xref: /freebsd/sys/kern/vfs_syscalls.c (revision f4b37ed0f8b307b1f3f0f630ca725d68f1dff30d)
1 /*-
2  * Copyright (c) 1989, 1993
3  *	The Regents of the University of California.  All rights reserved.
4  * (c) UNIX System Laboratories, Inc.
5  * All or some portions of this file are derived from material licensed
6  * to the University of California by American Telephone and Telegraph
7  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
8  * the permission of UNIX System Laboratories, Inc.
9  *
10  * Redistribution and use in source and binary forms, with or without
11  * modification, are permitted provided that the following conditions
12  * are met:
13  * 1. Redistributions of source code must retain the above copyright
14  *    notice, this list of conditions and the following disclaimer.
15  * 2. Redistributions in binary form must reproduce the above copyright
16  *    notice, this list of conditions and the following disclaimer in the
17  *    documentation and/or other materials provided with the distribution.
18  * 4. Neither the name of the University nor the names of its contributors
19  *    may be used to endorse or promote products derived from this software
20  *    without specific prior written permission.
21  *
22  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
23  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
24  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
25  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
26  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
27  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
28  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
29  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
30  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
31  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32  * SUCH DAMAGE.
33  *
34  *	@(#)vfs_syscalls.c	8.13 (Berkeley) 4/15/94
35  */
36 
37 #include <sys/cdefs.h>
38 __FBSDID("$FreeBSD$");
39 
40 #include "opt_capsicum.h"
41 #include "opt_compat.h"
42 #include "opt_ktrace.h"
43 
44 #include <sys/param.h>
45 #include <sys/systm.h>
46 #include <sys/bio.h>
47 #include <sys/buf.h>
48 #include <sys/capsicum.h>
49 #include <sys/disk.h>
50 #include <sys/sysent.h>
51 #include <sys/malloc.h>
52 #include <sys/mount.h>
53 #include <sys/mutex.h>
54 #include <sys/sysproto.h>
55 #include <sys/namei.h>
56 #include <sys/filedesc.h>
57 #include <sys/kernel.h>
58 #include <sys/fcntl.h>
59 #include <sys/file.h>
60 #include <sys/filio.h>
61 #include <sys/limits.h>
62 #include <sys/linker.h>
63 #include <sys/rwlock.h>
64 #include <sys/sdt.h>
65 #include <sys/stat.h>
66 #include <sys/sx.h>
67 #include <sys/unistd.h>
68 #include <sys/vnode.h>
69 #include <sys/priv.h>
70 #include <sys/proc.h>
71 #include <sys/dirent.h>
72 #include <sys/jail.h>
73 #include <sys/syscallsubr.h>
74 #include <sys/sysctl.h>
75 #ifdef KTRACE
76 #include <sys/ktrace.h>
77 #endif
78 
79 #include <machine/stdarg.h>
80 
81 #include <security/audit/audit.h>
82 #include <security/mac/mac_framework.h>
83 
84 #include <vm/vm.h>
85 #include <vm/vm_object.h>
86 #include <vm/vm_page.h>
87 #include <vm/uma.h>
88 
89 #include <ufs/ufs/quota.h>
90 
91 MALLOC_DEFINE(M_FADVISE, "fadvise", "posix_fadvise(2) information");
92 
93 SDT_PROVIDER_DEFINE(vfs);
94 SDT_PROBE_DEFINE2(vfs, , stat, mode, "char *", "int");
95 SDT_PROBE_DEFINE2(vfs, , stat, reg, "char *", "int");
96 
97 static int kern_chflagsat(struct thread *td, int fd, const char *path,
98     enum uio_seg pathseg, u_long flags, int atflag);
99 static int setfflags(struct thread *td, struct vnode *, u_long);
100 static int getutimes(const struct timeval *, enum uio_seg, struct timespec *);
101 static int getutimens(const struct timespec *, enum uio_seg,
102     struct timespec *, int *);
103 static int setutimes(struct thread *td, struct vnode *,
104     const struct timespec *, int, int);
105 static int vn_access(struct vnode *vp, int user_flags, struct ucred *cred,
106     struct thread *td);
107 
108 /*
109  * The module initialization routine for POSIX asynchronous I/O will
110  * set this to the version of AIO that it implements.  (Zero means
111  * that it is not implemented.)  This value is used here by pathconf()
112  * and in kern_descrip.c by fpathconf().
113  */
114 int async_io_version;
115 
116 /*
117  * Sync each mounted filesystem.
118  */
119 #ifndef _SYS_SYSPROTO_H_
120 struct sync_args {
121 	int     dummy;
122 };
123 #endif
124 /* ARGSUSED */
125 int
126 sys_sync(td, uap)
127 	struct thread *td;
128 	struct sync_args *uap;
129 {
130 	struct mount *mp, *nmp;
131 	int save;
132 
133 	mtx_lock(&mountlist_mtx);
134 	for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
135 		if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK)) {
136 			nmp = TAILQ_NEXT(mp, mnt_list);
137 			continue;
138 		}
139 		if ((mp->mnt_flag & MNT_RDONLY) == 0 &&
140 		    vn_start_write(NULL, &mp, V_NOWAIT) == 0) {
141 			save = curthread_pflags_set(TDP_SYNCIO);
142 			vfs_msync(mp, MNT_NOWAIT);
143 			VFS_SYNC(mp, MNT_NOWAIT);
144 			curthread_pflags_restore(save);
145 			vn_finished_write(mp);
146 		}
147 		mtx_lock(&mountlist_mtx);
148 		nmp = TAILQ_NEXT(mp, mnt_list);
149 		vfs_unbusy(mp);
150 	}
151 	mtx_unlock(&mountlist_mtx);
152 	return (0);
153 }
154 
155 /*
156  * Change filesystem quotas.
157  */
158 #ifndef _SYS_SYSPROTO_H_
159 struct quotactl_args {
160 	char *path;
161 	int cmd;
162 	int uid;
163 	caddr_t arg;
164 };
165 #endif
166 int
167 sys_quotactl(td, uap)
168 	struct thread *td;
169 	register struct quotactl_args /* {
170 		char *path;
171 		int cmd;
172 		int uid;
173 		caddr_t arg;
174 	} */ *uap;
175 {
176 	struct mount *mp;
177 	struct nameidata nd;
178 	int error;
179 
180 	AUDIT_ARG_CMD(uap->cmd);
181 	AUDIT_ARG_UID(uap->uid);
182 	if (!prison_allow(td->td_ucred, PR_ALLOW_QUOTAS))
183 		return (EPERM);
184 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
185 	    uap->path, td);
186 	if ((error = namei(&nd)) != 0)
187 		return (error);
188 	NDFREE(&nd, NDF_ONLY_PNBUF);
189 	mp = nd.ni_vp->v_mount;
190 	vfs_ref(mp);
191 	vput(nd.ni_vp);
192 	error = vfs_busy(mp, 0);
193 	vfs_rel(mp);
194 	if (error != 0)
195 		return (error);
196 	error = VFS_QUOTACTL(mp, uap->cmd, uap->uid, uap->arg);
197 
198 	/*
199 	 * Since quota on operation typically needs to open quota
200 	 * file, the Q_QUOTAON handler needs to unbusy the mount point
201 	 * before calling into namei.  Otherwise, unmount might be
202 	 * started between two vfs_busy() invocations (first is our,
203 	 * second is from mount point cross-walk code in lookup()),
204 	 * causing deadlock.
205 	 *
206 	 * Require that Q_QUOTAON handles the vfs_busy() reference on
207 	 * its own, always returning with ubusied mount point.
208 	 */
209 	if ((uap->cmd >> SUBCMDSHIFT) != Q_QUOTAON)
210 		vfs_unbusy(mp);
211 	return (error);
212 }
213 
214 /*
215  * Used by statfs conversion routines to scale the block size up if
216  * necessary so that all of the block counts are <= 'max_size'.  Note
217  * that 'max_size' should be a bitmask, i.e. 2^n - 1 for some non-zero
218  * value of 'n'.
219  */
220 void
221 statfs_scale_blocks(struct statfs *sf, long max_size)
222 {
223 	uint64_t count;
224 	int shift;
225 
226 	KASSERT(powerof2(max_size + 1), ("%s: invalid max_size", __func__));
227 
228 	/*
229 	 * Attempt to scale the block counts to give a more accurate
230 	 * overview to userland of the ratio of free space to used
231 	 * space.  To do this, find the largest block count and compute
232 	 * a divisor that lets it fit into a signed integer <= max_size.
233 	 */
234 	if (sf->f_bavail < 0)
235 		count = -sf->f_bavail;
236 	else
237 		count = sf->f_bavail;
238 	count = MAX(sf->f_blocks, MAX(sf->f_bfree, count));
239 	if (count <= max_size)
240 		return;
241 
242 	count >>= flsl(max_size);
243 	shift = 0;
244 	while (count > 0) {
245 		shift++;
246 		count >>=1;
247 	}
248 
249 	sf->f_bsize <<= shift;
250 	sf->f_blocks >>= shift;
251 	sf->f_bfree >>= shift;
252 	sf->f_bavail >>= shift;
253 }
254 
255 /*
256  * Get filesystem statistics.
257  */
258 #ifndef _SYS_SYSPROTO_H_
259 struct statfs_args {
260 	char *path;
261 	struct statfs *buf;
262 };
263 #endif
264 int
265 sys_statfs(td, uap)
266 	struct thread *td;
267 	register struct statfs_args /* {
268 		char *path;
269 		struct statfs *buf;
270 	} */ *uap;
271 {
272 	struct statfs sf;
273 	int error;
274 
275 	error = kern_statfs(td, uap->path, UIO_USERSPACE, &sf);
276 	if (error == 0)
277 		error = copyout(&sf, uap->buf, sizeof(sf));
278 	return (error);
279 }
280 
281 int
282 kern_statfs(struct thread *td, char *path, enum uio_seg pathseg,
283     struct statfs *buf)
284 {
285 	struct mount *mp;
286 	struct statfs *sp, sb;
287 	struct nameidata nd;
288 	int error;
289 
290 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
291 	    pathseg, path, td);
292 	error = namei(&nd);
293 	if (error != 0)
294 		return (error);
295 	mp = nd.ni_vp->v_mount;
296 	vfs_ref(mp);
297 	NDFREE(&nd, NDF_ONLY_PNBUF);
298 	vput(nd.ni_vp);
299 	error = vfs_busy(mp, 0);
300 	vfs_rel(mp);
301 	if (error != 0)
302 		return (error);
303 #ifdef MAC
304 	error = mac_mount_check_stat(td->td_ucred, mp);
305 	if (error != 0)
306 		goto out;
307 #endif
308 	/*
309 	 * Set these in case the underlying filesystem fails to do so.
310 	 */
311 	sp = &mp->mnt_stat;
312 	sp->f_version = STATFS_VERSION;
313 	sp->f_namemax = NAME_MAX;
314 	sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
315 	error = VFS_STATFS(mp, sp);
316 	if (error != 0)
317 		goto out;
318 	if (priv_check(td, PRIV_VFS_GENERATION)) {
319 		bcopy(sp, &sb, sizeof(sb));
320 		sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
321 		prison_enforce_statfs(td->td_ucred, mp, &sb);
322 		sp = &sb;
323 	}
324 	*buf = *sp;
325 out:
326 	vfs_unbusy(mp);
327 	return (error);
328 }
329 
330 /*
331  * Get filesystem statistics.
332  */
333 #ifndef _SYS_SYSPROTO_H_
334 struct fstatfs_args {
335 	int fd;
336 	struct statfs *buf;
337 };
338 #endif
339 int
340 sys_fstatfs(td, uap)
341 	struct thread *td;
342 	register struct fstatfs_args /* {
343 		int fd;
344 		struct statfs *buf;
345 	} */ *uap;
346 {
347 	struct statfs sf;
348 	int error;
349 
350 	error = kern_fstatfs(td, uap->fd, &sf);
351 	if (error == 0)
352 		error = copyout(&sf, uap->buf, sizeof(sf));
353 	return (error);
354 }
355 
356 int
357 kern_fstatfs(struct thread *td, int fd, struct statfs *buf)
358 {
359 	struct file *fp;
360 	struct mount *mp;
361 	struct statfs *sp, sb;
362 	struct vnode *vp;
363 	cap_rights_t rights;
364 	int error;
365 
366 	AUDIT_ARG_FD(fd);
367 	error = getvnode(td, fd, cap_rights_init(&rights, CAP_FSTATFS), &fp);
368 	if (error != 0)
369 		return (error);
370 	vp = fp->f_vnode;
371 	vn_lock(vp, LK_SHARED | LK_RETRY);
372 #ifdef AUDIT
373 	AUDIT_ARG_VNODE1(vp);
374 #endif
375 	mp = vp->v_mount;
376 	if (mp)
377 		vfs_ref(mp);
378 	VOP_UNLOCK(vp, 0);
379 	fdrop(fp, td);
380 	if (mp == NULL) {
381 		error = EBADF;
382 		goto out;
383 	}
384 	error = vfs_busy(mp, 0);
385 	vfs_rel(mp);
386 	if (error != 0)
387 		return (error);
388 #ifdef MAC
389 	error = mac_mount_check_stat(td->td_ucred, mp);
390 	if (error != 0)
391 		goto out;
392 #endif
393 	/*
394 	 * Set these in case the underlying filesystem fails to do so.
395 	 */
396 	sp = &mp->mnt_stat;
397 	sp->f_version = STATFS_VERSION;
398 	sp->f_namemax = NAME_MAX;
399 	sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
400 	error = VFS_STATFS(mp, sp);
401 	if (error != 0)
402 		goto out;
403 	if (priv_check(td, PRIV_VFS_GENERATION)) {
404 		bcopy(sp, &sb, sizeof(sb));
405 		sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
406 		prison_enforce_statfs(td->td_ucred, mp, &sb);
407 		sp = &sb;
408 	}
409 	*buf = *sp;
410 out:
411 	if (mp)
412 		vfs_unbusy(mp);
413 	return (error);
414 }
415 
416 /*
417  * Get statistics on all filesystems.
418  */
419 #ifndef _SYS_SYSPROTO_H_
420 struct getfsstat_args {
421 	struct statfs *buf;
422 	long bufsize;
423 	int flags;
424 };
425 #endif
426 int
427 sys_getfsstat(td, uap)
428 	struct thread *td;
429 	register struct getfsstat_args /* {
430 		struct statfs *buf;
431 		long bufsize;
432 		int flags;
433 	} */ *uap;
434 {
435 	size_t count;
436 	int error;
437 
438 	error = kern_getfsstat(td, &uap->buf, uap->bufsize, &count,
439 	    UIO_USERSPACE, uap->flags);
440 	if (error == 0)
441 		td->td_retval[0] = count;
442 	return (error);
443 }
444 
445 /*
446  * If (bufsize > 0 && bufseg == UIO_SYSSPACE)
447  *	The caller is responsible for freeing memory which will be allocated
448  *	in '*buf'.
449  */
450 int
451 kern_getfsstat(struct thread *td, struct statfs **buf, size_t bufsize,
452     size_t *countp, enum uio_seg bufseg, int flags)
453 {
454 	struct mount *mp, *nmp;
455 	struct statfs *sfsp, *sp, sb;
456 	size_t count, maxcount;
457 	int error;
458 
459 	maxcount = bufsize / sizeof(struct statfs);
460 	if (bufsize == 0)
461 		sfsp = NULL;
462 	else if (bufseg == UIO_USERSPACE)
463 		sfsp = *buf;
464 	else /* if (bufseg == UIO_SYSSPACE) */ {
465 		count = 0;
466 		mtx_lock(&mountlist_mtx);
467 		TAILQ_FOREACH(mp, &mountlist, mnt_list) {
468 			count++;
469 		}
470 		mtx_unlock(&mountlist_mtx);
471 		if (maxcount > count)
472 			maxcount = count;
473 		sfsp = *buf = malloc(maxcount * sizeof(struct statfs), M_TEMP,
474 		    M_WAITOK);
475 	}
476 	count = 0;
477 	mtx_lock(&mountlist_mtx);
478 	for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
479 		if (prison_canseemount(td->td_ucred, mp) != 0) {
480 			nmp = TAILQ_NEXT(mp, mnt_list);
481 			continue;
482 		}
483 #ifdef MAC
484 		if (mac_mount_check_stat(td->td_ucred, mp) != 0) {
485 			nmp = TAILQ_NEXT(mp, mnt_list);
486 			continue;
487 		}
488 #endif
489 		if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK)) {
490 			nmp = TAILQ_NEXT(mp, mnt_list);
491 			continue;
492 		}
493 		if (sfsp && count < maxcount) {
494 			sp = &mp->mnt_stat;
495 			/*
496 			 * Set these in case the underlying filesystem
497 			 * fails to do so.
498 			 */
499 			sp->f_version = STATFS_VERSION;
500 			sp->f_namemax = NAME_MAX;
501 			sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
502 			/*
503 			 * If MNT_NOWAIT or MNT_LAZY is specified, do not
504 			 * refresh the fsstat cache. MNT_NOWAIT or MNT_LAZY
505 			 * overrides MNT_WAIT.
506 			 */
507 			if (((flags & (MNT_LAZY|MNT_NOWAIT)) == 0 ||
508 			    (flags & MNT_WAIT)) &&
509 			    (error = VFS_STATFS(mp, sp))) {
510 				mtx_lock(&mountlist_mtx);
511 				nmp = TAILQ_NEXT(mp, mnt_list);
512 				vfs_unbusy(mp);
513 				continue;
514 			}
515 			if (priv_check(td, PRIV_VFS_GENERATION)) {
516 				bcopy(sp, &sb, sizeof(sb));
517 				sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
518 				prison_enforce_statfs(td->td_ucred, mp, &sb);
519 				sp = &sb;
520 			}
521 			if (bufseg == UIO_SYSSPACE)
522 				bcopy(sp, sfsp, sizeof(*sp));
523 			else /* if (bufseg == UIO_USERSPACE) */ {
524 				error = copyout(sp, sfsp, sizeof(*sp));
525 				if (error != 0) {
526 					vfs_unbusy(mp);
527 					return (error);
528 				}
529 			}
530 			sfsp++;
531 		}
532 		count++;
533 		mtx_lock(&mountlist_mtx);
534 		nmp = TAILQ_NEXT(mp, mnt_list);
535 		vfs_unbusy(mp);
536 	}
537 	mtx_unlock(&mountlist_mtx);
538 	if (sfsp && count > maxcount)
539 		*countp = maxcount;
540 	else
541 		*countp = count;
542 	return (0);
543 }
544 
545 #ifdef COMPAT_FREEBSD4
546 /*
547  * Get old format filesystem statistics.
548  */
549 static void cvtstatfs(struct statfs *, struct ostatfs *);
550 
551 #ifndef _SYS_SYSPROTO_H_
552 struct freebsd4_statfs_args {
553 	char *path;
554 	struct ostatfs *buf;
555 };
556 #endif
557 int
558 freebsd4_statfs(td, uap)
559 	struct thread *td;
560 	struct freebsd4_statfs_args /* {
561 		char *path;
562 		struct ostatfs *buf;
563 	} */ *uap;
564 {
565 	struct ostatfs osb;
566 	struct statfs sf;
567 	int error;
568 
569 	error = kern_statfs(td, uap->path, UIO_USERSPACE, &sf);
570 	if (error != 0)
571 		return (error);
572 	cvtstatfs(&sf, &osb);
573 	return (copyout(&osb, uap->buf, sizeof(osb)));
574 }
575 
576 /*
577  * Get filesystem statistics.
