xref: /netbsd-src/sys/kern/vfs_syscalls.c (revision 7fa608457b817eca6e0977b37f758ae064f3c99c)
1 /*	$NetBSD: vfs_syscalls.c,v 1.331 2007/10/24 15:28:55 pooka Exp $	*/
2 
3 /*
4  * Copyright (c) 1989, 1993
5  *	The Regents of the University of California.  All rights reserved.
6  * (c) UNIX System Laboratories, Inc.
7  * All or some portions of this file are derived from material licensed
8  * to the University of California by American Telephone and Telegraph
9  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
10  * the permission of UNIX System Laboratories, Inc.
11  *
12  * Redistribution and use in source and binary forms, with or without
13  * modification, are permitted provided that the following conditions
14  * are met:
15  * 1. Redistributions of source code must retain the above copyright
16  *    notice, this list of conditions and the following disclaimer.
17  * 2. Redistributions in binary form must reproduce the above copyright
18  *    notice, this list of conditions and the following disclaimer in the
19  *    documentation and/or other materials provided with the distribution.
20  * 3. Neither the name of the University nor the names of its contributors
21  *    may be used to endorse or promote products derived from this software
22  *    without specific prior written permission.
23  *
24  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
25  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
28  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
29  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
30  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
31  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
32  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
33  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
34  * SUCH DAMAGE.
35  *
36  *	@(#)vfs_syscalls.c	8.42 (Berkeley) 7/31/95
37  */
38 
39 #include <sys/cdefs.h>
40 __KERNEL_RCSID(0, "$NetBSD: vfs_syscalls.c,v 1.331 2007/10/24 15:28:55 pooka Exp $");
41 
42 #include "opt_compat_netbsd.h"
43 #include "opt_compat_43.h"
44 #include "opt_fileassoc.h"
45 #include "fss.h"
46 #include "veriexec.h"
47 
48 #include <sys/param.h>
49 #include <sys/systm.h>
50 #include <sys/namei.h>
51 #include <sys/filedesc.h>
52 #include <sys/kernel.h>
53 #include <sys/file.h>
54 #include <sys/stat.h>
55 #include <sys/vnode.h>
56 #include <sys/mount.h>
57 #include <sys/proc.h>
58 #include <sys/uio.h>
59 #include <sys/malloc.h>
60 #include <sys/kmem.h>
61 #include <sys/dirent.h>
62 #include <sys/sysctl.h>
63 #include <sys/syscallargs.h>
64 #include <sys/vfs_syscalls.h>
65 #include <sys/ktrace.h>
66 #ifdef FILEASSOC
67 #include <sys/fileassoc.h>
68 #endif /* FILEASSOC */
69 #include <sys/verified_exec.h>
70 #include <sys/kauth.h>
71 
72 #include <miscfs/genfs/genfs.h>
73 #include <miscfs/syncfs/syncfs.h>
74 
75 #ifdef COMPAT_30
76 #include "opt_nfsserver.h"
77 #include <nfs/rpcv2.h>
78 #endif
79 #include <nfs/nfsproto.h>
80 #ifdef COMPAT_30
81 #include <nfs/nfs.h>
82 #include <nfs/nfs_var.h>
83 #endif
84 
85 #if NFSS > 0
86 #include <dev/fssvar.h>
87 #endif
88 
89 MALLOC_DEFINE(M_MOUNT, "mount", "vfs mount struct");
90 
91 static int change_dir(struct nameidata *, struct lwp *);
92 static int change_flags(struct vnode *, u_long, struct lwp *);
93 static int change_mode(struct vnode *, int, struct lwp *l);
94 static int change_owner(struct vnode *, uid_t, gid_t, struct lwp *, int);
95 static int rename_files(const char *, const char *, struct lwp *, int);
96 
97 void checkdirs(struct vnode *);
98 
99 int dovfsusermount = 0;
100 
101 /*
102  * Virtual File System System Calls
103  */
104 
105 /*
106  * Mount a file system.
107  */
108 
109 #if defined(COMPAT_09) || defined(COMPAT_43)
110 /*
111  * This table is used to maintain compatibility with 4.3BSD
112  * and NetBSD 0.9 mount syscalls.  Note, the order is important!
113  *
114  * Do not modify this table. It should only contain filesystems
115  * supported by NetBSD 0.9 and 4.3BSD.
116  */
117 const char * const mountcompatnames[] = {
118 	NULL,		/* 0 = MOUNT_NONE */
119 	MOUNT_FFS,	/* 1 = MOUNT_UFS */
120 	MOUNT_NFS,	/* 2 */
121 	MOUNT_MFS,	/* 3 */
122 	MOUNT_MSDOS,	/* 4 */
123 	MOUNT_CD9660,	/* 5 = MOUNT_ISOFS */
124 	MOUNT_FDESC,	/* 6 */
125 	MOUNT_KERNFS,	/* 7 */
126 	NULL,		/* 8 = MOUNT_DEVFS */
127 	MOUNT_AFS,	/* 9 */
128 };
129 const int nmountcompatnames = sizeof(mountcompatnames) /
130     sizeof(mountcompatnames[0]);
131 #endif /* COMPAT_09 || COMPAT_43 */
132 
133 static int
134 mount_update(struct lwp *l, struct vnode *vp, const char *path, int flags,
135     void *data, size_t *data_len)
136 {
137 	struct mount *mp;
138 	int error = 0, saved_flags;
139 
140 	mp = vp->v_mount;
141 	saved_flags = mp->mnt_flag;
142 
143 	/* We can operate only on VV_ROOT nodes. */
144 	if ((vp->v_vflag & VV_ROOT) == 0) {
145 		error = EINVAL;
146 		goto out;
147 	}
148 
149 	/*
150 	 * We only allow the filesystem to be reloaded if it
151 	 * is currently mounted read-only.
152 	 */
153 	if (flags & MNT_RELOAD && !(mp->mnt_flag & MNT_RDONLY)) {
154 		error = EOPNOTSUPP;	/* Needs translation */
155 		goto out;
156 	}
157 
158 	error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT,
159 	    KAUTH_REQ_SYSTEM_MOUNT_UPDATE, mp, KAUTH_ARG(flags), data);
160 	if (error)
161 		goto out;
162 
163 	if (vfs_busy(mp, LK_NOWAIT, 0)) {
164 		error = EPERM;
165 		goto out;
166 	}
167 
168 	mp->mnt_flag &= ~MNT_OP_FLAGS;
169 	mp->mnt_flag |= flags & (MNT_RELOAD | MNT_FORCE | MNT_UPDATE);
170 
171 	/*
172 	 * Set the mount level flags.
173 	 */
174 	if (flags & MNT_RDONLY)
175 		mp->mnt_flag |= MNT_RDONLY;
176 	else if (mp->mnt_flag & MNT_RDONLY)
177 		mp->mnt_iflag |= IMNT_WANTRDWR;
178 	mp->mnt_flag &=
179 	  ~(MNT_NOSUID | MNT_NOEXEC | MNT_NODEV |
180 	    MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP |
181 	    MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP);
182 	mp->mnt_flag |= flags &
183 	   (MNT_NOSUID | MNT_NOEXEC | MNT_NODEV |
184 	    MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP |
185 	    MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP |
186 	    MNT_IGNORE);
187 
188 	error = VFS_MOUNT(mp, path, data, data_len, l);
189 
190 #if defined(COMPAT_30) && defined(NFSSERVER)
191 	if (error && data != NULL) {
192 		int error2;
193 
194 		/* Update failed; let's try and see if it was an
195 		 * export request. */
196 		error2 = nfs_update_exports_30(mp, path, data, l);
197 
198 		/* Only update error code if the export request was
199 		 * understood but some problem occurred while
200 		 * processing it. */
201 		if (error2 != EJUSTRETURN)
202 			error = error2;
203 	}
204 #endif
205 	if (mp->mnt_iflag & IMNT_WANTRDWR)
206 		mp->mnt_flag &= ~MNT_RDONLY;
207 	if (error)
208 		mp->mnt_flag = saved_flags;
209 	mp->mnt_flag &= ~MNT_OP_FLAGS;
210 	mp->mnt_iflag &= ~IMNT_WANTRDWR;
211 	if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) {
212 		if (mp->mnt_syncer == NULL)
213 			error = vfs_allocate_syncvnode(mp);
214 	} else {
215 		if (mp->mnt_syncer != NULL)
216 			vfs_deallocate_syncvnode(mp);
217 	}
218 	vfs_unbusy(mp);
219 
220  out:
221 	return (error);
222 }
223 
224 static int
225 mount_get_vfsops(const char *fstype, struct vfsops **vfsops)
226 {
227 	char fstypename[sizeof(((struct statvfs *)NULL)->f_fstypename)];
228 	int error;
229 
230 	/* Copy file-system type from userspace.  */
231 	error = copyinstr(fstype, fstypename, sizeof(fstypename), NULL);
232 	if (error) {
233 #if defined(COMPAT_09) || defined(COMPAT_43)
234 		/*
235 		 * Historically, filesystem types were identified by numbers.
236 		 * If we get an integer for the filesystem type instead of a
237 		 * string, we check to see if it matches one of the historic
238 		 * filesystem types.
239 		 */
240 		u_long fsindex = (u_long)fstype;
241 		if (fsindex >= nmountcompatnames ||
242 		    mountcompatnames[fsindex] == NULL)
243 			return ENODEV;
244 		strlcpy(fstypename, mountcompatnames[fsindex],
245 		    sizeof(fstypename));
246 #else
247 		return error;
248 #endif
249 	}
250 
251 #ifdef	COMPAT_10
252 	/* Accept `ufs' as an alias for `ffs'. */
253 	if (strcmp(fstypename, "ufs") == 0)
254 		fstypename[0] = 'f';
255 #endif
256 
257 	if ((*vfsops = vfs_getopsbyname(fstypename)) == NULL)
258 		return ENODEV;
259 	return 0;
260 }
261 
262 static int
263 mount_domount(struct lwp *l, struct vnode **vpp, struct vfsops *vfsops,
264     const char *path, int flags, void *data, size_t *data_len)
265 {
266 	struct mount *mp = NULL;
267 	struct vnode *vp = *vpp;
268 	struct vattr va;
269 	int error;
270 
271 	error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT,
272 	    KAUTH_REQ_SYSTEM_MOUNT_NEW, vp, KAUTH_ARG(flags), data);
273 	if (error)
274 		return error;
275 
276 	/* Can't make a non-dir a mount-point (from here anyway). */
277 	if (vp->v_type != VDIR)
278 		return ENOTDIR;
279 
280 	/*
281 	 * If the user is not root, ensure that they own the directory
282 	 * onto which we are attempting to mount.
283 	 */
284 	if ((error = VOP_GETATTR(vp, &va, l->l_cred, l)) != 0 ||
285 	    (va.va_uid != kauth_cred_geteuid(l->l_cred) &&
286 	    (error = kauth_authorize_generic(l->l_cred,
287 	    KAUTH_GENERIC_ISSUSER, NULL)) != 0)) {
288 		return error;
289 	}
290 
291 	if (flags & MNT_EXPORTED)
292 		return EINVAL;
293 
294 	if ((error = vinvalbuf(vp, V_SAVE, l->l_cred, l, 0, 0)) != 0)
295 		return error;
296 
297 	/*
298 	 * Check if a file-system is not already mounted on this vnode.
299 	 */
300 	if (vp->v_mountedhere != NULL)
301 		return EBUSY;
302 
303 	mp = malloc(sizeof(*mp), M_MOUNT, M_WAITOK|M_ZERO);
304 
305 	mp->mnt_op = vfsops;
306 
307 	TAILQ_INIT(&mp->mnt_vnodelist);
308 	lockinit(&mp->mnt_lock, PVFS, "vfslock", 0, 0);
309 	simple_lock_init(&mp->mnt_slock);
310 	(void)vfs_busy(mp, LK_NOWAIT, 0);
311 
312 	mp->mnt_vnodecovered = vp;
313 	mp->mnt_stat.f_owner = kauth_cred_geteuid(l->l_cred);
314 	mp->mnt_unmounter = NULL;
315 	mount_initspecific(mp);
316 
317 	/*
318 	 * The underlying file system may refuse the mount for
319 	 * various reasons.  Allow the user to force it to happen.
320 	 *
321 	 * Set the mount level flags.
322 	 */
323 	mp->mnt_flag = flags &
324 	   (MNT_FORCE | MNT_NOSUID | MNT_NOEXEC | MNT_NODEV |
325 	    MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP |
326 	    MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP |
327 	    MNT_IGNORE | MNT_RDONLY);
328 
329 	error = VFS_MOUNT(mp, path, data, data_len, l);
330 	mp->mnt_flag &= ~MNT_OP_FLAGS;
331 
332 	/*
333 	 * Put the new filesystem on the mount list after root.
334 	 */
335 	cache_purge(vp);
336 	if (error != 0) {
337 		vp->v_mountedhere = NULL;
338 		mp->mnt_op->vfs_refcount--;
339 		vfs_unbusy(mp);
340 		vfs_destroy(mp);
341 		return error;
342 	}
343 
344 	mp->mnt_iflag &= ~IMNT_WANTRDWR;
345 	vp->v_mountedhere = mp;
346 	mutex_enter(&mountlist_lock);
347 	CIRCLEQ_INSERT_TAIL(&mountlist, mp, mnt_list);
348 	mutex_exit(&mountlist_lock);
349 	VOP_UNLOCK(vp, 0);
350 	checkdirs(vp);
351 	if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0)
352 		error = vfs_allocate_syncvnode(mp);
353 	vfs_unbusy(mp);
354 	(void) VFS_STATVFS(mp, &mp->mnt_stat, l);
355 	error = VFS_START(mp, 0, l);
356 	if (error)
357 		vrele(vp);
358 	*vpp = NULL;
359 	return error;
360 }
361 
362 static int
363 mount_getargs(struct lwp *l, struct vnode *vp, const char *path, int flags,
364     void *data, size_t *data_len)
365 {
366 	struct mount *mp;
367 	int error;
368 
369 	/* If MNT_GETARGS is specified, it should be the only flag. */
370 	if (flags & ~MNT_GETARGS)
371 		return EINVAL;
372 
373 	mp = vp->v_mount;
374 
375 	/* XXX: probably some notion of "can see" here if we want isolation. */
376 	error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT,
377 	    KAUTH_REQ_SYSTEM_MOUNT_GET, mp, data, NULL);
378 	if (error)
379 		return error;
380 
381 	if ((vp->v_vflag & VV_ROOT) == 0)
382 		return EINVAL;
383 
384 	if (vfs_busy(mp, LK_NOWAIT, 0))
385 		return EPERM;
386 
387 	mp->mnt_flag &= ~MNT_OP_FLAGS;
388 	mp->mnt_flag |= MNT_GETARGS;
389 	error = VFS_MOUNT(mp, path, data, data_len, l);
390 	mp->mnt_flag &= ~MNT_OP_FLAGS;
391 
392 	vfs_unbusy(mp);
393 	return (error);
394 }
395 
396 #ifdef COMPAT_40
397 /* ARGSUSED */
398 int
399 compat_40_sys_mount(struct lwp *l, void *v, register_t *retval)
400 {
401 	struct compat_40_sys_mount_args /* {
402 		syscallarg(const char *) type;
403 		syscallarg(const char *) path;
404 		syscallarg(int) flags;
405 		syscallarg(void *) data;
406 	} */ *uap = v;
407 	register_t dummy;
408 
409 	return do_sys_mount(l, NULL, SCARG(uap, type), SCARG(uap, path),
410 	    SCARG(uap, flags), SCARG(uap, data), UIO_USERSPACE, 0, &dummy);
411 }
412 #endif
413 
414 int
415 sys___mount50(struct lwp *l, void *v, register_t *retval)
416 {
417 	struct sys___mount50_args /* {
418 		syscallarg(const char *) type;
419 		syscallarg(const char *) path;
420 		syscallarg(int) flags;
421 		syscallarg(void *) data;
422 		syscallarg(size_t) data_len;
423 	} */ *uap = v;
424 
425 	return do_sys_mount(l, NULL, SCARG(uap, type), SCARG(uap, path),
426 	    SCARG(uap, flags), SCARG(uap, data), UIO_USERSPACE,
427 	    SCARG(uap, data_len), retval);
428 }
429 
430 int
431 do_sys_mount(struct lwp *l, struct vfsops *vfsops, const char *type,
432     const char *path, int flags, void *data, enum uio_seg data_seg,
433     size_t data_len, register_t *retval)
434 {
435 	struct vnode *vp;
436 	struct nameidata nd;
437 	void *data_buf = data;
438 	int error;
439 
440 	/*
441 	 * Get vnode to be covered
442 	 */
443 	NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, path, l);
444 	if ((error = namei(&nd)) != 0)
445 		return (error);
446 	vp = nd.ni_vp;
447 
448 	/*
449 	 * A lookup in VFS_MOUNT might result in an attempt to
450 	 * lock this vnode again, so make the lock recursive.
