1 /*-
2 * SPDX-License-Identifier: BSD-3-Clause
3 *
4 * Copyright (c) 1989, 1993
5 * The Regents of the University of California. All rights reserved.
6 *
7 * This code is derived from software contributed
8 * to Berkeley by John Heidemann of the UCLA Ficus project.
9 *
10 * Source: * @(#)i405_init.c 2.10 92/04/27 UCLA Ficus project
11 *
12 * Redistribution and use in source and binary forms, with or without
13 * modification, are permitted provided that the following conditions
14 * are met:
15 * 1. Redistributions of source code must retain the above copyright
16 * notice, this list of conditions and the following disclaimer.
17 * 2. Redistributions in binary form must reproduce the above copyright
18 * notice, this list of conditions and the following disclaimer in the
19 * documentation and/or other materials provided with the distribution.
20 * 3. Neither the name of the University nor the names of its contributors
21 * may be used to endorse or promote products derived from this software
22 * without specific prior written permission.
23 *
24 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
25 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
28 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
29 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
30 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
31 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
32 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
33 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
34 * SUCH DAMAGE.
35 */
36
37 #include <sys/cdefs.h>
38 #include <sys/param.h>
39 #include <sys/systm.h>
40 #include <sys/bio.h>
41 #include <sys/buf.h>
42 #include <sys/conf.h>
43 #include <sys/event.h>
44 #include <sys/filio.h>
45 #include <sys/kernel.h>
46 #include <sys/limits.h>
47 #include <sys/lock.h>
48 #include <sys/lockf.h>
49 #include <sys/malloc.h>
50 #include <sys/mount.h>
51 #include <sys/namei.h>
52 #include <sys/rwlock.h>
53 #include <sys/fcntl.h>
54 #include <sys/unistd.h>
55 #include <sys/vnode.h>
56 #include <sys/dirent.h>
57 #include <sys/poll.h>
58 #include <sys/stat.h>
59 #include <security/audit/audit.h>
60 #include <sys/priv.h>
61
62 #include <security/mac/mac_framework.h>
63
64 #include <vm/vm.h>
65 #include <vm/vm_object.h>
66 #include <vm/vm_extern.h>
67 #include <vm/pmap.h>
68 #include <vm/vm_map.h>
69 #include <vm/vm_page.h>
70 #include <vm/vm_pager.h>
71 #include <vm/vnode_pager.h>
72
73 static int vop_nolookup(struct vop_lookup_args *);
74 static int vop_norename(struct vop_rename_args *);
75 static int vop_nostrategy(struct vop_strategy_args *);
76 static int dirent_exists(struct vnode *vp, const char *dirname,
77 struct thread *td);
78
79 static int vop_stdis_text(struct vop_is_text_args *ap);
80 static int vop_stdunset_text(struct vop_unset_text_args *ap);
81 static int vop_stdadd_writecount(struct vop_add_writecount_args *ap);
82 static int vop_stdcopy_file_range(struct vop_copy_file_range_args *ap);
83 static int vop_stdfdatasync(struct vop_fdatasync_args *ap);
84 static int vop_stdgetpages_async(struct vop_getpages_async_args *ap);
85 static int vop_stdread_pgcache(struct vop_read_pgcache_args *ap);
86 static int vop_stdstat(struct vop_stat_args *ap);
87 static int vop_stdvput_pair(struct vop_vput_pair_args *ap);
88 static int vop_stdgetlowvnode(struct vop_getlowvnode_args *ap);
89
90 /*
91 * This vnode table stores what we want to do if the filesystem doesn't
92 * implement a particular VOP.
93 *
94 * If there is no specific entry here, we will return EOPNOTSUPP.
95 *
96 * Note that every filesystem has to implement either vop_access
97 * or vop_accessx; failing to do so will result in immediate crash
98 * due to stack overflow, as vop_stdaccess() calls vop_stdaccessx(),
99 * which calls vop_stdaccess() etc.
100 */
101
102 struct vop_vector default_vnodeops = {
103 .vop_default = NULL,
104 .vop_bypass = VOP_EOPNOTSUPP,
105
106 .vop_access = vop_stdaccess,
107 .vop_accessx = vop_stdaccessx,
108 .vop_advise = vop_stdadvise,
109 .vop_advlock = vop_stdadvlock,
110 .vop_advlockasync = vop_stdadvlockasync,
111 .vop_advlockpurge = vop_stdadvlockpurge,
112 .vop_allocate = vop_stdallocate,
113 .vop_deallocate = vop_stddeallocate,
114 .vop_bmap = vop_stdbmap,
115 .vop_close = VOP_NULL,
116 .vop_fsync = VOP_NULL,
117 .vop_stat = vop_stdstat,
118 .vop_fdatasync = vop_stdfdatasync,
119 .vop_getlowvnode = vop_stdgetlowvnode,
120 .vop_getpages = vop_stdgetpages,
121 .vop_getpages_async = vop_stdgetpages_async,
122 .vop_getwritemount = vop_stdgetwritemount,
123 .vop_inactive = VOP_NULL,
124 .vop_need_inactive = vop_stdneed_inactive,
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 *ap->a_retval = 0;
458 return (0);
459 default:
460 return (EINVAL);
461 }
462 /* NOTREACHED */
463 }
464
465 /*
466 * Standard lock, unlock and islocked functions.
