xref: /freebsd/sys/kern/vfs_default.c (revision 38f0b757fd84d17d0fc24739a7cda160c4516d81)
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("$FreeBSD$");
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/namei.h>
51 #include <sys/rwlock.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 
69 static int	vop_nolookup(struct vop_lookup_args *);
70 static int	vop_norename(struct vop_rename_args *);
71 static int	vop_nostrategy(struct vop_strategy_args *);
72 static int	get_next_dirent(struct vnode *vp, struct dirent **dpp,
73 				char *dirbuf, int dirbuflen, off_t *off,
74 				char **cpos, int *len, int *eofflag,
75 				struct thread *td);
76 static int	dirent_exists(struct vnode *vp, const char *dirname,
77 			      struct thread *td);
78 
79 #define DIRENT_MINSIZE (sizeof(struct dirent) - (MAXNAMLEN+1) + 4)
80 
81 static int vop_stdis_text(struct vop_is_text_args *ap);
82 static int vop_stdset_text(struct vop_set_text_args *ap);
83 static int vop_stdunset_text(struct vop_unset_text_args *ap);
84 static int vop_stdget_writecount(struct vop_get_writecount_args *ap);
85 static int vop_stdadd_writecount(struct vop_add_writecount_args *ap);
86 
87 /*
88  * This vnode table stores what we want to do if the filesystem doesn't
89  * implement a particular VOP.
90  *
91  * If there is no specific entry here, we will return EOPNOTSUPP.
92  *
93  * Note that every filesystem has to implement either vop_access
94  * or vop_accessx; failing to do so will result in immediate crash
95  * due to stack overflow, as vop_stdaccess() calls vop_stdaccessx(),
96  * which calls vop_stdaccess() etc.
97  */
98 
99 struct vop_vector default_vnodeops = {
100 	.vop_default =		NULL,
101 	.vop_bypass =		VOP_EOPNOTSUPP,
102 
103 	.vop_access =		vop_stdaccess,
104 	.vop_accessx =		vop_stdaccessx,
105 	.vop_advise =		vop_stdadvise,
106 	.vop_advlock =		vop_stdadvlock,
107 	.vop_advlockasync =	vop_stdadvlockasync,
108 	.vop_advlockpurge =	vop_stdadvlockpurge,
109 	.vop_allocate =		vop_stdallocate,
110 	.vop_bmap =		vop_stdbmap,
111 	.vop_close =		VOP_NULL,
112 	.vop_fsync =		VOP_NULL,
113 	.vop_getpages =		vop_stdgetpages,
114 	.vop_getwritemount = 	vop_stdgetwritemount,
115 	.vop_inactive =		VOP_NULL,
116 	.vop_ioctl =		VOP_ENOTTY,
117 	.vop_kqfilter =		vop_stdkqfilter,
118 	.vop_islocked =		vop_stdislocked,
119 	.vop_lock1 =		vop_stdlock,
120 	.vop_lookup =		vop_nolookup,
121 	.vop_open =		VOP_NULL,
122 	.vop_pathconf =		VOP_EINVAL,
123 	.vop_poll =		vop_nopoll,
124 	.vop_putpages =		vop_stdputpages,
125 	.vop_readlink =		VOP_EINVAL,
126 	.vop_rename =		vop_norename,
127 	.vop_revoke =		VOP_PANIC,
128 	.vop_strategy =		vop_nostrategy,
129 	.vop_unlock =		vop_stdunlock,
130 	.vop_vptocnp =		vop_stdvptocnp,
131 	.vop_vptofh =		vop_stdvptofh,
132 	.vop_unp_bind =		vop_stdunp_bind,
133 	.vop_unp_connect =	vop_stdunp_connect,
134 	.vop_unp_detach =	vop_stdunp_detach,
135 	.vop_is_text =		vop_stdis_text,
136 	.vop_set_text =		vop_stdset_text,
137 	.vop_unset_text =	vop_stdunset_text,
138 	.vop_get_writecount =	vop_stdget_writecount,
139 	.vop_add_writecount =	vop_stdadd_writecount,
140 };
141 
142 /*
143  * Series of placeholder functions for various error returns for
144  * VOPs.
145  */
146 
147 int
148 vop_eopnotsupp(struct vop_generic_args *ap)
149 {
150 	/*
151 	printf("vop_notsupp[%s]\n", ap->a_desc->vdesc_name);
152 	*/
153 
154 	return (EOPNOTSUPP);
155 }
156 
157 int
158 vop_ebadf(struct vop_generic_args *ap)
159 {
160 
161 	return (EBADF);
162 }
163 
164 int
165 vop_enotty(struct vop_generic_args *ap)
166 {
167 
168 	return (ENOTTY);
169 }
170 
171 int
172 vop_einval(struct vop_generic_args *ap)
173 {
174 
175 	return (EINVAL);
176 }
177 
178 int
179 vop_enoent(struct vop_generic_args *ap)
180 {
181 
182 	return (ENOENT);
183 }
184 
185 int
186 vop_null(struct vop_generic_args *ap)
187 {
188 
189 	return (0);
190 }
191 
192 /*
193  * Helper function to panic on some bad VOPs in some filesystems.