451 	 */
452 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY | LK_SETRECURSE);
453 
454 	if (vfsops == NULL) {
455 		if (flags & (MNT_GETARGS | MNT_UPDATE))
456 			vfsops = vp->v_mount->mnt_op;
457 		else {
458 			/* 'type' is userspace */
459 			error = mount_get_vfsops(type, &vfsops);
460 			if (error != 0)
461 				goto done;
462 		}
463 	}
464 
465 	if (data != NULL && data_seg == UIO_USERSPACE) {
466 		if (data_len == 0) {
467 			/* No length supplied, use default for filesystem */
468 			data_len = vfsops->vfs_min_mount_data;
469 			if (data_len > VFS_MAX_MOUNT_DATA) {
470 				/* maybe a force loaded old LKM */
471 				error = EINVAL;
472 				goto done;
473 			}
474 #ifdef COMPAT_30
475 			/* Hopefully a longer buffer won't make copyin() fail */
476 			if (flags & MNT_UPDATE
477 			    && data_len < sizeof (struct mnt_export_args30))
478 				data_len = sizeof (struct mnt_export_args30);
479 #endif
480 		}
481 		data_buf = malloc(data_len, M_TEMP, M_WAITOK);
482 
483 		/* NFS needs the buffer even for mnt_getargs .... */
484 		error = copyin(data, data_buf, data_len);
485 		if (error != 0)
486 			goto done;
487 	}
488 
489 	if (flags & MNT_GETARGS) {
490 		if (data_len == 0) {
491 			error = EINVAL;
492 			goto done;
493 		}
494 		error = mount_getargs(l, vp, path, flags, data_buf, &data_len);
495 		if (error != 0)
496 			goto done;
497 		if (data_seg == UIO_USERSPACE)
498 			error = copyout(data_buf, data, data_len);
499 		*retval = data_len;
500 	} else if (flags & MNT_UPDATE) {
501 		error = mount_update(l, vp, path, flags, data_buf, &data_len);
502 	} else {
503 		/* Locking is handled internally in mount_domount(). */
504 		error = mount_domount(l, &vp, vfsops, path, flags, data_buf,
505 		    &data_len);
506 	}
507 
508     done:
509 	if (vp)
510 		vput(vp);
511 	if (data_buf != data)
512 		free(data_buf, M_TEMP);
513 	return (error);
514 }
515 
516 /*
517  * Scan all active processes to see if any of them have a current
518  * or root directory onto which the new filesystem has just been
519  * mounted. If so, replace them with the new mount point.
520  */
521 void
522 checkdirs(struct vnode *olddp)
523 {
524 	struct cwdinfo *cwdi;
525 	struct vnode *newdp;
526 	struct proc *p;
527 
528 	if (olddp->v_usecount == 1)
529 		return;
530 	if (VFS_ROOT(olddp->v_mountedhere, &newdp))
531 		panic("mount: lost mount");
532 	mutex_enter(&proclist_lock);
533 	PROCLIST_FOREACH(p, &allproc) {
534 		cwdi = p->p_cwdi;
535 		if (!cwdi)
536 			continue;
537 		rw_enter(&cwdi->cwdi_lock, RW_WRITER);
538 		if (cwdi->cwdi_cdir == olddp) {
539 			vrele(cwdi->cwdi_cdir);
540 			VREF(newdp);
541 			cwdi->cwdi_cdir = newdp;
542 		}
543 		if (cwdi->cwdi_rdir == olddp) {
544 			vrele(cwdi->cwdi_rdir);
545 			VREF(newdp);
546 			cwdi->cwdi_rdir = newdp;
547 		}
548 		rw_exit(&cwdi->cwdi_lock);
549 	}
550 	mutex_exit(&proclist_lock);
551 	if (rootvnode == olddp) {
552 		vrele(rootvnode);
553 		VREF(newdp);
554 		rootvnode = newdp;
555 	}
556 	vput(newdp);
557 }
558 
559 /*
560  * Unmount a file system.
561  *
562  * Note: unmount takes a path to the vnode mounted on as argument,
563  * not special file (as before).
564  */
565 /* ARGSUSED */
566 int
567 sys_unmount(struct lwp *l, void *v, register_t *retval)
568 {
569 	struct sys_unmount_args /* {
570 		syscallarg(const char *) path;
571 		syscallarg(int) flags;
572 	} */ *uap = v;
573 	struct vnode *vp;
574 	struct mount *mp;
575 	int error;
576 	struct nameidata nd;
577 
578 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE,
579 	    SCARG(uap, path), l);
580 	if ((error = namei(&nd)) != 0)
581 		return (error);
582 	vp = nd.ni_vp;
583 	mp = vp->v_mount;
584 
585 	error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT,
586 	    KAUTH_REQ_SYSTEM_MOUNT_UNMOUNT, mp, NULL, NULL);
587 	if (error) {
588 		vput(vp);
589 		return (error);
590 	}
591 
592 	/*
593 	 * Don't allow unmounting the root file system.
594 	 */
595 	if (mp->mnt_flag & MNT_ROOTFS) {
596 		vput(vp);
597 		return (EINVAL);
598 	}
599 
600 	/*
601 	 * Must be the root of the filesystem
602 	 */
603 	if ((vp->v_vflag & VV_ROOT) == 0) {
604 		vput(vp);
605 		return (EINVAL);
606 	}
607 	vput(vp);
608 
609 	/*
610 	 * XXX Freeze syncer.  Must do this before locking the
611 	 * mount point.  See dounmount() for details.
612 	 */
613 	mutex_enter(&syncer_mutex);
614 
615 	if (vfs_busy(mp, 0, 0)) {
616 		mutex_exit(&syncer_mutex);
617 		return (EBUSY);
618 	}
619 
620 	return (dounmount(mp, SCARG(uap, flags), l));
621 }
622 
623 /*
624  * Do the actual file system unmount. File system is assumed to have been
625  * marked busy by the caller.
626  */
627 int
628 dounmount(struct mount *mp, int flags, struct lwp *l)
629 {
630 	struct vnode *coveredvp;
631 	int error;
632 	int async;
633 	int used_syncer;
634 
635 #if NVERIEXEC > 0
636 	error = veriexec_unmountchk(mp);
637 	if (error)
638 		return (error);
639 #endif /* NVERIEXEC > 0 */
640 
641 	mutex_enter(&mountlist_lock);
642 	vfs_unbusy(mp);
643 	used_syncer = (mp->mnt_syncer != NULL);
644 
645 	/*
646 	 * XXX Syncer must be frozen when we get here.  This should really
647 	 * be done on a per-mountpoint basis, but especially the softdep
648 	 * code possibly called from the syncer doesn't exactly work on a
649 	 * per-mountpoint basis, so the softdep code would become a maze
650 	 * of vfs_busy() calls.
651 	 *
652 	 * The caller of dounmount() must acquire syncer_mutex because
653 	 * the syncer itself acquires locks in syncer_mutex -> vfs_busy
654 	 * order, and we must preserve that order to avoid deadlock.
655 	 *
656 	 * So, if the file system did not use the syncer, now is
657 	 * the time to release the syncer_mutex.
658 	 */
659 	if (used_syncer == 0)
660 		mutex_exit(&syncer_mutex);
661 
662 	mp->mnt_iflag |= IMNT_UNMOUNT;
663 	mp->mnt_unmounter = l;
664 	mutex_exit(&mountlist_lock);	/* XXX */
665 	lockmgr(&mp->mnt_lock, LK_DRAIN, NULL);
666 
667 	async = mp->mnt_flag & MNT_ASYNC;
668 	mp->mnt_flag &= ~MNT_ASYNC;
669 	cache_purgevfs(mp);	/* remove cache entries for this file sys */
670 	if (mp->mnt_syncer != NULL)
671 		vfs_deallocate_syncvnode(mp);
672 	error = 0;
673 	if ((mp->mnt_flag & MNT_RDONLY) == 0) {
674 #if NFSS > 0
675 		error = fss_umount_hook(mp, (flags & MNT_FORCE));
676 #endif
677 		if (error == 0)
678 			error = VFS_SYNC(mp, MNT_WAIT, l->l_cred, l);
679 	}
680 	if (error == 0 || (flags & MNT_FORCE))
681 		error = VFS_UNMOUNT(mp, flags, l);
682 	if (error) {
683 		if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0)
684 			(void) vfs_allocate_syncvnode(mp);
685 		mutex_enter(&mountlist_lock);
686 		mp->mnt_iflag &= ~IMNT_UNMOUNT;
687 		mp->mnt_unmounter = NULL;
688 		mp->mnt_flag |= async;
689 		mutex_exit(&mountlist_lock);	/* XXX */
690 		lockmgr(&mp->mnt_lock, LK_RELEASE | LK_REENABLE,
691 		    NULL);
692 		if (used_syncer)
693 			mutex_exit(&syncer_mutex);
694 		simple_lock(&mp->mnt_slock);
695 		while (mp->mnt_wcnt > 0) {
696 			wakeup(mp);
697 			ltsleep(&mp->mnt_wcnt, PVFS, "mntwcnt1",
698 				0, &mp->mnt_slock);
699 		}
700 		simple_unlock(&mp->mnt_slock);
701 		return (error);
702 	}
703 	mutex_enter(&mountlist_lock);
704 	CIRCLEQ_REMOVE(&mountlist, mp, mnt_list);
705 	if ((coveredvp = mp->mnt_vnodecovered) != NULLVP)
706 		coveredvp->v_mountedhere = NULL;
707 	if (TAILQ_FIRST(&mp->mnt_vnodelist) != NULL)
708 		panic("unmount: dangling vnode");
709 	mp->mnt_iflag |= IMNT_GONE;
710 	mutex_exit(&mountlist_lock);
711 	lockmgr(&mp->mnt_lock, LK_RELEASE, NULL);
712 	if (coveredvp != NULLVP)
713 		vrele(coveredvp);
714 	if (used_syncer)
715 		mutex_exit(&syncer_mutex);
716 	simple_lock(&mp->mnt_slock);
717 	while (mp->mnt_wcnt > 0) {
718 		wakeup(mp);
719 		ltsleep(&mp->mnt_wcnt, PVFS, "mntwcnt2", 0, &mp->mnt_slock);
720 	}
721 	simple_unlock(&mp->mnt_slock);
722 	vfs_hooks_unmount(mp);
723 	vfs_delref(mp->mnt_op);
724 	vfs_destroy(mp);
725 	return (0);
726 }
727 
728 /*
729  * Sync each mounted filesystem.
730  */
731 #ifdef DEBUG
732 int syncprt = 0;
733 struct ctldebug debug0 = { "syncprt", &syncprt };
734 #endif
735 
736 /* ARGSUSED */
737 int
738 sys_sync(struct lwp *l, void *v, register_t *retval)
739 {
740 	struct mount *mp, *nmp;
741 	int asyncflag;
742 
743 	if (l == NULL)
744 		l = &lwp0;
745 
746 	mutex_enter(&mountlist_lock);
747 	for (mp = mountlist.cqh_last; mp != (void *)&mountlist; mp = nmp) {
748 		if (vfs_busy(mp, LK_NOWAIT, &mountlist_lock)) {
749 			nmp = mp->mnt_list.cqe_prev;
750 			continue;
751 		}
752 		if ((mp->mnt_flag & MNT_RDONLY) == 0) {
753 			asyncflag = mp->mnt_flag & MNT_ASYNC;
754 			mp->mnt_flag &= ~MNT_ASYNC;
755 			VFS_SYNC(mp, MNT_NOWAIT, l->l_cred, l);
756 			if (asyncflag)
757 				 mp->mnt_flag |= MNT_ASYNC;
758 		}
759 		mutex_enter(&mountlist_lock);
760 		nmp = mp->mnt_list.cqe_prev;
761 		vfs_unbusy(mp);
762 
763 	}
764 	mutex_exit(&mountlist_lock);
765 #ifdef DEBUG
766 	if (syncprt)
767 		vfs_bufstats();
768 #endif /* DEBUG */
769 	return (0);
770 }
771 
772 /*
773  * Change filesystem quotas.
774  */
775 /* ARGSUSED */
776 int
777 sys_quotactl(struct lwp *l, void *v, register_t *retval)
778 {
779 	struct sys_quotactl_args /* {
780 		syscallarg(const char *) path;
781 		syscallarg(int) cmd;
782 		syscallarg(int) uid;
783 		syscallarg(void *) arg;
784 	} */ *uap = v;
785 	struct mount *mp;
786 	int error;
787 	struct nameidata nd;
788 
789 	NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE,
790 	    SCARG(uap, path), l);
791 	if ((error = namei(&nd)) != 0)
792 		return (error);
793 	mp = nd.ni_vp->v_mount;
794 	error = VFS_QUOTACTL(mp, SCARG(uap, cmd), SCARG(uap, uid),
795 	    SCARG(uap, arg), l);
796 	vrele(nd.ni_vp);
797 	return (error);
798 }
799 
800 int
801 dostatvfs(struct mount *mp, struct statvfs *sp, struct lwp *l, int flags,
802     int root)
803 {
804 	struct cwdinfo *cwdi = l->l_proc->p_cwdi;
805 	int error = 0;
806 
807 	/*
808 	 * If MNT_NOWAIT or MNT_LAZY is specified, do not
809 	 * refresh the fsstat cache. MNT_WAIT or MNT_LAZY
810 	 * overrides MNT_NOWAIT.
811 	 */
812 	if (flags == MNT_NOWAIT	|| flags == MNT_LAZY ||
813 	    (flags != MNT_WAIT && flags != 0)) {
814 		memcpy(sp, &mp->mnt_stat, sizeof(*sp));
815 		goto done;
816 	}
817 
818 	/* Get the filesystem stats now */
819 	memset(sp, 0, sizeof(*sp));
820 	if ((error = VFS_STATVFS(mp, sp, l)) != 0) {
821 		return error;
822 	}
823 
824 	if (cwdi->cwdi_rdir == NULL)
825 		(void)memcpy(&mp->mnt_stat, sp, sizeof(mp->mnt_stat));
826 done:
827 	if (cwdi->cwdi_rdir != NULL) {
828 		size_t len;
829 		char *bp;
830 		char *path = PNBUF_GET();
831 
832 		bp = path + MAXPATHLEN;
833 		*--bp = '\0';
834 		rw_enter(&cwdi->cwdi_lock, RW_READER);
835 		error = getcwd_common(cwdi->cwdi_rdir, rootvnode, &bp, path,
836 		    MAXPATHLEN / 2, 0, l);
837 		rw_exit(&cwdi->cwdi_lock);
838 		if (error) {
839 			PNBUF_PUT(path);
840 			return error;
841 		}
842 		len = strlen(bp);
843 		/*
844 		 * for mount points that are below our root, we can see
845 		 * them, so we fix up the pathname and return them. The
846 		 * rest we cannot see, so we don't allow viewing the
847 		 * data.
