xref: /freebsd/sys/kern/vfs_syscalls.c (revision a665699823a2099a451e1c9c1505ef1cbf29c914)
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 
109 /*
110  * Sync each mounted filesystem.
111  */
112 #ifndef _SYS_SYSPROTO_H_
113 struct sync_args {
114 	int     dummy;
115 };
116 #endif
117 /* ARGSUSED */
118 int
119 sys_sync(struct thread *td, struct sync_args *uap)
120 {
121 	struct mount *mp, *nmp;
122 	int save;
123 
124 	mtx_lock(&mountlist_mtx);
125 	for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
126 		if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK)) {
127 			nmp = TAILQ_NEXT(mp, mnt_list);
128 			continue;
129 		}
130 		if ((mp->mnt_flag & MNT_RDONLY) == 0 &&
131 		    vn_start_write(NULL, &mp, V_NOWAIT) == 0) {
132 			save = curthread_pflags_set(TDP_SYNCIO);
133 			vfs_msync(mp, MNT_NOWAIT);
134 			VFS_SYNC(mp, MNT_NOWAIT);
135 			curthread_pflags_restore(save);
136 			vn_finished_write(mp);
137 		}
138 		mtx_lock(&mountlist_mtx);
139 		nmp = TAILQ_NEXT(mp, mnt_list);
140 		vfs_unbusy(mp);
141 	}
142 	mtx_unlock(&mountlist_mtx);
143 	return (0);
144 }
145 
146 /*
147  * Change filesystem quotas.
148  */
149 #ifndef _SYS_SYSPROTO_H_
150 struct quotactl_args {
151 	char *path;
152 	int cmd;
153 	int uid;
154 	caddr_t arg;
155 };
156 #endif
157 int
158 sys_quotactl(struct thread *td, struct quotactl_args *uap)
159 {
160 	struct mount *mp;
161 	struct nameidata nd;
162 	int error;
163 
164 	AUDIT_ARG_CMD(uap->cmd);
165 	AUDIT_ARG_UID(uap->uid);
166 	if (!prison_allow(td->td_ucred, PR_ALLOW_QUOTAS))
167 		return (EPERM);
168 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
169 	    uap->path, td);
170 	if ((error = namei(&nd)) != 0)
171 		return (error);
172 	NDFREE(&nd, NDF_ONLY_PNBUF);
173 	mp = nd.ni_vp->v_mount;
174 	vfs_ref(mp);
175 	vput(nd.ni_vp);
176 	error = vfs_busy(mp, 0);
177 	vfs_rel(mp);
178 	if (error != 0)
179 		return (error);
180 	error = VFS_QUOTACTL(mp, uap->cmd, uap->uid, uap->arg);
181 
182 	/*
183 	 * Since quota on operation typically needs to open quota
184 	 * file, the Q_QUOTAON handler needs to unbusy the mount point
185 	 * before calling into namei.  Otherwise, unmount might be
186 	 * started between two vfs_busy() invocations (first is our,
187 	 * second is from mount point cross-walk code in lookup()),
188 	 * causing deadlock.
189 	 *
190 	 * Require that Q_QUOTAON handles the vfs_busy() reference on
191 	 * its own, always returning with ubusied mount point.
192 	 */
193 	if ((uap->cmd >> SUBCMDSHIFT) != Q_QUOTAON)
194 		vfs_unbusy(mp);
195 	return (error);
196 }
197 
198 /*
199  * Used by statfs conversion routines to scale the block size up if
200  * necessary so that all of the block counts are <= 'max_size'.  Note
201  * that 'max_size' should be a bitmask, i.e. 2^n - 1 for some non-zero
202  * value of 'n'.
203  */
204 void
205 statfs_scale_blocks(struct statfs *sf, long max_size)
206 {
207 	uint64_t count;
208 	int shift;
209 
210 	KASSERT(powerof2(max_size + 1), ("%s: invalid max_size", __func__));
211 
212 	/*
213 	 * Attempt to scale the block counts to give a more accurate
214 	 * overview to userland of the ratio of free space to used
215 	 * space.  To do this, find the largest block count and compute
216 	 * a divisor that lets it fit into a signed integer <= max_size.
217 	 */
218 	if (sf->f_bavail < 0)
219 		count = -sf->f_bavail;
220 	else
221 		count = sf->f_bavail;
222 	count = MAX(sf->f_blocks, MAX(sf->f_bfree, count));
223 	if (count <= max_size)
224 		return;
225 
226 	count >>= flsl(max_size);
227 	shift = 0;
228 	while (count > 0) {
229 		shift++;
230 		count >>=1;
231 	}
232 
233 	sf->f_bsize <<= shift;
234 	sf->f_blocks >>= shift;
235 	sf->f_bfree >>= shift;
236 	sf->f_bavail >>= shift;
237 }
238 
239 static int
240 kern_do_statfs(struct thread *td, struct mount *mp, struct statfs *buf)
241 {
242 	struct statfs *sp;
243 	int error;
244 
245 	if (mp == NULL)
246 		return (EBADF);
247 	error = vfs_busy(mp, 0);
248 	vfs_rel(mp);
249 	if (error != 0)
250 		return (error);
251 #ifdef MAC
252 	error = mac_mount_check_stat(td->td_ucred, mp);
253 	if (error != 0)
254 		goto out;
255 #endif
256 	/*
257 	 * Set these in case the underlying filesystem fails to do so.
258 	 */
259 	sp = &mp->mnt_stat;
260 	sp->f_version = STATFS_VERSION;
261 	sp->f_namemax = NAME_MAX;
262 	sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
263 	error = VFS_STATFS(mp, sp);
264 	if (error != 0)
265 		goto out;
266 	*buf = *sp;
267 	if (priv_check(td, PRIV_VFS_GENERATION)) {
268 		buf->f_fsid.val[0] = buf->f_fsid.val[1] = 0;
269 		prison_enforce_statfs(td->td_ucred, mp, buf);
270 	}
271 out:
272 	vfs_unbusy(mp);
273 	return (error);
274 }
275 
276 /*
277  * Get filesystem statistics.
278  */
279 #ifndef _SYS_SYSPROTO_H_
280 struct statfs_args {
281 	char *path;
282 	struct statfs *buf;
283 };
284 #endif
285 int
286 sys_statfs(struct thread *td, struct statfs_args *uap)
287 {
288 	struct statfs *sfp;
289 	int error;
290 
291 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
292 	error = kern_statfs(td, uap->path, UIO_USERSPACE, sfp);
293 	if (error == 0)
294 		error = copyout(sfp, uap->buf, sizeof(struct statfs));
295 	free(sfp, M_STATFS);
296 	return (error);
297 }
298 
299 int
300 kern_statfs(struct thread *td, char *path, enum uio_seg pathseg,
301     struct statfs *buf)
302 {
303 	struct mount *mp;
304 	struct nameidata nd;
305 	int error;
306 
307 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
308 	    pathseg, path, td);
309 	error = namei(&nd);
310 	if (error != 0)
311 		return (error);
312 	mp = nd.ni_vp->v_mount;
313 	vfs_ref(mp);
314 	NDFREE(&nd, NDF_ONLY_PNBUF);
315 	vput(nd.ni_vp);
316 	return (kern_do_statfs(td, mp, buf));
317 }
318 
319 /*
320  * Get filesystem statistics.
321  */
322 #ifndef _SYS_SYSPROTO_H_
323 struct fstatfs_args {
324 	int fd;
325 	struct statfs *buf;
326 };
327 #endif
328 int
329 sys_fstatfs(struct thread *td, struct fstatfs_args *uap)
330 {
331 	struct statfs *sfp;
332 	int error;
333 
334 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
335 	error = kern_fstatfs(td, uap->fd, sfp);
336 	if (error == 0)
337 		error = copyout(sfp, uap->buf, sizeof(struct statfs));
338 	free(sfp, M_STATFS);
339 	return (error);
340 }
341 
342 int
343 kern_fstatfs(struct thread *td, int fd, struct statfs *buf)
344 {
345 	struct file *fp;
346 	struct mount *mp;
347 	struct vnode *vp;
348 	int error;
349 
350 	AUDIT_ARG_FD(fd);
351 	error = getvnode(td, fd, &cap_fstatfs_rights, &fp);
352 	if (error != 0)
353 		return (error);
354 	vp = fp->f_vnode;
355 	vn_lock(vp, LK_SHARED | LK_RETRY);
356 #ifdef AUDIT
357 	AUDIT_ARG_VNODE1(vp);
358 #endif
359 	mp = vp->v_mount;
360 	if (mp != NULL)
361 		vfs_ref(mp);
362 	VOP_UNLOCK(vp, 0);
363 	fdrop(fp, td);
364 	return (kern_do_statfs(td, mp, buf));
365 }
366 
367 /*
368  * Get statistics on all filesystems.
369  */
370 #ifndef _SYS_SYSPROTO_H_
371 struct getfsstat_args {
372 	struct statfs *buf;
373 	long bufsize;
374 	int mode;
375 };
376 #endif
377 int
378 sys_getfsstat(struct thread *td, struct getfsstat_args *uap)
379 {
380 	size_t count;
381 	int error;
382 
383 	if (uap->bufsize < 0 || uap->bufsize > SIZE_MAX)
384 		return (EINVAL);
385 	error = kern_getfsstat(td, &uap->buf, uap->bufsize, &count,
386 	    UIO_USERSPACE, uap->mode);
387 	if (error == 0)
388 		td->td_retval[0] = count;
389 	return (error);
390 }
391 
392 /*
393  * If (bufsize > 0 && bufseg == UIO_SYSSPACE)
394  *	The caller is responsible for freeing memory which will be allocated
395  *	in '*buf'.
396  */
397 int
398 kern_getfsstat(struct thread *td, struct statfs **buf, size_t bufsize,
399     size_t *countp, enum uio_seg bufseg, int mode)
400 {
401 	struct mount *mp, *nmp;
402 	struct statfs *sfsp, *sp, *sptmp, *tofree;
403 	size_t count, maxcount;
404 	int error;
405 
406 	switch (mode) {
407 	case MNT_WAIT:
408 	case MNT_NOWAIT:
409 		break;
410 	default:
411 		if (bufseg == UIO_SYSSPACE)
412 			*buf = NULL;
413 		return (EINVAL);
414 	}
415 restart:
416 	maxcount = bufsize / sizeof(struct statfs);
417 	if (bufsize == 0) {
418 		sfsp = NULL;
419 		tofree = NULL;
420 	} else if (bufseg == UIO_USERSPACE) {
421 		sfsp = *buf;
422 		tofree = NULL;
423 	} else /* if (bufseg == UIO_SYSSPACE) */ {
424 		count = 0;
425 		mtx_lock(&mountlist_mtx);
426 		TAILQ_FOREACH(mp, &mountlist, mnt_list) {
427 			count++;
428 		}
429 		mtx_unlock(&mountlist_mtx);
430 		if (maxcount > count)
431 			maxcount = count;
432 		tofree = sfsp = *buf = malloc(maxcount * sizeof(struct statfs),
433 		    M_STATFS, M_WAITOK);
434 	}
435 	count = 0;
436 	mtx_lock(&mountlist_mtx);
437 	for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
438 		if (prison_canseemount(td->td_ucred, mp) != 0) {
439 			nmp = TAILQ_NEXT(mp, mnt_list);
440 			continue;
441 		}
442 #ifdef MAC
443 		if (mac_mount_check_stat(td->td_ucred, mp) != 0) {
444 			nmp = TAILQ_NEXT(mp, mnt_list);
445 			continue;
446 		}
447 #endif
448 		if (mode == MNT_WAIT) {
449 			if (vfs_busy(mp, MBF_MNTLSTLOCK) != 0) {
450 				/*
451 				 * If vfs_busy() failed, and MBF_NOWAIT
452 				 * wasn't passed, then the mp is gone.
453 				 * Furthermore, because of MBF_MNTLSTLOCK,
454 				 * the mountlist_mtx was dropped.  We have
455 				 * no other choice than to start over.
456 				 */
457 				mtx_unlock(&mountlist_mtx);
458 				free(tofree, M_STATFS);
459 				goto restart;
460 			}
461 		} else {
462 			if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK) != 0) {
463 				nmp = TAILQ_NEXT(mp, mnt_list);
464 				continue;
465 			}
466 		}
467 		if (sfsp != NULL && count < maxcount) {
468 			sp = &mp->mnt_stat;
469 			/*
470 			 * Set these in case the underlying filesystem
471 			 * fails to do so.
472 			 */
473 			sp->f_version = STATFS_VERSION;
474 			sp->f_namemax = NAME_MAX;
475 			sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
476 			/*
477 			 * If MNT_NOWAIT is specified, do not refresh
478 			 * the fsstat cache.
479 			 */
480 			if (mode != MNT_NOWAIT) {
481 				error = VFS_STATFS(mp, sp);
482 				if (error != 0) {
483 					mtx_lock(&mountlist_mtx);
484 					nmp = TAILQ_NEXT(mp, mnt_list);
485 					vfs_unbusy(mp);
486 					continue;
487 				}
488 			}
489 			if (priv_check(td, PRIV_VFS_GENERATION)) {
490 				sptmp = malloc(sizeof(struct statfs), M_STATFS,
491 				    M_WAITOK);
492 				*sptmp = *sp;
493 				sptmp->f_fsid.val[0] = sptmp->f_fsid.val[1] = 0;
494 				prison_enforce_statfs(td->td_ucred, mp, sptmp);
495 				sp = sptmp;
496 			} else
497 				sptmp = NULL;
498 			if (bufseg == UIO_SYSSPACE) {
499 				bcopy(sp, sfsp, sizeof(*sp));
500 				free(sptmp, M_STATFS);
501 			} else /* if (bufseg == UIO_USERSPACE) */ {
502 				error = copyout(sp, sfsp, sizeof(*sp));
503 				free(sptmp, M_STATFS);
504 				if (error != 0) {
505 					vfs_unbusy(mp);
506 					return (error);
507 				}
508 			}
509 			sfsp++;
510 		}
511 		count++;
512 		mtx_lock(&mountlist_mtx);
513 		nmp = TAILQ_NEXT(mp, mnt_list);
514 		vfs_unbusy(mp);
515 	}
516 	mtx_unlock(&mountlist_mtx);
517 	if (sfsp != NULL && count > maxcount)
518 		*countp = maxcount;
519 	else
520 		*countp = count;
521 	return (0);
522 }
523 
524 #ifdef COMPAT_FREEBSD4
525 /*
526  * Get old format filesystem statistics.
527  */
528 static void freebsd4_cvtstatfs(struct statfs *, struct ostatfs *);
529 
530 #ifndef _SYS_SYSPROTO_H_
531 struct freebsd4_statfs_args {
532 	char *path;
533 	struct ostatfs *buf;
534 };
535 #endif
536 int
537 freebsd4_statfs(struct thread *td, struct freebsd4_statfs_args *uap)
538 {
539 	struct ostatfs osb;
540 	struct statfs *sfp;
541 	int error;
542 
543 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
544 	error = kern_statfs(td, uap->path, UIO_USERSPACE, sfp);
545 	if (error == 0) {
546 		freebsd4_cvtstatfs(sfp, &osb);
547 		error = copyout(&osb, uap->buf, sizeof(osb));
548 	}
549 	free(sfp, M_STATFS);
550 	return (error);
551 }
552 
553 /*
554  * Get filesystem statistics.
555  */
556 #ifndef _SYS_SYSPROTO_H_
557 struct freebsd4_fstatfs_args {
558 	int fd;
559 	struct ostatfs *buf;
560 };
561 #endif
562 int
563 freebsd4_fstatfs(struct thread *td, struct freebsd4_fstatfs_args *uap)
564 {
565 	struct ostatfs osb;
566 	struct statfs *sfp;
567 	int error;
568 
569 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
570 	error = kern_fstatfs(td, uap->fd, sfp);
571 	if (error == 0) {
572 		freebsd4_cvtstatfs(sfp, &osb);
573 		error = copyout(&osb, uap->buf, sizeof(osb));
574 	}
575 	free(sfp, M_STATFS);
576 	return (error);
577 }
578 
579 /*
580  * Get statistics on all filesystems.
581  */
582 #ifndef _SYS_SYSPROTO_H_
583 struct freebsd4_getfsstat_args {
584 	struct ostatfs *buf;
585 	long bufsize;
586 	int mode;
587 };
588 #endif
589 int
590 freebsd4_getfsstat(struct thread *td, struct freebsd4_getfsstat_args *uap)
591 {
592 	struct statfs *buf, *sp;
593 	struct ostatfs osb;
594 	size_t count, size;
595 	int error;
596 
597 	if (uap->bufsize < 0)
598 		return (EINVAL);
599 	count = uap->bufsize / sizeof(struct ostatfs);
600 	if (count > SIZE_MAX / sizeof(struct statfs))
601 		return (EINVAL);
602 	size = count * sizeof(struct statfs);
603 	error = kern_getfsstat(td, &buf, size, &count, UIO_SYSSPACE,
604 	    uap->mode);
605 	if (error == 0)
606 		td->td_retval[0] = count;
607 	if (size != 0) {
608 		sp = buf;
609 		while (count != 0 && error == 0) {
610 			freebsd4_cvtstatfs(sp, &osb);
611 			error = copyout(&osb, uap->buf, sizeof(osb));
612 			sp++;
613 			uap->buf++;
614 			count--;
615 		}
616 		free(buf, M_STATFS);
617 	}
618 	return (error);
619 }
620 
621 /*
622  * Implement fstatfs() for (NFS) file handles.
623  */
624 #ifndef _SYS_SYSPROTO_H_
625 struct freebsd4_fhstatfs_args {
626 	struct fhandle *u_fhp;
627 	struct ostatfs *buf;
628 };
629 #endif
630 int
631 freebsd4_fhstatfs(struct thread *td, struct freebsd4_fhstatfs_args *uap)
632 {
633 	struct ostatfs osb;
634 	struct statfs *sfp;
635 	fhandle_t fh;
636 	int error;
637 
638 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
639 	if (error != 0)
640 		return (error);
641 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
642 	error = kern_fhstatfs(td, fh, sfp);
643 	if (error == 0) {
644 		freebsd4_cvtstatfs(sfp, &osb);
645 		error = copyout(&osb, uap->buf, sizeof(osb));
646 	}
647 	free(sfp, M_STATFS);
648 	return (error);
649 }
650 
651 /*
652  * Convert a new format statfs structure to an old format statfs structure.
653  */
654 static void
655 freebsd4_cvtstatfs(struct statfs *nsp, struct ostatfs *osp)
656 {
657 
658 	statfs_scale_blocks(nsp, LONG_MAX);
659 	bzero(osp, sizeof(*osp));
660 	osp->f_bsize = nsp->f_bsize;
661 	osp->f_iosize = MIN(nsp->f_iosize, LONG_MAX);
662 	osp->f_blocks = nsp->f_blocks;
663 	osp->f_bfree = nsp->f_bfree;
664 	osp->f_bavail = nsp->f_bavail;
665 	osp->f_files = MIN(nsp->f_files, LONG_MAX);
666 	osp->f_ffree = MIN(nsp->f_ffree, LONG_MAX);
667 	osp->f_owner = nsp->f_owner;
668 	osp->f_type = nsp->f_type;
669 	osp->f_flags = nsp->f_flags;
670 	osp->f_syncwrites = MIN(nsp->f_syncwrites, LONG_MAX);
671 	osp->f_asyncwrites = MIN(nsp->f_asyncwrites, LONG_MAX);
672 	osp->f_syncreads = MIN(nsp->f_syncreads, LONG_MAX);
673 	osp->f_asyncreads = MIN(nsp->f_asyncreads, LONG_MAX);
674 	strlcpy(osp->f_fstypename, nsp->f_fstypename,
675 	    MIN(MFSNAMELEN, OMFSNAMELEN));
676 	strlcpy(osp->f_mntonname, nsp->f_mntonname,
677 	    MIN(MNAMELEN, OMNAMELEN));
678 	strlcpy(osp->f_mntfromname, nsp->f_mntfromname,
679 	    MIN(MNAMELEN, OMNAMELEN));
680 	osp->f_fsid = nsp->f_fsid;
681 }
682 #endif /* COMPAT_FREEBSD4 */
683 
684 #if defined(COMPAT_FREEBSD11)
685 /*
686  * Get old format filesystem statistics.
