1 /*-
2 * Copyright (c) 1989, 1993
3 * The Regents of the University of California. All rights reserved.
4 *
5 * This code is derived from software contributed
6 * to Berkeley by John Heidemann of the UCLA Ficus project.
7 *
8 * Source: * @(#)i405_init.c 2.10 92/04/27 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 * 4. 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/cdefs.h>
36 __FBSDID("$FreeBSD: releng/10.4/sys/kern/vfs_default.c 304590 2016-08-22 07:26:43Z kib $");
37
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/kernel.h>
45 #include <sys/limits.h>
46 #include <sys/lock.h>
47 #include <sys/lockf.h>
48 #include <sys/malloc.h>
49 #include <sys/mount.h>
50 #include <sys/namei.h>
51 #include <sys/rwlock.h>
52 #include <sys/fcntl.h>
53 #include <sys/unistd.h>
54 #include <sys/vnode.h>
55 #include <sys/dirent.h>
56 #include <sys/poll.h>
57
58 #include <security/mac/mac_framework.h>
59
60 #include <vm/vm.h>
61 #include <vm/vm_object.h>
62 #include <vm/vm_extern.h>
63 #include <vm/pmap.h>
64 #include <vm/vm_map.h>
65 #include <vm/vm_page.h>
66 #include <vm/vm_pager.h>
67 #include <vm/vnode_pager.h>
68
69 static int vop_nolookup(struct vop_lookup_args *);
70 static int vop_norename(struct vop_rename_args *);
71 static int vop_nostrategy(struct vop_strategy_args *);
72 static int get_next_dirent(struct vnode *vp, struct dirent **dpp,
73 char *dirbuf, int dirbuflen, off_t *off,
74 char **cpos, int *len, int *eofflag,
75 struct thread *td);
76 static int dirent_exists(struct vnode *vp, const char *dirname,
77 struct thread *td);
78
79 #define DIRENT_MINSIZE (sizeof(struct dirent) - (MAXNAMLEN+1) + 4)
80
81 static int vop_stdis_text(struct vop_is_text_args *ap);
82 static int vop_stdset_text(struct vop_set_text_args *ap);
83 static int vop_stdunset_text(struct vop_unset_text_args *ap);
84 static int vop_stdget_writecount(struct vop_get_writecount_args *ap);
85 static int vop_stdadd_writecount(struct vop_add_writecount_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_bmap = vop_stdbmap,
111 .vop_close = VOP_NULL,
112 .vop_fsync = VOP_NULL,
113 .vop_getpages = vop_stdgetpages,
114 .vop_getwritemount = vop_stdgetwritemount,
115 .vop_inactive = VOP_NULL,
116 .vop_ioctl = VOP_ENOTTY,
117 .vop_kqfilter = vop_stdkqfilter,
118 .vop_islocked = vop_stdislocked,
119 .vop_lock1 = vop_stdlock,
120 .vop_lookup = vop_nolookup,
121 .vop_open = VOP_NULL,
122 .vop_pathconf = VOP_EINVAL,
123 .vop_poll = vop_nopoll,
124 .vop_putpages = vop_stdputpages,
125 .vop_readlink = VOP_EINVAL,
126 .vop_rename = vop_norename,
127 .vop_revoke = VOP_PANIC,
128 .vop_strategy = vop_nostrategy,
129 .vop_unlock = vop_stdunlock,
130 .vop_vptocnp = vop_stdvptocnp,
131 .vop_vptofh = vop_stdvptofh,
132 .vop_unp_bind = vop_stdunp_bind,
133 .vop_unp_connect = vop_stdunp_connect,
134 .vop_unp_detach = vop_stdunp_detach,
135 .vop_is_text = vop_stdis_text,
136 .vop_set_text = vop_stdset_text,
137 .vop_unset_text = vop_stdunset_text,
138 .vop_get_writecount = vop_stdget_writecount,
139 .vop_add_writecount = vop_stdadd_writecount,
140 };
141
142 /*
143 * Series of placeholder functions for various error returns for
144 * VOPs.
145 */
146
147 int
148 vop_eopnotsupp(struct vop_generic_args *ap)
149 {
150 /*
151 printf("vop_notsupp[%s]\n", ap->a_desc->vdesc_name);
152 */
153
154 return (EOPNOTSUPP);
155 }
156
157 int
158 vop_ebadf(struct vop_generic_args *ap)
159 {
160
161 return (EBADF);
162 }
163
164 int
165 vop_enotty(struct vop_generic_args *ap)
166 {
167
168 return (ENOTTY);
169 }
170
171 int
172 vop_einval(struct vop_generic_args *ap)
173 {
174
175 return (EINVAL);
176 }
177
178 int
179 vop_enoent(struct vop_generic_args *ap)
180 {
181
182 return (ENOENT);
183 }
184
185 int
186 vop_null(struct vop_generic_args *ap)
187 {
188
189 return (0);
190 }
191
192 /*
193 * Helper function to panic on some bad VOPs in some filesystems.
