xref: /freebsd/sys/kern/vfs_syscalls.c (revision b88cc53d4ddda4486683ee2121f131b10ed21c30)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (c) 1989, 1993
5  *	The Regents of the University of California.  All rights reserved.
6  * (c) UNIX System Laboratories, Inc.
7  * All or some portions of this file are derived from material licensed
8  * to the University of California by American Telephone and Telegraph
9  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
10  * the permission of UNIX System Laboratories, Inc.
11  *
12  * Redistribution and use in source and binary forms, with or without
13  * modification, are permitted provided that the following conditions
14  * are met:
15  * 1. Redistributions of source code must retain the above copyright
16  *    notice, this list of conditions and the following disclaimer.
17  * 2. Redistributions in binary form must reproduce the above copyright
18  *    notice, this list of conditions and the following disclaimer in the
19  *    documentation and/or other materials provided with the distribution.
20  * 3. Neither the name of the University nor the names of its contributors
21  *    may be used to endorse or promote products derived from this software
22  *    without specific prior written permission.
23  *
24  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
25  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
28  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
29  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
30  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
31  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
32  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
33  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
34  * SUCH DAMAGE.
35  *
36  *	@(#)vfs_syscalls.c	8.13 (Berkeley) 4/15/94
37  */
38 
39 #include <sys/cdefs.h>
40 __FBSDID("$FreeBSD$");
41 
42 #include "opt_capsicum.h"
43 #include "opt_ktrace.h"
44 
45 #include <sys/param.h>
46 #include <sys/systm.h>
47 #include <sys/bio.h>
48 #include <sys/buf.h>
49 #include <sys/capsicum.h>
50 #include <sys/disk.h>
51 #include <sys/sysent.h>
52 #include <sys/malloc.h>
53 #include <sys/mount.h>
54 #include <sys/mutex.h>
55 #include <sys/sysproto.h>
56 #include <sys/namei.h>
57 #include <sys/filedesc.h>
58 #include <sys/kernel.h>
59 #include <sys/fcntl.h>
60 #include <sys/file.h>
61 #include <sys/filio.h>
62 #include <sys/limits.h>
63 #include <sys/linker.h>
64 #include <sys/rwlock.h>
65 #include <sys/sdt.h>
66 #include <sys/stat.h>
67 #include <sys/sx.h>
68 #include <sys/unistd.h>
69 #include <sys/vnode.h>
70 #include <sys/priv.h>
71 #include <sys/proc.h>
72 #include <sys/dirent.h>
73 #include <sys/jail.h>
74 #include <sys/syscallsubr.h>
75 #include <sys/sysctl.h>
76 #ifdef KTRACE
77 #include <sys/ktrace.h>
78 #endif
79 
80 #include <machine/stdarg.h>
81 
82 #include <security/audit/audit.h>
83 #include <security/mac/mac_framework.h>
84 
85 #include <vm/vm.h>
86 #include <vm/vm_object.h>
87 #include <vm/vm_page.h>
88 #include <vm/uma.h>
89 
90 #include <ufs/ufs/quota.h>
91 
92 MALLOC_DEFINE(M_FADVISE, "fadvise", "posix_fadvise(2) information");
93 
94 SDT_PROVIDER_DEFINE(vfs);
95 SDT_PROBE_DEFINE2(vfs, , stat, mode, "char *", "int");
96 SDT_PROBE_DEFINE2(vfs, , stat, reg, "char *", "int");
97 
98 static int kern_chflagsat(struct thread *td, int fd, const char *path,
99     enum uio_seg pathseg, u_long flags, int atflag);
100 static int setfflags(struct thread *td, struct vnode *, u_long);
101 static int getutimes(const struct timeval *, enum uio_seg, struct timespec *);
102 static int getutimens(const struct timespec *, enum uio_seg,
103     struct timespec *, int *);
104 static int setutimes(struct thread *td, struct vnode *,
105     const struct timespec *, int, int);
106 static int vn_access(struct vnode *vp, int user_flags, struct ucred *cred,
107     struct thread *td);
108 static int kern_fhlinkat(struct thread *td, int fd, const char *path,
109     enum uio_seg pathseg, fhandle_t *fhp);
110 static int kern_getfhat(struct thread *td, int flags, int fd,
111     const char *path, enum uio_seg pathseg, fhandle_t *fhp);
112 static int kern_readlink_vp(struct vnode *vp, char *buf, enum uio_seg bufseg,
113     size_t count, struct thread *td);
114 static int kern_linkat_vp(struct thread *td, struct vnode *vp, int fd,
115     const char *path, enum uio_seg segflag);
116 
117 /*
118  * Sync each mounted filesystem.
119  */
120 #ifndef _SYS_SYSPROTO_H_
121 struct sync_args {
122 	int     dummy;
123 };
124 #endif
125 /* ARGSUSED */
126 int
127 sys_sync(struct thread *td, struct sync_args *uap)
128 {
129 	struct mount *mp, *nmp;
130 	int save;
131 
132 	mtx_lock(&mountlist_mtx);
133 	for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
134 		if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK)) {
135 			nmp = TAILQ_NEXT(mp, mnt_list);
136 			continue;
137 		}
138 		if ((mp->mnt_flag & MNT_RDONLY) == 0 &&
139 		    vn_start_write(NULL, &mp, V_NOWAIT) == 0) {
140 			save = curthread_pflags_set(TDP_SYNCIO);
141 			vfs_msync(mp, MNT_NOWAIT);
142 			VFS_SYNC(mp, MNT_NOWAIT);
143 			curthread_pflags_restore(save);
144 			vn_finished_write(mp);
145 		}
146 		mtx_lock(&mountlist_mtx);
147 		nmp = TAILQ_NEXT(mp, mnt_list);
148 		vfs_unbusy(mp);
149 	}
150 	mtx_unlock(&mountlist_mtx);
151 	return (0);
152 }
153 
154 /*
155  * Change filesystem quotas.
156  */
157 #ifndef _SYS_SYSPROTO_H_
158 struct quotactl_args {
159 	char *path;
160 	int cmd;
161 	int uid;
162 	caddr_t arg;
163 };
164 #endif
165 int
166 sys_quotactl(struct thread *td, struct quotactl_args *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 	    (uap->cmd >> SUBCMDSHIFT) != Q_QUOTAOFF)
203 		vfs_unbusy(mp);
204 	return (error);
205 }
206 
207 /*
208  * Used by statfs conversion routines to scale the block size up if
209  * necessary so that all of the block counts are <= 'max_size'.  Note
210  * that 'max_size' should be a bitmask, i.e. 2^n - 1 for some non-zero
211  * value of 'n'.
212  */
213 void
214 statfs_scale_blocks(struct statfs *sf, long max_size)
215 {
216 	uint64_t count;
217 	int shift;
218 
219 	KASSERT(powerof2(max_size + 1), ("%s: invalid max_size", __func__));
220 
221 	/*
222 	 * Attempt to scale the block counts to give a more accurate
223 	 * overview to userland of the ratio of free space to used
224 	 * space.  To do this, find the largest block count and compute
225 	 * a divisor that lets it fit into a signed integer <= max_size.
226 	 */
227 	if (sf->f_bavail < 0)
228 		count = -sf->f_bavail;
229 	else
230 		count = sf->f_bavail;
231 	count = MAX(sf->f_blocks, MAX(sf->f_bfree, count));
232 	if (count <= max_size)
233 		return;
234 
235 	count >>= flsl(max_size);
236 	shift = 0;
237 	while (count > 0) {
238 		shift++;
239 		count >>=1;
240 	}
241 
242 	sf->f_bsize <<= shift;
243 	sf->f_blocks >>= shift;
244 	sf->f_bfree >>= shift;
245 	sf->f_bavail >>= shift;
246 }
247 
248 static int
249 kern_do_statfs(struct thread *td, struct mount *mp, struct statfs *buf)
250 {
251 	struct statfs *sp;
252 	int error;
253 
254 	if (mp == NULL)
255 		return (EBADF);
256 	error = vfs_busy(mp, 0);
257 	vfs_rel(mp);
258 	if (error != 0)
259 		return (error);
260 #ifdef MAC
261 	error = mac_mount_check_stat(td->td_ucred, mp);
262 	if (error != 0)
263 		goto out;
264 #endif
265 	/*
266 	 * Set these in case the underlying filesystem fails to do so.
267 	 */
268 	sp = &mp->mnt_stat;
269 	sp->f_version = STATFS_VERSION;
270 	sp->f_namemax = NAME_MAX;
271 	sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
272 	error = VFS_STATFS(mp, sp);
273 	if (error != 0)
274 		goto out;
275 	*buf = *sp;
276 	if (priv_check(td, PRIV_VFS_GENERATION)) {
277 		buf->f_fsid.val[0] = buf->f_fsid.val[1] = 0;
278 		prison_enforce_statfs(td->td_ucred, mp, buf);
279 	}
280 out:
281 	vfs_unbusy(mp);
282 	return (error);
283 }
284 
285 /*
286  * Get filesystem statistics.
287  */
288 #ifndef _SYS_SYSPROTO_H_
289 struct statfs_args {
290 	char *path;
291 	struct statfs *buf;
292 };
293 #endif
294 int
295 sys_statfs(struct thread *td, struct statfs_args *uap)
296 {
297 	struct statfs *sfp;
298 	int error;
299 
300 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
301 	error = kern_statfs(td, uap->path, UIO_USERSPACE, sfp);
302 	if (error == 0)
303 		error = copyout(sfp, uap->buf, sizeof(struct statfs));
304 	free(sfp, M_STATFS);
305 	return (error);
306 }
307 
308 int
309 kern_statfs(struct thread *td, const char *path, enum uio_seg pathseg,
310     struct statfs *buf)
311 {
312 	struct mount *mp;
313 	struct nameidata nd;
314 	int error;
315 
316 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
317 	    pathseg, path, td);
318 	error = namei(&nd);
319 	if (error != 0)
320 		return (error);
321 	mp = nd.ni_vp->v_mount;
322 	vfs_ref(mp);
323 	NDFREE(&nd, NDF_ONLY_PNBUF);
324 	vput(nd.ni_vp);
325 	return (kern_do_statfs(td, mp, buf));
326 }
327 
328 /*
329  * Get filesystem statistics.
330  */
331 #ifndef _SYS_SYSPROTO_H_
332 struct fstatfs_args {
333 	int fd;
334 	struct statfs *buf;
335 };
336 #endif
337 int
338 sys_fstatfs(struct thread *td, struct fstatfs_args *uap)
339 {
340 	struct statfs *sfp;
341 	int error;
342 
343 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
344 	error = kern_fstatfs(td, uap->fd, sfp);
345 	if (error == 0)
346 		error = copyout(sfp, uap->buf, sizeof(struct statfs));
347 	free(sfp, M_STATFS);
348 	return (error);
349 }
350 
351 int
352 kern_fstatfs(struct thread *td, int fd, struct statfs *buf)
353 {
354 	struct file *fp;
355 	struct mount *mp;
356 	struct vnode *vp;
357 	int error;
358 
359 	AUDIT_ARG_FD(fd);
360 	error = getvnode(td, fd, &cap_fstatfs_rights, &fp);
361 	if (error != 0)
362 		return (error);
363 	vp = fp->f_vnode;
364 	vn_lock(vp, LK_SHARED | LK_RETRY);
365 #ifdef AUDIT
366 	AUDIT_ARG_VNODE1(vp);
367 #endif
368 	mp = vp->v_mount;
369 	if (mp != NULL)
370 		vfs_ref(mp);
371 	VOP_UNLOCK(vp, 0);
372 	fdrop(fp, td);
373 	return (kern_do_statfs(td, mp, buf));
374 }
375 
376 /*
377  * Get statistics on all filesystems.
378  */
379 #ifndef _SYS_SYSPROTO_H_
380 struct getfsstat_args {
381 	struct statfs *buf;
382 	long bufsize;
383 	int mode;
384 };
385 #endif
386 int
387 sys_getfsstat(struct thread *td, struct getfsstat_args *uap)
388 {
389 	size_t count;
390 	int error;
391 
392 	if (uap->bufsize < 0 || uap->bufsize > SIZE_MAX)
393 		return (EINVAL);
394 	error = kern_getfsstat(td, &uap->buf, uap->bufsize, &count,
395 	    UIO_USERSPACE, uap->mode);
396 	if (error == 0)
397 		td->td_retval[0] = count;
398 	return (error);
399 }
400 
401 /*
402  * If (bufsize > 0 && bufseg == UIO_SYSSPACE)
403  *	The caller is responsible for freeing memory which will be allocated
404  *	in '*buf'.
405  */
406 int
407 kern_getfsstat(struct thread *td, struct statfs **buf, size_t bufsize,
408     size_t *countp, enum uio_seg bufseg, int mode)
409 {
410 	struct mount *mp, *nmp;
411 	struct statfs *sfsp, *sp, *sptmp, *tofree;
412 	size_t count, maxcount;
413 	int error;
414 
415 	switch (mode) {
416 	case MNT_WAIT:
417 	case MNT_NOWAIT:
418 		break;
419 	default:
420 		if (bufseg == UIO_SYSSPACE)
421 			*buf = NULL;
422 		return (EINVAL);
423 	}
424 restart:
425 	maxcount = bufsize / sizeof(struct statfs);
426 	if (bufsize == 0) {
427 		sfsp = NULL;
428 		tofree = NULL;
429 	} else if (bufseg == UIO_USERSPACE) {
430 		sfsp = *buf;
431 		tofree = NULL;
432 	} else /* if (bufseg == UIO_SYSSPACE) */ {
433 		count = 0;
434 		mtx_lock(&mountlist_mtx);
435 		TAILQ_FOREACH(mp, &mountlist, mnt_list) {
436 			count++;
437 		}
438 		mtx_unlock(&mountlist_mtx);
439 		if (maxcount > count)
440 			maxcount = count;
441 		tofree = sfsp = *buf = malloc(maxcount * sizeof(struct statfs),
442 		    M_STATFS, M_WAITOK);
443 	}
444 	count = 0;
445 	mtx_lock(&mountlist_mtx);
446 	for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
447 		if (prison_canseemount(td->td_ucred, mp) != 0) {
448 			nmp = TAILQ_NEXT(mp, mnt_list);
449 			continue;
450 		}
451 #ifdef MAC
452 		if (mac_mount_check_stat(td->td_ucred, mp) != 0) {
453 			nmp = TAILQ_NEXT(mp, mnt_list);
454 			continue;
455 		}
456 #endif
457 		if (mode == MNT_WAIT) {
458 			if (vfs_busy(mp, MBF_MNTLSTLOCK) != 0) {
459 				/*
460 				 * If vfs_busy() failed, and MBF_NOWAIT
461 				 * wasn't passed, then the mp is gone.
462 				 * Furthermore, because of MBF_MNTLSTLOCK,
463 				 * the mountlist_mtx was dropped.  We have
464 				 * no other choice than to start over.
465 				 */
466 				mtx_unlock(&mountlist_mtx);
467 				free(tofree, M_STATFS);
468 				goto restart;
469 			}
470 		} else {
471 			if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK) != 0) {
472 				nmp = TAILQ_NEXT(mp, mnt_list);
473 				continue;
474 			}
475 		}
476 		if (sfsp != NULL && count < maxcount) {
477 			sp = &mp->mnt_stat;
478 			/*
479 			 * Set these in case the underlying filesystem
480 			 * fails to do so.
481 			 */
482 			sp->f_version = STATFS_VERSION;
483 			sp->f_namemax = NAME_MAX;
484 			sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
485 			/*
486 			 * If MNT_NOWAIT is specified, do not refresh
487 			 * the fsstat cache.
488 			 */
489 			if (mode != MNT_NOWAIT) {
490 				error = VFS_STATFS(mp, sp);
491 				if (error != 0) {
492 					mtx_lock(&mountlist_mtx);
493 					nmp = TAILQ_NEXT(mp, mnt_list);
494 					vfs_unbusy(mp);
495 					continue;
496 				}
497 			}
498 			if (priv_check(td, PRIV_VFS_GENERATION)) {
499 				sptmp = malloc(sizeof(struct statfs), M_STATFS,
500 				    M_WAITOK);
501 				*sptmp = *sp;
502 				sptmp->f_fsid.val[0] = sptmp->f_fsid.val[1] = 0;
503 				prison_enforce_statfs(td->td_ucred, mp, sptmp);
504 				sp = sptmp;
505 			} else
506 				sptmp = NULL;
507 			if (bufseg == UIO_SYSSPACE) {
508 				bcopy(sp, sfsp, sizeof(*sp));
509 				free(sptmp, M_STATFS);
510 			} else /* if (bufseg == UIO_USERSPACE) */ {
511 				error = copyout(sp, sfsp, sizeof(*sp));
512 				free(sptmp, M_STATFS);
513 				if (error != 0) {
514 					vfs_unbusy(mp);
515 					return (error);
516 				}
517 			}
518 			sfsp++;
519 		}
520 		count++;
521 		mtx_lock(&mountlist_mtx);
522 		nmp = TAILQ_NEXT(mp, mnt_list);
523 		vfs_unbusy(mp);
524 	}
525 	mtx_unlock(&mountlist_mtx);
526 	if (sfsp != NULL && count > maxcount)
527 		*countp = maxcount;
528 	else
529 		*countp = count;
530 	return (0);
531 }
532 
533 #ifdef COMPAT_FREEBSD4
534 /*
535  * Get old format filesystem statistics.
536  */
537 static void freebsd4_cvtstatfs(struct statfs *, struct ostatfs *);
538 
539 #ifndef _SYS_SYSPROTO_H_
540 struct freebsd4_statfs_args {
541 	char *path;
542 	struct ostatfs *buf;
543 };
544 #endif
545 int
546 freebsd4_statfs(struct thread *td, struct freebsd4_statfs_args *uap)
547 {
548 	struct ostatfs osb;
549 	struct statfs *sfp;
550 	int error;
551 
552 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
553 	error = kern_statfs(td, uap->path, UIO_USERSPACE, sfp);
554 	if (error == 0) {
555 		freebsd4_cvtstatfs(sfp, &osb);
556 		error = copyout(&osb, uap->buf, sizeof(osb));
557 	}
558 	free(sfp, M_STATFS);
559 	return (error);
560 }
561 
562 /*
563  * Get filesystem statistics.
564  */
565 #ifndef _SYS_SYSPROTO_H_
566 struct freebsd4_fstatfs_args {
567 	int fd;
568 	struct ostatfs *buf;
569 };
570 #endif
571 int
572 freebsd4_fstatfs(struct thread *td, struct freebsd4_fstatfs_args *uap)
573 {
574 	struct ostatfs osb;
575 	struct statfs *sfp;
576 	int error;
577 
578 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
579 	error = kern_fstatfs(td, uap->fd, sfp);
580 	if (error == 0) {
581 		freebsd4_cvtstatfs(sfp, &osb);
582 		error = copyout(&osb, uap->buf, sizeof(osb));
583 	}
584 	free(sfp, M_STATFS);
585 	return (error);
586 }
587 
588 /*
589  * Get statistics on all filesystems.
590  */
591 #ifndef _SYS_SYSPROTO_H_
592 struct freebsd4_getfsstat_args {
593 	struct ostatfs *buf;
594 	long bufsize;
595 	int mode;
596 };
597 #endif
598 int
599 freebsd4_getfsstat(struct thread *td, struct freebsd4_getfsstat_args *uap)
600 {
601 	struct statfs *buf, *sp;
602 	struct ostatfs osb;
603 	size_t count, size;
604 	int error;
605 
606 	if (uap->bufsize < 0)
607 		return (EINVAL);
608 	count = uap->bufsize / sizeof(struct ostatfs);
609 	if (count > SIZE_MAX / sizeof(struct statfs))
610 		return (EINVAL);
611 	size = count * sizeof(struct statfs);
612 	error = kern_getfsstat(td, &buf, size, &count, UIO_SYSSPACE,
613 	    uap->mode);
614 	if (error == 0)
615 		td->td_retval[0] = count;
616 	if (size != 0) {
617 		sp = buf;
618 		while (count != 0 && error == 0) {
619 			freebsd4_cvtstatfs(sp, &osb);
620 			error = copyout(&osb, uap->buf, sizeof(osb));
621 			sp++;
622 			uap->buf++;
623 			count--;
624 		}
625 		free(buf, M_STATFS);
626 	}
627 	return (error);
628 }
629 
630 /*
631  * Implement fstatfs() for (NFS) file handles.
632  */
633 #ifndef _SYS_SYSPROTO_H_
634 struct freebsd4_fhstatfs_args {
635 	struct fhandle *u_fhp;
636 	struct ostatfs *buf;
637 };
638 #endif
639 int
640 freebsd4_fhstatfs(struct thread *td, struct freebsd4_fhstatfs_args *uap)
641 {
642 	struct ostatfs osb;
643 	struct statfs *sfp;
644 	fhandle_t fh;
645 	int error;
646 
647 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
648 	if (error != 0)
649 		return (error);
650 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
651 	error = kern_fhstatfs(td, fh, sfp);
652 	if (error == 0) {
653 		freebsd4_cvtstatfs(sfp, &osb);
654 		error = copyout(&osb, uap->buf, sizeof(osb));
655 	}
656 	free(sfp, M_STATFS);
657 	return (error);
658 }
659 
660 /*
661  * Convert a new format statfs structure to an old format statfs structure.
