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_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_inotify = vop_stdinotify,
123 .vop_inotify_add_watch = vop_stdinotify_add_watch,
124 .vop_ioctl = vop_stdioctl,
125 .vop_kqfilter = vop_stdkqfilter,
126 .vop_islocked = vop_stdislocked,
127 .vop_lock1 = vop_stdlock,
128 .vop_lookup = vop_nolookup,
129 .vop_open = VOP_NULL,
130 .vop_pathconf = VOP_EINVAL,
131 .vop_poll = vop_nopoll,
132 .vop_putpages = vop_stdputpages,
133 .vop_readlink = VOP_EINVAL,
134 .vop_read_pgcache = vop_stdread_pgcache,
135 .vop_rename = vop_norename,
136 .vop_revoke = VOP_PANIC,
137 .vop_strategy = vop_nostrategy,
138 .vop_unlock = vop_stdunlock,
139 .vop_vptocnp = vop_stdvptocnp,
140 .vop_vptofh = vop_stdvptofh,
141 .vop_unp_bind = vop_stdunp_bind,
142 .vop_unp_connect = vop_stdunp_connect,
143 .vop_unp_detach = vop_stdunp_detach,
144 .vop_is_text = vop_stdis_text,
145 .vop_set_text = vop_stdset_text,
146 .vop_unset_text = vop_stdunset_text,
147 .vop_add_writecount = vop_stdadd_writecount,
148 .vop_copy_file_range = vop_stdcopy_file_range,
149 .vop_vput_pair = vop_stdvput_pair,
150 };
151 VFS_VOP_VECTOR_REGISTER(default_vnodeops);
152
153 /*
154 * Series of placeholder functions for various error returns for
155 * VOPs.
156 */
157
158 int
vop_eopnotsupp(struct vop_generic_args * ap)159 vop_eopnotsupp(struct vop_generic_args *ap)
160 {
161 /*
162 printf("vop_notsupp[%s]\n", ap->a_desc->vdesc_name);
163 */
164
165 return (EOPNOTSUPP);
166 }
167
168 int
vop_ebadf(struct vop_generic_args * ap)169 vop_ebadf(struct vop_generic_args *ap)
170 {
171
172 return (EBADF);
173 }
174
175 int
vop_enotty(struct vop_generic_args * ap)176 vop_enotty(struct vop_generic_args *ap)
177 {
178
179 return (ENOTTY);
180 }
181
182 int
vop_einval(struct vop_generic_args * ap)183 vop_einval(struct vop_generic_args *ap)
184 {
185
186 return (EINVAL);
187 }
188
189 int
vop_enoent(struct vop_generic_args * ap)190 vop_enoent(struct vop_generic_args *ap)
191 {
192
193 return (ENOENT);
194 }
195
196 int
vop_eagain(struct vop_generic_args * ap)197 vop_eagain(struct vop_generic_args *ap)
198 {
199
200 return (EAGAIN);
201 }
202
203 int
vop_null(struct vop_generic_args * ap)204 vop_null(struct vop_generic_args *ap)
205 {
206
207 return (0);
208 }
209
210 /*
211 * Helper function to panic on some bad VOPs in some filesystems.
212 */
213 int
vop_panic(struct vop_generic_args * ap)214 vop_panic(struct vop_generic_args *ap)
215 {
216
217 panic("filesystem goof: vop_panic[%s]", ap->a_desc->vdesc_name);
218 }
219
220 /*
221 * vop_std<something> and vop_no<something> are default functions for use by
222 * filesystems that need the "default reasonable" implementation for a
223 * particular operation.
224 *
225 * The documentation for the operations they implement exists (if it exists)
226 * in the VOP_<SOMETHING>(9) manpage (all uppercase).
227 */
228
229 /*
230 * Default vop for filesystems that do not support name lookup
231 */
232 static int
vop_nolookup(struct vop_lookup_args * ap)233 vop_nolookup(struct vop_lookup_args *ap)
234 {
235
236 *ap->a_vpp = NULL;
237 return (ENOTDIR);
238 }
239
240 /*
241 * vop_norename:
242 *
243 * Handle unlock and reference counting for arguments of vop_rename
244 * for filesystems that do not implement rename operation.
245 */
246 static int
vop_norename(struct vop_rename_args * ap)247 vop_norename(struct vop_rename_args *ap)
248 {
249
250 vop_rename_fail(ap);
251 return (EOPNOTSUPP);
252 }
253
254 /*
255 * vop_nostrategy:
256 *
257 * Strategy routine for VFS devices that have none.
258 *
259 * BIO_ERROR and B_INVAL must be cleared prior to calling any strategy
260 * routine. Typically this is done for a BIO_READ strategy call.
261 * Typically B_INVAL is assumed to already be clear prior to a write
262 * and should not be cleared manually unless you just made the buffer
263 * invalid. BIO_ERROR should be cleared either way.
264 */
265
266 static int
vop_nostrategy(struct vop_strategy_args * ap)267 vop_nostrategy (struct vop_strategy_args *ap)
268 {
269 printf("No strategy for buffer at %p\n", ap->a_bp);
270 vn_printf(ap->a_vp, "vnode ");
271 ap->a_bp->b_ioflags |= BIO_ERROR;
272 ap->a_bp->b_error = EOPNOTSUPP;
273 bufdone(ap->a_bp);
274 return (EOPNOTSUPP);
275 }
276
277 /*
278 * Check if a named file exists in a given directory vnode
279 *
280 * Returns 0 if the file exists, ENOENT if it doesn't, or errors returned by
281 * vn_dir_next_dirent().
