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 case _PC_HAS_HIDDENSYSTEM:
460 *ap->a_retval = 0;
461 return (0);
462 default:
463 return (EINVAL);
464 }
465 /* NOTREACHED */
466 }
467
468 /*
469 * Standard lock, unlock and islocked functions.
470 */
471 int
vop_stdlock(struct vop_lock1_args * ap)472 vop_stdlock(struct vop_lock1_args *ap)
473 {
474 struct vnode *vp = ap->a_vp;
475 struct mtx *ilk;
476
477 ilk = VI_MTX(vp);
478 return (lockmgr_lock_flags(vp->v_vnlock, ap->a_flags,
479 &ilk->lock_object, ap->a_file, ap->a_line));
480 }
481
482 /* See above. */
483 int
vop_stdunlock(struct vop_unlock_args * ap)484 vop_stdunlock(struct vop_unlock_args *ap)
485 {
486 struct vnode *vp = ap->a_vp;
487
488 return (lockmgr_unlock(vp->v_vnlock));
489 }
490
491 /* See above. */
492 int
vop_stdislocked(struct vop_islocked_args * ap)493 vop_stdislocked(struct vop_islocked_args *ap)
494 {
495
496 return (lockstatus(ap->a_vp->v_vnlock));
497 }
498
499 /*
500 * Variants of the above set.
501 *
502 * Differences are:
503 * - shared locking disablement is not supported
504 * - v_vnlock pointer is not honored
505 */
506 int
vop_lock(struct vop_lock1_args * ap)507 vop_lock(struct vop_lock1_args *ap)
508 {
509 struct vnode *vp = ap->a_vp;
510 int flags = ap->a_flags;
511 struct mtx *ilk;
512
513 MPASS(vp->v_vnlock == &vp->v_lock);
514
515 if (__predict_false((flags & ~(LK_TYPE_MASK | LK_NODDLKTREAT | LK_RETRY)) != 0))
516 goto other;
517
518 switch (flags & LK_TYPE_MASK) {
519 case LK_SHARED:
520 return (lockmgr_slock(&vp->v_lock, flags, ap->a_file, ap->a_line));
521 case LK_EXCLUSIVE:
522 return (lockmgr_xlock(&vp->v_lock, flags, ap->a_file, ap->a_line));
523 }
524 other:
525 ilk = VI_MTX(vp);
526 return (lockmgr_lock_flags(&vp->v_lock, flags,
527 &ilk->lock_object, ap->a_file, ap->a_line));
528 }
529
530 int
vop_unlock(struct vop_unlock_args * ap)531 vop_unlock(struct vop_unlock_args *ap)
532 {
533 struct vnode *vp = ap->a_vp;
534
535 MPASS(vp->v_vnlock == &vp->v_lock);
536
537 return (lockmgr_unlock(&vp->v_lock));
538 }
539
540 int
vop_islocked(struct vop_islocked_args * ap)541 vop_islocked(struct vop_islocked_args *ap)
542 {
543 struct vnode *vp = ap->a_vp;
544
545 MPASS(vp->v_vnlock == &vp->v_lock);
546
547 return (lockstatus(&vp->v_lock));
548 }
549
550 /*
551 * Return true for select/poll.
552 */
553 int
vop_nopoll(struct vop_poll_args * ap)554 vop_nopoll(struct vop_poll_args *ap)
555 {
556
557 if (ap->a_events & ~POLLSTANDARD)
558 return (POLLNVAL);
559 return (ap->a_events & (POLLIN | POLLOUT | POLLRDNORM | POLLWRNORM));
560 }
561
562 /*
563 * Implement poll for local filesystems that support it.
564 */
565 int
vop_stdpoll(struct vop_poll_args * ap)566 vop_stdpoll(struct vop_poll_args *ap)
567 {
568 if (ap->a_events & ~POLLSTANDARD)
569 return (vn_pollrecord(ap->a_vp, ap->a_td, ap->a_events));
570 return (ap->a_events & (POLLIN | POLLOUT | POLLRDNORM | POLLWRNORM));
571 }
572
573 /*
574 * Return our mount point, as we will take charge of the writes.
575 */
576 int
vop_stdgetwritemount(struct vop_getwritemount_args * ap)577 vop_stdgetwritemount(struct vop_getwritemount_args *ap)
578 {
579 struct mount *mp;
580 struct vnode *vp;
581
582 /*
583 * Note that having a reference does not prevent forced unmount from
584 * setting ->v_mount to NULL after the lock gets released. This is of
585 * no consequence for typical consumers (most notably vn_start_write)
586 * since in this case the vnode is VIRF_DOOMED. Unmount might have
587 * progressed far enough that its completion is only delayed by the
588 * reference obtained here. The consumer only needs to concern itself
589 * with releasing it.
590 */
591 vp = ap->a_vp;
592 mp = vfs_ref_from_vp(vp);
593 *(ap->a_mpp) = mp;
594 return (0);
595 }
596
597 /*
598 * If the file system doesn't implement VOP_BMAP, then return sensible defaults:
599 * - Return the vnode's bufobj instead of any underlying device's bufobj
600 * - Calculate the physical block number as if there were equal size
601 * consecutive blocks, but
602 * - Report no contiguous runs of blocks.
