1/*-
2 * SPDX-License-Identifier: BSD-3-Clause
3 *
4 * Copyright (c) 1989, 1993
5 *	The Regents of the University of California.  All rights reserved.
6 *
7 * This code is derived from software contributed
8 * to Berkeley by John Heidemann of the UCLA Ficus project.
9 *
10 * Source: * @(#)i405_init.c 2.10 92/04/27 UCLA Ficus project
11 *
12 * Redistribution and use in source and binary forms, with or without
13 * modification, are permitted provided that the following conditions
14 * are met:
15 * 1. Redistributions of source code must retain the above copyright
16 *    notice, this list of conditions and the following disclaimer.
17 * 2. Redistributions in binary form must reproduce the above copyright
18 *    notice, this list of conditions and the following disclaimer in the
19 *    documentation and/or other materials provided with the distribution.
20 * 3. Neither the name of the University nor the names of its contributors
21 *    may be used to endorse or promote products derived from this software
22 *    without specific prior written permission.
23 *
24 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
25 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27 * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
28 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
29 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
30 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
31 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
32 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
33 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
34 * SUCH DAMAGE.
35 */
36
37#include <sys/cdefs.h>
38__FBSDID("$FreeBSD$");
39
40#include <sys/param.h>
41#include <sys/systm.h>
42#include <sys/bio.h>
43#include <sys/buf.h>
44#include <sys/conf.h>
45#include <sys/event.h>
46#include <sys/kernel.h>
47#include <sys/limits.h>
48#include <sys/lock.h>
49#include <sys/lockf.h>
50#include <sys/malloc.h>
51#include <sys/mount.h>
52#include <sys/namei.h>
53#include <sys/rwlock.h>
54#include <sys/fcntl.h>
55#include <sys/unistd.h>
56#include <sys/vnode.h>
57#include <sys/dirent.h>
58#include <sys/poll.h>
59
60#include <security/mac/mac_framework.h>
61
62#include <vm/vm.h>
63#include <vm/vm_object.h>
64#include <vm/vm_extern.h>
65#include <vm/pmap.h>
66#include <vm/vm_map.h>
67#include <vm/vm_page.h>
68#include <vm/vm_pager.h>
69#include <vm/vnode_pager.h>
70
71static int	vop_nolookup(struct vop_lookup_args *);
72static int	vop_norename(struct vop_rename_args *);
73static int	vop_nostrategy(struct vop_strategy_args *);
74static int	get_next_dirent(struct vnode *vp, struct dirent **dpp,
75				char *dirbuf, int dirbuflen, off_t *off,
76				char **cpos, int *len, int *eofflag,
77				struct thread *td);
78static int	dirent_exists(struct vnode *vp, const char *dirname,
79			      struct thread *td);
80
81#define DIRENT_MINSIZE (sizeof(struct dirent) - (MAXNAMLEN+1) + 4)
82
83static int vop_stdis_text(struct vop_is_text_args *ap);
84static int vop_stdunset_text(struct vop_unset_text_args *ap);
85static int vop_stdadd_writecount(struct vop_add_writecount_args *ap);
86static int vop_stdfdatasync(struct vop_fdatasync_args *ap);
87static int vop_stdgetpages_async(struct vop_getpages_async_args *ap);
88
89/*
90 * This vnode table stores what we want to do if the filesystem doesn't
91 * implement a particular VOP.
92 *
93 * If there is no specific entry here, we will return EOPNOTSUPP.
94 *
95 * Note that every filesystem has to implement either vop_access
96 * or vop_accessx; failing to do so will result in immediate crash
97 * due to stack overflow, as vop_stdaccess() calls vop_stdaccessx(),
98 * which calls vop_stdaccess() etc.
99 */
100
101struct vop_vector default_vnodeops = {
102	.vop_default =		NULL,
103	.vop_bypass =		VOP_EOPNOTSUPP,
104
105	.vop_access =		vop_stdaccess,
106	.vop_accessx =		vop_stdaccessx,
107	.vop_advise =		vop_stdadvise,
108	.vop_advlock =		vop_stdadvlock,
109	.vop_advlockasync =	vop_stdadvlockasync,
110	.vop_advlockpurge =	vop_stdadvlockpurge,
111	.vop_allocate =		vop_stdallocate,
112	.vop_bmap =		vop_stdbmap,
113	.vop_close =		VOP_NULL,
114	.vop_fsync =		VOP_NULL,
115	.vop_fdatasync =	vop_stdfdatasync,
116	.vop_getpages =		vop_stdgetpages,
117	.vop_getpages_async =	vop_stdgetpages_async,
118	.vop_getwritemount = 	vop_stdgetwritemount,
119	.vop_inactive =		VOP_NULL,
120	.vop_ioctl =		VOP_ENOTTY,
121	.vop_kqfilter =		vop_stdkqfilter,
122	.vop_islocked =		vop_stdislocked,
123	.vop_lock1 =		vop_stdlock,
124	.vop_lookup =		vop_nolookup,
125	.vop_open =		VOP_NULL,
126	.vop_pathconf =		VOP_EINVAL,
127	.vop_poll =		vop_nopoll,
128	.vop_putpages =		vop_stdputpages,
129	.vop_readlink =		VOP_EINVAL,
130	.vop_rename =		vop_norename,
131	.vop_revoke =		VOP_PANIC,
132	.vop_strategy =		vop_nostrategy,
133	.vop_unlock =		vop_stdunlock,
134	.vop_vptocnp =		vop_stdvptocnp,
135	.vop_vptofh =		vop_stdvptofh,
136	.vop_unp_bind =		vop_stdunp_bind,
137	.vop_unp_connect =	vop_stdunp_connect,
138	.vop_unp_detach =	vop_stdunp_detach,
139	.vop_is_text =		vop_stdis_text,
140	.vop_set_text =		vop_stdset_text,
141	.vop_unset_text =	vop_stdunset_text,
142	.vop_add_writecount =	vop_stdadd_writecount,
143};
144
145/*
146 * Series of placeholder functions for various error returns for
147 * VOPs.
