vfs_default.c revision 204065 
1/*-
2 * Copyright (c) 1989, 1993
3 *	The Regents of the University of California.  All rights reserved.
4 *
5 * This code is derived from software contributed
6 * to Berkeley by John Heidemann of the UCLA Ficus project.
7 *
8 * Source: * @(#)i405_init.c 2.10 92/04/27 UCLA Ficus project
9 *
10 * Redistribution and use in source and binary forms, with or without
11 * modification, are permitted provided that the following conditions
12 * are met:
13 * 1. Redistributions of source code must retain the above copyright
14 *    notice, this list of conditions and the following disclaimer.
15 * 2. Redistributions in binary form must reproduce the above copyright
16 *    notice, this list of conditions and the following disclaimer in the
17 *    documentation and/or other materials provided with the distribution.
18 * 4. Neither the name of the University nor the names of its contributors
19 *    may be used to endorse or promote products derived from this software
20 *    without specific prior written permission.
21 *
22 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
23 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
24 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
25 * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
26 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
27 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
28 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
29 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
30 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
31 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32 * SUCH DAMAGE.
33 */
34
35#include <sys/cdefs.h>
36__FBSDID("$FreeBSD: head/sys/kern/vfs_default.c 204065 2010-02-18 22:14:44Z pjd $");
37
38#include <sys/param.h>
39#include <sys/systm.h>
40#include <sys/bio.h>
41#include <sys/buf.h>
42#include <sys/conf.h>
43#include <sys/event.h>
44#include <sys/kernel.h>
45#include <sys/limits.h>
46#include <sys/lock.h>
47#include <sys/lockf.h>
48#include <sys/malloc.h>
49#include <sys/mount.h>
50#include <sys/mutex.h>
51#include <sys/namei.h>
52#include <sys/fcntl.h>
53#include <sys/unistd.h>
54#include <sys/vnode.h>
55#include <sys/dirent.h>
56#include <sys/poll.h>
57
58#include <security/mac/mac_framework.h>
59
60#include <vm/vm.h>
61#include <vm/vm_object.h>
62#include <vm/vm_extern.h>
63#include <vm/pmap.h>
64#include <vm/vm_map.h>
65#include <vm/vm_page.h>
66#include <vm/vm_pager.h>
67#include <vm/vnode_pager.h>
68
69static int	vop_nolookup(struct vop_lookup_args *);
70static int	vop_nostrategy(struct vop_strategy_args *);
71static int	get_next_dirent(struct vnode *vp, struct dirent **dpp,
72				char *dirbuf, int dirbuflen, off_t *off,
73				char **cpos, int *len, int *eofflag,
74				struct thread *td);
75static int	dirent_exists(struct vnode *vp, const char *dirname,
76			      struct thread *td);
77
78#define DIRENT_MINSIZE (sizeof(struct dirent) - (MAXNAMLEN+1) + 4)
79
80/*
81 * This vnode table stores what we want to do if the filesystem doesn't
82 * implement a particular VOP.
83 *
84 * If there is no specific entry here, we will return EOPNOTSUPP.
85 *
86 * Note that every filesystem has to implement either vop_access
87 * or vop_accessx; failing to do so will result in immediate crash
88 * due to stack overflow, as vop_stdaccess() calls vop_stdaccessx(),
89 * which calls vop_stdaccess() etc.
90 */
91
92struct vop_vector default_vnodeops = {
93	.vop_default =		NULL,
94	.vop_bypass =		VOP_EOPNOTSUPP,
95
96	.vop_access =		vop_stdaccess,
97	.vop_accessx =		vop_stdaccessx,
98	.vop_advlock =		vop_stdadvlock,
99	.vop_advlockasync =	vop_stdadvlockasync,
100	.vop_bmap =		vop_stdbmap,
101	.vop_close =		VOP_NULL,
102	.vop_fsync =		VOP_NULL,
103	.vop_getpages =		vop_stdgetpages,
104	.vop_getwritemount = 	vop_stdgetwritemount,
105	.vop_inactive =		VOP_NULL,
106	.vop_ioctl =		VOP_ENOTTY,
107	.vop_kqfilter =		vop_stdkqfilter,
108	.vop_islocked =		vop_stdislocked,
109	.vop_lock1 =		vop_stdlock,
110	.vop_lookup =		vop_nolookup,
111	.vop_open =		VOP_NULL,
112	.vop_pathconf =		VOP_EINVAL,
113	.vop_poll =		vop_nopoll,
114	.vop_putpages =		vop_stdputpages,
115	.vop_readlink =		VOP_EINVAL,
116	.vop_revoke =		VOP_PANIC,
117	.vop_strategy =		vop_nostrategy,
118	.vop_unlock =		vop_stdunlock,
119	.vop_vptocnp =		vop_stdvptocnp,
120	.vop_vptofh =		vop_stdvptofh,
121};
122
123/*
124 * Series of placeholder functions for various error returns for
125 * VOPs.