282 */
283 static int
dirent_exists(struct vnode * vp,const char * dirname,struct thread * td)284 dirent_exists(struct vnode *vp, const char *dirname, struct thread *td)
285 {
286 char *dirbuf;
287 int error, eofflag;
288 size_t dirbuflen, len;
289 off_t off;
290 struct dirent *dp;
291 struct vattr va;
292
293 ASSERT_VOP_LOCKED(vp, "vnode not locked");
294 KASSERT(vp->v_type == VDIR, ("vp %p is not a directory", vp));
295
296 error = VOP_GETATTR(vp, &va, td->td_ucred);
297 if (error != 0)
298 return (error);
299
300 dirbuflen = MAX(DEV_BSIZE, GENERIC_MAXDIRSIZ);
301 if (dirbuflen < va.va_blocksize)
302 dirbuflen = va.va_blocksize;
303 dirbuf = malloc(dirbuflen, M_TEMP, M_WAITOK);
304
305 len = 0;
306 off = 0;
307 eofflag = 0;
308
309 for (;;) {
310 error = vn_dir_next_dirent(vp, td, dirbuf, dirbuflen,
311 &dp, &len, &off, &eofflag);
312 if (error != 0)
313 goto out;
314
315 if (len == 0)
316 break;
317
318 if (dp->d_type != DT_WHT && dp->d_fileno != 0 &&
319 strcmp(dp->d_name, dirname) == 0)
320 goto out;
321 }
322
323 error = ENOENT;
324
325 out:
326 free(dirbuf, M_TEMP);
327 return (error);
328 }
329
330 int
vop_stdaccess(struct vop_access_args * ap)331 vop_stdaccess(struct vop_access_args *ap)
332 {
333
334 KASSERT((ap->a_accmode & ~(VEXEC | VWRITE | VREAD | VADMIN |
335 VAPPEND)) == 0, ("invalid bit in accmode"));
336
337 return (VOP_ACCESSX(ap->a_vp, ap->a_accmode, ap->a_cred, ap->a_td));
338 }
339
340 int
vop_stdaccessx(struct vop_accessx_args * ap)341 vop_stdaccessx(struct vop_accessx_args *ap)
342 {
343 int error;
344 accmode_t accmode = ap->a_accmode;
345
346 error = vfs_unixify_accmode(&accmode);
347 if (error != 0)
348 return (error);
349
350 if (accmode == 0)
351 return (0);
352
353 return (VOP_ACCESS(ap->a_vp, accmode, ap->a_cred, ap->a_td));
354 }
355
356 /*
357 * Advisory record locking support
358 */
359 int
vop_stdadvlock(struct vop_advlock_args * ap)360 vop_stdadvlock(struct vop_advlock_args *ap)
361 {
362 struct vnode *vp;
363 struct mount *mp;
364 struct vattr vattr;
365 int error;
366
367 vp = ap->a_vp;
368
369 /*
370 * Provide atomicity of open(O_CREAT | O_EXCL | O_EXLOCK) for
371 * local filesystems. See vn_open_cred() for reciprocal part.
372 */
373 mp = vp->v_mount;
374 if (mp != NULL && (mp->mnt_flag & MNT_LOCAL) != 0 &&
375 ap->a_op == F_SETLK && (ap->a_flags & F_FIRSTOPEN) == 0) {
376 VI_LOCK(vp);
377 while ((vp->v_iflag & VI_FOPENING) != 0)
378 msleep(vp, VI_MTX(vp), PLOCK, "lockfo", 0);
379 VI_UNLOCK(vp);
380 }
381
382 if (ap->a_fl->l_whence == SEEK_END) {
383 /*
384 * The NFSv4 server must avoid doing a vn_lock() here, since it
385 * can deadlock the nfsd threads, due to a LOR. Fortunately
386 * the NFSv4 server always uses SEEK_SET and this code is
387 * only required for the SEEK_END case.
388 */
389 vn_lock(vp, LK_SHARED | LK_RETRY);
390 error = VOP_GETATTR(vp, &vattr, curthread->td_ucred);
391 VOP_UNLOCK(vp);
392 if (error)
393 return (error);
394 } else
395 vattr.va_size = 0;
396
397 return (lf_advlock(ap, &(vp->v_lockf), vattr.va_size));
398 }
399
400 int
vop_stdadvlockasync(struct vop_advlockasync_args * ap)401 vop_stdadvlockasync(struct vop_advlockasync_args *ap)
402 {
403 struct vnode *vp;
404 struct vattr vattr;
405 int error;
406
407 vp = ap->a_vp;
408 if (ap->a_fl->l_whence == SEEK_END) {
409 /* The size argument is only needed for SEEK_END. */
410 vn_lock(vp, LK_SHARED | LK_RETRY);
411 error = VOP_GETATTR(vp, &vattr, curthread->td_ucred);
412 VOP_UNLOCK(vp);
413 if (error)
414 return (error);
415 } else
416 vattr.va_size = 0;
417
418 return (lf_advlockasync(ap, &(vp->v_lockf), vattr.va_size));
419 }
420
421 int
vop_stdadvlockpurge(struct vop_advlockpurge_args * ap)422 vop_stdadvlockpurge(struct vop_advlockpurge_args *ap)
423 {
424 struct vnode *vp;
425
426 vp = ap->a_vp;
427 lf_purgelocks(vp, &vp->v_lockf);
428 return (0);
429 }
430
431 /*
432 * vop_stdpathconf:
433 *
434 * Standard implementation of POSIX pathconf, to get information about limits
435 * for a filesystem.
436 * Override per filesystem for the case where the filesystem has smaller
437 * limits.
438 */
439 int
vop_stdpathconf(struct vop_pathconf_args * ap)440 vop_stdpathconf(struct vop_pathconf_args *ap)
441 {
442
443 switch (ap->a_name) {
444 case _PC_ASYNC_IO:
445 *ap->a_retval = _POSIX_ASYNCHRONOUS_IO;
446 return (0);
447 case _PC_PATH_MAX:
448 *ap->a_retval = PATH_MAX;
449 return (0);
450 case _PC_ACL_EXTENDED:
451 case _PC_ACL_NFS4:
452 case _PC_CAP_PRESENT:
453 case _PC_DEALLOC_PRESENT:
454 case _PC_INF_PRESENT:
455 case _PC_MAC_PRESENT:
456 case _PC_NAMEDATTR_ENABLED:
457 case _PC_HAS_NAMEDATTR:
458 case _PC_HAS_HIDDENSYSTEM:
459 case _PC_CLONE_BLKSIZE:
460 case _PC_CASE_INSENSITIVE:
461 *ap->a_retval = 0;
462 return (0);
463 default:
464 return (EINVAL);
465 }
466 /* NOTREACHED */
467 }
468
469 /*
470 * Standard lock, unlock and islocked functions.
471 */
472 int
vop_stdlock(struct vop_lock1_args * ap)473 vop_stdlock(struct vop_lock1_args *ap)
474 {
475 struct vnode *vp = ap->a_vp;
476 struct mtx *ilk;
477
478 ilk = VI_MTX(vp);
479 return (lockmgr_lock_flags(vp->v_vnlock, ap->a_flags,
480 &ilk->lock_object, ap->a_file, ap->a_line));
481 }
482
483 /* See above. */
484 int
vop_stdunlock(struct vop_unlock_args * ap)485 vop_stdunlock(struct vop_unlock_args *ap)
486 {
487 struct vnode *vp = ap->a_vp;
488
489 return (lockmgr_unlock(vp->v_vnlock));
490 }
491
492 /* See above. */
493 int
vop_stdislocked(struct vop_islocked_args * ap)494 vop_stdislocked(struct vop_islocked_args *ap)
495 {
496
497 return (lockstatus(ap->a_vp->v_vnlock));
498 }
499
500 /*
501 * Variants of the above set.