467 */
468 int
vop_stdlock(struct vop_lock1_args * ap)469 vop_stdlock(struct vop_lock1_args *ap)
470 {
471 struct vnode *vp = ap->a_vp;
472 struct mtx *ilk;
473
474 ilk = VI_MTX(vp);
475 return (lockmgr_lock_flags(vp->v_vnlock, ap->a_flags,
476 &ilk->lock_object, ap->a_file, ap->a_line));
477 }
478
479 /* See above. */
480 int
vop_stdunlock(struct vop_unlock_args * ap)481 vop_stdunlock(struct vop_unlock_args *ap)
482 {
483 struct vnode *vp = ap->a_vp;
484
485 return (lockmgr_unlock(vp->v_vnlock));
486 }
487
488 /* See above. */
489 int
vop_stdislocked(struct vop_islocked_args * ap)490 vop_stdislocked(struct vop_islocked_args *ap)
491 {
492
493 return (lockstatus(ap->a_vp->v_vnlock));
494 }
495
496 /*
497 * Variants of the above set.
498 *
499 * Differences are:
500 * - shared locking disablement is not supported
501 * - v_vnlock pointer is not honored
502 */
503 int
vop_lock(struct vop_lock1_args * ap)504 vop_lock(struct vop_lock1_args *ap)
505 {
506 struct vnode *vp = ap->a_vp;
507 int flags = ap->a_flags;
508 struct mtx *ilk;
509
510 MPASS(vp->v_vnlock == &vp->v_lock);
511
512 if (__predict_false((flags & ~(LK_TYPE_MASK | LK_NODDLKTREAT | LK_RETRY)) != 0))
513 goto other;
514
515 switch (flags & LK_TYPE_MASK) {
516 case LK_SHARED:
517 return (lockmgr_slock(&vp->v_lock, flags, ap->a_file, ap->a_line));
518 case LK_EXCLUSIVE:
519 return (lockmgr_xlock(&vp->v_lock, flags, ap->a_file, ap->a_line));
520 }
521 other:
522 ilk = VI_MTX(vp);
523 return (lockmgr_lock_flags(&vp->v_lock, flags,
524 &ilk->lock_object, ap->a_file, ap->a_line));
525 }
526
527 int
vop_unlock(struct vop_unlock_args * ap)528 vop_unlock(struct vop_unlock_args *ap)
529 {
530 struct vnode *vp = ap->a_vp;
531
532 MPASS(vp->v_vnlock == &vp->v_lock);
533
534 return (lockmgr_unlock(&vp->v_lock));
535 }
536
537 int
vop_islocked(struct vop_islocked_args * ap)538 vop_islocked(struct vop_islocked_args *ap)
539 {
540 struct vnode *vp = ap->a_vp;
541
542 MPASS(vp->v_vnlock == &vp->v_lock);
543
544 return (lockstatus(&vp->v_lock));
545 }
546
547 /*
548 * Return true for select/poll.
549 */
550 int
vop_nopoll(struct vop_poll_args * ap)551 vop_nopoll(struct vop_poll_args *ap)
552 {
553
554 if (ap->a_events & ~POLLSTANDARD)
555 return (POLLNVAL);
556 return (ap->a_events & (POLLIN | POLLOUT | POLLRDNORM | POLLWRNORM));
557 }
558
559 /*
560 * Implement poll for local filesystems that support it.
561 */
562 int
vop_stdpoll(struct vop_poll_args * ap)563 vop_stdpoll(struct vop_poll_args *ap)
564 {
565 if (ap->a_events & ~POLLSTANDARD)
566 return (vn_pollrecord(ap->a_vp, ap->a_td, ap->a_events));
567 return (ap->a_events & (POLLIN | POLLOUT | POLLRDNORM | POLLWRNORM));
568 }
569
570 /*
571 * Return our mount point, as we will take charge of the writes.
572 */
573 int
vop_stdgetwritemount(struct vop_getwritemount_args * ap)574 vop_stdgetwritemount(struct vop_getwritemount_args *ap)
575 {
576 struct mount *mp;
577 struct vnode *vp;
578
579 /*
580 * Note that having a reference does not prevent forced unmount from
581 * setting ->v_mount to NULL after the lock gets released. This is of
582 * no consequence for typical consumers (most notably vn_start_write)
583 * since in this case the vnode is VIRF_DOOMED. Unmount might have
584 * progressed far enough that its completion is only delayed by the
585 * reference obtained here. The consumer only needs to concern itself
586 * with releasing it.
587 */
588 vp = ap->a_vp;
589 mp = vfs_ref_from_vp(vp);
590 *(ap->a_mpp) = mp;
591 return (0);
592 }
593
594 /*
595 * If the file system doesn't implement VOP_BMAP, then return sensible defaults:
596 * - Return the vnode's bufobj instead of any underlying device's bufobj
597 * - Calculate the physical block number as if there were equal size
598 * consecutive blocks, but
599 * - Report no contiguous runs of blocks.