687  */
688 static void freebsd11_cvtstatfs(struct statfs *, struct freebsd11_statfs *);
689 
690 int
691 freebsd11_statfs(struct thread *td, struct freebsd11_statfs_args *uap)
692 {
693 	struct freebsd11_statfs osb;
694 	struct statfs *sfp;
695 	int error;
696 
697 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
698 	error = kern_statfs(td, uap->path, UIO_USERSPACE, sfp);
699 	if (error == 0) {
700 		freebsd11_cvtstatfs(sfp, &osb);
701 		error = copyout(&osb, uap->buf, sizeof(osb));
702 	}
703 	free(sfp, M_STATFS);
704 	return (error);
705 }
706 
707 /*
708  * Get filesystem statistics.
709  */
710 int
711 freebsd11_fstatfs(struct thread *td, struct freebsd11_fstatfs_args *uap)
712 {
713 	struct freebsd11_statfs osb;
714 	struct statfs *sfp;
715 	int error;
716 
717 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
718 	error = kern_fstatfs(td, uap->fd, sfp);
719 	if (error == 0) {
720 		freebsd11_cvtstatfs(sfp, &osb);
721 		error = copyout(&osb, uap->buf, sizeof(osb));
722 	}
723 	free(sfp, M_STATFS);
724 	return (error);
725 }
726 
727 /*
728  * Get statistics on all filesystems.
729  */
730 int
731 freebsd11_getfsstat(struct thread *td, struct freebsd11_getfsstat_args *uap)
732 {
733 	struct freebsd11_statfs osb;
734 	struct statfs *buf, *sp;
735 	size_t count, size;
736 	int error;
737 
738 	count = uap->bufsize / sizeof(struct ostatfs);
739 	size = count * sizeof(struct statfs);
740 	error = kern_getfsstat(td, &buf, size, &count, UIO_SYSSPACE,
741 	    uap->mode);
742 	if (error == 0)
743 		td->td_retval[0] = count;
744 	if (size > 0) {
745 		sp = buf;
746 		while (count > 0 && error == 0) {
747 			freebsd11_cvtstatfs(sp, &osb);
748 			error = copyout(&osb, uap->buf, sizeof(osb));
749 			sp++;
750 			uap->buf++;
751 			count--;
752 		}
753 		free(buf, M_STATFS);
754 	}
755 	return (error);
756 }
757 
758 /*
759  * Implement fstatfs() for (NFS) file handles.
760  */
761 int
762 freebsd11_fhstatfs(struct thread *td, struct freebsd11_fhstatfs_args *uap)
763 {
764 	struct freebsd11_statfs osb;
765 	struct statfs *sfp;
766 	fhandle_t fh;
767 	int error;
768 
769 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
770 	if (error)
771 		return (error);
772 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
773 	error = kern_fhstatfs(td, fh, sfp);
774 	if (error == 0) {
775 		freebsd11_cvtstatfs(sfp, &osb);
776 		error = copyout(&osb, uap->buf, sizeof(osb));
777 	}
778 	free(sfp, M_STATFS);
779 	return (error);
780 }
781 
782 /*
783  * Convert a new format statfs structure to an old format statfs structure.
784  */
785 static void
786 freebsd11_cvtstatfs(struct statfs *nsp, struct freebsd11_statfs *osp)
787 {
788 
789 	bzero(osp, sizeof(*osp));
790 	osp->f_version = FREEBSD11_STATFS_VERSION;
791 	osp->f_type = nsp->f_type;
792 	osp->f_flags = nsp->f_flags;
793 	osp->f_bsize = nsp->f_bsize;
794 	osp->f_iosize = nsp->f_iosize;
795 	osp->f_blocks = nsp->f_blocks;
796 	osp->f_bfree = nsp->f_bfree;
797 	osp->f_bavail = nsp->f_bavail;
798 	osp->f_files = nsp->f_files;
799 	osp->f_ffree = nsp->f_ffree;
800 	osp->f_syncwrites = nsp->f_syncwrites;
801 	osp->f_asyncwrites = nsp->f_asyncwrites;
802 	osp->f_syncreads = nsp->f_syncreads;
803 	osp->f_asyncreads = nsp->f_asyncreads;
804 	osp->f_namemax = nsp->f_namemax;
805 	osp->f_owner = nsp->f_owner;
806 	osp->f_fsid = nsp->f_fsid;
807 	strlcpy(osp->f_fstypename, nsp->f_fstypename,
808 	    MIN(MFSNAMELEN, sizeof(osp->f_fstypename)));
809 	strlcpy(osp->f_mntonname, nsp->f_mntonname,
810 	    MIN(MNAMELEN, sizeof(osp->f_mntonname)));
811 	strlcpy(osp->f_mntfromname, nsp->f_mntfromname,
812 	    MIN(MNAMELEN, sizeof(osp->f_mntfromname)));
813 }
814 #endif /* COMPAT_FREEBSD11 */
815 
816 /*
817  * Change current working directory to a given file descriptor.
818  */
819 #ifndef _SYS_SYSPROTO_H_
820 struct fchdir_args {
821 	int	fd;
822 };
823 #endif
824 int
825 sys_fchdir(struct thread *td, struct fchdir_args *uap)
826 {
827 	struct vnode *vp, *tdp;
828 	struct mount *mp;
829 	struct file *fp;
830 	int error;
831 
832 	AUDIT_ARG_FD(uap->fd);
833 	error = getvnode(td, uap->fd, &cap_fchdir_rights,
834 	    &fp);
835 	if (error != 0)
836 		return (error);
837 	vp = fp->f_vnode;
838 	vrefact(vp);
839 	fdrop(fp, td);
840 	vn_lock(vp, LK_SHARED | LK_RETRY);
841 	AUDIT_ARG_VNODE1(vp);
842 	error = change_dir(vp, td);
843 	while (!error && (mp = vp->v_mountedhere) != NULL) {
844 		if (vfs_busy(mp, 0))
845 			continue;
846 		error = VFS_ROOT(mp, LK_SHARED, &tdp);
847 		vfs_unbusy(mp);
848 		if (error != 0)
849 			break;
850 		vput(vp);
851 		vp = tdp;
852 	}
853 	if (error != 0) {
854 		vput(vp);
855 		return (error);
856 	}
857 	VOP_UNLOCK(vp, 0);
858 	pwd_chdir(td, vp);
859 	return (0);
860 }
861 
862 /*
863  * Change current working directory (``.'').
864  */
865 #ifndef _SYS_SYSPROTO_H_
866 struct chdir_args {
867 	char	*path;
868 };
869 #endif
870 int
871 sys_chdir(struct thread *td, struct chdir_args *uap)
872 {
873 
874 	return (kern_chdir(td, uap->path, UIO_USERSPACE));
875 }
876 
877 int
878 kern_chdir(struct thread *td, char *path, enum uio_seg pathseg)
879 {
880 	struct nameidata nd;
881 	int error;
882 
883 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
884 	    pathseg, path, td);
885 	if ((error = namei(&nd)) != 0)
886 		return (error);
887 	if ((error = change_dir(nd.ni_vp, td)) != 0) {
888 		vput(nd.ni_vp);
889 		NDFREE(&nd, NDF_ONLY_PNBUF);
890 		return (error);
891 	}
892 	VOP_UNLOCK(nd.ni_vp, 0);
893 	NDFREE(&nd, NDF_ONLY_PNBUF);
894 	pwd_chdir(td, nd.ni_vp);
895 	return (0);
896 }
897 
898 /*
899  * Change notion of root (``/'') directory.
900  */
901 #ifndef _SYS_SYSPROTO_H_
902 struct chroot_args {
903 	char	*path;
904 };
905 #endif
906 int
907 sys_chroot(struct thread *td, struct chroot_args *uap)
908 {
909 	struct nameidata nd;
910 	int error;
911 
912 	error = priv_check(td, PRIV_VFS_CHROOT);
913 	if (error != 0)
914 		return (error);
915 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
916 	    UIO_USERSPACE, uap->path, td);
917 	error = namei(&nd);
918 	if (error != 0)
919 		goto error;
920 	error = change_dir(nd.ni_vp, td);
921 	if (error != 0)
922 		goto e_vunlock;
923 #ifdef MAC
924 	error = mac_vnode_check_chroot(td->td_ucred, nd.ni_vp);
925 	if (error != 0)
926 		goto e_vunlock;
927 #endif
928 	VOP_UNLOCK(nd.ni_vp, 0);
929 	error = pwd_chroot(td, nd.ni_vp);
930 	vrele(nd.ni_vp);
931 	NDFREE(&nd, NDF_ONLY_PNBUF);
932 	return (error);
933 e_vunlock:
934 	vput(nd.ni_vp);
935 error:
936 	NDFREE(&nd, NDF_ONLY_PNBUF);
937 	return (error);
938 }
939 
940 /*
941  * Common routine for chroot and chdir.  Callers must provide a locked vnode
942  * instance.
943  */
944 int
945 change_dir(struct vnode *vp, struct thread *td)
946 {
947 #ifdef MAC
948 	int error;
949 #endif
950 
951 	ASSERT_VOP_LOCKED(vp, "change_dir(): vp not locked");
952 	if (vp->v_type != VDIR)
953 		return (ENOTDIR);
954 #ifdef MAC
955 	error = mac_vnode_check_chdir(td->td_ucred, vp);
956 	if (error != 0)
957 		return (error);
958 #endif
959 	return (VOP_ACCESS(vp, VEXEC, td->td_ucred, td));
960 }
961 
962 static __inline void
963 flags_to_rights(int flags, cap_rights_t *rightsp)
964 {
965 
966 	if (flags & O_EXEC) {
967 		cap_rights_set(rightsp, CAP_FEXECVE);
968 	} else {
969 		switch ((flags & O_ACCMODE)) {
970 		case O_RDONLY:
971 			cap_rights_set(rightsp, CAP_READ);
972 			break;
973 		case O_RDWR:
974 			cap_rights_set(rightsp, CAP_READ);
975 			/* FALLTHROUGH */
976 		case O_WRONLY:
977 			cap_rights_set(rightsp, CAP_WRITE);
978 			if (!(flags & (O_APPEND | O_TRUNC)))
979 				cap_rights_set(rightsp, CAP_SEEK);
980 			break;
981 		}
982 	}
983 
984 	if (flags & O_CREAT)
985 		cap_rights_set(rightsp, CAP_CREATE);
986 
987 	if (flags & O_TRUNC)
988 		cap_rights_set(rightsp, CAP_FTRUNCATE);
989 
990 	if (flags & (O_SYNC | O_FSYNC))
991 		cap_rights_set(rightsp, CAP_FSYNC);
992 
993 	if (flags & (O_EXLOCK | O_SHLOCK))
994 		cap_rights_set(rightsp, CAP_FLOCK);
995 }
996 
997 /*
998  * Check permissions, allocate an open file structure, and call the device
999  * open routine if any.
1000  */
1001 #ifndef _SYS_SYSPROTO_H_
1002 struct open_args {
1003 	char	*path;
1004 	int	flags;
1005 	int	mode;
1006 };
1007 #endif
1008 int
1009 sys_open(struct thread *td, struct open_args *uap)
1010 {
1011 
1012 	return (kern_openat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1013 	    uap->flags, uap->mode));
1014 }
1015 
1016 #ifndef _SYS_SYSPROTO_H_
1017 struct openat_args {
1018 	int	fd;
1019 	char	*path;
1020 	int	flag;
1021 	int	mode;
1022 };
1023 #endif
1024 int
1025 sys_openat(struct thread *td, struct openat_args *uap)
1026 {
1027 
1028 	AUDIT_ARG_FD(uap->fd);
1029 	return (kern_openat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
1030 	    uap->mode));
1031 }
1032 
1033 int
1034 kern_openat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1035     int flags, int mode)
1036 {
1037 	struct proc *p = td->td_proc;
1038 	struct filedesc *fdp = p->p_fd;
1039 	struct file *fp;
1040 	struct vnode *vp;
1041 	struct nameidata nd;
1042 	cap_rights_t rights;
1043 	int cmode, error, indx;
1044 
1045 	indx = -1;
1046 
1047 	AUDIT_ARG_FFLAGS(flags);
1048 	AUDIT_ARG_MODE(mode);
1049 	cap_rights_init(&rights, CAP_LOOKUP);
1050 	flags_to_rights(flags, &rights);
1051 	/*
1052 	 * Only one of the O_EXEC, O_RDONLY, O_WRONLY and O_RDWR flags
1053 	 * may be specified.
1054 	 */
1055 	if (flags & O_EXEC) {
1056 		if (flags & O_ACCMODE)
1057 			return (EINVAL);
1058 	} else if ((flags & O_ACCMODE) == O_ACCMODE) {
1059 		return (EINVAL);
1060 	} else {
1061 		flags = FFLAGS(flags);
1062 	}
1063 
1064 	/*
1065 	 * Allocate a file structure. The descriptor to reference it
1066 	 * is allocated and set by finstall() below.
1067 	 */
1068 	error = falloc_noinstall(td, &fp);
1069 	if (error != 0)
1070 		return (error);
1071 	/*
1072 	 * An extra reference on `fp' has been held for us by
1073 	 * falloc_noinstall().
1074 	 */
1075 	/* Set the flags early so the finit in devfs can pick them up. */
1076 	fp->f_flag = flags & FMASK;
1077 	cmode = ((mode & ~fdp->fd_cmask) & ALLPERMS) & ~S_ISTXT;
1078 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, fd,
1079 	    &rights, td);
1080 	td->td_dupfd = -1;		/* XXX check for fdopen */
1081 	error = vn_open(&nd, &flags, cmode, fp);
1082 	if (error != 0) {
1083 		/*
1084 		 * If the vn_open replaced the method vector, something
1085 		 * wonderous happened deep below and we just pass it up
1086 		 * pretending we know what we do.
1087 		 */
1088 		if (error == ENXIO && fp->f_ops != &badfileops)
1089 			goto success;
1090 
1091 		/*
1092 		 * Handle special fdopen() case. bleh.
1093 		 *
1094 		 * Don't do this for relative (capability) lookups; we don't
1095 		 * understand exactly what would happen, and we don't think
1096 		 * that it ever should.
1097 		 */
1098 		if ((nd.ni_lcf & NI_LCF_STRICTRELATIVE) == 0 &&
1099 		    (error == ENODEV || error == ENXIO) &&
1100 		    td->td_dupfd >= 0) {
1101 			error = dupfdopen(td, fdp, td->td_dupfd, flags, error,
1102 			    &indx);
1103 			if (error == 0)
1104 				goto success;
1105 		}
1106 
1107 		goto bad;
1108 	}
1109 	td->td_dupfd = 0;
1110 	NDFREE(&nd, NDF_ONLY_PNBUF);
1111 	vp = nd.ni_vp;
1112 
1113 	/*
1114 	 * Store the vnode, for any f_type. Typically, the vnode use
1115 	 * count is decremented by direct call to vn_closefile() for
1116 	 * files that switched type in the cdevsw fdopen() method.
1117 	 */
1118 	fp->f_vnode = vp;
1119 	/*
1120 	 * If the file wasn't claimed by devfs bind it to the normal
1121 	 * vnode operations here.
1122 	 */
1123 	if (fp->f_ops == &badfileops) {
1124 		KASSERT(vp->v_type != VFIFO, ("Unexpected fifo."));
1125 		fp->f_seqcount = 1;
1126 		finit(fp, (flags & FMASK) | (fp->f_flag & FHASLOCK),
1127 		    DTYPE_VNODE, vp, &vnops);
1128 	}
1129 
1130 	VOP_UNLOCK(vp, 0);
1131 	if (flags & O_TRUNC) {
1132 		error = fo_truncate(fp, 0, td->td_ucred, td);
1133 		if (error != 0)
1134 			goto bad;
1135 	}
1136 success:
1137 	/*
1138 	 * If we haven't already installed the FD (for dupfdopen), do so now.
1139 	 */
1140 	if (indx == -1) {
1141 		struct filecaps *fcaps;
1142 
1143 #ifdef CAPABILITIES
1144 		if ((nd.ni_lcf & NI_LCF_STRICTRELATIVE) != 0)
1145 			fcaps = &nd.ni_filecaps;
1146 		else
1147 #endif
1148 			fcaps = NULL;
1149 		error = finstall(td, fp, &indx, flags, fcaps);
1150 		/* On success finstall() consumes fcaps. */
1151 		if (error != 0) {
1152 			filecaps_free(&nd.ni_filecaps);
1153 			goto bad;
1154 		}
1155 	} else {
1156 		filecaps_free(&nd.ni_filecaps);
1157 	}
1158 
1159 	/*
1160 	 * Release our private reference, leaving the one associated with
1161 	 * the descriptor table intact.
1162 	 */
1163 	fdrop(fp, td);
1164 	td->td_retval[0] = indx;
1165 	return (0);
1166 bad:
1167 	KASSERT(indx == -1, ("indx=%d, should be -1", indx));
1168 	fdrop(fp, td);
1169 	return (error);
1170 }
1171 
1172 #ifdef COMPAT_43
1173 /*
1174  * Create a file.
1175  */
1176 #ifndef _SYS_SYSPROTO_H_
1177 struct ocreat_args {
1178 	char	*path;
1179 	int	mode;
1180 };
1181 #endif
1182 int
1183 ocreat(struct thread *td, struct ocreat_args *uap)
1184 {
1185 
1186 	return (kern_openat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1187 	    O_WRONLY | O_CREAT | O_TRUNC, uap->mode));
1188 }
1189 #endif /* COMPAT_43 */
1190 
1191 /*
1192  * Create a special file.
