xref: /freebsd/sys/kern/vfs_default.c (revision f1f230439fa48581f40a57f095627f667a9713c3)
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  * Redistribution and use in source and binary forms, with or without
11  * modification, are permitted provided that the following conditions
12  * are met:
13  * 1. Redistributions of source code must retain the above copyright
14  *    notice, this list of conditions and the following disclaimer.
15  * 2. Redistributions in binary form must reproduce the above copyright
16  *    notice, this list of conditions and the following disclaimer in the
17  *    documentation and/or other materials provided with the distribution.
18  * 3. Neither the name of the University nor the names of its contributors
19  *    may be used to endorse or promote products derived from this software
20  *    without specific prior written permission.
21  *
22  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
23  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
24  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
25  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
26  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
27  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
28  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
29  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
30  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
31  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32  * SUCH DAMAGE.
33  */
34 
35 #include <sys/param.h>
36 #include <sys/systm.h>
37 #include <sys/bio.h>
38 #include <sys/buf.h>
39 #include <sys/conf.h>
40 #include <sys/event.h>
41 #include <sys/filio.h>
42 #include <sys/inotify.h>
43 #include <sys/kernel.h>
44 #include <sys/limits.h>
45 #include <sys/lock.h>
46 #include <sys/lockf.h>
47 #include <sys/malloc.h>
48 #include <sys/mount.h>
49 #include <sys/namei.h>
50 #include <sys/rwlock.h>
51 #include <sys/fcntl.h>
52 #include <sys/unistd.h>
53 #include <sys/vnode.h>
54 #include <sys/dirent.h>
55 #include <sys/poll.h>
56 #include <sys/stat.h>
57 #include <security/audit/audit.h>
58 #include <sys/priv.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	dirent_exists(struct vnode *vp, const char *dirname,
75 			      struct thread *td);
76 
77 static int vop_stdis_text(struct vop_is_text_args *ap);
78 static int vop_stdunset_text(struct vop_unset_text_args *ap);
79 static int vop_stdadd_writecount(struct vop_add_writecount_args *ap);
80 static int vop_stdcopy_file_range(struct vop_copy_file_range_args *ap);
81 static int vop_stdfdatasync(struct vop_fdatasync_args *ap);
82 static int vop_stdgetpages_async(struct vop_getpages_async_args *ap);
83 static int vop_stdread_pgcache(struct vop_read_pgcache_args *ap);
84 static int vop_stdstat(struct vop_stat_args *ap);
85 static int vop_stdvput_pair(struct vop_vput_pair_args *ap);
86 static int vop_stdgetlowvnode(struct vop_getlowvnode_args *ap);
87 
88 /*
89  * This vnode table stores what we want to do if the filesystem doesn't
90  * implement a particular VOP.
91  *
92  * If there is no specific entry here, we will return EOPNOTSUPP.
93  *
94  * Note that every filesystem has to implement either vop_access
95  * or vop_accessx; failing to do so will result in immediate crash
96  * due to stack overflow, as vop_stdaccess() calls vop_stdaccessx(),
97  * which calls vop_stdaccess() etc.
98  */
99 
100 struct vop_vector default_vnodeops = {
101 	.vop_default =		NULL,
102 	.vop_bypass =		VOP_EOPNOTSUPP,
103 
104 	.vop_access =		vop_stdaccess,
105 	.vop_accessx =		vop_stdaccessx,
106 	.vop_advise =		vop_stdadvise,
107 	.vop_advlock =		vop_stdadvlock,
108 	.vop_advlockasync =	vop_stdadvlockasync,
109 	.vop_advlockpurge =	vop_stdadvlockpurge,
110 	.vop_allocate =		vop_stdallocate,
111 	.vop_deallocate =	vop_stddeallocate,
112 	.vop_bmap =		vop_stdbmap,
113 	.vop_close =		VOP_NULL,
114 	.vop_fsync =		VOP_NULL,
115 	.vop_stat =		vop_stdstat,
116 	.vop_fdatasync =	vop_stdfdatasync,
117 	.vop_getlowvnode =	vop_stdgetlowvnode,
118 	.vop_getpages =		vop_stdgetpages,
119 	.vop_getpages_async =	vop_stdgetpages_async,
120 	.vop_getwritemount =	vop_stdgetwritemount,
121 	.vop_inactive =		VOP_NULL,
122 	.vop_need_inactive =	vop_stdneed_inactive,
123 	.vop_inotify =		vop_stdinotify,
124 	.vop_inotify_add_watch = vop_stdinotify_add_watch,
125 	.vop_ioctl =		vop_stdioctl,
126 	.vop_kqfilter =		vop_stdkqfilter,
127 	.vop_islocked =		vop_stdislocked,
128 	.vop_lock1 =		vop_stdlock,
129 	.vop_lookup =		vop_nolookup,
130 	.vop_open =		VOP_NULL,
131 	.vop_pathconf =		VOP_EINVAL,
132 	.vop_poll =		vop_nopoll,
133 	.vop_putpages =		vop_stdputpages,
134 	.vop_readlink =		VOP_EINVAL,
135 	.vop_read_pgcache =	vop_stdread_pgcache,
136 	.vop_rename =		vop_norename,
137 	.vop_revoke =		VOP_PANIC,
138 	.vop_strategy =		vop_nostrategy,
139 	.vop_unlock =		vop_stdunlock,
140 	.vop_vptocnp =		vop_stdvptocnp,
141 	.vop_vptofh =		vop_stdvptofh,
142 	.vop_unp_bind =		vop_stdunp_bind,
143 	.vop_unp_connect =	vop_stdunp_connect,
144 	.vop_unp_detach =	vop_stdunp_detach,
145 	.vop_is_text =		vop_stdis_text,
146 	.vop_set_text =		vop_stdset_text,
147 	.vop_unset_text =	vop_stdunset_text,
148 	.vop_add_writecount =	vop_stdadd_writecount,
149 	.vop_copy_file_range =	vop_stdcopy_file_range,
150 	.vop_vput_pair =	vop_stdvput_pair,
151 };
152 VFS_VOP_VECTOR_REGISTER(default_vnodeops);
153 
154 /*
155  * Series of placeholder functions for various error returns for
156  * VOPs.
157  */
158 
159 int
vop_eopnotsupp(struct vop_generic_args * ap)160 vop_eopnotsupp(struct vop_generic_args *ap)
161 {
162 	/*
163 	printf("vop_notsupp[%s]\n", ap->a_desc->vdesc_name);
164 	*/
165 
166 	return (EOPNOTSUPP);
167 }
168 
169 int
vop_ebadf(struct vop_generic_args * ap)170 vop_ebadf(struct vop_generic_args *ap)
171 {
172 
173 	return (EBADF);
174 }
175 
176 int
vop_enotty(struct vop_generic_args * ap)177 vop_enotty(struct vop_generic_args *ap)
178 {
179 
180 	return (ENOTTY);
181 }
182 
183 int
vop_einval(struct vop_generic_args * ap)184 vop_einval(struct vop_generic_args *ap)
185 {
186 
187 	return (EINVAL);
188 }
189 
190 int
vop_enoent(struct vop_generic_args * ap)191 vop_enoent(struct vop_generic_args *ap)
192 {
193 
194 	return (ENOENT);
195 }
196 
197 int
vop_eagain(struct vop_generic_args * ap)198 vop_eagain(struct vop_generic_args *ap)
199 {
200 
201 	return (EAGAIN);
202 }
203 
204 int
vop_null(struct vop_generic_args * ap)205 vop_null(struct vop_generic_args *ap)
206 {
207 
208 	return (0);
209 }
210 
211 /*
212  * Helper function to panic on some bad VOPs in some filesystems.