600 */
601 int
vop_stdbmap(struct vop_bmap_args * ap)602 vop_stdbmap(struct vop_bmap_args *ap)
603 {
604
605 if (ap->a_bop != NULL)
606 *ap->a_bop = &ap->a_vp->v_bufobj;
607 if (ap->a_bnp != NULL)
608 *ap->a_bnp = ap->a_bn * btodb(ap->a_vp->v_mount->mnt_stat.f_iosize);
609 if (ap->a_runp != NULL)
610 *ap->a_runp = 0;
611 if (ap->a_runb != NULL)
612 *ap->a_runb = 0;
613 return (0);
614 }
615
616 int
vop_stdfsync(struct vop_fsync_args * ap)617 vop_stdfsync(struct vop_fsync_args *ap)
618 {
619
620 return (vn_fsync_buf(ap->a_vp, ap->a_waitfor));
621 }
622
623 static int
vop_stdfdatasync(struct vop_fdatasync_args * ap)624 vop_stdfdatasync(struct vop_fdatasync_args *ap)
625 {
626
627 return (VOP_FSYNC(ap->a_vp, MNT_WAIT, ap->a_td));
628 }
629
630 int
vop_stdfdatasync_buf(struct vop_fdatasync_args * ap)631 vop_stdfdatasync_buf(struct vop_fdatasync_args *ap)
632 {
633
634 return (vn_fsync_buf(ap->a_vp, MNT_WAIT));
635 }
636
637 /* XXX Needs good comment and more info in the manpage (VOP_GETPAGES(9)). */
638 int
vop_stdgetpages(struct vop_getpages_args * ap)639 vop_stdgetpages(struct vop_getpages_args *ap)
640 {
641
642 return vnode_pager_generic_getpages(ap->a_vp, ap->a_m,
643 ap->a_count, ap->a_rbehind, ap->a_rahead, NULL, NULL);
644 }
645
646 static int
vop_stdgetpages_async(struct vop_getpages_async_args * ap)647 vop_stdgetpages_async(struct vop_getpages_async_args *ap)
648 {
649 int error;
650
651 error = VOP_GETPAGES(ap->a_vp, ap->a_m, ap->a_count, ap->a_rbehind,
652 ap->a_rahead);
653 if (ap->a_iodone != NULL)
654 ap->a_iodone(ap->a_arg, ap->a_m, ap->a_count, error);
655 return (error);
656 }
657
658 int
vop_stdkqfilter(struct vop_kqfilter_args * ap)659 vop_stdkqfilter(struct vop_kqfilter_args *ap)
660 {
661 return vfs_kqfilter(ap);
662 }
663
664 /* XXX Needs good comment and more info in the manpage (VOP_PUTPAGES(9)). */
665 int
vop_stdputpages(struct vop_putpages_args * ap)666 vop_stdputpages(struct vop_putpages_args *ap)
667 {
668
669 return vnode_pager_generic_putpages(ap->a_vp, ap->a_m, ap->a_count,
670 ap->a_sync, ap->a_rtvals);
671 }
672
673 int
vop_stdvptofh(struct vop_vptofh_args * ap)674 vop_stdvptofh(struct vop_vptofh_args *ap)
675 {
676 return (EOPNOTSUPP);
677 }
678
679 int
vop_stdvptocnp(struct vop_vptocnp_args * ap)680 vop_stdvptocnp(struct vop_vptocnp_args *ap)
681 {
682 struct vnode *const vp = ap->a_vp;
683 struct vnode **const dvp = ap->a_vpp;
684 char *buf = ap->a_buf;
685 size_t *buflen = ap->a_buflen;
686 char *dirbuf;
687 int i = *buflen;
688 int error = 0, covered = 0;
689 int eofflag, flags, locked;
690 size_t dirbuflen, len;
691 off_t off;
692 ino_t fileno;
693 struct vattr va;
694 struct nameidata nd;
695 struct thread *const td = curthread;
696 struct ucred *const cred = td->td_ucred;
697 struct dirent *dp;
698 struct vnode *mvp;
699
700 if (vp->v_type != VDIR)
701 return (ENOENT);
702
703 error = VOP_GETATTR(vp, &va, cred);
704 if (error)
705 return (error);
706
707 VREF(vp);
708 locked = VOP_ISLOCKED(vp);
709 VOP_UNLOCK(vp);
710 NDINIT_ATVP(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF, UIO_SYSSPACE,
711 "..", vp);
712 flags = FREAD;
713 error = vn_open_cred(&nd, &flags, 0, VN_OPEN_NOAUDIT, cred, NULL);
714 if (error) {
715 vn_lock(vp, locked | LK_RETRY);
716 return (error);
717 }
718 NDFREE_PNBUF(&nd);
719
720 mvp = *dvp = nd.ni_vp;
721
722 if (vp->v_mount != (*dvp)->v_mount &&
723 ((*dvp)->v_vflag & VV_ROOT) &&
724 ((*dvp)->v_mount->mnt_flag & MNT_UNION)) {
725 *dvp = (*dvp)->v_mount->mnt_vnodecovered;
726 VREF(mvp);
727 VOP_UNLOCK(mvp);
728 vn_close(mvp, FREAD, cred, td);
729 VREF(*dvp);
730 vn_lock(*dvp, LK_SHARED | LK_RETRY);
731 covered = 1;
732 }
733
734 fileno = va.va_fileid;
735
736 dirbuflen = MAX(DEV_BSIZE, GENERIC_MAXDIRSIZ);
737 if (dirbuflen < va.va_blocksize)
738 dirbuflen = va.va_blocksize;
739 dirbuf = malloc(dirbuflen, M_TEMP, M_WAITOK);
740
741 if ((*dvp)->v_type != VDIR) {
742 error = ENOENT;
743 goto out;
744 }
745
746 len = 0;
747 off = 0;
748 eofflag = 0;
749
750 for (;;) {
751 /* call VOP_READDIR of parent */
752 error = vn_dir_next_dirent(*dvp, td,
753 dirbuf, dirbuflen, &dp, &len, &off, &eofflag);
754 if (error != 0)
755 goto out;
756
757 if (len == 0) {
758 error = ENOENT;
759 goto out;
760 }
761