1193  */
1194 #ifndef _SYS_SYSPROTO_H_
1195 struct mknodat_args {
1196 	int	fd;
1197 	char	*path;
1198 	mode_t	mode;
1199 	dev_t	dev;
1200 };
1201 #endif
1202 int
1203 sys_mknodat(struct thread *td, struct mknodat_args *uap)
1204 {
1205 
1206 	return (kern_mknodat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode,
1207 	    uap->dev));
1208 }
1209 
1210 #if defined(COMPAT_FREEBSD11)
1211 int
1212 freebsd11_mknod(struct thread *td,
1213     struct freebsd11_mknod_args *uap)
1214 {
1215 
1216 	return (kern_mknodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1217 	    uap->mode, uap->dev));
1218 }
1219 
1220 int
1221 freebsd11_mknodat(struct thread *td,
1222     struct freebsd11_mknodat_args *uap)
1223 {
1224 
1225 	return (kern_mknodat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode,
1226 	    uap->dev));
1227 }
1228 #endif /* COMPAT_FREEBSD11 */
1229 
1230 int
1231 kern_mknodat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1232     int mode, dev_t dev)
1233 {
1234 	struct vnode *vp;
1235 	struct mount *mp;
1236 	struct vattr vattr;
1237 	struct nameidata nd;
1238 	int error, whiteout = 0;
1239 
1240 	AUDIT_ARG_MODE(mode);
1241 	AUDIT_ARG_DEV(dev);
1242 	switch (mode & S_IFMT) {
1243 	case S_IFCHR:
1244 	case S_IFBLK:
1245 		error = priv_check(td, PRIV_VFS_MKNOD_DEV);
1246 		if (error == 0 && dev == VNOVAL)
1247 			error = EINVAL;
1248 		break;
1249 	case S_IFWHT:
1250 		error = priv_check(td, PRIV_VFS_MKNOD_WHT);
1251 		break;
1252 	case S_IFIFO:
1253 		if (dev == 0)
1254 			return (kern_mkfifoat(td, fd, path, pathseg, mode));
1255 		/* FALLTHROUGH */
1256 	default:
1257 		error = EINVAL;
1258 		break;
1259 	}
1260 	if (error != 0)
1261 		return (error);
1262 restart:
1263 	bwillwrite();
1264 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
1265 	    NOCACHE, pathseg, path, fd, &cap_mknodat_rights,
1266 	    td);
1267 	if ((error = namei(&nd)) != 0)
1268 		return (error);
1269 	vp = nd.ni_vp;
1270 	if (vp != NULL) {
1271 		NDFREE(&nd, NDF_ONLY_PNBUF);
1272 		if (vp == nd.ni_dvp)
1273 			vrele(nd.ni_dvp);
1274 		else
1275 			vput(nd.ni_dvp);
1276 		vrele(vp);
1277 		return (EEXIST);
1278 	} else {
1279 		VATTR_NULL(&vattr);
1280 		vattr.va_mode = (mode & ALLPERMS) &
1281 		    ~td->td_proc->p_fd->fd_cmask;
1282 		vattr.va_rdev = dev;
1283 		whiteout = 0;
1284 
1285 		switch (mode & S_IFMT) {
1286 		case S_IFCHR:
1287 			vattr.va_type = VCHR;
1288 			break;
1289 		case S_IFBLK:
1290 			vattr.va_type = VBLK;
1291 			break;
1292 		case S_IFWHT:
1293 			whiteout = 1;
1294 			break;
1295 		default:
1296 			panic("kern_mknod: invalid mode");
1297 		}
1298 	}
1299 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1300 		NDFREE(&nd, NDF_ONLY_PNBUF);
1301 		vput(nd.ni_dvp);
1302 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1303 			return (error);
1304 		goto restart;
1305 	}
1306 #ifdef MAC
1307 	if (error == 0 && !whiteout)
1308 		error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp,
1309 		    &nd.ni_cnd, &vattr);
1310 #endif
1311 	if (error == 0) {
1312 		if (whiteout)
1313 			error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE);
1314 		else {
1315 			error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp,
1316 						&nd.ni_cnd, &vattr);
1317 			if (error == 0)
1318 				vput(nd.ni_vp);
1319 		}
1320 	}
1321 	NDFREE(&nd, NDF_ONLY_PNBUF);
1322 	vput(nd.ni_dvp);
1323 	vn_finished_write(mp);
1324 	return (error);
1325 }
1326 
1327 /*
1328  * Create a named pipe.
1329  */
1330 #ifndef _SYS_SYSPROTO_H_
1331 struct mkfifo_args {
1332 	char	*path;
1333 	int	mode;
1334 };
1335 #endif
1336 int
1337 sys_mkfifo(struct thread *td, struct mkfifo_args *uap)
1338 {
1339 
1340 	return (kern_mkfifoat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1341 	    uap->mode));
1342 }
1343 
1344 #ifndef _SYS_SYSPROTO_H_
1345 struct mkfifoat_args {
1346 	int	fd;
1347 	char	*path;
1348 	mode_t	mode;
1349 };
1350 #endif
1351 int
1352 sys_mkfifoat(struct thread *td, struct mkfifoat_args *uap)
1353 {
1354 
1355 	return (kern_mkfifoat(td, uap->fd, uap->path, UIO_USERSPACE,
1356 	    uap->mode));
1357 }
1358 
1359 int
1360 kern_mkfifoat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1361     int mode)
1362 {
1363 	struct mount *mp;
1364 	struct vattr vattr;
1365 	struct nameidata nd;
1366 	int error;
1367 
1368 	AUDIT_ARG_MODE(mode);
1369 restart:
1370 	bwillwrite();
1371 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
1372 	    NOCACHE, pathseg, path, fd, &cap_mkfifoat_rights,
1373 	    td);
1374 	if ((error = namei(&nd)) != 0)
1375 		return (error);
1376 	if (nd.ni_vp != NULL) {
1377 		NDFREE(&nd, NDF_ONLY_PNBUF);
1378 		if (nd.ni_vp == nd.ni_dvp)
1379 			vrele(nd.ni_dvp);
1380 		else
1381 			vput(nd.ni_dvp);
1382 		vrele(nd.ni_vp);
1383 		return (EEXIST);
1384 	}
1385 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1386 		NDFREE(&nd, NDF_ONLY_PNBUF);
1387 		vput(nd.ni_dvp);
1388 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1389 			return (error);
1390 		goto restart;
1391 	}
1392 	VATTR_NULL(&vattr);
1393 	vattr.va_type = VFIFO;
1394 	vattr.va_mode = (mode & ALLPERMS) & ~td->td_proc->p_fd->fd_cmask;
1395 #ifdef MAC
1396 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1397 	    &vattr);
1398 	if (error != 0)
1399 		goto out;
1400 #endif
1401 	error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
1402 	if (error == 0)
1403 		vput(nd.ni_vp);
1404 #ifdef MAC
1405 out:
1406 #endif
1407 	vput(nd.ni_dvp);
1408 	vn_finished_write(mp);
1409 	NDFREE(&nd, NDF_ONLY_PNBUF);
1410 	return (error);
1411 }
1412 
1413 /*
1414  * Make a hard file link.
1415  */
1416 #ifndef _SYS_SYSPROTO_H_
1417 struct link_args {
1418 	char	*path;
1419 	char	*link;
1420 };
1421 #endif
1422 int
1423 sys_link(struct thread *td, struct link_args *uap)
1424 {
1425 
1426 	return (kern_linkat(td, AT_FDCWD, AT_FDCWD, uap->path, uap->link,
1427 	    UIO_USERSPACE, FOLLOW));
1428 }
1429 
1430 #ifndef _SYS_SYSPROTO_H_
1431 struct linkat_args {
1432 	int	fd1;
1433 	char	*path1;
1434 	int	fd2;
1435 	char	*path2;
1436 	int	flag;
1437 };
1438 #endif
1439 int
1440 sys_linkat(struct thread *td, struct linkat_args *uap)
1441 {
1442 	int flag;
1443 
1444 	flag = uap->flag;
1445 	if (flag & ~AT_SYMLINK_FOLLOW)
1446 		return (EINVAL);
1447 
1448 	return (kern_linkat(td, uap->fd1, uap->fd2, uap->path1, uap->path2,
1449 	    UIO_USERSPACE, (flag & AT_SYMLINK_FOLLOW) ? FOLLOW : NOFOLLOW));
1450 }
1451 
1452 int hardlink_check_uid = 0;
1453 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_uid, CTLFLAG_RW,
1454     &hardlink_check_uid, 0,
1455     "Unprivileged processes cannot create hard links to files owned by other "
1456     "users");
1457 static int hardlink_check_gid = 0;
1458 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_gid, CTLFLAG_RW,
1459     &hardlink_check_gid, 0,
1460     "Unprivileged processes cannot create hard links to files owned by other "
1461     "groups");
1462 
1463 static int
1464 can_hardlink(struct vnode *vp, struct ucred *cred)
1465 {
1466 	struct vattr va;
1467 	int error;
1468 
1469 	if (!hardlink_check_uid && !hardlink_check_gid)
1470 		return (0);
1471 
1472 	error = VOP_GETATTR(vp, &va, cred);
1473 	if (error != 0)
1474 		return (error);
1475 
1476 	if (hardlink_check_uid && cred->cr_uid != va.va_uid) {
1477 		error = priv_check_cred(cred, PRIV_VFS_LINK, 0);
1478 		if (error != 0)
1479 			return (error);
1480 	}
1481 
1482 	if (hardlink_check_gid && !groupmember(va.va_gid, cred)) {
1483 		error = priv_check_cred(cred, PRIV_VFS_LINK, 0);
1484 		if (error != 0)
1485 			return (error);
1486 	}
1487 
1488 	return (0);
1489 }
1490 
1491 int
1492 kern_linkat(struct thread *td, int fd1, int fd2, char *path1, char *path2,
1493     enum uio_seg segflg, int follow)
1494 {
1495 	struct vnode *vp;
1496 	struct mount *mp;
1497 	struct nameidata nd;
1498 	int error;
1499 
1500 again:
1501 	bwillwrite();
1502 	NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, segflg, path1, fd1,
1503 	    &cap_linkat_source_rights, td);
1504 
1505 	if ((error = namei(&nd)) != 0)
1506 		return (error);
1507 	NDFREE(&nd, NDF_ONLY_PNBUF);
1508 	vp = nd.ni_vp;
1509 	if (vp->v_type == VDIR) {
1510 		vrele(vp);
1511 		return (EPERM);		/* POSIX */
1512 	}
1513 	NDINIT_ATRIGHTS(&nd, CREATE,
1514 	    LOCKPARENT | SAVENAME | AUDITVNODE2 | NOCACHE, segflg, path2, fd2,
1515 	    &cap_linkat_target_rights, td);
1516 	if ((error = namei(&nd)) == 0) {
1517 		if (nd.ni_vp != NULL) {
1518 			NDFREE(&nd, NDF_ONLY_PNBUF);
1519 			if (nd.ni_dvp == nd.ni_vp)
1520 				vrele(nd.ni_dvp);
1521 			else
1522 				vput(nd.ni_dvp);
1523 			vrele(nd.ni_vp);
1524 			vrele(vp);
1525 			return (EEXIST);
1526 		} else if (nd.ni_dvp->v_mount != vp->v_mount) {
1527 			/*
1528 			 * Cross-device link.  No need to recheck
1529 			 * vp->v_type, since it cannot change, except
1530 			 * to VBAD.
1531 			 */
1532 			NDFREE(&nd, NDF_ONLY_PNBUF);
1533 			vput(nd.ni_dvp);
1534 			vrele(vp);
1535 			return (EXDEV);
1536 		} else if ((error = vn_lock(vp, LK_EXCLUSIVE)) == 0) {
1537 			error = can_hardlink(vp, td->td_ucred);
1538 #ifdef MAC
1539 			if (error == 0)
1540 				error = mac_vnode_check_link(td->td_ucred,
1541 				    nd.ni_dvp, vp, &nd.ni_cnd);
1542 #endif
1543 			if (error != 0) {
1544 				vput(vp);
1545 				vput(nd.ni_dvp);
1546 				NDFREE(&nd, NDF_ONLY_PNBUF);
1547 				return (error);
1548 			}
1549 			error = vn_start_write(vp, &mp, V_NOWAIT);
1550 			if (error != 0) {
1551 				vput(vp);
1552 				vput(nd.ni_dvp);
1553 				NDFREE(&nd, NDF_ONLY_PNBUF);
1554 				error = vn_start_write(NULL, &mp,
1555 				    V_XSLEEP | PCATCH);
1556 				if (error != 0)
1557 					return (error);
1558 				goto again;
1559 			}
1560 			error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
1561 			VOP_UNLOCK(vp, 0);
1562 			vput(nd.ni_dvp);
1563 			vn_finished_write(mp);
1564 			NDFREE(&nd, NDF_ONLY_PNBUF);
1565 		} else {
1566 			vput(nd.ni_dvp);
1567 			NDFREE(&nd, NDF_ONLY_PNBUF);
1568 			vrele(vp);
1569 			goto again;
1570 		}
1571 	}
1572 	vrele(vp);
1573 	return (error);
1574 }
1575 
1576 /*
1577  * Make a symbolic link.
1578  */
1579 #ifndef _SYS_SYSPROTO_H_
1580 struct symlink_args {
1581 	char	*path;
1582 	char	*link;
1583 };
1584 #endif
1585 int
1586 sys_symlink(struct thread *td, struct symlink_args *uap)
1587 {
1588 
1589 	return (kern_symlinkat(td, uap->path, AT_FDCWD, uap->link,
1590 	    UIO_USERSPACE));
1591 }
1592 
1593 #ifndef _SYS_SYSPROTO_H_
1594 struct symlinkat_args {
1595 	char	*path;
1596 	int	fd;
1597 	char	*path2;
1598 };
1599 #endif
1600 int
1601 sys_symlinkat(struct thread *td, struct symlinkat_args *uap)
1602 {
1603 
1604 	return (kern_symlinkat(td, uap->path1, uap->fd, uap->path2,
1605 	    UIO_USERSPACE));
1606 }
1607 
1608 int
1609 kern_symlinkat(struct thread *td, char *path1, int fd, char *path2,
1610     enum uio_seg segflg)
1611 {
1612 	struct mount *mp;
1613 	struct vattr vattr;
1614 	char *syspath;
1615 	struct nameidata nd;
1616 	int error;
1617 
1618 	if (segflg == UIO_SYSSPACE) {
1619 		syspath = path1;
1620 	} else {
1621 		syspath = uma_zalloc(namei_zone, M_WAITOK);
1622 		if ((error = copyinstr(path1, syspath, MAXPATHLEN, NULL)) != 0)
1623 			goto out;
1624 	}
1625 	AUDIT_ARG_TEXT(syspath);
1626 restart:
1627 	bwillwrite();
1628 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
1629 	    NOCACHE, segflg, path2, fd, &cap_symlinkat_rights,
1630 	    td);
1631 	if ((error = namei(&nd)) != 0)
1632 		goto out;
1633 	if (nd.ni_vp) {
1634 		NDFREE(&nd, NDF_ONLY_PNBUF);
1635 		if (nd.ni_vp == nd.ni_dvp)
1636 			vrele(nd.ni_dvp);
1637 		else
1638 			vput(nd.ni_dvp);
1639 		vrele(nd.ni_vp);
1640 		error = EEXIST;
1641 		goto out;
1642 	}
1643 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1644 		NDFREE(&nd, NDF_ONLY_PNBUF);
1645 		vput(nd.ni_dvp);
1646 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1647 			goto out;
1648 		goto restart;
1649 	}
1650 	VATTR_NULL(&vattr);
1651 	vattr.va_mode = ACCESSPERMS &~ td->td_proc->p_fd->fd_cmask;
1652 #ifdef MAC
1653 	vattr.va_type = VLNK;
1654 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1655 	    &vattr);
1656 	if (error != 0)
1657 		goto out2;
1658 #endif
1659 	error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, syspath);
1660 	if (error == 0)
1661 		vput(nd.ni_vp);
1662 #ifdef MAC
1663 out2:
1664 #endif
1665 	NDFREE(&nd, NDF_ONLY_PNBUF);
1666 	vput(nd.ni_dvp);
1667 	vn_finished_write(mp);
1668 out:
1669 	if (segflg != UIO_SYSSPACE)
1670 		uma_zfree(namei_zone, syspath);
1671 	return (error);
1672 }
1673 
1674 /*
1675  * Delete a whiteout from the filesystem.
1676  */
1677 #ifndef _SYS_SYSPROTO_H_
1678 struct undelete_args {
1679 	char *path;
1680 };
1681 #endif
1682 int
1683 sys_undelete(struct thread *td, struct undelete_args *uap)
1684 {
1685 	struct mount *mp;
1686 	struct nameidata nd;
1687 	int error;
1688 
1689 restart:
1690 	bwillwrite();
1691 	NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | AUDITVNODE1,
1692 	    UIO_USERSPACE, uap->path, td);
1693 	error = namei(&nd);
1694 	if (error != 0)
1695 		return (error);
1696 
1697 	if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) {
1698 		NDFREE(&nd, NDF_ONLY_PNBUF);
1699 		if (nd.ni_vp == nd.ni_dvp)
1700 			vrele(nd.ni_dvp);
1701 		else
1702 			vput(nd.ni_dvp);
1703 		if (nd.ni_vp)
1704 			vrele(nd.ni_vp);
1705 		return (EEXIST);
1706 	}
1707 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1708 		NDFREE(&nd, NDF_ONLY_PNBUF);
1709 		vput(nd.ni_dvp);
1710 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1711 			return (error);
1712 		goto restart;
1713 	}
1714 	error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE);
1715 	NDFREE(&nd, NDF_ONLY_PNBUF);
1716 	vput(nd.ni_dvp);
1717 	vn_finished_write(mp);
1718 	return (error);
1719 }
1720 
1721 /*
1722  * Delete a name from the filesystem.
1723  */
1724 #ifndef _SYS_SYSPROTO_H_
1725 struct unlink_args {
1726 	char	*path;
1727 };
1728 #endif
1729 int
1730 sys_unlink(struct thread *td, struct unlink_args *uap)
1731 {
1732 
1733 	return (kern_unlinkat(td, AT_FDCWD, uap->path, UIO_USERSPACE, 0));
1734 }
1735 
1736 #ifndef _SYS_SYSPROTO_H_
1737 struct unlinkat_args {
1738 	int	fd;
1739 	char	*path;
1740 	int	flag;
1741 };
1742 #endif
1743 int
1744 sys_unlinkat(struct thread *td, struct unlinkat_args *uap)
1745 {
1746 	int flag = uap->flag;
1747 	int fd = uap->fd;
1748 	char *path = uap->path;
1749 
1750 	if (flag & ~AT_REMOVEDIR)
1751 		return (EINVAL);
1752 
1753 	if (flag & AT_REMOVEDIR)
1754 		return (kern_rmdirat(td, fd, path, UIO_USERSPACE));
1755 	else
1756 		return (kern_unlinkat(td, fd, path, UIO_USERSPACE, 0));
1757 }
1758 
1759 int
1760 kern_unlinkat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1761     ino_t oldinum)
1762 {
1763 	struct mount *mp;
1764 	struct vnode *vp;
1765 	struct nameidata nd;
1766 	struct stat sb;
1767 	int error;
1768 
1769 restart:
1770 	bwillwrite();
1771 	NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1,
1772 	    pathseg, path, fd, &cap_unlinkat_rights, td);
1773 	if ((error = namei(&nd)) != 0)
1774 		return (error == EINVAL ? EPERM : error);
1775 	vp = nd.ni_vp;
1776 	if (vp->v_type == VDIR && oldinum == 0) {
1777 		error = EPERM;		/* POSIX */
1778 	} else if (oldinum != 0 &&
1779 		  ((error = vn_stat(vp, &sb, td->td_ucred, NOCRED, td)) == 0) &&
1780 		  sb.st_ino != oldinum) {
1781 			error = EIDRM;	/* Identifier removed */
1782 	} else {
1783 		/*
1784 		 * The root of a mounted filesystem cannot be deleted.
1785 		 *
1786 		 * XXX: can this only be a VDIR case?