502 *
503 * Differences are:
504 * - shared locking disablement is not supported
505 * - v_vnlock pointer is not honored
506 */
507 int
vop_lock(struct vop_lock1_args * ap)508 vop_lock(struct vop_lock1_args *ap)
509 {
510 struct vnode *vp = ap->a_vp;
511 int flags = ap->a_flags;
512 struct mtx *ilk;
513
514 MPASS(vp->v_vnlock == &vp->v_lock);
515
516 if (__predict_false((flags & ~(LK_TYPE_MASK | LK_NODDLKTREAT | LK_RETRY)) != 0))
517 goto other;
518
519 switch (flags & LK_TYPE_MASK) {
520 case LK_SHARED:
521 return (lockmgr_slock(&vp->v_lock, flags, ap->a_file, ap->a_line));
522 case LK_EXCLUSIVE:
523 return (lockmgr_xlock(&vp->v_lock, flags, ap->a_file, ap->a_line));
524 }
525 other:
526 ilk = VI_MTX(vp);
527 return (lockmgr_lock_flags(&vp->v_lock, flags,
528 &ilk->lock_object, ap->a_file, ap->a_line));
529 }
530
531 int
vop_unlock(struct vop_unlock_args * ap)532 vop_unlock(struct vop_unlock_args *ap)
533 {
534 struct vnode *vp = ap->a_vp;
535
536 MPASS(vp->v_vnlock == &vp->v_lock);
537
538 return (lockmgr_unlock(&vp->v_lock));
539 }
540
541 int
vop_islocked(struct vop_islocked_args * ap)542 vop_islocked(struct vop_islocked_args *ap)
543 {
544 struct vnode *vp = ap->a_vp;
545
546 MPASS(vp->v_vnlock == &vp->v_lock);
547
548 return (lockstatus(&vp->v_lock));
549 }
550
551 /*
552 * Return true for select/poll.
553 */
554 int
vop_nopoll(struct vop_poll_args * ap)555 vop_nopoll(struct vop_poll_args *ap)
556 {
557
558 if (ap->a_events & ~POLLSTANDARD)
559 return (POLLNVAL);
560 return (ap->a_events & (POLLIN | POLLOUT | POLLRDNORM | POLLWRNORM));
561 }
562
563 /*
564 * Implement poll for local filesystems that support it.
565 */
566 int
vop_stdpoll(struct vop_poll_args * ap)567 vop_stdpoll(struct vop_poll_args *ap)
568 {
569 if (ap->a_events & ~POLLSTANDARD)
570 return (vn_pollrecord(ap->a_vp, ap->a_td, ap->a_events));
571 return (ap->a_events & (POLLIN | POLLOUT | POLLRDNORM | POLLWRNORM));
572 }
573
574 /*
575 * Return our mount point, as we will take charge of the writes.
576 */
577 int
vop_stdgetwritemount(struct vop_getwritemount_args * ap)578 vop_stdgetwritemount(struct vop_getwritemount_args *ap)
579 {
580 struct mount *mp;
581 struct vnode *vp;
582
583 /*
584 * Note that having a reference does not prevent forced unmount from
585 * setting ->v_mount to NULL after the lock gets released. This is of
586 * no consequence for typical consumers (most notably vn_start_write)
587 * since in this case the vnode is VIRF_DOOMED. Unmount might have
588 * progressed far enough that its completion is only delayed by the
589 * reference obtained here. The consumer only needs to concern itself
590 * with releasing it.
591 */
592 vp = ap->a_vp;
593 mp = vfs_ref_from_vp(vp);
594 *(ap->a_mpp) = mp;
595 return (0);
596 }
597
598 /*
599 * If the file system doesn't implement VOP_BMAP, then return sensible defaults:
600 * - Return the vnode's bufobj instead of any underlying device's bufobj
601 * - Calculate the physical block number as if there were equal size
602 * consecutive blocks, but
603 * - Report no contiguous runs of blocks.
604 */
605 int
vop_stdbmap(struct vop_bmap_args * ap)606 vop_stdbmap(struct vop_bmap_args *ap)
607 {
608
609 if (ap->a_bop != NULL)
610 *ap->a_bop = &ap->a_vp->v_bufobj;
611 if (ap->a_bnp != NULL)
612 *ap->a_bnp = ap->a_bn * btodb(ap->a_vp->v_mount->mnt_stat.f_iosize);
613 if (ap->a_runp != NULL)
614 *ap->a_runp = 0;
615 if (ap->a_runb != NULL)
616 *ap->a_runb = 0;
617 return (0);
618 }
619
620 int
vop_stdfsync(struct vop_fsync_args * ap)621 vop_stdfsync(struct vop_fsync_args *ap)
622 {
623
624 return (vn_fsync_buf(ap->a_vp, ap->a_waitfor));
625 }
626
627 static int
vop_stdfdatasync(struct vop_fdatasync_args * ap)628 vop_stdfdatasync(struct vop_fdatasync_args *ap)
629 {
630
631 return (VOP_FSYNC(ap->a_vp, MNT_WAIT, ap->a_td));
632 }
633
634 int
vop_stdfdatasync_buf(struct vop_fdatasync_args * ap)635 vop_stdfdatasync_buf(struct vop_fdatasync_args *ap)
636 {
637
638 return (vn_fsync_buf(ap->a_vp, MNT_WAIT));
639 }
640
641 /* XXX Needs good comment and more info in the manpage (VOP_GETPAGES(9)). */
642 int
vop_stdgetpages(struct vop_getpages_args * ap)643 vop_stdgetpages(struct vop_getpages_args *ap)
644 {
645
646 return vnode_pager_generic_getpages(ap->a_vp, ap->a_m,
647 ap->a_count, ap->a_rbehind, ap->a_rahead, NULL, NULL);
648 }
649
650 static int
vop_stdgetpages_async(struct vop_getpages_async_args * ap)651 vop_stdgetpages_async(struct vop_getpages_async_args *ap)
652 {
653 int error;
654
655 error = VOP_GETPAGES(ap->a_vp, ap->a_m, ap->a_count, ap->a_rbehind,
656 ap->a_rahead);
657 if (ap->a_iodone != NULL)
658 ap->a_iodone(ap->a_arg, ap->a_m, ap->a_count, error);
659 return (error);
660 }
661
662 int
vop_stdkqfilter(struct vop_kqfilter_args * ap)663 vop_stdkqfilter(struct vop_kqfilter_args *ap)
664 {
665 return vfs_kqfilter(ap);
666 }
667
668 /* XXX Needs good comment and more info in the manpage (VOP_PUTPAGES(9)). */
669 int
vop_stdputpages(struct vop_putpages_args * ap)670 vop_stdputpages(struct vop_putpages_args *ap)
671 {
672
673 return vnode_pager_generic_putpages(ap->a_vp, ap->a_m, ap->a_count,
