xref: /freebsd/sys/kern/vfs_syscalls.c (revision 23f6875a43f7ce365f2d52cf857da010c47fb03b)
1 /*-
2  * Copyright (c) 1989, 1993
3  *	The Regents of the University of California.  All rights reserved.
4  * (c) UNIX System Laboratories, Inc.
5  * All or some portions of this file are derived from material licensed
6  * to the University of California by American Telephone and Telegraph
7  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
8  * the permission of UNIX System Laboratories, Inc.
9  *
10  * Redistribution and use in source and binary forms, with or without
11  * modification, are permitted provided that the following conditions
12  * are met:
13  * 1. Redistributions of source code must retain the above copyright
14  *    notice, this list of conditions and the following disclaimer.
15  * 2. Redistributions in binary form must reproduce the above copyright
16  *    notice, this list of conditions and the following disclaimer in the
17  *    documentation and/or other materials provided with the distribution.
18  * 3. Neither the name of the University nor the names of its contributors
19  *    may be used to endorse or promote products derived from this software
20  *    without specific prior written permission.
21  *
22  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
23  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
24  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
25  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
26  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
27  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
28  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
29  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
30  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
31  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32  * SUCH DAMAGE.
33  *
34  *	@(#)vfs_syscalls.c	8.13 (Berkeley) 4/15/94
35  */
36 
37 #include <sys/cdefs.h>
38 __FBSDID("$FreeBSD$");
39 
40 #include "opt_capsicum.h"
41 #include "opt_compat.h"
42 #include "opt_ktrace.h"
43 
44 #include <sys/param.h>
45 #include <sys/systm.h>
46 #include <sys/bio.h>
47 #include <sys/buf.h>
48 #include <sys/capsicum.h>
49 #include <sys/disk.h>
50 #include <sys/sysent.h>
51 #include <sys/malloc.h>
52 #include <sys/mount.h>
53 #include <sys/mutex.h>
54 #include <sys/sysproto.h>
55 #include <sys/namei.h>
56 #include <sys/filedesc.h>
57 #include <sys/kernel.h>
58 #include <sys/fcntl.h>
59 #include <sys/file.h>
60 #include <sys/filio.h>
61 #include <sys/limits.h>
62 #include <sys/linker.h>
63 #include <sys/rwlock.h>
64 #include <sys/sdt.h>
65 #include <sys/stat.h>
66 #include <sys/sx.h>
67 #include <sys/unistd.h>
68 #include <sys/vnode.h>
69 #include <sys/priv.h>
70 #include <sys/proc.h>
71 #include <sys/dirent.h>
72 #include <sys/jail.h>
73 #include <sys/syscallsubr.h>
74 #include <sys/sysctl.h>
75 #ifdef KTRACE
76 #include <sys/ktrace.h>
77 #endif
78 
79 #include <machine/stdarg.h>
80 
81 #include <security/audit/audit.h>
82 #include <security/mac/mac_framework.h>
83 
84 #include <vm/vm.h>
85 #include <vm/vm_object.h>
86 #include <vm/vm_page.h>
87 #include <vm/uma.h>
88 
89 #include <ufs/ufs/quota.h>
90 
91 MALLOC_DEFINE(M_FADVISE, "fadvise", "posix_fadvise(2) information");
92 
93 SDT_PROVIDER_DEFINE(vfs);
94 SDT_PROBE_DEFINE2(vfs, , stat, mode, "char *", "int");
95 SDT_PROBE_DEFINE2(vfs, , stat, reg, "char *", "int");
96 
97 static int kern_chflagsat(struct thread *td, int fd, const char *path,
98     enum uio_seg pathseg, u_long flags, int atflag);
99 static int setfflags(struct thread *td, struct vnode *, u_long);
100 static int getutimes(const struct timeval *, enum uio_seg, struct timespec *);
101 static int getutimens(const struct timespec *, enum uio_seg,
102     struct timespec *, int *);
103 static int setutimes(struct thread *td, struct vnode *,
104     const struct timespec *, int, int);
105 static int vn_access(struct vnode *vp, int user_flags, struct ucred *cred,
106     struct thread *td);
107 
108 /*
109  * Sync each mounted filesystem.
110  */
111 #ifndef _SYS_SYSPROTO_H_
112 struct sync_args {
113 	int     dummy;
114 };
115 #endif
116 /* ARGSUSED */
117 int
118 sys_sync(td, uap)
119 	struct thread *td;
120 	struct sync_args *uap;
121 {
122 	struct mount *mp, *nmp;
123 	int save;
124 
125 	mtx_lock(&mountlist_mtx);
126 	for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
127 		if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK)) {
128 			nmp = TAILQ_NEXT(mp, mnt_list);
129 			continue;
130 		}
131 		if ((mp->mnt_flag & MNT_RDONLY) == 0 &&
132 		    vn_start_write(NULL, &mp, V_NOWAIT) == 0) {
133 			save = curthread_pflags_set(TDP_SYNCIO);
134 			vfs_msync(mp, MNT_NOWAIT);
135 			VFS_SYNC(mp, MNT_NOWAIT);
136 			curthread_pflags_restore(save);
137 			vn_finished_write(mp);
138 		}
139 		mtx_lock(&mountlist_mtx);
140 		nmp = TAILQ_NEXT(mp, mnt_list);
141 		vfs_unbusy(mp);
142 	}
143 	mtx_unlock(&mountlist_mtx);
144 	return (0);
145 }
146 
147 /*
148  * Change filesystem quotas.
149  */
150 #ifndef _SYS_SYSPROTO_H_
151 struct quotactl_args {
152 	char *path;
153 	int cmd;
154 	int uid;
155 	caddr_t arg;
156 };
157 #endif
158 int
159 sys_quotactl(td, uap)
160 	struct thread *td;
161 	register struct quotactl_args /* {
162 		char *path;
163 		int cmd;
164 		int uid;
165 		caddr_t arg;
166 	} */ *uap;
167 {
168 	struct mount *mp;
169 	struct nameidata nd;
170 	int error;
171 
172 	AUDIT_ARG_CMD(uap->cmd);
173 	AUDIT_ARG_UID(uap->uid);
174 	if (!prison_allow(td->td_ucred, PR_ALLOW_QUOTAS))
175 		return (EPERM);
176 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
177 	    uap->path, td);
178 	if ((error = namei(&nd)) != 0)
179 		return (error);
180 	NDFREE(&nd, NDF_ONLY_PNBUF);
181 	mp = nd.ni_vp->v_mount;
182 	vfs_ref(mp);
183 	vput(nd.ni_vp);
184 	error = vfs_busy(mp, 0);
185 	vfs_rel(mp);
186 	if (error != 0)
187 		return (error);
188 	error = VFS_QUOTACTL(mp, uap->cmd, uap->uid, uap->arg);
189 
190 	/*
191 	 * Since quota on operation typically needs to open quota
192 	 * file, the Q_QUOTAON handler needs to unbusy the mount point
193 	 * before calling into namei.  Otherwise, unmount might be
194 	 * started between two vfs_busy() invocations (first is our,
195 	 * second is from mount point cross-walk code in lookup()),
196 	 * causing deadlock.
197 	 *
198 	 * Require that Q_QUOTAON handles the vfs_busy() reference on
199 	 * its own, always returning with ubusied mount point.
200 	 */
201 	if ((uap->cmd >> SUBCMDSHIFT) != Q_QUOTAON)
202 		vfs_unbusy(mp);
203 	return (error);
204 }
205 
206 /*
207  * Used by statfs conversion routines to scale the block size up if
208  * necessary so that all of the block counts are <= 'max_size'.  Note
209  * that 'max_size' should be a bitmask, i.e. 2^n - 1 for some non-zero
210  * value of 'n'.
211  */
212 void
213 statfs_scale_blocks(struct statfs *sf, long max_size)
214 {
215 	uint64_t count;
216 	int shift;
217 
218 	KASSERT(powerof2(max_size + 1), ("%s: invalid max_size", __func__));
219 
220 	/*
221 	 * Attempt to scale the block counts to give a more accurate
222 	 * overview to userland of the ratio of free space to used
223 	 * space.  To do this, find the largest block count and compute
224 	 * a divisor that lets it fit into a signed integer <= max_size.
225 	 */
226 	if (sf->f_bavail < 0)
227 		count = -sf->f_bavail;
228 	else
229 		count = sf->f_bavail;
230 	count = MAX(sf->f_blocks, MAX(sf->f_bfree, count));
231 	if (count <= max_size)
232 		return;
233 
234 	count >>= flsl(max_size);
235 	shift = 0;
236 	while (count > 0) {
237 		shift++;
238 		count >>=1;
239 	}
240 
241 	sf->f_bsize <<= shift;
242 	sf->f_blocks >>= shift;
243 	sf->f_bfree >>= shift;
244 	sf->f_bavail >>= shift;
245 }
246 
247 static int
248 kern_do_statfs(struct thread *td, struct mount *mp, struct statfs *buf)
249 {
250 	struct statfs *sp;
251 	int error;
252 
253 	if (mp == NULL)
254 		return (EBADF);
255 	error = vfs_busy(mp, 0);
256 	vfs_rel(mp);
257 	if (error != 0)
258 		return (error);
259 #ifdef MAC
260 	error = mac_mount_check_stat(td->td_ucred, mp);
261 	if (error != 0)
262 		goto out;
263 #endif
264 	/*
265 	 * Set these in case the underlying filesystem fails to do so.
266 	 */
267 	sp = &mp->mnt_stat;
268 	sp->f_version = STATFS_VERSION;
269 	sp->f_namemax = NAME_MAX;
270 	sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
271 	error = VFS_STATFS(mp, sp);
272 	if (error != 0)
273 		goto out;
274 	*buf = *sp;
275 	if (priv_check(td, PRIV_VFS_GENERATION)) {
276 		buf->f_fsid.val[0] = buf->f_fsid.val[1] = 0;
277 		prison_enforce_statfs(td->td_ucred, mp, buf);
278 	}
279 out:
280 	vfs_unbusy(mp);
281 	return (error);
282 }
283 
284 /*
285  * Get filesystem statistics.
286  */
287 #ifndef _SYS_SYSPROTO_H_
288 struct statfs_args {
289 	char *path;
290 	struct statfs *buf;
291 };
292 #endif
293 int
294 sys_statfs(td, uap)
295 	struct thread *td;
296 	register struct statfs_args /* {
297 		char *path;
298 		struct statfs *buf;
299 	} */ *uap;
300 {
301 	struct statfs *sfp;
302 	int error;
303 
304 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
305 	error = kern_statfs(td, uap->path, UIO_USERSPACE, sfp);
306 	if (error == 0)
307 		error = copyout(sfp, uap->buf, sizeof(struct statfs));
308 	free(sfp, M_STATFS);
309 	return (error);
310 }
311 
312 int
313 kern_statfs(struct thread *td, char *path, enum uio_seg pathseg,
314     struct statfs *buf)
315 {
316 	struct mount *mp;
317 	struct nameidata nd;
318 	int error;
319 
320 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
321 	    pathseg, path, td);
322 	error = namei(&nd);
323 	if (error != 0)
324 		return (error);
325 	mp = nd.ni_vp->v_mount;
326 	vfs_ref(mp);
327 	NDFREE(&nd, NDF_ONLY_PNBUF);
328 	vput(nd.ni_vp);
329 	return (kern_do_statfs(td, mp, buf));
330 }
331 
332 /*
333  * Get filesystem statistics.
334  */
335 #ifndef _SYS_SYSPROTO_H_
336 struct fstatfs_args {
337 	int fd;
338 	struct statfs *buf;
339 };
340 #endif
341 int
342 sys_fstatfs(td, uap)
343 	struct thread *td;
344 	register struct fstatfs_args /* {
345 		int fd;
346 		struct statfs *buf;
347 	} */ *uap;
348 {
349 	struct statfs *sfp;
350 	int error;
351 
352 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
353 	error = kern_fstatfs(td, uap->fd, sfp);
354 	if (error == 0)
355 		error = copyout(sfp, uap->buf, sizeof(struct statfs));
356 	free(sfp, M_STATFS);
357 	return (error);
358 }
359 
360 int
361 kern_fstatfs(struct thread *td, int fd, struct statfs *buf)
362 {
363 	struct file *fp;
364 	struct mount *mp;
365 	struct vnode *vp;
366 	cap_rights_t rights;
367 	int error;
368 
369 	AUDIT_ARG_FD(fd);
370 	error = getvnode(td, fd, cap_rights_init(&rights, CAP_FSTATFS), &fp);
371 	if (error != 0)
372 		return (error);
373 	vp = fp->f_vnode;
374 	vn_lock(vp, LK_SHARED | LK_RETRY);
375 #ifdef AUDIT
376 	AUDIT_ARG_VNODE1(vp);
377 #endif
378 	mp = vp->v_mount;
379 	if (mp != NULL)
380 		vfs_ref(mp);
381 	VOP_UNLOCK(vp, 0);
382 	fdrop(fp, td);
383 	return (kern_do_statfs(td, mp, buf));
384 }
385 
386 /*
387  * Get statistics on all filesystems.
388  */
389 #ifndef _SYS_SYSPROTO_H_
390 struct getfsstat_args {
391 	struct statfs *buf;
392 	long bufsize;
393 	int mode;
394 };
395 #endif
396 int
397 sys_getfsstat(td, uap)
398 	struct thread *td;
399 	register struct getfsstat_args /* {
400 		struct statfs *buf;
401 		long bufsize;
402 		int mode;
403 	} */ *uap;
404 {
405 	size_t count;
406 	int error;
407 
408 	if (uap->bufsize < 0 || uap->bufsize > SIZE_MAX)
409 		return (EINVAL);
410 	error = kern_getfsstat(td, &uap->buf, uap->bufsize, &count,
411 	    UIO_USERSPACE, uap->mode);
412 	if (error == 0)
413 		td->td_retval[0] = count;
414 	return (error);
415 }
416 
417 /*
418  * If (bufsize > 0 && bufseg == UIO_SYSSPACE)
419  *	The caller is responsible for freeing memory which will be allocated
420  *	in '*buf'.
421  */
422 int
423 kern_getfsstat(struct thread *td, struct statfs **buf, size_t bufsize,
424     size_t *countp, enum uio_seg bufseg, int mode)
425 {
426 	struct mount *mp, *nmp;
427 	struct statfs *sfsp, *sp, *sptmp, *tofree;
428 	size_t count, maxcount;
429 	int error;
430 
431 	switch (mode) {
432 	case MNT_WAIT:
433 	case MNT_NOWAIT:
434 		break;
435 	default:
436 		if (bufseg == UIO_SYSSPACE)
437 			*buf = NULL;
438 		return (EINVAL);
439 	}
440 restart:
441 	maxcount = bufsize / sizeof(struct statfs);
442 	if (bufsize == 0) {
443 		sfsp = NULL;
444 		tofree = NULL;
445 	} else if (bufseg == UIO_USERSPACE) {
446 		sfsp = *buf;
447 		tofree = NULL;
448 	} else /* if (bufseg == UIO_SYSSPACE) */ {
449 		count = 0;
450 		mtx_lock(&mountlist_mtx);
451 		TAILQ_FOREACH(mp, &mountlist, mnt_list) {
452 			count++;
453 		}
454 		mtx_unlock(&mountlist_mtx);
455 		if (maxcount > count)
456 			maxcount = count;
457 		tofree = sfsp = *buf = malloc(maxcount * sizeof(struct statfs),
458 		    M_STATFS, M_WAITOK);
459 	}
460 	count = 0;
461 	mtx_lock(&mountlist_mtx);
462 	for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
463 		if (prison_canseemount(td->td_ucred, mp) != 0) {
464 			nmp = TAILQ_NEXT(mp, mnt_list);
465 			continue;
466 		}
467 #ifdef MAC
468 		if (mac_mount_check_stat(td->td_ucred, mp) != 0) {
469 			nmp = TAILQ_NEXT(mp, mnt_list);
470 			continue;
471 		}
472 #endif
473 		if (mode == MNT_WAIT) {
474 			if (vfs_busy(mp, MBF_MNTLSTLOCK) != 0) {
475 				/*
476 				 * If vfs_busy() failed, and MBF_NOWAIT
477 				 * wasn't passed, then the mp is gone.
478 				 * Furthermore, because of MBF_MNTLSTLOCK,
479 				 * the mountlist_mtx was dropped.  We have
480 				 * no other choice than to start over.
481 				 */
482 				mtx_unlock(&mountlist_mtx);
483 				free(tofree, M_STATFS);
484 				goto restart;
485 			}
486 		} else {
487 			if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK) != 0) {
488 				nmp = TAILQ_NEXT(mp, mnt_list);
489 				continue;
490 			}
491 		}
492 		if (sfsp != NULL && count < maxcount) {
493 			sp = &mp->mnt_stat;
494 			/*
495 			 * Set these in case the underlying filesystem
496 			 * fails to do so.
497 			 */
498 			sp->f_version = STATFS_VERSION;
499 			sp->f_namemax = NAME_MAX;
500 			sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
501 			/*
502 			 * If MNT_NOWAIT is specified, do not refresh
503 			 * the fsstat cache.
504 			 */
505 			if (mode != MNT_NOWAIT) {
506 				error = VFS_STATFS(mp, sp);
507 				if (error != 0) {
508 					mtx_lock(&mountlist_mtx);
509 					nmp = TAILQ_NEXT(mp, mnt_list);
510 					vfs_unbusy(mp);
511 					continue;
512 				}
513 			}
514 			if (priv_check(td, PRIV_VFS_GENERATION)) {
515 				sptmp = malloc(sizeof(struct statfs), M_STATFS,
516 				    M_WAITOK);
517 				*sptmp = *sp;
518 				sptmp->f_fsid.val[0] = sptmp->f_fsid.val[1] = 0;
519 				prison_enforce_statfs(td->td_ucred, mp, sptmp);
520 				sp = sptmp;
521 			} else
522 				sptmp = NULL;
523 			if (bufseg == UIO_SYSSPACE) {
524 				bcopy(sp, sfsp, sizeof(*sp));
525 				free(sptmp, M_STATFS);
526 			} else /* if (bufseg == UIO_USERSPACE) */ {
527 				error = copyout(sp, sfsp, sizeof(*sp));
528 				free(sptmp, M_STATFS);
529 				if (error != 0) {
530 					vfs_unbusy(mp);
531 					return (error);
532 				}
533 			}
534 			sfsp++;
535 		}
536 		count++;
537 		mtx_lock(&mountlist_mtx);
538 		nmp = TAILQ_NEXT(mp, mnt_list);
539 		vfs_unbusy(mp);
540 	}
541 	mtx_unlock(&mountlist_mtx);
542 	if (sfsp != NULL && count > maxcount)
543 		*countp = maxcount;
544 	else
545 		*countp = count;
546 	return (0);
547 }
548 
549 #ifdef COMPAT_FREEBSD4
550 /*
551  * Get old format filesystem statistics.
552  */
553 static void cvtstatfs(struct statfs *, struct ostatfs *);
554 
555 #ifndef _SYS_SYSPROTO_H_
556 struct freebsd4_statfs_args {
557 	char *path;
558 	struct ostatfs *buf;
559 };
560 #endif
561 int
562 freebsd4_statfs(td, uap)
563 	struct thread *td;
564 	struct freebsd4_statfs_args /* {
565 		char *path;
566 		struct ostatfs *buf;
567 	} */ *uap;
568 {
569 	struct ostatfs osb;
570 	struct statfs *sfp;
571 	int error;
572 
573 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
574 	error = kern_statfs(td, uap->path, UIO_USERSPACE, sfp);
575 	if (error == 0) {
576 		cvtstatfs(sfp, &osb);
577 		error = copyout(&osb, uap->buf, sizeof(osb));
578 	}
579 	free(sfp, M_STATFS);
580 	return (error);
581 }
582 
583 /*
584  * Get filesystem statistics.