1787 		 */
1788 		if (vp->v_vflag & VV_ROOT)
1789 			error = EBUSY;
1790 	}
1791 	if (error == 0) {
1792 		if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1793 			NDFREE(&nd, NDF_ONLY_PNBUF);
1794 			vput(nd.ni_dvp);
1795 			if (vp == nd.ni_dvp)
1796 				vrele(vp);
1797 			else
1798 				vput(vp);
1799 			if ((error = vn_start_write(NULL, &mp,
1800 			    V_XSLEEP | PCATCH)) != 0)
1801 				return (error);
1802 			goto restart;
1803 		}
1804 #ifdef MAC
1805 		error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
1806 		    &nd.ni_cnd);
1807 		if (error != 0)
1808 			goto out;
1809 #endif
1810 		vfs_notify_upper(vp, VFS_NOTIFY_UPPER_UNLINK);
1811 		error = VOP_REMOVE(nd.ni_dvp, vp, &nd.ni_cnd);
1812 #ifdef MAC
1813 out:
1814 #endif
1815 		vn_finished_write(mp);
1816 	}
1817 	NDFREE(&nd, NDF_ONLY_PNBUF);
1818 	vput(nd.ni_dvp);
1819 	if (vp == nd.ni_dvp)
1820 		vrele(vp);
1821 	else
1822 		vput(vp);
1823 	return (error);
1824 }
1825 
1826 /*
1827  * Reposition read/write file offset.
1828  */
1829 #ifndef _SYS_SYSPROTO_H_
1830 struct lseek_args {
1831 	int	fd;
1832 	int	pad;
1833 	off_t	offset;
1834 	int	whence;
1835 };
1836 #endif
1837 int
1838 sys_lseek(struct thread *td, struct lseek_args *uap)
1839 {
1840 
1841 	return (kern_lseek(td, uap->fd, uap->offset, uap->whence));
1842 }
1843 
1844 int
1845 kern_lseek(struct thread *td, int fd, off_t offset, int whence)
1846 {
1847 	struct file *fp;
1848 	int error;
1849 
1850 	AUDIT_ARG_FD(fd);
1851 	error = fget(td, fd, &cap_seek_rights, &fp);
1852 	if (error != 0)
1853 		return (error);
1854 	error = (fp->f_ops->fo_flags & DFLAG_SEEKABLE) != 0 ?
1855 	    fo_seek(fp, offset, whence, td) : ESPIPE;
1856 	fdrop(fp, td);
1857 	return (error);
1858 }
1859 
1860 #if defined(COMPAT_43)
1861 /*
1862  * Reposition read/write file offset.
1863  */
1864 #ifndef _SYS_SYSPROTO_H_
1865 struct olseek_args {
1866 	int	fd;
1867 	long	offset;
1868 	int	whence;
1869 };
1870 #endif
1871 int
1872 olseek(struct thread *td, struct olseek_args *uap)
1873 {
1874 
1875 	return (kern_lseek(td, uap->fd, uap->offset, uap->whence));
1876 }
1877 #endif /* COMPAT_43 */
1878 
1879 #if defined(COMPAT_FREEBSD6)
1880 /* Version with the 'pad' argument */
1881 int
1882 freebsd6_lseek(struct thread *td, struct freebsd6_lseek_args *uap)
1883 {
1884 
1885 	return (kern_lseek(td, uap->fd, uap->offset, uap->whence));
1886 }
1887 #endif
1888 
1889 /*
1890  * Check access permissions using passed credentials.
1891  */
1892 static int
1893 vn_access(struct vnode *vp, int user_flags, struct ucred *cred,
1894      struct thread *td)
1895 {
1896 	accmode_t accmode;
1897 	int error;
1898 
1899 	/* Flags == 0 means only check for existence. */
1900 	if (user_flags == 0)
1901 		return (0);
1902 
1903 	accmode = 0;
1904 	if (user_flags & R_OK)
1905 		accmode |= VREAD;
1906 	if (user_flags & W_OK)
1907 		accmode |= VWRITE;
1908 	if (user_flags & X_OK)
1909 		accmode |= VEXEC;
1910 #ifdef MAC
1911 	error = mac_vnode_check_access(cred, vp, accmode);
1912 	if (error != 0)
1913 		return (error);
1914 #endif
1915 	if ((accmode & VWRITE) == 0 || (error = vn_writechk(vp)) == 0)
1916 		error = VOP_ACCESS(vp, accmode, cred, td);
1917 	return (error);
1918 }
1919 
1920 /*
1921  * Check access permissions using "real" credentials.
1922  */
1923 #ifndef _SYS_SYSPROTO_H_
1924 struct access_args {
1925 	char	*path;
1926 	int	amode;
1927 };
1928 #endif
1929 int
1930 sys_access(struct thread *td, struct access_args *uap)
1931 {
1932 
1933 	return (kern_accessat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1934 	    0, uap->amode));
1935 }
1936 
1937 #ifndef _SYS_SYSPROTO_H_
1938 struct faccessat_args {
1939 	int	dirfd;
1940 	char	*path;
1941 	int	amode;
1942 	int	flag;
1943 }
1944 #endif
1945 int
1946 sys_faccessat(struct thread *td, struct faccessat_args *uap)
1947 {
1948 
1949 	return (kern_accessat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
1950 	    uap->amode));
1951 }
1952 
1953 int
1954 kern_accessat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1955     int flag, int amode)
1956 {
1957 	struct ucred *cred, *usecred;
1958 	struct vnode *vp;
1959 	struct nameidata nd;
1960 	int error;
1961 
1962 	if (flag & ~AT_EACCESS)
1963 		return (EINVAL);
1964 	if (amode != F_OK && (amode & ~(R_OK | W_OK | X_OK)) != 0)
1965 		return (EINVAL);
1966 
1967 	/*
1968 	 * Create and modify a temporary credential instead of one that
1969 	 * is potentially shared (if we need one).
1970 	 */
1971 	cred = td->td_ucred;
1972 	if ((flag & AT_EACCESS) == 0 &&
1973 	    ((cred->cr_uid != cred->cr_ruid ||
1974 	    cred->cr_rgid != cred->cr_groups[0]))) {
1975 		usecred = crdup(cred);
1976 		usecred->cr_uid = cred->cr_ruid;
1977 		usecred->cr_groups[0] = cred->cr_rgid;
1978 		td->td_ucred = usecred;
1979 	} else
1980 		usecred = cred;
1981 	AUDIT_ARG_VALUE(amode);
1982 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF |
1983 	    AUDITVNODE1, pathseg, path, fd, &cap_fstat_rights,
1984 	    td);
1985 	if ((error = namei(&nd)) != 0)
1986 		goto out;
1987 	vp = nd.ni_vp;
1988 
1989 	error = vn_access(vp, amode, usecred, td);
1990 	NDFREE(&nd, NDF_ONLY_PNBUF);
1991 	vput(vp);
1992 out:
1993 	if (usecred != cred) {
1994 		td->td_ucred = cred;
1995 		crfree(usecred);
1996 	}
1997 	return (error);
1998 }
1999 
2000 /*
2001  * Check access permissions using "effective" credentials.
2002  */
2003 #ifndef _SYS_SYSPROTO_H_
2004 struct eaccess_args {
2005 	char	*path;
2006 	int	amode;
2007 };
2008 #endif
2009 int
2010 sys_eaccess(struct thread *td, struct eaccess_args *uap)
2011 {
2012 
2013 	return (kern_accessat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2014 	    AT_EACCESS, uap->amode));
2015 }
2016 
2017 #if defined(COMPAT_43)
2018 /*
2019  * Get file status; this version follows links.
2020  */
2021 #ifndef _SYS_SYSPROTO_H_
2022 struct ostat_args {
2023 	char	*path;
2024 	struct ostat *ub;
2025 };
2026 #endif
2027 int
2028 ostat(struct thread *td, struct ostat_args *uap)
2029 {
2030 	struct stat sb;
2031 	struct ostat osb;
2032 	int error;
2033 
2034 	error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE,
2035 	    &sb, NULL);
2036 	if (error != 0)
2037 		return (error);
2038 	cvtstat(&sb, &osb);
2039 	return (copyout(&osb, uap->ub, sizeof (osb)));
2040 }
2041 
2042 /*
2043  * Get file status; this version does not follow links.
2044  */
2045 #ifndef _SYS_SYSPROTO_H_
2046 struct olstat_args {
2047 	char	*path;
2048 	struct ostat *ub;
2049 };
2050 #endif
2051 int
2052 olstat(struct thread *td, struct olstat_args *uap)
2053 {
2054 	struct stat sb;
2055 	struct ostat osb;
2056 	int error;
2057 
2058 	error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2059 	    UIO_USERSPACE, &sb, NULL);
2060 	if (error != 0)
2061 		return (error);
2062 	cvtstat(&sb, &osb);
2063 	return (copyout(&osb, uap->ub, sizeof (osb)));
2064 }
2065 
2066 /*
2067  * Convert from an old to a new stat structure.
2068  */
2069 void
2070 cvtstat(struct stat *st, struct ostat *ost)
2071 {
2072 
2073 	bzero(ost, sizeof(*ost));
2074 	ost->st_dev = st->st_dev;
2075 	ost->st_ino = st->st_ino;
2076 	ost->st_mode = st->st_mode;
2077 	ost->st_nlink = st->st_nlink;
2078 	ost->st_uid = st->st_uid;
2079 	ost->st_gid = st->st_gid;
2080 	ost->st_rdev = st->st_rdev;
2081 	if (st->st_size < (quad_t)1 << 32)
2082 		ost->st_size = st->st_size;
2083 	else
2084 		ost->st_size = -2;
2085 	ost->st_atim = st->st_atim;
2086 	ost->st_mtim = st->st_mtim;
2087 	ost->st_ctim = st->st_ctim;
2088 	ost->st_blksize = st->st_blksize;
2089 	ost->st_blocks = st->st_blocks;
2090 	ost->st_flags = st->st_flags;
2091 	ost->st_gen = st->st_gen;
2092 }
2093 #endif /* COMPAT_43 */
2094 
2095 #if defined(COMPAT_43) || defined(COMPAT_FREEBSD11)
2096 int ino64_trunc_error;
2097 SYSCTL_INT(_vfs, OID_AUTO, ino64_trunc_error, CTLFLAG_RW,
2098     &ino64_trunc_error, 0,
2099     "Error on truncation of inode number, device id or link count");
2100 int
2101 freebsd11_cvtstat(struct stat *st, struct freebsd11_stat *ost)
2102 {
2103 
2104 	ost->st_dev = st->st_dev;
2105 	ost->st_ino = st->st_ino;
2106 	if (ost->st_ino != st->st_ino) {
2107 		switch (ino64_trunc_error) {
2108 		default:
2109 		case 0:
2110 			break;
2111 		case 1:
2112 			return (EOVERFLOW);
2113 		case 2:
2114 			ost->st_ino = UINT32_MAX;
2115 			break;
2116 		}
2117 	}
2118 	ost->st_mode = st->st_mode;
2119 	ost->st_nlink = st->st_nlink;
2120 	if (ost->st_nlink != st->st_nlink) {
2121 		switch (ino64_trunc_error) {
2122 		default:
2123 		case 0:
2124 			break;
2125 		case 1:
2126 			return (EOVERFLOW);
2127 		case 2:
2128 			ost->st_nlink = UINT16_MAX;
2129 			break;
2130 		}
2131 	}
2132 	ost->st_uid = st->st_uid;
2133 	ost->st_gid = st->st_gid;
2134 	ost->st_rdev = st->st_rdev;
2135 	ost->st_atim = st->st_atim;
2136 	ost->st_mtim = st->st_mtim;
2137 	ost->st_ctim = st->st_ctim;
2138 	ost->st_size = st->st_size;
2139 	ost->st_blocks = st->st_blocks;
2140 	ost->st_blksize = st->st_blksize;
2141 	ost->st_flags = st->st_flags;
2142 	ost->st_gen = st->st_gen;
2143 	ost->st_lspare = 0;
2144 	ost->st_birthtim = st->st_birthtim;
2145 	bzero((char *)&ost->st_birthtim + sizeof(ost->st_birthtim),
2146 	    sizeof(*ost) - offsetof(struct freebsd11_stat,
2147 	    st_birthtim) - sizeof(ost->st_birthtim));
2148 	return (0);
2149 }
2150 
2151 int
2152 freebsd11_stat(struct thread *td, struct freebsd11_stat_args* uap)
2153 {
2154 	struct stat sb;
2155 	struct freebsd11_stat osb;
2156 	int error;
2157 
2158 	error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE,
2159 	    &sb, NULL);
2160 	if (error != 0)
2161 		return (error);
2162 	error = freebsd11_cvtstat(&sb, &osb);
2163 	if (error == 0)
2164 		error = copyout(&osb, uap->ub, sizeof(osb));
2165 	return (error);
2166 }
2167 
2168 int
2169 freebsd11_lstat(struct thread *td, struct freebsd11_lstat_args* uap)
2170 {
2171 	struct stat sb;
2172 	struct freebsd11_stat osb;
2173 	int error;
2174 
2175 	error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2176 	    UIO_USERSPACE, &sb, NULL);
2177 	if (error != 0)
2178 		return (error);
2179 	error = freebsd11_cvtstat(&sb, &osb);
2180 	if (error == 0)
2181 		error = copyout(&osb, uap->ub, sizeof(osb));
2182 	return (error);
2183 }
2184 
2185 int
2186 freebsd11_fhstat(struct thread *td, struct freebsd11_fhstat_args* uap)
2187 {
2188 	struct fhandle fh;
2189 	struct stat sb;
2190 	struct freebsd11_stat osb;
2191 	int error;
2192 
2193 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
2194 	if (error != 0)
2195 		return (error);
2196 	error = kern_fhstat(td, fh, &sb);
2197 	if (error != 0)
2198 		return (error);
2199 	error = freebsd11_cvtstat(&sb, &osb);
2200 	if (error == 0)
2201 		error = copyout(&osb, uap->sb, sizeof(osb));
2202 	return (error);
2203 }
2204 
2205 int
2206 freebsd11_fstatat(struct thread *td, struct freebsd11_fstatat_args* uap)
2207 {
2208 	struct stat sb;
2209 	struct freebsd11_stat osb;
2210 	int error;
2211 
2212 	error = kern_statat(td, uap->flag, uap->fd, uap->path,
2213 	    UIO_USERSPACE, &sb, NULL);
2214 	if (error != 0)
2215 		return (error);
2216 	error = freebsd11_cvtstat(&sb, &osb);
2217 	if (error == 0)
2218 		error = copyout(&osb, uap->buf, sizeof(osb));
2219 	return (error);
2220 }
2221 #endif	/* COMPAT_FREEBSD11 */
2222 
2223 /*
2224  * Get file status
2225  */
2226 #ifndef _SYS_SYSPROTO_H_
2227 struct fstatat_args {
2228 	int	fd;
2229 	char	*path;
2230 	struct stat	*buf;
2231 	int	flag;
2232 }
2233 #endif
2234 int
2235 sys_fstatat(struct thread *td, struct fstatat_args *uap)
2236 {
2237 	struct stat sb;
2238 	int error;
2239 
2240 	error = kern_statat(td, uap->flag, uap->fd, uap->path,
2241 	    UIO_USERSPACE, &sb, NULL);
2242 	if (error == 0)
2243 		error = copyout(&sb, uap->buf, sizeof (sb));
2244 	return (error);
2245 }
2246 
2247 int
2248 kern_statat(struct thread *td, int flag, int fd, char *path,
2249     enum uio_seg pathseg, struct stat *sbp,
2250     void (*hook)(struct vnode *vp, struct stat *sbp))
2251 {
2252 	struct nameidata nd;
2253 	struct stat sb;
2254 	int error;
2255 
2256 	if (flag & ~AT_SYMLINK_NOFOLLOW)
2257 		return (EINVAL);
2258 
2259 	NDINIT_ATRIGHTS(&nd, LOOKUP, ((flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW :
2260 	    FOLLOW) | LOCKSHARED | LOCKLEAF | AUDITVNODE1, pathseg, path, fd,
2261 	    &cap_fstat_rights, td);
2262 
2263 	if ((error = namei(&nd)) != 0)
2264 		return (error);
2265 	error = vn_stat(nd.ni_vp, &sb, td->td_ucred, NOCRED, td);
2266 	if (error == 0) {
2267 		SDT_PROBE2(vfs, , stat, mode, path, sb.st_mode);
2268 		if (S_ISREG(sb.st_mode))
2269 			SDT_PROBE2(vfs, , stat, reg, path, pathseg);
2270 		if (__predict_false(hook != NULL))
2271 			hook(nd.ni_vp, &sb);
2272 	}
2273 	NDFREE(&nd, NDF_ONLY_PNBUF);
2274 	vput(nd.ni_vp);
2275 	if (error != 0)
2276 		return (error);
2277 #ifdef __STAT_TIME_T_EXT
2278 	sb.st_atim_ext = 0;
2279 	sb.st_mtim_ext = 0;
2280 	sb.st_ctim_ext = 0;
2281 	sb.st_btim_ext = 0;
2282 #endif
2283 	*sbp = sb;
2284 #ifdef KTRACE
2285 	if (KTRPOINT(td, KTR_STRUCT))
2286 		ktrstat(&sb);
2287 #endif
2288 	return (0);
2289 }
2290 
2291 #if defined(COMPAT_FREEBSD11)
2292 /*
2293  * Implementation of the NetBSD [l]stat() functions.
2294  */
2295 void
2296 freebsd11_cvtnstat(struct stat *sb, struct nstat *nsb)
2297 {
2298 
2299 	bzero(nsb, sizeof(*nsb));
2300 	nsb->st_dev = sb->st_dev;
2301 	nsb->st_ino = sb->st_ino;
2302 	nsb->st_mode = sb->st_mode;
2303 	nsb->st_nlink = sb->st_nlink;
2304 	nsb->st_uid = sb->st_uid;
2305 	nsb->st_gid = sb->st_gid;
2306 	nsb->st_rdev = sb->st_rdev;
2307 	nsb->st_atim = sb->st_atim;
2308 	nsb->st_mtim = sb->st_mtim;
2309 	nsb->st_ctim = sb->st_ctim;
2310 	nsb->st_size = sb->st_size;
2311 	nsb->st_blocks = sb->st_blocks;
2312 	nsb->st_blksize = sb->st_blksize;
2313 	nsb->st_flags = sb->st_flags;
2314 	nsb->st_gen = sb->st_gen;
2315 	nsb->st_birthtim = sb->st_birthtim;
2316 }
2317 
2318 #ifndef _SYS_SYSPROTO_H_
2319 struct freebsd11_nstat_args {
2320 	char	*path;
2321 	struct nstat *ub;
2322 };
2323 #endif
2324 int
2325 freebsd11_nstat(struct thread *td, struct freebsd11_nstat_args *uap)
2326 {
2327 	struct stat sb;
2328 	struct nstat nsb;
2329 	int error;
2330 
2331 	error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE,
2332 	    &sb, NULL);
2333 	if (error != 0)
2334 		return (error);
2335 	freebsd11_cvtnstat(&sb, &nsb);
2336 	return (copyout(&nsb, uap->ub, sizeof (nsb)));
2337 }
2338 
2339 /*
2340  * NetBSD lstat.  Get file status; this version does not follow links.
2341  */
2342 #ifndef _SYS_SYSPROTO_H_
2343 struct freebsd11_nlstat_args {
2344 	char	*path;
2345 	struct nstat *ub;
2346 };
2347 #endif
2348 int
2349 freebsd11_nlstat(struct thread *td, struct freebsd11_nlstat_args *uap)
2350 {
2351 	struct stat sb;
2352 	struct nstat nsb;
2353 	int error;
2354 
2355 	error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2356 	    UIO_USERSPACE, &sb, NULL);
2357 	if (error != 0)
2358 		return (error);
2359 	freebsd11_cvtnstat(&sb, &nsb);
2360 	return (copyout(&nsb, uap->ub, sizeof (nsb)));
2361 }
2362 #endif /* COMPAT_FREEBSD11 */
2363 
2364 /*
2365  * Get configurable pathname variables.