578  */
579 #ifndef _SYS_SYSPROTO_H_
580 struct freebsd4_fstatfs_args {
581 	int fd;
582 	struct ostatfs *buf;
583 };
584 #endif
585 int
586 freebsd4_fstatfs(td, uap)
587 	struct thread *td;
588 	struct freebsd4_fstatfs_args /* {
589 		int fd;
590 		struct ostatfs *buf;
591 	} */ *uap;
592 {
593 	struct ostatfs osb;
594 	struct statfs sf;
595 	int error;
596 
597 	error = kern_fstatfs(td, uap->fd, &sf);
598 	if (error != 0)
599 		return (error);
600 	cvtstatfs(&sf, &osb);
601 	return (copyout(&osb, uap->buf, sizeof(osb)));
602 }
603 
604 /*
605  * Get statistics on all filesystems.
606  */
607 #ifndef _SYS_SYSPROTO_H_
608 struct freebsd4_getfsstat_args {
609 	struct ostatfs *buf;
610 	long bufsize;
611 	int flags;
612 };
613 #endif
614 int
615 freebsd4_getfsstat(td, uap)
616 	struct thread *td;
617 	register struct freebsd4_getfsstat_args /* {
618 		struct ostatfs *buf;
619 		long bufsize;
620 		int flags;
621 	} */ *uap;
622 {
623 	struct statfs *buf, *sp;
624 	struct ostatfs osb;
625 	size_t count, size;
626 	int error;
627 
628 	count = uap->bufsize / sizeof(struct ostatfs);
629 	size = count * sizeof(struct statfs);
630 	error = kern_getfsstat(td, &buf, size, &count, UIO_SYSSPACE,
631 	    uap->flags);
632 	if (size > 0) {
633 		sp = buf;
634 		while (count > 0 && error == 0) {
635 			cvtstatfs(sp, &osb);
636 			error = copyout(&osb, uap->buf, sizeof(osb));
637 			sp++;
638 			uap->buf++;
639 			count--;
640 		}
641 		free(buf, M_TEMP);
642 	}
643 	if (error == 0)
644 		td->td_retval[0] = count;
645 	return (error);
646 }
647 
648 /*
649  * Implement fstatfs() for (NFS) file handles.
650  */
651 #ifndef _SYS_SYSPROTO_H_
652 struct freebsd4_fhstatfs_args {
653 	struct fhandle *u_fhp;
654 	struct ostatfs *buf;
655 };
656 #endif
657 int
658 freebsd4_fhstatfs(td, uap)
659 	struct thread *td;
660 	struct freebsd4_fhstatfs_args /* {
661 		struct fhandle *u_fhp;
662 		struct ostatfs *buf;
663 	} */ *uap;
664 {
665 	struct ostatfs osb;
666 	struct statfs sf;
667 	fhandle_t fh;
668 	int error;
669 
670 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
671 	if (error != 0)
672 		return (error);
673 	error = kern_fhstatfs(td, fh, &sf);
674 	if (error != 0)
675 		return (error);
676 	cvtstatfs(&sf, &osb);
677 	return (copyout(&osb, uap->buf, sizeof(osb)));
678 }
679 
680 /*
681  * Convert a new format statfs structure to an old format statfs structure.
682  */
683 static void
684 cvtstatfs(nsp, osp)
685 	struct statfs *nsp;
686 	struct ostatfs *osp;
687 {
688 
689 	statfs_scale_blocks(nsp, LONG_MAX);
690 	bzero(osp, sizeof(*osp));
691 	osp->f_bsize = nsp->f_bsize;
692 	osp->f_iosize = MIN(nsp->f_iosize, LONG_MAX);
693 	osp->f_blocks = nsp->f_blocks;
694 	osp->f_bfree = nsp->f_bfree;
695 	osp->f_bavail = nsp->f_bavail;
696 	osp->f_files = MIN(nsp->f_files, LONG_MAX);
697 	osp->f_ffree = MIN(nsp->f_ffree, LONG_MAX);
698 	osp->f_owner = nsp->f_owner;
699 	osp->f_type = nsp->f_type;
700 	osp->f_flags = nsp->f_flags;
701 	osp->f_syncwrites = MIN(nsp->f_syncwrites, LONG_MAX);
702 	osp->f_asyncwrites = MIN(nsp->f_asyncwrites, LONG_MAX);
703 	osp->f_syncreads = MIN(nsp->f_syncreads, LONG_MAX);
704 	osp->f_asyncreads = MIN(nsp->f_asyncreads, LONG_MAX);
705 	strlcpy(osp->f_fstypename, nsp->f_fstypename,
706 	    MIN(MFSNAMELEN, OMFSNAMELEN));
707 	strlcpy(osp->f_mntonname, nsp->f_mntonname,
708 	    MIN(MNAMELEN, OMNAMELEN));
709 	strlcpy(osp->f_mntfromname, nsp->f_mntfromname,
710 	    MIN(MNAMELEN, OMNAMELEN));
711 	osp->f_fsid = nsp->f_fsid;
712 }
713 #endif /* COMPAT_FREEBSD4 */
714 
715 /*
716  * Change current working directory to a given file descriptor.
717  */
718 #ifndef _SYS_SYSPROTO_H_
719 struct fchdir_args {
720 	int	fd;
721 };
722 #endif
723 int
724 sys_fchdir(td, uap)
725 	struct thread *td;
726 	struct fchdir_args /* {
727 		int fd;
728 	} */ *uap;
729 {
730 	struct vnode *vp, *tdp;
731 	struct mount *mp;
732 	struct file *fp;
733 	cap_rights_t rights;
734 	int error;
735 
736 	AUDIT_ARG_FD(uap->fd);
737 	error = getvnode(td, uap->fd, cap_rights_init(&rights, CAP_FCHDIR),
738 	    &fp);
739 	if (error != 0)
740 		return (error);
741 	vp = fp->f_vnode;
742 	VREF(vp);
743 	fdrop(fp, td);
744 	vn_lock(vp, LK_SHARED | LK_RETRY);
745 	AUDIT_ARG_VNODE1(vp);
746 	error = change_dir(vp, td);
747 	while (!error && (mp = vp->v_mountedhere) != NULL) {
748 		if (vfs_busy(mp, 0))
749 			continue;
750 		error = VFS_ROOT(mp, LK_SHARED, &tdp);
751 		vfs_unbusy(mp);
752 		if (error != 0)
753 			break;
754 		vput(vp);
755 		vp = tdp;
756 	}
757 	if (error != 0) {
758 		vput(vp);
759 		return (error);
760 	}
761 	VOP_UNLOCK(vp, 0);
762 	pwd_chdir(td, vp);
763 	return (0);
764 }
765 
766 /*
767  * Change current working directory (``.'').
768  */
769 #ifndef _SYS_SYSPROTO_H_
770 struct chdir_args {
771 	char	*path;
772 };
773 #endif
774 int
775 sys_chdir(td, uap)
776 	struct thread *td;
777 	struct chdir_args /* {
778 		char *path;
779 	} */ *uap;
780 {
781 
782 	return (kern_chdir(td, uap->path, UIO_USERSPACE));
783 }
784 
785 int
786 kern_chdir(struct thread *td, char *path, enum uio_seg pathseg)
787 {
788 	struct nameidata nd;
789 	int error;
790 
791 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
792 	    pathseg, path, td);
793 	if ((error = namei(&nd)) != 0)
794 		return (error);
795 	if ((error = change_dir(nd.ni_vp, td)) != 0) {
796 		vput(nd.ni_vp);
797 		NDFREE(&nd, NDF_ONLY_PNBUF);
798 		return (error);
799 	}
800 	VOP_UNLOCK(nd.ni_vp, 0);
801 	NDFREE(&nd, NDF_ONLY_PNBUF);
802 	pwd_chdir(td, nd.ni_vp);
803 	return (0);
804 }
805 
806 /*
807  * Change notion of root (``/'') directory.
808  */
809 #ifndef _SYS_SYSPROTO_H_
810 struct chroot_args {
811 	char	*path;
812 };
813 #endif
814 int
815 sys_chroot(td, uap)
816 	struct thread *td;
817 	struct chroot_args /* {
818 		char *path;
819 	} */ *uap;
820 {
821 	struct nameidata nd;
822 	int error;
823 
824 	error = priv_check(td, PRIV_VFS_CHROOT);
825 	if (error != 0)
826 		return (error);
827 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
828 	    UIO_USERSPACE, uap->path, td);
829 	error = namei(&nd);
830 	if (error != 0)
831 		goto error;
832 	error = change_dir(nd.ni_vp, td);
833 	if (error != 0)
834 		goto e_vunlock;
835 #ifdef MAC
836 	error = mac_vnode_check_chroot(td->td_ucred, nd.ni_vp);
837 	if (error != 0)
838 		goto e_vunlock;
839 #endif
840 	VOP_UNLOCK(nd.ni_vp, 0);
841 	error = pwd_chroot(td, nd.ni_vp);
842 	vrele(nd.ni_vp);
843 	NDFREE(&nd, NDF_ONLY_PNBUF);
844 	return (error);
845 e_vunlock:
846 	vput(nd.ni_vp);
847 error:
848 	NDFREE(&nd, NDF_ONLY_PNBUF);
849 	return (error);
850 }
851 
852 /*
853  * Common routine for chroot and chdir.  Callers must provide a locked vnode
854  * instance.
855  */
856 int
857 change_dir(vp, td)
858 	struct vnode *vp;
859 	struct thread *td;
860 {
861 #ifdef MAC
862 	int error;
863 #endif
864 
865 	ASSERT_VOP_LOCKED(vp, "change_dir(): vp not locked");
866 	if (vp->v_type != VDIR)
867 		return (ENOTDIR);
868 #ifdef MAC
869 	error = mac_vnode_check_chdir(td->td_ucred, vp);
870 	if (error != 0)
871 		return (error);
872 #endif
873 	return (VOP_ACCESS(vp, VEXEC, td->td_ucred, td));
874 }
875 
876 static __inline void
877 flags_to_rights(int flags, cap_rights_t *rightsp)
878 {
879 
880 	if (flags & O_EXEC) {
881 		cap_rights_set(rightsp, CAP_FEXECVE);
882 	} else {
883 		switch ((flags & O_ACCMODE)) {
884 		case O_RDONLY:
885 			cap_rights_set(rightsp, CAP_READ);
886 			break;
887 		case O_RDWR:
888 			cap_rights_set(rightsp, CAP_READ);
889 			/* FALLTHROUGH */
890 		case O_WRONLY:
891 			cap_rights_set(rightsp, CAP_WRITE);
892 			if (!(flags & (O_APPEND | O_TRUNC)))
893 				cap_rights_set(rightsp, CAP_SEEK);
894 			break;
895 		}
896 	}
897 
898 	if (flags & O_CREAT)
899 		cap_rights_set(rightsp, CAP_CREATE);
900 
901 	if (flags & O_TRUNC)
902 		cap_rights_set(rightsp, CAP_FTRUNCATE);
903 
904 	if (flags & (O_SYNC | O_FSYNC))
905 		cap_rights_set(rightsp, CAP_FSYNC);
906 
907 	if (flags & (O_EXLOCK | O_SHLOCK))
908 		cap_rights_set(rightsp, CAP_FLOCK);
909 }
910 
911 /*
912  * Check permissions, allocate an open file structure, and call the device
913  * open routine if any.
914  */
915 #ifndef _SYS_SYSPROTO_H_
916 struct open_args {
917 	char	*path;
918 	int	flags;
919 	int	mode;
920 };
921 #endif
922 int
923 sys_open(td, uap)
924 	struct thread *td;
925 	register struct open_args /* {
926 		char *path;
927 		int flags;
928 		int mode;
929 	} */ *uap;
930 {
931 
932 	return (kern_openat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
933 	    uap->flags, uap->mode));
934 }
935 
936 #ifndef _SYS_SYSPROTO_H_
937 struct openat_args {
938 	int	fd;
939 	char	*path;
940 	int	flag;
941 	int	mode;
942 };
943 #endif
944 int
945 sys_openat(struct thread *td, struct openat_args *uap)
946 {
947 
948 	return (kern_openat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
949 	    uap->mode));
950 }
951 
952 int
953 kern_openat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
954     int flags, int mode)
955 {
956 	struct proc *p = td->td_proc;
957 	struct filedesc *fdp = p->p_fd;
958 	struct file *fp;
959 	struct vnode *vp;
960 	struct nameidata nd;
961 	cap_rights_t rights;
962 	int cmode, error, indx;
963 
964 	indx = -1;
965 
966 	AUDIT_ARG_FFLAGS(flags);
967 	AUDIT_ARG_MODE(mode);
968 	/* XXX: audit dirfd */
969 	cap_rights_init(&rights, CAP_LOOKUP);
970 	flags_to_rights(flags, &rights);
971 	/*
972 	 * Only one of the O_EXEC, O_RDONLY, O_WRONLY and O_RDWR flags
973 	 * may be specified.
974 	 */
975 	if (flags & O_EXEC) {
976 		if (flags & O_ACCMODE)
977 			return (EINVAL);
978 	} else if ((flags & O_ACCMODE) == O_ACCMODE) {
979 		return (EINVAL);
980 	} else {
981 		flags = FFLAGS(flags);
982 	}
983 
984 	/*
985 	 * Allocate the file descriptor, but don't install a descriptor yet.
986 	 */
987 	error = falloc_noinstall(td, &fp);
988 	if (error != 0)
989 		return (error);
990 	/*
991 	 * An extra reference on `fp' has been held for us by
992 	 * falloc_noinstall().
993 	 */
994 	/* Set the flags early so the finit in devfs can pick them up. */
995 	fp->f_flag = flags & FMASK;
996 	cmode = ((mode & ~fdp->fd_cmask) & ALLPERMS) & ~S_ISTXT;
997 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, fd,
998 	    &rights, td);
999 	td->td_dupfd = -1;		/* XXX check for fdopen */
1000 	error = vn_open(&nd, &flags, cmode, fp);
1001 	if (error != 0) {
1002 		/*
1003 		 * If the vn_open replaced the method vector, something
1004 		 * wonderous happened deep below and we just pass it up
1005 		 * pretending we know what we do.
1006 		 */
1007 		if (error == ENXIO && fp->f_ops != &badfileops)
1008 			goto success;
1009 
1010 		/*
1011 		 * Handle special fdopen() case. bleh.
1012 		 *
1013 		 * Don't do this for relative (capability) lookups; we don't
1014 		 * understand exactly what would happen, and we don't think
1015 		 * that it ever should.
1016 		 */
1017 		if (nd.ni_strictrelative == 0 &&
1018 		    (error == ENODEV || error == ENXIO) &&
1019 		    td->td_dupfd >= 0) {
1020 			error = dupfdopen(td, fdp, td->td_dupfd, flags, error,
1021 			    &indx);
1022 			if (error == 0)
1023 				goto success;
1024 		}
1025 
1026 		goto bad;
1027 	}
1028 	td->td_dupfd = 0;
1029 	NDFREE(&nd, NDF_ONLY_PNBUF);
1030 	vp = nd.ni_vp;
1031 
1032 	/*
1033 	 * Store the vnode, for any f_type. Typically, the vnode use
1034 	 * count is decremented by direct call to vn_closefile() for
1035 	 * files that switched type in the cdevsw fdopen() method.
1036 	 */
1037 	fp->f_vnode = vp;
1038 	/*
1039 	 * If the file wasn't claimed by devfs bind it to the normal
1040 	 * vnode operations here.
1041 	 */
1042 	if (fp->f_ops == &badfileops) {
1043 		KASSERT(vp->v_type != VFIFO, ("Unexpected fifo."));
1044 		fp->f_seqcount = 1;
1045 		finit(fp, (flags & FMASK) | (fp->f_flag & FHASLOCK),
1046 		    DTYPE_VNODE, vp, &vnops);
1047 	}
1048 
1049 	VOP_UNLOCK(vp, 0);
1050 	if (flags & O_TRUNC) {
1051 		error = fo_truncate(fp, 0, td->td_ucred, td);
1052 		if (error != 0)
1053 			goto bad;
1054 	}
1055 success:
1056 	/*
1057 	 * If we haven't already installed the FD (for dupfdopen), do so now.
1058 	 */
1059 	if (indx == -1) {
1060 		struct filecaps *fcaps;
1061 
1062 #ifdef CAPABILITIES
1063 		if (nd.ni_strictrelative == 1)
1064 			fcaps = &nd.ni_filecaps;
1065 		else
1066 #endif
1067 			fcaps = NULL;
1068 		error = finstall(td, fp, &indx, flags, fcaps);
1069 		/* On success finstall() consumes fcaps. */
1070 		if (error != 0) {
1071 			filecaps_free(&nd.ni_filecaps);
1072 			goto bad;
1073 		}
1074 	} else {
1075 		filecaps_free(&nd.ni_filecaps);
1076 	}
1077 
1078 	/*
1079 	 * Release our private reference, leaving the one associated with
1080 	 * the descriptor table intact.
1081 	 */
1082 	fdrop(fp, td);
1083 	td->td_retval[0] = indx;
1084 	return (0);
1085 bad:
1086 	KASSERT(indx == -1, ("indx=%d, should be -1", indx));
1087 	fdrop(fp, td);
1088 	return (error);
1089 }
1090 
1091 #ifdef COMPAT_43
1092 /*
1093  * Create a file.
1094  */
1095 #ifndef _SYS_SYSPROTO_H_
1096 struct ocreat_args {
1097 	char	*path;
1098 	int	mode;
1099 };
1100 #endif
1101 int
1102 ocreat(td, uap)
1103 	struct thread *td;
1104 	register struct ocreat_args /* {
1105 		char *path;
1106 		int mode;
1107 	} */ *uap;
1108 {
1109 
1110 	return (kern_openat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1111 	    O_WRONLY | O_CREAT | O_TRUNC, uap->mode));
1112 }
1113 #endif /* COMPAT_43 */
1114 
1115 /*
1116  * Create a special file.