662  */
663 static void
664 freebsd4_cvtstatfs(struct statfs *nsp, struct ostatfs *osp)
665 {
666 
667 	statfs_scale_blocks(nsp, LONG_MAX);
668 	bzero(osp, sizeof(*osp));
669 	osp->f_bsize = nsp->f_bsize;
670 	osp->f_iosize = MIN(nsp->f_iosize, LONG_MAX);
671 	osp->f_blocks = nsp->f_blocks;
672 	osp->f_bfree = nsp->f_bfree;
673 	osp->f_bavail = nsp->f_bavail;
674 	osp->f_files = MIN(nsp->f_files, LONG_MAX);
675 	osp->f_ffree = MIN(nsp->f_ffree, LONG_MAX);
676 	osp->f_owner = nsp->f_owner;
677 	osp->f_type = nsp->f_type;
678 	osp->f_flags = nsp->f_flags;
679 	osp->f_syncwrites = MIN(nsp->f_syncwrites, LONG_MAX);
680 	osp->f_asyncwrites = MIN(nsp->f_asyncwrites, LONG_MAX);
681 	osp->f_syncreads = MIN(nsp->f_syncreads, LONG_MAX);
682 	osp->f_asyncreads = MIN(nsp->f_asyncreads, LONG_MAX);
683 	strlcpy(osp->f_fstypename, nsp->f_fstypename,
684 	    MIN(MFSNAMELEN, OMFSNAMELEN));
685 	strlcpy(osp->f_mntonname, nsp->f_mntonname,
686 	    MIN(MNAMELEN, OMNAMELEN));
687 	strlcpy(osp->f_mntfromname, nsp->f_mntfromname,
688 	    MIN(MNAMELEN, OMNAMELEN));
689 	osp->f_fsid = nsp->f_fsid;
690 }
691 #endif /* COMPAT_FREEBSD4 */
692 
693 #if defined(COMPAT_FREEBSD11)
694 /*
695  * Get old format filesystem statistics.
696  */
697 static void freebsd11_cvtstatfs(struct statfs *, struct freebsd11_statfs *);
698 
699 int
700 freebsd11_statfs(struct thread *td, struct freebsd11_statfs_args *uap)
701 {
702 	struct freebsd11_statfs osb;
703 	struct statfs *sfp;
704 	int error;
705 
706 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
707 	error = kern_statfs(td, uap->path, UIO_USERSPACE, sfp);
708 	if (error == 0) {
709 		freebsd11_cvtstatfs(sfp, &osb);
710 		error = copyout(&osb, uap->buf, sizeof(osb));
711 	}
712 	free(sfp, M_STATFS);
713 	return (error);
714 }
715 
716 /*
717  * Get filesystem statistics.
718  */
719 int
720 freebsd11_fstatfs(struct thread *td, struct freebsd11_fstatfs_args *uap)
721 {
722 	struct freebsd11_statfs osb;
723 	struct statfs *sfp;
724 	int error;
725 
726 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
727 	error = kern_fstatfs(td, uap->fd, sfp);
728 	if (error == 0) {
729 		freebsd11_cvtstatfs(sfp, &osb);
730 		error = copyout(&osb, uap->buf, sizeof(osb));
731 	}
732 	free(sfp, M_STATFS);
733 	return (error);
734 }
735 
736 /*
737  * Get statistics on all filesystems.
738  */
739 int
740 freebsd11_getfsstat(struct thread *td, struct freebsd11_getfsstat_args *uap)
741 {
742 	struct freebsd11_statfs osb;
743 	struct statfs *buf, *sp;
744 	size_t count, size;
745 	int error;
746 
747 	count = uap->bufsize / sizeof(struct ostatfs);
748 	size = count * sizeof(struct statfs);
749 	error = kern_getfsstat(td, &buf, size, &count, UIO_SYSSPACE,
750 	    uap->mode);
751 	if (error == 0)
752 		td->td_retval[0] = count;
753 	if (size > 0) {
754 		sp = buf;
755 		while (count > 0 && error == 0) {
756 			freebsd11_cvtstatfs(sp, &osb);
757 			error = copyout(&osb, uap->buf, sizeof(osb));
758 			sp++;
759 			uap->buf++;
760 			count--;
761 		}
762 		free(buf, M_STATFS);
763 	}
764 	return (error);
765 }
766 
767 /*
768  * Implement fstatfs() for (NFS) file handles.
769  */
770 int
771 freebsd11_fhstatfs(struct thread *td, struct freebsd11_fhstatfs_args *uap)
772 {
773 	struct freebsd11_statfs osb;
774 	struct statfs *sfp;
775 	fhandle_t fh;
776 	int error;
777 
778 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
779 	if (error)
780 		return (error);
781 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
782 	error = kern_fhstatfs(td, fh, sfp);
783 	if (error == 0) {
784 		freebsd11_cvtstatfs(sfp, &osb);
785 		error = copyout(&osb, uap->buf, sizeof(osb));
786 	}
787 	free(sfp, M_STATFS);
788 	return (error);
789 }
790 
791 /*
792  * Convert a new format statfs structure to an old format statfs structure.
793  */
794 static void
795 freebsd11_cvtstatfs(struct statfs *nsp, struct freebsd11_statfs *osp)
796 {
797 
798 	bzero(osp, sizeof(*osp));
799 	osp->f_version = FREEBSD11_STATFS_VERSION;
800 	osp->f_type = nsp->f_type;
801 	osp->f_flags = nsp->f_flags;
802 	osp->f_bsize = nsp->f_bsize;
803 	osp->f_iosize = nsp->f_iosize;
804 	osp->f_blocks = nsp->f_blocks;
805 	osp->f_bfree = nsp->f_bfree;
806 	osp->f_bavail = nsp->f_bavail;
807 	osp->f_files = nsp->f_files;
808 	osp->f_ffree = nsp->f_ffree;
809 	osp->f_syncwrites = nsp->f_syncwrites;
810 	osp->f_asyncwrites = nsp->f_asyncwrites;
811 	osp->f_syncreads = nsp->f_syncreads;
812 	osp->f_asyncreads = nsp->f_asyncreads;
813 	osp->f_namemax = nsp->f_namemax;
814 	osp->f_owner = nsp->f_owner;
815 	osp->f_fsid = nsp->f_fsid;
816 	strlcpy(osp->f_fstypename, nsp->f_fstypename,
817 	    MIN(MFSNAMELEN, sizeof(osp->f_fstypename)));
818 	strlcpy(osp->f_mntonname, nsp->f_mntonname,
819 	    MIN(MNAMELEN, sizeof(osp->f_mntonname)));
820 	strlcpy(osp->f_mntfromname, nsp->f_mntfromname,
821 	    MIN(MNAMELEN, sizeof(osp->f_mntfromname)));
822 }
823 #endif /* COMPAT_FREEBSD11 */
824 
825 /*
826  * Change current working directory to a given file descriptor.
827  */
828 #ifndef _SYS_SYSPROTO_H_
829 struct fchdir_args {
830 	int	fd;
831 };
832 #endif
833 int
834 sys_fchdir(struct thread *td, struct fchdir_args *uap)
835 {
836 	struct vnode *vp, *tdp;
837 	struct mount *mp;
838 	struct file *fp;
839 	int error;
840 
841 	AUDIT_ARG_FD(uap->fd);
842 	error = getvnode(td, uap->fd, &cap_fchdir_rights,
843 	    &fp);
844 	if (error != 0)
845 		return (error);
846 	vp = fp->f_vnode;
847 	vrefact(vp);
848 	fdrop(fp, td);
849 	vn_lock(vp, LK_SHARED | LK_RETRY);
850 	AUDIT_ARG_VNODE1(vp);
851 	error = change_dir(vp, td);
852 	while (!error && (mp = vp->v_mountedhere) != NULL) {
853 		if (vfs_busy(mp, 0))
854 			continue;
855 		error = VFS_ROOT(mp, LK_SHARED, &tdp);
856 		vfs_unbusy(mp);
857 		if (error != 0)
858 			break;
859 		vput(vp);
860 		vp = tdp;
861 	}
862 	if (error != 0) {
863 		vput(vp);
864 		return (error);
865 	}
866 	VOP_UNLOCK(vp, 0);
867 	pwd_chdir(td, vp);
868 	return (0);
869 }
870 
871 /*
872  * Change current working directory (``.'').
873  */
874 #ifndef _SYS_SYSPROTO_H_
875 struct chdir_args {
876 	char	*path;
877 };
878 #endif
879 int
880 sys_chdir(struct thread *td, struct chdir_args *uap)
881 {
882 
883 	return (kern_chdir(td, uap->path, UIO_USERSPACE));
884 }
885 
886 int
887 kern_chdir(struct thread *td, const char *path, enum uio_seg pathseg)
888 {
889 	struct nameidata nd;
890 	int error;
891 
892 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
893 	    pathseg, path, td);
894 	if ((error = namei(&nd)) != 0)
895 		return (error);
896 	if ((error = change_dir(nd.ni_vp, td)) != 0) {
897 		vput(nd.ni_vp);
898 		NDFREE(&nd, NDF_ONLY_PNBUF);
899 		return (error);
900 	}
901 	VOP_UNLOCK(nd.ni_vp, 0);
902 	NDFREE(&nd, NDF_ONLY_PNBUF);
903 	pwd_chdir(td, nd.ni_vp);
904 	return (0);
905 }
906 
907 /*
908  * Change notion of root (``/'') directory.
909  */
910 #ifndef _SYS_SYSPROTO_H_
911 struct chroot_args {
912 	char	*path;
913 };
914 #endif
915 int
916 sys_chroot(struct thread *td, struct chroot_args *uap)
917 {
918 	struct nameidata nd;
919 	int error;
920 
921 	error = priv_check(td, PRIV_VFS_CHROOT);
922 	if (error != 0)
923 		return (error);
924 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
925 	    UIO_USERSPACE, uap->path, td);
926 	error = namei(&nd);
927 	if (error != 0)
928 		goto error;
929 	error = change_dir(nd.ni_vp, td);
930 	if (error != 0)
931 		goto e_vunlock;
932 #ifdef MAC
933 	error = mac_vnode_check_chroot(td->td_ucred, nd.ni_vp);
934 	if (error != 0)
935 		goto e_vunlock;
936 #endif
937 	VOP_UNLOCK(nd.ni_vp, 0);
938 	error = pwd_chroot(td, nd.ni_vp);
939 	vrele(nd.ni_vp);
940 	NDFREE(&nd, NDF_ONLY_PNBUF);
941 	return (error);
942 e_vunlock:
943 	vput(nd.ni_vp);
944 error:
945 	NDFREE(&nd, NDF_ONLY_PNBUF);
946 	return (error);
947 }
948 
949 /*
950  * Common routine for chroot and chdir.  Callers must provide a locked vnode
951  * instance.
952  */
953 int
954 change_dir(struct vnode *vp, struct thread *td)
955 {
956 #ifdef MAC
957 	int error;
958 #endif
959 
960 	ASSERT_VOP_LOCKED(vp, "change_dir(): vp not locked");
961 	if (vp->v_type != VDIR)
962 		return (ENOTDIR);
963 #ifdef MAC
964 	error = mac_vnode_check_chdir(td->td_ucred, vp);
965 	if (error != 0)
966 		return (error);
967 #endif
968 	return (VOP_ACCESS(vp, VEXEC, td->td_ucred, td));
969 }
970 
971 static __inline void
972 flags_to_rights(int flags, cap_rights_t *rightsp)
973 {
974 
975 	if (flags & O_EXEC) {
976 		cap_rights_set(rightsp, CAP_FEXECVE);
977 	} else {
978 		switch ((flags & O_ACCMODE)) {
979 		case O_RDONLY:
980 			cap_rights_set(rightsp, CAP_READ);
981 			break;
982 		case O_RDWR:
983 			cap_rights_set(rightsp, CAP_READ);
984 			/* FALLTHROUGH */
985 		case O_WRONLY:
986 			cap_rights_set(rightsp, CAP_WRITE);
987 			if (!(flags & (O_APPEND | O_TRUNC)))
988 				cap_rights_set(rightsp, CAP_SEEK);
989 			break;
990 		}
991 	}
992 
993 	if (flags & O_CREAT)
994 		cap_rights_set(rightsp, CAP_CREATE);
995 
996 	if (flags & O_TRUNC)
997 		cap_rights_set(rightsp, CAP_FTRUNCATE);
998 
999 	if (flags & (O_SYNC | O_FSYNC))
1000 		cap_rights_set(rightsp, CAP_FSYNC);
1001 
1002 	if (flags & (O_EXLOCK | O_SHLOCK))
1003 		cap_rights_set(rightsp, CAP_FLOCK);
1004 }
1005 
1006 /*
1007  * Check permissions, allocate an open file structure, and call the device
1008  * open routine if any.
1009  */
1010 #ifndef _SYS_SYSPROTO_H_
1011 struct open_args {
1012 	char	*path;
1013 	int	flags;
1014 	int	mode;
1015 };
1016 #endif
1017 int
1018 sys_open(struct thread *td, struct open_args *uap)
1019 {
1020 
1021 	return (kern_openat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1022 	    uap->flags, uap->mode));
1023 }
1024 
1025 #ifndef _SYS_SYSPROTO_H_
1026 struct openat_args {
1027 	int	fd;
1028 	char	*path;
1029 	int	flag;
1030 	int	mode;
1031 };
1032 #endif
1033 int
1034 sys_openat(struct thread *td, struct openat_args *uap)
1035 {
1036 
1037 	AUDIT_ARG_FD(uap->fd);
1038 	return (kern_openat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
1039 	    uap->mode));
1040 }
1041 
1042 int
1043 kern_openat(struct thread *td, int fd, const char *path, enum uio_seg pathseg,
1044     int flags, int mode)
1045 {
1046 	struct proc *p = td->td_proc;
1047 	struct filedesc *fdp = p->p_fd;
1048 	struct file *fp;
1049 	struct vnode *vp;
1050 	struct nameidata nd;
1051 	cap_rights_t rights;
1052 	int cmode, error, indx;
1053 
1054 	indx = -1;
1055 
1056 	AUDIT_ARG_FFLAGS(flags);
1057 	AUDIT_ARG_MODE(mode);
1058 	cap_rights_init(&rights, CAP_LOOKUP);
1059 	flags_to_rights(flags, &rights);
1060 	/*
1061 	 * Only one of the O_EXEC, O_RDONLY, O_WRONLY and O_RDWR flags
1062 	 * may be specified.
1063 	 */
1064 	if (flags & O_EXEC) {
1065 		if (flags & O_ACCMODE)
1066 			return (EINVAL);
1067 	} else if ((flags & O_ACCMODE) == O_ACCMODE) {
1068 		return (EINVAL);
1069 	} else {
1070 		flags = FFLAGS(flags);
1071 	}
1072 
1073 	/*
1074 	 * Allocate a file structure. The descriptor to reference it
1075 	 * is allocated and set by finstall() below.
1076 	 */
1077 	error = falloc_noinstall(td, &fp);
1078 	if (error != 0)
1079 		return (error);
1080 	/*
1081 	 * An extra reference on `fp' has been held for us by
1082 	 * falloc_noinstall().
1083 	 */
1084 	/* Set the flags early so the finit in devfs can pick them up. */
1085 	fp->f_flag = flags & FMASK;
1086 	cmode = ((mode & ~fdp->fd_cmask) & ALLPERMS) & ~S_ISTXT;
1087 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, fd,
1088 	    &rights, td);
1089 	td->td_dupfd = -1;		/* XXX check for fdopen */
1090 	error = vn_open(&nd, &flags, cmode, fp);
1091 	if (error != 0) {
1092 		/*
1093 		 * If the vn_open replaced the method vector, something
1094 		 * wonderous happened deep below and we just pass it up
1095 		 * pretending we know what we do.
1096 		 */
1097 		if (error == ENXIO && fp->f_ops != &badfileops)
1098 			goto success;
1099 
1100 		/*
1101 		 * Handle special fdopen() case. bleh.
1102 		 *
1103 		 * Don't do this for relative (capability) lookups; we don't
1104 		 * understand exactly what would happen, and we don't think
1105 		 * that it ever should.
1106 		 */
1107 		if ((nd.ni_lcf & NI_LCF_STRICTRELATIVE) == 0 &&
1108 		    (error == ENODEV || error == ENXIO) &&
1109 		    td->td_dupfd >= 0) {
1110 			error = dupfdopen(td, fdp, td->td_dupfd, flags, error,
1111 			    &indx);
1112 			if (error == 0)
1113 				goto success;
1114 		}
1115 
1116 		goto bad;
1117 	}
1118 	td->td_dupfd = 0;
1119 	NDFREE(&nd, NDF_ONLY_PNBUF);
1120 	vp = nd.ni_vp;
1121 
1122 	/*
1123 	 * Store the vnode, for any f_type. Typically, the vnode use
1124 	 * count is decremented by direct call to vn_closefile() for
1125 	 * files that switched type in the cdevsw fdopen() method.
1126 	 */
1127 	fp->f_vnode = vp;
1128 	/*
1129 	 * If the file wasn't claimed by devfs bind it to the normal
1130 	 * vnode operations here.
1131 	 */
1132 	if (fp->f_ops == &badfileops) {
1133 		KASSERT(vp->v_type != VFIFO, ("Unexpected fifo."));
1134 		fp->f_seqcount = 1;
1135 		finit(fp, (flags & FMASK) | (fp->f_flag & FHASLOCK),
1136 		    DTYPE_VNODE, vp, &vnops);
1137 	}
1138 
1139 	VOP_UNLOCK(vp, 0);
1140 	if (flags & O_TRUNC) {
1141 		error = fo_truncate(fp, 0, td->td_ucred, td);
1142 		if (error != 0)
1143 			goto bad;
1144 	}
1145 success:
1146 	/*
1147 	 * If we haven't already installed the FD (for dupfdopen), do so now.
1148 	 */
1149 	if (indx == -1) {
1150 		struct filecaps *fcaps;
1151 
1152 #ifdef CAPABILITIES
1153 		if ((nd.ni_lcf & NI_LCF_STRICTRELATIVE) != 0)
1154 			fcaps = &nd.ni_filecaps;
1155 		else
1156 #endif
1157 			fcaps = NULL;
1158 		error = finstall(td, fp, &indx, flags, fcaps);
1159 		/* On success finstall() consumes fcaps. */
1160 		if (error != 0) {
1161 			filecaps_free(&nd.ni_filecaps);
1162 			goto bad;
1163 		}
1164 	} else {
1165 		filecaps_free(&nd.ni_filecaps);
1166 	}
1167 
1168 	/*
1169 	 * Release our private reference, leaving the one associated with
1170 	 * the descriptor table intact.
1171 	 */
1172 	fdrop(fp, td);
1173 	td->td_retval[0] = indx;
1174 	return (0);
1175 bad:
1176 	KASSERT(indx == -1, ("indx=%d, should be -1", indx));
1177 	fdrop(fp, td);
1178 	return (error);
1179 }
1180 
1181 #ifdef COMPAT_43
1182 /*
1183  * Create a file.
1184  */
1185 #ifndef _SYS_SYSPROTO_H_
1186 struct ocreat_args {
1187 	char	*path;
1188 	int	mode;
1189 };
1190 #endif
1191 int
1192 ocreat(struct thread *td, struct ocreat_args *uap)
1193 {
1194 
1195 	return (kern_openat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1196 	    O_WRONLY | O_CREAT | O_TRUNC, uap->mode));
1197 }
1198 #endif /* COMPAT_43 */
1199 
1200 /*
1201  * Create a special file.
