xref: /freebsd/sys/kern/vfs_default.c (revision d9a42747950146bf03cda7f6e25d219253f8a57a)
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/filio.h>
47 #include <sys/kernel.h>
48 #include <sys/limits.h>
49 #include <sys/lock.h>
50 #include <sys/lockf.h>
51 #include <sys/malloc.h>
52 #include <sys/mount.h>
53 #include <sys/namei.h>
54 #include <sys/rwlock.h>
55 #include <sys/fcntl.h>
56 #include <sys/unistd.h>
57 #include <sys/vnode.h>
58 #include <sys/dirent.h>
59 #include <sys/poll.h>
60 #include <sys/stat.h>
61 #include <security/audit/audit.h>
62 #include <sys/priv.h>
63 
64 #include <security/mac/mac_framework.h>
65 
66 #include <vm/vm.h>
67 #include <vm/vm_object.h>
68 #include <vm/vm_extern.h>
69 #include <vm/pmap.h>
70 #include <vm/vm_map.h>
71 #include <vm/vm_page.h>
72 #include <vm/vm_pager.h>
73 #include <vm/vnode_pager.h>
74 
75 static int	vop_nolookup(struct vop_lookup_args *);
76 static int	vop_norename(struct vop_rename_args *);
77 static int	vop_nostrategy(struct vop_strategy_args *);
78 static int	get_next_dirent(struct vnode *vp, struct dirent **dpp,
79 				char *dirbuf, int dirbuflen, off_t *off,
80 				char **cpos, int *len, int *eofflag,
81 				struct thread *td);
82 static int	dirent_exists(struct vnode *vp, const char *dirname,
83 			      struct thread *td);
84 
85 #define DIRENT_MINSIZE (sizeof(struct dirent) - (MAXNAMLEN+1) + 4)
86 
87 static int vop_stdis_text(struct vop_is_text_args *ap);
88 static int vop_stdunset_text(struct vop_unset_text_args *ap);
89 static int vop_stdadd_writecount(struct vop_add_writecount_args *ap);
90 static int vop_stdcopy_file_range(struct vop_copy_file_range_args *ap);
91 static int vop_stdfdatasync(struct vop_fdatasync_args *ap);
92 static int vop_stdgetpages_async(struct vop_getpages_async_args *ap);
93 static int vop_stdread_pgcache(struct vop_read_pgcache_args *ap);
94 static int vop_stdstat(struct vop_stat_args *ap);
95 static int vop_stdvput_pair(struct vop_vput_pair_args *ap);
96 
97 /*
98  * This vnode table stores what we want to do if the filesystem doesn't
99  * implement a particular VOP.
100  *
101  * If there is no specific entry here, we will return EOPNOTSUPP.
102  *
103  * Note that every filesystem has to implement either vop_access
104  * or vop_accessx; failing to do so will result in immediate crash
105  * due to stack overflow, as vop_stdaccess() calls vop_stdaccessx(),
106  * which calls vop_stdaccess() etc.
107  */
108 
109 struct vop_vector default_vnodeops = {
110 	.vop_default =		NULL,
111 	.vop_bypass =		VOP_EOPNOTSUPP,
112 
113 	.vop_access =		vop_stdaccess,
114 	.vop_accessx =		vop_stdaccessx,
115 	.vop_advise =		vop_stdadvise,
116 	.vop_advlock =		vop_stdadvlock,
117 	.vop_advlockasync =	vop_stdadvlockasync,
118 	.vop_advlockpurge =	vop_stdadvlockpurge,
119 	.vop_allocate =		vop_stdallocate,
120 	.vop_deallocate =	vop_stddeallocate,
121 	.vop_bmap =		vop_stdbmap,
122 	.vop_close =		VOP_NULL,
123 	.vop_fsync =		VOP_NULL,
124 	.vop_stat =		vop_stdstat,
125 	.vop_fdatasync =	vop_stdfdatasync,
126 	.vop_getpages =		vop_stdgetpages,
127 	.vop_getpages_async =	vop_stdgetpages_async,
128 	.vop_getwritemount = 	vop_stdgetwritemount,
129 	.vop_inactive =		VOP_NULL,
130 	.vop_need_inactive =	vop_stdneed_inactive,
131 	.vop_ioctl =		vop_stdioctl,
132 	.vop_kqfilter =		vop_stdkqfilter,
133 	.vop_islocked =		vop_stdislocked,
134 	.vop_lock1 =		vop_stdlock,
135 	.vop_lookup =		vop_nolookup,
136 	.vop_open =		VOP_NULL,
137 	.vop_pathconf =		VOP_EINVAL,
138 	.vop_poll =		vop_nopoll,
139 	.vop_putpages =		vop_stdputpages,
140 	.vop_readlink =		VOP_EINVAL,
141 	.vop_read_pgcache =	vop_stdread_pgcache,
142 	.vop_rename =		vop_norename,
143 	.vop_revoke =		VOP_PANIC,
144 	.vop_strategy =		vop_nostrategy,
145 	.vop_unlock =		vop_stdunlock,
146 	.vop_vptocnp =		vop_stdvptocnp,
147 	.vop_vptofh =		vop_stdvptofh,
148 	.vop_unp_bind =		vop_stdunp_bind,
149 	.vop_unp_connect =	vop_stdunp_connect,
150 	.vop_unp_detach =	vop_stdunp_detach,
151 	.vop_is_text =		vop_stdis_text,
152 	.vop_set_text =		vop_stdset_text,
153 	.vop_unset_text =	vop_stdunset_text,
154 	.vop_add_writecount =	vop_stdadd_writecount,
155 	.vop_copy_file_range =	vop_stdcopy_file_range,
156 	.vop_vput_pair =	vop_stdvput_pair,
157 };
158 VFS_VOP_VECTOR_REGISTER(default_vnodeops);
159 
160 /*
161  * Series of placeholder functions for various error returns for
162  * VOPs.
163  */
164 
165 int
166 vop_eopnotsupp(struct vop_generic_args *ap)
167 {
168 	/*
169 	printf("vop_notsupp[%s]\n", ap->a_desc->vdesc_name);
170 	*/
171 
172 	return (EOPNOTSUPP);
173 }
174 
175 int
176 vop_ebadf(struct vop_generic_args *ap)
177 {
178 
179 	return (EBADF);
180 }
181 
182 int
183 vop_enotty(struct vop_generic_args *ap)
184 {
185 
186 	return (ENOTTY);
187 }
188 
189 int
190 vop_einval(struct vop_generic_args *ap)
191 {
192 
193 	return (EINVAL);
194 }
195 
196 int
197 vop_enoent(struct vop_generic_args *ap)
198 {
199 
200 	return (ENOENT);
201 }
202 
203 int
204 vop_eagain(struct vop_generic_args *ap)
205 {
206 
207 	return (EAGAIN);
208 }
209 
210 int
211 vop_null(struct vop_generic_args *ap)
212 {
213 
214 	return (0);
215 }
216 
217 /*
218  * Helper function to panic on some bad VOPs in some filesystems.
