xref: /openbsd-src/regress/usr.bin/diff/t9.2 (revision a28daedfc357b214be5c701aa8ba8adb29a7f1c2)
1/*	$OpenBSD: t9.2,v 1.1 2003/07/17 21:04:04 otto 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		CIRCLEQ_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	CIRCLEQ_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	for (mp = CIRCLEQ_LAST(&mountlist); mp != CIRCLEQ_END(&mountlist);
514	    mp = nmp) {
515		if (vfs_busy(mp, LK_NOWAIT, &mountlist_slock, p)) {
516			nmp = CIRCLEQ_PREV(mp, mnt_list);
517			continue;
518		}
519		if ((mp->mnt_flag & MNT_RDONLY) == 0) {
520			asyncflag = mp->mnt_flag & MNT_ASYNC;
521			mp->mnt_flag &= ~MNT_ASYNC;
522			uvm_vnp_sync(mp);
523			VFS_SYNC(mp, MNT_NOWAIT, p->p_ucred, p);
524			if (asyncflag)
525				mp->mnt_flag |= MNT_ASYNC;
526		}
527		simple_lock(&mountlist_slock);
528		nmp = CIRCLEQ_PREV(mp, mnt_list);
529		vfs_unbusy(mp, p);
530	}
531	simple_unlock(&mountlist_slock);
532
533#ifdef DEBUG
534	if (syncprt)
535		vfs_bufstats();
536#endif /* DEBUG */
537	return (0);
538}
539
540/*
541 * Change filesystem quotas.
542 */
543/* ARGSUSED */
544int
545sys_quotactl(p, v, retval)
546	struct proc *p;
547	void *v;
548	register_t *retval;
549{
550	register struct sys_quotactl_args /* {
551		syscallarg(char *) path;
552		syscallarg(int) cmd;
553		syscallarg(int) uid;
554		syscallarg(caddr_t) arg;
555	} */ *uap = v;
556	register struct mount *mp;
557	int error;
558	struct nameidata nd;
559
560	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
561	if ((error = namei(&nd)) != 0)
562		return (error);
563	mp = nd.ni_vp->v_mount;
564	vrele(nd.ni_vp);
565	return (VFS_QUOTACTL(mp, SCARG(uap, cmd), SCARG(uap, uid),
566	    SCARG(uap, arg), p));
567}
568
569/*
570 * Get filesystem statistics.
571 */
572/* ARGSUSED */
573int
574sys_statfs(p, v, retval)
575	struct proc *p;
576	void *v;
577	register_t *retval;
578{
579	register struct sys_statfs_args /* {
580		syscallarg(char *) path;
581		syscallarg(struct statfs *) buf;
582	} */ *uap = v;
583	register struct mount *mp;
584	register struct statfs *sp;
585	int error;
586	struct nameidata nd;
587	struct statfs sb;
588
589	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
590	if ((error = namei(&nd)) != 0)
591		return (error);
592	mp = nd.ni_vp->v_mount;
593	sp = &mp->mnt_stat;
594	vrele(nd.ni_vp);
595	if ((error = VFS_STATFS(mp, sp, p)) != 0)
596		return (error);
597	sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
598#if notyet
599	if (mp->mnt_flag & MNT_SOFTDEP)
600		sp->f_eflags = STATFS_SOFTUPD;
601#endif
602	/* Don't let non-root see filesystem id (for NFS security) */
603	if (suser(p->p_ucred, &p->p_acflag)) {
604		bcopy((caddr_t)sp, (caddr_t)&sb, sizeof(sb));
605		sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
606		sp = &sb;
607	}
608	return (copyout((caddr_t)sp, (caddr_t)SCARG(uap, buf), sizeof(*sp)));
609}
610
611/*
612 * Get filesystem statistics.
613 */
614/* ARGSUSED */
615int
616sys_fstatfs(p, v, retval)
617	struct proc *p;
618	void *v;
619	register_t *retval;
620{
621	struct sys_fstatfs_args /* {
622		syscallarg(int) fd;
623		syscallarg(struct statfs *) buf;
624	} */ *uap = v;
625	struct file *fp;
626	struct mount *mp;
627	struct statfs *sp;
628	int error;
629	struct statfs sb;
630
631	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
632		return (error);
633	mp = ((struct vnode *)fp->f_data)->v_mount;
634	sp = &mp->mnt_stat;
635	error = VFS_STATFS(mp, sp, p);
636	FRELE(fp);
637	if (error)
638		return (error);
639	sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
640#if notyet
641	if (mp->mnt_flag & MNT_SOFTDEP)
642		sp->f_eflags = STATFS_SOFTUPD;
643#endif
644	/* Don't let non-root see filesystem id (for NFS security) */
645	if (suser(p->p_ucred, &p->p_acflag)) {
646		bcopy((caddr_t)sp, (caddr_t)&sb, sizeof(sb));
647		sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
648		sp = &sb;
649	}
650	return (copyout((caddr_t)sp, (caddr_t)SCARG(uap, buf), sizeof(*sp)));
651}
652
653/*
654 * Get statistics on all filesystems.
655 */
656int
657sys_getfsstat(p, v, retval)
658	struct proc *p;
659	void *v;
660	register_t *retval;
661{
662	register struct sys_getfsstat_args /* {
663		syscallarg(struct statfs *) buf;
664		syscallarg(size_t) bufsize;
665		syscallarg(int) flags;
666	} */ *uap = v;
667	register struct mount *mp, *nmp;
668	register struct statfs *sp;
669	struct statfs sb;
670	caddr_t sfsp;
671	size_t count, maxcount;
672	int error, flags = SCARG(uap, flags);
673
674	maxcount = SCARG(uap, bufsize) / sizeof(struct statfs);
675	sfsp = (caddr_t)SCARG(uap, buf);
676	count = 0;
677	simple_lock(&mountlist_slock);
678	for (mp = CIRCLEQ_FIRST(&mountlist); mp != CIRCLEQ_END(&mountlist);
679	    mp = nmp) {
680		if (vfs_busy(mp, LK_NOWAIT, &mountlist_slock, p)) {
681			nmp = CIRCLEQ_NEXT(mp, mnt_list);
682			continue;
683		}
684		if (sfsp && count < maxcount) {
685			sp = &mp->mnt_stat;
686
687			/* Refresh stats unless MNT_NOWAIT is specified */
688			if (flags != MNT_NOWAIT &&
689			    flags != MNT_LAZY &&
690			    (flags == MNT_WAIT ||
691			    flags == 0) &&
692			    (error = VFS_STATFS(mp, sp, p))) {
693				simple_lock(&mountlist_slock);
694				nmp = CIRCLEQ_NEXT(mp, mnt_list);
695				vfs_unbusy(mp, p);
696 				continue;
697			}
698
699			sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
700#if notyet
701			if (mp->mnt_flag & MNT_SOFTDEP)
702				sp->f_eflags = STATFS_SOFTUPD;
703#endif
704			if (suser(p->p_ucred, &p->p_acflag)) {
705				bcopy((caddr_t)sp, (caddr_t)&sb, sizeof(sb));
706				sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
707				sp = &sb;
708			}
709			error = copyout((caddr_t)sp, sfsp, sizeof(*sp));
710			if (error) {
711				vfs_unbusy(mp, p);
712				return (error);
713			}
714			sfsp += sizeof(*sp);
715		}
716		count++;
717		simple_lock(&mountlist_slock);
718		nmp = CIRCLEQ_NEXT(mp, mnt_list);
719		vfs_unbusy(mp, p);
720	}
721	simple_unlock(&mountlist_slock);
722	if (sfsp && count > maxcount)
723		*retval = maxcount;
724	else
725		*retval = count;
726	return (0);
727}
728
729/*
730 * Change current working directory to a given file descriptor.
731 */
732/* ARGSUSED */
733int
734sys_fchdir(p, v, retval)
735	struct proc *p;
736	void *v;
737	register_t *retval;
738{
739	struct sys_fchdir_args /* {
740		syscallarg(int) fd;
741	} */ *uap = v;
742	struct filedesc *fdp = p->p_fd;
743	struct vnode *vp, *tdp;
744	struct mount *mp;
745	struct file *fp;
746	int error;
747
748	if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0)
749		return (error);
750	vp = (struct vnode *)fp->f_data;
751	VREF(vp);
752	FRELE(fp);
753	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, p);
754	if (vp->v_type != VDIR)
755		error = ENOTDIR;
756	else
757		error = VOP_ACCESS(vp, VEXEC, p->p_ucred, p);
758
759	while (!error && (mp = vp->v_mountedhere) != NULL) {
760		if (vfs_busy(mp, 0, 0, p))
761			continue;
762		error = VFS_ROOT(mp, &tdp);
763		vfs_unbusy(mp, p);
764		if (error)
765			break;
766		vput(vp);
767		vp = tdp;
768	}
769	if (error) {
770		vput(vp);
771		return (error);
772	}
773	VOP_UNLOCK(vp, 0, p);
774	vrele(fdp->fd_cdir);
775	fdp->fd_cdir = vp;
776	return (0);
777}
778
779/*
780 * Change current working directory (``.'').
781 */
782/* ARGSUSED */
783int
784sys_chdir(p, v, retval)
785	struct proc *p;
786	void *v;
787	register_t *retval;
788{
789	struct sys_chdir_args /* {
790		syscallarg(char *) path;
791	} */ *uap = v;
792	register struct filedesc *fdp = p->p_fd;
793	int error;
794	struct nameidata nd;
795
796	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
797	    SCARG(uap, path), p);
798	if ((error = change_dir(&nd, p)) != 0)
799		return (error);
800	vrele(fdp->fd_cdir);
801	fdp->fd_cdir = nd.ni_vp;
802	return (0);
803}
804
805/*
806 * Change notion of root (``/'') directory.
807 */
808/* ARGSUSED */
809int
810sys_chroot(p, v, retval)
811	struct proc *p;
812	void *v;
813	register_t *retval;
814{
815	struct sys_chroot_args /* {
816		syscallarg(char *) path;
817	} */ *uap = v;
818	register struct filedesc *fdp = p->p_fd;
819	int error;
820	struct nameidata nd;
821
822	if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
823		return (error);
824	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
825	    SCARG(uap, path), p);
826	if ((error = change_dir(&nd, p)) != 0)
827		return (error);
828	if (fdp->fd_rdir != NULL) {
829		/*
830		 * A chroot() done inside a changed root environment does
831		 * an automatic chdir to avoid the out-of-tree experience.
832		 */
833		vrele(fdp->fd_rdir);
834		vrele(fdp->fd_cdir);
835		VREF(nd.ni_vp);
836		fdp->fd_cdir = nd.ni_vp;
837	}
838	fdp->fd_rdir = nd.ni_vp;
839	return (0);
840}
841
842/*
843 * Common routine for chroot and chdir.