2366  */
2367 #ifndef _SYS_SYSPROTO_H_
2368 struct pathconf_args {
2369 	char	*path;
2370 	int	name;
2371 };
2372 #endif
2373 int
2374 sys_pathconf(struct thread *td, struct pathconf_args *uap)
2375 {
2376 	long value;
2377 	int error;
2378 
2379 	error = kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name, FOLLOW,
2380 	    &value);
2381 	if (error == 0)
2382 		td->td_retval[0] = value;
2383 	return (error);
2384 }
2385 
2386 #ifndef _SYS_SYSPROTO_H_
2387 struct lpathconf_args {
2388 	char	*path;
2389 	int	name;
2390 };
2391 #endif
2392 int
2393 sys_lpathconf(struct thread *td, struct lpathconf_args *uap)
2394 {
2395 	long value;
2396 	int error;
2397 
2398 	error = kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name,
2399 	    NOFOLLOW, &value);
2400 	if (error == 0)
2401 		td->td_retval[0] = value;
2402 	return (error);
2403 }
2404 
2405 int
2406 kern_pathconf(struct thread *td, char *path, enum uio_seg pathseg, int name,
2407     u_long flags, long *valuep)
2408 {
2409 	struct nameidata nd;
2410 	int error;
2411 
2412 	NDINIT(&nd, LOOKUP, LOCKSHARED | LOCKLEAF | AUDITVNODE1 | flags,
2413 	    pathseg, path, td);
2414 	if ((error = namei(&nd)) != 0)
2415 		return (error);
2416 	NDFREE(&nd, NDF_ONLY_PNBUF);
2417 
2418 	error = VOP_PATHCONF(nd.ni_vp, name, valuep);
2419 	vput(nd.ni_vp);
2420 	return (error);
2421 }
2422 
2423 /*
2424  * Return target name of a symbolic link.
2425  */
2426 #ifndef _SYS_SYSPROTO_H_
2427 struct readlink_args {
2428 	char	*path;
2429 	char	*buf;
2430 	size_t	count;
2431 };
2432 #endif
2433 int
2434 sys_readlink(struct thread *td, struct readlink_args *uap)
2435 {
2436 
2437 	return (kern_readlinkat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2438 	    uap->buf, UIO_USERSPACE, uap->count));
2439 }
2440 #ifndef _SYS_SYSPROTO_H_
2441 struct readlinkat_args {
2442 	int	fd;
2443 	char	*path;
2444 	char	*buf;
2445 	size_t	bufsize;
2446 };
2447 #endif
2448 int
2449 sys_readlinkat(struct thread *td, struct readlinkat_args *uap)
2450 {
2451 
2452 	return (kern_readlinkat(td, uap->fd, uap->path, UIO_USERSPACE,
2453 	    uap->buf, UIO_USERSPACE, uap->bufsize));
2454 }
2455 
2456 int
2457 kern_readlinkat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2458     char *buf, enum uio_seg bufseg, size_t count)
2459 {
2460 	struct vnode *vp;
2461 	struct iovec aiov;
2462 	struct uio auio;
2463 	struct nameidata nd;
2464 	int error;
2465 
2466 	if (count > IOSIZE_MAX)
2467 		return (EINVAL);
2468 
2469 	NDINIT_AT(&nd, LOOKUP, NOFOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
2470 	    pathseg, path, fd, td);
2471 
2472 	if ((error = namei(&nd)) != 0)
2473 		return (error);
2474 	NDFREE(&nd, NDF_ONLY_PNBUF);
2475 	vp = nd.ni_vp;
2476 #ifdef MAC
2477 	error = mac_vnode_check_readlink(td->td_ucred, vp);
2478 	if (error != 0) {
2479 		vput(vp);
2480 		return (error);
2481 	}
2482 #endif
2483 	if (vp->v_type != VLNK && (vp->v_vflag & VV_READLINK) == 0)
2484 		error = EINVAL;
2485 	else {
2486 		aiov.iov_base = buf;
2487 		aiov.iov_len = count;
2488 		auio.uio_iov = &aiov;
2489 		auio.uio_iovcnt = 1;
2490 		auio.uio_offset = 0;
2491 		auio.uio_rw = UIO_READ;
2492 		auio.uio_segflg = bufseg;
2493 		auio.uio_td = td;
2494 		auio.uio_resid = count;
2495 		error = VOP_READLINK(vp, &auio, td->td_ucred);
2496 		td->td_retval[0] = count - auio.uio_resid;
2497 	}
2498 	vput(vp);
2499 	return (error);
2500 }
2501 
2502 /*
2503  * Common implementation code for chflags() and fchflags().
2504  */
2505 static int
2506 setfflags(struct thread *td, struct vnode *vp, u_long flags)
2507 {
2508 	struct mount *mp;
2509 	struct vattr vattr;
2510 	int error;
2511 
2512 	/* We can't support the value matching VNOVAL. */
2513 	if (flags == VNOVAL)
2514 		return (EOPNOTSUPP);
2515 
2516 	/*
2517 	 * Prevent non-root users from setting flags on devices.  When
2518 	 * a device is reused, users can retain ownership of the device
2519 	 * if they are allowed to set flags and programs assume that
2520 	 * chown can't fail when done as root.
2521 	 */
2522 	if (vp->v_type == VCHR || vp->v_type == VBLK) {
2523 		error = priv_check(td, PRIV_VFS_CHFLAGS_DEV);
2524 		if (error != 0)
2525 			return (error);
2526 	}
2527 
2528 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2529 		return (error);
2530 	VATTR_NULL(&vattr);
2531 	vattr.va_flags = flags;
2532 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2533 #ifdef MAC
2534 	error = mac_vnode_check_setflags(td->td_ucred, vp, vattr.va_flags);
2535 	if (error == 0)
2536 #endif
2537 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
2538 	VOP_UNLOCK(vp, 0);
2539 	vn_finished_write(mp);
2540 	return (error);
2541 }
2542 
2543 /*
2544  * Change flags of a file given a path name.
2545  */
2546 #ifndef _SYS_SYSPROTO_H_
2547 struct chflags_args {
2548 	const char *path;
2549 	u_long	flags;
2550 };
2551 #endif
2552 int
2553 sys_chflags(struct thread *td, struct chflags_args *uap)
2554 {
2555 
2556 	return (kern_chflagsat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2557 	    uap->flags, 0));
2558 }
2559 
2560 #ifndef _SYS_SYSPROTO_H_
2561 struct chflagsat_args {
2562 	int	fd;
2563 	const char *path;
2564 	u_long	flags;
2565 	int	atflag;
2566 }
2567 #endif
2568 int
2569 sys_chflagsat(struct thread *td, struct chflagsat_args *uap)
2570 {
2571 	int fd = uap->fd;
2572 	const char *path = uap->path;
2573 	u_long flags = uap->flags;
2574 	int atflag = uap->atflag;
2575 
2576 	if (atflag & ~AT_SYMLINK_NOFOLLOW)
2577 		return (EINVAL);
2578 
2579 	return (kern_chflagsat(td, fd, path, UIO_USERSPACE, flags, atflag));
2580 }
2581 
2582 /*
2583  * Same as chflags() but doesn't follow symlinks.
2584  */
2585 #ifndef _SYS_SYSPROTO_H_
2586 struct lchflags_args {
2587 	const char *path;
2588 	u_long flags;
2589 };
2590 #endif
2591 int
2592 sys_lchflags(struct thread *td, struct lchflags_args *uap)
2593 {
2594 
2595 	return (kern_chflagsat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2596 	    uap->flags, AT_SYMLINK_NOFOLLOW));
2597 }
2598 
2599 static int
2600 kern_chflagsat(struct thread *td, int fd, const char *path,
2601     enum uio_seg pathseg, u_long flags, int atflag)
2602 {
2603 	struct nameidata nd;
2604 	int error, follow;
2605 
2606 	AUDIT_ARG_FFLAGS(flags);
2607 	follow = (atflag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
2608 	NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, pathseg, path, fd,
2609 	    &cap_fchflags_rights, td);
2610 	if ((error = namei(&nd)) != 0)
2611 		return (error);
2612 	NDFREE(&nd, NDF_ONLY_PNBUF);
2613 	error = setfflags(td, nd.ni_vp, flags);
2614 	vrele(nd.ni_vp);
2615 	return (error);
2616 }
2617 
2618 /*
2619  * Change flags of a file given a file descriptor.
2620  */
2621 #ifndef _SYS_SYSPROTO_H_
2622 struct fchflags_args {
2623 	int	fd;
2624 	u_long	flags;
2625 };
2626 #endif
2627 int
2628 sys_fchflags(struct thread *td, struct fchflags_args *uap)
2629 {
2630 	struct file *fp;
2631 	int error;
2632 
2633 	AUDIT_ARG_FD(uap->fd);
2634 	AUDIT_ARG_FFLAGS(uap->flags);
2635 	error = getvnode(td, uap->fd, &cap_fchflags_rights,
2636 	    &fp);
2637 	if (error != 0)
2638 		return (error);
2639 #ifdef AUDIT
2640 	vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
2641 	AUDIT_ARG_VNODE1(fp->f_vnode);
2642 	VOP_UNLOCK(fp->f_vnode, 0);
2643 #endif
2644 	error = setfflags(td, fp->f_vnode, uap->flags);
2645 	fdrop(fp, td);
2646 	return (error);
2647 }
2648 
2649 /*
2650  * Common implementation code for chmod(), lchmod() and fchmod().
2651  */
2652 int
2653 setfmode(struct thread *td, struct ucred *cred, struct vnode *vp, int mode)
2654 {
2655 	struct mount *mp;
2656 	struct vattr vattr;
2657 	int error;
2658 
2659 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2660 		return (error);
2661 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2662 	VATTR_NULL(&vattr);
2663 	vattr.va_mode = mode & ALLPERMS;
2664 #ifdef MAC
2665 	error = mac_vnode_check_setmode(cred, vp, vattr.va_mode);
2666 	if (error == 0)
2667 #endif
2668 		error = VOP_SETATTR(vp, &vattr, cred);
2669 	VOP_UNLOCK(vp, 0);
2670 	vn_finished_write(mp);
2671 	return (error);
2672 }
2673 
2674 /*
2675  * Change mode of a file given path name.
2676  */
2677 #ifndef _SYS_SYSPROTO_H_
2678 struct chmod_args {
2679 	char	*path;
2680 	int	mode;
2681 };
2682 #endif
2683 int
2684 sys_chmod(struct thread *td, struct chmod_args *uap)
2685 {
2686 
2687 	return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2688 	    uap->mode, 0));
2689 }
2690 
2691 #ifndef _SYS_SYSPROTO_H_
2692 struct fchmodat_args {
2693 	int	dirfd;
2694 	char	*path;
2695 	mode_t	mode;
2696 	int	flag;
2697 }
2698 #endif
2699 int
2700 sys_fchmodat(struct thread *td, struct fchmodat_args *uap)
2701 {
2702 	int flag = uap->flag;
2703 	int fd = uap->fd;
2704 	char *path = uap->path;
2705 	mode_t mode = uap->mode;
2706 
2707 	if (flag & ~AT_SYMLINK_NOFOLLOW)
2708 		return (EINVAL);
2709 
2710 	return (kern_fchmodat(td, fd, path, UIO_USERSPACE, mode, flag));
2711 }
2712 
2713 /*
2714  * Change mode of a file given path name (don't follow links.)
2715  */
2716 #ifndef _SYS_SYSPROTO_H_
2717 struct lchmod_args {
2718 	char	*path;
2719 	int	mode;
2720 };
2721 #endif
2722 int
2723 sys_lchmod(struct thread *td, struct lchmod_args *uap)
2724 {
2725 
2726 	return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2727 	    uap->mode, AT_SYMLINK_NOFOLLOW));
2728 }
2729 
2730 int
2731 kern_fchmodat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2732     mode_t mode, int flag)
2733 {
2734 	struct nameidata nd;
2735 	int error, follow;
2736 
2737 	AUDIT_ARG_MODE(mode);
2738 	follow = (flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
2739 	NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, pathseg, path, fd,
2740 	    &cap_fchmod_rights, td);
2741 	if ((error = namei(&nd)) != 0)
2742 		return (error);
2743 	NDFREE(&nd, NDF_ONLY_PNBUF);
2744 	error = setfmode(td, td->td_ucred, nd.ni_vp, mode);
2745 	vrele(nd.ni_vp);
2746 	return (error);
2747 }
2748 
2749 /*
2750  * Change mode of a file given a file descriptor.
2751  */
2752 #ifndef _SYS_SYSPROTO_H_
2753 struct fchmod_args {
2754 	int	fd;
2755 	int	mode;
2756 };
2757 #endif
2758 int
2759 sys_fchmod(struct thread *td, struct fchmod_args *uap)
2760 {
2761 	struct file *fp;
2762 	int error;
2763 
2764 	AUDIT_ARG_FD(uap->fd);
2765 	AUDIT_ARG_MODE(uap->mode);
2766 
2767 	error = fget(td, uap->fd, &cap_fchmod_rights, &fp);
2768 	if (error != 0)
2769 		return (error);
2770 	error = fo_chmod(fp, uap->mode, td->td_ucred, td);
2771 	fdrop(fp, td);
2772 	return (error);
2773 }
2774 
2775 /*
2776  * Common implementation for chown(), lchown(), and fchown()
2777  */
2778 int
2779 setfown(struct thread *td, struct ucred *cred, struct vnode *vp, uid_t uid,
2780     gid_t gid)
2781 {
2782 	struct mount *mp;
2783 	struct vattr vattr;
2784 	int error;
2785 
2786 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2787 		return (error);
2788 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2789 	VATTR_NULL(&vattr);
2790 	vattr.va_uid = uid;
2791 	vattr.va_gid = gid;
2792 #ifdef MAC
2793 	error = mac_vnode_check_setowner(cred, vp, vattr.va_uid,
2794 	    vattr.va_gid);
2795 	if (error == 0)
2796 #endif
2797 		error = VOP_SETATTR(vp, &vattr, cred);
2798 	VOP_UNLOCK(vp, 0);
2799 	vn_finished_write(mp);
2800 	return (error);
2801 }
2802 
2803 /*
2804  * Set ownership given a path name.
2805  */
2806 #ifndef _SYS_SYSPROTO_H_
2807 struct chown_args {
2808 	char	*path;
2809 	int	uid;
2810 	int	gid;
2811 };
2812 #endif
2813 int
2814 sys_chown(struct thread *td, struct chown_args *uap)
2815 {
2816 
2817 	return (kern_fchownat(td, AT_FDCWD, uap->path, UIO_USERSPACE, uap->uid,
2818 	    uap->gid, 0));
2819 }
2820 
2821 #ifndef _SYS_SYSPROTO_H_
2822 struct fchownat_args {
2823 	int fd;
2824 	const char * path;
2825 	uid_t uid;
2826 	gid_t gid;
2827 	int flag;
2828 };
2829 #endif
2830 int
2831 sys_fchownat(struct thread *td, struct fchownat_args *uap)
2832 {
2833 	int flag;
2834 
2835 	flag = uap->flag;
2836 	if (flag & ~AT_SYMLINK_NOFOLLOW)
2837 		return (EINVAL);
2838 
2839 	return (kern_fchownat(td, uap->fd, uap->path, UIO_USERSPACE, uap->uid,
2840 	    uap->gid, uap->flag));
2841 }
2842 
2843 int
2844 kern_fchownat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2845     int uid, int gid, int flag)
2846 {
2847 	struct nameidata nd;
2848 	int error, follow;
2849 
2850 	AUDIT_ARG_OWNER(uid, gid);
2851 	follow = (flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
2852 	NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, pathseg, path, fd,
2853 	    &cap_fchown_rights, td);
2854 
2855 	if ((error = namei(&nd)) != 0)
2856 		return (error);
2857 	NDFREE(&nd, NDF_ONLY_PNBUF);
2858 	error = setfown(td, td->td_ucred, nd.ni_vp, uid, gid);
2859 	vrele(nd.ni_vp);
2860 	return (error);
2861 }
2862 
2863 /*
2864  * Set ownership given a path name, do not cross symlinks.
2865  */
2866 #ifndef _SYS_SYSPROTO_H_
2867 struct lchown_args {
2868 	char	*path;
2869 	int	uid;
2870 	int	gid;
2871 };
2872 #endif
2873 int
2874 sys_lchown(struct thread *td, struct lchown_args *uap)
2875 {
2876 
2877 	return (kern_fchownat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2878 	    uap->uid, uap->gid, AT_SYMLINK_NOFOLLOW));
2879 }
2880 
2881 /*
2882  * Set ownership given a file descriptor.
2883  */
2884 #ifndef _SYS_SYSPROTO_H_
2885 struct fchown_args {
2886 	int	fd;
2887 	int	uid;
2888 	int	gid;
2889 };
2890 #endif
2891 int
2892 sys_fchown(struct thread *td, struct fchown_args *uap)
2893 {
2894 	struct file *fp;
2895 	int error;
2896 
2897 	AUDIT_ARG_FD(uap->fd);
2898 	AUDIT_ARG_OWNER(uap->uid, uap->gid);
2899 	error = fget(td, uap->fd, &cap_fchown_rights, &fp);
2900 	if (error != 0)
2901 		return (error);
2902 	error = fo_chown(fp, uap->uid, uap->gid, td->td_ucred, td);
2903 	fdrop(fp, td);
2904 	return (error);
2905 }
2906 
2907 /*
2908  * Common implementation code for utimes(), lutimes(), and futimes().
2909  */
2910 static int
2911 getutimes(const struct timeval *usrtvp, enum uio_seg tvpseg,
2912     struct timespec *tsp)
2913 {
2914 	struct timeval tv[2];
2915 	const struct timeval *tvp;
2916 	int error;
2917 
2918 	if (usrtvp == NULL) {
2919 		vfs_timestamp(&tsp[0]);
2920 		tsp[1] = tsp[0];
2921 	} else {
2922 		if (tvpseg == UIO_SYSSPACE) {
2923 			tvp = usrtvp;
2924 		} else {
2925 			if ((error = copyin(usrtvp, tv, sizeof(tv))) != 0)
2926 				return (error);
2927 			tvp = tv;
2928 		}
2929 
2930 		if (tvp[0].tv_usec < 0 || tvp[0].tv_usec >= 1000000 ||
2931 		    tvp[1].tv_usec < 0 || tvp[1].tv_usec >= 1000000)
2932 			return (EINVAL);
2933 		TIMEVAL_TO_TIMESPEC(&tvp[0], &tsp[0]);
2934 		TIMEVAL_TO_TIMESPEC(&tvp[1], &tsp[1]);
2935 	}
2936 	return (0);
2937 }
2938 
2939 /*
2940  * Common implementation code for futimens(), utimensat().