585  */
586 #ifndef _SYS_SYSPROTO_H_
587 struct freebsd4_fstatfs_args {
588 	int fd;
589 	struct ostatfs *buf;
590 };
591 #endif
592 int
593 freebsd4_fstatfs(td, uap)
594 	struct thread *td;
595 	struct freebsd4_fstatfs_args /* {
596 		int fd;
597 		struct ostatfs *buf;
598 	} */ *uap;
599 {
600 	struct ostatfs osb;
601 	struct statfs *sfp;
602 	int error;
603 
604 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
605 	error = kern_fstatfs(td, uap->fd, sfp);
606 	if (error == 0) {
607 		cvtstatfs(sfp, &osb);
608 		error = copyout(&osb, uap->buf, sizeof(osb));
609 	}
610 	free(sfp, M_STATFS);
611 	return (error);
612 }
613 
614 /*
615  * Get statistics on all filesystems.
616  */
617 #ifndef _SYS_SYSPROTO_H_
618 struct freebsd4_getfsstat_args {
619 	struct ostatfs *buf;
620 	long bufsize;
621 	int mode;
622 };
623 #endif
624 int
625 freebsd4_getfsstat(td, uap)
626 	struct thread *td;
627 	register struct freebsd4_getfsstat_args /* {
628 		struct ostatfs *buf;
629 		long bufsize;
630 		int mode;
631 	} */ *uap;
632 {
633 	struct statfs *buf, *sp;
634 	struct ostatfs osb;
635 	size_t count, size;
636 	int error;
637 
638 	if (uap->bufsize < 0)
639 		return (EINVAL);
640 	count = uap->bufsize / sizeof(struct ostatfs);
641 	if (count > SIZE_MAX / sizeof(struct statfs))
642 		return (EINVAL);
643 	size = count * sizeof(struct statfs);
644 	error = kern_getfsstat(td, &buf, size, &count, UIO_SYSSPACE,
645 	    uap->mode);
646 	td->td_retval[0] = count;
647 	if (size != 0) {
648 		sp = buf;
649 		while (count != 0 && error == 0) {
650 			cvtstatfs(sp, &osb);
651 			error = copyout(&osb, uap->buf, sizeof(osb));
652 			sp++;
653 			uap->buf++;
654 			count--;
655 		}
656 		free(buf, M_STATFS);
657 	}
658 	return (error);
659 }
660 
661 /*
662  * Implement fstatfs() for (NFS) file handles.
663  */
664 #ifndef _SYS_SYSPROTO_H_
665 struct freebsd4_fhstatfs_args {
666 	struct fhandle *u_fhp;
667 	struct ostatfs *buf;
668 };
669 #endif
670 int
671 freebsd4_fhstatfs(td, uap)
672 	struct thread *td;
673 	struct freebsd4_fhstatfs_args /* {
674 		struct fhandle *u_fhp;
675 		struct ostatfs *buf;
676 	} */ *uap;
677 {
678 	struct ostatfs osb;
679 	struct statfs *sfp;
680 	fhandle_t fh;
681 	int error;
682 
683 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
684 	if (error != 0)
685 		return (error);
686 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
687 	error = kern_fhstatfs(td, fh, sfp);
688 	if (error == 0) {
689 		cvtstatfs(sfp, &osb);
690 		error = copyout(&osb, uap->buf, sizeof(osb));
691 	}
692 	free(sfp, M_STATFS);
693 	return (error);
694 }
695 
696 /*
697  * Convert a new format statfs structure to an old format statfs structure.
698  */
699 static void
700 cvtstatfs(nsp, osp)
701 	struct statfs *nsp;
702 	struct ostatfs *osp;
703 {
704 
705 	statfs_scale_blocks(nsp, LONG_MAX);
706 	bzero(osp, sizeof(*osp));
707 	osp->f_bsize = nsp->f_bsize;
708 	osp->f_iosize = MIN(nsp->f_iosize, LONG_MAX);
709 	osp->f_blocks = nsp->f_blocks;
710 	osp->f_bfree = nsp->f_bfree;
711 	osp->f_bavail = nsp->f_bavail;
712 	osp->f_files = MIN(nsp->f_files, LONG_MAX);
713 	osp->f_ffree = MIN(nsp->f_ffree, LONG_MAX);
714 	osp->f_owner = nsp->f_owner;
715 	osp->f_type = nsp->f_type;
716 	osp->f_flags = nsp->f_flags;
717 	osp->f_syncwrites = MIN(nsp->f_syncwrites, LONG_MAX);
718 	osp->f_asyncwrites = MIN(nsp->f_asyncwrites, LONG_MAX);
719 	osp->f_syncreads = MIN(nsp->f_syncreads, LONG_MAX);
720 	osp->f_asyncreads = MIN(nsp->f_asyncreads, LONG_MAX);
721 	strlcpy(osp->f_fstypename, nsp->f_fstypename,
722 	    MIN(MFSNAMELEN, OMFSNAMELEN));
723 	strlcpy(osp->f_mntonname, nsp->f_mntonname,
724 	    MIN(MNAMELEN, OMNAMELEN));
725 	strlcpy(osp->f_mntfromname, nsp->f_mntfromname,
726 	    MIN(MNAMELEN, OMNAMELEN));
727 	osp->f_fsid = nsp->f_fsid;
728 }
729 #endif /* COMPAT_FREEBSD4 */
730 
731 /*
732  * Change current working directory to a given file descriptor.
733  */
734 #ifndef _SYS_SYSPROTO_H_
735 struct fchdir_args {
736 	int	fd;
737 };
738 #endif
739 int
740 sys_fchdir(td, uap)
741 	struct thread *td;
742 	struct fchdir_args /* {
743 		int fd;
744 	} */ *uap;
745 {
746 	struct vnode *vp, *tdp;
747 	struct mount *mp;
748 	struct file *fp;
749 	cap_rights_t rights;
750 	int error;
751 
752 	AUDIT_ARG_FD(uap->fd);
753 	error = getvnode(td, uap->fd, cap_rights_init(&rights, CAP_FCHDIR),
754 	    &fp);
755 	if (error != 0)
756 		return (error);
757 	vp = fp->f_vnode;
758 	vrefact(vp);
759 	fdrop(fp, td);
760 	vn_lock(vp, LK_SHARED | LK_RETRY);
761 	AUDIT_ARG_VNODE1(vp);
762 	error = change_dir(vp, td);
763 	while (!error && (mp = vp->v_mountedhere) != NULL) {
764 		if (vfs_busy(mp, 0))
765 			continue;
766 		error = VFS_ROOT(mp, LK_SHARED, &tdp);
767 		vfs_unbusy(mp);
768 		if (error != 0)
769 			break;
770 		vput(vp);
771 		vp = tdp;
772 	}
773 	if (error != 0) {
774 		vput(vp);
775 		return (error);
776 	}
777 	VOP_UNLOCK(vp, 0);
778 	pwd_chdir(td, vp);
779 	return (0);
780 }
781 
782 /*
783  * Change current working directory (``.'').
784  */
785 #ifndef _SYS_SYSPROTO_H_
786 struct chdir_args {
787 	char	*path;
788 };
789 #endif
790 int
791 sys_chdir(td, uap)
792 	struct thread *td;
793 	struct chdir_args /* {
794 		char *path;
795 	} */ *uap;
796 {
797 
798 	return (kern_chdir(td, uap->path, UIO_USERSPACE));
799 }
800 
801 int
802 kern_chdir(struct thread *td, char *path, enum uio_seg pathseg)
803 {
804 	struct nameidata nd;
805 	int error;
806 
807 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
808 	    pathseg, path, td);
809 	if ((error = namei(&nd)) != 0)
810 		return (error);
811 	if ((error = change_dir(nd.ni_vp, td)) != 0) {
812 		vput(nd.ni_vp);
813 		NDFREE(&nd, NDF_ONLY_PNBUF);
814 		return (error);
815 	}
816 	VOP_UNLOCK(nd.ni_vp, 0);
817 	NDFREE(&nd, NDF_ONLY_PNBUF);
818 	pwd_chdir(td, nd.ni_vp);
819 	return (0);
820 }
821 
822 /*
823  * Change notion of root (``/'') directory.
824  */
825 #ifndef _SYS_SYSPROTO_H_
826 struct chroot_args {
827 	char	*path;
828 };
829 #endif
830 int
831 sys_chroot(td, uap)
832 	struct thread *td;
833 	struct chroot_args /* {
834 		char *path;
835 	} */ *uap;
836 {
837 	struct nameidata nd;
838 	int error;
839 
840 	error = priv_check(td, PRIV_VFS_CHROOT);
841 	if (error != 0)
842 		return (error);
843 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
844 	    UIO_USERSPACE, uap->path, td);
845 	error = namei(&nd);
846 	if (error != 0)
847 		goto error;
848 	error = change_dir(nd.ni_vp, td);
849 	if (error != 0)
850 		goto e_vunlock;
851 #ifdef MAC
852 	error = mac_vnode_check_chroot(td->td_ucred, nd.ni_vp);
853 	if (error != 0)
854 		goto e_vunlock;
855 #endif
856 	VOP_UNLOCK(nd.ni_vp, 0);
857 	error = pwd_chroot(td, nd.ni_vp);
858 	vrele(nd.ni_vp);
859 	NDFREE(&nd, NDF_ONLY_PNBUF);
860 	return (error);
861 e_vunlock:
862 	vput(nd.ni_vp);
863 error:
864 	NDFREE(&nd, NDF_ONLY_PNBUF);
865 	return (error);
866 }
867 
868 /*
869  * Common routine for chroot and chdir.  Callers must provide a locked vnode
870  * instance.
871  */
872 int
873 change_dir(vp, td)
874 	struct vnode *vp;
875 	struct thread *td;
876 {
877 #ifdef MAC
878 	int error;
879 #endif
880 
881 	ASSERT_VOP_LOCKED(vp, "change_dir(): vp not locked");
882 	if (vp->v_type != VDIR)
883 		return (ENOTDIR);
884 #ifdef MAC
885 	error = mac_vnode_check_chdir(td->td_ucred, vp);
886 	if (error != 0)
887 		return (error);
888 #endif
889 	return (VOP_ACCESS(vp, VEXEC, td->td_ucred, td));
890 }
891 
892 static __inline void
893 flags_to_rights(int flags, cap_rights_t *rightsp)
894 {
895 
896 	if (flags & O_EXEC) {
897 		cap_rights_set(rightsp, CAP_FEXECVE);
898 	} else {
899 		switch ((flags & O_ACCMODE)) {
900 		case O_RDONLY:
901 			cap_rights_set(rightsp, CAP_READ);
902 			break;
903 		case O_RDWR:
904 			cap_rights_set(rightsp, CAP_READ);
905 			/* FALLTHROUGH */
906 		case O_WRONLY:
907 			cap_rights_set(rightsp, CAP_WRITE);
908 			if (!(flags & (O_APPEND | O_TRUNC)))
909 				cap_rights_set(rightsp, CAP_SEEK);
910 			break;
911 		}
912 	}
913 
914 	if (flags & O_CREAT)
915 		cap_rights_set(rightsp, CAP_CREATE);
916 
917 	if (flags & O_TRUNC)
918 		cap_rights_set(rightsp, CAP_FTRUNCATE);
919 
920 	if (flags & (O_SYNC | O_FSYNC))
921 		cap_rights_set(rightsp, CAP_FSYNC);
922 
923 	if (flags & (O_EXLOCK | O_SHLOCK))
924 		cap_rights_set(rightsp, CAP_FLOCK);
925 }
926 
927 /*
928  * Check permissions, allocate an open file structure, and call the device
929  * open routine if any.
930  */
931 #ifndef _SYS_SYSPROTO_H_
932 struct open_args {
933 	char	*path;
934 	int	flags;
935 	int	mode;
936 };
937 #endif
938 int
939 sys_open(td, uap)
940 	struct thread *td;
941 	register struct open_args /* {
942 		char *path;
943 		int flags;
944 		int mode;
945 	} */ *uap;
946 {
947 
948 	return (kern_openat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
949 	    uap->flags, uap->mode));
950 }
951 
952 #ifndef _SYS_SYSPROTO_H_
953 struct openat_args {
954 	int	fd;
955 	char	*path;
956 	int	flag;
957 	int	mode;
958 };
959 #endif
960 int
961 sys_openat(struct thread *td, struct openat_args *uap)
962 {
963 
964 	AUDIT_ARG_FD(uap->fd);
965 	return (kern_openat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
966 	    uap->mode));
967 }
968 
969 int
970 kern_openat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
971     int flags, int mode)
972 {
973 	struct proc *p = td->td_proc;
974 	struct filedesc *fdp = p->p_fd;
975 	struct file *fp;
976 	struct vnode *vp;
977 	struct nameidata nd;
978 	cap_rights_t rights;
979 	int cmode, error, indx;
980 
981 	indx = -1;
982 
983 	AUDIT_ARG_FFLAGS(flags);
984 	AUDIT_ARG_MODE(mode);
985 	cap_rights_init(&rights, CAP_LOOKUP);
986 	flags_to_rights(flags, &rights);
987 	/*
988 	 * Only one of the O_EXEC, O_RDONLY, O_WRONLY and O_RDWR flags
989 	 * may be specified.
990 	 */
991 	if (flags & O_EXEC) {
992 		if (flags & O_ACCMODE)
993 			return (EINVAL);
994 	} else if ((flags & O_ACCMODE) == O_ACCMODE) {
995 		return (EINVAL);
996 	} else {
997 		flags = FFLAGS(flags);
998 	}
999 
1000 	/*
1001 	 * Allocate a file structure. The descriptor to reference it
1002 	 * is allocated and set by finstall() below.
1003 	 */
1004 	error = falloc_noinstall(td, &fp);
1005 	if (error != 0)
1006 		return (error);
1007 	/*
1008 	 * An extra reference on `fp' has been held for us by
1009 	 * falloc_noinstall().
1010 	 */
1011 	/* Set the flags early so the finit in devfs can pick them up. */
1012 	fp->f_flag = flags & FMASK;
1013 	cmode = ((mode & ~fdp->fd_cmask) & ALLPERMS) & ~S_ISTXT;
1014 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, fd,
1015 	    &rights, td);
1016 	td->td_dupfd = -1;		/* XXX check for fdopen */
1017 	error = vn_open(&nd, &flags, cmode, fp);
1018 	if (error != 0) {
1019 		/*
1020 		 * If the vn_open replaced the method vector, something
1021 		 * wonderous happened deep below and we just pass it up
1022 		 * pretending we know what we do.
1023 		 */
1024 		if (error == ENXIO && fp->f_ops != &badfileops)
1025 			goto success;
1026 
1027 		/*
1028 		 * Handle special fdopen() case. bleh.
1029 		 *
1030 		 * Don't do this for relative (capability) lookups; we don't
1031 		 * understand exactly what would happen, and we don't think
1032 		 * that it ever should.
1033 		 */
1034 		if ((nd.ni_lcf & NI_LCF_STRICTRELATIVE) == 0 &&
1035 		    (error == ENODEV || error == ENXIO) &&
1036 		    td->td_dupfd >= 0) {
1037 			error = dupfdopen(td, fdp, td->td_dupfd, flags, error,
1038 			    &indx);
1039 			if (error == 0)
1040 				goto success;
1041 		}
1042 
1043 		goto bad;
1044 	}
1045 	td->td_dupfd = 0;
1046 	NDFREE(&nd, NDF_ONLY_PNBUF);
1047 	vp = nd.ni_vp;
1048 
1049 	/*
1050 	 * Store the vnode, for any f_type. Typically, the vnode use
1051 	 * count is decremented by direct call to vn_closefile() for
1052 	 * files that switched type in the cdevsw fdopen() method.
1053 	 */
1054 	fp->f_vnode = vp;
1055 	/*
1056 	 * If the file wasn't claimed by devfs bind it to the normal
1057 	 * vnode operations here.
1058 	 */
1059 	if (fp->f_ops == &badfileops) {
1060 		KASSERT(vp->v_type != VFIFO, ("Unexpected fifo."));
1061 		fp->f_seqcount = 1;
1062 		finit(fp, (flags & FMASK) | (fp->f_flag & FHASLOCK),
1063 		    DTYPE_VNODE, vp, &vnops);
1064 	}
1065 
1066 	VOP_UNLOCK(vp, 0);
1067 	if (flags & O_TRUNC) {
1068 		error = fo_truncate(fp, 0, td->td_ucred, td);
1069 		if (error != 0)
1070 			goto bad;
1071 	}
1072 success:
1073 	/*
1074 	 * If we haven't already installed the FD (for dupfdopen), do so now.
1075 	 */
1076 	if (indx == -1) {
1077 		struct filecaps *fcaps;
1078 
1079 #ifdef CAPABILITIES
1080 		if ((nd.ni_lcf & NI_LCF_STRICTRELATIVE) != 0)
1081 			fcaps = &nd.ni_filecaps;
1082 		else
1083 #endif
1084 			fcaps = NULL;
1085 		error = finstall(td, fp, &indx, flags, fcaps);
1086 		/* On success finstall() consumes fcaps. */
1087 		if (error != 0) {
1088 			filecaps_free(&nd.ni_filecaps);
1089 			goto bad;
1090 		}
1091 	} else {
1092 		filecaps_free(&nd.ni_filecaps);
1093 	}
1094 
1095 	/*
1096 	 * Release our private reference, leaving the one associated with
1097 	 * the descriptor table intact.
1098 	 */
1099 	fdrop(fp, td);
1100 	td->td_retval[0] = indx;
1101 	return (0);
1102 bad:
1103 	KASSERT(indx == -1, ("indx=%d, should be -1", indx));
1104 	fdrop(fp, td);
1105 	return (error);
1106 }
1107 
1108 #ifdef COMPAT_43
1109 /*
1110  * Create a file.
1111  */
1112 #ifndef _SYS_SYSPROTO_H_
1113 struct ocreat_args {
1114 	char	*path;
1115 	int	mode;
1116 };
1117 #endif
1118 int
1119 ocreat(td, uap)
1120 	struct thread *td;
1121 	register struct ocreat_args /* {
1122 		char *path;
1123 		int mode;
1124 	} */ *uap;
1125 {
1126 
1127 	return (kern_openat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1128 	    O_WRONLY | O_CREAT | O_TRUNC, uap->mode));
1129 }
1130 #endif /* COMPAT_43 */
1131 
1132 /*
1133  * Create a special file.