674 ap->a_sync, ap->a_rtvals);
675 }
676
677 int
vop_stdvptofh(struct vop_vptofh_args * ap)678 vop_stdvptofh(struct vop_vptofh_args *ap)
679 {
680 return (EOPNOTSUPP);
681 }
682
683 int
vop_stdvptocnp(struct vop_vptocnp_args * ap)684 vop_stdvptocnp(struct vop_vptocnp_args *ap)
685 {
686 struct vnode *const vp = ap->a_vp;
687 struct vnode **const dvp = ap->a_vpp;
688 char *buf = ap->a_buf;
689 size_t *buflen = ap->a_buflen;
690 char *dirbuf;
691 int i = *buflen;
692 int error = 0, covered = 0;
693 int eofflag, flags, locked;
694 size_t dirbuflen, len;
695 off_t off;
696 ino_t fileno;
697 struct vattr va;
698 struct nameidata nd;
699 struct thread *const td = curthread;
700 struct ucred *const cred = td->td_ucred;
701 struct dirent *dp;
702 struct vnode *mvp;
703
704 if (vp->v_type != VDIR)
705 return (ENOENT);
706
707 error = VOP_GETATTR(vp, &va, cred);
708 if (error)
709 return (error);
710
711 vref(vp);
712 locked = VOP_ISLOCKED(vp);
713 VOP_UNLOCK(vp);
714 NDINIT_ATVP(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF, UIO_SYSSPACE,
715 "..", vp);
716 flags = FREAD;
717 error = vn_open_cred(&nd, &flags, 0, VN_OPEN_NOAUDIT, cred, NULL);
718 if (error) {
719 vn_lock(vp, locked | LK_RETRY);
720 return (error);
721 }
722 NDFREE_PNBUF(&nd);
723
724 mvp = *dvp = nd.ni_vp;
725
726 if (vp->v_mount != (*dvp)->v_mount &&
727 ((*dvp)->v_vflag & VV_ROOT) &&
728 ((*dvp)->v_mount->mnt_flag & MNT_UNION)) {
729 *dvp = (*dvp)->v_mount->mnt_vnodecovered;
730 vref(mvp);
731 VOP_UNLOCK(mvp);
732 vn_close(mvp, FREAD, cred, td);
733 vref(*dvp);
734 vn_lock(*dvp, LK_SHARED | LK_RETRY);
735 covered = 1;
736 }
737
738 fileno = va.va_fileid;
739
740 dirbuflen = MAX(DEV_BSIZE, GENERIC_MAXDIRSIZ);
741 if (dirbuflen < va.va_blocksize)
742 dirbuflen = va.va_blocksize;
743 dirbuf = malloc(dirbuflen, M_TEMP, M_WAITOK);
744
745 if ((*dvp)->v_type != VDIR) {
746 error = ENOENT;
747 goto out;
748 }
749
750 len = 0;
751 off = 0;
752 eofflag = 0;
753
754 for (;;) {
755 /* call VOP_READDIR of parent */
756 error = vn_dir_next_dirent(*dvp, td,
757 dirbuf, dirbuflen, &dp, &len, &off, &eofflag);
758 if (error != 0)
759 goto out;
760
761 if (len == 0) {
762 error = ENOENT;
763 goto out;
764 }
765
766 if ((dp->d_type != DT_WHT) &&
767 (dp->d_fileno == fileno)) {
768 if (covered) {
769 VOP_UNLOCK(*dvp);
770 vn_lock(mvp, LK_SHARED | LK_RETRY);
771 if (dirent_exists(mvp, dp->d_name, td) == 0) {
772 error = ENOENT;
773 VOP_UNLOCK(mvp);
774 vn_lock(*dvp, LK_SHARED | LK_RETRY);
775 goto out;
776 }
777 VOP_UNLOCK(mvp);
778 vn_lock(*dvp, LK_SHARED | LK_RETRY);
779 }
780 i -= dp->d_namlen;
781
782 if (i < 0) {
783 error = ENOMEM;
784 goto out;
785 }
786 if (dp->d_namlen == 1 && dp->d_name[0] == '.') {
787 error = ENOENT;
788 } else {
789 bcopy(dp->d_name, buf + i, dp->d_namlen);
790 error = 0;
791 }
792 goto out;
793 }
794 }
795
796 out:
797 free(dirbuf, M_TEMP);
798 if (!error) {
799 *buflen = i;
800 vref(*dvp);
801 }
802 if (covered) {
803 vput(*dvp);
804 vrele(mvp);
805 } else {
806 VOP_UNLOCK(mvp);
807 vn_close(mvp, FREAD, cred, td);
808 }
809 vn_lock(vp, locked | LK_RETRY);
810 return (error);
811 }
812
813 int
vop_stdallocate(struct vop_allocate_args * ap)814 vop_stdallocate(struct vop_allocate_args *ap)
815 {
816 #ifdef __notyet__
817 struct statfs *sfs;
818 off_t maxfilesize = 0;
819 #endif
820 struct iovec aiov;
821 struct vattr vattr, *vap;
822 struct uio auio;
823 off_t fsize, len, cur, offset;
824 uint8_t *buf;
825 struct thread *td;
826 struct vnode *vp;
827 size_t iosize;
828 int error;
829
830 buf = NULL;
831 error = 0;
832 td = curthread;
833 vap = &vattr;
834 vp = ap->a_vp;
835 len = *ap->a_len;
836 offset = *ap->a_offset;
837
838 error = VOP_GETATTR(vp, vap, ap->a_cred);
839 if (error != 0)
840 goto out;
841 fsize = vap->va_size;
842 iosize = vap->va_blocksize;
843 if (iosize == 0)
844 iosize = BLKDEV_IOSIZE;
845 if (iosize > maxphys)
846 iosize = maxphys;
847 buf = malloc(iosize, M_TEMP, M_WAITOK);
848
849 #ifdef __notyet__
850 /*
851 * Check if the filesystem sets f_maxfilesize; if not use
852 * VOP_SETATTR to perform the check.
853 */
854 sfs = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
855 error = VFS_STATFS(vp->v_mount, sfs, td);
856 if (error == 0)
857 maxfilesize = sfs->f_maxfilesize;
858 free(sfs, M_STATFS);
859 if (error != 0)
860 goto out;
861 if (maxfilesize) {
862 if (offset > maxfilesize || len > maxfilesize ||
863 offset + len > maxfilesize) {
864 error = EFBIG;
865 goto out;
866 }
867 } else
868 #endif
869 if (offset + len > vap->va_size) {
870 /*
871 * Test offset + len against the filesystem's maxfilesize.
872 */
873 VATTR_NULL(vap);
874 vap->va_size = offset + len;
875 error = VOP_SETATTR(vp, vap, ap->a_cred);
876 if (error != 0)
877 goto out;
878 VATTR_NULL(vap);
879 vap->va_size = fsize;
880 error = VOP_SETATTR(vp, vap, ap->a_cred);
881 if (error != 0)
882 goto out;
883 }
884
885 for (;;) {
886 /*
887 * Read and write back anything below the nominal file
888 * size. There's currently no way outside the filesystem
889 * to know whether this area is sparse or not.