194  */
195 int
196 vop_panic(struct vop_generic_args *ap)
197 {
198 
199 	panic("filesystem goof: vop_panic[%s]", ap->a_desc->vdesc_name);
200 }
201 
202 /*
203  * vop_std<something> and vop_no<something> are default functions for use by
204  * filesystems that need the "default reasonable" implementation for a
205  * particular operation.
206  *
207  * The documentation for the operations they implement exists (if it exists)
208  * in the VOP_<SOMETHING>(9) manpage (all uppercase).
209  */
210 
211 /*
212  * Default vop for filesystems that do not support name lookup
213  */
214 static int
215 vop_nolookup(ap)
216 	struct vop_lookup_args /* {
217 		struct vnode *a_dvp;
218 		struct vnode **a_vpp;
219 		struct componentname *a_cnp;
220 	} */ *ap;
221 {
222 
223 	*ap->a_vpp = NULL;
224 	return (ENOTDIR);
225 }
226 
227 /*
228  * vop_norename:
229  *
230  * Handle unlock and reference counting for arguments of vop_rename
231  * for filesystems that do not implement rename operation.
232  */
233 static int
234 vop_norename(struct vop_rename_args *ap)
235 {
236 
237 	vop_rename_fail(ap);
238 	return (EOPNOTSUPP);
239 }
240 
241 /*
242  *	vop_nostrategy:
243  *
244  *	Strategy routine for VFS devices that have none.
245  *
246  *	BIO_ERROR and B_INVAL must be cleared prior to calling any strategy
247  *	routine.  Typically this is done for a BIO_READ strategy call.
248  *	Typically B_INVAL is assumed to already be clear prior to a write
249  *	and should not be cleared manually unless you just made the buffer
250  *	invalid.  BIO_ERROR should be cleared either way.
251  */
252 
253 static int
254 vop_nostrategy (struct vop_strategy_args *ap)
255 {
256 	printf("No strategy for buffer at %p\n", ap->a_bp);
257 	vprint("vnode", ap->a_vp);
258 	ap->a_bp->b_ioflags |= BIO_ERROR;
259 	ap->a_bp->b_error = EOPNOTSUPP;
260 	bufdone(ap->a_bp);
261 	return (EOPNOTSUPP);
262 }
263 
264 static int
265 get_next_dirent(struct vnode *vp, struct dirent **dpp, char *dirbuf,
266 		int dirbuflen, off_t *off, char **cpos, int *len,
267 		int *eofflag, struct thread *td)
268 {
269 	int error, reclen;
270 	struct uio uio;
271 	struct iovec iov;
272 	struct dirent *dp;
273 
274 	KASSERT(VOP_ISLOCKED(vp), ("vp %p is not locked", vp));
275 	KASSERT(vp->v_type == VDIR, ("vp %p is not a directory", vp));
276 
277 	if (*len == 0) {
278 		iov.iov_base = dirbuf;
279 		iov.iov_len = dirbuflen;
280 
281 		uio.uio_iov = &iov;
282 		uio.uio_iovcnt = 1;
283 		uio.uio_offset = *off;
284 		uio.uio_resid = dirbuflen;
285 		uio.uio_segflg = UIO_SYSSPACE;
286 		uio.uio_rw = UIO_READ;
287 		uio.uio_td = td;
288 
289 		*eofflag = 0;
290 
291 #ifdef MAC
292 		error = mac_vnode_check_readdir(td->td_ucred, vp);
293 		if (error == 0)
294 #endif
295 			error = VOP_READDIR(vp, &uio, td->td_ucred, eofflag,
296 		    		NULL, NULL);
297 		if (error)
298 			return (error);
299 
300 		*off = uio.uio_offset;
301 
302 		*cpos = dirbuf;
303 		*len = (dirbuflen - uio.uio_resid);
304 
305 		if (*len == 0)
306 			return (ENOENT);
307 	}
308 
309 	dp = (struct dirent *)(*cpos);
310 	reclen = dp->d_reclen;
311 	*dpp = dp;
312 
313 	/* check for malformed directory.. */
314 	if (reclen < DIRENT_MINSIZE)
315 		return (EINVAL);
316 
317 	*cpos += reclen;
318 	*len -= reclen;
319 
320 	return (0);
321 }
322 
323 /*
324  * Check if a named file exists in a given directory vnode.
