xref: /freebsd/sys/kern/vfs_default.c (revision 43d1e6ee299ad4e143d90d3ad374d1c24bd3306f)
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 
71 static int	vop_nolookup(struct vop_lookup_args *);
72 static int	vop_norename(struct vop_rename_args *);
73 static int	vop_nostrategy(struct vop_strategy_args *);
74 static 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);
78 static 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 
83 static int vop_stdis_text(struct vop_is_text_args *ap);
84 static int vop_stdset_text(struct vop_set_text_args *ap);
85 static int vop_stdunset_text(struct vop_unset_text_args *ap);
86 static int vop_stdget_writecount(struct vop_get_writecount_args *ap);
87 static int vop_stdadd_writecount(struct vop_add_writecount_args *ap);
88 static int vop_stdfdatasync(struct vop_fdatasync_args *ap);
89 static int vop_stdgetpages_async(struct vop_getpages_async_args *ap);
90 
91 /*
92  * This vnode table stores what we want to do if the filesystem doesn't
93  * implement a particular VOP.
94  *
95  * If there is no specific entry here, we will return EOPNOTSUPP.
96  *
97  * Note that every filesystem has to implement either vop_access
98  * or vop_accessx; failing to do so will result in immediate crash
99  * due to stack overflow, as vop_stdaccess() calls vop_stdaccessx(),
100  * which calls vop_stdaccess() etc.
101  */
102 
103 struct vop_vector default_vnodeops = {
104 	.vop_default =		NULL,
105 	.vop_bypass =		VOP_EOPNOTSUPP,
106 
107 	.vop_access =		vop_stdaccess,
108 	.vop_accessx =		vop_stdaccessx,
109 	.vop_advise =		vop_stdadvise,
110 	.vop_advlock =		vop_stdadvlock,
111 	.vop_advlockasync =	vop_stdadvlockasync,
112 	.vop_advlockpurge =	vop_stdadvlockpurge,
113 	.vop_allocate =		vop_stdallocate,
114 	.vop_bmap =		vop_stdbmap,
115 	.vop_close =		VOP_NULL,
116 	.vop_fsync =		VOP_NULL,
117 	.vop_fdatasync =	vop_stdfdatasync,
118 	.vop_getpages =		vop_stdgetpages,
119 	.vop_getpages_async =	vop_stdgetpages_async,
120 	.vop_getwritemount = 	vop_stdgetwritemount,
121 	.vop_inactive =		VOP_NULL,
122 	.vop_ioctl =		VOP_ENOTTY,
123 	.vop_kqfilter =		vop_stdkqfilter,
124 	.vop_islocked =		vop_stdislocked,
125 	.vop_lock1 =		vop_stdlock,
126 	.vop_lookup =		vop_nolookup,
127 	.vop_open =		VOP_NULL,
128 	.vop_pathconf =		VOP_EINVAL,
129 	.vop_poll =		vop_nopoll,
130 	.vop_putpages =		vop_stdputpages,
131 	.vop_readlink =		VOP_EINVAL,
132 	.vop_rename =		vop_norename,
133 	.vop_revoke =		VOP_PANIC,
134 	.vop_strategy =		vop_nostrategy,
135 	.vop_unlock =		vop_stdunlock,
136 	.vop_vptocnp =		vop_stdvptocnp,
137 	.vop_vptofh =		vop_stdvptofh,
138 	.vop_unp_bind =		vop_stdunp_bind,
139 	.vop_unp_connect =	vop_stdunp_connect,
140 	.vop_unp_detach =	vop_stdunp_detach,
141 	.vop_is_text =		vop_stdis_text,
142 	.vop_set_text =		vop_stdset_text,
143 	.vop_unset_text =	vop_stdunset_text,
144 	.vop_get_writecount =	vop_stdget_writecount,
145 	.vop_add_writecount =	vop_stdadd_writecount,
146 };
147 
148 /*
149  * Series of placeholder functions for various error returns for
150  * VOPs.
151  */
152 
153 int
154 vop_eopnotsupp(struct vop_generic_args *ap)
155 {
156 	/*
157 	printf("vop_notsupp[%s]\n", ap->a_desc->vdesc_name);
158 	*/
159 
160 	return (EOPNOTSUPP);
161 }
162 
163 int
164 vop_ebadf(struct vop_generic_args *ap)
165 {
166 
167 	return (EBADF);
168 }
169 
170 int
171 vop_enotty(struct vop_generic_args *ap)
172 {
173 
174 	return (ENOTTY);
175 }
176 
177 int
178 vop_einval(struct vop_generic_args *ap)
179 {
180 
181 	return (EINVAL);
182 }
183 
184 int
185 vop_enoent(struct vop_generic_args *ap)
186 {
187 
188 	return (ENOENT);
189 }
190 
191 int
192 vop_null(struct vop_generic_args *ap)
193 {
194 
195 	return (0);
196 }
197 
198 /*
199  * Helper function to panic on some bad VOPs in some filesystems.
