xref: /freebsd-13-stable/sys/kern/vfs_default.c (revision 62ed7f98a2db5018a5336398df37f60cb157ec6d)
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 #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/filio.h>
45 #include <sys/kernel.h>
46 #include <sys/limits.h>
47 #include <sys/lock.h>
48 #include <sys/lockf.h>
49 #include <sys/malloc.h>
50 #include <sys/mount.h>
51 #include <sys/namei.h>
52 #include <sys/rwlock.h>
53 #include <sys/fcntl.h>
54 #include <sys/unistd.h>
55 #include <sys/vnode.h>
56 #include <sys/dirent.h>
57 #include <sys/poll.h>
58 #include <sys/stat.h>
59 #include <security/audit/audit.h>
60 #include <sys/priv.h>
61 
62 #include <security/mac/mac_framework.h>
63 
64 #include <vm/vm.h>
65 #include <vm/vm_object.h>
66 #include <vm/vm_extern.h>
67 #include <vm/pmap.h>
68 #include <vm/vm_map.h>
69 #include <vm/vm_page.h>
70 #include <vm/vm_pager.h>
71 #include <vm/vnode_pager.h>
72 
73 static int	vop_nolookup(struct vop_lookup_args *);
74 static int	vop_norename(struct vop_rename_args *);
75 static int	vop_nostrategy(struct vop_strategy_args *);
76 static int	dirent_exists(struct vnode *vp, const char *dirname,
77 			      struct thread *td);
78 
79 static int vop_stdis_text(struct vop_is_text_args *ap);
80 static int vop_stdunset_text(struct vop_unset_text_args *ap);
81 static int vop_stdadd_writecount(struct vop_add_writecount_args *ap);
82 static int vop_stdcopy_file_range(struct vop_copy_file_range_args *ap);
83 static int vop_stdfdatasync(struct vop_fdatasync_args *ap);
84 static int vop_stdgetpages_async(struct vop_getpages_async_args *ap);
85 static int vop_stdread_pgcache(struct vop_read_pgcache_args *ap);
86 static int vop_stdstat(struct vop_stat_args *ap);
87 static int vop_stdvput_pair(struct vop_vput_pair_args *ap);
88 static int vop_stdgetlowvnode(struct vop_getlowvnode_args *ap);
89 
90 /*
91  * This vnode table stores what we want to do if the filesystem doesn't
92  * implement a particular VOP.
93  *
94  * If there is no specific entry here, we will return EOPNOTSUPP.
95  *
96  * Note that every filesystem has to implement either vop_access
97  * or vop_accessx; failing to do so will result in immediate crash
98  * due to stack overflow, as vop_stdaccess() calls vop_stdaccessx(),
99  * which calls vop_stdaccess() etc.
100  */
101 
102 struct vop_vector default_vnodeops = {
103 	.vop_default =		NULL,
104 	.vop_bypass =		VOP_EOPNOTSUPP,
105 
106 	.vop_access =		vop_stdaccess,
107 	.vop_accessx =		vop_stdaccessx,
108 	.vop_advise =		vop_stdadvise,
109 	.vop_advlock =		vop_stdadvlock,
110 	.vop_advlockasync =	vop_stdadvlockasync,
111 	.vop_advlockpurge =	vop_stdadvlockpurge,
112 	.vop_allocate =		vop_stdallocate,
113 	.vop_bmap =		vop_stdbmap,
114 	.vop_close =		VOP_NULL,
115 	.vop_fsync =		VOP_NULL,
116 	.vop_stat =		vop_stdstat,
117 	.vop_fdatasync =	vop_stdfdatasync,
118 	.vop_getlowvnode = 	vop_stdgetlowvnode,
119 	.vop_getpages =		vop_stdgetpages,
120 	.vop_getpages_async =	vop_stdgetpages_async,
121 	.vop_getwritemount = 	vop_stdgetwritemount,
122 	.vop_inactive =		VOP_NULL,
123 	.vop_need_inactive =	vop_stdneed_inactive,
124 	.vop_ioctl =		vop_stdioctl,
125 	.vop_kqfilter =		vop_stdkqfilter,
126 	.vop_islocked =		vop_stdislocked,
127 	.vop_lock1 =		vop_stdlock,
128 	.vop_lookup =		vop_nolookup,
129 	.vop_open =		VOP_NULL,
130 	.vop_pathconf =		VOP_EINVAL,
131 	.vop_poll =		vop_nopoll,
132 	.vop_putpages =		vop_stdputpages,
133 	.vop_readlink =		VOP_EINVAL,
134 	.vop_read_pgcache =	vop_stdread_pgcache,
135 	.vop_rename =		vop_norename,
136 	.vop_revoke =		VOP_PANIC,
137 	.vop_strategy =		vop_nostrategy,
138 	.vop_unlock =		vop_stdunlock,
139 	.vop_vptocnp =		vop_stdvptocnp,
140 	.vop_vptofh =		vop_stdvptofh,
141 	.vop_unp_bind =		vop_stdunp_bind,
142 	.vop_unp_connect =	vop_stdunp_connect,
143 	.vop_unp_detach =	vop_stdunp_detach,
144 	.vop_is_text =		vop_stdis_text,
145 	.vop_set_text =		vop_stdset_text,
146 	.vop_unset_text =	vop_stdunset_text,
147 	.vop_add_writecount =	vop_stdadd_writecount,
148 	.vop_copy_file_range =	vop_stdcopy_file_range,
149 	.vop_vput_pair =	vop_stdvput_pair,
150 };
151 VFS_VOP_VECTOR_REGISTER(default_vnodeops);
152 
153 /*
154  * Series of placeholder functions for various error returns for
155  * VOPs.
156  */
157 
158 int
vop_eopnotsupp(struct vop_generic_args * ap)159 vop_eopnotsupp(struct vop_generic_args *ap)
160 {
161 	/*
162 	printf("vop_notsupp[%s]\n", ap->a_desc->vdesc_name);
163 	*/
164 
165 	return (EOPNOTSUPP);
166 }
167 
168 int
vop_ebadf(struct vop_generic_args * ap)169 vop_ebadf(struct vop_generic_args *ap)
170 {
171 
172 	return (EBADF);
173 }
174 
175 int
vop_enotty(struct vop_generic_args * ap)176 vop_enotty(struct vop_generic_args *ap)
177 {
178 
179 	return (ENOTTY);
180 }
181 
182 int
vop_einval(struct vop_generic_args * ap)183 vop_einval(struct vop_generic_args *ap)
184 {
185 
186 	return (EINVAL);
187 }
188 
189 int
vop_enoent(struct vop_generic_args * ap)190 vop_enoent(struct vop_generic_args *ap)
191 {
192 
193 	return (ENOENT);
194 }
195 
196 int
vop_eagain(struct vop_generic_args * ap)197 vop_eagain(struct vop_generic_args *ap)
198 {
199 
200 	return (EAGAIN);
201 }
202 
203 int
vop_null(struct vop_generic_args * ap)204 vop_null(struct vop_generic_args *ap)
205 {
206 
207 	return (0);
208 }
209 
210 /*
211  * Helper function to panic on some bad VOPs in some filesystems.
