nd.ni_dvp);
1583 		}
1584 		NDFREE(&nd, NDF_ONLY_PNBUF);
1585 	}
1586 	vrele(vp);
1587 	vn_finished_write(mp);
1588 	return (error);
1589 }
1590 
1591 /*
1592  * Make a symbolic link.
1593  */
1594 #ifndef _SYS_SYSPROTO_H_
1595 struct symlink_args {
1596 	char	*path;
1597 	char	*link;
1598 };
1599 #endif
1600 int
1601 sys_symlink(td, uap)
1602 	struct thread *td;
1603 	register struct symlink_args /* {
1604 		char *path;
1605 		char *link;
1606 	} */ *uap;
1607 {
1608 
1609 	return (kern_symlink(td, uap->path, uap->link, UIO_USERSPACE));
1610 }
1611 
1612 #ifndef _SYS_SYSPROTO_H_
1613 struct symlinkat_args {
1614 	char	*path;
1615 	int	fd;
1616 	char	*path2;
1617 };
1618 #endif
1619 int
1620 sys_symlinkat(struct thread *td, struct symlinkat_args *uap)
1621 {
1622 
1623 	return (kern_symlinkat(td, uap->path1, uap->fd, uap->path2,
1624 	    UIO_USERSPACE));
1625 }
1626 
1627 int
1628 kern_symlink(struct thread *td, char *path, char *link, enum uio_seg segflg)
1629 {
1630 
1631 	return (kern_symlinkat(td, path, AT_FDCWD, link, segflg));
1632 }
1633 
1634 int
1635 kern_symlinkat(struct thread *td, char *path1, int fd, char *path2,
1636     enum uio_seg segflg)
1637 {
1638 	struct mount *mp;
1639 	struct vattr vattr;
1640 	char *syspath;
1641 	int error;
1642 	struct nameidata nd;
1643 
1644 	if (segflg == UIO_SYSSPACE) {
1645 		syspath = path1;
1646 	} else {
1647 		syspath = uma_zalloc(namei_zone, M_WAITOK);
1648 		if ((error = copyinstr(path1, syspath, MAXPATHLEN, NULL)) != 0)
1649 			goto out;
1650 	}
1651 	AUDIT_ARG_TEXT(syspath);
1652 restart:
1653 	bwillwrite();
1654 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1,
1655 	    segflg, path2, fd, CAP_SYMLINKAT, td);
1656 	if ((error = namei(&nd)) != 0)
1657 		goto out;
1658 	if (nd.ni_vp) {
1659 		NDFREE(&nd, NDF_ONLY_PNBUF);
1660 		if (nd.ni_vp == nd.ni_dvp)
1661 			vrele(nd.ni_dvp);
1662 		else
1663 			vput(nd.ni_dvp);
1664 		vrele(nd.ni_vp);
1665 		error = EEXIST;
1666 		goto out;
1667 	}
1668 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1669 		NDFREE(&nd, NDF_ONLY_PNBUF);
1670 		vput(nd.ni_dvp);
1671 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1672 			goto out;
1673 		goto restart;
1674 	}
1675 	VATTR_NULL(&vattr);
1676 	vattr.va_mode = ACCESSPERMS &~ td->td_proc->p_fd->fd_cmask;
1677 #ifdef MAC
1678 	vattr.va_type = VLNK;
1679 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1680 	    &vattr);
1681 	if (error)
1682 		goto out2;
1683 #endif
1684 	error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, syspath);
1685 	if (error == 0)
1686 		vput(nd.ni_vp);
1687 #ifdef MAC
1688 out2:
1689 #endif
1690 	NDFREE(&nd, NDF_ONLY_PNBUF);
1691 	vput(nd.ni_dvp);
1692 	vn_finished_write(mp);
1693 out:
1694 	if (segflg != UIO_SYSSPACE)
1695 		uma_zfree(namei_zone, syspath);
1696 	return (error);
1697 }
1698 
1699 /*
1700  * Delete a whiteout from the filesystem.
1701  */
1702 int
1703 sys_undelete(td, uap)
1704 	struct thread *td;
1705 	register struct undelete_args /* {
1706 		char *path;
1707 	} */ *uap;
1708 {
1709 	int error;
1710 	struct mount *mp;
1711 	struct nameidata nd;
1712 
1713 restart:
1714 	bwillwrite();
1715 	NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | AUDITVNODE1,
1716 	    UIO_USERSPACE, uap->path, td);
1717 	error = namei(&nd);
1718 	if (error)
1719 		return (error);
1720 
1721 	if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) {
1722 		NDFREE(&nd, NDF_ONLY_PNBUF);
1723 		if (nd.ni_vp == nd.ni_dvp)
1724 			vrele(nd.ni_dvp);
1725 		else
1726 			vput(nd.ni_dvp);
1727 		if (nd.ni_vp)
1728 			vrele(nd.ni_vp);
1729 		return (EEXIST);
1730 	}
1731 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1732 		NDFREE(&nd, NDF_ONLY_PNBUF);
1733 		vput(nd.ni_dvp);
1734 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1735 			return (error);
1736 		goto restart;
1737 	}
1738 	error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE);
1739 	NDFREE(&nd, NDF_ONLY_PNBUF);
1740 	vput(nd.ni_dvp);
1741 	vn_finished_write(mp);
1742 	return (error);
1743 }
1744 
1745 /*
1746  * Delete a name from the filesystem.
1747  */
1748 #ifndef _SYS_SYSPROTO_H_
1749 struct unlink_args {
1750 	char	*path;
1751 };
1752 #endif
1753 int
1754 sys_unlink(td, uap)
1755 	struct thread *td;
1756 	struct unlink_args /* {
1757 		char *path;
1758 	} */ *uap;
1759 {
1760 
1761 	return (kern_unlink(td, uap->path, UIO_USERSPACE));
1762 }
1763 
1764 #ifndef _SYS_SYSPROTO_H_
1765 struct unlinkat_args {
1766 	int	fd;
1767 	char	*path;
1768 	int	flag;
1769 };
1770 #endif
1771 int
1772 sys_unlinkat(struct thread *td, struct unlinkat_args *uap)
1773 {
1774 	int flag = uap->flag;
1775 	int fd = uap->fd;
1776 	char *path = uap->path;
1777 
1778 	if (flag & ~AT_REMOVEDIR)
1779 		return (EINVAL);
1780 
1781 	if (flag & AT_REMOVEDIR)
1782 		return (kern_rmdirat(td, fd, path, UIO_USERSPACE));
1783 	else
1784 		return (kern_unlinkat(td, fd, path, UIO_USERSPACE, 0));
1785 }
1786 
1787 int
1788 kern_unlink(struct thread *td, char *path, enum uio_seg pathseg)
1789 {
1790 
1791 	return (kern_unlinkat(td, AT_FDCWD, path, pathseg, 0));
1792 }
1793 
1794 int
1795 kern_unlinkat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1796     ino_t oldinum)
1797 {
1798 	struct mount *mp;
1799 	struct vnode *vp;
1800 	int error;
1801 	struct nameidata nd;
1802 	struct stat sb;
1803 
1804 restart:
1805 	bwillwrite();
1806 	NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1,
1807 	    pathseg, path, fd, CAP_UNLINKAT, td);
1808 	if ((error = namei(&nd)) != 0)
1809 		return (error == EINVAL ? EPERM : error);
1810 	vp = nd.ni_vp;
1811 	if (vp->v_type == VDIR && oldinum == 0) {
1812 		error = EPERM;		/* POSIX */
1813 	} else if (oldinum != 0 &&
1814 		  ((error = vn_stat(vp, &sb, td->td_ucred, NOCRED, td)) == 0) &&
1815 		  sb.st_ino != oldinum) {
1816 			error = EIDRM;	/* Identifier removed */
1817 	} else {
1818 		/*
1819 		 * The root of a mounted filesystem cannot be deleted.
1820 		 *
1821 		 * XXX: can this only be a VDIR case?
1822 		 */
1823 		if (vp->v_vflag & VV_ROOT)
1824 			error = EBUSY;
1825 	}
1826 	if (error == 0) {
1827 		if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1828 			NDFREE(&nd, NDF_ONLY_PNBUF);
1829 			vput(nd.ni_dvp);
1830 			if (vp == nd.ni_dvp)
1831 				vrele(vp);
1832 			else
1833 				vput(vp);
1834 			if ((error = vn_start_write(NULL, &mp,
1835 			    V_XSLEEP | PCATCH)) != 0)
1836 				return (error);
1837 			goto restart;
1838 		}
1839 #ifdef MAC
1840 		error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
1841 		    &nd.ni_cnd);
1842 		if (error)
1843 			goto out;
1844 #endif
1845 		vfs_notify_upper(vp, VFS_NOTIFY_UPPER_UNLINK);
1846 		error = VOP_REMOVE(nd.ni_dvp, vp, &nd.ni_cnd);
1847 #ifdef MAC
1848 out:
1849 #endif
1850 		vn_finished_write(mp);
1851 	}
1852 	NDFREE(&nd, NDF_ONLY_PNBUF);
1853 	vput(nd.ni_dvp);
1854 	if (vp == nd.ni_dvp)
1855 		vrele(vp);
1856 	else
1857 		vput(vp);
1858 	return (error);
1859 }
1860 
1861 /*
1862  * Reposition read/write file offset.
1863  */
1864 #ifndef _SYS_SYSPROTO_H_
1865 struct lseek_args {
1866 	int	fd;
1867 	int	pad;
1868 	off_t	offset;
1869 	int	whence;
1870 };
1871 #endif
1872 int
1873 sys_lseek(td, uap)
1874 	struct thread *td;
1875 	register struct lseek_args /* {
1876 		int fd;
1877 		int pad;
1878 		off_t offset;
1879 		int whence;
1880 	} */ *uap;
1881 {
1882 	struct file *fp;
1883 	int error;
1884 
1885 	AUDIT_ARG_FD(uap->fd);
1886 	if ((error = fget(td, uap->fd, CAP_SEEK, &fp)) != 0)
1887 		return (error);
1888 	error = (fp->f_ops->fo_flags & DFLAG_SEEKABLE) != 0 ?
1889 	    fo_seek(fp, uap->offset, uap->whence, td) : ESPIPE;
1890 	fdrop(fp, td);
1891 	return (error);
1892 }
1893 
1894 #if defined(COMPAT_43)
1895 /*
1896  * Reposition read/write file offset.
1897  */
1898 #ifndef _SYS_SYSPROTO_H_
1899 struct olseek_args {
1900 	int	fd;
1901 	long	offset;
1902 	int	whence;
1903 };
1904 #endif
1905 int
1906 olseek(td, uap)
1907 	struct thread *td;
1908 	register struct olseek_args /* {
1909 		int fd;
1910 		long offset;
1911 		int whence;
1912 	} */ *uap;
1913 {
1914 	struct lseek_args /* {
1915 		int fd;
1916 		int pad;
1917 		off_t offset;
1918 		int whence;
1919 	} */ nuap;
1920 
1921 	nuap.fd = uap->fd;
1922 	nuap.offset = uap->offset;
1923 	nuap.whence = uap->whence;
1924 	return (sys_lseek(td, &nuap));
1925 }
1926 #endif /* COMPAT_43 */
1927 
1928 /* Version with the 'pad' argument */
1929 int
1930 freebsd6_lseek(td, uap)
1931 	struct thread *td;
1932 	register struct freebsd6_lseek_args *uap;
1933 {
1934 	struct lseek_args ouap;
1935 
1936 	ouap.fd = uap->fd;
1937 	ouap.offset = uap->offset;
1938 	ouap.whence = uap->whence;
1939 	return (sys_lseek(td, &ouap));
1940 }
1941 
1942 /*
1943  * Check access permissions using passed credentials.
1944  */
1945 static int
1946 vn_access(vp, user_flags, cred, td)
1947 	struct vnode	*vp;
1948 	int		user_flags;
1949 	struct ucred	*cred;
1950 	struct thread	*td;
1951 {
1952 	int error;
1953 	accmode_t accmode;
1954 
1955 	/* Flags == 0 means only check for existence. */
1956 	error = 0;
1957 	if (user_flags) {
1958 		accmode = 0;
1959 		if (user_flags & R_OK)
1960 			accmode |= VREAD;
1961 		if (user_flags & W_OK)
1962 			accmode |= VWRITE;
1963 		if (user_flags & X_OK)
1964 			accmode |= VEXEC;
1965 #ifdef MAC
1966 		error = mac_vnode_check_access(cred, vp, accmode);
1967 		if (error)
1968 			return (error);
1969 #endif
1970 		if ((accmode & VWRITE) == 0 || (error = vn_writechk(vp)) == 0)
1971 			error = VOP_ACCESS(vp, accmode, cred, td);
1972 	}
1973 	return (error);
1974 }
1975 
1976 /*
1977  * Check access permissions using "real" credentials.