213  */
214 int
vop_panic(struct vop_generic_args * ap)215 vop_panic(struct vop_generic_args *ap)
216 {
217 
218 	panic("filesystem goof: vop_panic[%s]", ap->a_desc->vdesc_name);
219 }
220 
221 /*
222  * vop_std<something> and vop_no<something> are default functions for use by
223  * filesystems that need the "default reasonable" implementation for a
224  * particular operation.
225  *
226  * The documentation for the operations they implement exists (if it exists)
227  * in the VOP_<SOMETHING>(9) manpage (all uppercase).
228  */
229 
230 /*
231  * Default vop for filesystems that do not support name lookup
232  */
233 static int
vop_nolookup(struct vop_lookup_args * ap)234 vop_nolookup(struct vop_lookup_args *ap)
235 {
236 
237 	*ap->a_vpp = NULL;
238 	return (ENOTDIR);
239 }
240 
241 /*
242  * vop_norename:
243  *
244  * Handle unlock and reference counting for arguments of vop_rename
245  * for filesystems that do not implement rename operation.
246  */
247 static int
vop_norename(struct vop_rename_args * ap)248 vop_norename(struct vop_rename_args *ap)
249 {
250 
251 	vop_rename_fail(ap);
252 	return (EOPNOTSUPP);
253 }
254 
255 /*
256  *	vop_nostrategy:
257  *
258  *	Strategy routine for VFS devices that have none.
259  *
260  *	BIO_ERROR and B_INVAL must be cleared prior to calling any strategy
261  *	routine.  Typically this is done for a BIO_READ strategy call.
262  *	Typically B_INVAL is assumed to already be clear prior to a write
263  *	and should not be cleared manually unless you just made the buffer
264  *	invalid.  BIO_ERROR should be cleared either way.
265  */
266 
267 static int
vop_nostrategy(struct vop_strategy_args * ap)268 vop_nostrategy (struct vop_strategy_args *ap)
269 {
270 	printf("No strategy for buffer at %p\n", ap->a_bp);
271 	vn_printf(ap->a_vp, "vnode ");
272 	ap->a_bp->b_ioflags |= BIO_ERROR;
273 	ap->a_bp->b_error = EOPNOTSUPP;
274 	bufdone(ap->a_bp);
275 	return (EOPNOTSUPP);
276 }
277 
278 /*
279  * Check if a named file exists in a given directory vnode
280  *
281  * Returns 0 if the file exists, ENOENT if it doesn't, or errors returned by
282  * vn_dir_next_dirent().
283  */
284 static int
dirent_exists(struct vnode * vp,const char * dirname,struct thread * td)285 dirent_exists(struct vnode *vp, const char *dirname, struct thread *td)
286 {
287 	char *dirbuf;
288 	int error, eofflag;
289 	size_t dirbuflen, len;
290 	off_t off;
291 	struct dirent *dp;
292 	struct vattr va;
293 
294 	ASSERT_VOP_LOCKED(vp, "vnode not locked");
295 	KASSERT(vp->v_type == VDIR, ("vp %p is not a directory", vp));
296 
297 	error = VOP_GETATTR(vp, &va, td->td_ucred);
298 	if (error != 0)
299 		return (error);
300 
301 	dirbuflen = MAX(DEV_BSIZE, GENERIC_MAXDIRSIZ);
302 	if (dirbuflen < va.va_blocksize)
303 		dirbuflen = va.va_blocksize;
304 	dirbuf = malloc(dirbuflen, M_TEMP, M_WAITOK);
305 
306 	len = 0;
307 	off = 0;
308 	eofflag = 0;
309 
310 	for (;;) {
311 		error = vn_dir_next_dirent(vp, td, dirbuf, dirbuflen,
312 		    &dp, &len, &off, &eofflag);
313 		if (error != 0)
314 			goto out;
315 
316 		if (len == 0)
317 			break;
318 
319 		if (dp->d_type != DT_WHT && dp->d_fileno != 0 &&
320 		    strcmp(dp->d_name, dirname) == 0)
321 			goto out;
322 	}
323 
324 	error = ENOENT;
325 
326 out:
327 	free(dirbuf, M_TEMP);
328 	return (error);
329 }
330 
331 int
vop_stdaccess(struct vop_access_args * ap)332 vop_stdaccess(struct vop_access_args *ap)
333 {
334 
335 	KASSERT((ap->a_accmode & ~(VEXEC | VWRITE | VREAD | VADMIN |
336 	    VAPPEND)) == 0, ("invalid bit in accmode"));
337 
338 	return (VOP_ACCESSX(ap->a_vp, ap->a_accmode, ap->a_cred, ap->a_td));
339 }
340 
341 int
vop_stdaccessx(struct vop_accessx_args * ap)342 vop_stdaccessx(struct vop_accessx_args *ap)
343 {
344 	int error;
345 	accmode_t accmode = ap->a_accmode;
346 
347 	error = vfs_unixify_accmode(&accmode);
348 	if (error != 0)
349 		return (error);
350 
351 	if (accmode == 0)
352 		return (0);
353 
354 	return (VOP_ACCESS(ap->a_vp, accmode, ap->a_cred, ap->a_td));
355 }
356 
357 /*
358  * Advisory record locking support
359  */
360 int
vop_stdadvlock(struct vop_advlock_args * ap)361 vop_stdadvlock(struct vop_advlock_args *ap)
362 {
363 	struct vnode *vp;
364 	struct mount *mp;
365 	struct vattr vattr;
366 	int error;
367 
368 	vp = ap->a_vp;
369 
370 	/*
371 	 * Provide atomicity of open(O_CREAT | O_EXCL | O_EXLOCK) for
372 	 * local filesystems.  See vn_open_cred() for reciprocal part.
373 	 */
374 	mp = vp->v_mount;
375 	if (mp != NULL && (mp->mnt_flag & MNT_LOCAL) != 0 &&
376 	    ap->a_op == F_SETLK && (ap->a_flags & F_FIRSTOPEN) == 0) {
377 		VI_LOCK(vp);
378 		while ((vp->v_iflag & VI_FOPENING) != 0)
379 			msleep(vp, VI_MTX(vp), PLOCK, "lockfo", 0);
380 		VI_UNLOCK(vp);
381 	}
382 
383 	if (ap->a_fl->l_whence == SEEK_END) {
384 		/*
385 		 * The NFSv4 server must avoid doing a vn_lock() here, since it
386 		 * can deadlock the nfsd threads, due to a LOR.  Fortunately
387 		 * the NFSv4 server always uses SEEK_SET and this code is
388 		 * only required for the SEEK_END case.
389 		 */
390 		vn_lock(vp, LK_SHARED | LK_RETRY);
391 		error = VOP_GETATTR(vp, &vattr, curthread->td_ucred);
392 		VOP_UNLOCK(vp);
393 		if (error)
394 			return (error);
395 	} else
396 		vattr.va_size = 0;
397 
398 	return (lf_advlock(ap, &(vp->v_lockf), vattr.va_size));
399 }
400 
401 int
vop_stdadvlockasync(struct vop_advlockasync_args * ap)402 vop_stdadvlockasync(struct vop_advlockasync_args *ap)
403 {
404 	struct vnode *vp;
405 	struct vattr vattr;
406 	int error;
407 
408 	vp = ap->a_vp;
409 	if (ap->a_fl->l_whence == SEEK_END) {
410 		/* The size argument is only needed for SEEK_END. */
411 		vn_lock(vp, LK_SHARED | LK_RETRY);
412 		error = VOP_GETATTR(vp, &vattr, curthread->td_ucred);
413 		VOP_UNLOCK(vp);
414 		if (error)
415 			return (error);
416 	} else
417 		vattr.va_size = 0;
418 
419 	return (lf_advlockasync(ap, &(vp->v_lockf), vattr.va_size));
420 }
421 
422 int
vop_stdadvlockpurge(struct vop_advlockpurge_args * ap)423 vop_stdadvlockpurge(struct vop_advlockpurge_args *ap)
424 {
425 	struct vnode *vp;
426 
427 	vp = ap->a_vp;
428 	lf_purgelocks(vp, &vp->v_lockf);
429 	return (0);
430 }
431 
432 /*
433  * vop_stdpathconf:
434  *
435  * Standard implementation of POSIX pathconf, to get information about limits
436  * for a filesystem.