603 */
604 int
vop_stdbmap(struct vop_bmap_args * ap)605 vop_stdbmap(struct vop_bmap_args *ap)
606 {
607
608 if (ap->a_bop != NULL)
609 *ap->a_bop = &ap->a_vp->v_bufobj;
610 if (ap->a_bnp != NULL)
611 *ap->a_bnp = ap->a_bn * btodb(ap->a_vp->v_mount->mnt_stat.f_iosize);
612 if (ap->a_runp != NULL)
613 *ap->a_runp = 0;
614 if (ap->a_runb != NULL)
615 *ap->a_runb = 0;
616 return (0);
617 }
618
619 int
vop_stdfsync(struct vop_fsync_args * ap)620 vop_stdfsync(struct vop_fsync_args *ap)
621 {
622
623 return (vn_fsync_buf(ap->a_vp, ap->a_waitfor));
624 }
625
626 static int
vop_stdfdatasync(struct vop_fdatasync_args * ap)627 vop_stdfdatasync(struct vop_fdatasync_args *ap)
628 {
629
630 return (VOP_FSYNC(ap->a_vp, MNT_WAIT, ap->a_td));
631 }
632
633 int
vop_stdfdatasync_buf(struct vop_fdatasync_args * ap)634 vop_stdfdatasync_buf(struct vop_fdatasync_args *ap)
635 {
636
637 return (vn_fsync_buf(ap->a_vp, MNT_WAIT));
638 }
639
640 /* XXX Needs good comment and more info in the manpage (VOP_GETPAGES(9)). */
641 int
vop_stdgetpages(struct vop_getpages_args * ap)642 vop_stdgetpages(struct vop_getpages_args *ap)
643 {
644
645 return vnode_pager_generic_getpages(ap->a_vp, ap->a_m,
646 ap->a_count, ap->a_rbehind, ap->a_rahead, NULL, NULL);
647 }
648
649 static int
vop_stdgetpages_async(struct vop_getpages_async_args * ap)650 vop_stdgetpages_async(struct vop_getpages_async_args *ap)
651 {
652 int error;
653
654 error = VOP_GETPAGES(ap->a_vp, ap->a_m, ap->a_count, ap->a_rbehind,
655 ap->a_rahead);
656 if (ap->a_iodone != NULL)
657 ap->a_iodone(ap->a_arg, ap->a_m, ap->a_count, error);
658 return (error);
659 }
660
661 int
vop_stdkqfilter(struct vop_kqfilter_args * ap)662 vop_stdkqfilter(struct vop_kqfilter_args *ap)
663 {
664 return vfs_kqfilter(ap);
665 }
666
667 /* XXX Needs good comment and more info in the manpage (VOP_PUTPAGES(9)). */
668 int
vop_stdputpages(struct vop_putpages_args * ap)669 vop_stdputpages(struct vop_putpages_args *ap)
670 {
671
672 return vnode_pager_generic_putpages(ap->a_vp, ap->a_m, ap->a_count,
673 ap->a_sync, ap->a_rtvals);
674 }
675
676 int
vop_stdvptofh(struct vop_vptofh_args * ap)677 vop_stdvptofh(struct vop_vptofh_args *ap)
678 {
679 return (EOPNOTSUPP);
680 }
681
682 int
vop_stdvptocnp(struct vop_vptocnp_args * ap)683 vop_stdvptocnp(struct vop_vptocnp_args *ap)
684 {
685 struct vnode *const vp = ap->a_vp;
686 struct vnode **const dvp = ap->a_vpp;
687 char *buf = ap->a_buf;
688 size_t *buflen = ap->a_buflen;
689 char *dirbuf;
690 int i = *buflen;
691 int error = 0, covered = 0;
692 int eofflag, flags, locked;
693 size_t dirbuflen, len;
694 off_t off;
695 ino_t fileno;
696 struct vattr va;
697 struct nameidata nd;
698 struct thread *const td = curthread;
699 struct ucred *const cred = td->td_ucred;
700 struct dirent *dp;
701 struct vnode *mvp;
702
703 if (vp->v_type != VDIR)
704 return (ENOENT);
705
706 error = VOP_GETATTR(vp, &va, cred);
707 if (error)
708 return (error);
709
710 VREF(vp);
711 locked = VOP_ISLOCKED(vp);
712 VOP_UNLOCK(vp);
713 NDINIT_ATVP(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF, UIO_SYSSPACE,
714 "..", vp);
715 flags = FREAD;
716 error = vn_open_cred(&nd, &flags, 0, VN_OPEN_NOAUDIT, cred, NULL);
717 if (error) {
718 vn_lock(vp, locked | LK_RETRY);
719 return (error);
720 }
721 NDFREE_PNBUF(&nd);
722
723 mvp = *dvp = nd.ni_vp;
724
725 if (vp->v_mount != (*dvp)->v_mount &&
726 ((*dvp)->v_vflag & VV_ROOT) &&
727 ((*dvp)->v_mount->mnt_flag & MNT_UNION)) {
728 *dvp = (*dvp)->v_mount->mnt_vnodecovered;
729 VREF(mvp);
730 VOP_UNLOCK(mvp);