1978  */
1979 #ifndef _SYS_SYSPROTO_H_
1980 struct access_args {
1981 	char	*path;
1982 	int	amode;
1983 };
1984 #endif
1985 int
1986 sys_access(td, uap)
1987 	struct thread *td;
1988 	register struct access_args /* {
1989 		char *path;
1990 		int amode;
1991 	} */ *uap;
1992 {
1993 
1994 	return (kern_access(td, uap->path, UIO_USERSPACE, uap->amode));
1995 }
1996 
1997 #ifndef _SYS_SYSPROTO_H_
1998 struct faccessat_args {
1999 	int	dirfd;
2000 	char	*path;
2001 	int	amode;
2002 	int	flag;
2003 }
2004 #endif
2005 int
2006 sys_faccessat(struct thread *td, struct faccessat_args *uap)
2007 {
2008 
2009 	if (uap->flag & ~AT_EACCESS)
2010 		return (EINVAL);
2011 	return (kern_accessat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
2012 	    uap->amode));
2013 }
2014 
2015 int
2016 kern_access(struct thread *td, char *path, enum uio_seg pathseg, int amode)
2017 {
2018 
2019 	return (kern_accessat(td, AT_FDCWD, path, pathseg, 0, amode));
2020 }
2021 
2022 int
2023 kern_accessat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2024     int flag, int amode)
2025 {
2026 	struct ucred *cred, *tmpcred;
2027 	struct vnode *vp;
2028 	struct nameidata nd;
2029 	int error;
2030 
2031 	/*
2032 	 * Create and modify a temporary credential instead of one that
2033 	 * is potentially shared.
2034 	 */
2035 	if (!(flag & AT_EACCESS)) {
2036 		cred = td->td_ucred;
2037 		tmpcred = crdup(cred);
2038 		tmpcred->cr_uid = cred->cr_ruid;
2039 		tmpcred->cr_groups[0] = cred->cr_rgid;
2040 		td->td_ucred = tmpcred;
2041 	} else
2042 		cred = tmpcred = td->td_ucred;
2043 	AUDIT_ARG_VALUE(amode);
2044 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF |
2045 	    AUDITVNODE1, pathseg, path, fd, CAP_FSTAT, td);
2046 	if ((error = namei(&nd)) != 0)
2047 		goto out1;
2048 	vp = nd.ni_vp;
2049 
2050 	error = vn_access(vp, amode, tmpcred, td);
2051 	NDFREE(&nd, NDF_ONLY_PNBUF);
2052 	vput(vp);
2053 out1:
2054 	if (!(flag & AT_EACCESS)) {
2055 		td->td_ucred = cred;
2056 		crfree(tmpcred);
2057 	}
2058 	return (error);
2059 }
2060 
2061 /*
2062  * Check access permissions using "effective" credentials.
2063  */
2064 #ifndef _SYS_SYSPROTO_H_
2065 struct eaccess_args {
2066 	char	*path;
2067 	int	amode;
2068 };
2069 #endif
2070 int
2071 sys_eaccess(td, uap)
2072 	struct thread *td;
2073 	register struct eaccess_args /* {
2074 		char *path;
2075 		int amode;
2076 	} */ *uap;
2077 {
2078 
2079 	return (kern_eaccess(td, uap->path, UIO_USERSPACE, uap->amode));
2080 }
2081 
2082 int
2083 kern_eaccess(struct thread *td, char *path, enum uio_seg pathseg, int amode)
2084 {
2085 
2086 	return (kern_accessat(td, AT_FDCWD, path, pathseg, AT_EACCESS, amode));
2087 }
2088 
2089 #if defined(COMPAT_43)
2090 /*
2091  * Get file status; this version follows links.
2092  */
2093 #ifndef _SYS_SYSPROTO_H_
2094 struct ostat_args {
2095 	char	*path;
2096 	struct ostat *ub;
2097 };
2098 #endif
2099 int
2100 ostat(td, uap)
2101 	struct thread *td;
2102 	register struct ostat_args /* {
2103 		char *path;
2104 		struct ostat *ub;
2105 	} */ *uap;
2106 {
2107 	struct stat sb;
2108 	struct ostat osb;
2109 	int error;
2110 
2111 	error = kern_stat(td, uap->path, UIO_USERSPACE, &sb);
2112 	if (error)
2113 		return (error);
2114 	cvtstat(&sb, &osb);
2115 	error = copyout(&osb, uap->ub, sizeof (osb));
2116 	return (error);
2117 }
2118 
2119 /*
2120  * Get file status; this version does not follow links.
2121  */
2122 #ifndef _SYS_SYSPROTO_H_
2123 struct olstat_args {
2124 	char	*path;
2125 	struct ostat *ub;
2126 };
2127 #endif
2128 int
2129 olstat(td, uap)
2130 	struct thread *td;
2131 	register struct olstat_args /* {
2132 		char *path;
2133 		struct ostat *ub;
2134 	} */ *uap;
2135 {
2136 	struct stat sb;
2137 	struct ostat osb;
2138 	int error;
2139 
2140 	error = kern_lstat(td, uap->path, UIO_USERSPACE, &sb);
2141 	if (error)
2142 		return (error);
2143 	cvtstat(&sb, &osb);
2144 	error = copyout(&osb, uap->ub, sizeof (osb));
2145 	return (error);
2146 }
2147 
2148 /*
2149  * Convert from an old to a new stat structure.
2150  */
2151 void
2152 cvtstat(st, ost)
2153 	struct stat *st;
2154 	struct ostat *ost;
2155 {
2156 
2157 	ost->st_dev = st->st_dev;
2158 	ost->st_ino = st->st_ino;
2159 	ost->st_mode = st->st_mode;
2160 	ost->st_nlink = st->st_nlink;
2161 	ost->st_uid = st->st_uid;
2162 	ost->st_gid = st->st_gid;
2163 	ost->st_rdev = st->st_rdev;
2164 	if (st->st_size < (quad_t)1 << 32)
2165 		ost->st_size = st->st_size;
2166 	else
2167 		ost->st_size = -2;
2168 	ost->st_atim = st->st_atim;
2169 	ost->st_mtim = st->st_mtim;
2170 	ost->st_ctim = st->st_ctim;
2171 	ost->st_blksize = st->st_blksize;
2172 	ost->st_blocks = st->st_blocks;
2173 	ost->st_flags = st->st_flags;
2174 	ost->st_gen = st->st_gen;
2175 }
2176 #endif /* COMPAT_43 */
2177 
2178 /*
2179  * Get file status; this version follows links.
2180  */
2181 #ifndef _SYS_SYSPROTO_H_
2182 struct stat_args {
2183 	char	*path;
2184 	struct stat *ub;
2185 };
2186 #endif
2187 int
2188 sys_stat(td, uap)
2189 	struct thread *td;
2190 	register struct stat_args /* {
2191 		char *path;
2192 		struct stat *ub;
2193 	} */ *uap;
2194 {
2195 	struct stat sb;
2196 	int error;
2197 
2198 	error = kern_stat(td, uap->path, UIO_USERSPACE, &sb);
2199 	if (error == 0)
2200 		error = copyout(&sb, uap->ub, sizeof (sb));
2201 	return (error);
2202 }
2203 
2204 #ifndef _SYS_SYSPROTO_H_
2205 struct fstatat_args {
2206 	int	fd;
2207 	char	*path;
2208 	struct stat	*buf;
2209 	int	flag;
2210 }
2211 #endif
2212 int
2213 sys_fstatat(struct thread *td, struct fstatat_args *uap)
2214 {
2215 	struct stat sb;
2216 	int error;
2217 
2218 	error = kern_statat(td, uap->flag, uap->fd, uap->path,
2219 	    UIO_USERSPACE, &sb);
2220 	if (error == 0)
2221 		error = copyout(&sb, uap->buf, sizeof (sb));
2222 	return (error);
2223 }
2224 
2225 int
2226 kern_stat(struct thread *td, char *path, enum uio_seg pathseg, struct stat *sbp)
2227 {
2228 
2229 	return (kern_statat(td, 0, AT_FDCWD, path, pathseg, sbp));
2230 }
2231 
2232 int
2233 kern_statat(struct thread *td, int flag, int fd, char *path,
2234     enum uio_seg pathseg, struct stat *sbp)
2235 {
2236 
2237 	return (kern_statat_vnhook(td, flag, fd, path, pathseg, sbp, NULL));
2238 }
2239 
2240 int
2241 kern_statat_vnhook(struct thread *td, int flag, int fd, char *path,
2242     enum uio_seg pathseg, struct stat *sbp,
2243     void (*hook)(struct vnode *vp, struct stat *sbp))
2244 {
2245 	struct nameidata nd;
2246 	struct stat sb;
2247 	int error;
2248 
2249 	if (flag & ~AT_SYMLINK_NOFOLLOW)
2250 		return (EINVAL);
2251 
2252 	NDINIT_ATRIGHTS(&nd, LOOKUP, ((flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW :
2253 	    FOLLOW) | LOCKSHARED | LOCKLEAF | AUDITVNODE1, pathseg, path, fd,
2254 	    CAP_FSTAT, td);
2255 
2256 	if ((error = namei(&nd)) != 0)
2257 		return (error);
2258 	error = vn_stat(nd.ni_vp, &sb, td->td_ucred, NOCRED, td);
2259 	if (!error) {
2260 		SDT_PROBE(vfs, , stat, mode, path, sb.st_mode, 0, 0, 0);
2261 		if (S_ISREG(sb.st_mode))
2262 			SDT_PROBE(vfs, , stat, reg, path, pathseg, 0, 0, 0);
2263 		if (__predict_false(hook != NULL))
2264 			hook(nd.ni_vp, &sb);
2265 	}
2266 	NDFREE(&nd, NDF_ONLY_PNBUF);
2267 	vput(nd.ni_vp);
2268 	if (error)
2269 		return (error);
2270 	*sbp = sb;
2271 #ifdef KTRACE
2272 	if (KTRPOINT(td, KTR_STRUCT))
2273 		ktrstat(&sb);
2274 #endif
2275 	return (0);
2276 }
2277 
2278 /*
2279  * Get file status; this version does not follow links.
2280  */
2281 #ifndef _SYS_SYSPROTO_H_
2282 struct lstat_args {
2283 	char	*path;
2284 	struct stat *ub;
2285 };
2286 #endif
2287 int
2288 sys_lstat(td, uap)
2289 	struct thread *td;
2290 	register struct lstat_args /* {
2291 		char *path;
2292 		struct stat *ub;
2293 	} */ *uap;
2294 {
2295 	struct stat sb;
2296 	int error;
2297 
2298 	error = kern_lstat(td, uap->path, UIO_USERSPACE, &sb);
2299 	if (error == 0)
2300 		error = copyout(&sb, uap->ub, sizeof (sb));
2301 	return (error);
2302 }
2303 
2304 int
2305 kern_lstat(struct thread *td, char *path, enum uio_seg pathseg, struct stat *sbp)
2306 {
2307 
2308 	return (kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, path, pathseg,
2309 	    sbp));
2310 }
2311 
2312 /*
2313  * Implementation of the NetBSD [l]stat() functions.
2314  */
2315 void
2316 cvtnstat(sb, nsb)
2317 	struct stat *sb;
2318 	struct nstat *nsb;
2319 {
2320 	bzero(nsb, sizeof *nsb);
2321 	nsb->st_dev = sb->st_dev;
2322 	nsb->st_ino = sb->st_ino;
2323 	nsb->st_mode = sb->st_mode;
2324 	nsb->st_nlink = sb->st_nlink;
2325 	nsb->st_uid = sb->st_uid;
2326 	nsb->st_gid = sb->st_gid;
2327 	nsb->st_rdev = sb->st_rdev;
2328 	nsb->st_atim = sb->st_atim;
2329 	nsb->st_mtim = sb->st_mtim;
2330 	nsb->st_ctim = sb->st_ctim;
2331 	nsb->st_size = sb->st_size;
2332 	nsb->st_blocks = sb->st_blocks;
2333 	nsb->st_blksize = sb->st_blksize;
2334 	nsb->st_flags = sb->st_flags;
2335 	nsb->st_gen = sb->st_gen;
2336 	nsb->st_birthtim = sb->st_birthtim;
2337 }
2338 
2339 #ifndef _SYS_SYSPROTO_H_
2340 struct nstat_args {
2341 	char	*path;
2342 	struct nstat *ub;
2343 };
2344 #endif
2345 int
2346 sys_nstat(td, uap)
2347 	struct thread *td;
2348 	register struct nstat_args /* {
2349 		char *path;
2350 		struct nstat *ub;
2351 	} */ *uap;
2352 {
2353 	struct stat sb;
2354 	struct nstat nsb;
2355 	int error;
2356 
2357 	error = kern_stat(td, uap->path, UIO_USERSPACE, &sb);
2358 	if (error)
2359 		return (error);
2360 	cvtnstat(&sb, &nsb);
2361 	error = copyout(&nsb, uap->ub, sizeof (nsb));
2362 	return (error);
2363 }
2364 
2365 /*
2366  * NetBSD lstat.  Get file status; this version does not follow links.