212  */
213 int
vop_panic(struct vop_generic_args * ap)214 vop_panic(struct vop_generic_args *ap)
215 {
216 
217 	panic("filesystem goof: vop_panic[%s]", ap->a_desc->vdesc_name);
218 }
219 
220 /*
221  * vop_std<something> and vop_no<something> are default functions for use by
222  * filesystems that need the "default reasonable" implementation for a
223  * particular operation.
224  *
225  * The documentation for the operations they implement exists (if it exists)
226  * in the VOP_<SOMETHING>(9) manpage (all uppercase).
227  */
228 
229 /*
230  * Default vop for filesystems that do not support name lookup
231  */
232 static int
vop_nolookup(struct vop_lookup_args * ap)233 vop_nolookup(struct vop_lookup_args *ap)
234 {
235 
236 	*ap->a_vpp = NULL;
237 	return (ENOTDIR);
238 }
239 
240 /*
241  * vop_norename:
242  *
243  * Handle unlock and reference counting for arguments of vop_rename
244  * for filesystems that do not implement rename operation.
245  */
246 static int
vop_norename(struct vop_rename_args * ap)247 vop_norename(struct vop_rename_args *ap)
248 {
249 
250 	vop_rename_fail(ap);
251 	return (EOPNOTSUPP);
252 }
253 
254 /*
255  *	vop_nostrategy:
256  *
257  *	Strategy routine for VFS devices that have none.
258  *
259  *	BIO_ERROR and B_INVAL must be cleared prior to calling any strategy
260  *	routine.  Typically this is done for a BIO_READ strategy call.
261  *	Typically B_INVAL is assumed to already be clear prior to a write
262  *	and should not be cleared manually unless you just made the buffer
263  *	invalid.  BIO_ERROR should be cleared either way.
264  */
265 
266 static int
vop_nostrategy(struct vop_strategy_args * ap)267 vop_nostrategy (struct vop_strategy_args *ap)
268 {
269 	printf("No strategy for buffer at %p\n", ap->a_bp);
270 	vn_printf(ap->a_vp, "vnode ");
271 	ap->a_bp->b_ioflags |= BIO_ERROR;
272 	ap->a_bp->b_error = EOPNOTSUPP;
273 	bufdone(ap->a_bp);
274 	return (EOPNOTSUPP);
275 }
276 
277 /*
278  * Check if a named file exists in a given directory vnode
279  *
280  * Returns 0 if the file exists, ENOENT if it doesn't, or errors returned by
281  * vn_dir_next_dirent().
282  */
283 static int
dirent_exists(struct vnode * vp,const char * dirname,struct thread * td)284 dirent_exists(struct vnode *vp, const char *dirname, struct thread *td)
285 {
286 	char *dirbuf;
287 	int error, eofflag;
288 	size_t dirbuflen, len;
289 	off_t off;
290 	struct dirent *dp;
291 	struct vattr va;
292 
293 	ASSERT_VOP_LOCKED(vp, "vnode not locked");
294 	KASSERT(vp->v_type == VDIR, ("vp %p is not a directory", vp));
295 
296 	error = VOP_GETATTR(vp, &va, td->td_ucred);
297 	if (error != 0)
298 		return (error);
299 
300 	dirbuflen = MAX(DEV_BSIZE, GENERIC_MAXDIRSIZ);
301 	if (dirbuflen < va.va_blocksize)
302 		dirbuflen = va.va_blocksize;
303 	dirbuf = malloc(dirbuflen, M_TEMP, M_WAITOK);
304 
305 	len = 0;
306 	off = 0;
307 	eofflag = 0;
308 
309 	for (;;) {
310 		error = vn_dir_next_dirent(vp, td, dirbuf, dirbuflen,
311 		    &dp, &len, &off, &eofflag);
312 		if (error != 0)
313 			goto out;
314 
315 		if (len == 0)
316 			break;
317 
318 		if (dp->d_type != DT_WHT && dp->d_fileno != 0 &&
319 		    strcmp(dp->d_name, dirname) == 0)
320 			goto out;
321 	}
322 
323 	error = ENOENT;
324 
325 out:
326 	free(dirbuf, M_TEMP);
327 	return (error);
328 }
329 
330 int
vop_stdaccess(struct vop_access_args * ap)331 vop_stdaccess(struct vop_access_args *ap)
332 {
333 
334 	KASSERT((ap->a_accmode & ~(VEXEC | VWRITE | VREAD | VADMIN |
335 	    VAPPEND)) == 0, ("invalid bit in accmode"));
336 
337 	return (VOP_ACCESSX(ap->a_vp, ap->a_accmode, ap->a_cred, ap->a_td));
338 }
339 
340 int
vop_stdaccessx(struct vop_accessx_args * ap)341 vop_stdaccessx(struct vop_accessx_args *ap)
342 {
343 	int error;
344 	accmode_t accmode = ap->a_accmode;
345 
346 	error = vfs_unixify_accmode(&accmode);
347 	if (error != 0)
348 		return (error);
349 
350 	if (accmode == 0)
351 		return (0);
352 
353 	return (VOP_ACCESS(ap->a_vp, accmode, ap->a_cred, ap->a_td));
354 }
355 
356 /*
357  * Advisory record locking support
358  */
359 int
vop_stdadvlock(struct vop_advlock_args * ap)360 vop_stdadvlock(struct vop_advlock_args *ap)
361 {
362 	struct vnode *vp;
363 	struct mount *mp;
364 	struct vattr vattr;
365 	int error;
366 
367 	vp = ap->a_vp;
368 
369 	/*
370 	 * Provide atomicity of open(O_CREAT | O_EXCL | O_EXLOCK) for
371 	 * local filesystems.  See vn_open_cred() for reciprocal part.
372 	 */
373 	mp = vp->v_mount;
374 	if (mp != NULL && (mp->mnt_flag & MNT_LOCAL) != 0 &&
375 	    ap->a_op == F_SETLK && (ap->a_flags & F_FIRSTOPEN) == 0) {
376 		VI_LOCK(vp);
377 		while ((vp->v_iflag & VI_FOPENING) != 0)
378 			msleep(vp, VI_MTX(vp), PLOCK, "lockfo", 0);
379 		VI_UNLOCK(vp);
380 	}
381 
382 	if (ap->a_fl->l_whence == SEEK_END) {
383 		/*
384 		 * The NFSv4 server must avoid doing a vn_lock() here, since it
385 		 * can deadlock the nfsd threads, due to a LOR.  Fortunately
386 		 * the NFSv4 server always uses SEEK_SET and this code is
387 		 * only required for the SEEK_END case.