2941  */
2942 #define	UTIMENS_NULL	0x1
2943 #define	UTIMENS_EXIT	0x2
2944 static int
2945 getutimens(const struct timespec *usrtsp, enum uio_seg tspseg,
2946     struct timespec *tsp, int *retflags)
2947 {
2948 	struct timespec tsnow;
2949 	int error;
2950 
2951 	vfs_timestamp(&tsnow);
2952 	*retflags = 0;
2953 	if (usrtsp == NULL) {
2954 		tsp[0] = tsnow;
2955 		tsp[1] = tsnow;
2956 		*retflags |= UTIMENS_NULL;
2957 		return (0);
2958 	}
2959 	if (tspseg == UIO_SYSSPACE) {
2960 		tsp[0] = usrtsp[0];
2961 		tsp[1] = usrtsp[1];
2962 	} else if ((error = copyin(usrtsp, tsp, sizeof(*tsp) * 2)) != 0)
2963 		return (error);
2964 	if (tsp[0].tv_nsec == UTIME_OMIT && tsp[1].tv_nsec == UTIME_OMIT)
2965 		*retflags |= UTIMENS_EXIT;
2966 	if (tsp[0].tv_nsec == UTIME_NOW && tsp[1].tv_nsec == UTIME_NOW)
2967 		*retflags |= UTIMENS_NULL;
2968 	if (tsp[0].tv_nsec == UTIME_OMIT)
2969 		tsp[0].tv_sec = VNOVAL;
2970 	else if (tsp[0].tv_nsec == UTIME_NOW)
2971 		tsp[0] = tsnow;
2972 	else if (tsp[0].tv_nsec < 0 || tsp[0].tv_nsec >= 1000000000L)
2973 		return (EINVAL);
2974 	if (tsp[1].tv_nsec == UTIME_OMIT)
2975 		tsp[1].tv_sec = VNOVAL;
2976 	else if (tsp[1].tv_nsec == UTIME_NOW)
2977 		tsp[1] = tsnow;
2978 	else if (tsp[1].tv_nsec < 0 || tsp[1].tv_nsec >= 1000000000L)
2979 		return (EINVAL);
2980 
2981 	return (0);
2982 }
2983 
2984 /*
2985  * Common implementation code for utimes(), lutimes(), futimes(), futimens(),
2986  * and utimensat().
2987  */
2988 static int
2989 setutimes(struct thread *td, struct vnode *vp, const struct timespec *ts,
2990     int numtimes, int nullflag)
2991 {
2992 	struct mount *mp;
2993 	struct vattr vattr;
2994 	int error, setbirthtime;
2995 
2996 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2997 		return (error);
2998 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2999 	setbirthtime = 0;
3000 	if (numtimes < 3 && !VOP_GETATTR(vp, &vattr, td->td_ucred) &&
3001 	    timespeccmp(&ts[1], &vattr.va_birthtime, < ))
3002 		setbirthtime = 1;
3003 	VATTR_NULL(&vattr);
3004 	vattr.va_atime = ts[0];
3005 	vattr.va_mtime = ts[1];
3006 	if (setbirthtime)
3007 		vattr.va_birthtime = ts[1];
3008 	if (numtimes > 2)
3009 		vattr.va_birthtime = ts[2];
3010 	if (nullflag)
3011 		vattr.va_vaflags |= VA_UTIMES_NULL;
3012 #ifdef MAC
3013 	error = mac_vnode_check_setutimes(td->td_ucred, vp, vattr.va_atime,
3014 	    vattr.va_mtime);
3015 #endif
3016 	if (error == 0)
3017 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3018 	VOP_UNLOCK(vp, 0);
3019 	vn_finished_write(mp);
3020 	return (error);
3021 }
3022 
3023 /*
3024  * Set the access and modification times of a file.
3025  */
3026 #ifndef _SYS_SYSPROTO_H_
3027 struct utimes_args {
3028 	char	*path;
3029 	struct	timeval *tptr;
3030 };
3031 #endif
3032 int
3033 sys_utimes(struct thread *td, struct utimes_args *uap)
3034 {
3035 
3036 	return (kern_utimesat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
3037 	    uap->tptr, UIO_USERSPACE));
3038 }
3039 
3040 #ifndef _SYS_SYSPROTO_H_
3041 struct futimesat_args {
3042 	int fd;
3043 	const char * path;
3044 	const struct timeval * times;
3045 };
3046 #endif
3047 int
3048 sys_futimesat(struct thread *td, struct futimesat_args *uap)
3049 {
3050 
3051 	return (kern_utimesat(td, uap->fd, uap->path, UIO_USERSPACE,
3052 	    uap->times, UIO_USERSPACE));
3053 }
3054 
3055 int
3056 kern_utimesat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
3057     struct timeval *tptr, enum uio_seg tptrseg)
3058 {
3059 	struct nameidata nd;
3060 	struct timespec ts[2];
3061 	int error;
3062 
3063 	if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3064 		return (error);
3065 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, fd,
3066 	    &cap_futimes_rights, td);
3067 
3068 	if ((error = namei(&nd)) != 0)
3069 		return (error);
3070 	NDFREE(&nd, NDF_ONLY_PNBUF);
3071 	error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3072 	vrele(nd.ni_vp);
3073 	return (error);
3074 }
3075 
3076 /*
3077  * Set the access and modification times of a file.
3078  */
3079 #ifndef _SYS_SYSPROTO_H_
3080 struct lutimes_args {
3081 	char	*path;
3082 	struct	timeval *tptr;
3083 };
3084 #endif
3085 int
3086 sys_lutimes(struct thread *td, struct lutimes_args *uap)
3087 {
3088 
3089 	return (kern_lutimes(td, uap->path, UIO_USERSPACE, uap->tptr,
3090 	    UIO_USERSPACE));
3091 }
3092 
3093 int
3094 kern_lutimes(struct thread *td, char *path, enum uio_seg pathseg,
3095     struct timeval *tptr, enum uio_seg tptrseg)
3096 {
3097 	struct timespec ts[2];
3098 	struct nameidata nd;
3099 	int error;
3100 
3101 	if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3102 		return (error);
3103 	NDINIT(&nd, LOOKUP, NOFOLLOW | AUDITVNODE1, pathseg, path, td);
3104 	if ((error = namei(&nd)) != 0)
3105 		return (error);
3106 	NDFREE(&nd, NDF_ONLY_PNBUF);
3107 	error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3108 	vrele(nd.ni_vp);
3109 	return (error);
3110 }
3111 
3112 /*
3113  * Set the access and modification times of a file.
3114  */
3115 #ifndef _SYS_SYSPROTO_H_
3116 struct futimes_args {
3117 	int	fd;
3118 	struct	timeval *tptr;
3119 };
3120 #endif
3121 int
3122 sys_futimes(struct thread *td, struct futimes_args *uap)
3123 {
3124 
3125 	return (kern_futimes(td, uap->fd, uap->tptr, UIO_USERSPACE));
3126 }
3127 
3128 int
3129 kern_futimes(struct thread *td, int fd, struct timeval *tptr,
3130     enum uio_seg tptrseg)
3131 {
3132 	struct timespec ts[2];
3133 	struct file *fp;
3134 	int error;
3135 
3136 	AUDIT_ARG_FD(fd);
3137 	error = getutimes(tptr, tptrseg, ts);
3138 	if (error != 0)
3139 		return (error);
3140 	error = getvnode(td, fd, &cap_futimes_rights, &fp);
3141 	if (error != 0)
3142 		return (error);
3143 #ifdef AUDIT
3144 	vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
3145 	AUDIT_ARG_VNODE1(fp->f_vnode);
3146 	VOP_UNLOCK(fp->f_vnode, 0);
3147 #endif
3148 	error = setutimes(td, fp->f_vnode, ts, 2, tptr == NULL);
3149 	fdrop(fp, td);
3150 	return (error);
3151 }
3152 
3153 int
3154 sys_futimens(struct thread *td, struct futimens_args *uap)
3155 {
3156 
3157 	return (kern_futimens(td, uap->fd, uap->times, UIO_USERSPACE));
3158 }
3159 
3160 int
3161 kern_futimens(struct thread *td, int fd, struct timespec *tptr,
3162     enum uio_seg tptrseg)
3163 {
3164 	struct timespec ts[2];
3165 	struct file *fp;
3166 	int error, flags;
3167 
3168 	AUDIT_ARG_FD(fd);
3169 	error = getutimens(tptr, tptrseg, ts, &flags);
3170 	if (error != 0)
3171 		return (error);
3172 	if (flags & UTIMENS_EXIT)
3173 		return (0);
3174 	error = getvnode(td, fd, &cap_futimes_rights, &fp);
3175 	if (error != 0)
3176 		return (error);
3177 #ifdef AUDIT
3178 	vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
3179 	AUDIT_ARG_VNODE1(fp->f_vnode);
3180 	VOP_UNLOCK(fp->f_vnode, 0);
3181 #endif
3182 	error = setutimes(td, fp->f_vnode, ts, 2, flags & UTIMENS_NULL);
3183 	fdrop(fp, td);
3184 	return (error);
3185 }
3186 
3187 int
3188 sys_utimensat(struct thread *td, struct utimensat_args *uap)
3189 {
3190 
3191 	return (kern_utimensat(td, uap->fd, uap->path, UIO_USERSPACE,
3192 	    uap->times, UIO_USERSPACE, uap->flag));
3193 }
3194 
3195 int
3196 kern_utimensat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
3197     struct timespec *tptr, enum uio_seg tptrseg, int flag)
3198 {
3199 	struct nameidata nd;
3200 	struct timespec ts[2];
3201 	int error, flags;
3202 
3203 	if (flag & ~AT_SYMLINK_NOFOLLOW)
3204 		return (EINVAL);
3205 
3206 	if ((error = getutimens(tptr, tptrseg, ts, &flags)) != 0)
3207 		return (error);
3208 	NDINIT_ATRIGHTS(&nd, LOOKUP, ((flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW :
3209 	    FOLLOW) | AUDITVNODE1, pathseg, path, fd,
3210 	    &cap_futimes_rights, td);
3211 	if ((error = namei(&nd)) != 0)
3212 		return (error);
3213 	/*
3214 	 * We are allowed to call namei() regardless of 2xUTIME_OMIT.
3215 	 * POSIX states:
3216 	 * "If both tv_nsec fields are UTIME_OMIT... EACCESS may be detected."
3217 	 * "Search permission is denied by a component of the path prefix."
3218 	 */
3219 	NDFREE(&nd, NDF_ONLY_PNBUF);
3220 	if ((flags & UTIMENS_EXIT) == 0)
3221 		error = setutimes(td, nd.ni_vp, ts, 2, flags & UTIMENS_NULL);
3222 	vrele(nd.ni_vp);
3223 	return (error);
3224 }
3225 
3226 /*
3227  * Truncate a file given its path name.
3228  */
3229 #ifndef _SYS_SYSPROTO_H_
3230 struct truncate_args {
3231 	char	*path;
3232 	int	pad;
3233 	off_t	length;
3234 };
3235 #endif
3236 int
3237 sys_truncate(struct thread *td, struct truncate_args *uap)
3238 {
3239 
3240 	return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3241 }
3242 
3243 int
3244 kern_truncate(struct thread *td, char *path, enum uio_seg pathseg, off_t length)
3245 {
3246 	struct mount *mp;
3247 	struct vnode *vp;
3248 	void *rl_cookie;
3249 	struct vattr vattr;
3250 	struct nameidata nd;
3251 	int error;
3252 
3253 	if (length < 0)
3254 		return(EINVAL);
3255 	NDINIT(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, td);
3256 	if ((error = namei(&nd)) != 0)
3257 		return (error);
3258 	vp = nd.ni_vp;
3259 	rl_cookie = vn_rangelock_wlock(vp, 0, OFF_MAX);
3260 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) {
3261 		vn_rangelock_unlock(vp, rl_cookie);
3262 		vrele(vp);
3263 		return (error);
3264 	}
3265 	NDFREE(&nd, NDF_ONLY_PNBUF);
3266 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3267 	if (vp->v_type == VDIR)
3268 		error = EISDIR;
3269 #ifdef MAC
3270 	else if ((error = mac_vnode_check_write(td->td_ucred, NOCRED, vp))) {
3271 	}
3272 #endif
3273 	else if ((error = vn_writechk(vp)) == 0 &&
3274 	    (error = VOP_ACCESS(vp, VWRITE, td->td_ucred, td)) == 0) {
3275 		VATTR_NULL(&vattr);
3276 		vattr.va_size = length;
3277 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3278 	}
3279 	VOP_UNLOCK(vp, 0);
3280 	vn_finished_write(mp);
3281 	vn_rangelock_unlock(vp, rl_cookie);
3282 	vrele(vp);
3283 	return (error);
3284 }
3285 
3286 #if defined(COMPAT_43)
3287 /*
3288  * Truncate a file given its path name.
3289  */
3290 #ifndef _SYS_SYSPROTO_H_
3291 struct otruncate_args {
3292 	char	*path;
3293 	long	length;
3294 };
3295 #endif
3296 int
3297 otruncate(struct thread *td, struct otruncate_args *uap)
3298 {
3299 
3300 	return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3301 }
3302 #endif /* COMPAT_43 */
3303 
3304 #if defined(COMPAT_FREEBSD6)
3305 /* Versions with the pad argument */
3306 int
3307 freebsd6_truncate(struct thread *td, struct freebsd6_truncate_args *uap)
3308 {
3309 
3310 	return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3311 }
3312 
3313 int
3314 freebsd6_ftruncate(struct thread *td, struct freebsd6_ftruncate_args *uap)
3315 {
3316 
3317 	return (kern_ftruncate(td, uap->fd, uap->length));
3318 }
3319 #endif
3320 
3321 int
3322 kern_fsync(struct thread *td, int fd, bool fullsync)
3323 {
3324 	struct vnode *vp;
3325 	struct mount *mp;
3326 	struct file *fp;
3327 	int error, lock_flags;
3328 
3329 	AUDIT_ARG_FD(fd);
3330 	error = getvnode(td, fd, &cap_fsync_rights, &fp);
3331 	if (error != 0)
3332 		return (error);
3333 	vp = fp->f_vnode;
3334 #if 0
3335 	if (!fullsync)
3336 		/* XXXKIB: compete outstanding aio writes */;
3337 #endif
3338 	error = vn_start_write(vp, &mp, V_WAIT | PCATCH);
3339 	if (error != 0)
3340 		goto drop;
3341 	if (MNT_SHARED_WRITES(mp) ||
3342 	    ((mp == NULL) && MNT_SHARED_WRITES(vp->v_mount))) {
3343 		lock_flags = LK_SHARED;
3344 	} else {
3345 		lock_flags = LK_EXCLUSIVE;
3346 	}
3347 	vn_lock(vp, lock_flags | LK_RETRY);
3348 	AUDIT_ARG_VNODE1(vp);
3349 	if (vp->v_object != NULL) {
3350 		VM_OBJECT_WLOCK(vp->v_object);
3351 		vm_object_page_clean(vp->v_object, 0, 0, 0);
3352 		VM_OBJECT_WUNLOCK(vp->v_object);
3353 	}
3354 	error = fullsync ? VOP_FSYNC(vp, MNT_WAIT, td) : VOP_FDATASYNC(vp, td);
3355 	VOP_UNLOCK(vp, 0);
3356 	vn_finished_write(mp);
3357 drop:
3358 	fdrop(fp, td);
3359 	return (error);
3360 }
3361 
3362 /*
3363  * Sync an open file.
3364  */
3365 #ifndef _SYS_SYSPROTO_H_
3366 struct fsync_args {
3367 	int	fd;
3368 };
3369 #endif
3370 int
3371 sys_fsync(struct thread *td, struct fsync_args *uap)
3372 {
3373 
3374 	return (kern_fsync(td, uap->fd, true));
3375 }
3376 
3377 int
3378 sys_fdatasync(struct thread *td, struct fdatasync_args *uap)
3379 {
3380 
3381 	return (kern_fsync(td, uap->fd, false));
3382 }
3383 
3384 /*
3385  * Rename files.  Source and destination must either both be directories, or
3386  * both not be directories.  If target is a directory, it must be empty.
3387  */
3388 #ifndef _SYS_SYSPROTO_H_
3389 struct rename_args {
3390 	char	*from;
3391 	char	*to;
3392 };
3393 #endif
3394 int
3395 sys_rename(struct thread *td, struct rename_args *uap)
3396 {
3397 
3398 	return (kern_renameat(td, AT_FDCWD, uap->from, AT_FDCWD,
3399 	    uap->to, UIO_USERSPACE));
3400 }
3401 
3402 #ifndef _SYS_SYSPROTO_H_
3403 struct renameat_args {
3404 	int	oldfd;
3405 	char	*old;
3406 	int	newfd;
3407 	char	*new;
3408 };
3409 #endif
3410 int
3411 sys_renameat(struct thread *td, struct renameat_args *uap)
3412 {
3413 
3414 	return (kern_renameat(td, uap->oldfd, uap->old, uap->newfd, uap->new,
3415 	    UIO_USERSPACE));
3416 }
3417 
3418 int
3419 kern_renameat(struct thread *td, int oldfd, char *old, int newfd, char *new,
3420     enum uio_seg pathseg)
3421 {
3422 	struct mount *mp = NULL;
3423 	struct vnode *tvp, *fvp, *tdvp;
3424 	struct nameidata fromnd, tond;
3425 	int error;
3426 
3427 again:
3428 	bwillwrite();
3429 #ifdef MAC
3430 	NDINIT_ATRIGHTS(&fromnd, DELETE, LOCKPARENT | LOCKLEAF | SAVESTART |
3431 	    AUDITVNODE1, pathseg, old, oldfd,
3432 	    &cap_renameat_source_rights, td);
3433 #else
3434 	NDINIT_ATRIGHTS(&fromnd, DELETE, WANTPARENT | SAVESTART | AUDITVNODE1,
3435 	    pathseg, old, oldfd,
3436 	    &cap_renameat_source_rights, td);
3437 #endif
3438 
3439 	if ((error = namei(&fromnd)) != 0)
3440 		return (error);
3441 #ifdef MAC
3442 	error = mac_vnode_check_rename_from(td->td_ucred, fromnd.ni_dvp,
3443 	    fromnd.ni_vp, &fromnd.ni_cnd);
3444 	VOP_UNLOCK(fromnd.ni_dvp, 0);
3445 	if (fromnd.ni_dvp != fromnd.ni_vp)
3446 		VOP_UNLOCK(fromnd.ni_vp, 0);
3447 #endif
3448 	fvp = fromnd.ni_vp;
3449 	NDINIT_ATRIGHTS(&tond, RENAME, LOCKPARENT | LOCKLEAF | NOCACHE |
3450 	    SAVESTART | AUDITVNODE2, pathseg, new, newfd,
3451 	    &cap_renameat_target_rights, td);
3452 	if (fromnd.ni_vp->v_type == VDIR)
3453 		tond.ni_cnd.cn_flags |= WILLBEDIR;
3454 	if ((error = namei(&tond)) != 0) {
3455 		/* Translate error code for rename("dir1", "dir2/."). */
3456 		if (error == EISDIR && fvp->v_type == VDIR)
3457 			error = EINVAL;
3458 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3459 		vrele(fromnd.ni_dvp);
3460 		vrele(fvp);
3461 		goto out1;
3462 	}
3463 	tdvp = tond.ni_dvp;
3464 	tvp = tond.ni_vp;
3465 	error = vn_start_write(fvp, &mp, V_NOWAIT);
3466 	if (error != 0) {
3467 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3468 		NDFREE(&tond, NDF_ONLY_PNBUF);
3469 		if (tvp != NULL)
3470 			vput(tvp);
3471 		if (tdvp == tvp)
3472 			vrele(tdvp);
3473 		else
3474 			vput(tdvp);
3475 		vrele(fromnd.ni_dvp);
3476 		vrele(fvp);
3477 		vrele(tond.ni_startdir);
3478 		if (fromnd.ni_startdir != NULL)
3479 			vrele(fromnd.ni_startdir);
3480 		error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH);
3481 		if (error != 0)
3482 			return (error);
3483 		goto again;
3484 	}
3485 	if (tvp != NULL) {
3486 		if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
3487 			error = ENOTDIR;
3488 			goto out;
3489 		} else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
3490 			error = EISDIR;
3491 			goto out;
3492 		}
3493 #ifdef CAPABILITIES
3494 		if (newfd != AT_FDCWD) {
3495 			/*
3496 			 * If the target already exists we require CAP_UNLINKAT
3497 			 * from 'newfd'.