844 */
845static int
846change_dir(ndp, p)
847	register struct nameidata *ndp;
848	struct proc *p;
849{
850	struct vnode *vp;
851	int error;
852
853	if ((error = namei(ndp)) != 0)
854		return (error);
855	vp = ndp->ni_vp;
856	if (vp->v_type != VDIR)
857		error = ENOTDIR;
858	else
859		error = VOP_ACCESS(vp, VEXEC, p->p_ucred, p);
860	if (error)
861		vput(vp);
862	else
863		VOP_UNLOCK(vp, 0, p);
864	return (error);
865}
866
867/*
868 * Check permissions, allocate an open file structure,
869 * and call the device open routine if any.
870 */
871int
872sys_open(p, v, retval)
873	struct proc *p;
874	void *v;
875	register_t *retval;
876{
877	struct sys_open_args /* {
878		syscallarg(char *) path;
879		syscallarg(int) flags;
880		syscallarg(int) mode;
881	} */ *uap = v;
882	struct filedesc *fdp = p->p_fd;
883	struct file *fp;
884	struct vnode *vp;
885	struct vattr vattr;
886	int flags, cmode;
887	int type, indx, error, localtrunc = 0;
888	struct flock lf;
889	struct nameidata nd;
890
891	if ((error = falloc(p, &fp, &indx)) != 0)
892		return (error);
893
894	flags = FFLAGS(SCARG(uap, flags));
895	cmode = ((SCARG(uap, mode) &~ fdp->fd_cmask) & ALLPERMS) &~ S_ISTXT;
896	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
897	p->p_dupfd = -indx - 1;			/* XXX check for fdopen */
898	if ((flags & O_TRUNC) && (flags & (O_EXLOCK | O_SHLOCK))) {
899		localtrunc = 1;
900		flags &= ~O_TRUNC;	/* Must do truncate ourselves */
901	}
902	if ((error = vn_open(&nd, flags, cmode)) != 0) {
903		if ((error == ENODEV || error == ENXIO) &&
904		    p->p_dupfd >= 0 &&			/* XXX from fdopen */
905		    (error =
906			dupfdopen(fdp, indx, p->p_dupfd, flags, error)) == 0) {
907			closef(fp, p);
908			*retval = indx;
909			return (0);
910		}
911		if (error == ERESTART)
912			error = EINTR;
913		fdremove(fdp, indx);
914		closef(fp, p);
915		return (error);
916	}
917	p->p_dupfd = 0;
918	vp = nd.ni_vp;
919	fp->f_flag = flags & FMASK;
920	fp->f_type = DTYPE_VNODE;
921	fp->f_ops = &vnops;
922	fp->f_data = (caddr_t)vp;
923	if (flags & (O_EXLOCK | O_SHLOCK)) {
924		lf.l_whence = SEEK_SET;
925		lf.l_start = 0;
926		lf.l_len = 0;
927		if (flags & O_EXLOCK)
928			lf.l_type = F_WRLCK;
929		else
930			lf.l_type = F_RDLCK;
931		type = F_FLOCK;
932		if ((flags & FNONBLOCK) == 0)
933			type |= F_WAIT;
934		VOP_UNLOCK(vp, 0, p);
935		error = VOP_ADVLOCK(vp, (caddr_t)fp, F_SETLK, &lf, type);
936		if (error) {
937			/* closef will vn_close the file for us. */
938			fdremove(fdp, indx);
939			closef(fp, p);
940			return (error);
941		}
942		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, p);
943		fp->f_flag |= FHASLOCK;
944	}
945	if (localtrunc) {
946		VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
947		if ((fp->f_flag & FWRITE) == 0)
948			error = EACCES;
949		else if (vp->v_mount->mnt_flag & MNT_RDONLY)
950			error = EROFS;
951		else if (vp->v_type == VDIR)
952			error = EISDIR;
953		else if ((error = vn_writechk(vp)) == 0) {
954			VATTR_NULL(&vattr);
955			vattr.va_size = 0;
956			error = VOP_SETATTR(vp, &vattr, fp->f_cred, p);
957		}
958		if (error) {
959			VOP_UNLOCK(vp, 0, p);
960			/* closef will close the file for us. */
961			fdremove(fdp, indx);
962			closef(fp, p);
963			return (error);
964		}
965	}
966	VOP_UNLOCK(vp, 0, p);
967	*retval = indx;
968	FILE_SET_MATURE(fp);
969	return (0);
970}
971
972/*
973 * Get file handle system call
974 */
975int
976sys_getfh(p, v, retval)
977	struct proc *p;
978	register void *v;
979	register_t *retval;
980{
981	register struct sys_getfh_args /* {
982		syscallarg(char *) fname;
983		syscallarg(fhandle_t *) fhp;
984	} */ *uap = v;
985	register struct vnode *vp;
986	fhandle_t fh;
987	int error;
988	struct nameidata nd;
989
990	/*
991	 * Must be super user
992	 */
993	error = suser(p->p_ucred, &p->p_acflag);
994	if (error)
995		return (error);
996	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
997	    SCARG(uap, fname), p);
998	error = namei(&nd);
999	if (error)
1000		return (error);
1001	vp = nd.ni_vp;
1002	bzero((caddr_t)&fh, sizeof(fh));
1003	fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
1004	error = VFS_VPTOFH(vp, &fh.fh_fid);
1005	vput(vp);
1006	if (error)
1007		return (error);
1008	error = copyout((caddr_t)&fh, (caddr_t)SCARG(uap, fhp), sizeof (fh));
1009	return (error);
1010}
1011
1012/*
1013 * Open a file given a file handle.
1014 *
1015 * Check permissions, allocate an open file structure,
1016 * and call the device open routine if any.
1017 */
1018int
1019sys_fhopen(p, v, retval)
1020	struct proc *p;
1021	void *v;
1022	register_t *retval;
1023{
1024	register struct sys_fhopen_args /* {
1025		syscallarg(const fhandle_t *) fhp;
1026		syscallarg(int) flags;
1027	} */ *uap = v;
1028	struct filedesc *fdp = p->p_fd;
1029	struct file *fp;
1030	struct vnode *vp = NULL;
1031	struct mount *mp;
1032	struct ucred *cred = p->p_ucred;
1033	int flags;
1034	int type, indx, error=0;
1035	struct flock lf;
1036	struct vattr va;
1037	fhandle_t fh;
1038
1039	/*
1040	 * Must be super user
1041	 */
1042	if ((error = suser(p->p_ucred, &p->p_acflag)))
1043		return (error);
1044
1045	flags = FFLAGS(SCARG(uap, flags));
1046	if ((flags & (FREAD | FWRITE)) == 0)
1047		return (EINVAL);
1048	if ((flags & O_CREAT))
1049		return (EINVAL);
1050
1051	if ((error = falloc(p, &fp, &indx)) != 0)
1052		return (error);
1053
1054	if ((error = copyin(SCARG(uap, fhp), &fh, sizeof(fhandle_t))) != 0)
1055		goto bad;
1056
1057	if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL) {
1058		error = ESTALE;
1059		goto bad;
1060	}
1061
1062	if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp)) != 0) {
1063		vp = NULL;	/* most likely unnecessary sanity for bad: */
1064		goto bad;
1065	}
1066
1067	/* Now do an effective vn_open */
1068
1069	if (vp->v_type == VSOCK) {
1070		error = EOPNOTSUPP;
1071		goto bad;
1072	}
1073	if (flags & FREAD) {
1074		if ((error = VOP_ACCESS(vp, VREAD, cred, p)) != 0)
1075			goto bad;
1076	}
1077	if (flags & (FWRITE | O_TRUNC)) {
1078		if (vp->v_type == VDIR) {
1079			error = EISDIR;
1080			goto bad;
1081		}
1082		if ((error = vn_writechk(vp)) != 0 ||
1083		    (error = VOP_ACCESS(vp, VWRITE, cred, p)) != 0)
1084			goto bad;
1085	}
1086	if (flags & O_TRUNC) {
1087		VOP_UNLOCK(vp, 0, p);				/* XXX */
1088		VOP_LEASE(vp, p, cred, LEASE_WRITE);
1089		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, p);	/* XXX */
1090		VATTR_NULL(&va);
1091		va.va_size = 0;
1092		if ((error = VOP_SETATTR(vp, &va, cred, p)) != 0)
1093			goto bad;
1094	}
1095	if ((error = VOP_OPEN(vp, flags, cred, p)) != 0)
1096		goto bad;
1097	if (flags & FWRITE)
1098		vp->v_writecount++;
1099
1100	/* done with modified vn_open, now finish what sys_open does. */
1101
1102	fp->f_flag = flags & FMASK;
1103	fp->f_type = DTYPE_VNODE;
1104	fp->f_ops = &vnops;
1105	fp->f_data = (caddr_t)vp;
1106	if (flags & (O_EXLOCK | O_SHLOCK)) {
1107		lf.l_whence = SEEK_SET;
1108		lf.l_start = 0;
1109		lf.l_len = 0;
1110		if (flags & O_EXLOCK)
1111			lf.l_type = F_WRLCK;
1112		else
1113			lf.l_type = F_RDLCK;
1114		type = F_FLOCK;
1115		if ((flags & FNONBLOCK) == 0)
1116			type |= F_WAIT;
1117		VOP_UNLOCK(vp, 0, p);
1118		error = VOP_ADVLOCK(vp, (caddr_t)fp, F_SETLK, &lf, type);
1119		if (error) {
1120			/* closef will vn_close the file for us. */
1121			fdremove(fdp, indx);
1122			closef(fp, p);
1123			return (error);
1124		}
1125		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, p);
1126		fp->f_flag |= FHASLOCK;
1127	}
1128	VOP_UNLOCK(vp, 0, p);
1129	*retval = indx;
1130	FILE_SET_MATURE(fp);
1131	return (0);
1132
1133bad:
1134	fdremove(fdp, indx);
1135	closef(fp, p);
1136	if (vp != NULL)
1137		vput(vp);
1138	return (error);
1139}
1140
1141/* ARGSUSED */
1142int
1143sys_fhstat(p, v, retval)
1144	struct proc *p;
1145	void *v;
1146	register_t *retval;
1147{
1148	register struct sys_fhstat_args /* {
1149		syscallarg(const fhandle_t *) fhp;
1150		syscallarg(struct stat *) sb;
1151	} */ *uap = v;
1152	struct stat sb;
1153	int error;
1154	fhandle_t fh;
1155	struct mount *mp;
1156	struct vnode *vp;
1157
1158	/*
1159	 * Must be super user
1160	 */
1161	if ((error = suser(p->p_ucred, &p->p_acflag)))
1162		return (error);
1163
1164	if ((error = copyin(SCARG(uap, fhp), &fh, sizeof(fhandle_t))) != 0)
1165		return (error);
1166
1167	if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL)
1168		return (ESTALE);
1169	if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp)))
1170		return (error);
1171	error = vn_stat(vp, &sb, p);
1172	vput(vp);
1173	if (error)
1174		return (error);
1175	error = copyout(&sb, SCARG(uap, sb), sizeof(sb));
1176	return (error);
1177}
1178
1179/* ARGSUSED */
1180int
1181sys_fhstatfs(p, v, retval)
1182	struct proc *p;
1183	void *v;
1184	register_t *retval;
1185{
1186	register struct sys_fhstatfs_args /*
1187		syscallarg(const fhandle_t *) fhp;
1188		syscallarg(struct statfs *) buf;
1189	} */ *uap = v;
1190	struct statfs sp;
1191	fhandle_t fh;
1192	struct mount *mp;
1193	struct vnode *vp;
1194	int error;
1195
1196	/*
1197	 * Must be super user
1198	 */
1199	if ((error = suser(p->p_ucred, &p->p_acflag)))
1200		return (error);
1201
1202	if ((error = copyin(SCARG(uap, fhp), &fh, sizeof(fhandle_t))) != 0)
1203		return (error);
1204
1205	if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL)
1206		return (ESTALE);
1207	if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp)))
1208		return (error);
1209	mp = vp->v_mount;
1210	vput(vp);
1211	if ((error = VFS_STATFS(mp, &sp, p)) != 0)
1212		return (error);
1213	sp.f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
1214	return (copyout(&sp, SCARG(uap, buf), sizeof(sp)));
1215}
1216
1217/*
1218 * Create a special file.