325  */
326 static int
327 dirent_exists(struct vnode *vp, const char *dirname, struct thread *td)
328 {
329 	char *dirbuf, *cpos;
330 	int error, eofflag, dirbuflen, len, found;
331 	off_t off;
332 	struct dirent *dp;
333 	struct vattr va;
334 
335 	KASSERT(VOP_ISLOCKED(vp), ("vp %p is not locked", vp));
336 	KASSERT(vp->v_type == VDIR, ("vp %p is not a directory", vp));
337 
338 	found = 0;
339 
340 	error = VOP_GETATTR(vp, &va, td->td_ucred);
341 	if (error)
342 		return (found);
343 
344 	dirbuflen = DEV_BSIZE;
345 	if (dirbuflen < va.va_blocksize)
346 		dirbuflen = va.va_blocksize;
347 	dirbuf = (char *)malloc(dirbuflen, M_TEMP, M_WAITOK);
348 
349 	off = 0;
350 	len = 0;
351 	do {
352 		error = get_next_dirent(vp, &dp, dirbuf, dirbuflen, &off,
353 					&cpos, &len, &eofflag, td);
354 		if (error)
355 			goto out;
356 
357 		if (dp->d_type != DT_WHT && dp->d_fileno != 0 &&
358 		    strcmp(dp->d_name, dirname) == 0) {
359 			found = 1;
360 			goto out;
361 		}
362 	} while (len > 0 || !eofflag);
363 
364 out:
365 	free(dirbuf, M_TEMP);
366 	return (found);
367 }
368 
369 int
370 vop_stdaccess(struct vop_access_args *ap)
371 {
372 
373 	KASSERT((ap->a_accmode & ~(VEXEC | VWRITE | VREAD | VADMIN |
374 	    VAPPEND)) == 0, ("invalid bit in accmode"));
375 
376 	return (VOP_ACCESSX(ap->a_vp, ap->a_accmode, ap->a_cred, ap->a_td));
377 }
378 
379 int
380 vop_stdaccessx(struct vop_accessx_args *ap)
381 {
382 	int error;
383 	accmode_t accmode = ap->a_accmode;
384 
385 	error = vfs_unixify_accmode(&accmode);
386 	if (error != 0)
387 		return (error);
388 
389 	if (accmode == 0)
390 		return (0);
391 
392 	return (VOP_ACCESS(ap->a_vp, accmode, ap->a_cred, ap->a_td));
393 }
394 
395 /*
396  * Advisory record locking support
397  */
398 int
399 vop_stdadvlock(struct vop_advlock_args *ap)
400 {
401 	struct vnode *vp;
402 	struct ucred *cred;
403 	struct vattr vattr;
404 	int error;
405 
406 	vp = ap->a_vp;
407 	cred = curthread->td_ucred;
408 	vn_lock(vp, LK_SHARED | LK_RETRY);
409 	error = VOP_GETATTR(vp, &vattr, cred);
410 	VOP_UNLOCK(vp, 0);
411 	if (error)
412 		return (error);
413 
414 	return (lf_advlock(ap, &(vp->v_lockf), vattr.va_size));
415 }
416 
417 int
418 vop_stdadvlockasync(struct vop_advlockasync_args *ap)
419 {
420 	struct vnode *vp;
421 	struct ucred *cred;
422 	struct vattr vattr;
423 	int error;
424 
425 	vp = ap->a_vp;
426 	cred = curthread->td_ucred;
427 	vn_lock(vp, LK_SHARED | LK_RETRY);
428 	error = VOP_GETATTR(vp, &vattr, cred);
429 	VOP_UNLOCK(vp, 0);
430 	if (error)
431 		return (error);
432 
433 	return (lf_advlockasync(ap, &(vp->v_lockf), vattr.va_size));
434 }
435 
436 int
437 vop_stdadvlockpurge(struct vop_advlockpurge_args *ap)
438 {
439 	struct vnode *vp;
440 
441 	vp = ap->a_vp;
442 	lf_purgelocks(vp, &vp->v_lockf);
443 	return (0);
444 }
445 
446 /*
447  * vop_stdpathconf:
448  *
449  * Standard implementation of POSIX pathconf, to get information about limits
450  * for a filesystem.
451  * Override per filesystem for the case where the filesystem has smaller
452  * limits.
453  */
454 int
455 vop_stdpathconf(ap)
456 	struct vop_pathconf_args /* {
457 	struct vnode *a_vp;
458 	int a_name;
459 	int *a_retval;
460 	} */ *ap;
461 {
462 
463 	switch (ap->a_name) {
464 		case _PC_NAME_MAX:
465 			*ap->a_retval = NAME_MAX;
466 			return (0);
467 		case _PC_PATH_MAX:
468 			*ap->a_retval = PATH_MAX;
469 			return (0);
470 		case _PC_LINK_MAX:
471 			*ap->a_retval = LINK_MAX;
472 			return (0);
473 		case _PC_MAX_CANON:
474 			*ap->a_retval = MAX_CANON;
475 			return (0);
476 		case _PC_MAX_INPUT:
477 			*ap->a_retval = MAX_INPUT;
478 			return (0);
479 		case _PC_PIPE_BUF:
480 			*ap->a_retval = PIPE_BUF;
481 			return (0);
482 		case _PC_CHOWN_RESTRICTED:
483 			*ap->a_retval = 1;
484 			return (0);
485 		case _PC_VDISABLE:
486 			*ap->a_retval = _POSIX_VDISABLE;
487 			return (0);
488 		default:
489 			return (EINVAL);
490 	}
491 	/* NOTREACHED */
492 }
493 
494 /*
495  * Standard lock, unlock and islocked functions.