194 */
195 int
196 vop_panic(struct vop_generic_args *ap)
197 {
198
199 panic("filesystem goof: vop_panic[%s]", ap->a_desc->vdesc_name);
200 }
201
202 /*
203 * vop_std<something> and vop_no<something> are default functions for use by
204 * filesystems that need the "default reasonable" implementation for a
205 * particular operation.
206 *
207 * The documentation for the operations they implement exists (if it exists)
208 * in the VOP_<SOMETHING>(9) manpage (all uppercase).
209 */
210
211 /*
212 * Default vop for filesystems that do not support name lookup
213 */
214 static int
215 vop_nolookup(ap)
216 struct vop_lookup_args /* {
217 struct vnode *a_dvp;
218 struct vnode **a_vpp;
219 struct componentname *a_cnp;
220 } */ *ap;
221 {
222
223 *ap->a_vpp = NULL;
224 return (ENOTDIR);
225 }
226
227 /*
228 * vop_norename:
229 *
230 * Handle unlock and reference counting for arguments of vop_rename
231 * for filesystems that do not implement rename operation.
232 */
233 static int
234 vop_norename(struct vop_rename_args *ap)
235 {
236
237 vop_rename_fail(ap);
238 return (EOPNOTSUPP);
239 }
240
241 /*
242 * vop_nostrategy:
243 *
244 * Strategy routine for VFS devices that have none.
245 *
246 * BIO_ERROR and B_INVAL must be cleared prior to calling any strategy
247 * routine. Typically this is done for a BIO_READ strategy call.
248 * Typically B_INVAL is assumed to already be clear prior to a write
249 * and should not be cleared manually unless you just made the buffer
250 * invalid. BIO_ERROR should be cleared either way.
251 */
252
253 static int
254 vop_nostrategy (struct vop_strategy_args *ap)
255 {
256 printf("No strategy for buffer at %p\n", ap->a_bp);
257 vprint("vnode", ap->a_vp);
258 ap->a_bp->b_ioflags |= BIO_ERROR;
259 ap->a_bp->b_error = EOPNOTSUPP;
260 bufdone(ap->a_bp);
261 return (EOPNOTSUPP);
262 }
263
264 static int
265 get_next_dirent(struct vnode *vp, struct dirent **dpp, char *dirbuf,
266 int dirbuflen, off_t *off, char **cpos, int *len,
267 int *eofflag, struct thread *td)
268 {
269 int error, reclen;
270 struct uio uio;
271 struct iovec iov;
272 struct dirent *dp;
273
274 KASSERT(VOP_ISLOCKED(vp), ("vp %p is not locked", vp));
275 KASSERT(vp->v_type == VDIR, ("vp %p is not a directory", vp));
276
277 if (*len == 0) {
278 iov.iov_base = dirbuf;
279 iov.iov_len = dirbuflen;
280
281 uio.uio_iov = &iov;
282 uio.uio_iovcnt = 1;
283 uio.uio_offset = *off;
284 uio.uio_resid = dirbuflen;
285 uio.uio_segflg = UIO_SYSSPACE;
286 uio.uio_rw = UIO_READ;
287 uio.uio_td = td;
288
289 *eofflag = 0;
290
291 #ifdef MAC
292 error = mac_vnode_check_readdir(td->td_ucred, vp);
293 if (error == 0)
294 #endif
295 error = VOP_READDIR(vp, &uio, td->td_ucred, eofflag,
296 NULL, NULL);
297 if (error)
298 return (error);
299
300 *off = uio.uio_offset;
301
302 *cpos = dirbuf;
303 *len = (dirbuflen - uio.uio_resid);
304
305 if (*len == 0)
306 return (ENOENT);
307 }
308
309 dp = (struct dirent *)(*cpos);
310 reclen = dp->d_reclen;
311 *dpp = dp;
312
313 /* check for malformed directory.. */
314 if (reclen < DIRENT_MINSIZE)
315 return (EINVAL);
316
317 *cpos += reclen;
318 *len -= reclen;
319
320 return (0);
321 }
322
323 /*
324 * Check if a named file exists in a given directory vnode.