126 */
127
128int
129vop_eopnotsupp(struct vop_generic_args *ap)
130{
131	/*
132	printf("vop_notsupp[%s]\n", ap->a_desc->vdesc_name);
133	*/
134
135	return (EOPNOTSUPP);
136}
137
138int
139vop_ebadf(struct vop_generic_args *ap)
140{
141
142	return (EBADF);
143}
144
145int
146vop_enotty(struct vop_generic_args *ap)
147{
148
149	return (ENOTTY);
150}
151
152int
153vop_einval(struct vop_generic_args *ap)
154{
155
156	return (EINVAL);
157}
158
159int
160vop_enoent(struct vop_generic_args *ap)
161{
162
163	return (ENOENT);
164}
165
166int
167vop_null(struct vop_generic_args *ap)
168{
169
170	return (0);
171}
172
173/*
174 * Helper function to panic on some bad VOPs in some filesystems.
175 */
176int
177vop_panic(struct vop_generic_args *ap)
178{
179
180	panic("filesystem goof: vop_panic[%s]", ap->a_desc->vdesc_name);
181}
182
183/*
184 * vop_std<something> and vop_no<something> are default functions for use by
185 * filesystems that need the "default reasonable" implementation for a
186 * particular operation.
187 *
188 * The documentation for the operations they implement exists (if it exists)
189 * in the VOP_<SOMETHING>(9) manpage (all uppercase).
190 */
191
192/*
193 * Default vop for filesystems that do not support name lookup
194 */
195static int
196vop_nolookup(ap)
197	struct vop_lookup_args /* {
198		struct vnode *a_dvp;
199		struct vnode **a_vpp;
200		struct componentname *a_cnp;
201	} */ *ap;
202{
203
204	*ap->a_vpp = NULL;
205	return (ENOTDIR);
206}
207
208/*
209 *	vop_nostrategy:
210 *
211 *	Strategy routine for VFS devices that have none.
212 *
213 *	BIO_ERROR and B_INVAL must be cleared prior to calling any strategy
214 *	routine.  Typically this is done for a BIO_READ strategy call.
215 *	Typically B_INVAL is assumed to already be clear prior to a write
216 *	and should not be cleared manually unless you just made the buffer
217 *	invalid.  BIO_ERROR should be cleared either way.
218 */
219
220static int
221vop_nostrategy (struct vop_strategy_args *ap)
222{
223	printf("No strategy for buffer at %p\n", ap->a_bp);
224	vprint("vnode", ap->a_vp);
225	ap->a_bp->b_ioflags |= BIO_ERROR;
226	ap->a_bp->b_error = EOPNOTSUPP;
227	bufdone(ap->a_bp);
228	return (EOPNOTSUPP);
229}
230
231static int
232get_next_dirent(struct vnode *vp, struct dirent **dpp, char *dirbuf,
233		int dirbuflen, off_t *off, char **cpos, int *len,
234		int *eofflag, struct thread *td)
235{
236	int error, reclen;
237	struct uio uio;
238	struct iovec iov;
239	struct dirent *dp;
240
241	KASSERT(VOP_ISLOCKED(vp), ("vp %p is not locked", vp));
242	KASSERT(vp->v_type == VDIR, ("vp %p is not a directory", vp));
243
244	if (*len == 0) {
245		iov.iov_base = dirbuf;
246		iov.iov_len = dirbuflen;
247
248		uio.uio_iov = &iov;
249		uio.uio_iovcnt = 1;
250		uio.uio_offset = *off;
251		uio.uio_resid = dirbuflen;
252		uio.uio_segflg = UIO_SYSSPACE;
253		uio.uio_rw = UIO_READ;
254		uio.uio_td = td;
255
256		*eofflag = 0;
257
258#ifdef MAC
259		error = mac_vnode_check_readdir(td->td_ucred, vp);
260		if (error == 0)
261#endif
262			error = VOP_READDIR(vp, &uio, td->td_ucred, eofflag,
263		    		NULL, NULL);
264		if (error)
265			return (error);
266
267		*off = uio.uio_offset;
268
269		*cpos = dirbuf;
270		*len = (dirbuflen - uio.uio_resid);
271	}
272
273	dp = (struct dirent *)(*cpos);
274	reclen = dp->d_reclen;
275	*dpp = dp;
276
277	/* check for malformed directory.. */
278	if (reclen < DIRENT_MINSIZE)
279		return (EINVAL);
280
281	*cpos += reclen;
282	*len -= reclen;
283
284	return (0);
285}
286
287/*
288 * Check if a named file exists in a given directory vnode.