437  * Override per filesystem for the case where the filesystem has smaller
438  * limits.
439  */
440 int
vop_stdpathconf(struct vop_pathconf_args * ap)441 vop_stdpathconf(struct vop_pathconf_args *ap)
442 {
443 
444 	switch (ap->a_name) {
445 		case _PC_ASYNC_IO:
446 			*ap->a_retval = _POSIX_ASYNCHRONOUS_IO;
447 			return (0);
448 		case _PC_PATH_MAX:
449 			*ap->a_retval = PATH_MAX;
450 			return (0);
451 		case _PC_ACL_EXTENDED:
452 		case _PC_ACL_NFS4:
453 		case _PC_CAP_PRESENT:
454 		case _PC_DEALLOC_PRESENT:
455 		case _PC_INF_PRESENT:
456 		case _PC_MAC_PRESENT:
457 		case _PC_NAMEDATTR_ENABLED:
458 		case _PC_HAS_NAMEDATTR:
459 			*ap->a_retval = 0;
460 			return (0);
461 		default:
462 			return (EINVAL);
463 	}
464 	/* NOTREACHED */
465 }
466 
467 /*
468  * Standard lock, unlock and islocked functions.
469  */
470 int
vop_stdlock(struct vop_lock1_args * ap)471 vop_stdlock(struct vop_lock1_args *ap)
472 {
473 	struct vnode *vp = ap->a_vp;
474 	struct mtx *ilk;
475 
476 	ilk = VI_MTX(vp);
477 	return (lockmgr_lock_flags(vp->v_vnlock, ap->a_flags,
478 	    &ilk->lock_object, ap->a_file, ap->a_line));
479 }
480 
481 /* See above. */
482 int
vop_stdunlock(struct vop_unlock_args * ap)483 vop_stdunlock(struct vop_unlock_args *ap)
484 {
485 	struct vnode *vp = ap->a_vp;
486 
487 	return (lockmgr_unlock(vp->v_vnlock));
488 }
489 
490 /* See above. */
491 int
vop_stdislocked(struct vop_islocked_args * ap)492 vop_stdislocked(struct vop_islocked_args *ap)
493 {
494 
495 	return (lockstatus(ap->a_vp->v_vnlock));
496 }
497 
498 /*
499  * Variants of the above set.
500  *
501  * Differences are:
502  * - shared locking disablement is not supported
503  * - v_vnlock pointer is not honored
504  */
505 int
vop_lock(struct vop_lock1_args * ap)506 vop_lock(struct vop_lock1_args *ap)
507 {
508 	struct vnode *vp = ap->a_vp;
509 	int flags = ap->a_flags;
510 	struct mtx *ilk;
511 
512 	MPASS(vp->v_vnlock == &vp->v_lock);
513 
514 	if (__predict_false((flags & ~(LK_TYPE_MASK | LK_NODDLKTREAT | LK_RETRY)) != 0))
515 		goto other;
516 
517 	switch (flags & LK_TYPE_MASK) {
518 	case LK_SHARED:
519 		return (lockmgr_slock(&vp->v_lock, flags, ap->a_file, ap->a_line));
520 	case LK_EXCLUSIVE:
521 		return (lockmgr_xlock(&vp->v_lock, flags, ap->a_file, ap->a_line));
522 	}
523 other:
524 	ilk = VI_MTX(vp);
525 	return (lockmgr_lock_flags(&vp->v_lock, flags,
526 	    &ilk->lock_object, ap->a_file, ap->a_line));
527 }
528 
529 int
vop_unlock(struct vop_unlock_args * ap)530 vop_unlock(struct vop_unlock_args *ap)
531 {
532 	struct vnode *vp = ap->a_vp;
533 
534 	MPASS(vp->v_vnlock == &vp->v_lock);
535 
536 	return (lockmgr_unlock(&vp->v_lock));
537 }
538 
539 int
vop_islocked(struct vop_islocked_args * ap)540 vop_islocked(struct vop_islocked_args *ap)
541 {
542 	struct vnode *vp = ap->a_vp;
543 
544 	MPASS(vp->v_vnlock == &vp->v_lock);
545 
546 	return (lockstatus(&vp->v_lock));
547 }
548 
549 /*
550  * Return true for select/poll.
551  */
552 int
vop_nopoll(struct vop_poll_args * ap)553 vop_nopoll(struct vop_poll_args *ap)
554 {
555 
556 	if (ap->a_events & ~POLLSTANDARD)
557 		return (POLLNVAL);
558 	return (ap->a_events & (POLLIN | POLLOUT | POLLRDNORM | POLLWRNORM));
559 }
560 
561 /*
562  * Implement poll for local filesystems that support it.
563  */
564 int
vop_stdpoll(struct vop_poll_args * ap)565 vop_stdpoll(struct vop_poll_args *ap)
566 {
567 	if (ap->a_events & ~POLLSTANDARD)
568 		return (vn_pollrecord(ap->a_vp, ap->a_td, ap->a_events));
569 	return (ap->a_events & (POLLIN | POLLOUT | POLLRDNORM | POLLWRNORM));
570 }
571 
572 /*
573  * Return our mount point, as we will take charge of the writes.
574  */
575 int
vop_stdgetwritemount(struct vop_getwritemount_args * ap)576 vop_stdgetwritemount(struct vop_getwritemount_args *ap)
577 {
578 	struct mount *mp;
579 	struct vnode *vp;
580 
581 	/*
582 	 * Note that having a reference does not prevent forced unmount from
583 	 * setting ->v_mount to NULL after the lock gets released. This is of
584 	 * no consequence for typical consumers (most notably vn_start_write)
585 	 * since in this case the vnode is VIRF_DOOMED. Unmount might have
586 	 * progressed far enough that its completion is only delayed by the
587 	 * reference obtained here. The consumer only needs to concern itself
588 	 * with releasing it.
589 	 */
590 	vp = ap->a_vp;
591 	mp = vfs_ref_from_vp(vp);
592 	*(ap->a_mpp) = mp;
593 	return (0);
594 }
595 
596 /*
597  * If the file system doesn't implement VOP_BMAP, then return sensible defaults:
598  * - Return the vnode's bufobj instead of any underlying device's bufobj
599  * - Calculate the physical block number as if there were equal size
600  *   consecutive blocks, but
601  * - Report no contiguous runs of blocks.