496  */
497 int
498 vop_stdlock(ap)
499 	struct vop_lock1_args /* {
500 		struct vnode *a_vp;
501 		int a_flags;
502 		char *file;
503 		int line;
504 	} */ *ap;
505 {
506 	struct vnode *vp = ap->a_vp;
507 
508 	return (_lockmgr_args(vp->v_vnlock, ap->a_flags, VI_MTX(vp),
509 	    LK_WMESG_DEFAULT, LK_PRIO_DEFAULT, LK_TIMO_DEFAULT, ap->a_file,
510 	    ap->a_line));
511 }
512 
513 /* See above. */
514 int
515 vop_stdunlock(ap)
516 	struct vop_unlock_args /* {
517 		struct vnode *a_vp;
518 		int a_flags;
519 	} */ *ap;
520 {
521 	struct vnode *vp = ap->a_vp;
522 
523 	return (lockmgr(vp->v_vnlock, ap->a_flags | LK_RELEASE, VI_MTX(vp)));
524 }
525 
526 /* See above. */
527 int
528 vop_stdislocked(ap)
529 	struct vop_islocked_args /* {
530 		struct vnode *a_vp;
531 	} */ *ap;
532 {
533 
534 	return (lockstatus(ap->a_vp->v_vnlock));
535 }
536 
537 /*
538  * Return true for select/poll.
539  */
540 int
541 vop_nopoll(ap)
542 	struct vop_poll_args /* {
543 		struct vnode *a_vp;
544 		int  a_events;
545 		struct ucred *a_cred;
546 		struct thread *a_td;
547 	} */ *ap;
548 {
549 
550 	return (poll_no_poll(ap->a_events));
551 }
552 
553 /*
554  * Implement poll for local filesystems that support it.
555  */
556 int
557 vop_stdpoll(ap)
558 	struct vop_poll_args /* {
559 		struct vnode *a_vp;
560 		int  a_events;
561 		struct ucred *a_cred;
562 		struct thread *a_td;
563 	} */ *ap;
564 {
565 	if (ap->a_events & ~POLLSTANDARD)
566 		return (vn_pollrecord(ap->a_vp, ap->a_td, ap->a_events));
567 	return (ap->a_events & (POLLIN | POLLOUT | POLLRDNORM | POLLWRNORM));
568 }
569 
570 /*
571  * Return our mount point, as we will take charge of the writes.
572  */
573 int
574 vop_stdgetwritemount(ap)
575 	struct vop_getwritemount_args /* {
576 		struct vnode *a_vp;
577 		struct mount **a_mpp;
578 	} */ *ap;
579 {
580 	struct mount *mp;
581 
582 	/*
583 	 * XXX Since this is called unlocked we may be recycled while
584 	 * attempting to ref the mount.  If this is the case or mountpoint
585 	 * will be set to NULL.  We only have to prevent this call from
586 	 * returning with a ref to an incorrect mountpoint.  It is not
587 	 * harmful to return with a ref to our previous mountpoint.
588 	 */
589 	mp = ap->a_vp->v_mount;
590 	if (mp != NULL) {
591 		vfs_ref(mp);
592 		if (mp != ap->a_vp->v_mount) {
593 			vfs_rel(mp);
594 			mp = NULL;
595 		}
596 	}
597 	*(ap->a_mpp) = mp;
598 	return (0);
599 }
600 
601 /* XXX Needs good comment and VOP_BMAP(9) manpage */
602 int
603 vop_stdbmap(ap)
604 	struct vop_bmap_args /* {
605 		struct vnode *a_vp;
606 		daddr_t  a_bn;
607 		struct bufobj **a_bop;
608 		daddr_t *a_bnp;
609 		int *a_runp;
610 		int *a_runb;
611 	} */ *ap;
612 {
613 
614 	if (ap->a_bop != NULL)
615 		*ap->a_bop = &ap->a_vp->v_bufobj;
616 	if (ap->a_bnp != NULL)
617 		*ap->a_bnp = ap->a_bn * btodb(ap->a_vp->v_mount->mnt_stat.f_iosize);
618 	if (ap->a_runp != NULL)
619 		*ap->a_runp = 0;
620 	if (ap->a_runb != NULL)
621 		*ap->a_runb = 0;
622 	return (0);
623 }
624 
625 int
626 vop_stdfsync(ap)
627 	struct vop_fsync_args /* {
628 		struct vnode *a_vp;
629 		struct ucred *a_cred;
630 		int a_waitfor;
631 		struct thread *a_td;
632 	} */ *ap;
633 {
634 	struct vnode *vp = ap->a_vp;
635 	struct buf *bp;
636 	struct bufobj *bo;
637 	struct buf *nbp;
638 	int error = 0;
639 	int maxretry = 1000;     /* large, arbitrarily chosen */
640 
641 	bo = &vp->v_bufobj;
642 	BO_LOCK(bo);
643 loop1:
644 	/*
645 	 * MARK/SCAN initialization to avoid infinite loops.
646 	 */
647         TAILQ_FOREACH(bp, &bo->bo_dirty.bv_hd, b_bobufs) {
648                 bp->b_vflags &= ~BV_SCANNED;
649 		bp->b_error = 0;
650 	}
651 
652 	/*
653 	 * Flush all dirty buffers associated with a vnode.
