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/kernel.h>
43 #include <sys/limits.h>
44 #include <sys/lock.h>
45 #include <sys/lockf.h>
46 #include <sys/malloc.h>
47 #include <sys/mount.h>
48 #include <sys/namei.h>
49 #include <sys/rwlock.h>
50 #include <sys/fcntl.h>
51 #include <sys/unistd.h>
52 #include <sys/vnode.h>
53 #include <sys/dirent.h>
54 #include <sys/poll.h>
55 #include <sys/stat.h>
56 #include <security/audit/audit.h>
57 #include <sys/priv.h>
58
59 #include <security/mac/mac_framework.h>
60
61 #include <vm/vm.h>
62 #include <vm/vm_object.h>
63 #include <vm/vm_extern.h>
64 #include <vm/pmap.h>
65 #include <vm/vm_map.h>
66 #include <vm/vm_page.h>
67 #include <vm/vm_pager.h>
68 #include <vm/vnode_pager.h>
69
70 static int vop_nolookup(struct vop_lookup_args *);
71 static int vop_norename(struct vop_rename_args *);
72 static int vop_nostrategy(struct vop_strategy_args *);
73 static int dirent_exists(struct vnode *vp, const char *dirname,
74 struct thread *td);
75
76 static int vop_stdis_text(struct vop_is_text_args *ap);
77 static int vop_stdunset_text(struct vop_unset_text_args *ap);
78 static int vop_stdadd_writecount(struct vop_add_writecount_args *ap);
79 static int vop_stdcopy_file_range(struct vop_copy_file_range_args *ap);
80 static int vop_stdfdatasync(struct vop_fdatasync_args *ap);
81 static int vop_stdgetpages_async(struct vop_getpages_async_args *ap);
82 static int vop_stdread_pgcache(struct vop_read_pgcache_args *ap);
83 static int vop_stdstat(struct vop_stat_args *ap);
84 static int vop_stdvput_pair(struct vop_vput_pair_args *ap);
85 static int vop_stdgetlowvnode(struct vop_getlowvnode_args *ap);
86
87 /*
88 * This vnode table stores what we want to do if the filesystem doesn't
89 * implement a particular VOP.
90 *
91 * If there is no specific entry here, we will return EOPNOTSUPP.
92 *
93 * Note that every filesystem has to implement either vop_access
94 * or vop_accessx; failing to do so will result in immediate crash
95 * due to stack overflow, as vop_stdaccess() calls vop_stdaccessx(),
96 * which calls vop_stdaccess() etc.
97 */
98
99 struct vop_vector default_vnodeops = {
100 .vop_default = NULL,
101 .vop_bypass = VOP_EOPNOTSUPP,
102
103 .vop_access = vop_stdaccess,
104 .vop_accessx = vop_stdaccessx,
105 .vop_advise = vop_stdadvise,
106 .vop_advlock = vop_stdadvlock,
107 .vop_advlockasync = vop_stdadvlockasync,
108 .vop_advlockpurge = vop_stdadvlockpurge,
109 .vop_allocate = vop_stdallocate,
110 .vop_deallocate = vop_stddeallocate,
111 .vop_bmap = vop_stdbmap,
112 .vop_close = VOP_NULL,
113 .vop_fsync = VOP_NULL,
114 .vop_stat = vop_stdstat,
115 .vop_fdatasync = vop_stdfdatasync,
116 .vop_getlowvnode = vop_stdgetlowvnode,
117 .vop_getpages = vop_stdgetpages,
118 .vop_getpages_async = vop_stdgetpages_async,
119 .vop_getwritemount = vop_stdgetwritemount,
120 .vop_inactive = VOP_NULL,
121 .vop_need_inactive = vop_stdneed_inactive,
122 .vop_ioctl = vop_stdioctl,
123 .vop_kqfilter = vop_stdkqfilter,
124 .vop_islocked = vop_stdislocked,
125 .vop_lock1 = vop_stdlock,
126 .vop_lookup = vop_nolookup,
127 .vop_open = VOP_NULL,
128 .vop_pathconf = VOP_EINVAL,
129 .vop_poll = vop_nopoll,
130 .vop_putpages = vop_stdputpages,
131 .vop_readlink = VOP_EINVAL,
132 .vop_read_pgcache = vop_stdread_pgcache,
133 .vop_rename = vop_norename,
134 .vop_revoke = VOP_PANIC,
135 .vop_strategy = vop_nostrategy,
136 .vop_unlock = vop_stdunlock,
137 .vop_vptocnp = vop_stdvptocnp,
138 .vop_vptofh = vop_stdvptofh,
139 .vop_unp_bind = vop_stdunp_bind,
140 .vop_unp_connect = vop_stdunp_connect,
141 .vop_unp_detach = vop_stdunp_detach,
142 .vop_is_text = vop_stdis_text,
143 .vop_set_text = vop_stdset_text,
144 .vop_unset_text = vop_stdunset_text,
145 .vop_add_writecount = vop_stdadd_writecount,
146 .vop_copy_file_range = vop_stdcopy_file_range,
147 .vop_vput_pair = vop_stdvput_pair,
148 };
149 VFS_VOP_VECTOR_REGISTER(default_vnodeops);
150
151 /*
152 * Series of placeholder functions for various error returns for
153 * VOPs.
154 */
155
156 int
vop_eopnotsupp(struct vop_generic_args * ap)157 vop_eopnotsupp(struct vop_generic_args *ap)
158 {
159 /*
160 printf("vop_notsupp[%s]\n", ap->a_desc->vdesc_name);
161 */
162
163 return (EOPNOTSUPP);
164 }
165
166 int
vop_ebadf(struct vop_generic_args * ap)167 vop_ebadf(struct vop_generic_args *ap)
168 {
169
170 return (EBADF);
171 }
172
173 int
vop_enotty(struct vop_generic_args * ap)174 vop_enotty(struct vop_generic_args *ap)
175 {
176
177 return (ENOTTY);
178 }
179
180 int
vop_einval(struct vop_generic_args * ap)181 vop_einval(struct vop_generic_args *ap)
182 {
183
184 return (EINVAL);
185 }
186
187 int
vop_enoent(struct vop_generic_args * ap)188 vop_enoent(struct vop_generic_args *ap)
189 {
190
191 return (ENOENT);
192 }
193
194 int
vop_eagain(struct vop_generic_args * ap)195 vop_eagain(struct vop_generic_args *ap)
196 {
197
198 return (EAGAIN);
199 }
200
201 int
vop_null(struct vop_generic_args * ap)202 vop_null(struct vop_generic_args *ap)
203 {
204
205 return (0);
206 }
207
208 /*
209 * Helper function to panic on some bad VOPs in some filesystems.
210 */
211 int
vop_panic(struct vop_generic_args * ap)212 vop_panic(struct vop_generic_args *ap)
213 {
214
215 panic("filesystem goof: vop_panic[%s]", ap->a_desc->vdesc_name);
216 }
217
218 /*
219 * vop_std<something> and vop_no<something> are default functions for use by
220 * filesystems that need the "default reasonable" implementation for a
221 * particular operation.