148 */
149
150int
151vop_eopnotsupp(struct vop_generic_args *ap)
152{
153	/*
154	printf("vop_notsupp[%s]\n", ap->a_desc->vdesc_name);
155	*/
156
157	return (EOPNOTSUPP);
158}
159
160int
161vop_ebadf(struct vop_generic_args *ap)
162{
163
164	return (EBADF);
165}
166
167int
168vop_enotty(struct vop_generic_args *ap)
169{
170
171	return (ENOTTY);
172}
173
174int
175vop_einval(struct vop_generic_args *ap)
176{
177
178	return (EINVAL);
179}
180
181int
182vop_enoent(struct vop_generic_args *ap)
183{
184
185	return (ENOENT);
186}
187
188int
189vop_null(struct vop_generic_args *ap)
190{
191
192	return (0);
193}
194
195/*
196 * Helper function to panic on some bad VOPs in some filesystems.
197 */
198int
199vop_panic(struct vop_generic_args *ap)
200{
201
202	panic("filesystem goof: vop_panic[%s]", ap->a_desc->vdesc_name);
203}
204
205/*
206 * vop_std<something> and vop_no<something> are default functions for use by
207 * filesystems that need the "default reasonable" implementation for a
208 * particular operation.
209 *
210 * The documentation for the operations they implement exists (if it exists)
211 * in the VOP_<SOMETHING>(9) manpage (all uppercase).
212 */
213
214/*
215 * Default vop for filesystems that do not support name lookup
216 */
217static int
218vop_nolookup(ap)
219	struct vop_lookup_args /* {
220		struct vnode *a_dvp;
221		struct vnode **a_vpp;
222		struct componentname *a_cnp;
223	} */ *ap;
224{
225
226	*ap->a_vpp = NULL;
227	return (ENOTDIR);
228}
229
230/*
231 * vop_norename:
232 *
233 * Handle unlock and reference counting for arguments of vop_rename
234 * for filesystems that do not implement rename operation.
235 */
236static int
237vop_norename(struct vop_rename_args *ap)
238{
239
240	vop_rename_fail(ap);
241	return (EOPNOTSUPP);
242}
243
244/*
245 *	vop_nostrategy:
246 *
247 *	Strategy routine for VFS devices that have none.
248 *
249 *	BIO_ERROR and B_INVAL must be cleared prior to calling any strategy
250 *	routine.  Typically this is done for a BIO_READ strategy call.
251 *	Typically B_INVAL is assumed to already be clear prior to a write
252 *	and should not be cleared manually unless you just made the buffer
253 *	invalid.  BIO_ERROR should be cleared either way.
254 */
255
256static int
257vop_nostrategy (struct vop_strategy_args *ap)
258{
259	printf("No strategy for buffer at %p\n", ap->a_bp);
260	vn_printf(ap->a_vp, "vnode ");
261	ap->a_bp->b_ioflags |= BIO_ERROR;
262	ap->a_bp->b_error = EOPNOTSUPP;
263	bufdone(ap->a_bp);
264	return (EOPNOTSUPP);
265}
266
267static int
268get_next_dirent(struct vnode *vp, struct dirent **dpp, char *dirbuf,
269		int dirbuflen, off_t *off, char **cpos, int *len,
270		int *eofflag, struct thread *td)
271{
272	int error, reclen;
273	struct uio uio;
274	struct iovec iov;
275	struct dirent *dp;
276
277	KASSERT(VOP_ISLOCKED(vp), ("vp %p is not locked", vp));
278	KASSERT(vp->v_type == VDIR, ("vp %p is not a directory", vp));
279
280	if (*len == 0) {
281		iov.iov_base = dirbuf;
282		iov.iov_len = dirbuflen;
283
284		uio.uio_iov = &iov;
285		uio.uio_iovcnt = 1;
286		uio.uio_offset = *off;
287		uio.uio_resid = dirbuflen;
288		uio.uio_segflg = UIO_SYSSPACE;
289		uio.uio_rw = UIO_READ;
290		uio.uio_td = td;
291
292		*eofflag = 0;
293
294#ifdef MAC
295		error = mac_vnode_check_readdir(td->td_ucred, vp);
296		if (error == 0)
297#endif
298			error = VOP_READDIR(vp, &uio, td->td_ucred, eofflag,
299		    		NULL, NULL);
300		if (error)
301			return (error);
302
303		*off = uio.uio_offset;
304
305		*cpos = dirbuf;
306		*len = (dirbuflen - uio.uio_resid);
307
308		if (*len == 0)
309			return (ENOENT);
310	}
311
312	dp = (struct dirent *)(*cpos);
313	reclen = dp->d_reclen;
314	*dpp = dp;
315
316	/* check for malformed directory.. */
317	if (reclen < DIRENT_MINSIZE)
318		return (EINVAL);
319
320	*cpos += reclen;
321	*len -= reclen;
322
323	return (0);
324}
325
326/*
327 * Check if a named file exists in a given directory vnode.