762 if ((dp->d_type != DT_WHT) &&
763 (dp->d_fileno == fileno)) {
764 if (covered) {
765 VOP_UNLOCK(*dvp);
766 vn_lock(mvp, LK_SHARED | LK_RETRY);
767 if (dirent_exists(mvp, dp->d_name, td) == 0) {
768 error = ENOENT;
769 VOP_UNLOCK(mvp);
770 vn_lock(*dvp, LK_SHARED | LK_RETRY);
771 goto out;
772 }
773 VOP_UNLOCK(mvp);
774 vn_lock(*dvp, LK_SHARED | LK_RETRY);
775 }
776 i -= dp->d_namlen;
777
778 if (i < 0) {
779 error = ENOMEM;
780 goto out;
781 }
782 if (dp->d_namlen == 1 && dp->d_name[0] == '.') {
783 error = ENOENT;
784 } else {
785 bcopy(dp->d_name, buf + i, dp->d_namlen);
786 error = 0;
787 }
788 goto out;
789 }
790 }
791
792 out:
793 free(dirbuf, M_TEMP);
794 if (!error) {
795 *buflen = i;
796 vref(*dvp);
797 }
798 if (covered) {
799 vput(*dvp);
800 vrele(mvp);
801 } else {
802 VOP_UNLOCK(mvp);
803 vn_close(mvp, FREAD, cred, td);
804 }
805 vn_lock(vp, locked | LK_RETRY);
806 return (error);
807 }
808
809 int
vop_stdallocate(struct vop_allocate_args * ap)810 vop_stdallocate(struct vop_allocate_args *ap)
811 {
812 #ifdef __notyet__
813 struct statfs *sfs;
814 off_t maxfilesize = 0;
815 #endif
816 struct iovec aiov;
817 struct vattr vattr, *vap;
818 struct uio auio;
819 off_t fsize, len, cur, offset;
820 uint8_t *buf;
821 struct thread *td;
822 struct vnode *vp;
823 size_t iosize;
824 int error;
825
826 buf = NULL;
827 error = 0;
828 td = curthread;
829 vap = &vattr;
830 vp = ap->a_vp;
831 len = *ap->a_len;
832 offset = *ap->a_offset;
833
834 error = VOP_GETATTR(vp, vap, ap->a_cred);
835 if (error != 0)
836 goto out;
837 fsize = vap->va_size;
838 iosize = vap->va_blocksize;
839 if (iosize == 0)
840 iosize = BLKDEV_IOSIZE;
841 if (iosize > maxphys)
842 iosize = maxphys;
843 buf = malloc(iosize, M_TEMP, M_WAITOK);
844
845 #ifdef __notyet__
846 /*
847 * Check if the filesystem sets f_maxfilesize; if not use
848 * VOP_SETATTR to perform the check.
849 */
850 sfs = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
851 error = VFS_STATFS(vp->v_mount, sfs, td);
852 if (error == 0)
853 maxfilesize = sfs->f_maxfilesize;
854 free(sfs, M_STATFS);
855 if (error != 0)
856 goto out;
857 if (maxfilesize) {
858 if (offset > maxfilesize || len > maxfilesize ||
859 offset + len > maxfilesize) {
860 error = EFBIG;
861 goto out;
862 }
863 } else
864 #endif
865 if (offset + len > vap->va_size) {
866 /*
867 * Test offset + len against the filesystem's maxfilesize.
868 */
869 VATTR_NULL(vap);
870 vap->va_size = offset + len;
871 error = VOP_SETATTR(vp, vap, ap->a_cred);
872 if (error != 0)
873 goto out;
874 VATTR_NULL(vap);
875 vap->va_size = fsize;
876 error = VOP_SETATTR(vp, vap, ap->a_cred);
877 if (error != 0)
878 goto out;
879 }
880
881 for (;;) {
882 /*
883 * Read and write back anything below the nominal file
884 * size. There's currently no way outside the filesystem
885 * to know whether this area is sparse or not.
886 */
887 cur = iosize;
888 if ((offset % iosize) != 0)
889 cur -= (offset % iosize);
890 if (cur > len)
891 cur = len;
892 if (offset < fsize) {
893 aiov.iov_base = buf;
894 aiov.iov_len = cur;
895 auio.uio_iov = &aiov;
896 auio.uio_iovcnt = 1;
897 auio.uio_offset = offset;
898 auio.uio_resid = cur;
899 auio.uio_segflg = UIO_SYSSPACE;
900 auio.uio_rw = UIO_READ;
901 auio.uio_td = td;
902 error = VOP_READ(vp, &auio, ap->a_ioflag, ap->a_cred);
903 if (error != 0)
904 break;
905 if (auio.uio_resid > 0) {
906 bzero(buf + cur - auio.uio_resid,
907 auio.uio_resid);
908 }
909 } else {
910 bzero(buf, cur);
911 }
912
913 aiov.iov_base = buf;
914 aiov.iov_len = cur;
915 auio.uio_iov = &aiov;
916 auio.uio_iovcnt = 1;
917 auio.uio_offset = offset;
918 auio.uio_resid = cur;
919 auio.uio_segflg = UIO_SYSSPACE;
920 auio.uio_rw = UIO_WRITE;
921 auio.uio_td = td;
922
923 error = VOP_WRITE(vp, &auio, ap->a_ioflag, ap->a_cred);
924 if (error != 0)
925 break;
926
927 len -= cur;
928 offset += cur;
929 if (len == 0)
930 break;
931 if (should_yield())
932 break;
933 }
934
935 out:
936 *ap->a_len = len;
937 *ap->a_offset = offset;
938 free(buf, M_TEMP);
939 return (error);
940 }
941
942 static int
vp_zerofill(struct vnode * vp,struct vattr * vap,off_t * offsetp,off_t * lenp,int ioflag,struct ucred * cred)943 vp_zerofill(struct vnode *vp, struct vattr *vap, off_t *offsetp, off_t *lenp,
944 int ioflag, struct ucred *cred)