890 */
891 cur = iosize;
892 if ((offset % iosize) != 0)
893 cur -= (offset % iosize);
894 if (cur > len)
895 cur = len;
896 if (offset < fsize) {
897 aiov.iov_base = buf;
898 aiov.iov_len = cur;
899 auio.uio_iov = &aiov;
900 auio.uio_iovcnt = 1;
901 auio.uio_offset = offset;
902 auio.uio_resid = cur;
903 auio.uio_segflg = UIO_SYSSPACE;
904 auio.uio_rw = UIO_READ;
905 auio.uio_td = td;
906 error = VOP_READ(vp, &auio, ap->a_ioflag, ap->a_cred);
907 if (error != 0)
908 break;
909 if (auio.uio_resid > 0) {
910 bzero(buf + cur - auio.uio_resid,
911 auio.uio_resid);
912 }
913 } else {
914 bzero(buf, cur);
915 }
916
917 aiov.iov_base = buf;
918 aiov.iov_len = cur;
919 auio.uio_iov = &aiov;
920 auio.uio_iovcnt = 1;
921 auio.uio_offset = offset;
922 auio.uio_resid = cur;
923 auio.uio_segflg = UIO_SYSSPACE;
924 auio.uio_rw = UIO_WRITE;
925 auio.uio_td = td;
926
927 error = VOP_WRITE(vp, &auio, ap->a_ioflag, ap->a_cred);
928 if (error != 0)
929 break;
930
931 len -= cur;
932 offset += cur;
933 if (len == 0)
934 break;
935 if (should_yield())
936 break;
937 }
938
939 out:
940 *ap->a_len = len;
941 *ap->a_offset = offset;
942 free(buf, M_TEMP);
943 return (error);
944 }
945
946 static int
vp_zerofill(struct vnode * vp,struct vattr * vap,off_t * offsetp,off_t * lenp,int ioflag,struct ucred * cred)947 vp_zerofill(struct vnode *vp, struct vattr *vap, off_t *offsetp, off_t *lenp,
948 int ioflag, struct ucred *cred)
949 {
950 int iosize;
951 int error = 0;
952 struct iovec aiov;
953 struct uio auio;
954 struct thread *td;
955 off_t offset, len;
956
957 iosize = vap->va_blocksize;
958 td = curthread;
959 offset = *offsetp;
960 len = *lenp;
961
962 if (iosize == 0)
963 iosize = BLKDEV_IOSIZE;
964 /* If va_blocksize is 512 bytes, iosize will be 4 kilobytes */
965 iosize = min(iosize * 8, ZERO_REGION_SIZE);
966
967 while (len > 0) {
968 int xfersize = iosize;
969 if (offset % iosize != 0)
970 xfersize -= offset % iosize;
971 if (xfersize > len)
972 xfersize = len;
973
974 aiov.iov_base = __DECONST(void *, zero_region);
975 aiov.iov_len = xfersize;
976 auio.uio_iov = &aiov;
977 auio.uio_iovcnt = 1;
978 auio.uio_offset = offset;
979 auio.uio_resid = xfersize;
980 auio.uio_segflg = UIO_SYSSPACE;
981 auio.uio_rw = UIO_WRITE;
982 auio.uio_td = td;
983
984 error = VOP_WRITE(vp, &auio, ioflag, cred);
985 if (error != 0) {
986 len -= xfersize - auio.uio_resid;
987 offset += xfersize - auio.uio_resid;
988 break;
989 }
990
991 len -= xfersize;
992 offset += xfersize;
993 }
994
995 *offsetp = offset;
996 *lenp = len;
997 return (error);
998 }
999
1000 int
vop_stddeallocate(struct vop_deallocate_args * ap)1001 vop_stddeallocate(struct vop_deallocate_args *ap)
1002 {
1003 struct vnode *vp;
1004 off_t offset, len;
1005 struct ucred *cred;
1006 int error;
1007 struct vattr va;
1008 off_t noff, xfersize, rem;
1009
1010 vp = ap->a_vp;
1011 offset = *ap->a_offset;
1012 cred = ap->a_cred;
1013
1014 error = VOP_GETATTR(vp, &va, cred);
1015 if (error)
1016 return (error);
1017
1018 len = omin((off_t)va.va_size - offset, *ap->a_len);
1019 while (len > 0) {
1020 noff = offset;
1021 error = vn_bmap_seekhole_locked(vp, FIOSEEKDATA, &noff, cred);
1022 if (error) {
1023 if (error != ENXIO)
1024 /* XXX: Is it okay to fallback further? */
1025 goto out;
1026
1027 /*
1028 * No more data region to be filled
1029 */
1030 offset += len;
1031 len = 0;
1032 error = 0;
1033 break;
1034 }
1035 KASSERT(noff >= offset, ("FIOSEEKDATA going backward"));
1036 if (noff != offset) {
1037 xfersize = omin(noff - offset, len);
1038 len -= xfersize;
1039 offset += xfersize;
1040 if (len == 0)
1041 break;
1042 }
1043 error = vn_bmap_seekhole_locked(vp, FIOSEEKHOLE, &noff, cred);
1044 if (error)
1045 goto out;
1046
1047 /* Fill zeroes */
1048 xfersize = rem = omin(noff - offset, len);
1049 error = vp_zerofill(vp, &va, &offset, &rem, ap->a_ioflag, cred);
1050 if (error) {
1051 len -= xfersize - rem;
1052 goto out;
1053 }
1054
1055 len -= xfersize;
1056 if (should_yield())
1057 break;
1058 }
1059 /* Handle the case when offset is beyond EOF */
1060 if (len < 0)
1061 len = 0;
1062 out:
1063 *ap->a_offset = offset;
1064 *ap->a_len = len;
1065 return (error);
1066 }
1067
1068 int
vop_stdadvise(struct vop_advise_args * ap)1069 vop_stdadvise(struct vop_advise_args *ap)
1070 {
1071 struct vnode *vp;
1072 struct bufobj *bo;
1073 uintmax_t bstart, bend;
1074 daddr_t startn, endn;
1075 int bsize, error;
1076
1077 vp = ap->a_vp;
1078 switch (ap->a_advice) {
1079 case POSIX_FADV_WILLNEED:
1080 /*
1081 * Do nothing for now. Filesystems should provide a
1082 * custom method which starts an asynchronous read of
1083 * the requested region.