654 	 */
655 loop2:
656 	TAILQ_FOREACH_SAFE(bp, &bo->bo_dirty.bv_hd, b_bobufs, nbp) {
657 		if ((bp->b_vflags & BV_SCANNED) != 0)
658 			continue;
659 		bp->b_vflags |= BV_SCANNED;
660 		if (BUF_LOCK(bp, LK_EXCLUSIVE | LK_NOWAIT, NULL)) {
661 			if (ap->a_waitfor != MNT_WAIT)
662 				continue;
663 			if (BUF_LOCK(bp,
664 			    LK_EXCLUSIVE | LK_INTERLOCK | LK_SLEEPFAIL,
665 			    BO_LOCKPTR(bo)) != 0) {
666 				BO_LOCK(bo);
667 				goto loop1;
668 			}
669 			BO_LOCK(bo);
670 		}
671 		BO_UNLOCK(bo);
672 		KASSERT(bp->b_bufobj == bo,
673 		    ("bp %p wrong b_bufobj %p should be %p",
674 		    bp, bp->b_bufobj, bo));
675 		if ((bp->b_flags & B_DELWRI) == 0)
676 			panic("fsync: not dirty");
677 		if ((vp->v_object != NULL) && (bp->b_flags & B_CLUSTEROK)) {
678 			vfs_bio_awrite(bp);
679 		} else {
680 			bremfree(bp);
681 			bawrite(bp);
682 		}
683 		BO_LOCK(bo);
684 		goto loop2;
685 	}
686 
687 	/*
688 	 * If synchronous the caller expects us to completely resolve all
689 	 * dirty buffers in the system.  Wait for in-progress I/O to
690 	 * complete (which could include background bitmap writes), then
691 	 * retry if dirty blocks still exist.
692 	 */
693 	if (ap->a_waitfor == MNT_WAIT) {
694 		bufobj_wwait(bo, 0, 0);
695 		if (bo->bo_dirty.bv_cnt > 0) {
696 			/*
697 			 * If we are unable to write any of these buffers
698 			 * then we fail now rather than trying endlessly
699 			 * to write them out.
700 			 */
701 			TAILQ_FOREACH(bp, &bo->bo_dirty.bv_hd, b_bobufs)
702 				if ((error = bp->b_error) == 0)
703 					continue;
704 			if (error == 0 && --maxretry >= 0)
705 				goto loop1;
706 			error = EAGAIN;
707 		}
708 	}
709 	BO_UNLOCK(bo);
710 	if (error == EAGAIN)
711 		vprint("fsync: giving up on dirty", vp);
712 
713 	return (error);
714 }
715 
716 /* XXX Needs good comment and more info in the manpage (VOP_GETPAGES(9)). */
717 int
718 vop_stdgetpages(ap)
719 	struct vop_getpages_args /* {
720 		struct vnode *a_vp;
721 		vm_page_t *a_m;
722 		int a_count;
723 		int a_reqpage;
724 		vm_ooffset_t a_offset;
725 	} */ *ap;
726 {
727 
728 	return vnode_pager_generic_getpages(ap->a_vp, ap->a_m,
729 	    ap->a_count, ap->a_reqpage);
730 }
731 
732 int
733 vop_stdkqfilter(struct vop_kqfilter_args *ap)
734 {
735 	return vfs_kqfilter(ap);
736 }
737 
738 /* XXX Needs good comment and more info in the manpage (VOP_PUTPAGES(9)). */
739 int
740 vop_stdputpages(ap)
741 	struct vop_putpages_args /* {
742 		struct vnode *a_vp;
743 		vm_page_t *a_m;
744 		int a_count;
745 		int a_sync;
746 		int *a_rtvals;
747 		vm_ooffset_t a_offset;
748 	} */ *ap;
749 {
750 
751 	return vnode_pager_generic_putpages(ap->a_vp, ap->a_m, ap->a_count,
752 	     ap->a_sync, ap->a_rtvals);
753 }
754 
755 int
756 vop_stdvptofh(struct vop_vptofh_args *ap)
757 {
758 	return (EOPNOTSUPP);
759 }
760 
761 int
762 vop_stdvptocnp(struct vop_vptocnp_args *ap)
763 {
764 	struct vnode *vp = ap->a_vp;
765 	struct vnode **dvp = ap->a_vpp;
766 	struct ucred *cred = ap->a_cred;
767 	char *buf = ap->a_buf;
768 	int *buflen = ap->a_buflen;
769 	char *dirbuf, *cpos;
770 	int i, error, eofflag, dirbuflen, flags, locked, len, covered;
771 	off_t off;
772 	ino_t fileno;
773 	struct vattr va;
774 	struct nameidata nd;
775 	struct thread *td;
776 	struct dirent *dp;
777 	struct vnode *mvp;
778 
779 	i = *buflen;
780 	error = 0;
781 	covered = 0;
782 	td = curthread;
783 
784 	if (vp->v_type != VDIR)
785 		return (ENOENT);
786 
787 	error = VOP_GETATTR(vp, &va, cred);
788 	if (error)
789 		return (error);
790 
791 	VREF(vp);
792 	locked = VOP_ISLOCKED(vp);
793 	VOP_UNLOCK(vp, 0);
