xref: /openbsd-src/sys/kern/vfs_syscalls.c (revision 385dfb3d3203498d950451e384fe44afa9827fb1)
1 /*	$OpenBSD: vfs_syscalls.c,v 1.372 2025/01/29 14:57:19 mpi Exp $	*/
2 /*	$NetBSD: vfs_syscalls.c,v 1.71 1996/04/23 10:29:02 mycroft Exp $	*/
3 
4 /*
5  * Copyright (c) 1989, 1993
6  *	The Regents of the University of California.  All rights reserved.
7  * (c) UNIX System Laboratories, Inc.
8  * All or some portions of this file are derived from material licensed
9  * to the University of California by American Telephone and Telegraph
10  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
11  * the permission of UNIX System Laboratories, Inc.
12  *
13  * Redistribution and use in source and binary forms, with or without
14  * modification, are permitted provided that the following conditions
15  * are met:
16  * 1. Redistributions of source code must retain the above copyright
17  *    notice, this list of conditions and the following disclaimer.
18  * 2. Redistributions in binary form must reproduce the above copyright
19  *    notice, this list of conditions and the following disclaimer in the
20  *    documentation and/or other materials provided with the distribution.
21  * 3. Neither the name of the University nor the names of its contributors
22  *    may be used to endorse or promote products derived from this software
23  *    without specific prior written permission.
24  *
25  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
26  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
27  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
28  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
29  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
30  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
31  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
32  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
33  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
34  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
35  * SUCH DAMAGE.
36  *
37  *	@(#)vfs_syscalls.c	8.28 (Berkeley) 12/10/94
38  */
39 
40 #include <sys/param.h>
41 #include <sys/systm.h>
42 #include <sys/namei.h>
43 #include <sys/filedesc.h>
44 #include <sys/conf.h>
45 #include <sys/fcntl.h>
46 #include <sys/file.h>
47 #include <sys/stat.h>
48 #include <sys/lock.h>
49 #include <sys/vnode.h>
50 #include <sys/mount.h>
51 #include <sys/proc.h>
52 #include <sys/pledge.h>
53 #include <sys/uio.h>
54 #include <sys/malloc.h>
55 #include <sys/pool.h>
56 #include <sys/ktrace.h>
57 #include <sys/unistd.h>
58 #include <sys/specdev.h>
59 #include <sys/resourcevar.h>
60 #include <sys/signalvar.h>
61 
62 #include <sys/syscallargs.h>
63 
64 extern int suid_clear;
65 
66 static int change_dir(struct nameidata *, struct proc *);
67 
68 void checkdirs(struct vnode *);
69 
70 int copyout_statfs(struct statfs *, void *, struct proc *);
71 
72 int doopenat(struct proc *, int, const char *, int, mode_t, register_t *);
73 int domknodat(struct proc *, int, const char *, mode_t, dev_t);
74 int dolinkat(struct proc *, int, const char *, int, const char *, int);
75 int dosymlinkat(struct proc *, const char *, int, const char *);
76 int dounlinkat(struct proc *, int, const char *, int);
77 int dofaccessat(struct proc *, int, const char *, int, int);
78 int dofstatat(struct proc *, int, const char *, struct stat *, int);
79 int dopathconfat(struct proc *, int, const char *, int, int, register_t *);
80 int doreadlinkat(struct proc *, int, const char *, char *, size_t,
81     register_t *);
82 int dochflagsat(struct proc *, int, const char *, u_int, int);
83 int dovchflags(struct proc *, struct vnode *, u_int);
84 int dofchmodat(struct proc *, int, const char *, mode_t, int);
85 int dofchownat(struct proc *, int, const char *, uid_t, gid_t, int);
86 int dorenameat(struct proc *, int, const char *, int, const char *);
87 int domkdirat(struct proc *, int, const char *, mode_t);
88 int doutimensat(struct proc *, int, const char *, struct timespec [2], int);
89 int dovutimens(struct proc *, struct vnode *, struct timespec [2]);
90 int dofutimens(struct proc *, int, struct timespec [2]);
91 int dounmount_leaf(struct mount *, int, struct proc *);
92 
93 /*
94  * Virtual File System System Calls
95  */
96 
97 /*
98  * Mount a file system.
99  */
100 int
101 sys_mount(struct proc *p, void *v, register_t *retval)
102 {
103 	struct sys_mount_args /* {
104 		syscallarg(const char *) type;
105 		syscallarg(const char *) path;
106 		syscallarg(int) flags;
107 		syscallarg(void *) data;
108 	} */ *uap = v;
109 	struct vnode *vp;
110 	struct mount *mp;
111 	int error, mntflag = 0;
112 	char fstypename[MFSNAMELEN];
113 	char fspath[MNAMELEN];
114 	struct nameidata nd;
115 	struct vfsconf *vfsp;
116 	int flags = SCARG(uap, flags);
117 	void *args = NULL;
118 
119 	if ((error = suser(p)))
120 		return (error);
121 
122 	/*
123 	 * Mount points must fit in MNAMELEN, not MAXPATHLEN.
124 	 */
125 	error = copyinstr(SCARG(uap, path), fspath, MNAMELEN, NULL);
126 	if (error)
127 		return(error);
128 
129 	/*
130 	 * Get vnode to be covered
131 	 */
132 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE, fspath, p);
133 	if ((error = namei(&nd)) != 0)
134 		goto fail;
135 	vp = nd.ni_vp;
136 	if (flags & MNT_UPDATE) {
137 		if ((vp->v_flag & VROOT) == 0) {
138 			vput(vp);
139 			error = EINVAL;
140 			goto fail;
141 		}
142 		mp = vp->v_mount;
143 		vfsp = mp->mnt_vfc;
144 
145 		args = malloc(vfsp->vfc_datasize, M_TEMP, M_WAITOK | M_ZERO);
146 		error = copyin(SCARG(uap, data), args, vfsp->vfc_datasize);
147 		if (error) {
148 			vput(vp);
149 			goto fail;
150 		}
151 
152 		mntflag = mp->mnt_flag;
153 		/*
154 		 * We only allow the filesystem to be reloaded if it
155 		 * is currently mounted read-only.
156 		 */
157 		if ((flags & MNT_RELOAD) &&
158 		    ((mp->mnt_flag & MNT_RDONLY) == 0)) {
159 			vput(vp);
160 			error = EOPNOTSUPP;	/* Needs translation */
161 			goto fail;
162 		}
163 
164 		if ((error = vfs_busy(mp, VB_READ|VB_NOWAIT)) != 0) {
165 			vput(vp);
166 			goto fail;
167 		}
168 		mp->mnt_flag |= flags & (MNT_RELOAD | MNT_UPDATE);
169 		goto update;
170 	}
171 	/*
172 	 * Do not allow disabling of permission checks unless exec and access to
173 	 * device files is disabled too.
174 	 */
175 	if ((flags & MNT_NOPERM) &&
176 	    (flags & (MNT_NODEV | MNT_NOEXEC)) != (MNT_NODEV | MNT_NOEXEC)) {
177 		vput(vp);
178 		error = EPERM;
179 		goto fail;
180 	}
181 	if ((error = vinvalbuf(vp, V_SAVE, p->p_ucred, p, 0, INFSLP)) != 0) {
182 		vput(vp);
183 		goto fail;
184 	}
185 	if (vp->v_type != VDIR) {
186 		vput(vp);
187 		goto fail;
188 	}
189 	error = copyinstr(SCARG(uap, type), fstypename, MFSNAMELEN, NULL);
190 	if (error) {
191 		vput(vp);
192 		goto fail;
193 	}
194 	vfsp = vfs_byname(fstypename);
195 	if (vfsp == NULL) {
196 		vput(vp);
197 		error = EOPNOTSUPP;
198 		goto fail;
199 	}
200 
201 	args = malloc(vfsp->vfc_datasize, M_TEMP, M_WAITOK | M_ZERO);
202 	error = copyin(SCARG(uap, data), args, vfsp->vfc_datasize);
203 	if (error) {
204 		vput(vp);
205 		goto fail;
206 	}
207 
208 	if (vp->v_mountedhere != NULL) {
209 		vput(vp);
210 		error = EBUSY;
211 		goto fail;
212 	}
213 
214 	/*
215 	 * Allocate and initialize the file system.
216 	 */
217 	mp = vfs_mount_alloc(vp, vfsp);
218 	mp->mnt_stat.f_owner = p->p_ucred->cr_uid;
219 
220 update:
221 	/* Ensure that the parent mountpoint does not get unmounted. */
222 	error = vfs_busy(vp->v_mount, VB_READ|VB_NOWAIT|VB_DUPOK);
223 	if (error) {
224 		if (mp->mnt_flag & MNT_UPDATE) {
225 			mp->mnt_flag = mntflag;
226 			vfs_unbusy(mp);
227 		} else {
228 			vfs_unbusy(mp);
229 			vfs_mount_free(mp);
230 		}
231 		vput(vp);
232 		goto fail;
233 	}
234 
235 	/*
236 	 * Set the mount level flags.
237 	 */
238 	if (flags & MNT_RDONLY)
239 		mp->mnt_flag |= MNT_RDONLY;
240 	else if (mp->mnt_flag & MNT_RDONLY)
241 		mp->mnt_flag |= MNT_WANTRDWR;
242 	mp->mnt_flag &=~ (MNT_NOSUID | MNT_NOEXEC | MNT_WXALLOWED | MNT_NODEV |
243 	    MNT_SYNCHRONOUS | MNT_ASYNC | MNT_NOATIME | MNT_NOPERM | MNT_FORCE);
244 	mp->mnt_flag |= flags & (MNT_NOSUID | MNT_NOEXEC | MNT_WXALLOWED |
245 	    MNT_NODEV | MNT_SYNCHRONOUS | MNT_ASYNC | MNT_NOATIME | MNT_NOPERM |
246 	    MNT_FORCE);
247 	/*
248 	 * Mount the filesystem.
249 	 */
250 	error = VFS_MOUNT(mp, fspath, args, &nd, p);
251 	if (!error) {
252 		mp->mnt_stat.f_ctime = gettime();
253 	}
254 	if (mp->mnt_flag & MNT_UPDATE) {
255 		vfs_unbusy(vp->v_mount);
256 		vput(vp);
257 		if (mp->mnt_flag & MNT_WANTRDWR)
258 			mp->mnt_flag &= ~MNT_RDONLY;
259 		mp->mnt_flag &= ~MNT_OP_FLAGS;
260 		if (error)
261 			mp->mnt_flag = mntflag;
262 
263 		if ((mp->mnt_flag & MNT_RDONLY) == 0) {
264 			if (mp->mnt_syncer == NULL)
265 				error = vfs_allocate_syncvnode(mp);
266 		} else {
267 			if (mp->mnt_syncer != NULL)
268 				vgone(mp->mnt_syncer);
269 			mp->mnt_syncer = NULL;
270 		}
271 
272 		vfs_unbusy(mp);
273 		goto fail;
274 	}
275 
276 	mp->mnt_flag &= ~MNT_OP_FLAGS;
277 	vp->v_mountedhere = mp;
278 
279 	/*
280 	 * Put the new filesystem on the mount list after root.
281 	 */
282 	cache_purge(vp);
283 	if (!error) {
284 		TAILQ_INSERT_TAIL(&mountlist, mp, mnt_list);
285 		checkdirs(vp);
286 		vfs_unbusy(vp->v_mount);
287 		VOP_UNLOCK(vp);
288 		if ((mp->mnt_flag & MNT_RDONLY) == 0)
289 			error = vfs_allocate_syncvnode(mp);
290 		vfs_unbusy(mp);
291 		(void) VFS_STATFS(mp, &mp->mnt_stat, p);
292 		if ((error = VFS_START(mp, 0, p)) != 0)
293 			vrele(vp);
294 	} else {
295 		mp->mnt_vnodecovered->v_mountedhere = NULL;
296 		vfs_unbusy(mp);
297 		vfs_mount_free(mp);
298 		vfs_unbusy(vp->v_mount);
299 		vput(vp);
300 	}
301 fail:
302 	if (args)
303 		free(args, M_TEMP, vfsp->vfc_datasize);
304 	return (error);
305 }
306 
307 /*
308  * Scan all active processes to see if any of them have a current
309  * or root directory onto which the new filesystem has just been
310  * mounted. If so, replace them with the new mount point, keeping
311  * track of how many were replaced.  That's the number of references
312  * the old vnode had that we've replaced, so finish by vrele()'ing
313  * it that many times.  This puts off any possible sleeping until
314  * we've finished walking the allprocess list.
315  */
316 void
317 checkdirs(struct vnode *olddp)
318 {
319 	struct filedesc *fdp;
320 	struct vnode *newdp;
321 	struct process *pr;
322 	u_int  free_count = 0;
323 
324 	if (olddp->v_usecount == 1)
325 		return;
326 	if (VFS_ROOT(olddp->v_mountedhere, &newdp))
327 		panic("mount: lost mount");
328 	LIST_FOREACH(pr, &allprocess, ps_list) {
329 		fdp = pr->ps_fd;
330 		if (fdp->fd_cdir == olddp) {
331 			free_count++;
332 			vref(newdp);
333 			fdp->fd_cdir = newdp;
334 		}
335 		if (fdp->fd_rdir == olddp) {
336 			free_count++;
337 			vref(newdp);
338 			fdp->fd_rdir = newdp;
339 		}
340 	}
341 	if (rootvnode == olddp) {
342 		free_count++;
343 		vref(newdp);
344 		rootvnode = newdp;
345 	}
346 	while (free_count-- > 0)
347 		vrele(olddp);
348 	vput(newdp);
349 }
350 
351 /*
352  * Unmount a file system.
353  *
354  * Note: unmount takes a path to the vnode mounted on as argument,
355  * not special file (as before).
356  */
357 int
358 sys_unmount(struct proc *p, void *v, register_t *retval)
359 {
360 	struct sys_unmount_args /* {
361 		syscallarg(const char *) path;
362 		syscallarg(int) flags;
363 	} */ *uap = v;
364 	struct vnode *vp;
365 	struct mount *mp;
366 	int error;
367 	struct nameidata nd;
368 
369 	if ((error = suser(p)) != 0)
370 		return (error);
371 
372 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
373 	    SCARG(uap, path), p);
374 	if ((error = namei(&nd)) != 0)
375 		return (error);
376 	vp = nd.ni_vp;
377 	mp = vp->v_mount;
378 
379 	/*
380 	 * Don't allow unmounting the root file system.
381 	 */
382 	if (mp->mnt_flag & MNT_ROOTFS) {
383 		vput(vp);
384 		return (EINVAL);
385 	}
386 
387 	/*
388 	 * Must be the root of the filesystem
389 	 */
390 	if ((vp->v_flag & VROOT) == 0) {
391 		vput(vp);
392 		return (EINVAL);
393 	}
394 	vput(vp);
395 
396 	if (vfs_busy(mp, VB_WRITE|VB_WAIT))
397 		return (EBUSY);
398 
399 	return (dounmount(mp, SCARG(uap, flags) & MNT_FORCE, p));
400 }
401 
402 /*
403  * Do the actual file system unmount.
404  */
405 int
406 dounmount(struct mount *mp, int flags, struct proc *p)
407 {
408 	SLIST_HEAD(, mount) mplist;
409 	struct mount *nmp;
410 	int error;
411 
412 	SLIST_INIT(&mplist);
413 	SLIST_INSERT_HEAD(&mplist, mp, mnt_dounmount);
414 
415 	/*
416 	 * Collect nested mount points. This takes advantage of the mount list
417 	 * being ordered - nested mount points come after their parent.
