xref: /openbsd-src/regress/usr.bin/diff/t9.2 (revision 2889c8c8b2b0ea172361d6a867e520a23134027d)
1/*	$OpenBSD: t9.2,v 1.4 2024/07/10 09:24:03 krw Exp $	*/
2/*	$NetBSD: vfs_syscalls.c,v 1.71 1996/04/23 10:29:02 mycroft Exp $	*/
3
4/*
5 * Copyright (c) 1989, 1993
6 *	The Regents of the University of California.  All rights reserved.
7 * (c) UNIX System Laboratories, Inc.
8 * All or some portions of this file are derived from material licensed
9 * to the University of California by American Telephone and Telegraph
10 * Co. or Unix System Laboratories, Inc. and are reproduced herein with
11 * the permission of UNIX System Laboratories, Inc.
12 *
13 * Redistribution and use in source and binary forms, with or without
14 * modification, are permitted provided that the following conditions
15 * are met:
16 * 1. Redistributions of source code must retain the above copyright
17 *    notice, this list of conditions and the following disclaimer.
18 * 2. Redistributions in binary form must reproduce the above copyright
19 *    notice, this list of conditions and the following disclaimer in the
20 *    documentation and/or other materials provided with the distribution.
21 * 3. Neither the name of the University nor the names of its contributors
22 *    may be used to endorse or promote products derived from this software
23 *    without specific prior written permission.
24 *
25 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
26 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
27 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
28 * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
29 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
30 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
31 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
32 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
33 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
34 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
35 * SUCH DAMAGE.
36 *
37 *	@(#)vfs_syscalls.c	8.28 (Berkeley) 12/10/94
38 */
39
40#include <sys/param.h>
41#include <sys/systm.h>
42#include <sys/namei.h>
43#include <sys/filedesc.h>
44#include <sys/kernel.h>
45#include <sys/file.h>
46#include <sys/stat.h>
47#include <sys/vnode.h>
48#include <sys/mount.h>
49#include <sys/proc.h>
50#include <sys/uio.h>
51#include <sys/malloc.h>
52#include <sys/dirent.h>
53#include <sys/extattr.h>
54
55#include <sys/syscallargs.h>
56
57#include <uvm/uvm_extern.h>
58#include <sys/sysctl.h>
59
60extern int suid_clear;
61int	usermount = 0;		/* sysctl: by default, users may not mount */
62
63static int change_dir(struct nameidata *, struct proc *);
64
65void checkdirs(struct vnode *);
66
67/*
68 * Redirection info so we don't have to include the union fs routines in
69 * the kernel directly.  This way, we can build unionfs as an LKM.  The
70 * pointer gets filled in later, when we modload the LKM, or when the
71 * compiled-in unionfs code gets initialized.  For now, we just set
72 * it to a stub routine.
73 */
74
75int (*union_check_p)(struct proc *, struct vnode **,
76    struct file *, struct uio, int *) = NULL;
77
78/*
79 * Virtual File System System Calls
80 */
81
82/*
83 * Mount a file system.
84 */
85/* ARGSUSED */
86int
87sys_mount(p, v, retval)
88	struct proc *p;
89	void *v;
90	register_t *retval;
91{
92	register struct sys_mount_args /* {
93		syscallarg(char *) type;
94		syscallarg(char *) path;
95		syscallarg(int) flags;
96		syscallarg(void *) data;
97	} */ *uap = v;
98	register struct vnode *vp;
99	register struct mount *mp;
100	int error, flag = 0;
101#ifdef COMPAT_43
102	u_long fstypenum = 0;
103#endif
104	char fstypename[MFSNAMELEN];
105	char fspath[MNAMELEN];
106	struct vattr va;
107	struct nameidata nd;
108	struct vfsconf *vfsp;
109	struct timeval tv;
110
111	if (usermount == 0 && (error = suser(p->p_ucred, &p->p_acflag)))
112		return (error);
113
114	/*
115	 * Mount points must fit in MNAMELEN, not MAXPATHLEN.
116	 */
117	error = copyinstr(SCARG(uap, path), fspath, MNAMELEN, NULL);
118	if (error)
119		return(error);
120
121	/*
122	 * Get vnode to be covered
123	 */
124	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE, fspath, p);
125	if ((error = namei(&nd)) != 0)
126		return (error);
127	vp = nd.ni_vp;
128	if (SCARG(uap, flags) & MNT_UPDATE) {
129		if ((vp->v_flag & VROOT) == 0) {
130			vput(vp);
131			return (EINVAL);
132		}
133		mp = vp->v_mount;
134		flag = mp->mnt_flag;
135		/*
136		 * We only allow the filesystem to be reloaded if it
137		 * is currently mounted read-only.
138		 */
139		if ((SCARG(uap, flags) & MNT_RELOAD) &&
140		    ((mp->mnt_flag & MNT_RDONLY) == 0)) {
141			vput(vp);
142			return (EOPNOTSUPP);	/* Needs translation */
143		}
144		mp->mnt_flag |=
145		    SCARG(uap, flags) & (MNT_RELOAD | MNT_UPDATE);
146		/*
147		 * Only root, or the user that did the original mount is
148		 * permitted to update it.
149		 */
150		if (mp->mnt_stat.f_owner != p->p_ucred->cr_uid &&
151		    (error = suser(p->p_ucred, &p->p_acflag))) {
152			vput(vp);
153			return (error);
154		}
155		/*
156		 * Do not allow NFS export by non-root users. Silently
157		 * enforce MNT_NOSUID and MNT_NODEV for non-root users.
158		 */
159		if (p->p_ucred->cr_uid != 0) {
160			if (SCARG(uap, flags) & MNT_EXPORTED) {
161				vput(vp);
162				return (EPERM);
163			}
164			SCARG(uap, flags) |= MNT_NOSUID | MNT_NODEV;
165		}
166		if ((error = vfs_busy(mp, LK_NOWAIT, 0, p)) != 0) {
167			vput(vp);
168			return (error);
169		}
170		VOP_UNLOCK(vp, 0, p);
171		goto update;
172	}
173	/*
174	 * If the user is not root, ensure that they own the directory
175	 * onto which we are attempting to mount.
176	 */
177	if ((error = VOP_GETATTR(vp, &va, p->p_ucred, p)) ||
178	    (va.va_uid != p->p_ucred->cr_uid &&
179	    (error = suser(p->p_ucred, &p->p_acflag)))) {
180		vput(vp);
181		return (error);
182	}
183	/*
184	 * Do not allow NFS export by non-root users. Silently
185	 * enforce MNT_NOSUID and MNT_NODEV for non-root users.
186	 */
187	if (p->p_ucred->cr_uid != 0) {
188		if (SCARG(uap, flags) & MNT_EXPORTED) {
189			vput(vp);
190			return (EPERM);
191		}
192		SCARG(uap, flags) |= MNT_NOSUID | MNT_NODEV;
193	}
194	if ((error = vinvalbuf(vp, V_SAVE, p->p_ucred, p, 0, 0)) != 0)
195		return (error);
196	if (vp->v_type != VDIR) {
197		vput(vp);
198		return (ENOTDIR);
199	}
200	error = copyinstr(SCARG(uap, type), fstypename, MFSNAMELEN, NULL);
201	if (error) {
202#ifdef COMPAT_43
203		/*
204		 * Historically filesystem types were identified by number.
205		 * If we get an integer for the filesystem type instead of a
206		 * string, we check to see if it matches one of the historic
207		 * filesystem types.
208		 */
209		fstypenum = (u_long)SCARG(uap, type);
210
211		for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next)
212			if (vfsp->vfc_typenum == fstypenum)
213				break;
214		if (vfsp == NULL) {
215			vput(vp);
216			return (ENODEV);
217		}
218		strncpy(fstypename, vfsp->vfc_name, MFSNAMELEN);
219
220#else
221		vput(vp);
222		return (error);
223#endif
224	}
225	for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next) {
226		if (!strcmp(vfsp->vfc_name, fstypename))
227			break;
228	}
229
230	if (vfsp == NULL) {
231		vput(vp);
232		return (EOPNOTSUPP);
233	}
234
235	if (vp->v_mountedhere != NULL) {
236		vput(vp);
237		return (EBUSY);
238	}
239
240	/*
241	 * Allocate and initialize the file system.
242	 */
243	mp = (struct mount *)malloc((u_long)sizeof(struct mount),
244		M_MOUNT, M_WAITOK);
245	bzero((char *)mp, (u_long)sizeof(struct mount));
246	lockinit(&mp->mnt_lock, PVFS, "vfslock", 0, 0);
247	/* This error never happens, but it makes auditing easier */
248	if ((error = vfs_busy(mp, LK_NOWAIT, 0, p)))
249		return (error);
250	mp->mnt_op = vfsp->vfc_vfsops;
251	mp->mnt_vfc = vfsp;
252	mp->mnt_flag |= (vfsp->vfc_flags & MNT_VISFLAGMASK);
253	strncpy(mp->mnt_stat.f_fstypename, vfsp->vfc_name, MFSNAMELEN);
254	mp->mnt_vnodecovered = vp;
255	mp->mnt_stat.f_owner = p->p_ucred->cr_uid;
256update:
257	/*
258	 * Set the mount level flags.
259	 */
260	if (SCARG(uap, flags) & MNT_RDONLY)
261		mp->mnt_flag |= MNT_RDONLY;
262	else if (mp->mnt_flag & MNT_RDONLY)
263		mp->mnt_flag |= MNT_WANTRDWR;
264	mp->mnt_flag &=~ (MNT_NOSUID | MNT_NOEXEC | MNT_NODEV |
265	    MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_SOFTDEP |
266	    MNT_NOATIME | MNT_FORCE);
267	mp->mnt_flag |= SCARG(uap, flags) & (MNT_NOSUID | MNT_NOEXEC |
268	    MNT_NODEV | MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC |
269	    MNT_SOFTDEP | MNT_NOATIME | MNT_FORCE);
270	/*
271	 * Mount the filesystem.
272	 */
273	error = VFS_MOUNT(mp, SCARG(uap, path), SCARG(uap, data), &nd, p);
274	if (!error) {
275		microtime(&tv);
276		mp->mnt_stat.f_ctime = tv.tv_sec;
277	}
278	if (mp->mnt_flag & MNT_UPDATE) {
279		vrele(vp);
280		if (mp->mnt_flag & MNT_WANTRDWR)
281			mp->mnt_flag &= ~MNT_RDONLY;
282		mp->mnt_flag &=~
283		    (MNT_UPDATE | MNT_RELOAD | MNT_FORCE | MNT_WANTRDWR);
284		if (error)
285			mp->mnt_flag = flag;
286
287 		if ((mp->mnt_flag & MNT_RDONLY) == 0) {
288 			if (mp->mnt_syncer == NULL)
289 				error = vfs_allocate_syncvnode(mp);
290 		} else {
291 			if (mp->mnt_syncer != NULL)
292 				vgone(mp->mnt_syncer);
293 			mp->mnt_syncer = NULL;
294 		}
295
296		vfs_unbusy(mp, p);
297		return (error);
298	}
299
300	vp->v_mountedhere = mp;
301
302	/*
303	 * Put the new filesystem on the mount list after root.
304	 */
305	cache_purge(vp);
306	if (!error) {
307		vfsp->vfc_refcount++;
308		simple_lock(&mountlist_slock);
309		TAILQ_INSERT_TAIL(&mountlist, mp, mnt_list);
310		simple_unlock(&mountlist_slock);
311		checkdirs(vp);
312		VOP_UNLOCK(vp, 0, p);
313 		if ((mp->mnt_flag & MNT_RDONLY) == 0)
314 			error = vfs_allocate_syncvnode(mp);
315		vfs_unbusy(mp, p);
316		(void) VFS_STATFS(mp, &mp->mnt_stat, p);
317		if ((error = VFS_START(mp, 0, p)) != 0)
318			vrele(vp);
319	} else {
320		mp->mnt_vnodecovered->v_mountedhere = (struct mount *)0;
321		vfs_unbusy(mp, p);
322		free((caddr_t)mp, M_MOUNT);
323		vput(vp);
324	}
325	return (error);
326}
327
328/*
329 * Scan all active processes to see if any of them have a current
330 * or root directory onto which the new filesystem has just been
331 * mounted. If so, replace them with the new mount point.
332 */
333void
334checkdirs(olddp)
335	struct vnode *olddp;
336{
337	struct filedesc *fdp;
338	struct vnode *newdp;
339	struct proc *p;
340
341	if (olddp->v_usecount == 1)
342		return;
343	if (VFS_ROOT(olddp->v_mountedhere, &newdp))
344		panic("mount: lost mount");
345	for (p = LIST_FIRST(&allproc); p != 0; p = LIST_NEXT(p, p_list)) {
346		fdp = p->p_fd;
347		if (fdp->fd_cdir == olddp) {
348			vrele(fdp->fd_cdir);
349			VREF(newdp);
350			fdp->fd_cdir = newdp;
351		}
352		if (fdp->fd_rdir == olddp) {
353			vrele(fdp->fd_rdir);
354			VREF(newdp);
355			fdp->fd_rdir = newdp;
356		}
357	}
358	if (rootvnode == olddp) {
359		vrele(rootvnode);
360		VREF(newdp);
361		rootvnode = newdp;
362	}
363	vput(newdp);
364}
365
366/*
367 * Unmount a file system.
