xref: /freebsd/sys/kern/vfs_syscalls.c (revision 8f87df16d414c7b09739ba95030e1e5650207b73)
1 /*-
2  * Copyright (c) 1989, 1993
3  *	The Regents of the University of California.  All rights reserved.
4  * (c) UNIX System Laboratories, Inc.
5  * All or some portions of this file are derived from material licensed
6  * to the University of California by American Telephone and Telegraph
7  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
8  * the permission of UNIX System Laboratories, Inc.
9  *
10  * Redistribution and use in source and binary forms, with or without
11  * modification, are permitted provided that the following conditions
12  * are met:
13  * 1. Redistributions of source code must retain the above copyright
14  *    notice, this list of conditions and the following disclaimer.
15  * 2. Redistributions in binary form must reproduce the above copyright
16  *    notice, this list of conditions and the following disclaimer in the
17  *    documentation and/or other materials provided with the distribution.
18  * 3. Neither the name of the University nor the names of its contributors
19  *    may be used to endorse or promote products derived from this software
20  *    without specific prior written permission.
21  *
22  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
23  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
24  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
25  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
26  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
27  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
28  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
29  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
30  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
31  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32  * SUCH DAMAGE.
33  *
34  *	@(#)vfs_syscalls.c	8.13 (Berkeley) 4/15/94
35  */
36 
37 #include <sys/cdefs.h>
38 __FBSDID("$FreeBSD$");
39 
40 #include "opt_capsicum.h"
41 #include "opt_compat.h"
42 #include "opt_ktrace.h"
43 
44 #include <sys/param.h>
45 #include <sys/systm.h>
46 #include <sys/bio.h>
47 #include <sys/buf.h>
48 #include <sys/capsicum.h>
49 #include <sys/disk.h>
50 #include <sys/sysent.h>
51 #include <sys/malloc.h>
52 #include <sys/mount.h>
53 #include <sys/mutex.h>
54 #include <sys/sysproto.h>
55 #include <sys/namei.h>
56 #include <sys/filedesc.h>
57 #include <sys/kernel.h>
58 #include <sys/fcntl.h>
59 #include <sys/file.h>
60 #include <sys/filio.h>
61 #include <sys/limits.h>
62 #include <sys/linker.h>
63 #include <sys/rwlock.h>
64 #include <sys/sdt.h>
65 #include <sys/stat.h>
66 #include <sys/sx.h>
67 #include <sys/unistd.h>
68 #include <sys/vnode.h>
69 #include <sys/priv.h>
70 #include <sys/proc.h>
71 #include <sys/dirent.h>
72 #include <sys/jail.h>
73 #include <sys/syscallsubr.h>
74 #include <sys/sysctl.h>
75 #ifdef KTRACE
76 #include <sys/ktrace.h>
77 #endif
78 
79 #include <machine/stdarg.h>
80 
81 #include <security/audit/audit.h>
82 #include <security/mac/mac_framework.h>
83 
84 #include <vm/vm.h>
85 #include <vm/vm_object.h>
86 #include <vm/vm_page.h>
87 #include <vm/uma.h>
88 
89 #include <ufs/ufs/quota.h>
90 
91 MALLOC_DEFINE(M_FADVISE, "fadvise", "posix_fadvise(2) information");
92 
93 SDT_PROVIDER_DEFINE(vfs);
94 SDT_PROBE_DEFINE2(vfs, , stat, mode, "char *", "int");
95 SDT_PROBE_DEFINE2(vfs, , stat, reg, "char *", "int");
96 
97 static int kern_chflagsat(struct thread *td, int fd, const char *path,
98     enum uio_seg pathseg, u_long flags, int atflag);
99 static int setfflags(struct thread *td, struct vnode *, u_long);
100 static int getutimes(const struct timeval *, enum uio_seg, struct timespec *);
101 static int getutimens(const struct timespec *, enum uio_seg,
102     struct timespec *, int *);
103 static int setutimes(struct thread *td, struct vnode *,
104     const struct timespec *, int, int);
105 static int vn_access(struct vnode *vp, int user_flags, struct ucred *cred,
106     struct thread *td);
107 
108 /*
109  * Sync each mounted filesystem.
110  */
111 #ifndef _SYS_SYSPROTO_H_
112 struct sync_args {
113 	int     dummy;
114 };
115 #endif
116 /* ARGSUSED */
117 int
118 sys_sync(td, uap)
119 	struct thread *td;
120 	struct sync_args *uap;
121 {
122 	struct mount *mp, *nmp;
123 	int save;
124 
125 	mtx_lock(&mountlist_mtx);
126 	for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
127 		if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK)) {
128 			nmp = TAILQ_NEXT(mp, mnt_list);
129 			continue;
130 		}
131 		if ((mp->mnt_flag & MNT_RDONLY) == 0 &&
132 		    vn_start_write(NULL, &mp, V_NOWAIT) == 0) {
133 			save = curthread_pflags_set(TDP_SYNCIO);
134 			vfs_msync(mp, MNT_NOWAIT);
135 			VFS_SYNC(mp, MNT_NOWAIT);
136 			curthread_pflags_restore(save);
137 			vn_finished_write(mp);
138 		}
139 		mtx_lock(&mountlist_mtx);
140 		nmp = TAILQ_NEXT(mp, mnt_list);
141 		vfs_unbusy(mp);
142 	}
143 	mtx_unlock(&mountlist_mtx);
144 	return (0);
145 }
146 
147 /*
148  * Change filesystem quotas.
149  */
150 #ifndef _SYS_SYSPROTO_H_
151 struct quotactl_args {
152 	char *path;
153 	int cmd;
154 	int uid;
155 	caddr_t arg;
156 };
157 #endif
158 int
159 sys_quotactl(td, uap)
160 	struct thread *td;
161 	register struct quotactl_args /* {
162 		char *path;
163 		int cmd;
164 		int uid;
165 		caddr_t arg;
166 	} */ *uap;
167 {
168 	struct mount *mp;
169 	struct nameidata nd;
170 	int error;
171 
172 	AUDIT_ARG_CMD(uap->cmd);
173 	AUDIT_ARG_UID(uap->uid);
174 	if (!prison_allow(td->td_ucred, PR_ALLOW_QUOTAS))
175 		return (EPERM);
176 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
177 	    uap->path, td);
178 	if ((error = namei(&nd)) != 0)
179 		return (error);
180 	NDFREE(&nd, NDF_ONLY_PNBUF);
181 	mp = nd.ni_vp->v_mount;
182 	vfs_ref(mp);
183 	vput(nd.ni_vp);
184 	error = vfs_busy(mp, 0);
185 	vfs_rel(mp);
186 	if (error != 0)
187 		return (error);
188 	error = VFS_QUOTACTL(mp, uap->cmd, uap->uid, uap->arg);
189 
190 	/*
191 	 * Since quota on operation typically needs to open quota
192 	 * file, the Q_QUOTAON handler needs to unbusy the mount point
193 	 * before calling into namei.  Otherwise, unmount might be
194 	 * started between two vfs_busy() invocations (first is our,
195 	 * second is from mount point cross-walk code in lookup()),
196 	 * causing deadlock.
197 	 *
198 	 * Require that Q_QUOTAON handles the vfs_busy() reference on
199 	 * its own, always returning with ubusied mount point.
200 	 */
201 	if ((uap->cmd >> SUBCMDSHIFT) != Q_QUOTAON)
202 		vfs_unbusy(mp);
203 	return (error);
204 }
205 
206 /*
207  * Used by statfs conversion routines to scale the block size up if
208  * necessary so that all of the block counts are <= 'max_size'.  Note
209  * that 'max_size' should be a bitmask, i.e. 2^n - 1 for some non-zero
210  * value of 'n'.
211  */
212 void
213 statfs_scale_blocks(struct statfs *sf, long max_size)
214 {
215 	uint64_t count;
216 	int shift;
217 
218 	KASSERT(powerof2(max_size + 1), ("%s: invalid max_size", __func__));
219 
220 	/*
221 	 * Attempt to scale the block counts to give a more accurate
222 	 * overview to userland of the ratio of free space to used
223 	 * space.  To do this, find the largest block count and compute
224 	 * a divisor that lets it fit into a signed integer <= max_size.
225 	 */
226 	if (sf->f_bavail < 0)
227 		count = -sf->f_bavail;
228 	else
229 		count = sf->f_bavail;
230 	count = MAX(sf->f_blocks, MAX(sf->f_bfree, count));
231 	if (count <= max_size)
232 		return;
233 
234 	count >>= flsl(max_size);
235 	shift = 0;
236 	while (count > 0) {
237 		shift++;
238 		count >>=1;
239 	}
240 
241 	sf->f_bsize <<= shift;
242 	sf->f_blocks >>= shift;
243 	sf->f_bfree >>= shift;
244 	sf->f_bavail >>= shift;
245 }
246 
247 /*
248  * Get filesystem statistics.
249  */
250 #ifndef _SYS_SYSPROTO_H_
251 struct statfs_args {
252 	char *path;
253 	struct statfs *buf;
254 };
255 #endif
256 int
257 sys_statfs(td, uap)
258 	struct thread *td;
259 	register struct statfs_args /* {
260 		char *path;
261 		struct statfs *buf;
262 	} */ *uap;
263 {
264 	struct statfs sf;
265 	int error;
266 
267 	error = kern_statfs(td, uap->path, UIO_USERSPACE, &sf);
268 	if (error == 0)
269 		error = copyout(&sf, uap->buf, sizeof(sf));
270 	return (error);
271 }
272 
273 int
274 kern_statfs(struct thread *td, char *path, enum uio_seg pathseg,
275     struct statfs *buf)
276 {
277 	struct mount *mp;
278 	struct statfs *sp, sb;
279 	struct nameidata nd;
280 	int error;
281 
282 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
283 	    pathseg, path, td);
284 	error = namei(&nd);
285 	if (error != 0)
286 		return (error);
287 	mp = nd.ni_vp->v_mount;
288 	vfs_ref(mp);
289 	NDFREE(&nd, NDF_ONLY_PNBUF);
290 	vput(nd.ni_vp);
291 	error = vfs_busy(mp, 0);
292 	vfs_rel(mp);
293 	if (error != 0)
294 		return (error);
295 #ifdef MAC
296 	error = mac_mount_check_stat(td->td_ucred, mp);
297 	if (error != 0)
298 		goto out;
299 #endif
300 	/*
301 	 * Set these in case the underlying filesystem fails to do so.
302 	 */
303 	sp = &mp->mnt_stat;
304 	sp->f_version = STATFS_VERSION;
305 	sp->f_namemax = NAME_MAX;
306 	sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
307 	error = VFS_STATFS(mp, sp);
308 	if (error != 0)
309 		goto out;
310 	if (priv_check(td, PRIV_VFS_GENERATION)) {
311 		bcopy(sp, &sb, sizeof(sb));
312 		sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
313 		prison_enforce_statfs(td->td_ucred, mp, &sb);
314 		sp = &sb;
315 	}
316 	*buf = *sp;
317 out:
318 	vfs_unbusy(mp);
319 	return (error);
320 }
321 
322 /*
323  * Get filesystem statistics.
324  */
325 #ifndef _SYS_SYSPROTO_H_
326 struct fstatfs_args {
327 	int fd;
328 	struct statfs *buf;
329 };
330 #endif
331 int
332 sys_fstatfs(td, uap)
333 	struct thread *td;
334 	register struct fstatfs_args /* {
335 		int fd;
336 		struct statfs *buf;
337 	} */ *uap;
338 {
339 	struct statfs sf;
340 	int error;
341 
342 	error = kern_fstatfs(td, uap->fd, &sf);
343 	if (error == 0)
344 		error = copyout(&sf, uap->buf, sizeof(sf));
345 	return (error);
346 }
347 
348 int
349 kern_fstatfs(struct thread *td, int fd, struct statfs *buf)
350 {
351 	struct file *fp;
352 	struct mount *mp;
353 	struct statfs *sp, sb;
354 	struct vnode *vp;
355 	cap_rights_t rights;
356 	int error;
357 
358 	AUDIT_ARG_FD(fd);
359 	error = getvnode(td, fd, cap_rights_init(&rights, CAP_FSTATFS), &fp);
360 	if (error != 0)
361 		return (error);
362 	vp = fp->f_vnode;
363 	vn_lock(vp, LK_SHARED | LK_RETRY);
364 #ifdef AUDIT
365 	AUDIT_ARG_VNODE1(vp);
366 #endif
367 	mp = vp->v_mount;
368 	if (mp)
369 		vfs_ref(mp);
370 	VOP_UNLOCK(vp, 0);
371 	fdrop(fp, td);
372 	if (mp == NULL) {
373 		error = EBADF;
374 		goto out;
375 	}
376 	error = vfs_busy(mp, 0);
377 	vfs_rel(mp);
378 	if (error != 0)
379 		return (error);
380 #ifdef MAC
381 	error = mac_mount_check_stat(td->td_ucred, mp);
382 	if (error != 0)
383 		goto out;
384 #endif
385 	/*
386 	 * Set these in case the underlying filesystem fails to do so.
387 	 */
388 	sp = &mp->mnt_stat;
389 	sp->f_version = STATFS_VERSION;
390 	sp->f_namemax = NAME_MAX;
391 	sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
392 	error = VFS_STATFS(mp, sp);
393 	if (error != 0)
394 		goto out;
395 	if (priv_check(td, PRIV_VFS_GENERATION)) {
396 		bcopy(sp, &sb, sizeof(sb));
397 		sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
398 		prison_enforce_statfs(td->td_ucred, mp, &sb);
399 		sp = &sb;
400 	}
401 	*buf = *sp;
402 out:
403 	if (mp)
404 		vfs_unbusy(mp);
405 	return (error);
406 }
407 
408 /*
409  * Get statistics on all filesystems.
410  */
411 #ifndef _SYS_SYSPROTO_H_
412 struct getfsstat_args {
413 	struct statfs *buf;
414 	long bufsize;
415 	int mode;
416 };
417 #endif
418 int
419 sys_getfsstat(td, uap)
420 	struct thread *td;
421 	register struct getfsstat_args /* {
422 		struct statfs *buf;
423 		long bufsize;
424 		int mode;
425 	} */ *uap;
426 {
427 	size_t count;
428 	int error;
429 
430 	if (uap->bufsize < 0 || uap->bufsize > SIZE_MAX)
431 		return (EINVAL);
432 	error = kern_getfsstat(td, &uap->buf, uap->bufsize, &count,
433 	    UIO_USERSPACE, uap->mode);
434 	if (error == 0)
435 		td->td_retval[0] = count;
436 	return (error);
437 }
438 
439 /*
440  * If (bufsize > 0 && bufseg == UIO_SYSSPACE)
441  *	The caller is responsible for freeing memory which will be allocated
442  *	in '*buf'.
443  */
444 int
445 kern_getfsstat(struct thread *td, struct statfs **buf, size_t bufsize,
446     size_t *countp, enum uio_seg bufseg, int mode)
447 {
448 	struct mount *mp, *nmp;
449 	struct statfs *sfsp, *sp, sb, *tofree;
450 	size_t count, maxcount;
451 	int error;
452 
453 	switch (mode) {
454 	case MNT_WAIT:
455 	case MNT_NOWAIT:
456 		break;
457 	default:
458 		return (EINVAL);
459 	}
460 restart:
461 	maxcount = bufsize / sizeof(struct statfs);
462 	if (bufsize == 0) {
463 		sfsp = NULL;
464 		tofree = NULL;
465 	} else if (bufseg == UIO_USERSPACE) {
466 		sfsp = *buf;
467 		tofree = NULL;
468 	} else /* if (bufseg == UIO_SYSSPACE) */ {
469 		count = 0;
470 		mtx_lock(&mountlist_mtx);
471 		TAILQ_FOREACH(mp, &mountlist, mnt_list) {
472 			count++;
473 		}
474 		mtx_unlock(&mountlist_mtx);
475 		if (maxcount > count)
476 			maxcount = count;
477 		tofree = sfsp = *buf = malloc(maxcount * sizeof(struct statfs),
478 		    M_TEMP, M_WAITOK);
479 	}
480 	count = 0;
481 	mtx_lock(&mountlist_mtx);
482 	for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
483 		if (prison_canseemount(td->td_ucred, mp) != 0) {
484 			nmp = TAILQ_NEXT(mp, mnt_list);
485 			continue;
486 		}
487 #ifdef MAC
488 		if (mac_mount_check_stat(td->td_ucred, mp) != 0) {
489 			nmp = TAILQ_NEXT(mp, mnt_list);
490 			continue;
491 		}
492 #endif
493 		if (mode == MNT_WAIT) {
494 			if (vfs_busy(mp, MBF_MNTLSTLOCK) != 0) {
495 				/*
496 				 * If vfs_busy() failed, and MBF_NOWAIT
497 				 * wasn't passed, then the mp is gone.
498 				 * Furthermore, because of MBF_MNTLSTLOCK,
499 				 * the mountlist_mtx was dropped.  We have
500 				 * no other choice than to start over.
501 				 */
502 				mtx_unlock(&mountlist_mtx);
503 				free(tofree, M_TEMP);
504 				goto restart;
505 			}
506 		} else {
507 			if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK) != 0) {
508 				nmp = TAILQ_NEXT(mp, mnt_list);
509 				continue;
510 			}
511 		}
512 		if (sfsp && count < maxcount) {
513 			sp = &mp->mnt_stat;
514 			/*
515 			 * Set these in case the underlying filesystem
516 			 * fails to do so.
517 			 */
518 			sp->f_version = STATFS_VERSION;
519 			sp->f_namemax = NAME_MAX;
520 			sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
521 			/*
522 			 * If MNT_NOWAIT is specified, do not refresh
523 			 * the fsstat cache.
524 			 */
525 			if (mode != MNT_NOWAIT) {
526 				error = VFS_STATFS(mp, sp);
527 				if (error != 0) {
528 					mtx_lock(&mountlist_mtx);
529 					nmp = TAILQ_NEXT(mp, mnt_list);
530 					vfs_unbusy(mp);
531 					continue;
532 				}
533 			}
534 			if (priv_check(td, PRIV_VFS_GENERATION)) {
535 				bcopy(sp, &sb, sizeof(sb));
536 				sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
537 				prison_enforce_statfs(td->td_ucred, mp, &sb);
538 				sp = &sb;
539 			}
540 			if (bufseg == UIO_SYSSPACE)
541 				bcopy(sp, sfsp, sizeof(*sp));
542 			else /* if (bufseg == UIO_USERSPACE) */ {
543 				error = copyout(sp, sfsp, sizeof(*sp));
544 				if (error != 0) {
545 					vfs_unbusy(mp);
546 					return (error);
547 				}
548 			}
549 			sfsp++;
550 		}
551 		count++;
552 		mtx_lock(&mountlist_mtx);
553 		nmp = TAILQ_NEXT(mp, mnt_list);
554 		vfs_unbusy(mp);
555 	}
556 	mtx_unlock(&mountlist_mtx);
557 	if (sfsp && count > maxcount)
558 		*countp = maxcount;
559 	else
560 		*countp = count;
561 	return (0);
562 }
563 
564 #ifdef COMPAT_FREEBSD4
565 /*
566  * Get old format filesystem statistics.
567  */
568 static void cvtstatfs(struct statfs *, struct ostatfs *);
569 
570 #ifndef _SYS_SYSPROTO_H_
571 struct freebsd4_statfs_args {
572 	char *path;
573 	struct ostatfs *buf;
574 };
575 #endif
576 int
577 freebsd4_statfs(td, uap)
578 	struct thread *td;
579 	struct freebsd4_statfs_args /* {
580 		char *path;
581 		struct ostatfs *buf;
582 	} */ *uap;
583 {
584 	struct ostatfs osb;
585 	struct statfs sf;
586 	int error;
587 
588 	error = kern_statfs(td, uap->path, UIO_USERSPACE, &sf);
589 	if (error != 0)
590 		return (error);
591 	cvtstatfs(&sf, &osb);
592 	return (copyout(&osb, uap->buf, sizeof(osb)));
593 }
594 
595 /*
596  * Get filesystem statistics.