1134  */
1135 #ifndef _SYS_SYSPROTO_H_
1136 struct mknod_args {
1137 	char	*path;
1138 	int	mode;
1139 	int	dev;
1140 };
1141 #endif
1142 int
1143 sys_mknod(td, uap)
1144 	struct thread *td;
1145 	register struct mknod_args /* {
1146 		char *path;
1147 		int mode;
1148 		int dev;
1149 	} */ *uap;
1150 {
1151 
1152 	return (kern_mknodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1153 	    uap->mode, uap->dev));
1154 }
1155 
1156 #ifndef _SYS_SYSPROTO_H_
1157 struct mknodat_args {
1158 	int	fd;
1159 	char	*path;
1160 	mode_t	mode;
1161 	dev_t	dev;
1162 };
1163 #endif
1164 int
1165 sys_mknodat(struct thread *td, struct mknodat_args *uap)
1166 {
1167 
1168 	return (kern_mknodat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode,
1169 	    uap->dev));
1170 }
1171 
1172 int
1173 kern_mknodat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1174     int mode, int dev)
1175 {
1176 	struct vnode *vp;
1177 	struct mount *mp;
1178 	struct vattr vattr;
1179 	struct nameidata nd;
1180 	cap_rights_t rights;
1181 	int error, whiteout = 0;
1182 
1183 	AUDIT_ARG_MODE(mode);
1184 	AUDIT_ARG_DEV(dev);
1185 	switch (mode & S_IFMT) {
1186 	case S_IFCHR:
1187 	case S_IFBLK:
1188 		error = priv_check(td, PRIV_VFS_MKNOD_DEV);
1189 		if (error == 0 && dev == VNOVAL)
1190 			error = EINVAL;
1191 		break;
1192 	case S_IFMT:
1193 		error = priv_check(td, PRIV_VFS_MKNOD_BAD);
1194 		break;
1195 	case S_IFWHT:
1196 		error = priv_check(td, PRIV_VFS_MKNOD_WHT);
1197 		break;
1198 	case S_IFIFO:
1199 		if (dev == 0)
1200 			return (kern_mkfifoat(td, fd, path, pathseg, mode));
1201 		/* FALLTHROUGH */
1202 	default:
1203 		error = EINVAL;
1204 		break;
1205 	}
1206 	if (error != 0)
1207 		return (error);
1208 restart:
1209 	bwillwrite();
1210 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
1211 	    NOCACHE, pathseg, path, fd, cap_rights_init(&rights, CAP_MKNODAT),
1212 	    td);
1213 	if ((error = namei(&nd)) != 0)
1214 		return (error);
1215 	vp = nd.ni_vp;
1216 	if (vp != NULL) {
1217 		NDFREE(&nd, NDF_ONLY_PNBUF);
1218 		if (vp == nd.ni_dvp)
1219 			vrele(nd.ni_dvp);
1220 		else
1221 			vput(nd.ni_dvp);
1222 		vrele(vp);
1223 		return (EEXIST);
1224 	} else {
1225 		VATTR_NULL(&vattr);
1226 		vattr.va_mode = (mode & ALLPERMS) &
1227 		    ~td->td_proc->p_fd->fd_cmask;
1228 		vattr.va_rdev = dev;
1229 		whiteout = 0;
1230 
1231 		switch (mode & S_IFMT) {
1232 		case S_IFMT:	/* used by badsect to flag bad sectors */
1233 			vattr.va_type = VBAD;
1234 			break;
1235 		case S_IFCHR:
1236 			vattr.va_type = VCHR;
1237 			break;
1238 		case S_IFBLK:
1239 			vattr.va_type = VBLK;
1240 			break;
1241 		case S_IFWHT:
1242 			whiteout = 1;
1243 			break;
1244 		default:
1245 			panic("kern_mknod: invalid mode");
1246 		}
1247 	}
1248 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1249 		NDFREE(&nd, NDF_ONLY_PNBUF);
1250 		vput(nd.ni_dvp);
1251 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1252 			return (error);
1253 		goto restart;
1254 	}
1255 #ifdef MAC
1256 	if (error == 0 && !whiteout)
1257 		error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp,
1258 		    &nd.ni_cnd, &vattr);
1259 #endif
1260 	if (error == 0) {
1261 		if (whiteout)
1262 			error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE);
1263 		else {
1264 			error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp,
1265 						&nd.ni_cnd, &vattr);
1266 			if (error == 0)
1267 				vput(nd.ni_vp);
1268 		}
1269 	}
1270 	NDFREE(&nd, NDF_ONLY_PNBUF);
1271 	vput(nd.ni_dvp);
1272 	vn_finished_write(mp);
1273 	return (error);
1274 }
1275 
1276 /*
1277  * Create a named pipe.
1278  */
1279 #ifndef _SYS_SYSPROTO_H_
1280 struct mkfifo_args {
1281 	char	*path;
1282 	int	mode;
1283 };
1284 #endif
1285 int
1286 sys_mkfifo(td, uap)
1287 	struct thread *td;
1288 	register struct mkfifo_args /* {
1289 		char *path;
1290 		int mode;
1291 	} */ *uap;
1292 {
1293 
1294 	return (kern_mkfifoat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1295 	    uap->mode));
1296 }
1297 
1298 #ifndef _SYS_SYSPROTO_H_
1299 struct mkfifoat_args {
1300 	int	fd;
1301 	char	*path;
1302 	mode_t	mode;
1303 };
1304 #endif
1305 int
1306 sys_mkfifoat(struct thread *td, struct mkfifoat_args *uap)
1307 {
1308 
1309 	return (kern_mkfifoat(td, uap->fd, uap->path, UIO_USERSPACE,
1310 	    uap->mode));
1311 }
1312 
1313 int
1314 kern_mkfifoat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1315     int mode)
1316 {
1317 	struct mount *mp;
1318 	struct vattr vattr;
1319 	struct nameidata nd;
1320 	cap_rights_t rights;
1321 	int error;
1322 
1323 	AUDIT_ARG_MODE(mode);
1324 restart:
1325 	bwillwrite();
1326 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
1327 	    NOCACHE, pathseg, path, fd, cap_rights_init(&rights, CAP_MKFIFOAT),
1328 	    td);
1329 	if ((error = namei(&nd)) != 0)
1330 		return (error);
1331 	if (nd.ni_vp != NULL) {
1332 		NDFREE(&nd, NDF_ONLY_PNBUF);
1333 		if (nd.ni_vp == nd.ni_dvp)
1334 			vrele(nd.ni_dvp);
1335 		else
1336 			vput(nd.ni_dvp);
1337 		vrele(nd.ni_vp);
1338 		return (EEXIST);
1339 	}
1340 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1341 		NDFREE(&nd, NDF_ONLY_PNBUF);
1342 		vput(nd.ni_dvp);
1343 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1344 			return (error);
1345 		goto restart;
1346 	}
1347 	VATTR_NULL(&vattr);
1348 	vattr.va_type = VFIFO;
1349 	vattr.va_mode = (mode & ALLPERMS) & ~td->td_proc->p_fd->fd_cmask;
1350 #ifdef MAC
1351 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1352 	    &vattr);
1353 	if (error != 0)
1354 		goto out;
1355 #endif
1356 	error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
1357 	if (error == 0)
1358 		vput(nd.ni_vp);
1359 #ifdef MAC
1360 out:
1361 #endif
1362 	vput(nd.ni_dvp);
1363 	vn_finished_write(mp);
1364 	NDFREE(&nd, NDF_ONLY_PNBUF);
1365 	return (error);
1366 }
1367 
1368 /*
1369  * Make a hard file link.
1370  */
1371 #ifndef _SYS_SYSPROTO_H_
1372 struct link_args {
1373 	char	*path;
1374 	char	*link;
1375 };
1376 #endif
1377 int
1378 sys_link(td, uap)
1379 	struct thread *td;
1380 	register struct link_args /* {
1381 		char *path;
1382 		char *link;
1383 	} */ *uap;
1384 {
1385 
1386 	return (kern_linkat(td, AT_FDCWD, AT_FDCWD, uap->path, uap->link,
1387 	    UIO_USERSPACE, FOLLOW));
1388 }
1389 
1390 #ifndef _SYS_SYSPROTO_H_
1391 struct linkat_args {
1392 	int	fd1;
1393 	char	*path1;
1394 	int	fd2;
1395 	char	*path2;
1396 	int	flag;
1397 };
1398 #endif
1399 int
1400 sys_linkat(struct thread *td, struct linkat_args *uap)
1401 {
1402 	int flag;
1403 
1404 	flag = uap->flag;
1405 	if (flag & ~AT_SYMLINK_FOLLOW)
1406 		return (EINVAL);
1407 
1408 	return (kern_linkat(td, uap->fd1, uap->fd2, uap->path1, uap->path2,
1409 	    UIO_USERSPACE, (flag & AT_SYMLINK_FOLLOW) ? FOLLOW : NOFOLLOW));
1410 }
1411 
1412 int hardlink_check_uid = 0;
1413 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_uid, CTLFLAG_RW,
1414     &hardlink_check_uid, 0,
1415     "Unprivileged processes cannot create hard links to files owned by other "
1416     "users");
1417 static int hardlink_check_gid = 0;
1418 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_gid, CTLFLAG_RW,
1419     &hardlink_check_gid, 0,
1420     "Unprivileged processes cannot create hard links to files owned by other "
1421     "groups");
1422 
1423 static int
1424 can_hardlink(struct vnode *vp, struct ucred *cred)
1425 {
1426 	struct vattr va;
1427 	int error;
1428 
1429 	if (!hardlink_check_uid && !hardlink_check_gid)
1430 		return (0);
1431 
1432 	error = VOP_GETATTR(vp, &va, cred);
1433 	if (error != 0)
1434 		return (error);
1435 
1436 	if (hardlink_check_uid && cred->cr_uid != va.va_uid) {
1437 		error = priv_check_cred(cred, PRIV_VFS_LINK, 0);
1438 		if (error != 0)
1439 			return (error);
1440 	}
1441 
1442 	if (hardlink_check_gid && !groupmember(va.va_gid, cred)) {
1443 		error = priv_check_cred(cred, PRIV_VFS_LINK, 0);
1444 		if (error != 0)
1445 			return (error);
1446 	}
1447 
1448 	return (0);
1449 }
1450 
1451 int
1452 kern_linkat(struct thread *td, int fd1, int fd2, char *path1, char *path2,
1453     enum uio_seg segflg, int follow)
1454 {
1455 	struct vnode *vp;
1456 	struct mount *mp;
1457 	struct nameidata nd;
1458 	cap_rights_t rights;
1459 	int error;
1460 
1461 again:
1462 	bwillwrite();
1463 	NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, segflg, path1, fd1,
1464 	    cap_rights_init(&rights, CAP_LINKAT_SOURCE), td);
1465 
1466 	if ((error = namei(&nd)) != 0)
1467 		return (error);
1468 	NDFREE(&nd, NDF_ONLY_PNBUF);
1469 	vp = nd.ni_vp;
1470 	if (vp->v_type == VDIR) {
1471 		vrele(vp);
1472 		return (EPERM);		/* POSIX */
1473 	}
1474 	NDINIT_ATRIGHTS(&nd, CREATE,
1475 	    LOCKPARENT | SAVENAME | AUDITVNODE2 | NOCACHE, segflg, path2, fd2,
1476 	    cap_rights_init(&rights, CAP_LINKAT_TARGET), td);
1477 	if ((error = namei(&nd)) == 0) {
1478 		if (nd.ni_vp != NULL) {
1479 			NDFREE(&nd, NDF_ONLY_PNBUF);
1480 			if (nd.ni_dvp == nd.ni_vp)
1481 				vrele(nd.ni_dvp);
1482 			else
1483 				vput(nd.ni_dvp);
1484 			vrele(nd.ni_vp);
1485 			vrele(vp);
1486 			return (EEXIST);
1487 		} else if (nd.ni_dvp->v_mount != vp->v_mount) {
1488 			/*
1489 			 * Cross-device link.  No need to recheck
1490 			 * vp->v_type, since it cannot change, except
1491 			 * to VBAD.
1492 			 */
1493 			NDFREE(&nd, NDF_ONLY_PNBUF);
1494 			vput(nd.ni_dvp);
1495 			vrele(vp);
1496 			return (EXDEV);
1497 		} else if ((error = vn_lock(vp, LK_EXCLUSIVE)) == 0) {
1498 			error = can_hardlink(vp, td->td_ucred);
1499 #ifdef MAC
1500 			if (error == 0)
1501 				error = mac_vnode_check_link(td->td_ucred,
1502 				    nd.ni_dvp, vp, &nd.ni_cnd);
1503 #endif
1504 			if (error != 0) {
1505 				vput(vp);
1506 				vput(nd.ni_dvp);
1507 				NDFREE(&nd, NDF_ONLY_PNBUF);
1508 				return (error);
1509 			}
1510 			error = vn_start_write(vp, &mp, V_NOWAIT);
1511 			if (error != 0) {
1512 				vput(vp);
1513 				vput(nd.ni_dvp);
1514 				NDFREE(&nd, NDF_ONLY_PNBUF);
1515 				error = vn_start_write(NULL, &mp,
1516 				    V_XSLEEP | PCATCH);
1517 				if (error != 0)
1518 					return (error);
1519 				goto again;
1520 			}
1521 			error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
1522 			VOP_UNLOCK(vp, 0);
1523 			vput(nd.ni_dvp);
1524 			vn_finished_write(mp);
1525 			NDFREE(&nd, NDF_ONLY_PNBUF);
1526 		} else {
1527 			vput(nd.ni_dvp);
1528 			NDFREE(&nd, NDF_ONLY_PNBUF);
1529 			vrele(vp);
1530 			goto again;
1531 		}
1532 	}
1533 	vrele(vp);
1534 	return (error);
1535 }
1536 
1537 /*
1538  * Make a symbolic link.
1539  */
1540 #ifndef _SYS_SYSPROTO_H_
1541 struct symlink_args {
1542 	char	*path;
1543 	char	*link;
1544 };
1545 #endif
1546 int
1547 sys_symlink(td, uap)
1548 	struct thread *td;
1549 	register struct symlink_args /* {
1550 		char *path;
1551 		char *link;
1552 	} */ *uap;
1553 {
1554 
1555 	return (kern_symlinkat(td, uap->path, AT_FDCWD, uap->link,
1556 	    UIO_USERSPACE));
1557 }
1558 
1559 #ifndef _SYS_SYSPROTO_H_
1560 struct symlinkat_args {
1561 	char	*path;
1562 	int	fd;
1563 	char	*path2;
1564 };
1565 #endif
1566 int
1567 sys_symlinkat(struct thread *td, struct symlinkat_args *uap)
1568 {
1569 
1570 	return (kern_symlinkat(td, uap->path1, uap->fd, uap->path2,
1571 	    UIO_USERSPACE));
1572 }
1573 
1574 int
1575 kern_symlinkat(struct thread *td, char *path1, int fd, char *path2,
1576     enum uio_seg segflg)
1577 {
1578 	struct mount *mp;
1579 	struct vattr vattr;
1580 	char *syspath;
1581 	struct nameidata nd;
1582 	int error;
1583 	cap_rights_t rights;
1584 
1585 	if (segflg == UIO_SYSSPACE) {
1586 		syspath = path1;
1587 	} else {
1588 		syspath = uma_zalloc(namei_zone, M_WAITOK);
1589 		if ((error = copyinstr(path1, syspath, MAXPATHLEN, NULL)) != 0)
1590 			goto out;
1591 	}
1592 	AUDIT_ARG_TEXT(syspath);
1593 restart:
1594 	bwillwrite();
1595 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
1596 	    NOCACHE, segflg, path2, fd, cap_rights_init(&rights, CAP_SYMLINKAT),
1597 	    td);
1598 	if ((error = namei(&nd)) != 0)
1599 		goto out;
1600 	if (nd.ni_vp) {
1601 		NDFREE(&nd, NDF_ONLY_PNBUF);
1602 		if (nd.ni_vp == nd.ni_dvp)
1603 			vrele(nd.ni_dvp);
1604 		else
1605 			vput(nd.ni_dvp);
1606 		vrele(nd.ni_vp);
1607 		error = EEXIST;
1608 		goto out;
1609 	}
1610 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1611 		NDFREE(&nd, NDF_ONLY_PNBUF);
1612 		vput(nd.ni_dvp);
1613 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1614 			goto out;
1615 		goto restart;
1616 	}
1617 	VATTR_NULL(&vattr);
1618 	vattr.va_mode = ACCESSPERMS &~ td->td_proc->p_fd->fd_cmask;
1619 #ifdef MAC
1620 	vattr.va_type = VLNK;
1621 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1622 	    &vattr);
1623 	if (error != 0)
1624 		goto out2;
1625 #endif
1626 	error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, syspath);
1627 	if (error == 0)
1628 		vput(nd.ni_vp);
1629 #ifdef MAC
1630 out2:
1631 #endif
1632 	NDFREE(&nd, NDF_ONLY_PNBUF);
1633 	vput(nd.ni_dvp);
1634 	vn_finished_write(mp);
1635 out:
1636 	if (segflg != UIO_SYSSPACE)
1637 		uma_zfree(namei_zone, syspath);
1638 	return (error);
1639 }
1640 
1641 /*
1642  * Delete a whiteout from the filesystem.
1643  */
1644 int
1645 sys_undelete(td, uap)
1646 	struct thread *td;
1647 	register struct undelete_args /* {
1648 		char *path;
1649 	} */ *uap;
1650 {
1651 	struct mount *mp;
1652 	struct nameidata nd;
1653 	int error;
1654 
1655 restart:
1656 	bwillwrite();
1657 	NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | AUDITVNODE1,
1658 	    UIO_USERSPACE, uap->path, td);
1659 	error = namei(&nd);
1660 	if (error != 0)
1661 		return (error);
1662 
1663 	if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) {
1664 		NDFREE(&nd, NDF_ONLY_PNBUF);
1665 		if (nd.ni_vp == nd.ni_dvp)
1666 			vrele(nd.ni_dvp);
1667 		else
1668 			vput(nd.ni_dvp);
1669 		if (nd.ni_vp)
1670 			vrele(nd.ni_vp);
1671 		return (EEXIST);
1672 	}
1673 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1674 		NDFREE(&nd, NDF_ONLY_PNBUF);
1675 		vput(nd.ni_dvp);
1676 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1677 			return (error);
1678 		goto restart;
1679 	}
1680 	error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE);
1681 	NDFREE(&nd, NDF_ONLY_PNBUF);
1682 	vput(nd.ni_dvp);
1683 	vn_finished_write(mp);
1684 	return (error);
1685 }
1686 
1687 /*
1688  * Delete a name from the filesystem.
1689  */
1690 #ifndef _SYS_SYSPROTO_H_
1691 struct unlink_args {
1692 	char	*path;
1693 };
1694 #endif
1695 int
1696 sys_unlink(td, uap)
1697 	struct thread *td;
1698 	struct unlink_args /* {
1699 		char *path;
1700 	} */ *uap;
1701 {
1702 
1703 	return (kern_unlinkat(td, AT_FDCWD, uap->path, UIO_USERSPACE, 0));
1704 }
1705 
1706 #ifndef _SYS_SYSPROTO_H_
1707 struct unlinkat_args {
1708 	int	fd;
1709 	char	*path;
1710 	int	flag;
1711 };
1712 #endif
1713 int
1714 sys_unlinkat(struct thread *td, struct unlinkat_args *uap)
1715 {
1716 	int flag = uap->flag;
1717 	int fd = uap->fd;
1718 	char *path = uap->path;
1719 
1720 	if (flag & ~AT_REMOVEDIR)
1721 		return (EINVAL);
1722 
1723 	if (flag & AT_REMOVEDIR)
1724 		return (kern_rmdirat(td, fd, path, UIO_USERSPACE));
1725 	else
1726 		return (kern_unlinkat(td, fd, path, UIO_USERSPACE, 0));
1727 }
1728 
1729 int
1730 kern_unlinkat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1731     ino_t oldinum)
1732 {
1733 	struct mount *mp;
1734 	struct vnode *vp;
1735 	struct nameidata nd;
1736 	struct stat sb;
1737 	cap_rights_t rights;
1738 	int error;
1739 
1740 restart:
1741 	bwillwrite();
1742 	NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1,
1743 	    pathseg, path, fd, cap_rights_init(&rights, CAP_UNLINKAT), td);
1744 	if ((error = namei(&nd)) != 0)
1745 		return (error == EINVAL ? EPERM : error);
1746 	vp = nd.ni_vp;
1747 	if (vp->v_type == VDIR && oldinum == 0) {
1748 		error = EPERM;		/* POSIX */
1749 	} else if (oldinum != 0 &&
1750 		  ((error = vn_stat(vp, &sb, td->td_ucred, NOCRED, td)) == 0) &&
1751 		  sb.st_ino != oldinum) {
1752 			error = EIDRM;	/* Identifier removed */
1753 	} else {
1754 		/*
1755 		 * The root of a mounted filesystem cannot be deleted.
1756 		 *
1757 		 * XXX: can this only be a VDIR case?