388 		 */
389 		vn_lock(vp, LK_SHARED | LK_RETRY);
390 		error = VOP_GETATTR(vp, &vattr, curthread->td_ucred);
391 		VOP_UNLOCK(vp);
392 		if (error)
393 			return (error);
394 	} else
395 		vattr.va_size = 0;
396 
397 	return (lf_advlock(ap, &(vp->v_lockf), vattr.va_size));
398 }
399 
400 int
vop_stdadvlockasync(struct vop_advlockasync_args * ap)401 vop_stdadvlockasync(struct vop_advlockasync_args *ap)
402 {
403 	struct vnode *vp;
404 	struct vattr vattr;
405 	int error;
406 
407 	vp = ap->a_vp;
408 	if (ap->a_fl->l_whence == SEEK_END) {
409 		/* The size argument is only needed for SEEK_END. */
410 		vn_lock(vp, LK_SHARED | LK_RETRY);
411 		error = VOP_GETATTR(vp, &vattr, curthread->td_ucred);
412 		VOP_UNLOCK(vp);
413 		if (error)
414 			return (error);
415 	} else
416 		vattr.va_size = 0;
417 
418 	return (lf_advlockasync(ap, &(vp->v_lockf), vattr.va_size));
419 }
420 
421 int
vop_stdadvlockpurge(struct vop_advlockpurge_args * ap)422 vop_stdadvlockpurge(struct vop_advlockpurge_args *ap)
423 {
424 	struct vnode *vp;
425 
426 	vp = ap->a_vp;
427 	lf_purgelocks(vp, &vp->v_lockf);
428 	return (0);
429 }
430 
431 /*
432  * vop_stdpathconf:
433  *
434  * Standard implementation of POSIX pathconf, to get information about limits
435  * for a filesystem.
436  * Override per filesystem for the case where the filesystem has smaller
437  * limits.
438  */
439 int
vop_stdpathconf(struct vop_pathconf_args * ap)440 vop_stdpathconf(struct vop_pathconf_args *ap)
441 {
442 
443 	switch (ap->a_name) {
444 		case _PC_ASYNC_IO:
445 			*ap->a_retval = _POSIX_ASYNCHRONOUS_IO;
446 			return (0);
447 		case _PC_PATH_MAX:
448 			*ap->a_retval = PATH_MAX;
449 			return (0);
450 		case _PC_ACL_EXTENDED:
451 		case _PC_ACL_NFS4:
452 		case _PC_CAP_PRESENT:
453 		case _PC_INF_PRESENT:
454 		case _PC_MAC_PRESENT:
455 			*ap->a_retval = 0;
456 			return (0);
457 		default:
458 			return (EINVAL);
459 	}
460 	/* NOTREACHED */
461 }
462 
463 /*
464  * Standard lock, unlock and islocked functions.
465  */
466 int
vop_stdlock(struct vop_lock1_args * ap)467 vop_stdlock(struct vop_lock1_args *ap)
468 {
469 	struct vnode *vp = ap->a_vp;
470 	struct mtx *ilk;
471 
472 	ilk = VI_MTX(vp);
473 	return (lockmgr_lock_flags(vp->v_vnlock, ap->a_flags,
474 	    &ilk->lock_object, ap->a_file, ap->a_line));
475 }
476 
477 /* See above. */
478 int
vop_stdunlock(struct vop_unlock_args * ap)479 vop_stdunlock(struct vop_unlock_args *ap)
480 {
481 	struct vnode *vp = ap->a_vp;
482 
483 	return (lockmgr_unlock(vp->v_vnlock));
484 }
485 
486 /* See above. */
487 int
vop_stdislocked(struct vop_islocked_args * ap)488 vop_stdislocked(struct vop_islocked_args *ap)
489 {
490 
491 	return (lockstatus(ap->a_vp->v_vnlock));
492 }
493 
494 /*
495  * Variants of the above set.
496  *
497  * Differences are:
498  * - shared locking disablement is not supported
499  * - v_vnlock pointer is not honored
500  */
501 int
vop_lock(struct vop_lock1_args * ap)502 vop_lock(struct vop_lock1_args *ap)
503 {
504 	struct vnode *vp = ap->a_vp;
505 	int flags = ap->a_flags;
506 	struct mtx *ilk;
507 
508 	MPASS(vp->v_vnlock == &vp->v_lock);
509 
510 	if (__predict_false((flags & ~(LK_TYPE_MASK | LK_NODDLKTREAT | LK_RETRY)) != 0))
511 		goto other;
512 
513 	switch (flags & LK_TYPE_MASK) {
514 	case LK_SHARED:
515 		return (lockmgr_slock(&vp->v_lock, flags, ap->a_file, ap->a_line));
516 	case LK_EXCLUSIVE:
517 		return (lockmgr_xlock(&vp->v_lock, flags, ap->a_file, ap->a_line));
518 	}
519 other:
520 	ilk = VI_MTX(vp);
521 	return (lockmgr_lock_flags(&vp->v_lock, flags,
522 	    &ilk->lock_object, ap->a_file, ap->a_line));
523 }
524 
525 int
vop_unlock(struct vop_unlock_args * ap)526 vop_unlock(struct vop_unlock_args *ap)
527 {
528 	struct vnode *vp = ap->a_vp;
529 
530 	MPASS(vp->v_vnlock == &vp->v_lock);
531 
532 	return (lockmgr_unlock(&vp->v_lock));
533 }
534 
535 int
vop_islocked(struct vop_islocked_args * ap)536 vop_islocked(struct vop_islocked_args *ap)
537 {
538 	struct vnode *vp = ap->a_vp;
539 
540 	MPASS(vp->v_vnlock == &vp->v_lock);
541 
542 	return (lockstatus(&vp->v_lock));
543 }
544 
545 /*
546  * Return true for select/poll.
547  */
548 int
vop_nopoll(struct vop_poll_args * ap)549 vop_nopoll(struct vop_poll_args *ap)
550 {
551 
552 	if (ap->a_events & ~POLLSTANDARD)
553 		return (POLLNVAL);
554 	return (ap->a_events & (POLLIN | POLLOUT | POLLRDNORM | POLLWRNORM));
555 }
556 
557 /*
558  * Implement poll for local filesystems that support it.
559  */
560 int
vop_stdpoll(struct vop_poll_args * ap)561 vop_stdpoll(struct vop_poll_args *ap)
562 {
563 	if (ap->a_events & ~POLLSTANDARD)
564 		return (vn_pollrecord(ap->a_vp, ap->a_td, ap->a_events));
565 	return (ap->a_events & (POLLIN | POLLOUT | POLLRDNORM | POLLWRNORM));
566 }
567 
568 /*
569  * Return our mount point, as we will take charge of the writes.