1117  */
1118 #ifndef _SYS_SYSPROTO_H_
1119 struct mknod_args {
1120 	char	*path;
1121 	int	mode;
1122 	int	dev;
1123 };
1124 #endif
1125 int
1126 sys_mknod(td, uap)
1127 	struct thread *td;
1128 	register struct mknod_args /* {
1129 		char *path;
1130 		int mode;
1131 		int dev;
1132 	} */ *uap;
1133 {
1134 
1135 	return (kern_mknodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1136 	    uap->mode, uap->dev));
1137 }
1138 
1139 #ifndef _SYS_SYSPROTO_H_
1140 struct mknodat_args {
1141 	int	fd;
1142 	char	*path;
1143 	mode_t	mode;
1144 	dev_t	dev;
1145 };
1146 #endif
1147 int
1148 sys_mknodat(struct thread *td, struct mknodat_args *uap)
1149 {
1150 
1151 	return (kern_mknodat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode,
1152 	    uap->dev));
1153 }
1154 
1155 int
1156 kern_mknodat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1157     int mode, int dev)
1158 {
1159 	struct vnode *vp;
1160 	struct mount *mp;
1161 	struct vattr vattr;
1162 	struct nameidata nd;
1163 	cap_rights_t rights;
1164 	int error, whiteout = 0;
1165 
1166 	AUDIT_ARG_MODE(mode);
1167 	AUDIT_ARG_DEV(dev);
1168 	switch (mode & S_IFMT) {
1169 	case S_IFCHR:
1170 	case S_IFBLK:
1171 		error = priv_check(td, PRIV_VFS_MKNOD_DEV);
1172 		break;
1173 	case S_IFMT:
1174 		error = priv_check(td, PRIV_VFS_MKNOD_BAD);
1175 		break;
1176 	case S_IFWHT:
1177 		error = priv_check(td, PRIV_VFS_MKNOD_WHT);
1178 		break;
1179 	case S_IFIFO:
1180 		if (dev == 0)
1181 			return (kern_mkfifoat(td, fd, path, pathseg, mode));
1182 		/* FALLTHROUGH */
1183 	default:
1184 		error = EINVAL;
1185 		break;
1186 	}
1187 	if (error != 0)
1188 		return (error);
1189 restart:
1190 	bwillwrite();
1191 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
1192 	    NOCACHE, pathseg, path, fd, cap_rights_init(&rights, CAP_MKNODAT),
1193 	    td);
1194 	if ((error = namei(&nd)) != 0)
1195 		return (error);
1196 	vp = nd.ni_vp;
1197 	if (vp != NULL) {
1198 		NDFREE(&nd, NDF_ONLY_PNBUF);
1199 		if (vp == nd.ni_dvp)
1200 			vrele(nd.ni_dvp);
1201 		else
1202 			vput(nd.ni_dvp);
1203 		vrele(vp);
1204 		return (EEXIST);
1205 	} else {
1206 		VATTR_NULL(&vattr);
1207 		vattr.va_mode = (mode & ALLPERMS) &
1208 		    ~td->td_proc->p_fd->fd_cmask;
1209 		vattr.va_rdev = dev;
1210 		whiteout = 0;
1211 
1212 		switch (mode & S_IFMT) {
1213 		case S_IFMT:	/* used by badsect to flag bad sectors */
1214 			vattr.va_type = VBAD;
1215 			break;
1216 		case S_IFCHR:
1217 			vattr.va_type = VCHR;
1218 			break;
1219 		case S_IFBLK:
1220 			vattr.va_type = VBLK;
1221 			break;
1222 		case S_IFWHT:
1223 			whiteout = 1;
1224 			break;
1225 		default:
1226 			panic("kern_mknod: invalid mode");
1227 		}
1228 	}
1229 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1230 		NDFREE(&nd, NDF_ONLY_PNBUF);
1231 		vput(nd.ni_dvp);
1232 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1233 			return (error);
1234 		goto restart;
1235 	}
1236 #ifdef MAC
1237 	if (error == 0 && !whiteout)
1238 		error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp,
1239 		    &nd.ni_cnd, &vattr);
1240 #endif
1241 	if (error == 0) {
1242 		if (whiteout)
1243 			error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE);
1244 		else {
1245 			error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp,
1246 						&nd.ni_cnd, &vattr);
1247 			if (error == 0)
1248 				vput(nd.ni_vp);
1249 		}
1250 	}
1251 	NDFREE(&nd, NDF_ONLY_PNBUF);
1252 	vput(nd.ni_dvp);
1253 	vn_finished_write(mp);
1254 	return (error);
1255 }
1256 
1257 /*
1258  * Create a named pipe.
1259  */
1260 #ifndef _SYS_SYSPROTO_H_
1261 struct mkfifo_args {
1262 	char	*path;
1263 	int	mode;
1264 };
1265 #endif
1266 int
1267 sys_mkfifo(td, uap)
1268 	struct thread *td;
1269 	register struct mkfifo_args /* {
1270 		char *path;
1271 		int mode;
1272 	} */ *uap;
1273 {
1274 
1275 	return (kern_mkfifoat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1276 	    uap->mode));
1277 }
1278 
1279 #ifndef _SYS_SYSPROTO_H_
1280 struct mkfifoat_args {
1281 	int	fd;
1282 	char	*path;
1283 	mode_t	mode;
1284 };
1285 #endif
1286 int
1287 sys_mkfifoat(struct thread *td, struct mkfifoat_args *uap)
1288 {
1289 
1290 	return (kern_mkfifoat(td, uap->fd, uap->path, UIO_USERSPACE,
1291 	    uap->mode));
1292 }
1293 
1294 int
1295 kern_mkfifoat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1296     int mode)
1297 {
1298 	struct mount *mp;
1299 	struct vattr vattr;
1300 	struct nameidata nd;
1301 	cap_rights_t rights;
1302 	int error;
1303 
1304 	AUDIT_ARG_MODE(mode);
1305 restart:
1306 	bwillwrite();
1307 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
1308 	    NOCACHE, pathseg, path, fd, cap_rights_init(&rights, CAP_MKFIFOAT),
1309 	    td);
1310 	if ((error = namei(&nd)) != 0)
1311 		return (error);
1312 	if (nd.ni_vp != NULL) {
1313 		NDFREE(&nd, NDF_ONLY_PNBUF);
1314 		if (nd.ni_vp == nd.ni_dvp)
1315 			vrele(nd.ni_dvp);
1316 		else
1317 			vput(nd.ni_dvp);
1318 		vrele(nd.ni_vp);
1319 		return (EEXIST);
1320 	}
1321 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1322 		NDFREE(&nd, NDF_ONLY_PNBUF);
1323 		vput(nd.ni_dvp);
1324 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1325 			return (error);
1326 		goto restart;
1327 	}
1328 	VATTR_NULL(&vattr);
1329 	vattr.va_type = VFIFO;
1330 	vattr.va_mode = (mode & ALLPERMS) & ~td->td_proc->p_fd->fd_cmask;
1331 #ifdef MAC
1332 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1333 	    &vattr);
1334 	if (error != 0)
1335 		goto out;
1336 #endif
1337 	error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
1338 	if (error == 0)
1339 		vput(nd.ni_vp);
1340 #ifdef MAC
1341 out:
1342 #endif
1343 	vput(nd.ni_dvp);
1344 	vn_finished_write(mp);
1345 	NDFREE(&nd, NDF_ONLY_PNBUF);
1346 	return (error);
1347 }
1348 
1349 /*
1350  * Make a hard file link.
1351  */
1352 #ifndef _SYS_SYSPROTO_H_
1353 struct link_args {
1354 	char	*path;
1355 	char	*link;
1356 };
1357 #endif
1358 int
1359 sys_link(td, uap)
1360 	struct thread *td;
1361 	register struct link_args /* {
1362 		char *path;
1363 		char *link;
1364 	} */ *uap;
1365 {
1366 
1367 	return (kern_linkat(td, AT_FDCWD, AT_FDCWD, uap->path, uap->link,
1368 	    UIO_USERSPACE, FOLLOW));
1369 }
1370 
1371 #ifndef _SYS_SYSPROTO_H_
1372 struct linkat_args {
1373 	int	fd1;
1374 	char	*path1;
1375 	int	fd2;
1376 	char	*path2;
1377 	int	flag;
1378 };
1379 #endif
1380 int
1381 sys_linkat(struct thread *td, struct linkat_args *uap)
1382 {
1383 	int flag;
1384 
1385 	flag = uap->flag;
1386 	if (flag & ~AT_SYMLINK_FOLLOW)
1387 		return (EINVAL);
1388 
1389 	return (kern_linkat(td, uap->fd1, uap->fd2, uap->path1, uap->path2,
1390 	    UIO_USERSPACE, (flag & AT_SYMLINK_FOLLOW) ? FOLLOW : NOFOLLOW));
1391 }
1392 
1393 int hardlink_check_uid = 0;
1394 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_uid, CTLFLAG_RW,
1395     &hardlink_check_uid, 0,
1396     "Unprivileged processes cannot create hard links to files owned by other "
1397     "users");
1398 static int hardlink_check_gid = 0;
1399 SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_gid, CTLFLAG_RW,
1400     &hardlink_check_gid, 0,
1401     "Unprivileged processes cannot create hard links to files owned by other "
1402     "groups");
1403 
1404 static int
1405 can_hardlink(struct vnode *vp, struct ucred *cred)
1406 {
1407 	struct vattr va;
1408 	int error;
1409 
1410 	if (!hardlink_check_uid && !hardlink_check_gid)
1411 		return (0);
1412 
1413 	error = VOP_GETATTR(vp, &va, cred);
1414 	if (error != 0)
1415 		return (error);
1416 
1417 	if (hardlink_check_uid && cred->cr_uid != va.va_uid) {
1418 		error = priv_check_cred(cred, PRIV_VFS_LINK, 0);
1419 		if (error != 0)
1420 			return (error);
1421 	}
1422 
1423 	if (hardlink_check_gid && !groupmember(va.va_gid, cred)) {
1424 		error = priv_check_cred(cred, PRIV_VFS_LINK, 0);
1425 		if (error != 0)
1426 			return (error);
1427 	}
1428 
1429 	return (0);
1430 }
1431 
1432 int
1433 kern_linkat(struct thread *td, int fd1, int fd2, char *path1, char *path2,
1434     enum uio_seg segflg, int follow)
1435 {
1436 	struct vnode *vp;
1437 	struct mount *mp;
1438 	struct nameidata nd;
1439 	cap_rights_t rights;
1440 	int error;
1441 
1442 again:
1443 	bwillwrite();
1444 	NDINIT_AT(&nd, LOOKUP, follow | AUDITVNODE1, segflg, path1, fd1, td);
1445 
1446 	if ((error = namei(&nd)) != 0)
1447 		return (error);
1448 	NDFREE(&nd, NDF_ONLY_PNBUF);
1449 	vp = nd.ni_vp;
1450 	if (vp->v_type == VDIR) {
1451 		vrele(vp);
1452 		return (EPERM);		/* POSIX */
1453 	}
1454 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE2 |
1455 	    NOCACHE, segflg, path2, fd2, cap_rights_init(&rights, CAP_LINKAT),
1456 	    td);
1457 	if ((error = namei(&nd)) == 0) {
1458 		if (nd.ni_vp != NULL) {
1459 			NDFREE(&nd, NDF_ONLY_PNBUF);
1460 			if (nd.ni_dvp == nd.ni_vp)
1461 				vrele(nd.ni_dvp);
1462 			else
1463 				vput(nd.ni_dvp);
1464 			vrele(nd.ni_vp);
1465 			vrele(vp);
1466 			return (EEXIST);
1467 		} else if (nd.ni_dvp->v_mount != vp->v_mount) {
1468 			/*
1469 			 * Cross-device link.  No need to recheck
1470 			 * vp->v_type, since it cannot change, except
1471 			 * to VBAD.
1472 			 */
1473 			NDFREE(&nd, NDF_ONLY_PNBUF);
1474 			vput(nd.ni_dvp);
1475 			vrele(vp);
1476 			return (EXDEV);
1477 		} else if ((error = vn_lock(vp, LK_EXCLUSIVE)) == 0) {
1478 			error = can_hardlink(vp, td->td_ucred);
1479 #ifdef MAC
1480 			if (error == 0)
1481 				error = mac_vnode_check_link(td->td_ucred,
1482 				    nd.ni_dvp, vp, &nd.ni_cnd);
1483 #endif
1484 			if (error != 0) {
1485 				vput(vp);
1486 				vput(nd.ni_dvp);
1487 				NDFREE(&nd, NDF_ONLY_PNBUF);
1488 				return (error);
1489 			}
1490 			error = vn_start_write(vp, &mp, V_NOWAIT);
1491 			if (error != 0) {
1492 				vput(vp);
1493 				vput(nd.ni_dvp);
1494 				NDFREE(&nd, NDF_ONLY_PNBUF);
1495 				error = vn_start_write(NULL, &mp,
1496 				    V_XSLEEP | PCATCH);
1497 				if (error != 0)
1498 					return (error);
1499 				goto again;
1500 			}
1501 			error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
1502 			VOP_UNLOCK(vp, 0);
1503 			vput(nd.ni_dvp);
1504 			vn_finished_write(mp);
1505 			NDFREE(&nd, NDF_ONLY_PNBUF);
1506 		} else {
1507 			vput(nd.ni_dvp);
1508 			NDFREE(&nd, NDF_ONLY_PNBUF);
1509 			vrele(vp);
1510 			goto again;
1511 		}
1512 	}
1513 	vrele(vp);
1514 	return (error);
1515 }
1516 
1517 /*
1518  * Make a symbolic link.
1519  */
1520 #ifndef _SYS_SYSPROTO_H_
1521 struct symlink_args {
1522 	char	*path;
1523 	char	*link;
1524 };
1525 #endif
1526 int
1527 sys_symlink(td, uap)
1528 	struct thread *td;
1529 	register struct symlink_args /* {
1530 		char *path;
1531 		char *link;
1532 	} */ *uap;
1533 {
1534 
1535 	return (kern_symlinkat(td, uap->path, AT_FDCWD, uap->link,
1536 	    UIO_USERSPACE));
1537 }
1538 
1539 #ifndef _SYS_SYSPROTO_H_
1540 struct symlinkat_args {
1541 	char	*path;
1542 	int	fd;
1543 	char	*path2;
1544 };
1545 #endif
1546 int
1547 sys_symlinkat(struct thread *td, struct symlinkat_args *uap)
1548 {
1549 
1550 	return (kern_symlinkat(td, uap->path1, uap->fd, uap->path2,
1551 	    UIO_USERSPACE));
1552 }
1553 
1554 int
1555 kern_symlinkat(struct thread *td, char *path1, int fd, char *path2,
1556     enum uio_seg segflg)
1557 {
1558 	struct mount *mp;
1559 	struct vattr vattr;
1560 	char *syspath;
1561 	struct nameidata nd;
1562 	int error;
1563 	cap_rights_t rights;
1564 
1565 	if (segflg == UIO_SYSSPACE) {
1566 		syspath = path1;
1567 	} else {
1568 		syspath = uma_zalloc(namei_zone, M_WAITOK);
1569 		if ((error = copyinstr(path1, syspath, MAXPATHLEN, NULL)) != 0)
1570 			goto out;
1571 	}
1572 	AUDIT_ARG_TEXT(syspath);
1573 restart:
1574 	bwillwrite();
1575 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
1576 	    NOCACHE, segflg, path2, fd, cap_rights_init(&rights, CAP_SYMLINKAT),
1577 	    td);
1578 	if ((error = namei(&nd)) != 0)
1579 		goto out;
1580 	if (nd.ni_vp) {
1581 		NDFREE(&nd, NDF_ONLY_PNBUF);
1582 		if (nd.ni_vp == nd.ni_dvp)
1583 			vrele(nd.ni_dvp);
1584 		else
1585 			vput(nd.ni_dvp);
1586 		vrele(nd.ni_vp);
1587 		error = EEXIST;
1588 		goto out;
1589 	}
1590 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1591 		NDFREE(&nd, NDF_ONLY_PNBUF);
1592 		vput(nd.ni_dvp);
1593 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1594 			goto out;
1595 		goto restart;
1596 	}
1597 	VATTR_NULL(&vattr);
1598 	vattr.va_mode = ACCESSPERMS &~ td->td_proc->p_fd->fd_cmask;
1599 #ifdef MAC
1600 	vattr.va_type = VLNK;
1601 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
1602 	    &vattr);
1603 	if (error != 0)
1604 		goto out2;
1605 #endif
1606 	error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, syspath);
1607 	if (error == 0)
1608 		vput(nd.ni_vp);
1609 #ifdef MAC
1610 out2:
1611 #endif
1612 	NDFREE(&nd, NDF_ONLY_PNBUF);
1613 	vput(nd.ni_dvp);
1614 	vn_finished_write(mp);
1615 out:
1616 	if (segflg != UIO_SYSSPACE)
1617 		uma_zfree(namei_zone, syspath);
1618 	return (error);
1619 }
1620 
1621 /*
1622  * Delete a whiteout from the filesystem.
1623  */
1624 int
1625 sys_undelete(td, uap)
1626 	struct thread *td;
1627 	register struct undelete_args /* {
1628 		char *path;
1629 	} */ *uap;
1630 {
1631 	struct mount *mp;
1632 	struct nameidata nd;
1633 	int error;
1634 
1635 restart:
1636 	bwillwrite();
1637 	NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | AUDITVNODE1,
1638 	    UIO_USERSPACE, uap->path, td);
1639 	error = namei(&nd);
1640 	if (error != 0)
1641 		return (error);
1642 
1643 	if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) {
1644 		NDFREE(&nd, NDF_ONLY_PNBUF);
1645 		if (nd.ni_vp == nd.ni_dvp)
1646 			vrele(nd.ni_dvp);
1647 		else
1648 			vput(nd.ni_dvp);
1649 		if (nd.ni_vp)
1650 			vrele(nd.ni_vp);
1651 		return (EEXIST);
1652 	}
1653 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1654 		NDFREE(&nd, NDF_ONLY_PNBUF);
1655 		vput(nd.ni_dvp);
1656 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
1657 			return (error);
1658 		goto restart;
1659 	}
1660 	error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE);
1661 	NDFREE(&nd, NDF_ONLY_PNBUF);
1662 	vput(nd.ni_dvp);
1663 	vn_finished_write(mp);
1664 	return (error);
1665 }
1666 
1667 /*
1668  * Delete a name from the filesystem.
1669  */
1670 #ifndef _SYS_SYSPROTO_H_
1671 struct unlink_args {
1672 	char	*path;
1673 };
1674 #endif
1675 int
1676 sys_unlink(td, uap)
1677 	struct thread *td;
1678 	struct unlink_args /* {
1679 		char *path;
1680 	} */ *uap;
1681 {
1682 
1683 	return (kern_unlinkat(td, AT_FDCWD, uap->path, UIO_USERSPACE, 0));
1684 }
1685 
1686 #ifndef _SYS_SYSPROTO_H_
1687 struct unlinkat_args {
1688 	int	fd;
1689 	char	*path;
1690 	int	flag;
1691 };
1692 #endif
1693 int
1694 sys_unlinkat(struct thread *td, struct unlinkat_args *uap)
1695 {
1696 	int flag = uap->flag;
1697 	int fd = uap->fd;
1698 	char *path = uap->path;
1699 
1700 	if (flag & ~AT_REMOVEDIR)
1701 		return (EINVAL);
1702 
1703 	if (flag & AT_REMOVEDIR)
1704 		return (kern_rmdirat(td, fd, path, UIO_USERSPACE));
1705 	else
1706 		return (kern_unlinkat(td, fd, path, UIO_USERSPACE, 0));
1707 }
1708 
1709 int
1710 kern_unlinkat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1711     ino_t oldinum)
1712 {
1713 	struct mount *mp;
1714 	struct vnode *vp;
1715 	struct nameidata nd;
1716 	struct stat sb;
1717 	cap_rights_t rights;
1718 	int error;
1719 
1720 restart:
1721 	bwillwrite();
1722 	NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1,
1723 	    pathseg, path, fd, cap_rights_init(&rights, CAP_UNLINKAT), td);
1724 	if ((error = namei(&nd)) != 0)
1725 		return (error == EINVAL ? EPERM : error);
1726 	vp = nd.ni_vp;
1727 	if (vp->v_type == VDIR && oldinum == 0) {
1728 		error = EPERM;		/* POSIX */
1729 	} else if (oldinum != 0 &&
1730 		  ((error = vn_stat(vp, &sb, td->td_ucred, NOCRED, td)) == 0) &&
1731 		  sb.st_ino != oldinum) {
1732 			error = EIDRM;	/* Identifier removed */
1733 	} else {
1734 		/*
1735 		 * The root of a mounted filesystem cannot be deleted.
1736 		 *
1737 		 * XXX: can this only be a VDIR case?
1738 		 */
1739 		if (vp->v_vflag & VV_ROOT)
1740 			error = EBUSY;
1741 	}
1742 	if (error == 0) {
1743 		if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
1744 			NDFREE(&nd, NDF_ONLY_PNBUF);
1745 			vput(nd.ni_dvp);
1746 			if (vp == nd.ni_dvp)
1747 				vrele(vp);
1748 			else
1749 				vput(vp);
1750 			if ((error = vn_start_write(NULL, &mp,
1751 			    V_XSLEEP | PCATCH)) != 0)
1752 				return (error);
1753 			goto restart;
1754 		}
1755 #ifdef MAC
1756 		error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
1757 		    &nd.ni_cnd);
1758 		if (error != 0)
1759 			goto out;
1760 #endif
1761 		vfs_notify_upper(vp, VFS_NOTIFY_UPPER_UNLINK);
1762 		error = VOP_REMOVE(nd.ni_dvp, vp, &nd.ni_cnd);
1763 #ifdef MAC
1764 out:
1765 #endif
1766 		vn_finished_write(mp);
1767 	}
1768 	NDFREE(&nd, NDF_ONLY_PNBUF);
1769 	vput(nd.ni_dvp);
1770 	if (vp == nd.ni_dvp)
1771 		vrele(vp);
1772 	else
1773 		vput(vp);
1774 	return (error);
1775 }
1776 
1777 /*
1778  * Reposition read/write file offset.