602  */
603 int
vop_stdbmap(struct vop_bmap_args * ap)604 vop_stdbmap(struct vop_bmap_args *ap)
605 {
606 
607 	if (ap->a_bop != NULL)
608 		*ap->a_bop = &ap->a_vp->v_bufobj;
609 	if (ap->a_bnp != NULL)
610 		*ap->a_bnp = ap->a_bn * btodb(ap->a_vp->v_mount->mnt_stat.f_iosize);
611 	if (ap->a_runp != NULL)
612 		*ap->a_runp = 0;
613 	if (ap->a_runb != NULL)
614 		*ap->a_runb = 0;
615 	return (0);
616 }
617 
618 int
vop_stdfsync(struct vop_fsync_args * ap)619 vop_stdfsync(struct vop_fsync_args *ap)
620 {
621 
622 	return (vn_fsync_buf(ap->a_vp, ap->a_waitfor));
623 }
624 
625 static int
vop_stdfdatasync(struct vop_fdatasync_args * ap)626 vop_stdfdatasync(struct vop_fdatasync_args *ap)
627 {
628 
629 	return (VOP_FSYNC(ap->a_vp, MNT_WAIT, ap->a_td));
630 }
631 
632 int
vop_stdfdatasync_buf(struct vop_fdatasync_args * ap)633 vop_stdfdatasync_buf(struct vop_fdatasync_args *ap)
634 {
635 
636 	return (vn_fsync_buf(ap->a_vp, MNT_WAIT));
637 }
638 
639 /* XXX Needs good comment and more info in the manpage (VOP_GETPAGES(9)). */
640 int
vop_stdgetpages(struct vop_getpages_args * ap)641 vop_stdgetpages(struct vop_getpages_args *ap)
642 {
643 
644 	return vnode_pager_generic_getpages(ap->a_vp, ap->a_m,
645 	    ap->a_count, ap->a_rbehind, ap->a_rahead, NULL, NULL);
646 }
647 
648 static int
vop_stdgetpages_async(struct vop_getpages_async_args * ap)649 vop_stdgetpages_async(struct vop_getpages_async_args *ap)
650 {
651 	int error;
652 
653 	error = VOP_GETPAGES(ap->a_vp, ap->a_m, ap->a_count, ap->a_rbehind,
654 	    ap->a_rahead);
655 	if (ap->a_iodone != NULL)
656 		ap->a_iodone(ap->a_arg, ap->a_m, ap->a_count, error);
657 	return (error);
658 }
659 
660 int
vop_stdkqfilter(struct vop_kqfilter_args * ap)661 vop_stdkqfilter(struct vop_kqfilter_args *ap)
662 {
663 	return vfs_kqfilter(ap);
664 }
665 
666 /* XXX Needs good comment and more info in the manpage (VOP_PUTPAGES(9)). */
667 int
vop_stdputpages(struct vop_putpages_args * ap)668 vop_stdputpages(struct vop_putpages_args *ap)
669 {
670 
671 	return vnode_pager_generic_putpages(ap->a_vp, ap->a_m, ap->a_count,
672 	     ap->a_sync, ap->a_rtvals);
673 }
674 
675 int
vop_stdvptofh(struct vop_vptofh_args * ap)676 vop_stdvptofh(struct vop_vptofh_args *ap)
677 {
678 	return (EOPNOTSUPP);
679 }
680 
681 int
vop_stdvptocnp(struct vop_vptocnp_args * ap)682 vop_stdvptocnp(struct vop_vptocnp_args *ap)
683 {
684 	struct vnode *const vp = ap->a_vp;
685 	struct vnode **const dvp = ap->a_vpp;
686 	char *buf = ap->a_buf;
687 	size_t *buflen = ap->a_buflen;
688 	char *dirbuf;
689 	int i = *buflen;
690 	int error = 0, covered = 0;
691 	int eofflag, flags, locked;
692 	size_t dirbuflen, len;
693 	off_t off;
694 	ino_t fileno;
695 	struct vattr va;
696 	struct nameidata nd;
697 	struct thread *const td = curthread;
698 	struct ucred *const cred = td->td_ucred;
699 	struct dirent *dp;
700 	struct vnode *mvp;
701 
702 	if (vp->v_type != VDIR)
703 		return (ENOENT);
704 
705 	error = VOP_GETATTR(vp, &va, cred);
706 	if (error)
707 		return (error);
708 
709 	VREF(vp);
710 	locked = VOP_ISLOCKED(vp);
711 	VOP_UNLOCK(vp);
712 	NDINIT_ATVP(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF, UIO_SYSSPACE,
713 	    "..", vp);
714 	flags = FREAD;
715 	error = vn_open_cred(&nd, &flags, 0, VN_OPEN_NOAUDIT, cred, NULL);
716 	if (error) {
717 		vn_lock(vp, locked | LK_RETRY);
718 		return (error);
719 	}
720 	NDFREE_PNBUF(&nd);
721 
722 	mvp = *dvp = nd.ni_vp;
723 
724 	if (vp->v_mount != (*dvp)->v_mount &&
725 	    ((*dvp)->v_vflag & VV_ROOT) &&
726 	    ((*dvp)->v_mount->mnt_flag & MNT_UNION)) {
727 		*dvp = (*dvp)->v_mount->mnt_vnodecovered;
728 		VREF(mvp);
729 		VOP_UNLOCK(mvp);
730 		vn_close(mvp, FREAD, cred, td);
731 		VREF(*dvp);
732 		vn_lock(*dvp, LK_SHARED | LK_RETRY);
733 		covered = 1;
734 	}
735 
736 	fileno = va.va_fileid;
737 
738 	dirbuflen = MAX(DEV_BSIZE, GENERIC_MAXDIRSIZ);
739 	if (dirbuflen < va.va_blocksize)
740 		dirbuflen = va.va_blocksize;
741 	dirbuf = malloc(dirbuflen, M_TEMP, M_WAITOK);
742 
743 	if ((*dvp)->v_type != VDIR) {
744 		error = ENOENT;
745 		goto out;
746 	}
747 
748 	len = 0;
749 	off = 0;
750 	eofflag = 0;
751 
752 	for (;;) {
753 		/* call VOP_READDIR of parent */
754 		error = vn_dir_next_dirent(*dvp, td,
755 		    dirbuf, dirbuflen, &dp, &len, &off, &eofflag);
756 		if (error != 0)
757 			goto out;
758 
759 		if (len == 0) {
760 			error = ENOENT;
761 			goto out;
762 		}
763 
764 		if ((dp->d_type != DT_WHT) &&
765 		    (dp->d_fileno == fileno)) {
766 			if (covered) {
767 				VOP_UNLOCK(*dvp);
768 				vn_lock(mvp, LK_SHARED | LK_RETRY);
769 				if (dirent_exists(mvp, dp->d_name, td) == 0) {
770 					error = ENOENT;
771 					VOP_UNLOCK(mvp);
772 					vn_lock(*dvp, LK_SHARED | LK_RETRY);
773 					goto out;
774 				}
775 				VOP_UNLOCK(mvp);
776 				vn_lock(*dvp, LK_SHARED | LK_RETRY);
777 			}
778 			i -= dp->d_namlen;
779 
780 			if (i < 0) {
781 				error = ENOMEM;
782 				goto out;
783 			}
784 			if (dp->d_namlen == 1 && dp->d_name[0] == '.') {
785 				error = ENOENT;
786 			} else {
787 				bcopy(dp->d_name, buf + i, dp->d_namlen);
788 				error = 0;
789 			}
790 			goto out;
791 		}
792 	}
793 
794 out:
795 	free(dirbuf, M_TEMP);
796 	if (!error) {
797 		*buflen = i;
798 		vref(*dvp);
799 	}
800 	if (covered) {
801 		vput(*dvp);
802 		vrele(mvp);
803 	} else {
804 		VOP_UNLOCK(mvp);
805 		vn_close(mvp, FREAD, cred, td);
806 	}
807 	vn_lock(vp, locked | LK_RETRY);
808 	return (error);
809 }
810 
811 int
vop_stdallocate(struct vop_allocate_args * ap)812 vop_stdallocate(struct vop_allocate_args *ap)
813 {
814 #ifdef __notyet__
815 	struct statfs *sfs;
816 	off_t maxfilesize = 0;
817 #endif
818 	struct iovec aiov;
819 	struct vattr vattr, *vap;
820 	struct uio auio;
821 	off_t fsize, len, cur, offset;
822 	uint8_t *buf;
823 	struct thread *td;
824 	struct vnode *vp;
825 	size_t iosize;
826 	int error;
827 
828 	buf = NULL;
829 	error = 0;
830 	td = curthread;
831 	vap = &vattr;
832 	vp = ap->a_vp;
833 	len = *ap->a_len;
834 	offset = *ap->a_offset;
835 
836 	error = VOP_GETATTR(vp, vap, ap->a_cred);
837 	if (error != 0)
838 		goto out;
839 	fsize = vap->va_size;
840 	iosize = vap->va_blocksize;
841 	if (iosize == 0)
842 		iosize = BLKDEV_IOSIZE;
843 	if (iosize > maxphys)
844 		iosize = maxphys;
845 	buf = malloc(iosize, M_TEMP, M_WAITOK);
846 
847 #ifdef __notyet__
848 	/*
849 	 * Check if the filesystem sets f_maxfilesize; if not use
850 	 * VOP_SETATTR to perform the check.