570  */
571 int
vop_stdgetwritemount(struct vop_getwritemount_args * ap)572 vop_stdgetwritemount(struct vop_getwritemount_args *ap)
573 {
574 	struct mount *mp;
575 	struct vnode *vp;
576 
577 	/*
578 	 * Note that having a reference does not prevent forced unmount from
579 	 * setting ->v_mount to NULL after the lock gets released. This is of
580 	 * no consequence for typical consumers (most notably vn_start_write)
581 	 * since in this case the vnode is VIRF_DOOMED. Unmount might have
582 	 * progressed far enough that its completion is only delayed by the
583 	 * reference obtained here. The consumer only needs to concern itself
584 	 * with releasing it.
585 	 */
586 	vp = ap->a_vp;
587 	mp = vfs_ref_from_vp(vp);
588 	*(ap->a_mpp) = mp;
589 	return (0);
590 }
591 
592 /*
593  * If the file system doesn't implement VOP_BMAP, then return sensible defaults:
594  * - Return the vnode's bufobj instead of any underlying device's bufobj
595  * - Calculate the physical block number as if there were equal size
596  *   consecutive blocks, but
597  * - Report no contiguous runs of blocks.
598  */
599 int
vop_stdbmap(struct vop_bmap_args * ap)600 vop_stdbmap(struct vop_bmap_args *ap)
601 {
602 
603 	if (ap->a_bop != NULL)
604 		*ap->a_bop = &ap->a_vp->v_bufobj;
605 	if (ap->a_bnp != NULL)
606 		*ap->a_bnp = ap->a_bn * btodb(ap->a_vp->v_mount->mnt_stat.f_iosize);
607 	if (ap->a_runp != NULL)
608 		*ap->a_runp = 0;
609 	if (ap->a_runb != NULL)
610 		*ap->a_runb = 0;
611 	return (0);
612 }
613 
614 int
vop_stdfsync(struct vop_fsync_args * ap)615 vop_stdfsync(struct vop_fsync_args *ap)
616 {
617 
618 	return (vn_fsync_buf(ap->a_vp, ap->a_waitfor));
619 }
620 
621 static int
vop_stdfdatasync(struct vop_fdatasync_args * ap)622 vop_stdfdatasync(struct vop_fdatasync_args *ap)
623 {
624 
625 	return (VOP_FSYNC(ap->a_vp, MNT_WAIT, ap->a_td));
626 }
627 
628 int
vop_stdfdatasync_buf(struct vop_fdatasync_args * ap)629 vop_stdfdatasync_buf(struct vop_fdatasync_args *ap)
630 {
631 
632 	return (vn_fsync_buf(ap->a_vp, MNT_WAIT));
633 }
634 
635 /* XXX Needs good comment and more info in the manpage (VOP_GETPAGES(9)). */
636 int
vop_stdgetpages(struct vop_getpages_args * ap)637 vop_stdgetpages(struct vop_getpages_args *ap)
638 {
639 
640 	return vnode_pager_generic_getpages(ap->a_vp, ap->a_m,
641 	    ap->a_count, ap->a_rbehind, ap->a_rahead, NULL, NULL);
642 }
643 
644 static int
vop_stdgetpages_async(struct vop_getpages_async_args * ap)645 vop_stdgetpages_async(struct vop_getpages_async_args *ap)
646 {
647 	int error;
648 
649 	error = VOP_GETPAGES(ap->a_vp, ap->a_m, ap->a_count, ap->a_rbehind,
650 	    ap->a_rahead);
651 	if (ap->a_iodone != NULL)
652 		ap->a_iodone(ap->a_arg, ap->a_m, ap->a_count, error);
653 	return (error);
654 }
655 
656 int
vop_stdkqfilter(struct vop_kqfilter_args * ap)657 vop_stdkqfilter(struct vop_kqfilter_args *ap)
658 {
659 	return vfs_kqfilter(ap);
660 }
661 
662 /* XXX Needs good comment and more info in the manpage (VOP_PUTPAGES(9)). */
663 int
vop_stdputpages(struct vop_putpages_args * ap)664 vop_stdputpages(struct vop_putpages_args *ap)
665 {
666 
667 	return vnode_pager_generic_putpages(ap->a_vp, ap->a_m, ap->a_count,
668 	     ap->a_sync, ap->a_rtvals);
669 }
670 
671 int
vop_stdvptofh(struct vop_vptofh_args * ap)672 vop_stdvptofh(struct vop_vptofh_args *ap)
673 {
674 	return (EOPNOTSUPP);
675 }
676 
677 int
vop_stdvptocnp(struct vop_vptocnp_args * ap)678 vop_stdvptocnp(struct vop_vptocnp_args *ap)
679 {
680 	struct vnode *const vp = ap->a_vp;
681 	struct vnode **const dvp = ap->a_vpp;
682 	char *buf = ap->a_buf;
683 	size_t *buflen = ap->a_buflen;
684 	char *dirbuf;
685 	int i = *buflen;
686 	int error = 0, covered = 0;
687 	int eofflag, flags, locked;
688 	size_t dirbuflen, len;
689 	off_t off;
690 	ino_t fileno;
691 	struct vattr va;
692 	struct nameidata nd;
693 	struct thread *const td = curthread;
694 	struct ucred *const cred = td->td_ucred;
695 	struct dirent *dp;
696 	struct vnode *mvp;
697 
698 	if (vp->v_type != VDIR)
699 		return (ENOENT);
700 
701 	error = VOP_GETATTR(vp, &va, cred);
702 	if (error)
703 		return (error);
704 
705 	VREF(vp);
706 	locked = VOP_ISLOCKED(vp);
707 	VOP_UNLOCK(vp);
708 	NDINIT_ATVP(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF, UIO_SYSSPACE,
709 	    "..", vp, td);
710 	flags = FREAD;
711 	error = vn_open_cred(&nd, &flags, 0, VN_OPEN_NOAUDIT, cred, NULL);
712 	if (error) {
713 		vn_lock(vp, locked | LK_RETRY);
714 		return (error);
715 	}
716 	NDFREE(&nd, NDF_ONLY_PNBUF);
717 
718 	mvp = *dvp = nd.ni_vp;
719 
720 	if (vp->v_mount != (*dvp)->v_mount &&
721 	    ((*dvp)->v_vflag & VV_ROOT) &&
722 	    ((*dvp)->v_mount->mnt_flag & MNT_UNION)) {
723 		*dvp = (*dvp)->v_mount->mnt_vnodecovered;
724 		VREF(mvp);
725 		VOP_UNLOCK(mvp);
726 		vn_close(mvp, FREAD, cred, td);
727 		VREF(*dvp);
728 		vn_lock(*dvp, LK_SHARED | LK_RETRY);
729 		covered = 1;
730 	}
731 
732 	fileno = va.va_fileid;
733 
734 	dirbuflen = MAX(DEV_BSIZE, GENERIC_MAXDIRSIZ);
735 	if (dirbuflen < va.va_blocksize)
736 		dirbuflen = va.va_blocksize;
737 	dirbuf = malloc(dirbuflen, M_TEMP, M_WAITOK);
738 
739 	if ((*dvp)->v_type != VDIR) {
740 		error = ENOENT;
741 		goto out;
742 	}
743 