368 *
369 * Note: unmount takes a path to the vnode mounted on as argument,
370 * not special file (as before).
371 */
372/* ARGSUSED */
373int
374sys_unmount(p, v, retval)
375	struct proc *p;
376	void *v;
377	register_t *retval;
378{
379	register struct sys_unmount_args /* {
380		syscallarg(char *) path;
381		syscallarg(int) flags;
382	} */ *uap = v;
383	register struct vnode *vp;
384	struct mount *mp;
385	int error;
386	struct nameidata nd;
387
388	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
389	    SCARG(uap, path), p);
390	if ((error = namei(&nd)) != 0)
391		return (error);
392	vp = nd.ni_vp;
393	mp = vp->v_mount;
394
395	/*
396	 * Only root, or the user that did the original mount is
397	 * permitted to unmount this filesystem.
398	 */
399	if ((mp->mnt_stat.f_owner != p->p_ucred->cr_uid) &&
400	    (error = suser(p->p_ucred, &p->p_acflag))) {
401		vput(vp);
402		return (error);
403	}
404
405	/*
406	 * Don't allow unmounting the root file system.
407	 */
408	if (mp->mnt_flag & MNT_ROOTFS) {
409		vput(vp);
410		return (EINVAL);
411	}
412
413	/*
414	 * Must be the root of the filesystem
415	 */
416	if ((vp->v_flag & VROOT) == 0) {
417		vput(vp);
418		return (EINVAL);
419	}
420	vput(vp);
421
422	if (vfs_busy(mp, LK_EXCLUSIVE, NULL, p))
423		return (EBUSY);
424
425	return (dounmount(mp, SCARG(uap, flags), p, vp));
426}
427
428/*
429 * Do the actual file system unmount.
430 */
431int
432dounmount(struct mount *mp, int flags, struct proc *p, struct vnode *olddp)
433{
434	struct vnode *coveredvp;
435	struct proc *p2;
436	int error;
437	int hadsyncer = 0;
438
439 	mp->mnt_flag &=~ MNT_ASYNC;
440 	cache_purgevfs(mp);	/* remove cache entries for this file sys */
441 	if (mp->mnt_syncer != NULL) {
442		hadsyncer = 1;
443 		vgone(mp->mnt_syncer);
444		mp->mnt_syncer = NULL;
445	}
446	if (((mp->mnt_flag & MNT_RDONLY) ||
447	    (error = VFS_SYNC(mp, MNT_WAIT, p->p_ucred, p)) == 0) ||
448 	    (flags & MNT_FORCE))
449 		error = VFS_UNMOUNT(mp, flags, p);
450	simple_lock(&mountlist_slock);
451 	if (error) {
452 		if ((mp->mnt_flag & MNT_RDONLY) == 0 && hadsyncer)
453 			(void) vfs_allocate_syncvnode(mp);
454		lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK,
455		    &mountlist_slock, p);
456		return (error);
457	}
458	TAILQ_REMOVE(&mountlist, mp, mnt_list);
459	if ((coveredvp = mp->mnt_vnodecovered) != NULLVP) {
460		if (olddp) {
461			/*
462			 * Try to put processes back in a real directory
463			 * after a forced unmount.
464			 * XXX We're not holding a ref on olddp, which may
465			 * change, so compare id numbers.
466			 */
467			LIST_FOREACH(p2, &allproc, p_list) {
468				struct filedesc *fdp = p2->p_fd;
469				if (fdp->fd_cdir &&
470				    fdp->fd_cdir->v_id == olddp->v_id) {
471					vrele(fdp->fd_cdir);
472					vref(coveredvp);
473					fdp->fd_cdir = coveredvp;
474				}
475				if (fdp->fd_rdir &&
476				    fdp->fd_rdir->v_id == olddp->v_id) {
477					vrele(fdp->fd_rdir);
478					vref(coveredvp);
479					fdp->fd_rdir = coveredvp;
480				}
481			}
482		}
483		coveredvp->v_mountedhere = NULL;
484 		vrele(coveredvp);
485 	}
486	mp->mnt_vfc->vfc_refcount--;
487	if (mp->mnt_vnodelist.lh_first != NULL)
488		panic("unmount: dangling vnode");
489	lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK, &mountlist_slock, p);
490	free((caddr_t)mp, M_MOUNT);
491	return (0);
492}
493
494/*
495 * Sync each mounted filesystem.
496 */
497#ifdef DEBUG
498int syncprt = 0;
499struct ctldebug debug0 = { "syncprt", &syncprt };
500#endif
501
502/* ARGSUSED */
503int
504sys_sync(p, v, retval)
505	struct proc *p;
506	void *v;
507	register_t *retval;
508{
509	register struct mount *mp, *nmp;
510	int asyncflag;
511
512	simple_lock(&mountlist_slock);
513	TAILQ_FOREACH_REVERSE_SAFE(mp, &mountlist, mnt_list, nmp) {
514		if (vfs_busy(mp, LK_NOWAIT, &mountlist_slock, p))
515			continue;
516		if ((mp->mnt_flag & MNT_RDONLY) == 0) {
517			asyncflag = mp->mnt_flag & MNT_ASYNC;
518			mp->mnt_flag &= ~MNT_ASYNC;
519			uvm_vnp_sync(mp);
520			VFS_SYNC(mp, MNT_NOWAIT, p->p_ucred, p);
521			if (asyncflag)
522				mp->mnt_flag |= MNT_ASYNC;
523		}
524		simple_lock(&mountlist_slock);
525		vfs_unbusy(mp, p);
526	}
527	simple_unlock(&mountlist_slock);
528
529#ifdef DEBUG
530	if (syncprt)
531		vfs_bufstats();
532#endif /* DEBUG */
533	return (0);
534}
535
536/*
537 * Change filesystem quotas.
538 */
539/* ARGSUSED */
540int
541sys_quotactl(p, v, retval)
542	struct proc *p;
543	void *v;
544	register_t *retval;
545{
546	register struct sys_quotactl_args /* {
547		syscallarg(char *) path;
548		syscallarg(int) cmd;
549		syscallarg(int) uid;
550		syscallarg(caddr_t) arg;
551	} */ *uap = v;
552	register struct mount *mp;
553	int error;
554	struct nameidata nd;
555
556	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
557	if ((error = namei(&nd)) != 0)
558		return (error);
559	mp = nd.ni_vp->v_mount;
560	vrele(nd.ni_vp);
561	return (VFS_QUOTACTL(mp, SCARG(uap, cmd), SCARG(uap, uid),
562	    SCARG(uap, arg), p));
563}
564
565/*
566 * Get filesystem statistics.
567 */
568/* ARGSUSED */
569int
570sys_statfs(p, v, retval)
571	struct proc *p;
572	void *v;
573	register_t *retval;
574{
575	register struct sys_statfs_args /* {
576		syscallarg(char *) path;
577		syscallarg(struct statfs *) buf;
578	} */ *uap = v;
579	register struct mount *mp;
580	register struct statfs *sp;
581	int error;
582	struct nameidata nd;
583	struct statfs sb;
584
585	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
586	if ((error = namei(&nd)) != 0)
587		return (error);
588	mp = nd.ni_vp->v_mount;
589	sp = &mp->mnt_stat;
590	vrele(nd.ni_vp);
591	if ((error = VFS_STATFS(mp, sp, p)) != 0)
592		return (error);
593	sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
594	/* Don't let non-root see filesystem id (for NFS security) */
595	if (suser(p->p_ucred, &p->p_acflag)) {
596		bcopy((caddr_t)sp, (caddr_t)&sb, sizeof(sb));
597		sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
598		sp = &sb;
599	}
600	return (copyout((caddr_t)sp, (caddr_t)SCARG(uap, buf), sizeof(*sp)));
601}
602
603/*
604 * Get filesystem statistics.
605 */
606/* ARGSUSED */
607int
608sys_fstatfs(p, v, retval)
609	struct proc *p;
610	void *v;
611	register_t *retval;
612{
613	struct sys_fstatfs_args /* {
614		syscallarg(int) fd;
615		syscallarg(struct statfs *) buf;
616	} */ *uap = v;
617	struct file *fp;
618	struct mount *mp;
619	struct statfs *sp;
620	int error;
621	struct statfs sb;
622
623	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
624		return (error);
625	mp = ((struct vnode *)fp->f_data)->v_mount;
626	sp = &mp->mnt_stat;
627	error = VFS_STATFS(mp, sp, p);
628	FRELE(fp);
629	if (error)
630		return (error);
631	sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
632	/* Don't let non-root see filesystem id (for NFS security) */
633	if (suser(p->p_ucred, &p->p_acflag)) {
634		bcopy((caddr_t)sp, (caddr_t)&sb, sizeof(sb));
635		sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
636		sp = &sb;
637	}
638	return (copyout((caddr_t)sp, (caddr_t)SCARG(uap, buf), sizeof(*sp)));
639}
640
641/*
642 * Get statistics on all filesystems.
643 */
644int
645sys_getfsstat(p, v, retval)
646	struct proc *p;
647	void *v;
648	register_t *retval;
649{
650	register struct sys_getfsstat_args /* {
651		syscallarg(struct statfs *) buf;
652		syscallarg(size_t) bufsize;
653		syscallarg(int) flags;
654	} */ *uap = v;
655	register struct mount *mp *nmp;
656	register struct statfs *sp;
657	struct statfs sb;
658	caddr_t sfsp;
659	size_t count, maxcount;
660	int error, flags = SCARG(uap, flags);
661
662	maxcount = SCARG(uap, bufsize) / sizeof(struct statfs);
663	sfsp = (caddr_t)SCARG(uap, buf);
664	count = 0;
665	simple_lock(&mountlist_slock);
666	TAILQ_FOREACH_REVERSE_SAFE(mp, &mountlist, mnt_list, nmp) {
667		if (vfs_busy(mp, LK_NOWAIT, &mountlist_slock, p))
668			continue;
669		if (sfsp && count < maxcount) {
670			sp = &mp->mnt_stat;
671
672			/* Refresh stats unless MNT_NOWAIT is specified */
673			if (flags != MNT_NOWAIT &&
674			    flags != MNT_LAZY &&
675			    (flags == MNT_WAIT ||
676			    flags == 0) &&
677			    (error = VFS_STATFS(mp, sp, p))) {
678				simple_lock(&mountlist_slock);
679				vfs_unbusy(mp, p);
680 				continue;
681			}
682
683			sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
684			if (suser(p->p_ucred, &p->p_acflag)) {
685				bcopy((caddr_t)sp, (caddr_t)&sb, sizeof(sb));
686				sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
687				sp = &sb;
688			}
689			error = copyout((caddr_t)sp, sfsp, sizeof(*sp));
690			if (error) {
691				vfs_unbusy(mp, p);
692				return (error);
693			}
694			sfsp += sizeof(*sp);
695		}
696		count++;
697		simple_lock(&mountlist_slock);
698		vfs_unbusy(mp, p);
699	}
700	simple_unlock(&mountlist_slock);
701	if (sfsp && count > maxcount)
702		*retval = maxcount;
703	else
704		*retval = count;
705	return (0);
706}
707
708/*
709 * Change current working directory to a given file descriptor.
710 */
711/* ARGSUSED */
712int
713sys_fchdir(p, v, retval)
714	struct proc *p;
715	void *v;
716	register_t *retval;
717{
718	struct sys_fchdir_args /* {
719		syscallarg(int) fd;
720	} */ *uap = v;
721	struct filedesc *fdp = p->p_fd;
722	struct vnode *vp, *tdp;
723	struct mount *mp;
724	struct file *fp;
725	int error;
726
727	if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0)
728		return (error);
729	vp = (struct vnode *)fp->f_data;
730	VREF(vp);
731	FRELE(fp);
732	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, p);
733	if (vp->v_type != VDIR)
734		error = ENOTDIR;
735	else
736		error = VOP_ACCESS(vp, VEXEC, p->p_ucred, p);
737
738	while (!error && (mp = vp->v_mountedhere) != NULL) {
739		if (vfs_busy(mp, 0, 0, p))
740			continue;
741		error = VFS_ROOT(mp, &tdp);
742		vfs_unbusy(mp, p);
743		if (error)
744			break;
745		vput(vp);
746		vp = tdp;
747	}
748	if (error) {
749		vput(vp);
750		return (error);
751	}
752	VOP_UNLOCK(vp, 0, p);
753	vrele(fdp->fd_cdir);
754	fdp->fd_cdir = vp;
755	return (0);
756}
757
758/*
759 * Change current working directory (``.'').