418 	 */
419 	while ((mp = TAILQ_NEXT(mp, mnt_list)) != NULL) {
420 		SLIST_FOREACH(nmp, &mplist, mnt_dounmount) {
421 			if (mp->mnt_vnodecovered == NULLVP ||
422 			    mp->mnt_vnodecovered->v_mount != nmp)
423 				continue;
424 
425 			if ((flags & MNT_FORCE) == 0) {
426 				error = EBUSY;
427 				goto err;
428 			}
429 			error = vfs_busy(mp, VB_WRITE|VB_WAIT|VB_DUPOK);
430 			if (error) {
431 				if ((flags & MNT_DOOMED)) {
432 					/*
433 					 * If the mount point was busy due to
434 					 * being unmounted, it has been removed
435 					 * from the mount list already.
436 					 * Restart the iteration from the last
437 					 * collected busy entry.
438 					 */
439 					mp = SLIST_FIRST(&mplist);
440 					break;
441 				}
442 				goto err;
443 			}
444 			SLIST_INSERT_HEAD(&mplist, mp, mnt_dounmount);
445 			break;
446 		}
447 	}
448 
449 	/*
450 	 * Nested mount points cannot appear during this loop as mounting
451 	 * requires a read lock for the parent mount point.
452 	 */
453 	while ((mp = SLIST_FIRST(&mplist)) != NULL) {
454 		SLIST_REMOVE(&mplist, mp, mount, mnt_dounmount);
455 		error = dounmount_leaf(mp, flags, p);
456 		if (error)
457 			goto err;
458 	}
459 	return (0);
460 
461 err:
462 	while ((mp = SLIST_FIRST(&mplist)) != NULL) {
463 		SLIST_REMOVE(&mplist, mp, mount, mnt_dounmount);
464 		vfs_unbusy(mp);
465 	}
466 	return (error);
467 }
468 
469 int
470 dounmount_leaf(struct mount *mp, int flags, struct proc *p)
471 {
472 	struct vnode *coveredvp;
473 	struct vnode *vp, *nvp;
474 	int error;
475 	int hadsyncer = 0;
476 
477 	mp->mnt_flag &=~ MNT_ASYNC;
478 	cache_purgevfs(mp);	/* remove cache entries for this file sys */
479 	if (mp->mnt_syncer != NULL) {
480 		hadsyncer = 1;
481 		vgone(mp->mnt_syncer);
482 		mp->mnt_syncer = NULL;
483 	}
484 
485 	/*
486 	 * Before calling file system unmount, make sure
487 	 * all unveils to vnodes in here are dropped.
488 	 */
489 	TAILQ_FOREACH_SAFE(vp , &mp->mnt_vnodelist, v_mntvnodes, nvp) {
490 		unveil_removevnode(vp);
491 	}
492 
493 	if (((mp->mnt_flag & MNT_RDONLY) ||
494 	    (error = VFS_SYNC(mp, MNT_WAIT, 0, p->p_ucred, p)) == 0) ||
495 	    (flags & MNT_FORCE))
496 		error = VFS_UNMOUNT(mp, flags, p);
497 
498 	if (error && !(flags & MNT_DOOMED)) {
499 		if ((mp->mnt_flag & MNT_RDONLY) == 0 && hadsyncer)
500 			(void) vfs_allocate_syncvnode(mp);
501 		vfs_unbusy(mp);
502 		return (error);
503 	}
504 
505 	TAILQ_REMOVE(&mountlist, mp, mnt_list);
506 	if ((coveredvp = mp->mnt_vnodecovered) != NULLVP) {
507 		coveredvp->v_mountedhere = NULL;
508 		vrele(coveredvp);
509 	}
510 
511 	if (!TAILQ_EMPTY(&mp->mnt_vnodelist))
512 		panic("unmount: dangling vnode");
513 
514 	vfs_unbusy(mp);
515 	vfs_mount_free(mp);
516 
517 	return (0);
518 }
519 
520 /*
521  * Sync each mounted filesystem.
522  */
523 int
524 sys_sync(struct proc *p, void *v, register_t *retval)
525 {
526 	struct mount *mp;
527 	int asyncflag;
528 
529 	TAILQ_FOREACH_REVERSE(mp, &mountlist, mntlist, mnt_list) {
530 		if (vfs_busy(mp, VB_READ|VB_NOWAIT))
531 			continue;
532 		if ((mp->mnt_flag & MNT_RDONLY) == 0) {
533 			asyncflag = mp->mnt_flag & MNT_ASYNC;
534 			mp->mnt_flag &= ~MNT_ASYNC;
535 			uvm_vnp_sync(mp);
536 			VFS_SYNC(mp, MNT_NOWAIT, 0, p->p_ucred, p);
537 			if (asyncflag)
538 				mp->mnt_flag |= MNT_ASYNC;
539 		}
540 		vfs_unbusy(mp);
541 	}
542 
543 	return (0);
544 }
545 
546 /*
547  * Change filesystem quotas.
548  */
549 int
550 sys_quotactl(struct proc *p, void *v, register_t *retval)
551 {
552 	struct sys_quotactl_args /* {
553 		syscallarg(const char *) path;
554 		syscallarg(int) cmd;
555 		syscallarg(int) uid;
556 		syscallarg(char *) arg;
557 	} */ *uap = v;
558 	struct mount *mp;
559 	int error;
560 	struct nameidata nd;
561 
562 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
563 	if ((error = namei(&nd)) != 0)
564 		return (error);
565 	mp = nd.ni_vp->v_mount;
566 	vrele(nd.ni_vp);
567 	return (VFS_QUOTACTL(mp, SCARG(uap, cmd), SCARG(uap, uid),
568 	    SCARG(uap, arg), p));
569 }
570 
571 int
572 copyout_statfs(struct statfs *sp, void *uaddr, struct proc *p)
573 {
574 	size_t co_sz1 = offsetof(struct statfs, f_fsid);
575 	size_t co_off2 = co_sz1 + sizeof(fsid_t);
576 	size_t co_sz2 = sizeof(struct statfs) - co_off2;
577 	char *s, *d;
578 	int error;
579 
580 	/* Don't let non-root see filesystem id (for NFS security) */
581 	if (suser(p)) {
582 		fsid_t fsid;
583 
584 		s = (char *)sp;
585 		d = (char *)uaddr;
586 
587 		memset(&fsid, 0, sizeof(fsid));
588 
589 		if ((error = copyout(s, d, co_sz1)) != 0)
590 			return (error);
591 		if ((error = copyout(&fsid, d + co_sz1, sizeof(fsid))) != 0)
592 			return (error);
593 		return (copyout(s + co_off2, d + co_off2, co_sz2));
594 	}
595 
596 	return (copyout(sp, uaddr, sizeof(*sp)));
597 }
598 
599 /*
600  * Get filesystem statistics.
601  */
602 int
603 sys_statfs(struct proc *p, void *v, register_t *retval)
604 {
605 	struct sys_statfs_args /* {
606 		syscallarg(const char *) path;
607 		syscallarg(struct statfs *) buf;
608 	} */ *uap = v;
609 	struct mount *mp;
610 	struct statfs *sp;
611 	int error;
612 	struct nameidata nd;
613 
614 	NDINIT(&nd, LOOKUP, FOLLOW | BYPASSUNVEIL, UIO_USERSPACE,
615 	    SCARG(uap, path), p);
616 	nd.ni_pledge = PLEDGE_RPATH;
617 	nd.ni_unveil = UNVEIL_READ;
618 	if ((error = namei(&nd)) != 0)
619 		return (error);
620 	mp = nd.ni_vp->v_mount;
621 	sp = &mp->mnt_stat;
622 	vrele(nd.ni_vp);
623 	if ((error = VFS_STATFS(mp, sp, p)) != 0)
624 		return (error);
625 	sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
626 
627 	return (copyout_statfs(sp, SCARG(uap, buf), p));
628 }
629 
630 /*
631  * Get filesystem statistics.
632  */
633 int
634 sys_fstatfs(struct proc *p, void *v, register_t *retval)
635 {
636 	struct sys_fstatfs_args /* {
637 		syscallarg(int) fd;
638 		syscallarg(struct statfs *) buf;
639 	} */ *uap = v;
640 	struct file *fp;
641 	struct mount *mp;
642 	struct statfs *sp;
643 	int error;
644 
645 	if ((error = getvnode(p, SCARG(uap, fd), &fp)) != 0)
646 		return (error);
647 	mp = ((struct vnode *)fp->f_data)->v_mount;
648 	if (!mp) {
649 		FRELE(fp, p);
650 		return (ENOENT);
651 	}
652 	sp = &mp->mnt_stat;
653 	error = VFS_STATFS(mp, sp, p);
654 	FRELE(fp, p);
655 	if (error)
656 		return (error);
657 	sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
658 
659 	return (copyout_statfs(sp, SCARG(uap, buf), p));
660 }
661 
662 /*
663  * Get statistics on all filesystems.
664  */
665 int
666 sys_getfsstat(struct proc *p, void *v, register_t *retval)
667 {
668 	struct sys_getfsstat_args /* {
669 		syscallarg(struct statfs *) buf;
670 		syscallarg(size_t) bufsize;
671 		syscallarg(int) flags;
672 	} */ *uap = v;
673 	struct mount *mp;
674 	struct statfs *sp;
675 	struct statfs *sfsp;
676 	size_t count, maxcount;
677 	int error, flags = SCARG(uap, flags);
678 
679 	maxcount = SCARG(uap, bufsize) / sizeof(struct statfs);
680 	sfsp = SCARG(uap, buf);
681 	count = 0;
682 
683 	TAILQ_FOREACH(mp, &mountlist, mnt_list) {
684 		if (vfs_busy(mp, VB_READ|VB_NOWAIT))
685 			continue;
686 		if (sfsp && count < maxcount) {
687 			sp = &mp->mnt_stat;
688 
689 			/* Refresh stats unless MNT_NOWAIT is specified */
690 			if (flags != MNT_NOWAIT &&
691 			    flags != MNT_LAZY &&
692 			    (flags == MNT_WAIT ||
693 			    flags == 0) &&
694 			    (error = VFS_STATFS(mp, sp, p))) {
695 				vfs_unbusy(mp);
696 				continue;
697 			}
698 
699 			sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
700 			error = (copyout_statfs(sp, sfsp, p));
701 			if (error) {
702 				vfs_unbusy(mp);
703 				return (error);
704 			}
705 			sfsp++;
706 		}
707 		count++;
708 		vfs_unbusy(mp);
709 	}
710 
711 	if (sfsp && count > maxcount)
712 		*retval = maxcount;
713 	else
714 		*retval = count;
715 
716 	return (0);
717 }
718 
719 /*
720  * Change current working directory to a given file descriptor.
721  */
722 int
723 sys_fchdir(struct proc *p, void *v, register_t *retval)
724 {
725 	struct sys_fchdir_args /* {
726 		syscallarg(int) fd;
727 	} */ *uap = v;
728 	struct filedesc *fdp = p->p_fd;
729 	struct vnode *vp, *tdp, *old_cdir;
730 	struct mount *mp;
731 	struct file *fp;
732 	int error;
733 
734 	if ((fp = fd_getfile(fdp, SCARG(uap, fd))) == NULL)
735 		return (EBADF);
736 	vp = fp->f_data;
737 	if (fp->f_type != DTYPE_VNODE || vp->v_type != VDIR) {
738 		FRELE(fp, p);
739 		return (ENOTDIR);
740 	}
741 	vref(vp);
742 	FRELE(fp, p);
743 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
744 	error = VOP_ACCESS(vp, VEXEC, p->p_ucred, p);
745 
746 	while (!error && (mp = vp->v_mountedhere) != NULL) {
747 		if (vfs_busy(mp, VB_READ|VB_WAIT))
748 			continue;
749 		error = VFS_ROOT(mp, &tdp);
750 		vfs_unbusy(mp);
751 		if (error)
752 			break;
753 		vput(vp);
754 		vp = tdp;
755 	}
756 	if (error) {
757 		vput(vp);
758 		return (error);
759 	}
760 	VOP_UNLOCK(vp);
761 	old_cdir = fdp->fd_cdir;
762 	fdp->fd_cdir = vp;
763 	vrele(old_cdir);
764 	return (0);
765 }
766 
767 /*
768  * Change current working directory (``.'').
769  */
770 int
771 sys_chdir(struct proc *p, void *v, register_t *retval)
772 {
773 	struct sys_chdir_args /* {
774 		syscallarg(const char *) path;
775 	} */ *uap = v;
776 	struct filedesc *fdp = p->p_fd;
777 	struct vnode *old_cdir;
778 	int error;
779 	struct nameidata nd;
780 
781 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
782 	    SCARG(uap, path), p);
783 	nd.ni_pledge = PLEDGE_RPATH;
784 	nd.ni_unveil = UNVEIL_READ;
785 	if ((error = change_dir(&nd, p)) != 0)
786 		return (error);
787 	old_cdir = fdp->fd_cdir;
788 	fdp->fd_cdir = nd.ni_vp;
789 	vrele(old_cdir);
790 	return (0);
791 }
792 
793 /*
794  * Change notion of root (``/'') directory.
795  */
796 int
797 sys_chroot(struct proc *p, void *v, register_t *retval)
798 {
799 	struct sys_chroot_args /* {
800 		syscallarg(const char *) path;
801 	} */ *uap = v;
802 	struct filedesc *fdp = p->p_fd;
803 	struct vnode *old_cdir, *old_rdir;
804 	int error;
805 	struct nameidata nd;
806 
807 	if ((error = suser(p)) != 0)
808 		return (error);
809 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
810 	    SCARG(uap, path), p);
811 	if ((error = change_dir(&nd, p)) != 0)
812 		return (error);
813 	if (fdp->fd_rdir != NULL) {
814 		/*
815 		 * A chroot() done inside a changed root environment does
816 		 * an automatic chdir to avoid the out-of-tree experience.
817 		 */
818 		vref(nd.ni_vp);
819 		old_rdir = fdp->fd_rdir;
820 		old_cdir = fdp->fd_cdir;
821 		fdp->fd_rdir = fdp->fd_cdir = nd.ni_vp;
822 		vrele(old_rdir);
823 		vrele(old_cdir);
824 	} else
825 		fdp->fd_rdir = nd.ni_vp;
826 	atomic_setbits_int(&p->p_p->ps_flags, PS_CHROOT);
827 	return (0);
828 }
829 
830 /*
831  * Common routine for chroot and chdir.