1758 		 */
1759 		if (vp->v_vflag & VV_ROOT)
1760 			error = EBUSY;
1761 	}
1762 	if (error == 0) {
1763 		if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1764 			NDFREE(&nd, NDF_ONLY_PNBUF);
1765 			vput(nd.ni_dvp);
1766 			if (vp == nd.ni_dvp)
1767 				vrele(vp);
1768 			else
1769 				vput(vp);
1770 			if ((error = vn_start_write(NULL, &mp,
1771 			    V_XSLEEP | PCATCH)) != 0)
1772 				return (error);
1773 			goto restart;
1774 		}
1775 #ifdef MAC
1776 		error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
1777 		    &nd.ni_cnd);
1778 		if (error != 0)
1779 			goto out;
1780 #endif
1781 		vfs_notify_upper(vp, VFS_NOTIFY_UPPER_UNLINK);
1782 		error = VOP_REMOVE(nd.ni_dvp, vp, &nd.ni_cnd);
1783 #ifdef MAC
1784 out:
1785 #endif
1786 		vn_finished_write(mp);
1787 	}
1788 	NDFREE(&nd, NDF_ONLY_PNBUF);
1789 	vput(nd.ni_dvp);
1790 	if (vp == nd.ni_dvp)
1791 		vrele(vp);
1792 	else
1793 		vput(vp);
1794 	return (error);
1795 }
1796 
1797 /*
1798  * Reposition read/write file offset.
1799  */
1800 #ifndef _SYS_SYSPROTO_H_
1801 struct lseek_args {
1802 	int	fd;
1803 	int	pad;
1804 	off_t	offset;
1805 	int	whence;
1806 };
1807 #endif
1808 int
1809 sys_lseek(td, uap)
1810 	struct thread *td;
1811 	register struct lseek_args /* {
1812 		int fd;
1813 		int pad;
1814 		off_t offset;
1815 		int whence;
1816 	} */ *uap;
1817 {
1818 	struct file *fp;
1819 	cap_rights_t rights;
1820 	int error;
1821 
1822 	AUDIT_ARG_FD(uap->fd);
1823 	error = fget(td, uap->fd, cap_rights_init(&rights, CAP_SEEK), &fp);
1824 	if (error != 0)
1825 		return (error);
1826 	error = (fp->f_ops->fo_flags & DFLAG_SEEKABLE) != 0 ?
1827 	    fo_seek(fp, uap->offset, uap->whence, td) : ESPIPE;
1828 	fdrop(fp, td);
1829 	return (error);
1830 }
1831 
1832 #if defined(COMPAT_43)
1833 /*
1834  * Reposition read/write file offset.
1835  */
1836 #ifndef _SYS_SYSPROTO_H_
1837 struct olseek_args {
1838 	int	fd;
1839 	long	offset;
1840 	int	whence;
1841 };
1842 #endif
1843 int
1844 olseek(td, uap)
1845 	struct thread *td;
1846 	register struct olseek_args /* {
1847 		int fd;
1848 		long offset;
1849 		int whence;
1850 	} */ *uap;
1851 {
1852 	struct lseek_args /* {
1853 		int fd;
1854 		int pad;
1855 		off_t offset;
1856 		int whence;
1857 	} */ nuap;
1858 
1859 	nuap.fd = uap->fd;
1860 	nuap.offset = uap->offset;
1861 	nuap.whence = uap->whence;
1862 	return (sys_lseek(td, &nuap));
1863 }
1864 #endif /* COMPAT_43 */
1865 
1866 #if defined(COMPAT_FREEBSD6)
1867 /* Version with the 'pad' argument */
1868 int
1869 freebsd6_lseek(td, uap)
1870 	struct thread *td;
1871 	register struct freebsd6_lseek_args *uap;
1872 {
1873 	struct lseek_args ouap;
1874 
1875 	ouap.fd = uap->fd;
1876 	ouap.offset = uap->offset;
1877 	ouap.whence = uap->whence;
1878 	return (sys_lseek(td, &ouap));
1879 }
1880 #endif
1881 
1882 /*
1883  * Check access permissions using passed credentials.
1884  */
1885 static int
1886 vn_access(vp, user_flags, cred, td)
1887 	struct vnode	*vp;
1888 	int		user_flags;
1889 	struct ucred	*cred;
1890 	struct thread	*td;
1891 {
1892 	accmode_t accmode;
1893 	int error;
1894 
1895 	/* Flags == 0 means only check for existence. */
1896 	if (user_flags == 0)
1897 		return (0);
1898 
1899 	accmode = 0;
1900 	if (user_flags & R_OK)
1901 		accmode |= VREAD;
1902 	if (user_flags & W_OK)
1903 		accmode |= VWRITE;
1904 	if (user_flags & X_OK)
1905 		accmode |= VEXEC;
1906 #ifdef MAC
1907 	error = mac_vnode_check_access(cred, vp, accmode);
1908 	if (error != 0)
1909 		return (error);
1910 #endif
1911 	if ((accmode & VWRITE) == 0 || (error = vn_writechk(vp)) == 0)
1912 		error = VOP_ACCESS(vp, accmode, cred, td);
1913 	return (error);
1914 }
1915 
1916 /*
1917  * Check access permissions using "real" credentials.
1918  */
1919 #ifndef _SYS_SYSPROTO_H_
1920 struct access_args {
1921 	char	*path;
1922 	int	amode;
1923 };
1924 #endif
1925 int
1926 sys_access(td, uap)
1927 	struct thread *td;
1928 	register struct access_args /* {
1929 		char *path;
1930 		int amode;
1931 	} */ *uap;
1932 {
1933 
1934 	return (kern_accessat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1935 	    0, uap->amode));
1936 }
1937 
1938 #ifndef _SYS_SYSPROTO_H_
1939 struct faccessat_args {
1940 	int	dirfd;
1941 	char	*path;
1942 	int	amode;
1943 	int	flag;
1944 }
1945 #endif
1946 int
1947 sys_faccessat(struct thread *td, struct faccessat_args *uap)
1948 {
1949 
1950 	return (kern_accessat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
1951 	    uap->amode));
1952 }
1953 
1954 int
1955 kern_accessat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1956     int flag, int amode)
1957 {
1958 	struct ucred *cred, *usecred;
1959 	struct vnode *vp;
1960 	struct nameidata nd;
1961 	cap_rights_t rights;
1962 	int error;
1963 
1964 	if (flag & ~AT_EACCESS)
1965 		return (EINVAL);
1966 	if (amode != F_OK && (amode & ~(R_OK | W_OK | X_OK)) != 0)
1967 		return (EINVAL);
1968 
1969 	/*
1970 	 * Create and modify a temporary credential instead of one that
1971 	 * is potentially shared (if we need one).
1972 	 */
1973 	cred = td->td_ucred;
1974 	if ((flag & AT_EACCESS) == 0 &&
1975 	    ((cred->cr_uid != cred->cr_ruid ||
1976 	    cred->cr_rgid != cred->cr_groups[0]))) {
1977 		usecred = crdup(cred);
1978 		usecred->cr_uid = cred->cr_ruid;
1979 		usecred->cr_groups[0] = cred->cr_rgid;
1980 		td->td_ucred = usecred;
1981 	} else
1982 		usecred = cred;
1983 	AUDIT_ARG_VALUE(amode);
1984 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF |
1985 	    AUDITVNODE1, pathseg, path, fd, cap_rights_init(&rights, CAP_FSTAT),
1986 	    td);
1987 	if ((error = namei(&nd)) != 0)
1988 		goto out;
1989 	vp = nd.ni_vp;
1990 
1991 	error = vn_access(vp, amode, usecred, td);
1992 	NDFREE(&nd, NDF_ONLY_PNBUF);
1993 	vput(vp);
1994 out:
1995 	if (usecred != cred) {
1996 		td->td_ucred = cred;
1997 		crfree(usecred);
1998 	}
1999 	return (error);
2000 }
2001 
2002 /*
2003  * Check access permissions using "effective" credentials.
2004  */
2005 #ifndef _SYS_SYSPROTO_H_
2006 struct eaccess_args {
2007 	char	*path;
2008 	int	amode;
2009 };
2010 #endif
2011 int
2012 sys_eaccess(td, uap)
2013 	struct thread *td;
2014 	register struct eaccess_args /* {
2015 		char *path;
2016 		int amode;
2017 	} */ *uap;
2018 {
2019 
2020 	return (kern_accessat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2021 	    AT_EACCESS, uap->amode));
2022 }
2023 
2024 #if defined(COMPAT_43)
2025 /*
2026  * Get file status; this version follows links.
2027  */
2028 #ifndef _SYS_SYSPROTO_H_
2029 struct ostat_args {
2030 	char	*path;
2031 	struct ostat *ub;
2032 };
2033 #endif
2034 int
2035 ostat(td, uap)
2036 	struct thread *td;
2037 	register struct ostat_args /* {
2038 		char *path;
2039 		struct ostat *ub;
2040 	} */ *uap;
2041 {
2042 	struct stat sb;
2043 	struct ostat osb;
2044 	int error;
2045 
2046 	error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE,
2047 	    &sb, NULL);
2048 	if (error != 0)
2049 		return (error);
2050 	cvtstat(&sb, &osb);
2051 	return (copyout(&osb, uap->ub, sizeof (osb)));
2052 }
2053 
2054 /*
2055  * Get file status; this version does not follow links.
2056  */
2057 #ifndef _SYS_SYSPROTO_H_
2058 struct olstat_args {
2059 	char	*path;
2060 	struct ostat *ub;
2061 };
2062 #endif
2063 int
2064 olstat(td, uap)
2065 	struct thread *td;
2066 	register struct olstat_args /* {
2067 		char *path;
2068 		struct ostat *ub;
2069 	} */ *uap;
2070 {
2071 	struct stat sb;
2072 	struct ostat osb;
2073 	int error;
2074 
2075 	error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2076 	    UIO_USERSPACE, &sb, NULL);
2077 	if (error != 0)
2078 		return (error);
2079 	cvtstat(&sb, &osb);
2080 	return (copyout(&osb, uap->ub, sizeof (osb)));
2081 }
2082 
2083 /*
2084  * Convert from an old to a new stat structure.
2085  */
2086 void
2087 cvtstat(st, ost)
2088 	struct stat *st;
2089 	struct ostat *ost;
2090 {
2091 
2092 	bzero(ost, sizeof(*ost));
2093 	ost->st_dev = st->st_dev;
2094 	ost->st_ino = st->st_ino;
2095 	ost->st_mode = st->st_mode;
2096 	ost->st_nlink = st->st_nlink;
2097 	ost->st_uid = st->st_uid;
2098 	ost->st_gid = st->st_gid;
2099 	ost->st_rdev = st->st_rdev;
2100 	if (st->st_size < (quad_t)1 << 32)
2101 		ost->st_size = st->st_size;
2102 	else
2103 		ost->st_size = -2;
2104 	ost->st_atim = st->st_atim;
2105 	ost->st_mtim = st->st_mtim;
2106 	ost->st_ctim = st->st_ctim;
2107 	ost->st_blksize = st->st_blksize;
2108 	ost->st_blocks = st->st_blocks;
2109 	ost->st_flags = st->st_flags;
2110 	ost->st_gen = st->st_gen;
2111 }
2112 #endif /* COMPAT_43 */
2113 
2114 /*
2115  * Get file status; this version follows links.
2116  */
2117 #ifndef _SYS_SYSPROTO_H_
2118 struct stat_args {
2119 	char	*path;
2120 	struct stat *ub;
2121 };
2122 #endif
2123 int
2124 sys_stat(td, uap)
2125 	struct thread *td;
2126 	register struct stat_args /* {
2127 		char *path;
2128 		struct stat *ub;
2129 	} */ *uap;
2130 {
2131 	struct stat sb;
2132 	int error;
2133 
2134 	error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE,
2135 	    &sb, NULL);
2136 	if (error == 0)
2137 		error = copyout(&sb, uap->ub, sizeof (sb));
2138 	return (error);
2139 }
2140 
2141 #ifndef _SYS_SYSPROTO_H_
2142 struct fstatat_args {
2143 	int	fd;
2144 	char	*path;
2145 	struct stat	*buf;
2146 	int	flag;
2147 }
2148 #endif
2149 int
2150 sys_fstatat(struct thread *td, struct fstatat_args *uap)
2151 {
2152 	struct stat sb;
2153 	int error;
2154 
2155 	error = kern_statat(td, uap->flag, uap->fd, uap->path,
2156 	    UIO_USERSPACE, &sb, NULL);
2157 	if (error == 0)
2158 		error = copyout(&sb, uap->buf, sizeof (sb));
2159 	return (error);
2160 }
2161 
2162 int
2163 kern_statat(struct thread *td, int flag, int fd, char *path,
2164     enum uio_seg pathseg, struct stat *sbp,
2165     void (*hook)(struct vnode *vp, struct stat *sbp))
2166 {
2167 	struct nameidata nd;
2168 	struct stat sb;
2169 	cap_rights_t rights;
2170 	int error;
2171 
2172 	if (flag & ~AT_SYMLINK_NOFOLLOW)
2173 		return (EINVAL);
2174 
2175 	NDINIT_ATRIGHTS(&nd, LOOKUP, ((flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW :
2176 	    FOLLOW) | LOCKSHARED | LOCKLEAF | AUDITVNODE1, pathseg, path, fd,
2177 	    cap_rights_init(&rights, CAP_FSTAT), td);
2178 
2179 	if ((error = namei(&nd)) != 0)
2180 		return (error);
2181 	error = vn_stat(nd.ni_vp, &sb, td->td_ucred, NOCRED, td);
2182 	if (error == 0) {
2183 		SDT_PROBE2(vfs, , stat, mode, path, sb.st_mode);
2184 		if (S_ISREG(sb.st_mode))
2185 			SDT_PROBE2(vfs, , stat, reg, path, pathseg);
2186 		if (__predict_false(hook != NULL))
2187 			hook(nd.ni_vp, &sb);
2188 	}
2189 	NDFREE(&nd, NDF_ONLY_PNBUF);
2190 	vput(nd.ni_vp);
2191 	if (error != 0)
2192 		return (error);
2193 	*sbp = sb;
2194 #ifdef KTRACE
2195 	if (KTRPOINT(td, KTR_STRUCT))
2196 		ktrstat(&sb);
2197 #endif
2198 	return (0);
2199 }
2200 
2201 /*
2202  * Get file status; this version does not follow links.
2203  */
2204 #ifndef _SYS_SYSPROTO_H_
2205 struct lstat_args {
2206 	char	*path;
2207 	struct stat *ub;
2208 };
2209 #endif
2210 int
2211 sys_lstat(td, uap)
2212 	struct thread *td;
2213 	register struct lstat_args /* {
2214 		char *path;
2215 		struct stat *ub;
2216 	} */ *uap;
2217 {
2218 	struct stat sb;
2219 	int error;
2220 
2221 	error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2222 	    UIO_USERSPACE, &sb, NULL);
2223 	if (error == 0)
2224 		error = copyout(&sb, uap->ub, sizeof (sb));
2225 	return (error);
2226 }
2227 
2228 /*
2229  * Implementation of the NetBSD [l]stat() functions.
2230  */
2231 void
2232 cvtnstat(sb, nsb)
2233 	struct stat *sb;
2234 	struct nstat *nsb;
2235 {
2236 
2237 	bzero(nsb, sizeof *nsb);
2238 	nsb->st_dev = sb->st_dev;
2239 	nsb->st_ino = sb->st_ino;
2240 	nsb->st_mode = sb->st_mode;
2241 	nsb->st_nlink = sb->st_nlink;
2242 	nsb->st_uid = sb->st_uid;
2243 	nsb->st_gid = sb->st_gid;
2244 	nsb->st_rdev = sb->st_rdev;
2245 	nsb->st_atim = sb->st_atim;
2246 	nsb->st_mtim = sb->st_mtim;
2247 	nsb->st_ctim = sb->st_ctim;
2248 	nsb->st_size = sb->st_size;
2249 	nsb->st_blocks = sb->st_blocks;
2250 	nsb->st_blksize = sb->st_blksize;
2251 	nsb->st_flags = sb->st_flags;
2252 	nsb->st_gen = sb->st_gen;
2253 	nsb->st_birthtim = sb->st_birthtim;
2254 }
2255 
2256 #ifndef _SYS_SYSPROTO_H_
2257 struct nstat_args {
2258 	char	*path;
2259 	struct nstat *ub;
2260 };
2261 #endif
2262 int
2263 sys_nstat(td, uap)
2264 	struct thread *td;
2265 	register struct nstat_args /* {
2266 		char *path;
2267 		struct nstat *ub;
2268 	} */ *uap;
2269 {
2270 	struct stat sb;
2271 	struct nstat nsb;
2272 	int error;
2273 
2274 	error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE,
2275 	    &sb, NULL);
2276 	if (error != 0)
2277 		return (error);
2278 	cvtnstat(&sb, &nsb);
2279 	return (copyout(&nsb, uap->ub, sizeof (nsb)));
2280 }
2281 
2282 /*
2283  * NetBSD lstat.  Get file status; this version does not follow links.
2284  */
2285 #ifndef _SYS_SYSPROTO_H_
2286 struct lstat_args {
2287 	char	*path;
2288 	struct stat *ub;
2289 };
2290 #endif
2291 int
2292 sys_nlstat(td, uap)
2293 	struct thread *td;
2294 	register struct nlstat_args /* {
2295 		char *path;
2296 		struct nstat *ub;
2297 	} */ *uap;
2298 {
2299 	struct stat sb;
2300 	struct nstat nsb;
2301 	int error;
2302 
2303 	error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2304 	    UIO_USERSPACE, &sb, NULL);
2305 	if (error != 0)
2306 		return (error);
2307 	cvtnstat(&sb, &nsb);
2308 	return (copyout(&nsb, uap->ub, sizeof (nsb)));
2309 }
2310 
2311 /*
2312  * Get configurable pathname variables.
2313  */
2314 #ifndef _SYS_SYSPROTO_H_
2315 struct pathconf_args {
2316 	char	*path;
2317 	int	name;
2318 };
2319 #endif
2320 int
2321 sys_pathconf(td, uap)
2322 	struct thread *td;
2323 	register struct pathconf_args /* {
2324 		char *path;
2325 		int name;
2326 	} */ *uap;
2327 {
2328 
2329 	return (kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name, FOLLOW));
2330 }
2331 
2332 #ifndef _SYS_SYSPROTO_H_
2333 struct lpathconf_args {
2334 	char	*path;
2335 	int	name;
2336 };
2337 #endif
2338 int
2339 sys_lpathconf(td, uap)
2340 	struct thread *td;
2341 	register struct lpathconf_args /* {
2342 		char *path;
2343 		int name;
2344 	} */ *uap;
2345 {
2346 
2347 	return (kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name,
2348 	    NOFOLLOW));
2349 }
2350 
2351 int
2352 kern_pathconf(struct thread *td, char *path, enum uio_seg pathseg, int name,
2353     u_long flags)
2354 {
2355 	struct nameidata nd;
2356 	int error;
2357 
2358 	NDINIT(&nd, LOOKUP, LOCKSHARED | LOCKLEAF | AUDITVNODE1 | flags,
2359 	    pathseg, path, td);
2360 	if ((error = namei(&nd)) != 0)
2361 		return (error);
2362 	NDFREE(&nd, NDF_ONLY_PNBUF);
2363 
2364 	error = VOP_PATHCONF(nd.ni_vp, name, td->td_retval);
2365 	vput(nd.ni_vp);
2366 	return (error);
2367 }
2368 
2369 /*
2370  * Return target name of a symbolic link.