1202  */
1203 #ifndef _SYS_SYSPROTO_H_
1204 struct mknodat_args {
1205 	int	fd;
1206 	char	*path;
1207 	mode_t	mode;
1208 	dev_t	dev;
1209 };
1210 #endif
1211 int
1212 sys_mknodat(struct thread *td, struct mknodat_args *uap)
1213 {
1214 
1215 	return (kern_mknodat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode,
1216 	    uap->dev));
1217 }
1218 
1219 #if defined(COMPAT_FREEBSD11)
1220 int
1221 freebsd11_mknod(struct thread *td,
1222     struct freebsd11_mknod_args *uap)
1223 {
1224 
1225 	return (kern_mknodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1226 	    uap->mode, uap->dev));
1227 }
1228 
1229 int
1230 freebsd11_mknodat(struct thread *td,
1231     struct freebsd11_mknodat_args *uap)
1232 {
1233 
1234 	return (kern_mknodat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode,
1235 	    uap->dev));
1236 }
1237 #endif /* COMPAT_FREEBSD11 */
1238 
1239 int
1240 kern_mknodat(struct thread *td, int fd, const char *path, enum uio_seg pathseg,
1241     int mode, dev_t dev)
1242 {
1243 	struct vnode *vp;
1244 	struct mount *mp;
1245 	struct vattr vattr;
1246 	struct nameidata nd;
1247 	int error, whiteout = 0;
1248 
1249 	AUDIT_ARG_MODE(mode);
1250 	AUDIT_ARG_DEV(dev);
1251 	switch (mode & S_IFMT) {
1252 	case S_IFCHR:
1253 	case S_IFBLK:
1254 		error = priv_check(td, PRIV_VFS_MKNOD_DEV);
1255 		if (error == 0 && dev == VNOVAL)
1256 			error = EINVAL;
1257 		break;
1258 	case S_IFWHT:
1259 		error = priv_check(td, PRIV_VFS_MKNOD_WHT);
1260 		break;
1261 	case S_IFIFO:
1262 		if (dev == 0)
1263 			return (kern_mkfifoat(td, fd, path, pathseg, mode));
1264 		/* FALLTHROUGH */
1265 	default:
1266 		error = EINVAL;
1267 		break;
1268 	}
1269 	if (error != 0)
1270 		return (error);
1271 restart:
1272 	bwillwrite();
1273 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
1274 	    NOCACHE, pathseg, path, fd, &cap_mknodat_rights,
1275 	    td);
1276 	if ((error = namei(&nd)) != 0)
1277 		return (error);
1278 	vp = nd.ni_vp;
1279 	if (vp != NULL) {
1280 		NDFREE(&nd, NDF_ONLY_PNBUF);
1281 		if (vp == nd.ni_dvp)
1282 			vrele(nd.ni_dvp);
1283 		else
1284 			vput(nd.ni_dvp);
1285 		vrele(vp);
1286 		return (EEXIST);
1287 	} else {
1288 		VATTR_NULL(&vattr);
1289 		vattr.va_mode = (mode & ALLPERMS) &
1290 		    ~td->td_proc->p_fd->fd_cmask;
1291 		vattr.va_rdev = dev;
1292 		whiteout = 0;
1293 
1294 		switch (mode & S_IFMT) {
1295 		case S_IFCHR:
1296 			vattr.va_type = VCHR;
1297 			break;
1298 		case S_IFBLK:
1299 			vattr.va_type = VBLK;
1300 			break;
1301 		case S_IFWHT:
1302 			whiteout = 1;
1303 			break;
1304 		default:
1305 			panic("kern_mknod: invalid mode");
1306 		}
1307 	}
1308 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1309 		NDFREE(&nd, NDF_ONLY_PNBUF);
1310 		vput(nd.ni_dvp);
1311 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1312 			return (error);
1313 		goto restart;
1314 	}
1315 #ifdef MAC
1316 	if (error == 0 && !whiteout)
1317 		error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp,
1318 		    &nd.ni_cnd, &vattr);
1319 #endif
1320 	if (error == 0) {
1321 		if (whiteout)
1322 			error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE);
1323 		else {
1324 			error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp,
1325 						&nd.ni_cnd, &vattr);
1326 			if (error == 0)
1327 				vput(nd.ni_vp);
1328 		}
1329 	}
1330 	NDFREE(&nd, NDF_ONLY_PNBUF);
1331 	vput(nd.ni_dvp);
1332 	vn_finished_write(mp);
1333 	return (error);
1334 }
1335 
1336 /*
1337  * Create a named pipe.
1338  */
1339 #ifndef _SYS_SYSPROTO_H_
1340 struct mkfifo_args {
1341 	char	*path;
1342 	int	mode;
1343 };
1344 #endif
1345 int
1346 sys_mkfifo(struct thread *td, struct mkfifo_args *uap)
1347 {
1348 
1349 	return (kern_mkfifoat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1350 	    uap->mode));
1351 }
1352 
1353 #ifndef _SYS_SYSPROTO_H_
1354 struct mkfifoat_args {
1355 	int	fd;
1356 	char	*path;
1357 	mode_t	mode;
1358 };
1359 #endif
1360 int
1361 sys_mkfifoat(struct thread *td, struct mkfifoat_args *uap)
1362 {
1363 
1364 	return (kern_mkfifoat(td, uap->fd, uap->path, UIO_USERSPACE,
1365 	    uap->mode));
1366 }
1367 
1368 int
1369 kern_mkfifoat(struct thread *td, int fd, const char *path,
1370     enum uio_seg pathseg, int mode)
1371 {
1372 	struct mount *mp;
1373 	struct vattr vattr;
1374 	struct nameidata nd;
1375 	int error;
1376 
1377 	AUDIT_ARG_MODE(mode);
1378 restart:
1379 	bwillwrite();
1380 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
1381 	    NOCACHE, pathseg, path, fd, &cap_mkfifoat_rights,
1382 	    td);
1383 	if ((error = namei(&nd)) != 0)
1384 		return (error);
1385 	if (nd.ni_vp != NULL) {
1386 		NDFREE(&nd, NDF_ONLY_PNBUF);
1387 		if (nd.ni_vp == nd.ni_dvp)
1388 			vrele(nd.ni_dvp);
1389 		else
1390 			vput(nd.ni_dvp);
1391 		vrele(nd.ni_vp);
1392 		return (EEXIST);
1393 	}
1394 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1395 		NDFREE(&nd, NDF_ONLY_PNBUF);
1396 		vput(nd.ni_dvp);
1397 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1398 			return (error);
1399 		goto restart;
1400 	}
1401 	VATTR_NULL(&vattr);
1402 	vattr.va_type = VFIFO;
1403 	vattr.va_mode = (mode & ALLPERMS) & ~td->td_proc->p_fd->fd_cmask;
1404 #ifdef MAC
1405 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1406 	    &vattr);
1407 	if (error != 0)
1408 		goto out;
1409 #endif
1410 	error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
1411 	if (error == 0)
1412 		vput(nd.ni_vp);
1413 #ifdef MAC
1414 out:
1415 #endif
1416 	vput(nd.ni_dvp);
1417 	vn_finished_write(mp);
1418 	NDFREE(&nd, NDF_ONLY_PNBUF);
1419 	return (error);
1420 }
1421 
1422 /*
1423  * Make a hard file link.
1424  */
1425 #ifndef _SYS_SYSPROTO_H_
1426 struct link_args {
1427 	char	*path;
1428 	char	*link;
1429 };
1430 #endif
1431 int
1432 sys_link(struct thread *td, struct link_args *uap)
1433 {
1434 
1435 	return (kern_linkat(td, AT_FDCWD, AT_FDCWD, uap->path, uap->link,
1436 	    UIO_USERSPACE, FOLLOW));
1437 }
1438 
1439 #ifndef _SYS_SYSPROTO_H_
1440 struct linkat_args {
1441 	int	fd1;
1442 	char	*path1;
1443 	int	fd2;
1444 	char	*path2;
1445 	int	flag;
1446 };
1447 #endif
1448 int
1449 sys_linkat(struct thread *td, struct linkat_args *uap)
1450 {
1451 	int flag;
1452 
1453 	flag = uap->flag;
1454 	if ((flag & ~(AT_SYMLINK_FOLLOW | AT_BENEATH)) != 0)
1455 		return (EINVAL);
1456 
1457 	return (kern_linkat(td, uap->fd1, uap->fd2, uap->path1, uap->path2,
1458 	    UIO_USERSPACE, ((flag & AT_SYMLINK_FOLLOW) != 0 ? FOLLOW :
1459 	    NOFOLLOW) | ((flag & AT_BENEATH) != 0 ? BENEATH : 0)));
1460 }
1461 
1462 int hardlink_check_uid = 0;
1463 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_uid, CTLFLAG_RW,
1464     &hardlink_check_uid, 0,
1465     "Unprivileged processes cannot create hard links to files owned by other "
1466     "users");
1467 static int hardlink_check_gid = 0;
1468 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_gid, CTLFLAG_RW,
1469     &hardlink_check_gid, 0,
1470     "Unprivileged processes cannot create hard links to files owned by other "
1471     "groups");
1472 
1473 static int
1474 can_hardlink(struct vnode *vp, struct ucred *cred)
1475 {
1476 	struct vattr va;
1477 	int error;
1478 
1479 	if (!hardlink_check_uid && !hardlink_check_gid)
1480 		return (0);
1481 
1482 	error = VOP_GETATTR(vp, &va, cred);
1483 	if (error != 0)
1484 		return (error);
1485 
1486 	if (hardlink_check_uid && cred->cr_uid != va.va_uid) {
1487 		error = priv_check_cred(cred, PRIV_VFS_LINK);
1488 		if (error != 0)
1489 			return (error);
1490 	}
1491 
1492 	if (hardlink_check_gid && !groupmember(va.va_gid, cred)) {
1493 		error = priv_check_cred(cred, PRIV_VFS_LINK);
1494 		if (error != 0)
1495 			return (error);
1496 	}
1497 
1498 	return (0);
1499 }
1500 
1501 int
1502 kern_linkat(struct thread *td, int fd1, int fd2, const char *path1,
1503     const char *path2, enum uio_seg segflag, int follow)
1504 {
1505 	struct nameidata nd;
1506 	int error;
1507 
1508 	do {
1509 		bwillwrite();
1510 		NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, segflag,
1511 		    path1, fd1, &cap_linkat_source_rights, td);
1512 		if ((error = namei(&nd)) != 0)
1513 			return (error);
1514 		NDFREE(&nd, NDF_ONLY_PNBUF);
1515 		error = kern_linkat_vp(td, nd.ni_vp, fd2, path2, segflag);
1516 	} while (error ==  EAGAIN);
1517 	return (error);
1518 }
1519 
1520 static int
1521 kern_linkat_vp(struct thread *td, struct vnode *vp, int fd, const char *path,
1522     enum uio_seg segflag)
1523 {
1524 	struct nameidata nd;
1525 	struct mount *mp;
1526 	int error;
1527 
1528 	if (vp->v_type == VDIR) {
1529 		vrele(vp);
1530 		return (EPERM);		/* POSIX */
1531 	}
1532 	NDINIT_ATRIGHTS(&nd, CREATE,
1533 	    LOCKPARENT | SAVENAME | AUDITVNODE2 | NOCACHE, segflag, path, fd,
1534 	    &cap_linkat_target_rights, td);
1535 	if ((error = namei(&nd)) == 0) {
1536 		if (nd.ni_vp != NULL) {
1537 			NDFREE(&nd, NDF_ONLY_PNBUF);
1538 			if (nd.ni_dvp == nd.ni_vp)
1539 				vrele(nd.ni_dvp);
1540 			else
1541 				vput(nd.ni_dvp);
1542 			vrele(nd.ni_vp);
1543 			vrele(vp);
1544 			return (EEXIST);
1545 		} else if (nd.ni_dvp->v_mount != vp->v_mount) {
1546 			/*
1547 			 * Cross-device link.  No need to recheck
1548 			 * vp->v_type, since it cannot change, except
1549 			 * to VBAD.
1550 			 */
1551 			NDFREE(&nd, NDF_ONLY_PNBUF);
1552 			vput(nd.ni_dvp);
1553 			vrele(vp);
1554 			return (EXDEV);
1555 		} else if ((error = vn_lock(vp, LK_EXCLUSIVE)) == 0) {
1556 			error = can_hardlink(vp, td->td_ucred);
1557 #ifdef MAC
1558 			if (error == 0)
1559 				error = mac_vnode_check_link(td->td_ucred,
1560 				    nd.ni_dvp, vp, &nd.ni_cnd);
1561 #endif
1562 			if (error != 0) {
1563 				vput(vp);
1564 				vput(nd.ni_dvp);
1565 				NDFREE(&nd, NDF_ONLY_PNBUF);
1566 				return (error);
1567 			}
1568 			error = vn_start_write(vp, &mp, V_NOWAIT);
1569 			if (error != 0) {
1570 				vput(vp);
1571 				vput(nd.ni_dvp);
1572 				NDFREE(&nd, NDF_ONLY_PNBUF);
1573 				error = vn_start_write(NULL, &mp,
1574 				    V_XSLEEP | PCATCH);
1575 				if (error != 0)
1576 					return (error);
1577 				return (EAGAIN);
1578 			}
1579 			error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
1580 			VOP_UNLOCK(vp, 0);
1581 			vput(nd.ni_dvp);
1582 			vn_finished_write(mp);
1583 			NDFREE(&nd, NDF_ONLY_PNBUF);
1584 		} else {
1585 			vput(nd.ni_dvp);
1586 			NDFREE(&nd, NDF_ONLY_PNBUF);
1587 			vrele(vp);
1588 			return (EAGAIN);
1589 		}
1590 	}
1591 	vrele(vp);
1592 	return (error);
1593 }
1594 
1595 /*
1596  * Make a symbolic link.
1597  */
1598 #ifndef _SYS_SYSPROTO_H_
1599 struct symlink_args {
1600 	char	*path;
1601 	char	*link;
1602 };
1603 #endif
1604 int
1605 sys_symlink(struct thread *td, struct symlink_args *uap)
1606 {
1607 
1608 	return (kern_symlinkat(td, uap->path, AT_FDCWD, uap->link,
1609 	    UIO_USERSPACE));
1610 }
1611 
1612 #ifndef _SYS_SYSPROTO_H_
1613 struct symlinkat_args {
1614 	char	*path;
1615 	int	fd;
1616 	char	*path2;
1617 };
1618 #endif
1619 int
1620 sys_symlinkat(struct thread *td, struct symlinkat_args *uap)
1621 {
1622 
1623 	return (kern_symlinkat(td, uap->path1, uap->fd, uap->path2,
1624 	    UIO_USERSPACE));
1625 }
1626 
1627 int
1628 kern_symlinkat(struct thread *td, const char *path1, int fd, const char *path2,
1629     enum uio_seg segflg)
1630 {
1631 	struct mount *mp;
1632 	struct vattr vattr;
1633 	const char *syspath;
1634 	char *tmppath;
1635 	struct nameidata nd;
1636 	int error;
1637 
1638 	if (segflg == UIO_SYSSPACE) {
1639 		syspath = path1;
1640 	} else {
1641 		tmppath = uma_zalloc(namei_zone, M_WAITOK);
1642 		if ((error = copyinstr(path1, tmppath, MAXPATHLEN, NULL)) != 0)
1643 			goto out;
1644 		syspath = tmppath;
1645 	}
1646 	AUDIT_ARG_TEXT(syspath);
1647 restart:
1648 	bwillwrite();
1649 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
1650 	    NOCACHE, segflg, path2, fd, &cap_symlinkat_rights,
1651 	    td);
1652 	if ((error = namei(&nd)) != 0)
1653 		goto out;
1654 	if (nd.ni_vp) {
1655 		NDFREE(&nd, NDF_ONLY_PNBUF);
1656 		if (nd.ni_vp == nd.ni_dvp)
1657 			vrele(nd.ni_dvp);
1658 		else
1659 			vput(nd.ni_dvp);
1660 		vrele(nd.ni_vp);
1661 		error = EEXIST;
1662 		goto out;
1663 	}
1664 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1665 		NDFREE(&nd, NDF_ONLY_PNBUF);
1666 		vput(nd.ni_dvp);
1667 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1668 			goto out;
1669 		goto restart;
1670 	}
1671 	VATTR_NULL(&vattr);
1672 	vattr.va_mode = ACCESSPERMS &~ td->td_proc->p_fd->fd_cmask;
1673 #ifdef MAC
1674 	vattr.va_type = VLNK;
1675 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1676 	    &vattr);
1677 	if (error != 0)
1678 		goto out2;
1679 #endif
1680 	error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, syspath);
1681 	if (error == 0)
1682 		vput(nd.ni_vp);
1683 #ifdef MAC
1684 out2:
1685 #endif
1686 	NDFREE(&nd, NDF_ONLY_PNBUF);
1687 	vput(nd.ni_dvp);
1688 	vn_finished_write(mp);
1689 out:
1690 	if (segflg != UIO_SYSSPACE)
1691 		uma_zfree(namei_zone, tmppath);
1692 	return (error);
1693 }
1694 
1695 /*
1696  * Delete a whiteout from the filesystem.
1697  */
1698 #ifndef _SYS_SYSPROTO_H_
1699 struct undelete_args {
1700 	char *path;
1701 };
1702 #endif
1703 int
1704 sys_undelete(struct thread *td, struct undelete_args *uap)
1705 {
1706 	struct mount *mp;
1707 	struct nameidata nd;
1708 	int error;
1709 
1710 restart:
1711 	bwillwrite();
1712 	NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | AUDITVNODE1,
1713 	    UIO_USERSPACE, uap->path, td);
1714 	error = namei(&nd);
1715 	if (error != 0)
1716 		return (error);
1717 
1718 	if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) {
1719 		NDFREE(&nd, NDF_ONLY_PNBUF);
1720 		if (nd.ni_vp == nd.ni_dvp)
1721 			vrele(nd.ni_dvp);
1722 		else
1723 			vput(nd.ni_dvp);
1724 		if (nd.ni_vp)
1725 			vrele(nd.ni_vp);
1726 		return (EEXIST);
1727 	}
1728 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1729 		NDFREE(&nd, NDF_ONLY_PNBUF);
1730 		vput(nd.ni_dvp);
1731 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1732 			return (error);
1733 		goto restart;
1734 	}
1735 	error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE);
1736 	NDFREE(&nd, NDF_ONLY_PNBUF);
1737 	vput(nd.ni_dvp);
1738 	vn_finished_write(mp);
1739 	return (error);
1740 }
1741 
1742 /*
1743  * Delete a name from the filesystem.
1744  */
1745 #ifndef _SYS_SYSPROTO_H_
1746 struct unlink_args {
1747 	char	*path;
1748 };
1749 #endif
1750 int
1751 sys_unlink(struct thread *td, struct unlink_args *uap)
1752 {
1753 
1754 	return (kern_funlinkat(td, AT_FDCWD, uap->path, FD_NONE, UIO_USERSPACE,
1755 	    0, 0));
1756 }
1757 
1758 static int
1759 kern_funlinkat_ex(struct thread *td, int dfd, const char *path, int fd,
1760     int flag, enum uio_seg pathseg, ino_t oldinum)
1761 {
1762 
1763 	if ((flag & ~AT_REMOVEDIR) != 0)
1764 		return (EINVAL);
1765 
1766 	if ((flag & AT_REMOVEDIR) != 0)
1767 		return (kern_frmdirat(td, dfd, path, fd, UIO_USERSPACE, 0));
1768 
1769 	return (kern_funlinkat(td, dfd, path, fd, UIO_USERSPACE, 0, 0));
1770 }
1771 
1772 #ifndef _SYS_SYSPROTO_H_
1773 struct unlinkat_args {
1774 	int	fd;
1775 	char	*path;
1776 	int	flag;
1777 };
1778 #endif
1779 int
1780 sys_unlinkat(struct thread *td, struct unlinkat_args *uap)
1781 {
1782 
1783 	return (kern_funlinkat_ex(td, uap->fd, uap->path, FD_NONE, uap->flag,
1784 	    UIO_USERSPACE, 0));
1785 }
1786 
1787 #ifndef _SYS_SYSPROTO_H_
1788 struct funlinkat_args {
1789 	int		dfd;
1790 	const char	*path;
1791 	int		fd;
1792 	int		flag;
1793 };
1794 #endif
1795 int
1796 sys_funlinkat(struct thread *td, struct funlinkat_args *uap)
1797 {
1798 
1799 	return (kern_funlinkat_ex(td, uap->dfd, uap->path, uap->fd, uap->flag,
1800 	    UIO_USERSPACE, 0));
1801 }
1802 
1803 int
1804 kern_funlinkat(struct thread *td, int dfd, const char *path, int fd,
1805     enum uio_seg pathseg, int flag, ino_t oldinum)
1806 {
1807 	struct mount *mp;
1808 	struct file *fp;
1809 	struct vnode *vp;
1810 	struct nameidata nd;
1811 	struct stat sb;
1812 	int error;
1813 
1814 	fp = NULL;
1815 	if (fd != FD_NONE) {
1816 		error = getvnode(td, fd, &cap_no_rights, &fp);
1817 		if (error != 0)
1818 			return (error);
1819 	}
1820 
1821 restart:
1822 	bwillwrite();
1823 	NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1 |
1824 	    ((flag & AT_BENEATH) != 0 ? BENEATH : 0),
1825 	    pathseg, path, dfd, &cap_unlinkat_rights, td);
1826 	if ((error = namei(&nd)) != 0) {
1827 		if (error == EINVAL)
1828 			error = EPERM;
1829 		goto fdout;
1830 	}
1831 	vp = nd.ni_vp;
1832 	if (vp->v_type == VDIR && oldinum == 0) {
1833 		error = EPERM;		/* POSIX */
1834 	} else if (oldinum != 0 &&
1835 		  ((error = vn_stat(vp, &sb, td->td_ucred, NOCRED, td)) == 0) &&
1836 		  sb.st_ino != oldinum) {
1837 		error = EIDRM;	/* Identifier removed */
1838 	} else if (fp != NULL && fp->f_vnode != vp) {
1839 		if ((fp->f_vnode->v_iflag & VI_DOOMED) != 0)
1840 			error = EBADF;
1841 		else
1842 			error = EDEADLK;
1843 	} else {
1844 		/*
1845 		 * The root of a mounted filesystem cannot be deleted.