597  */
598 #ifndef _SYS_SYSPROTO_H_
599 struct freebsd4_fstatfs_args {
600 	int fd;
601 	struct ostatfs *buf;
602 };
603 #endif
604 int
605 freebsd4_fstatfs(td, uap)
606 	struct thread *td;
607 	struct freebsd4_fstatfs_args /* {
608 		int fd;
609 		struct ostatfs *buf;
610 	} */ *uap;
611 {
612 	struct ostatfs osb;
613 	struct statfs sf;
614 	int error;
615 
616 	error = kern_fstatfs(td, uap->fd, &sf);
617 	if (error != 0)
618 		return (error);
619 	cvtstatfs(&sf, &osb);
620 	return (copyout(&osb, uap->buf, sizeof(osb)));
621 }
622 
623 /*
624  * Get statistics on all filesystems.
625  */
626 #ifndef _SYS_SYSPROTO_H_
627 struct freebsd4_getfsstat_args {
628 	struct ostatfs *buf;
629 	long bufsize;
630 	int mode;
631 };
632 #endif
633 int
634 freebsd4_getfsstat(td, uap)
635 	struct thread *td;
636 	register struct freebsd4_getfsstat_args /* {
637 		struct ostatfs *buf;
638 		long bufsize;
639 		int mode;
640 	} */ *uap;
641 {
642 	struct statfs *buf, *sp;
643 	struct ostatfs osb;
644 	size_t count, size;
645 	int error;
646 
647 	if (uap->bufsize < 0)
648 		return (EINVAL);
649 	count = uap->bufsize / sizeof(struct ostatfs);
650 	if (count > SIZE_MAX / sizeof(struct statfs))
651 		return (EINVAL);
652 	size = count * sizeof(struct statfs);
653 	error = kern_getfsstat(td, &buf, size, &count, UIO_SYSSPACE,
654 	    uap->mode);
655 	td->td_retval[0] = count;
656 	if (size != 0) {
657 		sp = buf;
658 		while (count != 0 && error == 0) {
659 			cvtstatfs(sp, &osb);
660 			error = copyout(&osb, uap->buf, sizeof(osb));
661 			sp++;
662 			uap->buf++;
663 			count--;
664 		}
665 		free(buf, M_TEMP);
666 	}
667 	return (error);
668 }
669 
670 /*
671  * Implement fstatfs() for (NFS) file handles.
672  */
673 #ifndef _SYS_SYSPROTO_H_
674 struct freebsd4_fhstatfs_args {
675 	struct fhandle *u_fhp;
676 	struct ostatfs *buf;
677 };
678 #endif
679 int
680 freebsd4_fhstatfs(td, uap)
681 	struct thread *td;
682 	struct freebsd4_fhstatfs_args /* {
683 		struct fhandle *u_fhp;
684 		struct ostatfs *buf;
685 	} */ *uap;
686 {
687 	struct ostatfs osb;
688 	struct statfs sf;
689 	fhandle_t fh;
690 	int error;
691 
692 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
693 	if (error != 0)
694 		return (error);
695 	error = kern_fhstatfs(td, fh, &sf);
696 	if (error != 0)
697 		return (error);
698 	cvtstatfs(&sf, &osb);
699 	return (copyout(&osb, uap->buf, sizeof(osb)));
700 }
701 
702 /*
703  * Convert a new format statfs structure to an old format statfs structure.
704  */
705 static void
706 cvtstatfs(nsp, osp)
707 	struct statfs *nsp;
708 	struct ostatfs *osp;
709 {
710 
711 	statfs_scale_blocks(nsp, LONG_MAX);
712 	bzero(osp, sizeof(*osp));
713 	osp->f_bsize = nsp->f_bsize;
714 	osp->f_iosize = MIN(nsp->f_iosize, LONG_MAX);
715 	osp->f_blocks = nsp->f_blocks;
716 	osp->f_bfree = nsp->f_bfree;
717 	osp->f_bavail = nsp->f_bavail;
718 	osp->f_files = MIN(nsp->f_files, LONG_MAX);
719 	osp->f_ffree = MIN(nsp->f_ffree, LONG_MAX);
720 	osp->f_owner = nsp->f_owner;
721 	osp->f_type = nsp->f_type;
722 	osp->f_flags = nsp->f_flags;
723 	osp->f_syncwrites = MIN(nsp->f_syncwrites, LONG_MAX);
724 	osp->f_asyncwrites = MIN(nsp->f_asyncwrites, LONG_MAX);
725 	osp->f_syncreads = MIN(nsp->f_syncreads, LONG_MAX);
726 	osp->f_asyncreads = MIN(nsp->f_asyncreads, LONG_MAX);
727 	strlcpy(osp->f_fstypename, nsp->f_fstypename,
728 	    MIN(MFSNAMELEN, OMFSNAMELEN));
729 	strlcpy(osp->f_mntonname, nsp->f_mntonname,
730 	    MIN(MNAMELEN, OMNAMELEN));
731 	strlcpy(osp->f_mntfromname, nsp->f_mntfromname,
732 	    MIN(MNAMELEN, OMNAMELEN));
733 	osp->f_fsid = nsp->f_fsid;
734 }
735 #endif /* COMPAT_FREEBSD4 */
736 
737 /*
738  * Change current working directory to a given file descriptor.
739  */
740 #ifndef _SYS_SYSPROTO_H_
741 struct fchdir_args {
742 	int	fd;
743 };
744 #endif
745 int
746 sys_fchdir(td, uap)
747 	struct thread *td;
748 	struct fchdir_args /* {
749 		int fd;
750 	} */ *uap;
751 {
752 	struct vnode *vp, *tdp;
753 	struct mount *mp;
754 	struct file *fp;
755 	cap_rights_t rights;
756 	int error;
757 
758 	AUDIT_ARG_FD(uap->fd);
759 	error = getvnode(td, uap->fd, cap_rights_init(&rights, CAP_FCHDIR),
760 	    &fp);
761 	if (error != 0)
762 		return (error);
763 	vp = fp->f_vnode;
764 	vrefact(vp);
765 	fdrop(fp, td);
766 	vn_lock(vp, LK_SHARED | LK_RETRY);
767 	AUDIT_ARG_VNODE1(vp);
768 	error = change_dir(vp, td);
769 	while (!error && (mp = vp->v_mountedhere) != NULL) {
770 		if (vfs_busy(mp, 0))
771 			continue;
772 		error = VFS_ROOT(mp, LK_SHARED, &tdp);
773 		vfs_unbusy(mp);
774 		if (error != 0)
775 			break;
776 		vput(vp);
777 		vp = tdp;
778 	}
779 	if (error != 0) {
780 		vput(vp);
781 		return (error);
782 	}
783 	VOP_UNLOCK(vp, 0);
784 	pwd_chdir(td, vp);
785 	return (0);
786 }
787 
788 /*
789  * Change current working directory (``.'').
790  */
791 #ifndef _SYS_SYSPROTO_H_
792 struct chdir_args {
793 	char	*path;
794 };
795 #endif
796 int
797 sys_chdir(td, uap)
798 	struct thread *td;
799 	struct chdir_args /* {
800 		char *path;
801 	} */ *uap;
802 {
803 
804 	return (kern_chdir(td, uap->path, UIO_USERSPACE));
805 }
806 
807 int
808 kern_chdir(struct thread *td, char *path, enum uio_seg pathseg)
809 {
810 	struct nameidata nd;
811 	int error;
812 
813 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
814 	    pathseg, path, td);
815 	if ((error = namei(&nd)) != 0)
816 		return (error);
817 	if ((error = change_dir(nd.ni_vp, td)) != 0) {
818 		vput(nd.ni_vp);
819 		NDFREE(&nd, NDF_ONLY_PNBUF);
820 		return (error);
821 	}
822 	VOP_UNLOCK(nd.ni_vp, 0);
823 	NDFREE(&nd, NDF_ONLY_PNBUF);
824 	pwd_chdir(td, nd.ni_vp);
825 	return (0);
826 }
827 
828 /*
829  * Change notion of root (``/'') directory.
830  */
831 #ifndef _SYS_SYSPROTO_H_
832 struct chroot_args {
833 	char	*path;
834 };
835 #endif
836 int
837 sys_chroot(td, uap)
838 	struct thread *td;
839 	struct chroot_args /* {
840 		char *path;
841 	} */ *uap;
842 {
843 	struct nameidata nd;
844 	int error;
845 
846 	error = priv_check(td, PRIV_VFS_CHROOT);
847 	if (error != 0)
848 		return (error);
849 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
850 	    UIO_USERSPACE, uap->path, td);
851 	error = namei(&nd);
852 	if (error != 0)
853 		goto error;
854 	error = change_dir(nd.ni_vp, td);
855 	if (error != 0)
856 		goto e_vunlock;
857 #ifdef MAC
858 	error = mac_vnode_check_chroot(td->td_ucred, nd.ni_vp);
859 	if (error != 0)
860 		goto e_vunlock;
861 #endif
862 	VOP_UNLOCK(nd.ni_vp, 0);
863 	error = pwd_chroot(td, nd.ni_vp);
864 	vrele(nd.ni_vp);
865 	NDFREE(&nd, NDF_ONLY_PNBUF);
866 	return (error);
867 e_vunlock:
868 	vput(nd.ni_vp);
869 error:
870 	NDFREE(&nd, NDF_ONLY_PNBUF);
871 	return (error);
872 }
873 
874 /*
875  * Common routine for chroot and chdir.  Callers must provide a locked vnode
876  * instance.
877  */
878 int
879 change_dir(vp, td)
880 	struct vnode *vp;
881 	struct thread *td;
882 {
883 #ifdef MAC
884 	int error;
885 #endif
886 
887 	ASSERT_VOP_LOCKED(vp, "change_dir(): vp not locked");
888 	if (vp->v_type != VDIR)
889 		return (ENOTDIR);
890 #ifdef MAC
891 	error = mac_vnode_check_chdir(td->td_ucred, vp);
892 	if (error != 0)
893 		return (error);
894 #endif
895 	return (VOP_ACCESS(vp, VEXEC, td->td_ucred, td));
896 }
897 
898 static __inline void
899 flags_to_rights(int flags, cap_rights_t *rightsp)
900 {
901 
902 	if (flags & O_EXEC) {
903 		cap_rights_set(rightsp, CAP_FEXECVE);
904 	} else {
905 		switch ((flags & O_ACCMODE)) {
906 		case O_RDONLY:
907 			cap_rights_set(rightsp, CAP_READ);
908 			break;
909 		case O_RDWR:
910 			cap_rights_set(rightsp, CAP_READ);
911 			/* FALLTHROUGH */
912 		case O_WRONLY:
913 			cap_rights_set(rightsp, CAP_WRITE);
914 			if (!(flags & (O_APPEND | O_TRUNC)))
915 				cap_rights_set(rightsp, CAP_SEEK);
916 			break;
917 		}
918 	}
919 
920 	if (flags & O_CREAT)
921 		cap_rights_set(rightsp, CAP_CREATE);
922 
923 	if (flags & O_TRUNC)
924 		cap_rights_set(rightsp, CAP_FTRUNCATE);
925 
926 	if (flags & (O_SYNC | O_FSYNC))
927 		cap_rights_set(rightsp, CAP_FSYNC);
928 
929 	if (flags & (O_EXLOCK | O_SHLOCK))
930 		cap_rights_set(rightsp, CAP_FLOCK);
931 }
932 
933 /*
934  * Check permissions, allocate an open file structure, and call the device
935  * open routine if any.
936  */
937 #ifndef _SYS_SYSPROTO_H_
938 struct open_args {
939 	char	*path;
940 	int	flags;
941 	int	mode;
942 };
943 #endif
944 int
945 sys_open(td, uap)
946 	struct thread *td;
947 	register struct open_args /* {
948 		char *path;
949 		int flags;
950 		int mode;
951 	} */ *uap;
952 {
953 
954 	return (kern_openat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
955 	    uap->flags, uap->mode));
956 }
957 
958 #ifndef _SYS_SYSPROTO_H_
959 struct openat_args {
960 	int	fd;
961 	char	*path;
962 	int	flag;
963 	int	mode;
964 };
965 #endif
966 int
967 sys_openat(struct thread *td, struct openat_args *uap)
968 {
969 
970 	AUDIT_ARG_FD(uap->fd);
971 	return (kern_openat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
972 	    uap->mode));
973 }
974 
975 int
976 kern_openat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
977     int flags, int mode)
978 {
979 	struct proc *p = td->td_proc;
980 	struct filedesc *fdp = p->p_fd;
981 	struct file *fp;
982 	struct vnode *vp;
983 	struct nameidata nd;
984 	cap_rights_t rights;
985 	int cmode, error, indx;
986 
987 	indx = -1;
988 
989 	AUDIT_ARG_FFLAGS(flags);
990 	AUDIT_ARG_MODE(mode);
991 	cap_rights_init(&rights, CAP_LOOKUP);
992 	flags_to_rights(flags, &rights);
993 	/*
994 	 * Only one of the O_EXEC, O_RDONLY, O_WRONLY and O_RDWR flags
995 	 * may be specified.
996 	 */
997 	if (flags & O_EXEC) {
998 		if (flags & O_ACCMODE)
999 			return (EINVAL);
1000 	} else if ((flags & O_ACCMODE) == O_ACCMODE) {
1001 		return (EINVAL);
1002 	} else {
1003 		flags = FFLAGS(flags);
1004 	}
1005 
1006 	/*
1007 	 * Allocate a file structure. The descriptor to reference it
1008 	 * is allocated and set by finstall() below.
1009 	 */
1010 	error = falloc_noinstall(td, &fp);
1011 	if (error != 0)
1012 		return (error);
1013 	/*
1014 	 * An extra reference on `fp' has been held for us by
1015 	 * falloc_noinstall().
1016 	 */
1017 	/* Set the flags early so the finit in devfs can pick them up. */
1018 	fp->f_flag = flags & FMASK;
1019 	cmode = ((mode & ~fdp->fd_cmask) & ALLPERMS) & ~S_ISTXT;
1020 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, fd,
1021 	    &rights, td);
1022 	td->td_dupfd = -1;		/* XXX check for fdopen */
1023 	error = vn_open(&nd, &flags, cmode, fp);
1024 	if (error != 0) {
1025 		/*
1026 		 * If the vn_open replaced the method vector, something
1027 		 * wonderous happened deep below and we just pass it up
1028 		 * pretending we know what we do.
1029 		 */
1030 		if (error == ENXIO && fp->f_ops != &badfileops)
1031 			goto success;
1032 
1033 		/*
1034 		 * Handle special fdopen() case. bleh.
1035 		 *
1036 		 * Don't do this for relative (capability) lookups; we don't
1037 		 * understand exactly what would happen, and we don't think
1038 		 * that it ever should.
1039 		 */
1040 		if ((nd.ni_lcf & NI_LCF_STRICTRELATIVE) == 0 &&
1041 		    (error == ENODEV || error == ENXIO) &&
1042 		    td->td_dupfd >= 0) {
1043 			error = dupfdopen(td, fdp, td->td_dupfd, flags, error,
1044 			    &indx);
1045 			if (error == 0)
1046 				goto success;
1047 		}
1048 
1049 		goto bad;
1050 	}
1051 	td->td_dupfd = 0;
1052 	NDFREE(&nd, NDF_ONLY_PNBUF);
1053 	vp = nd.ni_vp;
1054 
1055 	/*
1056 	 * Store the vnode, for any f_type. Typically, the vnode use
1057 	 * count is decremented by direct call to vn_closefile() for
1058 	 * files that switched type in the cdevsw fdopen() method.
1059 	 */
1060 	fp->f_vnode = vp;
1061 	/*
1062 	 * If the file wasn't claimed by devfs bind it to the normal
1063 	 * vnode operations here.
1064 	 */
1065 	if (fp->f_ops == &badfileops) {
1066 		KASSERT(vp->v_type != VFIFO, ("Unexpected fifo."));
1067 		fp->f_seqcount = 1;
1068 		finit(fp, (flags & FMASK) | (fp->f_flag & FHASLOCK),
1069 		    DTYPE_VNODE, vp, &vnops);
1070 	}
1071 
1072 	VOP_UNLOCK(vp, 0);
1073 	if (flags & O_TRUNC) {
1074 		error = fo_truncate(fp, 0, td->td_ucred, td);
1075 		if (error != 0)
1076 			goto bad;
1077 	}
1078 success:
1079 	/*
1080 	 * If we haven't already installed the FD (for dupfdopen), do so now.
1081 	 */
1082 	if (indx == -1) {
1083 		struct filecaps *fcaps;
1084 
1085 #ifdef CAPABILITIES
1086 		if ((nd.ni_lcf & NI_LCF_STRICTRELATIVE) != 0)
1087 			fcaps = &nd.ni_filecaps;
1088 		else
1089 #endif
1090 			fcaps = NULL;
1091 		error = finstall(td, fp, &indx, flags, fcaps);
1092 		/* On success finstall() consumes fcaps. */
1093 		if (error != 0) {
1094 			filecaps_free(&nd.ni_filecaps);
1095 			goto bad;
1096 		}
1097 	} else {
1098 		filecaps_free(&nd.ni_filecaps);
1099 	}
1100 
1101 	/*
1102 	 * Release our private reference, leaving the one associated with
1103 	 * the descriptor table intact.
1104 	 */
1105 	fdrop(fp, td);
1106 	td->td_retval[0] = indx;
1107 	return (0);
1108 bad:
1109 	KASSERT(indx == -1, ("indx=%d, should be -1", indx));
1110 	fdrop(fp, td);
1111 	return (error);
1112 }
1113 
1114 #ifdef COMPAT_43
1115 /*
1116  * Create a file.
1117  */
1118 #ifndef _SYS_SYSPROTO_H_
1119 struct ocreat_args {
1120 	char	*path;
1121 	int	mode;
1122 };
1123 #endif
1124 int
1125 ocreat(td, uap)
1126 	struct thread *td;
1127 	register struct ocreat_args /* {
1128 		char *path;
1129 		int mode;
1130 	} */ *uap;
1131 {
1132 
1133 	return (kern_openat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1134 	    O_WRONLY | O_CREAT | O_TRUNC, uap->mode));
1135 }
1136 #endif /* COMPAT_43 */
1137 
1138 /*
1139  * Create a special file.