2371  */
2372 #ifndef _SYS_SYSPROTO_H_
2373 struct readlink_args {
2374 	char	*path;
2375 	char	*buf;
2376 	size_t	count;
2377 };
2378 #endif
2379 int
2380 sys_readlink(td, uap)
2381 	struct thread *td;
2382 	register struct readlink_args /* {
2383 		char *path;
2384 		char *buf;
2385 		size_t count;
2386 	} */ *uap;
2387 {
2388 
2389 	return (kern_readlinkat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2390 	    uap->buf, UIO_USERSPACE, uap->count));
2391 }
2392 #ifndef _SYS_SYSPROTO_H_
2393 struct readlinkat_args {
2394 	int	fd;
2395 	char	*path;
2396 	char	*buf;
2397 	size_t	bufsize;
2398 };
2399 #endif
2400 int
2401 sys_readlinkat(struct thread *td, struct readlinkat_args *uap)
2402 {
2403 
2404 	return (kern_readlinkat(td, uap->fd, uap->path, UIO_USERSPACE,
2405 	    uap->buf, UIO_USERSPACE, uap->bufsize));
2406 }
2407 
2408 int
2409 kern_readlinkat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2410     char *buf, enum uio_seg bufseg, size_t count)
2411 {
2412 	struct vnode *vp;
2413 	struct iovec aiov;
2414 	struct uio auio;
2415 	struct nameidata nd;
2416 	int error;
2417 
2418 	if (count > IOSIZE_MAX)
2419 		return (EINVAL);
2420 
2421 	NDINIT_AT(&nd, LOOKUP, NOFOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
2422 	    pathseg, path, fd, td);
2423 
2424 	if ((error = namei(&nd)) != 0)
2425 		return (error);
2426 	NDFREE(&nd, NDF_ONLY_PNBUF);
2427 	vp = nd.ni_vp;
2428 #ifdef MAC
2429 	error = mac_vnode_check_readlink(td->td_ucred, vp);
2430 	if (error != 0) {
2431 		vput(vp);
2432 		return (error);
2433 	}
2434 #endif
2435 	if (vp->v_type != VLNK)
2436 		error = EINVAL;
2437 	else {
2438 		aiov.iov_base = buf;
2439 		aiov.iov_len = count;
2440 		auio.uio_iov = &aiov;
2441 		auio.uio_iovcnt = 1;
2442 		auio.uio_offset = 0;
2443 		auio.uio_rw = UIO_READ;
2444 		auio.uio_segflg = bufseg;
2445 		auio.uio_td = td;
2446 		auio.uio_resid = count;
2447 		error = VOP_READLINK(vp, &auio, td->td_ucred);
2448 		td->td_retval[0] = count - auio.uio_resid;
2449 	}
2450 	vput(vp);
2451 	return (error);
2452 }
2453 
2454 /*
2455  * Common implementation code for chflags() and fchflags().
2456  */
2457 static int
2458 setfflags(td, vp, flags)
2459 	struct thread *td;
2460 	struct vnode *vp;
2461 	u_long flags;
2462 {
2463 	struct mount *mp;
2464 	struct vattr vattr;
2465 	int error;
2466 
2467 	/* We can't support the value matching VNOVAL. */
2468 	if (flags == VNOVAL)
2469 		return (EOPNOTSUPP);
2470 
2471 	/*
2472 	 * Prevent non-root users from setting flags on devices.  When
2473 	 * a device is reused, users can retain ownership of the device
2474 	 * if they are allowed to set flags and programs assume that
2475 	 * chown can't fail when done as root.
2476 	 */
2477 	if (vp->v_type == VCHR || vp->v_type == VBLK) {
2478 		error = priv_check(td, PRIV_VFS_CHFLAGS_DEV);
2479 		if (error != 0)
2480 			return (error);
2481 	}
2482 
2483 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2484 		return (error);
2485 	VATTR_NULL(&vattr);
2486 	vattr.va_flags = flags;
2487 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2488 #ifdef MAC
2489 	error = mac_vnode_check_setflags(td->td_ucred, vp, vattr.va_flags);
2490 	if (error == 0)
2491 #endif
2492 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
2493 	VOP_UNLOCK(vp, 0);
2494 	vn_finished_write(mp);
2495 	return (error);
2496 }
2497 
2498 /*
2499  * Change flags of a file given a path name.
2500  */
2501 #ifndef _SYS_SYSPROTO_H_
2502 struct chflags_args {
2503 	const char *path;
2504 	u_long	flags;
2505 };
2506 #endif
2507 int
2508 sys_chflags(td, uap)
2509 	struct thread *td;
2510 	register struct chflags_args /* {
2511 		const char *path;
2512 		u_long flags;
2513 	} */ *uap;
2514 {
2515 
2516 	return (kern_chflagsat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2517 	    uap->flags, 0));
2518 }
2519 
2520 #ifndef _SYS_SYSPROTO_H_
2521 struct chflagsat_args {
2522 	int	fd;
2523 	const char *path;
2524 	u_long	flags;
2525 	int	atflag;
2526 }
2527 #endif
2528 int
2529 sys_chflagsat(struct thread *td, struct chflagsat_args *uap)
2530 {
2531 	int fd = uap->fd;
2532 	const char *path = uap->path;
2533 	u_long flags = uap->flags;
2534 	int atflag = uap->atflag;
2535 
2536 	if (atflag & ~AT_SYMLINK_NOFOLLOW)
2537 		return (EINVAL);
2538 
2539 	return (kern_chflagsat(td, fd, path, UIO_USERSPACE, flags, atflag));
2540 }
2541 
2542 /*
2543  * Same as chflags() but doesn't follow symlinks.
2544  */
2545 int
2546 sys_lchflags(td, uap)
2547 	struct thread *td;
2548 	register struct lchflags_args /* {
2549 		const char *path;
2550 		u_long flags;
2551 	} */ *uap;
2552 {
2553 
2554 	return (kern_chflagsat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2555 	    uap->flags, AT_SYMLINK_NOFOLLOW));
2556 }
2557 
2558 static int
2559 kern_chflagsat(struct thread *td, int fd, const char *path,
2560     enum uio_seg pathseg, u_long flags, int atflag)
2561 {
2562 	struct nameidata nd;
2563 	cap_rights_t rights;
2564 	int error, follow;
2565 
2566 	AUDIT_ARG_FFLAGS(flags);
2567 	follow = (atflag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
2568 	NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, pathseg, path, fd,
2569 	    cap_rights_init(&rights, CAP_FCHFLAGS), td);
2570 	if ((error = namei(&nd)) != 0)
2571 		return (error);
2572 	NDFREE(&nd, NDF_ONLY_PNBUF);
2573 	error = setfflags(td, nd.ni_vp, flags);
2574 	vrele(nd.ni_vp);
2575 	return (error);
2576 }
2577 
2578 /*
2579  * Change flags of a file given a file descriptor.
2580  */
2581 #ifndef _SYS_SYSPROTO_H_
2582 struct fchflags_args {
2583 	int	fd;
2584 	u_long	flags;
2585 };
2586 #endif
2587 int
2588 sys_fchflags(td, uap)
2589 	struct thread *td;
2590 	register struct fchflags_args /* {
2591 		int fd;
2592 		u_long flags;
2593 	} */ *uap;
2594 {
2595 	struct file *fp;
2596 	cap_rights_t rights;
2597 	int error;
2598 
2599 	AUDIT_ARG_FD(uap->fd);
2600 	AUDIT_ARG_FFLAGS(uap->flags);
2601 	error = getvnode(td, uap->fd, cap_rights_init(&rights, CAP_FCHFLAGS),
2602 	    &fp);
2603 	if (error != 0)
2604 		return (error);
2605 #ifdef AUDIT
2606 	vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
2607 	AUDIT_ARG_VNODE1(fp->f_vnode);
2608 	VOP_UNLOCK(fp->f_vnode, 0);
2609 #endif
2610 	error = setfflags(td, fp->f_vnode, uap->flags);
2611 	fdrop(fp, td);
2612 	return (error);
2613 }
2614 
2615 /*
2616  * Common implementation code for chmod(), lchmod() and fchmod().
2617  */
2618 int
2619 setfmode(td, cred, vp, mode)
2620 	struct thread *td;
2621 	struct ucred *cred;
2622 	struct vnode *vp;
2623 	int mode;
2624 {
2625 	struct mount *mp;
2626 	struct vattr vattr;
2627 	int error;
2628 
2629 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2630 		return (error);
2631 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2632 	VATTR_NULL(&vattr);
2633 	vattr.va_mode = mode & ALLPERMS;
2634 #ifdef MAC
2635 	error = mac_vnode_check_setmode(cred, vp, vattr.va_mode);
2636 	if (error == 0)
2637 #endif
2638 		error = VOP_SETATTR(vp, &vattr, cred);
2639 	VOP_UNLOCK(vp, 0);
2640 	vn_finished_write(mp);
2641 	return (error);
2642 }
2643 
2644 /*
2645  * Change mode of a file given path name.
2646  */
2647 #ifndef _SYS_SYSPROTO_H_
2648 struct chmod_args {
2649 	char	*path;
2650 	int	mode;
2651 };
2652 #endif
2653 int
2654 sys_chmod(td, uap)
2655 	struct thread *td;
2656 	register struct chmod_args /* {
2657 		char *path;
2658 		int mode;
2659 	} */ *uap;
2660 {
2661 
2662 	return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2663 	    uap->mode, 0));
2664 }
2665 
2666 #ifndef _SYS_SYSPROTO_H_
2667 struct fchmodat_args {
2668 	int	dirfd;
2669 	char	*path;
2670 	mode_t	mode;
2671 	int	flag;
2672 }
2673 #endif
2674 int
2675 sys_fchmodat(struct thread *td, struct fchmodat_args *uap)
2676 {
2677 	int flag = uap->flag;
2678 	int fd = uap->fd;
2679 	char *path = uap->path;
2680 	mode_t mode = uap->mode;
2681 
2682 	if (flag & ~AT_SYMLINK_NOFOLLOW)
2683 		return (EINVAL);
2684 
2685 	return (kern_fchmodat(td, fd, path, UIO_USERSPACE, mode, flag));
2686 }
2687 
2688 /*
2689  * Change mode of a file given path name (don't follow links.)
2690  */
2691 #ifndef _SYS_SYSPROTO_H_
2692 struct lchmod_args {
2693 	char	*path;
2694 	int	mode;
2695 };
2696 #endif
2697 int
2698 sys_lchmod(td, uap)
2699 	struct thread *td;
2700 	register struct lchmod_args /* {
2701 		char *path;
2702 		int mode;
2703 	} */ *uap;
2704 {
2705 
2706 	return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2707 	    uap->mode, AT_SYMLINK_NOFOLLOW));
2708 }
2709 
2710 int
2711 kern_fchmodat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2712     mode_t mode, int flag)
2713 {
2714 	struct nameidata nd;
2715 	cap_rights_t rights;
2716 	int error, follow;
2717 
2718 	AUDIT_ARG_MODE(mode);
2719 	follow = (flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
2720 	NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, pathseg, path, fd,
2721 	    cap_rights_init(&rights, CAP_FCHMOD), td);
2722 	if ((error = namei(&nd)) != 0)
2723 		return (error);
2724 	NDFREE(&nd, NDF_ONLY_PNBUF);
2725 	error = setfmode(td, td->td_ucred, nd.ni_vp, mode);
2726 	vrele(nd.ni_vp);
2727 	return (error);
2728 }
2729 
2730 /*
2731  * Change mode of a file given a file descriptor.
2732  */
2733 #ifndef _SYS_SYSPROTO_H_
2734 struct fchmod_args {
2735 	int	fd;
2736 	int	mode;
2737 };
2738 #endif
2739 int
2740 sys_fchmod(struct thread *td, struct fchmod_args *uap)
2741 {
2742 	struct file *fp;
2743 	cap_rights_t rights;
2744 	int error;
2745 
2746 	AUDIT_ARG_FD(uap->fd);
2747 	AUDIT_ARG_MODE(uap->mode);
2748 
2749 	error = fget(td, uap->fd, cap_rights_init(&rights, CAP_FCHMOD), &fp);
2750 	if (error != 0)
2751 		return (error);
2752 	error = fo_chmod(fp, uap->mode, td->td_ucred, td);
2753 	fdrop(fp, td);
2754 	return (error);
2755 }
2756 
2757 /*
2758  * Common implementation for chown(), lchown(), and fchown()
2759  */
2760 int
2761 setfown(td, cred, vp, uid, gid)
2762 	struct thread *td;
2763 	struct ucred *cred;
2764 	struct vnode *vp;
2765 	uid_t uid;
2766 	gid_t gid;
2767 {
2768 	struct mount *mp;
2769 	struct vattr vattr;
2770 	int error;
2771 
2772 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2773 		return (error);
2774 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2775 	VATTR_NULL(&vattr);
2776 	vattr.va_uid = uid;
2777 	vattr.va_gid = gid;
2778 #ifdef MAC
2779 	error = mac_vnode_check_setowner(cred, vp, vattr.va_uid,
2780 	    vattr.va_gid);
2781 	if (error == 0)
2782 #endif
2783 		error = VOP_SETATTR(vp, &vattr, cred);
2784 	VOP_UNLOCK(vp, 0);
2785 	vn_finished_write(mp);
2786 	return (error);
2787 }
2788 
2789 /*
2790  * Set ownership given a path name.
2791  */
2792 #ifndef _SYS_SYSPROTO_H_
2793 struct chown_args {
2794 	char	*path;
2795 	int	uid;
2796 	int	gid;
2797 };
2798 #endif
2799 int
2800 sys_chown(td, uap)
2801 	struct thread *td;
2802 	register struct chown_args /* {
2803 		char *path;
2804 		int uid;
2805 		int gid;
2806 	} */ *uap;
2807 {
2808 
2809 	return (kern_fchownat(td, AT_FDCWD, uap->path, UIO_USERSPACE, uap->uid,
2810 	    uap->gid, 0));
2811 }
2812 
2813 #ifndef _SYS_SYSPROTO_H_
2814 struct fchownat_args {
2815 	int fd;
2816 	const char * path;
2817 	uid_t uid;
2818 	gid_t gid;
2819 	int flag;
2820 };
2821 #endif
2822 int
2823 sys_fchownat(struct thread *td, struct fchownat_args *uap)
2824 {
2825 	int flag;
2826 
2827 	flag = uap->flag;
2828 	if (flag & ~AT_SYMLINK_NOFOLLOW)
2829 		return (EINVAL);
2830 
2831 	return (kern_fchownat(td, uap->fd, uap->path, UIO_USERSPACE, uap->uid,
2832 	    uap->gid, uap->flag));
2833 }
2834 
2835 int
2836 kern_fchownat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2837     int uid, int gid, int flag)
2838 {
2839 	struct nameidata nd;
2840 	cap_rights_t rights;
2841 	int error, follow;
2842 
2843 	AUDIT_ARG_OWNER(uid, gid);
2844 	follow = (flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
2845 	NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, pathseg, path, fd,
2846 	    cap_rights_init(&rights, CAP_FCHOWN), td);
2847 
2848 	if ((error = namei(&nd)) != 0)
2849 		return (error);
2850 	NDFREE(&nd, NDF_ONLY_PNBUF);
2851 	error = setfown(td, td->td_ucred, nd.ni_vp, uid, gid);
2852 	vrele(nd.ni_vp);
2853 	return (error);
2854 }
2855 
2856 /*
2857  * Set ownership given a path name, do not cross symlinks.
2858  */
2859 #ifndef _SYS_SYSPROTO_H_
2860 struct lchown_args {
2861 	char	*path;
2862 	int	uid;
2863 	int	gid;
2864 };
2865 #endif
2866 int
2867 sys_lchown(td, uap)
2868 	struct thread *td;
2869 	register struct lchown_args /* {
2870 		char *path;
2871 		int uid;
2872 		int gid;
2873 	} */ *uap;
2874 {
2875 
2876 	return (kern_fchownat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2877 	    uap->uid, uap->gid, AT_SYMLINK_NOFOLLOW));
2878 }
2879 
2880 /*
2881  * Set ownership given a file descriptor.
2882  */
2883 #ifndef _SYS_SYSPROTO_H_
2884 struct fchown_args {
2885 	int	fd;
2886 	int	uid;
2887 	int	gid;
2888 };
2889 #endif
2890 int
2891 sys_fchown(td, uap)
2892 	struct thread *td;
2893 	register struct fchown_args /* {
2894 		int fd;
2895 		int uid;
2896 		int gid;
2897 	} */ *uap;
2898 {
2899 	struct file *fp;
2900 	cap_rights_t rights;
2901 	int error;
2902 
2903 	AUDIT_ARG_FD(uap->fd);
2904 	AUDIT_ARG_OWNER(uap->uid, uap->gid);
2905 	error = fget(td, uap->fd, cap_rights_init(&rights, CAP_FCHOWN), &fp);
2906 	if (error != 0)
2907 		return (error);
2908 	error = fo_chown(fp, uap->uid, uap->gid, td->td_ucred, td);
2909 	fdrop(fp, td);
2910 	return (error);
2911 }
2912 
2913 /*
2914  * Common implementation code for utimes(), lutimes(), and futimes().
2915  */
2916 static int
2917 getutimes(usrtvp, tvpseg, tsp)
2918 	const struct timeval *usrtvp;
2919 	enum uio_seg tvpseg;
2920 	struct timespec *tsp;
2921 {
2922 	struct timeval tv[2];
2923 	const struct timeval *tvp;
2924 	int error;
2925 
2926 	if (usrtvp == NULL) {
2927 		vfs_timestamp(&tsp[0]);
2928 		tsp[1] = tsp[0];
2929 	} else {
2930 		if (tvpseg == UIO_SYSSPACE) {
2931 			tvp = usrtvp;
2932 		} else {
2933 			if ((error = copyin(usrtvp, tv, sizeof(tv))) != 0)
2934 				return (error);
2935 			tvp = tv;
2936 		}
2937 
2938 		if (tvp[0].tv_usec < 0 || tvp[0].tv_usec >= 1000000 ||
2939 		    tvp[1].tv_usec < 0 || tvp[1].tv_usec >= 1000000)
2940 			return (EINVAL);
2941 		TIMEVAL_TO_TIMESPEC(&tvp[0], &tsp[0]);
2942 		TIMEVAL_TO_TIMESPEC(&tvp[1], &tsp[1]);
2943 	}
2944 	return (0);
2945 }
2946 
2947 /*
2948  * Common implementation code for futimens(), utimensat().
2949  */
2950 #define	UTIMENS_NULL	0x1
2951 #define	UTIMENS_EXIT	0x2
2952 static int
2953 getutimens(const struct timespec *usrtsp, enum uio_seg tspseg,
2954     struct timespec *tsp, int *retflags)
2955 {
2956 	struct timespec tsnow;
2957 	int error;
2958 
2959 	vfs_timestamp(&tsnow);
2960 	*retflags = 0;
2961 	if (usrtsp == NULL) {
2962 		tsp[0] = tsnow;
2963 		tsp[1] = tsnow;
2964 		*retflags |= UTIMENS_NULL;
2965 		return (0);
2966 	}
2967 	if (tspseg == UIO_SYSSPACE) {
2968 		tsp[0] = usrtsp[0];
2969 		tsp[1] = usrtsp[1];
2970 	} else if ((error = copyin(usrtsp, tsp, sizeof(*tsp) * 2)) != 0)
2971 		return (error);
2972 	if (tsp[0].tv_nsec == UTIME_OMIT && tsp[1].tv_nsec == UTIME_OMIT)
2973 		*retflags |= UTIMENS_EXIT;
2974 	if (tsp[0].tv_nsec == UTIME_NOW && tsp[1].tv_nsec == UTIME_NOW)
2975 		*retflags |= UTIMENS_NULL;
2976 	if (tsp[0].tv_nsec == UTIME_OMIT)
2977 		tsp[0].tv_sec = VNOVAL;
2978 	else if (tsp[0].tv_nsec == UTIME_NOW)
2979 		tsp[0] = tsnow;
2980 	else if (tsp[0].tv_nsec < 0 || tsp[0].tv_nsec >= 1000000000L)
2981 		return (EINVAL);
2982 	if (tsp[1].tv_nsec == UTIME_OMIT)
2983 		tsp[1].tv_sec = VNOVAL;
2984 	else if (tsp[1].tv_nsec == UTIME_NOW)
2985 		tsp[1] = tsnow;
2986 	else if (tsp[1].tv_nsec < 0 || tsp[1].tv_nsec >= 1000000000L)
2987 		return (EINVAL);
2988 
2989 	return (0);
2990 }
2991 
2992 /*
2993  * Common implementation code for utimes(), lutimes(), futimes(), futimens(),
2994  * and utimensat().