731 vn_close(mvp, FREAD, cred, td);
732 VREF(*dvp);
733 vn_lock(*dvp, LK_SHARED | LK_RETRY);
734 covered = 1;
735 }
736
737 fileno = va.va_fileid;
738
739 dirbuflen = MAX(DEV_BSIZE, GENERIC_MAXDIRSIZ);
740 if (dirbuflen < va.va_blocksize)
741 dirbuflen = va.va_blocksize;
742 dirbuf = malloc(dirbuflen, M_TEMP, M_WAITOK);
743
744 if ((*dvp)->v_type != VDIR) {
745 error = ENOENT;
746 goto out;
747 }
748
749 len = 0;
750 off = 0;
751 eofflag = 0;
752
753 for (;;) {
754 /* call VOP_READDIR of parent */
755 error = vn_dir_next_dirent(*dvp, td,
756 dirbuf, dirbuflen, &dp, &len, &off, &eofflag);
757 if (error != 0)
758 goto out;
759
760 if (len == 0) {
761 error = ENOENT;
762 goto out;
763 }
764
765 if ((dp->d_type != DT_WHT) &&
766 (dp->d_fileno == fileno)) {
767 if (covered) {
768 VOP_UNLOCK(*dvp);
769 vn_lock(mvp, LK_SHARED | LK_RETRY);
770 if (dirent_exists(mvp, dp->d_name, td) == 0) {
771 error = ENOENT;
772 VOP_UNLOCK(mvp);
773 vn_lock(*dvp, LK_SHARED | LK_RETRY);
774 goto out;
775 }
776 VOP_UNLOCK(mvp);
777 vn_lock(*dvp, LK_SHARED | LK_RETRY);
778 }
779 i -= dp->d_namlen;
780
781 if (i < 0) {
782 error = ENOMEM;
783 goto out;
784 }
785 if (dp->d_namlen == 1 && dp->d_name[0] == '.') {
786 error = ENOENT;
787 } else {
788 bcopy(dp->d_name, buf + i, dp->d_namlen);
789 error = 0;
790 }
791 goto out;
792 }
793 }
794
795 out:
796 free(dirbuf, M_TEMP);
797 if (!error) {
798 *buflen = i;
799 vref(*dvp);
800 }
801 if (covered) {
802 vput(*dvp);
803 vrele(mvp);
804 } else {
805 VOP_UNLOCK(mvp);
806 vn_close(mvp, FREAD, cred, td);
807 }
808 vn_lock(vp, locked | LK_RETRY);
809 return (error);
810 }
811
812 int
vop_stdallocate(struct vop_allocate_args * ap)813 vop_stdallocate(struct vop_allocate_args *ap)
814 {
815 #ifdef __notyet__
816 struct statfs *sfs;
817 off_t maxfilesize = 0;
818 #endif
819 struct iovec aiov;
820 struct vattr vattr, *vap;
821 struct uio auio;
822 off_t fsize, len, cur, offset;
823 uint8_t *buf;
824 struct thread *td;
825 struct vnode *vp;
826 size_t iosize;
827 int error;
828
829 buf = NULL;
830 error = 0;
831 td = curthread;
832 vap = &vattr;
833 vp = ap->a_vp;
834 len = *ap->a_len;
835 offset = *ap->a_offset;
836
837 error = VOP_GETATTR(vp, vap, ap->a_cred);
838 if (error != 0)
839 goto out;
840 fsize = vap->va_size;
841 iosize = vap->va_blocksize;
842 if (iosize == 0)
843 iosize = BLKDEV_IOSIZE;
844 if (iosize > maxphys)
845 iosize = maxphys;
846 buf = malloc(iosize, M_TEMP, M_WAITOK);
847
848 #ifdef __notyet__
849 /*
850 * Check if the filesystem sets f_maxfilesize; if not use
851 * VOP_SETATTR to perform the check.
852 */
853 sfs = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
854 error = VFS_STATFS(vp->v_mount, sfs, td);
855 if (error == 0)
856 maxfilesize = sfs->f_maxfilesize;
857 free(sfs, M_STATFS);
858 if (error != 0)
859 goto out;
860 if (maxfilesize) {
861 if (offset > maxfilesize || len > maxfilesize ||
862 offset + len > maxfilesize) {
863 error = EFBIG;
864 goto out;
865 }
866 } else
867 #endif
868 if (offset + len > vap->va_size) {
869 /*
870 * Test offset + len against the filesystem's maxfilesize.
871 */
872 VATTR_NULL(vap);
873 vap->va_size = offset + len;
874 error = VOP_SETATTR(vp, vap, ap->a_cred);
875 if (error != 0)
876 goto out;
877 VATTR_NULL(vap);
878 vap->va_size = fsize;
879 error = VOP_SETATTR(vp, vap, ap->a_cred);
880 if (error != 0)
881 goto out;
882 }
883
884 for (;;) {
885 /*
886 * Read and write back anything below the nominal file
887 * size. There's currently no way outside the filesystem
888 * to know whether this area is sparse or not.