222 *
223 * The documentation for the operations they implement exists (if it exists)
224 * in the VOP_<SOMETHING>(9) manpage (all uppercase).
225 */
226
227 /*
228 * Default vop for filesystems that do not support name lookup
229 */
230 static int
vop_nolookup(struct vop_lookup_args * ap)231 vop_nolookup(struct vop_lookup_args *ap)
232 {
233
234 *ap->a_vpp = NULL;
235 return (ENOTDIR);
236 }
237
238 /*
239 * vop_norename:
240 *
241 * Handle unlock and reference counting for arguments of vop_rename
242 * for filesystems that do not implement rename operation.
243 */
244 static int
vop_norename(struct vop_rename_args * ap)245 vop_norename(struct vop_rename_args *ap)
246 {
247
248 vop_rename_fail(ap);
249 return (EOPNOTSUPP);
250 }
251
252 /*
253 * vop_nostrategy:
254 *
255 * Strategy routine for VFS devices that have none.
256 *
257 * BIO_ERROR and B_INVAL must be cleared prior to calling any strategy
258 * routine. Typically this is done for a BIO_READ strategy call.
259 * Typically B_INVAL is assumed to already be clear prior to a write
260 * and should not be cleared manually unless you just made the buffer
261 * invalid. BIO_ERROR should be cleared either way.
262 */
263
264 static int
vop_nostrategy(struct vop_strategy_args * ap)265 vop_nostrategy (struct vop_strategy_args *ap)
266 {
267 printf("No strategy for buffer at %p\n", ap->a_bp);
268 vn_printf(ap->a_vp, "vnode ");
269 ap->a_bp->b_ioflags |= BIO_ERROR;
270 ap->a_bp->b_error = EOPNOTSUPP;
271 bufdone(ap->a_bp);
272 return (EOPNOTSUPP);
273 }
274
275 /*
276 * Check if a named file exists in a given directory vnode
277 *
278 * Returns 0 if the file exists, ENOENT if it doesn't, or errors returned by
279 * vn_dir_next_dirent().
280 */
281 static int
dirent_exists(struct vnode * vp,const char * dirname,struct thread * td)282 dirent_exists(struct vnode *vp, const char *dirname, struct thread *td)
283 {
284 char *dirbuf;
285 int error, eofflag;
286 size_t dirbuflen, len;
287 off_t off;
288 struct dirent *dp;
289 struct vattr va;
290
291 ASSERT_VOP_LOCKED(vp, "vnode not locked");
292 KASSERT(vp->v_type == VDIR, ("vp %p is not a directory", vp));
293
294 error = VOP_GETATTR(vp, &va, td->td_ucred);
295 if (error != 0)
296 return (error);
297
298 dirbuflen = MAX(DEV_BSIZE, GENERIC_MAXDIRSIZ);
299 if (dirbuflen < va.va_blocksize)
300 dirbuflen = va.va_blocksize;
301 dirbuf = malloc(dirbuflen, M_TEMP, M_WAITOK);
302
303 len = 0;
304 off = 0;
305 eofflag = 0;
306
307 for (;;) {
308 error = vn_dir_next_dirent(vp, td, dirbuf, dirbuflen,
309 &dp, &len, &off, &eofflag);
310 if (error != 0)
311 goto out;
312
313 if (len == 0)
314 break;
315
316 if (dp->d_type != DT_WHT && dp->d_fileno != 0 &&
317 strcmp(dp->d_name, dirname) == 0)
318 goto out;
319 }
320
321 error = ENOENT;
322
323 out:
324 free(dirbuf, M_TEMP);
325 return (error);
326 }
327
328 int
vop_stdaccess(struct vop_access_args * ap)329 vop_stdaccess(struct vop_access_args *ap)
330 {
331
332 KASSERT((ap->a_accmode & ~(VEXEC | VWRITE | VREAD | VADMIN |
333 VAPPEND)) == 0, ("invalid bit in accmode"));
334
335 return (VOP_ACCESSX(ap->a_vp, ap->a_accmode, ap->a_cred, ap->a_td));
336 }
337
338 int
vop_stdaccessx(struct vop_accessx_args * ap)339 vop_stdaccessx(struct vop_accessx_args *ap)
340 {
341 int error;
342 accmode_t accmode = ap->a_accmode;
343
344 error = vfs_unixify_accmode(&accmode);
345 if (error != 0)
346 return (error);
347
348 if (accmode == 0)
349 return (0);
350
351 return (VOP_ACCESS(ap->a_vp, accmode, ap->a_cred, ap->a_td));
352 }
353
354 /*
355 * Advisory record locking support
356 */
357 int
vop_stdadvlock(struct vop_advlock_args * ap)358 vop_stdadvlock(struct vop_advlock_args *ap)
359 {
360 struct vnode *vp;
361 struct mount *mp;
362 struct vattr vattr;
363 int error;
364
365 vp = ap->a_vp;
366
367 /*
368 * Provide atomicity of open(O_CREAT | O_EXCL | O_EXLOCK) for
369 * local filesystems. See vn_open_cred() for reciprocal part.
370 */
371 mp = vp->v_mount;
372 if (mp != NULL && (mp->mnt_flag & MNT_LOCAL) != 0 &&
373 ap->a_op == F_SETLK && (ap->a_flags & F_FIRSTOPEN) == 0) {
374 VI_LOCK(vp);
375 while ((vp->v_iflag & VI_FOPENING) != 0)
376 msleep(vp, VI_MTX(vp), PLOCK, "lockfo", 0);
377 VI_UNLOCK(vp);
378 }
379
380 if (ap->a_fl->l_whence == SEEK_END) {
381 /*
382 * The NFSv4 server must avoid doing a vn_lock() here, since it
383 * can deadlock the nfsd threads, due to a LOR. Fortunately
384 * the NFSv4 server always uses SEEK_SET and this code is
385 * only required for the SEEK_END case.
386 */
387 vn_lock(vp, LK_SHARED | LK_RETRY);
388 error = VOP_GETATTR(vp, &vattr, curthread->td_ucred);
389 VOP_UNLOCK(vp);
390 if (error)
391 return (error);
392 } else
393 vattr.va_size = 0;
394
395 return (lf_advlock(ap, &(vp->v_lockf), vattr.va_size));
396 }
397
398 int
vop_stdadvlockasync(struct vop_advlockasync_args * ap)399 vop_stdadvlockasync(struct vop_advlockasync_args *ap)
400 {
401 struct vnode *vp;
402 struct vattr vattr;
403 int error;
404
405 vp = ap->a_vp;
406 if (ap->a_fl->l_whence == SEEK_END) {
407 /* The size argument is only needed for SEEK_END. */
408 vn_lock(vp, LK_SHARED | LK_RETRY);
409 error = VOP_GETATTR(vp, &vattr, curthread->td_ucred);
410 VOP_UNLOCK(vp);
411 if (error)
412 return (error);
413 } else
414 vattr.va_size = 0;
415
416 return (lf_advlockasync(ap, &(vp->v_lockf), vattr.va_size));
417 }
418
419 int
vop_stdadvlockpurge(struct vop_advlockpurge_args * ap)420 vop_stdadvlockpurge(struct vop_advlockpurge_args *ap)
421 {
422 struct vnode *vp;
423
424 vp = ap->a_vp;
425 lf_purgelocks(vp, &vp->v_lockf);
426 return (0);
427 }
428
429 /*
430 * vop_stdpathconf:
431 *
432 * Standard implementation of POSIX pathconf, to get information about limits
433 * for a filesystem.