1219 */
1220/* ARGSUSED */
1221int
1222sys_mknod(p, v, retval)
1223	struct proc *p;
1224	void *v;
1225	register_t *retval;
1226{
1227	register struct sys_mknod_args /* {
1228		syscallarg(char *) path;
1229		syscallarg(int) mode;
1230		syscallarg(int) dev;
1231	} */ *uap = v;
1232	register struct vnode *vp;
1233	struct vattr vattr;
1234	int error;
1235	int whiteout = 0;
1236	struct nameidata nd;
1237
1238	if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
1239		return (error);
1240	if (p->p_fd->fd_rdir)
1241		return (EINVAL);
1242	NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), p);
1243	if ((error = namei(&nd)) != 0)
1244		return (error);
1245	vp = nd.ni_vp;
1246	if (vp != NULL)
1247		error = EEXIST;
1248	else {
1249		VATTR_NULL(&vattr);
1250		vattr.va_mode = (SCARG(uap, mode) & ALLPERMS) &~ p->p_fd->fd_cmask;
1251		vattr.va_rdev = SCARG(uap, dev);
1252		whiteout = 0;
1253
1254		switch (SCARG(uap, mode) & S_IFMT) {
1255		case S_IFMT:	/* used by badsect to flag bad sectors */
1256			vattr.va_type = VBAD;
1257			break;
1258		case S_IFCHR:
1259			vattr.va_type = VCHR;
1260			break;
1261		case S_IFBLK:
1262			vattr.va_type = VBLK;
1263			break;
1264		case S_IFWHT:
1265			whiteout = 1;
1266			break;
1267		default:
1268			error = EINVAL;
1269			break;
1270		}
1271	}
1272	if (!error) {
1273		VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
1274		if (whiteout) {
1275			error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE);
1276			if (error)
1277				VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1278			vput(nd.ni_dvp);
1279		} else {
1280			error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp,
1281						&nd.ni_cnd, &vattr);
1282		}
1283	} else {
1284		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1285		if (nd.ni_dvp == vp)
1286			vrele(nd.ni_dvp);
1287		else
1288			vput(nd.ni_dvp);
1289		if (vp)
1290			vrele(vp);
1291	}
1292	return (error);
1293}
1294
1295/*
1296 * Create a named pipe.
1297 */
1298/* ARGSUSED */
1299int
1300sys_mkfifo(p, v, retval)
1301	struct proc *p;
1302	void *v;
1303	register_t *retval;
1304{
1305#ifndef FIFO
1306	return (EOPNOTSUPP);
1307#else
1308	register struct sys_mkfifo_args /* {
1309		syscallarg(char *) path;
1310		syscallarg(int) mode;
1311	} */ *uap = v;
1312	struct vattr vattr;
1313	int error;
1314	struct nameidata nd;
1315
1316	NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), p);
1317	if ((error = namei(&nd)) != 0)
1318		return (error);
1319	if (nd.ni_vp != NULL) {
1320		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1321		if (nd.ni_dvp == nd.ni_vp)
1322			vrele(nd.ni_dvp);
1323		else
1324			vput(nd.ni_dvp);
1325		vrele(nd.ni_vp);
1326		return (EEXIST);
1327	}
1328	VATTR_NULL(&vattr);
1329	vattr.va_type = VFIFO;
1330	vattr.va_mode = (SCARG(uap, mode) & ALLPERMS) &~ p->p_fd->fd_cmask;
1331	VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
1332	return (VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr));
1333#endif /* FIFO */
1334}
1335
1336/*
1337 * Make a hard file link.
1338 */
1339/* ARGSUSED */
1340int
1341sys_link(p, v, retval)
1342	struct proc *p;
1343	void *v;
1344	register_t *retval;
1345{
1346	register struct sys_link_args /* {
1347		syscallarg(char *) path;
1348		syscallarg(char *) link;
1349	} */ *uap = v;
1350	register struct vnode *vp;
1351	struct nameidata nd;
1352	int error;
1353	int flags;
1354
1355	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
1356	if ((error = namei(&nd)) != 0)
1357		return (error);
1358	vp = nd.ni_vp;
1359
1360	flags = LOCKPARENT;
1361	if (vp->v_type == VDIR) {
1362		flags |= STRIPSLASHES;
1363	}
1364
1365	NDINIT(&nd, CREATE, flags, UIO_USERSPACE, SCARG(uap, link), p);
1366	if ((error = namei(&nd)) != 0)
1367		goto out;
1368	if (nd.ni_vp) {
1369		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1370		if (nd.ni_dvp == nd.ni_vp)
1371			vrele(nd.ni_dvp);
1372		else
1373			vput(nd.ni_dvp);
1374		vrele(nd.ni_vp);
1375		error = EEXIST;
1376		goto out;
1377	}
1378	VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
1379	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
1380	error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
1381out:
1382	vrele(vp);
1383	return (error);
1384}
1385
1386/*
1387 * Make a symbolic link.
1388 */
1389/* ARGSUSED */
1390int
1391sys_symlink(p, v, retval)
1392	struct proc *p;
1393	void *v;
1394	register_t *retval;
1395{
1396	register struct sys_symlink_args /* {
1397		syscallarg(char *) path;
1398		syscallarg(char *) link;
1399	} */ *uap = v;
1400	struct vattr vattr;
1401	char *path;
1402	int error;
1403	struct nameidata nd;
1404
1405	MALLOC(path, char *, MAXPATHLEN, M_NAMEI, M_WAITOK);
1406	error = copyinstr(SCARG(uap, path), path, MAXPATHLEN, NULL);
1407	if (error)
1408		goto out;
1409	NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, link), p);
1410	if ((error = namei(&nd)) != 0)
1411		goto out;
1412	if (nd.ni_vp) {
1413		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1414		if (nd.ni_dvp == nd.ni_vp)
1415			vrele(nd.ni_dvp);
1416		else
1417			vput(nd.ni_dvp);
1418		vrele(nd.ni_vp);
1419		error = EEXIST;
1420		goto out;
1421	}
1422	VATTR_NULL(&vattr);
1423	vattr.va_mode = ACCESSPERMS &~ p->p_fd->fd_cmask;
1424	VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
1425	error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, path);
1426out:
1427	FREE(path, M_NAMEI);
1428	return (error);
1429}
1430
1431/*
1432 * Delete a whiteout from the filesystem.
1433 */
1434/* ARGSUSED */
1435int
1436sys_undelete(p, v, retval)
1437	struct proc *p;
1438	void *v;
1439	register_t *retval;
1440{
1441	register struct sys_undelete_args /* {
1442		syscallarg(char *) path;
1443	} */ *uap = v;
1444	int error;
1445	struct nameidata nd;
1446
1447	NDINIT(&nd, DELETE, LOCKPARENT|DOWHITEOUT, UIO_USERSPACE,
1448	    SCARG(uap, path), p);
1449	error = namei(&nd);
1450	if (error)
1451		return (error);
1452
1453	if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) {
1454		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1455		if (nd.ni_dvp == nd.ni_vp)
1456			vrele(nd.ni_dvp);
1457		else
1458			vput(nd.ni_dvp);
1459		if (nd.ni_vp)
1460			vrele(nd.ni_vp);
1461		return (EEXIST);
1462	}
1463
1464	VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
1465	if ((error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE)) != 0)
1466		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1467	vput(nd.ni_dvp);
1468	return (error);
1469}
1470
1471/*
1472 * Delete a name from the filesystem.
1473 */
1474/* ARGSUSED */
1475int
1476sys_unlink(p, v, retval)
1477	struct proc *p;
1478	void *v;
1479	register_t *retval;
1480{
1481	struct sys_unlink_args /* {
1482		syscallarg(char *) path;
1483	} */ *uap = v;
1484	register struct vnode *vp;
1485	int error;
1486	struct nameidata nd;
1487
1488	NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF, UIO_USERSPACE,
1489	    SCARG(uap, path), p);
1490	if ((error = namei(&nd)) != 0)
1491		return (error);
1492	vp = nd.ni_vp;
1493
1494	/*
1495	 * The root of a mounted filesystem cannot be deleted.
1496	 */
1497	if (vp->v_flag & VROOT) {
1498		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1499		if (nd.ni_dvp == vp)
1500			vrele(nd.ni_dvp);
1501		else
1502			vput(nd.ni_dvp);
1503		vput(vp);
1504		error = EBUSY;
1505		goto out;
1506	}
1507
1508	(void)uvm_vnp_uncache(vp);
1509
1510	VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
1511	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
1512	error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
1513out:
1514	return (error);
1515}
1516
1517/*
1518 * Reposition read/write file offset.
1519 */
1520int
1521sys_lseek(p, v, retval)
1522	struct proc *p;
1523	void *v;
1524	register_t *retval;
1525{
1526	register struct sys_lseek_args /* {
1527		syscallarg(int) fd;
1528		syscallarg(int) pad;
1529		syscallarg(off_t) offset;
1530		syscallarg(int) whence;
1531	} */ *uap = v;
1532	struct ucred *cred = p->p_ucred;
1533	register struct filedesc *fdp = p->p_fd;
1534	register struct file *fp;
1535	struct vattr vattr;
1536	struct vnode *vp;
1537	int error, special;
1538
1539	if ((fp = fd_getfile(fdp, SCARG(uap, fd))) == NULL)
1540		return (EBADF);
1541	if (fp->f_type != DTYPE_VNODE)
1542		return (ESPIPE);
1543	vp = (struct vnode *)fp->f_data;
1544	if (vp->v_type == VFIFO)
1545		return (ESPIPE);
1546	if (vp->v_type == VCHR)
1547		special = 1;
1548	else
1549		special = 0;
1550	switch (SCARG(uap, whence)) {
1551	case SEEK_CUR:
1552		if (!special && fp->f_offset + SCARG(uap, offset) < 0)
1553			return (EINVAL);
1554		fp->f_offset += SCARG(uap, offset);
1555		break;
1556	case SEEK_END:
1557		error = VOP_GETATTR((struct vnode *)fp->f_data, &vattr,
1558				    cred, p);
1559		if (error)
1560			return (error);
1561		if (!special && (off_t)vattr.va_size + SCARG(uap, offset) < 0)
1562			return (EINVAL);
1563		fp->f_offset = SCARG(uap, offset) + vattr.va_size;
1564		break;
1565	case SEEK_SET:
1566		if (!special && SCARG(uap, offset) < 0)
1567			return (EINVAL);
1568		fp->f_offset = SCARG(uap, offset);
1569		break;
1570	default:
1571		return (EINVAL);
1572	}
1573	*(off_t *)retval = fp->f_offset;
1574	return (0);
1575}
1576
1577/*
1578 * Check access permissions.