2995  */
2996 static int
2997 setutimes(td, vp, ts, numtimes, nullflag)
2998 	struct thread *td;
2999 	struct vnode *vp;
3000 	const struct timespec *ts;
3001 	int numtimes;
3002 	int nullflag;
3003 {
3004 	struct mount *mp;
3005 	struct vattr vattr;
3006 	int error, setbirthtime;
3007 
3008 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
3009 		return (error);
3010 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3011 	setbirthtime = 0;
3012 	if (numtimes < 3 && !VOP_GETATTR(vp, &vattr, td->td_ucred) &&
3013 	    timespeccmp(&ts[1], &vattr.va_birthtime, < ))
3014 		setbirthtime = 1;
3015 	VATTR_NULL(&vattr);
3016 	vattr.va_atime = ts[0];
3017 	vattr.va_mtime = ts[1];
3018 	if (setbirthtime)
3019 		vattr.va_birthtime = ts[1];
3020 	if (numtimes > 2)
3021 		vattr.va_birthtime = ts[2];
3022 	if (nullflag)
3023 		vattr.va_vaflags |= VA_UTIMES_NULL;
3024 #ifdef MAC
3025 	error = mac_vnode_check_setutimes(td->td_ucred, vp, vattr.va_atime,
3026 	    vattr.va_mtime);
3027 #endif
3028 	if (error == 0)
3029 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3030 	VOP_UNLOCK(vp, 0);
3031 	vn_finished_write(mp);
3032 	return (error);
3033 }
3034 
3035 /*
3036  * Set the access and modification times of a file.
3037  */
3038 #ifndef _SYS_SYSPROTO_H_
3039 struct utimes_args {
3040 	char	*path;
3041 	struct	timeval *tptr;
3042 };
3043 #endif
3044 int
3045 sys_utimes(td, uap)
3046 	struct thread *td;
3047 	register struct utimes_args /* {
3048 		char *path;
3049 		struct timeval *tptr;
3050 	} */ *uap;
3051 {
3052 
3053 	return (kern_utimesat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
3054 	    uap->tptr, UIO_USERSPACE));
3055 }
3056 
3057 #ifndef _SYS_SYSPROTO_H_
3058 struct futimesat_args {
3059 	int fd;
3060 	const char * path;
3061 	const struct timeval * times;
3062 };
3063 #endif
3064 int
3065 sys_futimesat(struct thread *td, struct futimesat_args *uap)
3066 {
3067 
3068 	return (kern_utimesat(td, uap->fd, uap->path, UIO_USERSPACE,
3069 	    uap->times, UIO_USERSPACE));
3070 }
3071 
3072 int
3073 kern_utimesat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
3074     struct timeval *tptr, enum uio_seg tptrseg)
3075 {
3076 	struct nameidata nd;
3077 	struct timespec ts[2];
3078 	cap_rights_t rights;
3079 	int error;
3080 
3081 	if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3082 		return (error);
3083 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, fd,
3084 	    cap_rights_init(&rights, CAP_FUTIMES), td);
3085 
3086 	if ((error = namei(&nd)) != 0)
3087 		return (error);
3088 	NDFREE(&nd, NDF_ONLY_PNBUF);
3089 	error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3090 	vrele(nd.ni_vp);
3091 	return (error);
3092 }
3093 
3094 /*
3095  * Set the access and modification times of a file.
3096  */
3097 #ifndef _SYS_SYSPROTO_H_
3098 struct lutimes_args {
3099 	char	*path;
3100 	struct	timeval *tptr;
3101 };
3102 #endif
3103 int
3104 sys_lutimes(td, uap)
3105 	struct thread *td;
3106 	register struct lutimes_args /* {
3107 		char *path;
3108 		struct timeval *tptr;
3109 	} */ *uap;
3110 {
3111 
3112 	return (kern_lutimes(td, uap->path, UIO_USERSPACE, uap->tptr,
3113 	    UIO_USERSPACE));
3114 }
3115 
3116 int
3117 kern_lutimes(struct thread *td, char *path, enum uio_seg pathseg,
3118     struct timeval *tptr, enum uio_seg tptrseg)
3119 {
3120 	struct timespec ts[2];
3121 	struct nameidata nd;
3122 	int error;
3123 
3124 	if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3125 		return (error);
3126 	NDINIT(&nd, LOOKUP, NOFOLLOW | AUDITVNODE1, pathseg, path, td);
3127 	if ((error = namei(&nd)) != 0)
3128 		return (error);
3129 	NDFREE(&nd, NDF_ONLY_PNBUF);
3130 	error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3131 	vrele(nd.ni_vp);
3132 	return (error);
3133 }
3134 
3135 /*
3136  * Set the access and modification times of a file.
3137  */
3138 #ifndef _SYS_SYSPROTO_H_
3139 struct futimes_args {
3140 	int	fd;
3141 	struct	timeval *tptr;
3142 };
3143 #endif
3144 int
3145 sys_futimes(td, uap)
3146 	struct thread *td;
3147 	register struct futimes_args /* {
3148 		int  fd;
3149 		struct timeval *tptr;
3150 	} */ *uap;
3151 {
3152 
3153 	return (kern_futimes(td, uap->fd, uap->tptr, UIO_USERSPACE));
3154 }
3155 
3156 int
3157 kern_futimes(struct thread *td, int fd, struct timeval *tptr,
3158     enum uio_seg tptrseg)
3159 {
3160 	struct timespec ts[2];
3161 	struct file *fp;
3162 	cap_rights_t rights;
3163 	int error;
3164 
3165 	AUDIT_ARG_FD(fd);
3166 	error = getutimes(tptr, tptrseg, ts);
3167 	if (error != 0)
3168 		return (error);
3169 	error = getvnode(td, fd, cap_rights_init(&rights, CAP_FUTIMES), &fp);
3170 	if (error != 0)
3171 		return (error);
3172 #ifdef AUDIT
3173 	vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
3174 	AUDIT_ARG_VNODE1(fp->f_vnode);
3175 	VOP_UNLOCK(fp->f_vnode, 0);
3176 #endif
3177 	error = setutimes(td, fp->f_vnode, ts, 2, tptr == NULL);
3178 	fdrop(fp, td);
3179 	return (error);
3180 }
3181 
3182 int
3183 sys_futimens(struct thread *td, struct futimens_args *uap)
3184 {
3185 
3186 	return (kern_futimens(td, uap->fd, uap->times, UIO_USERSPACE));
3187 }
3188 
3189 int
3190 kern_futimens(struct thread *td, int fd, struct timespec *tptr,
3191     enum uio_seg tptrseg)
3192 {
3193 	struct timespec ts[2];
3194 	struct file *fp;
3195 	cap_rights_t rights;
3196 	int error, flags;
3197 
3198 	AUDIT_ARG_FD(fd);
3199 	error = getutimens(tptr, tptrseg, ts, &flags);
3200 	if (error != 0)
3201 		return (error);
3202 	if (flags & UTIMENS_EXIT)
3203 		return (0);
3204 	error = getvnode(td, fd, cap_rights_init(&rights, CAP_FUTIMES), &fp);
3205 	if (error != 0)
3206 		return (error);
3207 #ifdef AUDIT
3208 	vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
3209 	AUDIT_ARG_VNODE1(fp->f_vnode);
3210 	VOP_UNLOCK(fp->f_vnode, 0);
3211 #endif
3212 	error = setutimes(td, fp->f_vnode, ts, 2, flags & UTIMENS_NULL);
3213 	fdrop(fp, td);
3214 	return (error);
3215 }
3216 
3217 int
3218 sys_utimensat(struct thread *td, struct utimensat_args *uap)
3219 {
3220 
3221 	return (kern_utimensat(td, uap->fd, uap->path, UIO_USERSPACE,
3222 	    uap->times, UIO_USERSPACE, uap->flag));
3223 }
3224 
3225 int
3226 kern_utimensat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
3227     struct timespec *tptr, enum uio_seg tptrseg, int flag)
3228 {
3229 	struct nameidata nd;
3230 	struct timespec ts[2];
3231 	cap_rights_t rights;
3232 	int error, flags;
3233 
3234 	if (flag & ~AT_SYMLINK_NOFOLLOW)
3235 		return (EINVAL);
3236 
3237 	if ((error = getutimens(tptr, tptrseg, ts, &flags)) != 0)
3238 		return (error);
3239 	NDINIT_ATRIGHTS(&nd, LOOKUP, ((flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW :
3240 	    FOLLOW) | AUDITVNODE1, pathseg, path, fd,
3241 	    cap_rights_init(&rights, CAP_FUTIMES), td);
3242 	if ((error = namei(&nd)) != 0)
3243 		return (error);
3244 	/*
3245 	 * We are allowed to call namei() regardless of 2xUTIME_OMIT.
3246 	 * POSIX states:
3247 	 * "If both tv_nsec fields are UTIME_OMIT... EACCESS may be detected."
3248 	 * "Search permission is denied by a component of the path prefix."
3249 	 */
3250 	NDFREE(&nd, NDF_ONLY_PNBUF);
3251 	if ((flags & UTIMENS_EXIT) == 0)
3252 		error = setutimes(td, nd.ni_vp, ts, 2, flags & UTIMENS_NULL);
3253 	vrele(nd.ni_vp);
3254 	return (error);
3255 }
3256 
3257 /*
3258  * Truncate a file given its path name.
3259  */
3260 #ifndef _SYS_SYSPROTO_H_
3261 struct truncate_args {
3262 	char	*path;
3263 	int	pad;
3264 	off_t	length;
3265 };
3266 #endif
3267 int
3268 sys_truncate(td, uap)
3269 	struct thread *td;
3270 	register struct truncate_args /* {
3271 		char *path;
3272 		int pad;
3273 		off_t length;
3274 	} */ *uap;
3275 {
3276 
3277 	return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3278 }
3279 
3280 int
3281 kern_truncate(struct thread *td, char *path, enum uio_seg pathseg, off_t length)
3282 {
3283 	struct mount *mp;
3284 	struct vnode *vp;
3285 	void *rl_cookie;
3286 	struct vattr vattr;
3287 	struct nameidata nd;
3288 	int error;
3289 
3290 	if (length < 0)
3291 		return(EINVAL);
3292 	NDINIT(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, td);
3293 	if ((error = namei(&nd)) != 0)
3294 		return (error);
3295 	vp = nd.ni_vp;
3296 	rl_cookie = vn_rangelock_wlock(vp, 0, OFF_MAX);
3297 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) {
3298 		vn_rangelock_unlock(vp, rl_cookie);
3299 		vrele(vp);
3300 		return (error);
3301 	}
3302 	NDFREE(&nd, NDF_ONLY_PNBUF);
3303 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3304 	if (vp->v_type == VDIR)
3305 		error = EISDIR;
3306 #ifdef MAC
3307 	else if ((error = mac_vnode_check_write(td->td_ucred, NOCRED, vp))) {
3308 	}
3309 #endif
3310 	else if ((error = vn_writechk(vp)) == 0 &&
3311 	    (error = VOP_ACCESS(vp, VWRITE, td->td_ucred, td)) == 0) {
3312 		VATTR_NULL(&vattr);
3313 		vattr.va_size = length;
3314 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3315 	}
3316 	VOP_UNLOCK(vp, 0);
3317 	vn_finished_write(mp);
3318 	vn_rangelock_unlock(vp, rl_cookie);
3319 	vrele(vp);
3320 	return (error);
3321 }
3322 
3323 #if defined(COMPAT_43)
3324 /*
3325  * Truncate a file given its path name.
3326  */
3327 #ifndef _SYS_SYSPROTO_H_
3328 struct otruncate_args {
3329 	char	*path;
3330 	long	length;
3331 };
3332 #endif
3333 int
3334 otruncate(td, uap)
3335 	struct thread *td;
3336 	register struct otruncate_args /* {
3337 		char *path;
3338 		long length;
3339 	} */ *uap;
3340 {
3341 	struct truncate_args /* {
3342 		char *path;
3343 		int pad;
3344 		off_t length;
3345 	} */ nuap;
3346 
3347 	nuap.path = uap->path;
3348 	nuap.length = uap->length;
3349 	return (sys_truncate(td, &nuap));
3350 }
3351 #endif /* COMPAT_43 */
3352 
3353 #if defined(COMPAT_FREEBSD6)
3354 /* Versions with the pad argument */
3355 int
3356 freebsd6_truncate(struct thread *td, struct freebsd6_truncate_args *uap)
3357 {
3358 	struct truncate_args ouap;
3359 
3360 	ouap.path = uap->path;
3361 	ouap.length = uap->length;
3362 	return (sys_truncate(td, &ouap));
3363 }
3364 
3365 int
3366 freebsd6_ftruncate(struct thread *td, struct freebsd6_ftruncate_args *uap)
3367 {
3368 	struct ftruncate_args ouap;
3369 
3370 	ouap.fd = uap->fd;
3371 	ouap.length = uap->length;
3372 	return (sys_ftruncate(td, &ouap));
3373 }
3374 #endif
3375 
3376 int
3377 kern_fsync(struct thread *td, int fd, bool fullsync)
3378 {
3379 	struct vnode *vp;
3380 	struct mount *mp;
3381 	struct file *fp;
3382 	cap_rights_t rights;
3383 	int error, lock_flags;
3384 
3385 	AUDIT_ARG_FD(fd);
3386 	error = getvnode(td, fd, cap_rights_init(&rights, CAP_FSYNC), &fp);
3387 	if (error != 0)
3388 		return (error);
3389 	vp = fp->f_vnode;
3390 #if 0
3391 	if (!fullsync)
3392 		/* XXXKIB: compete outstanding aio writes */;
3393 #endif
3394 	error = vn_start_write(vp, &mp, V_WAIT | PCATCH);
3395 	if (error != 0)
3396 		goto drop;
3397 	if (MNT_SHARED_WRITES(mp) ||
3398 	    ((mp == NULL) && MNT_SHARED_WRITES(vp->v_mount))) {
3399 		lock_flags = LK_SHARED;
3400 	} else {
3401 		lock_flags = LK_EXCLUSIVE;
3402 	}
3403 	vn_lock(vp, lock_flags | LK_RETRY);
3404 	AUDIT_ARG_VNODE1(vp);
3405 	if (vp->v_object != NULL) {
3406 		VM_OBJECT_WLOCK(vp->v_object);
3407 		vm_object_page_clean(vp->v_object, 0, 0, 0);
3408 		VM_OBJECT_WUNLOCK(vp->v_object);
3409 	}
3410 	error = fullsync ? VOP_FSYNC(vp, MNT_WAIT, td) : VOP_FDATASYNC(vp, td);
3411 	VOP_UNLOCK(vp, 0);
3412 	vn_finished_write(mp);
3413 drop:
3414 	fdrop(fp, td);
3415 	return (error);
3416 }
3417 
3418 /*
3419  * Sync an open file.
3420  */
3421 #ifndef _SYS_SYSPROTO_H_
3422 struct fsync_args {
3423 	int	fd;
3424 };
3425 #endif
3426 int
3427 sys_fsync(struct thread *td, struct fsync_args *uap)
3428 {
3429 
3430 	return (kern_fsync(td, uap->fd, true));
3431 }
3432 
3433 int
3434 sys_fdatasync(struct thread *td, struct fdatasync_args *uap)
3435 {
3436 
3437 	return (kern_fsync(td, uap->fd, false));
3438 }
3439 
3440 /*
3441  * Rename files.  Source and destination must either both be directories, or
3442  * both not be directories.  If target is a directory, it must be empty.
3443  */
3444 #ifndef _SYS_SYSPROTO_H_
3445 struct rename_args {
3446 	char	*from;
3447 	char	*to;
3448 };
3449 #endif
3450 int
3451 sys_rename(td, uap)
3452 	struct thread *td;
3453 	register struct rename_args /* {
3454 		char *from;
3455 		char *to;
3456 	} */ *uap;
3457 {
3458 
3459 	return (kern_renameat(td, AT_FDCWD, uap->from, AT_FDCWD,
3460 	    uap->to, UIO_USERSPACE));
3461 }
3462 
3463 #ifndef _SYS_SYSPROTO_H_
3464 struct renameat_args {
3465 	int	oldfd;
3466 	char	*old;
3467 	int	newfd;
3468 	char	*new;
3469 };
3470 #endif
3471 int
3472 sys_renameat(struct thread *td, struct renameat_args *uap)
3473 {
3474 
3475 	return (kern_renameat(td, uap->oldfd, uap->old, uap->newfd, uap->new,
3476 	    UIO_USERSPACE));
3477 }
3478 
3479 int
3480 kern_renameat(struct thread *td, int oldfd, char *old, int newfd, char *new,
3481     enum uio_seg pathseg)
3482 {
3483 	struct mount *mp = NULL;
3484 	struct vnode *tvp, *fvp, *tdvp;
3485 	struct nameidata fromnd, tond;
3486 	cap_rights_t rights;
3487 	int error;
3488 
3489 again:
3490 	bwillwrite();
3491 #ifdef MAC
3492 	NDINIT_ATRIGHTS(&fromnd, DELETE, LOCKPARENT | LOCKLEAF | SAVESTART |
3493 	    AUDITVNODE1, pathseg, old, oldfd,
3494 	    cap_rights_init(&rights, CAP_RENAMEAT_SOURCE), td);
3495 #else
3496 	NDINIT_ATRIGHTS(&fromnd, DELETE, WANTPARENT | SAVESTART | AUDITVNODE1,
3497 	    pathseg, old, oldfd,
3498 	    cap_rights_init(&rights, CAP_RENAMEAT_SOURCE), td);
3499 #endif
3500 
3501 	if ((error = namei(&fromnd)) != 0)
3502 		return (error);
3503 #ifdef MAC
3504 	error = mac_vnode_check_rename_from(td->td_ucred, fromnd.ni_dvp,
3505 	    fromnd.ni_vp, &fromnd.ni_cnd);
3506 	VOP_UNLOCK(fromnd.ni_dvp, 0);
3507 	if (fromnd.ni_dvp != fromnd.ni_vp)
3508 		VOP_UNLOCK(fromnd.ni_vp, 0);
3509 #endif
3510 	fvp = fromnd.ni_vp;
3511 	NDINIT_ATRIGHTS(&tond, RENAME, LOCKPARENT | LOCKLEAF | NOCACHE |
3512 	    SAVESTART | AUDITVNODE2, pathseg, new, newfd,
3513 	    cap_rights_init(&rights, CAP_RENAMEAT_TARGET), td);
3514 	if (fromnd.ni_vp->v_type == VDIR)
3515 		tond.ni_cnd.cn_flags |= WILLBEDIR;
3516 	if ((error = namei(&tond)) != 0) {
3517 		/* Translate error code for rename("dir1", "dir2/."). */
3518 		if (error == EISDIR && fvp->v_type == VDIR)
3519 			error = EINVAL;
3520 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3521 		vrele(fromnd.ni_dvp);
3522 		vrele(fvp);
3523 		goto out1;
3524 	}
3525 	tdvp = tond.ni_dvp;
3526 	tvp = tond.ni_vp;
3527 	error = vn_start_write(fvp, &mp, V_NOWAIT);
3528 	if (error != 0) {
3529 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3530 		NDFREE(&tond, NDF_ONLY_PNBUF);
3531 		if (tvp != NULL)
3532 			vput(tvp);
3533 		if (tdvp == tvp)
3534 			vrele(tdvp);
3535 		else
3536 			vput(tdvp);
3537 		vrele(fromnd.ni_dvp);
3538 		vrele(fvp);
3539 		vrele(tond.ni_startdir);
3540 		if (fromnd.ni_startdir != NULL)
3541 			vrele(fromnd.ni_startdir);
3542 		error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH);
3543 		if (error != 0)
3544 			return (error);
3545 		goto again;
3546 	}
3547 	if (tvp != NULL) {
3548 		if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
3549 			error = ENOTDIR;
3550 			goto out;
3551 		} else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
3552 			error = EISDIR;
3553 			goto out;
3554 		}
3555 #ifdef CAPABILITIES
3556 		if (newfd != AT_FDCWD) {
3557 			/*
3558 			 * If the target already exists we require CAP_UNLINKAT
3559 			 * from 'newfd'.