1084 */
1085 error = 0;
1086 break;
1087 case POSIX_FADV_DONTNEED:
1088 error = 0;
1089 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1090 if (VN_IS_DOOMED(vp)) {
1091 VOP_UNLOCK(vp);
1092 break;
1093 }
1094
1095 /*
1096 * Round to block boundaries (and later possibly further to
1097 * page boundaries). Applications cannot reasonably be aware
1098 * of the boundaries, and the rounding must be to expand at
1099 * both extremities to cover enough. It still doesn't cover
1100 * read-ahead. For partial blocks, this gives unnecessary
1101 * discarding of buffers but is efficient enough since the
1102 * pages usually remain in VMIO for some time.
1103 */
1104 bsize = vp->v_bufobj.bo_bsize;
1105 bstart = rounddown(ap->a_start, bsize);
1106 bend = ap->a_end;
1107 bend = roundup(bend, bsize);
1108
1109 /*
1110 * Deactivate pages in the specified range from the backing VM
1111 * object. Pages that are resident in the buffer cache will
1112 * remain wired until their corresponding buffers are released
1113 * below.
1114 */
1115 if (vp->v_object != NULL) {
1116 VM_OBJECT_RLOCK(vp->v_object);
1117 vm_object_page_noreuse(vp->v_object,
1118 OFF_TO_IDX(trunc_page(bstart)),
1119 OFF_TO_IDX(round_page(bend)));
1120 VM_OBJECT_RUNLOCK(vp->v_object);
1121 }
1122
1123 bo = &vp->v_bufobj;
1124 startn = bstart / bsize;
1125 endn = bend / bsize;
1126 BO_RLOCK(bo);
1127 error = bnoreuselist(&bo->bo_clean, bo, startn, endn);
1128 if (error == 0)
1129 error = bnoreuselist(&bo->bo_dirty, bo, startn, endn);
1130 BO_RUNLOCK(bo);
1131 VOP_UNLOCK(vp);
1132 break;
1133 default:
1134 error = EINVAL;
1135 break;
1136 }
1137 return (error);
1138 }
1139
1140 int
vop_stdunp_bind(struct vop_unp_bind_args * ap)1141 vop_stdunp_bind(struct vop_unp_bind_args *ap)
1142 {
1143
1144 ap->a_vp->v_unpcb = ap->a_unpcb;
1145 return (0);
1146 }
1147
1148 int
vop_stdunp_connect(struct vop_unp_connect_args * ap)1149 vop_stdunp_connect(struct vop_unp_connect_args *ap)
1150 {
1151
1152 *ap->a_unpcb = ap->a_vp->v_unpcb;
1153 return (0);
1154 }
1155
1156 int
vop_stdunp_detach(struct vop_unp_detach_args * ap)1157 vop_stdunp_detach(struct vop_unp_detach_args *ap)
1158 {
1159
1160 ap->a_vp->v_unpcb = NULL;
1161 return (0);
1162 }
1163
1164 static int
vop_stdis_text(struct vop_is_text_args * ap)1165 vop_stdis_text(struct vop_is_text_args *ap)
1166 {
1167
1168 return ((int)atomic_load_int(&ap->a_vp->v_writecount) < 0);
1169 }
1170
1171 int
vop_stdset_text(struct vop_set_text_args * ap)1172 vop_stdset_text(struct vop_set_text_args *ap)
1173 {
1174 struct vnode *vp;
1175 int n;
1176 bool gotref;
1177
1178 vp = ap->a_vp;
1179
1180 n = atomic_load_int(&vp->v_writecount);
1181 for (;;) {
1182 if (__predict_false(n > 0)) {
1183 return (ETXTBSY);
1184 }
1185
1186 /*
1187 * Transition point, we may need to grab a reference on the vnode.
1188 *
1189 * Take the ref early As a safety measure against bogus calls
1190 * to vop_stdunset_text.
1191 */
1192 if (n == 0) {
1193 gotref = false;
1194 if ((vn_irflag_read(vp) & VIRF_TEXT_REF) != 0) {
1195 vref(vp);
1196 gotref = true;
1197 }
1198 if (atomic_fcmpset_int(&vp->v_writecount, &n, -1)) {
1199 return (0);
1200 }
1201 if (gotref) {
1202 vunref(vp);
1203 }
1204 continue;
1205 }
1206
1207 MPASS(n < 0);
1208 if (atomic_fcmpset_int(&vp->v_writecount, &n, n - 1)) {
1209 return (0);
1210 }
1211 }
1212 __assert_unreachable();
1213 }
1214
1215 static int
vop_stdunset_text(struct vop_unset_text_args * ap)1216 vop_stdunset_text(struct vop_unset_text_args *ap)
1217 {
1218 struct vnode *vp;
1219 int n;
1220
1221 vp = ap->a_vp;
1222
1223 n = atomic_load_int(&vp->v_writecount);
1224 for (;;) {
1225 if (__predict_false(n >= 0)) {
1226 return (EINVAL);
1227 }
1228
1229 /*
1230 * Transition point, we may need to release a reference on the vnode.