1779  */
1780 #ifndef _SYS_SYSPROTO_H_
1781 struct lseek_args {
1782 	int	fd;
1783 	int	pad;
1784 	off_t	offset;
1785 	int	whence;
1786 };
1787 #endif
1788 int
1789 sys_lseek(td, uap)
1790 	struct thread *td;
1791 	register struct lseek_args /* {
1792 		int fd;
1793 		int pad;
1794 		off_t offset;
1795 		int whence;
1796 	} */ *uap;
1797 {
1798 	struct file *fp;
1799 	cap_rights_t rights;
1800 	int error;
1801 
1802 	AUDIT_ARG_FD(uap->fd);
1803 	error = fget(td, uap->fd, cap_rights_init(&rights, CAP_SEEK), &fp);
1804 	if (error != 0)
1805 		return (error);
1806 	error = (fp->f_ops->fo_flags & DFLAG_SEEKABLE) != 0 ?
1807 	    fo_seek(fp, uap->offset, uap->whence, td) : ESPIPE;
1808 	fdrop(fp, td);
1809 	return (error);
1810 }
1811 
1812 #if defined(COMPAT_43)
1813 /*
1814  * Reposition read/write file offset.
1815  */
1816 #ifndef _SYS_SYSPROTO_H_
1817 struct olseek_args {
1818 	int	fd;
1819 	long	offset;
1820 	int	whence;
1821 };
1822 #endif
1823 int
1824 olseek(td, uap)
1825 	struct thread *td;
1826 	register struct olseek_args /* {
1827 		int fd;
1828 		long offset;
1829 		int whence;
1830 	} */ *uap;
1831 {
1832 	struct lseek_args /* {
1833 		int fd;
1834 		int pad;
1835 		off_t offset;
1836 		int whence;
1837 	} */ nuap;
1838 
1839 	nuap.fd = uap->fd;
1840 	nuap.offset = uap->offset;
1841 	nuap.whence = uap->whence;
1842 	return (sys_lseek(td, &nuap));
1843 }
1844 #endif /* COMPAT_43 */
1845 
1846 #if defined(COMPAT_FREEBSD6)
1847 /* Version with the 'pad' argument */
1848 int
1849 freebsd6_lseek(td, uap)
1850 	struct thread *td;
1851 	register struct freebsd6_lseek_args *uap;
1852 {
1853 	struct lseek_args ouap;
1854 
1855 	ouap.fd = uap->fd;
1856 	ouap.offset = uap->offset;
1857 	ouap.whence = uap->whence;
1858 	return (sys_lseek(td, &ouap));
1859 }
1860 #endif
1861 
1862 /*
1863  * Check access permissions using passed credentials.
1864  */
1865 static int
1866 vn_access(vp, user_flags, cred, td)
1867 	struct vnode	*vp;
1868 	int		user_flags;
1869 	struct ucred	*cred;
1870 	struct thread	*td;
1871 {
1872 	accmode_t accmode;
1873 	int error;
1874 
1875 	/* Flags == 0 means only check for existence. */
1876 	if (user_flags == 0)
1877 		return (0);
1878 
1879 	accmode = 0;
1880 	if (user_flags & R_OK)
1881 		accmode |= VREAD;
1882 	if (user_flags & W_OK)
1883 		accmode |= VWRITE;
1884 	if (user_flags & X_OK)
1885 		accmode |= VEXEC;
1886 #ifdef MAC
1887 	error = mac_vnode_check_access(cred, vp, accmode);
1888 	if (error != 0)
1889 		return (error);
1890 #endif
1891 	if ((accmode & VWRITE) == 0 || (error = vn_writechk(vp)) == 0)
1892 		error = VOP_ACCESS(vp, accmode, cred, td);
1893 	return (error);
1894 }
1895 
1896 /*
1897  * Check access permissions using "real" credentials.
1898  */
1899 #ifndef _SYS_SYSPROTO_H_
1900 struct access_args {
1901 	char	*path;
1902 	int	amode;
1903 };
1904 #endif
1905 int
1906 sys_access(td, uap)
1907 	struct thread *td;
1908 	register struct access_args /* {
1909 		char *path;
1910 		int amode;
1911 	} */ *uap;
1912 {
1913 
1914 	return (kern_accessat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
1915 	    0, uap->amode));
1916 }
1917 
1918 #ifndef _SYS_SYSPROTO_H_
1919 struct faccessat_args {
1920 	int	dirfd;
1921 	char	*path;
1922 	int	amode;
1923 	int	flag;
1924 }
1925 #endif
1926 int
1927 sys_faccessat(struct thread *td, struct faccessat_args *uap)
1928 {
1929 
1930 	return (kern_accessat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
1931 	    uap->amode));
1932 }
1933 
1934 int
1935 kern_accessat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
1936     int flag, int amode)
1937 {
1938 	struct ucred *cred, *usecred;
1939 	struct vnode *vp;
1940 	struct nameidata nd;
1941 	cap_rights_t rights;
1942 	int error;
1943 
1944 	if (flag & ~AT_EACCESS)
1945 		return (EINVAL);
1946 	if (amode != F_OK && (amode & ~(R_OK | W_OK | X_OK)) != 0)
1947 		return (EINVAL);
1948 
1949 	/*
1950 	 * Create and modify a temporary credential instead of one that
1951 	 * is potentially shared (if we need one).
1952 	 */
1953 	cred = td->td_ucred;
1954 	if ((flag & AT_EACCESS) == 0 &&
1955 	    ((cred->cr_uid != cred->cr_ruid ||
1956 	    cred->cr_rgid != cred->cr_groups[0]))) {
1957 		usecred = crdup(cred);
1958 		usecred->cr_uid = cred->cr_ruid;
1959 		usecred->cr_groups[0] = cred->cr_rgid;
1960 		td->td_ucred = usecred;
1961 	} else
1962 		usecred = cred;
1963 	AUDIT_ARG_VALUE(amode);
1964 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF |
1965 	    AUDITVNODE1, pathseg, path, fd, cap_rights_init(&rights, CAP_FSTAT),
1966 	    td);
1967 	if ((error = namei(&nd)) != 0)
1968 		goto out;
1969 	vp = nd.ni_vp;
1970 
1971 	error = vn_access(vp, amode, usecred, td);
1972 	NDFREE(&nd, NDF_ONLY_PNBUF);
1973 	vput(vp);
1974 out:
1975 	if (usecred != cred) {
1976 		td->td_ucred = cred;
1977 		crfree(usecred);
1978 	}
1979 	return (error);
1980 }
1981 
1982 /*
1983  * Check access permissions using "effective" credentials.
1984  */
1985 #ifndef _SYS_SYSPROTO_H_
1986 struct eaccess_args {
1987 	char	*path;
1988 	int	amode;
1989 };
1990 #endif
1991 int
1992 sys_eaccess(td, uap)
1993 	struct thread *td;
1994 	register struct eaccess_args /* {
1995 		char *path;
1996 		int amode;
1997 	} */ *uap;
1998 {
1999 
2000 	return (kern_accessat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2001 	    AT_EACCESS, uap->amode));
2002 }
2003 
2004 #if defined(COMPAT_43)
2005 /*
2006  * Get file status; this version follows links.
2007  */
2008 #ifndef _SYS_SYSPROTO_H_
2009 struct ostat_args {
2010 	char	*path;
2011 	struct ostat *ub;
2012 };
2013 #endif
2014 int
2015 ostat(td, uap)
2016 	struct thread *td;
2017 	register struct ostat_args /* {
2018 		char *path;
2019 		struct ostat *ub;
2020 	} */ *uap;
2021 {
2022 	struct stat sb;
2023 	struct ostat osb;
2024 	int error;
2025 
2026 	error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE,
2027 	    &sb, NULL);
2028 	if (error != 0)
2029 		return (error);
2030 	cvtstat(&sb, &osb);
2031 	return (copyout(&osb, uap->ub, sizeof (osb)));
2032 }
2033 
2034 /*
2035  * Get file status; this version does not follow links.
2036  */
2037 #ifndef _SYS_SYSPROTO_H_
2038 struct olstat_args {
2039 	char	*path;
2040 	struct ostat *ub;
2041 };
2042 #endif
2043 int
2044 olstat(td, uap)
2045 	struct thread *td;
2046 	register struct olstat_args /* {
2047 		char *path;
2048 		struct ostat *ub;
2049 	} */ *uap;
2050 {
2051 	struct stat sb;
2052 	struct ostat osb;
2053 	int error;
2054 
2055 	error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2056 	    UIO_USERSPACE, &sb, NULL);
2057 	if (error != 0)
2058 		return (error);
2059 	cvtstat(&sb, &osb);
2060 	return (copyout(&osb, uap->ub, sizeof (osb)));
2061 }
2062 
2063 /*
2064  * Convert from an old to a new stat structure.
2065  */
2066 void
2067 cvtstat(st, ost)
2068 	struct stat *st;
2069 	struct ostat *ost;
2070 {
2071 
2072 	ost->st_dev = st->st_dev;
2073 	ost->st_ino = st->st_ino;
2074 	ost->st_mode = st->st_mode;
2075 	ost->st_nlink = st->st_nlink;
2076 	ost->st_uid = st->st_uid;
2077 	ost->st_gid = st->st_gid;
2078 	ost->st_rdev = st->st_rdev;
2079 	if (st->st_size < (quad_t)1 << 32)
2080 		ost->st_size = st->st_size;
2081 	else
2082 		ost->st_size = -2;
2083 	ost->st_atim = st->st_atim;
2084 	ost->st_mtim = st->st_mtim;
2085 	ost->st_ctim = st->st_ctim;
2086 	ost->st_blksize = st->st_blksize;
2087 	ost->st_blocks = st->st_blocks;
2088 	ost->st_flags = st->st_flags;
2089 	ost->st_gen = st->st_gen;
2090 }
2091 #endif /* COMPAT_43 */
2092 
2093 /*
2094  * Get file status; this version follows links.
2095  */
2096 #ifndef _SYS_SYSPROTO_H_
2097 struct stat_args {
2098 	char	*path;
2099 	struct stat *ub;
2100 };
2101 #endif
2102 int
2103 sys_stat(td, uap)
2104 	struct thread *td;
2105 	register struct stat_args /* {
2106 		char *path;
2107 		struct stat *ub;
2108 	} */ *uap;
2109 {
2110 	struct stat sb;
2111 	int error;
2112 
2113 	error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE,
2114 	    &sb, NULL);
2115 	if (error == 0)
2116 		error = copyout(&sb, uap->ub, sizeof (sb));
2117 	return (error);
2118 }
2119 
2120 #ifndef _SYS_SYSPROTO_H_
2121 struct fstatat_args {
2122 	int	fd;
2123 	char	*path;
2124 	struct stat	*buf;
2125 	int	flag;
2126 }
2127 #endif
2128 int
2129 sys_fstatat(struct thread *td, struct fstatat_args *uap)
2130 {
2131 	struct stat sb;
2132 	int error;
2133 
2134 	error = kern_statat(td, uap->flag, uap->fd, uap->path,
2135 	    UIO_USERSPACE, &sb, NULL);
2136 	if (error == 0)
2137 		error = copyout(&sb, uap->buf, sizeof (sb));
2138 	return (error);
2139 }
2140 
2141 int
2142 kern_statat(struct thread *td, int flag, int fd, char *path,
2143     enum uio_seg pathseg, struct stat *sbp,
2144     void (*hook)(struct vnode *vp, struct stat *sbp))
2145 {
2146 	struct nameidata nd;
2147 	struct stat sb;
2148 	cap_rights_t rights;
2149 	int error;
2150 
2151 	if (flag & ~AT_SYMLINK_NOFOLLOW)
2152 		return (EINVAL);
2153 
2154 	NDINIT_ATRIGHTS(&nd, LOOKUP, ((flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW :
2155 	    FOLLOW) | LOCKSHARED | LOCKLEAF | AUDITVNODE1, pathseg, path, fd,
2156 	    cap_rights_init(&rights, CAP_FSTAT), td);
2157 
2158 	if ((error = namei(&nd)) != 0)
2159 		return (error);
2160 	error = vn_stat(nd.ni_vp, &sb, td->td_ucred, NOCRED, td);
2161 	if (error == 0) {
2162 		SDT_PROBE(vfs, , stat, mode, path, sb.st_mode, 0, 0, 0);
2163 		if (S_ISREG(sb.st_mode))
2164 			SDT_PROBE(vfs, , stat, reg, path, pathseg, 0, 0, 0);
2165 		if (__predict_false(hook != NULL))
2166 			hook(nd.ni_vp, &sb);
2167 	}
2168 	NDFREE(&nd, NDF_ONLY_PNBUF);
2169 	vput(nd.ni_vp);
2170 	if (error != 0)
2171 		return (error);
2172 	*sbp = sb;
2173 #ifdef KTRACE
2174 	if (KTRPOINT(td, KTR_STRUCT))
2175 		ktrstat(&sb);
2176 #endif
2177 	return (0);
2178 }
2179 
2180 /*
2181  * Get file status; this version does not follow links.
2182  */
2183 #ifndef _SYS_SYSPROTO_H_
2184 struct lstat_args {
2185 	char	*path;
2186 	struct stat *ub;
2187 };
2188 #endif
2189 int
2190 sys_lstat(td, uap)
2191 	struct thread *td;
2192 	register struct lstat_args /* {
2193 		char *path;
2194 		struct stat *ub;
2195 	} */ *uap;
2196 {
2197 	struct stat sb;
2198 	int error;
2199 
2200 	error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2201 	    UIO_USERSPACE, &sb, NULL);
2202 	if (error == 0)
2203 		error = copyout(&sb, uap->ub, sizeof (sb));
2204 	return (error);
2205 }
2206 
2207 /*
2208  * Implementation of the NetBSD [l]stat() functions.
2209  */
2210 void
2211 cvtnstat(sb, nsb)
2212 	struct stat *sb;
2213 	struct nstat *nsb;
2214 {
2215 
2216 	bzero(nsb, sizeof *nsb);
2217 	nsb->st_dev = sb->st_dev;
2218 	nsb->st_ino = sb->st_ino;
2219 	nsb->st_mode = sb->st_mode;
2220 	nsb->st_nlink = sb->st_nlink;
2221 	nsb->st_uid = sb->st_uid;
2222 	nsb->st_gid = sb->st_gid;
2223 	nsb->st_rdev = sb->st_rdev;
2224 	nsb->st_atim = sb->st_atim;
2225 	nsb->st_mtim = sb->st_mtim;
2226 	nsb->st_ctim = sb->st_ctim;
2227 	nsb->st_size = sb->st_size;
2228 	nsb->st_blocks = sb->st_blocks;
2229 	nsb->st_blksize = sb->st_blksize;
2230 	nsb->st_flags = sb->st_flags;
2231 	nsb->st_gen = sb->st_gen;
2232 	nsb->st_birthtim = sb->st_birthtim;
2233 }
2234 
2235 #ifndef _SYS_SYSPROTO_H_
2236 struct nstat_args {
2237 	char	*path;
2238 	struct nstat *ub;
2239 };
2240 #endif
2241 int
2242 sys_nstat(td, uap)
2243 	struct thread *td;
2244 	register struct nstat_args /* {
2245 		char *path;
2246 		struct nstat *ub;
2247 	} */ *uap;
2248 {
2249 	struct stat sb;
2250 	struct nstat nsb;
2251 	int error;
2252 
2253 	error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE,
2254 	    &sb, NULL);
2255 	if (error != 0)
2256 		return (error);
2257 	cvtnstat(&sb, &nsb);
2258 	return (copyout(&nsb, uap->ub, sizeof (nsb)));
2259 }
2260 
2261 /*
2262  * NetBSD lstat.  Get file status; this version does not follow links.
2263  */
2264 #ifndef _SYS_SYSPROTO_H_
2265 struct lstat_args {
2266 	char	*path;
2267 	struct stat *ub;
2268 };
2269 #endif
2270 int
2271 sys_nlstat(td, uap)
2272 	struct thread *td;
2273 	register struct nlstat_args /* {
2274 		char *path;
2275 		struct nstat *ub;
2276 	} */ *uap;
2277 {
2278 	struct stat sb;
2279 	struct nstat nsb;
2280 	int error;
2281 
2282 	error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
2283 	    UIO_USERSPACE, &sb, NULL);
2284 	if (error != 0)
2285 		return (error);
2286 	cvtnstat(&sb, &nsb);
2287 	return (copyout(&nsb, uap->ub, sizeof (nsb)));
2288 }
2289 
2290 /*
2291  * Get configurable pathname variables.
2292  */
2293 #ifndef _SYS_SYSPROTO_H_
2294 struct pathconf_args {
2295 	char	*path;
2296 	int	name;
2297 };
2298 #endif
2299 int
2300 sys_pathconf(td, uap)
2301 	struct thread *td;
2302 	register struct pathconf_args /* {
2303 		char *path;
2304 		int name;
2305 	} */ *uap;
2306 {
2307 
2308 	return (kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name, FOLLOW));
2309 }
2310 
2311 #ifndef _SYS_SYSPROTO_H_
2312 struct lpathconf_args {
2313 	char	*path;
2314 	int	name;
2315 };
2316 #endif
2317 int
2318 sys_lpathconf(td, uap)
2319 	struct thread *td;
2320 	register struct lpathconf_args /* {
2321 		char *path;
2322 		int name;
2323 	} */ *uap;
2324 {
2325 
2326 	return (kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name,
2327 	    NOFOLLOW));
2328 }
2329 
2330 int
2331 kern_pathconf(struct thread *td, char *path, enum uio_seg pathseg, int name,
2332     u_long flags)
2333 {
2334 	struct nameidata nd;
2335 	int error;
2336 
2337 	NDINIT(&nd, LOOKUP, LOCKSHARED | LOCKLEAF | AUDITVNODE1 | flags,
2338 	    pathseg, path, td);
2339 	if ((error = namei(&nd)) != 0)
2340 		return (error);
2341 	NDFREE(&nd, NDF_ONLY_PNBUF);
2342 
2343 	/* If asynchronous I/O is available, it works for all files. */
2344 	if (name == _PC_ASYNC_IO)
2345 		td->td_retval[0] = async_io_version;
2346 	else
2347 		error = VOP_PATHCONF(nd.ni_vp, name, td->td_retval);
2348 	vput(nd.ni_vp);
2349 	return (error);
2350 }
2351 
2352 /*
2353  * Return target name of a symbolic link.