2367  */
2368 #ifndef _SYS_SYSPROTO_H_
2369 struct lstat_args {
2370 	char	*path;
2371 	struct stat *ub;
2372 };
2373 #endif
2374 int
2375 sys_nlstat(td, uap)
2376 	struct thread *td;
2377 	register struct nlstat_args /* {
2378 		char *path;
2379 		struct nstat *ub;
2380 	} */ *uap;
2381 {
2382 	struct stat sb;
2383 	struct nstat nsb;
2384 	int error;
2385 
2386 	error = kern_lstat(td, uap->path, UIO_USERSPACE, &sb);
2387 	if (error)
2388 		return (error);
2389 	cvtnstat(&sb, &nsb);
2390 	error = copyout(&nsb, uap->ub, sizeof (nsb));
2391 	return (error);
2392 }
2393 
2394 /*
2395  * Get configurable pathname variables.
2396  */
2397 #ifndef _SYS_SYSPROTO_H_
2398 struct pathconf_args {
2399 	char	*path;
2400 	int	name;
2401 };
2402 #endif
2403 int
2404 sys_pathconf(td, uap)
2405 	struct thread *td;
2406 	register struct pathconf_args /* {
2407 		char *path;
2408 		int name;
2409 	} */ *uap;
2410 {
2411 
2412 	return (kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name, FOLLOW));
2413 }
2414 
2415 #ifndef _SYS_SYSPROTO_H_
2416 struct lpathconf_args {
2417 	char	*path;
2418 	int	name;
2419 };
2420 #endif
2421 int
2422 sys_lpathconf(td, uap)
2423 	struct thread *td;
2424 	register struct lpathconf_args /* {
2425 		char *path;
2426 		int name;
2427 	} */ *uap;
2428 {
2429 
2430 	return (kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name,
2431 	    NOFOLLOW));
2432 }
2433 
2434 int
2435 kern_pathconf(struct thread *td, char *path, enum uio_seg pathseg, int name,
2436     u_long flags)
2437 {
2438 	struct nameidata nd;
2439 	int error;
2440 
2441 	NDINIT(&nd, LOOKUP, LOCKSHARED | LOCKLEAF | AUDITVNODE1 | flags,
2442 	    pathseg, path, td);
2443 	if ((error = namei(&nd)) != 0)
2444 		return (error);
2445 	NDFREE(&nd, NDF_ONLY_PNBUF);
2446 
2447 	/* If asynchronous I/O is available, it works for all files. */
2448 	if (name == _PC_ASYNC_IO)
2449 		td->td_retval[0] = async_io_version;
2450 	else
2451 		error = VOP_PATHCONF(nd.ni_vp, name, td->td_retval);
2452 	vput(nd.ni_vp);
2453 	return (error);
2454 }
2455 
2456 /*
2457  * Return target name of a symbolic link.
2458  */
2459 #ifndef _SYS_SYSPROTO_H_
2460 struct readlink_args {
2461 	char	*path;
2462 	char	*buf;
2463 	size_t	count;
2464 };
2465 #endif
2466 int
2467 sys_readlink(td, uap)
2468 	struct thread *td;
2469 	register struct readlink_args /* {
2470 		char *path;
2471 		char *buf;
2472 		size_t count;
2473 	} */ *uap;
2474 {
2475 
2476 	return (kern_readlink(td, uap->path, UIO_USERSPACE, uap->buf,
2477 	    UIO_USERSPACE, uap->count));
2478 }
2479 #ifndef _SYS_SYSPROTO_H_
2480 struct readlinkat_args {
2481 	int	fd;
2482 	char	*path;
2483 	char	*buf;
2484 	size_t	bufsize;
2485 };
2486 #endif
2487 int
2488 sys_readlinkat(struct thread *td, struct readlinkat_args *uap)
2489 {
2490 
2491 	return (kern_readlinkat(td, uap->fd, uap->path, UIO_USERSPACE,
2492 	    uap->buf, UIO_USERSPACE, uap->bufsize));
2493 }
2494 
2495 int
2496 kern_readlink(struct thread *td, char *path, enum uio_seg pathseg, char *buf,
2497     enum uio_seg bufseg, size_t count)
2498 {
2499 
2500 	return (kern_readlinkat(td, AT_FDCWD, path, pathseg, buf, bufseg,
2501 	    count));
2502 }
2503 
2504 int
2505 kern_readlinkat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2506     char *buf, enum uio_seg bufseg, size_t count)
2507 {
2508 	struct vnode *vp;
2509 	struct iovec aiov;
2510 	struct uio auio;
2511 	int error;
2512 	struct nameidata nd;
2513 
2514 	if (count > IOSIZE_MAX)
2515 		return (EINVAL);
2516 
2517 	NDINIT_AT(&nd, LOOKUP, NOFOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
2518 	    pathseg, path, fd, td);
2519 
2520 	if ((error = namei(&nd)) != 0)
2521 		return (error);
2522 	NDFREE(&nd, NDF_ONLY_PNBUF);
2523 	vp = nd.ni_vp;
2524 #ifdef MAC
2525 	error = mac_vnode_check_readlink(td->td_ucred, vp);
2526 	if (error) {
2527 		vput(vp);
2528 		return (error);
2529 	}
2530 #endif
2531 	if (vp->v_type != VLNK)
2532 		error = EINVAL;
2533 	else {
2534 		aiov.iov_base = buf;
2535 		aiov.iov_len = count;
2536 		auio.uio_iov = &aiov;
2537 		auio.uio_iovcnt = 1;
2538 		auio.uio_offset = 0;
2539 		auio.uio_rw = UIO_READ;
2540 		auio.uio_segflg = bufseg;
2541 		auio.uio_td = td;
2542 		auio.uio_resid = count;
2543 		error = VOP_READLINK(vp, &auio, td->td_ucred);
2544 	}
2545 	vput(vp);
2546 	td->td_retval[0] = count - auio.uio_resid;
2547 	return (error);
2548 }
2549 
2550 /*
2551  * Common implementation code for chflags() and fchflags().
2552  */
2553 static int
2554 setfflags(td, vp, flags)
2555 	struct thread *td;
2556 	struct vnode *vp;
2557 	u_long flags;
2558 {
2559 	int error;
2560 	struct mount *mp;
2561 	struct vattr vattr;
2562 
2563 	/* We can't support the value matching VNOVAL. */
2564 	if (flags == VNOVAL)
2565 		return (EOPNOTSUPP);
2566 
2567 	/*
2568 	 * Prevent non-root users from setting flags on devices.  When
2569 	 * a device is reused, users can retain ownership of the device
2570 	 * if they are allowed to set flags and programs assume that
2571 	 * chown can't fail when done as root.
2572 	 */
2573 	if (vp->v_type == VCHR || vp->v_type == VBLK) {
2574 		error = priv_check(td, PRIV_VFS_CHFLAGS_DEV);
2575 		if (error)
2576 			return (error);
2577 	}
2578 
2579 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2580 		return (error);
2581 	VATTR_NULL(&vattr);
2582 	vattr.va_flags = flags;
2583 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2584 #ifdef MAC
2585 	error = mac_vnode_check_setflags(td->td_ucred, vp, vattr.va_flags);
2586 	if (error == 0)
2587 #endif
2588 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
2589 	VOP_UNLOCK(vp, 0);
2590 	vn_finished_write(mp);
2591 	return (error);
2592 }
2593 
2594 /*
2595  * Change flags of a file given a path name.
2596  */
2597 #ifndef _SYS_SYSPROTO_H_
2598 struct chflags_args {
2599 	const char *path;
2600 	u_long	flags;
2601 };
2602 #endif
2603 int
2604 sys_chflags(td, uap)
2605 	struct thread *td;
2606 	register struct chflags_args /* {
2607 		const char *path;
2608 		u_long flags;
2609 	} */ *uap;
2610 {
2611 
2612 	return (kern_chflags(td, uap->path, UIO_USERSPACE, uap->flags));
2613 }
2614 
2615 #ifndef _SYS_SYSPROTO_H_
2616 struct chflagsat_args {
2617 	int	fd;
2618 	const char *path;
2619 	u_long	flags;
2620 	int	atflag;
2621 }
2622 #endif
2623 int
2624 sys_chflagsat(struct thread *td, struct chflagsat_args *uap)
2625 {
2626 	int fd = uap->fd;
2627 	const char *path = uap->path;
2628 	u_long flags = uap->flags;
2629 	int atflag = uap->atflag;
2630 
2631 	if (atflag & ~AT_SYMLINK_NOFOLLOW)
2632 		return (EINVAL);
2633 
2634 	return (kern_chflagsat(td, fd, path, UIO_USERSPACE, flags, atflag));
2635 }
2636 
2637 static int
2638 kern_chflags(struct thread *td, const char *path, enum uio_seg pathseg,
2639     u_long flags)
2640 {
2641 
2642 	return (kern_chflagsat(td, AT_FDCWD, path, pathseg, flags, 0));
2643 }
2644 
2645 /*
2646  * Same as chflags() but doesn't follow symlinks.
2647  */
2648 int
2649 sys_lchflags(td, uap)
2650 	struct thread *td;
2651 	register struct lchflags_args /* {
2652 		const char *path;
2653 		u_long flags;
2654 	} */ *uap;
2655 {
2656 
2657 	return (kern_chflagsat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2658 	    uap->flags, AT_SYMLINK_NOFOLLOW));
2659 }
2660 
2661 static int
2662 kern_chflagsat(struct thread *td, int fd, const char *path,
2663     enum uio_seg pathseg, u_long flags, int atflag)
2664 {
2665 	struct nameidata nd;
2666 	int error, follow;
2667 
2668 	AUDIT_ARG_FFLAGS(flags);
2669 	follow = (atflag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
2670 	NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, pathseg, path, fd,
2671 	    CAP_FCHFLAGS, td);
2672 	if ((error = namei(&nd)) != 0)
2673 		return (error);
2674 	NDFREE(&nd, NDF_ONLY_PNBUF);
2675 	error = setfflags(td, nd.ni_vp, flags);
2676 	vrele(nd.ni_vp);
2677 	return (error);
2678 }
2679 
2680 /*
2681  * Change flags of a file given a file descriptor.
2682  */
2683 #ifndef _SYS_SYSPROTO_H_
2684 struct fchflags_args {
2685 	int	fd;
2686 	u_long	flags;
2687 };
2688 #endif
2689 int
2690 sys_fchflags(td, uap)
2691 	struct thread *td;
2692 	register struct fchflags_args /* {
2693 		int fd;
2694 		u_long flags;
2695 	} */ *uap;
2696 {
2697 	struct file *fp;
2698 	int error;
2699 
2700 	AUDIT_ARG_FD(uap->fd);
2701 	AUDIT_ARG_FFLAGS(uap->flags);
2702 	if ((error = getvnode(td->td_proc->p_fd, uap->fd, CAP_FCHFLAGS,
2703 	    &fp)) != 0)
2704 		return (error);
2705 #ifdef AUDIT
2706 	vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
2707 	AUDIT_ARG_VNODE1(fp->f_vnode);
2708 	VOP_UNLOCK(fp->f_vnode, 0);
2709 #endif
2710 	error = setfflags(td, fp->f_vnode, uap->flags);
2711 	fdrop(fp, td);
2712 	return (error);
2713 }
2714 
2715 /*
2716  * Common implementation code for chmod(), lchmod() and fchmod().
2717  */
2718 int
2719 setfmode(td, cred, vp, mode)
2720 	struct thread *td;
2721 	struct ucred *cred;
2722 	struct vnode *vp;
2723 	int mode;
2724 {
2725 	int error;
2726 	struct mount *mp;
2727 	struct vattr vattr;
2728 
2729 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2730 		return (error);
2731 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2732 	VATTR_NULL(&vattr);
2733 	vattr.va_mode = mode & ALLPERMS;
2734 #ifdef MAC
2735 	error = mac_vnode_check_setmode(cred, vp, vattr.va_mode);
2736 	if (error == 0)
2737 #endif
2738 		error = VOP_SETATTR(vp, &vattr, cred);
2739 	VOP_UNLOCK(vp, 0);
2740 	vn_finished_write(mp);
2741 	return (error);
2742 }
2743 
2744 /*
2745  * Change mode of a file given path name.
2746  */
2747 #ifndef _SYS_SYSPROTO_H_
2748 struct chmod_args {
2749 	char	*path;
2750 	int	mode;
2751 };
2752 #endif
2753 int
2754 sys_chmod(td, uap)
2755 	struct thread *td;
2756 	register struct chmod_args /* {
2757 		char *path;
2758 		int mode;
2759 	} */ *uap;
2760 {
2761 
2762 	return (kern_chmod(td, uap->path, UIO_USERSPACE, uap->mode));
2763 }
2764 
2765 #ifndef _SYS_SYSPROTO_H_
2766 struct fchmodat_args {
2767 	int	dirfd;
2768 	char	*path;
2769 	mode_t	mode;
2770 	int	flag;
2771 }
2772 #endif
2773 int
2774 sys_fchmodat(struct thread *td, struct fchmodat_args *uap)
2775 {
2776 	int flag = uap->flag;
2777 	int fd = uap->fd;
2778 	char *path = uap->path;
2779 	mode_t mode = uap->mode;
2780 
2781 	if (flag & ~AT_SYMLINK_NOFOLLOW)
2782 		return (EINVAL);
2783 
2784 	return (kern_fchmodat(td, fd, path, UIO_USERSPACE, mode, flag));
2785 }
2786 
2787 int
2788 kern_chmod(struct thread *td, char *path, enum uio_seg pathseg, int mode)
2789 {
2790 
2791 	return (kern_fchmodat(td, AT_FDCWD, path, pathseg, mode, 0));
2792 }
2793 
2794 /*
2795  * Change mode of a file given path name (don't follow links.)