219  */
220 int
221 vop_panic(struct vop_generic_args *ap)
222 {
223 
224 	panic("filesystem goof: vop_panic[%s]", ap->a_desc->vdesc_name);
225 }
226 
227 /*
228  * vop_std<something> and vop_no<something> are default functions for use by
229  * filesystems that need the "default reasonable" implementation for a
230  * particular operation.
231  *
232  * The documentation for the operations they implement exists (if it exists)
233  * in the VOP_<SOMETHING>(9) manpage (all uppercase).
234  */
235 
236 /*
237  * Default vop for filesystems that do not support name lookup
238  */
239 static int
240 vop_nolookup(struct vop_lookup_args *ap)
241 {
242 
243 	*ap->a_vpp = NULL;
244 	return (ENOTDIR);
245 }
246 
247 /*
248  * vop_norename:
249  *
250  * Handle unlock and reference counting for arguments of vop_rename
251  * for filesystems that do not implement rename operation.
252  */
253 static int
254 vop_norename(struct vop_rename_args *ap)
255 {
256 
257 	vop_rename_fail(ap);
258 	return (EOPNOTSUPP);
259 }
260 
261 /*
262  *	vop_nostrategy:
263  *
264  *	Strategy routine for VFS devices that have none.
265  *
266  *	BIO_ERROR and B_INVAL must be cleared prior to calling any strategy
267  *	routine.  Typically this is done for a BIO_READ strategy call.
268  *	Typically B_INVAL is assumed to already be clear prior to a write
269  *	and should not be cleared manually unless you just made the buffer
270  *	invalid.  BIO_ERROR should be cleared either way.
271  */
272 
273 static int
274 vop_nostrategy (struct vop_strategy_args *ap)
275 {
276 	printf("No strategy for buffer at %p\n", ap->a_bp);
277 	vn_printf(ap->a_vp, "vnode ");
278 	ap->a_bp->b_ioflags |= BIO_ERROR;
279 	ap->a_bp->b_error = EOPNOTSUPP;
280 	bufdone(ap->a_bp);
281 	return (EOPNOTSUPP);
282 }
283 
284 static int
285 get_next_dirent(struct vnode *vp, struct dirent **dpp, char *dirbuf,
286 		int dirbuflen, off_t *off, char **cpos, int *len,
287 		int *eofflag, struct thread *td)
288 {
289 	int error, reclen;
290 	struct uio uio;
291 	struct iovec iov;
292 	struct dirent *dp;
293 
294 	KASSERT(VOP_ISLOCKED(vp), ("vp %p is not locked", vp));
295 	KASSERT(vp->v_type == VDIR, ("vp %p is not a directory", vp));
296 
297 	if (*len == 0) {
298 		iov.iov_base = dirbuf;
299 		iov.iov_len = dirbuflen;
300 
301 		uio.uio_iov = &iov;
302 		uio.uio_iovcnt = 1;
303 		uio.uio_offset = *off;
304 		uio.uio_resid = dirbuflen;
305 		uio.uio_segflg = UIO_SYSSPACE;
306 		uio.uio_rw = UIO_READ;
307 		uio.uio_td = td;
308 
309 		*eofflag = 0;
310 
311 #ifdef MAC
312 		error = mac_vnode_check_readdir(td->td_ucred, vp);
313 		if (error == 0)
314 #endif
315 			error = VOP_READDIR(vp, &uio, td->td_ucred, eofflag,
316 		    		NULL, NULL);
317 		if (error)
318 			return (error);
319 
320 		*off = uio.uio_offset;
321 
322 		*cpos = dirbuf;
323 		*len = (dirbuflen - uio.uio_resid);
324 
325 		if (*len == 0)
326 			return (ENOENT);
327 	}
328 
329 	dp = (struct dirent *)(*cpos);
330 	reclen = dp->d_reclen;
331 	*dpp = dp;
332 
333 	/* check for malformed directory.. */
334 	if (reclen < DIRENT_MINSIZE)
335 		return (EINVAL);
336 
337 	*cpos += reclen;
338 	*len -= reclen;
339 
340 	return (0);
341 }
342 
343 /*
344  * Check if a named file exists in a given directory vnode.
345  */
346 static int
347 dirent_exists(struct vnode *vp, const char *dirname, struct thread *td)
348 {
349 	char *dirbuf, *cpos;
350 	int error, eofflag, dirbuflen, len, found;
351 	off_t off;
352 	struct dirent *dp;
353 	struct vattr va;
354 
355 	KASSERT(VOP_ISLOCKED(vp), ("vp %p is not locked", vp));
356 	KASSERT(vp->v_type == VDIR, ("vp %p is not a directory", vp));
357 
358 	found = 0;
359 
360 	error = VOP_GETATTR(vp, &va, td->td_ucred);
361 	if (error)
362 		return (found);
363 
364 	dirbuflen = DEV_BSIZE;
365 	if (dirbuflen < va.va_blocksize)
366 		dirbuflen = va.va_blocksize;
367 	dirbuf = (char *)malloc(dirbuflen, M_TEMP, M_WAITOK);
368 
369 	off = 0;
370 	len = 0;
371 	do {
372 		error = get_next_dirent(vp, &dp, dirbuf, dirbuflen, &off,
373 					&cpos, &len, &eofflag, td);
374 		if (error)
375 			goto out;
376 
377 		if (dp->d_type != DT_WHT && dp->d_fileno != 0 &&
378 		    strcmp(dp->d_name, dirname) == 0) {
379 			found = 1;
380 			goto out;
381 		}
382 	} while (len > 0 || !eofflag);
383 
384 out:
385 	free(dirbuf, M_TEMP);
386 	return (found);
387 }
388 
389 int
390 vop_stdaccess(struct vop_access_args *ap)
391 {
392 
393 	KASSERT((ap->a_accmode & ~(VEXEC | VWRITE | VREAD | VADMIN |
394 	    VAPPEND)) == 0, ("invalid bit in accmode"));
395 
396 	return (VOP_ACCESSX(ap->a_vp, ap->a_accmode, ap->a_cred, ap->a_td));
397 }
398 
399 int
400 vop_stdaccessx(struct vop_accessx_args *ap)
401 {
402 	int error;
403 	accmode_t accmode = ap->a_accmode;
404 
405 	error = vfs_unixify_accmode(&accmode);
406 	if (error != 0)
407 		return (error);
408 
409 	if (accmode == 0)
410 		return (0);
411 
412 	return (VOP_ACCESS(ap->a_vp, accmode, ap->a_cred, ap->a_td));
413 }
414 
415 /*
416  * Advisory record locking support
417  */
418 int
419 vop_stdadvlock(struct vop_advlock_args *ap)
420 {
421 	struct vnode *vp;
422 	struct mount *mp;
423 	struct vattr vattr;
424 	int error;
425 
426 	vp = ap->a_vp;
427 
428 	/*
429 	 * Provide atomicity of open(O_CREAT | O_EXCL | O_EXLOCK) for
430 	 * local filesystems.  See vn_open_cred() for reciprocal part.