851 	 */
852 	sfs = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
853 	error = VFS_STATFS(vp->v_mount, sfs, td);
854 	if (error == 0)
855 		maxfilesize = sfs->f_maxfilesize;
856 	free(sfs, M_STATFS);
857 	if (error != 0)
858 		goto out;
859 	if (maxfilesize) {
860 		if (offset > maxfilesize || len > maxfilesize ||
861 		    offset + len > maxfilesize) {
862 			error = EFBIG;
863 			goto out;
864 		}
865 	} else
866 #endif
867 	if (offset + len > vap->va_size) {
868 		/*
869 		 * Test offset + len against the filesystem's maxfilesize.
870 		 */
871 		VATTR_NULL(vap);
872 		vap->va_size = offset + len;
873 		error = VOP_SETATTR(vp, vap, ap->a_cred);
874 		if (error != 0)
875 			goto out;
876 		VATTR_NULL(vap);
877 		vap->va_size = fsize;
878 		error = VOP_SETATTR(vp, vap, ap->a_cred);
879 		if (error != 0)
880 			goto out;
881 	}
882 
883 	for (;;) {
884 		/*
885 		 * Read and write back anything below the nominal file
886 		 * size.  There's currently no way outside the filesystem
887 		 * to know whether this area is sparse or not.
888 		 */
889 		cur = iosize;
890 		if ((offset % iosize) != 0)
891 			cur -= (offset % iosize);
892 		if (cur > len)
893 			cur = len;
894 		if (offset < fsize) {
895 			aiov.iov_base = buf;
896 			aiov.iov_len = cur;
897 			auio.uio_iov = &aiov;
898 			auio.uio_iovcnt = 1;
899 			auio.uio_offset = offset;
900 			auio.uio_resid = cur;
901 			auio.uio_segflg = UIO_SYSSPACE;
902 			auio.uio_rw = UIO_READ;
903 			auio.uio_td = td;
904 			error = VOP_READ(vp, &auio, ap->a_ioflag, ap->a_cred);
905 			if (error != 0)
906 				break;
907 			if (auio.uio_resid > 0) {
908 				bzero(buf + cur - auio.uio_resid,
909 				    auio.uio_resid);
910 			}
911 		} else {
912 			bzero(buf, cur);
913 		}
914 
915 		aiov.iov_base = buf;
916 		aiov.iov_len = cur;
917 		auio.uio_iov = &aiov;
918 		auio.uio_iovcnt = 1;
919 		auio.uio_offset = offset;
920 		auio.uio_resid = cur;
921 		auio.uio_segflg = UIO_SYSSPACE;
922 		auio.uio_rw = UIO_WRITE;
923 		auio.uio_td = td;
924 
925 		error = VOP_WRITE(vp, &auio, ap->a_ioflag, ap->a_cred);
926 		if (error != 0)
927 			break;
928 
929 		len -= cur;
930 		offset += cur;
931 		if (len == 0)
932 			break;
933 		if (should_yield())
934 			break;
935 	}
936 
937  out:
938 	*ap->a_len = len;
939 	*ap->a_offset = offset;
940 	free(buf, M_TEMP);
941 	return (error);
942 }
943 
944 static int
vp_zerofill(struct vnode * vp,struct vattr * vap,off_t * offsetp,off_t * lenp,int ioflag,struct ucred * cred)945 vp_zerofill(struct vnode *vp, struct vattr *vap, off_t *offsetp, off_t *lenp,
946     int ioflag, struct ucred *cred)
947 {
948 	int iosize;
949 	int error = 0;
950 	struct iovec aiov;
951 	struct uio auio;
952 	struct thread *td;
953 	off_t offset, len;
954 
955 	iosize = vap->va_blocksize;
956 	td = curthread;
957 	offset = *offsetp;
958 	len = *lenp;
959 
960 	if (iosize == 0)
961 		iosize = BLKDEV_IOSIZE;
962 	/* If va_blocksize is 512 bytes, iosize will be 4 kilobytes */
963 	iosize = min(iosize * 8, ZERO_REGION_SIZE);
964 
965 	while (len > 0) {
966 		int xfersize = iosize;
967 		if (offset % iosize != 0)
968 			xfersize -= offset % iosize;
969 		if (xfersize > len)
970 			xfersize = len;
971 
972 		aiov.iov_base = __DECONST(void *, zero_region);
973 		aiov.iov_len = xfersize;
974 		auio.uio_iov = &aiov;
975 		auio.uio_iovcnt = 1;
976 		auio.uio_offset = offset;
977 		auio.uio_resid = xfersize;
978 		auio.uio_segflg = UIO_SYSSPACE;
979 		auio.uio_rw = UIO_WRITE;
980 		auio.uio_td = td;
981 
982 		error = VOP_WRITE(vp, &auio, ioflag, cred);
983 		if (error != 0) {
984 			len -= xfersize - auio.uio_resid;
985 			offset += xfersize - auio.uio_resid;
986 			break;
987 		}
988 
989 		len -= xfersize;
990 		offset += xfersize;
991 	}
992 
993 	*offsetp = offset;
994 	*lenp = len;
995 	return (error);
996 }
997 
998 int
vop_stddeallocate(struct vop_deallocate_args * ap)999 vop_stddeallocate(struct vop_deallocate_args *ap)
1000 {
1001 	struct vnode *vp;
1002 	off_t offset, len;
1003 	struct ucred *cred;
1004 	int error;
1005 	struct vattr va;
1006 	off_t noff, xfersize, rem;
1007 
1008 	vp = ap->a_vp;
1009 	offset = *ap->a_offset;
1010 	cred = ap->a_cred;
1011 
1012 	error = VOP_GETATTR(vp, &va, cred);
1013 	if (error)
1014 		return (error);
1015 
1016 	len = omin((off_t)va.va_size - offset, *ap->a_len);
1017 	while (len > 0) {
1018 		noff = offset;
1019 		error = vn_bmap_seekhole_locked(vp, FIOSEEKDATA, &noff, cred);
1020 		if (error) {
1021 			if (error != ENXIO)
1022 				/* XXX: Is it okay to fallback further? */
1023 				goto out;
1024 
1025 			/*
1026 			 * No more data region to be filled
1027 			 */
1028 			offset += len;
1029 			len = 0;
1030 			error = 0;
1031 			break;
1032 		}
1033 		KASSERT(noff >= offset, ("FIOSEEKDATA going backward"));
1034 		if (noff != offset) {
1035 			xfersize = omin(noff - offset, len);
1036 			len -= xfersize;
1037 			offset += xfersize;
1038 			if (len == 0)
1039 				break;
1040 		}
1041 		error = vn_bmap_seekhole_locked(vp, FIOSEEKHOLE, &noff, cred);
1042 		if (error)
1043 			goto out;
1044 
1045 		/* Fill zeroes */
1046 		xfersize = rem = omin(noff - offset, len);
1047 		error = vp_zerofill(vp, &va, &offset, &rem, ap->a_ioflag, cred);
1048 		if (error) {
1049 			len -= xfersize - rem;
1050 			goto out;
1051 		}
1052 
1053 		len -= xfersize;
1054 		if (should_yield())
1055 			break;
1056 	}
1057 	/* Handle the case when offset is beyond EOF */
1058 	if (len < 0)
1059 		len = 0;
1060 out:
1061 	*ap->a_offset = offset;
1062 	*ap->a_len = len;
1063 	return (error);
1064 }
1065 
1066 int
vop_stdadvise(struct vop_advise_args * ap)1067 vop_stdadvise(struct vop_advise_args *ap)
1068 {
1069 	struct vnode *vp;
1070 	struct bufobj *bo;
1071 	uintmax_t bstart, bend;
1072 	daddr_t startn, endn;
1073 	int bsize, error;
1074 
1075 	vp = ap->a_vp;
1076 	switch (ap->a_advice) {
1077 	case POSIX_FADV_WILLNEED:
1078 		/*
1079 		 * Do nothing for now.  Filesystems should provide a
1080 		 * custom method which starts an asynchronous read of
1081 		 * the requested region.