1846 		 *
1847 		 * XXX: can this only be a VDIR case?
1848 		 */
1849 		if (vp->v_vflag & VV_ROOT)
1850 			error = EBUSY;
1851 	}
1852 	if (error == 0) {
1853 		if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1854 			NDFREE(&nd, NDF_ONLY_PNBUF);
1855 			vput(nd.ni_dvp);
1856 			if (vp == nd.ni_dvp)
1857 				vrele(vp);
1858 			else
1859 				vput(vp);
1860 			if ((error = vn_start_write(NULL, &mp,
1861 			    V_XSLEEP | PCATCH)) != 0) {
1862 				goto fdout;
1863 			}
1864 			goto restart;
1865 		}
1866 #ifdef MAC
1867 		error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
1868 		    &nd.ni_cnd);
1869 		if (error != 0)
1870 			goto out;
1871 #endif
1872 		vfs_notify_upper(vp, VFS_NOTIFY_UPPER_UNLINK);
1873 		error = VOP_REMOVE(nd.ni_dvp, vp, &nd.ni_cnd);
1874 #ifdef MAC
1875 out:
1876 #endif
1877 		vn_finished_write(mp);
1878 	}
1879 	NDFREE(&nd, NDF_ONLY_PNBUF);
1880 	vput(nd.ni_dvp);
1881 	if (vp == nd.ni_dvp)
1882 		vrele(vp);
1883 	else
1884 		vput(vp);
1885 fdout:
1886 	if (fp != NULL)
1887 		fdrop(fp, td);
1888 	return (error);
1889 }
1890 
1891 /*
1892  * Reposition read/write file offset.
1893  */
1894 #ifndef _SYS_SYSPROTO_H_
1895 struct lseek_args {
1896 	int	fd;
1897 	int	pad;
1898 	off_t	offset;
1899 	int	whence;
1900 };
1901 #endif
1902 int
1903 sys_lseek(struct thread *td, struct lseek_args *uap)
1904 {
1905 
1906 	return (kern_lseek(td, uap->fd, uap->offset, uap->whence));
1907 }
1908 
1909 int
1910 kern_lseek(struct thread *td, int fd, off_t offset, int whence)
1911 {
1912 	struct file *fp;
1913 	int error;
1914 
1915 	AUDIT_ARG_FD(fd);
1916 	error = fget(td, fd, &cap_seek_rights, &fp);
1917 	if (error != 0)
1918 		return (error);
1919 	error = (fp->f_ops->fo_flags & DFLAG_SEEKABLE) != 0 ?
1920 	    fo_seek(fp, offset, whence, td) : ESPIPE;
1921 	fdrop(fp, td);
1922 	return (error);
1923 }
1924 
1925 #if defined(COMPAT_43)
1926 /*
1927  * Reposition read/write file offset.
1928  */
1929 #ifndef _SYS_SYSPROTO_H_
1930 struct olseek_args {
1931 	int	fd;
1932 	long	offset;
1933 	int	whence;
1934 };
1935 #endif
1936 int
1937 olseek(struct thread *td, struct olseek_args *uap)
1938 {
1939 
1940 	return (kern_lseek(td, uap->fd, uap->offset, uap->whence));
1941 }
1942 #endif /* COMPAT_43 */
1943 
1944 #if defined(COMPAT_FREEBSD6)
1945 /* Version with the 'pad' argument */
1946 int
1947 freebsd6_lseek(struct thread *td, struct freebsd6_lseek_args *uap)
1948 {
1949 
1950 	return (kern_lseek(td, uap->fd, uap->offset, uap->whence));
1951 }
1952 #endif
1953 
1954 /*
1955  * Check access permissions using passed credentials.
1956  */
1957 static int
1958 vn_access(struct vnode *vp, int user_flags, struct ucred *cred,
1959      struct thread *td)
1960 {
1961 	accmode_t accmode;
1962 	int error;
1963 
1964 	/* Flags == 0 means only check for existence. */
1965 	if (user_flags == 0)
1966 		return (0);
1967 
1968 	accmode = 0;
1969 	if (user_flags & R_OK)
1970 		accmode |= VREAD;
1971 	if (user_flags & W_OK)
1972 		accmode |= VWRITE;
1973 	if (user_flags & X_OK)
1974 		accmode |= VEXEC;
1975 #ifdef MAC
1976 	error = mac_vnode_check_access(cred, vp, accmode);
1977 	if (error != 0)
1978 		return (error);
1979 #endif
1980 	if ((accmode & VWRITE) == 0 || (error = vn_writechk(vp)) == 0)
1981 		error = VOP_ACCESS(vp, accmode, cred, td);
1982 	return (error);
1983 }
1984 
1985 /*
1986  * Check access permissions using "real" credentials.
1987  */
1988 #ifndef _SYS_SYSPROTO_H_
1989 struct access_args {
1990 	char	*path;
1991 	int	amode;
1992 };
1993 #endif
1994 int
1995 sys_access(struct thread *td, struct access_args *uap)
1996 {
1997 
1998 	return (kern_accessat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1999 	    0, uap->amode));
2000 }
2001 
2002 #ifndef _SYS_SYSPROTO_H_
2003 struct faccessat_args {
2004 	int	dirfd;
2005 	char	*path;
2006 	int	amode;
2007 	int	flag;
2008 }
2009 #endif
2010 int
2011 sys_faccessat(struct thread *td, struct faccessat_args *uap)
2012 {
2013 
2014 	return (kern_accessat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
2015 	    uap->amode));
2016 }
2017 
2018 int
2019 kern_accessat(struct thread *td, int fd, const char *path,
2020     enum uio_seg pathseg, int flag, int amode)
2021 {
2022 	struct ucred *cred, *usecred;
2023 	struct vnode *vp;
2024 	struct nameidata nd;
2025 	int error;
2026 
2027 	if ((flag & ~(AT_EACCESS | AT_BENEATH)) != 0)
2028 		return (EINVAL);
2029 	if (amode != F_OK && (amode & ~(R_OK | W_OK | X_OK)) != 0)
2030 		return (EINVAL);
2031 
2032 	/*
2033 	 * Create and modify a temporary credential instead of one that
2034 	 * is potentially shared (if we need one).
2035 	 */
2036 	cred = td->td_ucred;
2037 	if ((flag & AT_EACCESS) == 0 &&
2038 	    ((cred->cr_uid != cred->cr_ruid ||
2039 	    cred->cr_rgid != cred->cr_groups[0]))) {
2040 		usecred = crdup(cred);
2041 		usecred->cr_uid = cred->cr_ruid;
2042 		usecred->cr_groups[0] = cred->cr_rgid;
2043 		td->td_ucred = usecred;
2044 	} else
2045 		usecred = cred;
2046 	AUDIT_ARG_VALUE(amode);
2047 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF |
2048 	    AUDITVNODE1 | ((flag & AT_BENEATH) != 0 ? BENEATH : 0),
2049 	    pathseg, path, fd, &cap_fstat_rights, td);
2050 	if ((error = namei(&nd)) != 0)
2051 		goto out;
2052 	vp = nd.ni_vp;
2053 
2054 	error = vn_access(vp, amode, usecred, td);
2055 	NDFREE(&nd, NDF_ONLY_PNBUF);
2056 	vput(vp);
2057 out:
2058 	if (usecred != cred) {
2059 		td->td_ucred = cred;
2060 		crfree(usecred);
2061 	}
2062 	return (error);
2063 }
2064 
2065 /*
2066  * Check access permissions using "effective" credentials.
2067  */
2068 #ifndef _SYS_SYSPROTO_H_
2069 struct eaccess_args {
2070 	char	*path;
2071 	int	amode;
2072 };
2073 #endif
2074 int
2075 sys_eaccess(struct thread *td, struct eaccess_args *uap)
2076 {
2077 
2078 	return (kern_accessat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2079 	    AT_EACCESS, uap->amode));
2080 }
2081 
2082 #if defined(COMPAT_43)
2083 /*
2084  * Get file status; this version follows links.
2085  */
2086 #ifndef _SYS_SYSPROTO_H_
2087 struct ostat_args {
2088 	char	*path;
2089 	struct ostat *ub;
2090 };
2091 #endif
2092 int
2093 ostat(struct thread *td, struct ostat_args *uap)
2094 {
2095 	struct stat sb;
2096 	struct ostat osb;
2097 	int error;
2098 
2099 	error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE,
2100 	    &sb, NULL);
2101 	if (error != 0)
2102 		return (error);
2103 	cvtstat(&sb, &osb);
2104 	return (copyout(&osb, uap->ub, sizeof (osb)));
2105 }
2106 
2107 /*
2108  * Get file status; this version does not follow links.
2109  */
2110 #ifndef _SYS_SYSPROTO_H_
2111 struct olstat_args {
2112 	char	*path;
2113 	struct ostat *ub;
2114 };
2115 #endif
2116 int
2117 olstat(struct thread *td, struct olstat_args *uap)
2118 {
2119 	struct stat sb;
2120 	struct ostat osb;
2121 	int error;
2122 
2123 	error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2124 	    UIO_USERSPACE, &sb, NULL);
2125 	if (error != 0)
2126 		return (error);
2127 	cvtstat(&sb, &osb);
2128 	return (copyout(&osb, uap->ub, sizeof (osb)));
2129 }
2130 
2131 /*
2132  * Convert from an old to a new stat structure.
2133  * XXX: many values are blindly truncated.
2134  */
2135 void
2136 cvtstat(struct stat *st, struct ostat *ost)
2137 {
2138 
2139 	bzero(ost, sizeof(*ost));
2140 	ost->st_dev = st->st_dev;
2141 	ost->st_ino = st->st_ino;
2142 	ost->st_mode = st->st_mode;
2143 	ost->st_nlink = st->st_nlink;
2144 	ost->st_uid = st->st_uid;
2145 	ost->st_gid = st->st_gid;
2146 	ost->st_rdev = st->st_rdev;
2147 	ost->st_size = MIN(st->st_size, INT32_MAX);
2148 	ost->st_atim = st->st_atim;
2149 	ost->st_mtim = st->st_mtim;
2150 	ost->st_ctim = st->st_ctim;
2151 	ost->st_blksize = st->st_blksize;
2152 	ost->st_blocks = st->st_blocks;
2153 	ost->st_flags = st->st_flags;
2154 	ost->st_gen = st->st_gen;
2155 }
2156 #endif /* COMPAT_43 */
2157 
2158 #if defined(COMPAT_43) || defined(COMPAT_FREEBSD11)
2159 int ino64_trunc_error;
2160 SYSCTL_INT(_vfs, OID_AUTO, ino64_trunc_error, CTLFLAG_RW,
2161     &ino64_trunc_error, 0,
2162     "Error on truncation of device, file or inode number, or link count");
2163 
2164 int
2165 freebsd11_cvtstat(struct stat *st, struct freebsd11_stat *ost)
2166 {
2167 
2168 	ost->st_dev = st->st_dev;
2169 	if (ost->st_dev != st->st_dev) {
2170 		switch (ino64_trunc_error) {
2171 		default:
2172 			/*
2173 			 * Since dev_t is almost raw, don't clamp to the
2174 			 * maximum for case 2, but ignore the error.
2175 			 */
2176 			break;
2177 		case 1:
2178 			return (EOVERFLOW);
2179 		}
2180 	}
2181 	ost->st_ino = st->st_ino;
2182 	if (ost->st_ino != st->st_ino) {
2183 		switch (ino64_trunc_error) {
2184 		default:
2185 		case 0:
2186 			break;
2187 		case 1:
2188 			return (EOVERFLOW);
2189 		case 2:
2190 			ost->st_ino = UINT32_MAX;
2191 			break;
2192 		}
2193 	}
2194 	ost->st_mode = st->st_mode;
2195 	ost->st_nlink = st->st_nlink;
2196 	if (ost->st_nlink != st->st_nlink) {
2197 		switch (ino64_trunc_error) {
2198 		default:
2199 		case 0:
2200 			break;
2201 		case 1:
2202 			return (EOVERFLOW);
2203 		case 2:
2204 			ost->st_nlink = UINT16_MAX;
2205 			break;
2206 		}
2207 	}
2208 	ost->st_uid = st->st_uid;
2209 	ost->st_gid = st->st_gid;
2210 	ost->st_rdev = st->st_rdev;
2211 	if (ost->st_rdev != st->st_rdev) {
2212 		switch (ino64_trunc_error) {
2213 		default:
2214 			break;
2215 		case 1:
2216 			return (EOVERFLOW);
2217 		}
2218 	}
2219 	ost->st_atim = st->st_atim;
2220 	ost->st_mtim = st->st_mtim;
2221 	ost->st_ctim = st->st_ctim;
2222 	ost->st_size = st->st_size;
2223 	ost->st_blocks = st->st_blocks;
2224 	ost->st_blksize = st->st_blksize;
2225 	ost->st_flags = st->st_flags;
2226 	ost->st_gen = st->st_gen;
2227 	ost->st_lspare = 0;
2228 	ost->st_birthtim = st->st_birthtim;
2229 	bzero((char *)&ost->st_birthtim + sizeof(ost->st_birthtim),
2230 	    sizeof(*ost) - offsetof(struct freebsd11_stat,
2231 	    st_birthtim) - sizeof(ost->st_birthtim));
2232 	return (0);
2233 }
2234 
2235 int
2236 freebsd11_stat(struct thread *td, struct freebsd11_stat_args* uap)
2237 {
2238 	struct stat sb;
2239 	struct freebsd11_stat osb;
2240 	int error;
2241 
2242 	error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE,
2243 	    &sb, NULL);
2244 	if (error != 0)
2245 		return (error);
2246 	error = freebsd11_cvtstat(&sb, &osb);
2247 	if (error == 0)
2248 		error = copyout(&osb, uap->ub, sizeof(osb));
2249 	return (error);
2250 }
2251 
2252 int
2253 freebsd11_lstat(struct thread *td, struct freebsd11_lstat_args* uap)
2254 {
2255 	struct stat sb;
2256 	struct freebsd11_stat osb;
2257 	int error;
2258 
2259 	error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2260 	    UIO_USERSPACE, &sb, NULL);
2261 	if (error != 0)
2262 		return (error);
2263 	error = freebsd11_cvtstat(&sb, &osb);
2264 	if (error == 0)
2265 		error = copyout(&osb, uap->ub, sizeof(osb));
2266 	return (error);
2267 }
2268 
2269 int
2270 freebsd11_fhstat(struct thread *td, struct freebsd11_fhstat_args* uap)
2271 {
2272 	struct fhandle fh;
2273 	struct stat sb;
2274 	struct freebsd11_stat osb;
2275 	int error;
2276 
2277 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
2278 	if (error != 0)
2279 		return (error);
2280 	error = kern_fhstat(td, fh, &sb);
2281 	if (error != 0)
2282 		return (error);
2283 	error = freebsd11_cvtstat(&sb, &osb);
2284 	if (error == 0)
2285 		error = copyout(&osb, uap->sb, sizeof(osb));
2286 	return (error);
2287 }
2288 
2289 int
2290 freebsd11_fstatat(struct thread *td, struct freebsd11_fstatat_args* uap)
2291 {
2292 	struct stat sb;
2293 	struct freebsd11_stat osb;
2294 	int error;
2295 
2296 	error = kern_statat(td, uap->flag, uap->fd, uap->path,
2297 	    UIO_USERSPACE, &sb, NULL);
2298 	if (error != 0)
2299 		return (error);
2300 	error = freebsd11_cvtstat(&sb, &osb);
2301 	if (error == 0)
2302 		error = copyout(&osb, uap->buf, sizeof(osb));
2303 	return (error);
2304 }
2305 #endif	/* COMPAT_FREEBSD11 */
2306 
2307 /*
2308  * Get file status
2309  */
2310 #ifndef _SYS_SYSPROTO_H_
2311 struct fstatat_args {
2312 	int	fd;
2313 	char	*path;
2314 	struct stat	*buf;
2315 	int	flag;
2316 }
2317 #endif
2318 int
2319 sys_fstatat(struct thread *td, struct fstatat_args *uap)
2320 {
2321 	struct stat sb;
2322 	int error;
2323 
2324 	error = kern_statat(td, uap->flag, uap->fd, uap->path,
2325 	    UIO_USERSPACE, &sb, NULL);
2326 	if (error == 0)
2327 		error = copyout(&sb, uap->buf, sizeof (sb));
2328 	return (error);
2329 }
2330 
2331 int
2332 kern_statat(struct thread *td, int flag, int fd, const char *path,
2333     enum uio_seg pathseg, struct stat *sbp,
2334     void (*hook)(struct vnode *vp, struct stat *sbp))
2335 {
2336 	struct nameidata nd;
2337 	int error;
2338 
2339 	if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_BENEATH)) != 0)
2340 		return (EINVAL);
2341 
2342 	NDINIT_ATRIGHTS(&nd, LOOKUP, ((flag & AT_SYMLINK_NOFOLLOW) != 0 ?
2343 	    NOFOLLOW : FOLLOW) | ((flag & AT_BENEATH) != 0 ? BENEATH : 0) |
2344 	    LOCKSHARED | LOCKLEAF | AUDITVNODE1, pathseg, path, fd,
2345 	    &cap_fstat_rights, td);
2346 
2347 	if ((error = namei(&nd)) != 0)
2348 		return (error);
2349 	error = vn_stat(nd.ni_vp, sbp, td->td_ucred, NOCRED, td);
2350 	if (error == 0) {
2351 		SDT_PROBE2(vfs, , stat, mode, path, sbp->st_mode);
2352 		if (S_ISREG(sbp->st_mode))
2353 			SDT_PROBE2(vfs, , stat, reg, path, pathseg);
2354 		if (__predict_false(hook != NULL))
2355 			hook(nd.ni_vp, sbp);
2356 	}
2357 	NDFREE(&nd, NDF_ONLY_PNBUF);
2358 	vput(nd.ni_vp);
2359 	if (error != 0)
2360 		return (error);
2361 #ifdef __STAT_TIME_T_EXT
2362 	sbp->st_atim_ext = 0;
2363 	sbp->st_mtim_ext = 0;
2364 	sbp->st_ctim_ext = 0;
2365 	sbp->st_btim_ext = 0;
2366 #endif
2367 #ifdef KTRACE
2368 	if (KTRPOINT(td, KTR_STRUCT))
2369 		ktrstat(sbp);
2370 #endif
2371 	return (0);
2372 }
2373 
2374 #if defined(COMPAT_FREEBSD11)
2375 /*
2376  * Implementation of the NetBSD [l]stat() functions.
2377  */
2378 void
2379 freebsd11_cvtnstat(struct stat *sb, struct nstat *nsb)
2380 {
2381 
2382 	bzero(nsb, sizeof(*nsb));
2383 	nsb->st_dev = sb->st_dev;
2384 	nsb->st_ino = sb->st_ino;
2385 	nsb->st_mode = sb->st_mode;
2386 	nsb->st_nlink = sb->st_nlink;
2387 	nsb->st_uid = sb->st_uid;
2388 	nsb->st_gid = sb->st_gid;
2389 	nsb->st_rdev = sb->st_rdev;
2390 	nsb->st_atim = sb->st_atim;
2391 	nsb->st_mtim = sb->st_mtim;
2392 	nsb->st_ctim = sb->st_ctim;
2393 	nsb->st_size = sb->st_size;
2394 	nsb->st_blocks = sb->st_blocks;
2395 	nsb->st_blksize = sb->st_blksize;
2396 	nsb->st_flags = sb->st_flags;
2397 	nsb->st_gen = sb->st_gen;
2398 	nsb->st_birthtim = sb->st_birthtim;
2399 }
2400 
2401 #ifndef _SYS_SYSPROTO_H_
2402 struct freebsd11_nstat_args {
2403 	char	*path;
2404 	struct nstat *ub;
2405 };
2406 #endif
2407 int
2408 freebsd11_nstat(struct thread *td, struct freebsd11_nstat_args *uap)
2409 {
2410 	struct stat sb;
2411 	struct nstat nsb;
2412 	int error;
2413 
2414 	error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE,
2415 	    &sb, NULL);
2416 	if (error != 0)
2417 		return (error);
2418 	freebsd11_cvtnstat(&sb, &nsb);
2419 	return (copyout(&nsb, uap->ub, sizeof (nsb)));
2420 }
2421 
2422 /*
2423  * NetBSD lstat.  Get file status; this version does not follow links.
2424  */
2425 #ifndef _SYS_SYSPROTO_H_
2426 struct freebsd11_nlstat_args {
2427 	char	*path;
2428 	struct nstat *ub;
2429 };
2430 #endif
2431 int
2432 freebsd11_nlstat(struct thread *td, struct freebsd11_nlstat_args *uap)
2433 {
2434 	struct stat sb;
2435 	struct nstat nsb;
2436 	int error;
2437 
2438 	error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2439 	    UIO_USERSPACE, &sb, NULL);
2440 	if (error != 0)
2441 		return (error);
2442 	freebsd11_cvtnstat(&sb, &nsb);
2443 	return (copyout(&nsb, uap->ub, sizeof (nsb)));
2444 }
2445 #endif /* COMPAT_FREEBSD11 */
2446 
2447 /*
2448  * Get configurable pathname variables.