1140  */
1141 #ifndef _SYS_SYSPROTO_H_
1142 struct mknod_args {
1143 	char	*path;
1144 	int	mode;
1145 	int	dev;
1146 };
1147 #endif
1148 int
1149 sys_mknod(td, uap)
1150 	struct thread *td;
1151 	register struct mknod_args /* {
1152 		char *path;
1153 		int mode;
1154 		int dev;
1155 	} */ *uap;
1156 {
1157 
1158 	return (kern_mknodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1159 	    uap->mode, uap->dev));
1160 }
1161 
1162 #ifndef _SYS_SYSPROTO_H_
1163 struct mknodat_args {
1164 	int	fd;
1165 	char	*path;
1166 	mode_t	mode;
1167 	dev_t	dev;
1168 };
1169 #endif
1170 int
1171 sys_mknodat(struct thread *td, struct mknodat_args *uap)
1172 {
1173 
1174 	return (kern_mknodat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode,
1175 	    uap->dev));
1176 }
1177 
1178 int
1179 kern_mknodat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1180     int mode, int dev)
1181 {
1182 	struct vnode *vp;
1183 	struct mount *mp;
1184 	struct vattr vattr;
1185 	struct nameidata nd;
1186 	cap_rights_t rights;
1187 	int error, whiteout = 0;
1188 
1189 	AUDIT_ARG_MODE(mode);
1190 	AUDIT_ARG_DEV(dev);
1191 	switch (mode & S_IFMT) {
1192 	case S_IFCHR:
1193 	case S_IFBLK:
1194 		error = priv_check(td, PRIV_VFS_MKNOD_DEV);
1195 		if (error == 0 && dev == VNOVAL)
1196 			error = EINVAL;
1197 		break;
1198 	case S_IFMT:
1199 		error = priv_check(td, PRIV_VFS_MKNOD_BAD);
1200 		break;
1201 	case S_IFWHT:
1202 		error = priv_check(td, PRIV_VFS_MKNOD_WHT);
1203 		break;
1204 	case S_IFIFO:
1205 		if (dev == 0)
1206 			return (kern_mkfifoat(td, fd, path, pathseg, mode));
1207 		/* FALLTHROUGH */
1208 	default:
1209 		error = EINVAL;
1210 		break;
1211 	}
1212 	if (error != 0)
1213 		return (error);
1214 restart:
1215 	bwillwrite();
1216 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
1217 	    NOCACHE, pathseg, path, fd, cap_rights_init(&rights, CAP_MKNODAT),
1218 	    td);
1219 	if ((error = namei(&nd)) != 0)
1220 		return (error);
1221 	vp = nd.ni_vp;
1222 	if (vp != NULL) {
1223 		NDFREE(&nd, NDF_ONLY_PNBUF);
1224 		if (vp == nd.ni_dvp)
1225 			vrele(nd.ni_dvp);
1226 		else
1227 			vput(nd.ni_dvp);
1228 		vrele(vp);
1229 		return (EEXIST);
1230 	} else {
1231 		VATTR_NULL(&vattr);
1232 		vattr.va_mode = (mode & ALLPERMS) &
1233 		    ~td->td_proc->p_fd->fd_cmask;
1234 		vattr.va_rdev = dev;
1235 		whiteout = 0;
1236 
1237 		switch (mode & S_IFMT) {
1238 		case S_IFMT:	/* used by badsect to flag bad sectors */
1239 			vattr.va_type = VBAD;
1240 			break;
1241 		case S_IFCHR:
1242 			vattr.va_type = VCHR;
1243 			break;
1244 		case S_IFBLK:
1245 			vattr.va_type = VBLK;
1246 			break;
1247 		case S_IFWHT:
1248 			whiteout = 1;
1249 			break;
1250 		default:
1251 			panic("kern_mknod: invalid mode");
1252 		}
1253 	}
1254 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1255 		NDFREE(&nd, NDF_ONLY_PNBUF);
1256 		vput(nd.ni_dvp);
1257 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1258 			return (error);
1259 		goto restart;
1260 	}
1261 #ifdef MAC
1262 	if (error == 0 && !whiteout)
1263 		error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp,
1264 		    &nd.ni_cnd, &vattr);
1265 #endif
1266 	if (error == 0) {
1267 		if (whiteout)
1268 			error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE);
1269 		else {
1270 			error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp,
1271 						&nd.ni_cnd, &vattr);
1272 			if (error == 0)
1273 				vput(nd.ni_vp);
1274 		}
1275 	}
1276 	NDFREE(&nd, NDF_ONLY_PNBUF);
1277 	vput(nd.ni_dvp);
1278 	vn_finished_write(mp);
1279 	return (error);
1280 }
1281 
1282 /*
1283  * Create a named pipe.
1284  */
1285 #ifndef _SYS_SYSPROTO_H_
1286 struct mkfifo_args {
1287 	char	*path;
1288 	int	mode;
1289 };
1290 #endif
1291 int
1292 sys_mkfifo(td, uap)
1293 	struct thread *td;
1294 	register struct mkfifo_args /* {
1295 		char *path;
1296 		int mode;
1297 	} */ *uap;
1298 {
1299 
1300 	return (kern_mkfifoat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1301 	    uap->mode));
1302 }
1303 
1304 #ifndef _SYS_SYSPROTO_H_
1305 struct mkfifoat_args {
1306 	int	fd;
1307 	char	*path;
1308 	mode_t	mode;
1309 };
1310 #endif
1311 int
1312 sys_mkfifoat(struct thread *td, struct mkfifoat_args *uap)
1313 {
1314 
1315 	return (kern_mkfifoat(td, uap->fd, uap->path, UIO_USERSPACE,
1316 	    uap->mode));
1317 }
1318 
1319 int
1320 kern_mkfifoat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1321     int mode)
1322 {
1323 	struct mount *mp;
1324 	struct vattr vattr;
1325 	struct nameidata nd;
1326 	cap_rights_t rights;
1327 	int error;
1328 
1329 	AUDIT_ARG_MODE(mode);
1330 restart:
1331 	bwillwrite();
1332 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
1333 	    NOCACHE, pathseg, path, fd, cap_rights_init(&rights, CAP_MKFIFOAT),
1334 	    td);
1335 	if ((error = namei(&nd)) != 0)
1336 		return (error);
1337 	if (nd.ni_vp != NULL) {
1338 		NDFREE(&nd, NDF_ONLY_PNBUF);
1339 		if (nd.ni_vp == nd.ni_dvp)
1340 			vrele(nd.ni_dvp);
1341 		else
1342 			vput(nd.ni_dvp);
1343 		vrele(nd.ni_vp);
1344 		return (EEXIST);
1345 	}
1346 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1347 		NDFREE(&nd, NDF_ONLY_PNBUF);
1348 		vput(nd.ni_dvp);
1349 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1350 			return (error);
1351 		goto restart;
1352 	}
1353 	VATTR_NULL(&vattr);
1354 	vattr.va_type = VFIFO;
1355 	vattr.va_mode = (mode & ALLPERMS) & ~td->td_proc->p_fd->fd_cmask;
1356 #ifdef MAC
1357 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1358 	    &vattr);
1359 	if (error != 0)
1360 		goto out;
1361 #endif
1362 	error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
1363 	if (error == 0)
1364 		vput(nd.ni_vp);
1365 #ifdef MAC
1366 out:
1367 #endif
1368 	vput(nd.ni_dvp);
1369 	vn_finished_write(mp);
1370 	NDFREE(&nd, NDF_ONLY_PNBUF);
1371 	return (error);
1372 }
1373 
1374 /*
1375  * Make a hard file link.
1376  */
1377 #ifndef _SYS_SYSPROTO_H_
1378 struct link_args {
1379 	char	*path;
1380 	char	*link;
1381 };
1382 #endif
1383 int
1384 sys_link(td, uap)
1385 	struct thread *td;
1386 	register struct link_args /* {
1387 		char *path;
1388 		char *link;
1389 	} */ *uap;
1390 {
1391 
1392 	return (kern_linkat(td, AT_FDCWD, AT_FDCWD, uap->path, uap->link,
1393 	    UIO_USERSPACE, FOLLOW));
1394 }
1395 
1396 #ifndef _SYS_SYSPROTO_H_
1397 struct linkat_args {
1398 	int	fd1;
1399 	char	*path1;
1400 	int	fd2;
1401 	char	*path2;
1402 	int	flag;
1403 };
1404 #endif
1405 int
1406 sys_linkat(struct thread *td, struct linkat_args *uap)
1407 {
1408 	int flag;
1409 
1410 	flag = uap->flag;
1411 	if (flag & ~AT_SYMLINK_FOLLOW)
1412 		return (EINVAL);
1413 
1414 	return (kern_linkat(td, uap->fd1, uap->fd2, uap->path1, uap->path2,
1415 	    UIO_USERSPACE, (flag & AT_SYMLINK_FOLLOW) ? FOLLOW : NOFOLLOW));
1416 }
1417 
1418 int hardlink_check_uid = 0;
1419 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_uid, CTLFLAG_RW,
1420     &hardlink_check_uid, 0,
1421     "Unprivileged processes cannot create hard links to files owned by other "
1422     "users");
1423 static int hardlink_check_gid = 0;
1424 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_gid, CTLFLAG_RW,
1425     &hardlink_check_gid, 0,
1426     "Unprivileged processes cannot create hard links to files owned by other "
1427     "groups");
1428 
1429 static int
1430 can_hardlink(struct vnode *vp, struct ucred *cred)
1431 {
1432 	struct vattr va;
1433 	int error;
1434 
1435 	if (!hardlink_check_uid && !hardlink_check_gid)
1436 		return (0);
1437 
1438 	error = VOP_GETATTR(vp, &va, cred);
1439 	if (error != 0)
1440 		return (error);
1441 
1442 	if (hardlink_check_uid && cred->cr_uid != va.va_uid) {
1443 		error = priv_check_cred(cred, PRIV_VFS_LINK, 0);
1444 		if (error != 0)
1445 			return (error);
1446 	}
1447 
1448 	if (hardlink_check_gid && !groupmember(va.va_gid, cred)) {
1449 		error = priv_check_cred(cred, PRIV_VFS_LINK, 0);
1450 		if (error != 0)
1451 			return (error);
1452 	}
1453 
1454 	return (0);
1455 }
1456 
1457 int
1458 kern_linkat(struct thread *td, int fd1, int fd2, char *path1, char *path2,
1459     enum uio_seg segflg, int follow)
1460 {
1461 	struct vnode *vp;
1462 	struct mount *mp;
1463 	struct nameidata nd;
1464 	cap_rights_t rights;
1465 	int error;
1466 
1467 again:
1468 	bwillwrite();
1469 	NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, segflg, path1, fd1,
1470 	    cap_rights_init(&rights, CAP_LINKAT_SOURCE), td);
1471 
1472 	if ((error = namei(&nd)) != 0)
1473 		return (error);
1474 	NDFREE(&nd, NDF_ONLY_PNBUF);
1475 	vp = nd.ni_vp;
1476 	if (vp->v_type == VDIR) {
1477 		vrele(vp);
1478 		return (EPERM);		/* POSIX */
1479 	}
1480 	NDINIT_ATRIGHTS(&nd, CREATE,
1481 	    LOCKPARENT | SAVENAME | AUDITVNODE2 | NOCACHE, segflg, path2, fd2,
1482 	    cap_rights_init(&rights, CAP_LINKAT_TARGET), td);
1483 	if ((error = namei(&nd)) == 0) {
1484 		if (nd.ni_vp != NULL) {
1485 			NDFREE(&nd, NDF_ONLY_PNBUF);
1486 			if (nd.ni_dvp == nd.ni_vp)
1487 				vrele(nd.ni_dvp);
1488 			else
1489 				vput(nd.ni_dvp);
1490 			vrele(nd.ni_vp);
1491 			vrele(vp);
1492 			return (EEXIST);
1493 		} else if (nd.ni_dvp->v_mount != vp->v_mount) {
1494 			/*
1495 			 * Cross-device link.  No need to recheck
1496 			 * vp->v_type, since it cannot change, except
1497 			 * to VBAD.
1498 			 */
1499 			NDFREE(&nd, NDF_ONLY_PNBUF);
1500 			vput(nd.ni_dvp);
1501 			vrele(vp);
1502 			return (EXDEV);
1503 		} else if ((error = vn_lock(vp, LK_EXCLUSIVE)) == 0) {
1504 			error = can_hardlink(vp, td->td_ucred);
1505 #ifdef MAC
1506 			if (error == 0)
1507 				error = mac_vnode_check_link(td->td_ucred,
1508 				    nd.ni_dvp, vp, &nd.ni_cnd);
1509 #endif
1510 			if (error != 0) {
1511 				vput(vp);
1512 				vput(nd.ni_dvp);
1513 				NDFREE(&nd, NDF_ONLY_PNBUF);
1514 				return (error);
1515 			}
1516 			error = vn_start_write(vp, &mp, V_NOWAIT);
1517 			if (error != 0) {
1518 				vput(vp);
1519 				vput(nd.ni_dvp);
1520 				NDFREE(&nd, NDF_ONLY_PNBUF);
1521 				error = vn_start_write(NULL, &mp,
1522 				    V_XSLEEP | PCATCH);
1523 				if (error != 0)
1524 					return (error);
1525 				goto again;
1526 			}
1527 			error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
1528 			VOP_UNLOCK(vp, 0);
1529 			vput(nd.ni_dvp);
1530 			vn_finished_write(mp);
1531 			NDFREE(&nd, NDF_ONLY_PNBUF);
1532 		} else {
1533 			vput(nd.ni_dvp);
1534 			NDFREE(&nd, NDF_ONLY_PNBUF);
1535 			vrele(vp);
1536 			goto again;
1537 		}
1538 	}
1539 	vrele(vp);
1540 	return (error);
1541 }
1542 
1543 /*
1544  * Make a symbolic link.
1545  */
1546 #ifndef _SYS_SYSPROTO_H_
1547 struct symlink_args {
1548 	char	*path;
1549 	char	*link;
1550 };
1551 #endif
1552 int
1553 sys_symlink(td, uap)
1554 	struct thread *td;
1555 	register struct symlink_args /* {
1556 		char *path;
1557 		char *link;
1558 	} */ *uap;
1559 {
1560 
1561 	return (kern_symlinkat(td, uap->path, AT_FDCWD, uap->link,
1562 	    UIO_USERSPACE));
1563 }
1564 
1565 #ifndef _SYS_SYSPROTO_H_
1566 struct symlinkat_args {
1567 	char	*path;
1568 	int	fd;
1569 	char	*path2;
1570 };
1571 #endif
1572 int
1573 sys_symlinkat(struct thread *td, struct symlinkat_args *uap)
1574 {
1575 
1576 	return (kern_symlinkat(td, uap->path1, uap->fd, uap->path2,
1577 	    UIO_USERSPACE));
1578 }
1579 
1580 int
1581 kern_symlinkat(struct thread *td, char *path1, int fd, char *path2,
1582     enum uio_seg segflg)
1583 {
1584 	struct mount *mp;
1585 	struct vattr vattr;
1586 	char *syspath;
1587 	struct nameidata nd;
1588 	int error;
1589 	cap_rights_t rights;
1590 
1591 	if (segflg == UIO_SYSSPACE) {
1592 		syspath = path1;
1593 	} else {
1594 		syspath = uma_zalloc(namei_zone, M_WAITOK);
1595 		if ((error = copyinstr(path1, syspath, MAXPATHLEN, NULL)) != 0)
1596 			goto out;
1597 	}
1598 	AUDIT_ARG_TEXT(syspath);
1599 restart:
1600 	bwillwrite();
1601 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
1602 	    NOCACHE, segflg, path2, fd, cap_rights_init(&rights, CAP_SYMLINKAT),
1603 	    td);
1604 	if ((error = namei(&nd)) != 0)
1605 		goto out;
1606 	if (nd.ni_vp) {
1607 		NDFREE(&nd, NDF_ONLY_PNBUF);
1608 		if (nd.ni_vp == nd.ni_dvp)
1609 			vrele(nd.ni_dvp);
1610 		else
1611 			vput(nd.ni_dvp);
1612 		vrele(nd.ni_vp);
1613 		error = EEXIST;
1614 		goto out;
1615 	}
1616 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1617 		NDFREE(&nd, NDF_ONLY_PNBUF);
1618 		vput(nd.ni_dvp);
1619 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1620 			goto out;
1621 		goto restart;
1622 	}
1623 	VATTR_NULL(&vattr);
1624 	vattr.va_mode = ACCESSPERMS &~ td->td_proc->p_fd->fd_cmask;
1625 #ifdef MAC
1626 	vattr.va_type = VLNK;
1627 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1628 	    &vattr);
1629 	if (error != 0)
1630 		goto out2;
1631 #endif
1632 	error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, syspath);
1633 	if (error == 0)
1634 		vput(nd.ni_vp);
1635 #ifdef MAC
1636 out2:
1637 #endif
1638 	NDFREE(&nd, NDF_ONLY_PNBUF);
1639 	vput(nd.ni_dvp);
1640 	vn_finished_write(mp);
1641 out:
1642 	if (segflg != UIO_SYSSPACE)
1643 		uma_zfree(namei_zone, syspath);
1644 	return (error);
1645 }
1646 
1647 /*
1648  * Delete a whiteout from the filesystem.
1649  */
1650 int
1651 sys_undelete(td, uap)
1652 	struct thread *td;
1653 	register struct undelete_args /* {
1654 		char *path;
1655 	} */ *uap;
1656 {
1657 	struct mount *mp;
1658 	struct nameidata nd;
1659 	int error;
1660 
1661 restart:
1662 	bwillwrite();
1663 	NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | AUDITVNODE1,
1664 	    UIO_USERSPACE, uap->path, td);
1665 	error = namei(&nd);
1666 	if (error != 0)
1667 		return (error);
1668 
1669 	if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) {
1670 		NDFREE(&nd, NDF_ONLY_PNBUF);
1671 		if (nd.ni_vp == nd.ni_dvp)
1672 			vrele(nd.ni_dvp);
1673 		else
1674 			vput(nd.ni_dvp);
1675 		if (nd.ni_vp)
1676 			vrele(nd.ni_vp);
1677 		return (EEXIST);
1678 	}
1679 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1680 		NDFREE(&nd, NDF_ONLY_PNBUF);
1681 		vput(nd.ni_dvp);
1682 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1683 			return (error);
1684 		goto restart;
1685 	}
1686 	error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE);
1687 	NDFREE(&nd, NDF_ONLY_PNBUF);
1688 	vput(nd.ni_dvp);
1689 	vn_finished_write(mp);
1690 	return (error);
1691 }
1692 
1693 /*
1694  * Delete a name from the filesystem.
1695  */
1696 #ifndef _SYS_SYSPROTO_H_
1697 struct unlink_args {
1698 	char	*path;
1699 };
1700 #endif
1701 int
1702 sys_unlink(td, uap)
1703 	struct thread *td;
1704 	struct unlink_args /* {
1705 		char *path;
1706 	} */ *uap;
1707 {
1708 
1709 	return (kern_unlinkat(td, AT_FDCWD, uap->path, UIO_USERSPACE, 0));
1710 }
1711 
1712 #ifndef _SYS_SYSPROTO_H_
1713 struct unlinkat_args {
1714 	int	fd;
1715 	char	*path;
1716 	int	flag;
1717 };
1718 #endif
1719 int
1720 sys_unlinkat(struct thread *td, struct unlinkat_args *uap)
1721 {
1722 	int flag = uap->flag;
1723 	int fd = uap->fd;
1724 	char *path = uap->path;
1725 
1726 	if (flag & ~AT_REMOVEDIR)
1727 		return (EINVAL);
1728 
1729 	if (flag & AT_REMOVEDIR)
1730 		return (kern_rmdirat(td, fd, path, UIO_USERSPACE));
1731 	else
1732 		return (kern_unlinkat(td, fd, path, UIO_USERSPACE, 0));
1733 }
1734 
1735 int
1736 kern_unlinkat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1737     ino_t oldinum)
1738 {
1739 	struct mount *mp;
1740 	struct vnode *vp;
1741 	struct nameidata nd;
1742 	struct stat sb;
1743 	cap_rights_t rights;
1744 	int error;
1745 
1746 restart:
1747 	bwillwrite();
1748 	NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1,
1749 	    pathseg, path, fd, cap_rights_init(&rights, CAP_UNLINKAT), td);
1750 	if ((error = namei(&nd)) != 0)
1751 		return (error == EINVAL ? EPERM : error);
1752 	vp = nd.ni_vp;
1753 	if (vp->v_type == VDIR && oldinum == 0) {
1754 		error = EPERM;		/* POSIX */
1755 	} else if (oldinum != 0 &&
1756 		  ((error = vn_stat(vp, &sb, td->td_ucred, NOCRED, td)) == 0) &&
1757 		  sb.st_ino != oldinum) {
1758 			error = EIDRM;	/* Identifier removed */
1759 	} else {
1760 		/*
1761 		 * The root of a mounted filesystem cannot be deleted.
1762 		 *
1763 		 * XXX: can this only be a VDIR case?