200  */
201 int
202 vop_panic(struct vop_generic_args *ap)
203 {
204 
205 	panic("filesystem goof: vop_panic[%s]", ap->a_desc->vdesc_name);
206 }
207 
208 /*
209  * vop_std<something> and vop_no<something> are default functions for use by
210  * filesystems that need the "default reasonable" implementation for a
211  * particular operation.
212  *
213  * The documentation for the operations they implement exists (if it exists)
214  * in the VOP_<SOMETHING>(9) manpage (all uppercase).
215  */
216 
217 /*
218  * Default vop for filesystems that do not support name lookup
219  */
220 static int
221 vop_nolookup(ap)
222 	struct vop_lookup_args /* {
223 		struct vnode *a_dvp;
224 		struct vnode **a_vpp;
225 		struct componentname *a_cnp;
226 	} */ *ap;
227 {
228 
229 	*ap->a_vpp = NULL;
230 	return (ENOTDIR);
231 }
232 
233 /*
234  * vop_norename:
235  *
236  * Handle unlock and reference counting for arguments of vop_rename
237  * for filesystems that do not implement rename operation.
238  */
239 static int
240 vop_norename(struct vop_rename_args *ap)
241 {
242 
243 	vop_rename_fail(ap);
244 	return (EOPNOTSUPP);
245 }
246 
247 /*
248  *	vop_nostrategy:
249  *
250  *	Strategy routine for VFS devices that have none.
251  *
252  *	BIO_ERROR and B_INVAL must be cleared prior to calling any strategy
253  *	routine.  Typically this is done for a BIO_READ strategy call.
254  *	Typically B_INVAL is assumed to already be clear prior to a write
255  *	and should not be cleared manually unless you just made the buffer
256  *	invalid.  BIO_ERROR should be cleared either way.
257  */
258 
259 static int
260 vop_nostrategy (struct vop_strategy_args *ap)
261 {
262 	printf("No strategy for buffer at %p\n", ap->a_bp);
263 	vn_printf(ap->a_vp, "vnode ");
264 	ap->a_bp->b_ioflags |= BIO_ERROR;
265 	ap->a_bp->b_error = EOPNOTSUPP;
266 	bufdone(ap->a_bp);
267 	return (EOPNOTSUPP);
268 }
269 
270 static int
271 get_next_dirent(struct vnode *vp, struct dirent **dpp, char *dirbuf,
272 		int dirbuflen, off_t *off, char **cpos, int *len,
273 		int *eofflag, struct thread *td)
274 {
275 	int error, reclen;
276 	struct uio uio;
277 	struct iovec iov;
278 	struct dirent *dp;
279 
280 	KASSERT(VOP_ISLOCKED(vp), ("vp %p is not locked", vp));
281 	KASSERT(vp->v_type == VDIR, ("vp %p is not a directory", vp));
282 
283 	if (*len == 0) {
284 		iov.iov_base = dirbuf;
285 		iov.iov_len = dirbuflen;
286 
287 		uio.uio_iov = &iov;
288 		uio.uio_iovcnt = 1;
289 		uio.uio_offset = *off;
290 		uio.uio_resid = dirbuflen;
291 		uio.uio_segflg = UIO_SYSSPACE;
292 		uio.uio_rw = UIO_READ;
293 		uio.uio_td = td;
294 
295 		*eofflag = 0;
296 
297 #ifdef MAC
298 		error = mac_vnode_check_readdir(td->td_ucred, vp);
299 		if (error == 0)
300 #endif
301 			error = VOP_READDIR(vp, &uio, td->td_ucred, eofflag,
302 		    		NULL, NULL);
303 		if (error)
304 			return (error);
305 
306 		*off = uio.uio_offset;
307 
308 		*cpos = dirbuf;
309 		*len = (dirbuflen - uio.uio_resid);
310 
311 		if (*len == 0)
312 			return (ENOENT);
313 	}
314 
315 	dp = (struct dirent *)(*cpos);
316 	reclen = dp->d_reclen;
317 	*dpp = dp;
318 
319 	/* check for malformed directory.. */
320 	if (reclen < DIRENT_MINSIZE)
321 		return (EINVAL);
322 
323 	*cpos += reclen;
324 	*len -= reclen;
325 
326 	return (0);
327 }
328 
329 /*
330  * Check if a named file exists in a given directory vnode.