2449  */
2450 #ifndef _SYS_SYSPROTO_H_
2451 struct pathconf_args {
2452 	char	*path;
2453 	int	name;
2454 };
2455 #endif
2456 int
2457 sys_pathconf(struct thread *td, struct pathconf_args *uap)
2458 {
2459 	long value;
2460 	int error;
2461 
2462 	error = kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name, FOLLOW,
2463 	    &value);
2464 	if (error == 0)
2465 		td->td_retval[0] = value;
2466 	return (error);
2467 }
2468 
2469 #ifndef _SYS_SYSPROTO_H_
2470 struct lpathconf_args {
2471 	char	*path;
2472 	int	name;
2473 };
2474 #endif
2475 int
2476 sys_lpathconf(struct thread *td, struct lpathconf_args *uap)
2477 {
2478 	long value;
2479 	int error;
2480 
2481 	error = kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name,
2482 	    NOFOLLOW, &value);
2483 	if (error == 0)
2484 		td->td_retval[0] = value;
2485 	return (error);
2486 }
2487 
2488 int
2489 kern_pathconf(struct thread *td, const char *path, enum uio_seg pathseg,
2490     int name, u_long flags, long *valuep)
2491 {
2492 	struct nameidata nd;
2493 	int error;
2494 
2495 	NDINIT(&nd, LOOKUP, LOCKSHARED | LOCKLEAF | AUDITVNODE1 | flags,
2496 	    pathseg, path, td);
2497 	if ((error = namei(&nd)) != 0)
2498 		return (error);
2499 	NDFREE(&nd, NDF_ONLY_PNBUF);
2500 
2501 	error = VOP_PATHCONF(nd.ni_vp, name, valuep);
2502 	vput(nd.ni_vp);
2503 	return (error);
2504 }
2505 
2506 /*
2507  * Return target name of a symbolic link.
2508  */
2509 #ifndef _SYS_SYSPROTO_H_
2510 struct readlink_args {
2511 	char	*path;
2512 	char	*buf;
2513 	size_t	count;
2514 };
2515 #endif
2516 int
2517 sys_readlink(struct thread *td, struct readlink_args *uap)
2518 {
2519 
2520 	return (kern_readlinkat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2521 	    uap->buf, UIO_USERSPACE, uap->count));
2522 }
2523 #ifndef _SYS_SYSPROTO_H_
2524 struct readlinkat_args {
2525 	int	fd;
2526 	char	*path;
2527 	char	*buf;
2528 	size_t	bufsize;
2529 };
2530 #endif
2531 int
2532 sys_readlinkat(struct thread *td, struct readlinkat_args *uap)
2533 {
2534 
2535 	return (kern_readlinkat(td, uap->fd, uap->path, UIO_USERSPACE,
2536 	    uap->buf, UIO_USERSPACE, uap->bufsize));
2537 }
2538 
2539 int
2540 kern_readlinkat(struct thread *td, int fd, const char *path,
2541     enum uio_seg pathseg, char *buf, enum uio_seg bufseg, size_t count)
2542 {
2543 	struct vnode *vp;
2544 	struct nameidata nd;
2545 	int error;
2546 
2547 	if (count > IOSIZE_MAX)
2548 		return (EINVAL);
2549 
2550 	NDINIT_AT(&nd, LOOKUP, NOFOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
2551 	    pathseg, path, fd, td);
2552 
2553 	if ((error = namei(&nd)) != 0)
2554 		return (error);
2555 	NDFREE(&nd, NDF_ONLY_PNBUF);
2556 	vp = nd.ni_vp;
2557 
2558 	error = kern_readlink_vp(vp, buf, bufseg, count, td);
2559 	vput(vp);
2560 
2561 	return (error);
2562 }
2563 
2564 /*
2565  * Helper function to readlink from a vnode
2566  */
2567 static int
2568 kern_readlink_vp(struct vnode *vp, char *buf, enum uio_seg bufseg, size_t count,
2569     struct thread *td)
2570 {
2571 	struct iovec aiov;
2572 	struct uio auio;
2573 	int error;
2574 
2575 	ASSERT_VOP_LOCKED(vp, "kern_readlink_vp(): vp not locked");
2576 #ifdef MAC
2577 	error = mac_vnode_check_readlink(td->td_ucred, vp);
2578 	if (error != 0)
2579 		return (error);
2580 #endif
2581 	if (vp->v_type != VLNK && (vp->v_vflag & VV_READLINK) == 0)
2582 		return (EINVAL);
2583 
2584 	aiov.iov_base = buf;
2585 	aiov.iov_len = count;
2586 	auio.uio_iov = &aiov;
2587 	auio.uio_iovcnt = 1;
2588 	auio.uio_offset = 0;
2589 	auio.uio_rw = UIO_READ;
2590 	auio.uio_segflg = bufseg;
2591 	auio.uio_td = td;
2592 	auio.uio_resid = count;
2593 	error = VOP_READLINK(vp, &auio, td->td_ucred);
2594 	td->td_retval[0] = count - auio.uio_resid;
2595 	return (error);
2596 }
2597 
2598 /*
2599  * Common implementation code for chflags() and fchflags().
2600  */
2601 static int
2602 setfflags(struct thread *td, struct vnode *vp, u_long flags)
2603 {
2604 	struct mount *mp;
2605 	struct vattr vattr;
2606 	int error;
2607 
2608 	/* We can't support the value matching VNOVAL. */
2609 	if (flags == VNOVAL)
2610 		return (EOPNOTSUPP);
2611 
2612 	/*
2613 	 * Prevent non-root users from setting flags on devices.  When
2614 	 * a device is reused, users can retain ownership of the device
2615 	 * if they are allowed to set flags and programs assume that
2616 	 * chown can't fail when done as root.
2617 	 */
2618 	if (vp->v_type == VCHR || vp->v_type == VBLK) {
2619 		error = priv_check(td, PRIV_VFS_CHFLAGS_DEV);
2620 		if (error != 0)
2621 			return (error);
2622 	}
2623 
2624 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2625 		return (error);
2626 	VATTR_NULL(&vattr);
2627 	vattr.va_flags = flags;
2628 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2629 #ifdef MAC
2630 	error = mac_vnode_check_setflags(td->td_ucred, vp, vattr.va_flags);
2631 	if (error == 0)
2632 #endif
2633 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
2634 	VOP_UNLOCK(vp, 0);
2635 	vn_finished_write(mp);
2636 	return (error);
2637 }
2638 
2639 /*
2640  * Change flags of a file given a path name.
2641  */
2642 #ifndef _SYS_SYSPROTO_H_
2643 struct chflags_args {
2644 	const char *path;
2645 	u_long	flags;
2646 };
2647 #endif
2648 int
2649 sys_chflags(struct thread *td, struct chflags_args *uap)
2650 {
2651 
2652 	return (kern_chflagsat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2653 	    uap->flags, 0));
2654 }
2655 
2656 #ifndef _SYS_SYSPROTO_H_
2657 struct chflagsat_args {
2658 	int	fd;
2659 	const char *path;
2660 	u_long	flags;
2661 	int	atflag;
2662 }
2663 #endif
2664 int
2665 sys_chflagsat(struct thread *td, struct chflagsat_args *uap)
2666 {
2667 
2668 	if ((uap->atflag & ~(AT_SYMLINK_NOFOLLOW | AT_BENEATH)) != 0)
2669 		return (EINVAL);
2670 
2671 	return (kern_chflagsat(td, uap->fd, uap->path, UIO_USERSPACE,
2672 	    uap->flags, uap->atflag));
2673 }
2674 
2675 /*
2676  * Same as chflags() but doesn't follow symlinks.
2677  */
2678 #ifndef _SYS_SYSPROTO_H_
2679 struct lchflags_args {
2680 	const char *path;
2681 	u_long flags;
2682 };
2683 #endif
2684 int
2685 sys_lchflags(struct thread *td, struct lchflags_args *uap)
2686 {
2687 
2688 	return (kern_chflagsat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2689 	    uap->flags, AT_SYMLINK_NOFOLLOW));
2690 }
2691 
2692 static int
2693 kern_chflagsat(struct thread *td, int fd, const char *path,
2694     enum uio_seg pathseg, u_long flags, int atflag)
2695 {
2696 	struct nameidata nd;
2697 	int error, follow;
2698 
2699 	AUDIT_ARG_FFLAGS(flags);
2700 	follow = (atflag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
2701 	follow |= (atflag & AT_BENEATH) != 0 ? BENEATH : 0;
2702 	NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, pathseg, path, fd,
2703 	    &cap_fchflags_rights, td);
2704 	if ((error = namei(&nd)) != 0)
2705 		return (error);
2706 	NDFREE(&nd, NDF_ONLY_PNBUF);
2707 	error = setfflags(td, nd.ni_vp, flags);
2708 	vrele(nd.ni_vp);
2709 	return (error);
2710 }
2711 
2712 /*
2713  * Change flags of a file given a file descriptor.
2714  */
2715 #ifndef _SYS_SYSPROTO_H_
2716 struct fchflags_args {
2717 	int	fd;
2718 	u_long	flags;
2719 };
2720 #endif
2721 int
2722 sys_fchflags(struct thread *td, struct fchflags_args *uap)
2723 {
2724 	struct file *fp;
2725 	int error;
2726 
2727 	AUDIT_ARG_FD(uap->fd);
2728 	AUDIT_ARG_FFLAGS(uap->flags);
2729 	error = getvnode(td, uap->fd, &cap_fchflags_rights,
2730 	    &fp);
2731 	if (error != 0)
2732 		return (error);
2733 #ifdef AUDIT
2734 	vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
2735 	AUDIT_ARG_VNODE1(fp->f_vnode);
2736 	VOP_UNLOCK(fp->f_vnode, 0);
2737 #endif
2738 	error = setfflags(td, fp->f_vnode, uap->flags);
2739 	fdrop(fp, td);
2740 	return (error);
2741 }
2742 
2743 /*
2744  * Common implementation code for chmod(), lchmod() and fchmod().
2745  */
2746 int
2747 setfmode(struct thread *td, struct ucred *cred, struct vnode *vp, int mode)
2748 {
2749 	struct mount *mp;
2750 	struct vattr vattr;
2751 	int error;
2752 
2753 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2754 		return (error);
2755 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2756 	VATTR_NULL(&vattr);
2757 	vattr.va_mode = mode & ALLPERMS;
2758 #ifdef MAC
2759 	error = mac_vnode_check_setmode(cred, vp, vattr.va_mode);
2760 	if (error == 0)
2761 #endif
2762 		error = VOP_SETATTR(vp, &vattr, cred);
2763 	VOP_UNLOCK(vp, 0);
2764 	vn_finished_write(mp);
2765 	return (error);
2766 }
2767 
2768 /*
2769  * Change mode of a file given path name.
2770  */
2771 #ifndef _SYS_SYSPROTO_H_
2772 struct chmod_args {
2773 	char	*path;
2774 	int	mode;
2775 };
2776 #endif
2777 int
2778 sys_chmod(struct thread *td, struct chmod_args *uap)
2779 {
2780 
2781 	return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2782 	    uap->mode, 0));
2783 }
2784 
2785 #ifndef _SYS_SYSPROTO_H_
2786 struct fchmodat_args {
2787 	int	dirfd;
2788 	char	*path;
2789 	mode_t	mode;
2790 	int	flag;
2791 }
2792 #endif
2793 int
2794 sys_fchmodat(struct thread *td, struct fchmodat_args *uap)
2795 {
2796 
2797 	if ((uap->flag & ~(AT_SYMLINK_NOFOLLOW | AT_BENEATH)) != 0)
2798 		return (EINVAL);
2799 
2800 	return (kern_fchmodat(td, uap->fd, uap->path, UIO_USERSPACE,
2801 	    uap->mode, uap->flag));
2802 }
2803 
2804 /*
2805  * Change mode of a file given path name (don't follow links.)
2806  */
2807 #ifndef _SYS_SYSPROTO_H_
2808 struct lchmod_args {
2809 	char	*path;
2810 	int	mode;
2811 };
2812 #endif
2813 int
2814 sys_lchmod(struct thread *td, struct lchmod_args *uap)
2815 {
2816 
2817 	return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2818 	    uap->mode, AT_SYMLINK_NOFOLLOW));
2819 }
2820 
2821 int
2822 kern_fchmodat(struct thread *td, int fd, const char *path,
2823     enum uio_seg pathseg, mode_t mode, int flag)
2824 {
2825 	struct nameidata nd;
2826 	int error, follow;
2827 
2828 	AUDIT_ARG_MODE(mode);
2829 	follow = (flag & AT_SYMLINK_NOFOLLOW) != 0 ? NOFOLLOW : FOLLOW;
2830 	follow |= (flag & AT_BENEATH) != 0 ? BENEATH : 0;
2831 	NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, pathseg, path, fd,
2832 	    &cap_fchmod_rights, td);
2833 	if ((error = namei(&nd)) != 0)
2834 		return (error);
2835 	NDFREE(&nd, NDF_ONLY_PNBUF);
2836 	error = setfmode(td, td->td_ucred, nd.ni_vp, mode);
2837 	vrele(nd.ni_vp);
2838 	return (error);
2839 }
2840 
2841 /*
2842  * Change mode of a file given a file descriptor.
2843  */
2844 #ifndef _SYS_SYSPROTO_H_
2845 struct fchmod_args {
2846 	int	fd;
2847 	int	mode;
2848 };
2849 #endif
2850 int
2851 sys_fchmod(struct thread *td, struct fchmod_args *uap)
2852 {
2853 	struct file *fp;
2854 	int error;
2855 
2856 	AUDIT_ARG_FD(uap->fd);
2857 	AUDIT_ARG_MODE(uap->mode);
2858 
2859 	error = fget(td, uap->fd, &cap_fchmod_rights, &fp);
2860 	if (error != 0)
2861 		return (error);
2862 	error = fo_chmod(fp, uap->mode, td->td_ucred, td);
2863 	fdrop(fp, td);
2864 	return (error);
2865 }
2866 
2867 /*
2868  * Common implementation for chown(), lchown(), and fchown()
2869  */
2870 int
2871 setfown(struct thread *td, struct ucred *cred, struct vnode *vp, uid_t uid,
2872     gid_t gid)
2873 {
2874 	struct mount *mp;
2875 	struct vattr vattr;
2876 	int error;
2877 
2878 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2879 		return (error);
2880 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2881 	VATTR_NULL(&vattr);
2882 	vattr.va_uid = uid;
2883 	vattr.va_gid = gid;
2884 #ifdef MAC
2885 	error = mac_vnode_check_setowner(cred, vp, vattr.va_uid,
2886 	    vattr.va_gid);
2887 	if (error == 0)
2888 #endif
2889 		error = VOP_SETATTR(vp, &vattr, cred);
2890 	VOP_UNLOCK(vp, 0);
2891 	vn_finished_write(mp);
2892 	return (error);
2893 }
2894 
2895 /*
2896  * Set ownership given a path name.
2897  */
2898 #ifndef _SYS_SYSPROTO_H_
2899 struct chown_args {
2900 	char	*path;
2901 	int	uid;
2902 	int	gid;
2903 };
2904 #endif
2905 int
2906 sys_chown(struct thread *td, struct chown_args *uap)
2907 {
2908 
2909 	return (kern_fchownat(td, AT_FDCWD, uap->path, UIO_USERSPACE, uap->uid,
2910 	    uap->gid, 0));
2911 }
2912 
2913 #ifndef _SYS_SYSPROTO_H_
2914 struct fchownat_args {
2915 	int fd;
2916 	const char * path;
2917 	uid_t uid;
2918 	gid_t gid;
2919 	int flag;
2920 };
2921 #endif
2922 int
2923 sys_fchownat(struct thread *td, struct fchownat_args *uap)
2924 {
2925 
2926 	if ((uap->flag & ~(AT_SYMLINK_NOFOLLOW | AT_BENEATH)) != 0)
2927 		return (EINVAL);
2928 
2929 	return (kern_fchownat(td, uap->fd, uap->path, UIO_USERSPACE, uap->uid,
2930 	    uap->gid, uap->flag));
2931 }
2932 
2933 int
2934 kern_fchownat(struct thread *td, int fd, const char *path,
2935     enum uio_seg pathseg, int uid, int gid, int flag)
2936 {
2937 	struct nameidata nd;
2938 	int error, follow;
2939 
2940 	AUDIT_ARG_OWNER(uid, gid);
2941 	follow = (flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
2942 	follow |= (flag & AT_BENEATH) != 0 ? BENEATH : 0;
2943 	NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, pathseg, path, fd,
2944 	    &cap_fchown_rights, td);
2945 
2946 	if ((error = namei(&nd)) != 0)
2947 		return (error);
2948 	NDFREE(&nd, NDF_ONLY_PNBUF);
2949 	error = setfown(td, td->td_ucred, nd.ni_vp, uid, gid);
2950 	vrele(nd.ni_vp);
2951 	return (error);
2952 }
2953 
2954 /*
2955  * Set ownership given a path name, do not cross symlinks.
2956  */
2957 #ifndef _SYS_SYSPROTO_H_
2958 struct lchown_args {
2959 	char	*path;
2960 	int	uid;
2961 	int	gid;
2962 };
2963 #endif
2964 int
2965 sys_lchown(struct thread *td, struct lchown_args *uap)
2966 {
2967 
2968 	return (kern_fchownat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2969 	    uap->uid, uap->gid, AT_SYMLINK_NOFOLLOW));
2970 }
2971 
2972 /*
2973  * Set ownership given a file descriptor.
2974  */
2975 #ifndef _SYS_SYSPROTO_H_
2976 struct fchown_args {
2977 	int	fd;
2978 	int	uid;
2979 	int	gid;
2980 };
2981 #endif
2982 int
2983 sys_fchown(struct thread *td, struct fchown_args *uap)
2984 {
2985 	struct file *fp;
2986 	int error;
2987 
2988 	AUDIT_ARG_FD(uap->fd);
2989 	AUDIT_ARG_OWNER(uap->uid, uap->gid);
2990 	error = fget(td, uap->fd, &cap_fchown_rights, &fp);
2991 	if (error != 0)
2992 		return (error);
2993 	error = fo_chown(fp, uap->uid, uap->gid, td->td_ucred, td);
2994 	fdrop(fp, td);
2995 	return (error);
2996 }
2997 
2998 /*
2999  * Common implementation code for utimes(), lutimes(), and futimes().
3000  */
3001 static int
3002 getutimes(const struct timeval *usrtvp, enum uio_seg tvpseg,
3003     struct timespec *tsp)
3004 {
3005 	struct timeval tv[2];
3006 	const struct timeval *tvp;
3007 	int error;
3008 
3009 	if (usrtvp == NULL) {
3010 		vfs_timestamp(&tsp[0]);
3011 		tsp[1] = tsp[0];
3012 	} else {
3013 		if (tvpseg == UIO_SYSSPACE) {
3014 			tvp = usrtvp;
3015 		} else {
3016 			if ((error = copyin(usrtvp, tv, sizeof(tv))) != 0)
3017 				return (error);
3018 			tvp = tv;
3019 		}
3020 
3021 		if (tvp[0].tv_usec < 0 || tvp[0].tv_usec >= 1000000 ||
3022 		    tvp[1].tv_usec < 0 || tvp[1].tv_usec >= 1000000)
3023 			return (EINVAL);
3024 		TIMEVAL_TO_TIMESPEC(&tvp[0], &tsp[0]);
3025 		TIMEVAL_TO_TIMESPEC(&tvp[1], &tsp[1]);
3026 	}
3027 	return (0);
3028 }
3029 
3030 /*
3031  * Common implementation code for futimens(), utimensat().
3032  */
3033 #define	UTIMENS_NULL	0x1
3034 #define	UTIMENS_EXIT	0x2
3035 static int
3036 getutimens(const struct timespec *usrtsp, enum uio_seg tspseg,
3037     struct timespec *tsp, int *retflags)
3038 {
3039 	struct timespec tsnow;
3040 	int error;
3041 
3042 	vfs_timestamp(&tsnow);
3043 	*retflags = 0;
3044 	if (usrtsp == NULL) {
3045 		tsp[0] = tsnow;
3046 		tsp[1] = tsnow;
3047 		*retflags |= UTIMENS_NULL;
3048 		return (0);
3049 	}
3050 	if (tspseg == UIO_SYSSPACE) {
3051 		tsp[0] = usrtsp[0];
3052 		tsp[1] = usrtsp[1];
3053 	} else if ((error = copyin(usrtsp, tsp, sizeof(*tsp) * 2)) != 0)
3054 		return (error);
3055 	if (tsp[0].tv_nsec == UTIME_OMIT && tsp[1].tv_nsec == UTIME_OMIT)
3056 		*retflags |= UTIMENS_EXIT;
3057 	if (tsp[0].tv_nsec == UTIME_NOW && tsp[1].tv_nsec == UTIME_NOW)
3058 		*retflags |= UTIMENS_NULL;
3059 	if (tsp[0].tv_nsec == UTIME_OMIT)
3060 		tsp[0].tv_sec = VNOVAL;
3061 	else if (tsp[0].tv_nsec == UTIME_NOW)
3062 		tsp[0] = tsnow;
3063 	else if (tsp[0].tv_nsec < 0 || tsp[0].tv_nsec >= 1000000000L)
3064 		return (EINVAL);
3065 	if (tsp[1].tv_nsec == UTIME_OMIT)
3066 		tsp[1].tv_sec = VNOVAL;
3067 	else if (tsp[1].tv_nsec == UTIME_NOW)
3068 		tsp[1] = tsnow;
3069 	else if (tsp[1].tv_nsec < 0 || tsp[1].tv_nsec >= 1000000000L)
3070 		return (EINVAL);
3071 
3072 	return (0);
3073 }
3074 
3075 /*
3076  * Common implementation code for utimes(), lutimes(), futimes(), futimens(),
3077  * and utimensat().