760 */
761/* ARGSUSED */
762int
763sys_chdir(p, v, retval)
764	struct proc *p;
765	void *v;
766	register_t *retval;
767{
768	struct sys_chdir_args /* {
769		syscallarg(char *) path;
770	} */ *uap = v;
771	register struct filedesc *fdp = p->p_fd;
772	int error;
773	struct nameidata nd;
774
775	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
776	    SCARG(uap, path), p);
777	if ((error = change_dir(&nd, p)) != 0)
778		return (error);
779	vrele(fdp->fd_cdir);
780	fdp->fd_cdir = nd.ni_vp;
781	return (0);
782}
783
784/*
785 * Change notion of root (``/'') directory.
786 */
787/* ARGSUSED */
788int
789sys_chroot(p, v, retval)
790	struct proc *p;
791	void *v;
792	register_t *retval;
793{
794	struct sys_chroot_args /* {
795		syscallarg(char *) path;
796	} */ *uap = v;
797	register struct filedesc *fdp = p->p_fd;
798	int error;
799	struct nameidata nd;
800
801	if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
802		return (error);
803	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
804	    SCARG(uap, path), p);
805	if ((error = change_dir(&nd, p)) != 0)
806		return (error);
807	if (fdp->fd_rdir != NULL) {
808		/*
809		 * A chroot() done inside a changed root environment does
810		 * an automatic chdir to avoid the out-of-tree experience.
811		 */
812		vrele(fdp->fd_rdir);
813		vrele(fdp->fd_cdir);
814		VREF(nd.ni_vp);
815		fdp->fd_cdir = nd.ni_vp;
816	}
817	fdp->fd_rdir = nd.ni_vp;
818	return (0);
819}
820
821/*
822 * Common routine for chroot and chdir.
823 */
824static int
825change_dir(ndp, p)
826	register struct nameidata *ndp;
827	struct proc *p;
828{
829	struct vnode *vp;
830	int error;
831
832	if ((error = namei(ndp)) != 0)
833		return (error);
834	vp = ndp->ni_vp;
835	if (vp->v_type != VDIR)
836		error = ENOTDIR;
837	else
838		error = VOP_ACCESS(vp, VEXEC, p->p_ucred, p);
839	if (error)
840		vput(vp);
841	else
842		VOP_UNLOCK(vp, 0, p);
843	return (error);
844}
845
846/*
847 * Check permissions, allocate an open file structure,
848 * and call the device open routine if any.
849 */
850int
851sys_open(p, v, retval)
852	struct proc *p;
853	void *v;
854	register_t *retval;
855{
856	struct sys_open_args /* {
857		syscallarg(char *) path;
858		syscallarg(int) flags;
859		syscallarg(int) mode;
860	} */ *uap = v;
861	struct filedesc *fdp = p->p_fd;
862	struct file *fp;
863	struct vnode *vp;
864	struct vattr vattr;
865	int flags, cmode;
866	int type, indx, error, localtrunc = 0;
867	struct flock lf;
868	struct nameidata nd;
869
870	if ((error = falloc(p, &fp, &indx)) != 0)
871		return (error);
872
873	flags = FFLAGS(SCARG(uap, flags));
874	cmode = ((SCARG(uap, mode) &~ fdp->fd_cmask) & ALLPERMS) &~ S_ISTXT;
875	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
876	p->p_dupfd = -indx - 1;			/* XXX check for fdopen */
877	if ((flags & O_TRUNC) && (flags & (O_EXLOCK | O_SHLOCK))) {
878		localtrunc = 1;
879		flags &= ~O_TRUNC;	/* Must do truncate ourselves */
880	}
881	if ((error = vn_open(&nd, flags, cmode)) != 0) {
882		if ((error == ENODEV || error == ENXIO) &&
883		    p->p_dupfd >= 0 &&			/* XXX from fdopen */
884		    (error =
885			dupfdopen(fdp, indx, p->p_dupfd, flags, error)) == 0) {
886			closef(fp, p);
887			*retval = indx;
888			return (0);
889		}
890		if (error == ERESTART)
891			error = EINTR;
892		fdremove(fdp, indx);
893		closef(fp, p);
894		return (error);
895	}
896	p->p_dupfd = 0;
897	vp = nd.ni_vp;
898	fp->f_flag = flags & FMASK;
899	fp->f_type = DTYPE_VNODE;
900	fp->f_ops = &vnops;
901	fp->f_data = (caddr_t)vp;
902	if (flags & (O_EXLOCK | O_SHLOCK)) {
903		lf.l_whence = SEEK_SET;
904		lf.l_start = 0;
905		lf.l_len = 0;
906		if (flags & O_EXLOCK)
907			lf.l_type = F_WRLCK;
908		else
909			lf.l_type = F_RDLCK;
910		type = F_FLOCK;
911		if ((flags & FNONBLOCK) == 0)
912			type |= F_WAIT;
913		VOP_UNLOCK(vp, 0, p);
914		error = VOP_ADVLOCK(vp, (caddr_t)fp, F_SETLK, &lf, type);
915		if (error) {
916			/* closef will vn_close the file for us. */
917			fdremove(fdp, indx);
918			closef(fp, p);
919			return (error);
920		}
921		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, p);
922		fp->f_flag |= FHASLOCK;
923	}
924	if (localtrunc) {
925		VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
926		if ((fp->f_flag & FWRITE) == 0)
927			error = EACCES;
928		else if (vp->v_mount->mnt_flag & MNT_RDONLY)
929			error = EROFS;
930		else if (vp->v_type == VDIR)
931			error = EISDIR;
932		else if ((error = vn_writechk(vp)) == 0) {
933			VATTR_NULL(&vattr);
934			vattr.va_size = 0;
935			error = VOP_SETATTR(vp, &vattr, fp->f_cred, p);
936		}
937		if (error) {
938			VOP_UNLOCK(vp, 0, p);
939			/* closef will close the file for us. */
940			fdremove(fdp, indx);
941			closef(fp, p);
942			return (error);
943		}
944	}
945	VOP_UNLOCK(vp, 0, p);
946	*retval = indx;
947	FILE_SET_MATURE(fp);
948	return (0);
949}
950
951/*
952 * Get file handle system call
953 */
954int
955sys_getfh(p, v, retval)
956	struct proc *p;
957	register void *v;
958	register_t *retval;
959{
960	register struct sys_getfh_args /* {
961		syscallarg(char *) fname;
962		syscallarg(fhandle_t *) fhp;
963	} */ *uap = v;
964	register struct vnode *vp;
965	fhandle_t fh;
966	int error;
967	struct nameidata nd;
968
969	/*
970	 * Must be super user
971	 */
972	error = suser(p->p_ucred, &p->p_acflag);
973	if (error)
974		return (error);
975	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
976	    SCARG(uap, fname), p);
977	error = namei(&nd);
978	if (error)
979		return (error);
980	vp = nd.ni_vp;
981	bzero((caddr_t)&fh, sizeof(fh));
982	fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
983	error = VFS_VPTOFH(vp, &fh.fh_fid);
984	vput(vp);
985	if (error)
986		return (error);
987	error = copyout((caddr_t)&fh, (caddr_t)SCARG(uap, fhp), sizeof (fh));
988	return (error);
989}
990
991/*
992 * Open a file given a file handle.
993 *
994 * Check permissions, allocate an open file structure,
995 * and call the device open routine if any.
996 */
997int
998sys_fhopen(p, v, retval)
999	struct proc *p;
1000	void *v;
1001	register_t *retval;
1002{
1003	register struct sys_fhopen_args /* {
1004		syscallarg(const fhandle_t *) fhp;
1005		syscallarg(int) flags;
1006	} */ *uap = v;
1007	struct filedesc *fdp = p->p_fd;
1008	struct file *fp;
1009	struct vnode *vp = NULL;
1010	struct mount *mp;
1011	struct ucred *cred = p->p_ucred;
1012	int flags;
1013	int type, indx, error=0;
1014	struct flock lf;
1015	struct vattr va;
1016	fhandle_t fh;
1017
1018	/*
1019	 * Must be super user
1020	 */
1021	if ((error = suser(p->p_ucred, &p->p_acflag)))
1022		return (error);
1023
1024	flags = FFLAGS(SCARG(uap, flags));
1025	if ((flags & (FREAD | FWRITE)) == 0)
1026		return (EINVAL);
1027	if ((flags & O_CREAT))
1028		return (EINVAL);
1029
1030	if ((error = falloc(p, &fp, &indx)) != 0)
1031		return (error);
1032
1033	if ((error = copyin(SCARG(uap, fhp), &fh, sizeof(fhandle_t))) != 0)
1034		goto bad;
1035
1036	if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL) {
1037		error = ESTALE;
1038		goto bad;
1039	}
1040
1041	if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp)) != 0) {
1042		vp = NULL;	/* most likely unnecessary sanity for bad: */
1043		goto bad;
1044	}
1045
1046	/* Now do an effective vn_open */
1047
1048	if (vp->v_type == VSOCK) {
1049		error = EOPNOTSUPP;
1050		goto bad;
1051	}
1052	if (flags & FREAD) {
1053		if ((error = VOP_ACCESS(vp, VREAD, cred, p)) != 0)
1054			goto bad;
1055	}
1056	if (flags & (FWRITE | O_TRUNC)) {
1057		if (vp->v_type == VDIR) {
1058			error = EISDIR;
1059			goto bad;
1060		}
1061		if ((error = vn_writechk(vp)) != 0 ||
1062		    (error = VOP_ACCESS(vp, VWRITE, cred, p)) != 0)
1063			goto bad;
1064	}
1065	if (flags & O_TRUNC) {
1066		VOP_UNLOCK(vp, 0, p);				/* XXX */
1067		VOP_LEASE(vp, p, cred, LEASE_WRITE);
1068		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, p);	/* XXX */
1069		VATTR_NULL(&va);
1070		va.va_size = 0;
1071		if ((error = VOP_SETATTR(vp, &va, cred, p)) != 0)
1072			goto bad;
1073	}
1074	if ((error = VOP_OPEN(vp, flags, cred, p)) != 0)
1075		goto bad;
1076	if (flags & FWRITE)
1077		vp->v_writecount++;
1078
1079	/* done with modified vn_open, now finish what sys_open does. */
1080
1081	fp->f_flag = flags & FMASK;
1082	fp->f_type = DTYPE_VNODE;
1083	fp->f_ops = &vnops;
1084	fp->f_data = (caddr_t)vp;
1085	if (flags & (O_EXLOCK | O_SHLOCK)) {
1086		lf.l_whence = SEEK_SET;
1087		lf.l_start = 0;
1088		lf.l_len = 0;
1089		if (flags & O_EXLOCK)
1090			lf.l_type = F_WRLCK;
1091		else
1092			lf.l_type = F_RDLCK;
1093		type = F_FLOCK;
1094		if ((flags & FNONBLOCK) == 0)
1095			type |= F_WAIT;
1096		VOP_UNLOCK(vp, 0, p);
1097		error = VOP_ADVLOCK(vp, (caddr_t)fp, F_SETLK, &lf, type);
1098		if (error) {
1099			/* closef will vn_close the file for us. */
1100			fdremove(fdp, indx);
1101			closef(fp, p);
1102			return (error);
1103		}
1104		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, p);
1105		fp->f_flag |= FHASLOCK;
1106	}
1107	VOP_UNLOCK(vp, 0, p);
1108	*retval = indx;
1109	FILE_SET_MATURE(fp);
1110	return (0);
1111
1112bad:
1113	fdremove(fdp, indx);
1114	closef(fp, p);
1115	if (vp != NULL)
1116		vput(vp);
1117	return (error);
1118}
1119
1120/* ARGSUSED */
1121int
1122sys_fhstat(p, v, retval)
1123	struct proc *p;
1124	void *v;
1125	register_t *retval;
1126{
1127	register struct sys_fhstat_args /* {
1128		syscallarg(const fhandle_t *) fhp;
1129		syscallarg(struct stat *) sb;
1130	} */ *uap = v;
1131	struct stat sb;
1132	int error;
1133	fhandle_t fh;
1134	struct mount *mp;
1135	struct vnode *vp;
1136
1137	/*
1138	 * Must be super user
1139	 */
1140	if ((error = suser(p->p_ucred, &p->p_acflag)))
1141		return (error);
1142
1143	if ((error = copyin(SCARG(uap, fhp), &fh, sizeof(fhandle_t))) != 0)
1144		return (error);
1145
1146	if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL)
1147		return (ESTALE);
1148	if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp)))
1149		return (error);
1150	error = vn_stat(vp, &sb, p);
1151	vput(vp);
1152	if (error)
1153		return (error);
1154	error = copyout(&sb, SCARG(uap, sb), sizeof(sb));
1155	return (error);
1156}
1157
1158/* ARGSUSED */
1159int
1160sys_fhstatfs(p, v, retval)
1161	struct proc *p;
1162	void *v;
1163	register_t *retval;
1164{
1165	register struct sys_fhstatfs_args /*
1166		syscallarg(const fhandle_t *) fhp;
1167		syscallarg(struct statfs *) buf;
1168	} */ *uap = v;
1169	struct statfs sp;
1170	fhandle_t fh;
1171	struct mount *mp;
1172	struct vnode *vp;
1173	int error;
1174
1175	/*
1176	 * Must be super user
1177	 */
1178	if ((error = suser(p->p_ucred, &p->p_acflag)))
1179		return (error);
1180
1181	if ((error = copyin(SCARG(uap, fhp), &fh, sizeof(fhandle_t))) != 0)
1182		return (error);
1183
1184	if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL)
1185		return (ESTALE);
1186	if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp)))
1187		return (error);
1188	mp = vp->v_mount;
1189	vput(vp);
1190	if ((error = VFS_STATFS(mp, &sp, p)) != 0)
1191		return (error);
1192	sp.f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
1193	return (copyout(&sp, SCARG(uap, buf), sizeof(sp)));
1194}
1195
1196/*
1197 * Create a special file.