431 	 */
432 	mp = vp->v_mount;
433 	if (mp != NULL && (mp->mnt_flag & MNT_LOCAL) != 0 &&
434 	    ap->a_op == F_SETLK && (ap->a_flags & F_FIRSTOPEN) == 0) {
435 		VI_LOCK(vp);
436 		while ((vp->v_iflag & VI_FOPENING) != 0)
437 			msleep(vp, VI_MTX(vp), PLOCK, "lockfo", 0);
438 		VI_UNLOCK(vp);
439 	}
440 
441 	if (ap->a_fl->l_whence == SEEK_END) {
442 		/*
443 		 * The NFSv4 server must avoid doing a vn_lock() here, since it
444 		 * can deadlock the nfsd threads, due to a LOR.  Fortunately
445 		 * the NFSv4 server always uses SEEK_SET and this code is
446 		 * only required for the SEEK_END case.
447 		 */
448 		vn_lock(vp, LK_SHARED | LK_RETRY);
449 		error = VOP_GETATTR(vp, &vattr, curthread->td_ucred);
450 		VOP_UNLOCK(vp);
451 		if (error)
452 			return (error);
453 	} else
454 		vattr.va_size = 0;
455 
456 	return (lf_advlock(ap, &(vp->v_lockf), vattr.va_size));
457 }
458 
459 int
460 vop_stdadvlockasync(struct vop_advlockasync_args *ap)
461 {
462 	struct vnode *vp;
463 	struct vattr vattr;
464 	int error;
465 
466 	vp = ap->a_vp;
467 	if (ap->a_fl->l_whence == SEEK_END) {
468 		/* The size argument is only needed for SEEK_END. */
469 		vn_lock(vp, LK_SHARED | LK_RETRY);
470 		error = VOP_GETATTR(vp, &vattr, curthread->td_ucred);
471 		VOP_UNLOCK(vp);
472 		if (error)
473 			return (error);
474 	} else
475 		vattr.va_size = 0;
476 
477 	return (lf_advlockasync(ap, &(vp->v_lockf), vattr.va_size));
478 }
479 
480 int
481 vop_stdadvlockpurge(struct vop_advlockpurge_args *ap)
482 {
483 	struct vnode *vp;
484 
485 	vp = ap->a_vp;
486 	lf_purgelocks(vp, &vp->v_lockf);
487 	return (0);
488 }
489 
490 /*
491  * vop_stdpathconf:
492  *
493  * Standard implementation of POSIX pathconf, to get information about limits
494  * for a filesystem.
495  * Override per filesystem for the case where the filesystem has smaller
496  * limits.
497  */
498 int
499 vop_stdpathconf(struct vop_pathconf_args *ap)
500 {
501 
502 	switch (ap->a_name) {
503 		case _PC_ASYNC_IO:
504 			*ap->a_retval = _POSIX_ASYNCHRONOUS_IO;
505 			return (0);
506 		case _PC_PATH_MAX:
507 			*ap->a_retval = PATH_MAX;
508 			return (0);
509 		case _PC_ACL_EXTENDED:
510 		case _PC_ACL_NFS4:
511 		case _PC_CAP_PRESENT:
512 		case _PC_DEALLOC_PRESENT:
513 		case _PC_INF_PRESENT:
514 		case _PC_MAC_PRESENT:
515 			*ap->a_retval = 0;
516 			return (0);
517 		default:
518 			return (EINVAL);
519 	}
520 	/* NOTREACHED */
521 }
522 
523 /*
524  * Standard lock, unlock and islocked functions.
525  */
526 int
527 vop_stdlock(struct vop_lock1_args *ap)
528 {
529 	struct vnode *vp = ap->a_vp;
530 	struct mtx *ilk;
531 
532 	ilk = VI_MTX(vp);
533 	return (lockmgr_lock_flags(vp->v_vnlock, ap->a_flags,
534 	    &ilk->lock_object, ap->a_file, ap->a_line));
535 }
536 
537 /* See above. */
538 int
539 vop_stdunlock(struct vop_unlock_args *ap)
540 {
541 	struct vnode *vp = ap->a_vp;
542 
543 	return (lockmgr_unlock(vp->v_vnlock));
544 }
545 
546 /* See above. */
547 int
548 vop_stdislocked(struct vop_islocked_args *ap)
549 {
550 
551 	return (lockstatus(ap->a_vp->v_vnlock));
552 }
553 
554 /*
555  * Variants of the above set.
556  *
557  * Differences are:
558  * - shared locking disablement is not supported
559  * - v_vnlock pointer is not honored
560  */
561 int
562 vop_lock(struct vop_lock1_args *ap)
563 {
564 	struct vnode *vp = ap->a_vp;
565 	int flags = ap->a_flags;
566 	struct mtx *ilk;
567 
568 	MPASS(vp->v_vnlock == &vp->v_lock);
569 
570 	if (__predict_false((flags & ~(LK_TYPE_MASK | LK_NODDLKTREAT | LK_RETRY)) != 0))
571 		goto other;
572 
573 	switch (flags & LK_TYPE_MASK) {
574 	case LK_SHARED:
575 		return (lockmgr_slock(&vp->v_lock, flags, ap->a_file, ap->a_line));
576 	case LK_EXCLUSIVE:
577 		return (lockmgr_xlock(&vp->v_lock, flags, ap->a_file, ap->a_line));
578 	}
579 other:
580 	ilk = VI_MTX(vp);
581 	return (lockmgr_lock_flags(&vp->v_lock, flags,
582 	    &ilk->lock_object, ap->a_file, ap->a_line));
583 }
584 
585 int
586 vop_unlock(struct vop_unlock_args *ap)
587 {
588 	struct vnode *vp = ap->a_vp;
589 
590 	MPASS(vp->v_vnlock == &vp->v_lock);
591 
592 	return (lockmgr_unlock(&vp->v_lock));
593 }
594 
595 int
596 vop_islocked(struct vop_islocked_args *ap)
597 {
598 	struct vnode *vp = ap->a_vp;
599 
600 	MPASS(vp->v_vnlock == &vp->v_lock);
601 
602 	return (lockstatus(&vp->v_lock));
603 }
604 
605 /*
606  * Return true for select/poll.
607  */
608 int
609 vop_nopoll(struct vop_poll_args *ap)
610 {
611 
612 	if (ap->a_events & ~POLLSTANDARD)
613 		return (POLLNVAL);
614 	return (ap->a_events & (POLLIN | POLLOUT | POLLRDNORM | POLLWRNORM));
615 }
616 
617 /*
618  * Implement poll for local filesystems that support it.
619  */
620 int
621 vop_stdpoll(struct vop_poll_args *ap)
622 {
623 	if (ap->a_events & ~POLLSTANDARD)
624 		return (vn_pollrecord(ap->a_vp, ap->a_td, ap->a_events));
625 	return (ap->a_events & (POLLIN | POLLOUT | POLLRDNORM | POLLWRNORM));
626 }
627 
628 /*
629  * Return our mount point, as we will take charge of the writes.