794 	NDINIT_ATVP(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE,
795 	    "..", vp, td);
796 	flags = FREAD;
797 	error = vn_open_cred(&nd, &flags, 0, VN_OPEN_NOAUDIT, cred, NULL);
798 	if (error) {
799 		vn_lock(vp, locked | LK_RETRY);
800 		return (error);
801 	}
802 	NDFREE(&nd, NDF_ONLY_PNBUF);
803 
804 	mvp = *dvp = nd.ni_vp;
805 
806 	if (vp->v_mount != (*dvp)->v_mount &&
807 	    ((*dvp)->v_vflag & VV_ROOT) &&
808 	    ((*dvp)->v_mount->mnt_flag & MNT_UNION)) {
809 		*dvp = (*dvp)->v_mount->mnt_vnodecovered;
810 		VREF(mvp);
811 		VOP_UNLOCK(mvp, 0);
812 		vn_close(mvp, FREAD, cred, td);
813 		VREF(*dvp);
814 		vn_lock(*dvp, LK_EXCLUSIVE | LK_RETRY);
815 		covered = 1;
816 	}
817 
818 	fileno = va.va_fileid;
819 
820 	dirbuflen = DEV_BSIZE;
821 	if (dirbuflen < va.va_blocksize)
822 		dirbuflen = va.va_blocksize;
823 	dirbuf = (char *)malloc(dirbuflen, M_TEMP, M_WAITOK);
824 
825 	if ((*dvp)->v_type != VDIR) {
826 		error = ENOENT;
827 		goto out;
828 	}
829 
830 	off = 0;
831 	len = 0;
832 	do {
833 		/* call VOP_READDIR of parent */
834 		error = get_next_dirent(*dvp, &dp, dirbuf, dirbuflen, &off,
835 					&cpos, &len, &eofflag, td);
836 		if (error)
837 			goto out;
838 
839 		if ((dp->d_type != DT_WHT) &&
840 		    (dp->d_fileno == fileno)) {
841 			if (covered) {
842 				VOP_UNLOCK(*dvp, 0);
843 				vn_lock(mvp, LK_EXCLUSIVE | LK_RETRY);
844 				if (dirent_exists(mvp, dp->d_name, td)) {
845 					error = ENOENT;
846 					VOP_UNLOCK(mvp, 0);
847 					vn_lock(*dvp, LK_EXCLUSIVE | LK_RETRY);
848 					goto out;
849 				}
850 				VOP_UNLOCK(mvp, 0);
851 				vn_lock(*dvp, LK_EXCLUSIVE | LK_RETRY);
852 			}
853 			i -= dp->d_namlen;
854 
855 			if (i < 0) {
856 				error = ENOMEM;
857 				goto out;
858 			}
859 			if (dp->d_namlen == 1 && dp->d_name[0] == '.') {
860 				error = ENOENT;
861 			} else {
862 				bcopy(dp->d_name, buf + i, dp->d_namlen);
863 				error = 0;
864 			}
865 			goto out;
866 		}
867 	} while (len > 0 || !eofflag);
868 	error = ENOENT;
869 
870 out:
871 	free(dirbuf, M_TEMP);
872 	if (!error) {
873 		*buflen = i;
874 		vref(*dvp);
875 	}
876 	if (covered) {
877 		vput(*dvp);
878 		vrele(mvp);
879 	} else {
880 		VOP_UNLOCK(mvp, 0);
881 		vn_close(mvp, FREAD, cred, td);
882 	}
883 	vn_lock(vp, locked | LK_RETRY);
884 	return (error);
885 }
886 
887 int
888 vop_stdallocate(struct vop_allocate_args *ap)
889 {
890 #ifdef __notyet__
891 	struct statfs sfs;
892 #endif
893 	struct iovec aiov;
894 	struct vattr vattr, *vap;
895 	struct uio auio;
896 	off_t fsize, len, cur, offset;
897 	uint8_t *buf;
898 	struct thread *td;
899 	struct vnode *vp;
900 	size_t iosize;
901 	int error;
902 
903 	buf = NULL;
904 	error = 0;
905 	td = curthread;
906 	vap = &vattr;
907 	vp = ap->a_vp;
908 	len = *ap->a_len;
909 	offset = *ap->a_offset;
910 
911 	error = VOP_GETATTR(vp, vap, td->td_ucred);
912 	if (error != 0)
913 		goto out;
914 	fsize = vap->va_size;
915 	iosize = vap->va_blocksize;
916 	if (iosize == 0)
917 		iosize = BLKDEV_IOSIZE;
918 	if (iosize > MAXPHYS)
919 		iosize = MAXPHYS;
920 	buf = malloc(iosize, M_TEMP, M_WAITOK);
921 
922 #ifdef __notyet__
923 	/*
924 	 * Check if the filesystem sets f_maxfilesize; if not use
925 	 * VOP_SETATTR to perform the check.
926 	 */
927 	error = VFS_STATFS(vp->v_mount, &sfs, td);
928 	if (error != 0)
929 		goto out;
930 	if (sfs.f_maxfilesize) {
931 		if (offset > sfs.f_maxfilesize || len > sfs.f_maxfilesize ||
932 		    offset + len > sfs.f_maxfilesize) {
933 			error = EFBIG;
934 			goto out;
935 		}
936 	} else
937 #endif
938 	if (offset + len > vap->va_size) {
939 		/*
940 		 * Test offset + len against the filesystem's maxfilesize.