1198 */
1199/* ARGSUSED */
1200int
1201sys_mknod(p, v, retval)
1202	struct proc *p;
1203	void *v;
1204	register_t *retval;
1205{
1206	register struct sys_mknod_args /* {
1207		syscallarg(char *) path;
1208		syscallarg(int) mode;
1209		syscallarg(int) dev;
1210	} */ *uap = v;
1211	register struct vnode *vp;
1212	struct vattr vattr;
1213	int error;
1214	int whiteout = 0;
1215	struct nameidata nd;
1216
1217	if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
1218		return (error);
1219	if (p->p_fd->fd_rdir)
1220		return (EINVAL);
1221	NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), p);
1222	if ((error = namei(&nd)) != 0)
1223		return (error);
1224	vp = nd.ni_vp;
1225	if (vp != NULL)
1226		error = EEXIST;
1227	else {
1228		VATTR_NULL(&vattr);
1229		vattr.va_mode = (SCARG(uap, mode) & ALLPERMS) &~ p->p_fd->fd_cmask;
1230		vattr.va_rdev = SCARG(uap, dev);
1231		whiteout = 0;
1232
1233		switch (SCARG(uap, mode) & S_IFMT) {
1234		case S_IFMT:	/* used by badsect to flag bad sectors */
1235			vattr.va_type = VBAD;
1236			break;
1237		case S_IFCHR:
1238			vattr.va_type = VCHR;
1239			break;
1240		case S_IFBLK:
1241			vattr.va_type = VBLK;
1242			break;
1243		case S_IFWHT:
1244			whiteout = 1;
1245			break;
1246		default:
1247			error = EINVAL;
1248			break;
1249		}
1250	}
1251	if (!error) {
1252		VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
1253		if (whiteout) {
1254			error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE);
1255			if (error)
1256				VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1257			vput(nd.ni_dvp);
1258		} else {
1259			error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp,
1260						&nd.ni_cnd, &vattr);
1261		}
1262	} else {
1263		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1264		if (nd.ni_dvp == vp)
1265			vrele(nd.ni_dvp);
1266		else
1267			vput(nd.ni_dvp);
1268		if (vp)
1269			vrele(vp);
1270	}
1271	return (error);
1272}
1273
1274/*
1275 * Create a named pipe.
1276 */
1277/* ARGSUSED */
1278int
1279sys_mkfifo(p, v, retval)
1280	struct proc *p;
1281	void *v;
1282	register_t *retval;
1283{
1284#ifndef FIFO
1285	return (EOPNOTSUPP);
1286#else
1287	register struct sys_mkfifo_args /* {
1288		syscallarg(char *) path;
1289		syscallarg(int) mode;
1290	} */ *uap = v;
1291	struct vattr vattr;
1292	int error;
1293	struct nameidata nd;
1294
1295	NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), p);
1296	if ((error = namei(&nd)) != 0)
1297		return (error);
1298	if (nd.ni_vp != NULL) {
1299		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1300		if (nd.ni_dvp == nd.ni_vp)
1301			vrele(nd.ni_dvp);
1302		else
1303			vput(nd.ni_dvp);
1304		vrele(nd.ni_vp);
1305		return (EEXIST);
1306	}
1307	VATTR_NULL(&vattr);
1308	vattr.va_type = VFIFO;
1309	vattr.va_mode = (SCARG(uap, mode) & ALLPERMS) &~ p->p_fd->fd_cmask;
1310	VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
1311	return (VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr));
1312#endif /* FIFO */
1313}
1314
1315/*
1316 * Make a hard file link.
1317 */
1318/* ARGSUSED */
1319int
1320sys_link(p, v, retval)
1321	struct proc *p;
1322	void *v;
1323	register_t *retval;
1324{
1325	register struct sys_link_args /* {
1326		syscallarg(char *) path;
1327		syscallarg(char *) link;
1328	} */ *uap = v;
1329	register struct vnode *vp;
1330	struct nameidata nd;
1331	int error;
1332	int flags;
1333
1334	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
1335	if ((error = namei(&nd)) != 0)
1336		return (error);
1337	vp = nd.ni_vp;
1338
1339	flags = LOCKPARENT;
1340	if (vp->v_type == VDIR) {
1341		flags |= STRIPSLASHES;
1342	}
1343
1344	NDINIT(&nd, CREATE, flags, UIO_USERSPACE, SCARG(uap, link), p);
1345	if ((error = namei(&nd)) != 0)
1346		goto out;
1347	if (nd.ni_vp) {
1348		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1349		if (nd.ni_dvp == nd.ni_vp)
1350			vrele(nd.ni_dvp);
1351		else
1352			vput(nd.ni_dvp);
1353		vrele(nd.ni_vp);
1354		error = EEXIST;
1355		goto out;
1356	}
1357	VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
1358	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
1359	error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
1360out:
1361	vrele(vp);
1362	return (error);
1363}
1364
1365/*
1366 * Make a symbolic link.
1367 */
1368/* ARGSUSED */
1369int
1370sys_symlink(p, v, retval)
1371	struct proc *p;
1372	void *v;
1373	register_t *retval;
1374{
1375	register struct sys_symlink_args /* {
1376		syscallarg(char *) path;
1377		syscallarg(char *) link;
1378	} */ *uap = v;
1379	struct vattr vattr;
1380	char *path;
1381	int error;
1382	struct nameidata nd;
1383
1384	MALLOC(path, char *, MAXPATHLEN, M_NAMEI, M_WAITOK);
1385	error = copyinstr(SCARG(uap, path), path, MAXPATHLEN, NULL);
1386	if (error)
1387		goto out;
1388	NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, link), p);
1389	if ((error = namei(&nd)) != 0)
1390		goto out;
1391	if (nd.ni_vp) {
1392		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1393		if (nd.ni_dvp == nd.ni_vp)
1394			vrele(nd.ni_dvp);
1395		else
1396			vput(nd.ni_dvp);
1397		vrele(nd.ni_vp);
1398		error = EEXIST;
1399		goto out;
1400	}
1401	VATTR_NULL(&vattr);
1402	vattr.va_mode = ACCESSPERMS &~ p->p_fd->fd_cmask;
1403	VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
1404	error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, path);
1405out:
1406	FREE(path, M_NAMEI);
1407	return (error);
1408}
1409
1410/*
1411 * Delete a whiteout from the filesystem.
1412 */
1413/* ARGSUSED */
1414int
1415sys_undelete(p, v, retval)
1416	struct proc *p;
1417	void *v;
1418	register_t *retval;
1419{
1420	register struct sys_undelete_args /* {
1421		syscallarg(char *) path;
1422	} */ *uap = v;
1423	int error;
1424	struct nameidata nd;
1425
1426	NDINIT(&nd, DELETE, LOCKPARENT|DOWHITEOUT, UIO_USERSPACE,
1427	    SCARG(uap, path), p);
1428	error = namei(&nd);
1429	if (error)
1430		return (error);
1431
1432	if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) {
1433		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1434		if (nd.ni_dvp == nd.ni_vp)
1435			vrele(nd.ni_dvp);
1436		else
1437			vput(nd.ni_dvp);
1438		if (nd.ni_vp)
1439			vrele(nd.ni_vp);
1440		return (EEXIST);
1441	}
1442
1443	VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
1444	if ((error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE)) != 0)
1445		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1446	vput(nd.ni_dvp);
1447	return (error);
1448}
1449
1450/*
1451 * Delete a name from the filesystem.
1452 */
1453/* ARGSUSED */
1454int
1455sys_unlink(p, v, retval)
1456	struct proc *p;
1457	void *v;
1458	register_t *retval;
1459{
1460	struct sys_unlink_args /* {
1461		syscallarg(char *) path;
1462	} */ *uap = v;
1463	register struct vnode *vp;
1464	int error;
1465	struct nameidata nd;
1466
1467	NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF, UIO_USERSPACE,
1468	    SCARG(uap, path), p);
1469	if ((error = namei(&nd)) != 0)
1470		return (error);
1471	vp = nd.ni_vp;
1472
1473	/*
1474	 * The root of a mounted filesystem cannot be deleted.
1475	 */
1476	if (vp->v_flag & VROOT) {
1477		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1478		if (nd.ni_dvp == vp)
1479			vrele(nd.ni_dvp);
1480		else
1481			vput(nd.ni_dvp);
1482		vput(vp);
1483		error = EBUSY;
1484		goto out;
1485	}
1486
1487	(void)uvm_vnp_uncache(vp);
1488
1489	VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
1490	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
1491	error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
1492out:
1493	return (error);
1494}
1495
1496/*
1497 * Reposition read/write file offset.
1498 */
1499int
1500sys_lseek(p, v, retval)
1501	struct proc *p;
1502	void *v;
1503	register_t *retval;
1504{
1505	register struct sys_lseek_args /* {
1506		syscallarg(int) fd;
1507		syscallarg(int) pad;
1508		syscallarg(off_t) offset;
1509		syscallarg(int) whence;
1510	} */ *uap = v;
1511	struct ucred *cred = p->p_ucred;
1512	register struct filedesc *fdp = p->p_fd;
1513	register struct file *fp;
1514	struct vattr vattr;
1515	struct vnode *vp;
1516	int error, special;
1517
1518	if ((fp = fd_getfile(fdp, SCARG(uap, fd))) == NULL)
1519		return (EBADF);
1520	if (fp->f_type != DTYPE_VNODE)
1521		return (ESPIPE);
1522	vp = (struct vnode *)fp->f_data;
1523	if (vp->v_type == VFIFO)
1524		return (ESPIPE);
1525	if (vp->v_type == VCHR)
1526		special = 1;
1527	else
1528		special = 0;
1529	switch (SCARG(uap, whence)) {
1530	case SEEK_CUR:
1531		if (!special && fp->f_offset + SCARG(uap, offset) < 0)
1532			return (EINVAL);
1533		fp->f_offset += SCARG(uap, offset);
1534		break;
1535	case SEEK_END:
1536		error = VOP_GETATTR((struct vnode *)fp->f_data, &vattr,
1537				    cred, p);
1538		if (error)
1539			return (error);
1540		if (!special && (off_t)vattr.va_size + SCARG(uap, offset) < 0)
1541			return (EINVAL);
1542		fp->f_offset = SCARG(uap, offset) + vattr.va_size;
1543		break;
1544	case SEEK_SET:
1545		if (!special && SCARG(uap, offset) < 0)
1546			return (EINVAL);
1547		fp->f_offset = SCARG(uap, offset);
1548		break;
1549	default:
1550		return (EINVAL);
1551	}
1552	*(off_t *)retval = fp->f_offset;
1553	return (0);
1554}
1555
1556/*
1557 * Check access permissions.
1558 */
1559int
1560sys_access(p, v, retval)
1561	struct proc *p;
1562	void *v;
1563	register_t *retval;
1564{
1565	register struct sys_access_args /* {
1566		syscallarg(char *) path;
1567		syscallarg(int) flags;
1568	} */ *uap = v;
1569	register struct ucred *cred = p->p_ucred;
1570	register struct vnode *vp;
1571	int error, flags, t_gid, t_uid;
1572	struct nameidata nd;
1573
1574	if (SCARG(uap, flags) & ~(R_OK | W_OK | X_OK))
1575		return (EINVAL);
1576	t_uid = cred->cr_uid;
1577	t_gid = cred->cr_gid;
1578	cred->cr_uid = p->p_cred->p_ruid;
1579	cred->cr_gid = p->p_cred->p_rgid;
1580	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
1581	    SCARG(uap, path), p);
1582	if ((error = namei(&nd)) != 0)
1583		goto out1;
1584	vp = nd.ni_vp;
1585
1586	/* Flags == 0 means only check for existence. */
1587	if (SCARG(uap, flags)) {
1588		flags = 0;
1589		if (SCARG(uap, flags) & R_OK)
1590			flags |= VREAD;
1591		if (SCARG(uap, flags) & W_OK)
1592			flags |= VWRITE;
1593		if (SCARG(uap, flags) & X_OK)
1594			flags |= VEXEC;
1595		if ((flags & VWRITE) == 0 || (error = vn_writechk(vp)) == 0)
1596			error = VOP_ACCESS(vp, flags, cred, p);
1597	}
1598	vput(vp);
1599out1:
1600	cred->cr_uid = t_uid;
1601	cred->cr_gid = t_gid;
1602	return (error);
1603}
1604
1605/*
1606 * Get file status; this version follows links.