289 */
290static int
291dirent_exists(struct vnode *vp, const char *dirname, struct thread *td)
292{
293	char *dirbuf, *cpos;
294	int error, eofflag, dirbuflen, len, found;
295	off_t off;
296	struct dirent *dp;
297	struct vattr va;
298
299	KASSERT(VOP_ISLOCKED(vp), ("vp %p is not locked", vp));
300	KASSERT(vp->v_type == VDIR, ("vp %p is not a directory", vp));
301
302	found = 0;
303
304	error = VOP_GETATTR(vp, &va, td->td_ucred);
305	if (error)
306		return (found);
307
308	dirbuflen = DEV_BSIZE;
309	if (dirbuflen < va.va_blocksize)
310		dirbuflen = va.va_blocksize;
311	dirbuf = (char *)malloc(dirbuflen, M_TEMP, M_WAITOK);
312
313	off = 0;
314	len = 0;
315	do {
316		error = get_next_dirent(vp, &dp, dirbuf, dirbuflen, &off,
317					&cpos, &len, &eofflag, td);
318		if (error)
319			goto out;
320
321		if ((dp->d_type != DT_WHT) &&
322		    !strcmp(dp->d_name, dirname)) {
323			found = 1;
324			goto out;
325		}
326	} while (len > 0 || !eofflag);
327
328out:
329	free(dirbuf, M_TEMP);
330	return (found);
331}
332
333int
334vop_stdaccess(struct vop_access_args *ap)
335{
336
337	KASSERT((ap->a_accmode & ~(VEXEC | VWRITE | VREAD | VADMIN |
338	    VAPPEND)) == 0, ("invalid bit in accmode"));
339
340	return (VOP_ACCESSX(ap->a_vp, ap->a_accmode, ap->a_cred, ap->a_td));
341}
342
343int
344vop_stdaccessx(struct vop_accessx_args *ap)
345{
346	int error;
347	accmode_t accmode = ap->a_accmode;
348
349	error = vfs_unixify_accmode(&accmode);
350	if (error != 0)
351		return (error);
352
353	if (accmode == 0)
354		return (0);
355
356	return (VOP_ACCESS(ap->a_vp, accmode, ap->a_cred, ap->a_td));
357}
358
359/*
360 * Advisory record locking support
361 */
362int
363vop_stdadvlock(struct vop_advlock_args *ap)
364{
365	struct vnode *vp;
366	struct ucred *cred;
367	struct vattr vattr;
368	int error;
369
370	vp = ap->a_vp;
371	cred = curthread->td_ucred;
372	vn_lock(vp, LK_SHARED | LK_RETRY);
373	error = VOP_GETATTR(vp, &vattr, cred);
374	VOP_UNLOCK(vp, 0);
375	if (error)
376		return (error);
377
378	return (lf_advlock(ap, &(vp->v_lockf), vattr.va_size));
379}
380
381int
382vop_stdadvlockasync(struct vop_advlockasync_args *ap)
383{
384	struct vnode *vp;
385	struct ucred *cred;
386	struct vattr vattr;
387	int error;
388
389	vp = ap->a_vp;
390	cred = curthread->td_ucred;
391	vn_lock(vp, LK_SHARED | LK_RETRY);
392	error = VOP_GETATTR(vp, &vattr, cred);
393	VOP_UNLOCK(vp, 0);
394	if (error)
395		return (error);
396
397	return (lf_advlockasync(ap, &(vp->v_lockf), vattr.va_size));
398}
399
400/*
401 * vop_stdpathconf:
402 *
403 * Standard implementation of POSIX pathconf, to get information about limits
404 * for a filesystem.
405 * Override per filesystem for the case where the filesystem has smaller
406 * limits.