848 		 */
849 		if (strncmp(bp, sp->f_mntonname, len) == 0) {
850 			strlcpy(sp->f_mntonname, &sp->f_mntonname[len],
851 			    sizeof(sp->f_mntonname));
852 			if (sp->f_mntonname[0] == '\0')
853 				(void)strlcpy(sp->f_mntonname, "/",
854 				    sizeof(sp->f_mntonname));
855 		} else {
856 			if (root)
857 				(void)strlcpy(sp->f_mntonname, "/",
858 				    sizeof(sp->f_mntonname));
859 			else
860 				error = EPERM;
861 		}
862 		PNBUF_PUT(path);
863 	}
864 	sp->f_flag = mp->mnt_flag & MNT_VISFLAGMASK;
865 	return error;
866 }
867 
868 /*
869  * Get filesystem statistics by path.
870  */
871 int
872 do_sys_pstatvfs(struct lwp *l, const char *path, int flags, struct statvfs *sb)
873 {
874 	struct mount *mp;
875 	int error;
876 	struct nameidata nd;
877 
878 	NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, path, l);
879 	if ((error = namei(&nd)) != 0)
880 		return error;
881 	mp = nd.ni_vp->v_mount;
882 	error = dostatvfs(mp, sb, l, flags, 1);
883 	vrele(nd.ni_vp);
884 	return error;
885 }
886 
887 /* ARGSUSED */
888 int
889 sys_statvfs1(struct lwp *l, void *v, register_t *retval)
890 {
891 	struct sys_statvfs1_args /* {
892 		syscallarg(const char *) path;
893 		syscallarg(struct statvfs *) buf;
894 		syscallarg(int) flags;
895 	} */ *uap = v;
896 	struct statvfs *sb;
897 	int error;
898 
899 	sb = STATVFSBUF_GET();
900 	error = do_sys_pstatvfs(l, SCARG(uap, path), SCARG(uap, flags), sb);
901 	if (error == 0)
902 		error = copyout(sb, SCARG(uap, buf), sizeof(*sb));
903 	STATVFSBUF_PUT(sb);
904 	return error;
905 }
906 
907 /*
908  * Get filesystem statistics by fd.
909  */
910 int
911 do_sys_fstatvfs(struct lwp *l, int fd, int flags, struct statvfs *sb)
912 {
913 	struct proc *p = l->l_proc;
914 	struct file *fp;
915 	struct mount *mp;
916 	int error;
917 
918 	/* getvnode() will use the descriptor for us */
919 	if ((error = getvnode(p->p_fd, fd, &fp)) != 0)
920 		return (error);
921 	mp = ((struct vnode *)fp->f_data)->v_mount;
922 	error = dostatvfs(mp, sb, l, flags, 1);
923 	FILE_UNUSE(fp, l);
924 	return error;
925 }
926 
927 /* ARGSUSED */
928 int
929 sys_fstatvfs1(struct lwp *l, void *v, register_t *retval)
930 {
931 	struct sys_fstatvfs1_args /* {
932 		syscallarg(int) fd;
933 		syscallarg(struct statvfs *) buf;
934 		syscallarg(int) flags;
935 	} */ *uap = v;
936 	struct statvfs *sb;
937 	int error;
938 
939 	sb = STATVFSBUF_GET();
940 	error = do_sys_fstatvfs(l, SCARG(uap, fd), SCARG(uap, flags), sb);
941 	if (error == 0)
942 		error = copyout(sb, SCARG(uap, buf), sizeof(*sb));
943 	STATVFSBUF_PUT(sb);
944 	return error;
945 }
946 
947 
948 /*
949  * Get statistics on all filesystems.
950  */
951 int
952 do_sys_getvfsstat(struct lwp *l, void *sfsp, size_t bufsize, int flags,
953     int (*copyfn)(const void *, void *, size_t), size_t entry_sz,
954     register_t *retval)
955 {
956 	int root = 0;
957 	struct proc *p = l->l_proc;
958 	struct mount *mp, *nmp;
959 	struct statvfs *sb;
960 	size_t count, maxcount;
961 	int error = 0;
962 
963 	sb = STATVFSBUF_GET();
964 	maxcount = bufsize / entry_sz;
965 	mutex_enter(&mountlist_lock);
966 	count = 0;
967 	for (mp = CIRCLEQ_FIRST(&mountlist); mp != (void *)&mountlist;
968 	     mp = nmp) {
969 		if (vfs_busy(mp, LK_NOWAIT, &mountlist_lock)) {
970 			nmp = CIRCLEQ_NEXT(mp, mnt_list);
971 			continue;
972 		}
973 		if (sfsp && count < maxcount) {
974 			error = dostatvfs(mp, sb, l, flags, 0);
975 			if (error) {
976 				mutex_enter(&mountlist_lock);
977 				nmp = CIRCLEQ_NEXT(mp, mnt_list);
978 				vfs_unbusy(mp);
979 				continue;
980 			}
981 			error = copyfn(sb, sfsp, entry_sz);
982 			if (error) {
983 				vfs_unbusy(mp);
984 				goto out;
985 			}
986 			sfsp = (char *)sfsp + entry_sz;
987 			root |= strcmp(sb->f_mntonname, "/") == 0;
988 		}
989 		count++;
990 		mutex_enter(&mountlist_lock);
991 		nmp = CIRCLEQ_NEXT(mp, mnt_list);
992 		vfs_unbusy(mp);
993 	}
994 
995 	mutex_exit(&mountlist_lock);
996 	if (root == 0 && p->p_cwdi->cwdi_rdir) {
997 		/*
998 		 * fake a root entry
999 		 */
1000 		error = dostatvfs(p->p_cwdi->cwdi_rdir->v_mount,
1001 		    sb, l, flags, 1);
1002 		if (error != 0)
1003 			goto out;
1004 		if (sfsp)
1005 			error = copyfn(sb, sfsp, entry_sz);
1006 		count++;
1007 	}
1008 	if (sfsp && count > maxcount)
1009 		*retval = maxcount;
1010 	else
1011 		*retval = count;
1012 out:
1013 	STATVFSBUF_PUT(sb);
1014 	return error;
1015 }
1016 
1017 int
1018 sys_getvfsstat(struct lwp *l, void *v, register_t *retval)
1019 {
1020 	struct sys_getvfsstat_args /* {
1021 		syscallarg(struct statvfs *) buf;
1022 		syscallarg(size_t) bufsize;
1023 		syscallarg(int) flags;
1024 	} */ *uap = v;
1025 
1026 	return do_sys_getvfsstat(l, SCARG(uap, buf), SCARG(uap, bufsize),
1027 	    SCARG(uap, flags), copyout, sizeof (struct statvfs), retval);
1028 }
1029 
1030 /*
1031  * Change current working directory to a given file descriptor.
1032  */
1033 /* ARGSUSED */
1034 int
1035 sys_fchdir(struct lwp *l, void *v, register_t *retval)
1036 {
1037 	struct sys_fchdir_args /* {
1038 		syscallarg(int) fd;
1039 	} */ *uap = v;
1040 	struct proc *p = l->l_proc;
1041 	struct filedesc *fdp = p->p_fd;
1042 	struct cwdinfo *cwdi;
1043 	struct vnode *vp, *tdp;
1044 	struct mount *mp;
1045 	struct file *fp;
1046 	int error;
1047 
1048 	/* getvnode() will use the descriptor for us */
1049 	if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0)
1050 		return (error);
1051 	vp = (struct vnode *)fp->f_data;
1052 
1053 	VREF(vp);
1054 	vn_lock(vp,  LK_EXCLUSIVE | LK_RETRY);
1055 	if (vp->v_type != VDIR)
1056 		error = ENOTDIR;
1057 	else
1058 		error = VOP_ACCESS(vp, VEXEC, l->l_cred, l);
1059 	if (error) {
1060 		vput(vp);
1061 		goto out;
1062 	}
1063 	while ((mp = vp->v_mountedhere) != NULL) {
1064 		if (vfs_busy(mp, 0, 0))
1065 			continue;
1066 
1067 		vput(vp);
1068 		error = VFS_ROOT(mp, &tdp);
1069 		vfs_unbusy(mp);
1070 		if (error)
1071 			goto out;
1072 		vp = tdp;
1073 	}
1074 	VOP_UNLOCK(vp, 0);
1075 
1076 	/*
1077 	 * Disallow changing to a directory not under the process's
1078 	 * current root directory (if there is one).
1079 	 */
1080 	cwdi = p->p_cwdi;
1081 	rw_enter(&cwdi->cwdi_lock, RW_WRITER);
1082 	if (cwdi->cwdi_rdir && !vn_isunder(vp, NULL, l)) {
1083 		vrele(vp);
1084 		error = EPERM;	/* operation not permitted */
1085 	} else {
1086 		vrele(cwdi->cwdi_cdir);
1087 		cwdi->cwdi_cdir = vp;
1088 	}
1089 	rw_exit(&cwdi->cwdi_lock);
1090 
1091  out:
1092 	FILE_UNUSE(fp, l);
1093 	return (error);
1094 }
1095 
1096 /*
1097  * Change this process's notion of the root directory to a given file
1098  * descriptor.
1099  */
1100 int
1101 sys_fchroot(struct lwp *l, void *v, register_t *retval)
1102 {
1103 	struct sys_fchroot_args *uap = v;
1104 	struct proc *p = l->l_proc;
1105 	struct filedesc *fdp = p->p_fd;
1106 	struct cwdinfo *cwdi;
1107 	struct vnode	*vp;
1108 	struct file	*fp;
1109 	int		 error;
1110 
1111 	if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT,
1112  	    KAUTH_REQ_SYSTEM_CHROOT_FCHROOT, NULL, NULL, NULL)) != 0)
1113 		return error;
1114 	/* getvnode() will use the descriptor for us */
1115 	if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0)
1116 		return error;
1117 	vp = (struct vnode *) fp->f_data;
1118 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1119 	if (vp->v_type != VDIR)
1120 		error = ENOTDIR;
1121 	else
1122 		error = VOP_ACCESS(vp, VEXEC, l->l_cred, l);
1123 	VOP_UNLOCK(vp, 0);
1124 	if (error)
1125 		goto out;
1126 	VREF(vp);
1127 
1128 	/*
1129 	 * Prevent escaping from chroot by putting the root under
1130 	 * the working directory.  Silently chdir to / if we aren't
1131 	 * already there.
1132 	 */
1133 	cwdi = p->p_cwdi;
1134 	rw_enter(&cwdi->cwdi_lock, RW_WRITER);
1135 	if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) {
1136 		/*
1137 		 * XXX would be more failsafe to change directory to a
1138 		 * deadfs node here instead
1139 		 */
1140 		vrele(cwdi->cwdi_cdir);
1141 		VREF(vp);
1142 		cwdi->cwdi_cdir = vp;
1143 	}
1144 
1145 	if (cwdi->cwdi_rdir != NULL)
1146 		vrele(cwdi->cwdi_rdir);
1147 	cwdi->cwdi_rdir = vp;
1148 	rw_exit(&cwdi->cwdi_lock);
1149 
1150  out:
1151 	FILE_UNUSE(fp, l);
1152 	return (error);
1153 }
1154 
1155 /*
1156  * Change current working directory (``.'').
1157  */
1158 /* ARGSUSED */
1159 int
1160 sys_chdir(struct lwp *l, void *v, register_t *retval)
1161 {
1162 	struct sys_chdir_args /* {
1163 		syscallarg(const char *) path;
1164 	} */ *uap = v;
1165 	struct proc *p = l->l_proc;
1166 	struct cwdinfo *cwdi;
1167 	int error;
1168 	struct nameidata nd;
1169 
1170 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE,
1171 	    SCARG(uap, path), l);
1172 	if ((error = change_dir(&nd, l)) != 0)
1173 		return (error);
1174 	cwdi = p->p_cwdi;
1175 	rw_enter(&cwdi->cwdi_lock, RW_WRITER);
1176 	vrele(cwdi->cwdi_cdir);
1177 	cwdi->cwdi_cdir = nd.ni_vp;
1178 	rw_exit(&cwdi->cwdi_lock);
1179 	return (0);
1180 }
1181 
1182 /*
1183  * Change notion of root (``/'') directory.
1184  */
1185 /* ARGSUSED */
1186 int
1187 sys_chroot(struct lwp *l, void *v, register_t *retval)
1188 {
1189 	struct sys_chroot_args /* {
1190 		syscallarg(const char *) path;
1191 	} */ *uap = v;
1192 	struct proc *p = l->l_proc;
1193 	struct cwdinfo *cwdi;
1194 	struct vnode *vp;
1195 	int error;
1196 	struct nameidata nd;
1197 
1198 	if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT,
1199 	    KAUTH_REQ_SYSTEM_CHROOT_CHROOT, NULL, NULL, NULL)) != 0)
1200 		return (error);
1201 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE,
1202 	    SCARG(uap, path), l);
1203 	if ((error = change_dir(&nd, l)) != 0)
1204 		return (error);
1205 
1206 	cwdi = p->p_cwdi;
1207 	rw_enter(&cwdi->cwdi_lock, RW_WRITER);
1208 	if (cwdi->cwdi_rdir != NULL)
1209 		vrele(cwdi->cwdi_rdir);
1210 	vp = nd.ni_vp;
1211 	cwdi->cwdi_rdir = vp;
1212 
1213 	/*
1214 	 * Prevent escaping from chroot by putting the root under
1215 	 * the working directory.  Silently chdir to / if we aren't
1216 	 * already there.
1217 	 */
1218 	if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) {
1219 		/*
1220 		 * XXX would be more failsafe to change directory to a
1221 		 * deadfs node here instead
1222 		 */
1223 		vrele(cwdi->cwdi_cdir);
1224 		VREF(vp);
1225 		cwdi->cwdi_cdir = vp;
1226 	}
1227 	rw_exit(&cwdi->cwdi_lock);
1228 
1229 	return (0);
1230 }
1231 
1232 /*
1233  * Common routine for chroot and chdir.
1234  */
1235 static int
1236 change_dir(struct nameidata *ndp, struct lwp *l)
1237 {
1238 	struct vnode *vp;
1239 	int error;
1240 
1241 	if ((error = namei(ndp)) != 0)
1242 		return (error);
1243 	vp = ndp->ni_vp;
1244 	if (vp->v_type != VDIR)
1245 		error = ENOTDIR;
1246 	else
1247 		error = VOP_ACCESS(vp, VEXEC, l->l_cred, l);
1248 
1249 	if (error)
1250 		vput(vp);
1251 	else
1252 		VOP_UNLOCK(vp, 0);
1253 	return (error);
1254 }
1255 
1256 /*
1257  * Check permissions, allocate an open file structure,
1258  * and call the device open routine if any.