945 {
946 int iosize;
947 int error = 0;
948 struct iovec aiov;
949 struct uio auio;
950 struct thread *td;
951 off_t offset, len;
952
953 iosize = vap->va_blocksize;
954 td = curthread;
955 offset = *offsetp;
956 len = *lenp;
957
958 if (iosize == 0)
959 iosize = BLKDEV_IOSIZE;
960 /* If va_blocksize is 512 bytes, iosize will be 4 kilobytes */
961 iosize = min(iosize * 8, ZERO_REGION_SIZE);
962
963 while (len > 0) {
964 int xfersize = iosize;
965 if (offset % iosize != 0)
966 xfersize -= offset % iosize;
967 if (xfersize > len)
968 xfersize = len;
969
970 aiov.iov_base = __DECONST(void *, zero_region);
971 aiov.iov_len = xfersize;
972 auio.uio_iov = &aiov;
973 auio.uio_iovcnt = 1;
974 auio.uio_offset = offset;
975 auio.uio_resid = xfersize;
976 auio.uio_segflg = UIO_SYSSPACE;
977 auio.uio_rw = UIO_WRITE;
978 auio.uio_td = td;
979
980 error = VOP_WRITE(vp, &auio, ioflag, cred);
981 if (error != 0) {
982 len -= xfersize - auio.uio_resid;
983 offset += xfersize - auio.uio_resid;
984 break;
985 }
986
987 len -= xfersize;
988 offset += xfersize;
989 }
990
991 *offsetp = offset;
992 *lenp = len;
993 return (error);
994 }
995
996 int
vop_stddeallocate(struct vop_deallocate_args * ap)997 vop_stddeallocate(struct vop_deallocate_args *ap)
998 {
999 struct vnode *vp;
1000 off_t offset, len;
1001 struct ucred *cred;
1002 int error;
1003 struct vattr va;
1004 off_t noff, xfersize, rem;
1005
1006 vp = ap->a_vp;
1007 offset = *ap->a_offset;
1008 cred = ap->a_cred;
1009
1010 error = VOP_GETATTR(vp, &va, cred);
1011 if (error)
1012 return (error);
1013
1014 len = omin((off_t)va.va_size - offset, *ap->a_len);
1015 while (len > 0) {
1016 noff = offset;
1017 error = vn_bmap_seekhole_locked(vp, FIOSEEKDATA, &noff, cred);
1018 if (error) {
1019 if (error != ENXIO)
1020 /* XXX: Is it okay to fallback further? */
1021 goto out;
1022
1023 /*
1024 * No more data region to be filled
1025 */
1026 offset += len;
1027 len = 0;
1028 error = 0;
1029 break;
1030 }
1031 KASSERT(noff >= offset, ("FIOSEEKDATA going backward"));
1032 if (noff != offset) {
1033 xfersize = omin(noff - offset, len);
1034 len -= xfersize;
1035 offset += xfersize;
1036 if (len == 0)
1037 break;
1038 }
1039 error = vn_bmap_seekhole_locked(vp, FIOSEEKHOLE, &noff, cred);
1040 if (error)
1041 goto out;
1042
1043 /* Fill zeroes */
1044 xfersize = rem = omin(noff - offset, len);
1045 error = vp_zerofill(vp, &va, &offset, &rem, ap->a_ioflag, cred);
1046 if (error) {
1047 len -= xfersize - rem;
1048 goto out;
1049 }
1050
1051 len -= xfersize;
1052 if (should_yield())
1053 break;
1054 }
1055 /* Handle the case when offset is beyond EOF */
1056 if (len < 0)
1057 len = 0;
1058 out:
1059 *ap->a_offset = offset;
1060 *ap->a_len = len;
1061 return (error);
1062 }
1063
1064 int
vop_stdadvise(struct vop_advise_args * ap)1065 vop_stdadvise(struct vop_advise_args *ap)
1066 {
1067 struct vnode *vp;
1068 struct bufobj *bo;
1069 uintmax_t bstart, bend;
1070 daddr_t startn, endn;
1071 int bsize, error;
1072
1073 vp = ap->a_vp;
1074 switch (ap->a_advice) {
1075 case POSIX_FADV_WILLNEED:
1076 /*
1077 * Do nothing for now. Filesystems should provide a
1078 * custom method which starts an asynchronous read of
1079 * the requested region.
1080 */
1081 error = 0;
1082 break;
1083 case POSIX_FADV_DONTNEED:
1084 error = 0;
1085 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1086 if (VN_IS_DOOMED(vp)) {
1087 VOP_UNLOCK(vp);
1088 break;
1089 }
1090
1091 /*
1092 * Round to block boundaries (and later possibly further to
1093 * page boundaries). Applications cannot reasonably be aware
1094 * of the boundaries, and the rounding must be to expand at
1095 * both extremities to cover enough. It still doesn't cover
1096 * read-ahead. For partial blocks, this gives unnecessary
1097 * discarding of buffers but is efficient enough since the
1098 * pages usually remain in VMIO for some time.
1099 */
1100 bsize = vp->v_bufobj.bo_bsize;
1101 bstart = rounddown(ap->a_start, bsize);
1102 bend = ap->a_end;
1103 bend = roundup(bend, bsize);
1104
1105 /*
1106 * Deactivate pages in the specified range from the backing VM
1107 * object. Pages that are resident in the buffer cache will
1108 * remain wired until their corresponding buffers are released
1109 * below.