1607 */
1608/* ARGSUSED */
1609int
1610sys_stat(p, v, retval)
1611	struct proc *p;
1612	void *v;
1613	register_t *retval;
1614{
1615	register struct sys_stat_args /* {
1616		syscallarg(char *) path;
1617		syscallarg(struct stat *) ub;
1618	} */ *uap = v;
1619	struct stat sb;
1620	int error;
1621	struct nameidata nd;
1622
1623	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
1624	    SCARG(uap, path), p);
1625	if ((error = namei(&nd)) != 0)
1626		return (error);
1627	error = vn_stat(nd.ni_vp, &sb, p);
1628	vput(nd.ni_vp);
1629	if (error)
1630		return (error);
1631	/* Don't let non-root see generation numbers (for NFS security) */
1632	if (suser(p->p_ucred, &p->p_acflag))
1633		sb.st_gen = 0;
1634	error = copyout((caddr_t)&sb, (caddr_t)SCARG(uap, ub), sizeof (sb));
1635	return (error);
1636}
1637
1638/*
1639 * Get file status; this version does not follow links.
1640 */
1641/* ARGSUSED */
1642int
1643sys_lstat(p, v, retval)
1644	struct proc *p;
1645	void *v;
1646	register_t *retval;
1647{
1648	register struct sys_lstat_args /* {
1649		syscallarg(char *) path;
1650		syscallarg(struct stat *) ub;
1651	} */ *uap = v;
1652	struct stat sb;
1653	int error;
1654	struct nameidata nd;
1655
1656	NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF, UIO_USERSPACE,
1657	    SCARG(uap, path), p);
1658	if ((error = namei(&nd)) != 0)
1659		return (error);
1660	error = vn_stat(nd.ni_vp, &sb, p);
1661	vput(nd.ni_vp);
1662	if (error)
1663		return (error);
1664	/* Don't let non-root see generation numbers (for NFS security) */
1665	if (suser(p->p_ucred, &p->p_acflag))
1666		sb.st_gen = 0;
1667	error = copyout((caddr_t)&sb, (caddr_t)SCARG(uap, ub), sizeof (sb));
1668	return (error);
1669}
1670
1671/*
1672 * Get configurable pathname variables.
1673 */
1674/* ARGSUSED */
1675int
1676sys_pathconf(p, v, retval)
1677	struct proc *p;
1678	void *v;
1679	register_t *retval;
1680{
1681	register struct sys_pathconf_args /* {
1682		syscallarg(char *) path;
1683		syscallarg(int) name;
1684	} */ *uap = v;
1685	int error;
1686	struct nameidata nd;
1687
1688	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
1689	    SCARG(uap, path), p);
1690	if ((error = namei(&nd)) != 0)
1691		return (error);
1692	error = VOP_PATHCONF(nd.ni_vp, SCARG(uap, name), retval);
1693	vput(nd.ni_vp);
1694	return (error);
1695}
1696
1697/*
1698 * Return target name of a symbolic link.
1699 */
1700/* ARGSUSED */
1701int
1702sys_readlink(p, v, retval)
1703	struct proc *p;
1704	void *v;
1705	register_t *retval;
1706{
1707	register struct sys_readlink_args /* {
1708		syscallarg(char *) path;
1709		syscallarg(char *) buf;
1710		syscallarg(size_t) count;
1711	} */ *uap = v;
1712	register struct vnode *vp;
1713	struct iovec aiov;
1714	struct uio auio;
1715	int error;
1716	struct nameidata nd;
1717
1718	NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF, UIO_USERSPACE,
1719	    SCARG(uap, path), p);
1720	if ((error = namei(&nd)) != 0)
1721		return (error);
1722	vp = nd.ni_vp;
1723	if (vp->v_type != VLNK)
1724		error = EINVAL;
1725	else {
1726		aiov.iov_base = SCARG(uap, buf);
1727		aiov.iov_len = SCARG(uap, count);
1728		auio.uio_iov = &aiov;
1729		auio.uio_iovcnt = 1;
1730		auio.uio_offset = 0;
1731		auio.uio_rw = UIO_READ;
1732		auio.uio_segflg = UIO_USERSPACE;
1733		auio.uio_procp = p;
1734		auio.uio_resid = SCARG(uap, count);
1735		error = VOP_READLINK(vp, &auio, p->p_ucred);
1736	}
1737	vput(vp);
1738	*retval = SCARG(uap, count) - auio.uio_resid;
1739	return (error);
1740}
1741
1742/*
1743 * Change flags of a file given a path name.
1744 */
1745/* ARGSUSED */
1746int
1747sys_chflags(p, v, retval)
1748	struct proc *p;
1749	void *v;
1750	register_t *retval;
1751{
1752	register struct sys_chflags_args /* {
1753		syscallarg(char *) path;
1754		syscallarg(unsigned int) flags;
1755	} */ *uap = v;
1756	register struct vnode *vp;
1757	struct vattr vattr;
1758	int error;
1759	struct nameidata nd;
1760
1761	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
1762	if ((error = namei(&nd)) != 0)
1763		return (error);
1764	vp = nd.ni_vp;
1765	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
1766	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, p);
1767	if (vp->v_mount->mnt_flag & MNT_RDONLY)
1768		error = EROFS;
1769	else if (SCARG(uap, flags) == VNOVAL)
1770		error = EINVAL;
1771	else {
1772		if (suser(p->p_ucred, &p->p_acflag)) {
1773			if ((error = VOP_GETATTR(vp, &vattr, p->p_ucred, p)) != 0)
1774				goto out;
1775			if (vattr.va_type == VCHR || vattr.va_type == VBLK) {
1776				error = EINVAL;
1777				goto out;
1778			}
1779		}
1780		VATTR_NULL(&vattr);
1781		vattr.va_flags = SCARG(uap, flags);
1782		error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
1783	}
1784out:
1785	vput(vp);
1786	return (error);
1787}
1788
1789/*
1790 * Change flags of a file given a file descriptor.
1791 */
1792/* ARGSUSED */
1793int
1794sys_fchflags(p, v, retval)
1795	struct proc *p;
1796	void *v;
1797	register_t *retval;
1798{
1799	struct sys_fchflags_args /* {
1800		syscallarg(int) fd;
1801		syscallarg(unsigned int) flags;
1802	} */ *uap = v;
1803	struct vattr vattr;
1804	struct vnode *vp;
1805	struct file *fp;
1806	int error;
1807
1808	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
1809		return (error);
1810	vp = (struct vnode *)fp->f_data;
1811	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
1812	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, p);
1813	if (vp->v_mount->mnt_flag & MNT_RDONLY)
1814		error = EROFS;
1815	else if (SCARG(uap, flags) == VNOVAL)
1816		error = EINVAL;
1817	else {
1818		if (suser(p->p_ucred, &p->p_acflag)) {
1819			if ((error = VOP_GETATTR(vp, &vattr, p->p_ucred, p))
1820			    != 0)
1821				goto out;
1822			if (vattr.va_type == VCHR || vattr.va_type == VBLK) {
1823				error = EINVAL;
1824				goto out;
1825			}
1826		}
1827		VATTR_NULL(&vattr);
1828		vattr.va_flags = SCARG(uap, flags);
1829		error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
1830	}
1831out:
1832	VOP_UNLOCK(vp, 0, p);
1833	FRELE(fp);
1834	return (error);
1835}
1836
1837/*
1838 * Change mode of a file given path name.
1839 */
1840/* ARGSUSED */
1841int
1842sys_chmod(p, v, retval)
1843	struct proc *p;
1844	void *v;
1845	register_t *retval;
1846{
1847	register struct sys_chmod_args /* {
1848		syscallarg(char *) path;
1849		syscallarg(int) mode;
1850	} */ *uap = v;
1851	register struct vnode *vp;
1852	struct vattr vattr;
1853	int error;
1854	struct nameidata nd;
1855
1856	if (SCARG(uap, mode) & ~(S_IFMT | ALLPERMS))
1857		return (EINVAL);
1858
1859	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
1860	if ((error = namei(&nd)) != 0)
1861		return (error);
1862	vp = nd.ni_vp;
1863	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
1864	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, p);
1865	if (vp->v_mount->mnt_flag & MNT_RDONLY)
1866		error = EROFS;
1867	else {
1868		VATTR_NULL(&vattr);
1869		vattr.va_mode = SCARG(uap, mode) & ALLPERMS;
1870		error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
1871	}
1872	vput(vp);
1873	return (error);
1874}
1875
1876/*
1877 * Change mode of a file given a file descriptor.
1878 */
1879/* ARGSUSED */
1880int
1881sys_fchmod(p, v, retval)
1882	struct proc *p;
1883	void *v;
1884	register_t *retval;
1885{
1886	struct sys_fchmod_args /* {
1887		syscallarg(int) fd;
1888		syscallarg(int) mode;
1889	} */ *uap = v;
1890	struct vattr vattr;
1891	struct vnode *vp;
1892	struct file *fp;
1893	int error;
1894
1895	if (SCARG(uap, mode) & ~(S_IFMT | ALLPERMS))
1896		return (EINVAL);
1897
1898	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
1899		return (error);
1900	vp = (struct vnode *)fp->f_data;
1901	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
1902	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, p);
1903	if (vp->v_mount->mnt_flag & MNT_RDONLY)
1904		error = EROFS;
1905	else {
1906		VATTR_NULL(&vattr);
1907		vattr.va_mode = SCARG(uap, mode) & ALLPERMS;
1908		error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
1909	}
1910	VOP_UNLOCK(vp, 0, p);
1911	FRELE(fp);
1912	return (error);
1913}
1914
1915/*
1916 * Set ownership given a path name.
1917 */
1918/* ARGSUSED */
1919int
1920sys_chown(p, v, retval)
1921	struct proc *p;
1922	void *v;
1923	register_t *retval;
1924{
1925	register struct sys_chown_args /* {
1926		syscallarg(char *) path;
1927		syscallarg(int) uid;
1928		syscallarg(int) gid;
1929	} */ *uap = v;
1930	register struct vnode *vp;
1931	struct vattr vattr;
1932	int error;
1933	struct nameidata nd;
1934	u_short mode;
1935
1936	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
1937	if ((error = namei(&nd)) != 0)
1938		return (error);
1939	vp = nd.ni_vp;
1940	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
1941	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, p);
1942	if (vp->v_mount->mnt_flag & MNT_RDONLY)
1943		error = EROFS;
1944	else {
1945		if ((SCARG(uap, uid) != -1 || SCARG(uap, gid) != -1) &&
1946		    (suser(p->p_ucred, &p->p_acflag) || suid_clear)) {
1947			error = VOP_GETATTR(vp, &vattr, p->p_ucred, p);
1948			if (error)
1949				goto out;
1950			mode = vattr.va_mode & ~(VSUID | VSGID);
1951			if (mode == vattr.va_mode)
1952				mode = VNOVAL;
1953		}
1954		else
1955			mode = VNOVAL;
1956		VATTR_NULL(&vattr);
1957		vattr.va_uid = SCARG(uap, uid);
1958		vattr.va_gid = SCARG(uap, gid);
1959		vattr.va_mode = mode;
1960		error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
1961	}
1962out:
1963	vput(vp);
1964	return (error);
1965}
1966
1967/*
1968 * Set ownership given a path name, without following links.
1969 */
1970/* ARGSUSED */
1971int
1972sys_lchown(p, v, retval)
1973	struct proc *p;
1974	void *v;
1975	register_t *retval;
1976{
1977	register struct sys_lchown_args /* {
1978		syscallarg(char *) path;
1979		syscallarg(int) uid;
1980		syscallarg(int) gid;
1981	} */ *uap = v;
1982	register struct vnode *vp;
1983	struct vattr vattr;
1984	int error;
1985	struct nameidata nd;
1986	u_short mode;
1987
1988	NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
1989	if ((error = namei(&nd)) != 0)
1990		return (error);
1991	vp = nd.ni_vp;
1992	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
1993	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, p);
1994	if (vp->v_mount->mnt_flag & MNT_RDONLY)
1995		error = EROFS;
1996	else {
1997		if ((SCARG(uap, uid) != -1 || SCARG(uap, gid) != -1) &&
1998		    (suser(p->p_ucred, &p->p_acflag) || suid_clear)) {
1999			error = VOP_GETATTR(vp, &vattr, p->p_ucred, p);
2000			if (error)
2001				goto out;
2002			mode = vattr.va_mode & ~(VSUID | VSGID);
2003			if (mode == vattr.va_mode)
2004				mode = VNOVAL;
2005		}
2006		else
2007			mode = VNOVAL;
2008		VATTR_NULL(&vattr);
2009		vattr.va_uid = SCARG(uap, uid);
2010		vattr.va_gid = SCARG(uap, gid);
2011		vattr.va_mode = mode;
2012		error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
2013	}
2014out:
2015	vput(vp);
2016	return (error);
2017}
2018
2019/*
2020 * Set ownership given a file descriptor.