328 */
329static int
330dirent_exists(struct vnode *vp, const char *dirname, struct thread *td)
331{
332	char *dirbuf, *cpos;
333	int error, eofflag, dirbuflen, len, found;
334	off_t off;
335	struct dirent *dp;
336	struct vattr va;
337
338	KASSERT(VOP_ISLOCKED(vp), ("vp %p is not locked", vp));
339	KASSERT(vp->v_type == VDIR, ("vp %p is not a directory", vp));
340
341	found = 0;
342
343	error = VOP_GETATTR(vp, &va, td->td_ucred);
344	if (error)
345		return (found);
346
347	dirbuflen = DEV_BSIZE;
348	if (dirbuflen < va.va_blocksize)
349		dirbuflen = va.va_blocksize;
350	dirbuf = (char *)malloc(dirbuflen, M_TEMP, M_WAITOK);
351
352	off = 0;
353	len = 0;
354	do {
355		error = get_next_dirent(vp, &dp, dirbuf, dirbuflen, &off,
356					&cpos, &len, &eofflag, td);
357		if (error)
358			goto out;
359
360		if (dp->d_type != DT_WHT && dp->d_fileno != 0 &&
361		    strcmp(dp->d_name, dirname) == 0) {
362			found = 1;
363			goto out;
364		}
365	} while (len > 0 || !eofflag);
366
367out:
368	free(dirbuf, M_TEMP);
369	return (found);
370}
371
372int
373vop_stdaccess(struct vop_access_args *ap)
374{
375
376	KASSERT((ap->a_accmode & ~(VEXEC | VWRITE | VREAD | VADMIN |
377	    VAPPEND)) == 0, ("invalid bit in accmode"));
378
379	return (VOP_ACCESSX(ap->a_vp, ap->a_accmode, ap->a_cred, ap->a_td));
380}
381
382int
383vop_stdaccessx(struct vop_accessx_args *ap)
384{
385	int error;
386	accmode_t accmode = ap->a_accmode;
387
388	error = vfs_unixify_accmode(&accmode);
389	if (error != 0)
390		return (error);
391
392	if (accmode == 0)
393		return (0);
394
395	return (VOP_ACCESS(ap->a_vp, accmode, ap->a_cred, ap->a_td));
396}
397
398/*
399 * Advisory record locking support
400 */
401int
402vop_stdadvlock(struct vop_advlock_args *ap)
403{
404	struct vnode *vp;
405	struct vattr vattr;
406	int error;
407
408	vp = ap->a_vp;
409	if (ap->a_fl->l_whence == SEEK_END) {
410		/*
411		 * The NFSv4 server must avoid doing a vn_lock() here, since it
412		 * can deadlock the nfsd threads, due to a LOR.  Fortunately
413		 * the NFSv4 server always uses SEEK_SET and this code is
414		 * only required for the SEEK_END case.
415		 */
416		vn_lock(vp, LK_SHARED | LK_RETRY);
417		error = VOP_GETATTR(vp, &vattr, curthread->td_ucred);
418		VOP_UNLOCK(vp, 0);
419		if (error)
420			return (error);
421	} else
422		vattr.va_size = 0;
423
424	return (lf_advlock(ap, &(vp->v_lockf), vattr.va_size));
425}
426
427int
428vop_stdadvlockasync(struct vop_advlockasync_args *ap)
429{
430	struct vnode *vp;
431	struct vattr vattr;
432	int error;
433
434	vp = ap->a_vp;
435	if (ap->a_fl->l_whence == SEEK_END) {
436		/* The size argument is only needed for SEEK_END. */
437		vn_lock(vp, LK_SHARED | LK_RETRY);
438		error = VOP_GETATTR(vp, &vattr, curthread->td_ucred);
439		VOP_UNLOCK(vp, 0);
440		if (error)
441			return (error);
442	} else
443		vattr.va_size = 0;
444
445	return (lf_advlockasync(ap, &(vp->v_lockf), vattr.va_size));
446}
447
448int
449vop_stdadvlockpurge(struct vop_advlockpurge_args *ap)
450{
451	struct vnode *vp;
452
453	vp = ap->a_vp;
454	lf_purgelocks(vp, &vp->v_lockf);
455	return (0);
456}
457
458/*
459 * vop_stdpathconf:
460 *
461 * Standard implementation of POSIX pathconf, to get information about limits
462 * for a filesystem.