744 	len = 0;
745 	off = 0;
746 	eofflag = 0;
747 
748 	for (;;) {
749 		/* call VOP_READDIR of parent */
750 		error = vn_dir_next_dirent(*dvp, td,
751 		    dirbuf, dirbuflen, &dp, &len, &off, &eofflag);
752 		if (error != 0)
753 			goto out;
754 
755 		if (len == 0) {
756 			error = ENOENT;
757 			goto out;
758 		}
759 
760 		if ((dp->d_type != DT_WHT) &&
761 		    (dp->d_fileno == fileno)) {
762 			if (covered) {
763 				VOP_UNLOCK(*dvp);
764 				vn_lock(mvp, LK_SHARED | LK_RETRY);
765 				if (dirent_exists(mvp, dp->d_name, td) == 0) {
766 					error = ENOENT;
767 					VOP_UNLOCK(mvp);
768 					vn_lock(*dvp, LK_SHARED | LK_RETRY);
769 					goto out;
770 				}
771 				VOP_UNLOCK(mvp);
772 				vn_lock(*dvp, LK_SHARED | LK_RETRY);
773 			}
774 			i -= dp->d_namlen;
775 
776 			if (i < 0) {
777 				error = ENOMEM;
778 				goto out;
779 			}
780 			if (dp->d_namlen == 1 && dp->d_name[0] == '.') {
781 				error = ENOENT;
782 			} else {
783 				bcopy(dp->d_name, buf + i, dp->d_namlen);
784 				error = 0;
785 			}
786 			goto out;
787 		}
788 	}
789 
790 out:
791 	free(dirbuf, M_TEMP);
792 	if (!error) {
793 		*buflen = i;
794 		vref(*dvp);
795 	}
796 	if (covered) {
797 		vput(*dvp);
798 		vrele(mvp);
799 	} else {
800 		VOP_UNLOCK(mvp);
801 		vn_close(mvp, FREAD, cred, td);
802 	}
803 	vn_lock(vp, locked | LK_RETRY);
804 	return (error);
805 }
806 
807 int
vop_stdallocate(struct vop_allocate_args * ap)808 vop_stdallocate(struct vop_allocate_args *ap)
809 {
810 #ifdef __notyet__
811 	struct statfs *sfs;
812 	off_t maxfilesize = 0;
813 #endif
814 	struct iovec aiov;
815 	struct vattr vattr, *vap;
816 	struct uio auio;
817 	off_t fsize, len, cur, offset;
818 	uint8_t *buf;
819 	struct thread *td;
820 	struct vnode *vp;
821 	size_t iosize;
822 	int error;
823 
824 	buf = NULL;
825 	error = 0;
826 	td = curthread;
827 	vap = &vattr;
828 	vp = ap->a_vp;
829 	len = *ap->a_len;
830 	offset = *ap->a_offset;
831 
832 	error = VOP_GETATTR(vp, vap, ap->a_cred);
833 	if (error != 0)
834 		goto out;
835 	fsize = vap->va_size;
836 	iosize = vap->va_blocksize;
837 	if (iosize == 0)
838 		iosize = BLKDEV_IOSIZE;
839 	if (iosize > maxphys)
840 		iosize = maxphys;
841 	buf = malloc(iosize, M_TEMP, M_WAITOK);
842 
843 #ifdef __notyet__
844 	/*
845 	 * Check if the filesystem sets f_maxfilesize; if not use
846 	 * VOP_SETATTR to perform the check.
847 	 */
848 	sfs = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
849 	error = VFS_STATFS(vp->v_mount, sfs, td);
850 	if (error == 0)
851 		maxfilesize = sfs->f_maxfilesize;
852 	free(sfs, M_STATFS);
853 	if (error != 0)
854 		goto out;
855 	if (maxfilesize) {
856 		if (offset > maxfilesize || len > maxfilesize ||
857 		    offset + len > maxfilesize) {
858 			error = EFBIG;
859 			goto out;
860 		}
861 	} else
862 #endif
863 	if (offset + len > vap->va_size) {
864 		/*
865 		 * Test offset + len against the filesystem's maxfilesize.
866 		 */
867 		VATTR_NULL(vap);
868 		vap->va_size = offset + len;
869 		error = VOP_SETATTR(vp, vap, ap->a_cred);
870 		if (error != 0)
871 			goto out;
872 		VATTR_NULL(vap);
873 		vap->va_size = fsize;
874 		error = VOP_SETATTR(vp, vap, ap->a_cred);
875 		if (error != 0)
876 			goto out;
877 	}
878 
879 	for (;;) {
880 		/*
881 		 * Read and write back anything below the nominal file
882 		 * size.  There's currently no way outside the filesystem
883 		 * to know whether this area is sparse or not.
884 		 */
885 		cur = iosize;
886 		if ((offset % iosize) != 0)
887 			cur -= (offset % iosize);
888 		if (cur > len)
889 			cur = len;
890 		if (offset < fsize) {
891 			aiov.iov_base = buf;
892 			aiov.iov_len = cur;
893 			auio.uio_iov = &aiov;
894 			auio.uio_iovcnt = 1;
895 			auio.uio_offset = offset;
896 			auio.uio_resid = cur;
897 			auio.uio_segflg = UIO_SYSSPACE;
898 			auio.uio_rw = UIO_READ;
899 			auio.uio_td = td;
900 			error = VOP_READ(vp, &auio, ap->a_ioflag, ap->a_cred);
901 			if (error != 0)
902 				break;
903 			if (auio.uio_resid > 0) {
904 				bzero(buf + cur - auio.uio_resid,
905 				    auio.uio_resid);
906 			}
907 		} else {
908 			bzero(buf, cur);
909 		}
910 
911 		aiov.iov_base = buf;
912 		aiov.iov_len = cur;
913 		auio.uio_iov = &aiov;
914 		auio.uio_iovcnt = 1;
915 		auio.uio_offset = offset;
916 		auio.uio_resid = cur;
917 		auio.uio_segflg = UIO_SYSSPACE;
918 		auio.uio_rw = UIO_WRITE;
919 		auio.uio_td = td;
920 
921 		error = VOP_WRITE(vp, &auio, ap->a_ioflag, ap->a_cred);
922 		if (error != 0)
923 			break;
924 
925 		len -= cur;
926 		offset += cur;
927 		if (len == 0)
928 			break;
929 		if (should_yield())
930 			break;
931 	}
932 
933  out:
934 	*ap->a_len = len;
935 	*ap->a_offset = offset;
936 	free(buf, M_TEMP);
937 	return (error);
938 }
939 
940 int
vop_stdadvise(struct vop_advise_args * ap)941 vop_stdadvise(struct vop_advise_args *ap)
942 {
943 	struct vnode *vp;
944 	struct bufobj *bo;
945 	daddr_t startn, endn;
946 	off_t bstart, bend, start, end;
947 	int bsize, error;
948 
949 	vp = ap->a_vp;
950 	switch (ap->a_advice) {
951 	case POSIX_FADV_WILLNEED:
952 		/*
953 		 * Do nothing for now.  Filesystems should provide a
954 		 * custom method which starts an asynchronous read of
955 		 * the requested region.