434 * Override per filesystem for the case where the filesystem has smaller
435 * limits.
436 */
437 int
vop_stdpathconf(struct vop_pathconf_args * ap)438 vop_stdpathconf(struct vop_pathconf_args *ap)
439 {
440
441 switch (ap->a_name) {
442 case _PC_ASYNC_IO:
443 *ap->a_retval = _POSIX_ASYNCHRONOUS_IO;
444 return (0);
445 case _PC_PATH_MAX:
446 *ap->a_retval = PATH_MAX;
447 return (0);
448 case _PC_ACL_EXTENDED:
449 case _PC_ACL_NFS4:
450 case _PC_CAP_PRESENT:
451 case _PC_DEALLOC_PRESENT:
452 case _PC_INF_PRESENT:
453 case _PC_MAC_PRESENT:
454 *ap->a_retval = 0;
455 return (0);
456 default:
457 return (EINVAL);
458 }
459 /* NOTREACHED */
460 }
461
462 /*
463 * Standard lock, unlock and islocked functions.
464 */
465 int
vop_stdlock(struct vop_lock1_args * ap)466 vop_stdlock(struct vop_lock1_args *ap)
467 {
468 struct vnode *vp = ap->a_vp;
469 struct mtx *ilk;
470
471 ilk = VI_MTX(vp);
472 return (lockmgr_lock_flags(vp->v_vnlock, ap->a_flags,
473 &ilk->lock_object, ap->a_file, ap->a_line));
474 }
475
476 /* See above. */
477 int
vop_stdunlock(struct vop_unlock_args * ap)478 vop_stdunlock(struct vop_unlock_args *ap)
479 {
480 struct vnode *vp = ap->a_vp;
481
482 return (lockmgr_unlock(vp->v_vnlock));
483 }
484
485 /* See above. */
486 int
vop_stdislocked(struct vop_islocked_args * ap)487 vop_stdislocked(struct vop_islocked_args *ap)
488 {
489
490 return (lockstatus(ap->a_vp->v_vnlock));
491 }
492
493 /*
494 * Variants of the above set.
495 *
496 * Differences are:
497 * - shared locking disablement is not supported
498 * - v_vnlock pointer is not honored
499 */
500 int
vop_lock(struct vop_lock1_args * ap)501 vop_lock(struct vop_lock1_args *ap)
502 {
503 struct vnode *vp = ap->a_vp;
504 int flags = ap->a_flags;
505 struct mtx *ilk;
506
507 MPASS(vp->v_vnlock == &vp->v_lock);
508
509 if (__predict_false((flags & ~(LK_TYPE_MASK | LK_NODDLKTREAT | LK_RETRY)) != 0))
510 goto other;
511
512 switch (flags & LK_TYPE_MASK) {
513 case LK_SHARED:
514 return (lockmgr_slock(&vp->v_lock, flags, ap->a_file, ap->a_line));
515 case LK_EXCLUSIVE:
516 return (lockmgr_xlock(&vp->v_lock, flags, ap->a_file, ap->a_line));
517 }
518 other:
519 ilk = VI_MTX(vp);
520 return (lockmgr_lock_flags(&vp->v_lock, flags,
521 &ilk->lock_object, ap->a_file, ap->a_line));
522 }
523
524 int
vop_unlock(struct vop_unlock_args * ap)525 vop_unlock(struct vop_unlock_args *ap)
526 {
527 struct vnode *vp = ap->a_vp;
528
529 MPASS(vp->v_vnlock == &vp->v_lock);
530
531 return (lockmgr_unlock(&vp->v_lock));
532 }
533
534 int
vop_islocked(struct vop_islocked_args * ap)535 vop_islocked(struct vop_islocked_args *ap)
536 {
537 struct vnode *vp = ap->a_vp;
538
539 MPASS(vp->v_vnlock == &vp->v_lock);
540
541 return (lockstatus(&vp->v_lock));
542 }
543
544 /*
545 * Return true for select/poll.
546 */
547 int
vop_nopoll(struct vop_poll_args * ap)548 vop_nopoll(struct vop_poll_args *ap)
549 {
550
551 if (ap->a_events & ~POLLSTANDARD)
552 return (POLLNVAL);
553 return (ap->a_events & (POLLIN | POLLOUT | POLLRDNORM | POLLWRNORM));
554 }
555
556 /*
557 * Implement poll for local filesystems that support it.
558 */
559 int
vop_stdpoll(struct vop_poll_args * ap)560 vop_stdpoll(struct vop_poll_args *ap)
561 {
562 if (ap->a_events & ~POLLSTANDARD)
563 return (vn_pollrecord(ap->a_vp, ap->a_td, ap->a_events));
564 return (ap->a_events & (POLLIN | POLLOUT | POLLRDNORM | POLLWRNORM));
565 }
566
567 /*
568 * Return our mount point, as we will take charge of the writes.
569 */
570 int
vop_stdgetwritemount(struct vop_getwritemount_args * ap)571 vop_stdgetwritemount(struct vop_getwritemount_args *ap)
572 {
573 struct mount *mp;
574 struct vnode *vp;
575
576 /*
577 * Note that having a reference does not prevent forced unmount from
578 * setting ->v_mount to NULL after the lock gets released. This is of
579 * no consequence for typical consumers (most notably vn_start_write)
580 * since in this case the vnode is VIRF_DOOMED. Unmount might have
581 * progressed far enough that its completion is only delayed by the
582 * reference obtained here. The consumer only needs to concern itself
583 * with releasing it.
584 */
585 vp = ap->a_vp;
586 mp = vfs_ref_from_vp(vp);
587 *(ap->a_mpp) = mp;
588 return (0);
589 }
590
591 /*
592 * If the file system doesn't implement VOP_BMAP, then return sensible defaults:
593 * - Return the vnode's bufobj instead of any underlying device's bufobj
594 * - Calculate the physical block number as if there were equal size
595 * consecutive blocks, but
596 * - Report no contiguous runs of blocks.