1231 */
1232 if (n == -1) {
1233 if (atomic_fcmpset_int(&vp->v_writecount, &n, 0)) {
1234 if ((vn_irflag_read(vp) & VIRF_TEXT_REF) != 0) {
1235 vunref(vp);
1236 }
1237 return (0);
1238 }
1239 continue;
1240 }
1241
1242 MPASS(n < -1);
1243 if (atomic_fcmpset_int(&vp->v_writecount, &n, n + 1)) {
1244 return (0);
1245 }
1246 }
1247 __assert_unreachable();
1248 }
1249
1250 static __always_inline int
vop_stdadd_writecount_impl(struct vop_add_writecount_args * ap,bool handle_msync)1251 vop_stdadd_writecount_impl(struct vop_add_writecount_args *ap, bool handle_msync)
1252 {
1253 struct vnode *vp;
1254 struct mount *mp __diagused;
1255 int n;
1256
1257 vp = ap->a_vp;
1258
1259 #ifdef INVARIANTS
1260 mp = vp->v_mount;
1261 if (mp != NULL) {
1262 if (handle_msync) {
1263 VNPASS((mp->mnt_kern_flag & MNTK_NOMSYNC) == 0, vp);
1264 } else {
1265 VNPASS((mp->mnt_kern_flag & MNTK_NOMSYNC) != 0, vp);
1266 }
1267 }
1268 #endif
1269
1270 n = atomic_load_int(&vp->v_writecount);
1271 for (;;) {
1272 if (__predict_false(n < 0)) {
1273 return (ETXTBSY);
1274 }
1275
1276 VNASSERT(n + ap->a_inc >= 0, vp,
1277 ("neg writecount increment %d + %d = %d", n, ap->a_inc,
1278 n + ap->a_inc));
1279 if (n == 0) {
1280 if (handle_msync) {
1281 vlazy(vp);
1282 }
1283 }
1284
1285 if (atomic_fcmpset_int(&vp->v_writecount, &n, n + ap->a_inc)) {
1286 return (0);
1287 }
1288 }
1289 __assert_unreachable();
1290 }
1291
1292 int
vop_stdadd_writecount(struct vop_add_writecount_args * ap)1293 vop_stdadd_writecount(struct vop_add_writecount_args *ap)
1294 {
1295
1296 return (vop_stdadd_writecount_impl(ap, true));
1297 }
1298
1299 int
vop_stdadd_writecount_nomsync(struct vop_add_writecount_args * ap)1300 vop_stdadd_writecount_nomsync(struct vop_add_writecount_args *ap)
1301 {
1302
1303 return (vop_stdadd_writecount_impl(ap, false));
1304 }
1305
1306 int
vop_stdneed_inactive(struct vop_need_inactive_args * ap)1307 vop_stdneed_inactive(struct vop_need_inactive_args *ap)
1308 {
1309
1310 return (1);
1311 }
1312
1313 int
vop_stdinotify(struct vop_inotify_args * ap)1314 vop_stdinotify(struct vop_inotify_args *ap)
1315 {
1316 vn_inotify(ap->a_vp, ap->a_dvp, ap->a_cnp, ap->a_event, ap->a_cookie);
1317 return (0);
1318 }
1319
1320 int
vop_stdinotify_add_watch(struct vop_inotify_add_watch_args * ap)1321 vop_stdinotify_add_watch(struct vop_inotify_add_watch_args *ap)
1322 {
1323 return (vn_inotify_add_watch(ap->a_vp, ap->a_sc, ap->a_mask,
1324 ap->a_wdp, ap->a_td));
1325 }
1326
1327 int
vop_stdioctl(struct vop_ioctl_args * ap)1328 vop_stdioctl(struct vop_ioctl_args *ap)
1329 {
1330 struct vnode *vp;
1331 struct vattr va;
1332 off_t *offp;
1333 int error;
1334
1335 switch (ap->a_command) {
1336 case FIOSEEKDATA:
1337 case FIOSEEKHOLE:
1338 vp = ap->a_vp;
1339 error = vn_lock(vp, LK_SHARED);
1340 if (error != 0)
1341 return (EBADF);
1342 if (vp->v_type == VREG)
1343 error = VOP_GETATTR(vp, &va, ap->a_cred);
1344 else
1345 error = ENOTTY;
1346 if (error == 0) {
1347 offp = ap->a_data;
1348 if (*offp < 0 || *offp >= va.va_size)
1349 error = ENXIO;
1350 else if (ap->a_command == FIOSEEKHOLE)
1351 *offp = va.va_size;
1352 }
1353 VOP_UNLOCK(vp);
1354 break;
1355 default:
1356 error = ENOTTY;
1357 break;
1358 }
1359 return (error);
1360 }
1361
1362 /*
1363 * vfs default ops
1364 * used to fill the vfs function table to get reasonable default return values.
1365 */
1366 int
vfs_stdroot(struct mount * mp,int flags,struct vnode ** vpp)1367 vfs_stdroot(struct mount *mp, int flags, struct vnode **vpp)
1368 {
1369
1370 return (EOPNOTSUPP);
1371 }
1372
1373 int
vfs_stdstatfs(struct mount * mp,struct statfs * sbp)1374 vfs_stdstatfs(struct mount *mp, struct statfs *sbp)
1375 {
1376
1377 return (EOPNOTSUPP);
1378 }
1379
1380 int
vfs_stdquotactl(struct mount * mp,int cmds,uid_t uid,void * arg,bool * mp_busy)1381 vfs_stdquotactl(struct mount *mp, int cmds, uid_t uid, void *arg, bool *mp_busy)
1382 {
1383 return (EOPNOTSUPP);
1384 }
1385
1386 int
vfs_stdsync(struct mount * mp,int waitfor)1387 vfs_stdsync(struct mount *mp, int waitfor)
1388 {
1389 struct vnode *vp, *mvp;
1390 struct thread *td;
1391 int error, lockreq, allerror = 0;
1392
1393 td = curthread;
1394 lockreq = LK_EXCLUSIVE | LK_INTERLOCK;
1395 if (waitfor != MNT_WAIT)
1396 lockreq |= LK_NOWAIT;
1397 /*
1398 * Force stale buffer cache information to be flushed.
1399 */
1400 loop:
1401 MNT_VNODE_FOREACH_ALL(vp, mp, mvp) {
1402 if (vp->v_bufobj.bo_dirty.bv_cnt == 0) {
1403 VI_UNLOCK(vp);
1404 continue;
1405 }
1406 if ((error = vget(vp, lockreq)) != 0) {
1407 if (error == ENOENT) {
1408 MNT_VNODE_FOREACH_ALL_ABORT(mp, mvp);
1409 goto loop;
1410 }
1411 continue;
1412 }
1413 error = VOP_FSYNC(vp, waitfor, td);
1414 if (error)
1415 allerror = error;
1416 vput(vp);
1417 }
1418 return (allerror);
1419 }
1420
1421 int
vfs_stdnosync(struct mount * mp,int waitfor)1422 vfs_stdnosync(struct mount *mp, int waitfor)
1423 {
1424
1425 return (0);
1426 }
1427
1428 int
vop_stdcopy_file_range(struct vop_copy_file_range_args * ap)1429 vop_stdcopy_file_range(struct vop_copy_file_range_args *ap)
1430 {
1431 int error;
1432
1433 error = vn_generic_copy_file_range(ap->a_invp, ap->a_inoffp,
1434 ap->a_outvp, ap->a_outoffp, ap->a_lenp, ap->a_flags, ap->a_incred,
1435 ap->a_outcred, ap->a_fsizetd);
1436 return (error);
1437 }
1438
1439 int
vfs_stdvget(struct mount * mp,ino_t ino,int flags,struct vnode ** vpp)1440 vfs_stdvget(struct mount *mp, ino_t ino, int flags, struct vnode **vpp)