1764 		 */
1765 		if (vp->v_vflag & VV_ROOT)
1766 			error = EBUSY;
1767 	}
1768 	if (error == 0) {
1769 		if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1770 			NDFREE(&nd, NDF_ONLY_PNBUF);
1771 			vput(nd.ni_dvp);
1772 			if (vp == nd.ni_dvp)
1773 				vrele(vp);
1774 			else
1775 				vput(vp);
1776 			if ((error = vn_start_write(NULL, &mp,
1777 			    V_XSLEEP | PCATCH)) != 0)
1778 				return (error);
1779 			goto restart;
1780 		}
1781 #ifdef MAC
1782 		error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
1783 		    &nd.ni_cnd);
1784 		if (error != 0)
1785 			goto out;
1786 #endif
1787 		vfs_notify_upper(vp, VFS_NOTIFY_UPPER_UNLINK);
1788 		error = VOP_REMOVE(nd.ni_dvp, vp, &nd.ni_cnd);
1789 #ifdef MAC
1790 out:
1791 #endif
1792 		vn_finished_write(mp);
1793 	}
1794 	NDFREE(&nd, NDF_ONLY_PNBUF);
1795 	vput(nd.ni_dvp);
1796 	if (vp == nd.ni_dvp)
1797 		vrele(vp);
1798 	else
1799 		vput(vp);
1800 	return (error);
1801 }
1802 
1803 /*
1804  * Reposition read/write file offset.
1805  */
1806 #ifndef _SYS_SYSPROTO_H_
1807 struct lseek_args {
1808 	int	fd;
1809 	int	pad;
1810 	off_t	offset;
1811 	int	whence;
1812 };
1813 #endif
1814 int
1815 sys_lseek(td, uap)
1816 	struct thread *td;
1817 	register struct lseek_args /* {
1818 		int fd;
1819 		int pad;
1820 		off_t offset;
1821 		int whence;
1822 	} */ *uap;
1823 {
1824 	struct file *fp;
1825 	cap_rights_t rights;
1826 	int error;
1827 
1828 	AUDIT_ARG_FD(uap->fd);
1829 	error = fget(td, uap->fd, cap_rights_init(&rights, CAP_SEEK), &fp);
1830 	if (error != 0)
1831 		return (error);
1832 	error = (fp->f_ops->fo_flags & DFLAG_SEEKABLE) != 0 ?
1833 	    fo_seek(fp, uap->offset, uap->whence, td) : ESPIPE;
1834 	fdrop(fp, td);
1835 	return (error);
1836 }
1837 
1838 #if defined(COMPAT_43)
1839 /*
1840  * Reposition read/write file offset.
1841  */
1842 #ifndef _SYS_SYSPROTO_H_
1843 struct olseek_args {
1844 	int	fd;
1845 	long	offset;
1846 	int	whence;
1847 };
1848 #endif
1849 int
1850 olseek(td, uap)
1851 	struct thread *td;
1852 	register struct olseek_args /* {
1853 		int fd;
1854 		long offset;
1855 		int whence;
1856 	} */ *uap;
1857 {
1858 	struct lseek_args /* {
1859 		int fd;
1860 		int pad;
1861 		off_t offset;
1862 		int whence;
1863 	} */ nuap;
1864 
1865 	nuap.fd = uap->fd;
1866 	nuap.offset = uap->offset;
1867 	nuap.whence = uap->whence;
1868 	return (sys_lseek(td, &nuap));
1869 }
1870 #endif /* COMPAT_43 */
1871 
1872 #if defined(COMPAT_FREEBSD6)
1873 /* Version with the 'pad' argument */
1874 int
1875 freebsd6_lseek(td, uap)
1876 	struct thread *td;
1877 	register struct freebsd6_lseek_args *uap;
1878 {
1879 	struct lseek_args ouap;
1880 
1881 	ouap.fd = uap->fd;
1882 	ouap.offset = uap->offset;
1883 	ouap.whence = uap->whence;
1884 	return (sys_lseek(td, &ouap));
1885 }
1886 #endif
1887 
1888 /*
1889  * Check access permissions using passed credentials.
1890  */
1891 static int
1892 vn_access(vp, user_flags, cred, td)
1893 	struct vnode	*vp;
1894 	int		user_flags;
1895 	struct ucred	*cred;
1896 	struct thread	*td;
1897 {
1898 	accmode_t accmode;
1899 	int error;
1900 
1901 	/* Flags == 0 means only check for existence. */
1902 	if (user_flags == 0)
1903 		return (0);
1904 
1905 	accmode = 0;
1906 	if (user_flags & R_OK)
1907 		accmode |= VREAD;
1908 	if (user_flags & W_OK)
1909 		accmode |= VWRITE;
1910 	if (user_flags & X_OK)
1911 		accmode |= VEXEC;
1912 #ifdef MAC
1913 	error = mac_vnode_check_access(cred, vp, accmode);
1914 	if (error != 0)
1915 		return (error);
1916 #endif
1917 	if ((accmode & VWRITE) == 0 || (error = vn_writechk(vp)) == 0)
1918 		error = VOP_ACCESS(vp, accmode, cred, td);
1919 	return (error);
1920 }
1921 
1922 /*
1923  * Check access permissions using "real" credentials.
1924  */
1925 #ifndef _SYS_SYSPROTO_H_
1926 struct access_args {
1927 	char	*path;
1928 	int	amode;
1929 };
1930 #endif
1931 int
1932 sys_access(td, uap)
1933 	struct thread *td;
1934 	register struct access_args /* {
1935 		char *path;
1936 		int amode;
1937 	} */ *uap;
1938 {
1939 
1940 	return (kern_accessat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1941 	    0, uap->amode));
1942 }
1943 
1944 #ifndef _SYS_SYSPROTO_H_
1945 struct faccessat_args {
1946 	int	dirfd;
1947 	char	*path;
1948 	int	amode;
1949 	int	flag;
1950 }
1951 #endif
1952 int
1953 sys_faccessat(struct thread *td, struct faccessat_args *uap)
1954 {
1955 
1956 	return (kern_accessat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
1957 	    uap->amode));
1958 }
1959 
1960 int
1961 kern_accessat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1962     int flag, int amode)
1963 {
1964 	struct ucred *cred, *usecred;
1965 	struct vnode *vp;
1966 	struct nameidata nd;
1967 	cap_rights_t rights;
1968 	int error;
1969 
1970 	if (flag & ~AT_EACCESS)
1971 		return (EINVAL);
1972 	if (amode != F_OK && (amode & ~(R_OK | W_OK | X_OK)) != 0)
1973 		return (EINVAL);
1974 
1975 	/*
1976 	 * Create and modify a temporary credential instead of one that
1977 	 * is potentially shared (if we need one).
1978 	 */
1979 	cred = td->td_ucred;
1980 	if ((flag & AT_EACCESS) == 0 &&
1981 	    ((cred->cr_uid != cred->cr_ruid ||
1982 	    cred->cr_rgid != cred->cr_groups[0]))) {
1983 		usecred = crdup(cred);
1984 		usecred->cr_uid = cred->cr_ruid;
1985 		usecred->cr_groups[0] = cred->cr_rgid;
1986 		td->td_ucred = usecred;
1987 	} else
1988 		usecred = cred;
1989 	AUDIT_ARG_VALUE(amode);
1990 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF |
1991 	    AUDITVNODE1, pathseg, path, fd, cap_rights_init(&rights, CAP_FSTAT),
1992 	    td);
1993 	if ((error = namei(&nd)) != 0)
1994 		goto out;
1995 	vp = nd.ni_vp;
1996 
1997 	error = vn_access(vp, amode, usecred, td);
1998 	NDFREE(&nd, NDF_ONLY_PNBUF);
1999 	vput(vp);
2000 out:
2001 	if (usecred != cred) {
2002 		td->td_ucred = cred;
2003 		crfree(usecred);
2004 	}
2005 	return (error);
2006 }
2007 
2008 /*
2009  * Check access permissions using "effective" credentials.
2010  */
2011 #ifndef _SYS_SYSPROTO_H_
2012 struct eaccess_args {
2013 	char	*path;
2014 	int	amode;
2015 };
2016 #endif
2017 int
2018 sys_eaccess(td, uap)
2019 	struct thread *td;
2020 	register struct eaccess_args /* {
2021 		char *path;
2022 		int amode;
2023 	} */ *uap;
2024 {
2025 
2026 	return (kern_accessat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2027 	    AT_EACCESS, uap->amode));
2028 }
2029 
2030 #if defined(COMPAT_43)
2031 /*
2032  * Get file status; this version follows links.
2033  */
2034 #ifndef _SYS_SYSPROTO_H_
2035 struct ostat_args {
2036 	char	*path;
2037 	struct ostat *ub;
2038 };
2039 #endif
2040 int
2041 ostat(td, uap)
2042 	struct thread *td;
2043 	register struct ostat_args /* {
2044 		char *path;
2045 		struct ostat *ub;
2046 	} */ *uap;
2047 {
2048 	struct stat sb;
2049 	struct ostat osb;
2050 	int error;
2051 
2052 	error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE,
2053 	    &sb, NULL);
2054 	if (error != 0)
2055 		return (error);
2056 	cvtstat(&sb, &osb);
2057 	return (copyout(&osb, uap->ub, sizeof (osb)));
2058 }
2059 
2060 /*
2061  * Get file status; this version does not follow links.
2062  */
2063 #ifndef _SYS_SYSPROTO_H_
2064 struct olstat_args {
2065 	char	*path;
2066 	struct ostat *ub;
2067 };
2068 #endif
2069 int
2070 olstat(td, uap)
2071 	struct thread *td;
2072 	register struct olstat_args /* {
2073 		char *path;
2074 		struct ostat *ub;
2075 	} */ *uap;
2076 {
2077 	struct stat sb;
2078 	struct ostat osb;
2079 	int error;
2080 
2081 	error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2082 	    UIO_USERSPACE, &sb, NULL);
2083 	if (error != 0)
2084 		return (error);
2085 	cvtstat(&sb, &osb);
2086 	return (copyout(&osb, uap->ub, sizeof (osb)));
2087 }
2088 
2089 /*
2090  * Convert from an old to a new stat structure.
2091  */
2092 void
2093 cvtstat(st, ost)
2094 	struct stat *st;
2095 	struct ostat *ost;
2096 {
2097 
2098 	bzero(ost, sizeof(*ost));
2099 	ost->st_dev = st->st_dev;
2100 	ost->st_ino = st->st_ino;
2101 	ost->st_mode = st->st_mode;
2102 	ost->st_nlink = st->st_nlink;
2103 	ost->st_uid = st->st_uid;
2104 	ost->st_gid = st->st_gid;
2105 	ost->st_rdev = st->st_rdev;
2106 	if (st->st_size < (quad_t)1 << 32)
2107 		ost->st_size = st->st_size;
2108 	else
2109 		ost->st_size = -2;
2110 	ost->st_atim = st->st_atim;
2111 	ost->st_mtim = st->st_mtim;
2112 	ost->st_ctim = st->st_ctim;
2113 	ost->st_blksize = st->st_blksize;
2114 	ost->st_blocks = st->st_blocks;
2115 	ost->st_flags = st->st_flags;
2116 	ost->st_gen = st->st_gen;
2117 }
2118 #endif /* COMPAT_43 */
2119 
2120 /*
2121  * Get file status; this version follows links.
2122  */
2123 #ifndef _SYS_SYSPROTO_H_
2124 struct stat_args {
2125 	char	*path;
2126 	struct stat *ub;
2127 };
2128 #endif
2129 int
2130 sys_stat(td, uap)
2131 	struct thread *td;
2132 	register struct stat_args /* {
2133 		char *path;
2134 		struct stat *ub;
2135 	} */ *uap;
2136 {
2137 	struct stat sb;
2138 	int error;
2139 
2140 	error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE,
2141 	    &sb, NULL);
2142 	if (error == 0)
2143 		error = copyout(&sb, uap->ub, sizeof (sb));
2144 	return (error);
2145 }
2146 
2147 #ifndef _SYS_SYSPROTO_H_
2148 struct fstatat_args {
2149 	int	fd;
2150 	char	*path;
2151 	struct stat	*buf;
2152 	int	flag;
2153 }
2154 #endif
2155 int
2156 sys_fstatat(struct thread *td, struct fstatat_args *uap)
2157 {
2158 	struct stat sb;
2159 	int error;
2160 
2161 	error = kern_statat(td, uap->flag, uap->fd, uap->path,
2162 	    UIO_USERSPACE, &sb, NULL);
2163 	if (error == 0)
2164 		error = copyout(&sb, uap->buf, sizeof (sb));
2165 	return (error);
2166 }
2167 
2168 int
2169 kern_statat(struct thread *td, int flag, int fd, char *path,
2170     enum uio_seg pathseg, struct stat *sbp,
2171     void (*hook)(struct vnode *vp, struct stat *sbp))
2172 {
2173 	struct nameidata nd;
2174 	struct stat sb;
2175 	cap_rights_t rights;
2176 	int error;
2177 
2178 	if (flag & ~AT_SYMLINK_NOFOLLOW)
2179 		return (EINVAL);
2180 
2181 	NDINIT_ATRIGHTS(&nd, LOOKUP, ((flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW :
2182 	    FOLLOW) | LOCKSHARED | LOCKLEAF | AUDITVNODE1, pathseg, path, fd,
2183 	    cap_rights_init(&rights, CAP_FSTAT), td);
2184 
2185 	if ((error = namei(&nd)) != 0)
2186 		return (error);
2187 	error = vn_stat(nd.ni_vp, &sb, td->td_ucred, NOCRED, td);
2188 	if (error == 0) {
2189 		SDT_PROBE2(vfs, , stat, mode, path, sb.st_mode);
2190 		if (S_ISREG(sb.st_mode))
2191 			SDT_PROBE2(vfs, , stat, reg, path, pathseg);
2192 		if (__predict_false(hook != NULL))
2193 			hook(nd.ni_vp, &sb);
2194 	}
2195 	NDFREE(&nd, NDF_ONLY_PNBUF);
2196 	vput(nd.ni_vp);
2197 	if (error != 0)
2198 		return (error);
2199 	*sbp = sb;
2200 #ifdef KTRACE
2201 	if (KTRPOINT(td, KTR_STRUCT))
2202 		ktrstat(&sb);
2203 #endif
2204 	return (0);
2205 }
2206 
2207 /*
2208  * Get file status; this version does not follow links.
2209  */
2210 #ifndef _SYS_SYSPROTO_H_
2211 struct lstat_args {
2212 	char	*path;
2213 	struct stat *ub;
2214 };
2215 #endif
2216 int
2217 sys_lstat(td, uap)
2218 	struct thread *td;
2219 	register struct lstat_args /* {
2220 		char *path;
2221 		struct stat *ub;
2222 	} */ *uap;
2223 {
2224 	struct stat sb;
2225 	int error;
2226 
2227 	error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2228 	    UIO_USERSPACE, &sb, NULL);
2229 	if (error == 0)
2230 		error = copyout(&sb, uap->ub, sizeof (sb));
2231 	return (error);
2232 }
2233 
2234 /*
2235  * Implementation of the NetBSD [l]stat() functions.
2236  */
2237 void
2238 cvtnstat(sb, nsb)
2239 	struct stat *sb;
2240 	struct nstat *nsb;
2241 {
2242 
2243 	bzero(nsb, sizeof *nsb);
2244 	nsb->st_dev = sb->st_dev;
2245 	nsb->st_ino = sb->st_ino;
2246 	nsb->st_mode = sb->st_mode;
2247 	nsb->st_nlink = sb->st_nlink;
2248 	nsb->st_uid = sb->st_uid;
2249 	nsb->st_gid = sb->st_gid;
2250 	nsb->st_rdev = sb->st_rdev;
2251 	nsb->st_atim = sb->st_atim;
2252 	nsb->st_mtim = sb->st_mtim;
2253 	nsb->st_ctim = sb->st_ctim;
2254 	nsb->st_size = sb->st_size;
2255 	nsb->st_blocks = sb->st_blocks;
2256 	nsb->st_blksize = sb->st_blksize;
2257 	nsb->st_flags = sb->st_flags;
2258 	nsb->st_gen = sb->st_gen;
2259 	nsb->st_birthtim = sb->st_birthtim;
2260 }
2261 
2262 #ifndef _SYS_SYSPROTO_H_
2263 struct nstat_args {
2264 	char	*path;
2265 	struct nstat *ub;
2266 };
2267 #endif
2268 int
2269 sys_nstat(td, uap)
2270 	struct thread *td;
2271 	register struct nstat_args /* {
2272 		char *path;
2273 		struct nstat *ub;
2274 	} */ *uap;
2275 {
2276 	struct stat sb;
2277 	struct nstat nsb;
2278 	int error;
2279 
2280 	error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE,
2281 	    &sb, NULL);
2282 	if (error != 0)
2283 		return (error);
2284 	cvtnstat(&sb, &nsb);
2285 	return (copyout(&nsb, uap->ub, sizeof (nsb)));
2286 }
2287 
2288 /*
2289  * NetBSD lstat.  Get file status; this version does not follow links.
2290  */
2291 #ifndef _SYS_SYSPROTO_H_
2292 struct lstat_args {
2293 	char	*path;
2294 	struct stat *ub;
2295 };
2296 #endif
2297 int
2298 sys_nlstat(td, uap)
2299 	struct thread *td;
2300 	register struct nlstat_args /* {
2301 		char *path;
2302 		struct nstat *ub;
2303 	} */ *uap;
2304 {
2305 	struct stat sb;
2306 	struct nstat nsb;
2307 	int error;
2308 
2309 	error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2310 	    UIO_USERSPACE, &sb, NULL);
2311 	if (error != 0)
2312 		return (error);
2313 	cvtnstat(&sb, &nsb);
2314 	return (copyout(&nsb, uap->ub, sizeof (nsb)));
2315 }
2316 
2317 /*
2318  * Get configurable pathname variables.
2319  */
2320 #ifndef _SYS_SYSPROTO_H_
2321 struct pathconf_args {
2322 	char	*path;
2323 	int	name;
2324 };
2325 #endif
2326 int
2327 sys_pathconf(td, uap)
2328 	struct thread *td;
2329 	register struct pathconf_args /* {
2330 		char *path;
2331 		int name;
2332 	} */ *uap;
2333 {
2334 
2335 	return (kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name, FOLLOW));
2336 }
2337 
2338 #ifndef _SYS_SYSPROTO_H_
2339 struct lpathconf_args {
2340 	char	*path;
2341 	int	name;
2342 };
2343 #endif
2344 int
2345 sys_lpathconf(td, uap)
2346 	struct thread *td;
2347 	register struct lpathconf_args /* {
2348 		char *path;
2349 		int name;
2350 	} */ *uap;
2351 {
2352 
2353 	return (kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name,
2354 	    NOFOLLOW));
2355 }
2356 
2357 int
2358 kern_pathconf(struct thread *td, char *path, enum uio_seg pathseg, int name,
2359     u_long flags)
2360 {
2361 	struct nameidata nd;
2362 	int error;
2363 
2364 	NDINIT(&nd, LOOKUP, LOCKSHARED | LOCKLEAF | AUDITVNODE1 | flags,
2365 	    pathseg, path, td);
2366 	if ((error = namei(&nd)) != 0)
2367 		return (error);
2368 	NDFREE(&nd, NDF_ONLY_PNBUF);
2369 
2370 	error = VOP_PATHCONF(nd.ni_vp, name, td->td_retval);
2371 	vput(nd.ni_vp);
2372 	return (error);
2373 }
2374 
2375 /*
2376  * Return target name of a symbolic link.