1082 		 */
1083 		error = 0;
1084 		break;
1085 	case POSIX_FADV_DONTNEED:
1086 		error = 0;
1087 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1088 		if (VN_IS_DOOMED(vp)) {
1089 			VOP_UNLOCK(vp);
1090 			break;
1091 		}
1092 
1093 		/*
1094 		 * Round to block boundaries (and later possibly further to
1095 		 * page boundaries).  Applications cannot reasonably be aware
1096 		 * of the boundaries, and the rounding must be to expand at
1097 		 * both extremities to cover enough.  It still doesn't cover
1098 		 * read-ahead.  For partial blocks, this gives unnecessary
1099 		 * discarding of buffers but is efficient enough since the
1100 		 * pages usually remain in VMIO for some time.
1101 		 */
1102 		bsize = vp->v_bufobj.bo_bsize;
1103 		bstart = rounddown(ap->a_start, bsize);
1104 		bend = ap->a_end;
1105 		bend = roundup(bend, bsize);
1106 
1107 		/*
1108 		 * Deactivate pages in the specified range from the backing VM
1109 		 * object.  Pages that are resident in the buffer cache will
1110 		 * remain wired until their corresponding buffers are released
1111 		 * below.
1112 		 */
1113 		if (vp->v_object != NULL) {
1114 			VM_OBJECT_RLOCK(vp->v_object);
1115 			vm_object_page_noreuse(vp->v_object,
1116 			    OFF_TO_IDX(trunc_page(bstart)),
1117 			    OFF_TO_IDX(round_page(bend)));
1118 			VM_OBJECT_RUNLOCK(vp->v_object);
1119 		}
1120 
1121 		bo = &vp->v_bufobj;
1122 		startn = bstart / bsize;
1123 		endn = bend / bsize;
1124 		BO_RLOCK(bo);
1125 		error = bnoreuselist(&bo->bo_clean, bo, startn, endn);
1126 		if (error == 0)
1127 			error = bnoreuselist(&bo->bo_dirty, bo, startn, endn);
1128 		BO_RUNLOCK(bo);
1129 		VOP_UNLOCK(vp);
1130 		break;
1131 	default:
1132 		error = EINVAL;
1133 		break;
1134 	}
1135 	return (error);
1136 }
1137 
1138 int
vop_stdunp_bind(struct vop_unp_bind_args * ap)1139 vop_stdunp_bind(struct vop_unp_bind_args *ap)
1140 {
1141 
1142 	ap->a_vp->v_unpcb = ap->a_unpcb;
1143 	return (0);
1144 }
1145 
1146 int
vop_stdunp_connect(struct vop_unp_connect_args * ap)1147 vop_stdunp_connect(struct vop_unp_connect_args *ap)
1148 {
1149 
1150 	*ap->a_unpcb = ap->a_vp->v_unpcb;
1151 	return (0);
1152 }
1153 
1154 int
vop_stdunp_detach(struct vop_unp_detach_args * ap)1155 vop_stdunp_detach(struct vop_unp_detach_args *ap)
1156 {
1157 
1158 	ap->a_vp->v_unpcb = NULL;
1159 	return (0);
1160 }
1161 
1162 static int
vop_stdis_text(struct vop_is_text_args * ap)1163 vop_stdis_text(struct vop_is_text_args *ap)
1164 {
1165 
1166 	return ((int)atomic_load_int(&ap->a_vp->v_writecount) < 0);
1167 }
1168 
1169 int
vop_stdset_text(struct vop_set_text_args * ap)1170 vop_stdset_text(struct vop_set_text_args *ap)
1171 {
1172 	struct vnode *vp;
1173 	int n;
1174 	bool gotref;
1175 
1176 	vp = ap->a_vp;
1177 
1178 	n = atomic_load_int(&vp->v_writecount);
1179 	for (;;) {
1180 		if (__predict_false(n > 0)) {
1181 			return (ETXTBSY);
1182 		}
1183 
1184 		/*
1185 		 * Transition point, we may need to grab a reference on the vnode.
1186 		 *
1187 		 * Take the ref early As a safety measure against bogus calls
1188 		 * to vop_stdunset_text.
1189 		 */
1190 		if (n == 0) {
1191 			gotref = false;
1192 			if ((vn_irflag_read(vp) & VIRF_TEXT_REF) != 0) {
1193 				vref(vp);
1194 				gotref = true;
1195 			}
1196 			if (atomic_fcmpset_int(&vp->v_writecount, &n, -1)) {
1197 				return (0);
1198 			}
1199 			if (gotref) {
1200 				vunref(vp);
1201 			}
1202 			continue;
1203 		}
1204 
1205 		MPASS(n < 0);
1206 		if (atomic_fcmpset_int(&vp->v_writecount, &n, n - 1)) {
1207 			return (0);
1208 		}
1209 	}
1210 	__assert_unreachable();
1211 }
1212 
1213 static int
vop_stdunset_text(struct vop_unset_text_args * ap)1214 vop_stdunset_text(struct vop_unset_text_args *ap)
1215 {
1216 	struct vnode *vp;
1217 	int n;
1218 
1219 	vp = ap->a_vp;
1220 
1221 	n = atomic_load_int(&vp->v_writecount);
1222 	for (;;) {
1223 		if (__predict_false(n >= 0)) {
1224 			return (EINVAL);
1225 		}
1226 
1227 		/*
1228 		 * Transition point, we may need to release a reference on the vnode.