597 */
598 int
vop_stdbmap(struct vop_bmap_args * ap)599 vop_stdbmap(struct vop_bmap_args *ap)
600 {
601
602 if (ap->a_bop != NULL)
603 *ap->a_bop = &ap->a_vp->v_bufobj;
604 if (ap->a_bnp != NULL)
605 *ap->a_bnp = ap->a_bn * btodb(ap->a_vp->v_mount->mnt_stat.f_iosize);
606 if (ap->a_runp != NULL)
607 *ap->a_runp = 0;
608 if (ap->a_runb != NULL)
609 *ap->a_runb = 0;
610 return (0);
611 }
612
613 int
vop_stdfsync(struct vop_fsync_args * ap)614 vop_stdfsync(struct vop_fsync_args *ap)
615 {
616
617 return (vn_fsync_buf(ap->a_vp, ap->a_waitfor));
618 }
619
620 static int
vop_stdfdatasync(struct vop_fdatasync_args * ap)621 vop_stdfdatasync(struct vop_fdatasync_args *ap)
622 {
623
624 return (VOP_FSYNC(ap->a_vp, MNT_WAIT, ap->a_td));
625 }
626
627 int
vop_stdfdatasync_buf(struct vop_fdatasync_args * ap)628 vop_stdfdatasync_buf(struct vop_fdatasync_args *ap)
629 {
630
631 return (vn_fsync_buf(ap->a_vp, MNT_WAIT));
632 }
633
634 /* XXX Needs good comment and more info in the manpage (VOP_GETPAGES(9)). */
635 int
vop_stdgetpages(struct vop_getpages_args * ap)636 vop_stdgetpages(struct vop_getpages_args *ap)
637 {
638
639 return vnode_pager_generic_getpages(ap->a_vp, ap->a_m,
640 ap->a_count, ap->a_rbehind, ap->a_rahead, NULL, NULL);
641 }
642
643 static int
vop_stdgetpages_async(struct vop_getpages_async_args * ap)644 vop_stdgetpages_async(struct vop_getpages_async_args *ap)
645 {
646 int error;
647
648 error = VOP_GETPAGES(ap->a_vp, ap->a_m, ap->a_count, ap->a_rbehind,
649 ap->a_rahead);
650 if (ap->a_iodone != NULL)
651 ap->a_iodone(ap->a_arg, ap->a_m, ap->a_count, error);
652 return (error);
653 }
654
655 int
vop_stdkqfilter(struct vop_kqfilter_args * ap)656 vop_stdkqfilter(struct vop_kqfilter_args *ap)
657 {
658 return vfs_kqfilter(ap);
659 }
660
661 /* XXX Needs good comment and more info in the manpage (VOP_PUTPAGES(9)). */
662 int
vop_stdputpages(struct vop_putpages_args * ap)663 vop_stdputpages(struct vop_putpages_args *ap)
664 {
665
666 return vnode_pager_generic_putpages(ap->a_vp, ap->a_m, ap->a_count,
667 ap->a_sync, ap->a_rtvals);
668 }
669
670 int
vop_stdvptofh(struct vop_vptofh_args * ap)671 vop_stdvptofh(struct vop_vptofh_args *ap)
672 {
673 return (EOPNOTSUPP);
674 }
675
676 int
vop_stdvptocnp(struct vop_vptocnp_args * ap)677 vop_stdvptocnp(struct vop_vptocnp_args *ap)
678 {
679 struct vnode *const vp = ap->a_vp;
680 struct vnode **const dvp = ap->a_vpp;
681 char *buf = ap->a_buf;
682 size_t *buflen = ap->a_buflen;
683 char *dirbuf;
684 int i = *buflen;
685 int error = 0, covered = 0;
686 int eofflag, flags, locked;
687 size_t dirbuflen, len;
688 off_t off;
689 ino_t fileno;
690 struct vattr va;
691 struct nameidata nd;
692 struct thread *const td = curthread;
693 struct ucred *const cred = td->td_ucred;
694 struct dirent *dp;
695 struct vnode *mvp;
696
697 if (vp->v_type != VDIR)
698 return (ENOENT);
699
700 error = VOP_GETATTR(vp, &va, cred);
701 if (error)
702 return (error);
703
704 VREF(vp);
705 locked = VOP_ISLOCKED(vp);
706 VOP_UNLOCK(vp);
707 NDINIT_ATVP(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF, UIO_SYSSPACE,
708 "..", vp);
709 flags = FREAD;
710 error = vn_open_cred(&nd, &flags, 0, VN_OPEN_NOAUDIT, cred, NULL);
711 if (error) {
712 vn_lock(vp, locked | LK_RETRY);
713 return (error);
714 }
715 NDFREE_PNBUF(&nd);
716
717 mvp = *dvp = nd.ni_vp;
718
719 if (vp->v_mount != (*dvp)->v_mount &&
720 ((*dvp)->v_vflag & VV_ROOT) &&
721 ((*dvp)->v_mount->mnt_flag & MNT_UNION)) {
722 *dvp = (*dvp)->v_mount->mnt_vnodecovered;
723 VREF(mvp);
724 VOP_UNLOCK(mvp);
725 vn_close(mvp, FREAD, cred, td);
726 VREF(*dvp);
727 vn_lock(*dvp, LK_SHARED | LK_RETRY);
728 covered = 1;
729 }
730
731 fileno = va.va_fileid;
732
733 dirbuflen = MAX(DEV_BSIZE, GENERIC_MAXDIRSIZ);
734 if (dirbuflen < va.va_blocksize)
735 dirbuflen = va.va_blocksize;
736 dirbuf = malloc(dirbuflen, M_TEMP, M_WAITOK);
737
738 if ((*dvp)->v_type != VDIR) {
739 error = ENOENT;
740 goto out;
741 }
742
743 len = 0;
744 off = 0;
745 eofflag = 0;
746
747 for (;;) {
748 /* call VOP_READDIR of parent */
749 error = vn_dir_next_dirent(*dvp, td,
750 dirbuf, dirbuflen, &dp, &len, &off, &eofflag);
751 if (error != 0)
752 goto out;
753
754 if (len == 0) {
755 error = ENOENT;
756 goto out;
757 }
758
759 if ((dp->d_type != DT_WHT) &&
760 (dp->d_fileno == fileno)) {
761 if (covered) {
762 VOP_UNLOCK(*dvp);
763 vn_lock(mvp, LK_SHARED | LK_RETRY);
764 if (dirent_exists(mvp, dp->d_name, td) == 0) {
765 error = ENOENT;
766 VOP_UNLOCK(mvp);
767 vn_lock(*dvp, LK_SHARED | LK_RETRY);
768 goto out;
769 }
770 VOP_UNLOCK(mvp);
771 vn_lock(*dvp, LK_SHARED | LK_RETRY);
772 }
773 i -= dp->d_namlen;
774
775 if (i < 0) {
776 error = ENOMEM;
777 goto out;
778 }
779 if (dp->d_namlen == 1 && dp->d_name[0] == '.') {
780 error = ENOENT;
781 } else {
782 bcopy(dp->d_name, buf + i, dp->d_namlen);
783 error = 0;
784 }
785 goto out;
786 }
787 }
788
789 out:
790 free(dirbuf, M_TEMP);
791 if (!error) {
792 *buflen = i;
793 vref(*dvp);
794 }
795 if (covered) {
796 vput(*dvp);
797 vrele(mvp);
798 } else {
799 VOP_UNLOCK(mvp);
800 vn_close(mvp, FREAD, cred, td);
801 }
802 vn_lock(vp, locked | LK_RETRY);
803 return (error);
804 }
805
806 int
vop_stdallocate(struct vop_allocate_args * ap)807 vop_stdallocate(struct vop_allocate_args *ap)
808 {
809 #ifdef __notyet__
810 struct statfs *sfs;
811 off_t maxfilesize = 0;
812 #endif
813 struct iovec aiov;
814 struct vattr vattr, *vap;
815 struct uio auio;
816 off_t fsize, len, cur, offset;
817 uint8_t *buf;
818 struct thread *td;
819 struct vnode *vp;
820 size_t iosize;
821 int error;
822
823 buf = NULL;
824 error = 0;
825 td = curthread;
826 vap = &vattr;
827 vp = ap->a_vp;
828 len = *ap->a_len;
829 offset = *ap->a_offset;
830
831 error = VOP_GETATTR(vp, vap, ap->a_cred);
832 if (error != 0)
833 goto out;
834 fsize = vap->va_size;
835 iosize = vap->va_blocksize;
836 if (iosize == 0)
837 iosize = BLKDEV_IOSIZE;
838 if (iosize > maxphys)
839 iosize = maxphys;
840 buf = malloc(iosize, M_TEMP, M_WAITOK);
841
842 #ifdef __notyet__
843 /*
844 * Check if the filesystem sets f_maxfilesize; if not use
845 * VOP_SETATTR to perform the check.