832  */
833 static int
834 change_dir(struct nameidata *ndp, struct proc *p)
835 {
836 	struct vnode *vp;
837 	int error;
838 
839 	if ((error = namei(ndp)) != 0)
840 		return (error);
841 	vp = ndp->ni_vp;
842 	if (vp->v_type != VDIR)
843 		error = ENOTDIR;
844 	else
845 		error = VOP_ACCESS(vp, VEXEC, p->p_ucred, p);
846 	if (error)
847 		vput(vp);
848 	else
849 		VOP_UNLOCK(vp);
850 	return (error);
851 }
852 
853 int
854 sys___realpath(struct proc *p, void *v, register_t *retval)
855 {
856 	struct sys___realpath_args /* {
857 		syscallarg(const char *) pathname;
858 		syscallarg(char *) resolved;
859 	} */ *uap = v;
860 	char *pathname;
861 	char *rpbuf;
862 	struct nameidata nd;
863 	size_t pathlen;
864 	int error = 0;
865 
866 	if (SCARG(uap, pathname) == NULL)
867 		return (EINVAL);
868 
869 	pathname = pool_get(&namei_pool, PR_WAITOK);
870 	rpbuf = pool_get(&namei_pool, PR_WAITOK);
871 
872 	if ((error = copyinstr(SCARG(uap, pathname), pathname, MAXPATHLEN,
873 	    &pathlen)))
874 		goto end;
875 
876 	if (pathlen == 1) { /* empty string "" */
877 		error = ENOENT;
878 		goto end;
879 	}
880 	if (pathlen < 2) {
881 		error = EINVAL;
882 		goto end;
883 	}
884 
885 	/* Get cwd for relative path if needed, prepend to rpbuf */
886 	rpbuf[0] = '\0';
887 	if (pathname[0] != '/') {
888 		int cwdlen = MAXPATHLEN * 4; /* for vfs_getcwd_common */
889 		char *cwdbuf, *bp;
890 
891 		cwdbuf = malloc(cwdlen, M_TEMP, M_WAITOK);
892 
893 		/* vfs_getcwd_common fills this in backwards */
894 		bp = &cwdbuf[cwdlen - 1];
895 		*bp = '\0';
896 
897 		error = vfs_getcwd_common(p->p_fd->fd_cdir, NULL, &bp, cwdbuf,
898 		    cwdlen/2, GETCWD_CHECK_ACCESS, p);
899 
900 		if (error) {
901 			free(cwdbuf, M_TEMP, cwdlen);
902 			goto end;
903 		}
904 
905 		if (strlcpy(rpbuf, bp, MAXPATHLEN) >= MAXPATHLEN) {
906 			free(cwdbuf, M_TEMP, cwdlen);
907 			error = ENAMETOOLONG;
908 			goto end;
909 		}
910 
911 		free(cwdbuf, M_TEMP, cwdlen);
912 	}
913 
914 	NDINIT(&nd, LOOKUP, FOLLOW | SAVENAME | REALPATH, UIO_SYSSPACE,
915 	    pathname, p);
916 
917 	nd.ni_cnd.cn_rpbuf = rpbuf;
918 	nd.ni_cnd.cn_rpi = strlen(rpbuf);
919 
920 	nd.ni_pledge = PLEDGE_RPATH;
921 	nd.ni_unveil = UNVEIL_READ;
922 	if ((error = namei(&nd)) != 0)
923 		goto end;
924 
925 	/* release reference from namei */
926 	if (nd.ni_vp)
927 		vrele(nd.ni_vp);
928 
929 	error = copyoutstr(nd.ni_cnd.cn_rpbuf, SCARG(uap, resolved),
930 	    MAXPATHLEN, NULL);
931 
932 #ifdef KTRACE
933 	if (KTRPOINT(p, KTR_NAMEI))
934 		ktrnamei(p, nd.ni_cnd.cn_rpbuf);
935 #endif
936 	pool_put(&namei_pool, nd.ni_cnd.cn_pnbuf);
937 end:
938 	pool_put(&namei_pool, rpbuf);
939 	pool_put(&namei_pool, pathname);
940 	return (error);
941 }
942 
943 int
944 sys_unveil(struct proc *p, void *v, register_t *retval)
945 {
946 	struct sys_unveil_args /* {
947 		syscallarg(const char *) path;
948 		syscallarg(const char *) permissions;
949 	} */ *uap = v;
950 	struct process *pr = p->p_p;
951 	char *pathname, *c;
952 	struct nameidata nd;
953 	size_t pathlen;
954 	char permissions[5];
955 	int error, allow;
956 
957 	if (SCARG(uap, path) == NULL && SCARG(uap, permissions) == NULL) {
958 		pr->ps_uvdone = 1;
959 		return (0);
960 	}
961 
962 	if (pr->ps_uvdone != 0)
963 		return EPERM;
964 
965 	error = copyinstr(SCARG(uap, permissions), permissions,
966 	    sizeof(permissions), NULL);
967 	if (error)
968 		return (error);
969 	pathname = pool_get(&namei_pool, PR_WAITOK);
970 	error = copyinstr(SCARG(uap, path), pathname, MAXPATHLEN, &pathlen);
971 	if (error)
972 		goto end;
973 
974 #ifdef KTRACE
975 	if (KTRPOINT(p, KTR_STRUCT))
976 		ktrstruct(p, "unveil", permissions, strlen(permissions));
977 #endif
978 	if (pathlen < 2) {
979 		error = EINVAL;
980 		goto end;
981 	}
982 
983 	/* find root "/" or "//" */
984 	for (c = pathname; *c != '\0'; c++) {
985 		if (*c != '/')
986 			break;
987 	}
988 	if (*c == '\0')
989 		/* root directory */
990 		NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | SAVENAME,
991 		    UIO_SYSSPACE, pathname, p);
992 	else
993 		NDINIT(&nd, CREATE, FOLLOW | LOCKLEAF | LOCKPARENT | SAVENAME,
994 		    UIO_SYSSPACE, pathname, p);
995 
996 	nd.ni_pledge = PLEDGE_UNVEIL;
997 	if ((error = namei(&nd)) != 0)
998 		goto end;
999 
1000 	/*
1001 	 * XXX Any access to the file or directory will allow us to
1002 	 * pledge path it
1003 	 */
1004 	allow = ((nd.ni_vp &&
1005 	    (VOP_ACCESS(nd.ni_vp, VREAD, p->p_ucred, p) == 0 ||
1006 	    VOP_ACCESS(nd.ni_vp, VWRITE, p->p_ucred, p) == 0 ||
1007 	    VOP_ACCESS(nd.ni_vp, VEXEC, p->p_ucred, p) == 0)) ||
1008 	    (nd.ni_dvp &&
1009 	    (VOP_ACCESS(nd.ni_dvp, VREAD, p->p_ucred, p) == 0 ||
1010 	    VOP_ACCESS(nd.ni_dvp, VWRITE, p->p_ucred, p) == 0 ||
1011 	    VOP_ACCESS(nd.ni_dvp, VEXEC, p->p_ucred, p) == 0)));
1012 
1013 	/* release lock from namei, but keep ref */
1014 	if (nd.ni_vp)
1015 		VOP_UNLOCK(nd.ni_vp);
1016 	if (nd.ni_dvp && nd.ni_dvp != nd.ni_vp)
1017 		VOP_UNLOCK(nd.ni_dvp);
1018 
1019 	if (allow)
1020 		error = unveil_add(p, &nd, permissions);
1021 	else
1022 		error = EPERM;
1023 
1024 	/* release vref from namei, but not vref from unveil_add */
1025 	if (nd.ni_vp)
1026 		vrele(nd.ni_vp);
1027 	if (nd.ni_dvp)
1028 		vrele(nd.ni_dvp);
1029 
1030 	pool_put(&namei_pool, nd.ni_cnd.cn_pnbuf);
1031 end:
1032 	pool_put(&namei_pool, pathname);
1033 
1034 	return (error);
1035 }
1036 
1037 /*
1038  * Check permissions, allocate an open file structure,
1039  * and call the device open routine if any.
1040  */
1041 int
1042 sys_open(struct proc *p, void *v, register_t *retval)
1043 {
1044 	struct sys_open_args /* {
1045 		syscallarg(const char *) path;
1046 		syscallarg(int) flags;
1047 		syscallarg(mode_t) mode;
1048 	} */ *uap = v;
1049 
1050 	return (doopenat(p, AT_FDCWD, SCARG(uap, path), SCARG(uap, flags),
1051 	    SCARG(uap, mode), retval));
1052 }
1053 
1054 int
1055 sys_openat(struct proc *p, void *v, register_t *retval)
1056 {
1057 	struct sys_openat_args /* {
1058 		syscallarg(int) fd;
1059 		syscallarg(const char *) path;
1060 		syscallarg(int) flags;
1061 		syscallarg(mode_t) mode;
1062 	} */ *uap = v;
1063 
1064 	return (doopenat(p, SCARG(uap, fd), SCARG(uap, path),
1065 	    SCARG(uap, flags), SCARG(uap, mode), retval));
1066 }
1067 
1068 int
1069 doopenat(struct proc *p, int fd, const char *path, int oflags, mode_t mode,
1070     register_t *retval)
1071 {
1072 	struct filedesc *fdp = p->p_fd;
1073 	struct file *fp;
1074 	struct vnode *vp;
1075 	struct vattr vattr;
1076 	int flags, cloexec, cmode;
1077 	int type, indx, error, localtrunc = 0;
1078 	struct flock lf;
1079 	struct nameidata nd;
1080 	uint64_t ni_pledge = 0;
1081 	u_char ni_unveil = 0;
1082 
1083 	if (oflags & (O_EXLOCK | O_SHLOCK)) {
1084 		error = pledge_flock(p);
1085 		if (error != 0)
1086 			return (error);
1087 	}
1088 
1089 	cloexec = (oflags & O_CLOEXEC) ? UF_EXCLOSE : 0;
1090 
1091 	fdplock(fdp);
1092 	if ((error = falloc(p, &fp, &indx)) != 0) {
1093 		fdpunlock(fdp);
1094 		return (error);
1095 	}
1096 	fdpunlock(fdp);
1097 
1098 	flags = FFLAGS(oflags);
1099 	if (flags & FREAD) {
1100 		ni_pledge |= PLEDGE_RPATH;
1101 		ni_unveil |= UNVEIL_READ;
1102 	}
1103 	if (flags & FWRITE) {
1104 		ni_pledge |= PLEDGE_WPATH;
1105 		ni_unveil |= UNVEIL_WRITE;
1106 	}
1107 	if (oflags & O_CREAT) {
1108 		ni_pledge |= PLEDGE_CPATH;
1109 		ni_unveil |= UNVEIL_CREATE;
1110 	}
1111 
1112 	cmode = ((mode &~ fdp->fd_cmask) & ALLPERMS) &~ S_ISTXT;
1113 	if ((p->p_p->ps_flags & PS_PLEDGE))
1114 		cmode &= ACCESSPERMS;
1115 	NDINITAT(&nd, 0, 0, UIO_USERSPACE, fd, path, p);
1116 	nd.ni_pledge = ni_pledge;
1117 	nd.ni_unveil = ni_unveil;
1118 	p->p_dupfd = -1;			/* XXX check for fdopen */
1119 	if ((flags & O_TRUNC) && (flags & (O_EXLOCK | O_SHLOCK))) {
1120 		localtrunc = 1;
1121 		flags &= ~O_TRUNC;	/* Must do truncate ourselves */
1122 	}
1123 	KERNEL_LOCK();
1124 	if ((error = vn_open(&nd, flags, cmode)) != 0) {
1125 		fdplock(fdp);
1126 		if (error == ENODEV &&
1127 		    p->p_dupfd >= 0 &&			/* XXX from fdopen */
1128 		    (error =
1129 			dupfdopen(p, indx, flags)) == 0) {
1130 			*retval = indx;
1131 			goto error;
1132 		}
1133 		if (error == ERESTART)
1134 			error = EINTR;
1135 		fdremove(fdp, indx);
1136 		goto error;
1137 	}
1138 	p->p_dupfd = 0;
1139 	vp = nd.ni_vp;
1140 	fp->f_flag = flags & FMASK;
1141 	fp->f_type = DTYPE_VNODE;
1142 	fp->f_ops = &vnops;
1143 	fp->f_data = vp;
1144 	if (flags & (O_EXLOCK | O_SHLOCK)) {
1145 		lf.l_whence = SEEK_SET;
1146 		lf.l_start = 0;
1147 		lf.l_len = 0;
1148 		if (flags & O_EXLOCK)
1149 			lf.l_type = F_WRLCK;
1150 		else
1151 			lf.l_type = F_RDLCK;
1152 		type = F_FLOCK;
1153 		if ((flags & FNONBLOCK) == 0)
1154 			type |= F_WAIT;
1155 		VOP_UNLOCK(vp);
1156 		error = VOP_ADVLOCK(vp, (caddr_t)fp, F_SETLK, &lf, type);
1157 		if (error) {
1158 			fdplock(fdp);
1159 			/* closef will vn_close the file for us. */
1160 			fdremove(fdp, indx);
1161 			goto error;
1162 		}
1163 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1164 		atomic_setbits_int(&fp->f_iflags, FIF_HASLOCK);
1165 	}
1166 	if (localtrunc) {
1167 		if ((fp->f_flag & FWRITE) == 0)
1168 			error = EACCES;
1169 		else if (vp->v_mount && (vp->v_mount->mnt_flag & MNT_RDONLY))
1170 			error = EROFS;
1171 		else if (vp->v_type == VDIR)
1172 			error = EISDIR;
1173 		else if ((error = vn_writechk(vp)) == 0) {
1174 			vattr_null(&vattr);
1175 			vattr.va_size = 0;
1176 			error = VOP_SETATTR(vp, &vattr, fp->f_cred, p);
1177 		}
1178 		if (error) {
1179 			VOP_UNLOCK(vp);
1180 			fdplock(fdp);
1181 			/* closef will close the file for us. */
1182 			fdremove(fdp, indx);
1183 			goto error;
1184 		}
1185 	}
1186 	VOP_UNLOCK(vp);
1187 	KERNEL_UNLOCK();
1188 	*retval = indx;
1189 	fdplock(fdp);
1190 	fdinsert(fdp, indx, cloexec, fp);
1191 	fdpunlock(fdp);
1192 	FRELE(fp, p);
1193 	return (error);
1194 error:
1195 	KERNEL_UNLOCK();
1196 	fdpunlock(fdp);
1197 	closef(fp, p);
1198 	return (error);
1199 }
1200 
1201 /*
1202  * Open a new created file (in /tmp) suitable for mmaping.
1203  */
1204 int
1205 sys___tmpfd(struct proc *p, void *v, register_t *retval)
1206 {
1207 	struct sys___tmpfd_args /* {
1208 		syscallarg(int) flags;
1209 	} */ *uap = v;
1210 	struct filedesc *fdp = p->p_fd;
1211 	struct file *fp;
1212 	struct vnode *vp;
1213 	int oflags = SCARG(uap, flags);
1214 	int flags, cloexec, cmode;
1215 	int indx, error;
1216 	unsigned int i;
1217 	struct nameidata nd;
1218 	char path[64];
1219 	static const char *letters = "ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789_-";
1220 
1221 	/* most flags are hardwired */
1222 	oflags = O_RDWR | O_CREAT | O_EXCL | O_NOFOLLOW | (oflags & O_CLOEXEC);
1223 
1224 	cloexec = (oflags & O_CLOEXEC) ? UF_EXCLOSE : 0;
1225 
1226 	fdplock(fdp);
1227 	if ((error = falloc(p, &fp, &indx)) != 0) {
1228 		fdpunlock(fdp);
1229 		return (error);
1230 	}
1231 	fdpunlock(fdp);
1232 
1233 	flags = FFLAGS(oflags);
1234 
1235 	arc4random_buf(path, sizeof(path));
1236 	memcpy(path, "/tmp/", 5);
1237 	for (i = 5; i < sizeof(path) - 1; i++)
1238 		path[i] = letters[(unsigned char)path[i] & 63];
1239 	path[sizeof(path)-1] = 0;
1240 
1241 	cmode = 0600;
1242 	NDINITAT(&nd, 0, KERNELPATH, UIO_SYSSPACE, AT_FDCWD, path, p);
1243 	if ((error = vn_open(&nd, flags, cmode)) != 0) {
1244 		if (error == ERESTART)
1245 			error = EINTR;
1246 		fdplock(fdp);
1247 		fdremove(fdp, indx);
1248 		fdpunlock(fdp);
1249 		closef(fp, p);
1250 		return (error);
1251 	}
1252 	vp = nd.ni_vp;
1253 	fp->f_flag = flags & FMASK;
1254 	fp->f_type = DTYPE_VNODE;
1255 	fp->f_ops = &vnops;
1256 	fp->f_data = vp;
1257 	VOP_UNLOCK(vp);
1258 	*retval = indx;
1259 	fdplock(fdp);
1260 	fdinsert(fdp, indx, cloexec, fp);
1261 	fdpunlock(fdp);
1262 	FRELE(fp, p);
1263 
1264 	/* unlink it */
1265 	/* XXX
1266 	 * there is a wee race here, although it is mostly inconsequential.