331  */
332 static int
333 dirent_exists(struct vnode *vp, const char *dirname, struct thread *td)
334 {
335 	char *dirbuf, *cpos;
336 	int error, eofflag, dirbuflen, len, found;
337 	off_t off;
338 	struct dirent *dp;
339 	struct vattr va;
340 
341 	KASSERT(VOP_ISLOCKED(vp), ("vp %p is not locked", vp));
342 	KASSERT(vp->v_type == VDIR, ("vp %p is not a directory", vp));
343 
344 	found = 0;
345 
346 	error = VOP_GETATTR(vp, &va, td->td_ucred);
347 	if (error)
348 		return (found);
349 
350 	dirbuflen = DEV_BSIZE;
351 	if (dirbuflen < va.va_blocksize)
352 		dirbuflen = va.va_blocksize;
353 	dirbuf = (char *)malloc(dirbuflen, M_TEMP, M_WAITOK);
354 
355 	off = 0;
356 	len = 0;
357 	do {
358 		error = get_next_dirent(vp, &dp, dirbuf, dirbuflen, &off,
359 					&cpos, &len, &eofflag, td);
360 		if (error)
361 			goto out;
362 
363 		if (dp->d_type != DT_WHT && dp->d_fileno != 0 &&
364 		    strcmp(dp->d_name, dirname) == 0) {
365 			found = 1;
366 			goto out;
367 		}
368 	} while (len > 0 || !eofflag);
369 
370 out:
371 	free(dirbuf, M_TEMP);
372 	return (found);
373 }
374 
375 int
376 vop_stdaccess(struct vop_access_args *ap)
377 {
378 
379 	KASSERT((ap->a_accmode & ~(VEXEC | VWRITE | VREAD | VADMIN |
380 	    VAPPEND)) == 0, ("invalid bit in accmode"));
381 
382 	return (VOP_ACCESSX(ap->a_vp, ap->a_accmode, ap->a_cred, ap->a_td));
383 }
384 
385 int
386 vop_stdaccessx(struct vop_accessx_args *ap)
387 {
388 	int error;
389 	accmode_t accmode = ap->a_accmode;
390 
391 	error = vfs_unixify_accmode(&accmode);
392 	if (error != 0)
393 		return (error);
394 
395 	if (accmode == 0)
396 		return (0);
397 
398 	return (VOP_ACCESS(ap->a_vp, accmode, ap->a_cred, ap->a_td));
399 }
400 
401 /*
402  * Advisory record locking support
403  */
404 int
405 vop_stdadvlock(struct vop_advlock_args *ap)
406 {
407 	struct vnode *vp;
408 	struct vattr vattr;
409 	int error;
410 
411 	vp = ap->a_vp;
412 	if (ap->a_fl->l_whence == SEEK_END) {
413 		/*
414 		 * The NFSv4 server must avoid doing a vn_lock() here, since it
415 		 * can deadlock the nfsd threads, due to a LOR.  Fortunately
416 		 * the NFSv4 server always uses SEEK_SET and this code is
417 		 * only required for the SEEK_END case.
418 		 */
419 		vn_lock(vp, LK_SHARED | LK_RETRY);
420 		error = VOP_GETATTR(vp, &vattr, curthread->td_ucred);
421 		VOP_UNLOCK(vp, 0);
422 		if (error)
423 			return (error);
424 	} else
425 		vattr.va_size = 0;
426 
427 	return (lf_advlock(ap, &(vp->v_lockf), vattr.va_size));
428 }
429 
430 int
431 vop_stdadvlockasync(struct vop_advlockasync_args *ap)
432 {
433 	struct vnode *vp;
434 	struct vattr vattr;
435 	int error;
436 
437 	vp = ap->a_vp;
438 	if (ap->a_fl->l_whence == SEEK_END) {
439 		/* The size argument is only needed for SEEK_END. */
440 		vn_lock(vp, LK_SHARED | LK_RETRY);
441 		error = VOP_GETATTR(vp, &vattr, curthread->td_ucred);
442 		VOP_UNLOCK(vp, 0);
443 		if (error)
444 			return (error);
445 	} else
446 		vattr.va_size = 0;
447 
448 	return (lf_advlockasync(ap, &(vp->v_lockf), vattr.va_size));
449 }
450 
451 int
452 vop_stdadvlockpurge(struct vop_advlockpurge_args *ap)
453 {
454 	struct vnode *vp;
455 
456 	vp = ap->a_vp;
457 	lf_purgelocks(vp, &vp->v_lockf);
458 	return (0);
459 }
460 
461 /*
462  * vop_stdpathconf:
463  *
464  * Standard implementation of POSIX pathconf, to get information about limits
465  * for a filesystem.
466  * Override per filesystem for the case where the filesystem has smaller
467  * limits.
468  */
469 int
470 vop_stdpathconf(ap)
471 	struct vop_pathconf_args /* {
472 	struct vnode *a_vp;
473 	int a_name;
474 	int *a_retval;
475 	} */ *ap;
476 {
477 
478 	switch (ap->a_name) {
479 		case _PC_ASYNC_IO:
480 			*ap->a_retval = _POSIX_ASYNCHRONOUS_IO;
481 			return (0);
482 		case _PC_PATH_MAX:
483 			*ap->a_retval = PATH_MAX;
484 			return (0);
485 		case _PC_ACL_EXTENDED:
486 		case _PC_ACL_NFS4:
487 		case _PC_CAP_PRESENT:
488 		case _PC_INF_PRESENT:
489 		case _PC_MAC_PRESENT:
490 			*ap->a_retval = 0;
491 			return (0);
492 		default:
493 			return (EINVAL);
494 	}
495 	/* NOTREACHED */
496 }
497 
498 /*
499  * Standard lock, unlock and islocked functions.