3560 			 */
3561 			error = cap_check(&tond.ni_filecaps.fc_rights,
3562 			    cap_rights_init(&rights, CAP_UNLINKAT));
3563 			if (error != 0)
3564 				goto out;
3565 		}
3566 #endif
3567 	}
3568 	if (fvp == tdvp) {
3569 		error = EINVAL;
3570 		goto out;
3571 	}
3572 	/*
3573 	 * If the source is the same as the destination (that is, if they
3574 	 * are links to the same vnode), then there is nothing to do.
3575 	 */
3576 	if (fvp == tvp)
3577 		error = -1;
3578 #ifdef MAC
3579 	else
3580 		error = mac_vnode_check_rename_to(td->td_ucred, tdvp,
3581 		    tond.ni_vp, fromnd.ni_dvp == tdvp, &tond.ni_cnd);
3582 #endif
3583 out:
3584 	if (error == 0) {
3585 		error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
3586 		    tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
3587 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3588 		NDFREE(&tond, NDF_ONLY_PNBUF);
3589 	} else {
3590 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3591 		NDFREE(&tond, NDF_ONLY_PNBUF);
3592 		if (tvp != NULL)
3593 			vput(tvp);
3594 		if (tdvp == tvp)
3595 			vrele(tdvp);
3596 		else
3597 			vput(tdvp);
3598 		vrele(fromnd.ni_dvp);
3599 		vrele(fvp);
3600 	}
3601 	vrele(tond.ni_startdir);
3602 	vn_finished_write(mp);
3603 out1:
3604 	if (fromnd.ni_startdir)
3605 		vrele(fromnd.ni_startdir);
3606 	if (error == -1)
3607 		return (0);
3608 	return (error);
3609 }
3610 
3611 /*
3612  * Make a directory file.
3613  */
3614 #ifndef _SYS_SYSPROTO_H_
3615 struct mkdir_args {
3616 	char	*path;
3617 	int	mode;
3618 };
3619 #endif
3620 int
3621 sys_mkdir(td, uap)
3622 	struct thread *td;
3623 	register struct mkdir_args /* {
3624 		char *path;
3625 		int mode;
3626 	} */ *uap;
3627 {
3628 
3629 	return (kern_mkdirat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
3630 	    uap->mode));
3631 }
3632 
3633 #ifndef _SYS_SYSPROTO_H_
3634 struct mkdirat_args {
3635 	int	fd;
3636 	char	*path;
3637 	mode_t	mode;
3638 };
3639 #endif
3640 int
3641 sys_mkdirat(struct thread *td, struct mkdirat_args *uap)
3642 {
3643 
3644 	return (kern_mkdirat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode));
3645 }
3646 
3647 int
3648 kern_mkdirat(struct thread *td, int fd, char *path, enum uio_seg segflg,
3649     int mode)
3650 {
3651 	struct mount *mp;
3652 	struct vnode *vp;
3653 	struct vattr vattr;
3654 	struct nameidata nd;
3655 	cap_rights_t rights;
3656 	int error;
3657 
3658 	AUDIT_ARG_MODE(mode);
3659 restart:
3660 	bwillwrite();
3661 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
3662 	    NOCACHE, segflg, path, fd, cap_rights_init(&rights, CAP_MKDIRAT),
3663 	    td);
3664 	nd.ni_cnd.cn_flags |= WILLBEDIR;
3665 	if ((error = namei(&nd)) != 0)
3666 		return (error);
3667 	vp = nd.ni_vp;
3668 	if (vp != NULL) {
3669 		NDFREE(&nd, NDF_ONLY_PNBUF);
3670 		/*
3671 		 * XXX namei called with LOCKPARENT but not LOCKLEAF has
3672 		 * the strange behaviour of leaving the vnode unlocked
3673 		 * if the target is the same vnode as the parent.
3674 		 */
3675 		if (vp == nd.ni_dvp)
3676 			vrele(nd.ni_dvp);
3677 		else
3678 			vput(nd.ni_dvp);
3679 		vrele(vp);
3680 		return (EEXIST);
3681 	}
3682 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3683 		NDFREE(&nd, NDF_ONLY_PNBUF);
3684 		vput(nd.ni_dvp);
3685 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3686 			return (error);
3687 		goto restart;
3688 	}
3689 	VATTR_NULL(&vattr);
3690 	vattr.va_type = VDIR;
3691 	vattr.va_mode = (mode & ACCESSPERMS) &~ td->td_proc->p_fd->fd_cmask;
3692 #ifdef MAC
3693 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
3694 	    &vattr);
3695 	if (error != 0)
3696 		goto out;
3697 #endif
3698 	error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
3699 #ifdef MAC
3700 out:
3701 #endif
3702 	NDFREE(&nd, NDF_ONLY_PNBUF);
3703 	vput(nd.ni_dvp);
3704 	if (error == 0)
3705 		vput(nd.ni_vp);
3706 	vn_finished_write(mp);
3707 	return (error);
3708 }
3709 
3710 /*
3711  * Remove a directory file.
3712  */
3713 #ifndef _SYS_SYSPROTO_H_
3714 struct rmdir_args {
3715 	char	*path;
3716 };
3717 #endif
3718 int
3719 sys_rmdir(td, uap)
3720 	struct thread *td;
3721 	struct rmdir_args /* {
3722 		char *path;
3723 	} */ *uap;
3724 {
3725 
3726 	return (kern_rmdirat(td, AT_FDCWD, uap->path, UIO_USERSPACE));
3727 }
3728 
3729 int
3730 kern_rmdirat(struct thread *td, int fd, char *path, enum uio_seg pathseg)
3731 {
3732 	struct mount *mp;
3733 	struct vnode *vp;
3734 	struct nameidata nd;
3735 	cap_rights_t rights;
3736 	int error;
3737 
3738 restart:
3739 	bwillwrite();
3740 	NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1,
3741 	    pathseg, path, fd, cap_rights_init(&rights, CAP_UNLINKAT), td);
3742 	if ((error = namei(&nd)) != 0)
3743 		return (error);
3744 	vp = nd.ni_vp;
3745 	if (vp->v_type != VDIR) {
3746 		error = ENOTDIR;
3747 		goto out;
3748 	}
3749 	/*
3750 	 * No rmdir "." please.
3751 	 */
3752 	if (nd.ni_dvp == vp) {
3753 		error = EINVAL;
3754 		goto out;
3755 	}
3756 	/*
3757 	 * The root of a mounted filesystem cannot be deleted.
3758 	 */
3759 	if (vp->v_vflag & VV_ROOT) {
3760 		error = EBUSY;
3761 		goto out;
3762 	}
3763 #ifdef MAC
3764 	error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
3765 	    &nd.ni_cnd);
3766 	if (error != 0)
3767 		goto out;
3768 #endif
3769 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3770 		NDFREE(&nd, NDF_ONLY_PNBUF);
3771 		vput(vp);
3772 		if (nd.ni_dvp == vp)
3773 			vrele(nd.ni_dvp);
3774 		else
3775 			vput(nd.ni_dvp);
3776 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3777 			return (error);
3778 		goto restart;
3779 	}
3780 	vfs_notify_upper(vp, VFS_NOTIFY_UPPER_UNLINK);
3781 	error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
3782 	vn_finished_write(mp);
3783 out:
3784 	NDFREE(&nd, NDF_ONLY_PNBUF);
3785 	vput(vp);
3786 	if (nd.ni_dvp == vp)
3787 		vrele(nd.ni_dvp);
3788 	else
3789 		vput(nd.ni_dvp);
3790 	return (error);
3791 }
3792 
3793 #ifdef COMPAT_43
3794 /*
3795  * Read a block of directory entries in a filesystem independent format.
3796  */
3797 #ifndef _SYS_SYSPROTO_H_
3798 struct ogetdirentries_args {
3799 	int	fd;
3800 	char	*buf;
3801 	u_int	count;
3802 	long	*basep;
3803 };
3804 #endif
3805 int
3806 ogetdirentries(struct thread *td, struct ogetdirentries_args *uap)
3807 {
3808 	long loff;
3809 	int error;
3810 
3811 	error = kern_ogetdirentries(td, uap, &loff);
3812 	if (error == 0)
3813 		error = copyout(&loff, uap->basep, sizeof(long));
3814 	return (error);
3815 }
3816 
3817 int
3818 kern_ogetdirentries(struct thread *td, struct ogetdirentries_args *uap,
3819     long *ploff)
3820 {
3821 	struct vnode *vp;
3822 	struct file *fp;
3823 	struct uio auio, kuio;
3824 	struct iovec aiov, kiov;
3825 	struct dirent *dp, *edp;
3826 	cap_rights_t rights;
3827 	caddr_t dirbuf;
3828 	int error, eofflag, readcnt;
3829 	long loff;
3830 	off_t foffset;
3831 
3832 	/* XXX arbitrary sanity limit on `count'. */
3833 	if (uap->count > 64 * 1024)
3834 		return (EINVAL);
3835 	error = getvnode(td, uap->fd, cap_rights_init(&rights, CAP_READ), &fp);
3836 	if (error != 0)
3837 		return (error);
3838 	if ((fp->f_flag & FREAD) == 0) {
3839 		fdrop(fp, td);
3840 		return (EBADF);
3841 	}
3842 	vp = fp->f_vnode;
3843 	foffset = foffset_lock(fp, 0);
3844 unionread:
3845 	if (vp->v_type != VDIR) {
3846 		foffset_unlock(fp, foffset, 0);
3847 		fdrop(fp, td);
3848 		return (EINVAL);
3849 	}
3850 	aiov.iov_base = uap->buf;
3851 	aiov.iov_len = uap->count;
3852 	auio.uio_iov = &aiov;
3853 	auio.uio_iovcnt = 1;
3854 	auio.uio_rw = UIO_READ;
3855 	auio.uio_segflg = UIO_USERSPACE;
3856 	auio.uio_td = td;
3857 	auio.uio_resid = uap->count;
3858 	vn_lock(vp, LK_SHARED | LK_RETRY);
3859 	loff = auio.uio_offset = foffset;
3860 #ifdef MAC
3861 	error = mac_vnode_check_readdir(td->td_ucred, vp);
3862 	if (error != 0) {
3863 		VOP_UNLOCK(vp, 0);
3864 		foffset_unlock(fp, foffset, FOF_NOUPDATE);
3865 		fdrop(fp, td);
3866 		return (error);
3867 	}
3868 #endif
3869 #	if (BYTE_ORDER != LITTLE_ENDIAN)
3870 		if (vp->v_mount->mnt_maxsymlinklen <= 0) {
3871 			error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag,
3872 			    NULL, NULL);
3873 			foffset = auio.uio_offset;
3874 		} else
3875 #	endif
3876 	{
3877 		kuio = auio;
3878 		kuio.uio_iov = &kiov;
3879 		kuio.uio_segflg = UIO_SYSSPACE;
3880 		kiov.iov_len = uap->count;
3881 		dirbuf = malloc(uap->count, M_TEMP, M_WAITOK);
3882 		kiov.iov_base = dirbuf;
3883 		error = VOP_READDIR(vp, &kuio, fp->f_cred, &eofflag,
3884 			    NULL, NULL);
3885 		foffset = kuio.uio_offset;
3886 		if (error == 0) {
3887 			readcnt = uap->count - kuio.uio_resid;
3888 			edp = (struct dirent *)&dirbuf[readcnt];
3889 			for (dp = (struct dirent *)dirbuf; dp < edp; ) {
3890 #				if (BYTE_ORDER == LITTLE_ENDIAN)
3891 					/*
3892 					 * The expected low byte of
3893 					 * dp->d_namlen is our dp->d_type.
3894 					 * The high MBZ byte of dp->d_namlen
3895 					 * is our dp->d_namlen.
3896 					 */
3897 					dp->d_type = dp->d_namlen;
3898 					dp->d_namlen = 0;
3899 #				else
3900 					/*
3901 					 * The dp->d_type is the high byte
3902 					 * of the expected dp->d_namlen,
3903 					 * so must be zero'ed.
3904 					 */
3905 					dp->d_type = 0;
3906 #				endif
3907 				if (dp->d_reclen > 0) {
3908 					dp = (struct dirent *)
3909 					    ((char *)dp + dp->d_reclen);
3910 				} else {
3911 					error = EIO;
3912 					break;
3913 				}
3914 			}
3915 			if (dp >= edp)
3916 				error = uiomove(dirbuf, readcnt, &auio);
3917 		}
3918 		free(dirbuf, M_TEMP);
3919 	}
3920 	if (error != 0) {
3921 		VOP_UNLOCK(vp, 0);
3922 		foffset_unlock(fp, foffset, 0);
3923 		fdrop(fp, td);
3924 		return (error);
3925 	}
3926 	if (uap->count == auio.uio_resid &&
3927 	    (vp->v_vflag & VV_ROOT) &&
3928 	    (vp->v_mount->mnt_flag & MNT_UNION)) {
3929 		struct vnode *tvp = vp;
3930 		vp = vp->v_mount->mnt_vnodecovered;
3931 		VREF(vp);
3932 		fp->f_vnode = vp;
3933 		fp->f_data = vp;
3934 		foffset = 0;
3935 		vput(tvp);
3936 		goto unionread;
3937 	}
3938 	VOP_UNLOCK(vp, 0);
3939 	foffset_unlock(fp, foffset, 0);
3940 	fdrop(fp, td);
3941 	td->td_retval[0] = uap->count - auio.uio_resid;
3942 	if (error == 0)
3943 		*ploff = loff;
3944 	return (error);
3945 }
3946 #endif /* COMPAT_43 */
3947 
3948 /*
3949  * Read a block of directory entries in a filesystem independent format.
3950  */
3951 #ifndef _SYS_SYSPROTO_H_
3952 struct getdirentries_args {
3953 	int	fd;
3954 	char	*buf;
3955 	u_int	count;
3956 	long	*basep;
3957 };
3958 #endif
3959 int
3960 sys_getdirentries(td, uap)
3961 	struct thread *td;
3962 	register struct getdirentries_args /* {
3963 		int fd;
3964 		char *buf;
3965 		u_int count;
3966 		long *basep;
3967 	} */ *uap;
3968 {
3969 	long base;
3970 	int error;
3971 
3972 	error = kern_getdirentries(td, uap->fd, uap->buf, uap->count, &base,
3973 	    NULL, UIO_USERSPACE);
3974 	if (error != 0)
3975 		return (error);
3976 	if (uap->basep != NULL)
3977 		error = copyout(&base, uap->basep, sizeof(long));
3978 	return (error);
3979 }
3980 
3981 int
3982 kern_getdirentries(struct thread *td, int fd, char *buf, u_int count,
3983     long *basep, ssize_t *residp, enum uio_seg bufseg)
3984 {
3985 	struct vnode *vp;
3986 	struct file *fp;
3987 	struct uio auio;
3988 	struct iovec aiov;
3989 	cap_rights_t rights;
3990 	long loff;
3991 	int error, eofflag;
3992 	off_t foffset;
3993 
3994 	AUDIT_ARG_FD(fd);
3995 	if (count > IOSIZE_MAX)
3996 		return (EINVAL);
3997 	auio.uio_resid = count;
3998 	error = getvnode(td, fd, cap_rights_init(&rights, CAP_READ), &fp);
3999 	if (error != 0)
4000 		return (error);
4001 	if ((fp->f_flag & FREAD) == 0) {
4002 		fdrop(fp, td);
4003 		return (EBADF);
4004 	}
4005 	vp = fp->f_vnode;
4006 	foffset = foffset_lock(fp, 0);
4007 unionread:
4008 	if (vp->v_type != VDIR) {
4009 		error = EINVAL;
4010 		goto fail;
4011 	}
4012 	aiov.iov_base = buf;
4013 	aiov.iov_len = count;
4014 	auio.uio_iov = &aiov;
4015 	auio.uio_iovcnt = 1;
4016 	auio.uio_rw = UIO_READ;
4017 	auio.uio_segflg = bufseg;
4018 	auio.uio_td = td;
4019 	vn_lock(vp, LK_SHARED | LK_RETRY);
4020 	AUDIT_ARG_VNODE1(vp);
4021 	loff = auio.uio_offset = foffset;
4022 #ifdef MAC
4023 	error = mac_vnode_check_readdir(td->td_ucred, vp);
4024 	if (error == 0)
4025 #endif
4026 		error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, NULL,
4027 		    NULL);
4028 	foffset = auio.uio_offset;
4029 	if (error != 0) {
4030 		VOP_UNLOCK(vp, 0);
4031 		goto fail;
4032 	}
4033 	if (count == auio.uio_resid &&
4034 	    (vp->v_vflag & VV_ROOT) &&
4035 	    (vp->v_mount->mnt_flag & MNT_UNION)) {
4036 		struct vnode *tvp = vp;
4037 
4038 		vp = vp->v_mount->mnt_vnodecovered;
4039 		VREF(vp);
4040 		fp->f_vnode = vp;
4041 		fp->f_data = vp;
4042 		foffset = 0;
4043 		vput(tvp);
4044 		goto unionread;
4045 	}
4046 	VOP_UNLOCK(vp, 0);
4047 	*basep = loff;
4048 	if (residp != NULL)
4049 		*residp = auio.uio_resid;
4050 	td->td_retval[0] = count - auio.uio_resid;
4051 fail:
4052 	foffset_unlock(fp, foffset, 0);
4053 	fdrop(fp, td);
4054 	return (error);
4055 }
4056 
4057 #ifndef _SYS_SYSPROTO_H_
4058 struct getdents_args {
4059 	int fd;
4060 	char *buf;
4061 	size_t count;
4062 };
4063 #endif
4064 int
4065 sys_getdents(td, uap)
4066 	struct thread *td;
4067 	register struct getdents_args /* {
4068 		int fd;
4069 		char *buf;
4070 		u_int count;
4071 	} */ *uap;
4072 {
4073 	struct getdirentries_args ap;
4074 
4075 	ap.fd = uap->fd;
4076 	ap.buf = uap->buf;
4077 	ap.count = uap->count;
4078 	ap.basep = NULL;
4079 	return (sys_getdirentries(td, &ap));
4080 }
4081 
4082 /*
4083  * Set the mode mask for creation of filesystem nodes.
4084  */
4085 #ifndef _SYS_SYSPROTO_H_
4086 struct umask_args {
4087 	int	newmask;
4088 };
4089 #endif
4090 int
4091 sys_umask(td, uap)
4092 	struct thread *td;
4093 	struct umask_args /* {
4094 		int newmask;
4095 	} */ *uap;
4096 {
4097 	struct filedesc *fdp;
4098 
4099 	fdp = td->td_proc->p_fd;
4100 	FILEDESC_XLOCK(fdp);
4101 	td->td_retval[0] = fdp->fd_cmask;
4102 	fdp->fd_cmask = uap->newmask & ALLPERMS;
4103 	FILEDESC_XUNLOCK(fdp);
4104 	return (0);
4105 }
4106 
4107 /*
4108  * Void all references to file by ripping underlying filesystem away from
4109  * vnode.