463 * Override per filesystem for the case where the filesystem has smaller
464 * limits.
465 */
466int
467vop_stdpathconf(ap)
468	struct vop_pathconf_args /* {
469	struct vnode *a_vp;
470	int a_name;
471	int *a_retval;
472	} */ *ap;
473{
474
475	switch (ap->a_name) {
476		case _PC_ASYNC_IO:
477			*ap->a_retval = _POSIX_ASYNCHRONOUS_IO;
478			return (0);
479		case _PC_PATH_MAX:
480			*ap->a_retval = PATH_MAX;
481			return (0);
482		case _PC_ACL_EXTENDED:
483		case _PC_ACL_NFS4:
484		case _PC_CAP_PRESENT:
485		case _PC_INF_PRESENT:
486		case _PC_MAC_PRESENT:
487			*ap->a_retval = 0;
488			return (0);
489		default:
490			return (EINVAL);
491	}
492	/* NOTREACHED */
493}
494
495/*
496 * Standard lock, unlock and islocked functions.
497 */
498int
499vop_stdlock(ap)
500	struct vop_lock1_args /* {
501		struct vnode *a_vp;
502		int a_flags;
503		char *file;
504		int line;
505	} */ *ap;
506{
507	struct vnode *vp = ap->a_vp;
508	struct mtx *ilk;
509
510	ilk = VI_MTX(vp);
511	return (lockmgr_lock_fast_path(vp->v_vnlock, ap->a_flags,
512	    &ilk->lock_object, ap->a_file, ap->a_line));
513}
514
515/* See above. */
516int
517vop_stdunlock(ap)
518	struct vop_unlock_args /* {
519		struct vnode *a_vp;
520		int a_flags;
521	} */ *ap;
522{
523	struct vnode *vp = ap->a_vp;
524	struct mtx *ilk;
525
526	ilk = VI_MTX(vp);
527	return (lockmgr_unlock_fast_path(vp->v_vnlock, ap->a_flags,
528	    &ilk->lock_object));
529}
530
531/* See above. */
532int
533vop_stdislocked(ap)
534	struct vop_islocked_args /* {
535		struct vnode *a_vp;
536	} */ *ap;
537{
538
539	return (lockstatus(ap->a_vp->v_vnlock));
540}
541
542/*
543 * Return true for select/poll.
544 */
545int
546vop_nopoll(ap)
547	struct vop_poll_args /* {
548		struct vnode *a_vp;
549		int  a_events;
550		struct ucred *a_cred;
551		struct thread *a_td;
552	} */ *ap;
553{
554
555	return (poll_no_poll(ap->a_events));
556}
557
558/*
559 * Implement poll for local filesystems that support it.
560 */
561int
562vop_stdpoll(ap)
563	struct vop_poll_args /* {
564		struct vnode *a_vp;
565		int  a_events;
566		struct ucred *a_cred;
567		struct thread *a_td;
568	} */ *ap;
569{
570	if (ap->a_events & ~POLLSTANDARD)
571		return (vn_pollrecord(ap->a_vp, ap->a_td, ap->a_events));
572	return (ap->a_events & (POLLIN | POLLOUT | POLLRDNORM | POLLWRNORM));
573}
574
575/*
576 * Return our mount point, as we will take charge of the writes.
577 */
578int
579vop_stdgetwritemount(ap)
580	struct vop_getwritemount_args /* {
581		struct vnode *a_vp;
582		struct mount **a_mpp;
583	} */ *ap;
584{
585	struct mount *mp;
586
587	/*
588	 * XXX Since this is called unlocked we may be recycled while
589	 * attempting to ref the mount.  If this is the case or mountpoint
590	 * will be set to NULL.  We only have to prevent this call from
591	 * returning with a ref to an incorrect mountpoint.  It is not
592	 * harmful to return with a ref to our previous mountpoint.
593	 */
594	mp = ap->a_vp->v_mount;
595	if (mp != NULL) {
596		vfs_ref(mp);
597		if (mp != ap->a_vp->v_mount) {
598			vfs_rel(mp);
599			mp = NULL;
600		}
601	}
602	*(ap->a_mpp) = mp;
603	return (0);
604}
605
606/*
607 * If the file system doesn't implement VOP_BMAP, then return sensible defaults:
608 * - Return the vnode's bufobj instead of any underlying device's bufobj
609 * - Calculate the physical block number as if there were equal size
610 *   consecutive blocks, but
611 * - Report no contiguous runs of blocks.