2796  */
2797 #ifndef _SYS_SYSPROTO_H_
2798 struct lchmod_args {
2799 	char	*path;
2800 	int	mode;
2801 };
2802 #endif
2803 int
2804 sys_lchmod(td, uap)
2805 	struct thread *td;
2806 	register struct lchmod_args /* {
2807 		char *path;
2808 		int mode;
2809 	} */ *uap;
2810 {
2811 
2812 	return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2813 	    uap->mode, AT_SYMLINK_NOFOLLOW));
2814 }
2815 
2816 int
2817 kern_fchmodat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2818     mode_t mode, int flag)
2819 {
2820 	int error;
2821 	struct nameidata nd;
2822 	int follow;
2823 
2824 	AUDIT_ARG_MODE(mode);
2825 	follow = (flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
2826 	NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, pathseg, path, fd,
2827 	    CAP_FCHMOD, td);
2828 	if ((error = namei(&nd)) != 0)
2829 		return (error);
2830 	NDFREE(&nd, NDF_ONLY_PNBUF);
2831 	error = setfmode(td, td->td_ucred, nd.ni_vp, mode);
2832 	vrele(nd.ni_vp);
2833 	return (error);
2834 }
2835 
2836 /*
2837  * Change mode of a file given a file descriptor.
2838  */
2839 #ifndef _SYS_SYSPROTO_H_
2840 struct fchmod_args {
2841 	int	fd;
2842 	int	mode;
2843 };
2844 #endif
2845 int
2846 sys_fchmod(struct thread *td, struct fchmod_args *uap)
2847 {
2848 	struct file *fp;
2849 	int error;
2850 
2851 	AUDIT_ARG_FD(uap->fd);
2852 	AUDIT_ARG_MODE(uap->mode);
2853 
2854 	error = fget(td, uap->fd, CAP_FCHMOD, &fp);
2855 	if (error != 0)
2856 		return (error);
2857 	error = fo_chmod(fp, uap->mode, td->td_ucred, td);
2858 	fdrop(fp, td);
2859 	return (error);
2860 }
2861 
2862 /*
2863  * Common implementation for chown(), lchown(), and fchown()
2864  */
2865 int
2866 setfown(td, cred, vp, uid, gid)
2867 	struct thread *td;
2868 	struct ucred *cred;
2869 	struct vnode *vp;
2870 	uid_t uid;
2871 	gid_t gid;
2872 {
2873 	int error;
2874 	struct mount *mp;
2875 	struct vattr vattr;
2876 
2877 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2878 		return (error);
2879 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2880 	VATTR_NULL(&vattr);
2881 	vattr.va_uid = uid;
2882 	vattr.va_gid = gid;
2883 #ifdef MAC
2884 	error = mac_vnode_check_setowner(cred, vp, vattr.va_uid,
2885 	    vattr.va_gid);
2886 	if (error == 0)
2887 #endif
2888 		error = VOP_SETATTR(vp, &vattr, cred);
2889 	VOP_UNLOCK(vp, 0);
2890 	vn_finished_write(mp);
2891 	return (error);
2892 }
2893 
2894 /*
2895  * Set ownership given a path name.
2896  */
2897 #ifndef _SYS_SYSPROTO_H_
2898 struct chown_args {
2899 	char	*path;
2900 	int	uid;
2901 	int	gid;
2902 };
2903 #endif
2904 int
2905 sys_chown(td, uap)
2906 	struct thread *td;
2907 	register struct chown_args /* {
2908 		char *path;
2909 		int uid;
2910 		int gid;
2911 	} */ *uap;
2912 {
2913 
2914 	return (kern_chown(td, uap->path, UIO_USERSPACE, uap->uid, uap->gid));
2915 }
2916 
2917 #ifndef _SYS_SYSPROTO_H_
2918 struct fchownat_args {
2919 	int fd;
2920 	const char * path;
2921 	uid_t uid;
2922 	gid_t gid;
2923 	int flag;
2924 };
2925 #endif
2926 int
2927 sys_fchownat(struct thread *td, struct fchownat_args *uap)
2928 {
2929 	int flag;
2930 
2931 	flag = uap->flag;
2932 	if (flag & ~AT_SYMLINK_NOFOLLOW)
2933 		return (EINVAL);
2934 
2935 	return (kern_fchownat(td, uap->fd, uap->path, UIO_USERSPACE, uap->uid,
2936 	    uap->gid, uap->flag));
2937 }
2938 
2939 int
2940 kern_chown(struct thread *td, char *path, enum uio_seg pathseg, int uid,
2941     int gid)
2942 {
2943 
2944 	return (kern_fchownat(td, AT_FDCWD, path, pathseg, uid, gid, 0));
2945 }
2946 
2947 int
2948 kern_fchownat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2949     int uid, int gid, int flag)
2950 {
2951 	struct nameidata nd;
2952 	int error, follow;
2953 
2954 	AUDIT_ARG_OWNER(uid, gid);
2955 	follow = (flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
2956 	NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, pathseg, path, fd,
2957 	    CAP_FCHOWN, td);
2958 
2959 	if ((error = namei(&nd)) != 0)
2960 		return (error);
2961 	NDFREE(&nd, NDF_ONLY_PNBUF);
2962 	error = setfown(td, td->td_ucred, nd.ni_vp, uid, gid);
2963 	vrele(nd.ni_vp);
2964 	return (error);
2965 }
2966 
2967 /*
2968  * Set ownership given a path name, do not cross symlinks.
2969  */
2970 #ifndef _SYS_SYSPROTO_H_
2971 struct lchown_args {
2972 	char	*path;
2973 	int	uid;
2974 	int	gid;
2975 };
2976 #endif
2977 int
2978 sys_lchown(td, uap)
2979 	struct thread *td;
2980 	register struct lchown_args /* {
2981 		char *path;
2982 		int uid;
2983 		int gid;
2984 	} */ *uap;
2985 {
2986 
2987 	return (kern_lchown(td, uap->path, UIO_USERSPACE, uap->uid, uap->gid));
2988 }
2989 
2990 int
2991 kern_lchown(struct thread *td, char *path, enum uio_seg pathseg, int uid,
2992     int gid)
2993 {
2994 
2995 	return (kern_fchownat(td, AT_FDCWD, path, pathseg, uid, gid,
2996 	    AT_SYMLINK_NOFOLLOW));
2997 }
2998 
2999 /*
3000  * Set ownership given a file descriptor.
3001  */
3002 #ifndef _SYS_SYSPROTO_H_
3003 struct fchown_args {
3004 	int	fd;
3005 	int	uid;
3006 	int	gid;
3007 };
3008 #endif
3009 int
3010 sys_fchown(td, uap)
3011 	struct thread *td;
3012 	register struct fchown_args /* {
3013 		int fd;
3014 		int uid;
3015 		int gid;
3016 	} */ *uap;
3017 {
3018 	struct file *fp;
3019 	int error;
3020 
3021 	AUDIT_ARG_FD(uap->fd);
3022 	AUDIT_ARG_OWNER(uap->uid, uap->gid);
3023 	error = fget(td, uap->fd, CAP_FCHOWN, &fp);
3024 	if (error != 0)
3025 		return (error);
3026 	error = fo_chown(fp, uap->uid, uap->gid, td->td_ucred, td);
3027 	fdrop(fp, td);
3028 	return (error);
3029 }
3030 
3031 /*
3032  * Common implementation code for utimes(), lutimes(), and futimes().
3033  */
3034 static int
3035 getutimes(usrtvp, tvpseg, tsp)
3036 	const struct timeval *usrtvp;
3037 	enum uio_seg tvpseg;
3038 	struct timespec *tsp;
3039 {
3040 	struct timeval tv[2];
3041 	const struct timeval *tvp;
3042 	int error;
3043 
3044 	if (usrtvp == NULL) {
3045 		vfs_timestamp(&tsp[0]);
3046 		tsp[1] = tsp[0];
3047 	} else {
3048 		if (tvpseg == UIO_SYSSPACE) {
3049 			tvp = usrtvp;
3050 		} else {
3051 			if ((error = copyin(usrtvp, tv, sizeof(tv))) != 0)
3052 				return (error);
3053 			tvp = tv;
3054 		}
3055 
3056 		if (tvp[0].tv_usec < 0 || tvp[0].tv_usec >= 1000000 ||
3057 		    tvp[1].tv_usec < 0 || tvp[1].tv_usec >= 1000000)
3058 			return (EINVAL);
3059 		TIMEVAL_TO_TIMESPEC(&tvp[0], &tsp[0]);
3060 		TIMEVAL_TO_TIMESPEC(&tvp[1], &tsp[1]);
3061 	}
3062 	return (0);
3063 }
3064 
3065 /*
3066  * Common implementation code for utimes(), lutimes(), and futimes().
3067  */
3068 static int
3069 setutimes(td, vp, ts, numtimes, nullflag)
3070 	struct thread *td;
3071 	struct vnode *vp;
3072 	const struct timespec *ts;
3073 	int numtimes;
3074 	int nullflag;
3075 {
3076 	int error, setbirthtime;
3077 	struct mount *mp;
3078 	struct vattr vattr;
3079 
3080 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
3081 		return (error);
3082 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3083 	setbirthtime = 0;
3084 	if (numtimes < 3 && !VOP_GETATTR(vp, &vattr, td->td_ucred) &&
3085 	    timespeccmp(&ts[1], &vattr.va_birthtime, < ))
3086 		setbirthtime = 1;
3087 	VATTR_NULL(&vattr);
3088 	vattr.va_atime = ts[0];
3089 	vattr.va_mtime = ts[1];
3090 	if (setbirthtime)
3091 		vattr.va_birthtime = ts[1];
3092 	if (numtimes > 2)
3093 		vattr.va_birthtime = ts[2];
3094 	if (nullflag)
3095 		vattr.va_vaflags |= VA_UTIMES_NULL;
3096 #ifdef MAC
3097 	error = mac_vnode_check_setutimes(td->td_ucred, vp, vattr.va_atime,
3098 	    vattr.va_mtime);
3099 #endif
3100 	if (error == 0)
3101 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3102 	VOP_UNLOCK(vp, 0);
3103 	vn_finished_write(mp);
3104 	return (error);
3105 }
3106 
3107 /*
3108  * Set the access and modification times of a file.
3109  */
3110 #ifndef _SYS_SYSPROTO_H_
3111 struct utimes_args {
3112 	char	*path;
3113 	struct	timeval *tptr;
3114 };
3115 #endif
3116 int
3117 sys_utimes(td, uap)
3118 	struct thread *td;
3119 	register struct utimes_args /* {
3120 		char *path;
3121 		struct timeval *tptr;
3122 	} */ *uap;
3123 {
3124 
3125 	return (kern_utimes(td, uap->path, UIO_USERSPACE, uap->tptr,
3126 	    UIO_USERSPACE));
3127 }
3128 
3129 #ifndef _SYS_SYSPROTO_H_
3130 struct futimesat_args {
3131 	int fd;
3132 	const char * path;
3133 	const struct timeval * times;
3134 };
3135 #endif
3136 int
3137 sys_futimesat(struct thread *td, struct futimesat_args *uap)
3138 {
3139 
3140 	return (kern_utimesat(td, uap->fd, uap->path, UIO_USERSPACE,
3141 	    uap->times, UIO_USERSPACE));
3142 }
3143 
3144 int
3145 kern_utimes(struct thread *td, char *path, enum uio_seg pathseg,
3146     struct timeval *tptr, enum uio_seg tptrseg)
3147 {
3148 
3149 	return (kern_utimesat(td, AT_FDCWD, path, pathseg, tptr, tptrseg));
3150 }
3151 
3152 int
3153 kern_utimesat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
3154     struct timeval *tptr, enum uio_seg tptrseg)
3155 {
3156 	struct nameidata nd;
3157 	struct timespec ts[2];
3158 	int error;
3159 
3160 	if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3161 		return (error);
3162 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, fd,
3163 	    CAP_FUTIMES, td);
3164 
3165 	if ((error = namei(&nd)) != 0)
3166 		return (error);
3167 	NDFREE(&nd, NDF_ONLY_PNBUF);
3168 	error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3169 	vrele(nd.ni_vp);
3170 	return (error);
3171 }
3172 
3173 /*
3174  * Set the access and modification times of a file.