1110 */
1111 if (vp->v_object != NULL) {
1112 VM_OBJECT_RLOCK(vp->v_object);
1113 vm_object_page_noreuse(vp->v_object,
1114 OFF_TO_IDX(trunc_page(bstart)),
1115 OFF_TO_IDX(round_page(bend)));
1116 VM_OBJECT_RUNLOCK(vp->v_object);
1117 }
1118
1119 bo = &vp->v_bufobj;
1120 startn = bstart / bsize;
1121 endn = bend / bsize;
1122 BO_RLOCK(bo);
1123 error = bnoreuselist(&bo->bo_clean, bo, startn, endn);
1124 if (error == 0)
1125 error = bnoreuselist(&bo->bo_dirty, bo, startn, endn);
1126 BO_RUNLOCK(bo);
1127 VOP_UNLOCK(vp);
1128 break;
1129 default:
1130 error = EINVAL;
1131 break;
1132 }
1133 return (error);
1134 }
1135
1136 int
vop_stdunp_bind(struct vop_unp_bind_args * ap)1137 vop_stdunp_bind(struct vop_unp_bind_args *ap)
1138 {
1139
1140 ap->a_vp->v_unpcb = ap->a_unpcb;
1141 return (0);
1142 }
1143
1144 int
vop_stdunp_connect(struct vop_unp_connect_args * ap)1145 vop_stdunp_connect(struct vop_unp_connect_args *ap)
1146 {
1147
1148 *ap->a_unpcb = ap->a_vp->v_unpcb;
1149 return (0);
1150 }
1151
1152 int
vop_stdunp_detach(struct vop_unp_detach_args * ap)1153 vop_stdunp_detach(struct vop_unp_detach_args *ap)
1154 {
1155
1156 ap->a_vp->v_unpcb = NULL;
1157 return (0);
1158 }
1159
1160 static int
vop_stdis_text(struct vop_is_text_args * ap)1161 vop_stdis_text(struct vop_is_text_args *ap)
1162 {
1163
1164 return (atomic_load_int(&ap->a_vp->v_writecount) < 0);
1165 }
1166
1167 int
vop_stdset_text(struct vop_set_text_args * ap)1168 vop_stdset_text(struct vop_set_text_args *ap)
1169 {
1170 struct vnode *vp;
1171 int n;
1172 bool gotref;
1173
1174 vp = ap->a_vp;
1175
1176 n = atomic_load_int(&vp->v_writecount);
1177 for (;;) {
1178 if (__predict_false(n > 0)) {
1179 return (ETXTBSY);
1180 }
1181
1182 /*
1183 * Transition point, we may need to grab a reference on the vnode.
1184 *
1185 * Take the ref early As a safety measure against bogus calls
1186 * to vop_stdunset_text.
1187 */
1188 if (n == 0) {
1189 gotref = false;
1190 if ((vn_irflag_read(vp) & VIRF_TEXT_REF) != 0) {
1191 vref(vp);
1192 gotref = true;
1193 }
1194 if (atomic_fcmpset_int(&vp->v_writecount, &n, -1)) {
1195 return (0);
1196 }
1197 if (gotref) {
1198 vunref(vp);
1199 }
1200 continue;
1201 }
1202
1203 MPASS(n < 0);
1204 if (atomic_fcmpset_int(&vp->v_writecount, &n, n - 1)) {
1205 return (0);
1206 }
1207 }
1208 __assert_unreachable();
1209 }
1210
1211 static int
vop_stdunset_text(struct vop_unset_text_args * ap)1212 vop_stdunset_text(struct vop_unset_text_args *ap)
1213 {
1214 struct vnode *vp;
1215 int n;
1216
1217 vp = ap->a_vp;
1218
1219 n = atomic_load_int(&vp->v_writecount);
1220 for (;;) {
1221 if (__predict_false(n >= 0)) {
1222 return (EINVAL);
1223 }
1224
1225 /*
1226 * Transition point, we may need to release a reference on the vnode.
1227 */
1228 if (n == -1) {
1229 if (atomic_fcmpset_int(&vp->v_writecount, &n, 0)) {
1230 if ((vn_irflag_read(vp) & VIRF_TEXT_REF) != 0) {
1231 vunref(vp);
1232 }
1233 return (0);
1234 }
1235 continue;
1236 }
1237
1238 MPASS(n < -1);
1239 if (atomic_fcmpset_int(&vp->v_writecount, &n, n + 1)) {
1240 return (0);
1241 }
1242 }
1243 __assert_unreachable();
1244 }
1245
1246 static int __always_inline
vop_stdadd_writecount_impl(struct vop_add_writecount_args * ap,bool handle_msync)1247 vop_stdadd_writecount_impl(struct vop_add_writecount_args *ap, bool handle_msync)
1248 {
1249 struct vnode *vp;
1250 struct mount *mp __diagused;
1251 int n;
1252
1253 vp = ap->a_vp;
1254
1255 #ifdef INVARIANTS
1256 mp = vp->v_mount;
1257 if (mp != NULL) {
1258 if (handle_msync) {
1259 VNPASS((mp->mnt_kern_flag & MNTK_NOMSYNC) == 0, vp);
1260 } else {
1261 VNPASS((mp->mnt_kern_flag & MNTK_NOMSYNC) != 0, vp);
1262 }
1263 }
1264 #endif
1265
1266 n = atomic_load_int(&vp->v_writecount);
1267 for (;;) {
1268 if (__predict_false(n < 0)) {
1269 return (ETXTBSY);
1270 }
1271
1272 VNASSERT(n + ap->a_inc >= 0, vp,
1273 ("neg writecount increment %d + %d = %d", n, ap->a_inc,
1274 n + ap->a_inc));
1275 if (n == 0) {
1276 if (handle_msync) {
1277 vlazy(vp);
1278 }
1279 }
1280
1281 if (atomic_fcmpset_int(&vp->v_writecount, &n, n + ap->a_inc)) {
1282 return (0);
1283 }
1284 }
1285 __assert_unreachable();
1286 }
1287
1288 int
vop_stdadd_writecount(struct vop_add_writecount_args * ap)1289 vop_stdadd_writecount(struct vop_add_writecount_args *ap)
1290 {
1291
1292 return (vop_stdadd_writecount_impl(ap, true));
1293 }
1294
1295 int
vop_stdadd_writecount_nomsync(struct vop_add_writecount_args * ap)1296 vop_stdadd_writecount_nomsync(struct vop_add_writecount_args *ap)
1297 {
1298
1299 return (vop_stdadd_writecount_impl(ap, false));
1300 }
1301
1302 int
vop_stdneed_inactive(struct vop_need_inactive_args * ap)1303 vop_stdneed_inactive(struct vop_need_inactive_args *ap)
1304 {
1305
1306 return (1);
1307 }
1308
1309 int
vop_stdioctl(struct vop_ioctl_args * ap)1310 vop_stdioctl(struct vop_ioctl_args *ap)
1311 {
1312 struct vnode *vp;
1313 struct vattr va;
1314 off_t *offp;
1315 int error;
1316
1317 switch (ap->a_command) {
1318 case FIOSEEKDATA:
1319 case FIOSEEKHOLE:
1320 vp = ap->a_vp;
1321 error = vn_lock(vp, LK_SHARED);
1322 if (error != 0)
1323 return (EBADF);
1324 if (vp->v_type == VREG)
1325 error = VOP_GETATTR(vp, &va, ap->a_cred);
1326 else
1327 error = ENOTTY;
1328 if (error == 0) {
1329 offp = ap->a_data;
1330 if (*offp < 0 || *offp >= va.va_size)
1331 error = ENXIO;
1332 else if (ap->a_command == FIOSEEKHOLE)
1333 *offp = va.va_size;
1334 }
1335 VOP_UNLOCK(vp);
1336 break;
1337 default:
1338 error = ENOTTY;
1339 break;
1340 }
1341 return (error);
1342 }
1343
1344 /*
1345 * vfs default ops
1346 * used to fill the vfs function table to get reasonable default return values.