630  */
631 int
632 vop_stdgetwritemount(struct vop_getwritemount_args *ap)
633 {
634 	struct mount *mp;
635 	struct vnode *vp;
636 
637 	/*
638 	 * Note that having a reference does not prevent forced unmount from
639 	 * setting ->v_mount to NULL after the lock gets released. This is of
640 	 * no consequence for typical consumers (most notably vn_start_write)
641 	 * since in this case the vnode is VIRF_DOOMED. Unmount might have
642 	 * progressed far enough that its completion is only delayed by the
643 	 * reference obtained here. The consumer only needs to concern itself
644 	 * with releasing it.
645 	 */
646 	vp = ap->a_vp;
647 	mp = vfs_ref_from_vp(vp);
648 	*(ap->a_mpp) = mp;
649 	return (0);
650 }
651 
652 /*
653  * If the file system doesn't implement VOP_BMAP, then return sensible defaults:
654  * - Return the vnode's bufobj instead of any underlying device's bufobj
655  * - Calculate the physical block number as if there were equal size
656  *   consecutive blocks, but
657  * - Report no contiguous runs of blocks.
658  */
659 int
660 vop_stdbmap(struct vop_bmap_args *ap)
661 {
662 
663 	if (ap->a_bop != NULL)
664 		*ap->a_bop = &ap->a_vp->v_bufobj;
665 	if (ap->a_bnp != NULL)
666 		*ap->a_bnp = ap->a_bn * btodb(ap->a_vp->v_mount->mnt_stat.f_iosize);
667 	if (ap->a_runp != NULL)
668 		*ap->a_runp = 0;
669 	if (ap->a_runb != NULL)
670 		*ap->a_runb = 0;
671 	return (0);
672 }
673 
674 int
675 vop_stdfsync(struct vop_fsync_args *ap)
676 {
677 
678 	return (vn_fsync_buf(ap->a_vp, ap->a_waitfor));
679 }
680 
681 static int
682 vop_stdfdatasync(struct vop_fdatasync_args *ap)
683 {
684 
685 	return (VOP_FSYNC(ap->a_vp, MNT_WAIT, ap->a_td));
686 }
687 
688 int
689 vop_stdfdatasync_buf(struct vop_fdatasync_args *ap)
690 {
691 
692 	return (vn_fsync_buf(ap->a_vp, MNT_WAIT));
693 }
694 
695 /* XXX Needs good comment and more info in the manpage (VOP_GETPAGES(9)). */
696 int
697 vop_stdgetpages(struct vop_getpages_args *ap)
698 {
699 
700 	return vnode_pager_generic_getpages(ap->a_vp, ap->a_m,
701 	    ap->a_count, ap->a_rbehind, ap->a_rahead, NULL, NULL);
702 }
703 
704 static int
705 vop_stdgetpages_async(struct vop_getpages_async_args *ap)
706 {
707 	int error;
708 
709 	error = VOP_GETPAGES(ap->a_vp, ap->a_m, ap->a_count, ap->a_rbehind,
710 	    ap->a_rahead);
711 	if (ap->a_iodone != NULL)
712 		ap->a_iodone(ap->a_arg, ap->a_m, ap->a_count, error);
713 	return (error);
714 }
715 
716 int
717 vop_stdkqfilter(struct vop_kqfilter_args *ap)
718 {
719 	return vfs_kqfilter(ap);
720 }
721 
722 /* XXX Needs good comment and more info in the manpage (VOP_PUTPAGES(9)). */
723 int
724 vop_stdputpages(struct vop_putpages_args *ap)
725 {
726 
727 	return vnode_pager_generic_putpages(ap->a_vp, ap->a_m, ap->a_count,
728 	     ap->a_sync, ap->a_rtvals);
729 }
730 
731 int
732 vop_stdvptofh(struct vop_vptofh_args *ap)
733 {
734 	return (EOPNOTSUPP);
735 }
736 
737 int
738 vop_stdvptocnp(struct vop_vptocnp_args *ap)
739 {
740 	struct vnode *vp = ap->a_vp;
741 	struct vnode **dvp = ap->a_vpp;
742 	struct ucred *cred;
743 	char *buf = ap->a_buf;
744 	size_t *buflen = ap->a_buflen;
745 	char *dirbuf, *cpos;
746 	int i, error, eofflag, dirbuflen, flags, locked, len, covered;
747 	off_t off;
748 	ino_t fileno;
749 	struct vattr va;
750 	struct nameidata nd;
751 	struct thread *td;
752 	struct dirent *dp;
753 	struct vnode *mvp;
754 
755 	i = *buflen;
756 	error = 0;
757 	covered = 0;
758 	td = curthread;
759 	cred = td->td_ucred;
760 
761 	if (vp->v_type != VDIR)
762 		return (ENOENT);
763 
764 	error = VOP_GETATTR(vp, &va, cred);
765 	if (error)
766 		return (error);
767 
768 	VREF(vp);
769 	locked = VOP_ISLOCKED(vp);
770 	VOP_UNLOCK(vp);
771 	NDINIT_ATVP(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF, UIO_SYSSPACE,
772 	    "..", vp);
773 	flags = FREAD;
774 	error = vn_open_cred(&nd, &flags, 0, VN_OPEN_NOAUDIT, cred, NULL);
775 	if (error) {
776 		vn_lock(vp, locked | LK_RETRY);
777 		return (error);
778 	}
779 	NDFREE_PNBUF(&nd);
780 
781 	mvp = *dvp = nd.ni_vp;
782 
783 	if (vp->v_mount != (*dvp)->v_mount &&
784 	    ((*dvp)->v_vflag & VV_ROOT) &&
785 	    ((*dvp)->v_mount->mnt_flag & MNT_UNION)) {
786 		*dvp = (*dvp)->v_mount->mnt_vnodecovered;
787 		VREF(mvp);
788 		VOP_UNLOCK(mvp);
789 		vn_close(mvp, FREAD, cred, td);
790 		VREF(*dvp);
791 		vn_lock(*dvp, LK_SHARED | LK_RETRY);
792 		covered = 1;
793 	}
794 
795 	fileno = va.va_fileid;
796 
797 	dirbuflen = DEV_BSIZE;
798 	if (dirbuflen < va.va_blocksize)
799 		dirbuflen = va.va_blocksize;
800 	dirbuf = (char *)malloc(dirbuflen, M_TEMP, M_WAITOK);
801 
802 	if ((*dvp)->v_type != VDIR) {
803 		error = ENOENT;
804 		goto out;
805 	}
806 
807 	off = 0;
808 	len = 0;
809 	do {
810 		/* call VOP_READDIR of parent */
811 		error = get_next_dirent(*dvp, &dp, dirbuf, dirbuflen, &off,
812 					&cpos, &len, &eofflag, td);
813 		if (error)
814 			goto out;
815 
816 		if ((dp->d_type != DT_WHT) &&
817 		    (dp->d_fileno == fileno)) {
818 			if (covered) {
819 				VOP_UNLOCK(*dvp);
820 				vn_lock(mvp, LK_SHARED | LK_RETRY);
821 				if (dirent_exists(mvp, dp->d_name, td)) {
822 					error = ENOENT;
823 					VOP_UNLOCK(mvp);