500  */
501 int
502 vop_stdlock(ap)
503 	struct vop_lock1_args /* {
504 		struct vnode *a_vp;
505 		int a_flags;
506 		char *file;
507 		int line;
508 	} */ *ap;
509 {
510 	struct vnode *vp = ap->a_vp;
511 	struct mtx *ilk;
512 
513 	ilk = VI_MTX(vp);
514 	return (lockmgr_lock_fast_path(vp->v_vnlock, ap->a_flags,
515 	    &ilk->lock_object, ap->a_file, ap->a_line));
516 }
517 
518 /* See above. */
519 int
520 vop_stdunlock(ap)
521 	struct vop_unlock_args /* {
522 		struct vnode *a_vp;
523 		int a_flags;
524 	} */ *ap;
525 {
526 	struct vnode *vp = ap->a_vp;
527 	struct mtx *ilk;
528 
529 	ilk = VI_MTX(vp);
530 	return (lockmgr_unlock_fast_path(vp->v_vnlock, ap->a_flags,
531 	    &ilk->lock_object));
532 }
533 
534 /* See above. */
535 int
536 vop_stdislocked(ap)
537 	struct vop_islocked_args /* {
538 		struct vnode *a_vp;
539 	} */ *ap;
540 {
541 
542 	return (lockstatus(ap->a_vp->v_vnlock));
543 }
544 
545 /*
546  * Return true for select/poll.
547  */
548 int
549 vop_nopoll(ap)
550 	struct vop_poll_args /* {
551 		struct vnode *a_vp;
552 		int  a_events;
553 		struct ucred *a_cred;
554 		struct thread *a_td;
555 	} */ *ap;
556 {
557 
558 	return (poll_no_poll(ap->a_events));
559 }
560 
561 /*
562  * Implement poll for local filesystems that support it.
563  */
564 int
565 vop_stdpoll(ap)
566 	struct vop_poll_args /* {
567 		struct vnode *a_vp;
568 		int  a_events;
569 		struct ucred *a_cred;
570 		struct thread *a_td;
571 	} */ *ap;
572 {
573 	if (ap->a_events & ~POLLSTANDARD)
574 		return (vn_pollrecord(ap->a_vp, ap->a_td, ap->a_events));
575 	return (ap->a_events & (POLLIN | POLLOUT | POLLRDNORM | POLLWRNORM));
576 }
577 
578 /*
579  * Return our mount point, as we will take charge of the writes.
580  */
581 int
582 vop_stdgetwritemount(ap)
583 	struct vop_getwritemount_args /* {
584 		struct vnode *a_vp;
585 		struct mount **a_mpp;
586 	} */ *ap;
587 {
588 	struct mount *mp;
589 
590 	/*
591 	 * XXX Since this is called unlocked we may be recycled while
592 	 * attempting to ref the mount.  If this is the case or mountpoint
593 	 * will be set to NULL.  We only have to prevent this call from
594 	 * returning with a ref to an incorrect mountpoint.  It is not
595 	 * harmful to return with a ref to our previous mountpoint.
596 	 */
597 	mp = ap->a_vp->v_mount;
598 	if (mp != NULL) {
599 		vfs_ref(mp);
600 		if (mp != ap->a_vp->v_mount) {
601 			vfs_rel(mp);
602 			mp = NULL;
603 		}
604 	}
605 	*(ap->a_mpp) = mp;
606 	return (0);
607 }
608 
609 /* XXX Needs good comment and VOP_BMAP(9) manpage */
610 int
611 vop_stdbmap(ap)
612 	struct vop_bmap_args /* {
613 		struct vnode *a_vp;
614 		daddr_t  a_bn;
615 		struct bufobj **a_bop;
616 		daddr_t *a_bnp;
617 		int *a_runp;
618 		int *a_runb;
619 	} */ *ap;
620 {
621 
622 	if (ap->a_bop != NULL)
623 		*ap->a_bop = &ap->a_vp->v_bufobj;
624 	if (ap->a_bnp != NULL)
625 		*ap->a_bnp = ap->a_bn * btodb(ap->a_vp->v_mount->mnt_stat.f_iosize);
626 	if (ap->a_runp != NULL)
627 		*ap->a_runp = 0;
628 	if (ap->a_runb != NULL)
629 		*ap->a_runb = 0;
630 	return (0);
631 }
632 
633 int
634 vop_stdfsync(ap)
635 	struct vop_fsync_args /* {
636 		struct vnode *a_vp;
637 		int a_waitfor;
638 		struct thread *a_td;
639 	} */ *ap;
640 {
641 
642 	return (vn_fsync_buf(ap->a_vp, ap->a_waitfor));
643 }
644 
645 static int
646 vop_stdfdatasync(struct vop_fdatasync_args *ap)
647 {
648 
649 	return (VOP_FSYNC(ap->a_vp, MNT_WAIT, ap->a_td));
650 }
651 
652 int
653 vop_stdfdatasync_buf(struct vop_fdatasync_args *ap)
654 {
655 
656 	return (vn_fsync_buf(ap->a_vp, MNT_WAIT));
657 }
658 
659 /* XXX Needs good comment and more info in the manpage (VOP_GETPAGES(9)). */
660 int
661 vop_stdgetpages(ap)
662 	struct vop_getpages_args /* {
663 		struct vnode *a_vp;
664 		vm_page_t *a_m;
665 		int a_count;
666 		int *a_rbehind;
667 		int *a_rahead;
668 	} */ *ap;
669 {
670 
671 	return vnode_pager_generic_getpages(ap->a_vp, ap->a_m,