3175  */
3176 #ifndef _SYS_SYSPROTO_H_
3177 struct lutimes_args {
3178 	char	*path;
3179 	struct	timeval *tptr;
3180 };
3181 #endif
3182 int
3183 sys_lutimes(td, uap)
3184 	struct thread *td;
3185 	register struct lutimes_args /* {
3186 		char *path;
3187 		struct timeval *tptr;
3188 	} */ *uap;
3189 {
3190 
3191 	return (kern_lutimes(td, uap->path, UIO_USERSPACE, uap->tptr,
3192 	    UIO_USERSPACE));
3193 }
3194 
3195 int
3196 kern_lutimes(struct thread *td, char *path, enum uio_seg pathseg,
3197     struct timeval *tptr, enum uio_seg tptrseg)
3198 {
3199 	struct timespec ts[2];
3200 	int error;
3201 	struct nameidata nd;
3202 
3203 	if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3204 		return (error);
3205 	NDINIT(&nd, LOOKUP, NOFOLLOW | AUDITVNODE1, pathseg, path, td);
3206 	if ((error = namei(&nd)) != 0)
3207 		return (error);
3208 	NDFREE(&nd, NDF_ONLY_PNBUF);
3209 	error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3210 	vrele(nd.ni_vp);
3211 	return (error);
3212 }
3213 
3214 /*
3215  * Set the access and modification times of a file.
3216  */
3217 #ifndef _SYS_SYSPROTO_H_
3218 struct futimes_args {
3219 	int	fd;
3220 	struct	timeval *tptr;
3221 };
3222 #endif
3223 int
3224 sys_futimes(td, uap)
3225 	struct thread *td;
3226 	register struct futimes_args /* {
3227 		int  fd;
3228 		struct timeval *tptr;
3229 	} */ *uap;
3230 {
3231 
3232 	return (kern_futimes(td, uap->fd, uap->tptr, UIO_USERSPACE));
3233 }
3234 
3235 int
3236 kern_futimes(struct thread *td, int fd, struct timeval *tptr,
3237     enum uio_seg tptrseg)
3238 {
3239 	struct timespec ts[2];
3240 	struct file *fp;
3241 	int error;
3242 
3243 	AUDIT_ARG_FD(fd);
3244 	if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3245 		return (error);
3246 	if ((error = getvnode(td->td_proc->p_fd, fd, CAP_FUTIMES, &fp)) != 0)
3247 		return (error);
3248 #ifdef AUDIT
3249 	vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
3250 	AUDIT_ARG_VNODE1(fp->f_vnode);
3251 	VOP_UNLOCK(fp->f_vnode, 0);
3252 #endif
3253 	error = setutimes(td, fp->f_vnode, ts, 2, tptr == NULL);
3254 	fdrop(fp, td);
3255 	return (error);
3256 }
3257 
3258 /*
3259  * Truncate a file given its path name.
3260  */
3261 #ifndef _SYS_SYSPROTO_H_
3262 struct truncate_args {
3263 	char	*path;
3264 	int	pad;
3265 	off_t	length;
3266 };
3267 #endif
3268 int
3269 sys_truncate(td, uap)
3270 	struct thread *td;
3271 	register struct truncate_args /* {
3272 		char *path;
3273 		int pad;
3274 		off_t length;
3275 	} */ *uap;
3276 {
3277 
3278 	return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3279 }
3280 
3281 int
3282 kern_truncate(struct thread *td, char *path, enum uio_seg pathseg, off_t length)
3283 {
3284 	struct mount *mp;
3285 	struct vnode *vp;
3286 	void *rl_cookie;
3287 	struct vattr vattr;
3288 	struct nameidata nd;
3289 	int error;
3290 
3291 	if (length < 0)
3292 		return(EINVAL);
3293 	NDINIT(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, td);
3294 	if ((error = namei(&nd)) != 0)
3295 		return (error);
3296 	vp = nd.ni_vp;
3297 	rl_cookie = vn_rangelock_wlock(vp, 0, OFF_MAX);
3298 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) {
3299 		vn_rangelock_unlock(vp, rl_cookie);
3300 		vrele(vp);
3301 		return (error);
3302 	}
3303 	NDFREE(&nd, NDF_ONLY_PNBUF);
3304 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3305 	if (vp->v_type == VDIR)
3306 		error = EISDIR;
3307 #ifdef MAC
3308 	else if ((error = mac_vnode_check_write(td->td_ucred, NOCRED, vp))) {
3309 	}
3310 #endif
3311 	else if ((error = vn_writechk(vp)) == 0 &&
3312 	    (error = VOP_ACCESS(vp, VWRITE, td->td_ucred, td)) == 0) {
3313 		VATTR_NULL(&vattr);
3314 		vattr.va_size = length;
3315 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3316 	}
3317 	VOP_UNLOCK(vp, 0);
3318 	vn_finished_write(mp);
3319 	vn_rangelock_unlock(vp, rl_cookie);
3320 	vrele(vp);
3321 	return (error);
3322 }
3323 
3324 #if defined(COMPAT_43)
3325 /*
3326  * Truncate a file given its path name.
3327  */
3328 #ifndef _SYS_SYSPROTO_H_
3329 struct otruncate_args {
3330 	char	*path;
3331 	long	length;
3332 };
3333 #endif
3334 int
3335 otruncate(td, uap)
3336 	struct thread *td;
3337 	register struct otruncate_args /* {
3338 		char *path;
3339 		long length;
3340 	} */ *uap;
3341 {
3342 	struct truncate_args /* {
3343 		char *path;
3344 		int pad;
3345 		off_t length;
3346 	} */ nuap;
3347 
3348 	nuap.path = uap->path;
3349 	nuap.length = uap->length;
3350 	return (sys_truncate(td, &nuap));
3351 }
3352 #endif /* COMPAT_43 */
3353 
3354 /* Versions with the pad argument */
3355 int
3356 freebsd6_truncate(struct thread *td, struct freebsd6_truncate_args *uap)
3357 {
3358 	struct truncate_args ouap;
3359 
3360 	ouap.path = uap->path;
3361 	ouap.length = uap->length;
3362 	return (sys_truncate(td, &ouap));
3363 }
3364 
3365 int
3366 freebsd6_ftruncate(struct thread *td, struct freebsd6_ftruncate_args *uap)
3367 {
3368 	struct ftruncate_args ouap;
3369 
3370 	ouap.fd = uap->fd;
3371 	ouap.length = uap->length;
3372 	return (sys_ftruncate(td, &ouap));
3373 }
3374 
3375 /*
3376  * Sync an open file.
3377  */
3378 #ifndef _SYS_SYSPROTO_H_
3379 struct fsync_args {
3380 	int	fd;
3381 };
3382 #endif
3383 int
3384 sys_fsync(td, uap)
3385 	struct thread *td;
3386 	struct fsync_args /* {
3387 		int fd;
3388 	} */ *uap;
3389 {
3390 	struct vnode *vp;
3391 	struct mount *mp;
3392 	struct file *fp;
3393 	int error, lock_flags;
3394 
3395 	AUDIT_ARG_FD(uap->fd);
3396 	if ((error = getvnode(td->td_proc->p_fd, uap->fd, CAP_FSYNC, &fp)) != 0)
3397 		return (error);
3398 	vp = fp->f_vnode;
3399 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
3400 		goto drop;
3401 	if (MNT_SHARED_WRITES(mp) ||
3402 	    ((mp == NULL) && MNT_SHARED_WRITES(vp->v_mount))) {
3403 		lock_flags = LK_SHARED;
3404 	} else {
3405 		lock_flags = LK_EXCLUSIVE;
3406 	}
3407 	vn_lock(vp, lock_flags | LK_RETRY);
3408 	AUDIT_ARG_VNODE1(vp);
3409 	if (vp->v_object != NULL) {
3410 		VM_OBJECT_WLOCK(vp->v_object);
3411 		vm_object_page_clean(vp->v_object, 0, 0, 0);
3412 		VM_OBJECT_WUNLOCK(vp->v_object);
3413 	}
3414 	error = VOP_FSYNC(vp, MNT_WAIT, td);
3415 
3416 	VOP_UNLOCK(vp, 0);
3417 	vn_finished_write(mp);
3418 drop:
3419 	fdrop(fp, td);
3420 	return (error);
3421 }
3422 
3423 /*
3424  * Rename files.  Source and destination must either both be directories, or
3425  * both not be directories.  If target is a directory, it must be empty.
3426  */
3427 #ifndef _SYS_SYSPROTO_H_
3428 struct rename_args {
3429 	char	*from;
3430 	char	*to;
3431 };
3432 #endif
3433 int
3434 sys_rename(td, uap)
3435 	struct thread *td;
3436 	register struct rename_args /* {
3437 		char *from;
3438 		char *to;
3439 	} */ *uap;
3440 {
3441 
3442 	return (kern_rename(td, uap->from, uap->to, UIO_USERSPACE));
3443 }
3444 
3445 #ifndef _SYS_SYSPROTO_H_
3446 struct renameat_args {
3447 	int	oldfd;
3448 	char	*old;
3449 	int	newfd;
3450 	char	*new;
3451 };
3452 #endif
3453 int
3454 sys_renameat(struct thread *td, struct renameat_args *uap)
3455 {
3456 
3457 	return (kern_renameat(td, uap->oldfd, uap->old, uap->newfd, uap->new,
3458 	    UIO_USERSPACE));
3459 }
3460 
3461 int
3462 kern_rename(struct thread *td, char *from, char *to, enum uio_seg pathseg)
3463 {
3464 
3465 	return (kern_renameat(td, AT_FDCWD, from, AT_FDCWD, to, pathseg));
3466 }
3467 
3468 int
3469 kern_renameat(struct thread *td, int oldfd, char *old, int newfd, char *new,
3470     enum uio_seg pathseg)
3471 {
3472 	struct mount *mp = NULL;
3473 	struct vnode *tvp, *fvp, *tdvp;
3474 	struct nameidata fromnd, tond;
3475 	int error;
3476 
3477 	bwillwrite();
3478 #ifdef MAC
3479 	NDINIT_ATRIGHTS(&fromnd, DELETE, LOCKPARENT | LOCKLEAF | SAVESTART |
3480 	    AUDITVNODE1, pathseg, old, oldfd, CAP_RENAMEAT, td);
3481 #else
3482 	NDINIT_ATRIGHTS(&fromnd, DELETE, WANTPARENT | SAVESTART | AUDITVNODE1,
3483 	    pathseg, old, oldfd, CAP_RENAMEAT, td);
3484 #endif
3485 
3486 	if ((error = namei(&fromnd)) != 0)
3487 		return (error);
3488 #ifdef MAC
3489 	error = mac_vnode_check_rename_from(td->td_ucred, fromnd.ni_dvp,
3490 	    fromnd.ni_vp, &fromnd.ni_cnd);
3491 	VOP_UNLOCK(fromnd.ni_dvp, 0);
3492 	if (fromnd.ni_dvp != fromnd.ni_vp)
3493 		VOP_UNLOCK(fromnd.ni_vp, 0);
3494 #endif
3495 	fvp = fromnd.ni_vp;
3496 	if (error == 0)
3497 		error = vn_start_write(fvp, &mp, V_WAIT | PCATCH);
3498 	if (error != 0) {
3499 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3500 		vrele(fromnd.ni_dvp);
3501 		vrele(fvp);
3502 		goto out1;
3503 	}
3504 	NDINIT_ATRIGHTS(&tond, RENAME, LOCKPARENT | LOCKLEAF | NOCACHE |
3505 	    SAVESTART | AUDITVNODE2, pathseg, new, newfd, CAP_LINKAT, td);
3506 	if (fromnd.ni_vp->v_type == VDIR)
3507 		tond.ni_cnd.cn_flags |= WILLBEDIR;
3508 	if ((error = namei(&tond)) != 0) {
3509 		/* Translate error code for rename("dir1", "dir2/."). */
3510 		if (error == EISDIR && fvp->v_type == VDIR)
3511 			error = EINVAL;
3512 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3513 		vrele(fromnd.ni_dvp);
3514 		vrele(fvp);
3515 		vn_finished_write(mp);
3516 		goto out1;
3517 	}
3518 	tdvp = tond.ni_dvp;
3519 	tvp = tond.ni_vp;
3520 	if (tvp != NULL) {
3521 		if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
3522 			error = ENOTDIR;
3523 			goto out;
3524 		} else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
3525 			error = EISDIR;
3526 			goto out;
3527 		}
3528 #ifdef CAPABILITIES
3529 		if (newfd != AT_FDCWD) {
3530 			/*
3531 			 * If the target already exists we require CAP_UNLINKAT
3532 			 * from 'newfd'.
3533 			 */
3534 			error = cap_check(tond.ni_filecaps.fc_rights,
3535 			    CAP_UNLINKAT);
3536 			if (error != 0)
3537 				goto out;
3538 		}
3539 #endif
3540 	}
3541 	if (fvp == tdvp) {
3542 		error = EINVAL;
3543 		goto out;
3544 	}
3545 	/*
3546 	 * If the source is the same as the destination (that is, if they
3547 	 * are links to the same vnode), then there is nothing to do.