824 					vn_lock(*dvp, LK_SHARED | LK_RETRY);
825 					goto out;
826 				}
827 				VOP_UNLOCK(mvp);
828 				vn_lock(*dvp, LK_SHARED | LK_RETRY);
829 			}
830 			i -= dp->d_namlen;
831 
832 			if (i < 0) {
833 				error = ENOMEM;
834 				goto out;
835 			}
836 			if (dp->d_namlen == 1 && dp->d_name[0] == '.') {
837 				error = ENOENT;
838 			} else {
839 				bcopy(dp->d_name, buf + i, dp->d_namlen);
840 				error = 0;
841 			}
842 			goto out;
843 		}
844 	} while (len > 0 || !eofflag);
845 	error = ENOENT;
846 
847 out:
848 	free(dirbuf, M_TEMP);
849 	if (!error) {
850 		*buflen = i;
851 		vref(*dvp);
852 	}
853 	if (covered) {
854 		vput(*dvp);
855 		vrele(mvp);
856 	} else {
857 		VOP_UNLOCK(mvp);
858 		vn_close(mvp, FREAD, cred, td);
859 	}
860 	vn_lock(vp, locked | LK_RETRY);
861 	return (error);
862 }
863 
864 int
865 vop_stdallocate(struct vop_allocate_args *ap)
866 {
867 #ifdef __notyet__
868 	struct statfs *sfs;
869 	off_t maxfilesize = 0;
870 #endif
871 	struct iovec aiov;
872 	struct vattr vattr, *vap;
873 	struct uio auio;
874 	off_t fsize, len, cur, offset;
875 	uint8_t *buf;
876 	struct thread *td;
877 	struct vnode *vp;
878 	size_t iosize;
879 	int error;
880 
881 	buf = NULL;
882 	error = 0;
883 	td = curthread;
884 	vap = &vattr;
885 	vp = ap->a_vp;
886 	len = *ap->a_len;
887 	offset = *ap->a_offset;
888 
889 	error = VOP_GETATTR(vp, vap, ap->a_cred);
890 	if (error != 0)
891 		goto out;
892 	fsize = vap->va_size;
893 	iosize = vap->va_blocksize;
894 	if (iosize == 0)
895 		iosize = BLKDEV_IOSIZE;
896 	if (iosize > maxphys)
897 		iosize = maxphys;
898 	buf = malloc(iosize, M_TEMP, M_WAITOK);
899 
900 #ifdef __notyet__
901 	/*
902 	 * Check if the filesystem sets f_maxfilesize; if not use
903 	 * VOP_SETATTR to perform the check.
904 	 */
905 	sfs = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
906 	error = VFS_STATFS(vp->v_mount, sfs, td);
907 	if (error == 0)
908 		maxfilesize = sfs->f_maxfilesize;
909 	free(sfs, M_STATFS);
910 	if (error != 0)
911 		goto out;
912 	if (maxfilesize) {
913 		if (offset > maxfilesize || len > maxfilesize ||
914 		    offset + len > maxfilesize) {
915 			error = EFBIG;
916 			goto out;
917 		}
918 	} else
919 #endif
920 	if (offset + len > vap->va_size) {
921 		/*
922 		 * Test offset + len against the filesystem's maxfilesize.
923 		 */
924 		VATTR_NULL(vap);
925 		vap->va_size = offset + len;
926 		error = VOP_SETATTR(vp, vap, ap->a_cred);
927 		if (error != 0)
928 			goto out;
929 		VATTR_NULL(vap);
930 		vap->va_size = fsize;
931 		error = VOP_SETATTR(vp, vap, ap->a_cred);
932 		if (error != 0)
933 			goto out;
934 	}
935 
936 	for (;;) {
937 		/*
938 		 * Read and write back anything below the nominal file
939 		 * size.  There's currently no way outside the filesystem
940 		 * to know whether this area is sparse or not.
941 		 */
942 		cur = iosize;
943 		if ((offset % iosize) != 0)
944 			cur -= (offset % iosize);
945 		if (cur > len)
946 			cur = len;
947 		if (offset < fsize) {
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_READ;
956 			auio.uio_td = td;
957 			error = VOP_READ(vp, &auio, ap->a_ioflag, ap->a_cred);
958 			if (error != 0)
959 				break;
960 			if (auio.uio_resid > 0) {
961 				bzero(buf + cur - auio.uio_resid,
962 				    auio.uio_resid);
963 			}
964 		} else {
965 			bzero(buf, cur);
966 		}
967 
968 		aiov.iov_base = buf;
969 		aiov.iov_len = cur;
970 		auio.uio_iov = &aiov;
971 		auio.uio_iovcnt = 1;
972 		auio.uio_offset = offset;
973 		auio.uio_resid = cur;
974 		auio.uio_segflg = UIO_SYSSPACE;
975 		auio.uio_rw = UIO_WRITE;
976 		auio.uio_td = td;
977 
978 		error = VOP_WRITE(vp, &auio, ap->a_ioflag, ap->a_cred);
979 		if (error != 0)
980 			break;
981 
982 		len -= cur;
983 		offset += cur;
984 		if (len == 0)
985 			break;
986 		if (should_yield())
987 			break;
988 	}
989 
990  out:
991 	*ap->a_len = len;
992 	*ap->a_offset = offset;
993 	free(buf, M_TEMP);
994 	return (error);
995 }
996 
997 static int
998 vp_zerofill(struct vnode *vp, struct vattr *vap, off_t *offsetp, off_t *lenp,
999     int ioflag, struct ucred *cred)
1000 {
1001 	int iosize;
1002 	int error = 0;
1003 	struct iovec aiov;
1004 	struct uio auio;
1005 	struct thread *td;
1006 	off_t offset, len;
1007 
1008 	iosize = vap->va_blocksize;
1009 	td = curthread;
1010 	offset = *offsetp;
1011 	len = *lenp;
1012 
1013 	if (iosize == 0)
1014 		iosize = BLKDEV_IOSIZE;
1015 	/* If va_blocksize is 512 bytes, iosize will be 4 kilobytes */
1016 	iosize = min(iosize * 8, ZERO_REGION_SIZE);
1017 
1018 	while (len > 0) {
1019 		int xfersize = iosize;
1020 		if (offset % iosize != 0)
1021 			xfersize -= offset % iosize;
1022 		if (xfersize > len)
1023 			xfersize = len;
1024 
1025 		aiov.iov_base = __DECONST(void *, zero_region);
1026 		aiov.iov_len = xfersize;
1027 		auio.uio_iov = &aiov;
1028 		auio.uio_iovcnt = 1;
1029 		auio.uio_offset = offset;
1030 		auio.uio_resid = xfersize;
1031 		auio.uio_segflg = UIO_SYSSPACE;
1032 		auio.uio_rw = UIO_WRITE;
1033 		auio.uio_td = td;
1034 
1035 		error = VOP_WRITE(vp, &auio, ioflag, cred);
1036 		if (error != 0) {
1037 			len -= xfersize - auio.uio_resid;
1038 			offset += xfersize - auio.uio_resid;
1039 			break;