2377  */
2378 #ifndef _SYS_SYSPROTO_H_
2379 struct readlink_args {
2380 	char	*path;
2381 	char	*buf;
2382 	size_t	count;
2383 };
2384 #endif
2385 int
2386 sys_readlink(td, uap)
2387 	struct thread *td;
2388 	register struct readlink_args /* {
2389 		char *path;
2390 		char *buf;
2391 		size_t count;
2392 	} */ *uap;
2393 {
2394 
2395 	return (kern_readlinkat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2396 	    uap->buf, UIO_USERSPACE, uap->count));
2397 }
2398 #ifndef _SYS_SYSPROTO_H_
2399 struct readlinkat_args {
2400 	int	fd;
2401 	char	*path;
2402 	char	*buf;
2403 	size_t	bufsize;
2404 };
2405 #endif
2406 int
2407 sys_readlinkat(struct thread *td, struct readlinkat_args *uap)
2408 {
2409 
2410 	return (kern_readlinkat(td, uap->fd, uap->path, UIO_USERSPACE,
2411 	    uap->buf, UIO_USERSPACE, uap->bufsize));
2412 }
2413 
2414 int
2415 kern_readlinkat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2416     char *buf, enum uio_seg bufseg, size_t count)
2417 {
2418 	struct vnode *vp;
2419 	struct iovec aiov;
2420 	struct uio auio;
2421 	struct nameidata nd;
2422 	int error;
2423 
2424 	if (count > IOSIZE_MAX)
2425 		return (EINVAL);
2426 
2427 	NDINIT_AT(&nd, LOOKUP, NOFOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
2428 	    pathseg, path, fd, td);
2429 
2430 	if ((error = namei(&nd)) != 0)
2431 		return (error);
2432 	NDFREE(&nd, NDF_ONLY_PNBUF);
2433 	vp = nd.ni_vp;
2434 #ifdef MAC
2435 	error = mac_vnode_check_readlink(td->td_ucred, vp);
2436 	if (error != 0) {
2437 		vput(vp);
2438 		return (error);
2439 	}
2440 #endif
2441 	if (vp->v_type != VLNK)
2442 		error = EINVAL;
2443 	else {
2444 		aiov.iov_base = buf;
2445 		aiov.iov_len = count;
2446 		auio.uio_iov = &aiov;
2447 		auio.uio_iovcnt = 1;
2448 		auio.uio_offset = 0;
2449 		auio.uio_rw = UIO_READ;
2450 		auio.uio_segflg = bufseg;
2451 		auio.uio_td = td;
2452 		auio.uio_resid = count;
2453 		error = VOP_READLINK(vp, &auio, td->td_ucred);
2454 		td->td_retval[0] = count - auio.uio_resid;
2455 	}
2456 	vput(vp);
2457 	return (error);
2458 }
2459 
2460 /*
2461  * Common implementation code for chflags() and fchflags().
2462  */
2463 static int
2464 setfflags(td, vp, flags)
2465 	struct thread *td;
2466 	struct vnode *vp;
2467 	u_long flags;
2468 {
2469 	struct mount *mp;
2470 	struct vattr vattr;
2471 	int error;
2472 
2473 	/* We can't support the value matching VNOVAL. */
2474 	if (flags == VNOVAL)
2475 		return (EOPNOTSUPP);
2476 
2477 	/*
2478 	 * Prevent non-root users from setting flags on devices.  When
2479 	 * a device is reused, users can retain ownership of the device
2480 	 * if they are allowed to set flags and programs assume that
2481 	 * chown can't fail when done as root.
2482 	 */
2483 	if (vp->v_type == VCHR || vp->v_type == VBLK) {
2484 		error = priv_check(td, PRIV_VFS_CHFLAGS_DEV);
2485 		if (error != 0)
2486 			return (error);
2487 	}
2488 
2489 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2490 		return (error);
2491 	VATTR_NULL(&vattr);
2492 	vattr.va_flags = flags;
2493 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2494 #ifdef MAC
2495 	error = mac_vnode_check_setflags(td->td_ucred, vp, vattr.va_flags);
2496 	if (error == 0)
2497 #endif
2498 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
2499 	VOP_UNLOCK(vp, 0);
2500 	vn_finished_write(mp);
2501 	return (error);
2502 }
2503 
2504 /*
2505  * Change flags of a file given a path name.
2506  */
2507 #ifndef _SYS_SYSPROTO_H_
2508 struct chflags_args {
2509 	const char *path;
2510 	u_long	flags;
2511 };
2512 #endif
2513 int
2514 sys_chflags(td, uap)
2515 	struct thread *td;
2516 	register struct chflags_args /* {
2517 		const char *path;
2518 		u_long flags;
2519 	} */ *uap;
2520 {
2521 
2522 	return (kern_chflagsat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2523 	    uap->flags, 0));
2524 }
2525 
2526 #ifndef _SYS_SYSPROTO_H_
2527 struct chflagsat_args {
2528 	int	fd;
2529 	const char *path;
2530 	u_long	flags;
2531 	int	atflag;
2532 }
2533 #endif
2534 int
2535 sys_chflagsat(struct thread *td, struct chflagsat_args *uap)
2536 {
2537 	int fd = uap->fd;
2538 	const char *path = uap->path;
2539 	u_long flags = uap->flags;
2540 	int atflag = uap->atflag;
2541 
2542 	if (atflag & ~AT_SYMLINK_NOFOLLOW)
2543 		return (EINVAL);
2544 
2545 	return (kern_chflagsat(td, fd, path, UIO_USERSPACE, flags, atflag));
2546 }
2547 
2548 /*
2549  * Same as chflags() but doesn't follow symlinks.
2550  */
2551 int
2552 sys_lchflags(td, uap)
2553 	struct thread *td;
2554 	register struct lchflags_args /* {
2555 		const char *path;
2556 		u_long flags;
2557 	} */ *uap;
2558 {
2559 
2560 	return (kern_chflagsat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2561 	    uap->flags, AT_SYMLINK_NOFOLLOW));
2562 }
2563 
2564 static int
2565 kern_chflagsat(struct thread *td, int fd, const char *path,
2566     enum uio_seg pathseg, u_long flags, int atflag)
2567 {
2568 	struct nameidata nd;
2569 	cap_rights_t rights;
2570 	int error, follow;
2571 
2572 	AUDIT_ARG_FFLAGS(flags);
2573 	follow = (atflag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
2574 	NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, pathseg, path, fd,
2575 	    cap_rights_init(&rights, CAP_FCHFLAGS), td);
2576 	if ((error = namei(&nd)) != 0)
2577 		return (error);
2578 	NDFREE(&nd, NDF_ONLY_PNBUF);
2579 	error = setfflags(td, nd.ni_vp, flags);
2580 	vrele(nd.ni_vp);
2581 	return (error);
2582 }
2583 
2584 /*
2585  * Change flags of a file given a file descriptor.
2586  */
2587 #ifndef _SYS_SYSPROTO_H_
2588 struct fchflags_args {
2589 	int	fd;
2590 	u_long	flags;
2591 };
2592 #endif
2593 int
2594 sys_fchflags(td, uap)
2595 	struct thread *td;
2596 	register struct fchflags_args /* {
2597 		int fd;
2598 		u_long flags;
2599 	} */ *uap;
2600 {
2601 	struct file *fp;
2602 	cap_rights_t rights;
2603 	int error;
2604 
2605 	AUDIT_ARG_FD(uap->fd);
2606 	AUDIT_ARG_FFLAGS(uap->flags);
2607 	error = getvnode(td, uap->fd, cap_rights_init(&rights, CAP_FCHFLAGS),
2608 	    &fp);
2609 	if (error != 0)
2610 		return (error);
2611 #ifdef AUDIT
2612 	vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
2613 	AUDIT_ARG_VNODE1(fp->f_vnode);
2614 	VOP_UNLOCK(fp->f_vnode, 0);
2615 #endif
2616 	error = setfflags(td, fp->f_vnode, uap->flags);
2617 	fdrop(fp, td);
2618 	return (error);
2619 }
2620 
2621 /*
2622  * Common implementation code for chmod(), lchmod() and fchmod().
2623  */
2624 int
2625 setfmode(td, cred, vp, mode)
2626 	struct thread *td;
2627 	struct ucred *cred;
2628 	struct vnode *vp;
2629 	int mode;
2630 {
2631 	struct mount *mp;
2632 	struct vattr vattr;
2633 	int error;
2634 
2635 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2636 		return (error);
2637 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2638 	VATTR_NULL(&vattr);
2639 	vattr.va_mode = mode & ALLPERMS;
2640 #ifdef MAC
2641 	error = mac_vnode_check_setmode(cred, vp, vattr.va_mode);
2642 	if (error == 0)
2643 #endif
2644 		error = VOP_SETATTR(vp, &vattr, cred);
2645 	VOP_UNLOCK(vp, 0);
2646 	vn_finished_write(mp);
2647 	return (error);
2648 }
2649 
2650 /*
2651  * Change mode of a file given path name.
2652  */
2653 #ifndef _SYS_SYSPROTO_H_
2654 struct chmod_args {
2655 	char	*path;
2656 	int	mode;
2657 };
2658 #endif
2659 int
2660 sys_chmod(td, uap)
2661 	struct thread *td;
2662 	register struct chmod_args /* {
2663 		char *path;
2664 		int mode;
2665 	} */ *uap;
2666 {
2667 
2668 	return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2669 	    uap->mode, 0));
2670 }
2671 
2672 #ifndef _SYS_SYSPROTO_H_
2673 struct fchmodat_args {
2674 	int	dirfd;
2675 	char	*path;
2676 	mode_t	mode;
2677 	int	flag;
2678 }
2679 #endif
2680 int
2681 sys_fchmodat(struct thread *td, struct fchmodat_args *uap)
2682 {
2683 	int flag = uap->flag;
2684 	int fd = uap->fd;
2685 	char *path = uap->path;
2686 	mode_t mode = uap->mode;
2687 
2688 	if (flag & ~AT_SYMLINK_NOFOLLOW)
2689 		return (EINVAL);
2690 
2691 	return (kern_fchmodat(td, fd, path, UIO_USERSPACE, mode, flag));
2692 }
2693 
2694 /*
2695  * Change mode of a file given path name (don't follow links.)
2696  */
2697 #ifndef _SYS_SYSPROTO_H_
2698 struct lchmod_args {
2699 	char	*path;
2700 	int	mode;
2701 };
2702 #endif
2703 int
2704 sys_lchmod(td, uap)
2705 	struct thread *td;
2706 	register struct lchmod_args /* {
2707 		char *path;
2708 		int mode;
2709 	} */ *uap;
2710 {
2711 
2712 	return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2713 	    uap->mode, AT_SYMLINK_NOFOLLOW));
2714 }
2715 
2716 int
2717 kern_fchmodat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2718     mode_t mode, int flag)
2719 {
2720 	struct nameidata nd;
2721 	cap_rights_t rights;
2722 	int error, follow;
2723 
2724 	AUDIT_ARG_MODE(mode);
2725 	follow = (flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
2726 	NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, pathseg, path, fd,
2727 	    cap_rights_init(&rights, CAP_FCHMOD), td);
2728 	if ((error = namei(&nd)) != 0)
2729 		return (error);
2730 	NDFREE(&nd, NDF_ONLY_PNBUF);
2731 	error = setfmode(td, td->td_ucred, nd.ni_vp, mode);
2732 	vrele(nd.ni_vp);
2733 	return (error);
2734 }
2735 
2736 /*
2737  * Change mode of a file given a file descriptor.
2738  */
2739 #ifndef _SYS_SYSPROTO_H_
2740 struct fchmod_args {
2741 	int	fd;
2742 	int	mode;
2743 };
2744 #endif
2745 int
2746 sys_fchmod(struct thread *td, struct fchmod_args *uap)
2747 {
2748 	struct file *fp;
2749 	cap_rights_t rights;
2750 	int error;
2751 
2752 	AUDIT_ARG_FD(uap->fd);
2753 	AUDIT_ARG_MODE(uap->mode);
2754 
2755 	error = fget(td, uap->fd, cap_rights_init(&rights, CAP_FCHMOD), &fp);
2756 	if (error != 0)
2757 		return (error);
2758 	error = fo_chmod(fp, uap->mode, td->td_ucred, td);
2759 	fdrop(fp, td);
2760 	return (error);
2761 }
2762 
2763 /*
2764  * Common implementation for chown(), lchown(), and fchown()
2765  */
2766 int
2767 setfown(td, cred, vp, uid, gid)
2768 	struct thread *td;
2769 	struct ucred *cred;
2770 	struct vnode *vp;
2771 	uid_t uid;
2772 	gid_t gid;
2773 {
2774 	struct mount *mp;
2775 	struct vattr vattr;
2776 	int error;
2777 
2778 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2779 		return (error);
2780 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2781 	VATTR_NULL(&vattr);
2782 	vattr.va_uid = uid;
2783 	vattr.va_gid = gid;
2784 #ifdef MAC
2785 	error = mac_vnode_check_setowner(cred, vp, vattr.va_uid,
2786 	    vattr.va_gid);
2787 	if (error == 0)
2788 #endif
2789 		error = VOP_SETATTR(vp, &vattr, cred);
2790 	VOP_UNLOCK(vp, 0);
2791 	vn_finished_write(mp);
2792 	return (error);
2793 }
2794 
2795 /*
2796  * Set ownership given a path name.
2797  */
2798 #ifndef _SYS_SYSPROTO_H_
2799 struct chown_args {
2800 	char	*path;
2801 	int	uid;
2802 	int	gid;
2803 };
2804 #endif
2805 int
2806 sys_chown(td, uap)
2807 	struct thread *td;
2808 	register struct chown_args /* {
2809 		char *path;
2810 		int uid;
2811 		int gid;
2812 	} */ *uap;
2813 {
2814 
2815 	return (kern_fchownat(td, AT_FDCWD, uap->path, UIO_USERSPACE, uap->uid,
2816 	    uap->gid, 0));
2817 }
2818 
2819 #ifndef _SYS_SYSPROTO_H_
2820 struct fchownat_args {
2821 	int fd;
2822 	const char * path;
2823 	uid_t uid;
2824 	gid_t gid;
2825 	int flag;
2826 };
2827 #endif
2828 int
2829 sys_fchownat(struct thread *td, struct fchownat_args *uap)
2830 {
2831 	int flag;
2832 
2833 	flag = uap->flag;
2834 	if (flag & ~AT_SYMLINK_NOFOLLOW)
2835 		return (EINVAL);
2836 
2837 	return (kern_fchownat(td, uap->fd, uap->path, UIO_USERSPACE, uap->uid,
2838 	    uap->gid, uap->flag));
2839 }
2840 
2841 int
2842 kern_fchownat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2843     int uid, int gid, int flag)
2844 {
2845 	struct nameidata nd;
2846 	cap_rights_t rights;
2847 	int error, follow;
2848 
2849 	AUDIT_ARG_OWNER(uid, gid);
2850 	follow = (flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
2851 	NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, pathseg, path, fd,
2852 	    cap_rights_init(&rights, CAP_FCHOWN), td);
2853 
2854 	if ((error = namei(&nd)) != 0)
2855 		return (error);
2856 	NDFREE(&nd, NDF_ONLY_PNBUF);
2857 	error = setfown(td, td->td_ucred, nd.ni_vp, uid, gid);
2858 	vrele(nd.ni_vp);
2859 	return (error);
2860 }
2861 
2862 /*
2863  * Set ownership given a path name, do not cross symlinks.
2864  */
2865 #ifndef _SYS_SYSPROTO_H_
2866 struct lchown_args {
2867 	char	*path;
2868 	int	uid;
2869 	int	gid;
2870 };
2871 #endif
2872 int
2873 sys_lchown(td, uap)
2874 	struct thread *td;
2875 	register struct lchown_args /* {
2876 		char *path;
2877 		int uid;
2878 		int gid;
2879 	} */ *uap;
2880 {
2881 
2882 	return (kern_fchownat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2883 	    uap->uid, uap->gid, AT_SYMLINK_NOFOLLOW));
2884 }
2885 
2886 /*
2887  * Set ownership given a file descriptor.
2888  */
2889 #ifndef _SYS_SYSPROTO_H_
2890 struct fchown_args {
2891 	int	fd;
2892 	int	uid;
2893 	int	gid;
2894 };
2895 #endif
2896 int
2897 sys_fchown(td, uap)
2898 	struct thread *td;
2899 	register struct fchown_args /* {
2900 		int fd;
2901 		int uid;
2902 		int gid;
2903 	} */ *uap;
2904 {
2905 	struct file *fp;
2906 	cap_rights_t rights;
2907 	int error;
2908 
2909 	AUDIT_ARG_FD(uap->fd);
2910 	AUDIT_ARG_OWNER(uap->uid, uap->gid);
2911 	error = fget(td, uap->fd, cap_rights_init(&rights, CAP_FCHOWN), &fp);
2912 	if (error != 0)
2913 		return (error);
2914 	error = fo_chown(fp, uap->uid, uap->gid, td->td_ucred, td);
2915 	fdrop(fp, td);
2916 	return (error);
2917 }
2918 
2919 /*
2920  * Common implementation code for utimes(), lutimes(), and futimes().
2921  */
2922 static int
2923 getutimes(usrtvp, tvpseg, tsp)
2924 	const struct timeval *usrtvp;
2925 	enum uio_seg tvpseg;
2926 	struct timespec *tsp;
2927 {
2928 	struct timeval tv[2];
2929 	const struct timeval *tvp;
2930 	int error;
2931 
2932 	if (usrtvp == NULL) {
2933 		vfs_timestamp(&tsp[0]);
2934 		tsp[1] = tsp[0];
2935 	} else {
2936 		if (tvpseg == UIO_SYSSPACE) {
2937 			tvp = usrtvp;
2938 		} else {
2939 			if ((error = copyin(usrtvp, tv, sizeof(tv))) != 0)
2940 				return (error);
2941 			tvp = tv;
2942 		}
2943 
2944 		if (tvp[0].tv_usec < 0 || tvp[0].tv_usec >= 1000000 ||
2945 		    tvp[1].tv_usec < 0 || tvp[1].tv_usec >= 1000000)
2946 			return (EINVAL);
2947 		TIMEVAL_TO_TIMESPEC(&tvp[0], &tsp[0]);
2948 		TIMEVAL_TO_TIMESPEC(&tvp[1], &tsp[1]);
2949 	}
2950 	return (0);
2951 }
2952 
2953 /*
2954  * Common implementation code for futimens(), utimensat().
2955  */
2956 #define	UTIMENS_NULL	0x1
2957 #define	UTIMENS_EXIT	0x2
2958 static int
2959 getutimens(const struct timespec *usrtsp, enum uio_seg tspseg,
2960     struct timespec *tsp, int *retflags)
2961 {
2962 	struct timespec tsnow;
2963 	int error;
2964 
2965 	vfs_timestamp(&tsnow);
2966 	*retflags = 0;
2967 	if (usrtsp == NULL) {
2968 		tsp[0] = tsnow;
2969 		tsp[1] = tsnow;
2970 		*retflags |= UTIMENS_NULL;
2971 		return (0);
2972 	}
2973 	if (tspseg == UIO_SYSSPACE) {
2974 		tsp[0] = usrtsp[0];
2975 		tsp[1] = usrtsp[1];
2976 	} else if ((error = copyin(usrtsp, tsp, sizeof(*tsp) * 2)) != 0)
2977 		return (error);
2978 	if (tsp[0].tv_nsec == UTIME_OMIT && tsp[1].tv_nsec == UTIME_OMIT)
2979 		*retflags |= UTIMENS_EXIT;
2980 	if (tsp[0].tv_nsec == UTIME_NOW && tsp[1].tv_nsec == UTIME_NOW)
2981 		*retflags |= UTIMENS_NULL;
2982 	if (tsp[0].tv_nsec == UTIME_OMIT)
2983 		tsp[0].tv_sec = VNOVAL;
2984 	else if (tsp[0].tv_nsec == UTIME_NOW)
2985 		tsp[0] = tsnow;
2986 	else if (tsp[0].tv_nsec < 0 || tsp[0].tv_nsec >= 1000000000L)
2987 		return (EINVAL);
2988 	if (tsp[1].tv_nsec == UTIME_OMIT)
2989 		tsp[1].tv_sec = VNOVAL;
2990 	else if (tsp[1].tv_nsec == UTIME_NOW)
2991 		tsp[1] = tsnow;
2992 	else if (tsp[1].tv_nsec < 0 || tsp[1].tv_nsec >= 1000000000L)
2993 		return (EINVAL);
2994 
2995 	return (0);
2996 }
2997 
2998 /*
2999  * Common implementation code for utimes(), lutimes(), futimes(), futimens(),
3000  * and utimensat().