956 		 */
957 		error = 0;
958 		break;
959 	case POSIX_FADV_DONTNEED:
960 		error = 0;
961 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
962 		if (VN_IS_DOOMED(vp)) {
963 			VOP_UNLOCK(vp);
964 			break;
965 		}
966 
967 		/*
968 		 * Round to block boundaries (and later possibly further to
969 		 * page boundaries).  Applications cannot reasonably be aware
970 		 * of the boundaries, and the rounding must be to expand at
971 		 * both extremities to cover enough.  It still doesn't cover
972 		 * read-ahead.  For partial blocks, this gives unnecessary
973 		 * discarding of buffers but is efficient enough since the
974 		 * pages usually remain in VMIO for some time.
975 		 */
976 		bsize = vp->v_bufobj.bo_bsize;
977 		bstart = rounddown(ap->a_start, bsize);
978 		bend = roundup(ap->a_end, bsize);
979 
980 		/*
981 		 * Deactivate pages in the specified range from the backing VM
982 		 * object.  Pages that are resident in the buffer cache will
983 		 * remain wired until their corresponding buffers are released
984 		 * below.
985 		 */
986 		if (vp->v_object != NULL) {
987 			start = trunc_page(bstart);
988 			end = round_page(bend);
989 			VM_OBJECT_RLOCK(vp->v_object);
990 			vm_object_page_noreuse(vp->v_object, OFF_TO_IDX(start),
991 			    OFF_TO_IDX(end));
992 			VM_OBJECT_RUNLOCK(vp->v_object);
993 		}
994 
995 		bo = &vp->v_bufobj;
996 		BO_RLOCK(bo);
997 		startn = bstart / bsize;
998 		endn = bend / bsize;
999 		error = bnoreuselist(&bo->bo_clean, bo, startn, endn);
1000 		if (error == 0)
1001 			error = bnoreuselist(&bo->bo_dirty, bo, startn, endn);
1002 		BO_RUNLOCK(bo);
1003 		VOP_UNLOCK(vp);
1004 		break;
1005 	default:
1006 		error = EINVAL;
1007 		break;
1008 	}
1009 	return (error);
1010 }
1011 
1012 int
vop_stdunp_bind(struct vop_unp_bind_args * ap)1013 vop_stdunp_bind(struct vop_unp_bind_args *ap)
1014 {
1015 
1016 	ap->a_vp->v_unpcb = ap->a_unpcb;
1017 	return (0);
1018 }
1019 
1020 int
vop_stdunp_connect(struct vop_unp_connect_args * ap)1021 vop_stdunp_connect(struct vop_unp_connect_args *ap)
1022 {
1023 
1024 	*ap->a_unpcb = ap->a_vp->v_unpcb;
1025 	return (0);
1026 }
1027 
1028 int
vop_stdunp_detach(struct vop_unp_detach_args * ap)1029 vop_stdunp_detach(struct vop_unp_detach_args *ap)
1030 {
1031 
1032 	ap->a_vp->v_unpcb = NULL;
1033 	return (0);
1034 }
1035 
1036 static int
vop_stdis_text(struct vop_is_text_args * ap)1037 vop_stdis_text(struct vop_is_text_args *ap)
1038 {
1039 
1040 	return (ap->a_vp->v_writecount < 0);
1041 }
1042 
1043 int
vop_stdset_text(struct vop_set_text_args * ap)1044 vop_stdset_text(struct vop_set_text_args *ap)
1045 {
1046 	struct vnode *vp;
1047 	struct mount *mp;
1048 	int error, n;
1049 
1050 	vp = ap->a_vp;
1051 
1052 	/*
1053 	 * Avoid the interlock if execs are already present.
1054 	 */
1055 	n = atomic_load_int(&vp->v_writecount);
1056 	for (;;) {
1057 		if (n > -1) {
1058 			break;
1059 		}
1060 		if (atomic_fcmpset_int(&vp->v_writecount, &n, n - 1)) {
1061 			return (0);
1062 		}
1063 	}
1064 
1065 	VI_LOCK(vp);
1066 	if (vp->v_writecount > 0) {
1067 		error = ETXTBSY;
1068 	} else {
1069 		/*
1070 		 * If requested by fs, keep a use reference to the
1071 		 * vnode until the last text reference is released.
1072 		 */
1073 		mp = vp->v_mount;
1074 		if (mp != NULL && (mp->mnt_kern_flag & MNTK_TEXT_REFS) != 0 &&
1075 		    vp->v_writecount == 0) {
1076 			VNPASS((vp->v_iflag & VI_TEXT_REF) == 0, vp);
1077 			vp->v_iflag |= VI_TEXT_REF;
1078 			vrefl(vp);
1079 		}
1080 
1081 		atomic_subtract_int(&vp->v_writecount, 1);
1082 		error = 0;
1083 	}
1084 	VI_UNLOCK(vp);
1085 	return (error);
1086 }
1087 
1088 static int
vop_stdunset_text(struct vop_unset_text_args * ap)1089 vop_stdunset_text(struct vop_unset_text_args *ap)
1090 {
1091 	struct vnode *vp;
1092 	int error, n;
1093 	bool last;
1094 
1095 	vp = ap->a_vp;
1096 
1097 	/*
1098 	 * Avoid the interlock if this is not the last exec.