1040 		}
1041 
1042 		len -= xfersize;
1043 		offset += xfersize;
1044 	}
1045 
1046 	*offsetp = offset;
1047 	*lenp = len;
1048 	return (error);
1049 }
1050 
1051 int
1052 vop_stddeallocate(struct vop_deallocate_args *ap)
1053 {
1054 	struct vnode *vp;
1055 	off_t offset, len;
1056 	struct ucred *cred;
1057 	int error;
1058 	struct vattr va;
1059 	off_t noff, xfersize, rem;
1060 
1061 	vp = ap->a_vp;
1062 	offset = *ap->a_offset;
1063 	cred = ap->a_cred;
1064 
1065 	error = VOP_GETATTR(vp, &va, cred);
1066 	if (error)
1067 		return (error);
1068 
1069 	len = omin((off_t)va.va_size - offset, *ap->a_len);
1070 	while (len > 0) {
1071 		noff = offset;
1072 		error = vn_bmap_seekhole_locked(vp, FIOSEEKDATA, &noff, cred);
1073 		if (error) {
1074 			if (error != ENXIO)
1075 				/* XXX: Is it okay to fallback further? */
1076 				goto out;
1077 
1078 			/*
1079 			 * No more data region to be filled
1080 			 */
1081 			offset += len;
1082 			len = 0;
1083 			error = 0;
1084 			break;
1085 		}
1086 		KASSERT(noff >= offset, ("FIOSEEKDATA going backward"));
1087 		if (noff != offset) {
1088 			xfersize = omin(noff - offset, len);
1089 			len -= xfersize;
1090 			offset += xfersize;
1091 			if (len == 0)
1092 				break;
1093 		}
1094 		error = vn_bmap_seekhole_locked(vp, FIOSEEKHOLE, &noff, cred);
1095 		if (error)
1096 			goto out;
1097 
1098 		/* Fill zeroes */
1099 		xfersize = rem = omin(noff - offset, len);
1100 		error = vp_zerofill(vp, &va, &offset, &rem, ap->a_ioflag, cred);
1101 		if (error) {
1102 			len -= xfersize - rem;
1103 			goto out;
1104 		}
1105 
1106 		len -= xfersize;
1107 		if (should_yield())
1108 			break;
1109 	}
1110 	/* Handle the case when offset is beyond EOF */
1111 	if (len < 0)
1112 		len = 0;
1113 out:
1114 	*ap->a_offset = offset;
1115 	*ap->a_len = len;
1116 	return (error);
1117 }
1118 
1119 int
1120 vop_stdadvise(struct vop_advise_args *ap)
1121 {
1122 	struct vnode *vp;
1123 	struct bufobj *bo;
1124 	daddr_t startn, endn;
1125 	off_t bstart, bend, start, end;
1126 	int bsize, error;
1127 
1128 	vp = ap->a_vp;
1129 	switch (ap->a_advice) {
1130 	case POSIX_FADV_WILLNEED:
1131 		/*
1132 		 * Do nothing for now.  Filesystems should provide a
1133 		 * custom method which starts an asynchronous read of
1134 		 * the requested region.
1135 		 */
1136 		error = 0;
1137 		break;
1138 	case POSIX_FADV_DONTNEED:
1139 		error = 0;
1140 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1141 		if (VN_IS_DOOMED(vp)) {
1142 			VOP_UNLOCK(vp);
1143 			break;
1144 		}
1145 
1146 		/*
1147 		 * Round to block boundaries (and later possibly further to
1148 		 * page boundaries).  Applications cannot reasonably be aware
1149 		 * of the boundaries, and the rounding must be to expand at
1150 		 * both extremities to cover enough.  It still doesn't cover
1151 		 * read-ahead.  For partial blocks, this gives unnecessary
1152 		 * discarding of buffers but is efficient enough since the
1153 		 * pages usually remain in VMIO for some time.
1154 		 */
1155 		bsize = vp->v_bufobj.bo_bsize;
1156 		bstart = rounddown(ap->a_start, bsize);
1157 		bend = roundup(ap->a_end, bsize);
1158 
1159 		/*
1160 		 * Deactivate pages in the specified range from the backing VM
1161 		 * object.  Pages that are resident in the buffer cache will
1162 		 * remain wired until their corresponding buffers are released
1163 		 * below.
1164 		 */
1165 		if (vp->v_object != NULL) {
1166 			start = trunc_page(bstart);
1167 			end = round_page(bend);
1168 			VM_OBJECT_RLOCK(vp->v_object);
1169 			vm_object_page_noreuse(vp->v_object, OFF_TO_IDX(start),
1170 			    OFF_TO_IDX(end));
1171 			VM_OBJECT_RUNLOCK(vp->v_object);
1172 		}
1173 
1174 		bo = &vp->v_bufobj;
1175 		BO_RLOCK(bo);
1176 		startn = bstart / bsize;
1177 		endn = bend / bsize;
1178 		error = bnoreuselist(&bo->bo_clean, bo, startn, endn);
1179 		if (error == 0)
1180 			error = bnoreuselist(&bo->bo_dirty, bo, startn, endn);
1181 		BO_RUNLOCK(bo);
1182 		VOP_UNLOCK(vp);
1183 		break;
1184 	default:
1185 		error = EINVAL;
1186 		break;
1187 	}
1188 	return (error);
1189 }
1190 
1191 int
1192 vop_stdunp_bind(struct vop_unp_bind_args *ap)
1193 {
1194 
1195 	ap->a_vp->v_unpcb = ap->a_unpcb;
1196 	return (0);
1197 }
1198 
1199 int
1200 vop_stdunp_connect(struct vop_unp_connect_args *ap)
1201 {
1202 
1203 	*ap->a_unpcb = ap->a_vp->v_unpcb;
1204 	return (0);
1205 }
1206 
1207 int
1208 vop_stdunp_detach(struct vop_unp_detach_args *ap)
1209 {
1210 
1211 	ap->a_vp->v_unpcb = NULL;
1212 	return (0);
1213 }
1214 
1215 static int
1216 vop_stdis_text(struct vop_is_text_args *ap)
1217 {
1218 
1219 	return (atomic_load_int(&ap->a_vp->v_writecount) < 0);
1220 }
1221 
1222 int
1223 vop_stdset_text(struct vop_set_text_args *ap)
1224 {
1225 	struct vnode *vp;
1226 	int n;
1227 	bool gotref;
1228 
1229 	vp = ap->a_vp;
1230 
1231 	n = atomic_load_int(&vp->v_writecount);
1232 	for (;;) {
1233 		if (__predict_false(n > 0)) {
1234 			return (ETXTBSY);
1235 		}
1236 
1237 		/*
1238 		 * Transition point, we may need to grab a reference on the vnode.
1239 		 *
1240 		 * Take the ref early As a safety measure against bogus calls
1241 		 * to vop_stdunset_text.