407 */
408int
409vop_stdpathconf(ap)
410	struct vop_pathconf_args /* {
411	struct vnode *a_vp;
412	int a_name;
413	int *a_retval;
414	} */ *ap;
415{
416
417	switch (ap->a_name) {
418		case _PC_NAME_MAX:
419			*ap->a_retval = NAME_MAX;
420			return (0);
421		case _PC_PATH_MAX:
422			*ap->a_retval = PATH_MAX;
423			return (0);
424		case _PC_LINK_MAX:
425			*ap->a_retval = LINK_MAX;
426			return (0);
427		case _PC_MAX_CANON:
428			*ap->a_retval = MAX_CANON;
429			return (0);
430		case _PC_MAX_INPUT:
431			*ap->a_retval = MAX_INPUT;
432			return (0);
433		case _PC_PIPE_BUF:
434			*ap->a_retval = PIPE_BUF;
435			return (0);
436		case _PC_CHOWN_RESTRICTED:
437			*ap->a_retval = 1;
438			return (0);
439		case _PC_VDISABLE:
440			*ap->a_retval = _POSIX_VDISABLE;
441			return (0);
442		default:
443			return (EINVAL);
444	}
445	/* NOTREACHED */
446}
447
448/*
449 * Standard lock, unlock and islocked functions.
450 */
451int
452vop_stdlock(ap)
453	struct vop_lock1_args /* {
454		struct vnode *a_vp;
455		int a_flags;
456		char *file;
457		int line;
458	} */ *ap;
459{
460	struct vnode *vp = ap->a_vp;
461
462	return (_lockmgr_args(vp->v_vnlock, ap->a_flags, VI_MTX(vp),
463	    LK_WMESG_DEFAULT, LK_PRIO_DEFAULT, LK_TIMO_DEFAULT, ap->a_file,
464	    ap->a_line));
465}
466
467/* See above. */
468int
469vop_stdunlock(ap)
470	struct vop_unlock_args /* {
471		struct vnode *a_vp;
472		int a_flags;
473	} */ *ap;
474{
475	struct vnode *vp = ap->a_vp;
476
477	return (lockmgr(vp->v_vnlock, ap->a_flags | LK_RELEASE, VI_MTX(vp)));
478}
479
480/* See above. */
481int
482vop_stdislocked(ap)
483	struct vop_islocked_args /* {
484		struct vnode *a_vp;
485	} */ *ap;
486{
487
488	return (lockstatus(ap->a_vp->v_vnlock));
489}
490
491/*
492 * Return true for select/poll.
493 */
494int
495vop_nopoll(ap)
496	struct vop_poll_args /* {
497		struct vnode *a_vp;
498		int  a_events;
499		struct ucred *a_cred;
500		struct thread *a_td;
501	} */ *ap;
502{
503
504	return (poll_no_poll(ap->a_events));
505}
506
507/*
508 * Implement poll for local filesystems that support it.
509 */
510int
511vop_stdpoll(ap)
512	struct vop_poll_args /* {
513		struct vnode *a_vp;
514		int  a_events;
515		struct ucred *a_cred;
516		struct thread *a_td;
517	} */ *ap;
518{
519	if (ap->a_events & ~POLLSTANDARD)
520		return (vn_pollrecord(ap->a_vp, ap->a_td, ap->a_events));
521	return (ap->a_events & (POLLIN | POLLOUT | POLLRDNORM | POLLWRNORM));
522}
523
524/*
525 * Return our mount point, as we will take charge of the writes.
526 */
527int
528vop_stdgetwritemount(ap)
529	struct vop_getwritemount_args /* {
530		struct vnode *a_vp;
531		struct mount **a_mpp;
532	} */ *ap;
533{
534	struct mount *mp;
535
536	/*
537	 * XXX Since this is called unlocked we may be recycled while
538	 * attempting to ref the mount.  If this is the case or mountpoint
539	 * will be set to NULL.  We only have to prevent this call from
540	 * returning with a ref to an incorrect mountpoint.  It is not
541	 * harmful to return with a ref to our previous mountpoint.