1099 	 */
1100 	n = atomic_load_int(&vp->v_writecount);
1101 	for (;;) {
1102 		if (n >= -1) {
1103 			break;
1104 		}
1105 		if (atomic_fcmpset_int(&vp->v_writecount, &n, n + 1)) {
1106 			return (0);
1107 		}
1108 	}
1109 
1110 	last = false;
1111 	VI_LOCK(vp);
1112 	if (vp->v_writecount < 0) {
1113 		if ((vp->v_iflag & VI_TEXT_REF) != 0 &&
1114 		    vp->v_writecount == -1) {
1115 			last = true;
1116 			vp->v_iflag &= ~VI_TEXT_REF;
1117 		}
1118 		atomic_add_int(&vp->v_writecount, 1);
1119 		error = 0;
1120 	} else {
1121 		error = EINVAL;
1122 	}
1123 	VI_UNLOCK(vp);
1124 	if (last)
1125 		vunref(vp);
1126 	return (error);
1127 }
1128 
1129 static int
vop_stdadd_writecount(struct vop_add_writecount_args * ap)1130 vop_stdadd_writecount(struct vop_add_writecount_args *ap)
1131 {
1132 	struct vnode *vp;
1133 	struct mount *mp;
1134 	int error;
1135 
1136 	vp = ap->a_vp;
1137 	VI_LOCK_FLAGS(vp, MTX_DUPOK);
1138 	if (vp->v_writecount < 0) {
1139 		error = ETXTBSY;
1140 	} else {
1141 		VNASSERT(vp->v_writecount + ap->a_inc >= 0, vp,
1142 		    ("neg writecount increment %d", ap->a_inc));
1143 		if (vp->v_writecount == 0) {
1144 			mp = vp->v_mount;
1145 			if (mp != NULL && (mp->mnt_kern_flag & MNTK_NOMSYNC) == 0)
1146 				vlazy(vp);
1147 		}
1148 		vp->v_writecount += ap->a_inc;
1149 		error = 0;
1150 	}
1151 	VI_UNLOCK(vp);
1152 	return (error);
1153 }
1154 
1155 int
vop_stdneed_inactive(struct vop_need_inactive_args * ap)1156 vop_stdneed_inactive(struct vop_need_inactive_args *ap)
1157 {
1158 
1159 	return (1);
1160 }
1161 
1162 int
vop_stdioctl(struct vop_ioctl_args * ap)1163 vop_stdioctl(struct vop_ioctl_args *ap)
1164 {
1165 	struct vnode *vp;
1166 	struct vattr va;
1167 	off_t *offp;
1168 	int error;
1169 
1170 	switch (ap->a_command) {
1171 	case FIOSEEKDATA:
1172 	case FIOSEEKHOLE:
1173 		vp = ap->a_vp;
1174 		error = vn_lock(vp, LK_SHARED);
1175 		if (error != 0)
1176 			return (EBADF);
1177 		if (vp->v_type == VREG)
1178 			error = VOP_GETATTR(vp, &va, ap->a_cred);
1179 		else
1180 			error = ENOTTY;
1181 		if (error == 0) {
1182 			offp = ap->a_data;
1183 			if (*offp < 0 || *offp >= va.va_size)
1184 				error = ENXIO;
1185 			else if (ap->a_command == FIOSEEKHOLE)
1186 				*offp = va.va_size;
1187 		}
1188 		VOP_UNLOCK(vp);
1189 		break;
1190 	default:
1191 		error = ENOTTY;
1192 		break;
1193 	}
1194 	return (error);
1195 }
1196 
1197 /*
1198  * vfs default ops
1199  * used to fill the vfs function table to get reasonable default return values.
1200  */
1201 int
vfs_stdroot(struct mount * mp,int flags,struct vnode ** vpp)1202 vfs_stdroot(struct mount *mp, int flags, struct vnode **vpp)
1203 {
1204 
1205 	return (EOPNOTSUPP);
1206 }
1207 
1208 int
vfs_stdstatfs(struct mount * mp,struct statfs * sbp)1209 vfs_stdstatfs(struct mount *mp, struct statfs *sbp)
1210 {
1211 
1212 	return (EOPNOTSUPP);
1213 }
1214 
1215 int
vfs_stdquotactl(struct mount * mp,int cmds,uid_t uid,void * arg)1216 vfs_stdquotactl(struct mount *mp, int cmds, uid_t uid, void *arg)
1217 {
1218 
1219 	return (EOPNOTSUPP);
1220 }
1221 
1222 int
vfs_stdsync(struct mount * mp,int waitfor)1223 vfs_stdsync(struct mount *mp, int waitfor)
1224 {
1225 	struct vnode *vp, *mvp;
1226 	struct thread *td;
1227 	int error, lockreq, allerror = 0;
1228 
1229 	td = curthread;
1230 	lockreq = LK_EXCLUSIVE | LK_INTERLOCK;
1231 	if (waitfor != MNT_WAIT)
1232 		lockreq |= LK_NOWAIT;
1233 	/*
1234 	 * Force stale buffer cache information to be flushed.
1235 	 */
1236 loop:
1237 	MNT_VNODE_FOREACH_ALL(vp, mp, mvp) {
1238 		if (vp->v_bufobj.bo_dirty.bv_cnt == 0) {
1239 			VI_UNLOCK(vp);
1240 			continue;
1241 		}
1242 		if ((error = vget(vp, lockreq)) != 0) {
1243 			if (error == ENOENT) {
1244 				MNT_VNODE_FOREACH_ALL_ABORT(mp, mvp);
1245 				goto loop;
1246 			}
1247 			continue;
1248 		}
1249 		error = VOP_FSYNC(vp, waitfor, td);
1250 		if (error)
1251 			allerror = error;
1252 		vput(vp);
1253 	}
1254 	return (allerror);
1255 }
1256 
1257 int
vfs_stdnosync(struct mount * mp,int waitfor)1258 vfs_stdnosync(struct mount *mp, int waitfor)
1259 {
1260 
1261 	return (0);
1262 }
1263 
1264 static int
vop_stdcopy_file_range(struct vop_copy_file_range_args * ap)1265 vop_stdcopy_file_range(struct vop_copy_file_range_args *ap)
1266 {
1267 	int error;
1268 
1269 	error = vn_generic_copy_file_range(ap->a_invp, ap->a_inoffp,
1270 	    ap->a_outvp, ap->a_outoffp, ap->a_lenp, ap->a_flags, ap->a_incred,
1271 	    ap->a_outcred, ap->a_fsizetd);
1272 	return (error);
1273 }
1274 
1275 int
vfs_stdvget(struct mount * mp,ino_t ino,int flags,struct vnode ** vpp)1276 vfs_stdvget(struct mount *mp, ino_t ino, int flags, struct vnode **vpp)
1277 {
1278 
1279 	return (EOPNOTSUPP);
1280 }
1281 
1282 int
vfs_stdfhtovp(struct mount * mp,struct fid * fhp,int flags,struct vnode ** vpp)1283 vfs_stdfhtovp(struct mount *mp, struct fid *fhp, int flags, struct vnode **vpp)
1284 {
1285 
1286 	return (EOPNOTSUPP);
1287 }
1288 
1289 int
vfs_stdinit(struct vfsconf * vfsp)1290 vfs_stdinit(struct vfsconf *vfsp)
1291 {
1292 