846 */
847 sfs = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
848 error = VFS_STATFS(vp->v_mount, sfs, td);
849 if (error == 0)
850 maxfilesize = sfs->f_maxfilesize;
851 free(sfs, M_STATFS);
852 if (error != 0)
853 goto out;
854 if (maxfilesize) {
855 if (offset > maxfilesize || len > maxfilesize ||
856 offset + len > maxfilesize) {
857 error = EFBIG;
858 goto out;
859 }
860 } else
861 #endif
862 if (offset + len > vap->va_size) {
863 /*
864 * Test offset + len against the filesystem's maxfilesize.
865 */
866 VATTR_NULL(vap);
867 vap->va_size = offset + len;
868 error = VOP_SETATTR(vp, vap, ap->a_cred);
869 if (error != 0)
870 goto out;
871 VATTR_NULL(vap);
872 vap->va_size = fsize;
873 error = VOP_SETATTR(vp, vap, ap->a_cred);
874 if (error != 0)
875 goto out;
876 }
877
878 for (;;) {
879 /*
880 * Read and write back anything below the nominal file
881 * size. There's currently no way outside the filesystem
882 * to know whether this area is sparse or not.
883 */
884 cur = iosize;
885 if ((offset % iosize) != 0)
886 cur -= (offset % iosize);
887 if (cur > len)
888 cur = len;
889 if (offset < fsize) {
890 aiov.iov_base = buf;
891 aiov.iov_len = cur;
892 auio.uio_iov = &aiov;
893 auio.uio_iovcnt = 1;
894 auio.uio_offset = offset;
895 auio.uio_resid = cur;
896 auio.uio_segflg = UIO_SYSSPACE;
897 auio.uio_rw = UIO_READ;
898 auio.uio_td = td;
899 error = VOP_READ(vp, &auio, ap->a_ioflag, ap->a_cred);
900 if (error != 0)
901 break;
902 if (auio.uio_resid > 0) {
903 bzero(buf + cur - auio.uio_resid,
904 auio.uio_resid);
905 }
906 } else {
907 bzero(buf, cur);
908 }
909
910 aiov.iov_base = buf;
911 aiov.iov_len = cur;
912 auio.uio_iov = &aiov;
913 auio.uio_iovcnt = 1;
914 auio.uio_offset = offset;
915 auio.uio_resid = cur;
916 auio.uio_segflg = UIO_SYSSPACE;
917 auio.uio_rw = UIO_WRITE;
918 auio.uio_td = td;
919
920 error = VOP_WRITE(vp, &auio, ap->a_ioflag, ap->a_cred);
921 if (error != 0)
922 break;
923
924 len -= cur;
925 offset += cur;
926 if (len == 0)
927 break;
928 if (should_yield())
929 break;
930 }
931
932 out:
933 *ap->a_len = len;
934 *ap->a_offset = offset;
935 free(buf, M_TEMP);
936 return (error);
937 }
938
939 static int
vp_zerofill(struct vnode * vp,struct vattr * vap,off_t * offsetp,off_t * lenp,int ioflag,struct ucred * cred)940 vp_zerofill(struct vnode *vp, struct vattr *vap, off_t *offsetp, off_t *lenp,
941 int ioflag, struct ucred *cred)
942 {
943 int iosize;
944 int error = 0;
945 struct iovec aiov;
946 struct uio auio;
947 struct thread *td;
948 off_t offset, len;
949
950 iosize = vap->va_blocksize;
951 td = curthread;
952 offset = *offsetp;
953 len = *lenp;
954
955 if (iosize == 0)
956 iosize = BLKDEV_IOSIZE;
957 /* If va_blocksize is 512 bytes, iosize will be 4 kilobytes */
958 iosize = min(iosize * 8, ZERO_REGION_SIZE);
959
960 while (len > 0) {
961 int xfersize = iosize;
962 if (offset % iosize != 0)
963 xfersize -= offset % iosize;
964 if (xfersize > len)
965 xfersize = len;
966
967 aiov.iov_base = __DECONST(void *, zero_region);
968 aiov.iov_len = xfersize;
969 auio.uio_iov = &aiov;
970 auio.uio_iovcnt = 1;
971 auio.uio_offset = offset;
972 auio.uio_resid = xfersize;
973 auio.uio_segflg = UIO_SYSSPACE;
974 auio.uio_rw = UIO_WRITE;
975 auio.uio_td = td;
976
977 error = VOP_WRITE(vp, &auio, ioflag, cred);
978 if (error != 0) {
979 len -= xfersize - auio.uio_resid;
980 offset += xfersize - auio.uio_resid;
981 break;
982 }
983
984 len -= xfersize;
985 offset += xfersize;
986 }
987
988 *offsetp = offset;
989 *lenp = len;
990 return (error);
991 }
992
993 int
vop_stddeallocate(struct vop_deallocate_args * ap)994 vop_stddeallocate(struct vop_deallocate_args *ap)
995 {
996 struct vnode *vp;
997 off_t offset, len;
998 struct ucred *cred;
999 int error;
1000 struct vattr va;
1001 off_t noff, xfersize, rem;
1002
1003 vp = ap->a_vp;
1004 offset = *ap->a_offset;
1005 cred = ap->a_cred;
1006
1007 error = VOP_GETATTR(vp, &va, cred);
1008 if (error)
1009 return (error);
1010
1011 len = omin((off_t)va.va_size - offset, *ap->a_len);
1012 while (len > 0) {
1013 noff = offset;
1014 error = vn_bmap_seekhole_locked(vp, FIOSEEKDATA, &noff, cred);
1015 if (error) {
1016 if (error != ENXIO)
1017 /* XXX: Is it okay to fallback further? */
1018 goto out;
1019
1020 /*
1021 * No more data region to be filled
1022 */
1023 offset += len;
1024 len = 0;
1025 error = 0;
1026 break;
1027 }
1028 KASSERT(noff >= offset, ("FIOSEEKDATA going backward"));
1029 if (noff != offset) {
1030 xfersize = omin(noff - offset, len);
1031 len -= xfersize;
1032 offset += xfersize;
1033 if (len == 0)
1034 break;
1035 }
1036 error = vn_bmap_seekhole_locked(vp, FIOSEEKHOLE, &noff, cred);
1037 if (error)
1038 goto out;
1039
1040 /* Fill zeroes */
1041 xfersize = rem = omin(noff - offset, len);
1042 error = vp_zerofill(vp, &va, &offset, &rem, ap->a_ioflag, cred);
1043 if (error) {
1044 len -= xfersize - rem;
1045 goto out;
1046 }
1047
1048 len -= xfersize;
1049 if (should_yield())
1050 break;
1051 }
1052 /* Handle the case when offset is beyond EOF */
1053 if (len < 0)
1054 len = 0;
1055 out:
1056 *ap->a_offset = offset;
1057 *ap->a_len = len;
1058 return (error);
1059 }
1060
1061 int
vop_stdadvise(struct vop_advise_args * ap)1062 vop_stdadvise(struct vop_advise_args *ap)
1063 {
1064 struct vnode *vp;
1065 struct bufobj *bo;
1066 uintmax_t bstart, bend;
1067 daddr_t startn, endn;
1068 int bsize, error;
1069
1070 vp = ap->a_vp;
1071 switch (ap->a_advice) {
1072 case POSIX_FADV_WILLNEED:
1073 /*
1074 * Do nothing for now. Filesystems should provide a
1075 * custom method which starts an asynchronous read of
1076 * the requested region.