1259  */
1260 int
1261 sys_open(struct lwp *l, void *v, register_t *retval)
1262 {
1263 	struct sys_open_args /* {
1264 		syscallarg(const char *) path;
1265 		syscallarg(int) flags;
1266 		syscallarg(int) mode;
1267 	} */ *uap = v;
1268 	struct proc *p = l->l_proc;
1269 	struct cwdinfo *cwdi = p->p_cwdi;
1270 	struct filedesc *fdp = p->p_fd;
1271 	struct file *fp;
1272 	struct vnode *vp;
1273 	int flags, cmode;
1274 	int type, indx, error;
1275 	struct flock lf;
1276 	struct nameidata nd;
1277 
1278 	flags = FFLAGS(SCARG(uap, flags));
1279 	if ((flags & (FREAD | FWRITE)) == 0)
1280 		return (EINVAL);
1281 	/* falloc() will use the file descriptor for us */
1282 	if ((error = falloc(l, &fp, &indx)) != 0)
1283 		return (error);
1284 	/* We're going to read cwdi->cwdi_cmask unlocked here. */
1285 	cmode = ((SCARG(uap, mode) &~ cwdi->cwdi_cmask) & ALLPERMS) &~ S_ISTXT;
1286 	NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE,
1287 	    SCARG(uap, path), l);
1288 	l->l_dupfd = -indx - 1;			/* XXX check for fdopen */
1289 	if ((error = vn_open(&nd, flags, cmode)) != 0) {
1290 		rw_enter(&fdp->fd_lock, RW_WRITER);
1291 		FILE_UNUSE(fp, l);
1292 		fdp->fd_ofiles[indx] = NULL;
1293 		rw_exit(&fdp->fd_lock);
1294 		ffree(fp);
1295 		if ((error == EDUPFD || error == EMOVEFD) &&
1296 		    l->l_dupfd >= 0 &&			/* XXX from fdopen */
1297 		    (error =
1298 			dupfdopen(l, indx, l->l_dupfd, flags, error)) == 0) {
1299 			*retval = indx;
1300 			return (0);
1301 		}
1302 		if (error == ERESTART)
1303 			error = EINTR;
1304 		fdremove(fdp, indx);
1305 		return (error);
1306 	}
1307 
1308 	l->l_dupfd = 0;
1309 	vp = nd.ni_vp;
1310 	fp->f_flag = flags & FMASK;
1311 	fp->f_type = DTYPE_VNODE;
1312 	fp->f_ops = &vnops;
1313 	fp->f_data = vp;
1314 	if (flags & (O_EXLOCK | O_SHLOCK)) {
1315 		lf.l_whence = SEEK_SET;
1316 		lf.l_start = 0;
1317 		lf.l_len = 0;
1318 		if (flags & O_EXLOCK)
1319 			lf.l_type = F_WRLCK;
1320 		else
1321 			lf.l_type = F_RDLCK;
1322 		type = F_FLOCK;
1323 		if ((flags & FNONBLOCK) == 0)
1324 			type |= F_WAIT;
1325 		VOP_UNLOCK(vp, 0);
1326 		error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type);
1327 		if (error) {
1328 			(void) vn_close(vp, fp->f_flag, fp->f_cred, l);
1329 			FILE_UNUSE(fp, l);
1330 			ffree(fp);
1331 			fdremove(fdp, indx);
1332 			return (error);
1333 		}
1334 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1335 		fp->f_flag |= FHASLOCK;
1336 	}
1337 	VOP_UNLOCK(vp, 0);
1338 	*retval = indx;
1339 	FILE_SET_MATURE(fp);
1340 	FILE_UNUSE(fp, l);
1341 	return (0);
1342 }
1343 
1344 static void
1345 vfs__fhfree(fhandle_t *fhp)
1346 {
1347 	size_t fhsize;
1348 
1349 	if (fhp == NULL) {
1350 		return;
1351 	}
1352 	fhsize = FHANDLE_SIZE(fhp);
1353 	kmem_free(fhp, fhsize);
1354 }
1355 
1356 /*
1357  * vfs_composefh: compose a filehandle.
1358  */
1359 
1360 int
1361 vfs_composefh(struct vnode *vp, fhandle_t *fhp, size_t *fh_size)
1362 {
1363 	struct mount *mp;
1364 	struct fid *fidp;
1365 	int error;
1366 	size_t needfhsize;
1367 	size_t fidsize;
1368 
1369 	mp = vp->v_mount;
1370 	fidp = NULL;
1371 	if (*fh_size < FHANDLE_SIZE_MIN) {
1372 		fidsize = 0;
1373 	} else {
1374 		fidsize = *fh_size - offsetof(fhandle_t, fh_fid);
1375 		if (fhp != NULL) {
1376 			memset(fhp, 0, *fh_size);
1377 			fhp->fh_fsid = mp->mnt_stat.f_fsidx;
1378 			fidp = &fhp->fh_fid;
1379 		}
1380 	}
1381 	error = VFS_VPTOFH(vp, fidp, &fidsize);
1382 	needfhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize);
1383 	if (error == 0 && *fh_size < needfhsize) {
1384 		error = E2BIG;
1385 	}
1386 	*fh_size = needfhsize;
1387 	return error;
1388 }
1389 
1390 int
1391 vfs_composefh_alloc(struct vnode *vp, fhandle_t **fhpp)
1392 {
1393 	struct mount *mp;
1394 	fhandle_t *fhp;
1395 	size_t fhsize;
1396 	size_t fidsize;
1397 	int error;
1398 
1399 	*fhpp = NULL;
1400 	mp = vp->v_mount;
1401 	fidsize = 0;
1402 	error = VFS_VPTOFH(vp, NULL, &fidsize);
1403 	KASSERT(error != 0);
1404 	if (error != E2BIG) {
1405 		goto out;
1406 	}
1407 	fhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize);
1408 	fhp = kmem_zalloc(fhsize, KM_SLEEP);
1409 	if (fhp == NULL) {
1410 		error = ENOMEM;
1411 		goto out;
1412 	}
1413 	fhp->fh_fsid = mp->mnt_stat.f_fsidx;
1414 	error = VFS_VPTOFH(vp, &fhp->fh_fid, &fidsize);
1415 	if (error == 0) {
1416 		KASSERT((FHANDLE_SIZE(fhp) == fhsize &&
1417 		    FHANDLE_FILEID(fhp)->fid_len == fidsize));
1418 		*fhpp = fhp;
1419 	} else {
1420 		kmem_free(fhp, fhsize);
1421 	}
1422 out:
1423 	return error;
1424 }
1425 
1426 void
1427 vfs_composefh_free(fhandle_t *fhp)
1428 {
1429 
1430 	vfs__fhfree(fhp);
1431 }
1432 
1433 /*
1434  * vfs_fhtovp: lookup a vnode by a filehandle.
1435  */
1436 
1437 int
1438 vfs_fhtovp(fhandle_t *fhp, struct vnode **vpp)
1439 {
1440 	struct mount *mp;
1441 	int error;
1442 
1443 	*vpp = NULL;
1444 	mp = vfs_getvfs(FHANDLE_FSID(fhp));
1445 	if (mp == NULL) {
1446 		error = ESTALE;
1447 		goto out;
1448 	}
1449 	if (mp->mnt_op->vfs_fhtovp == NULL) {
1450 		error = EOPNOTSUPP;
1451 		goto out;
1452 	}
1453 	error = VFS_FHTOVP(mp, FHANDLE_FILEID(fhp), vpp);
1454 out:
1455 	return error;
1456 }
1457 
1458 /*
1459  * vfs_copyinfh_alloc: allocate and copyin a filehandle, given
1460  * the needed size.
1461  */
1462 
1463 int
1464 vfs_copyinfh_alloc(const void *ufhp, size_t fhsize, fhandle_t **fhpp)
1465 {
1466 	fhandle_t *fhp;
1467 	int error;
1468 
1469 	*fhpp = NULL;
1470 	if (fhsize > FHANDLE_SIZE_MAX) {
1471 		return EINVAL;
1472 	}
1473 	if (fhsize < FHANDLE_SIZE_MIN) {
1474 		return EINVAL;
1475 	}
1476 again:
1477 	fhp = kmem_alloc(fhsize, KM_SLEEP);
1478 	if (fhp == NULL) {
1479 		return ENOMEM;
1480 	}
1481 	error = copyin(ufhp, fhp, fhsize);
1482 	if (error == 0) {
1483 		/* XXX this check shouldn't be here */
1484 		if (FHANDLE_SIZE(fhp) == fhsize) {
1485 			*fhpp = fhp;
1486 			return 0;
1487 		} else if (fhsize == NFSX_V2FH && FHANDLE_SIZE(fhp) < fhsize) {
1488 			/*
1489 			 * a kludge for nfsv2 padded handles.
1490 			 */
1491 			size_t sz;
1492 
1493 			sz = FHANDLE_SIZE(fhp);
1494 			kmem_free(fhp, fhsize);
1495 			fhsize = sz;
1496 			goto again;
1497 		} else {
1498 			/*
1499 			 * userland told us wrong size.
1500 			 */
1501 		    	error = EINVAL;
1502 		}
1503 	}
1504 	kmem_free(fhp, fhsize);
1505 	return error;
1506 }
1507 
1508 void
1509 vfs_copyinfh_free(fhandle_t *fhp)
1510 {
1511 
1512 	vfs__fhfree(fhp);
1513 }
1514 
1515 /*
1516  * Get file handle system call
1517  */
1518 int
1519 sys___getfh30(struct lwp *l, void *v, register_t *retval)
1520 {
1521 	struct sys___getfh30_args /* {
1522 		syscallarg(char *) fname;
1523 		syscallarg(fhandle_t *) fhp;
1524 		syscallarg(size_t *) fh_size;
1525 	} */ *uap = v;
1526 	struct vnode *vp;
1527 	fhandle_t *fh;
1528 	int error;
1529 	struct nameidata nd;
1530 	size_t sz;
1531 	size_t usz;
1532 
1533 	/*
1534 	 * Must be super user
1535 	 */
1536 	error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE,
1537 	    0, NULL, NULL, NULL);
1538 	if (error)
1539 		return (error);
1540 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE,
1541 	    SCARG(uap, fname), l);
1542 	error = namei(&nd);
1543 	if (error)
1544 		return (error);
1545 	vp = nd.ni_vp;
1546 	error = vfs_composefh_alloc(vp, &fh);
1547 	vput(vp);
1548 	if (error != 0) {
1549 		goto out;
1550 	}
1551 	error = copyin(SCARG(uap, fh_size), &usz, sizeof(size_t));
1552 	if (error != 0) {
1553 		goto out;
1554 	}
1555 	sz = FHANDLE_SIZE(fh);
1556 	error = copyout(&sz, SCARG(uap, fh_size), sizeof(size_t));
1557 	if (error != 0) {
1558 		goto out;
1559 	}
1560 	if (usz >= sz) {
1561 		error = copyout(fh, SCARG(uap, fhp), sz);
1562 	} else {
1563 		error = E2BIG;
1564 	}
1565 out:
1566 	vfs_composefh_free(fh);
1567 	return (error);
1568 }
1569 
1570 /*
1571  * Open a file given a file handle.
1572  *
1573  * Check permissions, allocate an open file structure,
1574  * and call the device open routine if any.
1575  */
1576 
1577 int
1578 dofhopen(struct lwp *l, const void *ufhp, size_t fhsize, int oflags,
1579     register_t *retval)
1580 {
1581 	struct filedesc *fdp = l->l_proc->p_fd;
1582 	struct file *fp;
1583 	struct vnode *vp = NULL;
1584 	kauth_cred_t cred = l->l_cred;
1585 	struct file *nfp;
1586 	int type, indx, error=0;
1587 	struct flock lf;
1588 	struct vattr va;
1589 	fhandle_t *fh;
1590 	int flags;
1591 
1592 	/*
1593 	 * Must be super user
1594 	 */
1595 	if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE,
1596 	    0, NULL, NULL, NULL)))
1597 		return (error);
1598 
1599 	flags = FFLAGS(oflags);
1600 	if ((flags & (FREAD | FWRITE)) == 0)
1601 		return (EINVAL);
1602 	if ((flags & O_CREAT))
1603 		return (EINVAL);
1604 	/* falloc() will use the file descriptor for us */
1605 	if ((error = falloc(l, &nfp, &indx)) != 0)
1606 		return (error);
1607 	fp = nfp;
1608 	error = vfs_copyinfh_alloc(ufhp, fhsize, &fh);
1609 	if (error != 0) {
1610 		goto bad;
1611 	}
1612 	error = vfs_fhtovp(fh, &vp);
1613 	if (error != 0) {
1614 		goto bad;
1615 	}
1616 
1617 	/* Now do an effective vn_open */
1618 
1619 	if (vp->v_type == VSOCK) {
1620 		error = EOPNOTSUPP;
1621 		goto bad;
1622 	}
1623 	if (flags & FREAD) {
1624 		if ((error = VOP_ACCESS(vp, VREAD, cred, l)) != 0)
1625 			goto bad;
1626 	}
1627 	if (flags & (FWRITE | O_TRUNC)) {
1628 		if (vp->v_type == VDIR) {
1629 			error = EISDIR;
1630 			goto bad;
1631 		}
1632 		if ((error = vn_writechk(vp)) != 0 ||
1633 		    (error = VOP_ACCESS(vp, VWRITE, cred, l)) != 0)
1634 			goto bad;
1635 	}
1636 	if (flags & O_TRUNC) {
1637 		VOP_UNLOCK(vp, 0);			/* XXX */
1638 		VOP_LEASE(vp, l, cred, LEASE_WRITE);
1639 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);   /* XXX */
1640 		VATTR_NULL(&va);
1641 		va.va_size = 0;
1642 		error = VOP_SETATTR(vp, &va, cred, l);
1643 		if (error)
1644 			goto bad;
1645 	}
1646 	if ((error = VOP_OPEN(vp, flags, cred, l)) != 0)
1647 		goto bad;
1648 	if (flags & FWRITE)
1649 		vp->v_writecount++;
1650 
1651 	/* done with modified vn_open, now finish what sys_open does. */
1652 
1653 	fp->f_flag = flags & FMASK;
1654 	fp->f_type = DTYPE_VNODE;
1655 	fp->f_ops = &vnops;
1656 	fp->f_data = vp;
1657 	if (flags & (O_EXLOCK | O_SHLOCK)) {
1658 		lf.l_whence = SEEK_SET;
1659 		lf.l_start = 0;
1660 		lf.l_len = 0;
1661 		if (flags & O_EXLOCK)
1662 			lf.l_type = F_WRLCK;
1663 		else
1664 			lf.l_type = F_RDLCK;
1665 		type = F_FLOCK;
1666 		if ((flags & FNONBLOCK) == 0)
1667 			type |= F_WAIT;
1668 		VOP_UNLOCK(vp, 0);
1669 		error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type);
1670 		if (error) {
1671 			(void) vn_close(vp, fp->f_flag, fp->f_cred, l);
1672 			FILE_UNUSE(fp, l);
1673 			ffree(fp);
1674 			fdremove(fdp, indx);
1675 			return (error);
1676 		}
1677 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1678 		fp->f_flag |= FHASLOCK;
1679 	}
1680 	VOP_UNLOCK(vp, 0);
1681 	*retval = indx;
1682 	FILE_SET_MATURE(fp);
1683 	FILE_UNUSE(fp, l);
1684 	vfs_copyinfh_free(fh);
1685 	return (0);
1686 
1687 bad:
1688 	FILE_UNUSE(fp, l);
1689 	ffree(fp);
1690 	fdremove(fdp, indx);
1691 	if (vp != NULL)
1692 		vput(vp);
1693 	vfs_copyinfh_free(fh);
1694 	return (error);
1695 }
1696 
1697 int
1698 sys___fhopen40(struct lwp *l, void *v, register_t *retval)
1699 {
1700 	struct sys___fhopen40_args /* {
1701 		syscallarg(const void *) fhp;
1702 		syscallarg(size_t) fh_size;
1703 		syscallarg(int) flags;
1704 	} */ *uap = v;
1705 
1706 	return dofhopen(l, SCARG(uap, fhp), SCARG(uap, fh_size),
1707 	    SCARG(uap, flags), retval);
1708 }
1709 
1710 int
1711 do_fhstat(struct lwp *l, const void *ufhp, size_t fhsize, struct stat *sb)
1712 {
1713 	int error;
1714 	fhandle_t *fh;
1715 	struct vnode *vp;
1716 
1717 	/*
1718 	 * Must be super user
1719 	 */
1720 	if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE,
1721 	    0, NULL, NULL, NULL)))
1722 		return (error);
1723 
1724 	error = vfs_copyinfh_alloc(ufhp, fhsize, &fh);
1725 	if (error != 0)
1726 		return error;
1727 
1728 	error = vfs_fhtovp(fh, &vp);
1729 	vfs_copyinfh_free(fh);
1730 	if (error != 0)
1731 		return error;
1732 
1733 	error = vn_stat(vp, sb, l);
1734 	vput(vp);
1735 	return error;
1736 }
1737 
1738 
1739 /* ARGSUSED */
1740 int
1741 sys___fhstat40(struct lwp *l, void *v, register_t *retval)
1742 {
1743 	struct sys___fhstat40_args /* {
1744 		syscallarg(const void *) fhp;
1745 		syscallarg(size_t) fh_size;
1746 		syscallarg(struct stat *) sb;
1747 	} */ *uap = v;
1748 	struct stat sb;
1749 	int error;
1750 
1751 	error = do_fhstat(l, SCARG(uap, fhp), SCARG(uap, fh_size), &sb);
1752 	if (error)
1753 		return error;
1754 	return copyout(&sb, SCARG(uap, sb), sizeof(sb));
1755 }
1756 
1757 int
1758 do_fhstatvfs(struct lwp *l, const void *ufhp, size_t fhsize, struct statvfs *sb,
1759     int flags)
1760 {
1761 	fhandle_t *fh;
1762 	struct mount *mp;
1763 	struct vnode *vp;
1764 	int error;
1765 
1766 	/*
1767 	 * Must be super user
1768 	 */
1769 	if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE,
1770 	    0, NULL, NULL, NULL)))
1771 		return error;
1772 
1773 	error = vfs_copyinfh_alloc(ufhp, fhsize, &fh);
1774 	if (error != 0)
1775 		return error;
1776 
1777 	error = vfs_fhtovp(fh, &vp);
1778 	vfs_copyinfh_free(fh);
1779 	if (error != 0)
1780 		return error;
1781 
1782 	mp = vp->v_mount;
1783 	error = dostatvfs(mp, sb, l, flags, 1);
1784 	vput(vp);
1785 	return error;
1786 }
1787 
1788 /* ARGSUSED */
1789 int
1790 sys___fhstatvfs140(struct lwp *l, void *v, register_t *retval)
1791 {
1792 	struct sys___fhstatvfs140_args /* {
1793 		syscallarg(const void *) fhp;
1794 		syscallarg(size_t) fh_size;
1795 		syscallarg(struct statvfs *) buf;
1796 		syscallarg(int)	flags;
1797 	} */ *uap = v;
1798 	struct statvfs *sb = STATVFSBUF_GET();
1799 	int error;
1800 
1801 	error = do_fhstatvfs(l, SCARG(uap, fhp), SCARG(uap, fh_size), sb,
1802 	    SCARG(uap, flags));
1803 	if (error == 0)
1804 		error = copyout(sb, SCARG(uap, buf), sizeof(*sb));
1805 	STATVFSBUF_PUT(sb);
1806 	return error;
1807 }
1808 
1809 /*
1810  * Create a special file.