1579 */
1580int
1581sys_access(p, v, retval)
1582	struct proc *p;
1583	void *v;
1584	register_t *retval;
1585{
1586	register struct sys_access_args /* {
1587		syscallarg(char *) path;
1588		syscallarg(int) flags;
1589	} */ *uap = v;
1590	register struct ucred *cred = p->p_ucred;
1591	register struct vnode *vp;
1592	int error, flags, t_gid, t_uid;
1593	struct nameidata nd;
1594
1595	if (SCARG(uap, flags) & ~(R_OK | W_OK | X_OK))
1596		return (EINVAL);
1597	t_uid = cred->cr_uid;
1598	t_gid = cred->cr_gid;
1599	cred->cr_uid = p->p_cred->p_ruid;
1600	cred->cr_gid = p->p_cred->p_rgid;
1601	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
1602	    SCARG(uap, path), p);
1603	if ((error = namei(&nd)) != 0)
1604		goto out1;
1605	vp = nd.ni_vp;
1606
1607	/* Flags == 0 means only check for existence. */
1608	if (SCARG(uap, flags)) {
1609		flags = 0;
1610		if (SCARG(uap, flags) & R_OK)
1611			flags |= VREAD;
1612		if (SCARG(uap, flags) & W_OK)
1613			flags |= VWRITE;
1614		if (SCARG(uap, flags) & X_OK)
1615			flags |= VEXEC;
1616		if ((flags & VWRITE) == 0 || (error = vn_writechk(vp)) == 0)
1617			error = VOP_ACCESS(vp, flags, cred, p);
1618	}
1619	vput(vp);
1620out1:
1621	cred->cr_uid = t_uid;
1622	cred->cr_gid = t_gid;
1623	return (error);
1624}
1625
1626/*
1627 * Get file status; this version follows links.
1628 */
1629/* ARGSUSED */
1630int
1631sys_stat(p, v, retval)
1632	struct proc *p;
1633	void *v;
1634	register_t *retval;
1635{
1636	register struct sys_stat_args /* {
1637		syscallarg(char *) path;
1638		syscallarg(struct stat *) ub;
1639	} */ *uap = v;
1640	struct stat sb;
1641	int error;
1642	struct nameidata nd;
1643
1644	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
1645	    SCARG(uap, path), p);
1646	if ((error = namei(&nd)) != 0)
1647		return (error);
1648	error = vn_stat(nd.ni_vp, &sb, p);
1649	vput(nd.ni_vp);
1650	if (error)
1651		return (error);
1652	/* Don't let non-root see generation numbers (for NFS security) */
1653	if (suser(p->p_ucred, &p->p_acflag))
1654		sb.st_gen = 0;
1655	error = copyout((caddr_t)&sb, (caddr_t)SCARG(uap, ub), sizeof (sb));
1656	return (error);
1657}
1658
1659/*
1660 * Get file status; this version does not follow links.
1661 */
1662/* ARGSUSED */
1663int
1664sys_lstat(p, v, retval)
1665	struct proc *p;
1666	void *v;
1667	register_t *retval;
1668{
1669	register struct sys_lstat_args /* {
1670		syscallarg(char *) path;
1671		syscallarg(struct stat *) ub;
1672	} */ *uap = v;
1673	struct stat sb;
1674	int error;
1675	struct nameidata nd;
1676
1677	NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF, UIO_USERSPACE,
1678	    SCARG(uap, path), p);
1679	if ((error = namei(&nd)) != 0)
1680		return (error);
1681	error = vn_stat(nd.ni_vp, &sb, p);
1682	vput(nd.ni_vp);
1683	if (error)
1684		return (error);
1685	/* Don't let non-root see generation numbers (for NFS security) */
1686	if (suser(p->p_ucred, &p->p_acflag))
1687		sb.st_gen = 0;
1688	error = copyout((caddr_t)&sb, (caddr_t)SCARG(uap, ub), sizeof (sb));
1689	return (error);
1690}
1691
1692/*
1693 * Get configurable pathname variables.
1694 */
1695/* ARGSUSED */
1696int
1697sys_pathconf(p, v, retval)
1698	struct proc *p;
1699	void *v;
1700	register_t *retval;
1701{
1702	register struct sys_pathconf_args /* {
1703		syscallarg(char *) path;
1704		syscallarg(int) name;
1705	} */ *uap = v;
1706	int error;
1707	struct nameidata nd;
1708
1709	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
1710	    SCARG(uap, path), p);
1711	if ((error = namei(&nd)) != 0)
1712		return (error);
1713	error = VOP_PATHCONF(nd.ni_vp, SCARG(uap, name), retval);
1714	vput(nd.ni_vp);
1715	return (error);
1716}
1717
1718/*
1719 * Return target name of a symbolic link.
1720 */
1721/* ARGSUSED */
1722int
1723sys_readlink(p, v, retval)
1724	struct proc *p;
1725	void *v;
1726	register_t *retval;
1727{
1728	register struct sys_readlink_args /* {
1729		syscallarg(char *) path;
1730		syscallarg(char *) buf;
1731		syscallarg(size_t) count;
1732	} */ *uap = v;
1733	register struct vnode *vp;
1734	struct iovec aiov;
1735	struct uio auio;
1736	int error;
1737	struct nameidata nd;
1738
1739	NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF, UIO_USERSPACE,
1740	    SCARG(uap, path), p);
1741	if ((error = namei(&nd)) != 0)
1742		return (error);
1743	vp = nd.ni_vp;
1744	if (vp->v_type != VLNK)
1745		error = EINVAL;
1746	else {
1747		aiov.iov_base = SCARG(uap, buf);
1748		aiov.iov_len = SCARG(uap, count);
1749		auio.uio_iov = &aiov;
1750		auio.uio_iovcnt = 1;
1751		auio.uio_offset = 0;
1752		auio.uio_rw = UIO_READ;
1753		auio.uio_segflg = UIO_USERSPACE;
1754		auio.uio_procp = p;
1755		auio.uio_resid = SCARG(uap, count);
1756		error = VOP_READLINK(vp, &auio, p->p_ucred);
1757	}
1758	vput(vp);
1759	*retval = SCARG(uap, count) - auio.uio_resid;
1760	return (error);
1761}
1762
1763/*
1764 * Change flags of a file given a path name.
1765 */
1766/* ARGSUSED */
1767int
1768sys_chflags(p, v, retval)
1769	struct proc *p;
1770	void *v;
1771	register_t *retval;
1772{
1773	register struct sys_chflags_args /* {
1774		syscallarg(char *) path;
1775		syscallarg(unsigned int) flags;
1776	} */ *uap = v;
1777	register struct vnode *vp;
1778	struct vattr vattr;
1779	int error;
1780	struct nameidata nd;
1781
1782	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
1783	if ((error = namei(&nd)) != 0)
1784		return (error);
1785	vp = nd.ni_vp;
1786	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
1787	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, p);
1788	if (vp->v_mount->mnt_flag & MNT_RDONLY)
1789		error = EROFS;
1790	else if (SCARG(uap, flags) == VNOVAL)
1791		error = EINVAL;
1792	else {
1793		if (suser(p->p_ucred, &p->p_acflag)) {
1794			if ((error = VOP_GETATTR(vp, &vattr, p->p_ucred, p)) != 0)
1795				goto out;
1796			if (vattr.va_type == VCHR || vattr.va_type == VBLK) {
1797				error = EINVAL;
1798				goto out;
1799			}
1800		}
1801		VATTR_NULL(&vattr);
1802		vattr.va_flags = SCARG(uap, flags);
1803		error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
1804	}
1805out:
1806	vput(vp);
1807	return (error);
1808}
1809
1810/*
1811 * Change flags of a file given a file descriptor.
1812 */
1813/* ARGSUSED */
1814int
1815sys_fchflags(p, v, retval)
1816	struct proc *p;
1817	void *v;
1818	register_t *retval;
1819{
1820	struct sys_fchflags_args /* {
1821		syscallarg(int) fd;
1822		syscallarg(unsigned int) flags;
1823	} */ *uap = v;
1824	struct vattr vattr;
1825	struct vnode *vp;
1826	struct file *fp;
1827	int error;
1828
1829	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
1830		return (error);
1831	vp = (struct vnode *)fp->f_data;
1832	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
1833	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, p);
1834	if (vp->v_mount->mnt_flag & MNT_RDONLY)
1835		error = EROFS;
1836	else if (SCARG(uap, flags) == VNOVAL)
1837		error = EINVAL;
1838	else {
1839		if (suser(p->p_ucred, &p->p_acflag)) {
1840			if ((error = VOP_GETATTR(vp, &vattr, p->p_ucred, p))
1841			    != 0)
1842				goto out;
1843			if (vattr.va_type == VCHR || vattr.va_type == VBLK) {
1844				error = EINVAL;
1845				goto out;
1846			}
1847		}
1848		VATTR_NULL(&vattr);
1849		vattr.va_flags = SCARG(uap, flags);
1850		error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
1851	}
1852out:
1853	VOP_UNLOCK(vp, 0, p);
1854	FRELE(fp);
1855	return (error);
1856}
1857
1858/*
1859 * Change mode of a file given path name.
1860 */
1861/* ARGSUSED */
1862int
1863sys_chmod(p, v, retval)
1864	struct proc *p;
1865	void *v;
1866	register_t *retval;
1867{
1868	register struct sys_chmod_args /* {
1869		syscallarg(char *) path;
1870		syscallarg(int) mode;
1871	} */ *uap = v;
1872	register struct vnode *vp;
1873	struct vattr vattr;
1874	int error;
1875	struct nameidata nd;
1876
1877	if (SCARG(uap, mode) & ~(S_IFMT | ALLPERMS))
1878		return (EINVAL);
1879
1880	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
1881	if ((error = namei(&nd)) != 0)
1882		return (error);
1883	vp = nd.ni_vp;
1884	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
1885	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, p);
1886	if (vp->v_mount->mnt_flag & MNT_RDONLY)
1887		error = EROFS;
1888	else {
1889		VATTR_NULL(&vattr);
1890		vattr.va_mode = SCARG(uap, mode) & ALLPERMS;
1891		error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
1892	}
1893	vput(vp);
1894	return (error);
1895}
1896
1897/*
1898 * Change mode of a file given a file descriptor.