1077 */
1078 error = 0;
1079 break;
1080 case POSIX_FADV_DONTNEED:
1081 error = 0;
1082 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1083 if (VN_IS_DOOMED(vp)) {
1084 VOP_UNLOCK(vp);
1085 break;
1086 }
1087
1088 /*
1089 * Round to block boundaries (and later possibly further to
1090 * page boundaries). Applications cannot reasonably be aware
1091 * of the boundaries, and the rounding must be to expand at
1092 * both extremities to cover enough. It still doesn't cover
1093 * read-ahead. For partial blocks, this gives unnecessary
1094 * discarding of buffers but is efficient enough since the
1095 * pages usually remain in VMIO for some time.
1096 */
1097 bsize = vp->v_bufobj.bo_bsize;
1098 bstart = rounddown(ap->a_start, bsize);
1099 bend = ap->a_end;
1100 bend = roundup(bend, bsize);
1101
1102 /*
1103 * Deactivate pages in the specified range from the backing VM
1104 * object. Pages that are resident in the buffer cache will
1105 * remain wired until their corresponding buffers are released
1106 * below.
1107 */
1108 if (vp->v_object != NULL) {
1109 VM_OBJECT_RLOCK(vp->v_object);
1110 vm_object_page_noreuse(vp->v_object,
1111 OFF_TO_IDX(trunc_page(bstart)),
1112 OFF_TO_IDX(round_page(bend)));
1113 VM_OBJECT_RUNLOCK(vp->v_object);
1114 }
1115
1116 bo = &vp->v_bufobj;
1117 startn = bstart / bsize;
1118 endn = bend / bsize;
1119 BO_RLOCK(bo);
1120 error = bnoreuselist(&bo->bo_clean, bo, startn, endn);
1121 if (error == 0)
1122 error = bnoreuselist(&bo->bo_dirty, bo, startn, endn);
1123 BO_RUNLOCK(bo);
1124 VOP_UNLOCK(vp);
1125 break;
1126 default:
1127 error = EINVAL;
1128 break;
1129 }
1130 return (error);
1131 }
1132
1133 int
vop_stdunp_bind(struct vop_unp_bind_args * ap)1134 vop_stdunp_bind(struct vop_unp_bind_args *ap)
1135 {
1136
1137 ap->a_vp->v_unpcb = ap->a_unpcb;
1138 return (0);
1139 }
1140
1141 int
vop_stdunp_connect(struct vop_unp_connect_args * ap)1142 vop_stdunp_connect(struct vop_unp_connect_args *ap)
1143 {
1144
1145 *ap->a_unpcb = ap->a_vp->v_unpcb;
1146 return (0);
1147 }
1148
1149 int
vop_stdunp_detach(struct vop_unp_detach_args * ap)1150 vop_stdunp_detach(struct vop_unp_detach_args *ap)
1151 {
1152
1153 ap->a_vp->v_unpcb = NULL;
1154 return (0);
1155 }
1156
1157 static int
vop_stdis_text(struct vop_is_text_args * ap)1158 vop_stdis_text(struct vop_is_text_args *ap)
1159 {
1160
1161 return ((int)atomic_load_int(&ap->a_vp->v_writecount) < 0);
1162 }
1163
1164 int
vop_stdset_text(struct vop_set_text_args * ap)1165 vop_stdset_text(struct vop_set_text_args *ap)
1166 {
1167 struct vnode *vp;
1168 int n;
1169 bool gotref;
1170
1171 vp = ap->a_vp;
1172
1173 n = atomic_load_int(&vp->v_writecount);
1174 for (;;) {
1175 if (__predict_false(n > 0)) {
1176 return (ETXTBSY);
1177 }
1178
1179 /*
1180 * Transition point, we may need to grab a reference on the vnode.
1181 *
1182 * Take the ref early As a safety measure against bogus calls
1183 * to vop_stdunset_text.
1184 */
1185 if (n == 0) {
1186 gotref = false;
1187 if ((vn_irflag_read(vp) & VIRF_TEXT_REF) != 0) {
1188 vref(vp);
1189 gotref = true;
1190 }
1191 if (atomic_fcmpset_int(&vp->v_writecount, &n, -1)) {
1192 return (0);
1193 }
1194 if (gotref) {
1195 vunref(vp);
1196 }
1197 continue;
1198 }
1199
1200 MPASS(n < 0);
1201 if (atomic_fcmpset_int(&vp->v_writecount, &n, n - 1)) {
1202 return (0);
1203 }
1204 }
1205 __assert_unreachable();
1206 }
1207
1208 static int
vop_stdunset_text(struct vop_unset_text_args * ap)1209 vop_stdunset_text(struct vop_unset_text_args *ap)
1210 {
1211 struct vnode *vp;
1212 int n;
1213
1214 vp = ap->a_vp;
1215
1216 n = atomic_load_int(&vp->v_writecount);
1217 for (;;) {
1218 if (__predict_false(n >= 0)) {
1219 return (EINVAL);
1220 }
1221
1222 /*
1223 * Transition point, we may need to release a reference on the vnode.