889 */
890 cur = iosize;
891 if ((offset % iosize) != 0)
892 cur -= (offset % iosize);
893 if (cur > len)
894 cur = len;
895 if (offset < fsize) {
896 aiov.iov_base = buf;
897 aiov.iov_len = cur;
898 auio.uio_iov = &aiov;
899 auio.uio_iovcnt = 1;
900 auio.uio_offset = offset;
901 auio.uio_resid = cur;
902 auio.uio_segflg = UIO_SYSSPACE;
903 auio.uio_rw = UIO_READ;
904 auio.uio_td = td;
905 error = VOP_READ(vp, &auio, ap->a_ioflag, ap->a_cred);
906 if (error != 0)
907 break;
908 if (auio.uio_resid > 0) {
909 bzero(buf + cur - auio.uio_resid,
910 auio.uio_resid);
911 }
912 } else {
913 bzero(buf, cur);
914 }
915
916 aiov.iov_base = buf;
917 aiov.iov_len = cur;
918 auio.uio_iov = &aiov;
919 auio.uio_iovcnt = 1;
920 auio.uio_offset = offset;
921 auio.uio_resid = cur;
922 auio.uio_segflg = UIO_SYSSPACE;
923 auio.uio_rw = UIO_WRITE;
924 auio.uio_td = td;
925
926 error = VOP_WRITE(vp, &auio, ap->a_ioflag, ap->a_cred);
927 if (error != 0)
928 break;
929
930 len -= cur;
931 offset += cur;
932 if (len == 0)
933 break;
934 if (should_yield())
935 break;
936 }
937
938 out:
939 *ap->a_len = len;
940 *ap->a_offset = offset;
941 free(buf, M_TEMP);
942 return (error);
943 }
944
945 static int
vp_zerofill(struct vnode * vp,struct vattr * vap,off_t * offsetp,off_t * lenp,int ioflag,struct ucred * cred)946 vp_zerofill(struct vnode *vp, struct vattr *vap, off_t *offsetp, off_t *lenp,
947 int ioflag, struct ucred *cred)
948 {
949 int iosize;
950 int error = 0;
951 struct iovec aiov;
952 struct uio auio;
953 struct thread *td;
954 off_t offset, len;
955
956 iosize = vap->va_blocksize;
957 td = curthread;
958 offset = *offsetp;
959 len = *lenp;
960
961 if (iosize == 0)
962 iosize = BLKDEV_IOSIZE;
963 /* If va_blocksize is 512 bytes, iosize will be 4 kilobytes */
964 iosize = min(iosize * 8, ZERO_REGION_SIZE);
965
966 while (len > 0) {
967 int xfersize = iosize;
968 if (offset % iosize != 0)
969 xfersize -= offset % iosize;
970 if (xfersize > len)
971 xfersize = len;
972
973 aiov.iov_base = __DECONST(void *, zero_region);
974 aiov.iov_len = xfersize;
975 auio.uio_iov = &aiov;
976 auio.uio_iovcnt = 1;
977 auio.uio_offset = offset;
978 auio.uio_resid = xfersize;
979 auio.uio_segflg = UIO_SYSSPACE;
980 auio.uio_rw = UIO_WRITE;
981 auio.uio_td = td;
982
983 error = VOP_WRITE(vp, &auio, ioflag, cred);
984 if (error != 0) {
985 len -= xfersize - auio.uio_resid;
986 offset += xfersize - auio.uio_resid;
987 break;
988 }
989
990 len -= xfersize;
991 offset += xfersize;
992 }
993
994 *offsetp = offset;
995 *lenp = len;
996 return (error);
997 }
998
999 int
vop_stddeallocate(struct vop_deallocate_args * ap)1000 vop_stddeallocate(struct vop_deallocate_args *ap)
1001 {
1002 struct vnode *vp;
1003 off_t offset, len;
1004 struct ucred *cred;
1005 int error;
1006 struct vattr va;
1007 off_t noff, xfersize, rem;
1008
1009 vp = ap->a_vp;
1010 offset = *ap->a_offset;
1011 cred = ap->a_cred;
1012
1013 error = VOP_GETATTR(vp, &va, cred);
1014 if (error)
1015 return (error);
1016
1017 len = omin((off_t)va.va_size - offset, *ap->a_len);
1018 while (len > 0) {
1019 noff = offset;
1020 error = vn_bmap_seekhole_locked(vp, FIOSEEKDATA, &noff, cred);
1021 if (error) {
1022 if (error != ENXIO)
1023 /* XXX: Is it okay to fallback further? */
1024 goto out;
1025
1026 /*
1027 * No more data region to be filled
1028 */
1029 offset += len;
1030 len = 0;
1031 error = 0;
1032 break;
1033 }
1034 KASSERT(noff >= offset, ("FIOSEEKDATA going backward"));
1035 if (noff != offset) {
1036 xfersize = omin(noff - offset, len);
1037 len -= xfersize;
1038 offset += xfersize;
1039 if (len == 0)
1040 break;
1041 }
1042 error = vn_bmap_seekhole_locked(vp, FIOSEEKHOLE, &noff, cred);
1043 if (error)
1044 goto out;
1045
1046 /* Fill zeroes */
1047 xfersize = rem = omin(noff - offset, len);
1048 error = vp_zerofill(vp, &va, &offset, &rem, ap->a_ioflag, cred);
1049 if (error) {
1050 len -= xfersize - rem;
1051 goto out;
1052 }
1053
1054 len -= xfersize;
1055 if (should_yield())
1056 break;
1057 }
1058 /* Handle the case when offset is beyond EOF */
1059 if (len < 0)
1060 len = 0;
1061 out:
1062 *ap->a_offset = offset;
1063 *ap->a_len = len;
1064 return (error);
1065 }
1066
1067 int
vop_stdadvise(struct vop_advise_args * ap)1068 vop_stdadvise(struct vop_advise_args *ap)
1069 {
1070 struct vnode *vp;
1071 struct bufobj *bo;
1072 uintmax_t bstart, bend;
1073 daddr_t startn, endn;
1074 int bsize, error;
1075
1076 vp = ap->a_vp;
1077 switch (ap->a_advice) {
1078 case POSIX_FADV_WILLNEED:
1079 /*
1080 * Do nothing for now. Filesystems should provide a
1081 * custom method which starts an asynchronous read of
1082 * the requested region.