1899 */
1900/* ARGSUSED */
1901int
1902sys_fchmod(p, v, retval)
1903	struct proc *p;
1904	void *v;
1905	register_t *retval;
1906{
1907	struct sys_fchmod_args /* {
1908		syscallarg(int) fd;
1909		syscallarg(int) mode;
1910	} */ *uap = v;
1911	struct vattr vattr;
1912	struct vnode *vp;
1913	struct file *fp;
1914	int error;
1915
1916	if (SCARG(uap, mode) & ~(S_IFMT | ALLPERMS))
1917		return (EINVAL);
1918
1919	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
1920		return (error);
1921	vp = (struct vnode *)fp->f_data;
1922	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
1923	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, p);
1924	if (vp->v_mount->mnt_flag & MNT_RDONLY)
1925		error = EROFS;
1926	else {
1927		VATTR_NULL(&vattr);
1928		vattr.va_mode = SCARG(uap, mode) & ALLPERMS;
1929		error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
1930	}
1931	VOP_UNLOCK(vp, 0, p);
1932	FRELE(fp);
1933	return (error);
1934}
1935
1936/*
1937 * Set ownership given a path name.
1938 */
1939/* ARGSUSED */
1940int
1941sys_chown(p, v, retval)
1942	struct proc *p;
1943	void *v;
1944	register_t *retval;
1945{
1946	register struct sys_chown_args /* {
1947		syscallarg(char *) path;
1948		syscallarg(int) uid;
1949		syscallarg(int) gid;
1950	} */ *uap = v;
1951	register struct vnode *vp;
1952	struct vattr vattr;
1953	int error;
1954	struct nameidata nd;
1955	u_short mode;
1956
1957	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
1958	if ((error = namei(&nd)) != 0)
1959		return (error);
1960	vp = nd.ni_vp;
1961	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
1962	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, p);
1963	if (vp->v_mount->mnt_flag & MNT_RDONLY)
1964		error = EROFS;
1965	else {
1966		if ((SCARG(uap, uid) != -1 || SCARG(uap, gid) != -1) &&
1967		    (suser(p->p_ucred, &p->p_acflag) || suid_clear)) {
1968			error = VOP_GETATTR(vp, &vattr, p->p_ucred, p);
1969			if (error)
1970				goto out;
1971			mode = vattr.va_mode & ~(VSUID | VSGID);
1972			if (mode == vattr.va_mode)
1973				mode = VNOVAL;
1974		}
1975		else
1976			mode = VNOVAL;
1977		VATTR_NULL(&vattr);
1978		vattr.va_uid = SCARG(uap, uid);
1979		vattr.va_gid = SCARG(uap, gid);
1980		vattr.va_mode = mode;
1981		error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
1982	}
1983out:
1984	vput(vp);
1985	return (error);
1986}
1987
1988/*
1989 * Set ownership given a path name, without following links.
1990 */
1991/* ARGSUSED */
1992int
1993sys_lchown(p, v, retval)
1994	struct proc *p;
1995	void *v;
1996	register_t *retval;
1997{
1998	register struct sys_lchown_args /* {
1999		syscallarg(char *) path;
2000		syscallarg(int) uid;
2001		syscallarg(int) gid;
2002	} */ *uap = v;
2003	register struct vnode *vp;
2004	struct vattr vattr;
2005	int error;
2006	struct nameidata nd;
2007	u_short mode;
2008
2009	NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2010	if ((error = namei(&nd)) != 0)
2011		return (error);
2012	vp = nd.ni_vp;
2013	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2014	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, p);
2015	if (vp->v_mount->mnt_flag & MNT_RDONLY)
2016		error = EROFS;
2017	else {
2018		if ((SCARG(uap, uid) != -1 || SCARG(uap, gid) != -1) &&
2019		    (suser(p->p_ucred, &p->p_acflag) || suid_clear)) {
2020			error = VOP_GETATTR(vp, &vattr, p->p_ucred, p);
2021			if (error)
2022				goto out;
2023			mode = vattr.va_mode & ~(VSUID | VSGID);
2024			if (mode == vattr.va_mode)
2025				mode = VNOVAL;
2026		}
2027		else
2028			mode = VNOVAL;
2029		VATTR_NULL(&vattr);
2030		vattr.va_uid = SCARG(uap, uid);
2031		vattr.va_gid = SCARG(uap, gid);
2032		vattr.va_mode = mode;
2033		error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
2034	}
2035out:
2036	vput(vp);
2037	return (error);
2038}
2039
2040/*
2041 * Set ownership given a file descriptor.
2042 */
2043/* ARGSUSED */
2044int
2045sys_fchown(p, v, retval)
2046	struct proc *p;
2047	void *v;
2048	register_t *retval;
2049{
2050	struct sys_fchown_args /* {
2051		syscallarg(int) fd;
2052		syscallarg(int) uid;
2053		syscallarg(int) gid;
2054	} */ *uap = v;
2055	struct vnode *vp;
2056	struct vattr vattr;
2057	int error;
2058	struct file *fp;
2059	u_short mode;
2060
2061	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2062		return (error);
2063	vp = (struct vnode *)fp->f_data;
2064	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2065	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, p);
2066	if (vp->v_mount->mnt_flag & MNT_RDONLY)
2067		error = EROFS;
2068	else {
2069		if ((SCARG(uap, uid) != -1 || SCARG(uap, gid) != -1) &&
2070		    (suser(p->p_ucred, &p->p_acflag) || suid_clear)) {
2071			error = VOP_GETATTR(vp, &vattr, p->p_ucred, p);
2072			if (error)
2073				goto out;
2074			mode = vattr.va_mode & ~(VSUID | VSGID);
2075			if (mode == vattr.va_mode)
2076				mode = VNOVAL;
2077		} else
2078			mode = VNOVAL;
2079		VATTR_NULL(&vattr);
2080		vattr.va_uid = SCARG(uap, uid);
2081		vattr.va_gid = SCARG(uap, gid);
2082		vattr.va_mode = mode;
2083		error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
2084	}
2085out:
2086	VOP_UNLOCK(vp, 0, p);
2087	FRELE(fp);
2088	return (error);
2089}
2090
2091/*
2092 * Set the access and modification times given a path name.
2093 */
2094/* ARGSUSED */
2095int
2096sys_utimes(p, v, retval)
2097	struct proc *p;
2098	void *v;
2099	register_t *retval;
2100{
2101	register struct sys_utimes_args /* {
2102		syscallarg(char *) path;
2103		syscallarg(struct timeval *) tptr;
2104	} */ *uap = v;
2105	register struct vnode *vp;
2106	struct timeval tv[2];
2107	struct vattr vattr;
2108	int error;
2109	struct nameidata nd;
2110
2111	VATTR_NULL(&vattr);
2112	if (SCARG(uap, tptr) == NULL) {
2113		microtime(&tv[0]);
2114		tv[1] = tv[0];
2115		vattr.va_vaflags |= VA_UTIMES_NULL;
2116	} else {
2117		error = copyin((caddr_t)SCARG(uap, tptr), (caddr_t)tv,
2118		    sizeof (tv));
2119		if (error)
2120			return (error);
2121		/* XXX workaround timeval matching the VFS constant VNOVAL */
2122		if (tv[0].tv_sec == VNOVAL)
2123			tv[0].tv_sec = VNOVAL - 1;
2124		if (tv[1].tv_sec == VNOVAL)
2125			tv[1].tv_sec = VNOVAL - 1;
2126	}
2127	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2128	if ((error = namei(&nd)) != 0)
2129		return (error);
2130	vp = nd.ni_vp;
2131	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2132	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, p);
2133	if (vp->v_mount->mnt_flag & MNT_RDONLY)
2134		error = EROFS;
2135	else {
2136		vattr.va_atime.tv_sec = tv[0].tv_sec;
2137		vattr.va_atime.tv_nsec = tv[0].tv_usec * 1000;
2138		vattr.va_mtime.tv_sec = tv[1].tv_sec;
2139		vattr.va_mtime.tv_nsec = tv[1].tv_usec * 1000;
2140		error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
2141	}
2142	vput(vp);
2143	return (error);
2144}
2145
2146
2147/*
2148 * Set the access and modification times given a file descriptor.
2149 */
2150/* ARGSUSED */
2151int
2152sys_futimes(p, v, retval)
2153	struct proc *p;
2154	void *v;
2155	register_t *retval;
2156{
2157	register struct sys_futimes_args /* {
2158		syscallarg(int) fd;
2159		syscallarg(struct timeval *) tptr;
2160	} */ *uap = v;
2161	struct vnode *vp;
2162	struct timeval tv[2];
2163	struct vattr vattr;
2164	int error;
2165	struct file *fp;
2166
2167	VATTR_NULL(&vattr);
2168	if (SCARG(uap, tptr) == NULL) {
2169		microtime(&tv[0]);
2170		tv[1] = tv[0];
2171		vattr.va_vaflags |= VA_UTIMES_NULL;
2172	} else {
2173		error = copyin((caddr_t)SCARG(uap, tptr), (caddr_t)tv,
2174		    sizeof (tv));
2175		if (error)
2176			return (error);
2177		/* XXX workaround timeval matching the VFS constant VNOVAL */
2178		if (tv[0].tv_sec == VNOVAL)
2179			tv[0].tv_sec = VNOVAL - 1;
2180		if (tv[1].tv_sec == VNOVAL)
2181			tv[1].tv_sec = VNOVAL - 1;
2182	}
2183	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2184		return (error);
2185	vp = (struct vnode *)fp->f_data;
2186	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2187	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, p);
2188	if (vp->v_mount->mnt_flag & MNT_RDONLY)
2189		error = EROFS;
2190	else {
2191		vattr.va_atime.tv_sec = tv[0].tv_sec;
2192		vattr.va_atime.tv_nsec = tv[0].tv_usec * 1000;
2193		vattr.va_mtime.tv_sec = tv[1].tv_sec;
2194		vattr.va_mtime.tv_nsec = tv[1].tv_usec * 1000;
2195		error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
2196	}
2197	VOP_UNLOCK(vp, 0, p);
2198	FRELE(fp);
2199	return (error);
2200}
2201
2202/*
2203 * Truncate a file given its path name.