542	 */
543	mp = ap->a_vp->v_mount;
544	if (mp != NULL) {
545		vfs_ref(mp);
546		if (mp != ap->a_vp->v_mount) {
547			vfs_rel(mp);
548			mp = NULL;
549		}
550	}
551	*(ap->a_mpp) = mp;
552	return (0);
553}
554
555/* XXX Needs good comment and VOP_BMAP(9) manpage */
556int
557vop_stdbmap(ap)
558	struct vop_bmap_args /* {
559		struct vnode *a_vp;
560		daddr_t  a_bn;
561		struct bufobj **a_bop;
562		daddr_t *a_bnp;
563		int *a_runp;
564		int *a_runb;
565	} */ *ap;
566{
567
568	if (ap->a_bop != NULL)
569		*ap->a_bop = &ap->a_vp->v_bufobj;
570	if (ap->a_bnp != NULL)
571		*ap->a_bnp = ap->a_bn * btodb(ap->a_vp->v_mount->mnt_stat.f_iosize);
572	if (ap->a_runp != NULL)
573		*ap->a_runp = 0;
574	if (ap->a_runb != NULL)
575		*ap->a_runb = 0;
576	return (0);
577}
578
579int
580vop_stdfsync(ap)
581	struct vop_fsync_args /* {
582		struct vnode *a_vp;
583		struct ucred *a_cred;
584		int a_waitfor;
585		struct thread *a_td;
586	} */ *ap;
587{
588	struct vnode *vp = ap->a_vp;
589	struct buf *bp;
590	struct bufobj *bo;
591	struct buf *nbp;
592	int error = 0;
593	int maxretry = 1000;     /* large, arbitrarily chosen */
594
595	bo = &vp->v_bufobj;
596	BO_LOCK(bo);
597loop1:
598	/*
599	 * MARK/SCAN initialization to avoid infinite loops.
600	 */
601        TAILQ_FOREACH(bp, &bo->bo_dirty.bv_hd, b_bobufs) {
602                bp->b_vflags &= ~BV_SCANNED;
603		bp->b_error = 0;
604	}
605
606	/*
607	 * Flush all dirty buffers associated with a vnode.
608	 */
609loop2:
610	TAILQ_FOREACH_SAFE(bp, &bo->bo_dirty.bv_hd, b_bobufs, nbp) {
611		if ((bp->b_vflags & BV_SCANNED) != 0)
612			continue;
613		bp->b_vflags |= BV_SCANNED;
614		if (BUF_LOCK(bp, LK_EXCLUSIVE | LK_NOWAIT, NULL))
615			continue;
616		BO_UNLOCK(bo);
617		KASSERT(bp->b_bufobj == bo,
618		    ("bp %p wrong b_bufobj %p should be %p",
619		    bp, bp->b_bufobj, bo));
620		if ((bp->b_flags & B_DELWRI) == 0)
621			panic("fsync: not dirty");
622		if ((vp->v_object != NULL) && (bp->b_flags & B_CLUSTEROK)) {
623			vfs_bio_awrite(bp);
624		} else {
625			bremfree(bp);
626			bawrite(bp);
627		}
628		BO_LOCK(bo);
629		goto loop2;
630	}
631
632	/*
633	 * If synchronous the caller expects us to completely resolve all
634	 * dirty buffers in the system.  Wait for in-progress I/O to
635	 * complete (which could include background bitmap writes), then
636	 * retry if dirty blocks still exist.
637	 */
638	if (ap->a_waitfor == MNT_WAIT) {
639		bufobj_wwait(bo, 0, 0);
640		if (bo->bo_dirty.bv_cnt > 0) {
641			/*
642			 * If we are unable to write any of these buffers
643			 * then we fail now rather than trying endlessly
644			 * to write them out.