3548 	 */
3549 	if (fvp == tvp)
3550 		error = -1;
3551 #ifdef MAC
3552 	else
3553 		error = mac_vnode_check_rename_to(td->td_ucred, tdvp,
3554 		    tond.ni_vp, fromnd.ni_dvp == tdvp, &tond.ni_cnd);
3555 #endif
3556 out:
3557 	if (!error) {
3558 		error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
3559 				   tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
3560 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3561 		NDFREE(&tond, NDF_ONLY_PNBUF);
3562 	} else {
3563 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3564 		NDFREE(&tond, NDF_ONLY_PNBUF);
3565 		if (tvp)
3566 			vput(tvp);
3567 		if (tdvp == tvp)
3568 			vrele(tdvp);
3569 		else
3570 			vput(tdvp);
3571 		vrele(fromnd.ni_dvp);
3572 		vrele(fvp);
3573 	}
3574 	vrele(tond.ni_startdir);
3575 	vn_finished_write(mp);
3576 out1:
3577 	if (fromnd.ni_startdir)
3578 		vrele(fromnd.ni_startdir);
3579 	if (error == -1)
3580 		return (0);
3581 	return (error);
3582 }
3583 
3584 /*
3585  * Make a directory file.
3586  */
3587 #ifndef _SYS_SYSPROTO_H_
3588 struct mkdir_args {
3589 	char	*path;
3590 	int	mode;
3591 };
3592 #endif
3593 int
3594 sys_mkdir(td, uap)
3595 	struct thread *td;
3596 	register struct mkdir_args /* {
3597 		char *path;
3598 		int mode;
3599 	} */ *uap;
3600 {
3601 
3602 	return (kern_mkdir(td, uap->path, UIO_USERSPACE, uap->mode));
3603 }
3604 
3605 #ifndef _SYS_SYSPROTO_H_
3606 struct mkdirat_args {
3607 	int	fd;
3608 	char	*path;
3609 	mode_t	mode;
3610 };
3611 #endif
3612 int
3613 sys_mkdirat(struct thread *td, struct mkdirat_args *uap)
3614 {
3615 
3616 	return (kern_mkdirat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode));
3617 }
3618 
3619 int
3620 kern_mkdir(struct thread *td, char *path, enum uio_seg segflg, int mode)
3621 {
3622 
3623 	return (kern_mkdirat(td, AT_FDCWD, path, segflg, mode));
3624 }
3625 
3626 int
3627 kern_mkdirat(struct thread *td, int fd, char *path, enum uio_seg segflg,
3628     int mode)
3629 {
3630 	struct mount *mp;
3631 	struct vnode *vp;
3632 	struct vattr vattr;
3633 	int error;
3634 	struct nameidata nd;
3635 
3636 	AUDIT_ARG_MODE(mode);
3637 restart:
3638 	bwillwrite();
3639 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1,
3640 	    segflg, path, fd, CAP_MKDIRAT, td);
3641 	nd.ni_cnd.cn_flags |= WILLBEDIR;
3642 	if ((error = namei(&nd)) != 0)
3643 		return (error);
3644 	vp = nd.ni_vp;
3645 	if (vp != NULL) {
3646 		NDFREE(&nd, NDF_ONLY_PNBUF);
3647 		/*
3648 		 * XXX namei called with LOCKPARENT but not LOCKLEAF has
3649 		 * the strange behaviour of leaving the vnode unlocked
3650 		 * if the target is the same vnode as the parent.
3651 		 */
3652 		if (vp == nd.ni_dvp)
3653 			vrele(nd.ni_dvp);
3654 		else
3655 			vput(nd.ni_dvp);
3656 		vrele(vp);
3657 		return (EEXIST);
3658 	}
3659 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3660 		NDFREE(&nd, NDF_ONLY_PNBUF);
3661 		vput(nd.ni_dvp);
3662 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3663 			return (error);
3664 		goto restart;
3665 	}
3666 	VATTR_NULL(&vattr);
3667 	vattr.va_type = VDIR;
3668 	vattr.va_mode = (mode & ACCESSPERMS) &~ td->td_proc->p_fd->fd_cmask;
3669 #ifdef MAC
3670 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
3671 	    &vattr);
3672 	if (error)
3673 		goto out;
3674 #endif
3675 	error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
3676 #ifdef MAC
3677 out:
3678 #endif
3679 	NDFREE(&nd, NDF_ONLY_PNBUF);
3680 	vput(nd.ni_dvp);
3681 	if (!error)
3682 		vput(nd.ni_vp);
3683 	vn_finished_write(mp);
3684 	return (error);
3685 }
3686 
3687 /*
3688  * Remove a directory file.
3689  */
3690 #ifndef _SYS_SYSPROTO_H_
3691 struct rmdir_args {
3692 	char	*path;
3693 };
3694 #endif
3695 int
3696 sys_rmdir(td, uap)
3697 	struct thread *td;
3698 	struct rmdir_args /* {
3699 		char *path;
3700 	} */ *uap;
3701 {
3702 
3703 	return (kern_rmdir(td, uap->path, UIO_USERSPACE));
3704 }
3705 
3706 int
3707 kern_rmdir(struct thread *td, char *path, enum uio_seg pathseg)
3708 {
3709 
3710 	return (kern_rmdirat(td, AT_FDCWD, path, pathseg));
3711 }
3712 
3713 int
3714 kern_rmdirat(struct thread *td, int fd, char *path, enum uio_seg pathseg)
3715 {
3716 	struct mount *mp;
3717 	struct vnode *vp;
3718 	int error;
3719 	struct nameidata nd;
3720 
3721 restart:
3722 	bwillwrite();
3723 	NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1,
3724 	    pathseg, path, fd, CAP_UNLINKAT, td);
3725 	if ((error = namei(&nd)) != 0)
3726 		return (error);
3727 	vp = nd.ni_vp;
3728 	if (vp->v_type != VDIR) {
3729 		error = ENOTDIR;
3730 		goto out;
3731 	}
3732 	/*
3733 	 * No rmdir "." please.
3734 	 */
3735 	if (nd.ni_dvp == vp) {
3736 		error = EINVAL;
3737 		goto out;
3738 	}
3739 	/*
3740 	 * The root of a mounted filesystem cannot be deleted.
3741 	 */
3742 	if (vp->v_vflag & VV_ROOT) {
3743 		error = EBUSY;
3744 		goto out;
3745 	}
3746 #ifdef MAC
3747 	error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
3748 	    &nd.ni_cnd);
3749 	if (error)
3750 		goto out;
3751 #endif
3752 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3753 		NDFREE(&nd, NDF_ONLY_PNBUF);
3754 		vput(vp);
3755 		if (nd.ni_dvp == vp)
3756 			vrele(nd.ni_dvp);
3757 		else
3758 			vput(nd.ni_dvp);
3759 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3760 			return (error);
3761 		goto restart;
3762 	}
3763 	vfs_notify_upper(vp, VFS_NOTIFY_UPPER_UNLINK);
3764 	error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
3765 	vn_finished_write(mp);
3766 out:
3767 	NDFREE(&nd, NDF_ONLY_PNBUF);
3768 	vput(vp);
3769 	if (nd.ni_dvp == vp)
3770 		vrele(nd.ni_dvp);
3771 	else
3772 		vput(nd.ni_dvp);
3773 	return (error);
3774 }
3775 
3776 #ifdef COMPAT_43
3777 /*
3778  * Read a block of directory entries in a filesystem independent format.
3779  */
3780 #ifndef _SYS_SYSPROTO_H_
3781 struct ogetdirentries_args {
3782 	int	fd;
3783 	char	*buf;
3784 	u_int	count;
3785 	long	*basep;
3786 };
3787 #endif
3788 int
3789 ogetdirentries(struct thread *td, struct ogetdirentries_args *uap)
3790 {
3791 	long loff;
3792 	int error;
3793 
3794 	error = kern_ogetdirentries(td, uap, &loff);
3795 	if (error == 0)
3796 		error = copyout(&loff, uap->basep, sizeof(long));
3797 	return (error);
3798 }
3799 
3800 int
3801 kern_ogetdirentries(struct thread *td, struct ogetdirentries_args *uap,
3802     long *ploff)
3803 {
3804 	struct vnode *vp;
3805 	struct file *fp;
3806 	struct uio auio, kuio;
3807 	struct iovec aiov, kiov;
3808 	struct dirent *dp, *edp;
3809 	caddr_t dirbuf;
3810 	int error, eofflag, readcnt;
3811 	long loff;
3812 	off_t foffset;
3813 
3814 	/* XXX arbitrary sanity limit on `count'. */
3815 	if (uap->count > 64 * 1024)
3816 		return (EINVAL);
3817 	if ((error = getvnode(td->td_proc->p_fd, uap->fd, CAP_READ, &fp)) != 0)
3818 		return (error);
3819 	if ((fp->f_flag & FREAD) == 0) {
3820 		fdrop(fp, td);
3821 		return (EBADF);
3822 	}
3823 	vp = fp->f_vnode;
3824 	foffset = foffset_lock(fp, 0);
3825 unionread:
3826 	if (vp->v_type != VDIR) {
3827 		foffset_unlock(fp, foffset, 0);
3828 		fdrop(fp, td);
3829 		return (EINVAL);
3830 	}
3831 	aiov.iov_base = uap->buf;
3832 	aiov.iov_len = uap->count;
3833 	auio.uio_iov = &aiov;
3834 	auio.uio_iovcnt = 1;
3835 	auio.uio_rw = UIO_READ;
3836 	auio.uio_segflg = UIO_USERSPACE;
3837 	auio.uio_td = td;
3838 	auio.uio_resid = uap->count;
3839 	vn_lock(vp, LK_SHARED | LK_RETRY);
3840 	loff = auio.uio_offset = foffset;
3841 #ifdef MAC
3842 	error = mac_vnode_check_readdir(td->td_ucred, vp);
3843 	if (error) {
3844 		VOP_UNLOCK(vp, 0);
3845 		foffset_unlock(fp, foffset, FOF_NOUPDATE);
3846 		fdrop(fp, td);
3847 		return (error);
3848 	}
3849 #endif
3850 #	if (BYTE_ORDER != LITTLE_ENDIAN)
3851 		if (vp->v_mount->mnt_maxsymlinklen <= 0) {
3852 			error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag,
3853 			    NULL, NULL);
3854 			foffset = auio.uio_offset;
3855 		} else
3856 #	endif
3857 	{
3858 		kuio = auio;
3859 		kuio.uio_iov = &kiov;
3860 		kuio.uio_segflg = UIO_SYSSPACE;
3861 		kiov.iov_len = uap->count;
3862 		dirbuf = malloc(uap->count, M_TEMP, M_WAITOK);
3863 		kiov.iov_base = dirbuf;
3864 		error = VOP_READDIR(vp, &kuio, fp->f_cred, &eofflag,
3865 			    NULL, NULL);
3866 		foffset = kuio.uio_offset;
3867 		if (error == 0) {
3868 			readcnt = uap->count - kuio.uio_resid;
3869 			edp = (struct dirent *)&dirbuf[readcnt];
3870 			for (dp = (struct dirent *)dirbuf; dp < edp; ) {
3871 #				if (BYTE_ORDER == LITTLE_ENDIAN)
3872 					/*
3873 					 * The expected low byte of
3874 					 * dp->d_namlen is our dp->d_type.
3875 					 * The high MBZ byte of dp->d_namlen
3876 					 * is our dp->d_namlen.
3877 					 */
3878 					dp->d_type = dp->d_namlen;
3879 					dp->d_namlen = 0;
3880 #				else
3881 					/*
3882 					 * The dp->d_type is the high byte
3883 					 * of the expected dp->d_namlen,
3884 					 * so must be zero'ed.
3885 					 */
3886 					dp->d_type = 0;
3887 #				endif
3888 				if (dp->d_reclen > 0) {
3889 					dp = (struct dirent *)
3890 					    ((char *)dp + dp->d_reclen);
3891 				} else {
3892 					error = EIO;
3893 					break;
3894 				}
3895 			}
3896 			if (dp >= edp)
3897 				error = uiomove(dirbuf, readcnt, &auio);
3898 		}
3899 		free(dirbuf, M_TEMP);
3900 	}
3901 	if (error) {
3902 		VOP_UNLOCK(vp, 0);
3903 		foffset_unlock(fp, foffset, 0);
3904 		fdrop(fp, td);
3905 		return (error);
3906 	}
3907 	if (uap->count == auio.uio_resid &&
3908 	    (vp->v_vflag & VV_ROOT) &&
3909 	    (vp->v_mount->mnt_flag & MNT_UNION)) {
3910 		struct vnode *tvp = vp;
3911 		vp = vp->v_mount->mnt_vnodecovered;
3912 		VREF(vp);
3913 		fp->f_vnode = vp;
3914 		fp->f_data = vp;
3915 		foffset = 0;
3916 		vput(tvp);
3917 		goto unionread;
3918 	}
3919 	VOP_UNLOCK(vp, 0);
3920 	foffset_unlock(fp, foffset, 0);
3921 	fdrop(fp, td);
3922 	td->td_retval[0] = uap->count - auio.uio_resid;
3923 	if (error == 0)
3924 		*ploff = loff;
3925 	return (error);
3926 }
3927 #endif /* COMPAT_43 */
3928 
3929 /*
3930  * Read a block of directory entries in a filesystem independent format.