1083 */
1084 error = 0;
1085 break;
1086 case POSIX_FADV_DONTNEED:
1087 error = 0;
1088 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1089 if (VN_IS_DOOMED(vp)) {
1090 VOP_UNLOCK(vp);
1091 break;
1092 }
1093
1094 /*
1095 * Round to block boundaries (and later possibly further to
1096 * page boundaries). Applications cannot reasonably be aware
1097 * of the boundaries, and the rounding must be to expand at
1098 * both extremities to cover enough. It still doesn't cover
1099 * read-ahead. For partial blocks, this gives unnecessary
1100 * discarding of buffers but is efficient enough since the
1101 * pages usually remain in VMIO for some time.
1102 */
1103 bsize = vp->v_bufobj.bo_bsize;
1104 bstart = rounddown(ap->a_start, bsize);
1105 bend = ap->a_end;
1106 bend = roundup(bend, bsize);
1107
1108 /*
1109 * Deactivate pages in the specified range from the backing VM
1110 * object. Pages that are resident in the buffer cache will
1111 * remain wired until their corresponding buffers are released
1112 * below.
1113 */
1114 if (vp->v_object != NULL) {
1115 VM_OBJECT_RLOCK(vp->v_object);
1116 vm_object_page_noreuse(vp->v_object,
1117 OFF_TO_IDX(trunc_page(bstart)),
1118 OFF_TO_IDX(round_page(bend)));
1119 VM_OBJECT_RUNLOCK(vp->v_object);
1120 }
1121
1122 bo = &vp->v_bufobj;
1123 startn = bstart / bsize;
1124 endn = bend / bsize;
1125 BO_RLOCK(bo);
1126 error = bnoreuselist(&bo->bo_clean, bo, startn, endn);
1127 if (error == 0)
1128 error = bnoreuselist(&bo->bo_dirty, bo, startn, endn);
1129 BO_RUNLOCK(bo);
1130 VOP_UNLOCK(vp);
1131 break;
1132 default:
1133 error = EINVAL;
1134 break;
1135 }
1136 return (error);
1137 }
1138
1139 int
vop_stdunp_bind(struct vop_unp_bind_args * ap)1140 vop_stdunp_bind(struct vop_unp_bind_args *ap)
1141 {
1142
1143 ap->a_vp->v_unpcb = ap->a_unpcb;
1144 return (0);
1145 }
1146
1147 int
vop_stdunp_connect(struct vop_unp_connect_args * ap)1148 vop_stdunp_connect(struct vop_unp_connect_args *ap)
1149 {
1150
1151 *ap->a_unpcb = ap->a_vp->v_unpcb;
1152 return (0);
1153 }
1154
1155 int
vop_stdunp_detach(struct vop_unp_detach_args * ap)1156 vop_stdunp_detach(struct vop_unp_detach_args *ap)
1157 {
1158
1159 ap->a_vp->v_unpcb = NULL;
1160 return (0);
1161 }
1162
1163 static int
vop_stdis_text(struct vop_is_text_args * ap)1164 vop_stdis_text(struct vop_is_text_args *ap)
1165 {
1166
1167 return ((int)atomic_load_int(&ap->a_vp->v_writecount) < 0);
1168 }
1169
1170 int
vop_stdset_text(struct vop_set_text_args * ap)1171 vop_stdset_text(struct vop_set_text_args *ap)
1172 {
1173 struct vnode *vp;
1174 int n;
1175 bool gotref;
1176
1177 vp = ap->a_vp;
1178
1179 n = atomic_load_int(&vp->v_writecount);
1180 for (;;) {
1181 if (__predict_false(n > 0)) {
1182 return (ETXTBSY);
1183 }
1184
1185 /*
1186 * Transition point, we may need to grab a reference on the vnode.
1187 *
1188 * Take the ref early As a safety measure against bogus calls
1189 * to vop_stdunset_text.
1190 */
1191 if (n == 0) {
1192 gotref = false;
1193 if ((vn_irflag_read(vp) & VIRF_TEXT_REF) != 0) {
1194 vref(vp);
1195 gotref = true;
1196 }
1197 if (atomic_fcmpset_int(&vp->v_writecount, &n, -1)) {
1198 return (0);
1199 }
1200 if (gotref) {
1201 vunref(vp);
1202 }
1203 continue;
1204 }
1205
1206 MPASS(n < 0);
1207 if (atomic_fcmpset_int(&vp->v_writecount, &n, n - 1)) {
1208 return (0);
1209 }
1210 }
1211 __assert_unreachable();
1212 }
1213
1214 static int
vop_stdunset_text(struct vop_unset_text_args * ap)1215 vop_stdunset_text(struct vop_unset_text_args *ap)
1216 {
1217 struct vnode *vp;
1218 int n;
1219
1220 vp = ap->a_vp;
1221
1222 n = atomic_load_int(&vp->v_writecount);
1223 for (;;) {
1224 if (__predict_false(n >= 0)) {
1225 return (EINVAL);
1226 }
1227
1228 /*
1229 * Transition point, we may need to release a reference on the vnode.