2354  */
2355 #ifndef _SYS_SYSPROTO_H_
2356 struct readlink_args {
2357 	char	*path;
2358 	char	*buf;
2359 	size_t	count;
2360 };
2361 #endif
2362 int
2363 sys_readlink(td, uap)
2364 	struct thread *td;
2365 	register struct readlink_args /* {
2366 		char *path;
2367 		char *buf;
2368 		size_t count;
2369 	} */ *uap;
2370 {
2371 
2372 	return (kern_readlinkat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2373 	    uap->buf, UIO_USERSPACE, uap->count));
2374 }
2375 #ifndef _SYS_SYSPROTO_H_
2376 struct readlinkat_args {
2377 	int	fd;
2378 	char	*path;
2379 	char	*buf;
2380 	size_t	bufsize;
2381 };
2382 #endif
2383 int
2384 sys_readlinkat(struct thread *td, struct readlinkat_args *uap)
2385 {
2386 
2387 	return (kern_readlinkat(td, uap->fd, uap->path, UIO_USERSPACE,
2388 	    uap->buf, UIO_USERSPACE, uap->bufsize));
2389 }
2390 
2391 int
2392 kern_readlinkat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2393     char *buf, enum uio_seg bufseg, size_t count)
2394 {
2395 	struct vnode *vp;
2396 	struct iovec aiov;
2397 	struct uio auio;
2398 	struct nameidata nd;
2399 	int error;
2400 
2401 	if (count > IOSIZE_MAX)
2402 		return (EINVAL);
2403 
2404 	NDINIT_AT(&nd, LOOKUP, NOFOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
2405 	    pathseg, path, fd, td);
2406 
2407 	if ((error = namei(&nd)) != 0)
2408 		return (error);
2409 	NDFREE(&nd, NDF_ONLY_PNBUF);
2410 	vp = nd.ni_vp;
2411 #ifdef MAC
2412 	error = mac_vnode_check_readlink(td->td_ucred, vp);
2413 	if (error != 0) {
2414 		vput(vp);
2415 		return (error);
2416 	}
2417 #endif
2418 	if (vp->v_type != VLNK)
2419 		error = EINVAL;
2420 	else {
2421 		aiov.iov_base = buf;
2422 		aiov.iov_len = count;
2423 		auio.uio_iov = &aiov;
2424 		auio.uio_iovcnt = 1;
2425 		auio.uio_offset = 0;
2426 		auio.uio_rw = UIO_READ;
2427 		auio.uio_segflg = bufseg;
2428 		auio.uio_td = td;
2429 		auio.uio_resid = count;
2430 		error = VOP_READLINK(vp, &auio, td->td_ucred);
2431 		td->td_retval[0] = count - auio.uio_resid;
2432 	}
2433 	vput(vp);
2434 	return (error);
2435 }
2436 
2437 /*
2438  * Common implementation code for chflags() and fchflags().
2439  */
2440 static int
2441 setfflags(td, vp, flags)
2442 	struct thread *td;
2443 	struct vnode *vp;
2444 	u_long flags;
2445 {
2446 	struct mount *mp;
2447 	struct vattr vattr;
2448 	int error;
2449 
2450 	/* We can't support the value matching VNOVAL. */
2451 	if (flags == VNOVAL)
2452 		return (EOPNOTSUPP);
2453 
2454 	/*
2455 	 * Prevent non-root users from setting flags on devices.  When
2456 	 * a device is reused, users can retain ownership of the device
2457 	 * if they are allowed to set flags and programs assume that
2458 	 * chown can't fail when done as root.
2459 	 */
2460 	if (vp->v_type == VCHR || vp->v_type == VBLK) {
2461 		error = priv_check(td, PRIV_VFS_CHFLAGS_DEV);
2462 		if (error != 0)
2463 			return (error);
2464 	}
2465 
2466 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2467 		return (error);
2468 	VATTR_NULL(&vattr);
2469 	vattr.va_flags = flags;
2470 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2471 #ifdef MAC
2472 	error = mac_vnode_check_setflags(td->td_ucred, vp, vattr.va_flags);
2473 	if (error == 0)
2474 #endif
2475 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
2476 	VOP_UNLOCK(vp, 0);
2477 	vn_finished_write(mp);
2478 	return (error);
2479 }
2480 
2481 /*
2482  * Change flags of a file given a path name.
2483  */
2484 #ifndef _SYS_SYSPROTO_H_
2485 struct chflags_args {
2486 	const char *path;
2487 	u_long	flags;
2488 };
2489 #endif
2490 int
2491 sys_chflags(td, uap)
2492 	struct thread *td;
2493 	register struct chflags_args /* {
2494 		const char *path;
2495 		u_long flags;
2496 	} */ *uap;
2497 {
2498 
2499 	return (kern_chflagsat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2500 	    uap->flags, 0));
2501 }
2502 
2503 #ifndef _SYS_SYSPROTO_H_
2504 struct chflagsat_args {
2505 	int	fd;
2506 	const char *path;
2507 	u_long	flags;
2508 	int	atflag;
2509 }
2510 #endif
2511 int
2512 sys_chflagsat(struct thread *td, struct chflagsat_args *uap)
2513 {
2514 	int fd = uap->fd;
2515 	const char *path = uap->path;
2516 	u_long flags = uap->flags;
2517 	int atflag = uap->atflag;
2518 
2519 	if (atflag & ~AT_SYMLINK_NOFOLLOW)
2520 		return (EINVAL);
2521 
2522 	return (kern_chflagsat(td, fd, path, UIO_USERSPACE, flags, atflag));
2523 }
2524 
2525 /*
2526  * Same as chflags() but doesn't follow symlinks.
2527  */
2528 int
2529 sys_lchflags(td, uap)
2530 	struct thread *td;
2531 	register struct lchflags_args /* {
2532 		const char *path;
2533 		u_long flags;
2534 	} */ *uap;
2535 {
2536 
2537 	return (kern_chflagsat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2538 	    uap->flags, AT_SYMLINK_NOFOLLOW));
2539 }
2540 
2541 static int
2542 kern_chflagsat(struct thread *td, int fd, const char *path,
2543     enum uio_seg pathseg, u_long flags, int atflag)
2544 {
2545 	struct nameidata nd;
2546 	cap_rights_t rights;
2547 	int error, follow;
2548 
2549 	AUDIT_ARG_FFLAGS(flags);
2550 	follow = (atflag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
2551 	NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, pathseg, path, fd,
2552 	    cap_rights_init(&rights, CAP_FCHFLAGS), td);
2553 	if ((error = namei(&nd)) != 0)
2554 		return (error);
2555 	NDFREE(&nd, NDF_ONLY_PNBUF);
2556 	error = setfflags(td, nd.ni_vp, flags);
2557 	vrele(nd.ni_vp);
2558 	return (error);
2559 }
2560 
2561 /*
2562  * Change flags of a file given a file descriptor.
2563  */
2564 #ifndef _SYS_SYSPROTO_H_
2565 struct fchflags_args {
2566 	int	fd;
2567 	u_long	flags;
2568 };
2569 #endif
2570 int
2571 sys_fchflags(td, uap)
2572 	struct thread *td;
2573 	register struct fchflags_args /* {
2574 		int fd;
2575 		u_long flags;
2576 	} */ *uap;
2577 {
2578 	struct file *fp;
2579 	cap_rights_t rights;
2580 	int error;
2581 
2582 	AUDIT_ARG_FD(uap->fd);
2583 	AUDIT_ARG_FFLAGS(uap->flags);
2584 	error = getvnode(td, uap->fd, cap_rights_init(&rights, CAP_FCHFLAGS),
2585 	    &fp);
2586 	if (error != 0)
2587 		return (error);
2588 #ifdef AUDIT
2589 	vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
2590 	AUDIT_ARG_VNODE1(fp->f_vnode);
2591 	VOP_UNLOCK(fp->f_vnode, 0);
2592 #endif
2593 	error = setfflags(td, fp->f_vnode, uap->flags);
2594 	fdrop(fp, td);
2595 	return (error);
2596 }
2597 
2598 /*
2599  * Common implementation code for chmod(), lchmod() and fchmod().
2600  */
2601 int
2602 setfmode(td, cred, vp, mode)
2603 	struct thread *td;
2604 	struct ucred *cred;
2605 	struct vnode *vp;
2606 	int mode;
2607 {
2608 	struct mount *mp;
2609 	struct vattr vattr;
2610 	int error;
2611 
2612 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2613 		return (error);
2614 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2615 	VATTR_NULL(&vattr);
2616 	vattr.va_mode = mode & ALLPERMS;
2617 #ifdef MAC
2618 	error = mac_vnode_check_setmode(cred, vp, vattr.va_mode);
2619 	if (error == 0)
2620 #endif
2621 		error = VOP_SETATTR(vp, &vattr, cred);
2622 	VOP_UNLOCK(vp, 0);
2623 	vn_finished_write(mp);
2624 	return (error);
2625 }
2626 
2627 /*
2628  * Change mode of a file given path name.
2629  */
2630 #ifndef _SYS_SYSPROTO_H_
2631 struct chmod_args {
2632 	char	*path;
2633 	int	mode;
2634 };
2635 #endif
2636 int
2637 sys_chmod(td, uap)
2638 	struct thread *td;
2639 	register struct chmod_args /* {
2640 		char *path;
2641 		int mode;
2642 	} */ *uap;
2643 {
2644 
2645 	return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2646 	    uap->mode, 0));
2647 }
2648 
2649 #ifndef _SYS_SYSPROTO_H_
2650 struct fchmodat_args {
2651 	int	dirfd;
2652 	char	*path;
2653 	mode_t	mode;
2654 	int	flag;
2655 }
2656 #endif
2657 int
2658 sys_fchmodat(struct thread *td, struct fchmodat_args *uap)
2659 {
2660 	int flag = uap->flag;
2661 	int fd = uap->fd;
2662 	char *path = uap->path;
2663 	mode_t mode = uap->mode;
2664 
2665 	if (flag & ~AT_SYMLINK_NOFOLLOW)
2666 		return (EINVAL);
2667 
2668 	return (kern_fchmodat(td, fd, path, UIO_USERSPACE, mode, flag));
2669 }
2670 
2671 /*
2672  * Change mode of a file given path name (don't follow links.)
2673  */
2674 #ifndef _SYS_SYSPROTO_H_
2675 struct lchmod_args {
2676 	char	*path;
2677 	int	mode;
2678 };
2679 #endif
2680 int
2681 sys_lchmod(td, uap)
2682 	struct thread *td;
2683 	register struct lchmod_args /* {
2684 		char *path;
2685 		int mode;
2686 	} */ *uap;
2687 {
2688 
2689 	return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2690 	    uap->mode, AT_SYMLINK_NOFOLLOW));
2691 }
2692 
2693 int
2694 kern_fchmodat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2695     mode_t mode, int flag)
2696 {
2697 	struct nameidata nd;
2698 	cap_rights_t rights;
2699 	int error, follow;
2700 
2701 	AUDIT_ARG_MODE(mode);
2702 	follow = (flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
2703 	NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, pathseg, path, fd,
2704 	    cap_rights_init(&rights, CAP_FCHMOD), td);
2705 	if ((error = namei(&nd)) != 0)
2706 		return (error);
2707 	NDFREE(&nd, NDF_ONLY_PNBUF);
2708 	error = setfmode(td, td->td_ucred, nd.ni_vp, mode);
2709 	vrele(nd.ni_vp);
2710 	return (error);
2711 }
2712 
2713 /*
2714  * Change mode of a file given a file descriptor.
2715  */
2716 #ifndef _SYS_SYSPROTO_H_
2717 struct fchmod_args {
2718 	int	fd;
2719 	int	mode;
2720 };
2721 #endif
2722 int
2723 sys_fchmod(struct thread *td, struct fchmod_args *uap)
2724 {
2725 	struct file *fp;
2726 	cap_rights_t rights;
2727 	int error;
2728 
2729 	AUDIT_ARG_FD(uap->fd);
2730 	AUDIT_ARG_MODE(uap->mode);
2731 
2732 	error = fget(td, uap->fd, cap_rights_init(&rights, CAP_FCHMOD), &fp);
2733 	if (error != 0)
2734 		return (error);
2735 	error = fo_chmod(fp, uap->mode, td->td_ucred, td);
2736 	fdrop(fp, td);
2737 	return (error);
2738 }
2739 
2740 /*
2741  * Common implementation for chown(), lchown(), and fchown()
2742  */
2743 int
2744 setfown(td, cred, vp, uid, gid)
2745 	struct thread *td;
2746 	struct ucred *cred;
2747 	struct vnode *vp;
2748 	uid_t uid;
2749 	gid_t gid;
2750 {
2751 	struct mount *mp;
2752 	struct vattr vattr;
2753 	int error;
2754 
2755 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2756 		return (error);
2757 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2758 	VATTR_NULL(&vattr);
2759 	vattr.va_uid = uid;
2760 	vattr.va_gid = gid;
2761 #ifdef MAC
2762 	error = mac_vnode_check_setowner(cred, vp, vattr.va_uid,
2763 	    vattr.va_gid);
2764 	if (error == 0)
2765 #endif
2766 		error = VOP_SETATTR(vp, &vattr, cred);
2767 	VOP_UNLOCK(vp, 0);
2768 	vn_finished_write(mp);
2769 	return (error);
2770 }
2771 
2772 /*
2773  * Set ownership given a path name.
2774  */
2775 #ifndef _SYS_SYSPROTO_H_
2776 struct chown_args {
2777 	char	*path;
2778 	int	uid;
2779 	int	gid;
2780 };
2781 #endif
2782 int
2783 sys_chown(td, uap)
2784 	struct thread *td;
2785 	register struct chown_args /* {
2786 		char *path;
2787 		int uid;
2788 		int gid;
2789 	} */ *uap;
2790 {
2791 
2792 	return (kern_fchownat(td, AT_FDCWD, uap->path, UIO_USERSPACE, uap->uid,
2793 	    uap->gid, 0));
2794 }
2795 
2796 #ifndef _SYS_SYSPROTO_H_
2797 struct fchownat_args {
2798 	int fd;
2799 	const char * path;
2800 	uid_t uid;
2801 	gid_t gid;
2802 	int flag;
2803 };
2804 #endif
2805 int
2806 sys_fchownat(struct thread *td, struct fchownat_args *uap)
2807 {
2808 	int flag;
2809 
2810 	flag = uap->flag;
2811 	if (flag & ~AT_SYMLINK_NOFOLLOW)
2812 		return (EINVAL);
2813 
2814 	return (kern_fchownat(td, uap->fd, uap->path, UIO_USERSPACE, uap->uid,
2815 	    uap->gid, uap->flag));
2816 }
2817 
2818 int
2819 kern_fchownat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
2820     int uid, int gid, int flag)
2821 {
2822 	struct nameidata nd;
2823 	cap_rights_t rights;
2824 	int error, follow;
2825 
2826 	AUDIT_ARG_OWNER(uid, gid);
2827 	follow = (flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
2828 	NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, pathseg, path, fd,
2829 	    cap_rights_init(&rights, CAP_FCHOWN), td);
2830 
2831 	if ((error = namei(&nd)) != 0)
2832 		return (error);
2833 	NDFREE(&nd, NDF_ONLY_PNBUF);
2834 	error = setfown(td, td->td_ucred, nd.ni_vp, uid, gid);
2835 	vrele(nd.ni_vp);
2836 	return (error);
2837 }
2838 
2839 /*
2840  * Set ownership given a path name, do not cross symlinks.
2841  */
2842 #ifndef _SYS_SYSPROTO_H_
2843 struct lchown_args {
2844 	char	*path;
2845 	int	uid;
2846 	int	gid;
2847 };
2848 #endif
2849 int
2850 sys_lchown(td, uap)
2851 	struct thread *td;
2852 	register struct lchown_args /* {
2853 		char *path;
2854 		int uid;
2855 		int gid;
2856 	} */ *uap;
2857 {
2858 
2859 	return (kern_fchownat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
2860 	    uap->uid, uap->gid, AT_SYMLINK_NOFOLLOW));
2861 }
2862 
2863 /*
2864  * Set ownership given a file descriptor.
2865  */
2866 #ifndef _SYS_SYSPROTO_H_
2867 struct fchown_args {
2868 	int	fd;
2869 	int	uid;
2870 	int	gid;
2871 };
2872 #endif
2873 int
2874 sys_fchown(td, uap)
2875 	struct thread *td;
2876 	register struct fchown_args /* {
2877 		int fd;
2878 		int uid;
2879 		int gid;
2880 	} */ *uap;
2881 {
2882 	struct file *fp;
2883 	cap_rights_t rights;
2884 	int error;
2885 
2886 	AUDIT_ARG_FD(uap->fd);
2887 	AUDIT_ARG_OWNER(uap->uid, uap->gid);
2888 	error = fget(td, uap->fd, cap_rights_init(&rights, CAP_FCHOWN), &fp);
2889 	if (error != 0)
2890 		return (error);
2891 	error = fo_chown(fp, uap->uid, uap->gid, td->td_ucred, td);
2892 	fdrop(fp, td);
2893 	return (error);
2894 }
2895 
2896 /*
2897  * Common implementation code for utimes(), lutimes(), and futimes().
2898  */
2899 static int
2900 getutimes(usrtvp, tvpseg, tsp)
2901 	const struct timeval *usrtvp;
2902 	enum uio_seg tvpseg;
2903 	struct timespec *tsp;
2904 {
2905 	struct timeval tv[2];
2906 	const struct timeval *tvp;
2907 	int error;
2908 
2909 	if (usrtvp == NULL) {
2910 		vfs_timestamp(&tsp[0]);
2911 		tsp[1] = tsp[0];
2912 	} else {
2913 		if (tvpseg == UIO_SYSSPACE) {
2914 			tvp = usrtvp;
2915 		} else {
2916 			if ((error = copyin(usrtvp, tv, sizeof(tv))) != 0)
2917 				return (error);
2918 			tvp = tv;
2919 		}
2920 
2921 		if (tvp[0].tv_usec < 0 || tvp[0].tv_usec >= 1000000 ||
2922 		    tvp[1].tv_usec < 0 || tvp[1].tv_usec >= 1000000)
2923 			return (EINVAL);
2924 		TIMEVAL_TO_TIMESPEC(&tvp[0], &tsp[0]);
2925 		TIMEVAL_TO_TIMESPEC(&tvp[1], &tsp[1]);
2926 	}
2927 	return (0);
2928 }
2929 
2930 /*
2931  * Common implementation code for futimens(), utimensat().
2932  */
2933 #define	UTIMENS_NULL	0x1
2934 #define	UTIMENS_EXIT	0x2
2935 static int
2936 getutimens(const struct timespec *usrtsp, enum uio_seg tspseg,
2937     struct timespec *tsp, int *retflags)
2938 {
2939 	struct timespec tsnow;
2940 	int error;
2941 
2942 	vfs_timestamp(&tsnow);
2943 	*retflags = 0;
2944 	if (usrtsp == NULL) {
2945 		tsp[0] = tsnow;
2946 		tsp[1] = tsnow;
2947 		*retflags |= UTIMENS_NULL;
2948 		return (0);
2949 	}
2950 	if (tspseg == UIO_SYSSPACE) {
2951 		tsp[0] = usrtsp[0];
2952 		tsp[1] = usrtsp[1];
2953 	} else if ((error = copyin(usrtsp, tsp, sizeof(*tsp) * 2)) != 0)
2954 		return (error);
2955 	if (tsp[0].tv_nsec == UTIME_OMIT && tsp[1].tv_nsec == UTIME_OMIT)
2956 		*retflags |= UTIMENS_EXIT;
2957 	if (tsp[0].tv_nsec == UTIME_NOW && tsp[1].tv_nsec == UTIME_NOW)
2958 		*retflags |= UTIMENS_NULL;
2959 	if (tsp[0].tv_nsec == UTIME_OMIT)
2960 		tsp[0].tv_sec = VNOVAL;
2961 	else if (tsp[0].tv_nsec == UTIME_NOW)
2962 		tsp[0] = tsnow;
2963 	else if (tsp[0].tv_nsec < 0 || tsp[0].tv_nsec >= 1000000000L)
2964 		return (EINVAL);
2965 	if (tsp[1].tv_nsec == UTIME_OMIT)
2966 		tsp[1].tv_sec = VNOVAL;
2967 	else if (tsp[1].tv_nsec == UTIME_NOW)
2968 		tsp[1] = tsnow;
2969 	else if (tsp[1].tv_nsec < 0 || tsp[1].tv_nsec >= 1000000000L)
2970 		return (EINVAL);
2971 
2972 	return (0);
2973 }
2974 
2975 /*
2976  * Common implementation code for utimes(), lutimes(), futimes(), futimens(),
2977  * and utimensat().