2204 */
2205/* ARGSUSED */
2206int
2207sys_truncate(p, v, retval)
2208	struct proc *p;
2209	void *v;
2210	register_t *retval;
2211{
2212	register struct sys_truncate_args /* {
2213		syscallarg(char *) path;
2214		syscallarg(int) pad;
2215		syscallarg(off_t) length;
2216	} */ *uap = v;
2217	register struct vnode *vp;
2218	struct vattr vattr;
2219	int error;
2220	struct nameidata nd;
2221
2222	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2223	if ((error = namei(&nd)) != 0)
2224		return (error);
2225	vp = nd.ni_vp;
2226	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2227	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, p);
2228	if (vp->v_type == VDIR)
2229		error = EISDIR;
2230	else if ((error = vn_writechk(vp)) == 0 &&
2231	    (error = VOP_ACCESS(vp, VWRITE, p->p_ucred, p)) == 0) {
2232		VATTR_NULL(&vattr);
2233		vattr.va_size = SCARG(uap, length);
2234		error = VOP_SETATTR(vp, &vattr, p->p_ucred, p);
2235	}
2236	vput(vp);
2237	return (error);
2238}
2239
2240/*
2241 * Truncate a file given a file descriptor.
2242 */
2243/* ARGSUSED */
2244int
2245sys_ftruncate(p, v, retval)
2246	struct proc *p;
2247	void *v;
2248	register_t *retval;
2249{
2250	struct sys_ftruncate_args /* {
2251		syscallarg(int) fd;
2252		syscallarg(int) pad;
2253		syscallarg(off_t) length;
2254	} */ *uap = v;
2255	struct vattr vattr;
2256	struct vnode *vp;
2257	struct file *fp;
2258	int error;
2259
2260	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2261		return (error);
2262	if ((fp->f_flag & FWRITE) == 0) {
2263		error = EINVAL;
2264		goto bad;
2265	}
2266	vp = (struct vnode *)fp->f_data;
2267	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2268	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, p);
2269	if (vp->v_type == VDIR)
2270		error = EISDIR;
2271	else if ((error = vn_writechk(vp)) == 0) {
2272		VATTR_NULL(&vattr);
2273		vattr.va_size = SCARG(uap, length);
2274		error = VOP_SETATTR(vp, &vattr, fp->f_cred, p);
2275	}
2276	VOP_UNLOCK(vp, 0, p);
2277bad:
2278	FRELE(fp);
2279	return (error);
2280}
2281
2282/*
2283 * Sync an open file.
2284 */
2285/* ARGSUSED */
2286int
2287sys_fsync(p, v, retval)
2288	struct proc *p;
2289	void *v;
2290	register_t *retval;
2291{
2292	struct sys_fsync_args /* {
2293		syscallarg(int) fd;
2294	} */ *uap = v;
2295	struct vnode *vp;
2296	struct file *fp;
2297	int error;
2298
2299	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2300		return (error);
2301	vp = (struct vnode *)fp->f_data;
2302	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, p);
2303	error = VOP_FSYNC(vp, fp->f_cred, MNT_WAIT, p);
2304#ifdef FFS_SOFTUPDATES
2305	if (error == 0 && vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP))
2306		error = softdep_fsync(vp);
2307#endif
2308
2309	VOP_UNLOCK(vp, 0, p);
2310	FRELE(fp);
2311	return (error);
2312}
2313
2314/*
2315 * Rename files.  Source and destination must either both be directories,
2316 * or both not be directories.  If target is a directory, it must be empty.
2317 */
2318/* ARGSUSED */
2319int
2320sys_rename(p, v, retval)
2321	struct proc *p;
2322	void *v;
2323	register_t *retval;
2324{
2325	register struct sys_rename_args /* {
2326		syscallarg(char *) from;
2327		syscallarg(char *) to;
2328	} */ *uap = v;
2329	register struct vnode *tvp, *fvp, *tdvp;
2330	struct nameidata fromnd, tond;
2331	int error;
2332	int flags;
2333
2334	NDINIT(&fromnd, DELETE, WANTPARENT | SAVESTART, UIO_USERSPACE,
2335	    SCARG(uap, from), p);
2336	if ((error = namei(&fromnd)) != 0)
2337		return (error);
2338	fvp = fromnd.ni_vp;
2339
2340	flags = LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART;
2341	/*
2342	 * rename("foo/", "bar/");  is  OK
2343	 */
2344	if (fvp->v_type == VDIR)
2345		flags |= STRIPSLASHES;
2346
2347	NDINIT(&tond, RENAME, flags,
2348	    UIO_USERSPACE, SCARG(uap, to), p);
2349	if ((error = namei(&tond)) != 0) {
2350		VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
2351		vrele(fromnd.ni_dvp);
2352		vrele(fvp);
2353		goto out1;
2354	}
2355	tdvp = tond.ni_dvp;
2356	tvp = tond.ni_vp;
2357	if (tvp != NULL) {
2358		if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
2359			error = ENOTDIR;
2360			goto out;
2361		} else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
2362			error = EISDIR;
2363			goto out;
2364		}
2365	}
2366	if (fvp == tdvp)
2367		error = EINVAL;
2368	/*
2369	 * If source is the same as the destination (that is the
2370	 * same inode number)
2371	 */
2372	if (fvp == tvp)
2373		error = -1;
2374out:
2375	if (!error) {
2376		VOP_LEASE(tdvp, p, p->p_ucred, LEASE_WRITE);
2377		if (fromnd.ni_dvp != tdvp)
2378			VOP_LEASE(fromnd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
2379		if (tvp) {
2380			(void)uvm_vnp_uncache(tvp);
2381			VOP_LEASE(tvp, p, p->p_ucred, LEASE_WRITE);
2382		}
2383		error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
2384				   tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
2385	} else {
2386		VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd);
2387		if (tdvp == tvp)
2388			vrele(tdvp);
2389		else
2390			vput(tdvp);
2391		if (tvp)
2392			vput(tvp);
2393		VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
2394		vrele(fromnd.ni_dvp);
2395		vrele(fvp);
2396	}
2397	vrele(tond.ni_startdir);
2398	FREE(tond.ni_cnd.cn_pnbuf, M_NAMEI);
2399out1:
2400	if (fromnd.ni_startdir)
2401		vrele(fromnd.ni_startdir);
2402	FREE(fromnd.ni_cnd.cn_pnbuf, M_NAMEI);
2403	if (error == -1)
2404		return (0);
2405	return (error);
2406}
2407
2408/*
2409 * Make a directory file.
2410 */
2411/* ARGSUSED */
2412int
2413sys_mkdir(p, v, retval)
2414	struct proc *p;
2415	void *v;
2416	register_t *retval;
2417{
2418	register struct sys_mkdir_args /* {
2419		syscallarg(char *) path;
2420		syscallarg(int) mode;
2421	} */ *uap = v;
2422	register struct vnode *vp;
2423	struct vattr vattr;
2424	int error;
2425	struct nameidata nd;
2426
2427	NDINIT(&nd, CREATE, LOCKPARENT | STRIPSLASHES,
2428	    UIO_USERSPACE, SCARG(uap, path), p);
2429	if ((error = namei(&nd)) != 0)
2430		return (error);
2431	vp = nd.ni_vp;
2432	if (vp != NULL) {
2433		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2434		if (nd.ni_dvp == vp)
2435			vrele(nd.ni_dvp);
2436		else
2437			vput(nd.ni_dvp);
2438		vrele(vp);
2439		return (EEXIST);
2440	}
2441	VATTR_NULL(&vattr);
2442	vattr.va_type = VDIR;
2443	vattr.va_mode = (SCARG(uap, mode) & ACCESSPERMS) &~ p->p_fd->fd_cmask;
2444	VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
2445	error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
2446	if (!error)
2447		vput(nd.ni_vp);
2448	return (error);
2449}
2450
2451/*
2452 * Remove a directory file.
2453 */
2454/* ARGSUSED */
2455int
2456sys_rmdir(p, v, retval)
2457	struct proc *p;
2458	void *v;
2459	register_t *retval;
2460{
2461	struct sys_rmdir_args /* {
2462		syscallarg(char *) path;
2463	} */ *uap = v;
2464	register struct vnode *vp;
2465	int error;
2466	struct nameidata nd;
2467
2468	NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF, UIO_USERSPACE,
2469	    SCARG(uap, path), p);
2470	if ((error = namei(&nd)) != 0)
2471		return (error);
2472	vp = nd.ni_vp;
2473	if (vp->v_type != VDIR) {
2474		error = ENOTDIR;
2475		goto out;
2476	}
2477	/*
2478	 * No rmdir "." please.
2479	 */
2480	if (nd.ni_dvp == vp) {
2481		error = EBUSY;
2482		goto out;
2483	}
2484	/*
2485	 * The root of a mounted filesystem cannot be deleted.
2486	 */
2487	if (vp->v_flag & VROOT)
2488		error = EBUSY;
2489out:
2490	if (!error) {
2491		VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
2492		VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2493		error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
2494	} else {
2495		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2496		if (nd.ni_dvp == vp)
2497			vrele(nd.ni_dvp);
2498		else
2499			vput(nd.ni_dvp);
2500		vput(vp);
2501	}
2502	return (error);
2503}
2504
2505/*
2506 * Read a block of directory entries in a file system independent format.
2507 */
2508int
2509sys_getdirentries(p, v, retval)
2510	struct proc *p;
2511	void *v;
2512	register_t *retval;
2513{
2514	struct sys_getdirentries_args /* {
2515		syscallarg(int) fd;
2516		syscallarg(char *) buf;
2517		syscallarg(int) count;
2518		syscallarg(long *) basep;
2519	} */ *uap = v;
2520	struct vnode *vp;
2521	struct file *fp;
2522	struct uio auio;
2523	struct iovec aiov;
2524	long loff;
2525	int error, eofflag;
2526
2527	if (SCARG(uap, count) < 0)
2528		return EINVAL;
2529	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
2530		return (error);
2531	if ((fp->f_flag & FREAD) == 0) {
2532		error = EBADF;
2533		goto bad;
2534	}
2535	vp = (struct vnode *)fp->f_data;
2536unionread:
2537	if (vp->v_type != VDIR) {
2538		error = EINVAL;
2539		goto bad;
2540	}
2541	aiov.iov_base = SCARG(uap, buf);
2542	aiov.iov_len = SCARG(uap, count);
2543	auio.uio_iov = &aiov;
2544	auio.uio_iovcnt = 1;
2545	auio.uio_rw = UIO_READ;
2546	auio.uio_segflg = UIO_USERSPACE;
2547	auio.uio_procp = p;
2548	auio.uio_resid = SCARG(uap, count);
2549	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, p);
2550	loff = auio.uio_offset = fp->f_offset;
2551	error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, 0, 0);
2552	fp->f_offset = auio.uio_offset;
2553	VOP_UNLOCK(vp, 0, p);
2554	if (error)
2555		goto bad;
2556	if ((SCARG(uap, count) == auio.uio_resid) &&
2557	    union_check_p &&
2558	    (union_check_p(p, &vp, fp, auio, &error) != 0))
2559		goto unionread;
2560	if (error)
2561		goto bad;
2562
2563	if ((SCARG(uap, count) == auio.uio_resid) &&
2564	    (vp->v_flag & VROOT) &&
2565	    (vp->v_mount->mnt_flag & MNT_UNION)) {
2566		struct vnode *tvp = vp;
2567		vp = vp->v_mount->mnt_vnodecovered;
2568		VREF(vp);
2569		fp->f_data = (caddr_t) vp;
2570		fp->f_offset = 0;
2571		vrele(tvp);
2572		goto unionread;
2573	}
2574	error = copyout((caddr_t)&loff, (caddr_t)SCARG(uap, basep),
2575	    sizeof(long));
2576	*retval = SCARG(uap, count) - auio.uio_resid;
2577bad:
2578	FRELE(fp);
2579	return (error);
2580}
2581
2582/*
2583 * Set the mode mask for creation of filesystem nodes.