1811  */
1812 /* ARGSUSED */
1813 int
1814 sys_mknod(struct lwp *l, void *v, register_t *retval)
1815 {
1816 	struct sys_mknod_args /* {
1817 		syscallarg(const char *) path;
1818 		syscallarg(int) mode;
1819 		syscallarg(int) dev;
1820 	} */ *uap = v;
1821 	struct proc *p = l->l_proc;
1822 	struct vnode *vp;
1823 	struct vattr vattr;
1824 	int error, optype;
1825 	struct nameidata nd;
1826 	char *path;
1827 	const char *cpath;
1828 	enum uio_seg seg = UIO_USERSPACE;
1829 
1830 	if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MKNOD,
1831 	    0, NULL, NULL, NULL)) != 0)
1832 		return (error);
1833 
1834 	optype = VOP_MKNOD_DESCOFFSET;
1835 
1836 	VERIEXEC_PATH_GET(SCARG(uap, path), seg, cpath, path);
1837 	NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, seg, cpath, l);
1838 
1839 	if ((error = namei(&nd)) != 0)
1840 		goto out;
1841 	vp = nd.ni_vp;
1842 	if (vp != NULL)
1843 		error = EEXIST;
1844 	else {
1845 		VATTR_NULL(&vattr);
1846 		/* We will read cwdi->cwdi_cmask unlocked. */
1847 		vattr.va_mode =
1848 		    (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask;
1849 		vattr.va_rdev = SCARG(uap, dev);
1850 
1851 		switch (SCARG(uap, mode) & S_IFMT) {
1852 		case S_IFMT:	/* used by badsect to flag bad sectors */
1853 			vattr.va_type = VBAD;
1854 			break;
1855 		case S_IFCHR:
1856 			vattr.va_type = VCHR;
1857 			break;
1858 		case S_IFBLK:
1859 			vattr.va_type = VBLK;
1860 			break;
1861 		case S_IFWHT:
1862 			optype = VOP_WHITEOUT_DESCOFFSET;
1863 			break;
1864 		case S_IFREG:
1865 #if NVERIEXEC > 0
1866 			error = veriexec_openchk(l, nd.ni_vp, nd.ni_dirp,
1867 			    O_CREAT);
1868 #endif /* NVERIEXEC > 0 */
1869 			vattr.va_type = VREG;
1870 			vattr.va_rdev = VNOVAL;
1871 			optype = VOP_CREATE_DESCOFFSET;
1872 			break;
1873 		default:
1874 			error = EINVAL;
1875 			break;
1876 		}
1877 	}
1878 	if (!error) {
1879 		VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE);
1880 		switch (optype) {
1881 		case VOP_WHITEOUT_DESCOFFSET:
1882 			error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE);
1883 			if (error)
1884 				VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1885 			vput(nd.ni_dvp);
1886 			break;
1887 
1888 		case VOP_MKNOD_DESCOFFSET:
1889 			error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp,
1890 						&nd.ni_cnd, &vattr);
1891 			if (error == 0)
1892 				vput(nd.ni_vp);
1893 			break;
1894 
1895 		case VOP_CREATE_DESCOFFSET:
1896 			error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp,
1897 						&nd.ni_cnd, &vattr);
1898 			if (error == 0)
1899 				vput(nd.ni_vp);
1900 			break;
1901 		}
1902 	} else {
1903 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1904 		if (nd.ni_dvp == vp)
1905 			vrele(nd.ni_dvp);
1906 		else
1907 			vput(nd.ni_dvp);
1908 		if (vp)
1909 			vrele(vp);
1910 	}
1911 out:
1912 	VERIEXEC_PATH_PUT(path);
1913 	return (error);
1914 }
1915 
1916 /*
1917  * Create a named pipe.
1918  */
1919 /* ARGSUSED */
1920 int
1921 sys_mkfifo(struct lwp *l, void *v, register_t *retval)
1922 {
1923 	struct sys_mkfifo_args /* {
1924 		syscallarg(const char *) path;
1925 		syscallarg(int) mode;
1926 	} */ *uap = v;
1927 	struct proc *p = l->l_proc;
1928 	struct vattr vattr;
1929 	int error;
1930 	struct nameidata nd;
1931 
1932 	NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l);
1933 	if ((error = namei(&nd)) != 0)
1934 		return (error);
1935 	if (nd.ni_vp != NULL) {
1936 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1937 		if (nd.ni_dvp == nd.ni_vp)
1938 			vrele(nd.ni_dvp);
1939 		else
1940 			vput(nd.ni_dvp);
1941 		vrele(nd.ni_vp);
1942 		return (EEXIST);
1943 	}
1944 	VATTR_NULL(&vattr);
1945 	vattr.va_type = VFIFO;
1946 	/* We will read cwdi->cwdi_cmask unlocked. */
1947 	vattr.va_mode = (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask;
1948 	VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE);
1949 	error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
1950 	if (error == 0)
1951 		vput(nd.ni_vp);
1952 	return (error);
1953 }
1954 
1955 /*
1956  * Make a hard file link.
1957  */
1958 /* ARGSUSED */
1959 int
1960 sys_link(struct lwp *l, void *v, register_t *retval)
1961 {
1962 	struct sys_link_args /* {
1963 		syscallarg(const char *) path;
1964 		syscallarg(const char *) link;
1965 	} */ *uap = v;
1966 	struct vnode *vp;
1967 	struct nameidata nd;
1968 	int error;
1969 
1970 	NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE,
1971 	    SCARG(uap, path), l);
1972 	if ((error = namei(&nd)) != 0)
1973 		return (error);
1974 	vp = nd.ni_vp;
1975 	NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, UIO_USERSPACE,
1976 	    SCARG(uap, link), l);
1977 	if ((error = namei(&nd)) != 0)
1978 		goto out;
1979 	if (nd.ni_vp) {
1980 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1981 		if (nd.ni_dvp == nd.ni_vp)
1982 			vrele(nd.ni_dvp);
1983 		else
1984 			vput(nd.ni_dvp);
1985 		vrele(nd.ni_vp);
1986 		error = EEXIST;
1987 		goto out;
1988 	}
1989 	VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE);
1990 	VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE);
1991 	error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
1992 out:
1993 	vrele(vp);
1994 	return (error);
1995 }
1996 
1997 /*
1998  * Make a symbolic link.
1999  */
2000 /* ARGSUSED */
2001 int
2002 sys_symlink(struct lwp *l, void *v, register_t *retval)
2003 {
2004 	struct sys_symlink_args /* {
2005 		syscallarg(const char *) path;
2006 		syscallarg(const char *) link;
2007 	} */ *uap = v;
2008 	struct proc *p = l->l_proc;
2009 	struct vattr vattr;
2010 	char *path;
2011 	int error;
2012 	struct nameidata nd;
2013 
2014 	path = PNBUF_GET();
2015 	error = copyinstr(SCARG(uap, path), path, MAXPATHLEN, NULL);
2016 	if (error)
2017 		goto out;
2018 	NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, UIO_USERSPACE,
2019 	    SCARG(uap, link), l);
2020 	if ((error = namei(&nd)) != 0)
2021 		goto out;
2022 	if (nd.ni_vp) {
2023 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2024 		if (nd.ni_dvp == nd.ni_vp)
2025 			vrele(nd.ni_dvp);
2026 		else
2027 			vput(nd.ni_dvp);
2028 		vrele(nd.ni_vp);
2029 		error = EEXIST;
2030 		goto out;
2031 	}
2032 	VATTR_NULL(&vattr);
2033 	vattr.va_type = VLNK;
2034 	/* We will read cwdi->cwdi_cmask unlocked. */
2035 	vattr.va_mode = ACCESSPERMS &~ p->p_cwdi->cwdi_cmask;
2036 	VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE);
2037 	error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, path);
2038 	if (error == 0)
2039 		vput(nd.ni_vp);
2040 out:
2041 	PNBUF_PUT(path);
2042 	return (error);
2043 }
2044 
2045 /*
2046  * Delete a whiteout from the filesystem.
2047  */
2048 /* ARGSUSED */
2049 int
2050 sys_undelete(struct lwp *l, void *v, register_t *retval)
2051 {
2052 	struct sys_undelete_args /* {
2053 		syscallarg(const char *) path;
2054 	} */ *uap = v;
2055 	int error;
2056 	struct nameidata nd;
2057 
2058 	NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | TRYEMULROOT,
2059 	    UIO_USERSPACE, SCARG(uap, path), l);
2060 	error = namei(&nd);
2061 	if (error)
2062 		return (error);
2063 
2064 	if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) {
2065 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2066 		if (nd.ni_dvp == nd.ni_vp)
2067 			vrele(nd.ni_dvp);
2068 		else
2069 			vput(nd.ni_dvp);
2070 		if (nd.ni_vp)
2071 			vrele(nd.ni_vp);
2072 		return (EEXIST);
2073 	}
2074 	VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE);
2075 	if ((error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE)) != 0)
2076 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2077 	vput(nd.ni_dvp);
2078 	return (error);
2079 }
2080 
2081 /*
2082  * Delete a name from the filesystem.
2083  */
2084 /* ARGSUSED */
2085 int
2086 sys_unlink(struct lwp *l, void *v, register_t *retval)
2087 {
2088 	struct sys_unlink_args /* {
2089 		syscallarg(const char *) path;
2090 	} */ *uap = v;
2091 	struct vnode *vp;
2092 	int error;
2093 	struct nameidata nd;
2094 	char *path;
2095 	const char *cpath;
2096 	enum uio_seg seg = UIO_USERSPACE;
2097 
2098 	VERIEXEC_PATH_GET(SCARG(uap, path), seg, cpath, path);
2099 	NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF | TRYEMULROOT, seg, cpath, l);
2100 
2101 	if ((error = namei(&nd)) != 0)
2102 		goto out;
2103 	vp = nd.ni_vp;
2104 
2105 	/*
2106 	 * The root of a mounted filesystem cannot be deleted.
2107 	 */
2108 	if (vp->v_vflag & VV_ROOT) {
2109 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2110 		if (nd.ni_dvp == vp)
2111 			vrele(nd.ni_dvp);
2112 		else
2113 			vput(nd.ni_dvp);
2114 		vput(vp);
2115 		error = EBUSY;
2116 		goto out;
2117 	}
2118 
2119 #if NVERIEXEC > 0
2120 	/* Handle remove requests for veriexec entries. */
2121 	if ((error = veriexec_removechk(l, nd.ni_vp, nd.ni_dirp)) != 0) {
2122 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2123 		if (nd.ni_dvp == vp)
2124 			vrele(nd.ni_dvp);
2125 		else
2126 			vput(nd.ni_dvp);
2127 		vput(vp);
2128 		goto out;
2129 	}
2130 #endif /* NVERIEXEC > 0 */
2131 
2132 	VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE);
2133 	VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE);
2134 #ifdef FILEASSOC
2135 	(void)fileassoc_file_delete(vp);
2136 #endif /* FILEASSOC */
2137 	error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
2138 out:
2139 	VERIEXEC_PATH_PUT(path);
2140 	return (error);
2141 }
2142 
2143 /*
2144  * Reposition read/write file offset.
2145  */
2146 int
2147 sys_lseek(struct lwp *l, void *v, register_t *retval)
2148 {
2149 	struct sys_lseek_args /* {
2150 		syscallarg(int) fd;
2151 		syscallarg(int) pad;
2152 		syscallarg(off_t) offset;
2153 		syscallarg(int) whence;
2154 	} */ *uap = v;
2155 	struct proc *p = l->l_proc;
2156 	kauth_cred_t cred = l->l_cred;
2157 	struct filedesc *fdp = p->p_fd;
2158 	struct file *fp;
2159 	struct vnode *vp;
2160 	struct vattr vattr;
2161 	off_t newoff;
2162 	int error;
2163 
2164 	if ((fp = fd_getfile(fdp, SCARG(uap, fd))) == NULL)
2165 		return (EBADF);
2166 
2167 	vp = (struct vnode *)fp->f_data;
2168 	if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
2169 		error = ESPIPE;
2170 		mutex_exit(&fp->f_lock);
2171 		goto out;
2172 	}
2173 
2174 	switch (SCARG(uap, whence)) {
2175 	case SEEK_CUR:
2176 		newoff = fp->f_offset + SCARG(uap, offset);
2177 		FILE_USE(fp);
2178 		break;
2179 	case SEEK_END:
2180 		FILE_USE(fp);
2181 		error = VOP_GETATTR(vp, &vattr, cred, l);
2182 		if (error) {
2183 			FILE_UNUSE(fp, l);
2184 			goto out;
2185 		}
2186 		newoff = SCARG(uap, offset) + vattr.va_size;
2187 		break;
2188 	case SEEK_SET:
2189 		FILE_USE(fp);
2190 		newoff = SCARG(uap, offset);
2191 		break;
2192 	default:
2193 		mutex_exit(&fp->f_lock);
2194 		error = EINVAL;
2195 		goto out;
2196 	}
2197 	if ((error = VOP_SEEK(vp, fp->f_offset, newoff, cred)) == 0) {
2198 		mutex_enter(&fp->f_lock);
2199 		*(off_t *)retval = fp->f_offset = newoff;
2200 		mutex_exit(&fp->f_lock);
2201 	}
2202 	FILE_UNUSE(fp, l);
2203  out:
2204 	return (error);
2205 }
2206 
2207 /*
2208  * Positional read system call.
2209  */
2210 int
2211 sys_pread(struct lwp *l, void *v, register_t *retval)
2212 {
2213 	struct sys_pread_args /* {
2214 		syscallarg(int) fd;
2215 		syscallarg(void *) buf;
2216 		syscallarg(size_t) nbyte;
2217 		syscallarg(off_t) offset;
2218 	} */ *uap = v;
2219 	struct proc *p = l->l_proc;
2220 	struct filedesc *fdp = p->p_fd;
2221 	struct file *fp;
2222 	struct vnode *vp;
2223 	off_t offset;
2224 	int error, fd = SCARG(uap, fd);
2225 
2226 	if ((fp = fd_getfile(fdp, fd)) == NULL)
2227 		return (EBADF);
2228 
2229 	if ((fp->f_flag & FREAD) == 0) {
2230 		mutex_exit(&fp->f_lock);
2231 		return (EBADF);
2232 	}
2233 
2234 	FILE_USE(fp);
2235 
2236 	vp = (struct vnode *)fp->f_data;
2237 	if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
2238 		error = ESPIPE;
2239 		goto out;
2240 	}
2241 
2242 	offset = SCARG(uap, offset);
2243 
2244 	/*
2245 	 * XXX This works because no file systems actually
2246 	 * XXX take any action on the seek operation.