1224 */
1225 if (n == -1) {
1226 if (atomic_fcmpset_int(&vp->v_writecount, &n, 0)) {
1227 if ((vn_irflag_read(vp) & VIRF_TEXT_REF) != 0) {
1228 vunref(vp);
1229 }
1230 return (0);
1231 }
1232 continue;
1233 }
1234
1235 MPASS(n < -1);
1236 if (atomic_fcmpset_int(&vp->v_writecount, &n, n + 1)) {
1237 return (0);
1238 }
1239 }
1240 __assert_unreachable();
1241 }
1242
1243 static __always_inline int
vop_stdadd_writecount_impl(struct vop_add_writecount_args * ap,bool handle_msync)1244 vop_stdadd_writecount_impl(struct vop_add_writecount_args *ap, bool handle_msync)
1245 {
1246 struct vnode *vp;
1247 struct mount *mp __diagused;
1248 int n;
1249
1250 vp = ap->a_vp;
1251
1252 #ifdef INVARIANTS
1253 mp = vp->v_mount;
1254 if (mp != NULL) {
1255 if (handle_msync) {
1256 VNPASS((mp->mnt_kern_flag & MNTK_NOMSYNC) == 0, vp);
1257 } else {
1258 VNPASS((mp->mnt_kern_flag & MNTK_NOMSYNC) != 0, vp);
1259 }
1260 }
1261 #endif
1262
1263 n = atomic_load_int(&vp->v_writecount);
1264 for (;;) {
1265 if (__predict_false(n < 0)) {
1266 return (ETXTBSY);
1267 }
1268
1269 VNASSERT(n + ap->a_inc >= 0, vp,
1270 ("neg writecount increment %d + %d = %d", n, ap->a_inc,
1271 n + ap->a_inc));
1272 if (n == 0) {
1273 if (handle_msync) {
1274 vlazy(vp);
1275 }
1276 }
1277
1278 if (atomic_fcmpset_int(&vp->v_writecount, &n, n + ap->a_inc)) {
1279 return (0);
1280 }
1281 }
1282 __assert_unreachable();
1283 }
1284
1285 int
vop_stdadd_writecount(struct vop_add_writecount_args * ap)1286 vop_stdadd_writecount(struct vop_add_writecount_args *ap)
1287 {
1288
1289 return (vop_stdadd_writecount_impl(ap, true));
1290 }
1291
1292 int
vop_stdadd_writecount_nomsync(struct vop_add_writecount_args * ap)1293 vop_stdadd_writecount_nomsync(struct vop_add_writecount_args *ap)
1294 {
1295
1296 return (vop_stdadd_writecount_impl(ap, false));
1297 }
1298
1299 int
vop_stdneed_inactive(struct vop_need_inactive_args * ap)1300 vop_stdneed_inactive(struct vop_need_inactive_args *ap)
1301 {
1302
1303 return (1);
1304 }
1305
1306 int
vop_stdioctl(struct vop_ioctl_args * ap)1307 vop_stdioctl(struct vop_ioctl_args *ap)
1308 {
1309 struct vnode *vp;
1310 struct vattr va;
1311 off_t *offp;
1312 int error;
1313
1314 switch (ap->a_command) {
1315 case FIOSEEKDATA:
1316 case FIOSEEKHOLE:
1317 vp = ap->a_vp;
1318 error = vn_lock(vp, LK_SHARED);
1319 if (error != 0)
1320 return (EBADF);
1321 if (vp->v_type == VREG)
1322 error = VOP_GETATTR(vp, &va, ap->a_cred);
1323 else
1324 error = ENOTTY;
1325 if (error == 0) {
1326 offp = ap->a_data;
1327 if (*offp < 0 || *offp >= va.va_size)
1328 error = ENXIO;
1329 else if (ap->a_command == FIOSEEKHOLE)
1330 *offp = va.va_size;
1331 }
1332 VOP_UNLOCK(vp);
1333 break;
1334 default:
1335 error = ENOTTY;
1336 break;
1337 }
1338 return (error);
1339 }
1340
1341 /*
1342 * vfs default ops
1343 * used to fill the vfs function table to get reasonable default return values.
1344 */
1345 int
vfs_stdroot(struct mount * mp,int flags,struct vnode ** vpp)1346 vfs_stdroot(struct mount *mp, int flags, struct vnode **vpp)
1347 {
1348
1349 return (EOPNOTSUPP);
1350 }
1351
1352 int
vfs_stdstatfs(struct mount * mp,struct statfs * sbp)1353 vfs_stdstatfs(struct mount *mp, struct statfs *sbp)
1354 {
1355
1356 return (EOPNOTSUPP);
1357 }
1358
1359 int
vfs_stdquotactl(struct mount * mp,int cmds,uid_t uid,void * arg,bool * mp_busy)1360 vfs_stdquotactl(struct mount *mp, int cmds, uid_t uid, void *arg, bool *mp_busy)
1361 {
1362 return (EOPNOTSUPP);
1363 }
1364
1365 int
vfs_stdsync(struct mount * mp,int waitfor)1366 vfs_stdsync(struct mount *mp, int waitfor)
1367 {
1368 struct vnode *vp, *mvp;
1369 struct thread *td;
1370 int error, lockreq, allerror = 0;
1371
1372 td = curthread;
1373 lockreq = LK_EXCLUSIVE | LK_INTERLOCK;
1374 if (waitfor != MNT_WAIT)
1375 lockreq |= LK_NOWAIT;
1376 /*
1377 * Force stale buffer cache information to be flushed.
1378 */
1379 loop:
1380 MNT_VNODE_FOREACH_ALL(vp, mp, mvp) {
1381 if (vp->v_bufobj.bo_dirty.bv_cnt == 0) {
1382 VI_UNLOCK(vp);
1383 continue;
1384 }
1385 if ((error = vget(vp, lockreq)) != 0) {
1386 if (error == ENOENT) {
1387 MNT_VNODE_FOREACH_ALL_ABORT(mp, mvp);
1388 goto loop;
1389 }
1390 continue;
1391 }
1392 error = VOP_FSYNC(vp, waitfor, td);
1393 if (error)
1394 allerror = error;
1395 vput(vp);
1396 }
1397 return (allerror);
1398 }
1399
1400 int
vfs_stdnosync(struct mount * mp,int waitfor)1401 vfs_stdnosync(struct mount *mp, int waitfor)
1402 {
1403
1404 return (0);
1405 }
1406
1407 static int
vop_stdcopy_file_range(struct vop_copy_file_range_args * ap)1408 vop_stdcopy_file_range(struct vop_copy_file_range_args *ap)
1409 {
1410 int error;
1411
1412 error = vn_generic_copy_file_range(ap->a_invp, ap->a_inoffp,
1413 ap->a_outvp, ap->a_outoffp, ap->a_lenp, ap->a_flags, ap->a_incred,
1414 ap->a_outcred, ap->a_fsizetd);
1415 return (error);
1416 }
1417
1418 int
vfs_stdvget(struct mount * mp,ino_t ino,int flags,struct vnode ** vpp)1419 vfs_stdvget(struct mount *mp, ino_t ino, int flags, struct vnode **vpp)
1420 {
1421
1422 return (EOPNOTSUPP);
1423 }
1424
1425 int