2978  */
2979 static int
2980 setutimes(td, vp, ts, numtimes, nullflag)
2981 	struct thread *td;
2982 	struct vnode *vp;
2983 	const struct timespec *ts;
2984 	int numtimes;
2985 	int nullflag;
2986 {
2987 	struct mount *mp;
2988 	struct vattr vattr;
2989 	int error, setbirthtime;
2990 
2991 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
2992 		return (error);
2993 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2994 	setbirthtime = 0;
2995 	if (numtimes < 3 && !VOP_GETATTR(vp, &vattr, td->td_ucred) &&
2996 	    timespeccmp(&ts[1], &vattr.va_birthtime, < ))
2997 		setbirthtime = 1;
2998 	VATTR_NULL(&vattr);
2999 	vattr.va_atime = ts[0];
3000 	vattr.va_mtime = ts[1];
3001 	if (setbirthtime)
3002 		vattr.va_birthtime = ts[1];
3003 	if (numtimes > 2)
3004 		vattr.va_birthtime = ts[2];
3005 	if (nullflag)
3006 		vattr.va_vaflags |= VA_UTIMES_NULL;
3007 #ifdef MAC
3008 	error = mac_vnode_check_setutimes(td->td_ucred, vp, vattr.va_atime,
3009 	    vattr.va_mtime);
3010 #endif
3011 	if (error == 0)
3012 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3013 	VOP_UNLOCK(vp, 0);
3014 	vn_finished_write(mp);
3015 	return (error);
3016 }
3017 
3018 /*
3019  * Set the access and modification times of a file.
3020  */
3021 #ifndef _SYS_SYSPROTO_H_
3022 struct utimes_args {
3023 	char	*path;
3024 	struct	timeval *tptr;
3025 };
3026 #endif
3027 int
3028 sys_utimes(td, uap)
3029 	struct thread *td;
3030 	register struct utimes_args /* {
3031 		char *path;
3032 		struct timeval *tptr;
3033 	} */ *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 	cap_rights_t rights;
3062 	int error;
3063 
3064 	if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3065 		return (error);
3066 	NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, fd,
3067 	    cap_rights_init(&rights, CAP_FUTIMES), td);
3068 
3069 	if ((error = namei(&nd)) != 0)
3070 		return (error);
3071 	NDFREE(&nd, NDF_ONLY_PNBUF);
3072 	error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3073 	vrele(nd.ni_vp);
3074 	return (error);
3075 }
3076 
3077 /*
3078  * Set the access and modification times of a file.
3079  */
3080 #ifndef _SYS_SYSPROTO_H_
3081 struct lutimes_args {
3082 	char	*path;
3083 	struct	timeval *tptr;
3084 };
3085 #endif
3086 int
3087 sys_lutimes(td, uap)
3088 	struct thread *td;
3089 	register struct lutimes_args /* {
3090 		char *path;
3091 		struct timeval *tptr;
3092 	} */ *uap;
3093 {
3094 
3095 	return (kern_lutimes(td, uap->path, UIO_USERSPACE, uap->tptr,
3096 	    UIO_USERSPACE));
3097 }
3098 
3099 int
3100 kern_lutimes(struct thread *td, char *path, enum uio_seg pathseg,
3101     struct timeval *tptr, enum uio_seg tptrseg)
3102 {
3103 	struct timespec ts[2];
3104 	struct nameidata nd;
3105 	int error;
3106 
3107 	if ((error = getutimes(tptr, tptrseg, ts)) != 0)
3108 		return (error);
3109 	NDINIT(&nd, LOOKUP, NOFOLLOW | AUDITVNODE1, pathseg, path, td);
3110 	if ((error = namei(&nd)) != 0)
3111 		return (error);
3112 	NDFREE(&nd, NDF_ONLY_PNBUF);
3113 	error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
3114 	vrele(nd.ni_vp);
3115 	return (error);
3116 }
3117 
3118 /*
3119  * Set the access and modification times of a file.
3120  */
3121 #ifndef _SYS_SYSPROTO_H_
3122 struct futimes_args {
3123 	int	fd;
3124 	struct	timeval *tptr;
3125 };
3126 #endif
3127 int
3128 sys_futimes(td, uap)
3129 	struct thread *td;
3130 	register struct futimes_args /* {
3131 		int  fd;
3132 		struct timeval *tptr;
3133 	} */ *uap;
3134 {
3135 
3136 	return (kern_futimes(td, uap->fd, uap->tptr, UIO_USERSPACE));
3137 }
3138 
3139 int
3140 kern_futimes(struct thread *td, int fd, struct timeval *tptr,
3141     enum uio_seg tptrseg)
3142 {
3143 	struct timespec ts[2];
3144 	struct file *fp;
3145 	cap_rights_t rights;
3146 	int error;
3147 
3148 	AUDIT_ARG_FD(fd);
3149 	error = getutimes(tptr, tptrseg, ts);
3150 	if (error != 0)
3151 		return (error);
3152 	error = getvnode(td, fd, cap_rights_init(&rights, CAP_FUTIMES), &fp);
3153 	if (error != 0)
3154 		return (error);
3155 #ifdef AUDIT
3156 	vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
3157 	AUDIT_ARG_VNODE1(fp->f_vnode);
3158 	VOP_UNLOCK(fp->f_vnode, 0);
3159 #endif
3160 	error = setutimes(td, fp->f_vnode, ts, 2, tptr == NULL);
3161 	fdrop(fp, td);
3162 	return (error);
3163 }
3164 
3165 int
3166 sys_futimens(struct thread *td, struct futimens_args *uap)
3167 {
3168 
3169 	return (kern_futimens(td, uap->fd, uap->times, UIO_USERSPACE));
3170 }
3171 
3172 int
3173 kern_futimens(struct thread *td, int fd, struct timespec *tptr,
3174     enum uio_seg tptrseg)
3175 {
3176 	struct timespec ts[2];
3177 	struct file *fp;
3178 	cap_rights_t rights;
3179 	int error, flags;
3180 
3181 	AUDIT_ARG_FD(fd);
3182 	error = getutimens(tptr, tptrseg, ts, &flags);
3183 	if (error != 0)
3184 		return (error);
3185 	if (flags & UTIMENS_EXIT)
3186 		return (0);
3187 	error = getvnode(td, fd, cap_rights_init(&rights, CAP_FUTIMES), &fp);
3188 	if (error != 0)
3189 		return (error);
3190 #ifdef AUDIT
3191 	vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
3192 	AUDIT_ARG_VNODE1(fp->f_vnode);
3193 	VOP_UNLOCK(fp->f_vnode, 0);
3194 #endif
3195 	error = setutimes(td, fp->f_vnode, ts, 2, flags & UTIMENS_NULL);
3196 	fdrop(fp, td);
3197 	return (error);
3198 }
3199 
3200 int
3201 sys_utimensat(struct thread *td, struct utimensat_args *uap)
3202 {
3203 
3204 	return (kern_utimensat(td, uap->fd, uap->path, UIO_USERSPACE,
3205 	    uap->times, UIO_USERSPACE, uap->flag));
3206 }
3207 
3208 int
3209 kern_utimensat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
3210     struct timespec *tptr, enum uio_seg tptrseg, int flag)
3211 {
3212 	struct nameidata nd;
3213 	struct timespec ts[2];
3214 	cap_rights_t rights;
3215 	int error, flags;
3216 
3217 	if (flag & ~AT_SYMLINK_NOFOLLOW)
3218 		return (EINVAL);
3219 
3220 	if ((error = getutimens(tptr, tptrseg, ts, &flags)) != 0)
3221 		return (error);
3222 	NDINIT_ATRIGHTS(&nd, LOOKUP, ((flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW :
3223 	    FOLLOW) | AUDITVNODE1, pathseg, path, fd,
3224 	    cap_rights_init(&rights, CAP_FUTIMES), td);
3225 	if ((error = namei(&nd)) != 0)
3226 		return (error);
3227 	/*
3228 	 * We are allowed to call namei() regardless of 2xUTIME_OMIT.
3229 	 * POSIX states:
3230 	 * "If both tv_nsec fields are UTIME_OMIT... EACCESS may be detected."
3231 	 * "Search permission is denied by a component of the path prefix."
3232 	 */
3233 	NDFREE(&nd, NDF_ONLY_PNBUF);
3234 	if ((flags & UTIMENS_EXIT) == 0)
3235 		error = setutimes(td, nd.ni_vp, ts, 2, flags & UTIMENS_NULL);
3236 	vrele(nd.ni_vp);
3237 	return (error);
3238 }
3239 
3240 /*
3241  * Truncate a file given its path name.
3242  */
3243 #ifndef _SYS_SYSPROTO_H_
3244 struct truncate_args {
3245 	char	*path;
3246 	int	pad;
3247 	off_t	length;
3248 };
3249 #endif
3250 int
3251 sys_truncate(td, uap)
3252 	struct thread *td;
3253 	register struct truncate_args /* {
3254 		char *path;
3255 		int pad;
3256 		off_t length;
3257 	} */ *uap;
3258 {
3259 
3260 	return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
3261 }
3262 
3263 int
3264 kern_truncate(struct thread *td, char *path, enum uio_seg pathseg, off_t length)
3265 {
3266 	struct mount *mp;
3267 	struct vnode *vp;
3268 	void *rl_cookie;
3269 	struct vattr vattr;
3270 	struct nameidata nd;
3271 	int error;
3272 
3273 	if (length < 0)
3274 		return(EINVAL);
3275 	NDINIT(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, td);
3276 	if ((error = namei(&nd)) != 0)
3277 		return (error);
3278 	vp = nd.ni_vp;
3279 	rl_cookie = vn_rangelock_wlock(vp, 0, OFF_MAX);
3280 	if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) {
3281 		vn_rangelock_unlock(vp, rl_cookie);
3282 		vrele(vp);
3283 		return (error);
3284 	}
3285 	NDFREE(&nd, NDF_ONLY_PNBUF);
3286 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3287 	if (vp->v_type == VDIR)
3288 		error = EISDIR;
3289 #ifdef MAC
3290 	else if ((error = mac_vnode_check_write(td->td_ucred, NOCRED, vp))) {
3291 	}
3292 #endif
3293 	else if ((error = vn_writechk(vp)) == 0 &&
3294 	    (error = VOP_ACCESS(vp, VWRITE, td->td_ucred, td)) == 0) {
3295 		VATTR_NULL(&vattr);
3296 		vattr.va_size = length;
3297 		error = VOP_SETATTR(vp, &vattr, td->td_ucred);
3298 	}
3299 	VOP_UNLOCK(vp, 0);
3300 	vn_finished_write(mp);
3301 	vn_rangelock_unlock(vp, rl_cookie);
3302 	vrele(vp);
3303 	return (error);
3304 }
3305 
3306 #if defined(COMPAT_43)
3307 /*
3308  * Truncate a file given its path name.
3309  */
3310 #ifndef _SYS_SYSPROTO_H_
3311 struct otruncate_args {
3312 	char	*path;
3313 	long	length;
3314 };
3315 #endif
3316 int
3317 otruncate(td, uap)
3318 	struct thread *td;
3319 	register struct otruncate_args /* {
3320 		char *path;
3321 		long length;
3322 	} */ *uap;
3323 {
3324 	struct truncate_args /* {
3325 		char *path;
3326 		int pad;
3327 		off_t length;
3328 	} */ nuap;
3329 
3330 	nuap.path = uap->path;
3331 	nuap.length = uap->length;
3332 	return (sys_truncate(td, &nuap));
3333 }
3334 #endif /* COMPAT_43 */
3335 
3336 #if defined(COMPAT_FREEBSD6)
3337 /* Versions with the pad argument */
3338 int
3339 freebsd6_truncate(struct thread *td, struct freebsd6_truncate_args *uap)
3340 {
3341 	struct truncate_args ouap;
3342 
3343 	ouap.path = uap->path;
3344 	ouap.length = uap->length;
3345 	return (sys_truncate(td, &ouap));
3346 }
3347 
3348 int
3349 freebsd6_ftruncate(struct thread *td, struct freebsd6_ftruncate_args *uap)
3350 {
3351 	struct ftruncate_args ouap;
3352 
3353 	ouap.fd = uap->fd;
3354 	ouap.length = uap->length;
3355 	return (sys_ftruncate(td, &ouap));
3356 }
3357 #endif
3358 
3359 /*
3360  * Sync an open file.
3361  */
3362 #ifndef _SYS_SYSPROTO_H_
3363 struct fsync_args {
3364 	int	fd;
3365 };
3366 #endif
3367 int
3368 sys_fsync(td, uap)
3369 	struct thread *td;
3370 	struct fsync_args /* {
3371 		int fd;
3372 	} */ *uap;
3373 {
3374 	struct vnode *vp;
3375 	struct mount *mp;
3376 	struct file *fp;
3377 	cap_rights_t rights;
3378 	int error, lock_flags;
3379 
3380 	AUDIT_ARG_FD(uap->fd);
3381 	error = getvnode(td, uap->fd, cap_rights_init(&rights, CAP_FSYNC), &fp);
3382 	if (error != 0)
3383 		return (error);
3384 	vp = fp->f_vnode;
3385 	error = vn_start_write(vp, &mp, V_WAIT | PCATCH);
3386 	if (error != 0)
3387 		goto drop;
3388 	if (MNT_SHARED_WRITES(mp) ||
3389 	    ((mp == NULL) && MNT_SHARED_WRITES(vp->v_mount))) {
3390 		lock_flags = LK_SHARED;
3391 	} else {
3392 		lock_flags = LK_EXCLUSIVE;
3393 	}
3394 	vn_lock(vp, lock_flags | LK_RETRY);
3395 	AUDIT_ARG_VNODE1(vp);
3396 	if (vp->v_object != NULL) {
3397 		VM_OBJECT_WLOCK(vp->v_object);
3398 		vm_object_page_clean(vp->v_object, 0, 0, 0);
3399 		VM_OBJECT_WUNLOCK(vp->v_object);
3400 	}
3401 	error = VOP_FSYNC(vp, MNT_WAIT, td);
3402 
3403 	VOP_UNLOCK(vp, 0);
3404 	vn_finished_write(mp);
3405 drop:
3406 	fdrop(fp, td);
3407 	return (error);
3408 }
3409 
3410 /*
3411  * Rename files.  Source and destination must either both be directories, or
3412  * both not be directories.  If target is a directory, it must be empty.
3413  */
3414 #ifndef _SYS_SYSPROTO_H_
3415 struct rename_args {
3416 	char	*from;
3417 	char	*to;
3418 };
3419 #endif
3420 int
3421 sys_rename(td, uap)
3422 	struct thread *td;
3423 	register struct rename_args /* {
3424 		char *from;
3425 		char *to;
3426 	} */ *uap;
3427 {
3428 
3429 	return (kern_renameat(td, AT_FDCWD, uap->from, AT_FDCWD,
3430 	    uap->to, UIO_USERSPACE));
3431 }
3432 
3433 #ifndef _SYS_SYSPROTO_H_
3434 struct renameat_args {
3435 	int	oldfd;
3436 	char	*old;
3437 	int	newfd;
3438 	char	*new;
3439 };
3440 #endif
3441 int
3442 sys_renameat(struct thread *td, struct renameat_args *uap)
3443 {
3444 
3445 	return (kern_renameat(td, uap->oldfd, uap->old, uap->newfd, uap->new,
3446 	    UIO_USERSPACE));
3447 }
3448 
3449 int
3450 kern_renameat(struct thread *td, int oldfd, char *old, int newfd, char *new,
3451     enum uio_seg pathseg)
3452 {
3453 	struct mount *mp = NULL;
3454 	struct vnode *tvp, *fvp, *tdvp;
3455 	struct nameidata fromnd, tond;
3456 	cap_rights_t rights;
3457 	int error;
3458 
3459 again:
3460 	bwillwrite();
3461 #ifdef MAC
3462 	NDINIT_ATRIGHTS(&fromnd, DELETE, LOCKPARENT | LOCKLEAF | SAVESTART |
3463 	    AUDITVNODE1, pathseg, old, oldfd,
3464 	    cap_rights_init(&rights, CAP_RENAMEAT), td);
3465 #else
3466 	NDINIT_ATRIGHTS(&fromnd, DELETE, WANTPARENT | SAVESTART | AUDITVNODE1,
3467 	    pathseg, old, oldfd, cap_rights_init(&rights, CAP_RENAMEAT), td);
3468 #endif
3469 
3470 	if ((error = namei(&fromnd)) != 0)
3471 		return (error);
3472 #ifdef MAC
3473 	error = mac_vnode_check_rename_from(td->td_ucred, fromnd.ni_dvp,
3474 	    fromnd.ni_vp, &fromnd.ni_cnd);
3475 	VOP_UNLOCK(fromnd.ni_dvp, 0);
3476 	if (fromnd.ni_dvp != fromnd.ni_vp)
3477 		VOP_UNLOCK(fromnd.ni_vp, 0);
3478 #endif
3479 	fvp = fromnd.ni_vp;
3480 	NDINIT_ATRIGHTS(&tond, RENAME, LOCKPARENT | LOCKLEAF | NOCACHE |
3481 	    SAVESTART | AUDITVNODE2, pathseg, new, newfd,
3482 	    cap_rights_init(&rights, CAP_LINKAT), td);
3483 	if (fromnd.ni_vp->v_type == VDIR)
3484 		tond.ni_cnd.cn_flags |= WILLBEDIR;
3485 	if ((error = namei(&tond)) != 0) {
3486 		/* Translate error code for rename("dir1", "dir2/."). */
3487 		if (error == EISDIR && fvp->v_type == VDIR)
3488 			error = EINVAL;
3489 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3490 		vrele(fromnd.ni_dvp);
3491 		vrele(fvp);
3492 		goto out1;
3493 	}
3494 	tdvp = tond.ni_dvp;
3495 	tvp = tond.ni_vp;
3496 	error = vn_start_write(fvp, &mp, V_NOWAIT);
3497 	if (error != 0) {
3498 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3499 		NDFREE(&tond, NDF_ONLY_PNBUF);
3500 		if (tvp != NULL)
3501 			vput(tvp);
3502 		if (tdvp == tvp)
3503 			vrele(tdvp);
3504 		else
3505 			vput(tdvp);
3506 		vrele(fromnd.ni_dvp);
3507 		vrele(fvp);
3508 		vrele(tond.ni_startdir);
3509 		if (fromnd.ni_startdir != NULL)
3510 			vrele(fromnd.ni_startdir);
3511 		error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH);
3512 		if (error != 0)
3513 			return (error);
3514 		goto again;
3515 	}
3516 	if (tvp != NULL) {
3517 		if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
3518 			error = ENOTDIR;
3519 			goto out;
3520 		} else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
3521 			error = EISDIR;
3522 			goto out;
3523 		}
3524 #ifdef CAPABILITIES
3525 		if (newfd != AT_FDCWD) {
3526 			/*
3527 			 * If the target already exists we require CAP_UNLINKAT
3528 			 * from 'newfd'.
3529 			 */
3530 			error = cap_check(&tond.ni_filecaps.fc_rights,
3531 			    cap_rights_init(&rights, CAP_UNLINKAT));
3532 			if (error != 0)
3533 				goto out;
3534 		}
3535 #endif
3536 	}
3537 	if (fvp == tdvp) {
3538 		error = EINVAL;
3539 		goto out;
3540 	}
3541 	/*
3542 	 * If the source is the same as the destination (that is, if they
3543 	 * are links to the same vnode), then there is nothing to do.