1229 		 */
1230 		if (n == -1) {
1231 			if (atomic_fcmpset_int(&vp->v_writecount, &n, 0)) {
1232 				if ((vn_irflag_read(vp) & VIRF_TEXT_REF) != 0) {
1233 					vunref(vp);
1234 				}
1235 				return (0);
1236 			}
1237 			continue;
1238 		}
1239 
1240 		MPASS(n < -1);
1241 		if (atomic_fcmpset_int(&vp->v_writecount, &n, n + 1)) {
1242 			return (0);
1243 		}
1244 	}
1245 	__assert_unreachable();
1246 }
1247 
1248 static __always_inline int
vop_stdadd_writecount_impl(struct vop_add_writecount_args * ap,bool handle_msync)1249 vop_stdadd_writecount_impl(struct vop_add_writecount_args *ap, bool handle_msync)
1250 {
1251 	struct vnode *vp;
1252 	struct mount *mp __diagused;
1253 	int n;
1254 
1255 	vp = ap->a_vp;
1256 
1257 #ifdef INVARIANTS
1258 	mp = vp->v_mount;
1259 	if (mp != NULL) {
1260 		if (handle_msync) {
1261 			VNPASS((mp->mnt_kern_flag & MNTK_NOMSYNC) == 0, vp);
1262 		} else {
1263 			VNPASS((mp->mnt_kern_flag & MNTK_NOMSYNC) != 0, vp);
1264 		}
1265 	}
1266 #endif
1267 
1268 	n = atomic_load_int(&vp->v_writecount);
1269 	for (;;) {
1270 		if (__predict_false(n < 0)) {
1271 			return (ETXTBSY);
1272 		}
1273 
1274 		VNASSERT(n + ap->a_inc >= 0, vp,
1275 		    ("neg writecount increment %d + %d = %d", n, ap->a_inc,
1276 		    n + ap->a_inc));
1277 		if (n == 0) {
1278 			if (handle_msync) {
1279 				vlazy(vp);
1280 			}
1281 		}
1282 
1283 		if (atomic_fcmpset_int(&vp->v_writecount, &n, n + ap->a_inc)) {
1284 			return (0);
1285 		}
1286 	}
1287 	__assert_unreachable();
1288 }
1289 
1290 int
vop_stdadd_writecount(struct vop_add_writecount_args * ap)1291 vop_stdadd_writecount(struct vop_add_writecount_args *ap)
1292 {
1293 
1294 	return (vop_stdadd_writecount_impl(ap, true));
1295 }
1296 
1297 int
vop_stdadd_writecount_nomsync(struct vop_add_writecount_args * ap)1298 vop_stdadd_writecount_nomsync(struct vop_add_writecount_args *ap)
1299 {
1300 
1301 	return (vop_stdadd_writecount_impl(ap, false));
1302 }
1303 
1304 int
vop_stdneed_inactive(struct vop_need_inactive_args * ap)1305 vop_stdneed_inactive(struct vop_need_inactive_args *ap)
1306 {
1307 
1308 	return (1);
1309 }
1310 
1311 int
vop_stdinotify(struct vop_inotify_args * ap)1312 vop_stdinotify(struct vop_inotify_args *ap)
1313 {
1314 	vn_inotify(ap->a_vp, ap->a_dvp, ap->a_cnp, ap->a_event, ap->a_cookie);
1315 	return (0);
1316 }
1317 
1318 int
vop_stdinotify_add_watch(struct vop_inotify_add_watch_args * ap)1319 vop_stdinotify_add_watch(struct vop_inotify_add_watch_args *ap)
1320 {
1321 	return (vn_inotify_add_watch(ap->a_vp, ap->a_sc, ap->a_mask,
1322 	    ap->a_wdp, ap->a_td));
1323 }
1324 
1325 int
vop_stdioctl(struct vop_ioctl_args * ap)1326 vop_stdioctl(struct vop_ioctl_args *ap)
1327 {
1328 	struct vnode *vp;
1329 	struct vattr va;
1330 	off_t *offp;
1331 	int error;
1332 
1333 	switch (ap->a_command) {
1334 	case FIOSEEKDATA:
1335 	case FIOSEEKHOLE:
1336 		vp = ap->a_vp;
1337 		error = vn_lock(vp, LK_SHARED);
1338 		if (error != 0)
1339 			return (EBADF);
1340 		if (vp->v_type == VREG)
1341 			error = VOP_GETATTR(vp, &va, ap->a_cred);
1342 		else
1343 			error = ENOTTY;
1344 		if (error == 0) {
1345 			offp = ap->a_data;
1346 			if (*offp < 0 || *offp >= va.va_size)
1347 				error = ENXIO;
1348 			else if (ap->a_command == FIOSEEKHOLE)
1349 				*offp = va.va_size;
1350 		}
1351 		VOP_UNLOCK(vp);
1352 		break;
1353 	default:
1354 		error = ENOTTY;
1355 		break;
1356 	}
1357 	return (error);
1358 }
1359 
1360 /*
1361  * vfs default ops
1362  * used to fill the vfs function table to get reasonable default return values.
1363  */
1364 int
vfs_stdroot(struct mount * mp,int flags,struct vnode ** vpp)1365 vfs_stdroot(struct mount *mp, int flags, struct vnode **vpp)
1366 {
1367 
1368 	return (EOPNOTSUPP);
1369 }
1370 
1371 int
vfs_stdstatfs(struct mount * mp,struct statfs * sbp)1372 vfs_stdstatfs(struct mount *mp, struct statfs *sbp)
1373 {
1374 
1375 	return (EOPNOTSUPP);
1376 }
1377 
1378 int
vfs_stdquotactl(struct mount * mp,int cmds,uid_t uid,void * arg,bool * mp_busy)1379 vfs_stdquotactl(struct mount *mp, int cmds, uid_t uid, void *arg, bool *mp_busy)
1380 {
1381 	return (EOPNOTSUPP);
1382 }
1383 
1384 int
vfs_stdsync(struct mount * mp,int waitfor)1385 vfs_stdsync(struct mount *mp, int waitfor)
1386 {
1387 	struct vnode *vp, *mvp;
1388 	struct thread *td;
1389 	int error, lockreq, allerror = 0;
1390 
1391 	td = curthread;
1392 	lockreq = LK_EXCLUSIVE | LK_INTERLOCK;
1393 	if (waitfor != MNT_WAIT)
1394 		lockreq |= LK_NOWAIT;
1395 	/*
1396 	 * Force stale buffer cache information to be flushed.