2247 	 */
2248 	if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0)
2249 		goto out;
2250 
2251 	/* dofileread() will unuse the descriptor for us */
2252 	return (dofileread(fd, fp, SCARG(uap, buf), SCARG(uap, nbyte),
2253 	    &offset, 0, retval));
2254 
2255  out:
2256 	FILE_UNUSE(fp, l);
2257 	return (error);
2258 }
2259 
2260 /*
2261  * Positional scatter read system call.
2262  */
2263 int
2264 sys_preadv(struct lwp *l, void *v, register_t *retval)
2265 {
2266 	struct sys_preadv_args /* {
2267 		syscallarg(int) fd;
2268 		syscallarg(const struct iovec *) iovp;
2269 		syscallarg(int) iovcnt;
2270 		syscallarg(off_t) offset;
2271 	} */ *uap = v;
2272 
2273 	return do_filereadv(SCARG(uap, fd), SCARG(uap, iovp),
2274 	    SCARG(uap, iovcnt), &SCARG(uap, offset), 0, retval);
2275 }
2276 
2277 /*
2278  * Positional write system call.
2279  */
2280 int
2281 sys_pwrite(struct lwp *l, void *v, register_t *retval)
2282 {
2283 	struct sys_pwrite_args /* {
2284 		syscallarg(int) fd;
2285 		syscallarg(const void *) buf;
2286 		syscallarg(size_t) nbyte;
2287 		syscallarg(off_t) offset;
2288 	} */ *uap = v;
2289 	struct proc *p = l->l_proc;
2290 	struct filedesc *fdp = p->p_fd;
2291 	struct file *fp;
2292 	struct vnode *vp;
2293 	off_t offset;
2294 	int error, fd = SCARG(uap, fd);
2295 
2296 	if ((fp = fd_getfile(fdp, fd)) == NULL)
2297 		return (EBADF);
2298 
2299 	if ((fp->f_flag & FWRITE) == 0) {
2300 		mutex_exit(&fp->f_lock);
2301 		return (EBADF);
2302 	}
2303 
2304 	FILE_USE(fp);
2305 
2306 	vp = (struct vnode *)fp->f_data;
2307 	if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
2308 		error = ESPIPE;
2309 		goto out;
2310 	}
2311 
2312 	offset = SCARG(uap, offset);
2313 
2314 	/*
2315 	 * XXX This works because no file systems actually
2316 	 * XXX take any action on the seek operation.
2317 	 */
2318 	if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0)
2319 		goto out;
2320 
2321 	/* dofilewrite() will unuse the descriptor for us */
2322 	return (dofilewrite(fd, fp, SCARG(uap, buf), SCARG(uap, nbyte),
2323 	    &offset, 0, retval));
2324 
2325  out:
2326 	FILE_UNUSE(fp, l);
2327 	return (error);
2328 }
2329 
2330 /*
2331  * Positional gather write system call.
2332  */
2333 int
2334 sys_pwritev(struct lwp *l, void *v, register_t *retval)
2335 {
2336 	struct sys_pwritev_args /* {
2337 		syscallarg(int) fd;
2338 		syscallarg(const struct iovec *) iovp;
2339 		syscallarg(int) iovcnt;
2340 		syscallarg(off_t) offset;
2341 	} */ *uap = v;
2342 
2343 	return do_filewritev(SCARG(uap, fd), SCARG(uap, iovp),
2344 	    SCARG(uap, iovcnt), &SCARG(uap, offset), 0, retval);
2345 }
2346 
2347 /*
2348  * Check access permissions.
2349  */
2350 int
2351 sys_access(struct lwp *l, void *v, register_t *retval)
2352 {
2353 	struct sys_access_args /* {
2354 		syscallarg(const char *) path;
2355 		syscallarg(int) flags;
2356 	} */ *uap = v;
2357 	kauth_cred_t cred;
2358 	struct vnode *vp;
2359 	int error, flags;
2360 	struct nameidata nd;
2361 
2362 	cred = kauth_cred_dup(l->l_cred);
2363 	kauth_cred_seteuid(cred, kauth_cred_getuid(l->l_cred));
2364 	kauth_cred_setegid(cred, kauth_cred_getgid(l->l_cred));
2365 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE,
2366 	    SCARG(uap, path), l);
2367 	/* Override default credentials */
2368 	nd.ni_cnd.cn_cred = cred;
2369 	if ((error = namei(&nd)) != 0)
2370 		goto out;
2371 	vp = nd.ni_vp;
2372 
2373 	/* Flags == 0 means only check for existence. */
2374 	if (SCARG(uap, flags)) {
2375 		flags = 0;
2376 		if (SCARG(uap, flags) & R_OK)
2377 			flags |= VREAD;
2378 		if (SCARG(uap, flags) & W_OK)
2379 			flags |= VWRITE;
2380 		if (SCARG(uap, flags) & X_OK)
2381 			flags |= VEXEC;
2382 
2383 		error = VOP_ACCESS(vp, flags, cred, l);
2384 		if (!error && (flags & VWRITE))
2385 			error = vn_writechk(vp);
2386 	}
2387 	vput(vp);
2388 out:
2389 	kauth_cred_free(cred);
2390 	return (error);
2391 }
2392 
2393 /*
2394  * Common code for all sys_stat functions, including compat versions.
2395  */
2396 int
2397 do_sys_stat(struct lwp *l, const char *path, unsigned int nd_flags,
2398     struct stat *sb)
2399 {
2400 	int error;
2401 	struct nameidata nd;
2402 
2403 	NDINIT(&nd, LOOKUP, nd_flags | LOCKLEAF | TRYEMULROOT,
2404 	    UIO_USERSPACE, path, l);
2405 	error = namei(&nd);
2406 	if (error != 0)
2407 		return error;
2408 	error = vn_stat(nd.ni_vp, sb, l);
2409 	vput(nd.ni_vp);
2410 	return error;
2411 }
2412 
2413 /*
2414  * Get file status; this version follows links.
2415  */
2416 /* ARGSUSED */
2417 int
2418 sys___stat30(struct lwp *l, void *v, register_t *retval)
2419 {
2420 	struct sys___stat30_args /* {
2421 		syscallarg(const char *) path;
2422 		syscallarg(struct stat *) ub;
2423 	} */ *uap = v;
2424 	struct stat sb;
2425 	int error;
2426 
2427 	error = do_sys_stat(l, SCARG(uap, path), FOLLOW, &sb);
2428 	if (error)
2429 		return error;
2430 	return copyout(&sb, SCARG(uap, ub), sizeof(sb));
2431 }
2432 
2433 /*
2434  * Get file status; this version does not follow links.
2435  */
2436 /* ARGSUSED */
2437 int
2438 sys___lstat30(struct lwp *l, void *v, register_t *retval)
2439 {
2440 	struct sys___lstat30_args /* {
2441 		syscallarg(const char *) path;
2442 		syscallarg(struct stat *) ub;
2443 	} */ *uap = v;
2444 	struct stat sb;
2445 	int error;
2446 
2447 	error = do_sys_stat(l, SCARG(uap, path), NOFOLLOW, &sb);
2448 	if (error)
2449 		return error;
2450 	return copyout(&sb, SCARG(uap, ub), sizeof(sb));
2451 }
2452 
2453 /*
2454  * Get configurable pathname variables.
2455  */
2456 /* ARGSUSED */
2457 int
2458 sys_pathconf(struct lwp *l, void *v, register_t *retval)
2459 {
2460 	struct sys_pathconf_args /* {
2461 		syscallarg(const char *) path;
2462 		syscallarg(int) name;
2463 	} */ *uap = v;
2464 	int error;
2465 	struct nameidata nd;
2466 
2467 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE,
2468 	    SCARG(uap, path), l);
2469 	if ((error = namei(&nd)) != 0)
2470 		return (error);
2471 	error = VOP_PATHCONF(nd.ni_vp, SCARG(uap, name), retval);
2472 	vput(nd.ni_vp);
2473 	return (error);
2474 }
2475 
2476 /*
2477  * Return target name of a symbolic link.
2478  */
2479 /* ARGSUSED */
2480 int
2481 sys_readlink(struct lwp *l, void *v, register_t *retval)
2482 {
2483 	struct sys_readlink_args /* {
2484 		syscallarg(const char *) path;
2485 		syscallarg(char *) buf;
2486 		syscallarg(size_t) count;
2487 	} */ *uap = v;
2488 	struct vnode *vp;
2489 	struct iovec aiov;
2490 	struct uio auio;
2491 	int error;
2492 	struct nameidata nd;
2493 
2494 	NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE,
2495 	    SCARG(uap, path), l);
2496 	if ((error = namei(&nd)) != 0)
2497 		return (error);
2498 	vp = nd.ni_vp;
2499 	if (vp->v_type != VLNK)
2500 		error = EINVAL;
2501 	else if (!(vp->v_mount->mnt_flag & MNT_SYMPERM) ||
2502 	    (error = VOP_ACCESS(vp, VREAD, l->l_cred, l)) == 0) {
2503 		aiov.iov_base = SCARG(uap, buf);
2504 		aiov.iov_len = SCARG(uap, count);
2505 		auio.uio_iov = &aiov;
2506 		auio.uio_iovcnt = 1;
2507 		auio.uio_offset = 0;
2508 		auio.uio_rw = UIO_READ;
2509 		KASSERT(l == curlwp);
2510 		auio.uio_vmspace = l->l_proc->p_vmspace;
2511 		auio.uio_resid = SCARG(uap, count);
2512 		error = VOP_READLINK(vp, &auio, l->l_cred);
2513 	}
2514 	vput(vp);
2515 	*retval = SCARG(uap, count) - auio.uio_resid;
2516 	return (error);
2517 }
2518 
2519 /*
2520  * Change flags of a file given a path name.
2521  */
2522 /* ARGSUSED */
2523 int
2524 sys_chflags(struct lwp *l, void *v, register_t *retval)
2525 {
2526 	struct sys_chflags_args /* {
2527 		syscallarg(const char *) path;
2528 		syscallarg(u_long) flags;
2529 	} */ *uap = v;
2530 	struct vnode *vp;
2531 	int error;
2532 	struct nameidata nd;
2533 
2534 	NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE,
2535 	    SCARG(uap, path), l);
2536 	if ((error = namei(&nd)) != 0)
2537 		return (error);
2538 	vp = nd.ni_vp;
2539 	error = change_flags(vp, SCARG(uap, flags), l);
2540 	vput(vp);
2541 	return (error);
2542 }
2543 
2544 /*
2545  * Change flags of a file given a file descriptor.
2546  */
2547 /* ARGSUSED */
2548 int
2549 sys_fchflags(struct lwp *l, void *v, register_t *retval)
2550 {
2551 	struct sys_fchflags_args /* {
2552 		syscallarg(int) fd;
2553 		syscallarg(u_long) flags;
2554 	} */ *uap = v;
2555 	struct proc *p = l->l_proc;
2556 	struct vnode *vp;
2557 	struct file *fp;
2558 	int error;
2559 
2560 	/* getvnode() will use the descriptor for us */
2561 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2562 		return (error);
2563 	vp = (struct vnode *)fp->f_data;
2564 	error = change_flags(vp, SCARG(uap, flags), l);
2565 	VOP_UNLOCK(vp, 0);
2566 	FILE_UNUSE(fp, l);
2567 	return (error);
2568 }
2569 
2570 /*
2571  * Change flags of a file given a path name; this version does
2572  * not follow links.
2573  */
2574 int
2575 sys_lchflags(struct lwp *l, void *v, register_t *retval)
2576 {
2577 	struct sys_lchflags_args /* {
2578 		syscallarg(const char *) path;
2579 		syscallarg(u_long) flags;
2580 	} */ *uap = v;
2581 	struct vnode *vp;
2582 	int error;
2583 	struct nameidata nd;
2584 
2585 	NDINIT(&nd, LOOKUP, NOFOLLOW | TRYEMULROOT, UIO_USERSPACE,
2586 	    SCARG(uap, path), l);
2587 	if ((error = namei(&nd)) != 0)
2588 		return (error);
2589 	vp = nd.ni_vp;
2590 	error = change_flags(vp, SCARG(uap, flags), l);
2591 	vput(vp);
2592 	return (error);
2593 }
2594 
2595 /*
2596  * Common routine to change flags of a file.
2597  */
2598 int
2599 change_flags(struct vnode *vp, u_long flags, struct lwp *l)
2600 {
2601 	struct vattr vattr;
2602 	int error;
2603 
2604 	VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE);
2605 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2606 	/*
2607 	 * Non-superusers cannot change the flags on devices, even if they
2608 	 * own them.
2609 	 */
2610 	if (kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER, NULL)) {
2611 		if ((error = VOP_GETATTR(vp, &vattr, l->l_cred, l)) != 0)
2612 			goto out;
2613 		if (vattr.va_type == VCHR || vattr.va_type == VBLK) {
2614 			error = EINVAL;
2615 			goto out;
2616 		}
2617 	}
2618 	VATTR_NULL(&vattr);
2619 	vattr.va_flags = flags;
2620 	error = VOP_SETATTR(vp, &vattr, l->l_cred, l);
2621 out:
2622 	return (error);
2623 }
2624 
2625 /*
2626  * Change mode of a file given path name; this version follows links.
2627  */
2628 /* ARGSUSED */
2629 int
2630 sys_chmod(struct lwp *l, void *v, register_t *retval)
2631 {
2632 	struct sys_chmod_args /* {
2633 		syscallarg(const char *) path;
2634 		syscallarg(int) mode;
2635 	} */ *uap = v;
2636 	int error;
2637 	struct nameidata nd;
2638 
2639 	NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE,
2640 	    SCARG(uap, path), l);
2641 	if ((error = namei(&nd)) != 0)
2642 		return (error);
2643 
2644 	error = change_mode(nd.ni_vp, SCARG(uap, mode), l);
2645 
2646 	vrele(nd.ni_vp);
2647 	return (error);
2648 }
2649 
2650 /*
2651  * Change mode of a file given a file descriptor.
2652  */
2653 /* ARGSUSED */
2654 int
2655 sys_fchmod(struct lwp *l, void *v, register_t *retval)
2656 {
2657 	struct sys_fchmod_args /* {
2658 		syscallarg(int) fd;
2659 		syscallarg(int) mode;
2660 	} */ *uap = v;
2661 	struct proc *p = l->l_proc;
2662 	struct file *fp;
2663 	int error;
2664 
2665 	/* getvnode() will use the descriptor for us */
2666 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2667 		return (error);
2668 
2669 	error = change_mode((struct vnode *)fp->f_data, SCARG(uap, mode), l);
2670 	FILE_UNUSE(fp, l);
2671 	return (error);
2672 }
2673 
2674 /*
2675  * Change mode of a file given path name; this version does not follow links.
2676  */
2677 /* ARGSUSED */
2678 int
2679 sys_lchmod(struct lwp *l, void *v, register_t *retval)
2680 {
2681 	struct sys_lchmod_args /* {
2682 		syscallarg(const char *) path;
2683 		syscallarg(int) mode;
2684 	} */ *uap = v;
2685 	int error;
2686 	struct nameidata nd;
2687 
2688 	NDINIT(&nd, LOOKUP, NOFOLLOW | TRYEMULROOT, UIO_USERSPACE,
2689 	    SCARG(uap, path), l);
2690 	if ((error = namei(&nd)) != 0)
2691 		return (error);
2692 
2693 	error = change_mode(nd.ni_vp, SCARG(uap, mode), l);
2694 
2695 	vrele(nd.ni_vp);
2696 	return (error);
2697 }
2698 
2699 /*
2700  * Common routine to set mode given a vnode.