2021 */
2022/* ARGSUSED */
2023int
2024sys_fchown(p, v, retval)
2025	struct proc *p;
2026	void *v;
2027	register_t *retval;
2028{
2029	struct sys_fchown_args /* {
2030		syscallarg(int) fd;
2031		syscallarg(int) uid;
2032		syscallarg(int) gid;
2033	} */ *uap = v;
2034	struct vnode *vp;
2035	struct vattr vattr;
2036	int error;
2037	struct file *fp;
2038	u_short mode;
2039
2040	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2041		return (error);
2042	vp = (struct vnode *)fp->f_data;
2043	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2044	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, p);
2045	if (vp->v_mount->mnt_flag & MNT_RDONLY)
2046		error = EROFS;
2047	else {
2048		if ((SCARG(uap, uid) != -1 || SCARG(uap, gid) != -1) &&
2049		    (suser(p->p_ucred, &p->p_acflag) || suid_clear)) {
2050			error = VOP_GETATTR(vp, &vattr, p->p_ucred, p);
2051			if (error)
2052				goto out;
2053			mode = vattr.va_mode & ~(VSUID | VSGID);
2054			if (mode == vattr.va_mode)
2055				mode = VNOVAL;
2056		} else
2057			mode = VNOVAL;
2058		VATTR_NULL(&vattr);
2059		vattr.va_uid = SCARG(uap, uid);
2060		vattr.va_gid = SCARG(uap, gid);
2061		vattr.va_mode = mode;
2062		error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
2063	}
2064out:
2065	VOP_UNLOCK(vp, 0, p);
2066	FRELE(fp);
2067	return (error);
2068}
2069
2070/*
2071 * Set the access and modification times given a path name.
2072 */
2073/* ARGSUSED */
2074int
2075sys_utimes(p, v, retval)
2076	struct proc *p;
2077	void *v;
2078	register_t *retval;
2079{
2080	register struct sys_utimes_args /* {
2081		syscallarg(char *) path;
2082		syscallarg(struct timeval *) tptr;
2083	} */ *uap = v;
2084	register struct vnode *vp;
2085	struct timeval tv[2];
2086	struct vattr vattr;
2087	int error;
2088	struct nameidata nd;
2089
2090	VATTR_NULL(&vattr);
2091	if (SCARG(uap, tptr) == NULL) {
2092		microtime(&tv[0]);
2093		tv[1] = tv[0];
2094		vattr.va_vaflags |= VA_UTIMES_NULL;
2095	} else {
2096		error = copyin((caddr_t)SCARG(uap, tptr), (caddr_t)tv,
2097		    sizeof (tv));
2098		if (error)
2099			return (error);
2100		/* XXX workaround timeval matching the VFS constant VNOVAL */
2101		if (tv[0].tv_sec == VNOVAL)
2102			tv[0].tv_sec = VNOVAL - 1;
2103		if (tv[1].tv_sec == VNOVAL)
2104			tv[1].tv_sec = VNOVAL - 1;
2105	}
2106	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2107	if ((error = namei(&nd)) != 0)
2108		return (error);
2109	vp = nd.ni_vp;
2110	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2111	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, p);
2112	if (vp->v_mount->mnt_flag & MNT_RDONLY)
2113		error = EROFS;
2114	else {
2115		vattr.va_atime.tv_sec = tv[0].tv_sec;
2116		vattr.va_atime.tv_nsec = tv[0].tv_usec * 1000;
2117		vattr.va_mtime.tv_sec = tv[1].tv_sec;
2118		vattr.va_mtime.tv_nsec = tv[1].tv_usec * 1000;
2119		error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
2120	}
2121	vput(vp);
2122	return (error);
2123}
2124
2125
2126/*
2127 * Set the access and modification times given a file descriptor.
2128 */
2129/* ARGSUSED */
2130int
2131sys_futimes(p, v, retval)
2132	struct proc *p;
2133	void *v;
2134	register_t *retval;
2135{
2136	register struct sys_futimes_args /* {
2137		syscallarg(int) fd;
2138		syscallarg(struct timeval *) tptr;
2139	} */ *uap = v;
2140	struct vnode *vp;
2141	struct timeval tv[2];
2142	struct vattr vattr;
2143	int error;
2144	struct file *fp;
2145
2146	VATTR_NULL(&vattr);
2147	if (SCARG(uap, tptr) == NULL) {
2148		microtime(&tv[0]);
2149		tv[1] = tv[0];
2150		vattr.va_vaflags |= VA_UTIMES_NULL;
2151	} else {
2152		error = copyin((caddr_t)SCARG(uap, tptr), (caddr_t)tv,
2153		    sizeof (tv));
2154		if (error)
2155			return (error);
2156		/* XXX workaround timeval matching the VFS constant VNOVAL */
2157		if (tv[0].tv_sec == VNOVAL)
2158			tv[0].tv_sec = VNOVAL - 1;
2159		if (tv[1].tv_sec == VNOVAL)
2160			tv[1].tv_sec = VNOVAL - 1;
2161	}
2162	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2163		return (error);
2164	vp = (struct vnode *)fp->f_data;
2165	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2166	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, p);
2167	if (vp->v_mount->mnt_flag & MNT_RDONLY)
2168		error = EROFS;
2169	else {
2170		vattr.va_atime.tv_sec = tv[0].tv_sec;
2171		vattr.va_atime.tv_nsec = tv[0].tv_usec * 1000;
2172		vattr.va_mtime.tv_sec = tv[1].tv_sec;
2173		vattr.va_mtime.tv_nsec = tv[1].tv_usec * 1000;
2174		error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
2175	}
2176	VOP_UNLOCK(vp, 0, p);
2177	FRELE(fp);
2178	return (error);
2179}
2180
2181/*
2182 * Truncate a file given its path name.
2183 */
2184/* ARGSUSED */
2185int
2186sys_truncate(p, v, retval)
2187	struct proc *p;
2188	void *v;
2189	register_t *retval;
2190{
2191	register struct sys_truncate_args /* {
2192		syscallarg(char *) path;
2193		syscallarg(int) pad;
2194		syscallarg(off_t) length;
2195	} */ *uap = v;
2196	register struct vnode *vp;
2197	struct vattr vattr;
2198	int error;
2199	struct nameidata nd;
2200
2201	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2202	if ((error = namei(&nd)) != 0)
2203		return (error);
2204	vp = nd.ni_vp;
2205	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2206	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, p);
2207	if (vp->v_type == VDIR)
2208		error = EISDIR;
2209	else if ((error = vn_writechk(vp)) == 0 &&
2210	    (error = VOP_ACCESS(vp, VWRITE, p->p_ucred, p)) == 0) {
2211		VATTR_NULL(&vattr);
2212		vattr.va_size = SCARG(uap, length);
2213		error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
2214	}
2215	vput(vp);
2216	return (error);
2217}
2218
2219/*
2220 * Truncate a file given a file descriptor.
2221 */
2222/* ARGSUSED */
2223int
2224sys_ftruncate(p, v, retval)
2225	struct proc *p;
2226	void *v;
2227	register_t *retval;
2228{
2229	struct sys_ftruncate_args /* {
2230		syscallarg(int) fd;
2231		syscallarg(int) pad;
2232		syscallarg(off_t) length;
2233	} */ *uap = v;
2234	struct vattr vattr;
2235	struct vnode *vp;
2236	struct file *fp;
2237	int error;
2238
2239	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2240		return (error);
2241	if ((fp->f_flag & FWRITE) == 0) {
2242		error = EINVAL;
2243		goto bad;
2244	}
2245	vp = (struct vnode *)fp->f_data;
2246	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2247	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, p);
2248	if (vp->v_type == VDIR)
2249		error = EISDIR;
2250	else if ((error = vn_writechk(vp)) == 0) {
2251		VATTR_NULL(&vattr);
2252		vattr.va_size = SCARG(uap, length);
2253		error = VOP_SETATTR(vp, &vattr, fp->f_cred, p);
2254	}
2255	VOP_UNLOCK(vp, 0, p);
2256bad:
2257	FRELE(fp);
2258	return (error);
2259}
2260
2261/*
2262 * Sync an open file.
2263 */
2264/* ARGSUSED */
2265int
2266sys_fsync(p, v, retval)
2267	struct proc *p;
2268	void *v;
2269	register_t *retval;
2270{
2271	struct sys_fsync_args /* {
2272		syscallarg(int) fd;
2273	} */ *uap = v;
2274	struct vnode *vp;
2275	struct file *fp;
2276	int error;
2277
2278	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2279		return (error);
2280	vp = (struct vnode *)fp->f_data;
2281	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, p);
2282	error = VOP_FSYNC(vp, fp->f_cred, MNT_WAIT, p);
2283
2284	VOP_UNLOCK(vp, 0, p);
2285	FRELE(fp);
2286	return (error);
2287}
2288
2289/*
2290 * Rename files.  Source and destination must either both be directories,
2291 * or both not be directories.  If target is a directory, it must be empty.
2292 */
2293/* ARGSUSED */
2294int
2295sys_rename(p, v, retval)
2296	struct proc *p;
2297	void *v;
2298	register_t *retval;
2299{
2300	register struct sys_rename_args /* {
2301		syscallarg(char *) from;
2302		syscallarg(char *) to;
2303	} */ *uap = v;
2304	register struct vnode *tvp, *fvp, *tdvp;
2305	struct nameidata fromnd, tond;
2306	int error;
2307	int flags;
2308
2309	NDINIT(&fromnd, DELETE, WANTPARENT | SAVESTART, UIO_USERSPACE,
2310	    SCARG(uap, from), p);
2311	if ((error = namei(&fromnd)) != 0)
2312		return (error);
2313	fvp = fromnd.ni_vp;
2314
2315	flags = LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART;
2316	/*
2317	 * rename("foo/", "bar/");  is  OK
2318	 */
2319	if (fvp->v_type == VDIR)
2320		flags |= STRIPSLASHES;
2321
2322	NDINIT(&tond, RENAME, flags,
2323	    UIO_USERSPACE, SCARG(uap, to), p);
2324	if ((error = namei(&tond)) != 0) {
2325		VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
2326		vrele(fromnd.ni_dvp);
2327		vrele(fvp);
2328		goto out1;
2329	}
2330	tdvp = tond.ni_dvp;
2331	tvp = tond.ni_vp;
2332	if (tvp != NULL) {
2333		if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
2334			error = ENOTDIR;
2335			goto out;
2336		} else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
2337			error = EISDIR;
2338			goto out;
2339		}
2340	}
2341	if (fvp == tdvp)
2342		error = EINVAL;
2343	/*
2344	 * If source is the same as the destination (that is the
2345	 * same inode number)
2346	 */
2347	if (fvp == tvp)
2348		error = -1;
2349out:
2350	if (!error) {
2351		VOP_LEASE(tdvp, p, p->p_ucred, LEASE_WRITE);
2352		if (fromnd.ni_dvp != tdvp)
2353			VOP_LEASE(fromnd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
2354		if (tvp) {
2355			(void)uvm_vnp_uncache(tvp);
2356			VOP_LEASE(tvp, p, p->p_ucred, LEASE_WRITE);
2357		}
2358		error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
2359				   tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
2360	} else {
2361		VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd);
2362		if (tdvp == tvp)
2363			vrele(tdvp);
2364		else
2365			vput(tdvp);
2366		if (tvp)
2367			vput(tvp);
2368		VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
2369		vrele(fromnd.ni_dvp);
2370		vrele(fvp);
2371	}
2372	vrele(tond.ni_startdir);
2373	FREE(tond.ni_cnd.cn_pnbuf, M_NAMEI);
2374out1:
2375	if (fromnd.ni_startdir)
2376		vrele(fromnd.ni_startdir);
2377	FREE(fromnd.ni_cnd.cn_pnbuf, M_NAMEI);
2378	if (error == -1)
2379		return (0);
2380	return (error);
2381}
2382
2383/*
2384 * Make a directory file.
2385 */
2386/* ARGSUSED */
2387int
2388sys_mkdir(p, v, retval)
2389	struct proc *p;
2390	void *v;
2391	register_t *retval;
2392{
2393	register struct sys_mkdir_args /* {
2394		syscallarg(char *) path;
2395		syscallarg(int) mode;
2396	} */ *uap = v;
2397	register struct vnode *vp;
2398	struct vattr vattr;
2399	int error;
2400	struct nameidata nd;
2401
2402	NDINIT(&nd, CREATE, LOCKPARENT | STRIPSLASHES,
2403	    UIO_USERSPACE, SCARG(uap, path), p);
2404	if ((error = namei(&nd)) != 0)
2405		return (error);
2406	vp = nd.ni_vp;
2407	if (vp != NULL) {
2408		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2409		if (nd.ni_dvp == vp)
2410			vrele(nd.ni_dvp);
2411		else
2412			vput(nd.ni_dvp);
2413		vrele(vp);
2414		return (EEXIST);
2415	}
2416	VATTR_NULL(&vattr);
2417	vattr.va_type = VDIR;
2418	vattr.va_mode = (SCARG(uap, mode) & ACCESSPERMS) &~ p->p_fd->fd_cmask;
2419	VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
2420	error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
2421	if (!error)
2422		vput(nd.ni_vp);
2423	return (error);
2424}
2425
2426/*
2427 * Remove a directory file.