612 */
613int
614vop_stdbmap(ap)
615	struct vop_bmap_args /* {
616		struct vnode *a_vp;
617		daddr_t  a_bn;
618		struct bufobj **a_bop;
619		daddr_t *a_bnp;
620		int *a_runp;
621		int *a_runb;
622	} */ *ap;
623{
624
625	if (ap->a_bop != NULL)
626		*ap->a_bop = &ap->a_vp->v_bufobj;
627	if (ap->a_bnp != NULL)
628		*ap->a_bnp = ap->a_bn * btodb(ap->a_vp->v_mount->mnt_stat.f_iosize);
629	if (ap->a_runp != NULL)
630		*ap->a_runp = 0;
631	if (ap->a_runb != NULL)
632		*ap->a_runb = 0;
633	return (0);
634}
635
636int
637vop_stdfsync(ap)
638	struct vop_fsync_args /* {
639		struct vnode *a_vp;
640		int a_waitfor;
641		struct thread *a_td;
642	} */ *ap;
643{
644
645	return (vn_fsync_buf(ap->a_vp, ap->a_waitfor));
646}
647
648static int
649vop_stdfdatasync(struct vop_fdatasync_args *ap)
650{
651
652	return (VOP_FSYNC(ap->a_vp, MNT_WAIT, ap->a_td));
653}
654
655int
656vop_stdfdatasync_buf(struct vop_fdatasync_args *ap)
657{
658
659	return (vn_fsync_buf(ap->a_vp, MNT_WAIT));
660}
661
662/* XXX Needs good comment and more info in the manpage (VOP_GETPAGES(9)). */
663int
664vop_stdgetpages(ap)
665	struct vop_getpages_args /* {
666		struct vnode *a_vp;
667		vm_page_t *a_m;
668		int a_count;
669		int *a_rbehind;
670		int *a_rahead;
671	} */ *ap;
672{
673
674	return vnode_pager_generic_getpages(ap->a_vp, ap->a_m,
675	    ap->a_count, ap->a_rbehind, ap->a_rahead, NULL, NULL);
676}
677
678static int
679vop_stdgetpages_async(struct vop_getpages_async_args *ap)
680{
681	int error;
682
683	error = VOP_GETPAGES(ap->a_vp, ap->a_m, ap->a_count, ap->a_rbehind,
684	    ap->a_rahead);
685	if (ap->a_iodone != NULL)
686		ap->a_iodone(ap->a_arg, ap->a_m, ap->a_count, error);
687	return (error);
688}
689
690int
691vop_stdkqfilter(struct vop_kqfilter_args *ap)
692{
693	return vfs_kqfilter(ap);
694}
695
696/* XXX Needs good comment and more info in the manpage (VOP_PUTPAGES(9)). */
697int
698vop_stdputpages(ap)
699	struct vop_putpages_args /* {
700		struct vnode *a_vp;
701		vm_page_t *a_m;
702		int a_count;
703		int a_sync;
704		int *a_rtvals;
705	} */ *ap;
706{
707
708	return vnode_pager_generic_putpages(ap->a_vp, ap->a_m, ap->a_count,
709	     ap->a_sync, ap->a_rtvals);
710}
711
712int
713vop_stdvptofh(struct vop_vptofh_args *ap)
714{
715	return (EOPNOTSUPP);
716}
717
718int
719vop_stdvptocnp(struct vop_vptocnp_args *ap)
720{
721	struct vnode *vp = ap->a_vp;
722	struct vnode **dvp = ap->a_vpp;
723	struct ucred *cred = ap->a_cred;
724	char *buf = ap->a_buf;
725	int *buflen = ap->a_buflen;
726	char *dirbuf, *cpos;
727	int i, error, eofflag, dirbuflen, flags, locked, len, covered;
728	off_t off;
729	ino_t fileno;
730	struct vattr va;
731	struct nameidata nd;
732	struct thread *td;
733	struct dirent *dp;
734	struct vnode *mvp;
735
736	i = *buflen;
737	error = 0;
738	covered = 0;
739	td = curthread;
740
741	if (vp->v_type != VDIR)
742		return (ENOENT);
743
744	error = VOP_GETATTR(vp, &va, cred);
745	if (error)
746		return (error);
747
748	VREF(vp);
749	locked = VOP_ISLOCKED(vp);
750	VOP_UNLOCK(vp, 0);
751	NDINIT_ATVP(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF, UIO_SYSSPACE,
752	    "..", vp, td);
753	flags = FREAD;
754	error = vn_open_cred(&nd, &flags, 0, VN_OPEN_NOAUDIT, cred, NULL);
755	if (error) {
756		vn_lock(vp, locked | LK_RETRY);
757		return (error);
758	}
759	NDFREE(&nd, NDF_ONLY_PNBUF);
760
761	mvp = *dvp = nd.ni_vp;
762
763	if (vp->v_mount != (*dvp)->v_mount &&
764	    ((*dvp)->v_vflag & VV_ROOT) &&
765	    ((*dvp)->v_mount->mnt_flag & MNT_UNION)) {
766		*dvp = (*dvp)->v_mount->mnt_vnodecovered;
767		VREF(mvp);
768		VOP_UNLOCK(mvp, 0);