1267 	 * perhaps someday we can create a file like object without a name...
1268 	 */
1269 	NDINITAT(&nd, DELETE, KERNELPATH | LOCKPARENT | LOCKLEAF, UIO_SYSSPACE,
1270 	    AT_FDCWD, path, p);
1271 	if ((error = namei(&nd)) != 0) {
1272 		printf("can't unlink temp file! %d\n", error);
1273 		error = 0;
1274 	} else {
1275 		vp = nd.ni_vp;
1276 		uvm_vnp_uncache(vp);
1277 		error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
1278 		if (error) {
1279 			printf("error removing vop: %d\n", error);
1280 			error = 0;
1281 		}
1282 	}
1283 
1284 	return (error);
1285 }
1286 
1287 /*
1288  * Get file handle system call
1289  */
1290 int
1291 sys_getfh(struct proc *p, void *v, register_t *retval)
1292 {
1293 	struct sys_getfh_args /* {
1294 		syscallarg(const char *) fname;
1295 		syscallarg(fhandle_t *) fhp;
1296 	} */ *uap = v;
1297 	struct vnode *vp;
1298 	fhandle_t fh;
1299 	int error;
1300 	struct nameidata nd;
1301 
1302 	/*
1303 	 * Must be super user
1304 	 */
1305 	error = suser(p);
1306 	if (error)
1307 		return (error);
1308 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
1309 	    SCARG(uap, fname), p);
1310 	error = namei(&nd);
1311 	if (error)
1312 		return (error);
1313 	vp = nd.ni_vp;
1314 	memset(&fh, 0, sizeof(fh));
1315 	fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
1316 	error = VFS_VPTOFH(vp, &fh.fh_fid);
1317 	vput(vp);
1318 	if (error)
1319 		return (error);
1320 	error = copyout(&fh, SCARG(uap, fhp), sizeof(fh));
1321 	return (error);
1322 }
1323 
1324 /*
1325  * Open a file given a file handle.
1326  *
1327  * Check permissions, allocate an open file structure,
1328  * and call the device open routine if any.
1329  */
1330 int
1331 sys_fhopen(struct proc *p, void *v, register_t *retval)
1332 {
1333 	struct sys_fhopen_args /* {
1334 		syscallarg(const fhandle_t *) fhp;
1335 		syscallarg(int) flags;
1336 	} */ *uap = v;
1337 	struct filedesc *fdp = p->p_fd;
1338 	struct file *fp;
1339 	struct vnode *vp = NULL;
1340 	struct mount *mp;
1341 	struct ucred *cred = p->p_ucred;
1342 	int flags, cloexec;
1343 	int type, indx, error=0;
1344 	struct flock lf;
1345 	struct vattr va;
1346 	fhandle_t fh;
1347 
1348 	/*
1349 	 * Must be super user
1350 	 */
1351 	if ((error = suser(p)))
1352 		return (error);
1353 
1354 	flags = FFLAGS(SCARG(uap, flags));
1355 	if ((flags & (FREAD | FWRITE)) == 0)
1356 		return (EINVAL);
1357 	if ((flags & O_CREAT))
1358 		return (EINVAL);
1359 
1360 	cloexec = (flags & O_CLOEXEC) ? UF_EXCLOSE : 0;
1361 
1362 	fdplock(fdp);
1363 	if ((error = falloc(p, &fp, &indx)) != 0) {
1364 		fdpunlock(fdp);
1365 		fp = NULL;
1366 		goto bad;
1367 	}
1368 	fdpunlock(fdp);
1369 
1370 	if ((error = copyin(SCARG(uap, fhp), &fh, sizeof(fhandle_t))) != 0)
1371 		goto bad;
1372 
1373 	if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL) {
1374 		error = ESTALE;
1375 		goto bad;
1376 	}
1377 
1378 	if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp)) != 0) {
1379 		vp = NULL;	/* most likely unnecessary sanity for bad: */
1380 		goto bad;
1381 	}
1382 
1383 	/* Now do an effective vn_open */
1384 
1385 	if (vp->v_type == VSOCK) {
1386 		error = EOPNOTSUPP;
1387 		goto bad;
1388 	}
1389 	if ((flags & O_DIRECTORY) && vp->v_type != VDIR) {
1390 		error = ENOTDIR;
1391 		goto bad;
1392 	}
1393 	if (flags & FREAD) {
1394 		if ((error = VOP_ACCESS(vp, VREAD, cred, p)) != 0)
1395 			goto bad;
1396 	}
1397 	if (flags & (FWRITE | O_TRUNC)) {
1398 		if (vp->v_type == VDIR) {
1399 			error = EISDIR;
1400 			goto bad;
1401 		}
1402 		if ((error = VOP_ACCESS(vp, VWRITE, cred, p)) != 0 ||
1403 		    (error = vn_writechk(vp)) != 0)
1404 			goto bad;
1405 	}
1406 	if (flags & O_TRUNC) {
1407 		vattr_null(&va);
1408 		va.va_size = 0;
1409 		if ((error = VOP_SETATTR(vp, &va, cred, p)) != 0)
1410 			goto bad;
1411 	}
1412 	if ((error = VOP_OPEN(vp, flags, cred, p)) != 0)
1413 		goto bad;
1414 	if (flags & FWRITE)
1415 		vp->v_writecount++;
1416 
1417 	/* done with modified vn_open, now finish what sys_open does. */
1418 
1419 	fp->f_flag = flags & FMASK;
1420 	fp->f_type = DTYPE_VNODE;
1421 	fp->f_ops = &vnops;
1422 	fp->f_data = vp;
1423 	if (flags & (O_EXLOCK | O_SHLOCK)) {
1424 		lf.l_whence = SEEK_SET;
1425 		lf.l_start = 0;
1426 		lf.l_len = 0;
1427 		if (flags & O_EXLOCK)
1428 			lf.l_type = F_WRLCK;
1429 		else
1430 			lf.l_type = F_RDLCK;
1431 		type = F_FLOCK;
1432 		if ((flags & FNONBLOCK) == 0)
1433 			type |= F_WAIT;
1434 		VOP_UNLOCK(vp);
1435 		error = VOP_ADVLOCK(vp, (caddr_t)fp, F_SETLK, &lf, type);
1436 		if (error) {
1437 			vp = NULL;	/* closef will vn_close the file */
1438 			goto bad;
1439 		}
1440 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1441 		atomic_setbits_int(&fp->f_iflags, FIF_HASLOCK);
1442 	}
1443 	VOP_UNLOCK(vp);
1444 	*retval = indx;
1445 	fdplock(fdp);
1446 	fdinsert(fdp, indx, cloexec, fp);
1447 	fdpunlock(fdp);
1448 	FRELE(fp, p);
1449 	return (0);
1450 
1451 bad:
1452 	if (fp) {
1453 		fdplock(fdp);
1454 		fdremove(fdp, indx);
1455 		fdpunlock(fdp);
1456 		closef(fp, p);
1457 		if (vp != NULL)
1458 			vput(vp);
1459 	}
1460 	return (error);
1461 }
1462 
1463 int
1464 sys_fhstat(struct proc *p, void *v, register_t *retval)
1465 {
1466 	struct sys_fhstat_args /* {
1467 		syscallarg(const fhandle_t *) fhp;
1468 		syscallarg(struct stat *) sb;
1469 	} */ *uap = v;
1470 	struct stat sb;
1471 	int error;
1472 	fhandle_t fh;
1473 	struct mount *mp;
1474 	struct vnode *vp;
1475 
1476 	/*
1477 	 * Must be super user
1478 	 */
1479 	if ((error = suser(p)))
1480 		return (error);
1481 
1482 	if ((error = copyin(SCARG(uap, fhp), &fh, sizeof(fhandle_t))) != 0)
1483 		return (error);
1484 
1485 	if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL)
1486 		return (ESTALE);
1487 	if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp)))
1488 		return (error);
1489 	error = vn_stat(vp, &sb, p);
1490 	vput(vp);
1491 	if (error)
1492 		return (error);
1493 	error = copyout(&sb, SCARG(uap, sb), sizeof(sb));
1494 	return (error);
1495 }
1496 
1497 int
1498 sys_fhstatfs(struct proc *p, void *v, register_t *retval)
1499 {
1500 	struct sys_fhstatfs_args /* {
1501 		syscallarg(const fhandle_t *) fhp;
1502 		syscallarg(struct statfs *) buf;
1503 	} */ *uap = v;
1504 	struct statfs *sp;
1505 	fhandle_t fh;
1506 	struct mount *mp;
1507 	struct vnode *vp;
1508 	int error;
1509 
1510 	/*
1511 	 * Must be super user
1512 	 */
1513 	if ((error = suser(p)))
1514 		return (error);
1515 
1516 	if ((error = copyin(SCARG(uap, fhp), &fh, sizeof(fhandle_t))) != 0)
1517 		return (error);
1518 
1519 	if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL)
1520 		return (ESTALE);
1521 	if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp)))
1522 		return (error);
1523 	mp = vp->v_mount;
1524 	sp = &mp->mnt_stat;
1525 	vput(vp);
1526 	if ((error = VFS_STATFS(mp, sp, p)) != 0)
1527 		return (error);
1528 	sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
1529 	return (copyout(sp, SCARG(uap, buf), sizeof(*sp)));
1530 }
1531 
1532 /*
1533  * Create a special file or named pipe.
1534  */
1535 int
1536 sys_mknod(struct proc *p, void *v, register_t *retval)
1537 {
1538 	struct sys_mknod_args /* {
1539 		syscallarg(const char *) path;
1540 		syscallarg(mode_t) mode;
1541 		syscallarg(int) dev;
1542 	} */ *uap = v;
1543 
1544 	return (domknodat(p, AT_FDCWD, SCARG(uap, path), SCARG(uap, mode),
1545 	    SCARG(uap, dev)));
1546 }
1547 
1548 int
1549 sys_mknodat(struct proc *p, void *v, register_t *retval)
1550 {
1551 	struct sys_mknodat_args /* {
1552 		syscallarg(int) fd;
1553 		syscallarg(const char *) path;
1554 		syscallarg(mode_t) mode;
1555 		syscallarg(dev_t) dev;
1556 	} */ *uap = v;
1557 
1558 	return (domknodat(p, SCARG(uap, fd), SCARG(uap, path),
1559 	    SCARG(uap, mode), SCARG(uap, dev)));
1560 }
1561 
1562 int
1563 domknodat(struct proc *p, int fd, const char *path, mode_t mode, dev_t dev)
1564 {
1565 	struct vnode *vp;
1566 	struct vattr vattr;
1567 	int error;
1568 	struct nameidata nd;
1569 
1570 	if (dev == VNOVAL)
1571 		return (EINVAL);
1572 	NDINITAT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, fd, path, p);
1573 	nd.ni_pledge = PLEDGE_DPATH;
1574 	nd.ni_unveil = UNVEIL_CREATE;
1575 	if ((error = namei(&nd)) != 0)
1576 		return (error);
1577 	vp = nd.ni_vp;
1578 	if (!S_ISFIFO(mode) || dev != 0) {
1579 		if (!vnoperm(nd.ni_dvp) && (error = suser(p)) != 0)
1580 			goto out;
1581 		if (p->p_fd->fd_rdir) {
1582 			error = EINVAL;
1583 			goto out;
1584 		}
1585 	}
1586 	if (vp != NULL)
1587 		error = EEXIST;
1588 	else {
1589 		vattr_null(&vattr);
1590 		vattr.va_mode = (mode & ALLPERMS) &~ p->p_fd->fd_cmask;
1591 		if ((p->p_p->ps_flags & PS_PLEDGE))
1592 			vattr.va_mode &= ACCESSPERMS;
1593 		vattr.va_rdev = dev;
1594 
1595 		switch (mode & S_IFMT) {
1596 		case S_IFMT:	/* used by badsect to flag bad sectors */
1597 			vattr.va_type = VBAD;
1598 			break;
1599 		case S_IFCHR:
1600 			vattr.va_type = VCHR;
1601 			break;
1602 		case S_IFBLK:
1603 			vattr.va_type = VBLK;
1604 			break;
1605 		case S_IFIFO:
1606 #ifndef FIFO
1607 			error = EOPNOTSUPP;
1608 			break;
1609 #else
1610 			if (dev == 0) {
1611 				vattr.va_type = VFIFO;
1612 				break;
1613 			}
1614 			/* FALLTHROUGH */
1615 #endif /* FIFO */
1616 		default:
1617 			error = EINVAL;
1618 			break;
1619 		}
1620 	}
1621 out:
1622 	if (!error) {
1623 		error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
1624 		vput(nd.ni_dvp);
1625 	} else {
1626 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1627 		if (nd.ni_dvp == vp)
1628 			vrele(nd.ni_dvp);
1629 		else
1630 			vput(nd.ni_dvp);
1631 		if (vp)
1632 			vrele(vp);
1633 	}
1634 	return (error);
1635 }
1636 
1637 /*
1638  * Create a named pipe.
1639  */
1640 int
1641 sys_mkfifo(struct proc *p, void *v, register_t *retval)
1642 {
1643 	struct sys_mkfifo_args /* {
1644 		syscallarg(const char *) path;
1645 		syscallarg(mode_t) mode;
1646 	} */ *uap = v;
1647 
1648 	return (domknodat(p, AT_FDCWD, SCARG(uap, path),
1649 	    (SCARG(uap, mode) & ALLPERMS) | S_IFIFO, 0));
1650 }
1651 
1652 int
1653 sys_mkfifoat(struct proc *p, void *v, register_t *retval)
1654 {
1655 	struct sys_mkfifoat_args /* {
1656 		syscallarg(int) fd;
1657 		syscallarg(const char *) path;
1658 		syscallarg(mode_t) mode;
1659 	} */ *uap = v;
1660 
1661 	return (domknodat(p, SCARG(uap, fd), SCARG(uap, path),
1662 	    (SCARG(uap, mode) & ALLPERMS) | S_IFIFO, 0));
1663 }
1664 
1665 /*
1666  * Make a hard file link.