3001  */
3002 static int
3003 setutimes(td, vp, ts, numtimes, nullflag)
3004 	struct thread *td;
3005 	struct vnode *vp;
3006 	const struct timespec *ts;
3007 	int numtimes;
3008 	int nullflag;
3009 {
3010 	struct mount *mp;
3011 	struct vattr vattr;
3012 	int error, setbirthtime;
3013 
3014 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
3015 		return (error);
3016 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3017 	setbirthtime = 0;
3018 	if (numtimes < 3 && !VOP_GETATTR(vp, &vattr, td->td_ucred) &&
3019 	    timespeccmp(&ts[1], &vattr.va_birthtime, < ))
3020 		setbirthtime = 1;
3021 	VATTR_NULL(&vattr);
3022 	vattr.va_atime = ts[0];
3023 	vattr.va_mtime = ts[1];
3024 	if (setbirthtime)
3025 		vattr.va_birthtime = ts[1];
3026 	if (numtimes > 2)
3027 		vattr.va_birthtime = ts[2];
3028 	if (nullflag)
3029 		vattr.va_vaflags |= VA_UTIMES_NULL;
3030 #ifdef MAC
3031 	error = mac_vnode_check_setutimes(td->td_ucred, vp, vattr.va_atime,
3032 	    vattr.va_mtime);
3033 #endif
3034 	if (error == 0)
3035 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3036 	VOP_UNLOCK(vp, 0);
3037 	vn_finished_write(mp);
3038 	return (error);
3039 }
3040 
3041 /*
3042  * Set the access and modification times of a file.
3043  */
3044 #ifndef _SYS_SYSPROTO_H_
3045 struct utimes_args {
3046 	char	*path;
3047 	struct	timeval *tptr;
3048 };
3049 #endif
3050 int
3051 sys_utimes(td, uap)
3052 	struct thread *td;
3053 	register struct utimes_args /* {
3054 		char *path;
3055 		struct timeval *tptr;
3056 	} */ *uap;
3057 {
3058 
3059 	return (kern_utimesat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
3060 	    uap->tptr, UIO_USERSPACE));
3061 }
3062 
3063 #ifndef _SYS_SYSPROTO_H_
3064 struct futimesat_args {
3065 	int fd;
3066 	const char * path;
3067 	const struct timeval * times;
3068 };
3069 #endif
3070 int
3071 sys_futimesat(struct thread *td, struct futimesat_args *uap)
3072 {
3073 
3074 	return (kern_utimesat(td, uap->fd, uap->path, UIO_USERSPACE,
3075 	    uap->times, UIO_USERSPACE));
3076 }
3077 
3078 int
3079 kern_utimesat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
3080     struct timeval *tptr, enum uio_seg tptrseg)
3081 {
3082 	struct nameidata nd;
3083 	struct timespec ts[2];
3084 	cap_rights_t rights;
3085 	int error;
3086 
3087 	if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3088 		return (error);
3089 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, fd,
3090 	    cap_rights_init(&rights, CAP_FUTIMES), td);
3091 
3092 	if ((error = namei(&nd)) != 0)
3093 		return (error);
3094 	NDFREE(&nd, NDF_ONLY_PNBUF);
3095 	error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3096 	vrele(nd.ni_vp);
3097 	return (error);
3098 }
3099 
3100 /*
3101  * Set the access and modification times of a file.
3102  */
3103 #ifndef _SYS_SYSPROTO_H_
3104 struct lutimes_args {
3105 	char	*path;
3106 	struct	timeval *tptr;
3107 };
3108 #endif
3109 int
3110 sys_lutimes(td, uap)
3111 	struct thread *td;
3112 	register struct lutimes_args /* {
3113 		char *path;
3114 		struct timeval *tptr;
3115 	} */ *uap;
3116 {
3117 
3118 	return (kern_lutimes(td, uap->path, UIO_USERSPACE, uap->tptr,
3119 	    UIO_USERSPACE));
3120 }
3121 
3122 int
3123 kern_lutimes(struct thread *td, char *path, enum uio_seg pathseg,
3124     struct timeval *tptr, enum uio_seg tptrseg)
3125 {
3126 	struct timespec ts[2];
3127 	struct nameidata nd;
3128 	int error;
3129 
3130 	if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3131 		return (error);
3132 	NDINIT(&nd, LOOKUP, NOFOLLOW | AUDITVNODE1, pathseg, path, td);
3133 	if ((error = namei(&nd)) != 0)
3134 		return (error);
3135 	NDFREE(&nd, NDF_ONLY_PNBUF);
3136 	error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3137 	vrele(nd.ni_vp);
3138 	return (error);
3139 }
3140 
3141 /*
3142  * Set the access and modification times of a file.
3143  */
3144 #ifndef _SYS_SYSPROTO_H_
3145 struct futimes_args {
3146 	int	fd;
3147 	struct	timeval *tptr;
3148 };
3149 #endif
3150 int
3151 sys_futimes(td, uap)
3152 	struct thread *td;
3153 	register struct futimes_args /* {
3154 		int  fd;
3155 		struct timeval *tptr;
3156 	} */ *uap;
3157 {
3158 
3159 	return (kern_futimes(td, uap->fd, uap->tptr, UIO_USERSPACE));
3160 }
3161 
3162 int
3163 kern_futimes(struct thread *td, int fd, struct timeval *tptr,
3164     enum uio_seg tptrseg)
3165 {
3166 	struct timespec ts[2];
3167 	struct file *fp;
3168 	cap_rights_t rights;
3169 	int error;
3170 
3171 	AUDIT_ARG_FD(fd);
3172 	error = getutimes(tptr, tptrseg, ts);
3173 	if (error != 0)
3174 		return (error);
3175 	error = getvnode(td, fd, cap_rights_init(&rights, CAP_FUTIMES), &fp);
3176 	if (error != 0)
3177 		return (error);
3178 #ifdef AUDIT
3179 	vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
3180 	AUDIT_ARG_VNODE1(fp->f_vnode);
3181 	VOP_UNLOCK(fp->f_vnode, 0);
3182 #endif
3183 	error = setutimes(td, fp->f_vnode, ts, 2, tptr == NULL);
3184 	fdrop(fp, td);
3185 	return (error);
3186 }
3187 
3188 int
3189 sys_futimens(struct thread *td, struct futimens_args *uap)
3190 {
3191 
3192 	return (kern_futimens(td, uap->fd, uap->times, UIO_USERSPACE));
3193 }
3194 
3195 int
3196 kern_futimens(struct thread *td, int fd, struct timespec *tptr,
3197     enum uio_seg tptrseg)
3198 {
3199 	struct timespec ts[2];
3200 	struct file *fp;
3201 	cap_rights_t rights;
3202 	int error, flags;
3203 
3204 	AUDIT_ARG_FD(fd);
3205 	error = getutimens(tptr, tptrseg, ts, &flags);
3206 	if (error != 0)
3207 		return (error);
3208 	if (flags & UTIMENS_EXIT)
3209 		return (0);
3210 	error = getvnode(td, fd, cap_rights_init(&rights, CAP_FUTIMES), &fp);
3211 	if (error != 0)
3212 		return (error);
3213 #ifdef AUDIT
3214 	vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
3215 	AUDIT_ARG_VNODE1(fp->f_vnode);
3216 	VOP_UNLOCK(fp->f_vnode, 0);
3217 #endif
3218 	error = setutimes(td, fp->f_vnode, ts, 2, flags & UTIMENS_NULL);
3219 	fdrop(fp, td);
3220 	return (error);
3221 }
3222 
3223 int
3224 sys_utimensat(struct thread *td, struct utimensat_args *uap)
3225 {
3226 
3227 	return (kern_utimensat(td, uap->fd, uap->path, UIO_USERSPACE,
3228 	    uap->times, UIO_USERSPACE, uap->flag));
3229 }
3230 
3231 int
3232 kern_utimensat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
3233     struct timespec *tptr, enum uio_seg tptrseg, int flag)
3234 {
3235 	struct nameidata nd;
3236 	struct timespec ts[2];
3237 	cap_rights_t rights;
3238 	int error, flags;
3239 
3240 	if (flag & ~AT_SYMLINK_NOFOLLOW)
3241 		return (EINVAL);
3242 
3243 	if ((error = getutimens(tptr, tptrseg, ts, &flags)) != 0)
3244 		return (error);
3245 	NDINIT_ATRIGHTS(&nd, LOOKUP, ((flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW :
3246 	    FOLLOW) | AUDITVNODE1, pathseg, path, fd,
3247 	    cap_rights_init(&rights, CAP_FUTIMES), td);
3248 	if ((error = namei(&nd)) != 0)
3249 		return (error);
3250 	/*
3251 	 * We are allowed to call namei() regardless of 2xUTIME_OMIT.
3252 	 * POSIX states:
3253 	 * "If both tv_nsec fields are UTIME_OMIT... EACCESS may be detected."
3254 	 * "Search permission is denied by a component of the path prefix."
3255 	 */
3256 	NDFREE(&nd, NDF_ONLY_PNBUF);
3257 	if ((flags & UTIMENS_EXIT) == 0)
3258 		error = setutimes(td, nd.ni_vp, ts, 2, flags & UTIMENS_NULL);
3259 	vrele(nd.ni_vp);
3260 	return (error);
3261 }
3262 
3263 /*
3264  * Truncate a file given its path name.
3265  */
3266 #ifndef _SYS_SYSPROTO_H_
3267 struct truncate_args {
3268 	char	*path;
3269 	int	pad;
3270 	off_t	length;
3271 };
3272 #endif
3273 int
3274 sys_truncate(td, uap)
3275 	struct thread *td;
3276 	register struct truncate_args /* {
3277 		char *path;
3278 		int pad;
3279 		off_t length;
3280 	} */ *uap;
3281 {
3282 
3283 	return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3284 }
3285 
3286 int
3287 kern_truncate(struct thread *td, char *path, enum uio_seg pathseg, off_t length)
3288 {
3289 	struct mount *mp;
3290 	struct vnode *vp;
3291 	void *rl_cookie;
3292 	struct vattr vattr;
3293 	struct nameidata nd;
3294 	int error;
3295 
3296 	if (length < 0)
3297 		return(EINVAL);
3298 	NDINIT(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, td);
3299 	if ((error = namei(&nd)) != 0)
3300 		return (error);
3301 	vp = nd.ni_vp;
3302 	rl_cookie = vn_rangelock_wlock(vp, 0, OFF_MAX);
3303 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) {
3304 		vn_rangelock_unlock(vp, rl_cookie);
3305 		vrele(vp);
3306 		return (error);
3307 	}
3308 	NDFREE(&nd, NDF_ONLY_PNBUF);
3309 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3310 	if (vp->v_type == VDIR)
3311 		error = EISDIR;
3312 #ifdef MAC
3313 	else if ((error = mac_vnode_check_write(td->td_ucred, NOCRED, vp))) {
3314 	}
3315 #endif
3316 	else if ((error = vn_writechk(vp)) == 0 &&
3317 	    (error = VOP_ACCESS(vp, VWRITE, td->td_ucred, td)) == 0) {
3318 		VATTR_NULL(&vattr);
3319 		vattr.va_size = length;
3320 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3321 	}
3322 	VOP_UNLOCK(vp, 0);
3323 	vn_finished_write(mp);
3324 	vn_rangelock_unlock(vp, rl_cookie);
3325 	vrele(vp);
3326 	return (error);
3327 }
3328 
3329 #if defined(COMPAT_43)
3330 /*
3331  * Truncate a file given its path name.
3332  */
3333 #ifndef _SYS_SYSPROTO_H_
3334 struct otruncate_args {
3335 	char	*path;
3336 	long	length;
3337 };
3338 #endif
3339 int
3340 otruncate(td, uap)
3341 	struct thread *td;
3342 	register struct otruncate_args /* {
3343 		char *path;
3344 		long length;
3345 	} */ *uap;
3346 {
3347 	struct truncate_args /* {
3348 		char *path;
3349 		int pad;
3350 		off_t length;
3351 	} */ nuap;
3352 
3353 	nuap.path = uap->path;
3354 	nuap.length = uap->length;
3355 	return (sys_truncate(td, &nuap));
3356 }
3357 #endif /* COMPAT_43 */
3358 
3359 #if defined(COMPAT_FREEBSD6)
3360 /* Versions with the pad argument */
3361 int
3362 freebsd6_truncate(struct thread *td, struct freebsd6_truncate_args *uap)
3363 {
3364 	struct truncate_args ouap;
3365 
3366 	ouap.path = uap->path;
3367 	ouap.length = uap->length;
3368 	return (sys_truncate(td, &ouap));
3369 }
3370 
3371 int
3372 freebsd6_ftruncate(struct thread *td, struct freebsd6_ftruncate_args *uap)
3373 {
3374 	struct ftruncate_args ouap;
3375 
3376 	ouap.fd = uap->fd;
3377 	ouap.length = uap->length;
3378 	return (sys_ftruncate(td, &ouap));
3379 }
3380 #endif
3381 
3382 int
3383 kern_fsync(struct thread *td, int fd, bool fullsync)
3384 {
3385 	struct vnode *vp;
3386 	struct mount *mp;
3387 	struct file *fp;
3388 	cap_rights_t rights;
3389 	int error, lock_flags;
3390 
3391 	AUDIT_ARG_FD(fd);
3392 	error = getvnode(td, fd, cap_rights_init(&rights, CAP_FSYNC), &fp);
3393 	if (error != 0)
3394 		return (error);
3395 	vp = fp->f_vnode;
3396 #if 0
3397 	if (!fullsync)
3398 		/* XXXKIB: compete outstanding aio writes */;
3399 #endif
3400 	error = vn_start_write(vp, &mp, V_WAIT | PCATCH);
3401 	if (error != 0)
3402 		goto drop;
3403 	if (MNT_SHARED_WRITES(mp) ||
3404 	    ((mp == NULL) && MNT_SHARED_WRITES(vp->v_mount))) {
3405 		lock_flags = LK_SHARED;
3406 	} else {
3407 		lock_flags = LK_EXCLUSIVE;
3408 	}
3409 	vn_lock(vp, lock_flags | LK_RETRY);
3410 	AUDIT_ARG_VNODE1(vp);
3411 	if (vp->v_object != NULL) {
3412 		VM_OBJECT_WLOCK(vp->v_object);
3413 		vm_object_page_clean(vp->v_object, 0, 0, 0);
3414 		VM_OBJECT_WUNLOCK(vp->v_object);
3415 	}
3416 	error = fullsync ? VOP_FSYNC(vp, MNT_WAIT, td) : VOP_FDATASYNC(vp, td);
3417 	VOP_UNLOCK(vp, 0);
3418 	vn_finished_write(mp);
3419 drop:
3420 	fdrop(fp, td);
3421 	return (error);
3422 }
3423 
3424 /*
3425  * Sync an open file.
3426  */
3427 #ifndef _SYS_SYSPROTO_H_
3428 struct fsync_args {
3429 	int	fd;
3430 };
3431 #endif
3432 int
3433 sys_fsync(struct thread *td, struct fsync_args *uap)
3434 {
3435 
3436 	return (kern_fsync(td, uap->fd, true));
3437 }
3438 
3439 int
3440 sys_fdatasync(struct thread *td, struct fdatasync_args *uap)
3441 {
3442 
3443 	return (kern_fsync(td, uap->fd, false));
3444 }
3445 
3446 /*
3447  * Rename files.  Source and destination must either both be directories, or
3448  * both not be directories.  If target is a directory, it must be empty.
3449  */
3450 #ifndef _SYS_SYSPROTO_H_
3451 struct rename_args {
3452 	char	*from;
3453 	char	*to;
3454 };
3455 #endif
3456 int
3457 sys_rename(td, uap)
3458 	struct thread *td;
3459 	register struct rename_args /* {
3460 		char *from;
3461 		char *to;
3462 	} */ *uap;
3463 {
3464 
3465 	return (kern_renameat(td, AT_FDCWD, uap->from, AT_FDCWD,
3466 	    uap->to, UIO_USERSPACE));
3467 }
3468 
3469 #ifndef _SYS_SYSPROTO_H_
3470 struct renameat_args {
3471 	int	oldfd;
3472 	char	*old;
3473 	int	newfd;
3474 	char	*new;
3475 };
3476 #endif
3477 int
3478 sys_renameat(struct thread *td, struct renameat_args *uap)
3479 {
3480 
3481 	return (kern_renameat(td, uap->oldfd, uap->old, uap->newfd, uap->new,
3482 	    UIO_USERSPACE));
3483 }
3484 
3485 int
3486 kern_renameat(struct thread *td, int oldfd, char *old, int newfd, char *new,
3487     enum uio_seg pathseg)
3488 {
3489 	struct mount *mp = NULL;
3490 	struct vnode *tvp, *fvp, *tdvp;
3491 	struct nameidata fromnd, tond;
3492 	cap_rights_t rights;
3493 	int error;
3494 
3495 again:
3496 	bwillwrite();
3497 #ifdef MAC
3498 	NDINIT_ATRIGHTS(&fromnd, DELETE, LOCKPARENT | LOCKLEAF | SAVESTART |
3499 	    AUDITVNODE1, pathseg, old, oldfd,
3500 	    cap_rights_init(&rights, CAP_RENAMEAT_SOURCE), td);
3501 #else
3502 	NDINIT_ATRIGHTS(&fromnd, DELETE, WANTPARENT | SAVESTART | AUDITVNODE1,
3503 	    pathseg, old, oldfd,
3504 	    cap_rights_init(&rights, CAP_RENAMEAT_SOURCE), td);
3505 #endif
3506 
3507 	if ((error = namei(&fromnd)) != 0)
3508 		return (error);
3509 #ifdef MAC
3510 	error = mac_vnode_check_rename_from(td->td_ucred, fromnd.ni_dvp,
3511 	    fromnd.ni_vp, &fromnd.ni_cnd);
3512 	VOP_UNLOCK(fromnd.ni_dvp, 0);
3513 	if (fromnd.ni_dvp != fromnd.ni_vp)
3514 		VOP_UNLOCK(fromnd.ni_vp, 0);
3515 #endif
3516 	fvp = fromnd.ni_vp;
3517 	NDINIT_ATRIGHTS(&tond, RENAME, LOCKPARENT | LOCKLEAF | NOCACHE |
3518 	    SAVESTART | AUDITVNODE2, pathseg, new, newfd,
3519 	    cap_rights_init(&rights, CAP_RENAMEAT_TARGET), td);
3520 	if (fromnd.ni_vp->v_type == VDIR)
3521 		tond.ni_cnd.cn_flags |= WILLBEDIR;
3522 	if ((error = namei(&tond)) != 0) {
3523 		/* Translate error code for rename("dir1", "dir2/."). */
3524 		if (error == EISDIR && fvp->v_type == VDIR)
3525 			error = EINVAL;
3526 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3527 		vrele(fromnd.ni_dvp);
3528 		vrele(fvp);
3529 		goto out1;
3530 	}
3531 	tdvp = tond.ni_dvp;
3532 	tvp = tond.ni_vp;
3533 	error = vn_start_write(fvp, &mp, V_NOWAIT);
3534 	if (error != 0) {
3535 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3536 		NDFREE(&tond, NDF_ONLY_PNBUF);
3537 		if (tvp != NULL)
3538 			vput(tvp);
3539 		if (tdvp == tvp)
3540 			vrele(tdvp);
3541 		else
3542 			vput(tdvp);
3543 		vrele(fromnd.ni_dvp);
3544 		vrele(fvp);
3545 		vrele(tond.ni_startdir);
3546 		if (fromnd.ni_startdir != NULL)
3547 			vrele(fromnd.ni_startdir);
3548 		error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH);
3549 		if (error != 0)
3550 			return (error);
3551 		goto again;
3552 	}
3553 	if (tvp != NULL) {
3554 		if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
3555 			error = ENOTDIR;
3556 			goto out;
3557 		} else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
3558 			error = EISDIR;
3559 			goto out;
3560 		}
3561 #ifdef CAPABILITIES
3562 		if (newfd != AT_FDCWD) {
3563 			/*
3564 			 * If the target already exists we require CAP_UNLINKAT
3565 			 * from 'newfd'.