3931  */
3932 #ifndef _SYS_SYSPROTO_H_
3933 struct getdirentries_args {
3934 	int	fd;
3935 	char	*buf;
3936 	u_int	count;
3937 	long	*basep;
3938 };
3939 #endif
3940 int
3941 sys_getdirentries(td, uap)
3942 	struct thread *td;
3943 	register struct getdirentries_args /* {
3944 		int fd;
3945 		char *buf;
3946 		u_int count;
3947 		long *basep;
3948 	} */ *uap;
3949 {
3950 	long base;
3951 	int error;
3952 
3953 	error = kern_getdirentries(td, uap->fd, uap->buf, uap->count, &base,
3954 	    NULL, UIO_USERSPACE);
3955 	if (error)
3956 		return (error);
3957 	if (uap->basep != NULL)
3958 		error = copyout(&base, uap->basep, sizeof(long));
3959 	return (error);
3960 }
3961 
3962 int
3963 kern_getdirentries(struct thread *td, int fd, char *buf, u_int count,
3964     long *basep, ssize_t *residp, enum uio_seg bufseg)
3965 {
3966 	struct vnode *vp;
3967 	struct file *fp;
3968 	struct uio auio;
3969 	struct iovec aiov;
3970 	long loff;
3971 	int error, eofflag;
3972 	off_t foffset;
3973 
3974 	AUDIT_ARG_FD(fd);
3975 	if (count > IOSIZE_MAX)
3976 		return (EINVAL);
3977 	auio.uio_resid = count;
3978 	if ((error = getvnode(td->td_proc->p_fd, fd, CAP_READ, &fp)) != 0)
3979 		return (error);
3980 	if ((fp->f_flag & FREAD) == 0) {
3981 		fdrop(fp, td);
3982 		return (EBADF);
3983 	}
3984 	vp = fp->f_vnode;
3985 	foffset = foffset_lock(fp, 0);
3986 unionread:
3987 	if (vp->v_type != VDIR) {
3988 		error = EINVAL;
3989 		goto fail;
3990 	}
3991 	aiov.iov_base = buf;
3992 	aiov.iov_len = count;
3993 	auio.uio_iov = &aiov;
3994 	auio.uio_iovcnt = 1;
3995 	auio.uio_rw = UIO_READ;
3996 	auio.uio_segflg = bufseg;
3997 	auio.uio_td = td;
3998 	vn_lock(vp, LK_SHARED | LK_RETRY);
3999 	AUDIT_ARG_VNODE1(vp);
4000 	loff = auio.uio_offset = foffset;
4001 #ifdef MAC
4002 	error = mac_vnode_check_readdir(td->td_ucred, vp);
4003 	if (error == 0)
4004 #endif
4005 		error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, NULL,
4006 		    NULL);
4007 	foffset = auio.uio_offset;
4008 	if (error) {
4009 		VOP_UNLOCK(vp, 0);
4010 		goto fail;
4011 	}
4012 	if (count == auio.uio_resid &&
4013 	    (vp->v_vflag & VV_ROOT) &&
4014 	    (vp->v_mount->mnt_flag & MNT_UNION)) {
4015 		struct vnode *tvp = vp;
4016 		vp = vp->v_mount->mnt_vnodecovered;
4017 		VREF(vp);
4018 		fp->f_vnode = vp;
4019 		fp->f_data = vp;
4020 		foffset = 0;
4021 		vput(tvp);
4022 		goto unionread;
4023 	}
4024 	VOP_UNLOCK(vp, 0);
4025 	*basep = loff;
4026 	if (residp != NULL)
4027 		*residp = auio.uio_resid;
4028 	td->td_retval[0] = count - auio.uio_resid;
4029 fail:
4030 	foffset_unlock(fp, foffset, 0);
4031 	fdrop(fp, td);
4032 	return (error);
4033 }
4034 
4035 #ifndef _SYS_SYSPROTO_H_
4036 struct getdents_args {
4037 	int fd;
4038 	char *buf;
4039 	size_t count;
4040 };
4041 #endif
4042 int
4043 sys_getdents(td, uap)
4044 	struct thread *td;
4045 	register struct getdents_args /* {
4046 		int fd;
4047 		char *buf;
4048 		u_int count;
4049 	} */ *uap;
4050 {
4051 	struct getdirentries_args ap;
4052 	ap.fd = uap->fd;
4053 	ap.buf = uap->buf;
4054 	ap.count = uap->count;
4055 	ap.basep = NULL;
4056 	return (sys_getdirentries(td, &ap));
4057 }
4058 
4059 /*
4060  * Set the mode mask for creation of filesystem nodes.
4061  */
4062 #ifndef _SYS_SYSPROTO_H_
4063 struct umask_args {
4064 	int	newmask;
4065 };
4066 #endif
4067 int
4068 sys_umask(td, uap)
4069 	struct thread *td;
4070 	struct umask_args /* {
4071 		int newmask;
4072 	} */ *uap;
4073 {
4074 	register struct filedesc *fdp;
4075 
4076 	FILEDESC_XLOCK(td->td_proc->p_fd);
4077 	fdp = td->td_proc->p_fd;
4078 	td->td_retval[0] = fdp->fd_cmask;
4079 	fdp->fd_cmask = uap->newmask & ALLPERMS;
4080 	FILEDESC_XUNLOCK(td->td_proc->p_fd);
4081 	return (0);
4082 }
4083 
4084 /*
4085  * Void all references to file by ripping underlying filesystem away from
4086  * vnode.
4087  */
4088 #ifndef _SYS_SYSPROTO_H_
4089 struct revoke_args {
4090 	char	*path;
4091 };
4092 #endif
4093 int
4094 sys_revoke(td, uap)
4095 	struct thread *td;
4096 	register struct revoke_args /* {
4097 		char *path;
4098 	} */ *uap;
4099 {
4100 	struct vnode *vp;
4101 	struct vattr vattr;
4102 	int error;
4103 	struct nameidata nd;
4104 
4105 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
4106 	    uap->path, td);
4107 	if ((error = namei(&nd)) != 0)
4108 		return (error);
4109 	vp = nd.ni_vp;
4110 	NDFREE(&nd, NDF_ONLY_PNBUF);
4111 	if (vp->v_type != VCHR || vp->v_rdev == NULL) {
4112 		error = EINVAL;
4113 		goto out;
4114 	}
4115 #ifdef MAC
4116 	error = mac_vnode_check_revoke(td->td_ucred, vp);
4117 	if (error)
4118 		goto out;
4119 #endif
4120 	error = VOP_GETATTR(vp, &vattr, td->td_ucred);
4121 	if (error)
4122 		goto out;
4123 	if (td->td_ucred->cr_uid != vattr.va_uid) {
4124 		error = priv_check(td, PRIV_VFS_ADMIN);
4125 		if (error)
4126 			goto out;
4127 	}
4128 	if (vcount(vp) > 1)
4129 		VOP_REVOKE(vp, REVOKEALL);
4130 out:
4131 	vput(vp);
4132 	return (error);
4133 }
4134 
4135 /*
4136  * Convert a user file descriptor to a kernel file entry and check that, if it
4137  * is a capability, the correct rights are present. A reference on the file
4138  * entry is held upon returning.
4139  */
4140 int
4141 getvnode(struct filedesc *fdp, int fd, cap_rights_t rights, struct file **fpp)
4142 {
4143 	struct file *fp;
4144 	int error;
4145 
4146 	error = fget_unlocked(fdp, fd, rights, 0, &fp, NULL);
4147 	if (error != 0)
4148 		return (error);
4149 
4150 	/*
4151 	 * The file could be not of the vnode type, or it may be not
4152 	 * yet fully initialized, in which case the f_vnode pointer
4153 	 * may be set, but f_ops is still badfileops.  E.g.,
4154 	 * devfs_open() transiently create such situation to
4155 	 * facilitate csw d_fdopen().
4156 	 *
4157 	 * Dupfdopen() handling in kern_openat() installs the
4158 	 * half-baked file into the process descriptor table, allowing
4159 	 * other thread to dereference it. Guard against the race by
4160 	 * checking f_ops.
4161 	 */
4162 	if (fp->f_vnode == NULL || fp->f_ops == &badfileops) {
4163 		fdrop(fp, curthread);
4164 		return (EINVAL);
4165 	}
4166 	*fpp = fp;
4167 	return (0);
4168 }
4169 
4170 
4171 /*
4172  * Get an (NFS) file handle.
4173  */
4174 #ifndef _SYS_SYSPROTO_H_
4175 struct lgetfh_args {
4176 	char	*fname;
4177 	fhandle_t *fhp;
4178 };
4179 #endif
4180 int
4181 sys_lgetfh(td, uap)
4182 	struct thread *td;
4183 	register struct lgetfh_args *uap;
4184 {
4185 	struct nameidata nd;
4186 	fhandle_t fh;
4187 	register struct vnode *vp;
4188 	int error;
4189 
4190 	error = priv_check(td, PRIV_VFS_GETFH);
4191 	if (error)
4192 		return (error);
4193 	NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
4194 	    uap->fname, td);
4195 	error = namei(&nd);
4196 	if (error)
4197 		return (error);
4198 	NDFREE(&nd, NDF_ONLY_PNBUF);
4199 	vp = nd.ni_vp;
4200 	bzero(&fh, sizeof(fh));
4201 	fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
4202 	error = VOP_VPTOFH(vp, &fh.fh_fid);
4203 	vput(vp);
4204 	if (error)
4205 		return (error);
4206 	error = copyout(&fh, uap->fhp, sizeof (fh));
4207 	return (error);
4208 }
4209 
4210 #ifndef _SYS_SYSPROTO_H_
4211 struct getfh_args {
4212 	char	*fname;
4213 	fhandle_t *fhp;
4214 };
4215 #endif
4216 int
4217 sys_getfh(td, uap)
4218 	struct thread *td;
4219 	register struct getfh_args *uap;
4220 {
4221 	struct nameidata nd;
4222 	fhandle_t fh;
4223 	register struct vnode *vp;
4224 	int error;
4225 
4226 	error = priv_check(td, PRIV_VFS_GETFH);
4227 	if (error)
4228 		return (error);
4229 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
4230 	    uap->fname, td);
4231 	error = namei(&nd);
4232 	if (error)
4233 		return (error);
4234 	NDFREE(&nd, NDF_ONLY_PNBUF);
4235 	vp = nd.ni_vp;
4236 	bzero(&fh, sizeof(fh));
4237 	fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
4238 	error = VOP_VPTOFH(vp, &fh.fh_fid);
4239 	vput(vp);
4240 	if (error)
4241 		return (error);
4242 	error = copyout(&fh, uap->fhp, sizeof (fh));
4243 	return (error);
4244 }
4245 
4246 /*
4247  * syscall for the rpc.lockd to use to translate a NFS file handle into an
4248  * open descriptor.
4249  *
4250  * warning: do not remove the priv_check() call or this becomes one giant
4251  * security hole.
4252  */
4253 #ifndef _SYS_SYSPROTO_H_
4254 struct fhopen_args {
4255 	const struct fhandle *u_fhp;
4256 	int flags;
4257 };
4258 #endif
4259 int
4260 sys_fhopen(td, uap)
4261 	struct thread *td;
4262 	struct fhopen_args /* {
4263 		const struct fhandle *u_fhp;
4264 		int flags;
4265 	} */ *uap;
4266 {
4267 	struct mount *mp;
4268 	struct vnode *vp;
4269 	struct fhandle fhp;
4270 	struct file *fp;
4271 	int fmode, error;
4272 	int indx;
4273 
4274 	error = priv_check(td, PRIV_VFS_FHOPEN);
4275 	if (error)
4276 		return (error);
4277 	indx = -1;
4278 	fmode = FFLAGS(uap->flags);
4279 	/* why not allow a non-read/write open for our lockd? */
4280 	if (((fmode & (FREAD | FWRITE)) == 0) || (fmode & O_CREAT))
4281 		return (EINVAL);
4282 	error = copyin(uap->u_fhp, &fhp, sizeof(fhp));
4283 	if (error)
4284 		return(error);
4285 	/* find the mount point */
4286 	mp = vfs_busyfs(&fhp.fh_fsid);
4287 	if (mp == NULL)
4288 		return (ESTALE);
4289 	/* now give me my vnode, it gets returned to me locked */
4290 	error = VFS_FHTOVP(mp, &fhp.fh_fid, LK_EXCLUSIVE, &vp);
4291 	vfs_unbusy(mp);
4292 	if (error)
4293 		return (error);
4294 
4295 	error = falloc_noinstall(td, &fp);
4296 	if (error) {
4297 		vput(vp);
4298 		return (error);
4299 	}
4300 	/*
4301 	 * An extra reference on `fp' has been held for us by
4302 	 * falloc_noinstall().