941 		 */
942 		VATTR_NULL(vap);
943 		vap->va_size = offset + len;
944 		error = VOP_SETATTR(vp, vap, td->td_ucred);
945 		if (error != 0)
946 			goto out;
947 		VATTR_NULL(vap);
948 		vap->va_size = fsize;
949 		error = VOP_SETATTR(vp, vap, td->td_ucred);
950 		if (error != 0)
951 			goto out;
952 	}
953 
954 	for (;;) {
955 		/*
956 		 * Read and write back anything below the nominal file
957 		 * size.  There's currently no way outside the filesystem
958 		 * to know whether this area is sparse or not.
959 		 */
960 		cur = iosize;
961 		if ((offset % iosize) != 0)
962 			cur -= (offset % iosize);
963 		if (cur > len)
964 			cur = len;
965 		if (offset < fsize) {
966 			aiov.iov_base = buf;
967 			aiov.iov_len = cur;
968 			auio.uio_iov = &aiov;
969 			auio.uio_iovcnt = 1;
970 			auio.uio_offset = offset;
971 			auio.uio_resid = cur;
972 			auio.uio_segflg = UIO_SYSSPACE;
973 			auio.uio_rw = UIO_READ;
974 			auio.uio_td = td;
975 			error = VOP_READ(vp, &auio, 0, td->td_ucred);
976 			if (error != 0)
977 				break;
978 			if (auio.uio_resid > 0) {
979 				bzero(buf + cur - auio.uio_resid,
980 				    auio.uio_resid);
981 			}
982 		} else {
983 			bzero(buf, cur);
984 		}
985 
986 		aiov.iov_base = buf;
987 		aiov.iov_len = cur;
988 		auio.uio_iov = &aiov;
989 		auio.uio_iovcnt = 1;
990 		auio.uio_offset = offset;
991 		auio.uio_resid = cur;
992 		auio.uio_segflg = UIO_SYSSPACE;
993 		auio.uio_rw = UIO_WRITE;
994 		auio.uio_td = td;
995 
996 		error = VOP_WRITE(vp, &auio, 0, td->td_ucred);
997 		if (error != 0)
998 			break;
999 
1000 		len -= cur;
1001 		offset += cur;
1002 		if (len == 0)
1003 			break;
1004 		if (should_yield())
1005 			break;
1006 	}
1007 
1008  out:
1009 	*ap->a_len = len;
1010 	*ap->a_offset = offset;
1011 	free(buf, M_TEMP);
1012 	return (error);
1013 }
1014 
1015 int
1016 vop_stdadvise(struct vop_advise_args *ap)
1017 {
1018 	struct vnode *vp;
1019 	off_t start, end;
1020 	int error;
1021 
1022 	vp = ap->a_vp;
1023 	switch (ap->a_advice) {
1024 	case POSIX_FADV_WILLNEED:
1025 		/*
1026 		 * Do nothing for now.  Filesystems should provide a
1027 		 * custom method which starts an asynchronous read of
1028 		 * the requested region.
1029 		 */
1030 		error = 0;
1031 		break;
1032 	case POSIX_FADV_DONTNEED:
1033 		/*
1034 		 * Flush any open FS buffers and then remove pages
1035 		 * from the backing VM object.  Using vinvalbuf() here
1036 		 * is a bit heavy-handed as it flushes all buffers for
1037 		 * the given vnode, not just the buffers covering the
1038 		 * requested range.
1039 		 */
1040 		error = 0;
1041 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1042 		if (vp->v_iflag & VI_DOOMED) {
1043 			VOP_UNLOCK(vp, 0);
1044 			break;
1045 		}
1046 		vinvalbuf(vp, V_CLEANONLY, 0, 0);
1047 		if (vp->v_object != NULL) {
1048 			start = trunc_page(ap->a_start);
1049 			end = round_page(ap->a_end);
1050 			VM_OBJECT_WLOCK(vp->v_object);
1051 			vm_object_page_cache(vp->v_object, OFF_TO_IDX(start),
1052 			    OFF_TO_IDX(end));
1053 			VM_OBJECT_WUNLOCK(vp->v_object);
1054 		}
1055 		VOP_UNLOCK(vp, 0);
1056 		break;
1057 	default:
1058 		error = EINVAL;
1059 		break;
1060 	}
1061 	return (error);
1062 }
1063 
1064 int
1065 vop_stdunp_bind(struct vop_unp_bind_args *ap)
1066 {
1067 
1068 	ap->a_vp->v_socket = ap->a_socket;
1069 	return (0);
1070 }
1071 
1072 int
1073 vop_stdunp_connect(struct vop_unp_connect_args *ap)
1074 {
1075 
1076 	*ap->a_socket = ap->a_vp->v_socket;
1077 	return (0);
1078 }
1079 
1080 int
1081 vop_stdunp_detach(struct vop_unp_detach_args *ap)
1082 {
1083 
1084 	ap->a_vp->v_socket = NULL;
1085 	return (0);
1086 }
1087 
1088 static int
1089 vop_stdis_text(struct vop_is_text_args *ap)
1090 {
1091 
1092 	return ((ap->a_vp->v_vflag & VV_TEXT) != 0);
1093 }
1094 
1095 static int
1096 vop_stdset_text(struct vop_set_text_args *ap)
1097 {
1098 
1099 	ap->a_vp->v_vflag |= VV_TEXT;
1100 	return (0);
1101 }
1102 
1103 static int
1104 vop_stdunset_text(struct vop_unset_text_args *ap)
1105 {
1106 
1107 	ap->a_vp->v_vflag &= ~VV_TEXT;
1108 	return (0);
1109 }
1110 
1111 static int
1112 vop_stdget_writecount(struct vop_get_writecount_args *ap)
1113 {
1114 
1115 	*ap->a_writecount = ap->a_vp->v_writecount;
1116 	return (0);
1117 }
1118 
1119 static int
1120 vop_stdadd_writecount(struct vop_add_writecount_args *ap)
1121 {
1122 
1123 	ap->a_vp->v_writecount += ap->a_inc;
1124 	return (0);
1125 }
1126 
1127 /*
1128  * vfs default ops
1129  * used to fill the vfs function table to get reasonable default return values.