672 	    ap->a_count, ap->a_rbehind, ap->a_rahead, NULL, NULL);
673 }
674 
675 static int
676 vop_stdgetpages_async(struct vop_getpages_async_args *ap)
677 {
678 	int error;
679 
680 	error = VOP_GETPAGES(ap->a_vp, ap->a_m, ap->a_count, ap->a_rbehind,
681 	    ap->a_rahead);
682 	ap->a_iodone(ap->a_arg, ap->a_m, ap->a_count, error);
683 	return (error);
684 }
685 
686 int
687 vop_stdkqfilter(struct vop_kqfilter_args *ap)
688 {
689 	return vfs_kqfilter(ap);
690 }
691 
692 /* XXX Needs good comment and more info in the manpage (VOP_PUTPAGES(9)). */
693 int
694 vop_stdputpages(ap)
695 	struct vop_putpages_args /* {
696 		struct vnode *a_vp;
697 		vm_page_t *a_m;
698 		int a_count;
699 		int a_sync;
700 		int *a_rtvals;
701 	} */ *ap;
702 {
703 
704 	return vnode_pager_generic_putpages(ap->a_vp, ap->a_m, ap->a_count,
705 	     ap->a_sync, ap->a_rtvals);
706 }
707 
708 int
709 vop_stdvptofh(struct vop_vptofh_args *ap)
710 {
711 	return (EOPNOTSUPP);
712 }
713 
714 int
715 vop_stdvptocnp(struct vop_vptocnp_args *ap)
716 {
717 	struct vnode *vp = ap->a_vp;
718 	struct vnode **dvp = ap->a_vpp;
719 	struct ucred *cred = ap->a_cred;
720 	char *buf = ap->a_buf;
721 	int *buflen = ap->a_buflen;
722 	char *dirbuf, *cpos;
723 	int i, error, eofflag, dirbuflen, flags, locked, len, covered;
724 	off_t off;
725 	ino_t fileno;
726 	struct vattr va;
727 	struct nameidata nd;
728 	struct thread *td;
729 	struct dirent *dp;
730 	struct vnode *mvp;
731 
732 	i = *buflen;
733 	error = 0;
734 	covered = 0;
735 	td = curthread;
736 
737 	if (vp->v_type != VDIR)
738 		return (ENOENT);
739 
740 	error = VOP_GETATTR(vp, &va, cred);
741 	if (error)
742 		return (error);
743 
744 	VREF(vp);
745 	locked = VOP_ISLOCKED(vp);
746 	VOP_UNLOCK(vp, 0);
747 	NDINIT_ATVP(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF, UIO_SYSSPACE,
748 	    "..", vp, td);
749 	flags = FREAD;
750 	error = vn_open_cred(&nd, &flags, 0, VN_OPEN_NOAUDIT, cred, NULL);
751 	if (error) {
752 		vn_lock(vp, locked | LK_RETRY);
753 		return (error);
754 	}
755 	NDFREE(&nd, NDF_ONLY_PNBUF);
756 
757 	mvp = *dvp = nd.ni_vp;
758 
759 	if (vp->v_mount != (*dvp)->v_mount &&
760 	    ((*dvp)->v_vflag & VV_ROOT) &&
761 	    ((*dvp)->v_mount->mnt_flag & MNT_UNION)) {
762 		*dvp = (*dvp)->v_mount->mnt_vnodecovered;
763 		VREF(mvp);
764 		VOP_UNLOCK(mvp, 0);
765 		vn_close(mvp, FREAD, cred, td);
766 		VREF(*dvp);
767 		vn_lock(*dvp, LK_SHARED | LK_RETRY);
768 		covered = 1;
769 	}
770 
771 	fileno = va.va_fileid;
772 
773 	dirbuflen = DEV_BSIZE;
774 	if (dirbuflen < va.va_blocksize)
775 		dirbuflen = va.va_blocksize;
776 	dirbuf = (char *)malloc(dirbuflen, M_TEMP, M_WAITOK);
777 
778 	if ((*dvp)->v_type != VDIR) {
779 		error = ENOENT;
780 		goto out;
781 	}
782 
783 	off = 0;
784 	len = 0;
785 	do {
786 		/* call VOP_READDIR of parent */
787 		error = get_next_dirent(*dvp, &dp, dirbuf, dirbuflen, &off,
788 					&cpos, &len, &eofflag, td);
789 		if (error)
790 			goto out;
791 
792 		if ((dp->d_type != DT_WHT) &&
793 		    (dp->d_fileno == fileno)) {
794 			if (covered) {
795 				VOP_UNLOCK(*dvp, 0);
796 				vn_lock(mvp, LK_SHARED | LK_RETRY);
797 				if (dirent_exists(mvp, dp->d_name, td)) {
798 					error = ENOENT;
799 					VOP_UNLOCK(mvp, 0);
800 					vn_lock(*dvp, LK_SHARED | LK_RETRY);
801 					goto out;
802 				}
803 				VOP_UNLOCK(mvp, 0);
804 				vn_lock(*dvp, LK_SHARED | LK_RETRY);
805 			}
806 			i -= dp->d_namlen;
807 
808 			if (i < 0) {
809 				error = ENOMEM;
810 				goto out;
811 			}
812 			if (dp->d_namlen == 1 && dp->d_name[0] == '.') {
813 				error = ENOENT;
814 			} else {
815 				bcopy(dp->d_name, buf + i, dp->d_namlen);
816 				error = 0;
817 			}
818 			goto out;
819 		}
820 	} while (len > 0 || !eofflag);
821 	error = ENOENT;
822 
823 out:
824 	free(dirbuf, M_TEMP);
825 	if (!error) {
826 		*buflen = i;
827 		vref(*dvp);
828 	}
829 	if (covered) {
830 		vput(*dvp);