2701  */
2702 static int
2703 change_mode(struct vnode *vp, int mode, struct lwp *l)
2704 {
2705 	struct vattr vattr;
2706 	int error;
2707 
2708 	VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE);
2709 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2710 	VATTR_NULL(&vattr);
2711 	vattr.va_mode = mode & ALLPERMS;
2712 	error = VOP_SETATTR(vp, &vattr, l->l_cred, l);
2713 	VOP_UNLOCK(vp, 0);
2714 	return (error);
2715 }
2716 
2717 /*
2718  * Set ownership given a path name; this version follows links.
2719  */
2720 /* ARGSUSED */
2721 int
2722 sys_chown(struct lwp *l, void *v, register_t *retval)
2723 {
2724 	struct sys_chown_args /* {
2725 		syscallarg(const char *) path;
2726 		syscallarg(uid_t) uid;
2727 		syscallarg(gid_t) gid;
2728 	} */ *uap = v;
2729 	int error;
2730 	struct nameidata nd;
2731 
2732 	NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE,
2733 	    SCARG(uap, path), l);
2734 	if ((error = namei(&nd)) != 0)
2735 		return (error);
2736 
2737 	error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 0);
2738 
2739 	vrele(nd.ni_vp);
2740 	return (error);
2741 }
2742 
2743 /*
2744  * Set ownership given a path name; this version follows links.
2745  * Provides POSIX semantics.
2746  */
2747 /* ARGSUSED */
2748 int
2749 sys___posix_chown(struct lwp *l, void *v, register_t *retval)
2750 {
2751 	struct sys_chown_args /* {
2752 		syscallarg(const char *) path;
2753 		syscallarg(uid_t) uid;
2754 		syscallarg(gid_t) gid;
2755 	} */ *uap = v;
2756 	int error;
2757 	struct nameidata nd;
2758 
2759 	NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE,
2760 	    SCARG(uap, path), l);
2761 	if ((error = namei(&nd)) != 0)
2762 		return (error);
2763 
2764 	error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 1);
2765 
2766 	vrele(nd.ni_vp);
2767 	return (error);
2768 }
2769 
2770 /*
2771  * Set ownership given a file descriptor.
2772  */
2773 /* ARGSUSED */
2774 int
2775 sys_fchown(struct lwp *l, void *v, register_t *retval)
2776 {
2777 	struct sys_fchown_args /* {
2778 		syscallarg(int) fd;
2779 		syscallarg(uid_t) uid;
2780 		syscallarg(gid_t) gid;
2781 	} */ *uap = v;
2782 	struct proc *p = l->l_proc;
2783 	int error;
2784 	struct file *fp;
2785 
2786 	/* getvnode() will use the descriptor for us */
2787 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2788 		return (error);
2789 
2790 	error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid),
2791 	    SCARG(uap, gid), l, 0);
2792 	FILE_UNUSE(fp, l);
2793 	return (error);
2794 }
2795 
2796 /*
2797  * Set ownership given a file descriptor, providing POSIX/XPG semantics.
2798  */
2799 /* ARGSUSED */
2800 int
2801 sys___posix_fchown(struct lwp *l, void *v, register_t *retval)
2802 {
2803 	struct sys_fchown_args /* {
2804 		syscallarg(int) fd;
2805 		syscallarg(uid_t) uid;
2806 		syscallarg(gid_t) gid;
2807 	} */ *uap = v;
2808 	struct proc *p = l->l_proc;
2809 	int error;
2810 	struct file *fp;
2811 
2812 	/* getvnode() will use the descriptor for us */
2813 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2814 		return (error);
2815 
2816 	error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid),
2817 	    SCARG(uap, gid), l, 1);
2818 	FILE_UNUSE(fp, l);
2819 	return (error);
2820 }
2821 
2822 /*
2823  * Set ownership given a path name; this version does not follow links.
2824  */
2825 /* ARGSUSED */
2826 int
2827 sys_lchown(struct lwp *l, void *v, register_t *retval)
2828 {
2829 	struct sys_lchown_args /* {
2830 		syscallarg(const char *) path;
2831 		syscallarg(uid_t) uid;
2832 		syscallarg(gid_t) gid;
2833 	} */ *uap = v;
2834 	int error;
2835 	struct nameidata nd;
2836 
2837 	NDINIT(&nd, LOOKUP, NOFOLLOW | TRYEMULROOT, UIO_USERSPACE,
2838 	    SCARG(uap, path), l);
2839 	if ((error = namei(&nd)) != 0)
2840 		return (error);
2841 
2842 	error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 0);
2843 
2844 	vrele(nd.ni_vp);
2845 	return (error);
2846 }
2847 
2848 /*
2849  * Set ownership given a path name; this version does not follow links.
2850  * Provides POSIX/XPG semantics.
2851  */
2852 /* ARGSUSED */
2853 int
2854 sys___posix_lchown(struct lwp *l, void *v, register_t *retval)
2855 {
2856 	struct sys_lchown_args /* {
2857 		syscallarg(const char *) path;
2858 		syscallarg(uid_t) uid;
2859 		syscallarg(gid_t) gid;
2860 	} */ *uap = v;
2861 	int error;
2862 	struct nameidata nd;
2863 
2864 	NDINIT(&nd, LOOKUP, NOFOLLOW | TRYEMULROOT, UIO_USERSPACE,
2865 	    SCARG(uap, path), l);
2866 	if ((error = namei(&nd)) != 0)
2867 		return (error);
2868 
2869 	error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 1);
2870 
2871 	vrele(nd.ni_vp);
2872 	return (error);
2873 }
2874 
2875 /*
2876  * Common routine to set ownership given a vnode.
2877  */
2878 static int
2879 change_owner(struct vnode *vp, uid_t uid, gid_t gid, struct lwp *l,
2880     int posix_semantics)
2881 {
2882 	struct vattr vattr;
2883 	mode_t newmode;
2884 	int error;
2885 
2886 	VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE);
2887 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2888 	if ((error = VOP_GETATTR(vp, &vattr, l->l_cred, l)) != 0)
2889 		goto out;
2890 
2891 #define CHANGED(x) ((int)(x) != -1)
2892 	newmode = vattr.va_mode;
2893 	if (posix_semantics) {
2894 		/*
2895 		 * POSIX/XPG semantics: if the caller is not the super-user,
2896 		 * clear set-user-id and set-group-id bits.  Both POSIX and
2897 		 * the XPG consider the behaviour for calls by the super-user
2898 		 * implementation-defined; we leave the set-user-id and set-
2899 		 * group-id settings intact in that case.
2900 		 */
2901 		if (kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER,
2902 				      NULL) != 0)
2903 			newmode &= ~(S_ISUID | S_ISGID);
2904 	} else {
2905 		/*
2906 		 * NetBSD semantics: when changing owner and/or group,
2907 		 * clear the respective bit(s).
2908 		 */
2909 		if (CHANGED(uid))
2910 			newmode &= ~S_ISUID;
2911 		if (CHANGED(gid))
2912 			newmode &= ~S_ISGID;
2913 	}
2914 	/* Update va_mode iff altered. */
2915 	if (vattr.va_mode == newmode)
2916 		newmode = VNOVAL;
2917 
2918 	VATTR_NULL(&vattr);
2919 	vattr.va_uid = CHANGED(uid) ? uid : (uid_t)VNOVAL;
2920 	vattr.va_gid = CHANGED(gid) ? gid : (gid_t)VNOVAL;
2921 	vattr.va_mode = newmode;
2922 	error = VOP_SETATTR(vp, &vattr, l->l_cred, l);
2923 #undef CHANGED
2924 
2925 out:
2926 	VOP_UNLOCK(vp, 0);
2927 	return (error);
2928 }
2929 
2930 /*
2931  * Set the access and modification times given a path name; this
2932  * version follows links.
2933  */
2934 /* ARGSUSED */
2935 int
2936 sys_utimes(struct lwp *l, void *v, register_t *retval)
2937 {
2938 	struct sys_utimes_args /* {
2939 		syscallarg(const char *) path;
2940 		syscallarg(const struct timeval *) tptr;
2941 	} */ *uap = v;
2942 
2943 	return do_sys_utimes(l, NULL, SCARG(uap, path), FOLLOW,
2944 			SCARG(uap, tptr), UIO_USERSPACE);
2945 }
2946 
2947 /*
2948  * Set the access and modification times given a file descriptor.
2949  */
2950 /* ARGSUSED */
2951 int
2952 sys_futimes(struct lwp *l, void *v, register_t *retval)
2953 {
2954 	struct sys_futimes_args /* {
2955 		syscallarg(int) fd;
2956 		syscallarg(const struct timeval *) tptr;
2957 	} */ *uap = v;
2958 	int error;
2959 	struct file *fp;
2960 
2961 	/* getvnode() will use the descriptor for us */
2962 	if ((error = getvnode(l->l_proc->p_fd, SCARG(uap, fd), &fp)) != 0)
2963 		return (error);
2964 
2965 	error = do_sys_utimes(l, fp->f_data, NULL, 0,
2966 			SCARG(uap, tptr), UIO_USERSPACE);
2967 
2968 	FILE_UNUSE(fp, l);
2969 	return (error);
2970 }
2971 
2972 /*
2973  * Set the access and modification times given a path name; this
2974  * version does not follow links.
2975  */
2976 int
2977 sys_lutimes(struct lwp *l, void *v, register_t *retval)
2978 {
2979 	struct sys_lutimes_args /* {
2980 		syscallarg(const char *) path;
2981 		syscallarg(const struct timeval *) tptr;
2982 	} */ *uap = v;
2983 
2984 	return do_sys_utimes(l, NULL, SCARG(uap, path), NOFOLLOW,
2985 			SCARG(uap, tptr), UIO_USERSPACE);
2986 }
2987 
2988 /*
2989  * Common routine to set access and modification times given a vnode.
2990  */
2991 int
2992 do_sys_utimes(struct lwp *l, struct vnode *vp, const char *path, int flag,
2993     const struct timeval *tptr, enum uio_seg seg)
2994 {
2995 	struct vattr vattr;
2996 	struct nameidata nd;
2997 	int error;
2998 
2999 	VATTR_NULL(&vattr);
3000 	if (tptr == NULL) {
3001 		nanotime(&vattr.va_atime);
3002 		vattr.va_mtime = vattr.va_atime;
3003 		vattr.va_vaflags |= VA_UTIMES_NULL;
3004 	} else {
3005 		struct timeval tv[2];
3006 
3007 		if (seg != UIO_SYSSPACE) {
3008 			error = copyin(tptr, &tv, sizeof (tv));
3009 			if (error != 0)
3010 				return error;
3011 			tptr = tv;
3012 		}
3013 		TIMEVAL_TO_TIMESPEC(tptr, &vattr.va_atime);
3014 		TIMEVAL_TO_TIMESPEC(tptr + 1, &vattr.va_mtime);
3015 	}
3016 
3017 	if (vp == NULL) {
3018 		NDINIT(&nd, LOOKUP, flag | TRYEMULROOT, UIO_USERSPACE, path, l);
3019 		if ((error = namei(&nd)) != 0)
3020 			return (error);
3021 		vp = nd.ni_vp;
3022 	} else
3023 		nd.ni_vp = NULL;
3024 
3025 	VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE);
3026 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3027 	error = VOP_SETATTR(vp, &vattr, l->l_cred, l);
3028 	VOP_UNLOCK(vp, 0);
3029 
3030 	if (nd.ni_vp != NULL)
3031 		vrele(nd.ni_vp);
3032 
3033 	return (error);
3034 }
3035 
3036 /*
3037  * Truncate a file given its path name.
3038  */
3039 /* ARGSUSED */
3040 int
3041 sys_truncate(struct lwp *l, void *v, register_t *retval)
3042 {
3043 	struct sys_truncate_args /* {
3044 		syscallarg(const char *) path;
3045 		syscallarg(int) pad;
3046 		syscallarg(off_t) length;
3047 	} */ *uap = v;
3048 	struct vnode *vp;
3049 	struct vattr vattr;
3050 	int error;
3051 	struct nameidata nd;
3052 
3053 	NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE,
3054 	    SCARG(uap, path), l);
3055 	if ((error = namei(&nd)) != 0)
3056 		return (error);
3057 	vp = nd.ni_vp;
3058 	VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE);
3059 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3060 	if (vp->v_type == VDIR)
3061 		error = EISDIR;
3062 	else if ((error = vn_writechk(vp)) == 0 &&
3063 	    (error = VOP_ACCESS(vp, VWRITE, l->l_cred, l)) == 0) {
3064 		VATTR_NULL(&vattr);
3065 		vattr.va_size = SCARG(uap, length);
3066 		error = VOP_SETATTR(vp, &vattr, l->l_cred, l);
3067 	}
3068 	vput(vp);
3069 	return (error);
3070 }
3071 
3072 /*
3073  * Truncate a file given a file descriptor.
3074  */
3075 /* ARGSUSED */
3076 int
3077 sys_ftruncate(struct lwp *l, void *v, register_t *retval)
3078 {
3079 	struct sys_ftruncate_args /* {
3080 		syscallarg(int) fd;
3081 		syscallarg(int) pad;
3082 		syscallarg(off_t) length;
3083 	} */ *uap = v;
3084 	struct proc *p = l->l_proc;
3085 	struct vattr vattr;
3086 	struct vnode *vp;
3087 	struct file *fp;
3088 	int error;
3089 
3090 	/* getvnode() will use the descriptor for us */
3091 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
3092 		return (error);
3093 	if ((fp->f_flag & FWRITE) == 0) {
3094 		error = EINVAL;
3095 		goto out;
3096 	}
3097 	vp = (struct vnode *)fp->f_data;
3098 	VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE);
3099 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3100 	if (vp->v_type == VDIR)
3101 		error = EISDIR;
3102 	else if ((error = vn_writechk(vp)) == 0) {
3103 		VATTR_NULL(&vattr);
3104 		vattr.va_size = SCARG(uap, length);
3105 		error = VOP_SETATTR(vp, &vattr, fp->f_cred, l);
3106 	}
3107 	VOP_UNLOCK(vp, 0);
3108  out:
3109 	FILE_UNUSE(fp, l);
3110 	return (error);
3111 }
3112 
3113 /*
3114  * Sync an open file.
3115  */
3116 /* ARGSUSED */
3117 int
3118 sys_fsync(struct lwp *l, void *v, register_t *retval)
3119 {
3120 	struct sys_fsync_args /* {
3121 		syscallarg(int) fd;
3122 	} */ *uap = v;
3123 	struct proc *p = l->l_proc;
3124 	struct vnode *vp;
3125 	struct file *fp;
3126 	int error;
3127 
3128 	/* getvnode() will use the descriptor for us */
3129 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
3130 		return (error);
3131 	vp = (struct vnode *)fp->f_data;
3132 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3133 	error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT, 0, 0, l);
3134 	if (error == 0 && bioopsp != NULL &&
3135 	    vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP))
3136 		(*bioopsp->io_fsync)(vp, 0);
3137 	VOP_UNLOCK(vp, 0);
3138 	FILE_UNUSE(fp, l);
3139 	return (error);
3140 }
3141 
3142 /*
3143  * Sync a range of file data.  API modeled after that found in AIX.
3144  *
3145  * FDATASYNC indicates that we need only save enough metadata to be able
3146  * to re-read the written data.  Note we duplicate AIX's requirement that
3147  * the file be open for writing.