2428 */
2429/* ARGSUSED */
2430int
2431sys_rmdir(p, v, retval)
2432	struct proc *p;
2433	void *v;
2434	register_t *retval;
2435{
2436	struct sys_rmdir_args /* {
2437		syscallarg(char *) path;
2438	} */ *uap = v;
2439	register struct vnode *vp;
2440	int error;
2441	struct nameidata nd;
2442
2443	NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF, UIO_USERSPACE,
2444	    SCARG(uap, path), p);
2445	if ((error = namei(&nd)) != 0)
2446		return (error);
2447	vp = nd.ni_vp;
2448	if (vp->v_type != VDIR) {
2449		error = ENOTDIR;
2450		goto out;
2451	}
2452	/*
2453	 * No rmdir "." please.
2454	 */
2455	if (nd.ni_dvp == vp) {
2456		error = EBUSY;
2457		goto out;
2458	}
2459	/*
2460	 * The root of a mounted filesystem cannot be deleted.
2461	 */
2462	if (vp->v_flag & VROOT)
2463		error = EBUSY;
2464out:
2465	if (!error) {
2466		VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
2467		VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2468		error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
2469	} else {
2470		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2471		if (nd.ni_dvp == vp)
2472			vrele(nd.ni_dvp);
2473		else
2474			vput(nd.ni_dvp);
2475		vput(vp);
2476	}
2477	return (error);
2478}
2479
2480/*
2481 * Read a block of directory entries in a file system independent format.
2482 */
2483int
2484sys_getdirentries(p, v, retval)
2485	struct proc *p;
2486	void *v;
2487	register_t *retval;
2488{
2489	struct sys_getdirentries_args /* {
2490		syscallarg(int) fd;
2491		syscallarg(char *) buf;
2492		syscallarg(int) count;
2493		syscallarg(long *) basep;
2494	} */ *uap = v;
2495	struct vnode *vp;
2496	struct file *fp;
2497	struct uio auio;
2498	struct iovec aiov;
2499	long loff;
2500	int error, eofflag;
2501
2502	if (SCARG(uap, count) < 0)
2503		return EINVAL;
2504	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2505		return (error);
2506	if ((fp->f_flag & FREAD) == 0) {
2507		error = EBADF;
2508		goto bad;
2509	}
2510	vp = (struct vnode *)fp->f_data;
2511unionread:
2512	if (vp->v_type != VDIR) {
2513		error = EINVAL;
2514		goto bad;
2515	}
2516	aiov.iov_base = SCARG(uap, buf);
2517	aiov.iov_len = SCARG(uap, count);
2518	auio.uio_iov = &aiov;
2519	auio.uio_iovcnt = 1;
2520	auio.uio_rw = UIO_READ;
2521	auio.uio_segflg = UIO_USERSPACE;
2522	auio.uio_procp = p;
2523	auio.uio_resid = SCARG(uap, count);
2524	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, p);
2525	loff = auio.uio_offset = fp->f_offset;
2526	error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, 0, 0);
2527	fp->f_offset = auio.uio_offset;
2528	VOP_UNLOCK(vp, 0, p);
2529	if (error)
2530		goto bad;
2531	if ((SCARG(uap, count) == auio.uio_resid) &&
2532	    union_check_p &&
2533	    (union_check_p(p, &vp, fp, auio, &error) != 0))
2534		goto unionread;
2535	if (error)
2536		goto bad;
2537
2538	if ((SCARG(uap, count) == auio.uio_resid) &&
2539	    (vp->v_flag & VROOT) &&
2540	    (vp->v_mount->mnt_flag & MNT_UNION)) {
2541		struct vnode *tvp = vp;
2542		vp = vp->v_mount->mnt_vnodecovered;
2543		VREF(vp);
2544		fp->f_data = (caddr_t) vp;
2545		fp->f_offset = 0;
2546		vrele(tvp);
2547		goto unionread;
2548	}
2549	error = copyout((caddr_t)&loff, (caddr_t)SCARG(uap, basep),
2550	    sizeof(long));
2551	*retval = SCARG(uap, count) - auio.uio_resid;
2552bad:
2553	FRELE(fp);
2554	return (error);
2555}
2556
2557/*
2558 * Set the mode mask for creation of filesystem nodes.
2559 */
2560int
2561sys_umask(p, v, retval)
2562	struct proc *p;
2563	void *v;
2564	register_t *retval;
2565{
2566	struct sys_umask_args /* {
2567		syscallarg(int) newmask;
2568	} */ *uap = v;
2569	register struct filedesc *fdp;
2570
2571	fdp = p->p_fd;
2572	*retval = fdp->fd_cmask;
2573	fdp->fd_cmask = SCARG(uap, newmask) & ACCESSPERMS;
2574	return (0);
2575}
2576
2577/*
2578 * Void all references to file by ripping underlying filesystem
2579 * away from vnode.
2580 */
2581/* ARGSUSED */
2582int
2583sys_revoke(p, v, retval)
2584	struct proc *p;
2585	void *v;
2586	register_t *retval;
2587{
2588	register struct sys_revoke_args /* {
2589		syscallarg(char *) path;
2590	} */ *uap = v;
2591	register struct vnode *vp;
2592	struct vattr vattr;
2593	int error;
2594	struct nameidata nd;
2595
2596	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2597	if ((error = namei(&nd)) != 0)
2598		return (error);
2599	vp = nd.ni_vp;
2600	if ((error = VOP_GETATTR(vp, &vattr, p->p_ucred, p)) != 0)
2601		goto out;
2602	if (p->p_ucred->cr_uid != vattr.va_uid &&
2603	    (error = suser(p->p_ucred, &p->p_acflag)))
2604		goto out;
2605	if (vp->v_usecount > 1 || (vp->v_flag & (VALIASED | VLAYER)))
2606		VOP_REVOKE(vp, REVOKEALL);
2607out:
2608	vrele(vp);
2609	return (error);
2610}
2611
2612/*
2613 * Convert a user file descriptor to a kernel file entry.
2614 *
2615 * On return *fpp is FREF:ed.
2616 */
2617int
2618getvnode(fdp, fd, fpp)
2619	struct filedesc *fdp;
2620	struct file **fpp;
2621	int fd;
2622{
2623	struct file *fp;
2624
2625	if ((fp = fd_getfile(fdp, fd)) == NULL)
2626		return (EBADF);
2627	if (fp->f_type != DTYPE_VNODE)
2628		return (EINVAL);
2629	FREF(fp);
2630	*fpp = fp;
2631
2632	return (0);
2633}
2634
2635/*
2636 * Positional read system call.
2637 */
2638int
2639sys_pread(p, v, retval)
2640	struct proc *p;
2641	void *v;
2642	register_t *retval;
2643{
2644	struct sys_pread_args /* {
2645		syscallarg(int) fd;
2646		syscallarg(void *) buf;
2647		syscallarg(size_t) nbyte;
2648		syscallarg(int) pad;
2649		syscallarg(off_t) offset;
2650	} */ *uap = v;
2651	struct filedesc *fdp = p->p_fd;
2652	struct file *fp;
2653	struct vnode *vp;
2654	off_t offset;
2655	int fd = SCARG(uap, fd);
2656
2657	if ((fp = fd_getfile(fdp, fd)) == NULL)
2658		return (EBADF);
2659	if ((fp->f_flag & FREAD) == 0)
2660		return (EBADF);
2661
2662	vp = (struct vnode *)fp->f_data;
2663	if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
2664		return (ESPIPE);
2665	}
2666
2667	offset = SCARG(uap, offset);
2668
2669	FREF(fp);
2670
2671	/* dofileread() will FRELE the descriptor for us */
2672	return (dofileread(p, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte),
2673	    &offset, retval));
2674}
2675
2676/*
2677 * Positional scatter read system call.
2678 */
2679int
2680sys_preadv(p, v, retval)
2681	struct proc *p;
2682	void *v;
2683	register_t *retval;
2684{
2685	struct sys_preadv_args /* {
2686		syscallarg(int) fd;
2687		syscallarg(const struct iovec *) iovp;
2688		syscallarg(int) iovcnt;
2689		syscallarg(int) pad;
2690		syscallarg(off_t) offset;
2691	} */ *uap = v;
2692	struct filedesc *fdp = p->p_fd;
2693	struct file *fp;
2694	struct vnode *vp;
2695	off_t offset;
2696	int fd = SCARG(uap, fd);
2697
2698	if ((fp = fd_getfile(fdp, fd)) == NULL)
2699		return (EBADF);
2700	if ((fp->f_flag & FREAD) == 0)
2701		return (EBADF);
2702
2703	vp = (struct vnode *)fp->f_data;
2704	if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
2705		return (ESPIPE);
2706	}
2707
2708	FREF(fp);
2709
2710	offset = SCARG(uap, offset);
2711
2712	/* dofilereadv() will FRELE the descriptor for us */
2713	return (dofilereadv(p, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt),
2714	    &offset, retval));
2715}
2716
2717/*
2718 * Positional write system call.
2719 */
2720int
2721sys_pwrite(p, v, retval)
2722	struct proc *p;
2723	void *v;
2724	register_t *retval;
2725{
2726	struct sys_pwrite_args /* {
2727		syscallarg(int) fd;
2728		syscallarg(const void *) buf;
2729		syscallarg(size_t) nbyte;
2730		syscallarg(int) pad;
2731		syscallarg(off_t) offset;
2732	} */ *uap = v;
2733	struct filedesc *fdp = p->p_fd;
2734	struct file *fp;
2735	struct vnode *vp;
2736	off_t offset;
2737	int fd = SCARG(uap, fd);
2738
2739	if ((fp = fd_getfile(fdp, fd)) == NULL)
2740		return (EBADF);
2741	if ((fp->f_flag & FWRITE) == 0)
2742		return (EBADF);
2743
2744	vp = (struct vnode *)fp->f_data;
2745	if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
2746		return (ESPIPE);
2747	}
2748
2749	FREF(fp);
2750
2751	offset = SCARG(uap, offset);
2752
2753	/* dofilewrite() will FRELE the descriptor for us */
2754	return (dofilewrite(p, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte),
2755	    &offset, retval));
2756}
2757
2758
2759/*
2760 * Positional gather write system call.
2761 */
2762int
2763sys_pwritev(p, v, retval)
2764	struct proc *p;
2765	void *v;
2766	register_t *retval;
2767{
2768	struct sys_pwritev_args /* {
2769		syscallarg(int) fd;
2770		syscallarg(const struct iovec *) iovp;
2771		syscallarg(int) iovcnt;
2772		syscallarg(int) pad;
2773		syscallarg(off_t) offset;
2774	} */ *uap = v;
2775	struct filedesc *fdp = p->p_fd;
2776	struct file *fp;
2777	struct vnode *vp;
2778	off_t offset;
2779	int fd = SCARG(uap, fd);
2780
2781	if ((fp = fd_getfile(fdp, fd)) == NULL)
2782		return (EBADF);
2783	if ((fp->f_flag & FWRITE) == 0)
2784		return (EBADF);
2785
2786	vp = (struct vnode *)fp->f_data;
2787	if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
2788		return (ESPIPE);
2789	}
2790
2791	FREF(fp);
2792
2793	offset = SCARG(uap, offset);
2794
2795	/* dofilewritev() will FRELE the descriptor for us */
2796	return (dofilewritev(p, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt),
2797	    &offset, retval));
2798}
2799
2800#ifdef UFS_EXTATTR
2801/*
2802 * Syscall to push extended attribute configuration information into the
2803 * VFS.  Accepts a path, which it converts to a mountpoint, as well as
2804 * a command (int cmd), and attribute name and misc data.  For now, the
2805 * attribute name is left in userspace for consumption by the VFS_op.
2806 * It will probably be changed to be copied into sysspace by the
2807 * syscall in the future, once issues with various consumers of the
2808 * attribute code have raised their hands.
2809 *
2810 * Currently this is used only by UFS Extended Attributes.
2811 */
2812int
2813sys_extattrctl(struct proc *p, void *v, register_t *reval)
2814{
2815	struct sys_extattrctl_args /* {
2816		syscallarg(const char *) path;
2817		syscallarg(int) cmd;
2818		syscallarg(const char *) filename;
2819		syscallarg(int) attrnamespace;
2820		syscallarg(const char *) attrname;
2821	} */ *uap = v;
2822	struct vnode *filename_vp;
2823	struct nameidata nd;
2824	struct mount *mp;
2825	char attrname[EXTATTR_MAXNAMELEN];
2826	int error;
2827
2828	/*
2829	 * SCARG(uap, attrname) not always defined.  We check again later
2830	 * when we invoke the VFS call so as to pass in NULL there if needed.