831 		vrele(mvp);
832 	} else {
833 		VOP_UNLOCK(mvp, 0);
834 		vn_close(mvp, FREAD, cred, td);
835 	}
836 	vn_lock(vp, locked | LK_RETRY);
837 	return (error);
838 }
839 
840 int
841 vop_stdallocate(struct vop_allocate_args *ap)
842 {
843 #ifdef __notyet__
844 	struct statfs *sfs;
845 	off_t maxfilesize = 0;
846 #endif
847 	struct iovec aiov;
848 	struct vattr vattr, *vap;
849 	struct uio auio;
850 	off_t fsize, len, cur, offset;
851 	uint8_t *buf;
852 	struct thread *td;
853 	struct vnode *vp;
854 	size_t iosize;
855 	int error;
856 
857 	buf = NULL;
858 	error = 0;
859 	td = curthread;
860 	vap = &vattr;
861 	vp = ap->a_vp;
862 	len = *ap->a_len;
863 	offset = *ap->a_offset;
864 
865 	error = VOP_GETATTR(vp, vap, td->td_ucred);
866 	if (error != 0)
867 		goto out;
868 	fsize = vap->va_size;
869 	iosize = vap->va_blocksize;
870 	if (iosize == 0)
871 		iosize = BLKDEV_IOSIZE;
872 	if (iosize > MAXPHYS)
873 		iosize = MAXPHYS;
874 	buf = malloc(iosize, M_TEMP, M_WAITOK);
875 
876 #ifdef __notyet__
877 	/*
878 	 * Check if the filesystem sets f_maxfilesize; if not use
879 	 * VOP_SETATTR to perform the check.
880 	 */
881 	sfs = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
882 	error = VFS_STATFS(vp->v_mount, sfs, td);
883 	if (error == 0)
884 		maxfilesize = sfs->f_maxfilesize;
885 	free(sfs, M_STATFS);
886 	if (error != 0)
887 		goto out;
888 	if (maxfilesize) {
889 		if (offset > maxfilesize || len > maxfilesize ||
890 		    offset + len > maxfilesize) {
891 			error = EFBIG;
892 			goto out;
893 		}
894 	} else
895 #endif
896 	if (offset + len > vap->va_size) {
897 		/*
898 		 * Test offset + len against the filesystem's maxfilesize.
899 		 */
900 		VATTR_NULL(vap);
901 		vap->va_size = offset + len;
902 		error = VOP_SETATTR(vp, vap, td->td_ucred);
903 		if (error != 0)
904 			goto out;
905 		VATTR_NULL(vap);
906 		vap->va_size = fsize;
907 		error = VOP_SETATTR(vp, vap, td->td_ucred);
908 		if (error != 0)
909 			goto out;
910 	}
911 
912 	for (;;) {
913 		/*
914 		 * Read and write back anything below the nominal file
915 		 * size.  There's currently no way outside the filesystem
916 		 * to know whether this area is sparse or not.
917 		 */
918 		cur = iosize;
919 		if ((offset % iosize) != 0)
920 			cur -= (offset % iosize);
921 		if (cur > len)
922 			cur = len;
923 		if (offset < fsize) {
924 			aiov.iov_base = buf;
925 			aiov.iov_len = cur;
926 			auio.uio_iov = &aiov;
927 			auio.uio_iovcnt = 1;
928 			auio.uio_offset = offset;
929 			auio.uio_resid = cur;
930 			auio.uio_segflg = UIO_SYSSPACE;
931 			auio.uio_rw = UIO_READ;
932 			auio.uio_td = td;
933 			error = VOP_READ(vp, &auio, 0, td->td_ucred);
934 			if (error != 0)
935 				break;
936 			if (auio.uio_resid > 0) {
937 				bzero(buf + cur - auio.uio_resid,
938 				    auio.uio_resid);
939 			}
940 		} else {
941 			bzero(buf, cur);
942 		}
943 
944 		aiov.iov_base = buf;
945 		aiov.iov_len = cur;
946 		auio.uio_iov = &aiov;
947 		auio.uio_iovcnt = 1;
948 		auio.uio_offset = offset;
949 		auio.uio_resid = cur;
950 		auio.uio_segflg = UIO_SYSSPACE;
951 		auio.uio_rw = UIO_WRITE;
952 		auio.uio_td = td;
953 
954 		error = VOP_WRITE(vp, &auio, 0, td->td_ucred);
955 		if (error != 0)
956 			break;
957 
958 		len -= cur;
959 		offset += cur;
960 		if (len == 0)
961 			break;
962 		if (should_yield())
963 			break;
964 	}
965 
966  out:
967 	*ap->a_len = len;
968 	*ap->a_offset = offset;
969 	free(buf, M_TEMP);
970 	return (error);
971 }
972 
973 int
974 vop_stdadvise(struct vop_advise_args *ap)
975 {
976 	struct vnode *vp;
977 	struct bufobj *bo;
978 	daddr_t startn, endn;
979 	off_t bstart, bend, start, end;
980 	int bsize, error;
981 
982 	vp = ap->a_vp;
983 	switch (ap->a_advice) {
984 	case POSIX_FADV_WILLNEED:
985 		/*
986 		 * Do nothing for now.  Filesystems should provide a
987 		 * custom method which starts an asynchronous read of
988 		 * the requested region.