645			 */
646			TAILQ_FOREACH(bp, &bo->bo_dirty.bv_hd, b_bobufs)
647				if ((error = bp->b_error) == 0)
648					continue;
649			if (error == 0 && --maxretry >= 0)
650				goto loop1;
651			error = EAGAIN;
652		}
653	}
654	BO_UNLOCK(bo);
655	if (error == EAGAIN)
656		vprint("fsync: giving up on dirty", vp);
657
658	return (error);
659}
660
661/* XXX Needs good comment and more info in the manpage (VOP_GETPAGES(9)). */
662int
663vop_stdgetpages(ap)
664	struct vop_getpages_args /* {
665		struct vnode *a_vp;
666		vm_page_t *a_m;
667		int a_count;
668		int a_reqpage;
669		vm_ooffset_t a_offset;
670	} */ *ap;
671{
672
673	return vnode_pager_generic_getpages(ap->a_vp, ap->a_m,
674	    ap->a_count, ap->a_reqpage);
675}
676
677int
678vop_stdkqfilter(struct vop_kqfilter_args *ap)
679{
680	return vfs_kqfilter(ap);
681}
682
683/* XXX Needs good comment and more info in the manpage (VOP_PUTPAGES(9)). */
684int
685vop_stdputpages(ap)
686	struct vop_putpages_args /* {
687		struct vnode *a_vp;
688		vm_page_t *a_m;
689		int a_count;
690		int a_sync;
691		int *a_rtvals;
692		vm_ooffset_t a_offset;
693	} */ *ap;
694{
695
696	return vnode_pager_generic_putpages(ap->a_vp, ap->a_m, ap->a_count,
697	     ap->a_sync, ap->a_rtvals);
698}
699
700int
701vop_stdvptofh(struct vop_vptofh_args *ap)
702{
703	return (EOPNOTSUPP);
704}
705
706int
707vop_stdvptocnp(struct vop_vptocnp_args *ap)
708{
709	struct vnode *vp = ap->a_vp;
710	struct vnode **dvp = ap->a_vpp;
711	struct ucred *cred = ap->a_cred;
712	char *buf = ap->a_buf;
713	int *buflen = ap->a_buflen;
714	char *dirbuf, *cpos;
715	int i, error, eofflag, dirbuflen, flags, locked, len, covered;
716	off_t off;
717	ino_t fileno;
718	struct vattr va;
719	struct nameidata nd;
720	struct thread *td;
721	struct dirent *dp;
722	struct vnode *mvp;
723
724	i = *buflen;
725	error = 0;
726	covered = 0;
727	td = curthread;
728
729	if (vp->v_type != VDIR)
730		return (ENOENT);
731
732	error = VOP_GETATTR(vp, &va, cred);
733	if (error)
734		return (error);
735
736	VREF(vp);
737	locked = VOP_ISLOCKED(vp);
738	VOP_UNLOCK(vp, 0);
739	NDINIT_ATVP(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE,
740	    "..", vp, td);
741	flags = FREAD;
742	error = vn_open_cred(&nd, &flags, 0, VN_OPEN_NOAUDIT, cred, NULL);
743	if (error) {
744		vn_lock(vp, locked | LK_RETRY);
745		return (error);
746	}
747	NDFREE(&nd, NDF_ONLY_PNBUF);
748
749	mvp = *dvp = nd.ni_vp;
750
751	if (vp->v_mount != (*dvp)->v_mount &&
752	    ((*dvp)->v_vflag & VV_ROOT) &&
753	    ((*dvp)->v_mount->mnt_flag & MNT_UNION)) {
754		*dvp = (*dvp)->v_mount->mnt_vnodecovered;
755		VREF(mvp);
756		VOP_UNLOCK(mvp, 0);
757		vn_close(mvp, FREAD, cred, td);
758		VREF(*dvp);
759		vn_lock(*dvp, LK_EXCLUSIVE | LK_RETRY);
760		covered = 1;
761	}
762
763	fileno = va.va_fileid;
764
765	dirbuflen = DEV_BSIZE;
766	if (dirbuflen < va.va_blocksize)
767		dirbuflen = va.va_blocksize;
768	dirbuf = (char *)malloc(dirbuflen, M_TEMP, M_WAITOK);
769
770	if ((*dvp)->v_type != VDIR) {
771		error = ENOENT;
772		goto out;
773	}
774
775	off = 0;
776	len = 0;
777	do {
778		/* call VOP_READDIR of parent */
779		error = get_next_dirent(*dvp, &dp, dirbuf, dirbuflen, &off,
780					&cpos, &len, &eofflag, td);
781		if (error)
782			goto out;
783
784		if ((dp->d_type != DT_WHT) &&
785		    (dp->d_fileno == fileno)) {
786			if (covered) {
787				VOP_UNLOCK(*dvp, 0);
788				vn_lock(mvp, LK_EXCLUSIVE | LK_RETRY);
789				if (dirent_exists(mvp, dp->d_name, td)) {
790					error = ENOENT;
791					VOP_UNLOCK(mvp, 0);
792					vn_lock(*dvp, LK_EXCLUSIVE | LK_RETRY);
793					goto out;
794				}
795				VOP_UNLOCK(mvp, 0);
796				vn_lock(*dvp, LK_EXCLUSIVE | LK_RETRY);
797			}
798			i -= dp->d_namlen;
799
800			if (i < 0) {
801				error = ENOMEM;
802				goto out;
803			}
804			bcopy(dp->d_name, buf + i, dp->d_namlen);
805			error = 0;
806			goto out;
807		}
808	} while (len > 0 || !eofflag);
809	error = ENOENT;
810
811out:
812	free(dirbuf, M_TEMP);
813	if (!error) {
814		*buflen = i;
815		vhold(*dvp);
816	}
817	if (covered) {
818		vput(*dvp);
819		vrele(mvp);
820	} else {
821		VOP_UNLOCK(mvp, 0);
822		vn_close(mvp, FREAD, cred, td);
823	}
824	vn_lock(vp, locked | LK_RETRY);
825	return (error);
826}
827
828/*
829 * vfs default ops
830 * used to fill the vfs function table to get reasonable default return values.