1230 */
1231 if (n == -1) {
1232 if (atomic_fcmpset_int(&vp->v_writecount, &n, 0)) {
1233 if ((vn_irflag_read(vp) & VIRF_TEXT_REF) != 0) {
1234 vunref(vp);
1235 }
1236 return (0);
1237 }
1238 continue;
1239 }
1240
1241 MPASS(n < -1);
1242 if (atomic_fcmpset_int(&vp->v_writecount, &n, n + 1)) {
1243 return (0);
1244 }
1245 }
1246 __assert_unreachable();
1247 }
1248
1249 static __always_inline int
vop_stdadd_writecount_impl(struct vop_add_writecount_args * ap,bool handle_msync)1250 vop_stdadd_writecount_impl(struct vop_add_writecount_args *ap, bool handle_msync)
1251 {
1252 struct vnode *vp;
1253 struct mount *mp __diagused;
1254 int n;
1255
1256 vp = ap->a_vp;
1257
1258 #ifdef INVARIANTS
1259 mp = vp->v_mount;
1260 if (mp != NULL) {
1261 if (handle_msync) {
1262 VNPASS((mp->mnt_kern_flag & MNTK_NOMSYNC) == 0, vp);
1263 } else {
1264 VNPASS((mp->mnt_kern_flag & MNTK_NOMSYNC) != 0, vp);
1265 }
1266 }
1267 #endif
1268
1269 n = atomic_load_int(&vp->v_writecount);
1270 for (;;) {
1271 if (__predict_false(n < 0)) {
1272 return (ETXTBSY);
1273 }
1274
1275 VNASSERT(n + ap->a_inc >= 0, vp,
1276 ("neg writecount increment %d + %d = %d", n, ap->a_inc,
1277 n + ap->a_inc));
1278 if (n == 0) {
1279 if (handle_msync) {
1280 vlazy(vp);
1281 }
1282 }
1283
1284 if (atomic_fcmpset_int(&vp->v_writecount, &n, n + ap->a_inc)) {
1285 return (0);
1286 }
1287 }
1288 __assert_unreachable();
1289 }
1290
1291 int
vop_stdadd_writecount(struct vop_add_writecount_args * ap)1292 vop_stdadd_writecount(struct vop_add_writecount_args *ap)
1293 {
1294
1295 return (vop_stdadd_writecount_impl(ap, true));
1296 }
1297
1298 int
vop_stdadd_writecount_nomsync(struct vop_add_writecount_args * ap)1299 vop_stdadd_writecount_nomsync(struct vop_add_writecount_args *ap)
1300 {
1301
1302 return (vop_stdadd_writecount_impl(ap, false));
1303 }
1304
1305 int
vop_stdneed_inactive(struct vop_need_inactive_args * ap)1306 vop_stdneed_inactive(struct vop_need_inactive_args *ap)
1307 {
1308
1309 return (1);
1310 }
1311
1312 int
vop_stdinotify(struct vop_inotify_args * ap)1313 vop_stdinotify(struct vop_inotify_args *ap)
1314 {
1315 vn_inotify(ap->a_vp, ap->a_dvp, ap->a_cnp, ap->a_event, ap->a_cookie);
1316 return (0);
1317 }
1318
1319 int
vop_stdinotify_add_watch(struct vop_inotify_add_watch_args * ap)1320 vop_stdinotify_add_watch(struct vop_inotify_add_watch_args *ap)
1321 {
1322 return (vn_inotify_add_watch(ap->a_vp, ap->a_sc, ap->a_mask,
1323 ap->a_wdp, ap->a_td));
1324 }
1325
1326 int
vop_stdioctl(struct vop_ioctl_args * ap)1327 vop_stdioctl(struct vop_ioctl_args *ap)
1328 {
1329 struct vnode *vp;
1330 struct vattr va;
1331 off_t *offp;
1332 int error;
1333
1334 switch (ap->a_command) {
1335 case FIOSEEKDATA:
1336 case FIOSEEKHOLE:
1337 vp = ap->a_vp;
1338 error = vn_lock(vp, LK_SHARED);
1339 if (error != 0)
1340 return (EBADF);
1341 if (vp->v_type == VREG)
1342 error = VOP_GETATTR(vp, &va, ap->a_cred);
1343 else
1344 error = ENOTTY;
1345 if (error == 0) {
1346 offp = ap->a_data;
1347 if (*offp < 0 || *offp >= va.va_size)
1348 error = ENXIO;
1349 else if (ap->a_command == FIOSEEKHOLE)
1350 *offp = va.va_size;
1351 }
1352 VOP_UNLOCK(vp);
1353 break;
1354 default:
1355 error = ENOTTY;
1356 break;
1357 }
1358 return (error);
1359 }
1360
1361 /*
1362 * vfs default ops
1363 * used to fill the vfs function table to get reasonable default return values.