3566 			 */
3567 			error = cap_check(&tond.ni_filecaps.fc_rights,
3568 			    cap_rights_init(&rights, CAP_UNLINKAT));
3569 			if (error != 0)
3570 				goto out;
3571 		}
3572 #endif
3573 	}
3574 	if (fvp == tdvp) {
3575 		error = EINVAL;
3576 		goto out;
3577 	}
3578 	/*
3579 	 * If the source is the same as the destination (that is, if they
3580 	 * are links to the same vnode), then there is nothing to do.
3581 	 */
3582 	if (fvp == tvp)
3583 		error = -1;
3584 #ifdef MAC
3585 	else
3586 		error = mac_vnode_check_rename_to(td->td_ucred, tdvp,
3587 		    tond.ni_vp, fromnd.ni_dvp == tdvp, &tond.ni_cnd);
3588 #endif
3589 out:
3590 	if (error == 0) {
3591 		error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
3592 		    tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
3593 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3594 		NDFREE(&tond, NDF_ONLY_PNBUF);
3595 	} else {
3596 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3597 		NDFREE(&tond, NDF_ONLY_PNBUF);
3598 		if (tvp != NULL)
3599 			vput(tvp);
3600 		if (tdvp == tvp)
3601 			vrele(tdvp);
3602 		else
3603 			vput(tdvp);
3604 		vrele(fromnd.ni_dvp);
3605 		vrele(fvp);
3606 	}
3607 	vrele(tond.ni_startdir);
3608 	vn_finished_write(mp);
3609 out1:
3610 	if (fromnd.ni_startdir)
3611 		vrele(fromnd.ni_startdir);
3612 	if (error == -1)
3613 		return (0);
3614 	return (error);
3615 }
3616 
3617 /*
3618  * Make a directory file.
3619  */
3620 #ifndef _SYS_SYSPROTO_H_
3621 struct mkdir_args {
3622 	char	*path;
3623 	int	mode;
3624 };
3625 #endif
3626 int
3627 sys_mkdir(td, uap)
3628 	struct thread *td;
3629 	register struct mkdir_args /* {
3630 		char *path;
3631 		int mode;
3632 	} */ *uap;
3633 {
3634 
3635 	return (kern_mkdirat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
3636 	    uap->mode));
3637 }
3638 
3639 #ifndef _SYS_SYSPROTO_H_
3640 struct mkdirat_args {
3641 	int	fd;
3642 	char	*path;
3643 	mode_t	mode;
3644 };
3645 #endif
3646 int
3647 sys_mkdirat(struct thread *td, struct mkdirat_args *uap)
3648 {
3649 
3650 	return (kern_mkdirat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode));
3651 }
3652 
3653 int
3654 kern_mkdirat(struct thread *td, int fd, char *path, enum uio_seg segflg,
3655     int mode)
3656 {
3657 	struct mount *mp;
3658 	struct vnode *vp;
3659 	struct vattr vattr;
3660 	struct nameidata nd;
3661 	cap_rights_t rights;
3662 	int error;
3663 
3664 	AUDIT_ARG_MODE(mode);
3665 restart:
3666 	bwillwrite();
3667 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
3668 	    NOCACHE, segflg, path, fd, cap_rights_init(&rights, CAP_MKDIRAT),
3669 	    td);
3670 	nd.ni_cnd.cn_flags |= WILLBEDIR;
3671 	if ((error = namei(&nd)) != 0)
3672 		return (error);
3673 	vp = nd.ni_vp;
3674 	if (vp != NULL) {
3675 		NDFREE(&nd, NDF_ONLY_PNBUF);
3676 		/*
3677 		 * XXX namei called with LOCKPARENT but not LOCKLEAF has
3678 		 * the strange behaviour of leaving the vnode unlocked
3679 		 * if the target is the same vnode as the parent.
3680 		 */
3681 		if (vp == nd.ni_dvp)
3682 			vrele(nd.ni_dvp);
3683 		else
3684 			vput(nd.ni_dvp);
3685 		vrele(vp);
3686 		return (EEXIST);
3687 	}
3688 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3689 		NDFREE(&nd, NDF_ONLY_PNBUF);
3690 		vput(nd.ni_dvp);
3691 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3692 			return (error);
3693 		goto restart;
3694 	}
3695 	VATTR_NULL(&vattr);
3696 	vattr.va_type = VDIR;
3697 	vattr.va_mode = (mode & ACCESSPERMS) &~ td->td_proc->p_fd->fd_cmask;
3698 #ifdef MAC
3699 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
3700 	    &vattr);
3701 	if (error != 0)
3702 		goto out;
3703 #endif
3704 	error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
3705 #ifdef MAC
3706 out:
3707 #endif
3708 	NDFREE(&nd, NDF_ONLY_PNBUF);
3709 	vput(nd.ni_dvp);
3710 	if (error == 0)
3711 		vput(nd.ni_vp);
3712 	vn_finished_write(mp);
3713 	return (error);
3714 }
3715 
3716 /*
3717  * Remove a directory file.
3718  */
3719 #ifndef _SYS_SYSPROTO_H_
3720 struct rmdir_args {
3721 	char	*path;
3722 };
3723 #endif
3724 int
3725 sys_rmdir(td, uap)
3726 	struct thread *td;
3727 	struct rmdir_args /* {
3728 		char *path;
3729 	} */ *uap;
3730 {
3731 
3732 	return (kern_rmdirat(td, AT_FDCWD, uap->path, UIO_USERSPACE));
3733 }
3734 
3735 int
3736 kern_rmdirat(struct thread *td, int fd, char *path, enum uio_seg pathseg)
3737 {
3738 	struct mount *mp;
3739 	struct vnode *vp;
3740 	struct nameidata nd;
3741 	cap_rights_t rights;
3742 	int error;
3743 
3744 restart:
3745 	bwillwrite();
3746 	NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1,
3747 	    pathseg, path, fd, cap_rights_init(&rights, CAP_UNLINKAT), td);
3748 	if ((error = namei(&nd)) != 0)
3749 		return (error);
3750 	vp = nd.ni_vp;
3751 	if (vp->v_type != VDIR) {
3752 		error = ENOTDIR;
3753 		goto out;
3754 	}
3755 	/*
3756 	 * No rmdir "." please.
3757 	 */
3758 	if (nd.ni_dvp == vp) {
3759 		error = EINVAL;
3760 		goto out;
3761 	}
3762 	/*
3763 	 * The root of a mounted filesystem cannot be deleted.
3764 	 */
3765 	if (vp->v_vflag & VV_ROOT) {
3766 		error = EBUSY;
3767 		goto out;
3768 	}
3769 #ifdef MAC
3770 	error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
3771 	    &nd.ni_cnd);
3772 	if (error != 0)
3773 		goto out;
3774 #endif
3775 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3776 		NDFREE(&nd, NDF_ONLY_PNBUF);
3777 		vput(vp);
3778 		if (nd.ni_dvp == vp)
3779 			vrele(nd.ni_dvp);
3780 		else
3781 			vput(nd.ni_dvp);
3782 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3783 			return (error);
3784 		goto restart;
3785 	}
3786 	vfs_notify_upper(vp, VFS_NOTIFY_UPPER_UNLINK);
3787 	error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
3788 	vn_finished_write(mp);
3789 out:
3790 	NDFREE(&nd, NDF_ONLY_PNBUF);
3791 	vput(vp);
3792 	if (nd.ni_dvp == vp)
3793 		vrele(nd.ni_dvp);
3794 	else
3795 		vput(nd.ni_dvp);
3796 	return (error);
3797 }
3798 
3799 #ifdef COMPAT_43
3800 /*
3801  * Read a block of directory entries in a filesystem independent format.
3802  */
3803 #ifndef _SYS_SYSPROTO_H_
3804 struct ogetdirentries_args {
3805 	int	fd;
3806 	char	*buf;
3807 	u_int	count;
3808 	long	*basep;
3809 };
3810 #endif
3811 int
3812 ogetdirentries(struct thread *td, struct ogetdirentries_args *uap)
3813 {
3814 	long loff;
3815 	int error;
3816 
3817 	error = kern_ogetdirentries(td, uap, &loff);
3818 	if (error == 0)
3819 		error = copyout(&loff, uap->basep, sizeof(long));
3820 	return (error);
3821 }
3822 
3823 int
3824 kern_ogetdirentries(struct thread *td, struct ogetdirentries_args *uap,
3825     long *ploff)
3826 {
3827 	struct vnode *vp;
3828 	struct file *fp;
3829 	struct uio auio, kuio;
3830 	struct iovec aiov, kiov;
3831 	struct dirent *dp, *edp;
3832 	cap_rights_t rights;
3833 	caddr_t dirbuf;
3834 	int error, eofflag, readcnt;
3835 	long loff;
3836 	off_t foffset;
3837 
3838 	/* XXX arbitrary sanity limit on `count'. */
3839 	if (uap->count > 64 * 1024)
3840 		return (EINVAL);
3841 	error = getvnode(td, uap->fd, cap_rights_init(&rights, CAP_READ), &fp);
3842 	if (error != 0)
3843 		return (error);
3844 	if ((fp->f_flag & FREAD) == 0) {
3845 		fdrop(fp, td);
3846 		return (EBADF);
3847 	}
3848 	vp = fp->f_vnode;
3849 	foffset = foffset_lock(fp, 0);
3850 unionread:
3851 	if (vp->v_type != VDIR) {
3852 		foffset_unlock(fp, foffset, 0);
3853 		fdrop(fp, td);
3854 		return (EINVAL);
3855 	}
3856 	aiov.iov_base = uap->buf;
3857 	aiov.iov_len = uap->count;
3858 	auio.uio_iov = &aiov;
3859 	auio.uio_iovcnt = 1;
3860 	auio.uio_rw = UIO_READ;
3861 	auio.uio_segflg = UIO_USERSPACE;
3862 	auio.uio_td = td;
3863 	auio.uio_resid = uap->count;
3864 	vn_lock(vp, LK_SHARED | LK_RETRY);
3865 	loff = auio.uio_offset = foffset;
3866 #ifdef MAC
3867 	error = mac_vnode_check_readdir(td->td_ucred, vp);
3868 	if (error != 0) {
3869 		VOP_UNLOCK(vp, 0);
3870 		foffset_unlock(fp, foffset, FOF_NOUPDATE);
3871 		fdrop(fp, td);
3872 		return (error);
3873 	}
3874 #endif
3875 #	if (BYTE_ORDER != LITTLE_ENDIAN)
3876 		if (vp->v_mount->mnt_maxsymlinklen <= 0) {
3877 			error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag,
3878 			    NULL, NULL);
3879 			foffset = auio.uio_offset;
3880 		} else
3881 #	endif
3882 	{
3883 		kuio = auio;
3884 		kuio.uio_iov = &kiov;
3885 		kuio.uio_segflg = UIO_SYSSPACE;
3886 		kiov.iov_len = uap->count;
3887 		dirbuf = malloc(uap->count, M_TEMP, M_WAITOK);
3888 		kiov.iov_base = dirbuf;
3889 		error = VOP_READDIR(vp, &kuio, fp->f_cred, &eofflag,
3890 			    NULL, NULL);
3891 		foffset = kuio.uio_offset;
3892 		if (error == 0) {
3893 			readcnt = uap->count - kuio.uio_resid;
3894 			edp = (struct dirent *)&dirbuf[readcnt];
3895 			for (dp = (struct dirent *)dirbuf; dp < edp; ) {
3896 #				if (BYTE_ORDER == LITTLE_ENDIAN)
3897 					/*
3898 					 * The expected low byte of
3899 					 * dp->d_namlen is our dp->d_type.
3900 					 * The high MBZ byte of dp->d_namlen
3901 					 * is our dp->d_namlen.
3902 					 */
3903 					dp->d_type = dp->d_namlen;
3904 					dp->d_namlen = 0;
3905 #				else
3906 					/*
3907 					 * The dp->d_type is the high byte
3908 					 * of the expected dp->d_namlen,
3909 					 * so must be zero'ed.
3910 					 */
3911 					dp->d_type = 0;
3912 #				endif
3913 				if (dp->d_reclen > 0) {
3914 					dp = (struct dirent *)
3915 					    ((char *)dp + dp->d_reclen);
3916 				} else {
3917 					error = EIO;
3918 					break;
3919 				}
3920 			}
3921 			if (dp >= edp)
3922 				error = uiomove(dirbuf, readcnt, &auio);
3923 		}
3924 		free(dirbuf, M_TEMP);
3925 	}
3926 	if (error != 0) {
3927 		VOP_UNLOCK(vp, 0);
3928 		foffset_unlock(fp, foffset, 0);
3929 		fdrop(fp, td);
3930 		return (error);
3931 	}
3932 	if (uap->count == auio.uio_resid &&
3933 	    (vp->v_vflag & VV_ROOT) &&
3934 	    (vp->v_mount->mnt_flag & MNT_UNION)) {
3935 		struct vnode *tvp = vp;
3936 		vp = vp->v_mount->mnt_vnodecovered;
3937 		VREF(vp);
3938 		fp->f_vnode = vp;
3939 		fp->f_data = vp;
3940 		foffset = 0;
3941 		vput(tvp);
3942 		goto unionread;
3943 	}
3944 	VOP_UNLOCK(vp, 0);
3945 	foffset_unlock(fp, foffset, 0);
3946 	fdrop(fp, td);
3947 	td->td_retval[0] = uap->count - auio.uio_resid;
3948 	if (error == 0)
3949 		*ploff = loff;
3950 	return (error);
3951 }
3952 #endif /* COMPAT_43 */
3953 
3954 /*
3955  * Read a block of directory entries in a filesystem independent format.
3956  */
3957 #ifndef _SYS_SYSPROTO_H_
3958 struct getdirentries_args {
3959 	int	fd;
3960 	char	*buf;
3961 	u_int	count;
3962 	long	*basep;
3963 };
3964 #endif
3965 int
3966 sys_getdirentries(td, uap)
3967 	struct thread *td;
3968 	register struct getdirentries_args /* {
3969 		int fd;
3970 		char *buf;
3971 		u_int count;
3972 		long *basep;
3973 	} */ *uap;
3974 {
3975 	long base;
3976 	int error;
3977 
3978 	error = kern_getdirentries(td, uap->fd, uap->buf, uap->count, &base,
3979 	    NULL, UIO_USERSPACE);
3980 	if (error != 0)
3981 		return (error);
3982 	if (uap->basep != NULL)
3983 		error = copyout(&base, uap->basep, sizeof(long));
3984 	return (error);
3985 }
3986 
3987 int
3988 kern_getdirentries(struct thread *td, int fd, char *buf, u_int count,
3989     long *basep, ssize_t *residp, enum uio_seg bufseg)
3990 {
3991 	struct vnode *vp;
3992 	struct file *fp;
3993 	struct uio auio;
3994 	struct iovec aiov;
3995 	cap_rights_t rights;
3996 	long loff;
3997 	int error, eofflag;
3998 	off_t foffset;
3999 
4000 	AUDIT_ARG_FD(fd);
4001 	if (count > IOSIZE_MAX)
4002 		return (EINVAL);
4003 	auio.uio_resid = count;
4004 	error = getvnode(td, fd, cap_rights_init(&rights, CAP_READ), &fp);
4005 	if (error != 0)
4006 		return (error);
4007 	if ((fp->f_flag & FREAD) == 0) {
4008 		fdrop(fp, td);
4009 		return (EBADF);
4010 	}
4011 	vp = fp->f_vnode;
4012 	foffset = foffset_lock(fp, 0);
4013 unionread:
4014 	if (vp->v_type != VDIR) {
4015 		error = EINVAL;
4016 		goto fail;
4017 	}
4018 	aiov.iov_base = buf;
4019 	aiov.iov_len = count;
4020 	auio.uio_iov = &aiov;
4021 	auio.uio_iovcnt = 1;
4022 	auio.uio_rw = UIO_READ;
4023 	auio.uio_segflg = bufseg;
4024 	auio.uio_td = td;
4025 	vn_lock(vp, LK_SHARED | LK_RETRY);
4026 	AUDIT_ARG_VNODE1(vp);
4027 	loff = auio.uio_offset = foffset;
4028 #ifdef MAC
4029 	error = mac_vnode_check_readdir(td->td_ucred, vp);
4030 	if (error == 0)
4031 #endif
4032 		error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, NULL,
4033 		    NULL);
4034 	foffset = auio.uio_offset;
4035 	if (error != 0) {
4036 		VOP_UNLOCK(vp, 0);
4037 		goto fail;
4038 	}
4039 	if (count == auio.uio_resid &&
4040 	    (vp->v_vflag & VV_ROOT) &&
4041 	    (vp->v_mount->mnt_flag & MNT_UNION)) {
4042 		struct vnode *tvp = vp;
4043 
4044 		vp = vp->v_mount->mnt_vnodecovered;
4045 		VREF(vp);
4046 		fp->f_vnode = vp;
4047 		fp->f_data = vp;
4048 		foffset = 0;
4049 		vput(tvp);
4050 		goto unionread;
4051 	}
4052 	VOP_UNLOCK(vp, 0);
4053 	*basep = loff;
4054 	if (residp != NULL)
4055 		*residp = auio.uio_resid;
4056 	td->td_retval[0] = count - auio.uio_resid;
4057 fail:
4058 	foffset_unlock(fp, foffset, 0);
4059 	fdrop(fp, td);
4060 	return (error);
4061 }
4062 
4063 #ifndef _SYS_SYSPROTO_H_
4064 struct getdents_args {
4065 	int fd;
4066 	char *buf;
4067 	size_t count;
4068 };
4069 #endif
4070 int
4071 sys_getdents(td, uap)
4072 	struct thread *td;
4073 	register struct getdents_args /* {
4074 		int fd;
4075 		char *buf;
4076 		u_int count;
4077 	} */ *uap;
4078 {
4079 	struct getdirentries_args ap;
4080 
4081 	ap.fd = uap->fd;
4082 	ap.buf = uap->buf;
4083 	ap.count = uap->count;
4084 	ap.basep = NULL;
4085 	return (sys_getdirentries(td, &ap));
4086 }
4087 
4088 /*
4089  * Set the mode mask for creation of filesystem nodes.
4090  */
4091 #ifndef _SYS_SYSPROTO_H_
4092 struct umask_args {
4093 	int	newmask;
4094 };
4095 #endif
4096 int
4097 sys_umask(td, uap)
4098 	struct thread *td;
4099 	struct umask_args /* {
4100 		int newmask;
4101 	} */ *uap;
4102 {
4103 	struct filedesc *fdp;
4104 
4105 	fdp = td->td_proc->p_fd;
4106 	FILEDESC_XLOCK(fdp);
4107 	td->td_retval[0] = fdp->fd_cmask;
4108 	fdp->fd_cmask = uap->newmask & ALLPERMS;
4109 	FILEDESC_XUNLOCK(fdp);
4110 	return (0);
4111 }
4112 
4113 /*
4114  * Void all references to file by ripping underlying filesystem away from
4115  * vnode.