3544 	 */
3545 	if (fvp == tvp)
3546 		error = -1;
3547 #ifdef MAC
3548 	else
3549 		error = mac_vnode_check_rename_to(td->td_ucred, tdvp,
3550 		    tond.ni_vp, fromnd.ni_dvp == tdvp, &tond.ni_cnd);
3551 #endif
3552 out:
3553 	if (error == 0) {
3554 		error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
3555 		    tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
3556 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3557 		NDFREE(&tond, NDF_ONLY_PNBUF);
3558 	} else {
3559 		NDFREE(&fromnd, NDF_ONLY_PNBUF);
3560 		NDFREE(&tond, NDF_ONLY_PNBUF);
3561 		if (tvp != NULL)
3562 			vput(tvp);
3563 		if (tdvp == tvp)
3564 			vrele(tdvp);
3565 		else
3566 			vput(tdvp);
3567 		vrele(fromnd.ni_dvp);
3568 		vrele(fvp);
3569 	}
3570 	vrele(tond.ni_startdir);
3571 	vn_finished_write(mp);
3572 out1:
3573 	if (fromnd.ni_startdir)
3574 		vrele(fromnd.ni_startdir);
3575 	if (error == -1)
3576 		return (0);
3577 	return (error);
3578 }
3579 
3580 /*
3581  * Make a directory file.
3582  */
3583 #ifndef _SYS_SYSPROTO_H_
3584 struct mkdir_args {
3585 	char	*path;
3586 	int	mode;
3587 };
3588 #endif
3589 int
3590 sys_mkdir(td, uap)
3591 	struct thread *td;
3592 	register struct mkdir_args /* {
3593 		char *path;
3594 		int mode;
3595 	} */ *uap;
3596 {
3597 
3598 	return (kern_mkdirat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
3599 	    uap->mode));
3600 }
3601 
3602 #ifndef _SYS_SYSPROTO_H_
3603 struct mkdirat_args {
3604 	int	fd;
3605 	char	*path;
3606 	mode_t	mode;
3607 };
3608 #endif
3609 int
3610 sys_mkdirat(struct thread *td, struct mkdirat_args *uap)
3611 {
3612 
3613 	return (kern_mkdirat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode));
3614 }
3615 
3616 int
3617 kern_mkdirat(struct thread *td, int fd, char *path, enum uio_seg segflg,
3618     int mode)
3619 {
3620 	struct mount *mp;
3621 	struct vnode *vp;
3622 	struct vattr vattr;
3623 	struct nameidata nd;
3624 	cap_rights_t rights;
3625 	int error;
3626 
3627 	AUDIT_ARG_MODE(mode);
3628 restart:
3629 	bwillwrite();
3630 	NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
3631 	    NOCACHE, segflg, path, fd, cap_rights_init(&rights, CAP_MKDIRAT),
3632 	    td);
3633 	nd.ni_cnd.cn_flags |= WILLBEDIR;
3634 	if ((error = namei(&nd)) != 0)
3635 		return (error);
3636 	vp = nd.ni_vp;
3637 	if (vp != NULL) {
3638 		NDFREE(&nd, NDF_ONLY_PNBUF);
3639 		/*
3640 		 * XXX namei called with LOCKPARENT but not LOCKLEAF has
3641 		 * the strange behaviour of leaving the vnode unlocked
3642 		 * if the target is the same vnode as the parent.
3643 		 */
3644 		if (vp == nd.ni_dvp)
3645 			vrele(nd.ni_dvp);
3646 		else
3647 			vput(nd.ni_dvp);
3648 		vrele(vp);
3649 		return (EEXIST);
3650 	}
3651 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3652 		NDFREE(&nd, NDF_ONLY_PNBUF);
3653 		vput(nd.ni_dvp);
3654 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3655 			return (error);
3656 		goto restart;
3657 	}
3658 	VATTR_NULL(&vattr);
3659 	vattr.va_type = VDIR;
3660 	vattr.va_mode = (mode & ACCESSPERMS) &~ td->td_proc->p_fd->fd_cmask;
3661 #ifdef MAC
3662 	error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
3663 	    &vattr);
3664 	if (error != 0)
3665 		goto out;
3666 #endif
3667 	error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
3668 #ifdef MAC
3669 out:
3670 #endif
3671 	NDFREE(&nd, NDF_ONLY_PNBUF);
3672 	vput(nd.ni_dvp);
3673 	if (error == 0)
3674 		vput(nd.ni_vp);
3675 	vn_finished_write(mp);
3676 	return (error);
3677 }
3678 
3679 /*
3680  * Remove a directory file.
3681  */
3682 #ifndef _SYS_SYSPROTO_H_
3683 struct rmdir_args {
3684 	char	*path;
3685 };
3686 #endif
3687 int
3688 sys_rmdir(td, uap)
3689 	struct thread *td;
3690 	struct rmdir_args /* {
3691 		char *path;
3692 	} */ *uap;
3693 {
3694 
3695 	return (kern_rmdirat(td, AT_FDCWD, uap->path, UIO_USERSPACE));
3696 }
3697 
3698 int
3699 kern_rmdirat(struct thread *td, int fd, char *path, enum uio_seg pathseg)
3700 {
3701 	struct mount *mp;
3702 	struct vnode *vp;
3703 	struct nameidata nd;
3704 	cap_rights_t rights;
3705 	int error;
3706 
3707 restart:
3708 	bwillwrite();
3709 	NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1,
3710 	    pathseg, path, fd, cap_rights_init(&rights, CAP_UNLINKAT), td);
3711 	if ((error = namei(&nd)) != 0)
3712 		return (error);
3713 	vp = nd.ni_vp;
3714 	if (vp->v_type != VDIR) {
3715 		error = ENOTDIR;
3716 		goto out;
3717 	}
3718 	/*
3719 	 * No rmdir "." please.
3720 	 */
3721 	if (nd.ni_dvp == vp) {
3722 		error = EINVAL;
3723 		goto out;
3724 	}
3725 	/*
3726 	 * The root of a mounted filesystem cannot be deleted.
3727 	 */
3728 	if (vp->v_vflag & VV_ROOT) {
3729 		error = EBUSY;
3730 		goto out;
3731 	}
3732 #ifdef MAC
3733 	error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
3734 	    &nd.ni_cnd);
3735 	if (error != 0)
3736 		goto out;
3737 #endif
3738 	if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
3739 		NDFREE(&nd, NDF_ONLY_PNBUF);
3740 		vput(vp);
3741 		if (nd.ni_dvp == vp)
3742 			vrele(nd.ni_dvp);
3743 		else
3744 			vput(nd.ni_dvp);
3745 		if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
3746 			return (error);
3747 		goto restart;
3748 	}
3749 	vfs_notify_upper(vp, VFS_NOTIFY_UPPER_UNLINK);
3750 	error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
3751 	vn_finished_write(mp);
3752 out:
3753 	NDFREE(&nd, NDF_ONLY_PNBUF);
3754 	vput(vp);
3755 	if (nd.ni_dvp == vp)
3756 		vrele(nd.ni_dvp);
3757 	else
3758 		vput(nd.ni_dvp);
3759 	return (error);
3760 }
3761 
3762 #ifdef COMPAT_43
3763 /*
3764  * Read a block of directory entries in a filesystem independent format.
3765  */
3766 #ifndef _SYS_SYSPROTO_H_
3767 struct ogetdirentries_args {
3768 	int	fd;
3769 	char	*buf;
3770 	u_int	count;
3771 	long	*basep;
3772 };
3773 #endif
3774 int
3775 ogetdirentries(struct thread *td, struct ogetdirentries_args *uap)
3776 {
3777 	long loff;
3778 	int error;
3779 
3780 	error = kern_ogetdirentries(td, uap, &loff);
3781 	if (error == 0)
3782 		error = copyout(&loff, uap->basep, sizeof(long));
3783 	return (error);
3784 }
3785 
3786 int
3787 kern_ogetdirentries(struct thread *td, struct ogetdirentries_args *uap,
3788     long *ploff)
3789 {
3790 	struct vnode *vp;
3791 	struct file *fp;
3792 	struct uio auio, kuio;
3793 	struct iovec aiov, kiov;
3794 	struct dirent *dp, *edp;
3795 	cap_rights_t rights;
3796 	caddr_t dirbuf;
3797 	int error, eofflag, readcnt;
3798 	long loff;
3799 	off_t foffset;
3800 
3801 	/* XXX arbitrary sanity limit on `count'. */
3802 	if (uap->count > 64 * 1024)
3803 		return (EINVAL);
3804 	error = getvnode(td, uap->fd, cap_rights_init(&rights, CAP_READ), &fp);
3805 	if (error != 0)
3806 		return (error);
3807 	if ((fp->f_flag & FREAD) == 0) {
3808 		fdrop(fp, td);
3809 		return (EBADF);
3810 	}
3811 	vp = fp->f_vnode;
3812 	foffset = foffset_lock(fp, 0);
3813 unionread:
3814 	if (vp->v_type != VDIR) {
3815 		foffset_unlock(fp, foffset, 0);
3816 		fdrop(fp, td);
3817 		return (EINVAL);
3818 	}
3819 	aiov.iov_base = uap->buf;
3820 	aiov.iov_len = uap->count;
3821 	auio.uio_iov = &aiov;
3822 	auio.uio_iovcnt = 1;
3823 	auio.uio_rw = UIO_READ;
3824 	auio.uio_segflg = UIO_USERSPACE;
3825 	auio.uio_td = td;
3826 	auio.uio_resid = uap->count;
3827 	vn_lock(vp, LK_SHARED | LK_RETRY);
3828 	loff = auio.uio_offset = foffset;
3829 #ifdef MAC
3830 	error = mac_vnode_check_readdir(td->td_ucred, vp);
3831 	if (error != 0) {
3832 		VOP_UNLOCK(vp, 0);
3833 		foffset_unlock(fp, foffset, FOF_NOUPDATE);
3834 		fdrop(fp, td);
3835 		return (error);
3836 	}
3837 #endif
3838 #	if (BYTE_ORDER != LITTLE_ENDIAN)
3839 		if (vp->v_mount->mnt_maxsymlinklen <= 0) {
3840 			error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag,
3841 			    NULL, NULL);
3842 			foffset = auio.uio_offset;
3843 		} else
3844 #	endif
3845 	{
3846 		kuio = auio;
3847 		kuio.uio_iov = &kiov;
3848 		kuio.uio_segflg = UIO_SYSSPACE;
3849 		kiov.iov_len = uap->count;
3850 		dirbuf = malloc(uap->count, M_TEMP, M_WAITOK);
3851 		kiov.iov_base = dirbuf;
3852 		error = VOP_READDIR(vp, &kuio, fp->f_cred, &eofflag,
3853 			    NULL, NULL);
3854 		foffset = kuio.uio_offset;
3855 		if (error == 0) {
3856 			readcnt = uap->count - kuio.uio_resid;
3857 			edp = (struct dirent *)&dirbuf[readcnt];
3858 			for (dp = (struct dirent *)dirbuf; dp < edp; ) {
3859 #				if (BYTE_ORDER == LITTLE_ENDIAN)
3860 					/*
3861 					 * The expected low byte of
3862 					 * dp->d_namlen is our dp->d_type.
3863 					 * The high MBZ byte of dp->d_namlen
3864 					 * is our dp->d_namlen.
3865 					 */
3866 					dp->d_type = dp->d_namlen;
3867 					dp->d_namlen = 0;
3868 #				else
3869 					/*
3870 					 * The dp->d_type is the high byte
3871 					 * of the expected dp->d_namlen,
3872 					 * so must be zero'ed.
3873 					 */
3874 					dp->d_type = 0;
3875 #				endif
3876 				if (dp->d_reclen > 0) {
3877 					dp = (struct dirent *)
3878 					    ((char *)dp + dp->d_reclen);
3879 				} else {
3880 					error = EIO;
3881 					break;
3882 				}
3883 			}
3884 			if (dp >= edp)
3885 				error = uiomove(dirbuf, readcnt, &auio);
3886 		}
3887 		free(dirbuf, M_TEMP);
3888 	}
3889 	if (error != 0) {
3890 		VOP_UNLOCK(vp, 0);
3891 		foffset_unlock(fp, foffset, 0);
3892 		fdrop(fp, td);
3893 		return (error);
3894 	}
3895 	if (uap->count == auio.uio_resid &&
3896 	    (vp->v_vflag & VV_ROOT) &&
3897 	    (vp->v_mount->mnt_flag & MNT_UNION)) {
3898 		struct vnode *tvp = vp;
3899 		vp = vp->v_mount->mnt_vnodecovered;
3900 		VREF(vp);
3901 		fp->f_vnode = vp;
3902 		fp->f_data = vp;
3903 		foffset = 0;
3904 		vput(tvp);
3905 		goto unionread;
3906 	}
3907 	VOP_UNLOCK(vp, 0);
3908 	foffset_unlock(fp, foffset, 0);
3909 	fdrop(fp, td);
3910 	td->td_retval[0] = uap->count - auio.uio_resid;
3911 	if (error == 0)
3912 		*ploff = loff;
3913 	return (error);
3914 }
3915 #endif /* COMPAT_43 */
3916 
3917 /*
3918  * Read a block of directory entries in a filesystem independent format.
3919  */
3920 #ifndef _SYS_SYSPROTO_H_
3921 struct getdirentries_args {
3922 	int	fd;
3923 	char	*buf;
3924 	u_int	count;
3925 	long	*basep;
3926 };
3927 #endif
3928 int
3929 sys_getdirentries(td, uap)
3930 	struct thread *td;
3931 	register struct getdirentries_args /* {
3932 		int fd;
3933 		char *buf;
3934 		u_int count;
3935 		long *basep;
3936 	} */ *uap;
3937 {
3938 	long base;
3939 	int error;
3940 
3941 	error = kern_getdirentries(td, uap->fd, uap->buf, uap->count, &base,
3942 	    NULL, UIO_USERSPACE);
3943 	if (error != 0)
3944 		return (error);
3945 	if (uap->basep != NULL)
3946 		error = copyout(&base, uap->basep, sizeof(long));
3947 	return (error);
3948 }
3949 
3950 int
3951 kern_getdirentries(struct thread *td, int fd, char *buf, u_int count,
3952     long *basep, ssize_t *residp, enum uio_seg bufseg)
3953 {
3954 	struct vnode *vp;
3955 	struct file *fp;
3956 	struct uio auio;
3957 	struct iovec aiov;
3958 	cap_rights_t rights;
3959 	long loff;
3960 	int error, eofflag;
3961 	off_t foffset;
3962 
3963 	AUDIT_ARG_FD(fd);
3964 	if (count > IOSIZE_MAX)
3965 		return (EINVAL);
3966 	auio.uio_resid = count;
3967 	error = getvnode(td, fd, cap_rights_init(&rights, CAP_READ), &fp);
3968 	if (error != 0)
3969 		return (error);
3970 	if ((fp->f_flag & FREAD) == 0) {
3971 		fdrop(fp, td);
3972 		return (EBADF);
3973 	}
3974 	vp = fp->f_vnode;
3975 	foffset = foffset_lock(fp, 0);
3976 unionread:
3977 	if (vp->v_type != VDIR) {
3978 		error = EINVAL;
3979 		goto fail;
3980 	}
3981 	aiov.iov_base = buf;
3982 	aiov.iov_len = count;
3983 	auio.uio_iov = &aiov;
3984 	auio.uio_iovcnt = 1;
3985 	auio.uio_rw = UIO_READ;
3986 	auio.uio_segflg = bufseg;
3987 	auio.uio_td = td;
3988 	vn_lock(vp, LK_SHARED | LK_RETRY);
3989 	AUDIT_ARG_VNODE1(vp);
3990 	loff = auio.uio_offset = foffset;
3991 #ifdef MAC
3992 	error = mac_vnode_check_readdir(td->td_ucred, vp);
3993 	if (error == 0)
3994 #endif
3995 		error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, NULL,
3996 		    NULL);
3997 	foffset = auio.uio_offset;
3998 	if (error != 0) {
3999 		VOP_UNLOCK(vp, 0);
4000 		goto fail;
4001 	}
4002 	if (count == auio.uio_resid &&
4003 	    (vp->v_vflag & VV_ROOT) &&
4004 	    (vp->v_mount->mnt_flag & MNT_UNION)) {
4005 		struct vnode *tvp = vp;
4006 
4007 		vp = vp->v_mount->mnt_vnodecovered;
4008 		VREF(vp);
4009 		fp->f_vnode = vp;
4010 		fp->f_data = vp;
4011 		foffset = 0;
4012 		vput(tvp);
4013 		goto unionread;
4014 	}
4015 	VOP_UNLOCK(vp, 0);
4016 	*basep = loff;
4017 	if (residp != NULL)
4018 		*residp = auio.uio_resid;
4019 	td->td_retval[0] = count - auio.uio_resid;
4020 fail:
4021 	foffset_unlock(fp, foffset, 0);
4022 	fdrop(fp, td);
4023 	return (error);
4024 }
4025 
4026 #ifndef _SYS_SYSPROTO_H_
4027 struct getdents_args {
4028 	int fd;
4029 	char *buf;
4030 	size_t count;
4031 };
4032 #endif
4033 int
4034 sys_getdents(td, uap)
4035 	struct thread *td;
4036 	register struct getdents_args /* {
4037 		int fd;
4038 		char *buf;
4039 		u_int count;
4040 	} */ *uap;
4041 {
4042 	struct getdirentries_args ap;
4043 
4044 	ap.fd = uap->fd;
4045 	ap.buf = uap->buf;
4046 	ap.count = uap->count;
4047 	ap.basep = NULL;
4048 	return (sys_getdirentries(td, &ap));
4049 }
4050 
4051 /*
4052  * Set the mode mask for creation of filesystem nodes.
4053  */
4054 #ifndef _SYS_SYSPROTO_H_
4055 struct umask_args {
4056 	int	newmask;
4057 };
4058 #endif
4059 int
4060 sys_umask(td, uap)
4061 	struct thread *td;
4062 	struct umask_args /* {
4063 		int newmask;
4064 	} */ *uap;
4065 {
4066 	struct filedesc *fdp;
4067 
4068 	fdp = td->td_proc->p_fd;
4069 	FILEDESC_XLOCK(fdp);
4070 	td->td_retval[0] = fdp->fd_cmask;
4071 	fdp->fd_cmask = uap->newmask & ALLPERMS;
4072 	FILEDESC_XUNLOCK(fdp);
4073 	return (0);
4074 }
4075 
4076 /*
4077  * Void all references to file by ripping underlying filesystem away from
4078  * vnode.
4079  */
4080 #ifndef _SYS_SYSPROTO_H_
4081 struct revoke_args {
4082 	char	*path;
4083 };
4084 #endif
4085 int
4086 sys_revoke(td, uap)
4087 	struct thread *td;
4088 	register struct revoke_args /* {
4089 		char *path;
4090 	} */ *uap;
4091 {
4092 	struct vnode *vp;
4093 	struct vattr vattr;
4094 	struct nameidata nd;
4095 	int error;
4096 
4097 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
4098 	    uap->path, td);
4099 	if ((error = namei(&nd)) != 0)
4100 		return (error);
4101 	vp = nd.ni_vp;
4102 	NDFREE(&nd, NDF_ONLY_PNBUF);
4103 	if (vp->v_type != VCHR || vp->v_rdev == NULL) {
4104 		error = EINVAL;
4105 		goto out;
4106 	}
4107 #ifdef MAC
4108 	error = mac_vnode_check_revoke(td->td_ucred, vp);
4109 	if (error != 0)
4110 		goto out;
4111 #endif
4112 	error = VOP_GETATTR(vp, &vattr, td->td_ucred);
4113 	if (error != 0)
4114 		goto out;
4115 	if (td->td_ucred->cr_uid != vattr.va_uid) {
4116 		error = priv_check(td, PRIV_VFS_ADMIN);
4117 		if (error != 0)
4118 			goto out;
4119 	}
4120 	if (vcount(vp) > 1)
4121 		VOP_REVOKE(vp, REVOKEALL);
4122 out:
4123 	vput(vp);
4124 	return (error);
4125 }
4126 
4127 /*
4128  * Convert a user file descriptor to a kernel file entry and check that, if it
4129  * is a capability, the correct rights are present. A reference on the file
4130  * entry is held upon returning.