1293 	return (0);
1294 }
1295 
1296 int
vfs_stduninit(struct vfsconf * vfsp)1297 vfs_stduninit(struct vfsconf *vfsp)
1298 {
1299 
1300 	return(0);
1301 }
1302 
1303 int
vfs_stdextattrctl(struct mount * mp,int cmd,struct vnode * filename_vp,int attrnamespace,const char * attrname)1304 vfs_stdextattrctl(struct mount *mp, int cmd, struct vnode *filename_vp,
1305     int attrnamespace, const char *attrname)
1306 {
1307 
1308 	if (filename_vp != NULL)
1309 		VOP_UNLOCK(filename_vp);
1310 	return (EOPNOTSUPP);
1311 }
1312 
1313 int
vfs_stdsysctl(struct mount * mp,fsctlop_t op,struct sysctl_req * req)1314 vfs_stdsysctl(struct mount *mp, fsctlop_t op, struct sysctl_req *req)
1315 {
1316 
1317 	return (EOPNOTSUPP);
1318 }
1319 
1320 static vop_bypass_t *
bp_by_off(struct vop_vector * vop,struct vop_generic_args * a)1321 bp_by_off(struct vop_vector *vop, struct vop_generic_args *a)
1322 {
1323 
1324 	return (*(vop_bypass_t **)((char *)vop + a->a_desc->vdesc_vop_offset));
1325 }
1326 
1327 int
vop_sigdefer(struct vop_vector * vop,struct vop_generic_args * a)1328 vop_sigdefer(struct vop_vector *vop, struct vop_generic_args *a)
1329 {
1330 	vop_bypass_t *bp;
1331 	int prev_stops, rc;
1332 
1333 	bp = bp_by_off(vop, a);
1334 	MPASS(bp != NULL);
1335 
1336 	prev_stops = sigdeferstop(SIGDEFERSTOP_SILENT);
1337 	rc = bp(a);
1338 	sigallowstop(prev_stops);
1339 	return (rc);
1340 }
1341 
1342 static int
vop_stdstat(struct vop_stat_args * a)1343 vop_stdstat(struct vop_stat_args *a)
1344 {
1345 	struct vattr vattr;
1346 	struct vattr *vap;
1347 	struct vnode *vp;
1348 	struct stat *sb;
1349 	int error;
1350 	u_short mode;
1351 
1352 	vp = a->a_vp;
1353 	sb = a->a_sb;
1354 
1355 	error = vop_stat_helper_pre(a);
1356 	if (error != 0)
1357 		return (error);
1358 
1359 	vap = &vattr;
1360 
1361 	/*
1362 	 * Initialize defaults for new and unusual fields, so that file
1363 	 * systems which don't support these fields don't need to know
1364 	 * about them.
1365 	 */
1366 	vap->va_birthtime.tv_sec = -1;
1367 	vap->va_birthtime.tv_nsec = 0;
1368 	vap->va_fsid = VNOVAL;
1369 	vap->va_rdev = NODEV;
1370 
1371 	error = VOP_GETATTR(vp, vap, a->a_active_cred);
1372 	if (error)
1373 		goto out;
1374 
1375 	/*
1376 	 * Zero the spare stat fields
1377 	 */
1378 	bzero(sb, sizeof *sb);
1379 
1380 	/*
1381 	 * Copy from vattr table
1382 	 */
1383 	if (vap->va_fsid != VNOVAL)
1384 		sb->st_dev = vap->va_fsid;
1385 	else
1386 		sb->st_dev = vp->v_mount->mnt_stat.f_fsid.val[0];
1387 	sb->st_ino = vap->va_fileid;
1388 	mode = vap->va_mode;
1389 	switch (vap->va_type) {
1390 	case VREG:
1391 		mode |= S_IFREG;
1392 		break;
1393 	case VDIR:
1394 		mode |= S_IFDIR;
1395 		break;
1396 	case VBLK:
1397 		mode |= S_IFBLK;
1398 		break;
1399 	case VCHR:
1400 		mode |= S_IFCHR;
1401 		break;
1402 	case VLNK:
1403 		mode |= S_IFLNK;
1404 		break;
1405 	case VSOCK:
1406 		mode |= S_IFSOCK;
1407 		break;
1408 	case VFIFO:
1409 		mode |= S_IFIFO;
1410 		break;
1411 	default:
1412 		error = EBADF;
1413 		goto out;
1414 	}
1415 	sb->st_mode = mode;
1416 	sb->st_nlink = vap->va_nlink;
1417 	sb->st_uid = vap->va_uid;
1418 	sb->st_gid = vap->va_gid;
1419 	sb->st_rdev = vap->va_rdev;
1420 	if (vap->va_size > OFF_MAX) {
1421 		error = EOVERFLOW;
1422 		goto out;
1423 	}
1424 	sb->st_size = vap->va_size;
1425 	sb->st_atim.tv_sec = vap->va_atime.tv_sec;
1426 	sb->st_atim.tv_nsec = vap->va_atime.tv_nsec;
1427 	sb->st_mtim.tv_sec = vap->va_mtime.tv_sec;
1428 	sb->st_mtim.tv_nsec = vap->va_mtime.tv_nsec;
1429 	sb->st_ctim.tv_sec = vap->va_ctime.tv_sec;
1430 	sb->st_ctim.tv_nsec = vap->va_ctime.tv_nsec;
1431 	sb->st_birthtim.tv_sec = vap->va_birthtime.tv_sec;
1432 	sb->st_birthtim.tv_nsec = vap->va_birthtime.tv_nsec;
1433 
1434 	/*
1435 	 * According to www.opengroup.org, the meaning of st_blksize is
1436 	 *   "a filesystem-specific preferred I/O block size for this
1437 	 *    object.  In some filesystem types, this may vary from file
1438 	 *    to file"
1439 	 * Use minimum/default of PAGE_SIZE (e.g. for VCHR).
1440 	 */
1441 
1442 	sb->st_blksize = max(PAGE_SIZE, vap->va_blocksize);
1443 	sb->st_flags = vap->va_flags;
1444 	sb->st_blocks = vap->va_bytes / S_BLKSIZE;
1445 	sb->st_gen = vap->va_gen;
1446 out:
1447 	return (vop_stat_helper_post(a, error));
1448 }
1449 
1450 static int
vop_stdread_pgcache(struct vop_read_pgcache_args * ap __unused)1451 vop_stdread_pgcache(struct vop_read_pgcache_args *ap __unused)
1452 {
1453 	return (EJUSTRETURN);
1454 }
1455 
1456 static int
vop_stdvput_pair(struct vop_vput_pair_args * ap)1457 vop_stdvput_pair(struct vop_vput_pair_args *ap)
1458 {
1459 	struct vnode *dvp, *vp, **vpp;
1460 
1461 	dvp = ap->a_dvp;
1462 	vpp = ap->a_vpp;
1463 	vput(dvp);
1464 	if (vpp != NULL && ap->a_unlock_vp && (vp = *vpp) != NULL)
1465 		vput(vp);
1466 	return (0);
1467 }
1468 
1469 static int
vop_stdgetlowvnode(struct vop_getlowvnode_args * ap)1470 vop_stdgetlowvnode(struct vop_getlowvnode_args *ap)
1471 {
1472 	vref(ap->a_vp);
1473 	*ap->a_vplp = ap->a_vp;
1474 	return (0);
1475 }
1476