1667  */
1668 int
1669 sys_link(struct proc *p, void *v, register_t *retval)
1670 {
1671 	struct sys_link_args /* {
1672 		syscallarg(const char *) path;
1673 		syscallarg(const char *) link;
1674 	} */ *uap = v;
1675 
1676 	return (dolinkat(p, AT_FDCWD, SCARG(uap, path), AT_FDCWD,
1677 	    SCARG(uap, link), AT_SYMLINK_FOLLOW));
1678 }
1679 
1680 int
1681 sys_linkat(struct proc *p, void *v, register_t *retval)
1682 {
1683 	struct sys_linkat_args /* {
1684 		syscallarg(int) fd1;
1685 		syscallarg(const char *) path1;
1686 		syscallarg(int) fd2;
1687 		syscallarg(const char *) path2;
1688 		syscallarg(int) flag;
1689 	} */ *uap = v;
1690 
1691 	return (dolinkat(p, SCARG(uap, fd1), SCARG(uap, path1),
1692 	    SCARG(uap, fd2), SCARG(uap, path2), SCARG(uap, flag)));
1693 }
1694 
1695 int
1696 dolinkat(struct proc *p, int fd1, const char *path1, int fd2,
1697     const char *path2, int flag)
1698 {
1699 	struct vnode *vp;
1700 	struct nameidata nd;
1701 	int error, follow;
1702 
1703 	if (flag & ~AT_SYMLINK_FOLLOW)
1704 		return (EINVAL);
1705 
1706 	follow = (flag & AT_SYMLINK_FOLLOW) ? FOLLOW : NOFOLLOW;
1707 	NDINITAT(&nd, LOOKUP, follow, UIO_USERSPACE, fd1, path1, p);
1708 	nd.ni_pledge = PLEDGE_RPATH;
1709 	nd.ni_unveil = UNVEIL_READ;
1710 	if ((error = namei(&nd)) != 0)
1711 		return (error);
1712 	vp = nd.ni_vp;
1713 
1714 	if (vp->v_type == VDIR) {
1715 		error = EPERM;
1716 		goto out;
1717 	}
1718 
1719 	NDINITAT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, fd2, path2, p);
1720 	nd.ni_pledge = PLEDGE_CPATH;
1721 	nd.ni_unveil = UNVEIL_CREATE;
1722 	if ((error = namei(&nd)) != 0)
1723 		goto out;
1724 	if (nd.ni_vp) {
1725 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1726 		if (nd.ni_dvp == nd.ni_vp)
1727 			vrele(nd.ni_dvp);
1728 		else
1729 			vput(nd.ni_dvp);
1730 		vrele(nd.ni_vp);
1731 		error = EEXIST;
1732 		goto out;
1733 	}
1734 
1735 	/* No cross-mount links! */
1736 	if (nd.ni_dvp->v_mount != vp->v_mount) {
1737 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1738 		vput(nd.ni_dvp);
1739 		error = EXDEV;
1740 		goto out;
1741 	}
1742 
1743 	error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
1744 out:
1745 	vrele(vp);
1746 	return (error);
1747 }
1748 
1749 /*
1750  * Make a symbolic link.
1751  */
1752 int
1753 sys_symlink(struct proc *p, void *v, register_t *retval)
1754 {
1755 	struct sys_symlink_args /* {
1756 		syscallarg(const char *) path;
1757 		syscallarg(const char *) link;
1758 	} */ *uap = v;
1759 
1760 	return (dosymlinkat(p, SCARG(uap, path), AT_FDCWD, SCARG(uap, link)));
1761 }
1762 
1763 int
1764 sys_symlinkat(struct proc *p, void *v, register_t *retval)
1765 {
1766 	struct sys_symlinkat_args /* {
1767 		syscallarg(const char *) path;
1768 		syscallarg(int) fd;
1769 		syscallarg(const char *) link;
1770 	} */ *uap = v;
1771 
1772 	return (dosymlinkat(p, SCARG(uap, path), SCARG(uap, fd),
1773 	    SCARG(uap, link)));
1774 }
1775 
1776 int
1777 dosymlinkat(struct proc *p, const char *upath, int fd, const char *link)
1778 {
1779 	struct vattr vattr;
1780 	char *path;
1781 	int error;
1782 	struct nameidata nd;
1783 
1784 	path = pool_get(&namei_pool, PR_WAITOK);
1785 	error = copyinstr(upath, path, MAXPATHLEN, NULL);
1786 	if (error)
1787 		goto out;
1788 	NDINITAT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, fd, link, p);
1789 	nd.ni_pledge = PLEDGE_CPATH;
1790 	nd.ni_unveil = UNVEIL_CREATE;
1791 	if ((error = namei(&nd)) != 0)
1792 		goto out;
1793 	if (nd.ni_vp) {
1794 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1795 		if (nd.ni_dvp == nd.ni_vp)
1796 			vrele(nd.ni_dvp);
1797 		else
1798 			vput(nd.ni_dvp);
1799 		vrele(nd.ni_vp);
1800 		error = EEXIST;
1801 		goto out;
1802 	}
1803 	vattr_null(&vattr);
1804 	vattr.va_mode = ACCESSPERMS &~ p->p_fd->fd_cmask;
1805 	error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, path);
1806 out:
1807 	pool_put(&namei_pool, path);
1808 	return (error);
1809 }
1810 
1811 /*
1812  * Delete a name from the filesystem.
1813  */
1814 int
1815 sys_unlink(struct proc *p, void *v, register_t *retval)
1816 {
1817 	struct sys_unlink_args /* {
1818 		syscallarg(const char *) path;
1819 	} */ *uap = v;
1820 
1821 	return (dounlinkat(p, AT_FDCWD, SCARG(uap, path), 0));
1822 }
1823 
1824 int
1825 sys_unlinkat(struct proc *p, void *v, register_t *retval)
1826 {
1827 	struct sys_unlinkat_args /* {
1828 		syscallarg(int) fd;
1829 		syscallarg(const char *) path;
1830 		syscallarg(int) flag;
1831 	} */ *uap = v;
1832 
1833 	return (dounlinkat(p, SCARG(uap, fd), SCARG(uap, path),
1834 	    SCARG(uap, flag)));
1835 }
1836 
1837 int
1838 dounlinkat(struct proc *p, int fd, const char *path, int flag)
1839 {
1840 	struct vnode *vp;
1841 	int error;
1842 	struct nameidata nd;
1843 
1844 	if (flag & ~AT_REMOVEDIR)
1845 		return (EINVAL);
1846 
1847 	NDINITAT(&nd, DELETE, LOCKPARENT | LOCKLEAF, UIO_USERSPACE,
1848 	    fd, path, p);
1849 	nd.ni_pledge = PLEDGE_CPATH;
1850 	nd.ni_unveil = UNVEIL_CREATE;
1851 	if ((error = namei(&nd)) != 0)
1852 		return (error);
1853 	vp = nd.ni_vp;
1854 
1855 	if (flag & AT_REMOVEDIR) {
1856 		if (vp->v_type != VDIR) {
1857 			error = ENOTDIR;
1858 			goto out;
1859 		}
1860 		/*
1861 		 * No rmdir "." please.
1862 		 */
1863 		if (nd.ni_dvp == vp) {
1864 			error = EINVAL;
1865 			goto out;
1866 		}
1867 		/*
1868 		 * A mounted on directory cannot be deleted.
1869 		 */
1870 		if (vp->v_mountedhere != NULL) {
1871 			error = EBUSY;
1872 			goto out;
1873 		}
1874 	}
1875 
1876 	/*
1877 	 * The root of a mounted filesystem cannot be deleted.
1878 	 */
1879 	if (vp->v_flag & VROOT)
1880 		error = EBUSY;
1881 out:
1882 	if (!error) {
1883 		if (flag & AT_REMOVEDIR) {
1884 			error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
1885 		} else {
1886 			(void)uvm_vnp_uncache(vp);
1887 			error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
1888 		}
1889 	} else {
1890 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1891 		if (nd.ni_dvp == vp)
1892 			vrele(nd.ni_dvp);
1893 		else
1894 			vput(nd.ni_dvp);
1895 		vput(vp);
1896 	}
1897 	return (error);
1898 }
1899 
1900 /*
1901  * Reposition read/write file offset.
1902  */
1903 int
1904 sys_lseek(struct proc *p, void *v, register_t *retval)
1905 {
1906 	struct sys_lseek_args /* {
1907 		syscallarg(int) fd;
1908 		syscallarg(off_t) offset;
1909 		syscallarg(int) whence;
1910 	} */ *uap = v;
1911 	struct filedesc *fdp = p->p_fd;
1912 	struct file *fp;
1913 	off_t offset;
1914 	int error;
1915 
1916 	if ((fp = fd_getfile(fdp, SCARG(uap, fd))) == NULL)
1917 		return (EBADF);
1918 	if (fp->f_ops->fo_seek == NULL) {
1919 		error = ESPIPE;
1920 		goto bad;
1921 	}
1922 	offset = SCARG(uap, offset);
1923 
1924 	error = (*fp->f_ops->fo_seek)(fp, &offset, SCARG(uap, whence), p);
1925 	if (error)
1926 		goto bad;
1927 
1928 	*(off_t *)retval = offset;
1929 	mtx_enter(&fp->f_mtx);
1930 	fp->f_seek++;
1931 	mtx_leave(&fp->f_mtx);
1932 	error = 0;
1933  bad:
1934 	FRELE(fp, p);
1935 	return (error);
1936 }
1937 
1938 /*
1939  * Check access permissions.
1940  */
1941 int
1942 sys_access(struct proc *p, void *v, register_t *retval)
1943 {
1944 	struct sys_access_args /* {
1945 		syscallarg(const char *) path;
1946 		syscallarg(int) amode;
1947 	} */ *uap = v;
1948 
1949 	return (dofaccessat(p, AT_FDCWD, SCARG(uap, path),
1950 	    SCARG(uap, amode), 0));
1951 }
1952 
1953 int
1954 sys_faccessat(struct proc *p, void *v, register_t *retval)
1955 {
1956 	struct sys_faccessat_args /* {
1957 		syscallarg(int) fd;
1958 		syscallarg(const char *) path;
1959 		syscallarg(int) amode;
1960 		syscallarg(int) flag;
1961 	} */ *uap = v;
1962 
1963 	return (dofaccessat(p, SCARG(uap, fd), SCARG(uap, path),
1964 	    SCARG(uap, amode), SCARG(uap, flag)));
1965 }
1966 
1967 int
1968 dofaccessat(struct proc *p, int fd, const char *path, int amode, int flag)
1969 {
1970 	struct vnode *vp;
1971 	struct ucred *newcred, *oldcred;
1972 	struct nameidata nd;
1973 	int vflags = 0, error;
1974 
1975 	if (amode & ~(R_OK | W_OK | X_OK))
1976 		return (EINVAL);
1977 	if (flag & ~AT_EACCESS)
1978 		return (EINVAL);
1979 
1980 	newcred = NULL;
1981 	oldcred = p->p_ucred;
1982 
1983 	/*
1984 	 * If access as real ids was requested and they really differ,
1985 	 * give the thread new creds with them reset
1986 	 */
1987 	if ((flag & AT_EACCESS) == 0 &&
1988 	    (oldcred->cr_uid != oldcred->cr_ruid ||
1989 	    (oldcred->cr_gid != oldcred->cr_rgid))) {
1990 		p->p_ucred = newcred = crdup(oldcred);
1991 		newcred->cr_uid = newcred->cr_ruid;
1992 		newcred->cr_gid = newcred->cr_rgid;
1993 	}
1994 
1995 	NDINITAT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, fd, path, p);
1996 	nd.ni_pledge = PLEDGE_RPATH;
1997 	nd.ni_unveil = UNVEIL_READ;
1998 	if (amode & R_OK)
1999 		vflags |= VREAD;
2000 	if (amode & W_OK) {
2001 		vflags |= VWRITE;
2002 		nd.ni_unveil |= UNVEIL_WRITE;
2003 	}
2004 	if (amode & X_OK)
2005 		vflags |= VEXEC;
2006 	if ((error = namei(&nd)) != 0)
2007 		goto out;
2008 	vp = nd.ni_vp;
2009 
2010 	/* Flags == 0 means only check for existence. */
2011 	if (amode) {
2012 		error = VOP_ACCESS(vp, vflags, p->p_ucred, p);
2013 		if (!error && (vflags & VWRITE))
2014 			error = vn_writechk(vp);
2015 	}
2016 	vput(vp);
2017 out:
2018 	if (newcred != NULL) {
2019 		p->p_ucred = oldcred;
2020 		crfree(newcred);
2021 	}
2022 	return (error);
2023 }
2024 
2025 /*
2026  * Get file status; this version follows links.
2027  */
2028 int
2029 sys_stat(struct proc *p, void *v, register_t *retval)
2030 {
2031 	struct sys_stat_args /* {
2032 		syscallarg(const char *) path;
2033 		syscallarg(struct stat *) ub;
2034 	} */ *uap = v;
2035 
2036 	return (dofstatat(p, AT_FDCWD, SCARG(uap, path), SCARG(uap, ub), 0));
2037 }
2038 
2039 int
2040 sys_fstatat(struct proc *p, void *v, register_t *retval)
2041 {
2042 	struct sys_fstatat_args /* {
2043 		syscallarg(int) fd;
2044 		syscallarg(const char *) path;
2045 		syscallarg(struct stat *) buf;
2046 		syscallarg(int) flag;
2047 	} */ *uap = v;
2048 
2049 	return (dofstatat(p, SCARG(uap, fd), SCARG(uap, path),
2050 	    SCARG(uap, buf), SCARG(uap, flag)));
2051 }
2052 
2053 int
2054 dofstatat(struct proc *p, int fd, const char *path, struct stat *buf, int flag)
2055 {
2056 	struct stat sb;
2057 	int error, follow;
2058 	struct nameidata nd;
2059 
2060 	if (flag & ~AT_SYMLINK_NOFOLLOW)
2061 		return (EINVAL);
2062 
2063 
2064 	follow = (flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
2065 	NDINITAT(&nd, LOOKUP, follow | LOCKLEAF, UIO_USERSPACE, fd, path, p);
2066 	nd.ni_pledge = PLEDGE_RPATH;
2067 	nd.ni_unveil = UNVEIL_READ;
2068 	KERNEL_LOCK();
2069 	if ((error = namei(&nd)) != 0) {
2070 		KERNEL_UNLOCK();
2071 		return (error);
2072 	}
2073 	error = vn_stat(nd.ni_vp, &sb, p);
2074 	vput(nd.ni_vp);
2075 	KERNEL_UNLOCK();
2076 	if (error)
2077 		return (error);
2078 	/* Don't let non-root see generation numbers (for NFS security) */
2079 	if (suser(p))
2080 		sb.st_gen = 0;
2081 	error = copyout(&sb, buf, sizeof(sb));
2082 #ifdef KTRACE
2083 	if (error == 0 && KTRPOINT(p, KTR_STRUCT))
2084 		ktrstat(p, &sb);
2085 #endif
2086 	return (error);
2087 }
2088 
2089 /*
2090  * Get file status; this version does not follow links.
2091  */
2092 int
2093 sys_lstat(struct proc *p, void *v, register_t *retval)
2094 {
2095 	struct sys_lstat_args /* {
2096 		syscallarg(const char *) path;
2097 		syscallarg(struct stat *) ub;
2098 	} */ *uap = v;
2099 
2100 	return (dofstatat(p, AT_FDCWD, SCARG(uap, path), SCARG(uap, ub),
2101 	    AT_SYMLINK_NOFOLLOW));
2102 }
2103 
2104 /*
2105  * Get configurable pathname variables.