4110  */
4111 #ifndef _SYS_SYSPROTO_H_
4112 struct revoke_args {
4113 	char	*path;
4114 };
4115 #endif
4116 int
4117 sys_revoke(td, uap)
4118 	struct thread *td;
4119 	register struct revoke_args /* {
4120 		char *path;
4121 	} */ *uap;
4122 {
4123 	struct vnode *vp;
4124 	struct vattr vattr;
4125 	struct nameidata nd;
4126 	int error;
4127 
4128 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
4129 	    uap->path, td);
4130 	if ((error = namei(&nd)) != 0)
4131 		return (error);
4132 	vp = nd.ni_vp;
4133 	NDFREE(&nd, NDF_ONLY_PNBUF);
4134 	if (vp->v_type != VCHR || vp->v_rdev == NULL) {
4135 		error = EINVAL;
4136 		goto out;
4137 	}
4138 #ifdef MAC
4139 	error = mac_vnode_check_revoke(td->td_ucred, vp);
4140 	if (error != 0)
4141 		goto out;
4142 #endif
4143 	error = VOP_GETATTR(vp, &vattr, td->td_ucred);
4144 	if (error != 0)
4145 		goto out;
4146 	if (td->td_ucred->cr_uid != vattr.va_uid) {
4147 		error = priv_check(td, PRIV_VFS_ADMIN);
4148 		if (error != 0)
4149 			goto out;
4150 	}
4151 	if (vcount(vp) > 1)
4152 		VOP_REVOKE(vp, REVOKEALL);
4153 out:
4154 	vput(vp);
4155 	return (error);
4156 }
4157 
4158 /*
4159  * Convert a user file descriptor to a kernel file entry and check that, if it
4160  * is a capability, the correct rights are present. A reference on the file
4161  * entry is held upon returning.
4162  */
4163 int
4164 getvnode(struct thread *td, int fd, cap_rights_t *rightsp, struct file **fpp)
4165 {
4166 	struct file *fp;
4167 	int error;
4168 
4169 	error = fget_unlocked(td->td_proc->p_fd, fd, rightsp, &fp, NULL);
4170 	if (error != 0)
4171 		return (error);
4172 
4173 	/*
4174 	 * The file could be not of the vnode type, or it may be not
4175 	 * yet fully initialized, in which case the f_vnode pointer
4176 	 * may be set, but f_ops is still badfileops.  E.g.,
4177 	 * devfs_open() transiently create such situation to
4178 	 * facilitate csw d_fdopen().
4179 	 *
4180 	 * Dupfdopen() handling in kern_openat() installs the
4181 	 * half-baked file into the process descriptor table, allowing
4182 	 * other thread to dereference it. Guard against the race by
4183 	 * checking f_ops.
4184 	 */
4185 	if (fp->f_vnode == NULL || fp->f_ops == &badfileops) {
4186 		fdrop(fp, td);
4187 		return (EINVAL);
4188 	}
4189 	*fpp = fp;
4190 	return (0);
4191 }
4192 
4193 
4194 /*
4195  * Get an (NFS) file handle.
4196  */
4197 #ifndef _SYS_SYSPROTO_H_
4198 struct lgetfh_args {
4199 	char	*fname;
4200 	fhandle_t *fhp;
4201 };
4202 #endif
4203 int
4204 sys_lgetfh(td, uap)
4205 	struct thread *td;
4206 	register struct lgetfh_args *uap;
4207 {
4208 	struct nameidata nd;
4209 	fhandle_t fh;
4210 	register struct vnode *vp;
4211 	int error;
4212 
4213 	error = priv_check(td, PRIV_VFS_GETFH);
4214 	if (error != 0)
4215 		return (error);
4216 	NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
4217 	    uap->fname, td);
4218 	error = namei(&nd);
4219 	if (error != 0)
4220 		return (error);
4221 	NDFREE(&nd, NDF_ONLY_PNBUF);
4222 	vp = nd.ni_vp;
4223 	bzero(&fh, sizeof(fh));
4224 	fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
4225 	error = VOP_VPTOFH(vp, &fh.fh_fid);
4226 	vput(vp);
4227 	if (error == 0)
4228 		error = copyout(&fh, uap->fhp, sizeof (fh));
4229 	return (error);
4230 }
4231 
4232 #ifndef _SYS_SYSPROTO_H_
4233 struct getfh_args {
4234 	char	*fname;
4235 	fhandle_t *fhp;
4236 };
4237 #endif
4238 int
4239 sys_getfh(td, uap)
4240 	struct thread *td;
4241 	register struct getfh_args *uap;
4242 {
4243 	struct nameidata nd;
4244 	fhandle_t fh;
4245 	register struct vnode *vp;
4246 	int error;
4247 
4248 	error = priv_check(td, PRIV_VFS_GETFH);
4249 	if (error != 0)
4250 		return (error);
4251 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
4252 	    uap->fname, td);
4253 	error = namei(&nd);
4254 	if (error != 0)
4255 		return (error);
4256 	NDFREE(&nd, NDF_ONLY_PNBUF);
4257 	vp = nd.ni_vp;
4258 	bzero(&fh, sizeof(fh));
4259 	fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
4260 	error = VOP_VPTOFH(vp, &fh.fh_fid);
4261 	vput(vp);
4262 	if (error == 0)
4263 		error = copyout(&fh, uap->fhp, sizeof (fh));
4264 	return (error);
4265 }
4266 
4267 /*
4268  * syscall for the rpc.lockd to use to translate a NFS file handle into an
4269  * open descriptor.
4270  *
4271  * warning: do not remove the priv_check() call or this becomes one giant
4272  * security hole.
4273  */
4274 #ifndef _SYS_SYSPROTO_H_
4275 struct fhopen_args {
4276 	const struct fhandle *u_fhp;
4277 	int flags;
4278 };
4279 #endif
4280 int
4281 sys_fhopen(td, uap)
4282 	struct thread *td;
4283 	struct fhopen_args /* {
4284 		const struct fhandle *u_fhp;
4285 		int flags;
4286 	} */ *uap;
4287 {
4288 	struct mount *mp;
4289 	struct vnode *vp;
4290 	struct fhandle fhp;
4291 	struct file *fp;
4292 	int fmode, error;
4293 	int indx;
4294 
4295 	error = priv_check(td, PRIV_VFS_FHOPEN);
4296 	if (error != 0)
4297 		return (error);
4298 	indx = -1;
4299 	fmode = FFLAGS(uap->flags);
4300 	/* why not allow a non-read/write open for our lockd? */
4301 	if (((fmode & (FREAD | FWRITE)) == 0) || (fmode & O_CREAT))
4302 		return (EINVAL);
4303 	error = copyin(uap->u_fhp, &fhp, sizeof(fhp));
4304 	if (error != 0)
4305 		return(error);
4306 	/* find the mount point */
4307 	mp = vfs_busyfs(&fhp.fh_fsid);
4308 	if (mp == NULL)
4309 		return (ESTALE);
4310 	/* now give me my vnode, it gets returned to me locked */
4311 	error = VFS_FHTOVP(mp, &fhp.fh_fid, LK_EXCLUSIVE, &vp);
4312 	vfs_unbusy(mp);
4313 	if (error != 0)
4314 		return (error);
4315 
4316 	error = falloc_noinstall(td, &fp);
4317 	if (error != 0) {
4318 		vput(vp);
4319 		return (error);
4320 	}
4321 	/*
4322 	 * An extra reference on `fp' has been held for us by
4323 	 * falloc_noinstall().
4324 	 */
4325 
4326 #ifdef INVARIANTS
4327 	td->td_dupfd = -1;
4328 #endif
4329 	error = vn_open_vnode(vp, fmode, td->td_ucred, td, fp);
4330 	if (error != 0) {
4331 		KASSERT(fp->f_ops == &badfileops,
4332 		    ("VOP_OPEN in fhopen() set f_ops"));
4333 		KASSERT(td->td_dupfd < 0,
4334 		    ("fhopen() encountered fdopen()"));
4335 
4336 		vput(vp);
4337 		goto bad;
4338 	}
4339 #ifdef INVARIANTS
4340 	td->td_dupfd = 0;
4341 #endif
4342 	fp->f_vnode = vp;
4343 	fp->f_seqcount = 1;
4344 	finit(fp, (fmode & FMASK) | (fp->f_flag & FHASLOCK), DTYPE_VNODE, vp,
4345 	    &vnops);
4346 	VOP_UNLOCK(vp, 0);
4347 	if ((fmode & O_TRUNC) != 0) {
4348 		error = fo_truncate(fp, 0, td->td_ucred, td);
4349 		if (error != 0)
4350 			goto bad;
4351 	}
4352 
4353 	error = finstall(td, fp, &indx, fmode, NULL);
4354 bad:
4355 	fdrop(fp, td);
4356 	td->td_retval[0] = indx;
4357 	return (error);
4358 }
4359 
4360 /*
4361  * Stat an (NFS) file handle.
4362  */
4363 #ifndef _SYS_SYSPROTO_H_
4364 struct fhstat_args {
4365 	struct fhandle *u_fhp;
4366 	struct stat *sb;
4367 };
4368 #endif
4369 int
4370 sys_fhstat(td, uap)
4371 	struct thread *td;
4372 	register struct fhstat_args /* {
4373 		struct fhandle *u_fhp;
4374 		struct stat *sb;
4375 	} */ *uap;
4376 {
4377 	struct stat sb;
4378 	struct fhandle fh;
4379 	int error;
4380 
4381 	error = copyin(uap->u_fhp, &fh, sizeof(fh));
4382 	if (error != 0)
4383 		return (error);
4384 	error = kern_fhstat(td, fh, &sb);
4385 	if (error == 0)
4386 		error = copyout(&sb, uap->sb, sizeof(sb));
4387 	return (error);
4388 }
4389 
4390 int
4391 kern_fhstat(struct thread *td, struct fhandle fh, struct stat *sb)
4392 {
4393 	struct mount *mp;
4394 	struct vnode *vp;
4395 	int error;
4396 
4397 	error = priv_check(td, PRIV_VFS_FHSTAT);
4398 	if (error != 0)
4399 		return (error);
4400 	if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4401 		return (ESTALE);
4402 	error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4403 	vfs_unbusy(mp);
4404 	if (error != 0)
4405 		return (error);
4406 	error = vn_stat(vp, sb, td->td_ucred, NOCRED, td);
4407 	vput(vp);
4408 	return (error);
4409 }
4410 
4411 /*
4412  * Implement fstatfs() for (NFS) file handles.
4413  */
4414 #ifndef _SYS_SYSPROTO_H_
4415 struct fhstatfs_args {
4416 	struct fhandle *u_fhp;
4417 	struct statfs *buf;
4418 };
4419 #endif
4420 int
4421 sys_fhstatfs(td, uap)
4422 	struct thread *td;
4423 	struct fhstatfs_args /* {
4424 		struct fhandle *u_fhp;
4425 		struct statfs *buf;
4426 	} */ *uap;
4427 {
4428 	struct statfs *sfp;
4429 	fhandle_t fh;
4430 	int error;
4431 
4432 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
4433 	if (error != 0)
4434 		return (error);
4435 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
4436 	error = kern_fhstatfs(td, fh, sfp);
4437 	if (error == 0)
4438 		error = copyout(sfp, uap->buf, sizeof(*sfp));
4439 	free(sfp, M_STATFS);
4440 	return (error);
4441 }
4442 
4443 int
4444 kern_fhstatfs(struct thread *td, fhandle_t fh, struct statfs *buf)
4445 {
4446 	struct statfs *sp;
4447 	struct mount *mp;
4448 	struct vnode *vp;
4449 	int error;
4450 
4451 	error = priv_check(td, PRIV_VFS_FHSTATFS);
4452 	if (error != 0)
4453 		return (error);
4454 	if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4455 		return (ESTALE);
4456 	error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4457 	if (error != 0) {
4458 		vfs_unbusy(mp);
4459 		return (error);
4460 	}
4461 	vput(vp);
4462 	error = prison_canseemount(td->td_ucred, mp);
4463 	if (error != 0)
4464 		goto out;
4465 #ifdef MAC
4466 	error = mac_mount_check_stat(td->td_ucred, mp);
4467 	if (error != 0)
4468 		goto out;
4469 #endif
4470 	/*
4471 	 * Set these in case the underlying filesystem fails to do so.
4472 	 */
4473 	sp = &mp->mnt_stat;
4474 	sp->f_version = STATFS_VERSION;
4475 	sp->f_namemax = NAME_MAX;
4476 	sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
4477 	error = VFS_STATFS(mp, sp);
4478 	if (error == 0)
4479 		*buf = *sp;
4480 out:
4481 	vfs_unbusy(mp);
4482 	return (error);
4483 }
4484 
4485 int
4486 kern_posix_fallocate(struct thread *td, int fd, off_t offset, off_t len)
4487 {
4488 	struct file *fp;
4489 	struct mount *mp;
4490 	struct vnode *vp;
4491 	cap_rights_t rights;
4492 	off_t olen, ooffset;
4493 	int error;
4494 
4495 	if (offset < 0 || len <= 0)
4496 		return (EINVAL);
4497 	/* Check for wrap. */
4498 	if (offset > OFF_MAX - len)
4499 		return (EFBIG);
4500 	error = fget(td, fd, cap_rights_init(&rights, CAP_WRITE), &fp);
4501 	if (error != 0)
4502 		return (error);
4503 	if ((fp->f_ops->fo_flags & DFLAG_SEEKABLE) == 0) {
4504 		error = ESPIPE;
4505 		goto out;
4506 	}
4507 	if ((fp->f_flag & FWRITE) == 0) {
4508 		error = EBADF;
4509 		goto out;
4510 	}
4511 	if (fp->f_type != DTYPE_VNODE) {
4512 		error = ENODEV;
4513 		goto out;
4514 	}
4515 	vp = fp->f_vnode;
4516 	if (vp->v_type != VREG) {
4517 		error = ENODEV;
4518 		goto out;
4519 	}
4520 
4521 	/* Allocating blocks may take a long time, so iterate. */
4522 	for (;;) {
4523 		olen = len;
4524 		ooffset = offset;
4525 
4526 		bwillwrite();
4527 		mp = NULL;
4528 		error = vn_start_write(vp, &mp, V_WAIT | PCATCH);
4529 		if (error != 0)
4530 			break;
4531 		error = vn_lock(vp, LK_EXCLUSIVE);
4532 		if (error != 0) {
4533 			vn_finished_write(mp);
4534 			break;
4535 		}
4536 #ifdef MAC
4537 		error = mac_vnode_check_write(td->td_ucred, fp->f_cred, vp);
4538 		if (error == 0)
4539 #endif
4540 			error = VOP_ALLOCATE(vp, &offset, &len);
4541 		VOP_UNLOCK(vp, 0);
4542 		vn_finished_write(mp);
4543 
4544 		if (olen + ooffset != offset + len) {
4545 			panic("offset + len changed from %jx/%jx to %jx/%jx",
4546 			    ooffset, olen, offset, len);
4547 		}
4548 		if (error != 0 || len == 0)
4549 			break;
4550 		KASSERT(olen > len, ("Iteration did not make progress?"));
4551 		maybe_yield();
4552 	}
4553  out:
4554 	fdrop(fp, td);
4555 	return (error);
4556 }
4557 
4558 int
4559 sys_posix_fallocate(struct thread *td, struct posix_fallocate_args *uap)
4560 {
4561 	int error;
4562 
4563 	error = kern_posix_fallocate(td, uap->fd, uap->offset, uap->len);
4564 	return (kern_posix_error(td, error));
4565 }
4566 
4567 /*
4568  * Unlike madvise(2), we do not make a best effort to remember every
4569  * possible caching hint.  Instead, we remember the last setting with
4570  * the exception that we will allow POSIX_FADV_NORMAL to adjust the
4571  * region of any current setting.
4572  */
4573 int
4574 kern_posix_fadvise(struct thread *td, int fd, off_t offset, off_t len,
4575     int advice)
4576 {
4577 	struct fadvise_info *fa, *new;
4578 	struct file *fp;
4579 	struct vnode *vp;
4580 	cap_rights_t rights;
4581 	off_t end;
4582 	int error;
4583 
4584 	if (offset < 0 || len < 0 || offset > OFF_MAX - len)
4585 		return (EINVAL);
4586 	switch (advice) {
4587 	case POSIX_FADV_SEQUENTIAL:
4588 	case POSIX_FADV_RANDOM:
4589 	case POSIX_FADV_NOREUSE:
4590 		new = malloc(sizeof(*fa), M_FADVISE, M_WAITOK);
4591 		break;
4592 	case POSIX_FADV_NORMAL:
4593 	case POSIX_FADV_WILLNEED:
4594 	case POSIX_FADV_DONTNEED:
4595 		new = NULL;
4596 		break;
4597 	default:
4598 		return (EINVAL);
4599 	}
4600 	/* XXX: CAP_POSIX_FADVISE? */
4601 	error = fget(td, fd, cap_rights_init(&rights), &fp);
4602 	if (error != 0)
4603 		goto out;
4604 	if ((fp->f_ops->fo_flags & DFLAG_SEEKABLE) == 0) {
4605 		error = ESPIPE;
4606 		goto out;
4607 	}
4608 	if (fp->f_type != DTYPE_VNODE) {
4609 		error = ENODEV;
4610 		goto out;
4611 	}
4612 	vp = fp->f_vnode;
4613 	if (vp->v_type != VREG) {
4614 		error = ENODEV;
4615 		goto out;
4616 	}
4617 	if (len == 0)
4618 		end = OFF_MAX;
4619 	else
4620 		end = offset + len - 1;
4621 	switch (advice) {
4622 	case POSIX_FADV_SEQUENTIAL:
4623 	case POSIX_FADV_RANDOM:
4624 	case POSIX_FADV_NOREUSE:
4625 		/*
4626 		 * Try to merge any existing non-standard region with
4627 		 * this new region if possible, otherwise create a new
4628 		 * non-standard region for this request.
4629 		 */
4630 		mtx_pool_lock(mtxpool_sleep, fp);
4631 		fa = fp->f_advice;
4632 		if (fa != NULL && fa->fa_advice == advice &&
4633 		    ((fa->fa_start <= end && fa->fa_end >= offset) ||
4634 		    (end != OFF_MAX && fa->fa_start == end + 1) ||
4635 		    (fa->fa_end != OFF_MAX && fa->fa_end + 1 == offset))) {
4636 			if (offset < fa->fa_start)
4637 				fa->fa_start = offset;
4638 			if (end > fa->fa_end)
4639 				fa->fa_end = end;
4640 		} else {
4641 			new->fa_advice = advice;
4642 			new->fa_start = offset;
4643 			new->fa_end = end;
4644 			fp->f_advice = new;
4645 			new = fa;
4646 		}
4647 		mtx_pool_unlock(mtxpool_sleep, fp);
4648 		break;
4649 	case POSIX_FADV_NORMAL:
4650 		/*
4651 		 * If a the "normal" region overlaps with an existing
4652 		 * non-standard region, trim or remove the
4653 		 * non-standard region.
4654 		 */
4655 		mtx_pool_lock(mtxpool_sleep, fp);
4656 		fa = fp->f_advice;
4657 		if (fa != NULL) {
4658 			if (offset <= fa->fa_start && end >= fa->fa_end) {
4659 				new = fa;
4660 				fp->f_advice = NULL;
4661 			} else if (offset <= fa->fa_start &&
4662 			    end >= fa->fa_start)
4663 				fa->fa_start = end + 1;
4664 			else if (offset <= fa->fa_end && end >= fa->fa_end)
4665 				fa->fa_end = offset - 1;
4666 			else if (offset >= fa->fa_start && end <= fa->fa_end) {
4667 				/*
4668 				 * If the "normal" region is a middle
4669 				 * portion of the existing
4670 				 * non-standard region, just remove
4671 				 * the whole thing rather than picking
4672 				 * one side or the other to
4673 				 * preserve.
4674 				 */
4675 				new = fa;
4676 				fp->f_advice = NULL;
4677 			}
4678 		}
4679 		mtx_pool_unlock(mtxpool_sleep, fp);
4680 		break;
4681 	case POSIX_FADV_WILLNEED:
4682 	case POSIX_FADV_DONTNEED:
4683 		error = VOP_ADVISE(vp, offset, end, advice);
4684 		break;
4685 	}
4686 out:
4687 	if (fp != NULL)
4688 		fdrop(fp, td);
4689 	free(new, M_FADVISE);
4690 	return (error);
4691 }
4692 
4693 int
4694 sys_posix_fadvise(struct thread *td, struct posix_fadvise_args *uap)
4695 {
4696 	int error;
4697 
4698 	error = kern_posix_fadvise(td, uap->fd, uap->offset, uap->len,
4699 	    uap->advice);
4700 	return (kern_posix_error(td, error));
4701 }
4702