1441 {
1442
1443 return (EOPNOTSUPP);
1444 }
1445
1446 int
vfs_stdfhtovp(struct mount * mp,struct fid * fhp,int flags,struct vnode ** vpp)1447 vfs_stdfhtovp(struct mount *mp, struct fid *fhp, int flags, struct vnode **vpp)
1448 {
1449
1450 return (EOPNOTSUPP);
1451 }
1452
1453 int
vfs_stdinit(struct vfsconf * vfsp)1454 vfs_stdinit(struct vfsconf *vfsp)
1455 {
1456
1457 return (0);
1458 }
1459
1460 int
vfs_stduninit(struct vfsconf * vfsp)1461 vfs_stduninit(struct vfsconf *vfsp)
1462 {
1463
1464 return(0);
1465 }
1466
1467 int
vfs_stdextattrctl(struct mount * mp,int cmd,struct vnode * filename_vp,int attrnamespace,const char * attrname)1468 vfs_stdextattrctl(struct mount *mp, int cmd, struct vnode *filename_vp,
1469 int attrnamespace, const char *attrname)
1470 {
1471
1472 if (filename_vp != NULL)
1473 VOP_UNLOCK(filename_vp);
1474 return (EOPNOTSUPP);
1475 }
1476
1477 int
vfs_stdsysctl(struct mount * mp,fsctlop_t op,struct sysctl_req * req)1478 vfs_stdsysctl(struct mount *mp, fsctlop_t op, struct sysctl_req *req)
1479 {
1480
1481 return (EOPNOTSUPP);
1482 }
1483
1484 static vop_bypass_t *
bp_by_off(struct vop_vector * vop,struct vop_generic_args * a)1485 bp_by_off(struct vop_vector *vop, struct vop_generic_args *a)
1486 {
1487
1488 return (*(vop_bypass_t **)((char *)vop + a->a_desc->vdesc_vop_offset));
1489 }
1490
1491 int
vop_sigdefer(struct vop_vector * vop,struct vop_generic_args * a)1492 vop_sigdefer(struct vop_vector *vop, struct vop_generic_args *a)
1493 {
1494 vop_bypass_t *bp;
1495 int prev_stops, rc;
1496
1497 bp = bp_by_off(vop, a);
1498 MPASS(bp != NULL);
1499
1500 prev_stops = sigdeferstop(SIGDEFERSTOP_SILENT);
1501 rc = bp(a);
1502 sigallowstop(prev_stops);
1503 return (rc);
1504 }
1505
1506 static int
vop_stdstat(struct vop_stat_args * a)1507 vop_stdstat(struct vop_stat_args *a)
1508 {
1509 struct vattr vattr;
1510 struct vattr *vap;
1511 struct vnode *vp;
1512 struct stat *sb;
1513 int error;
1514 u_short mode;
1515
1516 vp = a->a_vp;
1517 sb = a->a_sb;
1518
1519 error = vop_stat_helper_pre(a);
1520 if (error != 0)
1521 return (error);
1522
1523 vap = &vattr;
1524
1525 /*
1526 * Initialize defaults for new and unusual fields, so that file
1527 * systems which don't support these fields don't need to know
1528 * about them.
1529 */
1530 vap->va_birthtime.tv_sec = -1;
1531 vap->va_birthtime.tv_nsec = 0;
1532 vap->va_fsid = VNOVAL;
1533 vap->va_gen = 0;
1534 vap->va_rdev = NODEV;
1535 vap->va_filerev = 0;
1536 vap->va_bsdflags = 0;
1537
1538 error = VOP_GETATTR(vp, vap, a->a_active_cred);
1539 if (error)
1540 goto out;
1541
1542 /*
1543 * Zero the spare stat fields
1544 */
1545 bzero(sb, sizeof *sb);
1546
1547 /*
1548 * Copy from vattr table
1549 */
1550 if (vap->va_fsid != VNOVAL)
1551 sb->st_dev = vap->va_fsid;
1552 else
1553 sb->st_dev = vp->v_mount->mnt_stat.f_fsid.val[0];
1554 sb->st_ino = vap->va_fileid;
1555 mode = vap->va_mode;
1556 switch (vap->va_type) {
1557 case VREG:
1558 mode |= S_IFREG;
1559 break;
1560 case VDIR:
1561 mode |= S_IFDIR;
1562 break;
1563 case VBLK:
1564 mode |= S_IFBLK;
1565 break;
1566 case VCHR:
1567 mode |= S_IFCHR;
1568 break;
1569 case VLNK:
1570 mode |= S_IFLNK;
1571 break;
1572 case VSOCK:
1573 mode |= S_IFSOCK;
1574 break;
1575 case VFIFO:
1576 mode |= S_IFIFO;
1577 break;
1578 default:
1579 error = EBADF;
1580 goto out;
1581 }
1582 sb->st_mode = mode;
1583 sb->st_nlink = vap->va_nlink;
1584 sb->st_uid = vap->va_uid;
1585 sb->st_gid = vap->va_gid;
1586 sb->st_rdev = vap->va_rdev;
1587 if (vap->va_size > OFF_MAX) {
1588 error = EOVERFLOW;
1589 goto out;
1590 }
1591 sb->st_size = vap->va_size;
1592 sb->st_atim.tv_sec = vap->va_atime.tv_sec;
1593 sb->st_atim.tv_nsec = vap->va_atime.tv_nsec;
1594 sb->st_mtim.tv_sec = vap->va_mtime.tv_sec;
1595 sb->st_mtim.tv_nsec = vap->va_mtime.tv_nsec;
1596 sb->st_ctim.tv_sec = vap->va_ctime.tv_sec;
1597 sb->st_ctim.tv_nsec = vap->va_ctime.tv_nsec;
1598 sb->st_birthtim.tv_sec = vap->va_birthtime.tv_sec;
1599 sb->st_birthtim.tv_nsec = vap->va_birthtime.tv_nsec;
1600
1601 /*
1602 * According to www.opengroup.org, the meaning of st_blksize is
1603 * "a filesystem-specific preferred I/O block size for this
1604 * object. In some filesystem types, this may vary from file
1605 * to file"
1606 * Use minimum/default of PAGE_SIZE (e.g. for VCHR).
1607 */
1608
1609 sb->st_blksize = max(PAGE_SIZE, vap->va_blocksize);
1610 sb->st_flags = vap->va_flags;
1611 sb->st_blocks = vap->va_bytes / S_BLKSIZE;
1612 sb->st_gen = vap->va_gen;
1613 sb->st_filerev = vap->va_filerev;
1614 sb->st_bsdflags = vap->va_bsdflags;
1615 out:
1616 return (vop_stat_helper_post(a, error));
1617 }
1618
1619 static int
vop_stdread_pgcache(struct vop_read_pgcache_args * ap __unused)1620 vop_stdread_pgcache(struct vop_read_pgcache_args *ap __unused)
1621 {
1622 return (EJUSTRETURN);
1623 }
1624
1625 static int
vop_stdvput_pair(struct vop_vput_pair_args * ap)1626 vop_stdvput_pair(struct vop_vput_pair_args *ap)
1627 {
1628 struct vnode *dvp, *vp, **vpp;
1629
1630 dvp = ap->a_dvp;
1631 vpp = ap->a_vpp;
1632 vput(dvp);
1633 if (vpp != NULL && ap->a_unlock_vp && (vp = *vpp) != NULL)
1634 vput(vp);
1635 return (0);
1636 }
1637
1638 static int
vop_stdgetlowvnode(struct vop_getlowvnode_args * ap)1639 vop_stdgetlowvnode(struct vop_getlowvnode_args *ap)
1640 {
1641 vref(ap->a_vp);
1642 *ap->a_vplp = ap->a_vp;
1643 return (0);
1644 }
1645