831 */
832int
833vfs_stdroot (mp, flags, vpp)
834	struct mount *mp;
835	int flags;
836	struct vnode **vpp;
837{
838
839	return (EOPNOTSUPP);
840}
841
842int
843vfs_stdstatfs (mp, sbp)
844	struct mount *mp;
845	struct statfs *sbp;
846{
847
848	return (EOPNOTSUPP);
849}
850
851int
852vfs_stdquotactl (mp, cmds, uid, arg)
853	struct mount *mp;
854	int cmds;
855	uid_t uid;
856	void *arg;
857{
858
859	return (EOPNOTSUPP);
860}
861
862int
863vfs_stdsync(mp, waitfor)
864	struct mount *mp;
865	int waitfor;
866{
867	struct vnode *vp, *mvp;
868	struct thread *td;
869	int error, lockreq, allerror = 0;
870
871	td = curthread;
872	lockreq = LK_EXCLUSIVE | LK_INTERLOCK;
873	if (waitfor != MNT_WAIT)
874		lockreq |= LK_NOWAIT;
875	/*
876	 * Force stale buffer cache information to be flushed.
877	 */
878	MNT_ILOCK(mp);
879loop:
880	MNT_VNODE_FOREACH(vp, mp, mvp) {
881		/* bv_cnt is an acceptable race here. */
882		if (vp->v_bufobj.bo_dirty.bv_cnt == 0)
883			continue;
884		VI_LOCK(vp);
885		MNT_IUNLOCK(mp);
886		if ((error = vget(vp, lockreq, td)) != 0) {
887			MNT_ILOCK(mp);
888			if (error == ENOENT) {
889				MNT_VNODE_FOREACH_ABORT_ILOCKED(mp, mvp);
890				goto loop;
891			}
892			continue;
893		}
894		error = VOP_FSYNC(vp, waitfor, td);
895		if (error)
896			allerror = error;
897		vput(vp);
898		MNT_ILOCK(mp);
899	}
900	MNT_IUNLOCK(mp);
901	return (allerror);
902}
903
904int
905vfs_stdnosync (mp, waitfor)
906	struct mount *mp;
907	int waitfor;
908{
909
910	return (0);
911}
912
913int
914vfs_stdvget (mp, ino, flags, vpp)
915	struct mount *mp;
916	ino_t ino;
917	int flags;
918	struct vnode **vpp;
919{
920
921	return (EOPNOTSUPP);
922}
923
924int
925vfs_stdfhtovp (mp, fhp, vpp)
926	struct mount *mp;
927	struct fid *fhp;
928	struct vnode **vpp;
929{
930
931	return (EOPNOTSUPP);
932}
933
934int
935vfs_stdinit (vfsp)
936	struct vfsconf *vfsp;
937{
938
939	return (0);
940}
941
942int
943vfs_stduninit (vfsp)
944	struct vfsconf *vfsp;
945{
946
947	return(0);
948}
949
950int
951vfs_stdextattrctl(mp, cmd, filename_vp, attrnamespace, attrname)
952	struct mount *mp;
953	int cmd;
954	struct vnode *filename_vp;
955	int attrnamespace;
956	const char *attrname;
957{
958
959	if (filename_vp != NULL)
960		VOP_UNLOCK(filename_vp, 0);
961	return (EOPNOTSUPP);
962}
963
964int
965vfs_stdsysctl(mp, op, req)
966	struct mount *mp;
967	fsctlop_t op;
968	struct sysctl_req *req;
969{
970
971	return (EOPNOTSUPP);
972}
973
974/* end of vfs default ops */
975