2106  */
2107 int
2108 sys_pathconf(struct proc *p, void *v, register_t *retval)
2109 {
2110 	struct sys_pathconf_args /* {
2111 		syscallarg(const char *) path;
2112 		syscallarg(int) name;
2113 	} */ *uap = v;
2114 
2115 	return dopathconfat(p, AT_FDCWD, SCARG(uap, path), SCARG(uap, name),
2116 	    0, retval);
2117 }
2118 
2119 int
2120 sys_pathconfat(struct proc *p, void *v, register_t *retval)
2121 {
2122 	struct sys_pathconfat_args /* {
2123 		syscallarg(int) fd;
2124 		syscallarg(const char *) path;
2125 		syscallarg(int) name;
2126 		syscallarg(int) flag;
2127 	} */ *uap = v;
2128 
2129 	return dopathconfat(p, SCARG(uap, fd), SCARG(uap, path),
2130 	    SCARG(uap, name), SCARG(uap, flag), retval);
2131 }
2132 
2133 int
2134 dopathconfat(struct proc *p, int fd, const char *path, int name, int flag,
2135     register_t *retval)
2136 {
2137 	int follow, error;
2138 	struct nameidata nd;
2139 
2140 	if (flag & ~AT_SYMLINK_NOFOLLOW)
2141 		return EINVAL;
2142 
2143 	follow = (flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
2144 	NDINITAT(&nd, LOOKUP, follow | LOCKLEAF, UIO_USERSPACE, fd, path, p);
2145 	nd.ni_pledge = PLEDGE_RPATH;
2146 	nd.ni_unveil = UNVEIL_READ;
2147 	if ((error = namei(&nd)) != 0)
2148 		return (error);
2149 	error = VOP_PATHCONF(nd.ni_vp, name, retval);
2150 	vput(nd.ni_vp);
2151 	return (error);
2152 }
2153 
2154 /*
2155  * Return target name of a symbolic link.
2156  */
2157 int
2158 sys_readlink(struct proc *p, void *v, register_t *retval)
2159 {
2160 	struct sys_readlink_args /* {
2161 		syscallarg(const char *) path;
2162 		syscallarg(char *) buf;
2163 		syscallarg(size_t) count;
2164 	} */ *uap = v;
2165 
2166 	return (doreadlinkat(p, AT_FDCWD, SCARG(uap, path), SCARG(uap, buf),
2167 	    SCARG(uap, count), retval));
2168 }
2169 
2170 int
2171 sys_readlinkat(struct proc *p, void *v, register_t *retval)
2172 {
2173 	struct sys_readlinkat_args /* {
2174 		syscallarg(int) fd;
2175 		syscallarg(const char *) path;
2176 		syscallarg(char *) buf;
2177 		syscallarg(size_t) count;
2178 	} */ *uap = v;
2179 
2180 	return (doreadlinkat(p, SCARG(uap, fd), SCARG(uap, path),
2181 	    SCARG(uap, buf), SCARG(uap, count), retval));
2182 }
2183 
2184 int
2185 doreadlinkat(struct proc *p, int fd, const char *path, char *buf,
2186     size_t count, register_t *retval)
2187 {
2188 	struct vnode *vp;
2189 	struct iovec aiov;
2190 	struct uio auio;
2191 	int error;
2192 	struct nameidata nd;
2193 
2194 	NDINITAT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF, UIO_USERSPACE, fd, path, p);
2195 	nd.ni_pledge = PLEDGE_RPATH;
2196 	nd.ni_unveil = UNVEIL_READ;
2197 	if ((error = namei(&nd)) != 0)
2198 		return (error);
2199 	vp = nd.ni_vp;
2200 	if (vp->v_type != VLNK)
2201 		error = EINVAL;
2202 	else {
2203 		aiov.iov_base = buf;
2204 		aiov.iov_len = count;
2205 		auio.uio_iov = &aiov;
2206 		auio.uio_iovcnt = 1;
2207 		auio.uio_offset = 0;
2208 		auio.uio_rw = UIO_READ;
2209 		auio.uio_segflg = UIO_USERSPACE;
2210 		auio.uio_procp = p;
2211 		auio.uio_resid = count;
2212 		error = VOP_READLINK(vp, &auio, p->p_ucred);
2213 		*retval = count - auio.uio_resid;
2214 	}
2215 	vput(vp);
2216 	return (error);
2217 }
2218 
2219 /*
2220  * Change flags of a file given a path name.
2221  */
2222 int
2223 sys_chflags(struct proc *p, void *v, register_t *retval)
2224 {
2225 	struct sys_chflags_args /* {
2226 		syscallarg(const char *) path;
2227 		syscallarg(u_int) flags;
2228 	} */ *uap = v;
2229 
2230 	return (dochflagsat(p, AT_FDCWD, SCARG(uap, path),
2231 	    SCARG(uap, flags), 0));
2232 }
2233 
2234 int
2235 sys_chflagsat(struct proc *p, void *v, register_t *retval)
2236 {
2237 	struct sys_chflagsat_args /* {
2238 		syscallarg(int) fd;
2239 		syscallarg(const char *) path;
2240 		syscallarg(u_int) flags;
2241 		syscallarg(int) atflags;
2242 	} */ *uap = v;
2243 
2244 	return (dochflagsat(p, SCARG(uap, fd), SCARG(uap, path),
2245 	    SCARG(uap, flags), SCARG(uap, atflags)));
2246 }
2247 
2248 int
2249 dochflagsat(struct proc *p, int fd, const char *path, u_int flags, int atflags)
2250 {
2251 	struct nameidata nd;
2252 	int error, follow;
2253 
2254 	if (atflags & ~AT_SYMLINK_NOFOLLOW)
2255 		return (EINVAL);
2256 
2257 	follow = (atflags & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
2258 	NDINITAT(&nd, LOOKUP, follow, UIO_USERSPACE, fd, path, p);
2259 	nd.ni_pledge = PLEDGE_FATTR | PLEDGE_RPATH;
2260 	nd.ni_unveil = UNVEIL_WRITE;
2261 	if ((error = namei(&nd)) != 0)
2262 		return (error);
2263 	return (dovchflags(p, nd.ni_vp, flags));
2264 }
2265 
2266 /*
2267  * Change flags of a file given a file descriptor.
2268  */
2269 int
2270 sys_fchflags(struct proc *p, void *v, register_t *retval)
2271 {
2272 	struct sys_fchflags_args /* {
2273 		syscallarg(int) fd;
2274 		syscallarg(u_int) flags;
2275 	} */ *uap = v;
2276 	struct file *fp;
2277 	struct vnode *vp;
2278 	int error;
2279 
2280 	if ((error = getvnode(p, SCARG(uap, fd), &fp)) != 0)
2281 		return (error);
2282 	vp = fp->f_data;
2283 	vref(vp);
2284 	FRELE(fp, p);
2285 	return (dovchflags(p, vp, SCARG(uap, flags)));
2286 }
2287 
2288 int
2289 dovchflags(struct proc *p, struct vnode *vp, u_int flags)
2290 {
2291 	struct vattr vattr;
2292 	int error;
2293 
2294 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2295 	if (vp->v_mount && vp->v_mount->mnt_flag & MNT_RDONLY)
2296 		error = EROFS;
2297 	else if (flags == VNOVAL)
2298 		error = EINVAL;
2299 	else {
2300 		if (suser(p)) {
2301 			if ((error = VOP_GETATTR(vp, &vattr, p->p_ucred, p))
2302 			    != 0)
2303 				goto out;
2304 			if (vattr.va_type == VCHR || vattr.va_type == VBLK) {
2305 				error = EINVAL;
2306 				goto out;
2307 			}
2308 		}
2309 		vattr_null(&vattr);
2310 		vattr.va_flags = flags;
2311 		error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
2312 	}
2313 out:
2314 	vput(vp);
2315 	return (error);
2316 }
2317 
2318 /*
2319  * Change mode of a file given path name.
2320  */
2321 int
2322 sys_chmod(struct proc *p, void *v, register_t *retval)
2323 {
2324 	struct sys_chmod_args /* {
2325 		syscallarg(const char *) path;
2326 		syscallarg(mode_t) mode;
2327 	} */ *uap = v;
2328 
2329 	return (dofchmodat(p, AT_FDCWD, SCARG(uap, path), SCARG(uap, mode), 0));
2330 }
2331 
2332 int
2333 sys_fchmodat(struct proc *p, void *v, register_t *retval)
2334 {
2335 	struct sys_fchmodat_args /* {
2336 		syscallarg(int) fd;
2337 		syscallarg(const char *) path;
2338 		syscallarg(mode_t) mode;
2339 		syscallarg(int) flag;
2340 	} */ *uap = v;
2341 
2342 	return (dofchmodat(p, SCARG(uap, fd), SCARG(uap, path),
2343 	    SCARG(uap, mode), SCARG(uap, flag)));
2344 }
2345 
2346 int
2347 dofchmodat(struct proc *p, int fd, const char *path, mode_t mode, int flag)
2348 {
2349 	struct vnode *vp;
2350 	struct vattr vattr;
2351 	int error, follow;
2352 	struct nameidata nd;
2353 
2354 	if (mode & ~(S_IFMT | ALLPERMS))
2355 		return (EINVAL);
2356 	if ((p->p_p->ps_flags & PS_PLEDGE))
2357 		mode &= ACCESSPERMS;
2358 	if (flag & ~AT_SYMLINK_NOFOLLOW)
2359 		return (EINVAL);
2360 
2361 	follow = (flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
2362 	NDINITAT(&nd, LOOKUP, follow, UIO_USERSPACE, fd, path, p);
2363 	nd.ni_pledge = PLEDGE_FATTR | PLEDGE_RPATH;
2364 	nd.ni_unveil = UNVEIL_WRITE;
2365 	if ((error = namei(&nd)) != 0)
2366 		return (error);
2367 	vp = nd.ni_vp;
2368 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2369 	if (vp->v_mount->mnt_flag & MNT_RDONLY)
2370 		error = EROFS;
2371 	else {
2372 		vattr_null(&vattr);
2373 		vattr.va_mode = mode & ALLPERMS;
2374 		error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
2375 	}
2376 	vput(vp);
2377 	return (error);
2378 }
2379 
2380 /*
2381  * Change mode of a file given a file descriptor.
2382  */
2383 int
2384 sys_fchmod(struct proc *p, void *v, register_t *retval)
2385 {
2386 	struct sys_fchmod_args /* {
2387 		syscallarg(int) fd;
2388 		syscallarg(mode_t) mode;
2389 	} */ *uap = v;
2390 	struct vattr vattr;
2391 	struct vnode *vp;
2392 	struct file *fp;
2393 	mode_t mode = SCARG(uap, mode);
2394 	int error;
2395 
2396 	if (mode & ~(S_IFMT | ALLPERMS))
2397 		return (EINVAL);
2398 	if ((p->p_p->ps_flags & PS_PLEDGE))
2399 		mode &= ACCESSPERMS;
2400 
2401 	if ((error = getvnode(p, SCARG(uap, fd), &fp)) != 0)
2402 		return (error);
2403 	vp = fp->f_data;
2404 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2405 	if (vp->v_mount && vp->v_mount->mnt_flag & MNT_RDONLY)
2406 		error = EROFS;
2407 	else {
2408 		vattr_null(&vattr);
2409 		vattr.va_mode = mode & ALLPERMS;
2410 		error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
2411 	}
2412 	VOP_UNLOCK(vp);
2413 	FRELE(fp, p);
2414 	return (error);
2415 }
2416 
2417 /*
2418  * Set ownership given a path name.
2419  */
2420 int
2421 sys_chown(struct proc *p, void *v, register_t *retval)
2422 {
2423 	struct sys_chown_args /* {
2424 		syscallarg(const char *) path;
2425 		syscallarg(uid_t) uid;
2426 		syscallarg(gid_t) gid;
2427 	} */ *uap = v;
2428 
2429 	return (dofchownat(p, AT_FDCWD, SCARG(uap, path), SCARG(uap, uid),
2430 	    SCARG(uap, gid), 0));
2431 }
2432 
2433 int
2434 sys_fchownat(struct proc *p, void *v, register_t *retval)
2435 {
2436 	struct sys_fchownat_args /* {
2437 		syscallarg(int) fd;
2438 		syscallarg(const char *) path;
2439 		syscallarg(uid_t) uid;
2440 		syscallarg(gid_t) gid;
2441 		syscallarg(int) flag;
2442 	} */ *uap = v;
2443 
2444 	return (dofchownat(p, SCARG(uap, fd), SCARG(uap, path),
2445 	    SCARG(uap, uid), SCARG(uap, gid), SCARG(uap, flag)));
2446 }
2447 
2448 int
2449 dofchownat(struct proc *p, int fd, const char *path, uid_t uid, gid_t gid,
2450     int flag)
2451 {
2452 	struct vnode *vp;
2453 	struct vattr vattr;
2454 	int error, follow;
2455 	struct nameidata nd;
2456 	mode_t mode;
2457 
2458 	if (flag & ~AT_SYMLINK_NOFOLLOW)
2459 		return (EINVAL);
2460 
2461 	follow = (flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
2462 	NDINITAT(&nd, LOOKUP, follow, UIO_USERSPACE, fd, path, p);
2463 	nd.ni_pledge = PLEDGE_CHOWN | PLEDGE_RPATH;
2464 	nd.ni_unveil = UNVEIL_WRITE;
2465 	if ((error = namei(&nd)) != 0)
2466 		return (error);
2467 	vp = nd.ni_vp;
2468 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2469 	if (vp->v_mount->mnt_flag & MNT_RDONLY)
2470 		error = EROFS;
2471 	else {
2472 		if ((error = pledge_chown(p, uid, gid)))
2473 			goto out;
2474 		if ((uid != -1 || gid != -1) &&
2475 		    !vnoperm(vp) &&
2476 		    (suser(p) || atomic_load_int(&suid_clear))) {
2477 			error = VOP_GETATTR(vp, &vattr, p->p_ucred, p);
2478 			if (error)
2479 				goto out;
2480 			mode = vattr.va_mode & ~(VSUID | VSGID);
2481 			if (mode == vattr.va_mode)
2482 				mode = VNOVAL;
2483 		} else
2484 			mode = VNOVAL;
2485 		vattr_null(&vattr);
2486 		vattr.va_uid = uid;
2487 		vattr.va_gid = gid;
2488 		vattr.va_mode = mode;
2489 		error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
2490 	}
2491 out:
2492 	vput(vp);
2493 	return (error);
2494 }
2495 
2496 /*
2497  * Set ownership given a path name, without following links.
2498  */
2499 int
2500 sys_lchown(struct proc *p, void *v, register_t *retval)
2501 {
2502 	struct sys_lchown_args /* {
2503 		syscallarg(const char *) path;
2504 		syscallarg(uid_t) uid;
2505 		syscallarg(gid_t) gid;
2506 	} */ *uap = v;
2507 	struct vnode *vp;
2508 	struct vattr vattr;
2509 	int error;
2510 	struct nameidata nd;
2511 	mode_t mode;
2512 	uid_t uid = SCARG(uap, uid);
2513 	gid_t gid = SCARG(uap, gid);
2514 
2515 	NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2516 	nd.ni_pledge = PLEDGE_CHOWN | PLEDGE_RPATH;
2517 	nd.ni_unveil = UNVEIL_WRITE;
2518 	if ((error = namei(&nd)) != 0)
2519 		return (error);
2520 	vp = nd.ni_vp;
2521 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2522 	if (vp->v_mount->mnt_flag & MNT_RDONLY)
2523 		error = EROFS;
2524 	else {
2525 		if ((error = pledge_chown(p, uid, gid)))
2526 			goto out;
2527 		if ((uid != -1 || gid != -1) &&
2528 		    !vnoperm(vp) &&
2529 		    (suser(p) || atomic_load_int(&suid_clear))) {
2530 			error = VOP_GETATTR(vp, &vattr, p->p_ucred, p);
2531 			if (error)
2532 				goto out;
2533 			mode = vattr.va_mode & ~(VSUID | VSGID);
2534 			if (mode == vattr.va_mode)
2535 				mode = VNOVAL;
2536 		} else
2537 			mode = VNOVAL;
2538 		vattr_null(&vattr);
2539 		vattr.va_uid = uid;
2540 		vattr.va_gid = gid;
2541 		vattr.va_mode = mode;
2542 		error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
2543 	}
2544 out:
2545 	vput(vp);
2546 	return (error);
2547 }
2548 
2549 /*
2550  * Set ownership given a file descriptor.