2831	 */
2832	if (SCARG(uap, attrname) != NULL) {
2833		error = copyinstr(SCARG(uap, attrname), attrname,
2834		    EXTATTR_MAXNAMELEN, NULL);
2835		if (error)
2836			return (error);
2837	}
2838
2839	/*
2840	 * SCARG(uap, filename) not always defined.  If it is, grab
2841	 * a vnode lock, which VFS_EXTATTRCTL() will later release.
2842	 */
2843	filename_vp = NULL;
2844	if (SCARG(uap, filename) != NULL) {
2845		NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
2846		    SCARG(uap, filename), p);
2847		if ((error = namei(&nd)) != 0)
2848			return (error);
2849		filename_vp = nd.ni_vp;
2850	}
2851
2852	/* SCARG(uap, path) always defined. */
2853	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2854	if ((error = namei(&nd)) != 0) {
2855		if (filename_vp != NULL)
2856			vput(filename_vp);
2857		return (error);
2858	}
2859
2860	mp = nd.ni_vp->v_mount;
2861	if (error) {
2862		if (filename_vp != NULL)
2863			vput(filename_vp);
2864		return (error);
2865	}
2866
2867	if (SCARG(uap, attrname) != NULL) {
2868		error = VFS_EXTATTRCTL(mp, SCARG(uap, cmd), filename_vp,
2869		    SCARG(uap, attrnamespace), attrname, p);
2870	} else {
2871		error = VFS_EXTATTRCTL(mp, SCARG(uap, cmd), filename_vp,
2872		    SCARG(uap, attrnamespace), NULL, p);
2873	}
2874
2875	/*
2876	 * VFS_EXTATTRCTL will have unlocked, but not de-ref'd,
2877	 * filename_vp, so vrele it if it is defined.
2878	 */
2879	if (filename_vp != NULL)
2880		vrele(filename_vp);
2881
2882	return (error);
2883}
2884
2885/*-
2886 * Set a named extended attribute on a file or directory
2887 *
2888 * Arguments: unlocked vnode "vp", attribute namespace "attrnamespace",
2889 *            kernelspace string pointer "attrname", userspace buffer
2890 *            pointer "data", buffer length "nbytes", thread "td".
2891 * Returns: 0 on success, an error number otherwise
2892 * Locks: none
2893 * References: vp must be a valid reference for the duration of the call
2894 */
2895static int
2896extattr_set_vp(struct vnode *vp, int attrnamespace, const char *attrname,
2897    void *data, size_t nbytes, struct proc *p, register_t *retval)
2898{
2899	struct uio auio;
2900	struct iovec aiov;
2901	ssize_t cnt;
2902	int error;
2903
2904	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2905	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, p);
2906
2907	aiov.iov_base = data;
2908	aiov.iov_len = nbytes;
2909	auio.uio_iov = &aiov;
2910	auio.uio_iovcnt = 1;
2911	auio.uio_offset = 0;
2912	if (nbytes > INT_MAX) {
2913		error = EINVAL;
2914		goto done;
2915	}
2916	auio.uio_resid = nbytes;
2917	auio.uio_rw = UIO_WRITE;
2918	auio.uio_segflg = UIO_USERSPACE;
2919	auio.uio_procp = p;
2920	cnt = nbytes;
2921
2922	error = VOP_SETEXTATTR(vp, attrnamespace, attrname, &auio,
2923	    p->p_ucred, p);
2924	cnt -= auio.uio_resid;
2925	retval[0] = cnt;
2926
2927done:
2928	VOP_UNLOCK(vp, 0, p);
2929	return (error);
2930}
2931
2932int
2933sys_extattr_set_file(struct proc *p, void *v, register_t *retval)
2934{
2935	struct sys_extattr_set_file_args /* {
2936		syscallarg(const char *) path;
2937		syscallarg(int) attrnamespace;
2938		syscallarg(const char *) attrname;
2939		syscallarg(void *) data;
2940		syscallarg(size_t) nbytes;
2941	} */ *uap = v;
2942	struct nameidata nd;
2943	char attrname[EXTATTR_MAXNAMELEN];
2944	int error;
2945
2946	error = copyinstr(SCARG(uap, attrname), attrname, EXTATTR_MAXNAMELEN,
2947	    NULL);
2948	if (error)
2949		return (error);
2950
2951	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2952	if ((error = namei(&nd)) != 0)
2953		return (error);
2954
2955	error = extattr_set_vp(nd.ni_vp, SCARG(uap, attrnamespace), attrname,
2956	    SCARG(uap, data), SCARG(uap, nbytes), p, retval);
2957
2958	vrele(nd.ni_vp);
2959	return (error);
2960}
2961
2962int
2963sys_extattr_set_fd(struct proc *p, void *v, register_t *retval)
2964{
2965	struct sys_extattr_set_fd_args /* {
2966		syscallarg(int) fd;
2967		syscallarg(int) attrnamespace;
2968		syscallarg(const char *) attrname;
2969		syscallarg(struct iovec *) iovp;
2970		syscallarg(int) iovcnt;
2971	} */ *uap = v;
2972	struct file *fp;
2973	char attrname[EXTATTR_MAXNAMELEN];
2974	int error;
2975
2976	error = copyinstr(SCARG(uap, attrname), attrname, EXTATTR_MAXNAMELEN,
2977	    NULL);
2978	if (error)
2979		return (error);
2980
2981	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2982		return (error);
2983
2984	error = extattr_set_vp((struct vnode *)fp->f_data,
2985	    SCARG(uap, attrnamespace), attrname, SCARG(uap, data),
2986	    SCARG(uap, nbytes), p, retval);
2987	FRELE(fp);
2988
2989	return (error);
2990}
2991
2992/*-
2993 * Get a named extended attribute on a file or directory
2994 *
2995 * Arguments: unlocked vnode "vp", attribute namespace "attrnamespace",
2996 *            kernelspace string pointer "attrname", userspace buffer
2997 *            pointer "data", buffer length "nbytes", thread "td".
2998 * Returns: 0 on success, an error number otherwise
2999 * Locks: none
3000 * References: vp must be a valid reference for the duration of the call
3001 */
3002static int
3003extattr_get_vp(struct vnode *vp, int attrnamespace, const char *attrname,
3004    void *data, size_t nbytes, struct proc *p, register_t *retval)
3005{
3006	struct uio auio;
3007	struct iovec aiov;
3008	ssize_t cnt;
3009	size_t size;
3010	int error;
3011
3012	VOP_LEASE(vp, p, p->p_ucred, LEASE_READ);
3013	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, p);
3014
3015	/*
3016	 * Slightly unusual semantics: if the user provides a NULL data
3017	 * pointer, they don't want to receive the data, just the
3018	 * maximum read length.
3019	 */
3020	if (data != NULL) {
3021		aiov.iov_base = data;
3022		aiov.iov_len = nbytes;
3023		auio.uio_iov = &aiov;
3024		auio.uio_offset = 0;
3025		if (nbytes > INT_MAX) {
3026			error = EINVAL;
3027			goto done;
3028		}
3029		auio.uio_resid = nbytes;
3030		auio.uio_rw = UIO_READ;
3031		auio.uio_segflg = UIO_USERSPACE;
3032		auio.uio_procp = p;
3033		cnt = nbytes;
3034		error = VOP_GETEXTATTR(vp, attrnamespace, attrname, &auio,
3035		    NULL, p->p_ucred, p);
3036		cnt -= auio.uio_resid;
3037		retval[0] = cnt;
3038	} else {
3039		error = VOP_GETEXTATTR(vp, attrnamespace, attrname, NULL,
3040		    &size, p->p_ucred, p);
3041		retval[0] = size;
3042	}
3043done:
3044	VOP_UNLOCK(vp, 0, p);
3045	return (error);
3046}
3047
3048int
3049sys_extattr_get_file(p, v, retval)
3050	struct proc *p;
3051	void *v;
3052	register_t *retval;
3053{
3054	struct sys_extattr_get_file_args /* {
3055		syscallarg(const char *) path;
3056		syscallarg(int) attrnamespace;
3057		syscallarg(const char *) attrname;
3058		syscallarg(void *) data;
3059		syscallarg(size_t) nbytes;
3060	} */ *uap = v;
3061	struct nameidata nd;
3062	char attrname[EXTATTR_MAXNAMELEN];
3063	int error;
3064
3065	error = copyinstr(SCARG(uap, attrname), attrname, EXTATTR_MAXNAMELEN,
3066	    NULL);
3067	if (error)
3068		return (error);
3069
3070	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
3071	if ((error = namei(&nd)) != 0)
3072		return (error);
3073
3074	error = extattr_get_vp(nd.ni_vp, SCARG(uap, attrnamespace), attrname,
3075	    SCARG(uap, data), SCARG(uap, nbytes), p, retval);
3076
3077	vrele(nd.ni_vp);
3078	return (error);
3079}
3080
3081int
3082sys_extattr_get_fd(p, v, retval)
3083	struct proc *p;
3084	void *v;
3085	register_t *retval;
3086{
3087	struct sys_extattr_get_fd_args /* {
3088		syscallarg(int) fd;
3089		syscallarg(int) attrnamespace;
3090		syscallarg(const char *) attrname;
3091		syscallarg(void *) data;
3092		syscallarg(size_t) nbytes;
3093	} */ *uap = v;
3094	struct file *fp;
3095	char attrname[EXTATTR_MAXNAMELEN];
3096	int error;
3097
3098	error = copyinstr(SCARG(uap, attrname), attrname, EXTATTR_MAXNAMELEN,
3099	    NULL);
3100	if (error)
3101		return (error);
3102
3103	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
3104		return (error);
3105
3106	error = extattr_get_vp((struct vnode *)fp->f_data,
3107	    SCARG(uap, attrnamespace), attrname, SCARG(uap, data),
3108	    SCARG(uap, nbytes), p, retval);
3109	FRELE(fp);
3110
3111	return (error);
3112}
3113
3114/*
3115 * extattr_delete_vp(): Delete a named extended attribute on a file or
3116 *                      directory
3117 *
3118 * Arguments: unlocked vnode "vp", attribute namespace "attrnamespace",
3119 *            kernelspace string pointer "attrname", proc "p"
3120 * Returns: 0 on success, an error number otherwise
3121 * Locks: none
3122 * References: vp must be a valid reference for the duration of the call
3123 */
3124static int
3125extattr_delete_vp(struct vnode *vp, int attrnamespace, const char *attrname,
3126    struct proc *p)
3127{
3128	int error;
3129
3130	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
3131	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, p);
3132
3133	error = VOP_SETEXTATTR(vp, attrnamespace, attrname, NULL,
3134	    p->p_ucred, p);
3135
3136	VOP_UNLOCK(vp, 0, p);
3137	return (error);
3138}
3139
3140int
3141sys_extattr_delete_file(p, v, retval)
3142	struct proc *p;
3143	void *v;
3144	register_t *retval;
3145{
3146	struct sys_extattr_delete_file_args /* {
3147		syscallarg(int) fd;
3148		syscallarg(int) attrnamespace;
3149		syscallarg(const char *) attrname;
3150	} */ *uap = v;
3151	struct nameidata nd;
3152	char attrname[EXTATTR_MAXNAMELEN];
3153	int error;
3154
3155	error = copyinstr(SCARG(uap, attrname), attrname, EXTATTR_MAXNAMELEN,
3156	     NULL);
3157	if (error)
3158		return(error);
3159
3160	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
3161	if ((error = namei(&nd)) != 0)
3162		return(error);
3163
3164	error = extattr_delete_vp(nd.ni_vp, SCARG(uap, attrnamespace),
3165	    attrname, p);
3166
3167	vrele(nd.ni_vp);
3168	return(error);
3169}
3170
3171int
3172sys_extattr_delete_fd(p, v, retval)
3173	struct proc *p;
3174	void *v;
3175	register_t *retval;
3176{
3177	struct sys_extattr_delete_fd_args /* {
3178		syscallarg(int) fd;
3179		syscallarg(int) attrnamespace;
3180		syscallarg(const char *) attrname;
3181	} */ *uap = v;
3182	struct file *fp;
3183	char attrname[EXTATTR_MAXNAMELEN];
3184	int error;
3185
3186	error = copyinstr(SCARG(uap, attrname), attrname, EXTATTR_MAXNAMELEN,
3187	    NULL);
3188	if (error)
3189		return (error);
3190
3191	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
3192		return (error);
3193
3194	error = extattr_delete_vp((struct vnode *)fp->f_data,
3195	    SCARG(uap, attrnamespace), attrname, p);
3196	FRELE(fp);
3197
3198	return (error);
3199}
3200#endif
3201