1397 	 */
1398 loop:
1399 	MNT_VNODE_FOREACH_ALL(vp, mp, mvp) {
1400 		if (vp->v_bufobj.bo_dirty.bv_cnt == 0) {
1401 			VI_UNLOCK(vp);
1402 			continue;
1403 		}
1404 		if ((error = vget(vp, lockreq)) != 0) {
1405 			if (error == ENOENT) {
1406 				MNT_VNODE_FOREACH_ALL_ABORT(mp, mvp);
1407 				goto loop;
1408 			}
1409 			continue;
1410 		}
1411 		error = VOP_FSYNC(vp, waitfor, td);
1412 		if (error)
1413 			allerror = error;
1414 		vput(vp);
1415 	}
1416 	return (allerror);
1417 }
1418 
1419 int
vfs_stdnosync(struct mount * mp,int waitfor)1420 vfs_stdnosync(struct mount *mp, int waitfor)
1421 {
1422 
1423 	return (0);
1424 }
1425 
1426 static int
vop_stdcopy_file_range(struct vop_copy_file_range_args * ap)1427 vop_stdcopy_file_range(struct vop_copy_file_range_args *ap)
1428 {
1429 	int error;
1430 
1431 	error = vn_generic_copy_file_range(ap->a_invp, ap->a_inoffp,
1432 	    ap->a_outvp, ap->a_outoffp, ap->a_lenp, ap->a_flags, ap->a_incred,
1433 	    ap->a_outcred, ap->a_fsizetd);
1434 	return (error);
1435 }
1436 
1437 int
vfs_stdvget(struct mount * mp,ino_t ino,int flags,struct vnode ** vpp)1438 vfs_stdvget(struct mount *mp, ino_t ino, int flags, struct vnode **vpp)
1439 {
1440 
1441 	return (EOPNOTSUPP);
1442 }
1443 
1444 int
vfs_stdfhtovp(struct mount * mp,struct fid * fhp,int flags,struct vnode ** vpp)1445 vfs_stdfhtovp(struct mount *mp, struct fid *fhp, int flags, struct vnode **vpp)
1446 {
1447 
1448 	return (EOPNOTSUPP);
1449 }
1450 
1451 int
vfs_stdinit(struct vfsconf * vfsp)1452 vfs_stdinit(struct vfsconf *vfsp)
1453 {
1454 
1455 	return (0);
1456 }
1457 
1458 int
vfs_stduninit(struct vfsconf * vfsp)1459 vfs_stduninit(struct vfsconf *vfsp)
1460 {
1461 
1462 	return(0);
1463 }
1464 
1465 int
vfs_stdextattrctl(struct mount * mp,int cmd,struct vnode * filename_vp,int attrnamespace,const char * attrname)1466 vfs_stdextattrctl(struct mount *mp, int cmd, struct vnode *filename_vp,
1467     int attrnamespace, const char *attrname)
1468 {
1469 
1470 	if (filename_vp != NULL)
1471 		VOP_UNLOCK(filename_vp);
1472 	return (EOPNOTSUPP);
1473 }
1474 
1475 int
vfs_stdsysctl(struct mount * mp,fsctlop_t op,struct sysctl_req * req)1476 vfs_stdsysctl(struct mount *mp, fsctlop_t op, struct sysctl_req *req)
1477 {
1478 
1479 	return (EOPNOTSUPP);
1480 }
1481 
1482 static vop_bypass_t *
bp_by_off(struct vop_vector * vop,struct vop_generic_args * a)1483 bp_by_off(struct vop_vector *vop, struct vop_generic_args *a)
1484 {
1485 
1486 	return (*(vop_bypass_t **)((char *)vop + a->a_desc->vdesc_vop_offset));
1487 }
1488 
1489 int
vop_sigdefer(struct vop_vector * vop,struct vop_generic_args * a)1490 vop_sigdefer(struct vop_vector *vop, struct vop_generic_args *a)
1491 {
1492 	vop_bypass_t *bp;
1493 	int prev_stops, rc;
1494 
1495 	bp = bp_by_off(vop, a);
1496 	MPASS(bp != NULL);
1497 
1498 	prev_stops = sigdeferstop(SIGDEFERSTOP_SILENT);
1499 	rc = bp(a);
1500 	sigallowstop(prev_stops);
1501 	return (rc);
1502 }
1503 
1504 static int
vop_stdstat(struct vop_stat_args * a)1505 vop_stdstat(struct vop_stat_args *a)
1506 {
1507 	struct vattr vattr;
1508 	struct vattr *vap;
1509 	struct vnode *vp;
1510 	struct stat *sb;
1511 	int error;
1512 	u_short mode;
1513 
1514 	vp = a->a_vp;
1515 	sb = a->a_sb;
1516 
1517 	error = vop_stat_helper_pre(a);
1518 	if (error != 0)
1519 		return (error);
1520 
1521 	vap = &vattr;
1522 
1523 	/*
1524 	 * Initialize defaults for new and unusual fields, so that file
1525 	 * systems which don't support these fields don't need to know
1526 	 * about them.
1527 	 */
1528 	vap->va_birthtime.tv_sec = -1;
1529 	vap->va_birthtime.tv_nsec = 0;
1530 	vap->va_fsid = VNOVAL;
1531 	vap->va_gen = 0;
1532 	vap->va_rdev = NODEV;
1533 	vap->va_filerev = 0;
1534 	vap->va_bsdflags = 0;
1535 
1536 	error = VOP_GETATTR(vp, vap, a->a_active_cred);
1537 	if (error)
1538 		goto out;
1539 
1540 	/*
1541 	 * Zero the spare stat fields
1542 	 */
1543 	bzero(sb, sizeof *sb);
1544 
1545 	/*
1546 	 * Copy from vattr table
1547 	 */
1548 	if (vap->va_fsid != VNOVAL)
1549 		sb->st_dev = vap->va_fsid;
1550 	else
1551 		sb->st_dev = vp->v_mount->mnt_stat.f_fsid.val[0];
1552 	sb->st_ino = vap->va_fileid;
1553 	mode = vap->va_mode;
1554 	switch (vap->va_type) {
1555 	case VREG:
1556 		mode |= S_IFREG;
1557 		break;
1558 	case VDIR:
1559 		mode |= S_IFDIR;
1560 		break;
1561 	case VBLK:
1562 		mode |= S_IFBLK;
1563 		break;
1564 	case VCHR:
1565 		mode |= S_IFCHR;
1566 		break;
1567 	case VLNK:
1568 		mode |= S_IFLNK;
1569 		break;
1570 	case VSOCK:
1571 		mode |= S_IFSOCK;
1572 		break;
1573 	case VFIFO:
1574 		mode |= S_IFIFO;
1575 		break;
1576 	default:
1577 		error = EBADF;
1578 		goto out;
1579 	}
1580 	sb->st_mode = mode;
1581 	sb->st_nlink = vap->va_nlink;
1582 	sb->st_uid = vap->va_uid;
1583 	sb->st_gid = vap->va_gid;
1584 	sb->st_rdev = vap->va_rdev;
1585 	if (vap->va_size > OFF_MAX) {
1586 		error = EOVERFLOW;
1587 		goto out;
1588 	}
1589 	sb->st_size = vap->va_size;
1590 	sb->st_atim.tv_sec = vap->va_atime.tv_sec;
1591 	sb->st_atim.tv_nsec = vap->va_atime.tv_nsec;
1592 	sb->st_mtim.tv_sec = vap->va_mtime.tv_sec;
1593 	sb->st_mtim.tv_nsec = vap->va_mtime.tv_nsec;
1594 	sb->st_ctim.tv_sec = vap->va_ctime.tv_sec;
1595 	sb->st_ctim.tv_nsec = vap->va_ctime.tv_nsec;
1596 	sb->st_birthtim.tv_sec = vap->va_birthtime.tv_sec;
1597 	sb->st_birthtim.tv_nsec = vap->va_birthtime.tv_nsec;
1598 
1599 	/*
1600 	 * According to www.opengroup.org, the meaning of st_blksize is
1601 	 *   "a filesystem-specific preferred I/O block size for this
1602 	 *    object.  In some filesystem types, this may vary from file
1603 	 *    to file"
1604 	 * Use minimum/default of PAGE_SIZE (e.g. for VCHR).
1605 	 */
1606 
1607 	sb->st_blksize = max(PAGE_SIZE, vap->va_blocksize);
1608 	sb->st_flags = vap->va_flags;
1609 	sb->st_blocks = vap->va_bytes / S_BLKSIZE;
1610 	sb->st_gen = vap->va_gen;
1611 	sb->st_filerev = vap->va_filerev;
1612 	sb->st_bsdflags = vap->va_bsdflags;
1613 out:
1614 	return (vop_stat_helper_post(a, error));
1615 }
1616 
1617 static int
vop_stdread_pgcache(struct vop_read_pgcache_args * ap __unused)1618 vop_stdread_pgcache(struct vop_read_pgcache_args *ap __unused)
1619 {
1620 	return (EJUSTRETURN);
1621 }
1622 
1623 static int
vop_stdvput_pair(struct vop_vput_pair_args * ap)1624 vop_stdvput_pair(struct vop_vput_pair_args *ap)
1625 {
1626 	struct vnode *dvp, *vp, **vpp;
1627 
1628 	dvp = ap->a_dvp;
1629 	vpp = ap->a_vpp;
1630 	vput(dvp);
1631 	if (vpp != NULL && ap->a_unlock_vp && (vp = *vpp) != NULL)
1632 		vput(vp);
1633 	return (0);
1634 }
1635 
1636 static int
vop_stdgetlowvnode(struct vop_getlowvnode_args * ap)1637 vop_stdgetlowvnode(struct vop_getlowvnode_args *ap)
1638 {
1639 	vref(ap->a_vp);
1640 	*ap->a_vplp = ap->a_vp;
1641 	return (0);
1642 }
1643