3078  */
3079 static int
3080 setutimes(struct thread *td, struct vnode *vp, const struct timespec *ts,
3081     int numtimes, int nullflag)
3082 {
3083 	struct mount *mp;
3084 	struct vattr vattr;
3085 	int error, setbirthtime;
3086 
3087 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
3088 		return (error);
3089 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3090 	setbirthtime = 0;
3091 	if (numtimes < 3 && !VOP_GETATTR(vp, &vattr, td->td_ucred) &&
3092 	    timespeccmp(&ts[1], &vattr.va_birthtime, < ))
3093 		setbirthtime = 1;
3094 	VATTR_NULL(&vattr);
3095 	vattr.va_atime = ts[0];
3096 	vattr.va_mtime = ts[1];
3097 	if (setbirthtime)
3098 		vattr.va_birthtime = ts[1];
3099 	if (numtimes > 2)
3100 		vattr.va_birthtime = ts[2];
3101 	if (nullflag)
3102 		vattr.va_vaflags |= VA_UTIMES_NULL;
3103 #ifdef MAC
3104 	error = mac_vnode_check_setutimes(td->td_ucred, vp, vattr.va_atime,
3105 	    vattr.va_mtime);
3106 #endif
3107 	if (error == 0)
3108 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3109 	VOP_UNLOCK(vp, 0);
3110 	vn_finished_write(mp);
3111 	return (error);
3112 }
3113 
3114 /*
3115  * Set the access and modification times of a file.
3116  */
3117 #ifndef _SYS_SYSPROTO_H_
3118 struct utimes_args {
3119 	char	*path;
3120 	struct	timeval *tptr;
3121 };
3122 #endif
3123 int
3124 sys_utimes(struct thread *td, struct utimes_args *uap)
3125 {
3126 
3127 	return (kern_utimesat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
3128 	    uap->tptr, UIO_USERSPACE));
3129 }
3130 
3131 #ifndef _SYS_SYSPROTO_H_
3132 struct futimesat_args {
3133 	int fd;
3134 	const char * path;
3135 	const struct timeval * times;
3136 };
3137 #endif
3138 int
3139 sys_futimesat(struct thread *td, struct futimesat_args *uap)
3140 {
3141 
3142 	return (kern_utimesat(td, uap->fd, uap->path, UIO_USERSPACE,
3143 	    uap->times, UIO_USERSPACE));
3144 }
3145 
3146 int
3147 kern_utimesat(struct thread *td, int fd, const char *path,
3148     enum uio_seg pathseg, struct timeval *tptr, enum uio_seg tptrseg)
3149 {
3150 	struct nameidata nd;
3151 	struct timespec ts[2];
3152 	int error;
3153 
3154 	if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3155 		return (error);
3156 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, fd,
3157 	    &cap_futimes_rights, td);
3158 
3159 	if ((error = namei(&nd)) != 0)
3160 		return (error);
3161 	NDFREE(&nd, NDF_ONLY_PNBUF);
3162 	error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3163 	vrele(nd.ni_vp);
3164 	return (error);
3165 }
3166 
3167 /*
3168  * Set the access and modification times of a file.
3169  */
3170 #ifndef _SYS_SYSPROTO_H_
3171 struct lutimes_args {
3172 	char	*path;
3173 	struct	timeval *tptr;
3174 };
3175 #endif
3176 int
3177 sys_lutimes(struct thread *td, struct lutimes_args *uap)
3178 {
3179 
3180 	return (kern_lutimes(td, uap->path, UIO_USERSPACE, uap->tptr,
3181 	    UIO_USERSPACE));
3182 }
3183 
3184 int
3185 kern_lutimes(struct thread *td, const char *path, enum uio_seg pathseg,
3186     struct timeval *tptr, enum uio_seg tptrseg)
3187 {
3188 	struct timespec ts[2];
3189 	struct nameidata nd;
3190 	int error;
3191 
3192 	if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3193 		return (error);
3194 	NDINIT(&nd, LOOKUP, NOFOLLOW | AUDITVNODE1, pathseg, path, td);
3195 	if ((error = namei(&nd)) != 0)
3196 		return (error);
3197 	NDFREE(&nd, NDF_ONLY_PNBUF);
3198 	error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3199 	vrele(nd.ni_vp);
3200 	return (error);
3201 }
3202 
3203 /*
3204  * Set the access and modification times of a file.
3205  */
3206 #ifndef _SYS_SYSPROTO_H_
3207 struct futimes_args {
3208 	int	fd;
3209 	struct	timeval *tptr;
3210 };
3211 #endif
3212 int
3213 sys_futimes(struct thread *td, struct futimes_args *uap)
3214 {
3215 
3216 	return (kern_futimes(td, uap->fd, uap->tptr, UIO_USERSPACE));
3217 }
3218 
3219 int
3220 kern_futimes(struct thread *td, int fd, struct timeval *tptr,
3221     enum uio_seg tptrseg)
3222 {
3223 	struct timespec ts[2];
3224 	struct file *fp;
3225 	int error;
3226 
3227 	AUDIT_ARG_FD(fd);
3228 	error = getutimes(tptr, tptrseg, ts);
3229 	if (error != 0)
3230 		return (error);
3231 	error = getvnode(td, fd, &cap_futimes_rights, &fp);
3232 	if (error != 0)
3233 		return (error);
3234 #ifdef AUDIT
3235 	vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
3236 	AUDIT_ARG_VNODE1(fp->f_vnode);
3237 	VOP_UNLOCK(fp->f_vnode, 0);
3238 #endif
3239 	error = setutimes(td, fp->f_vnode, ts, 2, tptr == NULL);
3240 	fdrop(fp, td);
3241 	return (error);
3242 }
3243 
3244 int
3245 sys_futimens(struct thread *td, struct futimens_args *uap)
3246 {
3247 
3248 	return (kern_futimens(td, uap->fd, uap->times, UIO_USERSPACE));
3249 }
3250 
3251 int
3252 kern_futimens(struct thread *td, int fd, struct timespec *tptr,
3253     enum uio_seg tptrseg)
3254 {
3255 	struct timespec ts[2];
3256 	struct file *fp;
3257 	int error, flags;
3258 
3259 	AUDIT_ARG_FD(fd);
3260 	error = getutimens(tptr, tptrseg, ts, &flags);
3261 	if (error != 0)
3262 		return (error);
3263 	if (flags & UTIMENS_EXIT)
3264 		return (0);
3265 	error = getvnode(td, fd, &cap_futimes_rights, &fp);
3266 	if (error != 0)
3267 		return (error);
3268 #ifdef AUDIT
3269 	vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
3270 	AUDIT_ARG_VNODE1(fp->f_vnode);
3271 	VOP_UNLOCK(fp->f_vnode, 0);
3272 #endif
3273 	error = setutimes(td, fp->f_vnode, ts, 2, flags & UTIMENS_NULL);
3274 	fdrop(fp, td);
3275 	return (error);
3276 }
3277 
3278 int
3279 sys_utimensat(struct thread *td, struct utimensat_args *uap)
3280 {
3281 
3282 	return (kern_utimensat(td, uap->fd, uap->path, UIO_USERSPACE,
3283 	    uap->times, UIO_USERSPACE, uap->flag));
3284 }
3285 
3286 int
3287 kern_utimensat(struct thread *td, int fd, const char *path,
3288     enum uio_seg pathseg, struct timespec *tptr, enum uio_seg tptrseg,
3289     int flag)
3290 {
3291 	struct nameidata nd;
3292 	struct timespec ts[2];
3293 	int error, flags;
3294 
3295 	if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_BENEATH)) != 0)
3296 		return (EINVAL);
3297 
3298 	if ((error = getutimens(tptr, tptrseg, ts, &flags)) != 0)
3299 		return (error);
3300 	NDINIT_ATRIGHTS(&nd, LOOKUP, ((flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW :
3301 	    FOLLOW) |  ((flag & AT_BENEATH) != 0 ? BENEATH : 0) | AUDITVNODE1,
3302 	    pathseg, path, fd, &cap_futimes_rights, td);
3303 	if ((error = namei(&nd)) != 0)
3304 		return (error);
3305 	/*
3306 	 * We are allowed to call namei() regardless of 2xUTIME_OMIT.
3307 	 * POSIX states:
3308 	 * "If both tv_nsec fields are UTIME_OMIT... EACCESS may be detected."
3309 	 * "Search permission is denied by a component of the path prefix."
3310 	 */
3311 	NDFREE(&nd, NDF_ONLY_PNBUF);
3312 	if ((flags & UTIMENS_EXIT) == 0)
3313 		error = setutimes(td, nd.ni_vp, ts, 2, flags & UTIMENS_NULL);
3314 	vrele(nd.ni_vp);
3315 	return (error);
3316 }
3317 
3318 /*
3319  * Truncate a file given its path name.
3320  */
3321 #ifndef _SYS_SYSPROTO_H_
3322 struct truncate_args {
3323 	char	*path;
3324 	int	pad;
3325 	off_t	length;
3326 };
3327 #endif
3328 int
3329 sys_truncate(struct thread *td, struct truncate_args *uap)
3330 {
3331 
3332 	return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3333 }
3334 
3335 int
3336 kern_truncate(struct thread *td, const char *path, enum uio_seg pathseg,
3337     off_t length)
3338 {
3339 	struct mount *mp;
3340 	struct vnode *vp;
3341 	void *rl_cookie;
3342 	struct vattr vattr;
3343 	struct nameidata nd;
3344 	int error;
3345 
3346 	if (length < 0)
3347 		return(EINVAL);
3348 	NDINIT(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, td);
3349 	if ((error = namei(&nd)) != 0)
3350 		return (error);
3351 	vp = nd.ni_vp;
3352 	rl_cookie = vn_rangelock_wlock(vp, 0, OFF_MAX);
3353 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) {
3354 		vn_rangelock_unlock(vp, rl_cookie);
3355 		vrele(vp);
3356 		return (error);
3357 	}
3358 	NDFREE(&nd, NDF_ONLY_PNBUF);
3359 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3360 	if (vp->v_type == VDIR)
3361 		error = EISDIR;
3362 #ifdef MAC
3363 	else if ((error = mac_vnode_check_write(td->td_ucred, NOCRED, vp))) {
3364 	}
3365 #endif
3366 	else if ((error = vn_writechk(vp)) == 0 &&
3367 	    (error = VOP_ACCESS(vp, VWRITE, td->td_ucred, td)) == 0) {
3368 		VATTR_NULL(&vattr);
3369 		vattr.va_size = length;
3370 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3371 	}
3372 	VOP_UNLOCK(vp, 0);
3373 	vn_finished_write(mp);
3374 	vn_rangelock_unlock(vp, rl_cookie);
3375 	vrele(vp);
3376 	return (error);
3377 }
3378 
3379 #if defined(COMPAT_43)
3380 /*
3381  * Truncate a file given its path name.
3382  */
3383 #ifndef _SYS_SYSPROTO_H_
3384 struct otruncate_args {
3385 	char	*path;
3386 	long	length;
3387 };
3388 #endif
3389 int
3390 otruncate(struct thread *td, struct otruncate_args *uap)
3391 {
3392 
3393 	return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3394 }
3395 #endif /* COMPAT_43 */
3396 
3397 #if defined(COMPAT_FREEBSD6)
3398 /* Versions with the pad argument */
3399 int
3400 freebsd6_truncate(struct thread *td, struct freebsd6_truncate_args *uap)
3401 {
3402 
3403 	return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3404 }
3405 
3406 int
3407 freebsd6_ftruncate(struct thread *td, struct freebsd6_ftruncate_args *uap)
3408 {
3409 
3410 	return (kern_ftruncate(td, uap->fd, uap->length));
3411 }
3412 #endif
3413 
3414 int
3415 kern_fsync(struct thread *td, int fd, bool fullsync)
3416 {
3417 	struct vnode *vp;
3418 	struct mount *mp;
3419 	struct file *fp;
3420 	int error, lock_flags;
3421 
3422 	AUDIT_ARG_FD(fd);
3423 	error = getvnode(td, fd, &cap_fsync_rights, &fp);
3424 	if (error != 0)
3425 		return (error);
3426 	vp = fp->f_vnode;
3427 #if 0
3428 	if (!fullsync)
3429 		/* XXXKIB: compete outstanding aio writes */;
3430 #endif
3431 	error = vn_start_write(vp, &mp, V_WAIT | PCATCH);
3432 	if (error != 0)
3433 		goto drop;
3434 	if (MNT_SHARED_WRITES(mp) ||
3435 	    ((mp == NULL) && MNT_SHARED_WRITES(vp->v_mount))) {
3436 		lock_flags = LK_SHARED;
3437 	} else {
3438 		lock_flags = LK_EXCLUSIVE;
3439 	}
3440 	vn_lock(vp, lock_flags | LK_RETRY);
3441 	AUDIT_ARG_VNODE1(vp);
3442 	if (vp->v_object != NULL) {
3443 		VM_OBJECT_WLOCK(vp->v_object);
3444 		vm_object_page_clean(vp->v_object, 0, 0, 0);
3445 		VM_OBJECT_WUNLOCK(vp->v_object);
3446 	}
3447 	error = fullsync ? VOP_FSYNC(vp, MNT_WAIT, td) : VOP_FDATASYNC(vp, td);
3448 	VOP_UNLOCK(vp, 0);
3449 	vn_finished_write(mp);
3450 drop:
3451 	fdrop(fp, td);
3452 	return (error);
3453 }
3454 
3455 /*
3456  * Sync an open file.
3457  */
3458 #ifndef _SYS_SYSPROTO_H_
3459 struct fsync_args {
3460 	int	fd;
3461 };
3462 #endif
3463 int
3464 sys_fsync(struct thread *td, struct fsync_args *uap)
3465 {
3466 
3467 	return (kern_fsync(td, uap->fd, true));
3468 }
3469 
3470 int
3471 sys_fdatasync(struct thread *td, struct fdatasync_args *uap)
3472 {
3473 
3474 	return (kern_fsync(td, uap->fd, false));
3475 }
3476 
3477 /*
3478  * Rename files.  Source and destination must either both be directories, or
3479  * both not be directories.  If target is a directory, it must be empty.
3480  */
3481 #ifndef _SYS_SYSPROTO_H_
3482 struct rename_args {
3483 	char	*from;
3484 	char	*to;
3485 };
3486 #endif
3487 int
3488 sys_rename(struct thread *td, struct rename_args *uap)
3489 {
3490 
3491 	return (kern_renameat(td, AT_FDCWD, uap->from, AT_FDCWD,
3492 	    uap->to, UIO_USERSPACE));
3493 }
3494 
3495 #ifndef _SYS_SYSPROTO_H_
3496 struct renameat_args {
3497 	int	oldfd;
3498 	char	*old;
3499 	int	newfd;
3500 	char	*new;
3501 };
3502 #endif
3503 int
3504 sys_renameat(struct thread *td, struct renameat_args *uap)
3505 {
3506 
3507 	return (kern_renameat(td, uap->oldfd, uap->old, uap->newfd, uap->new,
3508 	    UIO_USERSPACE));
3509 }
3510 
3511 int
3512 kern_renameat(struct thread *td, int oldfd, const char *old, int newfd,
3513     const char *new, enum uio_seg pathseg)
3514 {
3515 	struct mount *mp = NULL;
3516 	struct vnode *tvp, *fvp, *tdvp;
3517 	struct nameidata fromnd, tond;
3518 	int error;
3519 
3520 again:
3521 	bwillwrite();
3522 #ifdef MAC
3523 	NDINIT_ATRIGHTS(&fromnd, DELETE, LOCKPARENT | LOCKLEAF | SAVESTART |
3524 	    AUDITVNODE1, pathseg, old, oldfd,
3525 	    &cap_renameat_source_rights, td);
3526 #else
3527 	NDINIT_ATRIGHTS(&fromnd, DELETE, WANTPARENT | SAVESTART | AUDITVNODE1,
3528 	    pathseg, old, oldfd,
3529 	    &cap_renameat_source_rights, td);
3530 #endif
3531 
3532 	if ((error = namei(&fromnd)) != 0)
3533 		return (error);
3534 #ifdef MAC
3535 	error = mac_vnode_check_rename_from(td->td_ucred, fromnd.ni_dvp,
3536 	    fromnd.ni_vp, &fromnd.ni_cnd);
3537 	VOP_UNLOCK(fromnd.ni_dvp, 0);
3538 	if (fromnd.ni_dvp != fromnd.ni_vp)
3539 		VOP_UNLOCK(fromnd.ni_vp, 0);
3540 #endif
3541 	fvp = fromnd.ni_vp;
3542 	NDINIT_ATRIGHTS(&tond, RENAME, LOCKPARENT | LOCKLEAF | NOCACHE |
3543 	    SAVESTART | AUDITVNODE2, pathseg, new, newfd,
3544 	    &cap_renameat_target_rights, td);
3545 	if (fromnd.ni_vp->v_type == VDIR)
3546 		tond.ni_cnd.cn_flags |= WILLBEDIR;
3547 	if ((error = namei(&tond)) != 0) {
3548 		/* Translate error code for rename("dir1", "dir2/."). */
3549 		if (error == EISDIR && fvp->v_type == VDIR)
3550 			error = EINVAL;
3551 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3552 		vrele(fromnd.ni_dvp);
3553 		vrele(fvp);
3554 		goto out1;
3555 	}
3556 	tdvp = tond.ni_dvp;
3557 	tvp = tond.ni_vp;
3558 	error = vn_start_write(fvp, &mp, V_NOWAIT);
3559 	if (error != 0) {
3560 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3561 		NDFREE(&tond, NDF_ONLY_PNBUF);
3562 		if (tvp != NULL)
3563 			vput(tvp);
3564 		if (tdvp == tvp)
3565 			vrele(tdvp);
3566 		else
3567 			vput(tdvp);
3568 		vrele(fromnd.ni_dvp);
3569 		vrele(fvp);
3570 		vrele(tond.ni_startdir);
3571 		if (fromnd.ni_startdir != NULL)
3572 			vrele(fromnd.ni_startdir);
3573 		error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH);
3574 		if (error != 0)
3575 			return (error);
3576 		goto again;
3577 	}
3578 	if (tvp != NULL) {
3579 		if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
3580 			error = ENOTDIR;
3581 			goto out;
3582 		} else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
3583 			error = EISDIR;
3584 			goto out;
3585 		}
3586 #ifdef CAPABILITIES
3587 		if (newfd != AT_FDCWD && (tond.ni_resflags & NIRES_ABS) == 0) {
3588 			/*
3589 			 * If the target already exists we require CAP_UNLINKAT
3590 			 * from 'newfd', when newfd was used for the lookup.
3591 			 */
3592 			error = cap_check(&tond.ni_filecaps.fc_rights,
3593 			    &cap_unlinkat_rights);
3594 			if (error != 0)
3595 				goto out;
3596 		}
3597 #endif
3598 	}
3599 	if (fvp == tdvp) {
3600 		error = EINVAL;
3601 		goto out;
3602 	}
3603 	/*
3604 	 * If the source is the same as the destination (that is, if they
3605 	 * are links to the same vnode), then there is nothing to do.
3606 	 */
3607 	if (fvp == tvp)
3608 		error = -1;
3609 #ifdef MAC
3610 	else
3611 		error = mac_vnode_check_rename_to(td->td_ucred, tdvp,
3612 		    tond.ni_vp, fromnd.ni_dvp == tdvp, &tond.ni_cnd);
3613 #endif
3614 out:
3615 	if (error == 0) {
3616 		error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
3617 		    tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
3618 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3619 		NDFREE(&tond, NDF_ONLY_PNBUF);
3620 	} else {
3621 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3622 		NDFREE(&tond, NDF_ONLY_PNBUF);
3623 		if (tvp != NULL)
3624 			vput(tvp);
3625 		if (tdvp == tvp)
3626 			vrele(tdvp);
3627 		else
3628 			vput(tdvp);
3629 		vrele(fromnd.ni_dvp);
3630 		vrele(fvp);
3631 	}
3632 	vrele(tond.ni_startdir);
3633 	vn_finished_write(mp);
3634 out1:
3635 	if (fromnd.ni_startdir)
3636 		vrele(fromnd.ni_startdir);
3637 	if (error == -1)
3638 		return (0);
3639 	return (error);
3640 }
3641 
3642 /*
3643  * Make a directory file.