1242 		 */
1243 		if (n == 0) {
1244 			gotref = false;
1245 			if ((vn_irflag_read(vp) & VIRF_TEXT_REF) != 0) {
1246 				vref(vp);
1247 				gotref = true;
1248 			}
1249 			if (atomic_fcmpset_int(&vp->v_writecount, &n, -1)) {
1250 				return (0);
1251 			}
1252 			if (gotref) {
1253 				vunref(vp);
1254 			}
1255 			continue;
1256 		}
1257 
1258 		MPASS(n < 0);
1259 		if (atomic_fcmpset_int(&vp->v_writecount, &n, n - 1)) {
1260 			return (0);
1261 		}
1262 	}
1263 	__assert_unreachable();
1264 }
1265 
1266 static int
1267 vop_stdunset_text(struct vop_unset_text_args *ap)
1268 {
1269 	struct vnode *vp;
1270 	int n;
1271 
1272 	vp = ap->a_vp;
1273 
1274 	n = atomic_load_int(&vp->v_writecount);
1275 	for (;;) {
1276 		if (__predict_false(n >= 0)) {
1277 			return (EINVAL);
1278 		}
1279 
1280 		/*
1281 		 * Transition point, we may need to release a reference on the vnode.
1282 		 */
1283 		if (n == -1) {
1284 			if (atomic_fcmpset_int(&vp->v_writecount, &n, 0)) {
1285 				if ((vn_irflag_read(vp) & VIRF_TEXT_REF) != 0) {
1286 					vunref(vp);
1287 				}
1288 				return (0);
1289 			}
1290 			continue;
1291 		}
1292 
1293 		MPASS(n < -1);
1294 		if (atomic_fcmpset_int(&vp->v_writecount, &n, n + 1)) {
1295 			return (0);
1296 		}
1297 	}
1298 	__assert_unreachable();
1299 }
1300 
1301 static int __always_inline
1302 vop_stdadd_writecount_impl(struct vop_add_writecount_args *ap, bool handle_msync)
1303 {
1304 	struct vnode *vp;
1305 	struct mount *mp __diagused;
1306 	int n;
1307 
1308 	vp = ap->a_vp;
1309 
1310 #ifdef INVARIANTS
1311 	mp = vp->v_mount;
1312 	if (mp != NULL) {
1313 		if (handle_msync) {
1314 			VNPASS((mp->mnt_kern_flag & MNTK_NOMSYNC) == 0, vp);
1315 		} else {
1316 			VNPASS((mp->mnt_kern_flag & MNTK_NOMSYNC) != 0, vp);
1317 		}
1318 	}
1319 #endif
1320 
1321 	n = atomic_load_int(&vp->v_writecount);
1322 	for (;;) {
1323 		if (__predict_false(n < 0)) {
1324 			return (ETXTBSY);
1325 		}
1326 
1327 		VNASSERT(n + ap->a_inc >= 0, vp,
1328 		    ("neg writecount increment %d + %d = %d", n, ap->a_inc,
1329 		    n + ap->a_inc));
1330 		if (n == 0) {
1331 			if (handle_msync) {
1332 				vlazy(vp);
1333 			}
1334 		}
1335 
1336 		if (atomic_fcmpset_int(&vp->v_writecount, &n, n + ap->a_inc)) {
1337 			return (0);
1338 		}
1339 	}
1340 	__assert_unreachable();
1341 }
1342 
1343 int
1344 vop_stdadd_writecount(struct vop_add_writecount_args *ap)
1345 {
1346 
1347 	return (vop_stdadd_writecount_impl(ap, true));
1348 }
1349 
1350 int
1351 vop_stdadd_writecount_nomsync(struct vop_add_writecount_args *ap)
1352 {
1353 
1354 	return (vop_stdadd_writecount_impl(ap, false));
1355 }
1356 
1357 int
1358 vop_stdneed_inactive(struct vop_need_inactive_args *ap)
1359 {
1360 
1361 	return (1);
1362 }
1363 
1364 int
1365 vop_stdioctl(struct vop_ioctl_args *ap)
1366 {
1367 	struct vnode *vp;
1368 	struct vattr va;
1369 	off_t *offp;
1370 	int error;
1371 
1372 	switch (ap->a_command) {
1373 	case FIOSEEKDATA:
1374 	case FIOSEEKHOLE:
1375 		vp = ap->a_vp;
1376 		error = vn_lock(vp, LK_SHARED);
1377 		if (error != 0)
1378 			return (EBADF);
1379 		if (vp->v_type == VREG)
1380 			error = VOP_GETATTR(vp, &va, ap->a_cred);
1381 		else
1382 			error = ENOTTY;
1383 		if (error == 0) {
1384 			offp = ap->a_data;
1385 			if (*offp < 0 || *offp >= va.va_size)
1386 				error = ENXIO;
1387 			else if (ap->a_command == FIOSEEKHOLE)
1388 				*offp = va.va_size;
1389 		}
1390 		VOP_UNLOCK(vp);
1391 		break;
1392 	default:
1393 		error = ENOTTY;
1394 		break;
1395 	}
1396 	return (error);
1397 }
1398 
1399 /*
1400  * vfs default ops
1401  * used to fill the vfs function table to get reasonable default return values.
1402  */
1403 int
1404 vfs_stdroot(struct mount *mp, int flags, struct vnode **vpp)
1405 {
1406 
1407 	return (EOPNOTSUPP);
1408 }
1409 
1410 int
1411 vfs_stdstatfs(struct mount *mp, struct statfs *sbp)
1412 {
1413 
1414 	return (EOPNOTSUPP);
1415 }
1416 
1417 int
1418 vfs_stdquotactl(struct mount *mp, int cmds, uid_t uid, void *arg, bool *mp_busy)
1419 {
1420 	return (EOPNOTSUPP);
1421 }
1422 
1423 int
1424 vfs_stdsync(struct mount *mp, int waitfor)
1425 {
1426 	struct vnode *vp, *mvp;
1427 	struct thread *td;
1428 	int error, lockreq, allerror = 0;
1429 
1430 	td = curthread;
1431 	lockreq = LK_EXCLUSIVE | LK_INTERLOCK;
1432 	if (waitfor != MNT_WAIT)
1433 		lockreq |= LK_NOWAIT;
1434 	/*
1435 	 * Force stale buffer cache information to be flushed.