4131  */
4132 int
4133 getvnode(struct thread *td, int fd, cap_rights_t *rightsp, struct file **fpp)
4134 {
4135 	struct file *fp;
4136 	int error;
4137 
4138 	error = fget_unlocked(td->td_proc->p_fd, fd, rightsp, &fp, NULL);
4139 	if (error != 0)
4140 		return (error);
4141 
4142 	/*
4143 	 * The file could be not of the vnode type, or it may be not
4144 	 * yet fully initialized, in which case the f_vnode pointer
4145 	 * may be set, but f_ops is still badfileops.  E.g.,
4146 	 * devfs_open() transiently create such situation to
4147 	 * facilitate csw d_fdopen().
4148 	 *
4149 	 * Dupfdopen() handling in kern_openat() installs the
4150 	 * half-baked file into the process descriptor table, allowing
4151 	 * other thread to dereference it. Guard against the race by
4152 	 * checking f_ops.
4153 	 */
4154 	if (fp->f_vnode == NULL || fp->f_ops == &badfileops) {
4155 		fdrop(fp, td);
4156 		return (EINVAL);
4157 	}
4158 	*fpp = fp;
4159 	return (0);
4160 }
4161 
4162 
4163 /*
4164  * Get an (NFS) file handle.
4165  */
4166 #ifndef _SYS_SYSPROTO_H_
4167 struct lgetfh_args {
4168 	char	*fname;
4169 	fhandle_t *fhp;
4170 };
4171 #endif
4172 int
4173 sys_lgetfh(td, uap)
4174 	struct thread *td;
4175 	register struct lgetfh_args *uap;
4176 {
4177 	struct nameidata nd;
4178 	fhandle_t fh;
4179 	register struct vnode *vp;
4180 	int error;
4181 
4182 	error = priv_check(td, PRIV_VFS_GETFH);
4183 	if (error != 0)
4184 		return (error);
4185 	NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
4186 	    uap->fname, td);
4187 	error = namei(&nd);
4188 	if (error != 0)
4189 		return (error);
4190 	NDFREE(&nd, NDF_ONLY_PNBUF);
4191 	vp = nd.ni_vp;
4192 	bzero(&fh, sizeof(fh));
4193 	fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
4194 	error = VOP_VPTOFH(vp, &fh.fh_fid);
4195 	vput(vp);
4196 	if (error == 0)
4197 		error = copyout(&fh, uap->fhp, sizeof (fh));
4198 	return (error);
4199 }
4200 
4201 #ifndef _SYS_SYSPROTO_H_
4202 struct getfh_args {
4203 	char	*fname;
4204 	fhandle_t *fhp;
4205 };
4206 #endif
4207 int
4208 sys_getfh(td, uap)
4209 	struct thread *td;
4210 	register struct getfh_args *uap;
4211 {
4212 	struct nameidata nd;
4213 	fhandle_t fh;
4214 	register struct vnode *vp;
4215 	int error;
4216 
4217 	error = priv_check(td, PRIV_VFS_GETFH);
4218 	if (error != 0)
4219 		return (error);
4220 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
4221 	    uap->fname, td);
4222 	error = namei(&nd);
4223 	if (error != 0)
4224 		return (error);
4225 	NDFREE(&nd, NDF_ONLY_PNBUF);
4226 	vp = nd.ni_vp;
4227 	bzero(&fh, sizeof(fh));
4228 	fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
4229 	error = VOP_VPTOFH(vp, &fh.fh_fid);
4230 	vput(vp);
4231 	if (error == 0)
4232 		error = copyout(&fh, uap->fhp, sizeof (fh));
4233 	return (error);
4234 }
4235 
4236 /*
4237  * syscall for the rpc.lockd to use to translate a NFS file handle into an
4238  * open descriptor.
4239  *
4240  * warning: do not remove the priv_check() call or this becomes one giant
4241  * security hole.
4242  */
4243 #ifndef _SYS_SYSPROTO_H_
4244 struct fhopen_args {
4245 	const struct fhandle *u_fhp;
4246 	int flags;
4247 };
4248 #endif
4249 int
4250 sys_fhopen(td, uap)
4251 	struct thread *td;
4252 	struct fhopen_args /* {
4253 		const struct fhandle *u_fhp;
4254 		int flags;
4255 	} */ *uap;
4256 {
4257 	struct mount *mp;
4258 	struct vnode *vp;
4259 	struct fhandle fhp;
4260 	struct file *fp;
4261 	int fmode, error;
4262 	int indx;
4263 
4264 	error = priv_check(td, PRIV_VFS_FHOPEN);
4265 	if (error != 0)
4266 		return (error);
4267 	indx = -1;
4268 	fmode = FFLAGS(uap->flags);
4269 	/* why not allow a non-read/write open for our lockd? */
4270 	if (((fmode & (FREAD | FWRITE)) == 0) || (fmode & O_CREAT))
4271 		return (EINVAL);
4272 	error = copyin(uap->u_fhp, &fhp, sizeof(fhp));
4273 	if (error != 0)
4274 		return(error);
4275 	/* find the mount point */
4276 	mp = vfs_busyfs(&fhp.fh_fsid);
4277 	if (mp == NULL)
4278 		return (ESTALE);
4279 	/* now give me my vnode, it gets returned to me locked */
4280 	error = VFS_FHTOVP(mp, &fhp.fh_fid, LK_EXCLUSIVE, &vp);
4281 	vfs_unbusy(mp);
4282 	if (error != 0)
4283 		return (error);
4284 
4285 	error = falloc_noinstall(td, &fp);
4286 	if (error != 0) {
4287 		vput(vp);
4288 		return (error);
4289 	}
4290 	/*
4291 	 * An extra reference on `fp' has been held for us by
4292 	 * falloc_noinstall().
4293 	 */
4294 
4295 #ifdef INVARIANTS
4296 	td->td_dupfd = -1;
4297 #endif
4298 	error = vn_open_vnode(vp, fmode, td->td_ucred, td, fp);
4299 	if (error != 0) {
4300 		KASSERT(fp->f_ops == &badfileops,
4301 		    ("VOP_OPEN in fhopen() set f_ops"));
4302 		KASSERT(td->td_dupfd < 0,
4303 		    ("fhopen() encountered fdopen()"));
4304 
4305 		vput(vp);
4306 		goto bad;
4307 	}
4308 #ifdef INVARIANTS
4309 	td->td_dupfd = 0;
4310 #endif
4311 	fp->f_vnode = vp;
4312 	fp->f_seqcount = 1;
4313 	finit(fp, (fmode & FMASK) | (fp->f_flag & FHASLOCK), DTYPE_VNODE, vp,
4314 	    &vnops);
4315 	VOP_UNLOCK(vp, 0);
4316 	if ((fmode & O_TRUNC) != 0) {
4317 		error = fo_truncate(fp, 0, td->td_ucred, td);
4318 		if (error != 0)
4319 			goto bad;
4320 	}
4321 
4322 	error = finstall(td, fp, &indx, fmode, NULL);
4323 bad:
4324 	fdrop(fp, td);
4325 	td->td_retval[0] = indx;
4326 	return (error);
4327 }
4328 
4329 /*
4330  * Stat an (NFS) file handle.
4331  */
4332 #ifndef _SYS_SYSPROTO_H_
4333 struct fhstat_args {
4334 	struct fhandle *u_fhp;
4335 	struct stat *sb;
4336 };
4337 #endif
4338 int
4339 sys_fhstat(td, uap)
4340 	struct thread *td;
4341 	register struct fhstat_args /* {
4342 		struct fhandle *u_fhp;
4343 		struct stat *sb;
4344 	} */ *uap;
4345 {
4346 	struct stat sb;
4347 	struct fhandle fh;
4348 	int error;
4349 
4350 	error = copyin(uap->u_fhp, &fh, sizeof(fh));
4351 	if (error != 0)
4352 		return (error);
4353 	error = kern_fhstat(td, fh, &sb);
4354 	if (error == 0)
4355 		error = copyout(&sb, uap->sb, sizeof(sb));
4356 	return (error);
4357 }
4358 
4359 int
4360 kern_fhstat(struct thread *td, struct fhandle fh, struct stat *sb)
4361 {
4362 	struct mount *mp;
4363 	struct vnode *vp;
4364 	int error;
4365 
4366 	error = priv_check(td, PRIV_VFS_FHSTAT);
4367 	if (error != 0)
4368 		return (error);
4369 	if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4370 		return (ESTALE);
4371 	error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4372 	vfs_unbusy(mp);
4373 	if (error != 0)
4374 		return (error);
4375 	error = vn_stat(vp, sb, td->td_ucred, NOCRED, td);
4376 	vput(vp);
4377 	return (error);
4378 }
4379 
4380 /*
4381  * Implement fstatfs() for (NFS) file handles.
4382  */
4383 #ifndef _SYS_SYSPROTO_H_
4384 struct fhstatfs_args {
4385 	struct fhandle *u_fhp;
4386 	struct statfs *buf;
4387 };
4388 #endif
4389 int
4390 sys_fhstatfs(td, uap)
4391 	struct thread *td;
4392 	struct fhstatfs_args /* {
4393 		struct fhandle *u_fhp;
4394 		struct statfs *buf;
4395 	} */ *uap;
4396 {
4397 	struct statfs sf;
4398 	fhandle_t fh;
4399 	int error;
4400 
4401 	error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
4402 	if (error != 0)
4403 		return (error);
4404 	error = kern_fhstatfs(td, fh, &sf);
4405 	if (error != 0)
4406 		return (error);
4407 	return (copyout(&sf, uap->buf, sizeof(sf)));
4408 }
4409 
4410 int
4411 kern_fhstatfs(struct thread *td, fhandle_t fh, struct statfs *buf)
4412 {
4413 	struct statfs *sp;
4414 	struct mount *mp;
4415 	struct vnode *vp;
4416 	int error;
4417 
4418 	error = priv_check(td, PRIV_VFS_FHSTATFS);
4419 	if (error != 0)
4420 		return (error);
4421 	if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
4422 		return (ESTALE);
4423 	error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
4424 	if (error != 0) {
4425 		vfs_unbusy(mp);
4426 		return (error);
4427 	}
4428 	vput(vp);
4429 	error = prison_canseemount(td->td_ucred, mp);
4430 	if (error != 0)
4431 		goto out;
4432 #ifdef MAC
4433 	error = mac_mount_check_stat(td->td_ucred, mp);
4434 	if (error != 0)
4435 		goto out;
4436 #endif
4437 	/*
4438 	 * Set these in case the underlying filesystem fails to do so.
4439 	 */
4440 	sp = &mp->mnt_stat;
4441 	sp->f_version = STATFS_VERSION;
4442 	sp->f_namemax = NAME_MAX;
4443 	sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
4444 	error = VFS_STATFS(mp, sp);
4445 	if (error == 0)
4446 		*buf = *sp;
4447 out:
4448 	vfs_unbusy(mp);
4449 	return (error);
4450 }
4451 
4452 int
4453 kern_posix_fallocate(struct thread *td, int fd, off_t offset, off_t len)
4454 {
4455 	struct file *fp;
4456 	struct mount *mp;
4457 	struct vnode *vp;
4458 	cap_rights_t rights;
4459 	off_t olen, ooffset;
4460 	int error;
4461 
4462 	if (offset < 0 || len <= 0)
4463 		return (EINVAL);
4464 	/* Check for wrap. */
4465 	if (offset > OFF_MAX - len)
4466 		return (EFBIG);
4467 	error = fget(td, fd, cap_rights_init(&rights, CAP_WRITE), &fp);
4468 	if (error != 0)
4469 		return (error);
4470 	if ((fp->f_ops->fo_flags & DFLAG_SEEKABLE) == 0) {
4471 		error = ESPIPE;
4472 		goto out;
4473 	}
4474 	if ((fp->f_flag & FWRITE) == 0) {
4475 		error = EBADF;
4476 		goto out;
4477 	}
4478 	if (fp->f_type != DTYPE_VNODE) {
4479 		error = ENODEV;
4480 		goto out;
4481 	}
4482 	vp = fp->f_vnode;
4483 	if (vp->v_type != VREG) {
4484 		error = ENODEV;
4485 		goto out;
4486 	}
4487 
4488 	/* Allocating blocks may take a long time, so iterate. */
4489 	for (;;) {
4490 		olen = len;
4491 		ooffset = offset;
4492 
4493 		bwillwrite();
4494 		mp = NULL;
4495 		error = vn_start_write(vp, &mp, V_WAIT | PCATCH);
4496 		if (error != 0)
4497 			break;
4498 		error = vn_lock(vp, LK_EXCLUSIVE);
4499 		if (error != 0) {
4500 			vn_finished_write(mp);
4501 			break;
4502 		}
4503 #ifdef MAC
4504 		error = mac_vnode_check_write(td->td_ucred, fp->f_cred, vp);
4505 		if (error == 0)
4506 #endif
4507 			error = VOP_ALLOCATE(vp, &offset, &len);
4508 		VOP_UNLOCK(vp, 0);
4509 		vn_finished_write(mp);
4510 
4511 		if (olen + ooffset != offset + len) {
4512 			panic("offset + len changed from %jx/%jx to %jx/%jx",
4513 			    ooffset, olen, offset, len);
4514 		}
4515 		if (error != 0 || len == 0)
4516 			break;
4517 		KASSERT(olen > len, ("Iteration did not make progress?"));
4518 		maybe_yield();
4519 	}
4520  out:
4521 	fdrop(fp, td);
4522 	return (error);
4523 }
4524 
4525 int
4526 sys_posix_fallocate(struct thread *td, struct posix_fallocate_args *uap)
4527 {
4528 
4529 	td->td_retval[0] = kern_posix_fallocate(td, uap->fd, uap->offset,
4530 	    uap->len);
4531 	return (0);
4532 }
4533 
4534 /*
4535  * Unlike madvise(2), we do not make a best effort to remember every
4536  * possible caching hint.  Instead, we remember the last setting with
4537  * the exception that we will allow POSIX_FADV_NORMAL to adjust the
4538  * region of any current setting.
4539  */
4540 int
4541 kern_posix_fadvise(struct thread *td, int fd, off_t offset, off_t len,
4542     int advice)
4543 {
4544 	struct fadvise_info *fa, *new;
4545 	struct file *fp;
4546 	struct vnode *vp;
4547 	cap_rights_t rights;
4548 	off_t end;
4549 	int error;
4550 
4551 	if (offset < 0 || len < 0 || offset > OFF_MAX - len)
4552 		return (EINVAL);
4553 	switch (advice) {
4554 	case POSIX_FADV_SEQUENTIAL:
4555 	case POSIX_FADV_RANDOM:
4556 	case POSIX_FADV_NOREUSE:
4557 		new = malloc(sizeof(*fa), M_FADVISE, M_WAITOK);
4558 		break;
4559 	case POSIX_FADV_NORMAL:
4560 	case POSIX_FADV_WILLNEED:
4561 	case POSIX_FADV_DONTNEED:
4562 		new = NULL;
4563 		break;
4564 	default:
4565 		return (EINVAL);
4566 	}
4567 	/* XXX: CAP_POSIX_FADVISE? */
4568 	error = fget(td, fd, cap_rights_init(&rights), &fp);
4569 	if (error != 0)
4570 		goto out;
4571 	if ((fp->f_ops->fo_flags & DFLAG_SEEKABLE) == 0) {
4572 		error = ESPIPE;
4573 		goto out;
4574 	}
4575 	if (fp->f_type != DTYPE_VNODE) {
4576 		error = ENODEV;
4577 		goto out;
4578 	}
4579 	vp = fp->f_vnode;
4580 	if (vp->v_type != VREG) {
4581 		error = ENODEV;
4582 		goto out;
4583 	}
4584 	if (len == 0)
4585 		end = OFF_MAX;
4586 	else
4587 		end = offset + len - 1;
4588 	switch (advice) {
4589 	case POSIX_FADV_SEQUENTIAL:
4590 	case POSIX_FADV_RANDOM:
4591 	case POSIX_FADV_NOREUSE:
4592 		/*
4593 		 * Try to merge any existing non-standard region with
4594 		 * this new region if possible, otherwise create a new
4595 		 * non-standard region for this request.
4596 		 */
4597 		mtx_pool_lock(mtxpool_sleep, fp);
4598 		fa = fp->f_advice;
4599 		if (fa != NULL && fa->fa_advice == advice &&
4600 		    ((fa->fa_start <= end && fa->fa_end >= offset) ||
4601 		    (end != OFF_MAX && fa->fa_start == end + 1) ||
4602 		    (fa->fa_end != OFF_MAX && fa->fa_end + 1 == offset))) {
4603 			if (offset < fa->fa_start)
4604 				fa->fa_start = offset;
4605 			if (end > fa->fa_end)
4606 				fa->fa_end = end;
4607 		} else {
4608 			new->fa_advice = advice;
4609 			new->fa_start = offset;
4610 			new->fa_end = end;
4611 			new->fa_prevstart = 0;
4612 			new->fa_prevend = 0;
4613 			fp->f_advice = new;
4614 			new = fa;
4615 		}
4616 		mtx_pool_unlock(mtxpool_sleep, fp);
4617 		break;
4618 	case POSIX_FADV_NORMAL:
4619 		/*
4620 		 * If a the "normal" region overlaps with an existing
4621 		 * non-standard region, trim or remove the
4622 		 * non-standard region.
4623 		 */
4624 		mtx_pool_lock(mtxpool_sleep, fp);
4625 		fa = fp->f_advice;
4626 		if (fa != NULL) {
4627 			if (offset <= fa->fa_start && end >= fa->fa_end) {
4628 				new = fa;
4629 				fp->f_advice = NULL;
4630 			} else if (offset <= fa->fa_start &&
4631 			    end >= fa->fa_start)
4632 				fa->fa_start = end + 1;
4633 			else if (offset <= fa->fa_end && end >= fa->fa_end)
4634 				fa->fa_end = offset - 1;
4635 			else if (offset >= fa->fa_start && end <= fa->fa_end) {
4636 				/*
4637 				 * If the "normal" region is a middle
4638 				 * portion of the existing
4639 				 * non-standard region, just remove
4640 				 * the whole thing rather than picking
4641 				 * one side or the other to
4642 				 * preserve.
4643 				 */
4644 				new = fa;
4645 				fp->f_advice = NULL;
4646 			}
4647 		}
4648 		mtx_pool_unlock(mtxpool_sleep, fp);
4649 		break;
4650 	case POSIX_FADV_WILLNEED:
4651 	case POSIX_FADV_DONTNEED:
4652 		error = VOP_ADVISE(vp, offset, end, advice);
4653 		break;
4654 	}
4655 out:
4656 	if (fp != NULL)
4657 		fdrop(fp, td);
4658 	free(new, M_FADVISE);
4659 	return (error);
4660 }
4661 
4662 int
4663 sys_posix_fadvise(struct thread *td, struct posix_fadvise_args *uap)
4664 {
4665 
4666 	td->td_retval[0] = kern_posix_fadvise(td, uap->fd, uap->offset,
4667 	    uap->len, uap->advice);
4668 	return (0);
4669 }
4670