3498 			 */
3499 			error = cap_check(&tond.ni_filecaps.fc_rights,
3500 			    &cap_unlinkat_rights);
3501 			if (error != 0)
3502 				goto out;
3503 		}
3504 #endif
3505 	}
3506 	if (fvp == tdvp) {
3507 		error = EINVAL;
3508 		goto out;
3509 	}
3510 	/*
3511 	 * If the source is the same as the destination (that is, if they
3512 	 * are links to the same vnode), then there is nothing to do.
3513 	 */
3514 	if (fvp == tvp)
3515 		error = -1;
3516 #ifdef MAC
3517 	else
3518 		error = mac_vnode_check_rename_to(td->td_ucred, tdvp,
3519 		    tond.ni_vp, fromnd.ni_dvp == tdvp, &tond.ni_cnd);
3520 #endif
3521 out:
3522 	if (error == 0) {
3523 		error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
3524 		    tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
3525 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3526 		NDFREE(&tond, NDF_ONLY_PNBUF);
3527 	} else {
3528 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3529 		NDFREE(&tond, NDF_ONLY_PNBUF);
3530 		if (tvp != NULL)
3531 			vput(tvp);
3532 		if (tdvp == tvp)
3533 			vrele(tdvp);
3534 		else
3535 			vput(tdvp);
3536 		vrele(fromnd.ni_dvp);
3537 		vrele(fvp);
3538 	}
3539 	vrele(tond.ni_startdir);
3540 	vn_finished_write(mp);
3541 out1:
3542 	if (fromnd.ni_startdir)
3543 		vrele(fromnd.ni_startdir);
3544 	if (error == -1)
3545 		return (0);
3546 	return (error);
3547 }
3548 
3549 /*
3550  * Make a directory file.
3551  */
3552 #ifndef _SYS_SYSPROTO_H_
3553 struct mkdir_args {
3554 	char	*path;
3555 	int	mode;
3556 };
3557 #endif
3558 int
3559 sys_mkdir(struct thread *td, struct mkdir_args *uap)
3560 {
3561 
3562 	return (kern_mkdirat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
3563 	    uap->mode));
3564 }
3565 
3566 #ifndef _SYS_SYSPROTO_H_
3567 struct mkdirat_args {
3568 	int	fd;
3569 	char	*path;
3570 	mode_t	mode;
3571 };
3572 #endif
3573 int
3574 sys_mkdirat(struct thread *td, struct mkdirat_args *uap)
3575 {
3576 
3577 	return (kern_mkdirat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode));
3578 }
3579 
3580 int
3581 kern_mkdirat(struct thread *td, int fd, char *path, enum uio_seg segflg,
3582     int mode)
3583 {
3584 	struct mount *mp;
3585 	struct vnode *vp;
3586 	struct vattr vattr;
3587 	struct nameidata nd;
3588 	int error;
3589 
3590 	AUDIT_ARG_MODE(mode);
3591 restart:
3592 	bwillwrite();
3593 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
3594 	    NOCACHE, segflg, path, fd, &cap_mkdirat_rights,
3595 	    td);
3596 	nd.ni_cnd.cn_flags |= WILLBEDIR;
3597 	if ((error = namei(&nd)) != 0)
3598 		return (error);
3599 	vp = nd.ni_vp;
3600 	if (vp != NULL) {
3601 		NDFREE(&nd, NDF_ONLY_PNBUF);
3602 		/*
3603 		 * XXX namei called with LOCKPARENT but not LOCKLEAF has
3604 		 * the strange behaviour of leaving the vnode unlocked
3605 		 * if the target is the same vnode as the parent.
3606 		 */
3607 		if (vp == nd.ni_dvp)
3608 			vrele(nd.ni_dvp);
3609 		else
3610 			vput(nd.ni_dvp);
3611 		vrele(vp);
3612 		return (EEXIST);
3613 	}
3614 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3615 		NDFREE(&nd, NDF_ONLY_PNBUF);
3616 		vput(nd.ni_dvp);
3617 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3618 			return (error);
3619 		goto restart;
3620 	}
3621 	VATTR_NULL(&vattr);
3622 	vattr.va_type = VDIR;
3623 	vattr.va_mode = (mode & ACCESSPERMS) &~ td->td_proc->p_fd->fd_cmask;
3624 #ifdef MAC
3625 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
3626 	    &vattr);
3627 	if (error != 0)
3628 		goto out;
3629 #endif
3630 	error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
3631 #ifdef MAC
3632 out:
3633 #endif
3634 	NDFREE(&nd, NDF_ONLY_PNBUF);
3635 	vput(nd.ni_dvp);
3636 	if (error == 0)
3637 		vput(nd.ni_vp);
3638 	vn_finished_write(mp);
3639 	return (error);
3640 }
3641 
3642 /*
3643  * Remove a directory file.
3644  */
3645 #ifndef _SYS_SYSPROTO_H_
3646 struct rmdir_args {
3647 	char	*path;
3648 };
3649 #endif
3650 int
3651 sys_rmdir(struct thread *td, struct rmdir_args *uap)
3652 {
3653 
3654 	return (kern_rmdirat(td, AT_FDCWD, uap->path, UIO_USERSPACE));
3655 }
3656 
3657 int
3658 kern_rmdirat(struct thread *td, int fd, char *path, enum uio_seg pathseg)
3659 {
3660 	struct mount *mp;
3661 	struct vnode *vp;
3662 	struct nameidata nd;
3663 	int error;
3664 
3665 restart:
3666 	bwillwrite();
3667 	NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1,
3668 	    pathseg, path, fd, &cap_unlinkat_rights, td);
3669 	if ((error = namei(&nd)) != 0)
3670 		return (error);
3671 	vp = nd.ni_vp;
3672 	if (vp->v_type != VDIR) {
3673 		error = ENOTDIR;
3674 		goto out;
3675 	}
3676 	/*
3677 	 * No rmdir "." please.
3678 	 */
3679 	if (nd.ni_dvp == vp) {
3680 		error = EINVAL;
3681 		goto out;
3682 	}
3683 	/*
3684 	 * The root of a mounted filesystem cannot be deleted.
3685 	 */
3686 	if (vp->v_vflag & VV_ROOT) {
3687 		error = EBUSY;
3688 		goto out;
3689 	}
3690 #ifdef MAC
3691 	error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
3692 	    &nd.ni_cnd);
3693 	if (error != 0)
3694 		goto out;
3695 #endif
3696 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3697 		NDFREE(&nd, NDF_ONLY_PNBUF);
3698 		vput(vp);
3699 		if (nd.ni_dvp == vp)
3700 			vrele(nd.ni_dvp);
3701 		else
3702 			vput(nd.ni_dvp);
3703 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3704 			return (error);
3705 		goto restart;
3706 	}
3707 	vfs_notify_upper(vp, VFS_NOTIFY_UPPER_UNLINK);
3708 	error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
3709 	vn_finished_write(mp);
3710 out:
3711 	NDFREE(&nd, NDF_ONLY_PNBUF);
3712 	vput(vp);
3713 	if (nd.ni_dvp == vp)
3714 		vrele(nd.ni_dvp);
3715 	else
3716 		vput(nd.ni_dvp);
3717 	return (error);
3718 }
3719 
3720 #if defined(COMPAT_43) || defined(COMPAT_FREEBSD11)
3721 int
3722 freebsd11_kern_getdirentries(struct thread *td, int fd, char *ubuf, u_int count,
3723     long *basep, void (*func)(struct freebsd11_dirent *))
3724 {
3725 	struct freebsd11_dirent dstdp;
3726 	struct dirent *dp, *edp;
3727 	char *dirbuf;
3728 	off_t base;
3729 	ssize_t resid, ucount;
3730 	int error;
3731 
3732 	/* XXX arbitrary sanity limit on `count'. */
3733 	count = min(count, 64 * 1024);
3734 
3735 	dirbuf = malloc(count, M_TEMP, M_WAITOK);
3736 
3737 	error = kern_getdirentries(td, fd, dirbuf, count, &base, &resid,
3738 	    UIO_SYSSPACE);
3739 	if (error != 0)
3740 		goto done;
3741 	if (basep != NULL)
3742 		*basep = base;
3743 
3744 	ucount = 0;
3745 	for (dp = (struct dirent *)dirbuf,
3746 	    edp = (struct dirent *)&dirbuf[count - resid];
3747 	    ucount < count && dp < edp; ) {
3748 		if (dp->d_reclen == 0)
3749 			break;
3750 		MPASS(dp->d_reclen >= _GENERIC_DIRLEN(0));
3751 		if (dp->d_namlen >= sizeof(dstdp.d_name))
3752 			continue;
3753 		dstdp.d_type = dp->d_type;
3754 		dstdp.d_namlen = dp->d_namlen;
3755 		dstdp.d_fileno = dp->d_fileno;		/* truncate */
3756 		if (dstdp.d_fileno != dp->d_fileno) {
3757 			switch (ino64_trunc_error) {
3758 			default:
3759 			case 0:
3760 				break;
3761 			case 1:
3762 				error = EOVERFLOW;
3763 				goto done;
3764 			case 2:
3765 				dstdp.d_fileno = UINT32_MAX;
3766 				break;
3767 			}
3768 		}
3769 		dstdp.d_reclen = sizeof(dstdp) - sizeof(dstdp.d_name) +
3770 		    ((dp->d_namlen + 1 + 3) &~ 3);
3771 		bcopy(dp->d_name, dstdp.d_name, dstdp.d_namlen);
3772 		bzero(dstdp.d_name + dstdp.d_namlen,
3773 		    dstdp.d_reclen - offsetof(struct freebsd11_dirent, d_name) -
3774 		    dstdp.d_namlen);
3775 		MPASS(dstdp.d_reclen <= dp->d_reclen);
3776 		MPASS(ucount + dstdp.d_reclen <= count);
3777 		if (func != NULL)
3778 			func(&dstdp);
3779 		error = copyout(&dstdp, ubuf + ucount, dstdp.d_reclen);
3780 		if (error != 0)
3781 			break;
3782 		dp = (struct dirent *)((char *)dp + dp->d_reclen);
3783 		ucount += dstdp.d_reclen;
3784 	}
3785 
3786 done:
3787 	free(dirbuf, M_TEMP);
3788 	if (error == 0)
3789 		td->td_retval[0] = ucount;
3790 	return (error);
3791 }
3792 #endif /* COMPAT */
3793 
3794 #ifdef COMPAT_43
3795 static void
3796 ogetdirentries_cvt(struct freebsd11_dirent *dp)
3797 {
3798 #if (BYTE_ORDER == LITTLE_ENDIAN)
3799 	/*
3800 	 * The expected low byte of dp->d_namlen is our dp->d_type.
3801 	 * The high MBZ byte of dp->d_namlen is our dp->d_namlen.
3802 	 */
3803 	dp->d_type = dp->d_namlen;
3804 	dp->d_namlen = 0;
3805 #else
3806 	/*
3807 	 * The dp->d_type is the high byte of the expected dp->d_namlen,
3808 	 * so must be zero'ed.
3809 	 */
3810 	dp->d_type = 0;
3811 #endif
3812 }
3813 
3814 /*
3815  * Read a block of directory entries in a filesystem independent format.
3816  */
3817 #ifndef _SYS_SYSPROTO_H_
3818 struct ogetdirentries_args {
3819 	int	fd;
3820 	char	*buf;
3821 	u_int	count;
3822 	long	*basep;
3823 };
3824 #endif
3825 int
3826 ogetdirentries(struct thread *td, struct ogetdirentries_args *uap)
3827 {
3828 	long loff;
3829 	int error;
3830 
3831 	error = kern_ogetdirentries(td, uap, &loff);
3832 	if (error == 0)
3833 		error = copyout(&loff, uap->basep, sizeof(long));
3834 	return (error);
3835 }
3836 
3837 int
3838 kern_ogetdirentries(struct thread *td, struct ogetdirentries_args *uap,
3839     long *ploff)
3840 {
3841 	long base;
3842 	int error;
3843 
3844 	/* XXX arbitrary sanity limit on `count'. */
3845 	if (uap->count > 64 * 1024)
3846 		return (EINVAL);
3847 
3848 	error = freebsd11_kern_getdirentries(td, uap->fd, uap->buf, uap->count,
3849 	    &base, ogetdirentries_cvt);
3850 
3851 	if (error == 0 && uap->basep != NULL)
3852 		error = copyout(&base, uap->basep, sizeof(long));
3853 
3854 	return (error);
3855 }
3856 #endif /* COMPAT_43 */
3857 
3858 #if defined(COMPAT_FREEBSD11)
3859 #ifndef _SYS_SYSPROTO_H_
3860 struct freebsd11_getdirentries_args {
3861 	int	fd;
3862 	char	*buf;
3863 	u_int	count;
3864 	long	*basep;
3865 };
3866 #endif
3867 int
3868 freebsd11_getdirentries(struct thread *td,
3869     struct freebsd11_getdirentries_args *uap)
3870 {
3871 	long base;
3872 	int error;
3873 
3874 	error = freebsd11_kern_getdirentries(td, uap->fd, uap->buf, uap->count,
3875 	    &base, NULL);
3876 
3877 	if (error == 0 && uap->basep != NULL)
3878 		error = copyout(&base, uap->basep, sizeof(long));
3879 	return (error);
3880 }
3881 
3882 int
3883 freebsd11_getdents(struct thread *td, struct freebsd11_getdents_args *uap)
3884 {
3885 	struct freebsd11_getdirentries_args ap;
3886 
3887 	ap.fd = uap->fd;
3888 	ap.buf = uap->buf;
3889 	ap.count = uap->count;
3890 	ap.basep = NULL;
3891 	return (freebsd11_getdirentries(td, &ap));
3892 }
3893 #endif /* COMPAT_FREEBSD11 */
3894 
3895 /*
3896  * Read a block of directory entries in a filesystem independent format.
3897  */
3898 int
3899 sys_getdirentries(struct thread *td, struct getdirentries_args *uap)
3900 {
3901 	off_t base;
3902 	int error;
3903 
3904 	error = kern_getdirentries(td, uap->fd, uap->buf, uap->count, &base,
3905 	    NULL, UIO_USERSPACE);
3906 	if (error != 0)
3907 		return (error);
3908 	if (uap->basep != NULL)
3909 		error = copyout(&base, uap->basep, sizeof(off_t));
3910 	return (error);
3911 }
3912 
3913 int
3914 kern_getdirentries(struct thread *td, int fd, char *buf, size_t count,
3915     off_t *basep, ssize_t *residp, enum uio_seg bufseg)
3916 {
3917 	struct vnode *vp;
3918 	struct file *fp;
3919 	struct uio auio;
3920 	struct iovec aiov;
3921 	off_t loff;
3922 	int error, eofflag;
3923 	off_t foffset;
3924 
3925 	AUDIT_ARG_FD(fd);
3926 	if (count > IOSIZE_MAX)
3927 		return (EINVAL);
3928 	auio.uio_resid = count;
3929 	error = getvnode(td, fd, &cap_read_rights, &fp);
3930 	if (error != 0)
3931 		return (error);
3932 	if ((fp->f_flag & FREAD) == 0) {
3933 		fdrop(fp, td);
3934 		return (EBADF);
3935 	}
3936 	vp = fp->f_vnode;
3937 	foffset = foffset_lock(fp, 0);
3938 unionread:
3939 	if (vp->v_type != VDIR) {
3940 		error = EINVAL;
3941 		goto fail;
3942 	}
3943 	aiov.iov_base = buf;
3944 	aiov.iov_len = count;
3945 	auio.uio_iov = &aiov;
3946 	auio.uio_iovcnt = 1;
3947 	auio.uio_rw = UIO_READ;
3948 	auio.uio_segflg = bufseg;
3949 	auio.uio_td = td;
3950 	vn_lock(vp, LK_SHARED | LK_RETRY);
3951 	AUDIT_ARG_VNODE1(vp);
3952 	loff = auio.uio_offset = foffset;
3953 #ifdef MAC
3954 	error = mac_vnode_check_readdir(td->td_ucred, vp);
3955 	if (error == 0)
3956 #endif
3957 		error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, NULL,
3958 		    NULL);
3959 	foffset = auio.uio_offset;
3960 	if (error != 0) {
3961 		VOP_UNLOCK(vp, 0);
3962 		goto fail;
3963 	}
3964 	if (count == auio.uio_resid &&
3965 	    (vp->v_vflag & VV_ROOT) &&
3966 	    (vp->v_mount->mnt_flag & MNT_UNION)) {
3967 		struct vnode *tvp = vp;
3968 
3969 		vp = vp->v_mount->mnt_vnodecovered;
3970 		VREF(vp);
3971 		fp->f_vnode = vp;
3972 		fp->f_data = vp;
3973 		foffset = 0;
3974 		vput(tvp);
3975 		goto unionread;
3976 	}
3977 	VOP_UNLOCK(vp, 0);
3978 	*basep = loff;
3979 	if (residp != NULL)
3980 		*residp = auio.uio_resid;
3981 	td->td_retval[0] = count - auio.uio_resid;
3982 fail:
3983 	foffset_unlock(fp, foffset, 0);
3984 	fdrop(fp, td);
3985 	return (error);
3986 }
3987 
3988 /*
3989  * Set the mode mask for creation of filesystem nodes.
3990  */
3991 #ifndef _SYS_SYSPROTO_H_
3992 struct umask_args {
3993 	int	newmask;
3994 };
3995 #endif
3996 int
3997 sys_umask(struct thread *td, struct umask_args *uap)
3998 {
3999 	struct filedesc *fdp;
4000 
4001 	fdp = td->td_proc->p_fd;
4002 	FILEDESC_XLOCK(fdp);
4003 	td->td_retval[0] = fdp->fd_cmask;
4004 	fdp->fd_cmask = uap->newmask & ALLPERMS;
4005 	FILEDESC_XUNLOCK(fdp);
4006 	return (0);
4007 }
4008 
4009 /*
4010  * Void all references to file by ripping underlying filesystem away from
4011  * vnode.
4012  */
4013 #ifndef _SYS_SYSPROTO_H_
4014 struct revoke_args {
4015 	char	*path;
4016 };
4017 #endif
4018 int
4019 sys_revoke(struct thread *td, struct revoke_args *uap)
4020 {
4021 	struct vnode *vp;
4022 	struct vattr vattr;
4023 	struct nameidata nd;
4024 	int error;
4025 
4026 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
4027 	    uap->path, td);
4028 	if ((error = namei(&nd)) != 0)
4029 		return (error);
4030 	vp = nd.ni_vp;
4031 	NDFREE(&nd, NDF_ONLY_PNBUF);
4032 	if (vp->v_type != VCHR || vp->v_rdev == NULL) {
4033 		error = EINVAL;
4034 		goto out;
4035 	}
4036 #ifdef MAC
4037 	error = mac_vnode_check_revoke(td->td_ucred, vp);
4038 	if (error != 0)
4039 		goto out;
4040 #endif
4041 	error = VOP_GETATTR(vp, &vattr, td->td_ucred);
4042 	if (error != 0)
4043 		goto out;
4044 	if (td->td_ucred->cr_uid != vattr.va_uid) {
4045 		error = priv_check(td, PRIV_VFS_ADMIN);
4046 		if (error != 0)
4047 			goto out;
4048 	}
4049 	if (vcount(vp) > 1)
4050 		VOP_REVOKE(vp, REVOKEALL);
4051 out:
4052 	vput(vp);
4053 	return (error);
4054 }
4055 
4056 /*
4057  * Convert a user file descriptor to a kernel file entry and check that, if it
4058  * is a capability, the correct rights are present. A reference on the file
4059  * entry is held upon returning.