2551  */
2552 int
2553 sys_fchown(struct proc *p, void *v, register_t *retval)
2554 {
2555 	struct sys_fchown_args /* {
2556 		syscallarg(int) fd;
2557 		syscallarg(uid_t) uid;
2558 		syscallarg(gid_t) gid;
2559 	} */ *uap = v;
2560 	struct vnode *vp;
2561 	struct vattr vattr;
2562 	int error;
2563 	struct file *fp;
2564 	mode_t mode;
2565 	uid_t uid = SCARG(uap, uid);
2566 	gid_t gid = SCARG(uap, gid);
2567 
2568 	if ((error = getvnode(p, SCARG(uap, fd), &fp)) != 0)
2569 		return (error);
2570 	vp = fp->f_data;
2571 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2572 	if (vp->v_mount && (vp->v_mount->mnt_flag & MNT_RDONLY))
2573 		error = EROFS;
2574 	else {
2575 		if ((error = pledge_chown(p, uid, gid)))
2576 			goto out;
2577 		if ((uid != -1 || gid != -1) &&
2578 		    !vnoperm(vp) &&
2579 		    (suser(p) || atomic_load_int(&suid_clear))) {
2580 			error = VOP_GETATTR(vp, &vattr, p->p_ucred, p);
2581 			if (error)
2582 				goto out;
2583 			mode = vattr.va_mode & ~(VSUID | VSGID);
2584 			if (mode == vattr.va_mode)
2585 				mode = VNOVAL;
2586 		} else
2587 			mode = VNOVAL;
2588 		vattr_null(&vattr);
2589 		vattr.va_uid = uid;
2590 		vattr.va_gid = gid;
2591 		vattr.va_mode = mode;
2592 		error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
2593 	}
2594 out:
2595 	VOP_UNLOCK(vp);
2596 	FRELE(fp, p);
2597 	return (error);
2598 }
2599 
2600 /*
2601  * Set the access and modification times given a path name.
2602  */
2603 int
2604 sys_utimes(struct proc *p, void *v, register_t *retval)
2605 {
2606 	struct sys_utimes_args /* {
2607 		syscallarg(const char *) path;
2608 		syscallarg(const struct timeval *) tptr;
2609 	} */ *uap = v;
2610 
2611 	struct timespec ts[2];
2612 	struct timeval tv[2];
2613 	const struct timeval *tvp;
2614 	int error;
2615 
2616 	tvp = SCARG(uap, tptr);
2617 	if (tvp != NULL) {
2618 		error = copyin(tvp, tv, sizeof(tv));
2619 		if (error)
2620 			return (error);
2621 #ifdef KTRACE
2622 		if (KTRPOINT(p, KTR_STRUCT))
2623 			ktrabstimeval(p, &tv);
2624 #endif
2625 		if (!timerisvalid(&tv[0]) || !timerisvalid(&tv[1]))
2626 			return (EINVAL);
2627 		TIMEVAL_TO_TIMESPEC(&tv[0], &ts[0]);
2628 		TIMEVAL_TO_TIMESPEC(&tv[1], &ts[1]);
2629 	} else
2630 		ts[0].tv_nsec = ts[1].tv_nsec = UTIME_NOW;
2631 
2632 	return (doutimensat(p, AT_FDCWD, SCARG(uap, path), ts, 0));
2633 }
2634 
2635 int
2636 sys_utimensat(struct proc *p, void *v, register_t *retval)
2637 {
2638 	struct sys_utimensat_args /* {
2639 		syscallarg(int) fd;
2640 		syscallarg(const char *) path;
2641 		syscallarg(const struct timespec *) times;
2642 		syscallarg(int) flag;
2643 	} */ *uap = v;
2644 
2645 	struct timespec ts[2];
2646 	const struct timespec *tsp;
2647 	int error, i;
2648 
2649 	tsp = SCARG(uap, times);
2650 	if (tsp != NULL) {
2651 		error = copyin(tsp, ts, sizeof(ts));
2652 		if (error)
2653 			return (error);
2654 		for (i = 0; i < nitems(ts); i++) {
2655 			if (ts[i].tv_nsec == UTIME_NOW)
2656 				continue;
2657 			if (ts[i].tv_nsec == UTIME_OMIT)
2658 				continue;
2659 #ifdef KTRACE
2660 			if (KTRPOINT(p, KTR_STRUCT))
2661 				ktrabstimespec(p, &ts[i]);
2662 #endif
2663 			if (!timespecisvalid(&ts[i]))
2664 				return (EINVAL);
2665 		}
2666 	} else
2667 		ts[0].tv_nsec = ts[1].tv_nsec = UTIME_NOW;
2668 
2669 	return (doutimensat(p, SCARG(uap, fd), SCARG(uap, path), ts,
2670 	    SCARG(uap, flag)));
2671 }
2672 
2673 int
2674 doutimensat(struct proc *p, int fd, const char *path,
2675     struct timespec ts[2], int flag)
2676 {
2677 	struct vnode *vp;
2678 	int error, follow;
2679 	struct nameidata nd;
2680 
2681 	if (flag & ~AT_SYMLINK_NOFOLLOW)
2682 		return (EINVAL);
2683 
2684 	follow = (flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
2685 	NDINITAT(&nd, LOOKUP, follow, UIO_USERSPACE, fd, path, p);
2686 	nd.ni_pledge = PLEDGE_FATTR | PLEDGE_RPATH;
2687 	nd.ni_unveil = UNVEIL_WRITE;
2688 	if ((error = namei(&nd)) != 0)
2689 		return (error);
2690 	vp = nd.ni_vp;
2691 
2692 	return (dovutimens(p, vp, ts));
2693 }
2694 
2695 int
2696 dovutimens(struct proc *p, struct vnode *vp, struct timespec ts[2])
2697 {
2698 	struct vattr vattr;
2699 	struct timespec now;
2700 	int error;
2701 
2702 #ifdef KTRACE
2703 	/* if they're both UTIME_NOW, then don't report either */
2704 	if ((ts[0].tv_nsec != UTIME_NOW || ts[1].tv_nsec != UTIME_NOW) &&
2705 	    KTRPOINT(p, KTR_STRUCT)) {
2706 		ktrabstimespec(p, &ts[0]);
2707 		ktrabstimespec(p, &ts[1]);
2708 	}
2709 #endif
2710 
2711 	vattr_null(&vattr);
2712 
2713 	/*  make sure ctime is updated even if neither mtime nor atime is */
2714 	vattr.va_vaflags = VA_UTIMES_CHANGE;
2715 
2716 	if (ts[0].tv_nsec == UTIME_NOW || ts[1].tv_nsec == UTIME_NOW) {
2717 		if (ts[0].tv_nsec == UTIME_NOW && ts[1].tv_nsec == UTIME_NOW)
2718 			vattr.va_vaflags |= VA_UTIMES_NULL;
2719 
2720 		getnanotime(&now);
2721 		if (ts[0].tv_nsec == UTIME_NOW)
2722 			ts[0] = now;
2723 		if (ts[1].tv_nsec == UTIME_NOW)
2724 			ts[1] = now;
2725 	}
2726 
2727 	if (ts[0].tv_nsec != UTIME_OMIT)
2728 		vattr.va_atime = ts[0];
2729 	if (ts[1].tv_nsec != UTIME_OMIT)
2730 		vattr.va_mtime = ts[1];
2731 
2732 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2733 	if (vp->v_mount->mnt_flag & MNT_RDONLY)
2734 		error = EROFS;
2735 	else
2736 		error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
2737 	vput(vp);
2738 	return (error);
2739 }
2740 
2741 /*
2742  * Set the access and modification times given a file descriptor.
2743  */
2744 int
2745 sys_futimes(struct proc *p, void *v, register_t *retval)
2746 {
2747 	struct sys_futimes_args /* {
2748 		syscallarg(int) fd;
2749 		syscallarg(const struct timeval *) tptr;
2750 	} */ *uap = v;
2751 	struct timeval tv[2];
2752 	struct timespec ts[2];
2753 	const struct timeval *tvp;
2754 	int error;
2755 
2756 	tvp = SCARG(uap, tptr);
2757 	if (tvp != NULL) {
2758 		error = copyin(tvp, tv, sizeof(tv));
2759 		if (error)
2760 			return (error);
2761 #ifdef KTRACE
2762 		if (KTRPOINT(p, KTR_STRUCT)) {
2763 			ktrabstimeval(p, &tv[0]);
2764 			ktrabstimeval(p, &tv[1]);
2765 		}
2766 #endif
2767 		if (!timerisvalid(&tv[0]) || !timerisvalid(&tv[1]))
2768 			return (EINVAL);
2769 		TIMEVAL_TO_TIMESPEC(&tv[0], &ts[0]);
2770 		TIMEVAL_TO_TIMESPEC(&tv[1], &ts[1]);
2771 	} else
2772 		ts[0].tv_nsec = ts[1].tv_nsec = UTIME_NOW;
2773 
2774 	return (dofutimens(p, SCARG(uap, fd), ts));
2775 }
2776 
2777 int
2778 sys_futimens(struct proc *p, void *v, register_t *retval)
2779 {
2780 	struct sys_futimens_args /* {
2781 		syscallarg(int) fd;
2782 		syscallarg(const struct timespec *) times;
2783 	} */ *uap = v;
2784 	struct timespec ts[2];
2785 	const struct timespec *tsp;
2786 	int error, i;
2787 
2788 	tsp = SCARG(uap, times);
2789 	if (tsp != NULL) {
2790 		error = copyin(tsp, ts, sizeof(ts));
2791 		if (error)
2792 			return (error);
2793 		for (i = 0; i < nitems(ts); i++) {
2794 			if (ts[i].tv_nsec == UTIME_NOW)
2795 				continue;
2796 			if (ts[i].tv_nsec == UTIME_OMIT)
2797 				continue;
2798 #ifdef KTRACE
2799 			if (KTRPOINT(p, KTR_STRUCT))
2800 				ktrabstimespec(p, &ts[i]);
2801 #endif
2802 			if (!timespecisvalid(&ts[i]))
2803 				return (EINVAL);
2804 		}
2805 	} else
2806 		ts[0].tv_nsec = ts[1].tv_nsec = UTIME_NOW;
2807 
2808 	return (dofutimens(p, SCARG(uap, fd), ts));
2809 }
2810 
2811 int
2812 dofutimens(struct proc *p, int fd, struct timespec ts[2])
2813 {
2814 	struct file *fp;
2815 	struct vnode *vp;
2816 	int error;
2817 
2818 	if ((error = getvnode(p, fd, &fp)) != 0)
2819 		return (error);
2820 	vp = fp->f_data;
2821 	vref(vp);
2822 	FRELE(fp, p);
2823 
2824 	return (dovutimens(p, vp, ts));
2825 }
2826 
2827 /*
2828  * Truncate a file given a vnode.
2829  */
2830 int
2831 dotruncate(struct proc *p, struct vnode *vp, off_t len)
2832 {
2833 	struct vattr vattr;
2834 	int error;
2835 
2836 	if (len < 0)
2837 		return EINVAL;
2838 	if (vp->v_type == VDIR)
2839 		return EISDIR;
2840 	if ((error = vn_writechk(vp)) != 0)
2841 		return error;
2842 	if (vp->v_type == VREG && len > lim_cur_proc(p, RLIMIT_FSIZE)) {
2843 		if ((error = VOP_GETATTR(vp, &vattr, p->p_ucred, p)) != 0)
2844 			return error;
2845 		if (len > vattr.va_size) {
2846 			/* if extending over the limit, send signal and fail */
2847 			psignal(p, SIGXFSZ);
2848 			return EFBIG;
2849 		}
2850 	}
2851 	vattr_null(&vattr);
2852 	vattr.va_size = len;
2853 	return VOP_SETATTR(vp, &vattr, p->p_ucred, p);
2854 }
2855 
2856 /*
2857  * Truncate a file given its path name.
2858  */
2859 int
2860 sys_truncate(struct proc *p, void *v, register_t *retval)
2861 {
2862 	struct sys_truncate_args /* {
2863 		syscallarg(const char *) path;
2864 		syscallarg(off_t) length;
2865 	} */ *uap = v;
2866 	struct vnode *vp;
2867 	int error;
2868 	struct nameidata nd;
2869 
2870 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2871 	nd.ni_pledge = PLEDGE_FATTR | PLEDGE_RPATH;
2872 	nd.ni_unveil = UNVEIL_WRITE;
2873 	if ((error = namei(&nd)) != 0)
2874 		return (error);
2875 	vp = nd.ni_vp;
2876 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2877 	if ((error = VOP_ACCESS(vp, VWRITE, p->p_ucred, p)) == 0)
2878 		error = dotruncate(p, vp, SCARG(uap, length));
2879 	vput(vp);
2880 	return (error);
2881 }
2882 
2883 /*
2884  * Truncate a file given a file descriptor.
2885  */
2886 int
2887 sys_ftruncate(struct proc *p, void *v, register_t *retval)
2888 {
2889 	struct sys_ftruncate_args /* {
2890 		syscallarg(int) fd;
2891 		syscallarg(off_t) length;
2892 	} */ *uap = v;
2893 	struct vnode *vp;
2894 	struct file *fp;
2895 	int error;
2896 
2897 	if ((error = getvnode(p, SCARG(uap, fd), &fp)) != 0)
2898 		return (error);
2899 	if ((fp->f_flag & FWRITE) == 0) {
2900 		error = EINVAL;
2901 		goto bad;
2902 	}
2903 	vp = fp->f_data;
2904 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2905 	error = dotruncate(p, vp, SCARG(uap, length));
2906 	VOP_UNLOCK(vp);
2907 bad:
2908 	FRELE(fp, p);
2909 	return (error);
2910 }
2911 
2912 /*
2913  * Sync an open file.
2914  */
2915 int
2916 sys_fsync(struct proc *p, void *v, register_t *retval)
2917 {
2918 	struct sys_fsync_args /* {
2919 		syscallarg(int) fd;
2920 	} */ *uap = v;
2921 	struct vnode *vp;
2922 	struct file *fp;
2923 	int error;
2924 
2925 	if ((error = getvnode(p, SCARG(uap, fd), &fp)) != 0)
2926 		return (error);
2927 	vp = fp->f_data;
2928 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2929 	error = VOP_FSYNC(vp, fp->f_cred, MNT_WAIT, p);
2930 
2931 	VOP_UNLOCK(vp);
2932 	FRELE(fp, p);
2933 	return (error);
2934 }
2935 
2936 /*
2937  * Rename files.  Source and destination must either both be directories,
2938  * or both not be directories.  If target is a directory, it must be empty.