3644  */
3645 #ifndef _SYS_SYSPROTO_H_
3646 struct mkdir_args {
3647 	char	*path;
3648 	int	mode;
3649 };
3650 #endif
3651 int
3652 sys_mkdir(struct thread *td, struct mkdir_args *uap)
3653 {
3654 
3655 	return (kern_mkdirat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
3656 	    uap->mode));
3657 }
3658 
3659 #ifndef _SYS_SYSPROTO_H_
3660 struct mkdirat_args {
3661 	int	fd;
3662 	char	*path;
3663 	mode_t	mode;
3664 };
3665 #endif
3666 int
3667 sys_mkdirat(struct thread *td, struct mkdirat_args *uap)
3668 {
3669 
3670 	return (kern_mkdirat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode));
3671 }
3672 
3673 int
3674 kern_mkdirat(struct thread *td, int fd, const char *path, enum uio_seg segflg,
3675     int mode)
3676 {
3677 	struct mount *mp;
3678 	struct vnode *vp;
3679 	struct vattr vattr;
3680 	struct nameidata nd;
3681 	int error;
3682 
3683 	AUDIT_ARG_MODE(mode);
3684 restart:
3685 	bwillwrite();
3686 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
3687 	    NOCACHE, segflg, path, fd, &cap_mkdirat_rights,
3688 	    td);
3689 	nd.ni_cnd.cn_flags |= WILLBEDIR;
3690 	if ((error = namei(&nd)) != 0)
3691 		return (error);
3692 	vp = nd.ni_vp;
3693 	if (vp != NULL) {
3694 		NDFREE(&nd, NDF_ONLY_PNBUF);
3695 		/*
3696 		 * XXX namei called with LOCKPARENT but not LOCKLEAF has
3697 		 * the strange behaviour of leaving the vnode unlocked
3698 		 * if the target is the same vnode as the parent.
3699 		 */
3700 		if (vp == nd.ni_dvp)
3701 			vrele(nd.ni_dvp);
3702 		else
3703 			vput(nd.ni_dvp);
3704 		vrele(vp);
3705 		return (EEXIST);
3706 	}
3707 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3708 		NDFREE(&nd, NDF_ONLY_PNBUF);
3709 		vput(nd.ni_dvp);
3710 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3711 			return (error);
3712 		goto restart;
3713 	}
3714 	VATTR_NULL(&vattr);
3715 	vattr.va_type = VDIR;
3716 	vattr.va_mode = (mode & ACCESSPERMS) &~ td->td_proc->p_fd->fd_cmask;
3717 #ifdef MAC
3718 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
3719 	    &vattr);
3720 	if (error != 0)
3721 		goto out;
3722 #endif
3723 	error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
3724 #ifdef MAC
3725 out:
3726 #endif
3727 	NDFREE(&nd, NDF_ONLY_PNBUF);
3728 	vput(nd.ni_dvp);
3729 	if (error == 0)
3730 		vput(nd.ni_vp);
3731 	vn_finished_write(mp);
3732 	return (error);
3733 }
3734 
3735 /*
3736  * Remove a directory file.
3737  */
3738 #ifndef _SYS_SYSPROTO_H_
3739 struct rmdir_args {
3740 	char	*path;
3741 };
3742 #endif
3743 int
3744 sys_rmdir(struct thread *td, struct rmdir_args *uap)
3745 {
3746 
3747 	return (kern_frmdirat(td, AT_FDCWD, uap->path, FD_NONE, UIO_USERSPACE,
3748 	    0));
3749 }
3750 
3751 int
3752 kern_frmdirat(struct thread *td, int dfd, const char *path, int fd,
3753     enum uio_seg pathseg, int flag)
3754 {
3755 	struct mount *mp;
3756 	struct vnode *vp;
3757 	struct file *fp;
3758 	struct nameidata nd;
3759 	cap_rights_t rights;
3760 	int error;
3761 
3762 	fp = NULL;
3763 	if (fd != FD_NONE) {
3764 		error = getvnode(td, fd, cap_rights_init(&rights, CAP_LOOKUP),
3765 		    &fp);
3766 		if (error != 0)
3767 			return (error);
3768 	}
3769 
3770 restart:
3771 	bwillwrite();
3772 	NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1 |
3773 	    ((flag & AT_BENEATH) != 0 ? BENEATH : 0),
3774 	    pathseg, path, dfd, &cap_unlinkat_rights, td);
3775 	if ((error = namei(&nd)) != 0)
3776 		goto fdout;
3777 	vp = nd.ni_vp;
3778 	if (vp->v_type != VDIR) {
3779 		error = ENOTDIR;
3780 		goto out;
3781 	}
3782 	/*
3783 	 * No rmdir "." please.
3784 	 */
3785 	if (nd.ni_dvp == vp) {
3786 		error = EINVAL;
3787 		goto out;
3788 	}
3789 	/*
3790 	 * The root of a mounted filesystem cannot be deleted.
3791 	 */
3792 	if (vp->v_vflag & VV_ROOT) {
3793 		error = EBUSY;
3794 		goto out;
3795 	}
3796 
3797 	if (fp != NULL && fp->f_vnode != vp) {
3798 		if ((fp->f_vnode->v_iflag & VI_DOOMED) != 0)
3799 			error = EBADF;
3800 		else
3801 			error = EDEADLK;
3802 		goto out;
3803 	}
3804 
3805 #ifdef MAC
3806 	error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
3807 	    &nd.ni_cnd);
3808 	if (error != 0)
3809 		goto out;
3810 #endif
3811 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3812 		NDFREE(&nd, NDF_ONLY_PNBUF);
3813 		vput(vp);
3814 		if (nd.ni_dvp == vp)
3815 			vrele(nd.ni_dvp);
3816 		else
3817 			vput(nd.ni_dvp);
3818 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3819 			goto fdout;
3820 		goto restart;
3821 	}
3822 	vfs_notify_upper(vp, VFS_NOTIFY_UPPER_UNLINK);
3823 	error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
3824 	vn_finished_write(mp);
3825 out:
3826 	NDFREE(&nd, NDF_ONLY_PNBUF);
3827 	vput(vp);
3828 	if (nd.ni_dvp == vp)
3829 		vrele(nd.ni_dvp);
3830 	else
3831 		vput(nd.ni_dvp);
3832 fdout:
3833 	if (fp != NULL)
3834 		fdrop(fp, td);
3835 	return (error);
3836 }
3837 
3838 #if defined(COMPAT_43) || defined(COMPAT_FREEBSD11)
3839 int
3840 freebsd11_kern_getdirentries(struct thread *td, int fd, char *ubuf, u_int count,
3841     long *basep, void (*func)(struct freebsd11_dirent *))
3842 {
3843 	struct freebsd11_dirent dstdp;
3844 	struct dirent *dp, *edp;
3845 	char *dirbuf;
3846 	off_t base;
3847 	ssize_t resid, ucount;
3848 	int error;
3849 
3850 	/* XXX arbitrary sanity limit on `count'. */
3851 	count = min(count, 64 * 1024);
3852 
3853 	dirbuf = malloc(count, M_TEMP, M_WAITOK);
3854 
3855 	error = kern_getdirentries(td, fd, dirbuf, count, &base, &resid,
3856 	    UIO_SYSSPACE);
3857 	if (error != 0)
3858 		goto done;
3859 	if (basep != NULL)
3860 		*basep = base;
3861 
3862 	ucount = 0;
3863 	for (dp = (struct dirent *)dirbuf,
3864 	    edp = (struct dirent *)&dirbuf[count - resid];
3865 	    ucount < count && dp < edp; ) {
3866 		if (dp->d_reclen == 0)
3867 			break;
3868 		MPASS(dp->d_reclen >= _GENERIC_DIRLEN(0));
3869 		if (dp->d_namlen >= sizeof(dstdp.d_name))
3870 			continue;
3871 		dstdp.d_type = dp->d_type;
3872 		dstdp.d_namlen = dp->d_namlen;
3873 		dstdp.d_fileno = dp->d_fileno;		/* truncate */
3874 		if (dstdp.d_fileno != dp->d_fileno) {
3875 			switch (ino64_trunc_error) {
3876 			default:
3877 			case 0:
3878 				break;
3879 			case 1:
3880 				error = EOVERFLOW;
3881 				goto done;
3882 			case 2:
3883 				dstdp.d_fileno = UINT32_MAX;
3884 				break;
3885 			}
3886 		}
3887 		dstdp.d_reclen = sizeof(dstdp) - sizeof(dstdp.d_name) +
3888 		    ((dp->d_namlen + 1 + 3) &~ 3);
3889 		bcopy(dp->d_name, dstdp.d_name, dstdp.d_namlen);
3890 		bzero(dstdp.d_name + dstdp.d_namlen,
3891 		    dstdp.d_reclen - offsetof(struct freebsd11_dirent, d_name) -
3892 		    dstdp.d_namlen);
3893 		MPASS(dstdp.d_reclen <= dp->d_reclen);
3894 		MPASS(ucount + dstdp.d_reclen <= count);
3895 		if (func != NULL)
3896 			func(&dstdp);
3897 		error = copyout(&dstdp, ubuf + ucount, dstdp.d_reclen);
3898 		if (error != 0)
3899 			break;
3900 		dp = (struct dirent *)((char *)dp + dp->d_reclen);
3901 		ucount += dstdp.d_reclen;
3902 	}
3903 
3904 done:
3905 	free(dirbuf, M_TEMP);
3906 	if (error == 0)
3907 		td->td_retval[0] = ucount;
3908 	return (error);
3909 }
3910 #endif /* COMPAT */
3911 
3912 #ifdef COMPAT_43
3913 static void
3914 ogetdirentries_cvt(struct freebsd11_dirent *dp)
3915 {
3916 #if (BYTE_ORDER == LITTLE_ENDIAN)
3917 	/*
3918 	 * The expected low byte of dp->d_namlen is our dp->d_type.
3919 	 * The high MBZ byte of dp->d_namlen is our dp->d_namlen.
3920 	 */
3921 	dp->d_type = dp->d_namlen;
3922 	dp->d_namlen = 0;
3923 #else
3924 	/*
3925 	 * The dp->d_type is the high byte of the expected dp->d_namlen,
3926 	 * so must be zero'ed.
3927 	 */
3928 	dp->d_type = 0;
3929 #endif
3930 }
3931 
3932 /*
3933  * Read a block of directory entries in a filesystem independent format.
3934  */
3935 #ifndef _SYS_SYSPROTO_H_
3936 struct ogetdirentries_args {
3937 	int	fd;
3938 	char	*buf;
3939 	u_int	count;
3940 	long	*basep;
3941 };
3942 #endif
3943 int
3944 ogetdirentries(struct thread *td, struct ogetdirentries_args *uap)
3945 {
3946 	long loff;
3947 	int error;
3948 
3949 	error = kern_ogetdirentries(td, uap, &loff);
3950 	if (error == 0)
3951 		error = copyout(&loff, uap->basep, sizeof(long));
3952 	return (error);
3953 }
3954 
3955 int
3956 kern_ogetdirentries(struct thread *td, struct ogetdirentries_args *uap,
3957     long *ploff)
3958 {
3959 	long base;
3960 	int error;
3961 
3962 	/* XXX arbitrary sanity limit on `count'. */
3963 	if (uap->count > 64 * 1024)
3964 		return (EINVAL);
3965 
3966 	error = freebsd11_kern_getdirentries(td, uap->fd, uap->buf, uap->count,
3967 	    &base, ogetdirentries_cvt);
3968 
3969 	if (error == 0 && uap->basep != NULL)
3970 		error = copyout(&base, uap->basep, sizeof(long));
3971 
3972 	return (error);
3973 }
3974 #endif /* COMPAT_43 */
3975 
3976 #if defined(COMPAT_FREEBSD11)
3977 #ifndef _SYS_SYSPROTO_H_
3978 struct freebsd11_getdirentries_args {
3979 	int	fd;
3980 	char	*buf;
3981 	u_int	count;
3982 	long	*basep;
3983 };
3984 #endif
3985 int
3986 freebsd11_getdirentries(struct thread *td,
3987     struct freebsd11_getdirentries_args *uap)
3988 {
3989 	long base;
3990 	int error;
3991 
3992 	error = freebsd11_kern_getdirentries(td, uap->fd, uap->buf, uap->count,
3993 	    &base, NULL);
3994 
3995 	if (error == 0 && uap->basep != NULL)
3996 		error = copyout(&base, uap->basep, sizeof(long));
3997 	return (error);
3998 }
3999 
4000 int
4001 freebsd11_getdents(struct thread *td, struct freebsd11_getdents_args *uap)
4002 {
4003 	struct freebsd11_getdirentries_args ap;
4004 
4005 	ap.fd = uap->fd;
4006 	ap.buf = uap->buf;
4007 	ap.count = uap->count;
4008 	ap.basep = NULL;
4009 	return (freebsd11_getdirentries(td, &ap));
4010 }
4011 #endif /* COMPAT_FREEBSD11 */
4012 
4013 /*
4014  * Read a block of directory entries in a filesystem independent format.
4015  */
4016 int
4017 sys_getdirentries(struct thread *td, struct getdirentries_args *uap)
4018 {
4019 	off_t base;
4020 	int error;
4021 
4022 	error = kern_getdirentries(td, uap->fd, uap->buf, uap->count, &base,
4023 	    NULL, UIO_USERSPACE);
4024 	if (error != 0)
4025 		return (error);
4026 	if (uap->basep != NULL)
4027 		error = copyout(&base, uap->basep, sizeof(off_t));
4028 	return (error);
4029 }
4030 
4031 int
4032 kern_getdirentries(struct thread *td, int fd, char *buf, size_t count,
4033     off_t *basep, ssize_t *residp, enum uio_seg bufseg)
4034 {
4035 	struct vnode *vp;
4036 	struct file *fp;
4037 	struct uio auio;
4038 	struct iovec aiov;
4039 	off_t loff;
4040 	int error, eofflag;
4041 	off_t foffset;
4042 
4043 	AUDIT_ARG_FD(fd);
4044 	if (count > IOSIZE_MAX)
4045 		return (EINVAL);
4046 	auio.uio_resid = count;
4047 	error = getvnode(td, fd, &cap_read_rights, &fp);
4048 	if (error != 0)
4049 		return (error);
4050 	if ((fp->f_flag & FREAD) == 0) {
4051 		fdrop(fp, td);
4052 		return (EBADF);
4053 	}
4054 	vp = fp->f_vnode;
4055 	foffset = foffset_lock(fp, 0);
4056 unionread:
4057 	if (vp->v_type != VDIR) {
4058 		error = EINVAL;
4059 		goto fail;
4060 	}
4061 	aiov.iov_base = buf;
4062 	aiov.iov_len = count;
4063 	auio.uio_iov = &aiov;
4064 	auio.uio_iovcnt = 1;
4065 	auio.uio_rw = UIO_READ;
4066 	auio.uio_segflg = bufseg;
4067 	auio.uio_td = td;
4068 	vn_lock(vp, LK_SHARED | LK_RETRY);
4069 	AUDIT_ARG_VNODE1(vp);
4070 	loff = auio.uio_offset = foffset;
4071 #ifdef MAC
4072 	error = mac_vnode_check_readdir(td->td_ucred, vp);
4073 	if (error == 0)
4074 #endif
4075 		error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, NULL,
4076 		    NULL);
4077 	foffset = auio.uio_offset;
4078 	if (error != 0) {
4079 		VOP_UNLOCK(vp, 0);
4080 		goto fail;
4081 	}
4082 	if (count == auio.uio_resid &&
4083 	    (vp->v_vflag & VV_ROOT) &&
4084 	    (vp->v_mount->mnt_flag & MNT_UNION)) {
4085 		struct vnode *tvp = vp;
4086 
4087 		vp = vp->v_mount->mnt_vnodecovered;
4088 		VREF(vp);
4089 		fp->f_vnode = vp;
4090 		fp->f_data = vp;
4091 		foffset = 0;
4092 		vput(tvp);
4093 		goto unionread;
4094 	}
4095 	VOP_UNLOCK(vp, 0);
4096 	*basep = loff;
4097 	if (residp != NULL)
4098 		*residp = auio.uio_resid;
4099 	td->td_retval[0] = count - auio.uio_resid;
4100 fail:
4101 	foffset_unlock(fp, foffset, 0);
4102 	fdrop(fp, td);
4103 	return (error);
4104 }
4105 
4106 /*
4107  * Set the mode mask for creation of filesystem nodes.
4108  */
4109 #ifndef _SYS_SYSPROTO_H_
4110 struct umask_args {
4111 	int	newmask;
4112 };
4113 #endif
4114 int
4115 sys_umask(struct thread *td, struct umask_args *uap)
4116 {
4117 	struct filedesc *fdp;
4118 
4119 	fdp = td->td_proc->p_fd;
4120 	FILEDESC_XLOCK(fdp);
4121 	td->td_retval[0] = fdp->fd_cmask;
4122 	fdp->fd_cmask = uap->newmask & ALLPERMS;
4123 	FILEDESC_XUNLOCK(fdp);
4124 	return (0);
4125 }
4126 
4127 /*
4128  * Void all references to file by ripping underlying filesystem away from
4129  * vnode.
4130  */
4131 #ifndef _SYS_SYSPROTO_H_
4132 struct revoke_args {
4133 	char	*path;
4134 };
4135 #endif
4136 int
4137 sys_revoke(struct thread *td, struct revoke_args *uap)
4138 {
4139 	struct vnode *vp;
4140 	struct vattr vattr;
4141 	struct nameidata nd;
4142 	int error;
4143 
4144 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
4145 	    uap->path, td);
4146 	if ((error = namei(&nd)) != 0)
4147 		return (error);
4148 	vp = nd.ni_vp;
4149 	NDFREE(&nd, NDF_ONLY_PNBUF);
4150 	if (vp->v_type != VCHR || vp->v_rdev == NULL) {
4151 		error = EINVAL;
4152 		goto out;
4153 	}
4154 #ifdef MAC
4155 	error = mac_vnode_check_revoke(td->td_ucred, vp);
4156 	if (error != 0)
4157 		goto out;
4158 #endif
4159 	error = VOP_GETATTR(vp, &vattr, td->td_ucred);
4160 	if (error != 0)
4161 		goto out;
4162 	if (td->td_ucred->cr_uid != vattr.va_uid) {
4163 		error = priv_check(td, PRIV_VFS_ADMIN);
4164 		if (error != 0)
4165 			goto out;
4166 	}
4167 	if (vcount(vp) > 1)
4168 		VOP_REVOKE(vp, REVOKEALL);
4169 out:
4170 	vput(vp);
4171 	return (error);
4172 }
4173 
4174 /*
4175  * Convert a user file descriptor to a kernel file entry and check that, if it
4176  * is a capability, the correct rights are present. A reference on the file
4177  * entry is held upon returning.
4178  */
4179 int
4180 getvnode(struct thread *td, int fd, cap_rights_t *rightsp, struct file **fpp)
4181 {
4182 	struct file *fp;
4183 	int error;
4184 
4185 	error = fget_unlocked(td->td_proc->p_fd, fd, rightsp, &fp, NULL);
4186 	if (error != 0)
4187 		return (error);
4188 
4189 	/*
4190 	 * The file could be not of the vnode type, or it may be not
4191 	 * yet fully initialized, in which case the f_vnode pointer
4192 	 * may be set, but f_ops is still badfileops.  E.g.,
4193 	 * devfs_open() transiently create such situation to
4194 	 * facilitate csw d_fdopen().
4195 	 *
4196 	 * Dupfdopen() handling in kern_openat() installs the
4197 	 * half-baked file into the process descriptor table, allowing
4198 	 * other thread to dereference it. Guard against the race by
4199 	 * checking f_ops.
4200 	 */
4201 	if (fp->f_vnode == NULL || fp->f_ops == &badfileops) {
4202 		fdrop(fp, td);
4203 		return (EINVAL);
4204 	}
4205 	*fpp = fp;
4206 	return (0);
4207 }
4208 
4209 
4210 /*
4211  * Get an (NFS) file handle.
4212  */
4213 #ifndef _SYS_SYSPROTO_H_
4214 struct lgetfh_args {
4215 	char *fname;
4216 	fhandle_t *fhp;
4217 };
4218 #endif
4219 int
4220 sys_lgetfh(struct thread *td, struct lgetfh_args *uap)
4221 {
4222 
4223 	return (kern_getfhat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->fname,
4224 	    UIO_USERSPACE, uap->fhp));
4225 }
4226 
4227 #ifndef _SYS_SYSPROTO_H_
4228 struct getfh_args {
4229 	char *fname;
4230 	fhandle_t *fhp;
4231 };
4232 #endif
4233 int
4234 sys_getfh(struct thread *td, struct getfh_args *uap)
4235 {
4236 
4237 	return (kern_getfhat(td, 0, AT_FDCWD, uap->fname, UIO_USERSPACE,
4238 	    uap->fhp));
4239 }
4240 
4241 /*
4242  * syscall for the rpc.lockd to use to translate an open descriptor into
4243  * a NFS file handle.