4060  */
4061 int
4062 getvnode(struct thread *td, int fd, cap_rights_t *rightsp, struct file **fpp)
4063 {
4064 	struct file *fp;
4065 	int error;
4066 
4067 	error = fget_unlocked(td->td_proc->p_fd, fd, rightsp, &fp, NULL);
4068 	if (error != 0)
4069 		return (error);
4070 
4071 	/*
4072 	 * The file could be not of the vnode type, or it may be not
4073 	 * yet fully initialized, in which case the f_vnode pointer
4074 	 * may be set, but f_ops is still badfileops.  E.g.,
4075 	 * devfs_open() transiently create such situation to
4076 	 * facilitate csw d_fdopen().
4077 	 *
4078 	 * Dupfdopen() handling in kern_openat() installs the
4079 	 * half-baked file into the process descriptor table, allowing
4080 	 * other thread to dereference it. Guard against the race by
4081 	 * checking f_ops.
4082 	 */
4083 	if (fp->f_vnode == NULL || fp->f_ops == &badfileops) {
4084 		fdrop(fp, td);
4085 		return (EINVAL);
4086 	}
4087 	*fpp = fp;
4088 	return (0);
4089 }
4090 
4091 
4092 /*
4093  * Get an (NFS) file handle.
4094  */
4095 #ifndef _SYS_SYSPROTO_H_
4096 struct lgetfh_args {
4097 	char	*fname;
4098 	fhandle_t *fhp;
4099 };
4100 #endif
4101 int
4102 sys_lgetfh(struct thread *td, struct lgetfh_args *uap)
4103 {
4104 	struct nameidata nd;
4105 	fhandle_t fh;
4106 	struct vnode *vp;
4107 	int error;
4108 
4109 	error = priv_check(td, PRIV_VFS_GETFH);
4110 	if (error != 0)
4111 		return (error);
4112 	NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
4113 	    uap->fname, td);
4114 	error = namei(&nd);
4115 	if (error != 0)
4116 		return (error);
4117 	NDFREE(&nd, NDF_ONLY_PNBUF);
4118 	vp = nd.ni_vp;
4119 	bzero(&fh, sizeof(fh));
4120 	fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
4121 	error = VOP_VPTOFH(vp, &fh.fh_fid);
4122 	vput(vp);
4123 	if (error == 0)
4124 		error = copyout(&fh, uap->fhp, sizeof (fh));
4125 	return (error);
4126 }
4127 
4128 #ifndef _SYS_SYSPROTO_H_
4129 struct getfh_args {
4130 	char	*fname;
4131 	fhandle_t *fhp;
4132 };
4133 #endif
4134 int
4135 sys_getfh(struct thread *td, struct getfh_args *uap)
4136 {
4137 	struct nameidata nd;
4138 	fhandle_t fh;
4139 	struct vnode *vp;
4140 	int error;
4141 
4142 	error = priv_check(td, PRIV_VFS_GETFH);
4143 	if (error != 0)
4144 		return (error);
4145 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
4146 	    uap->fname, td);
4147 	error = namei(&nd);
4148 	if (error != 0)
4149 		return (error);
4150 	NDFREE(&nd, NDF_ONLY_PNBUF);
4151 	vp = nd.ni_vp;
4152 	bzero(&fh, sizeof(fh));
4153 	fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
4154 	error = VOP_VPTOFH(vp, &fh.fh_fid);
4155 	vput(vp);
4156 	if (error == 0)
4157 		error = copyout(&fh, uap->fhp, sizeof (fh));
4158 	return (error);
4159 }
4160 
4161 /*
4162  * syscall for the rpc.lockd to use to translate a NFS file handle into an
4163  * open descriptor.
4164  *
4165  * warning: do not remove the priv_check() call or this becomes one giant
4166  * security hole.
4167  */
4168 #ifndef _SYS_SYSPROTO_H_
4169 struct fhopen_args {
4170 	const struct fhandle *u_fhp;
4171 	int flags;
4172 };
4173 #endif
4174 int
4175 sys_fhopen(struct thread *td, struct fhopen_args *uap)
4176 {
4177 	struct mount *mp;
4178 	struct vnode *vp;
4179 	struct fhandle fhp;
4180 	struct file *fp;
4181 	int fmode, error;
4182 	int indx;
4183 
4184 	error = priv_check(td, PRIV_VFS_FHOPEN);
4185 	if (error != 0)
4186 		return (error);
4187 	indx = -1;
4188 	fmode = FFLAGS(uap->flags);
4189 	/* why not allow a non-read/write open for our lockd? */
4190 	if (((fmode & (FREAD | FWRITE)) == 0) || (fmode & O_CREAT))
4191 		return (EINVAL);
4192 	error = copyin(uap->u_fhp, &fhp, sizeof(fhp));
4193 	if (error != 0)
4194 		return(error);
4195 	/* find the mount point */
4196 	mp = vfs_busyfs(&fhp.fh_fsid);
4197 	if (mp == NULL)
4198 		return (ESTALE);
4199 	/* now give me my vnode, it gets returned to me locked */
4200 	error = VFS_FHTOVP(mp, &fhp.fh_fid, LK_EXCLUSIVE, &vp);
4201 	vfs_unbusy(mp);
4202 	if (error != 0)
4203 		return (error);
4204 
4205 	error = falloc_noinstall(td, &fp);
4206 	if (error != 0) {
4207 		vput(vp);
4208 		return (error);
4209 	}
4210 	/*
4211 	 * An extra reference on `fp' has been held for us by
4212 	 * falloc_noinstall().
4213 	 */
4214 
4215 #ifdef INVARIANTS
4216 	td->td_dupfd = -1;
4217 #endif
4218 	error = vn_open_vnode(vp, fmode, td->td_ucred, td, fp);
4219 	if (error != 0) {
4220 		KASSERT(fp->f_ops == &badfileops,
4221 		    ("VOP_OPEN in fhopen() set f_ops"));
4222 		KASSERT(td->td_dupfd < 0,
4223 		    ("fhopen() encountered fdopen()"));
4224 
4225 		vput(vp);
4226 		goto bad;
4227 	}
4228 #ifdef INVARIANTS
4229 	td->td_dupfd = 0;
4230 #endif
4231 	fp->f_vnode = vp;
4232 	fp->f_seqcount = 1;
4233 	finit(fp, (fmode & FMASK) | (fp->f_flag & FHASLOCK), DTYPE_VNODE, vp,
4234 	    &vnops);
4235 	VOP_UNLOCK(vp, 0);
4236 	if ((fmode & O_TRUNC) != 0) {
4237 		error = fo_truncate(fp, 0, td->td_ucred, td);
4238 		if (error != 0)
4239 			goto bad;
4240 	}
4241 
4242 	error = finstall(td, fp, &indx, fmode, NULL);
4243 bad:
4244 	fdrop(fp, td);
4245 	td->td_retval[0] = indx;
4246 	return (error);
4247 }
4248 
4249 /*
4250  * Stat an (NFS) file handle.
4251  */
4252 #ifndef _SYS_SYSPROTO_H_
4253 struct fhstat_args {
4254 	struct fhandle *u_fhp;
4255 	struct stat *sb;
4256 };
4257 #endif
4258 int
4259 sys_fhstat(struct thread *td, struct fhstat_args *uap)
4260 {
4261 	struct stat sb;
4262 	struct fhandle fh;
4263 	int error;
4264 
4265 	error = copyin(uap->u_fhp, &fh, sizeof(fh));
4266 	if (error != 0)
4267 		return (error);
4268 	error = kern_fhstat(td, fh, &sb);
4269 	if (error == 0)
4270 		error = copyout(&sb, uap->sb, sizeof(sb));
4271 	return (error);
4272 }
4273 
4274 int
4275 kern_fhstat(struct thread *td, struct fhandle fh, struct stat *sb)
4276 {
4277 	struct mount *mp;
4278 	struct vnode *vp;
4279 	int error;
4280 
4281 	error = priv_check(td, PRIV_VFS_FHSTAT);
4282 	if (error != 0)
4283 		return (error);
4284 	if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4285 		return (ESTALE);
4286 	error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4287 	vfs_unbusy(mp);
4288 	if (error != 0)
4289 		return (error);
4290 	error = vn_stat(vp, sb, td->td_ucred, NOCRED, td);
4291 	vput(vp);
4292 	return (error);
4293 }
4294 
4295 /*
4296  * Implement fstatfs() for (NFS) file handles.
4297  */
4298 #ifndef _SYS_SYSPROTO_H_
4299 struct fhstatfs_args {
4300 	struct fhandle *u_fhp;
4301 	struct statfs *buf;
4302 };
4303 #endif
4304 int
4305 sys_fhstatfs(struct thread *td, struct fhstatfs_args *uap)
4306 {
4307 	struct statfs *sfp;
4308 	fhandle_t fh;
4309 	int error;
4310 
4311 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
4312 	if (error != 0)
4313 		return (error);
4314 	sfp = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
4315 	error = kern_fhstatfs(td, fh, sfp);
4316 	if (error == 0)
4317 		error = copyout(sfp, uap->buf, sizeof(*sfp));
4318 	free(sfp, M_STATFS);
4319 	return (error);
4320 }
4321 
4322 int
4323 kern_fhstatfs(struct thread *td, fhandle_t fh, struct statfs *buf)
4324 {
4325 	struct statfs *sp;
4326 	struct mount *mp;
4327 	struct vnode *vp;
4328 	int error;
4329 
4330 	error = priv_check(td, PRIV_VFS_FHSTATFS);
4331 	if (error != 0)
4332 		return (error);
4333 	if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4334 		return (ESTALE);
4335 	error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4336 	if (error != 0) {
4337 		vfs_unbusy(mp);
4338 		return (error);
4339 	}
4340 	vput(vp);
4341 	error = prison_canseemount(td->td_ucred, mp);
4342 	if (error != 0)
4343 		goto out;
4344 #ifdef MAC
4345 	error = mac_mount_check_stat(td->td_ucred, mp);
4346 	if (error != 0)
4347 		goto out;
4348 #endif
4349 	/*
4350 	 * Set these in case the underlying filesystem fails to do so.
4351 	 */
4352 	sp = &mp->mnt_stat;
4353 	sp->f_version = STATFS_VERSION;
4354 	sp->f_namemax = NAME_MAX;
4355 	sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
4356 	error = VFS_STATFS(mp, sp);
4357 	if (error == 0)
4358 		*buf = *sp;
4359 out:
4360 	vfs_unbusy(mp);
4361 	return (error);
4362 }
4363 
4364 int
4365 kern_posix_fallocate(struct thread *td, int fd, off_t offset, off_t len)
4366 {
4367 	struct file *fp;
4368 	struct mount *mp;
4369 	struct vnode *vp;
4370 	off_t olen, ooffset;
4371 	int error;
4372 #ifdef AUDIT
4373 	int audited_vnode1 = 0;
4374 #endif
4375 
4376 	AUDIT_ARG_FD(fd);
4377 	if (offset < 0 || len <= 0)
4378 		return (EINVAL);
4379 	/* Check for wrap. */
4380 	if (offset > OFF_MAX - len)
4381 		return (EFBIG);
4382 	AUDIT_ARG_FD(fd);
4383 	error = fget(td, fd, &cap_pwrite_rights, &fp);
4384 	if (error != 0)
4385 		return (error);
4386 	AUDIT_ARG_FILE(td->td_proc, fp);
4387 	if ((fp->f_ops->fo_flags & DFLAG_SEEKABLE) == 0) {
4388 		error = ESPIPE;
4389 		goto out;
4390 	}
4391 	if ((fp->f_flag & FWRITE) == 0) {
4392 		error = EBADF;
4393 		goto out;
4394 	}
4395 	if (fp->f_type != DTYPE_VNODE) {
4396 		error = ENODEV;
4397 		goto out;
4398 	}
4399 	vp = fp->f_vnode;
4400 	if (vp->v_type != VREG) {
4401 		error = ENODEV;
4402 		goto out;
4403 	}
4404 
4405 	/* Allocating blocks may take a long time, so iterate. */
4406 	for (;;) {
4407 		olen = len;
4408 		ooffset = offset;
4409 
4410 		bwillwrite();
4411 		mp = NULL;
4412 		error = vn_start_write(vp, &mp, V_WAIT | PCATCH);
4413 		if (error != 0)
4414 			break;
4415 		error = vn_lock(vp, LK_EXCLUSIVE);
4416 		if (error != 0) {
4417 			vn_finished_write(mp);
4418 			break;
4419 		}
4420 #ifdef AUDIT
4421 		if (!audited_vnode1) {
4422 			AUDIT_ARG_VNODE1(vp);
4423 			audited_vnode1 = 1;
4424 		}
4425 #endif
4426 #ifdef MAC
4427 		error = mac_vnode_check_write(td->td_ucred, fp->f_cred, vp);
4428 		if (error == 0)
4429 #endif
4430 			error = VOP_ALLOCATE(vp, &offset, &len);
4431 		VOP_UNLOCK(vp, 0);
4432 		vn_finished_write(mp);
4433 
4434 		if (olen + ooffset != offset + len) {
4435 			panic("offset + len changed from %jx/%jx to %jx/%jx",
4436 			    ooffset, olen, offset, len);
4437 		}
4438 		if (error != 0 || len == 0)
4439 			break;
4440 		KASSERT(olen > len, ("Iteration did not make progress?"));
4441 		maybe_yield();
4442 	}
4443  out:
4444 	fdrop(fp, td);
4445 	return (error);
4446 }
4447 
4448 int
4449 sys_posix_fallocate(struct thread *td, struct posix_fallocate_args *uap)
4450 {
4451 	int error;
4452 
4453 	error = kern_posix_fallocate(td, uap->fd, uap->offset, uap->len);
4454 	return (kern_posix_error(td, error));
4455 }
4456 
4457 /*
4458  * Unlike madvise(2), we do not make a best effort to remember every
4459  * possible caching hint.  Instead, we remember the last setting with
4460  * the exception that we will allow POSIX_FADV_NORMAL to adjust the
4461  * region of any current setting.
4462  */
4463 int
4464 kern_posix_fadvise(struct thread *td, int fd, off_t offset, off_t len,
4465     int advice)
4466 {
4467 	struct fadvise_info *fa, *new;
4468 	struct file *fp;
4469 	struct vnode *vp;
4470 	off_t end;
4471 	int error;
4472 
4473 	if (offset < 0 || len < 0 || offset > OFF_MAX - len)
4474 		return (EINVAL);
4475 	AUDIT_ARG_VALUE(advice);
4476 	switch (advice) {
4477 	case POSIX_FADV_SEQUENTIAL:
4478 	case POSIX_FADV_RANDOM:
4479 	case POSIX_FADV_NOREUSE:
4480 		new = malloc(sizeof(*fa), M_FADVISE, M_WAITOK);
4481 		break;
4482 	case POSIX_FADV_NORMAL:
4483 	case POSIX_FADV_WILLNEED:
4484 	case POSIX_FADV_DONTNEED:
4485 		new = NULL;
4486 		break;
4487 	default:
4488 		return (EINVAL);
4489 	}
4490 	/* XXX: CAP_POSIX_FADVISE? */
4491 	AUDIT_ARG_FD(fd);
4492 	error = fget(td, fd, &cap_no_rights, &fp);
4493 	if (error != 0)
4494 		goto out;
4495 	AUDIT_ARG_FILE(td->td_proc, fp);
4496 	if ((fp->f_ops->fo_flags & DFLAG_SEEKABLE) == 0) {
4497 		error = ESPIPE;
4498 		goto out;
4499 	}
4500 	if (fp->f_type != DTYPE_VNODE) {
4501 		error = ENODEV;
4502 		goto out;
4503 	}
4504 	vp = fp->f_vnode;
4505 	if (vp->v_type != VREG) {
4506 		error = ENODEV;
4507 		goto out;
4508 	}
4509 	if (len == 0)
4510 		end = OFF_MAX;
4511 	else
4512 		end = offset + len - 1;
4513 	switch (advice) {
4514 	case POSIX_FADV_SEQUENTIAL:
4515 	case POSIX_FADV_RANDOM:
4516 	case POSIX_FADV_NOREUSE:
4517 		/*
4518 		 * Try to merge any existing non-standard region with
4519 		 * this new region if possible, otherwise create a new
4520 		 * non-standard region for this request.
4521 		 */
4522 		mtx_pool_lock(mtxpool_sleep, fp);
4523 		fa = fp->f_advice;
4524 		if (fa != NULL && fa->fa_advice == advice &&
4525 		    ((fa->fa_start <= end && fa->fa_end >= offset) ||
4526 		    (end != OFF_MAX && fa->fa_start == end + 1) ||
4527 		    (fa->fa_end != OFF_MAX && fa->fa_end + 1 == offset))) {
4528 			if (offset < fa->fa_start)
4529 				fa->fa_start = offset;
4530 			if (end > fa->fa_end)
4531 				fa->fa_end = end;
4532 		} else {
4533 			new->fa_advice = advice;
4534 			new->fa_start = offset;
4535 			new->fa_end = end;
4536 			fp->f_advice = new;
4537 			new = fa;
4538 		}
4539 		mtx_pool_unlock(mtxpool_sleep, fp);
4540 		break;
4541 	case POSIX_FADV_NORMAL:
4542 		/*
4543 		 * If a the "normal" region overlaps with an existing
4544 		 * non-standard region, trim or remove the
4545 		 * non-standard region.
4546 		 */
4547 		mtx_pool_lock(mtxpool_sleep, fp);
4548 		fa = fp->f_advice;
4549 		if (fa != NULL) {
4550 			if (offset <= fa->fa_start && end >= fa->fa_end) {
4551 				new = fa;
4552 				fp->f_advice = NULL;
4553 			} else if (offset <= fa->fa_start &&
4554 			    end >= fa->fa_start)
4555 				fa->fa_start = end + 1;
4556 			else if (offset <= fa->fa_end && end >= fa->fa_end)
4557 				fa->fa_end = offset - 1;
4558 			else if (offset >= fa->fa_start && end <= fa->fa_end) {
4559 				/*
4560 				 * If the "normal" region is a middle
4561 				 * portion of the existing
4562 				 * non-standard region, just remove
4563 				 * the whole thing rather than picking
4564 				 * one side or the other to
4565 				 * preserve.
4566 				 */
4567 				new = fa;
4568 				fp->f_advice = NULL;
4569 			}
4570 		}
4571 		mtx_pool_unlock(mtxpool_sleep, fp);
4572 		break;
4573 	case POSIX_FADV_WILLNEED:
4574 	case POSIX_FADV_DONTNEED:
4575 		error = VOP_ADVISE(vp, offset, end, advice);
4576 		break;
4577 	}
4578 out:
4579 	if (fp != NULL)
4580 		fdrop(fp, td);
4581 	free(new, M_FADVISE);
4582 	return (error);
4583 }
4584 
4585 int
4586 sys_posix_fadvise(struct thread *td, struct posix_fadvise_args *uap)
4587 {
4588 	int error;
4589 
4590 	error = kern_posix_fadvise(td, uap->fd, uap->offset, uap->len,
4591 	    uap->advice);
4592 	return (kern_posix_error(td, error));
4593 }
4594