769		vn_close(mvp, FREAD, cred, td);
770		VREF(*dvp);
771		vn_lock(*dvp, LK_SHARED | LK_RETRY);
772		covered = 1;
773	}
774
775	fileno = va.va_fileid;
776
777	dirbuflen = DEV_BSIZE;
778	if (dirbuflen < va.va_blocksize)
779		dirbuflen = va.va_blocksize;
780	dirbuf = (char *)malloc(dirbuflen, M_TEMP, M_WAITOK);
781
782	if ((*dvp)->v_type != VDIR) {
783		error = ENOENT;
784		goto out;
785	}
786
787	off = 0;
788	len = 0;
789	do {
790		/* call VOP_READDIR of parent */
791		error = get_next_dirent(*dvp, &dp, dirbuf, dirbuflen, &off,
792					&cpos, &len, &eofflag, td);
793		if (error)
794			goto out;
795
796		if ((dp->d_type != DT_WHT) &&
797		    (dp->d_fileno == fileno)) {
798			if (covered) {
799				VOP_UNLOCK(*dvp, 0);
800				vn_lock(mvp, LK_SHARED | LK_RETRY);
801				if (dirent_exists(mvp, dp->d_name, td)) {
802					error = ENOENT;
803					VOP_UNLOCK(mvp, 0);
804					vn_lock(*dvp, LK_SHARED | LK_RETRY);
805					goto out;
806				}
807				VOP_UNLOCK(mvp, 0);
808				vn_lock(*dvp, LK_SHARED | LK_RETRY);
809			}
810			i -= dp->d_namlen;
811
812			if (i < 0) {
813				error = ENOMEM;
814				goto out;
815			}
816			if (dp->d_namlen == 1 && dp->d_name[0] == '.') {
817				error = ENOENT;
818			} else {
819				bcopy(dp->d_name, buf + i, dp->d_namlen);
820				error = 0;
821			}
822			goto out;
823		}
824	} while (len > 0 || !eofflag);
825	error = ENOENT;
826
827out:
828	free(dirbuf, M_TEMP);
829	if (!error) {
830		*buflen = i;
831		vref(*dvp);
832	}
833	if (covered) {
834		vput(*dvp);
835		vrele(mvp);
836	} else {
837		VOP_UNLOCK(mvp, 0);
838		vn_close(mvp, FREAD, cred, td);
839	}
840	vn_lock(vp, locked | LK_RETRY);
841	return (error);
842}
843
844int
845vop_stdallocate(struct vop_allocate_args *ap)
846{
847#ifdef __notyet__
848	struct statfs *sfs;
849	off_t maxfilesize = 0;
850#endif
851	struct iovec aiov;
852	struct vattr vattr, *vap;
853	struct uio auio;
854	off_t fsize, len, cur, offset;
855	uint8_t *buf;
856	struct thread *td;
857	struct vnode *vp;
858	size_t iosize;
859	int error;
860
861	buf = NULL;
862	error = 0;
863	td = curthread;
864	vap = &vattr;
865	vp = ap->a_vp;
866	len = *ap->a_len;
867	offset = *ap->a_offset;
868
869	error = VOP_GETATTR(vp, vap, td->td_ucred);
870	if (error != 0)
871		goto out;
872	fsize = vap->va_size;
873	iosize = vap->va_blocksize;
874	if (iosize == 0)
875		iosize = BLKDEV_IOSIZE;
876	if (iosize > MAXPHYS)
877		iosize = MAXPHYS;
878	buf = malloc(iosize, M_TEMP, M_WAITOK);
879
880#ifdef __notyet__
881	/*
882	 * Check if the filesystem sets f_maxfilesize; if not use
883	 * VOP_SETATTR to perform the check.
884	 */
885	sfs = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
886	error = VFS_STATFS(vp->v_mount, sfs, td);
887	if (error == 0)
888		maxfilesize = sfs->f_maxfilesize;
889	free(sfs, M_STATFS);
890	if (error != 0)
891		goto out;
892	if (maxfilesize) {
893		if (offset > maxfilesize || len > maxfilesize ||
894		    offset + len > maxfilesize) {
895			error = EFBIG;
896			goto out;
897		}
898	} else
899#endif
900	if (offset + len > vap->va_size) {
901		/*
902		 * Test offset + len against the filesystem's maxfilesize.
903		 */
904		VATTR_NULL(vap);
905		vap->va_size = offset + len;
906		error = VOP_SETATTR(vp, vap, td->td_ucred);
907		if (error != 0)
908			goto out;
909		VATTR_NULL(vap);
910		vap->va_size = fsize;
911		error = VOP_SETATTR(vp, vap, td->td_ucred);
912		if (error != 0)
913			goto out;
914	}
915
916	for (;;) {
917		/*
918		 * Read and write back anything below the nominal file
919		 * size.  There's currently no way outside the filesystem
920		 * to know whether this area is sparse or not.