1347 */
1348 int
vfs_stdroot(struct mount * mp,int flags,struct vnode ** vpp)1349 vfs_stdroot(struct mount *mp, int flags, struct vnode **vpp)
1350 {
1351
1352 return (EOPNOTSUPP);
1353 }
1354
1355 int
vfs_stdstatfs(struct mount * mp,struct statfs * sbp)1356 vfs_stdstatfs(struct mount *mp, struct statfs *sbp)
1357 {
1358
1359 return (EOPNOTSUPP);
1360 }
1361
1362 int
vfs_stdquotactl(struct mount * mp,int cmds,uid_t uid,void * arg,bool * mp_busy)1363 vfs_stdquotactl(struct mount *mp, int cmds, uid_t uid, void *arg, bool *mp_busy)
1364 {
1365 return (EOPNOTSUPP);
1366 }
1367
1368 int
vfs_stdsync(struct mount * mp,int waitfor)1369 vfs_stdsync(struct mount *mp, int waitfor)
1370 {
1371 struct vnode *vp, *mvp;
1372 struct thread *td;
1373 int error, lockreq, allerror = 0;
1374
1375 td = curthread;
1376 lockreq = LK_EXCLUSIVE | LK_INTERLOCK;
1377 if (waitfor != MNT_WAIT)
1378 lockreq |= LK_NOWAIT;
1379 /*
1380 * Force stale buffer cache information to be flushed.
1381 */
1382 loop:
1383 MNT_VNODE_FOREACH_ALL(vp, mp, mvp) {
1384 if (vp->v_bufobj.bo_dirty.bv_cnt == 0) {
1385 VI_UNLOCK(vp);
1386 continue;
1387 }
1388 if ((error = vget(vp, lockreq)) != 0) {
1389 if (error == ENOENT) {
1390 MNT_VNODE_FOREACH_ALL_ABORT(mp, mvp);
1391 goto loop;
1392 }
1393 continue;
1394 }
1395 error = VOP_FSYNC(vp, waitfor, td);
1396 if (error)
1397 allerror = error;
1398 vput(vp);
1399 }
1400 return (allerror);
1401 }
1402
1403 int
vfs_stdnosync(struct mount * mp,int waitfor)1404 vfs_stdnosync(struct mount *mp, int waitfor)
1405 {
1406
1407 return (0);
1408 }
1409
1410 static int
vop_stdcopy_file_range(struct vop_copy_file_range_args * ap)1411 vop_stdcopy_file_range(struct vop_copy_file_range_args *ap)
1412 {
1413 int error;
1414
1415 error = vn_generic_copy_file_range(ap->a_invp, ap->a_inoffp,
1416 ap->a_outvp, ap->a_outoffp, ap->a_lenp, ap->a_flags, ap->a_incred,
1417 ap->a_outcred, ap->a_fsizetd);
1418 return (error);
1419 }
1420
1421 int
vfs_stdvget(struct mount * mp,ino_t ino,int flags,struct vnode ** vpp)1422 vfs_stdvget(struct mount *mp, ino_t ino, int flags, struct vnode **vpp)
1423 {
1424
1425 return (EOPNOTSUPP);
1426 }
1427
1428 int
vfs_stdfhtovp(struct mount * mp,struct fid * fhp,int flags,struct vnode ** vpp)1429 vfs_stdfhtovp(struct mount *mp, struct fid *fhp, int flags, struct vnode **vpp)
1430 {
1431
1432 return (EOPNOTSUPP);
1433 }
1434
1435 int
vfs_stdinit(struct vfsconf * vfsp)1436 vfs_stdinit(struct vfsconf *vfsp)
1437 {
1438
1439 return (0);
1440 }
1441
1442 int
vfs_stduninit(struct vfsconf * vfsp)1443 vfs_stduninit(struct vfsconf *vfsp)
1444 {
1445
1446 return(0);
1447 }
1448
1449 int
vfs_stdextattrctl(struct mount * mp,int cmd,struct vnode * filename_vp,int attrnamespace,const char * attrname)1450 vfs_stdextattrctl(struct mount *mp, int cmd, struct vnode *filename_vp,
1451 int attrnamespace, const char *attrname)
1452 {
1453
1454 if (filename_vp != NULL)
1455 VOP_UNLOCK(filename_vp);
1456 return (EOPNOTSUPP);
1457 }
1458
1459 int
vfs_stdsysctl(struct mount * mp,fsctlop_t op,struct sysctl_req * req)1460 vfs_stdsysctl(struct mount *mp, fsctlop_t op, struct sysctl_req *req)
1461 {
1462
1463 return (EOPNOTSUPP);
1464 }
1465
1466 static vop_bypass_t *
bp_by_off(struct vop_vector * vop,struct vop_generic_args * a)1467 bp_by_off(struct vop_vector *vop, struct vop_generic_args *a)
1468 {
1469
1470 return (*(vop_bypass_t **)((char *)vop + a->a_desc->vdesc_vop_offset));
1471 }
1472
1473 int
vop_sigdefer(struct vop_vector * vop,struct vop_generic_args * a)1474 vop_sigdefer(struct vop_vector *vop, struct vop_generic_args *a)
1475 {
1476 vop_bypass_t *bp;
1477 int prev_stops, rc;
1478
1479 bp = bp_by_off(vop, a);
1480 MPASS(bp != NULL);
1481
1482 prev_stops = sigdeferstop(SIGDEFERSTOP_SILENT);
1483 rc = bp(a);
1484 sigallowstop(prev_stops);
1485 return (rc);
1486 }
1487
1488 static int
vop_stdstat(struct vop_stat_args * a)1489 vop_stdstat(struct vop_stat_args *a)
1490 {
1491 struct vattr vattr;
1492 struct vattr *vap;
1493 struct vnode *vp;
1494 struct stat *sb;
1495 int error;
1496 u_short mode;
1497
1498 vp = a->a_vp;
1499 sb = a->a_sb;
1500
1501 error = vop_stat_helper_pre(a);
1502 if (error != 0)
1503 return (error);
1504
1505 vap = &vattr;
1506
1507 /*
1508 * Initialize defaults for new and unusual fields, so that file
1509 * systems which don't support these fields don't need to know
1510 * about them.