2939  */
2940 int
2941 sys_rename(struct proc *p, void *v, register_t *retval)
2942 {
2943 	struct sys_rename_args /* {
2944 		syscallarg(const char *) from;
2945 		syscallarg(const char *) to;
2946 	} */ *uap = v;
2947 
2948 	return (dorenameat(p, AT_FDCWD, SCARG(uap, from), AT_FDCWD,
2949 	    SCARG(uap, to)));
2950 }
2951 
2952 int
2953 sys_renameat(struct proc *p, void *v, register_t *retval)
2954 {
2955 	struct sys_renameat_args /* {
2956 		syscallarg(int) fromfd;
2957 		syscallarg(const char *) from;
2958 		syscallarg(int) tofd;
2959 		syscallarg(const char *) to;
2960 	} */ *uap = v;
2961 
2962 	return (dorenameat(p, SCARG(uap, fromfd), SCARG(uap, from),
2963 	    SCARG(uap, tofd), SCARG(uap, to)));
2964 }
2965 
2966 int
2967 dorenameat(struct proc *p, int fromfd, const char *from, int tofd,
2968     const char *to)
2969 {
2970 	struct vnode *tvp, *fvp, *tdvp;
2971 	struct nameidata fromnd, tond;
2972 	int error;
2973 	int flags;
2974 
2975 	NDINITAT(&fromnd, DELETE, WANTPARENT | SAVESTART, UIO_USERSPACE,
2976 	    fromfd, from, p);
2977 	fromnd.ni_pledge = PLEDGE_RPATH | PLEDGE_CPATH;
2978 	fromnd.ni_unveil = UNVEIL_READ | UNVEIL_CREATE;
2979 	if ((error = namei(&fromnd)) != 0)
2980 		return (error);
2981 	fvp = fromnd.ni_vp;
2982 
2983 	flags = LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART;
2984 	/*
2985 	 * rename("foo/", "bar/");  is  OK
2986 	 */
2987 	if (fvp->v_type == VDIR)
2988 		flags |= STRIPSLASHES;
2989 
2990 	NDINITAT(&tond, RENAME, flags, UIO_USERSPACE, tofd, to, p);
2991 	tond.ni_pledge = PLEDGE_CPATH;
2992 	tond.ni_unveil = UNVEIL_CREATE;
2993 	if ((error = namei(&tond)) != 0) {
2994 		VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
2995 		vrele(fromnd.ni_dvp);
2996 		vrele(fvp);
2997 		goto out1;
2998 	}
2999 	tdvp = tond.ni_dvp;
3000 	tvp = tond.ni_vp;
3001 	if (tvp != NULL) {
3002 		if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
3003 			error = ENOTDIR;
3004 			goto out;
3005 		} else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
3006 			error = EISDIR;
3007 			goto out;
3008 		}
3009 	}
3010 	if (fvp == tdvp)
3011 		error = EINVAL;
3012 	/*
3013 	 * If source is the same as the destination (that is the
3014 	 * same inode number)
3015 	 */
3016 	if (fvp == tvp)
3017 		error = -1;
3018 out:
3019 	if (!error) {
3020 		if (tvp) {
3021 			(void)uvm_vnp_uncache(tvp);
3022 		}
3023 		error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
3024 				   tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
3025 	} else {
3026 		VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd);
3027 		if (tdvp == tvp)
3028 			vrele(tdvp);
3029 		else
3030 			vput(tdvp);
3031 		if (tvp)
3032 			vput(tvp);
3033 		VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
3034 		vrele(fromnd.ni_dvp);
3035 		vrele(fvp);
3036 	}
3037 	vrele(tond.ni_startdir);
3038 	pool_put(&namei_pool, tond.ni_cnd.cn_pnbuf);
3039 out1:
3040 	if (fromnd.ni_startdir)
3041 		vrele(fromnd.ni_startdir);
3042 	pool_put(&namei_pool, fromnd.ni_cnd.cn_pnbuf);
3043 	if (error == -1)
3044 		return (0);
3045 	return (error);
3046 }
3047 
3048 /*
3049  * Make a directory file.
3050  */
3051 int
3052 sys_mkdir(struct proc *p, void *v, register_t *retval)
3053 {
3054 	struct sys_mkdir_args /* {
3055 		syscallarg(const char *) path;
3056 		syscallarg(mode_t) mode;
3057 	} */ *uap = v;
3058 
3059 	return (domkdirat(p, AT_FDCWD, SCARG(uap, path), SCARG(uap, mode)));
3060 }
3061 
3062 int
3063 sys_mkdirat(struct proc *p, void *v, register_t *retval)
3064 {
3065 	struct sys_mkdirat_args /* {
3066 		syscallarg(int) fd;
3067 		syscallarg(const char *) path;
3068 		syscallarg(mode_t) mode;
3069 	} */ *uap = v;
3070 
3071 	return (domkdirat(p, SCARG(uap, fd), SCARG(uap, path),
3072 	    SCARG(uap, mode)));
3073 }
3074 
3075 int
3076 domkdirat(struct proc *p, int fd, const char *path, mode_t mode)
3077 {
3078 	struct vnode *vp;
3079 	struct vattr vattr;
3080 	int error;
3081 	struct nameidata nd;
3082 
3083 	NDINITAT(&nd, CREATE, LOCKPARENT | STRIPSLASHES, UIO_USERSPACE,
3084 	    fd, path, p);
3085 	nd.ni_pledge = PLEDGE_CPATH;
3086 	nd.ni_unveil = UNVEIL_CREATE;
3087 	if ((error = namei(&nd)) != 0)
3088 		return (error);
3089 	vp = nd.ni_vp;
3090 	if (vp != NULL) {
3091 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
3092 		if (nd.ni_dvp == vp)
3093 			vrele(nd.ni_dvp);
3094 		else
3095 			vput(nd.ni_dvp);
3096 		vrele(vp);
3097 		return (EEXIST);
3098 	}
3099 	vattr_null(&vattr);
3100 	vattr.va_type = VDIR;
3101 	vattr.va_mode = (mode & ACCESSPERMS) &~ p->p_fd->fd_cmask;
3102 	error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
3103 	if (!error)
3104 		vput(nd.ni_vp);
3105 	return (error);
3106 }
3107 
3108 /*
3109  * Remove a directory file.
3110  */
3111 int
3112 sys_rmdir(struct proc *p, void *v, register_t *retval)
3113 {
3114 	struct sys_rmdir_args /* {
3115 		syscallarg(const char *) path;
3116 	} */ *uap = v;
3117 
3118 	return (dounlinkat(p, AT_FDCWD, SCARG(uap, path), AT_REMOVEDIR));
3119 }
3120 
3121 /*
3122  * Read a block of directory entries in a file system independent format.
3123  */
3124 int
3125 sys_getdents(struct proc *p, void *v, register_t *retval)
3126 {
3127 	struct sys_getdents_args /* {
3128 		syscallarg(int) fd;
3129 		syscallarg(void *) buf;
3130 		syscallarg(size_t) buflen;
3131 	} */ *uap = v;
3132 	struct vnode *vp;
3133 	struct file *fp;
3134 	struct uio auio;
3135 	struct iovec aiov;
3136 	size_t buflen;
3137 	int error, eofflag;
3138 
3139 	buflen = SCARG(uap, buflen);
3140 
3141 	if (buflen > INT_MAX)
3142 		return (EINVAL);
3143 	if ((error = getvnode(p, SCARG(uap, fd), &fp)) != 0)
3144 		return (error);
3145 	if ((fp->f_flag & FREAD) == 0) {
3146 		error = EBADF;
3147 		goto bad;
3148 	}
3149 	vp = fp->f_data;
3150 	if (vp->v_type != VDIR) {
3151 		error = EINVAL;
3152 		goto bad;
3153 	}
3154 
3155 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3156 
3157 	if (fp->f_offset < 0) {
3158 		VOP_UNLOCK(vp);
3159 		error = EINVAL;
3160 		goto bad;
3161 	}
3162 
3163 	aiov.iov_base = SCARG(uap, buf);
3164 	aiov.iov_len = buflen;
3165 	auio.uio_iov = &aiov;
3166 	auio.uio_iovcnt = 1;
3167 	auio.uio_rw = UIO_READ;
3168 	auio.uio_segflg = UIO_USERSPACE;
3169 	auio.uio_procp = p;
3170 	auio.uio_resid = buflen;
3171 	auio.uio_offset = fp->f_offset;
3172 	error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag);
3173 	mtx_enter(&fp->f_mtx);
3174 	fp->f_offset = auio.uio_offset;
3175 	mtx_leave(&fp->f_mtx);
3176 	VOP_UNLOCK(vp);
3177 	if (error)
3178 		goto bad;
3179 	*retval = buflen - auio.uio_resid;
3180 bad:
3181 	FRELE(fp, p);
3182 	return (error);
3183 }
3184 
3185 /*
3186  * Set the mode mask for creation of filesystem nodes.
3187  */
3188 int
3189 sys_umask(struct proc *p, void *v, register_t *retval)
3190 {
3191 	struct sys_umask_args /* {
3192 		syscallarg(mode_t) newmask;
3193 	} */ *uap = v;
3194 	struct filedesc *fdp = p->p_fd;
3195 
3196 	fdplock(fdp);
3197 	*retval = fdp->fd_cmask;
3198 	fdp->fd_cmask = SCARG(uap, newmask) & ACCESSPERMS;
3199 	fdpunlock(fdp);
3200 	return (0);
3201 }
3202 
3203 /*
3204  * Void all references to file by ripping underlying filesystem
3205  * away from vnode.
3206  */
3207 int
3208 sys_revoke(struct proc *p, void *v, register_t *retval)
3209 {
3210 	struct sys_revoke_args /* {
3211 		syscallarg(const char *) path;
3212 	} */ *uap = v;
3213 	struct vnode *vp;
3214 	struct vattr vattr;
3215 	int error;
3216 	struct nameidata nd;
3217 
3218 	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
3219 	nd.ni_pledge = PLEDGE_RPATH | PLEDGE_TTY;
3220 	nd.ni_unveil = UNVEIL_READ;
3221 	if ((error = namei(&nd)) != 0)
3222 		return (error);
3223 	vp = nd.ni_vp;
3224 	if (vp->v_type != VCHR || (u_int)major(vp->v_rdev) >= nchrdev ||
3225 	    cdevsw[major(vp->v_rdev)].d_type != D_TTY) {
3226 		error = ENOTTY;
3227 		goto out;
3228 	}
3229 	if ((error = VOP_GETATTR(vp, &vattr, p->p_ucred, p)) != 0)
3230 		goto out;
3231 	if (p->p_ucred->cr_uid != vattr.va_uid &&
3232 	    (error = suser(p)))
3233 		goto out;
3234 	if (vp->v_usecount > 1 || (vp->v_flag & (VALIASED)))
3235 		VOP_REVOKE(vp, REVOKEALL);
3236 out:
3237 	vrele(vp);
3238 	return (error);
3239 }
3240 
3241 /*
3242  * Convert a user file descriptor to a kernel file entry.
3243  *
3244  * On return *fpp is FREF:ed.
3245  */
3246 int
3247 getvnode(struct proc *p, int fd, struct file **fpp)
3248 {
3249 	struct file *fp;
3250 	struct vnode *vp;
3251 
3252 	if ((fp = fd_getfile(p->p_fd, fd)) == NULL)
3253 		return (EBADF);
3254 
3255 	if (fp->f_type != DTYPE_VNODE) {
3256 		FRELE(fp, p);
3257 		return (EINVAL);
3258 	}
3259 
3260 	vp = fp->f_data;
3261 	if (vp->v_type == VBAD) {
3262 		FRELE(fp, p);
3263 		return (EBADF);
3264 	}
3265 
3266 	*fpp = fp;
3267 
3268 	return (0);
3269 }
3270 
3271 /*
3272  * Positional read system call.
3273  */
3274 int
3275 sys_pread(struct proc *p, void *v, register_t *retval)
3276 {
3277 	struct sys_pread_args /* {
3278 		syscallarg(int) fd;
3279 		syscallarg(void *) buf;
3280 		syscallarg(size_t) nbyte;
3281 		syscallarg(off_t) offset;
3282 	} */ *uap = v;
3283 	struct iovec iov;
3284 	struct uio auio;
3285 
3286 	iov.iov_base = SCARG(uap, buf);
3287 	iov.iov_len = SCARG(uap, nbyte);
3288 	if (iov.iov_len > SSIZE_MAX)
3289 		return (EINVAL);
3290 
3291 	auio.uio_iov = &iov;
3292 	auio.uio_iovcnt = 1;
3293 	auio.uio_resid = iov.iov_len;
3294 	auio.uio_offset = SCARG(uap, offset);
3295 
3296 	return (dofilereadv(p, SCARG(uap, fd), &auio, FO_POSITION, retval));
3297 }
3298 
3299 /*
3300  * Positional scatter read system call.
3301  */
3302 int
3303 sys_preadv(struct proc *p, void *v, register_t *retval)
3304 {
3305 	struct sys_preadv_args /* {
3306 		syscallarg(int) fd;
3307 		syscallarg(const struct iovec *) iovp;
3308 		syscallarg(int) iovcnt;
3309 		syscallarg(off_t) offset;
3310 	} */ *uap = v;
3311 	struct iovec aiov[UIO_SMALLIOV], *iov = NULL;
3312 	int error, iovcnt = SCARG(uap, iovcnt);
3313 	struct uio auio;
3314 	size_t resid;
3315 
3316 	error = iovec_copyin(SCARG(uap, iovp), &iov, aiov, iovcnt, &resid);
3317 	if (error)
3318 		goto done;
3319 
3320 	auio.uio_iov = iov;
3321 	auio.uio_iovcnt = iovcnt;
3322 	auio.uio_resid = resid;
3323 	auio.uio_offset = SCARG(uap, offset);
3324 
3325 	error = dofilereadv(p, SCARG(uap, fd), &auio, FO_POSITION, retval);
3326  done:
3327 	iovec_free(iov, iovcnt);
3328 	return (error);
3329 }
3330 
3331 /*
3332  * Positional write system call.
3333  */
3334 int
3335 sys_pwrite(struct proc *p, void *v, register_t *retval)
3336 {
3337 	struct sys_pwrite_args /* {
3338 		syscallarg(int) fd;
3339 		syscallarg(const void *) buf;
3340 		syscallarg(size_t) nbyte;
3341 		syscallarg(off_t) offset;
3342 	} */ *uap = v;
3343 	struct iovec iov;
3344 	struct uio auio;
3345 
3346 	iov.iov_base = (void *)SCARG(uap, buf);
3347 	iov.iov_len = SCARG(uap, nbyte);
3348 	if (iov.iov_len > SSIZE_MAX)
3349 		return (EINVAL);
3350 
3351 	auio.uio_iov = &iov;
3352 	auio.uio_iovcnt = 1;
3353 	auio.uio_resid = iov.iov_len;
3354 	auio.uio_offset = SCARG(uap, offset);
3355 
3356 	return (dofilewritev(p, SCARG(uap, fd), &auio, FO_POSITION, retval));
3357 }
3358 
3359 /*
3360  * Positional gather write system call.
3361  */
3362 int
3363 sys_pwritev(struct proc *p, void *v, register_t *retval)
3364 {
3365 	struct sys_pwritev_args /* {
3366 		syscallarg(int) fd;
3367 		syscallarg(const struct iovec *) iovp;
3368 		syscallarg(int) iovcnt;
3369 		syscallarg(off_t) offset;
3370 	} */ *uap = v;
3371 	struct iovec aiov[UIO_SMALLIOV], *iov = NULL;
3372 	int error, iovcnt = SCARG(uap, iovcnt);
3373 	struct uio auio;
3374 	size_t resid;
3375 
3376 	error = iovec_copyin(SCARG(uap, iovp), &iov, aiov, iovcnt, &resid);
3377 	if (error)
3378 		goto done;
3379 
3380 	auio.uio_iov = iov;
3381 	auio.uio_iovcnt = iovcnt;
3382 	auio.uio_resid = resid;
3383 	auio.uio_offset = SCARG(uap, offset);
3384 
3385 	error = dofilewritev(p, SCARG(uap, fd), &auio, FO_POSITION, retval);
3386  done:
3387 	iovec_free(iov, iovcnt);
3388 	return (error);
3389 }
3390