3148  */
3149 /* ARGSUSED */
3150 int
3151 sys_fsync_range(struct lwp *l, void *v, register_t *retval)
3152 {
3153 	struct sys_fsync_range_args /* {
3154 		syscallarg(int) fd;
3155 		syscallarg(int) flags;
3156 		syscallarg(off_t) start;
3157 		syscallarg(off_t) length;
3158 	} */ *uap = v;
3159 	struct proc *p = l->l_proc;
3160 	struct vnode *vp;
3161 	struct file *fp;
3162 	int flags, nflags;
3163 	off_t s, e, len;
3164 	int error;
3165 
3166 	/* getvnode() will use the descriptor for us */
3167 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
3168 		return (error);
3169 
3170 	if ((fp->f_flag & FWRITE) == 0) {
3171 		error = EBADF;
3172 		goto out;
3173 	}
3174 
3175 	flags = SCARG(uap, flags);
3176 	if (((flags & (FDATASYNC | FFILESYNC)) == 0) ||
3177 	    ((~flags & (FDATASYNC | FFILESYNC)) == 0)) {
3178 		error = EINVAL;
3179 		goto out;
3180 	}
3181 	/* Now set up the flags for value(s) to pass to VOP_FSYNC() */
3182 	if (flags & FDATASYNC)
3183 		nflags = FSYNC_DATAONLY | FSYNC_WAIT;
3184 	else
3185 		nflags = FSYNC_WAIT;
3186 	if (flags & FDISKSYNC)
3187 		nflags |= FSYNC_CACHE;
3188 
3189 	len = SCARG(uap, length);
3190 	/* If length == 0, we do the whole file, and s = l = 0 will do that */
3191 	if (len) {
3192 		s = SCARG(uap, start);
3193 		e = s + len;
3194 		if (e < s) {
3195 			error = EINVAL;
3196 			goto out;
3197 		}
3198 	} else {
3199 		e = 0;
3200 		s = 0;
3201 	}
3202 
3203 	vp = (struct vnode *)fp->f_data;
3204 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3205 	error = VOP_FSYNC(vp, fp->f_cred, nflags, s, e, l);
3206 
3207 	if (error == 0 && bioopsp != NULL &&
3208 	    vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP))
3209 		(*bioopsp->io_fsync)(vp, nflags);
3210 
3211 	VOP_UNLOCK(vp, 0);
3212 out:
3213 	FILE_UNUSE(fp, l);
3214 	return (error);
3215 }
3216 
3217 /*
3218  * Sync the data of an open file.
3219  */
3220 /* ARGSUSED */
3221 int
3222 sys_fdatasync(struct lwp *l, void *v, register_t *retval)
3223 {
3224 	struct sys_fdatasync_args /* {
3225 		syscallarg(int) fd;
3226 	} */ *uap = v;
3227 	struct proc *p = l->l_proc;
3228 	struct vnode *vp;
3229 	struct file *fp;
3230 	int error;
3231 
3232 	/* getvnode() will use the descriptor for us */
3233 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
3234 		return (error);
3235 	if ((fp->f_flag & FWRITE) == 0) {
3236 		FILE_UNUSE(fp, l);
3237 		return (EBADF);
3238 	}
3239 	vp = (struct vnode *)fp->f_data;
3240 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3241 	error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT|FSYNC_DATAONLY, 0, 0, l);
3242 	VOP_UNLOCK(vp, 0);
3243 	FILE_UNUSE(fp, l);
3244 	return (error);
3245 }
3246 
3247 /*
3248  * Rename files, (standard) BSD semantics frontend.
3249  */
3250 /* ARGSUSED */
3251 int
3252 sys_rename(struct lwp *l, void *v, register_t *retval)
3253 {
3254 	struct sys_rename_args /* {
3255 		syscallarg(const char *) from;
3256 		syscallarg(const char *) to;
3257 	} */ *uap = v;
3258 
3259 	return (rename_files(SCARG(uap, from), SCARG(uap, to), l, 0));
3260 }
3261 
3262 /*
3263  * Rename files, POSIX semantics frontend.
3264  */
3265 /* ARGSUSED */
3266 int
3267 sys___posix_rename(struct lwp *l, void *v, register_t *retval)
3268 {
3269 	struct sys___posix_rename_args /* {
3270 		syscallarg(const char *) from;
3271 		syscallarg(const char *) to;
3272 	} */ *uap = v;
3273 
3274 	return (rename_files(SCARG(uap, from), SCARG(uap, to), l, 1));
3275 }
3276 
3277 /*
3278  * Rename files.  Source and destination must either both be directories,
3279  * or both not be directories.  If target is a directory, it must be empty.
3280  * If `from' and `to' refer to the same object, the value of the `retain'
3281  * argument is used to determine whether `from' will be
3282  *
3283  * (retain == 0)	deleted unless `from' and `to' refer to the same
3284  *			object in the file system's name space (BSD).
3285  * (retain == 1)	always retained (POSIX).
3286  */
3287 static int
3288 rename_files(const char *from, const char *to, struct lwp *l, int retain)
3289 {
3290 	struct vnode *tvp, *fvp, *tdvp;
3291 	struct nameidata fromnd, tond;
3292 	struct proc *p;
3293 	int error;
3294 
3295 	NDINIT(&fromnd, DELETE, LOCKPARENT | SAVESTART | TRYEMULROOT,
3296 	    UIO_USERSPACE, from, l);
3297 	if ((error = namei(&fromnd)) != 0)
3298 		return (error);
3299 	if (fromnd.ni_dvp != fromnd.ni_vp)
3300 		VOP_UNLOCK(fromnd.ni_dvp, 0);
3301 	fvp = fromnd.ni_vp;
3302 	NDINIT(&tond, RENAME,
3303 	    LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART | TRYEMULROOT
3304 	      | (fvp->v_type == VDIR ? CREATEDIR : 0),
3305 	    UIO_USERSPACE, to, l);
3306 	if ((error = namei(&tond)) != 0) {
3307 		VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
3308 		vrele(fromnd.ni_dvp);
3309 		vrele(fvp);
3310 		goto out1;
3311 	}
3312 	tdvp = tond.ni_dvp;
3313 	tvp = tond.ni_vp;
3314 
3315 	if (tvp != NULL) {
3316 		if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
3317 			error = ENOTDIR;
3318 			goto out;
3319 		} else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
3320 			error = EISDIR;
3321 			goto out;
3322 		}
3323 	}
3324 
3325 	if (fvp == tdvp)
3326 		error = EINVAL;
3327 
3328 	/*
3329 	 * Source and destination refer to the same object.
3330 	 */
3331 	if (fvp == tvp) {
3332 		if (retain)
3333 			error = -1;
3334 		else if (fromnd.ni_dvp == tdvp &&
3335 		    fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen &&
3336 		    !memcmp(fromnd.ni_cnd.cn_nameptr,
3337 		          tond.ni_cnd.cn_nameptr,
3338 		          fromnd.ni_cnd.cn_namelen))
3339 		error = -1;
3340 	}
3341 
3342 #if NVERIEXEC > 0
3343 	if (!error) {
3344 		char *f1, *f2;
3345 
3346 		f1 = malloc(fromnd.ni_cnd.cn_namelen + 1, M_TEMP, M_WAITOK);
3347 		strlcpy(f1, fromnd.ni_cnd.cn_nameptr, fromnd.ni_cnd.cn_namelen);
3348 
3349 		f2 = malloc(tond.ni_cnd.cn_namelen + 1, M_TEMP, M_WAITOK);
3350 		strlcpy(f2, tond.ni_cnd.cn_nameptr, tond.ni_cnd.cn_namelen);
3351 
3352 		error = veriexec_renamechk(l, fvp, f1, tvp, f2);
3353 
3354 		free(f1, M_TEMP);
3355 		free(f2, M_TEMP);
3356 	}
3357 #endif /* NVERIEXEC > 0 */
3358 
3359 out:
3360 	p = l->l_proc;
3361 	if (!error) {
3362 		VOP_LEASE(tdvp, l, l->l_cred, LEASE_WRITE);
3363 		if (fromnd.ni_dvp != tdvp)
3364 			VOP_LEASE(fromnd.ni_dvp, l, l->l_cred, LEASE_WRITE);
3365 		if (tvp) {
3366 			VOP_LEASE(tvp, l, l->l_cred, LEASE_WRITE);
3367 		}
3368 		error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
3369 				   tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
3370 	} else {
3371 		VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd);
3372 		if (tdvp == tvp)
3373 			vrele(tdvp);
3374 		else
3375 			vput(tdvp);
3376 		if (tvp)
3377 			vput(tvp);
3378 		VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
3379 		vrele(fromnd.ni_dvp);
3380 		vrele(fvp);
3381 	}
3382 	vrele(tond.ni_startdir);
3383 	PNBUF_PUT(tond.ni_cnd.cn_pnbuf);
3384 out1:
3385 	if (fromnd.ni_startdir)
3386 		vrele(fromnd.ni_startdir);
3387 	PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf);
3388 	return (error == -1 ? 0 : error);
3389 }
3390 
3391 /*
3392  * Make a directory file.
3393  */
3394 /* ARGSUSED */
3395 int
3396 sys_mkdir(struct lwp *l, void *v, register_t *retval)
3397 {
3398 	struct sys_mkdir_args /* {
3399 		syscallarg(const char *) path;
3400 		syscallarg(int) mode;
3401 	} */ *uap = v;
3402 	struct proc *p = l->l_proc;
3403 	struct vnode *vp;
3404 	struct vattr vattr;
3405 	int error;
3406 	struct nameidata nd;
3407 
3408 	NDINIT(&nd, CREATE, LOCKPARENT | CREATEDIR | TRYEMULROOT, UIO_USERSPACE,
3409 	    SCARG(uap, path), l);
3410 	if ((error = namei(&nd)) != 0)
3411 		return (error);
3412 	vp = nd.ni_vp;
3413 	if (vp != NULL) {
3414 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
3415 		if (nd.ni_dvp == vp)
3416 			vrele(nd.ni_dvp);
3417 		else
3418 			vput(nd.ni_dvp);
3419 		vrele(vp);
3420 		return (EEXIST);
3421 	}
3422 	VATTR_NULL(&vattr);
3423 	vattr.va_type = VDIR;
3424 	/* We will read cwdi->cwdi_cmask unlocked. */
3425 	vattr.va_mode =
3426 	    (SCARG(uap, mode) & ACCESSPERMS) &~ p->p_cwdi->cwdi_cmask;
3427 	VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE);
3428 	error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
3429 	if (!error)
3430 		vput(nd.ni_vp);
3431 	return (error);
3432 }
3433 
3434 /*
3435  * Remove a directory file.
3436  */
3437 /* ARGSUSED */
3438 int
3439 sys_rmdir(struct lwp *l, void *v, register_t *retval)
3440 {
3441 	struct sys_rmdir_args /* {
3442 		syscallarg(const char *) path;
3443 	} */ *uap = v;
3444 	struct vnode *vp;
3445 	int error;
3446 	struct nameidata nd;
3447 
3448 	NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE,
3449 	    SCARG(uap, path), l);
3450 	if ((error = namei(&nd)) != 0)
3451 		return (error);
3452 	vp = nd.ni_vp;
3453 	if (vp->v_type != VDIR) {
3454 		error = ENOTDIR;
3455 		goto out;
3456 	}
3457 	/*
3458 	 * No rmdir "." please.
3459 	 */
3460 	if (nd.ni_dvp == vp) {
3461 		error = EINVAL;
3462 		goto out;
3463 	}
3464 	/*
3465 	 * The root of a mounted filesystem cannot be deleted.
3466 	 */
3467 	if (vp->v_vflag & VV_ROOT) {
3468 		error = EBUSY;
3469 		goto out;
3470 	}
3471 	VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE);
3472 	VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE);
3473 	error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
3474 	return (error);
3475 
3476 out:
3477 	VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
3478 	if (nd.ni_dvp == vp)
3479 		vrele(nd.ni_dvp);
3480 	else
3481 		vput(nd.ni_dvp);
3482 	vput(vp);
3483 	return (error);
3484 }
3485 
3486 /*
3487  * Read a block of directory entries in a file system independent format.
3488  */
3489 int
3490 sys___getdents30(struct lwp *l, void *v, register_t *retval)
3491 {
3492 	struct sys___getdents30_args /* {
3493 		syscallarg(int) fd;
3494 		syscallarg(char *) buf;
3495 		syscallarg(size_t) count;
3496 	} */ *uap = v;
3497 	struct proc *p = l->l_proc;
3498 	struct file *fp;
3499 	int error, done;
3500 
3501 	/* getvnode() will use the descriptor for us */
3502 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
3503 		return (error);
3504 	if ((fp->f_flag & FREAD) == 0) {
3505 		error = EBADF;
3506 		goto out;
3507 	}
3508 	error = vn_readdir(fp, SCARG(uap, buf), UIO_USERSPACE,
3509 			SCARG(uap, count), &done, l, 0, 0);
3510 	ktrgenio(SCARG(uap, fd), UIO_READ, SCARG(uap, buf), done, error);
3511 	*retval = done;
3512  out:
3513 	FILE_UNUSE(fp, l);
3514 	return (error);
3515 }
3516 
3517 /*
3518  * Set the mode mask for creation of filesystem nodes.
3519  */
3520 int
3521 sys_umask(struct lwp *l, void *v, register_t *retval)
3522 {
3523 	struct sys_umask_args /* {
3524 		syscallarg(mode_t) newmask;
3525 	} */ *uap = v;
3526 	struct proc *p = l->l_proc;
3527 	struct cwdinfo *cwdi;
3528 
3529 	/*
3530 	 * cwdi->cwdi_cmask will be read unlocked elsewhere.  What's
3531 	 * important is that we serialize changes to the mask.  The
3532 	 * rw_exit() will issue a write memory barrier on our behalf,
3533 	 * and force the changes out to other CPUs (as it must use an
3534 	 * atomic operation, draining the local CPU's store buffers).
3535 	 */
3536 	cwdi = p->p_cwdi;
3537 	rw_enter(&cwdi->cwdi_lock, RW_WRITER);
3538 	*retval = cwdi->cwdi_cmask;
3539 	cwdi->cwdi_cmask = SCARG(uap, newmask) & ALLPERMS;
3540 	rw_exit(&cwdi->cwdi_lock);
3541 
3542 	return (0);
3543 }
3544 
3545 /*
3546  * Void all references to file by ripping underlying filesystem
3547  * away from vnode.
3548  */
3549 /* ARGSUSED */
3550 int
3551 sys_revoke(struct lwp *l, void *v, register_t *retval)
3552 {
3553 	struct sys_revoke_args /* {
3554 		syscallarg(const char *) path;
3555 	} */ *uap = v;
3556 	struct vnode *vp;
3557 	struct vattr vattr;
3558 	int error;
3559 	bool revoke;
3560 	struct nameidata nd;
3561 
3562 	NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE,
3563 	    SCARG(uap, path), l);
3564 	if ((error = namei(&nd)) != 0)
3565 		return (error);
3566 	vp = nd.ni_vp;
3567 	if ((error = VOP_GETATTR(vp, &vattr, l->l_cred, l)) != 0)
3568 		goto out;
3569 	if (kauth_cred_geteuid(l->l_cred) != vattr.va_uid &&
3570 	    (error = kauth_authorize_generic(l->l_cred,
3571 	    KAUTH_GENERIC_ISSUSER, NULL)) != 0)
3572 		goto out;
3573 	simple_lock(&vp->v_interlock);
3574 	revoke = (vp->v_usecount > 1 || (vp->v_iflag & (VI_ALIASED|VI_LAYER)));
3575 	simple_unlock(&vp->v_interlock);
3576 	if (revoke)
3577 		VOP_REVOKE(vp, REVOKEALL);
3578 out:
3579 	vrele(vp);
3580 	return (error);
3581 }
3582 
3583 /*
3584  * Convert a user file descriptor to a kernel file entry.
3585  */
3586 int
3587 getvnode(struct filedesc *fdp, int fd, struct file **fpp)
3588 {
3589 	struct vnode *vp;
3590 	struct file *fp;
3591 
3592 	if ((fp = fd_getfile(fdp, fd)) == NULL)
3593 		return (EBADF);
3594 
3595 	FILE_USE(fp);
3596 
3597 	if (fp->f_type != DTYPE_VNODE) {
3598 		FILE_UNUSE(fp, NULL);
3599 		return (EINVAL);
3600 	}
3601 
3602 	vp = (struct vnode *)fp->f_data;
3603 	if (vp->v_type == VBAD) {
3604 		FILE_UNUSE(fp, NULL);
3605 		return (EBADF);
3606 	}
3607 
3608 	*fpp = fp;
3609 	return (0);
3610 }
3611