1364 */
1365 int
vfs_stdroot(struct mount * mp,int flags,struct vnode ** vpp)1366 vfs_stdroot(struct mount *mp, int flags, struct vnode **vpp)
1367 {
1368
1369 return (EOPNOTSUPP);
1370 }
1371
1372 int
vfs_stdstatfs(struct mount * mp,struct statfs * sbp)1373 vfs_stdstatfs(struct mount *mp, struct statfs *sbp)
1374 {
1375
1376 return (EOPNOTSUPP);
1377 }
1378
1379 int
vfs_stdquotactl(struct mount * mp,int cmds,uid_t uid,void * arg,bool * mp_busy)1380 vfs_stdquotactl(struct mount *mp, int cmds, uid_t uid, void *arg, bool *mp_busy)
1381 {
1382 return (EOPNOTSUPP);
1383 }
1384
1385 int
vfs_stdsync(struct mount * mp,int waitfor)1386 vfs_stdsync(struct mount *mp, int waitfor)
1387 {
1388 struct vnode *vp, *mvp;
1389 struct thread *td;
1390 int error, lockreq, allerror = 0;
1391
1392 td = curthread;
1393 lockreq = LK_EXCLUSIVE | LK_INTERLOCK;
1394 if (waitfor != MNT_WAIT)
1395 lockreq |= LK_NOWAIT;
1396 /*
1397 * Force stale buffer cache information to be flushed.
1398 */
1399 loop:
1400 MNT_VNODE_FOREACH_ALL(vp, mp, mvp) {
1401 if (vp->v_bufobj.bo_dirty.bv_cnt == 0) {
1402 VI_UNLOCK(vp);
1403 continue;
1404 }
1405 if ((error = vget(vp, lockreq)) != 0) {
1406 if (error == ENOENT) {
1407 MNT_VNODE_FOREACH_ALL_ABORT(mp, mvp);
1408 goto loop;
1409 }
1410 continue;
1411 }
1412 error = VOP_FSYNC(vp, waitfor, td);
1413 if (error)
1414 allerror = error;
1415 vput(vp);
1416 }
1417 return (allerror);
1418 }
1419
1420 int
vfs_stdnosync(struct mount * mp,int waitfor)1421 vfs_stdnosync(struct mount *mp, int waitfor)
1422 {
1423
1424 return (0);
1425 }
1426
1427 static int
vop_stdcopy_file_range(struct vop_copy_file_range_args * ap)1428 vop_stdcopy_file_range(struct vop_copy_file_range_args *ap)
1429 {
1430 int error;
1431
1432 error = vn_generic_copy_file_range(ap->a_invp, ap->a_inoffp,
1433 ap->a_outvp, ap->a_outoffp, ap->a_lenp, ap->a_flags, ap->a_incred,
1434 ap->a_outcred, ap->a_fsizetd);
1435 return (error);
1436 }
1437
1438 int
vfs_stdvget(struct mount * mp,ino_t ino,int flags,struct vnode ** vpp)1439 vfs_stdvget(struct mount *mp, ino_t ino, int flags, struct vnode **vpp)
1440 {
1441
1442 return (EOPNOTSUPP);
1443 }
1444
1445 int
vfs_stdfhtovp(struct mount * mp,struct fid * fhp,int flags,struct vnode ** vpp)1446 vfs_stdfhtovp(struct mount *mp, struct fid *fhp, int flags, struct vnode **vpp)
1447 {
1448
1449 return (EOPNOTSUPP);
1450 }
1451
1452 int
vfs_stdinit(struct vfsconf * vfsp)1453 vfs_stdinit(struct vfsconf *vfsp)
1454 {
1455
1456 return (0);
1457 }
1458
1459 int
vfs_stduninit(struct vfsconf * vfsp)1460 vfs_stduninit(struct vfsconf *vfsp)
1461 {
1462
1463 return(0);
1464 }
1465
1466 int
vfs_stdextattrctl(struct mount * mp,int cmd,struct vnode * filename_vp,int attrnamespace,const char * attrname)1467 vfs_stdextattrctl(struct mount *mp, int cmd, struct vnode *filename_vp,
1468 int attrnamespace, const char *attrname)
1469 {
1470
1471 if (filename_vp != NULL)
1472 VOP_UNLOCK(filename_vp);
1473 return (EOPNOTSUPP);
1474 }
1475
1476 int
vfs_stdsysctl(struct mount * mp,fsctlop_t op,struct sysctl_req * req)1477 vfs_stdsysctl(struct mount *mp, fsctlop_t op, struct sysctl_req *req)
1478 {
1479
1480 return (EOPNOTSUPP);
1481 }
1482
1483 static vop_bypass_t *
bp_by_off(struct vop_vector * vop,struct vop_generic_args * a)1484 bp_by_off(struct vop_vector *vop, struct vop_generic_args *a)
1485 {
1486
1487 return (*(vop_bypass_t **)((char *)vop + a->a_desc->vdesc_vop_offset));
1488 }
1489
1490 int
vop_sigdefer(struct vop_vector * vop,struct vop_generic_args * a)1491 vop_sigdefer(struct vop_vector *vop, struct vop_generic_args *a)
1492 {
1493 vop_bypass_t *bp;
1494 int prev_stops, rc;
1495
1496 bp = bp_by_off(vop, a);
1497 MPASS(bp != NULL);
1498
1499 prev_stops = sigdeferstop(SIGDEFERSTOP_SILENT);
1500 rc = bp(a);
1501 sigallowstop(prev_stops);
1502 return (rc);
1503 }
1504
1505 static int
vop_stdstat(struct vop_stat_args * a)1506 vop_stdstat(struct vop_stat_args *a)
1507 {
1508 struct vattr vattr;
1509 struct vattr *vap;
1510 struct vnode *vp;
1511 struct stat *sb;
1512 int error;
1513 u_short mode;
1514
1515 vp = a->a_vp;
1516 sb = a->a_sb;
1517
1518 error = vop_stat_helper_pre(a);
1519 if (error != 0)
1520 return (error);
1521
1522 vap = &vattr;
1523
1524 /*
1525 * Initialize defaults for new and unusual fields, so that file
1526 * systems which don't support these fields don't need to know
1527 * about them.