4116  */
4117 #ifndef _SYS_SYSPROTO_H_
4118 struct revoke_args {
4119 	char	*path;
4120 };
4121 #endif
4122 int
4123 sys_revoke(td, uap)
4124 	struct thread *td;
4125 	register struct revoke_args /* {
4126 		char *path;
4127 	} */ *uap;
4128 {
4129 	struct vnode *vp;
4130 	struct vattr vattr;
4131 	struct nameidata nd;
4132 	int error;
4133 
4134 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
4135 	    uap->path, td);
4136 	if ((error = namei(&nd)) != 0)
4137 		return (error);
4138 	vp = nd.ni_vp;
4139 	NDFREE(&nd, NDF_ONLY_PNBUF);
4140 	if (vp->v_type != VCHR || vp->v_rdev == NULL) {
4141 		error = EINVAL;
4142 		goto out;
4143 	}
4144 #ifdef MAC
4145 	error = mac_vnode_check_revoke(td->td_ucred, vp);
4146 	if (error != 0)
4147 		goto out;
4148 #endif
4149 	error = VOP_GETATTR(vp, &vattr, td->td_ucred);
4150 	if (error != 0)
4151 		goto out;
4152 	if (td->td_ucred->cr_uid != vattr.va_uid) {
4153 		error = priv_check(td, PRIV_VFS_ADMIN);
4154 		if (error != 0)
4155 			goto out;
4156 	}
4157 	if (vcount(vp) > 1)
4158 		VOP_REVOKE(vp, REVOKEALL);
4159 out:
4160 	vput(vp);
4161 	return (error);
4162 }
4163 
4164 /*
4165  * Convert a user file descriptor to a kernel file entry and check that, if it
4166  * is a capability, the correct rights are present. A reference on the file
4167  * entry is held upon returning.
4168  */
4169 int
4170 getvnode(struct thread *td, int fd, cap_rights_t *rightsp, struct file **fpp)
4171 {
4172 	struct file *fp;
4173 	int error;
4174 
4175 	error = fget_unlocked(td->td_proc->p_fd, fd, rightsp, &fp, NULL);
4176 	if (error != 0)
4177 		return (error);
4178 
4179 	/*
4180 	 * The file could be not of the vnode type, or it may be not
4181 	 * yet fully initialized, in which case the f_vnode pointer
4182 	 * may be set, but f_ops is still badfileops.  E.g.,
4183 	 * devfs_open() transiently create such situation to
4184 	 * facilitate csw d_fdopen().
4185 	 *
4186 	 * Dupfdopen() handling in kern_openat() installs the
4187 	 * half-baked file into the process descriptor table, allowing
4188 	 * other thread to dereference it. Guard against the race by
4189 	 * checking f_ops.
4190 	 */
4191 	if (fp->f_vnode == NULL || fp->f_ops == &badfileops) {
4192 		fdrop(fp, td);
4193 		return (EINVAL);
4194 	}
4195 	*fpp = fp;
4196 	return (0);
4197 }
4198 
4199 
4200 /*
4201  * Get an (NFS) file handle.
4202  */
4203 #ifndef _SYS_SYSPROTO_H_
4204 struct lgetfh_args {
4205 	char	*fname;
4206 	fhandle_t *fhp;
4207 };
4208 #endif
4209 int
4210 sys_lgetfh(td, uap)
4211 	struct thread *td;
4212 	register struct lgetfh_args *uap;
4213 {
4214 	struct nameidata nd;
4215 	fhandle_t fh;
4216 	register struct vnode *vp;
4217 	int error;
4218 
4219 	error = priv_check(td, PRIV_VFS_GETFH);
4220 	if (error != 0)
4221 		return (error);
4222 	NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
4223 	    uap->fname, td);
4224 	error = namei(&nd);
4225 	if (error != 0)
4226 		return (error);
4227 	NDFREE(&nd, NDF_ONLY_PNBUF);
4228 	vp = nd.ni_vp;
4229 	bzero(&fh, sizeof(fh));
4230 	fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
4231 	error = VOP_VPTOFH(vp, &fh.fh_fid);
4232 	vput(vp);
4233 	if (error == 0)
4234 		error = copyout(&fh, uap->fhp, sizeof (fh));
4235 	return (error);
4236 }
4237 
4238 #ifndef _SYS_SYSPROTO_H_
4239 struct getfh_args {
4240 	char	*fname;
4241 	fhandle_t *fhp;
4242 };
4243 #endif
4244 int
4245 sys_getfh(td, uap)
4246 	struct thread *td;
4247 	register struct getfh_args *uap;
4248 {
4249 	struct nameidata nd;
4250 	fhandle_t fh;
4251 	register struct vnode *vp;
4252 	int error;
4253 
4254 	error = priv_check(td, PRIV_VFS_GETFH);
4255 	if (error != 0)
4256 		return (error);
4257 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
4258 	    uap->fname, td);
4259 	error = namei(&nd);
4260 	if (error != 0)
4261 		return (error);
4262 	NDFREE(&nd, NDF_ONLY_PNBUF);
4263 	vp = nd.ni_vp;
4264 	bzero(&fh, sizeof(fh));
4265 	fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
4266 	error = VOP_VPTOFH(vp, &fh.fh_fid);
4267 	vput(vp);
4268 	if (error == 0)
4269 		error = copyout(&fh, uap->fhp, sizeof (fh));
4270 	return (error);
4271 }
4272 
4273 /*
4274  * syscall for the rpc.lockd to use to translate a NFS file handle into an
4275  * open descriptor.
4276  *
4277  * warning: do not remove the priv_check() call or this becomes one giant
4278  * security hole.
4279  */
4280 #ifndef _SYS_SYSPROTO_H_
4281 struct fhopen_args {
4282 	const struct fhandle *u_fhp;
4283 	int flags;
4284 };
4285 #endif
4286 int
4287 sys_fhopen(td, uap)
4288 	struct thread *td;
4289 	struct fhopen_args /* {
4290 		const struct fhandle *u_fhp;
4291 		int flags;
4292 	} */ *uap;
4293 {
4294 	struct mount *mp;
4295 	struct vnode *vp;
4296 	struct fhandle fhp;
4297 	struct file *fp;
4298 	int fmode, error;
4299 	int indx;
4300 
4301 	error = priv_check(td, PRIV_VFS_FHOPEN);
4302 	if (error != 0)
4303 		return (error);
4304 	indx = -1;
4305 	fmode = FFLAGS(uap->flags);
4306 	/* why not allow a non-read/write open for our lockd? */
4307 	if (((fmode & (FREAD | FWRITE)) == 0) || (fmode & O_CREAT))
4308 		return (EINVAL);
4309 	error = copyin(uap->u_fhp, &fhp, sizeof(fhp));
4310 	if (error != 0)
4311 		return(error);
4312 	/* find the mount point */
4313 	mp = vfs_busyfs(&fhp.fh_fsid);
4314 	if (mp == NULL)
4315 		return (ESTALE);
4316 	/* now give me my vnode, it gets returned to me locked */
4317 	error = VFS_FHTOVP(mp, &fhp.fh_fid, LK_EXCLUSIVE, &vp);
4318 	vfs_unbusy(mp);
4319 	if (error != 0)
4320 		return (error);
4321 
4322 	error = falloc_noinstall(td, &fp);
4323 	if (error != 0) {
4324 		vput(vp);
4325 		return (error);
4326 	}
4327 	/*
4328 	 * An extra reference on `fp' has been held for us by
4329 	 * falloc_noinstall().
4330 	 */
4331 
4332 #ifdef INVARIANTS
4333 	td->td_dupfd = -1;
4334 #endif
4335 	error = vn_open_vnode(vp, fmode, td->td_ucred, td, fp);
4336 	if (error != 0) {
4337 		KASSERT(fp->f_ops == &badfileops,
4338 		    ("VOP_OPEN in fhopen() set f_ops"));
4339 		KASSERT(td->td_dupfd < 0,
4340 		    ("fhopen() encountered fdopen()"));
4341 
4342 		vput(vp);
4343 		goto bad;
4344 	}
4345 #ifdef INVARIANTS
4346 	td->td_dupfd = 0;
4347 #endif
4348 	fp->f_vnode = vp;
4349 	fp->f_seqcount = 1;
4350 	finit(fp, (fmode & FMASK) | (fp->f_flag & FHASLOCK), DTYPE_VNODE, vp,
4351 	    &vnops);
4352 	VOP_UNLOCK(vp, 0);
4353 	if ((fmode & O_TRUNC) != 0) {
4354 		error = fo_truncate(fp, 0, td->td_ucred, td);
4355 		if (error != 0)
4356 			goto bad;
4357 	}
4358 
4359 	error = finstall(td, fp, &indx, fmode, NULL);
4360 bad:
4361 	fdrop(fp, td);
4362 	td->td_retval[0] = indx;
4363 	return (error);
4364 }
4365 
4366 /*
4367  * Stat an (NFS) file handle.
4368  */
4369 #ifndef _SYS_SYSPROTO_H_
4370 struct fhstat_args {
4371 	struct fhandle *u_fhp;
4372 	struct stat *sb;
4373 };
4374 #endif
4375 int
4376 sys_fhstat(td, uap)
4377 	struct thread *td;
4378 	register struct fhstat_args /* {
4379 		struct fhandle *u_fhp;
4380 		struct stat *sb;
4381 	} */ *uap;
4382 {
4383 	struct stat sb;
4384 	struct fhandle fh;
4385 	int error;
4386 
4387 	error = copyin(uap->u_fhp, &fh, sizeof(fh));
4388 	if (error != 0)
4389 		return (error);
4390 	error = kern_fhstat(td, fh, &sb);
4391 	if (error == 0)
4392 		error = copyout(&sb, uap->sb, sizeof(sb));
4393 	return (error);
4394 }
4395 
4396 int
4397 kern_fhstat(struct thread *td, struct fhandle fh, struct stat *sb)
4398 {
4399 	struct mount *mp;
4400 	struct vnode *vp;
4401 	int error;
4402 
4403 	error = priv_check(td, PRIV_VFS_FHSTAT);
4404 	if (error != 0)
4405 		return (error);
4406 	if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4407 		return (ESTALE);
4408 	error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4409 	vfs_unbusy(mp);
4410 	if (error != 0)
4411 		return (error);
4412 	error = vn_stat(vp, sb, td->td_ucred, NOCRED, td);
4413 	vput(vp);
4414 	return (error);
4415 }
4416 
4417 /*
4418  * Implement fstatfs() for (NFS) file handles.
4419  */
4420 #ifndef _SYS_SYSPROTO_H_
4421 struct fhstatfs_args {
4422 	struct fhandle *u_fhp;
4423 	struct statfs *buf;
4424 };
4425 #endif
4426 int
4427 sys_fhstatfs(td, uap)
4428 	struct thread *td;
4429 	struct fhstatfs_args /* {
4430 		struct fhandle *u_fhp;
4431 		struct statfs *buf;
4432 	} */ *uap;
4433 {
4434 	struct statfs sf;
4435 	fhandle_t fh;
4436 	int error;
4437 
4438 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
4439 	if (error != 0)
4440 		return (error);
4441 	error = kern_fhstatfs(td, fh, &sf);
4442 	if (error != 0)
4443 		return (error);
4444 	return (copyout(&sf, uap->buf, sizeof(sf)));
4445 }
4446 
4447 int
4448 kern_fhstatfs(struct thread *td, fhandle_t fh, struct statfs *buf)
4449 {
4450 	struct statfs *sp;
4451 	struct mount *mp;
4452 	struct vnode *vp;
4453 	int error;
4454 
4455 	error = priv_check(td, PRIV_VFS_FHSTATFS);
4456 	if (error != 0)
4457 		return (error);
4458 	if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4459 		return (ESTALE);
4460 	error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4461 	if (error != 0) {
4462 		vfs_unbusy(mp);
4463 		return (error);
4464 	}
4465 	vput(vp);
4466 	error = prison_canseemount(td->td_ucred, mp);
4467 	if (error != 0)
4468 		goto out;
4469 #ifdef MAC
4470 	error = mac_mount_check_stat(td->td_ucred, mp);
4471 	if (error != 0)
4472 		goto out;
4473 #endif
4474 	/*
4475 	 * Set these in case the underlying filesystem fails to do so.
4476 	 */
4477 	sp = &mp->mnt_stat;
4478 	sp->f_version = STATFS_VERSION;
4479 	sp->f_namemax = NAME_MAX;
4480 	sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
4481 	error = VFS_STATFS(mp, sp);
4482 	if (error == 0)
4483 		*buf = *sp;
4484 out:
4485 	vfs_unbusy(mp);
4486 	return (error);
4487 }
4488 
4489 int
4490 kern_posix_fallocate(struct thread *td, int fd, off_t offset, off_t len)
4491 {
4492 	struct file *fp;
4493 	struct mount *mp;
4494 	struct vnode *vp;
4495 	cap_rights_t rights;
4496 	off_t olen, ooffset;
4497 	int error;
4498 
4499 	if (offset < 0 || len <= 0)
4500 		return (EINVAL);
4501 	/* Check for wrap. */
4502 	if (offset > OFF_MAX - len)
4503 		return (EFBIG);
4504 	error = fget(td, fd, cap_rights_init(&rights, CAP_WRITE), &fp);
4505 	if (error != 0)
4506 		return (error);
4507 	if ((fp->f_ops->fo_flags & DFLAG_SEEKABLE) == 0) {
4508 		error = ESPIPE;
4509 		goto out;
4510 	}
4511 	if ((fp->f_flag & FWRITE) == 0) {
4512 		error = EBADF;
4513 		goto out;
4514 	}
4515 	if (fp->f_type != DTYPE_VNODE) {
4516 		error = ENODEV;
4517 		goto out;
4518 	}
4519 	vp = fp->f_vnode;
4520 	if (vp->v_type != VREG) {
4521 		error = ENODEV;
4522 		goto out;
4523 	}
4524 
4525 	/* Allocating blocks may take a long time, so iterate. */
4526 	for (;;) {
4527 		olen = len;
4528 		ooffset = offset;
4529 
4530 		bwillwrite();
4531 		mp = NULL;
4532 		error = vn_start_write(vp, &mp, V_WAIT | PCATCH);
4533 		if (error != 0)
4534 			break;
4535 		error = vn_lock(vp, LK_EXCLUSIVE);
4536 		if (error != 0) {
4537 			vn_finished_write(mp);
4538 			break;
4539 		}
4540 #ifdef MAC
4541 		error = mac_vnode_check_write(td->td_ucred, fp->f_cred, vp);
4542 		if (error == 0)
4543 #endif
4544 			error = VOP_ALLOCATE(vp, &offset, &len);
4545 		VOP_UNLOCK(vp, 0);
4546 		vn_finished_write(mp);
4547 
4548 		if (olen + ooffset != offset + len) {
4549 			panic("offset + len changed from %jx/%jx to %jx/%jx",
4550 			    ooffset, olen, offset, len);
4551 		}
4552 		if (error != 0 || len == 0)
4553 			break;
4554 		KASSERT(olen > len, ("Iteration did not make progress?"));
4555 		maybe_yield();
4556 	}
4557  out:
4558 	fdrop(fp, td);
4559 	return (error);
4560 }
4561 
4562 int
4563 sys_posix_fallocate(struct thread *td, struct posix_fallocate_args *uap)
4564 {
4565 	int error;
4566 
4567 	error = kern_posix_fallocate(td, uap->fd, uap->offset, uap->len);
4568 	return (kern_posix_error(td, error));
4569 }
4570 
4571 /*
4572  * Unlike madvise(2), we do not make a best effort to remember every
4573  * possible caching hint.  Instead, we remember the last setting with
4574  * the exception that we will allow POSIX_FADV_NORMAL to adjust the
4575  * region of any current setting.
4576  */
4577 int
4578 kern_posix_fadvise(struct thread *td, int fd, off_t offset, off_t len,
4579     int advice)
4580 {
4581 	struct fadvise_info *fa, *new;
4582 	struct file *fp;
4583 	struct vnode *vp;
4584 	cap_rights_t rights;
4585 	off_t end;
4586 	int error;
4587 
4588 	if (offset < 0 || len < 0 || offset > OFF_MAX - len)
4589 		return (EINVAL);
4590 	switch (advice) {
4591 	case POSIX_FADV_SEQUENTIAL:
4592 	case POSIX_FADV_RANDOM:
4593 	case POSIX_FADV_NOREUSE:
4594 		new = malloc(sizeof(*fa), M_FADVISE, M_WAITOK);
4595 		break;
4596 	case POSIX_FADV_NORMAL:
4597 	case POSIX_FADV_WILLNEED:
4598 	case POSIX_FADV_DONTNEED:
4599 		new = NULL;
4600 		break;
4601 	default:
4602 		return (EINVAL);
4603 	}
4604 	/* XXX: CAP_POSIX_FADVISE? */
4605 	error = fget(td, fd, cap_rights_init(&rights), &fp);
4606 	if (error != 0)
4607 		goto out;
4608 	if ((fp->f_ops->fo_flags & DFLAG_SEEKABLE) == 0) {
4609 		error = ESPIPE;
4610 		goto out;
4611 	}
4612 	if (fp->f_type != DTYPE_VNODE) {
4613 		error = ENODEV;
4614 		goto out;
4615 	}
4616 	vp = fp->f_vnode;
4617 	if (vp->v_type != VREG) {
4618 		error = ENODEV;
4619 		goto out;
4620 	}
4621 	if (len == 0)
4622 		end = OFF_MAX;
4623 	else
4624 		end = offset + len - 1;
4625 	switch (advice) {
4626 	case POSIX_FADV_SEQUENTIAL:
4627 	case POSIX_FADV_RANDOM:
4628 	case POSIX_FADV_NOREUSE:
4629 		/*
4630 		 * Try to merge any existing non-standard region with
4631 		 * this new region if possible, otherwise create a new
4632 		 * non-standard region for this request.
4633 		 */
4634 		mtx_pool_lock(mtxpool_sleep, fp);
4635 		fa = fp->f_advice;
4636 		if (fa != NULL && fa->fa_advice == advice &&
4637 		    ((fa->fa_start <= end && fa->fa_end >= offset) ||
4638 		    (end != OFF_MAX && fa->fa_start == end + 1) ||
4639 		    (fa->fa_end != OFF_MAX && fa->fa_end + 1 == offset))) {
4640 			if (offset < fa->fa_start)
4641 				fa->fa_start = offset;
4642 			if (end > fa->fa_end)
4643 				fa->fa_end = end;
4644 		} else {
4645 			new->fa_advice = advice;
4646 			new->fa_start = offset;
4647 			new->fa_end = end;
4648 			fp->f_advice = new;
4649 			new = fa;
4650 		}
4651 		mtx_pool_unlock(mtxpool_sleep, fp);
4652 		break;
4653 	case POSIX_FADV_NORMAL:
4654 		/*
4655 		 * If a the "normal" region overlaps with an existing
4656 		 * non-standard region, trim or remove the
4657 		 * non-standard region.
4658 		 */
4659 		mtx_pool_lock(mtxpool_sleep, fp);
4660 		fa = fp->f_advice;
4661 		if (fa != NULL) {
4662 			if (offset <= fa->fa_start && end >= fa->fa_end) {
4663 				new = fa;
4664 				fp->f_advice = NULL;
4665 			} else if (offset <= fa->fa_start &&
4666 			    end >= fa->fa_start)
4667 				fa->fa_start = end + 1;
4668 			else if (offset <= fa->fa_end && end >= fa->fa_end)
4669 				fa->fa_end = offset - 1;
4670 			else if (offset >= fa->fa_start && end <= fa->fa_end) {
4671 				/*
4672 				 * If the "normal" region is a middle
4673 				 * portion of the existing
4674 				 * non-standard region, just remove
4675 				 * the whole thing rather than picking
4676 				 * one side or the other to
4677 				 * preserve.
4678 				 */
4679 				new = fa;
4680 				fp->f_advice = NULL;
4681 			}
4682 		}
4683 		mtx_pool_unlock(mtxpool_sleep, fp);
4684 		break;
4685 	case POSIX_FADV_WILLNEED:
4686 	case POSIX_FADV_DONTNEED:
4687 		error = VOP_ADVISE(vp, offset, end, advice);
4688 		break;
4689 	}
4690 out:
4691 	if (fp != NULL)
4692 		fdrop(fp, td);
4693 	free(new, M_FADVISE);
4694 	return (error);
4695 }
4696 
4697 int
4698 sys_posix_fadvise(struct thread *td, struct posix_fadvise_args *uap)
4699 {
4700 	int error;
4701 
4702 	error = kern_posix_fadvise(td, uap->fd, uap->offset, uap->len,
4703 	    uap->advice);
4704 	return (kern_posix_error(td, error));
4705 }
4706