989 		 */
990 		error = 0;
991 		break;
992 	case POSIX_FADV_DONTNEED:
993 		error = 0;
994 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
995 		if (vp->v_iflag & VI_DOOMED) {
996 			VOP_UNLOCK(vp, 0);
997 			break;
998 		}
999 
1000 		/*
1001 		 * Round to block boundaries (and later possibly further to
1002 		 * page boundaries).  Applications cannot reasonably be aware
1003 		 * of the boundaries, and the rounding must be to expand at
1004 		 * both extremities to cover enough.  It still doesn't cover
1005 		 * read-ahead.  For partial blocks, this gives unnecessary
1006 		 * discarding of buffers but is efficient enough since the
1007 		 * pages usually remain in VMIO for some time.
1008 		 */
1009 		bsize = vp->v_bufobj.bo_bsize;
1010 		bstart = rounddown(ap->a_start, bsize);
1011 		bend = roundup(ap->a_end, bsize);
1012 
1013 		/*
1014 		 * Deactivate pages in the specified range from the backing VM
1015 		 * object.  Pages that are resident in the buffer cache will
1016 		 * remain wired until their corresponding buffers are released
1017 		 * below.
1018 		 */
1019 		if (vp->v_object != NULL) {
1020 			start = trunc_page(bstart);
1021 			end = round_page(bend);
1022 			VM_OBJECT_RLOCK(vp->v_object);
1023 			vm_object_page_noreuse(vp->v_object, OFF_TO_IDX(start),
1024 			    OFF_TO_IDX(end));
1025 			VM_OBJECT_RUNLOCK(vp->v_object);
1026 		}
1027 
1028 		bo = &vp->v_bufobj;
1029 		BO_RLOCK(bo);
1030 		startn = bstart / bsize;
1031 		endn = bend / bsize;
1032 		error = bnoreuselist(&bo->bo_clean, bo, startn, endn);
1033 		if (error == 0)
1034 			error = bnoreuselist(&bo->bo_dirty, bo, startn, endn);
1035 		BO_RUNLOCK(bo);
1036 		VOP_UNLOCK(vp, 0);
1037 		break;
1038 	default:
1039 		error = EINVAL;
1040 		break;
1041 	}
1042 	return (error);
1043 }
1044 
1045 int
1046 vop_stdunp_bind(struct vop_unp_bind_args *ap)
1047 {
1048 
1049 	ap->a_vp->v_unpcb = ap->a_unpcb;
1050 	return (0);
1051 }
1052 
1053 int
1054 vop_stdunp_connect(struct vop_unp_connect_args *ap)
1055 {
1056 
1057 	*ap->a_unpcb = ap->a_vp->v_unpcb;
1058 	return (0);
1059 }
1060 
1061 int
1062 vop_stdunp_detach(struct vop_unp_detach_args *ap)
1063 {
1064 
1065 	ap->a_vp->v_unpcb = NULL;
1066 	return (0);
1067 }
1068 
1069 static int
1070 vop_stdis_text(struct vop_is_text_args *ap)
1071 {
1072 
1073 	return ((ap->a_vp->v_vflag & VV_TEXT) != 0);
1074 }
1075 
1076 static int
1077 vop_stdset_text(struct vop_set_text_args *ap)
1078 {
1079 
1080 	ap->a_vp->v_vflag |= VV_TEXT;
1081 	return (0);
1082 }
1083 
1084 static int
1085 vop_stdunset_text(struct vop_unset_text_args *ap)
1086 {
1087 
1088 	ap->a_vp->v_vflag &= ~VV_TEXT;
1089 	return (0);
1090 }
1091 
1092 static int
1093 vop_stdget_writecount(struct vop_get_writecount_args *ap)
1094 {
1095 
1096 	*ap->a_writecount = ap->a_vp->v_writecount;
1097 	return (0);
1098 }
1099 
1100 static int
1101 vop_stdadd_writecount(struct vop_add_writecount_args *ap)
1102 {
1103 
1104 	ap->a_vp->v_writecount += ap->a_inc;
1105 	return (0);
1106 }
1107 
1108 /*
1109  * vfs default ops
1110  * used to fill the vfs function table to get reasonable default return values.