1436 	 */
1437 loop:
1438 	MNT_VNODE_FOREACH_ALL(vp, mp, mvp) {
1439 		if (vp->v_bufobj.bo_dirty.bv_cnt == 0) {
1440 			VI_UNLOCK(vp);
1441 			continue;
1442 		}
1443 		if ((error = vget(vp, lockreq)) != 0) {
1444 			if (error == ENOENT) {
1445 				MNT_VNODE_FOREACH_ALL_ABORT(mp, mvp);
1446 				goto loop;
1447 			}
1448 			continue;
1449 		}
1450 		error = VOP_FSYNC(vp, waitfor, td);
1451 		if (error)
1452 			allerror = error;
1453 		vput(vp);
1454 	}
1455 	return (allerror);
1456 }
1457 
1458 int
1459 vfs_stdnosync(struct mount *mp, int waitfor)
1460 {
1461 
1462 	return (0);
1463 }
1464 
1465 static int
1466 vop_stdcopy_file_range(struct vop_copy_file_range_args *ap)
1467 {
1468 	int error;
1469 
1470 	error = vn_generic_copy_file_range(ap->a_invp, ap->a_inoffp,
1471 	    ap->a_outvp, ap->a_outoffp, ap->a_lenp, ap->a_flags, ap->a_incred,
1472 	    ap->a_outcred, ap->a_fsizetd);
1473 	return (error);
1474 }
1475 
1476 int
1477 vfs_stdvget(struct mount *mp, ino_t ino, int flags, struct vnode **vpp)
1478 {
1479 
1480 	return (EOPNOTSUPP);
1481 }
1482 
1483 int
1484 vfs_stdfhtovp(struct mount *mp, struct fid *fhp, int flags, struct vnode **vpp)
1485 {
1486 
1487 	return (EOPNOTSUPP);
1488 }
1489 
1490 int
1491 vfs_stdinit(struct vfsconf *vfsp)
1492 {
1493 
1494 	return (0);
1495 }
1496 
1497 int
1498 vfs_stduninit(struct vfsconf *vfsp)
1499 {
1500 
1501 	return(0);
1502 }
1503 
1504 int
1505 vfs_stdextattrctl(struct mount *mp, int cmd, struct vnode *filename_vp,
1506     int attrnamespace, const char *attrname)
1507 {
1508 
1509 	if (filename_vp != NULL)
1510 		VOP_UNLOCK(filename_vp);
1511 	return (EOPNOTSUPP);
1512 }
1513 
1514 int
1515 vfs_stdsysctl(struct mount *mp, fsctlop_t op, struct sysctl_req *req)
1516 {
1517 
1518 	return (EOPNOTSUPP);
1519 }
1520 
1521 static vop_bypass_t *
1522 bp_by_off(struct vop_vector *vop, struct vop_generic_args *a)
1523 {
1524 
1525 	return (*(vop_bypass_t **)((char *)vop + a->a_desc->vdesc_vop_offset));
1526 }
1527 
1528 int
1529 vop_sigdefer(struct vop_vector *vop, struct vop_generic_args *a)
1530 {
1531 	vop_bypass_t *bp;
1532 	int prev_stops, rc;
1533 
1534 	bp = bp_by_off(vop, a);
1535 	MPASS(bp != NULL);
1536 
1537 	prev_stops = sigdeferstop(SIGDEFERSTOP_SILENT);
1538 	rc = bp(a);
1539 	sigallowstop(prev_stops);
1540 	return (rc);
1541 }
1542 
1543 static int
1544 vop_stdstat(struct vop_stat_args *a)
1545 {
1546 	struct vattr vattr;
1547 	struct vattr *vap;
1548 	struct vnode *vp;
1549 	struct stat *sb;
1550 	int error;
1551 	u_short mode;
1552 
1553 	vp = a->a_vp;
1554 	sb = a->a_sb;
1555 
1556 	error = vop_stat_helper_pre(a);
1557 	if (error != 0)
1558 		return (error);
1559 
1560 	vap = &vattr;
1561 
1562 	/*
1563 	 * Initialize defaults for new and unusual fields, so that file
1564 	 * systems which don't support these fields don't need to know
1565 	 * about them.
1566 	 */
1567 	vap->va_birthtime.tv_sec = -1;
1568 	vap->va_birthtime.tv_nsec = 0;
1569 	vap->va_fsid = VNOVAL;
1570 	vap->va_gen = 0;
1571 	vap->va_rdev = NODEV;
1572 
1573 	error = VOP_GETATTR(vp, vap, a->a_active_cred);
1574 	if (error)
1575 		goto out;
1576 
1577 	/*
1578 	 * Zero the spare stat fields
1579 	 */
1580 	bzero(sb, sizeof *sb);
1581 
1582 	/*
1583 	 * Copy from vattr table
1584 	 */
1585 	if (vap->va_fsid != VNOVAL)
1586 		sb->st_dev = vap->va_fsid;
1587 	else
1588 		sb->st_dev = vp->v_mount->mnt_stat.f_fsid.val[0];
1589 	sb->st_ino = vap->va_fileid;
1590 	mode = vap->va_mode;
1591 	switch (vap->va_type) {
1592 	case VREG:
1593 		mode |= S_IFREG;
1594 		break;
1595 	case VDIR:
1596 		mode |= S_IFDIR;
1597 		break;
1598 	case VBLK:
1599 		mode |= S_IFBLK;
1600 		break;
1601 	case VCHR:
1602 		mode |= S_IFCHR;
1603 		break;
1604 	case VLNK:
1605 		mode |= S_IFLNK;
1606 		break;
1607 	case VSOCK:
1608 		mode |= S_IFSOCK;
1609 		break;
1610 	case VFIFO:
1611 		mode |= S_IFIFO;
1612 		break;
1613 	default:
1614 		error = EBADF;
1615 		goto out;
1616 	}
1617 	sb->st_mode = mode;
1618 	sb->st_nlink = vap->va_nlink;
1619 	sb->st_uid = vap->va_uid;
1620 	sb->st_gid = vap->va_gid;
1621 	sb->st_rdev = vap->va_rdev;
1622 	if (vap->va_size > OFF_MAX) {
1623 		error = EOVERFLOW;
1624 		goto out;
1625 	}
1626 	sb->st_size = vap->va_size;
1627 	sb->st_atim.tv_sec = vap->va_atime.tv_sec;
1628 	sb->st_atim.tv_nsec = vap->va_atime.tv_nsec;
1629 	sb->st_mtim.tv_sec = vap->va_mtime.tv_sec;
1630 	sb->st_mtim.tv_nsec = vap->va_mtime.tv_nsec;
1631 	sb->st_ctim.tv_sec = vap->va_ctime.tv_sec;
1632 	sb->st_ctim.tv_nsec = vap->va_ctime.tv_nsec;
1633 	sb->st_birthtim.tv_sec = vap->va_birthtime.tv_sec;
1634 	sb->st_birthtim.tv_nsec = vap->va_birthtime.tv_nsec;
1635 
1636 	/*
1637 	 * According to www.opengroup.org, the meaning of st_blksize is
1638 	 *   "a filesystem-specific preferred I/O block size for this
1639 	 *    object.  In some filesystem types, this may vary from file
1640 	 *    to file"
1641 	 * Use minimum/default of PAGE_SIZE (e.g. for VCHR).
1642 	 */
1643 
1644 	sb->st_blksize = max(PAGE_SIZE, vap->va_blocksize);
1645 	sb->st_flags = vap->va_flags;
1646 	sb->st_blocks = vap->va_bytes / S_BLKSIZE;
1647 	sb->st_gen = vap->va_gen;
1648 out:
1649 	return (vop_stat_helper_post(a, error));
1650 }
1651 
1652 static int
1653 vop_stdread_pgcache(struct vop_read_pgcache_args *ap __unused)
1654 {
1655 	return (EJUSTRETURN);
1656 }
1657 
1658 static int
1659 vop_stdvput_pair(struct vop_vput_pair_args *ap)
1660 {
1661 	struct vnode *dvp, *vp, **vpp;
1662 
1663 	dvp = ap->a_dvp;
1664 	vpp = ap->a_vpp;
1665 	vput(dvp);
1666 	if (vpp != NULL && ap->a_unlock_vp && (vp = *vpp) != NULL)
1667 		vput(vp);
1668 	return (0);
1669 }
1670