2584 */
2585int
2586sys_umask(p, v, retval)
2587	struct proc *p;
2588	void *v;
2589	register_t *retval;
2590{
2591	struct sys_umask_args /* {
2592		syscallarg(int) newmask;
2593	} */ *uap = v;
2594	register struct filedesc *fdp;
2595
2596	fdp = p->p_fd;
2597	*retval = fdp->fd_cmask;
2598	fdp->fd_cmask = SCARG(uap, newmask) & ACCESSPERMS;
2599	return (0);
2600}
2601
2602/*
2603 * Void all references to file by ripping underlying filesystem
2604 * away from vnode.
2605 */
2606/* ARGSUSED */
2607int
2608sys_revoke(p, v, retval)
2609	struct proc *p;
2610	void *v;
2611	register_t *retval;
2612{
2613	register struct sys_revoke_args /* {
2614		syscallarg(char *) path;
2615	} */ *uap = v;
2616	register struct vnode *vp;
2617	struct vattr vattr;
2618	int error;
2619	struct nameidata nd;
2620
2621	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2622	if ((error = namei(&nd)) != 0)
2623		return (error);
2624	vp = nd.ni_vp;
2625	if ((error = VOP_GETATTR(vp, &vattr, p->p_ucred, p)) != 0)
2626		goto out;
2627	if (p->p_ucred->cr_uid != vattr.va_uid &&
2628	    (error = suser(p->p_ucred, &p->p_acflag)))
2629		goto out;
2630	if (vp->v_usecount > 1 || (vp->v_flag & (VALIASED | VLAYER)))
2631		VOP_REVOKE(vp, REVOKEALL);
2632out:
2633	vrele(vp);
2634	return (error);
2635}
2636
2637/*
2638 * Convert a user file descriptor to a kernel file entry.
2639 *
2640 * On return *fpp is FREF:ed.
2641 */
2642int
2643getvnode(fdp, fd, fpp)
2644	struct filedesc *fdp;
2645	struct file **fpp;
2646	int fd;
2647{
2648	struct file *fp;
2649
2650	if ((fp = fd_getfile(fdp, fd)) == NULL)
2651		return (EBADF);
2652	if (fp->f_type != DTYPE_VNODE)
2653		return (EINVAL);
2654	FREF(fp);
2655	*fpp = fp;
2656
2657	return (0);
2658}
2659
2660/*
2661 * Positional read system call.
2662 */
2663int
2664sys_pread(p, v, retval)
2665	struct proc *p;
2666	void *v;
2667	register_t *retval;
2668{
2669	struct sys_pread_args /* {
2670		syscallarg(int) fd;
2671		syscallarg(void *) buf;
2672		syscallarg(size_t) nbyte;
2673		syscallarg(int) pad;
2674		syscallarg(off_t) offset;
2675	} */ *uap = v;
2676	struct filedesc *fdp = p->p_fd;
2677	struct file *fp;
2678	struct vnode *vp;
2679	off_t offset;
2680	int fd = SCARG(uap, fd);
2681
2682	if ((fp = fd_getfile(fdp, fd)) == NULL)
2683		return (EBADF);
2684	if ((fp->f_flag & FREAD) == 0)
2685		return (EBADF);
2686
2687	vp = (struct vnode *)fp->f_data;
2688	if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
2689		return (ESPIPE);
2690	}
2691
2692	offset = SCARG(uap, offset);
2693
2694	FREF(fp);
2695
2696	/* dofileread() will FRELE the descriptor for us */
2697	return (dofileread(p, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte),
2698	    &offset, retval));
2699}
2700
2701/*
2702 * Positional scatter read system call.
2703 */
2704int
2705sys_preadv(p, v, retval)
2706	struct proc *p;
2707	void *v;
2708	register_t *retval;
2709{
2710	struct sys_preadv_args /* {
2711		syscallarg(int) fd;
2712		syscallarg(const struct iovec *) iovp;
2713		syscallarg(int) iovcnt;
2714		syscallarg(int) pad;
2715		syscallarg(off_t) offset;
2716	} */ *uap = v;
2717	struct filedesc *fdp = p->p_fd;
2718	struct file *fp;
2719	struct vnode *vp;
2720	off_t offset;
2721	int fd = SCARG(uap, fd);
2722
2723	if ((fp = fd_getfile(fdp, fd)) == NULL)
2724		return (EBADF);
2725	if ((fp->f_flag & FREAD) == 0)
2726		return (EBADF);
2727
2728	vp = (struct vnode *)fp->f_data;
2729	if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
2730		return (ESPIPE);
2731	}
2732
2733	FREF(fp);
2734
2735	offset = SCARG(uap, offset);
2736
2737	/* dofilereadv() will FRELE the descriptor for us */
2738	return (dofilereadv(p, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt),
2739	    &offset, retval));
2740}
2741
2742/*
2743 * Positional write system call.
2744 */
2745int
2746sys_pwrite(p, v, retval)
2747	struct proc *p;
2748	void *v;
2749	register_t *retval;
2750{
2751	struct sys_pwrite_args /* {
2752		syscallarg(int) fd;
2753		syscallarg(const void *) buf;
2754		syscallarg(size_t) nbyte;
2755		syscallarg(int) pad;
2756		syscallarg(off_t) offset;
2757	} */ *uap = v;
2758	struct filedesc *fdp = p->p_fd;
2759	struct file *fp;
2760	struct vnode *vp;
2761	off_t offset;
2762	int fd = SCARG(uap, fd);
2763
2764	if ((fp = fd_getfile(fdp, fd)) == NULL)
2765		return (EBADF);
2766	if ((fp->f_flag & FWRITE) == 0)
2767		return (EBADF);
2768
2769	vp = (struct vnode *)fp->f_data;
2770	if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
2771		return (ESPIPE);
2772	}
2773
2774	FREF(fp);
2775
2776	offset = SCARG(uap, offset);
2777
2778	/* dofilewrite() will FRELE the descriptor for us */
2779	return (dofilewrite(p, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte),
2780	    &offset, retval));
2781}
2782
2783
2784/*
2785 * Positional gather write system call.
2786 */
2787int
2788sys_pwritev(p, v, retval)
2789	struct proc *p;
2790	void *v;
2791	register_t *retval;
2792{
2793	struct sys_pwritev_args /* {
2794		syscallarg(int) fd;
2795		syscallarg(const struct iovec *) iovp;
2796		syscallarg(int) iovcnt;
2797		syscallarg(int) pad;
2798		syscallarg(off_t) offset;
2799	} */ *uap = v;
2800	struct filedesc *fdp = p->p_fd;
2801	struct file *fp;
2802	struct vnode *vp;
2803	off_t offset;
2804	int fd = SCARG(uap, fd);
2805
2806	if ((fp = fd_getfile(fdp, fd)) == NULL)
2807		return (EBADF);
2808	if ((fp->f_flag & FWRITE) == 0)
2809		return (EBADF);
2810
2811	vp = (struct vnode *)fp->f_data;
2812	if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
2813		return (ESPIPE);
2814	}
2815
2816	FREF(fp);
2817
2818	offset = SCARG(uap, offset);
2819
2820	/* dofilewritev() will FRELE the descriptor for us */
2821	return (dofilewritev(p, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt),
2822	    &offset, retval));
2823}
2824
2825#ifdef UFS_EXTATTR
2826/*
2827 * Syscall to push extended attribute configuration information into the
2828 * VFS.  Accepts a path, which it converts to a mountpoint, as well as
2829 * a command (int cmd), and attribute name and misc data.  For now, the
2830 * attribute name is left in userspace for consumption by the VFS_op.
2831 * It will probably be changed to be copied into sysspace by the
2832 * syscall in the future, once issues with various consumers of the
2833 * attribute code have raised their hands.
2834 *
2835 * Currently this is used only by UFS Extended Attributes.
2836 */
2837int
2838sys_extattrctl(struct proc *p, void *v, register_t *reval)
2839{
2840	struct sys_extattrctl_args /* {
2841		syscallarg(const char *) path;
2842		syscallarg(int) cmd;
2843		syscallarg(const char *) filename;
2844		syscallarg(int) attrnamespace;
2845		syscallarg(const char *) attrname;
2846	} */ *uap = v;
2847	struct vnode *filename_vp;
2848	struct nameidata nd;
2849	struct mount *mp;
2850	char attrname[EXTATTR_MAXNAMELEN];
2851	int error;
2852
2853	/*
2854	 * SCARG(uap, attrname) not always defined.  We check again later
2855	 * when we invoke the VFS call so as to pass in NULL there if needed.
2856	 */
2857	if (SCARG(uap, attrname) != NULL) {
2858		error = copyinstr(SCARG(uap, attrname), attrname,
2859		    EXTATTR_MAXNAMELEN, NULL);
2860		if (error)
2861			return (error);
2862	}
2863
2864	/*
2865	 * SCARG(uap, filename) not always defined.  If it is, grab
2866	 * a vnode lock, which VFS_EXTATTRCTL() will later release.
2867	 */
2868	filename_vp = NULL;
2869	if (SCARG(uap, filename) != NULL) {
2870		NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
2871		    SCARG(uap, filename), p);
2872		if ((error = namei(&nd)) != 0)
2873			return (error);
2874		filename_vp = nd.ni_vp;
2875	}
2876
2877	/* SCARG(uap, path) always defined. */
2878	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2879	if ((error = namei(&nd)) != 0) {
2880		if (filename_vp != NULL)
2881			vput(filename_vp);
2882		return (error);
2883	}
2884
2885	mp = nd.ni_vp->v_mount;
2886	if (error) {
2887		if (filename_vp != NULL)
2888			vput(filename_vp);
2889		return (error);
2890	}
2891
2892	if (SCARG(uap, attrname) != NULL) {
2893		error = VFS_EXTATTRCTL(mp, SCARG(uap, cmd), filename_vp,
2894		    SCARG(uap, attrnamespace), attrname, p);
2895	} else {
2896		error = VFS_EXTATTRCTL(mp, SCARG(uap, cmd), filename_vp,
2897		    SCARG(uap, attrnamespace), NULL, p);
2898	}
2899
2900	/*
2901	 * VFS_EXTATTRCTL will have unlocked, but not de-ref'd,
2902	 * filename_vp, so vrele it if it is defined.