4303 	 */
4304 
4305 #ifdef INVARIANTS
4306 	td->td_dupfd = -1;
4307 #endif
4308 	error = vn_open_vnode(vp, fmode, td->td_ucred, td, fp);
4309 	if (error) {
4310 		KASSERT(fp->f_ops == &badfileops,
4311 		    ("VOP_OPEN in fhopen() set f_ops"));
4312 		KASSERT(td->td_dupfd < 0,
4313 		    ("fhopen() encountered fdopen()"));
4314 
4315 		vput(vp);
4316 		goto bad;
4317 	}
4318 #ifdef INVARIANTS
4319 	td->td_dupfd = 0;
4320 #endif
4321 	fp->f_vnode = vp;
4322 	fp->f_seqcount = 1;
4323 	finit(fp, (fmode & FMASK) | (fp->f_flag & FHASLOCK), DTYPE_VNODE, vp,
4324 	    &vnops);
4325 	VOP_UNLOCK(vp, 0);
4326 	if (fmode & O_TRUNC) {
4327 		error = fo_truncate(fp, 0, td->td_ucred, td);
4328 		if (error)
4329 			goto bad;
4330 	}
4331 
4332 	error = finstall(td, fp, &indx, fmode, NULL);
4333 bad:
4334 	fdrop(fp, td);
4335 	td->td_retval[0] = indx;
4336 	return (error);
4337 }
4338 
4339 /*
4340  * Stat an (NFS) file handle.
4341  */
4342 #ifndef _SYS_SYSPROTO_H_
4343 struct fhstat_args {
4344 	struct fhandle *u_fhp;
4345 	struct stat *sb;
4346 };
4347 #endif
4348 int
4349 sys_fhstat(td, uap)
4350 	struct thread *td;
4351 	register struct fhstat_args /* {
4352 		struct fhandle *u_fhp;
4353 		struct stat *sb;
4354 	} */ *uap;
4355 {
4356 	struct stat sb;
4357 	struct fhandle fh;
4358 	int error;
4359 
4360 	error = copyin(uap->u_fhp, &fh, sizeof(fh));
4361 	if (error != 0)
4362 		return (error);
4363 	error = kern_fhstat(td, fh, &sb);
4364 	if (error != 0)
4365 		return (error);
4366 	error = copyout(&sb, uap->sb, sizeof(sb));
4367 	return (error);
4368 }
4369 
4370 int
4371 kern_fhstat(struct thread *td, struct fhandle fh, struct stat *sb)
4372 {
4373 	struct mount *mp;
4374 	struct vnode *vp;
4375 	int error;
4376 
4377 	error = priv_check(td, PRIV_VFS_FHSTAT);
4378 	if (error)
4379 		return (error);
4380 	if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4381 		return (ESTALE);
4382 	error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4383 	vfs_unbusy(mp);
4384 	if (error)
4385 		return (error);
4386 	error = vn_stat(vp, sb, td->td_ucred, NOCRED, td);
4387 	vput(vp);
4388 	return (error);
4389 }
4390 
4391 /*
4392  * Implement fstatfs() for (NFS) file handles.
4393  */
4394 #ifndef _SYS_SYSPROTO_H_
4395 struct fhstatfs_args {
4396 	struct fhandle *u_fhp;
4397 	struct statfs *buf;
4398 };
4399 #endif
4400 int
4401 sys_fhstatfs(td, uap)
4402 	struct thread *td;
4403 	struct fhstatfs_args /* {
4404 		struct fhandle *u_fhp;
4405 		struct statfs *buf;
4406 	} */ *uap;
4407 {
4408 	struct statfs sf;
4409 	fhandle_t fh;
4410 	int error;
4411 
4412 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
4413 	if (error)
4414 		return (error);
4415 	error = kern_fhstatfs(td, fh, &sf);
4416 	if (error)
4417 		return (error);
4418 	return (copyout(&sf, uap->buf, sizeof(sf)));
4419 }
4420 
4421 int
4422 kern_fhstatfs(struct thread *td, fhandle_t fh, struct statfs *buf)
4423 {
4424 	struct statfs *sp;
4425 	struct mount *mp;
4426 	struct vnode *vp;
4427 	int error;
4428 
4429 	error = priv_check(td, PRIV_VFS_FHSTATFS);
4430 	if (error)
4431 		return (error);
4432 	if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4433 		return (ESTALE);
4434 	error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4435 	if (error) {
4436 		vfs_unbusy(mp);
4437 		return (error);
4438 	}
4439 	vput(vp);
4440 	error = prison_canseemount(td->td_ucred, mp);
4441 	if (error)
4442 		goto out;
4443 #ifdef MAC
4444 	error = mac_mount_check_stat(td->td_ucred, mp);
4445 	if (error)
4446 		goto out;
4447 #endif
4448 	/*
4449 	 * Set these in case the underlying filesystem fails to do so.
4450 	 */
4451 	sp = &mp->mnt_stat;
4452 	sp->f_version = STATFS_VERSION;
4453 	sp->f_namemax = NAME_MAX;
4454 	sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
4455 	error = VFS_STATFS(mp, sp);
4456 	if (error == 0)
4457 		*buf = *sp;
4458 out:
4459 	vfs_unbusy(mp);
4460 	return (error);
4461 }
4462 
4463 int
4464 kern_posix_fallocate(struct thread *td, int fd, off_t offset, off_t len)
4465 {
4466 	struct file *fp;
4467 	struct mount *mp;
4468 	struct vnode *vp;
4469 	off_t olen, ooffset;
4470 	int error;
4471 
4472 	fp = NULL;
4473 	error = fget(td, fd, CAP_WRITE, &fp);
4474 	if (error != 0)
4475 		goto out;
4476 
4477 	switch (fp->f_type) {
4478 	case DTYPE_VNODE:
4479 		break;
4480 	case DTYPE_PIPE:
4481 	case DTYPE_FIFO:
4482 		error = ESPIPE;
4483 		goto out;
4484 	default:
4485 		error = ENODEV;
4486 		goto out;
4487 	}
4488 	if ((fp->f_flag & FWRITE) == 0) {
4489 		error = EBADF;
4490 		goto out;
4491 	}
4492 	vp = fp->f_vnode;
4493 	if (vp->v_type != VREG) {
4494 		error = ENODEV;
4495 		goto out;
4496 	}
4497 	if (offset < 0 || len <= 0) {
4498 		error = EINVAL;
4499 		goto out;
4500 	}
4501 	/* Check for wrap. */
4502 	if (offset > OFF_MAX - len) {
4503 		error = EFBIG;
4504 		goto out;
4505 	}
4506 
4507 	/* Allocating blocks may take a long time, so iterate. */
4508 	for (;;) {
4509 		olen = len;
4510 		ooffset = offset;
4511 
4512 		bwillwrite();
4513 		mp = NULL;
4514 		error = vn_start_write(vp, &mp, V_WAIT | PCATCH);
4515 		if (error != 0)
4516 			break;
4517 		error = vn_lock(vp, LK_EXCLUSIVE);
4518 		if (error != 0) {
4519 			vn_finished_write(mp);
4520 			break;
4521 		}
4522 #ifdef MAC
4523 		error = mac_vnode_check_write(td->td_ucred, fp->f_cred, vp);
4524 		if (error == 0)
4525 #endif
4526 			error = VOP_ALLOCATE(vp, &offset, &len);
4527 		VOP_UNLOCK(vp, 0);
4528 		vn_finished_write(mp);
4529 
4530 		if (olen + ooffset != offset + len) {
4531 			panic("offset + len changed from %jx/%jx to %jx/%jx",
4532 			    ooffset, olen, offset, len);
4533 		}
4534 		if (error != 0 || len == 0)
4535 			break;
4536 		KASSERT(olen > len, ("Iteration did not make progress?"));
4537 		maybe_yield();
4538 	}
4539  out:
4540 	if (fp != NULL)
4541 		fdrop(fp, td);
4542 	return (error);
4543 }
4544 
4545 int
4546 sys_posix_fallocate(struct thread *td, struct posix_fallocate_args *uap)
4547 {
4548 
4549 	return (kern_posix_fallocate(td, uap->fd, uap->offset, uap->len));
4550 }
4551 
4552 /*
4553  * Unlike madvise(2), we do not make a best effort to remember every
4554  * possible caching hint.  Instead, we remember the last setting with
4555  * the exception that we will allow POSIX_FADV_NORMAL to adjust the
4556  * region of any current setting.
4557  */
4558 int
4559 kern_posix_fadvise(struct thread *td, int fd, off_t offset, off_t len,
4560     int advice)
4561 {
4562 	struct fadvise_info *fa, *new;
4563 	struct file *fp;
4564 	struct vnode *vp;
4565 	off_t end;
4566 	int error;
4567 
4568 	if (offset < 0 || len < 0 || offset > OFF_MAX - len)
4569 		return (EINVAL);
4570 	switch (advice) {
4571 	case POSIX_FADV_SEQUENTIAL:
4572 	case POSIX_FADV_RANDOM:
4573 	case POSIX_FADV_NOREUSE:
4574 		new = malloc(sizeof(*fa), M_FADVISE, M_WAITOK);
4575 		break;
4576 	case POSIX_FADV_NORMAL:
4577 	case POSIX_FADV_WILLNEED:
4578 	case POSIX_FADV_DONTNEED:
4579 		new = NULL;
4580 		break;
4581 	default:
4582 		return (EINVAL);
4583 	}
4584 	/* XXX: CAP_POSIX_FADVISE? */
4585 	error = fget(td, fd, CAP_NONE, &fp);
4586 	if (error != 0)
4587 		goto out;
4588 
4589 	switch (fp->f_type) {
4590 	case DTYPE_VNODE:
4591 		break;
4592 	case DTYPE_PIPE:
4593 	case DTYPE_FIFO:
4594 		error = ESPIPE;
4595 		goto out;
4596 	default:
4597 		error = ENODEV;
4598 		goto out;
4599 	}
4600 	vp = fp->f_vnode;
4601 	if (vp->v_type != VREG) {
4602 		error = ENODEV;
4603 		goto out;
4604 	}
4605 	if (len == 0)
4606 		end = OFF_MAX;
4607 	else
4608 		end = offset + len - 1;
4609 	switch (advice) {
4610 	case POSIX_FADV_SEQUENTIAL:
4611 	case POSIX_FADV_RANDOM:
4612 	case POSIX_FADV_NOREUSE:
4613 		/*
4614 		 * Try to merge any existing non-standard region with
4615 		 * this new region if possible, otherwise create a new
4616 		 * non-standard region for this request.
4617 		 */
4618 		mtx_pool_lock(mtxpool_sleep, fp);
4619 		fa = fp->f_advice;
4620 		if (fa != NULL && fa->fa_advice == advice &&
4621 		    ((fa->fa_start <= end && fa->fa_end >= offset) ||
4622 		    (end != OFF_MAX && fa->fa_start == end + 1) ||
4623 		    (fa->fa_end != OFF_MAX && fa->fa_end + 1 == offset))) {
4624 			if (offset < fa->fa_start)
4625 				fa->fa_start = offset;
4626 			if (end > fa->fa_end)
4627 				fa->fa_end = end;
4628 		} else {
4629 			new->fa_advice = advice;
4630 			new->fa_start = offset;
4631 			new->fa_end = end;
4632 			new->fa_prevstart = 0;
4633 			new->fa_prevend = 0;
4634 			fp->f_advice = new;
4635 			new = fa;
4636 		}
4637 		mtx_pool_unlock(mtxpool_sleep, fp);
4638 		break;
4639 	case POSIX_FADV_NORMAL:
4640 		/*
4641 		 * If a the "normal" region overlaps with an existing
4642 		 * non-standard region, trim or remove the
4643 		 * non-standard region.
4644 		 */
4645 		mtx_pool_lock(mtxpool_sleep, fp);
4646 		fa = fp->f_advice;
4647 		if (fa != NULL) {
4648 			if (offset <= fa->fa_start && end >= fa->fa_end) {
4649 				new = fa;
4650 				fp->f_advice = NULL;
4651 			} else if (offset <= fa->fa_start &&
4652 			    end >= fa->fa_start)
4653 				fa->fa_start = end + 1;
4654 			else if (offset <= fa->fa_end && end >= fa->fa_end)
4655 				fa->fa_end = offset - 1;
4656 			else if (offset >= fa->fa_start && end <= fa->fa_end) {
4657 				/*
4658 				 * If the "normal" region is a middle
4659 				 * portion of the existing
4660 				 * non-standard region, just remove
4661 				 * the whole thing rather than picking
4662 				 * one side or the other to
4663 				 * preserve.
4664 				 */
4665 				new = fa;
4666 				fp->f_advice = NULL;
4667 			}
4668 		}
4669 		mtx_pool_unlock(mtxpool_sleep, fp);
4670 		break;
4671 	case POSIX_FADV_WILLNEED:
4672 	case POSIX_FADV_DONTNEED:
4673 		error = VOP_ADVISE(vp, offset, end, advice);
4674 		break;
4675 	}
4676 out:
4677 	if (fp != NULL)
4678 		fdrop(fp, td);
4679 	free(new, M_FADVISE);
4680 	return (error);
4681 }
4682 
4683 int
4684 sys_posix_fadvise(struct thread *td, struct posix_fadvise_args *uap)
4685 {
4686 
4687 	return (kern_posix_fadvise(td, uap->fd, uap->offset, uap->len,
4688 	    uap->advice));
4689 }
4690