1528 */
1529 vap->va_birthtime.tv_sec = -1;
1530 vap->va_birthtime.tv_nsec = 0;
1531 vap->va_fsid = VNOVAL;
1532 vap->va_gen = 0;
1533 vap->va_rdev = NODEV;
1534 vap->va_filerev = 0;
1535 vap->va_bsdflags = 0;
1536
1537 error = VOP_GETATTR(vp, vap, a->a_active_cred);
1538 if (error)
1539 goto out;
1540
1541 /*
1542 * Zero the spare stat fields
1543 */
1544 bzero(sb, sizeof *sb);
1545
1546 /*
1547 * Copy from vattr table
1548 */
1549 if (vap->va_fsid != VNOVAL)
1550 sb->st_dev = vap->va_fsid;
1551 else
1552 sb->st_dev = vp->v_mount->mnt_stat.f_fsid.val[0];
1553 sb->st_ino = vap->va_fileid;
1554 mode = vap->va_mode;
1555 switch (vap->va_type) {
1556 case VREG:
1557 mode |= S_IFREG;
1558 break;
1559 case VDIR:
1560 mode |= S_IFDIR;
1561 break;
1562 case VBLK:
1563 mode |= S_IFBLK;
1564 break;
1565 case VCHR:
1566 mode |= S_IFCHR;
1567 break;
1568 case VLNK:
1569 mode |= S_IFLNK;
1570 break;
1571 case VSOCK:
1572 mode |= S_IFSOCK;
1573 break;
1574 case VFIFO:
1575 mode |= S_IFIFO;
1576 break;
1577 default:
1578 error = EBADF;
1579 goto out;
1580 }
1581 sb->st_mode = mode;
1582 sb->st_nlink = vap->va_nlink;
1583 sb->st_uid = vap->va_uid;
1584 sb->st_gid = vap->va_gid;
1585 sb->st_rdev = vap->va_rdev;
1586 if (vap->va_size > OFF_MAX) {
1587 error = EOVERFLOW;
1588 goto out;
1589 }
1590 sb->st_size = vap->va_size;
1591 sb->st_atim.tv_sec = vap->va_atime.tv_sec;
1592 sb->st_atim.tv_nsec = vap->va_atime.tv_nsec;
1593 sb->st_mtim.tv_sec = vap->va_mtime.tv_sec;
1594 sb->st_mtim.tv_nsec = vap->va_mtime.tv_nsec;
1595 sb->st_ctim.tv_sec = vap->va_ctime.tv_sec;
1596 sb->st_ctim.tv_nsec = vap->va_ctime.tv_nsec;
1597 sb->st_birthtim.tv_sec = vap->va_birthtime.tv_sec;
1598 sb->st_birthtim.tv_nsec = vap->va_birthtime.tv_nsec;
1599
1600 /*
1601 * According to www.opengroup.org, the meaning of st_blksize is
1602 * "a filesystem-specific preferred I/O block size for this
1603 * object. In some filesystem types, this may vary from file
1604 * to file"
1605 * Use minimum/default of PAGE_SIZE (e.g. for VCHR).
1606 */
1607
1608 sb->st_blksize = max(PAGE_SIZE, vap->va_blocksize);
1609 sb->st_flags = vap->va_flags;
1610 sb->st_blocks = vap->va_bytes / S_BLKSIZE;
1611 sb->st_gen = vap->va_gen;
1612 sb->st_filerev = vap->va_filerev;
1613 sb->st_bsdflags = vap->va_bsdflags;
1614 out:
1615 return (vop_stat_helper_post(a, error));
1616 }
1617
1618 static int
vop_stdread_pgcache(struct vop_read_pgcache_args * ap __unused)1619 vop_stdread_pgcache(struct vop_read_pgcache_args *ap __unused)
1620 {
1621 return (EJUSTRETURN);
1622 }
1623
1624 static int
vop_stdvput_pair(struct vop_vput_pair_args * ap)1625 vop_stdvput_pair(struct vop_vput_pair_args *ap)
1626 {
1627 struct vnode *dvp, *vp, **vpp;
1628
1629 dvp = ap->a_dvp;
1630 vpp = ap->a_vpp;
1631 vput(dvp);
1632 if (vpp != NULL && ap->a_unlock_vp && (vp = *vpp) != NULL)
1633 vput(vp);
1634 return (0);
1635 }
1636
1637 static int
vop_stdgetlowvnode(struct vop_getlowvnode_args * ap)1638 vop_stdgetlowvnode(struct vop_getlowvnode_args *ap)
1639 {
1640 vref(ap->a_vp);
1641 *ap->a_vplp = ap->a_vp;
1642 return (0);
1643 }
1644