921		 */
922		cur = iosize;
923		if ((offset % iosize) != 0)
924			cur -= (offset % iosize);
925		if (cur > len)
926			cur = len;
927		if (offset < fsize) {
928			aiov.iov_base = buf;
929			aiov.iov_len = cur;
930			auio.uio_iov = &aiov;
931			auio.uio_iovcnt = 1;
932			auio.uio_offset = offset;
933			auio.uio_resid = cur;
934			auio.uio_segflg = UIO_SYSSPACE;
935			auio.uio_rw = UIO_READ;
936			auio.uio_td = td;
937			error = VOP_READ(vp, &auio, 0, td->td_ucred);
938			if (error != 0)
939				break;
940			if (auio.uio_resid > 0) {
941				bzero(buf + cur - auio.uio_resid,
942				    auio.uio_resid);
943			}
944		} else {
945			bzero(buf, cur);
946		}
947
948		aiov.iov_base = buf;
949		aiov.iov_len = cur;
950		auio.uio_iov = &aiov;
951		auio.uio_iovcnt = 1;
952		auio.uio_offset = offset;
953		auio.uio_resid = cur;
954		auio.uio_segflg = UIO_SYSSPACE;
955		auio.uio_rw = UIO_WRITE;
956		auio.uio_td = td;
957
958		error = VOP_WRITE(vp, &auio, 0, td->td_ucred);
959		if (error != 0)
960			break;
961
962		len -= cur;
963		offset += cur;
964		if (len == 0)
965			break;
966		if (should_yield())
967			break;
968	}
969
970 out:
971	*ap->a_len = len;
972	*ap->a_offset = offset;
973	free(buf, M_TEMP);
974	return (error);
975}
976
977int
978vop_stdadvise(struct vop_advise_args *ap)
979{
980	struct vnode *vp;
981	struct bufobj *bo;
982	daddr_t startn, endn;
983	off_t start, end;
984	int bsize, error;
985
986	vp = ap->a_vp;
987	switch (ap->a_advice) {
988	case POSIX_FADV_WILLNEED:
989		/*
990		 * Do nothing for now.  Filesystems should provide a
991		 * custom method which starts an asynchronous read of
992		 * the requested region.
993		 */
994		error = 0;
995		break;
996	case POSIX_FADV_DONTNEED:
997		error = 0;
998		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
999		if (vp->v_iflag & VI_DOOMED) {
1000			VOP_UNLOCK(vp, 0);
1001			break;
1002		}
1003
1004		/*
1005		 * Deactivate pages in the specified range from the backing VM
1006		 * object.  Pages that are resident in the buffer cache will
1007		 * remain wired until their corresponding buffers are released
1008		 * below.
1009		 */
1010		if (vp->v_object != NULL) {
1011			start = trunc_page(ap->a_start);
1012			end = round_page(ap->a_end);
1013			VM_OBJECT_RLOCK(vp->v_object);
1014			vm_object_page_noreuse(vp->v_object, OFF_TO_IDX(start),
1015			    OFF_TO_IDX(end));
1016			VM_OBJECT_RUNLOCK(vp->v_object);
1017		}
1018
1019		bo = &vp->v_bufobj;
1020		BO_RLOCK(bo);
1021		bsize = vp->v_bufobj.bo_bsize;
1022		startn = ap->a_start / bsize;
1023		endn = ap->a_end / bsize;
1024		error = bnoreuselist(&bo->bo_clean, bo, startn, endn);
1025		if (error == 0)
1026			error = bnoreuselist(&bo->bo_dirty, bo, startn, endn);
1027		BO_RUNLOCK(bo);
1028		VOP_UNLOCK(vp, 0);
1029		break;
1030	default:
1031		error = EINVAL;
1032		break;
1033	}
1034	return (error);
1035}
1036
1037int
1038vop_stdunp_bind(struct vop_unp_bind_args *ap)
1039{
1040
1041	ap->a_vp->v_unpcb = ap->a_unpcb;
1042	return (0);
1043}
1044
1045int
1046vop_stdunp_connect(struct vop_unp_connect_args *ap)
1047{
1048
1049	*ap->a_unpcb = ap->a_vp->v_unpcb;
1050	return (0);
1051}
1052
1053int
1054vop_stdunp_detach(struct vop_unp_detach_args *ap)
1055{
1056
1057	ap->a_vp->v_unpcb = NULL;
1058	return (0);
1059}
1060
1061static int
1062vop_stdis_text(struct vop_is_text_args *ap)
1063{
1064
1065	return (ap->a_vp->v_writecount < 0);
1066}
1067
1068int
1069vop_stdset_text(struct vop_set_text_args *ap)
1070{
1071	struct vnode *vp;
1072	struct mount *mp;
1073	int error;
1074
1075	vp = ap->a_vp;
1076	VI_LOCK(vp);
1077	if (vp->v_writecount > 0) {
1078		error = ETXTBSY;
1079	} else {
1080		/*
1081		 * If requested by fs, keep a use reference to the
1082		 * vnode until the last text reference is released.