1511 */
1512 vap->va_birthtime.tv_sec = -1;
1513 vap->va_birthtime.tv_nsec = 0;
1514 vap->va_fsid = VNOVAL;
1515 vap->va_gen = 0;
1516 vap->va_rdev = NODEV;
1517
1518 error = VOP_GETATTR(vp, vap, a->a_active_cred);
1519 if (error)
1520 goto out;
1521
1522 /*
1523 * Zero the spare stat fields
1524 */
1525 bzero(sb, sizeof *sb);
1526
1527 /*
1528 * Copy from vattr table
1529 */
1530 if (vap->va_fsid != VNOVAL)
1531 sb->st_dev = vap->va_fsid;
1532 else
1533 sb->st_dev = vp->v_mount->mnt_stat.f_fsid.val[0];
1534 sb->st_ino = vap->va_fileid;
1535 mode = vap->va_mode;
1536 switch (vap->va_type) {
1537 case VREG:
1538 mode |= S_IFREG;
1539 break;
1540 case VDIR:
1541 mode |= S_IFDIR;
1542 break;
1543 case VBLK:
1544 mode |= S_IFBLK;
1545 break;
1546 case VCHR:
1547 mode |= S_IFCHR;
1548 break;
1549 case VLNK:
1550 mode |= S_IFLNK;
1551 break;
1552 case VSOCK:
1553 mode |= S_IFSOCK;
1554 break;
1555 case VFIFO:
1556 mode |= S_IFIFO;
1557 break;
1558 default:
1559 error = EBADF;
1560 goto out;
1561 }
1562 sb->st_mode = mode;
1563 sb->st_nlink = vap->va_nlink;
1564 sb->st_uid = vap->va_uid;
1565 sb->st_gid = vap->va_gid;
1566 sb->st_rdev = vap->va_rdev;
1567 if (vap->va_size > OFF_MAX) {
1568 error = EOVERFLOW;
1569 goto out;
1570 }
1571 sb->st_size = vap->va_size;
1572 sb->st_atim.tv_sec = vap->va_atime.tv_sec;
1573 sb->st_atim.tv_nsec = vap->va_atime.tv_nsec;
1574 sb->st_mtim.tv_sec = vap->va_mtime.tv_sec;
1575 sb->st_mtim.tv_nsec = vap->va_mtime.tv_nsec;
1576 sb->st_ctim.tv_sec = vap->va_ctime.tv_sec;
1577 sb->st_ctim.tv_nsec = vap->va_ctime.tv_nsec;
1578 sb->st_birthtim.tv_sec = vap->va_birthtime.tv_sec;
1579 sb->st_birthtim.tv_nsec = vap->va_birthtime.tv_nsec;
1580
1581 /*
1582 * According to www.opengroup.org, the meaning of st_blksize is
1583 * "a filesystem-specific preferred I/O block size for this
1584 * object. In some filesystem types, this may vary from file
1585 * to file"
1586 * Use minimum/default of PAGE_SIZE (e.g. for VCHR).
1587 */
1588
1589 sb->st_blksize = max(PAGE_SIZE, vap->va_blocksize);
1590 sb->st_flags = vap->va_flags;
1591 sb->st_blocks = vap->va_bytes / S_BLKSIZE;
1592 sb->st_gen = vap->va_gen;
1593 out:
1594 return (vop_stat_helper_post(a, error));
1595 }
1596
1597 static int
vop_stdread_pgcache(struct vop_read_pgcache_args * ap __unused)1598 vop_stdread_pgcache(struct vop_read_pgcache_args *ap __unused)
1599 {
1600 return (EJUSTRETURN);
1601 }
1602
1603 static int
vop_stdvput_pair(struct vop_vput_pair_args * ap)1604 vop_stdvput_pair(struct vop_vput_pair_args *ap)
1605 {
1606 struct vnode *dvp, *vp, **vpp;
1607
1608 dvp = ap->a_dvp;
1609 vpp = ap->a_vpp;
1610 vput(dvp);
1611 if (vpp != NULL && ap->a_unlock_vp && (vp = *vpp) != NULL)
1612 vput(vp);
1613 return (0);
1614 }
1615
1616 static int
vop_stdgetlowvnode(struct vop_getlowvnode_args * ap)1617 vop_stdgetlowvnode(struct vop_getlowvnode_args *ap)
1618 {
1619 vref(ap->a_vp);
1620 *ap->a_vplp = ap->a_vp;
1621 return (0);
1622 }
1623