1130  */
1131 int
1132 vfs_stdroot (mp, flags, vpp)
1133 	struct mount *mp;
1134 	int flags;
1135 	struct vnode **vpp;
1136 {
1137 
1138 	return (EOPNOTSUPP);
1139 }
1140 
1141 int
1142 vfs_stdstatfs (mp, sbp)
1143 	struct mount *mp;
1144 	struct statfs *sbp;
1145 {
1146 
1147 	return (EOPNOTSUPP);
1148 }
1149 
1150 int
1151 vfs_stdquotactl (mp, cmds, uid, arg)
1152 	struct mount *mp;
1153 	int cmds;
1154 	uid_t uid;
1155 	void *arg;
1156 {
1157 
1158 	return (EOPNOTSUPP);
1159 }
1160 
1161 int
1162 vfs_stdsync(mp, waitfor)
1163 	struct mount *mp;
1164 	int waitfor;
1165 {
1166 	struct vnode *vp, *mvp;
1167 	struct thread *td;
1168 	int error, lockreq, allerror = 0;
1169 
1170 	td = curthread;
1171 	lockreq = LK_EXCLUSIVE | LK_INTERLOCK;
1172 	if (waitfor != MNT_WAIT)
1173 		lockreq |= LK_NOWAIT;
1174 	/*
1175 	 * Force stale buffer cache information to be flushed.
1176 	 */
1177 loop:
1178 	MNT_VNODE_FOREACH_ALL(vp, mp, mvp) {
1179 		if (vp->v_bufobj.bo_dirty.bv_cnt == 0) {
1180 			VI_UNLOCK(vp);
1181 			continue;
1182 		}
1183 		if ((error = vget(vp, lockreq, td)) != 0) {
1184 			if (error == ENOENT) {
1185 				MNT_VNODE_FOREACH_ALL_ABORT(mp, mvp);
1186 				goto loop;
1187 			}
1188 			continue;
1189 		}
1190 		error = VOP_FSYNC(vp, waitfor, td);
1191 		if (error)
1192 			allerror = error;
1193 		vput(vp);
1194 	}
1195 	return (allerror);
1196 }
1197 
1198 int
1199 vfs_stdnosync (mp, waitfor)
1200 	struct mount *mp;
1201 	int waitfor;
1202 {
1203 
1204 	return (0);
1205 }
1206 
1207 int
1208 vfs_stdvget (mp, ino, flags, vpp)
1209 	struct mount *mp;
1210 	ino_t ino;
1211 	int flags;
1212 	struct vnode **vpp;
1213 {
1214 
1215 	return (EOPNOTSUPP);
1216 }
1217 
1218 int
1219 vfs_stdfhtovp (mp, fhp, flags, vpp)
1220 	struct mount *mp;
1221 	struct fid *fhp;
1222 	int flags;
1223 	struct vnode **vpp;
1224 {
1225 
1226 	return (EOPNOTSUPP);
1227 }
1228 
1229 int
1230 vfs_stdinit (vfsp)
1231 	struct vfsconf *vfsp;
1232 {
1233 
1234 	return (0);
1235 }
1236 
1237 int
1238 vfs_stduninit (vfsp)
1239 	struct vfsconf *vfsp;
1240 {
1241 
1242 	return(0);
1243 }
1244 
1245 int
1246 vfs_stdextattrctl(mp, cmd, filename_vp, attrnamespace, attrname)
1247 	struct mount *mp;
1248 	int cmd;
1249 	struct vnode *filename_vp;
1250 	int attrnamespace;
1251 	const char *attrname;
1252 {
1253 
1254 	if (filename_vp != NULL)
1255 		VOP_UNLOCK(filename_vp, 0);
1256 	return (EOPNOTSUPP);
1257 }
1258 
1259 int
1260 vfs_stdsysctl(mp, op, req)
1261 	struct mount *mp;
1262 	fsctlop_t op;
1263 	struct sysctl_req *req;
1264 {
1265 
1266 	return (EOPNOTSUPP);
1267 }
1268 
1269 /* end of vfs default ops */
1270