1083		 */
1084		mp = vp->v_mount;
1085		if (mp != NULL && (mp->mnt_kern_flag & MNTK_TEXT_REFS) != 0 &&
1086		    vp->v_writecount == 0) {
1087			vp->v_iflag |= VI_TEXT_REF;
1088			vrefl(vp);
1089		}
1090
1091		vp->v_writecount--;
1092		error = 0;
1093	}
1094	VI_UNLOCK(vp);
1095	return (error);
1096}
1097
1098static int
1099vop_stdunset_text(struct vop_unset_text_args *ap)
1100{
1101	struct vnode *vp;
1102	int error;
1103	bool last;
1104
1105	vp = ap->a_vp;
1106	last = false;
1107	VI_LOCK(vp);
1108	if (vp->v_writecount < 0) {
1109		if ((vp->v_iflag & VI_TEXT_REF) != 0 &&
1110		    vp->v_writecount == -1) {
1111			last = true;
1112			vp->v_iflag &= ~VI_TEXT_REF;
1113		}
1114		vp->v_writecount++;
1115		error = 0;
1116	} else {
1117		error = EINVAL;
1118	}
1119	VI_UNLOCK(vp);
1120	if (last)
1121		vunref(vp);
1122	return (error);
1123}
1124
1125static int
1126vop_stdadd_writecount(struct vop_add_writecount_args *ap)
1127{
1128	struct vnode *vp;
1129	int error;
1130
1131	vp = ap->a_vp;
1132	VI_LOCK_FLAGS(vp, MTX_DUPOK);
1133	if (vp->v_writecount < 0) {
1134		error = ETXTBSY;
1135	} else {
1136		VNASSERT(vp->v_writecount + ap->a_inc >= 0, vp,
1137		    ("neg writecount increment %d", ap->a_inc));
1138		vp->v_writecount += ap->a_inc;
1139		error = 0;
1140	}
1141	VI_UNLOCK(vp);
1142	return (error);
1143}
1144
1145/*
1146 * vfs default ops
1147 * used to fill the vfs function table to get reasonable default return values.
1148 */
1149int
1150vfs_stdroot (mp, flags, vpp)
1151	struct mount *mp;
1152	int flags;
1153	struct vnode **vpp;
1154{
1155
1156	return (EOPNOTSUPP);
1157}
1158
1159int
1160vfs_stdstatfs (mp, sbp)
1161	struct mount *mp;
1162	struct statfs *sbp;
1163{
1164
1165	return (EOPNOTSUPP);
1166}
1167
1168int
1169vfs_stdquotactl (mp, cmds, uid, arg)
1170	struct mount *mp;
1171	int cmds;
1172	uid_t uid;
1173	void *arg;
1174{
1175
1176	return (EOPNOTSUPP);
1177}
1178
1179int
1180vfs_stdsync(mp, waitfor)
1181	struct mount *mp;
1182	int waitfor;
1183{
1184	struct vnode *vp, *mvp;
1185	struct thread *td;
1186	int error, lockreq, allerror = 0;
1187
1188	td = curthread;
1189	lockreq = LK_EXCLUSIVE | LK_INTERLOCK;
1190	if (waitfor != MNT_WAIT)
1191		lockreq |= LK_NOWAIT;
1192	/*
1193	 * Force stale buffer cache information to be flushed.
1194	 */
1195loop:
1196	MNT_VNODE_FOREACH_ALL(vp, mp, mvp) {
1197		if (vp->v_bufobj.bo_dirty.bv_cnt == 0) {
1198			VI_UNLOCK(vp);
1199			continue;
1200		}
1201		if ((error = vget(vp, lockreq, td)) != 0) {
1202			if (error == ENOENT) {
1203				MNT_VNODE_FOREACH_ALL_ABORT(mp, mvp);
1204				goto loop;
1205			}
1206			continue;
1207		}
1208		error = VOP_FSYNC(vp, waitfor, td);
1209		if (error)
1210			allerror = error;
1211		vput(vp);
1212	}
1213	return (allerror);
1214}
1215
1216int
1217vfs_stdnosync (mp, waitfor)
1218	struct mount *mp;
1219	int waitfor;
1220{
1221
1222	return (0);
1223}
1224
1225int
1226vfs_stdvget (mp, ino, flags, vpp)
1227	struct mount *mp;
1228	ino_t ino;
1229	int flags;
1230	struct vnode **vpp;
1231{
1232
1233	return (EOPNOTSUPP);
1234}
1235
1236int
1237vfs_stdfhtovp (mp, fhp, flags, vpp)
1238	struct mount *mp;
1239	struct fid *fhp;
1240	int flags;
1241	struct vnode **vpp;
1242{
1243
1244	return (EOPNOTSUPP);
1245}
1246
1247int
1248vfs_stdinit (vfsp)
1249	struct vfsconf *vfsp;
1250{
1251
1252	return (0);
1253}
1254
1255int
1256vfs_stduninit (vfsp)
1257	struct vfsconf *vfsp;
1258{
1259
1260	return(0);
1261}
1262
1263int
1264vfs_stdextattrctl(mp, cmd, filename_vp, attrnamespace, attrname)
1265	struct mount *mp;
1266	int cmd;
1267	struct vnode *filename_vp;
1268	int attrnamespace;
1269	const char *attrname;
1270{
1271
1272	if (filename_vp != NULL)
1273		VOP_UNLOCK(filename_vp, 0);
1274	return (EOPNOTSUPP);
1275}
1276
1277int
1278vfs_stdsysctl(mp, op, req)
1279	struct mount *mp;
1280	fsctlop_t op;
1281	struct sysctl_req *req;
1282{
1283
1284	return (EOPNOTSUPP);
1285}
1286
1287/* end of vfs default ops */
1288