1111  */
1112 int
1113 vfs_stdroot (mp, flags, vpp)
1114 	struct mount *mp;
1115 	int flags;
1116 	struct vnode **vpp;
1117 {
1118 
1119 	return (EOPNOTSUPP);
1120 }
1121 
1122 int
1123 vfs_stdstatfs (mp, sbp)
1124 	struct mount *mp;
1125 	struct statfs *sbp;
1126 {
1127 
1128 	return (EOPNOTSUPP);
1129 }
1130 
1131 int
1132 vfs_stdquotactl (mp, cmds, uid, arg)
1133 	struct mount *mp;
1134 	int cmds;
1135 	uid_t uid;
1136 	void *arg;
1137 {
1138 
1139 	return (EOPNOTSUPP);
1140 }
1141 
1142 int
1143 vfs_stdsync(mp, waitfor)
1144 	struct mount *mp;
1145 	int waitfor;
1146 {
1147 	struct vnode *vp, *mvp;
1148 	struct thread *td;
1149 	int error, lockreq, allerror = 0;
1150 
1151 	td = curthread;
1152 	lockreq = LK_EXCLUSIVE | LK_INTERLOCK;
1153 	if (waitfor != MNT_WAIT)
1154 		lockreq |= LK_NOWAIT;
1155 	/*
1156 	 * Force stale buffer cache information to be flushed.
1157 	 */
1158 loop:
1159 	MNT_VNODE_FOREACH_ALL(vp, mp, mvp) {
1160 		if (vp->v_bufobj.bo_dirty.bv_cnt == 0) {
1161 			VI_UNLOCK(vp);
1162 			continue;
1163 		}
1164 		if ((error = vget(vp, lockreq, td)) != 0) {
1165 			if (error == ENOENT) {
1166 				MNT_VNODE_FOREACH_ALL_ABORT(mp, mvp);
1167 				goto loop;
1168 			}
1169 			continue;
1170 		}
1171 		error = VOP_FSYNC(vp, waitfor, td);
1172 		if (error)
1173 			allerror = error;
1174 		vput(vp);
1175 	}
1176 	return (allerror);
1177 }
1178 
1179 int
1180 vfs_stdnosync (mp, waitfor)
1181 	struct mount *mp;
1182 	int waitfor;
1183 {
1184 
1185 	return (0);
1186 }
1187 
1188 int
1189 vfs_stdvget (mp, ino, flags, vpp)
1190 	struct mount *mp;
1191 	ino_t ino;
1192 	int flags;
1193 	struct vnode **vpp;
1194 {
1195 
1196 	return (EOPNOTSUPP);
1197 }
1198 
1199 int
1200 vfs_stdfhtovp (mp, fhp, flags, vpp)
1201 	struct mount *mp;
1202 	struct fid *fhp;
1203 	int flags;
1204 	struct vnode **vpp;
1205 {
1206 
1207 	return (EOPNOTSUPP);
1208 }
1209 
1210 int
1211 vfs_stdinit (vfsp)
1212 	struct vfsconf *vfsp;
1213 {
1214 
1215 	return (0);
1216 }
1217 
1218 int
1219 vfs_stduninit (vfsp)
1220 	struct vfsconf *vfsp;
1221 {
1222 
1223 	return(0);
1224 }
1225 
1226 int
1227 vfs_stdextattrctl(mp, cmd, filename_vp, attrnamespace, attrname)
1228 	struct mount *mp;
1229 	int cmd;
1230 	struct vnode *filename_vp;
1231 	int attrnamespace;
1232 	const char *attrname;
1233 {
1234 
1235 	if (filename_vp != NULL)
1236 		VOP_UNLOCK(filename_vp, 0);
1237 	return (EOPNOTSUPP);
1238 }
1239 
1240 int
1241 vfs_stdsysctl(mp, op, req)
1242 	struct mount *mp;
1243 	fsctlop_t op;
1244 	struct sysctl_req *req;
1245 {
1246 
1247 	return (EOPNOTSUPP);
1248 }
1249 
1250 static vop_bypass_t *
1251 bp_by_off(struct vop_vector *vop, struct vop_generic_args *a)
1252 {
1253 
1254 	return (*(vop_bypass_t **)((char *)vop + a->a_desc->vdesc_vop_offset));
1255 }
1256 
1257 int
1258 vop_sigdefer(struct vop_vector *vop, struct vop_generic_args *a)
1259 {
1260 	vop_bypass_t *bp;
1261 	int prev_stops, rc;
1262 
1263 	for (; vop != NULL; vop = vop->vop_default) {
1264 		bp = bp_by_off(vop, a);
1265 		if (bp != NULL)
1266 			break;
1267 
1268 		/*
1269 		 * Bypass is not really supported.  It is done for
1270 		 * fallback to unimplemented vops in the default
1271 		 * vector.
1272 		 */
1273 		bp = vop->vop_bypass;
1274 		if (bp != NULL)
1275 			break;
1276 	}
1277 	MPASS(bp != NULL);
1278 
1279 	prev_stops = sigdeferstop(SIGDEFERSTOP_SILENT);
1280 	rc = bp(a);
1281 	sigallowstop(prev_stops);
1282 	return (rc);
1283 }
1284