vfs_stdfhtovp(struct mount * mp,struct fid * fhp,int flags,struct vnode ** vpp)1426 vfs_stdfhtovp(struct mount *mp, struct fid *fhp, int flags, struct vnode **vpp)
1427 {
1428
1429 return (EOPNOTSUPP);
1430 }
1431
1432 int
vfs_stdinit(struct vfsconf * vfsp)1433 vfs_stdinit(struct vfsconf *vfsp)
1434 {
1435
1436 return (0);
1437 }
1438
1439 int
vfs_stduninit(struct vfsconf * vfsp)1440 vfs_stduninit(struct vfsconf *vfsp)
1441 {
1442
1443 return(0);
1444 }
1445
1446 int
vfs_stdextattrctl(struct mount * mp,int cmd,struct vnode * filename_vp,int attrnamespace,const char * attrname)1447 vfs_stdextattrctl(struct mount *mp, int cmd, struct vnode *filename_vp,
1448 int attrnamespace, const char *attrname)
1449 {
1450
1451 if (filename_vp != NULL)
1452 VOP_UNLOCK(filename_vp);
1453 return (EOPNOTSUPP);
1454 }
1455
1456 int
vfs_stdsysctl(struct mount * mp,fsctlop_t op,struct sysctl_req * req)1457 vfs_stdsysctl(struct mount *mp, fsctlop_t op, struct sysctl_req *req)
1458 {
1459
1460 return (EOPNOTSUPP);
1461 }
1462
1463 static vop_bypass_t *
bp_by_off(struct vop_vector * vop,struct vop_generic_args * a)1464 bp_by_off(struct vop_vector *vop, struct vop_generic_args *a)
1465 {
1466
1467 return (*(vop_bypass_t **)((char *)vop + a->a_desc->vdesc_vop_offset));
1468 }
1469
1470 int
vop_sigdefer(struct vop_vector * vop,struct vop_generic_args * a)1471 vop_sigdefer(struct vop_vector *vop, struct vop_generic_args *a)
1472 {
1473 vop_bypass_t *bp;
1474 int prev_stops, rc;
1475
1476 bp = bp_by_off(vop, a);
1477 MPASS(bp != NULL);
1478
1479 prev_stops = sigdeferstop(SIGDEFERSTOP_SILENT);
1480 rc = bp(a);
1481 sigallowstop(prev_stops);
1482 return (rc);
1483 }
1484
1485 static int
vop_stdstat(struct vop_stat_args * a)1486 vop_stdstat(struct vop_stat_args *a)
1487 {
1488 struct vattr vattr;
1489 struct vattr *vap;
1490 struct vnode *vp;
1491 struct stat *sb;
1492 int error;
1493 u_short mode;
1494
1495 vp = a->a_vp;
1496 sb = a->a_sb;
1497
1498 error = vop_stat_helper_pre(a);
1499 if (error != 0)
1500 return (error);
1501
1502 vap = &vattr;
1503
1504 /*
1505 * Initialize defaults for new and unusual fields, so that file
1506 * systems which don't support these fields don't need to know
1507 * about them.
1508 */
1509 vap->va_birthtime.tv_sec = -1;
1510 vap->va_birthtime.tv_nsec = 0;
1511 vap->va_fsid = VNOVAL;
1512 vap->va_gen = 0;
1513 vap->va_rdev = NODEV;
1514
1515 error = VOP_GETATTR(vp, vap, a->a_active_cred);
1516 if (error)
1517 goto out;
1518
1519 /*
1520 * Zero the spare stat fields
1521 */
1522 bzero(sb, sizeof *sb);
1523
1524 /*
1525 * Copy from vattr table
1526 */
1527 if (vap->va_fsid != VNOVAL)
1528 sb->st_dev = vap->va_fsid;
1529 else
1530 sb->st_dev = vp->v_mount->mnt_stat.f_fsid.val[0];
1531 sb->st_ino = vap->va_fileid;
1532 mode = vap->va_mode;
1533 switch (vap->va_type) {
1534 case VREG:
1535 mode |= S_IFREG;
1536 break;
1537 case VDIR:
1538 mode |= S_IFDIR;
1539 break;
1540 case VBLK:
1541 mode |= S_IFBLK;
1542 break;
1543 case VCHR:
1544 mode |= S_IFCHR;
1545 break;
1546 case VLNK:
1547 mode |= S_IFLNK;
1548 break;
1549 case VSOCK:
1550 mode |= S_IFSOCK;
1551 break;
1552 case VFIFO:
1553 mode |= S_IFIFO;
1554 break;
1555 default:
1556 error = EBADF;
1557 goto out;
1558 }
1559 sb->st_mode = mode;
1560 sb->st_nlink = vap->va_nlink;
1561 sb->st_uid = vap->va_uid;
1562 sb->st_gid = vap->va_gid;
1563 sb->st_rdev = vap->va_rdev;
1564 if (vap->va_size > OFF_MAX) {
1565 error = EOVERFLOW;
1566 goto out;
1567 }
1568 sb->st_size = vap->va_size;
1569 sb->st_atim.tv_sec = vap->va_atime.tv_sec;
1570 sb->st_atim.tv_nsec = vap->va_atime.tv_nsec;
1571 sb->st_mtim.tv_sec = vap->va_mtime.tv_sec;
1572 sb->st_mtim.tv_nsec = vap->va_mtime.tv_nsec;
1573 sb->st_ctim.tv_sec = vap->va_ctime.tv_sec;
1574 sb->st_ctim.tv_nsec = vap->va_ctime.tv_nsec;
1575 sb->st_birthtim.tv_sec = vap->va_birthtime.tv_sec;
1576 sb->st_birthtim.tv_nsec = vap->va_birthtime.tv_nsec;
1577
1578 /*
1579 * According to www.opengroup.org, the meaning of st_blksize is
1580 * "a filesystem-specific preferred I/O block size for this
1581 * object. In some filesystem types, this may vary from file
1582 * to file"
1583 * Use minimum/default of PAGE_SIZE (e.g. for VCHR).
1584 */
1585
1586 sb->st_blksize = max(PAGE_SIZE, vap->va_blocksize);
1587 sb->st_flags = vap->va_flags;
1588 sb->st_blocks = vap->va_bytes / S_BLKSIZE;
1589 sb->st_gen = vap->va_gen;
1590 out:
1591 return (vop_stat_helper_post(a, error));
1592 }
1593
1594 static int
vop_stdread_pgcache(struct vop_read_pgcache_args * ap __unused)1595 vop_stdread_pgcache(struct vop_read_pgcache_args *ap __unused)
1596 {
1597 return (EJUSTRETURN);
1598 }
1599
1600 static int
vop_stdvput_pair(struct vop_vput_pair_args * ap)1601 vop_stdvput_pair(struct vop_vput_pair_args *ap)
1602 {
1603 struct vnode *dvp, *vp, **vpp;
1604
1605 dvp = ap->a_dvp;
1606 vpp = ap->a_vpp;
1607 vput(dvp);
1608 if (vpp != NULL && ap->a_unlock_vp && (vp = *vpp) != NULL)
1609 vput(vp);
1610 return (0);
1611 }
1612
1613 static int
vop_stdgetlowvnode(struct vop_getlowvnode_args * ap)1614 vop_stdgetlowvnode(struct vop_getlowvnode_args *ap)
1615 {
1616 vref(ap->a_vp);
1617 *ap->a_vplp = ap->a_vp;
1618 return (0);
1619 }
1620