4244  *
4245  * warning: do not remove the priv_check() call or this becomes one giant
4246  * security hole.
4247  */
4248 #ifndef _SYS_SYSPROTO_H_
4249 struct getfhat_args {
4250 	int fd;
4251 	char *path;
4252 	fhandle_t *fhp;
4253 	int flags;
4254 };
4255 #endif
4256 int
4257 sys_getfhat(struct thread *td, struct getfhat_args *uap)
4258 {
4259 
4260 	if ((uap->flags & ~(AT_SYMLINK_NOFOLLOW | AT_BENEATH)) != 0)
4261 		return (EINVAL);
4262 	return (kern_getfhat(td, uap->flags, uap->fd, uap->path, UIO_USERSPACE,
4263 	    uap->fhp));
4264 }
4265 
4266 static int
4267 kern_getfhat(struct thread *td, int flags, int fd, const char *path,
4268     enum uio_seg pathseg, fhandle_t *fhp)
4269 {
4270 	struct nameidata nd;
4271 	fhandle_t fh;
4272 	struct vnode *vp;
4273 	int error;
4274 
4275 	error = priv_check(td, PRIV_VFS_GETFH);
4276 	if (error != 0)
4277 		return (error);
4278 	NDINIT_AT(&nd, LOOKUP, ((flags & AT_SYMLINK_NOFOLLOW) != 0 ? NOFOLLOW :
4279 	    FOLLOW) | ((flags & AT_BENEATH) != 0 ? BENEATH : 0) | LOCKLEAF |
4280 	    AUDITVNODE1, pathseg, path, fd, td);
4281 	error = namei(&nd);
4282 	if (error != 0)
4283 		return (error);
4284 	NDFREE(&nd, NDF_ONLY_PNBUF);
4285 	vp = nd.ni_vp;
4286 	bzero(&fh, sizeof(fh));
4287 	fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
4288 	error = VOP_VPTOFH(vp, &fh.fh_fid);
4289 	vput(vp);
4290 	if (error == 0)
4291 		error = copyout(&fh, fhp, sizeof (fh));
4292 	return (error);
4293 }
4294 
4295 #ifndef _SYS_SYSPROTO_H_
4296 struct fhlink_args {
4297 	fhandle_t *fhp;
4298 	const char *to;
4299 };
4300 #endif
4301 int
4302 sys_fhlink(struct thread *td, struct fhlink_args *uap)
4303 {
4304 
4305 	return (kern_fhlinkat(td, AT_FDCWD, uap->to, UIO_USERSPACE, uap->fhp));
4306 }
4307 
4308 #ifndef _SYS_SYSPROTO_H_
4309 struct fhlinkat_args {
4310 	fhandle_t *fhp;
4311 	int tofd;
4312 	const char *to;
4313 };
4314 #endif
4315 int
4316 sys_fhlinkat(struct thread *td, struct fhlinkat_args *uap)
4317 {
4318 
4319 	return (kern_fhlinkat(td, uap->tofd, uap->to, UIO_USERSPACE, uap->fhp));
4320 }
4321 
4322 static int
4323 kern_fhlinkat(struct thread *td, int fd, const char *path,
4324     enum uio_seg pathseg, fhandle_t *fhp)
4325 {
4326 	fhandle_t fh;
4327 	struct mount *mp;
4328 	struct vnode *vp;
4329 	int error;
4330 
4331 	error = priv_check(td, PRIV_VFS_GETFH);
4332 	if (error != 0)
4333 		return (error);
4334 	error = copyin(fhp, &fh, sizeof(fh));
4335 	if (error != 0)
4336 		return (error);
4337 	do {
4338 		bwillwrite();
4339 		if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4340 			return (ESTALE);
4341 		error = VFS_FHTOVP(mp, &fh.fh_fid, LK_SHARED, &vp);
4342 		vfs_unbusy(mp);
4343 		if (error != 0)
4344 			return (error);
4345 		VOP_UNLOCK(vp, 0);
4346 	} while ((error = kern_linkat_vp(td, vp, fd, path, pathseg)) == EAGAIN);
4347 	return (error);
4348 }
4349 
4350 #ifndef _SYS_SYSPROTO_H_
4351 struct fhreadlink_args {
4352 	fhandle_t *fhp;
4353 	char *buf;
4354 	size_t bufsize;
4355 };
4356 #endif
4357 int
4358 sys_fhreadlink(struct thread *td, struct fhreadlink_args *uap)
4359 {
4360 	fhandle_t fh;
4361 	struct mount *mp;
4362 	struct vnode *vp;
4363 	int error;
4364 
4365 	error = priv_check(td, PRIV_VFS_GETFH);
4366 	if (error != 0)
4367 		return (error);
4368 	if (uap->bufsize > IOSIZE_MAX)
4369 		return (EINVAL);
4370 	error = copyin(uap->fhp, &fh, sizeof(fh));
4371 	if (error != 0)
4372 		return (error);
4373 	if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4374 		return (ESTALE);
4375 	error = VFS_FHTOVP(mp, &fh.fh_fid, LK_SHARED, &vp);
4376 	vfs_unbusy(mp);
4377 	if (error != 0)
4378 		return (error);
4379 	error = kern_readlink_vp(vp, uap->buf, UIO_USERSPACE, uap->bufsize, td);
4380 	vput(vp);
4381 	return (error);
4382 }
4383 
4384 /*
4385  * syscall for the rpc.lockd to use to translate a NFS file handle into an
4386  * open descriptor.
4387  *
4388  * warning: do not remove the priv_check() call or this becomes one giant
4389  * security hole.
4390  */
4391 #ifndef _SYS_SYSPROTO_H_
4392 struct fhopen_args {
4393 	const struct fhandle *u_fhp;
4394 	int flags;
4395 };
4396 #endif
4397 int
4398 sys_fhopen(struct thread *td, struct fhopen_args *uap)
4399 {
4400 	struct mount *mp;
4401 	struct vnode *vp;
4402 	struct fhandle fhp;
4403 	struct file *fp;
4404 	int fmode, error;
4405 	int indx;
4406 
4407 	error = priv_check(td, PRIV_VFS_FHOPEN);
4408 	if (error != 0)
4409 		return (error);
4410 	indx = -1;
4411 	fmode = FFLAGS(uap->flags);
4412 	/* why not allow a non-read/write open for our lockd? */
4413 	if (((fmode & (FREAD | FWRITE)) == 0) || (fmode & O_CREAT))
4414 		return (EINVAL);
4415 	error = copyin(uap->u_fhp, &fhp, sizeof(fhp));
4416 	if (error != 0)
4417 		return(error);
4418 	/* find the mount point */
4419 	mp = vfs_busyfs(&fhp.fh_fsid);
4420 	if (mp == NULL)
4421 		return (ESTALE);
4422 	/* now give me my vnode, it gets returned to me locked */
4423 	error = VFS_FHTOVP(mp, &fhp.fh_fid, LK_EXCLUSIVE, &vp);
4424 	vfs_unbusy(mp);
4425 	if (error != 0)
4426 		return (error);
4427 
4428 	error = falloc_noinstall(td, &fp);
4429 	if (error != 0) {
4430 		vput(vp);
4431 		return (error);
4432 	}
4433 	/*
4434 	 * An extra reference on `fp' has been held for us by
4435 	 * falloc_noinstall().
4436 	 */
4437 
4438 #ifdef INVARIANTS
4439 	td->td_dupfd = -1;
4440 #endif
4441 	error = vn_open_vnode(vp, fmode, td->td_ucred, td, fp);
4442 	if (error != 0) {
4443 		KASSERT(fp->f_ops == &badfileops,
4444 		    ("VOP_OPEN in fhopen() set f_ops"));
4445 		KASSERT(td->td_dupfd < 0,
4446 		    ("fhopen() encountered fdopen()"));
4447 
4448 		vput(vp);
4449 		goto bad;
4450 	}
4451 #ifdef INVARIANTS
4452 	td->td_dupfd = 0;
4453 #endif
4454 	fp->f_vnode = vp;
4455 	fp->f_seqcount = 1;
4456 	finit(fp, (fmode & FMASK) | (fp->f_flag & FHASLOCK), DTYPE_VNODE, vp,
4457 	    &vnops);
4458 	VOP_UNLOCK(vp, 0);
4459 	if ((fmode & O_TRUNC) != 0) {
4460 		error = fo_truncate(fp, 0, td->td_ucred, td);
4461 		if (error != 0)
4462 			goto bad;
4463 	}
4464 
4465 	error = finstall(td, fp, &indx, fmode, NULL);
4466 bad:
4467 	fdrop(fp, td);
4468 	td->td_retval[0] = indx;
4469 	return (error);
4470 }
4471 
4472 /*
4473  * Stat an (NFS) file handle.
4474  */
4475 #ifndef _SYS_SYSPROTO_H_
4476 struct fhstat_args {
4477 	struct fhandle *u_fhp;
4478 	struct stat *sb;
4479 };
4480 #endif
4481 int
4482 sys_fhstat(struct thread *td, struct fhstat_args *uap)
4483 {
4484 	struct stat sb;
4485 	struct fhandle fh;
4486 	int error;
4487 
4488 	error = copyin(uap->u_fhp, &fh, sizeof(fh));
4489 	if (error != 0)
4490 		return (error);
4491 	error = kern_fhstat(td, fh, &sb);
4492 	if (error == 0)
4493 		error = copyout(&sb, uap->sb, sizeof(sb));
4494 	return (error);
4495 }
4496 
4497 int
4498 kern_fhstat(struct thread *td, struct fhandle fh, struct stat *sb)
4499 {
4500 	struct mount *mp;
4501 	struct vnode *vp;
4502 	int error;
4503 
4504 	error = priv_check(td, PRIV_VFS_FHSTAT);
4505 	if (error != 0)
4506 		return (error);
4507 	if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4508 		return (ESTALE);
4509 	error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4510 	vfs_unbusy(mp);
4511 	if (error != 0)
4512 		return (error);
4513 	error = vn_stat(vp, sb, td->td_ucred, NOCRED, td);
4514 	vput(vp);
4515 	return (error);
4516 }
4517 
4518 /*
4519  * Implement fstatfs() for (NFS) file handles.
4520  */
4521 #ifndef _SYS_SYSPROTO_H_
4522 struct fhstatfs_args {
4523 	struct fhandle *u_fhp;
4524 	struct statfs *buf;
4525 };
4526 #endif
4527 int
4528 sys_fhstatfs(struct thread *td, struct fhstatfs_args *uap)
4529 {
4530 	struct statfs *sfp;
4531 	fhandle_t fh;
4532 	int error;
4533 
4534 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
4535 	if (error != 0)
4536 		return (error);
4537 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
4538 	error = kern_fhstatfs(td, fh, sfp);
4539 	if (error == 0)
4540 		error = copyout(sfp, uap->buf, sizeof(*sfp));
4541 	free(sfp, M_STATFS);
4542 	return (error);
4543 }
4544 
4545 int
4546 kern_fhstatfs(struct thread *td, fhandle_t fh, struct statfs *buf)
4547 {
4548 	struct statfs *sp;
4549 	struct mount *mp;
4550 	struct vnode *vp;
4551 	int error;
4552 
4553 	error = priv_check(td, PRIV_VFS_FHSTATFS);
4554 	if (error != 0)
4555 		return (error);
4556 	if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4557 		return (ESTALE);
4558 	error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4559 	if (error != 0) {
4560 		vfs_unbusy(mp);
4561 		return (error);
4562 	}
4563 	vput(vp);
4564 	error = prison_canseemount(td->td_ucred, mp);
4565 	if (error != 0)
4566 		goto out;
4567 #ifdef MAC
4568 	error = mac_mount_check_stat(td->td_ucred, mp);
4569 	if (error != 0)
4570 		goto out;
4571 #endif
4572 	/*
4573 	 * Set these in case the underlying filesystem fails to do so.
4574 	 */
4575 	sp = &mp->mnt_stat;
4576 	sp->f_version = STATFS_VERSION;
4577 	sp->f_namemax = NAME_MAX;
4578 	sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
4579 	error = VFS_STATFS(mp, sp);
4580 	if (error == 0)
4581 		*buf = *sp;
4582 out:
4583 	vfs_unbusy(mp);
4584 	return (error);
4585 }
4586 
4587 int
4588 kern_posix_fallocate(struct thread *td, int fd, off_t offset, off_t len)
4589 {
4590 	struct file *fp;
4591 	struct mount *mp;
4592 	struct vnode *vp;
4593 	off_t olen, ooffset;
4594 	int error;
4595 #ifdef AUDIT
4596 	int audited_vnode1 = 0;
4597 #endif
4598 
4599 	AUDIT_ARG_FD(fd);
4600 	if (offset < 0 || len <= 0)
4601 		return (EINVAL);
4602 	/* Check for wrap. */
4603 	if (offset > OFF_MAX - len)
4604 		return (EFBIG);
4605 	AUDIT_ARG_FD(fd);
4606 	error = fget(td, fd, &cap_pwrite_rights, &fp);
4607 	if (error != 0)
4608 		return (error);
4609 	AUDIT_ARG_FILE(td->td_proc, fp);
4610 	if ((fp->f_ops->fo_flags & DFLAG_SEEKABLE) == 0) {
4611 		error = ESPIPE;
4612 		goto out;
4613 	}
4614 	if ((fp->f_flag & FWRITE) == 0) {
4615 		error = EBADF;
4616 		goto out;
4617 	}
4618 	if (fp->f_type != DTYPE_VNODE) {
4619 		error = ENODEV;
4620 		goto out;
4621 	}
4622 	vp = fp->f_vnode;
4623 	if (vp->v_type != VREG) {
4624 		error = ENODEV;
4625 		goto out;
4626 	}
4627 
4628 	/* Allocating blocks may take a long time, so iterate. */
4629 	for (;;) {
4630 		olen = len;
4631 		ooffset = offset;
4632 
4633 		bwillwrite();
4634 		mp = NULL;
4635 		error = vn_start_write(vp, &mp, V_WAIT | PCATCH);
4636 		if (error != 0)
4637 			break;
4638 		error = vn_lock(vp, LK_EXCLUSIVE);
4639 		if (error != 0) {
4640 			vn_finished_write(mp);
4641 			break;
4642 		}
4643 #ifdef AUDIT
4644 		if (!audited_vnode1) {
4645 			AUDIT_ARG_VNODE1(vp);
4646 			audited_vnode1 = 1;
4647 		}
4648 #endif
4649 #ifdef MAC
4650 		error = mac_vnode_check_write(td->td_ucred, fp->f_cred, vp);
4651 		if (error == 0)
4652 #endif
4653 			error = VOP_ALLOCATE(vp, &offset, &len);
4654 		VOP_UNLOCK(vp, 0);
4655 		vn_finished_write(mp);
4656 
4657 		if (olen + ooffset != offset + len) {
4658 			panic("offset + len changed from %jx/%jx to %jx/%jx",
4659 			    ooffset, olen, offset, len);
4660 		}
4661 		if (error != 0 || len == 0)
4662 			break;
4663 		KASSERT(olen > len, ("Iteration did not make progress?"));
4664 		maybe_yield();
4665 	}
4666  out:
4667 	fdrop(fp, td);
4668 	return (error);
4669 }
4670 
4671 int
4672 sys_posix_fallocate(struct thread *td, struct posix_fallocate_args *uap)
4673 {
4674 	int error;
4675 
4676 	error = kern_posix_fallocate(td, uap->fd, uap->offset, uap->len);
4677 	return (kern_posix_error(td, error));
4678 }
4679 
4680 /*
4681  * Unlike madvise(2), we do not make a best effort to remember every
4682  * possible caching hint.  Instead, we remember the last setting with
4683  * the exception that we will allow POSIX_FADV_NORMAL to adjust the
4684  * region of any current setting.
4685  */
4686 int
4687 kern_posix_fadvise(struct thread *td, int fd, off_t offset, off_t len,
4688     int advice)
4689 {
4690 	struct fadvise_info *fa, *new;
4691 	struct file *fp;
4692 	struct vnode *vp;
4693 	off_t end;
4694 	int error;
4695 
4696 	if (offset < 0 || len < 0 || offset > OFF_MAX - len)
4697 		return (EINVAL);
4698 	AUDIT_ARG_VALUE(advice);
4699 	switch (advice) {
4700 	case POSIX_FADV_SEQUENTIAL:
4701 	case POSIX_FADV_RANDOM:
4702 	case POSIX_FADV_NOREUSE:
4703 		new = malloc(sizeof(*fa), M_FADVISE, M_WAITOK);
4704 		break;
4705 	case POSIX_FADV_NORMAL:
4706 	case POSIX_FADV_WILLNEED:
4707 	case POSIX_FADV_DONTNEED:
4708 		new = NULL;
4709 		break;
4710 	default:
4711 		return (EINVAL);
4712 	}
4713 	/* XXX: CAP_POSIX_FADVISE? */
4714 	AUDIT_ARG_FD(fd);
4715 	error = fget(td, fd, &cap_no_rights, &fp);
4716 	if (error != 0)
4717 		goto out;
4718 	AUDIT_ARG_FILE(td->td_proc, fp);
4719 	if ((fp->f_ops->fo_flags & DFLAG_SEEKABLE) == 0) {
4720 		error = ESPIPE;
4721 		goto out;
4722 	}
4723 	if (fp->f_type != DTYPE_VNODE) {
4724 		error = ENODEV;
4725 		goto out;
4726 	}
4727 	vp = fp->f_vnode;
4728 	if (vp->v_type != VREG) {
4729 		error = ENODEV;
4730 		goto out;
4731 	}
4732 	if (len == 0)
4733 		end = OFF_MAX;
4734 	else
4735 		end = offset + len - 1;
4736 	switch (advice) {
4737 	case POSIX_FADV_SEQUENTIAL:
4738 	case POSIX_FADV_RANDOM:
4739 	case POSIX_FADV_NOREUSE:
4740 		/*
4741 		 * Try to merge any existing non-standard region with
4742 		 * this new region if possible, otherwise create a new
4743 		 * non-standard region for this request.
4744 		 */
4745 		mtx_pool_lock(mtxpool_sleep, fp);
4746 		fa = fp->f_advice;
4747 		if (fa != NULL && fa->fa_advice == advice &&
4748 		    ((fa->fa_start <= end && fa->fa_end >= offset) ||
4749 		    (end != OFF_MAX && fa->fa_start == end + 1) ||
4750 		    (fa->fa_end != OFF_MAX && fa->fa_end + 1 == offset))) {
4751 			if (offset < fa->fa_start)
4752 				fa->fa_start = offset;
4753 			if (end > fa->fa_end)
4754 				fa->fa_end = end;
4755 		} else {
4756 			new->fa_advice = advice;
4757 			new->fa_start = offset;
4758 			new->fa_end = end;
4759 			fp->f_advice = new;
4760 			new = fa;
4761 		}
4762 		mtx_pool_unlock(mtxpool_sleep, fp);
4763 		break;
4764 	case POSIX_FADV_NORMAL:
4765 		/*
4766 		 * If a the "normal" region overlaps with an existing
4767 		 * non-standard region, trim or remove the
4768 		 * non-standard region.
4769 		 */
4770 		mtx_pool_lock(mtxpool_sleep, fp);
4771 		fa = fp->f_advice;
4772 		if (fa != NULL) {
4773 			if (offset <= fa->fa_start && end >= fa->fa_end) {
4774 				new = fa;
4775 				fp->f_advice = NULL;
4776 			} else if (offset <= fa->fa_start &&
4777 			    end >= fa->fa_start)
4778 				fa->fa_start = end + 1;
4779 			else if (offset <= fa->fa_end && end >= fa->fa_end)
4780 				fa->fa_end = offset - 1;
4781 			else if (offset >= fa->fa_start && end <= fa->fa_end) {
4782 				/*
4783 				 * If the "normal" region is a middle
4784 				 * portion of the existing
4785 				 * non-standard region, just remove
4786 				 * the whole thing rather than picking
4787 				 * one side or the other to
4788 				 * preserve.
4789 				 */
4790 				new = fa;
4791 				fp->f_advice = NULL;
4792 			}
4793 		}
4794 		mtx_pool_unlock(mtxpool_sleep, fp);
4795 		break;
4796 	case POSIX_FADV_WILLNEED:
4797 	case POSIX_FADV_DONTNEED:
4798 		error = VOP_ADVISE(vp, offset, end, advice);
4799 		break;
4800 	}
4801 out:
4802 	if (fp != NULL)
4803 		fdrop(fp, td);
4804 	free(new, M_FADVISE);
4805 	return (error);
4806 }
4807 
4808 int
4809 sys_posix_fadvise(struct thread *td, struct posix_fadvise_args *uap)
4810 {
4811 	int error;
4812 
4813 	error = kern_posix_fadvise(td, uap->fd, uap->offset, uap->len,
4814 	    uap->advice);
4815 	return (kern_posix_error(td, error));
4816 }
4817