2903	 */
2904	if (filename_vp != NULL)
2905		vrele(filename_vp);
2906
2907	return (error);
2908}
2909
2910/*-
2911 * Set a named extended attribute on a file or directory
2912 *
2913 * Arguments: unlocked vnode "vp", attribute namespace "attrnamespace",
2914 *            kernelspace string pointer "attrname", userspace buffer
2915 *            pointer "data", buffer length "nbytes", thread "td".
2916 * Returns: 0 on success, an error number otherwise
2917 * Locks: none
2918 * References: vp must be a valid reference for the duration of the call
2919 */
2920static int
2921extattr_set_vp(struct vnode *vp, int attrnamespace, const char *attrname,
2922    void *data, size_t nbytes, struct proc *p, register_t *retval)
2923{
2924	struct uio auio;
2925	struct iovec aiov;
2926	ssize_t cnt;
2927	int error;
2928
2929	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
2930	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, p);
2931
2932	aiov.iov_base = data;
2933	aiov.iov_len = nbytes;
2934	auio.uio_iov = &aiov;
2935	auio.uio_iovcnt = 1;
2936	auio.uio_offset = 0;
2937	if (nbytes > INT_MAX) {
2938		error = EINVAL;
2939		goto done;
2940	}
2941	auio.uio_resid = nbytes;
2942	auio.uio_rw = UIO_WRITE;
2943	auio.uio_segflg = UIO_USERSPACE;
2944	auio.uio_procp = p;
2945	cnt = nbytes;
2946
2947	error = VOP_SETEXTATTR(vp, attrnamespace, attrname, &auio,
2948	    p->p_ucred, p);
2949	cnt -= auio.uio_resid;
2950	retval[0] = cnt;
2951
2952done:
2953	VOP_UNLOCK(vp, 0, p);
2954	return (error);
2955}
2956
2957int
2958sys_extattr_set_file(struct proc *p, void *v, register_t *retval)
2959{
2960	struct sys_extattr_set_file_args /* {
2961		syscallarg(const char *) path;
2962		syscallarg(int) attrnamespace;
2963		syscallarg(const char *) attrname;
2964		syscallarg(void *) data;
2965		syscallarg(size_t) nbytes;
2966	} */ *uap = v;
2967	struct nameidata nd;
2968	char attrname[EXTATTR_MAXNAMELEN];
2969	int error;
2970
2971	error = copyinstr(SCARG(uap, attrname), attrname, EXTATTR_MAXNAMELEN,
2972	    NULL);
2973	if (error)
2974		return (error);
2975
2976	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
2977	if ((error = namei(&nd)) != 0)
2978		return (error);
2979
2980	error = extattr_set_vp(nd.ni_vp, SCARG(uap, attrnamespace), attrname,
2981	    SCARG(uap, data), SCARG(uap, nbytes), p, retval);
2982
2983	vrele(nd.ni_vp);
2984	return (error);
2985}
2986
2987int
2988sys_extattr_set_fd(struct proc *p, void *v, register_t *retval)
2989{
2990	struct sys_extattr_set_fd_args /* {
2991		syscallarg(int) fd;
2992		syscallarg(int) attrnamespace;
2993		syscallarg(const char *) attrname;
2994		syscallarg(struct iovec *) iovp;
2995		syscallarg(int) iovcnt;
2996	} */ *uap = v;
2997	struct file *fp;
2998	char attrname[EXTATTR_MAXNAMELEN];
2999	int error;
3000
3001	error = copyinstr(SCARG(uap, attrname), attrname, EXTATTR_MAXNAMELEN,
3002	    NULL);
3003	if (error)
3004		return (error);
3005
3006	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
3007		return (error);
3008
3009	error = extattr_set_vp((struct vnode *)fp->f_data,
3010	    SCARG(uap, attrnamespace), attrname, SCARG(uap, data),
3011	    SCARG(uap, nbytes), p, retval);
3012	FRELE(fp);
3013
3014	return (error);
3015}
3016
3017/*-
3018 * Get a named extended attribute on a file or directory
3019 *
3020 * Arguments: unlocked vnode "vp", attribute namespace "attrnamespace",
3021 *            kernelspace string pointer "attrname", userspace buffer
3022 *            pointer "data", buffer length "nbytes", thread "td".
3023 * Returns: 0 on success, an error number otherwise
3024 * Locks: none
3025 * References: vp must be a valid reference for the duration of the call
3026 */
3027static int
3028extattr_get_vp(struct vnode *vp, int attrnamespace, const char *attrname,
3029    void *data, size_t nbytes, struct proc *p, register_t *retval)
3030{
3031	struct uio auio;
3032	struct iovec aiov;
3033	ssize_t cnt;
3034	size_t size;
3035	int error;
3036
3037	VOP_LEASE(vp, p, p->p_ucred, LEASE_READ);
3038	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, p);
3039
3040	/*
3041	 * Slightly unusual semantics: if the user provides a NULL data
3042	 * pointer, they don't want to receive the data, just the
3043	 * maximum read length.
3044	 */
3045	if (data != NULL) {
3046		aiov.iov_base = data;
3047		aiov.iov_len = nbytes;
3048		auio.uio_iov = &aiov;
3049		auio.uio_offset = 0;
3050		if (nbytes > INT_MAX) {
3051			error = EINVAL;
3052			goto done;
3053		}
3054		auio.uio_resid = nbytes;
3055		auio.uio_rw = UIO_READ;
3056		auio.uio_segflg = UIO_USERSPACE;
3057		auio.uio_procp = p;
3058		cnt = nbytes;
3059		error = VOP_GETEXTATTR(vp, attrnamespace, attrname, &auio,
3060		    NULL, p->p_ucred, p);
3061		cnt -= auio.uio_resid;
3062		retval[0] = cnt;
3063	} else {
3064		error = VOP_GETEXTATTR(vp, attrnamespace, attrname, NULL,
3065		    &size, p->p_ucred, p);
3066		retval[0] = size;
3067	}
3068done:
3069	VOP_UNLOCK(vp, 0, p);
3070	return (error);
3071}
3072
3073int
3074sys_extattr_get_file(p, v, retval)
3075	struct proc *p;
3076	void *v;
3077	register_t *retval;
3078{
3079	struct sys_extattr_get_file_args /* {
3080		syscallarg(const char *) path;
3081		syscallarg(int) attrnamespace;
3082		syscallarg(const char *) attrname;
3083		syscallarg(void *) data;
3084		syscallarg(size_t) nbytes;
3085	} */ *uap = v;
3086	struct nameidata nd;
3087	char attrname[EXTATTR_MAXNAMELEN];
3088	int error;
3089
3090	error = copyinstr(SCARG(uap, attrname), attrname, EXTATTR_MAXNAMELEN,
3091	    NULL);
3092	if (error)
3093		return (error);
3094
3095	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
3096	if ((error = namei(&nd)) != 0)
3097		return (error);
3098
3099	error = extattr_get_vp(nd.ni_vp, SCARG(uap, attrnamespace), attrname,
3100	    SCARG(uap, data), SCARG(uap, nbytes), p, retval);
3101
3102	vrele(nd.ni_vp);
3103	return (error);
3104}
3105
3106int
3107sys_extattr_get_fd(p, v, retval)
3108	struct proc *p;
3109	void *v;
3110	register_t *retval;
3111{
3112	struct sys_extattr_get_fd_args /* {
3113		syscallarg(int) fd;
3114		syscallarg(int) attrnamespace;
3115		syscallarg(const char *) attrname;
3116		syscallarg(void *) data;
3117		syscallarg(size_t) nbytes;
3118	} */ *uap = v;
3119	struct file *fp;
3120	char attrname[EXTATTR_MAXNAMELEN];
3121	int error;
3122
3123	error = copyinstr(SCARG(uap, attrname), attrname, EXTATTR_MAXNAMELEN,
3124	    NULL);
3125	if (error)
3126		return (error);
3127
3128	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
3129		return (error);
3130
3131	error = extattr_get_vp((struct vnode *)fp->f_data,
3132	    SCARG(uap, attrnamespace), attrname, SCARG(uap, data),
3133	    SCARG(uap, nbytes), p, retval);
3134	FRELE(fp);
3135
3136	return (error);
3137}
3138
3139/*
3140 * extattr_delete_vp(): Delete a named extended attribute on a file or
3141 *                      directory
3142 *
3143 * Arguments: unlocked vnode "vp", attribute namespace "attrnamespace",
3144 *            kernelspace string pointer "attrname", proc "p"
3145 * Returns: 0 on success, an error number otherwise
3146 * Locks: none
3147 * References: vp must be a valid reference for the duration of the call
3148 */
3149static int
3150extattr_delete_vp(struct vnode *vp, int attrnamespace, const char *attrname,
3151    struct proc *p)
3152{
3153	int error;
3154
3155	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
3156	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, p);
3157
3158	error = VOP_SETEXTATTR(vp, attrnamespace, attrname, NULL,
3159	    p->p_ucred, p);
3160
3161	VOP_UNLOCK(vp, 0, p);
3162	return (error);
3163}
3164
3165int
3166sys_extattr_delete_file(p, v, retval)
3167	struct proc *p;
3168	void *v;
3169	register_t *retval;
3170{
3171	struct sys_extattr_delete_file_args /* {
3172		syscallarg(int) fd;
3173		syscallarg(int) attrnamespace;
3174		syscallarg(const char *) attrname;
3175	} */ *uap = v;
3176	struct nameidata nd;
3177	char attrname[EXTATTR_MAXNAMELEN];
3178	int error;
3179
3180	error = copyinstr(SCARG(uap, attrname), attrname, EXTATTR_MAXNAMELEN,
3181	     NULL);
3182	if (error)
3183		return(error);
3184
3185	NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
3186	if ((error = namei(&nd)) != 0)
3187		return(error);
3188
3189	error = extattr_delete_vp(nd.ni_vp, SCARG(uap, attrnamespace),
3190	    attrname, p);
3191
3192	vrele(nd.ni_vp);
3193	return(error);
3194}
3195
3196int
3197sys_extattr_delete_fd(p, v, retval)
3198	struct proc *p;
3199	void *v;
3200	register_t *retval;
3201{
3202	struct sys_extattr_delete_fd_args /* {
3203		syscallarg(int) fd;
3204		syscallarg(int) attrnamespace;
3205		syscallarg(const char *) attrname;
3206	} */ *uap = v;
3207	struct file *fp;
3208	char attrname[EXTATTR_MAXNAMELEN];
3209	int error;
3210
3211	error = copyinstr(SCARG(uap, attrname), attrname, EXTATTR_MAXNAMELEN,
3212	    NULL);
3213	if (error)
3214		return (error);
3215
3216	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
3217		return (error);
3218
3219	error = extattr_delete_vp((struct vnode *)fp->f_data,
3220	    SCARG(uap, attrnamespace), attrname, p);
3221	FRELE(fp);
3222
3223	return (error);
3224}
3225#endif
3226