325 */
326 static int
327 dirent_exists(struct vnode *vp, const char *dirname, struct thread *td)
328 {
329 char *dirbuf, *cpos;
330 int error, eofflag, dirbuflen, len, found;
331 off_t off;
332 struct dirent *dp;
333 struct vattr va;
334
335 KASSERT(VOP_ISLOCKED(vp), ("vp %p is not locked", vp));
336 KASSERT(vp->v_type == VDIR, ("vp %p is not a directory", vp));
337
338 found = 0;
339
340 error = VOP_GETATTR(vp, &va, td->td_ucred);
341 if (error)
342 return (found);
343
344 dirbuflen = DEV_BSIZE;
345 if (dirbuflen < va.va_blocksize)
346 dirbuflen = va.va_blocksize;
347 dirbuf = (char *)malloc(dirbuflen, M_TEMP, M_WAITOK);
348
349 off = 0;
350 len = 0;
351 do {
352 error = get_next_dirent(vp, &dp, dirbuf, dirbuflen, &off,
353 &cpos, &len, &eofflag, td);
354 if (error)
355 goto out;
356
357 if (dp->d_type != DT_WHT && dp->d_fileno != 0 &&
358 strcmp(dp->d_name, dirname) == 0) {
359 found = 1;
360 goto out;
361 }
362 } while (len > 0 || !eofflag);
363
364 out:
365 free(dirbuf, M_TEMP);
366 return (found);
367 }
368
369 int
370 vop_stdaccess(struct vop_access_args *ap)
371 {
372
373 KASSERT((ap->a_accmode & ~(VEXEC | VWRITE | VREAD | VADMIN |
374 VAPPEND)) == 0, ("invalid bit in accmode"));
375
376 return (VOP_ACCESSX(ap->a_vp, ap->a_accmode, ap->a_cred, ap->a_td));
377 }
378
379 int
380 vop_stdaccessx(struct vop_accessx_args *ap)
381 {
382 int error;
383 accmode_t accmode = ap->a_accmode;
384
385 error = vfs_unixify_accmode(&accmode);
386 if (error != 0)
387 return (error);
388
389 if (accmode == 0)
390 return (0);
391
392 return (VOP_ACCESS(ap->a_vp, accmode, ap->a_cred, ap->a_td));
393 }
394
395 /*
396 * Advisory record locking support
397 */
398 int
399 vop_stdadvlock(struct vop_advlock_args *ap)
400 {
401 struct vnode *vp;
402 struct vattr vattr;
403 int error;
404
405 vp = ap->a_vp;
406 if (ap->a_fl->l_whence == SEEK_END) {
407 /*
408 * The NFSv4 server must avoid doing a vn_lock() here, since it
409 * can deadlock the nfsd threads, due to a LOR. Fortunately
410 * the NFSv4 server always uses SEEK_SET and this code is
411 * only required for the SEEK_END case.
412 */
413 vn_lock(vp, LK_SHARED | LK_RETRY);
414 error = VOP_GETATTR(vp, &vattr, curthread->td_ucred);
415 VOP_UNLOCK(vp, 0);
416 if (error)
417 return (error);
418 } else
419 vattr.va_size = 0;
420
421 return (lf_advlock(ap, &(vp->v_lockf), vattr.va_size));
422 }
423
424 int
425 vop_stdadvlockasync(struct vop_advlockasync_args *ap)
426 {
427 struct vnode *vp;
428 struct vattr vattr;
429 int error;
430
431 vp = ap->a_vp;
432 if (ap->a_fl->l_whence == SEEK_END) {
433 /* The size argument is only needed for SEEK_END. */
434 vn_lock(vp, LK_SHARED | LK_RETRY);
435 error = VOP_GETATTR(vp, &vattr, curthread->td_ucred);
436 VOP_UNLOCK(vp, 0);
437 if (error)
438 return (error);
439 } else
440 vattr.va_size = 0;
441
442 return (lf_advlockasync(ap, &(vp->v_lockf), vattr.va_size));
443 }
444
445 int
446 vop_stdadvlockpurge(struct vop_advlockpurge_args *ap)
447 {
448 struct vnode *vp;
449
450 vp = ap->a_vp;
451 lf_purgelocks(vp, &vp->v_lockf);
452 return (0);
453 }
454
455 /*
456 * vop_stdpathconf:
457 *
458 * Standard implementation of POSIX pathconf, to get information about limits
459 * for a filesystem.
460 * Override per filesystem for the case where the filesystem has smaller
461 * limits.
462 */
463 int
464 vop_stdpathconf(ap)
465 struct vop_pathconf_args /* {
466 struct vnode *a_vp;
467 int a_name;
468 int *a_retval;
469 } */ *ap;
470 {
471
472 switch (ap->a_name) {
473 case _PC_NAME_MAX:
474 *ap->a_retval = NAME_MAX;
475 return (0);
476 case _PC_PATH_MAX:
477 *ap->a_retval = PATH_MAX;
478 return (0);
479 case _PC_LINK_MAX:
480 *ap->a_retval = LINK_MAX;
481 return (0);
482 case _PC_MAX_CANON:
483 *ap->a_retval = MAX_CANON;
484 return (0);
485 case _PC_MAX_INPUT:
486 *ap->a_retval = MAX_INPUT;
487 return (0);
488 case _PC_PIPE_BUF:
489 *ap->a_retval = PIPE_BUF;
490 return (0);
491 case _PC_CHOWN_RESTRICTED:
492 *ap->a_retval = 1;
493 return (0);
494 case _PC_VDISABLE:
495 *ap->a_retval = _POSIX_VDISABLE;
496 return (0);
497 default:
498 return (EINVAL);
499 }
500 /* NOTREACHED */
501 }
502
503 /*
504 * Standard lock, unlock and islocked functions.
505 */
506 int
507 vop_stdlock(ap)
508 struct vop_lock1_args /* {
509 struct vnode *a_vp;
510 int a_flags;
511 char *file;
512 int line;
513 } */ *ap;
514 {
515 struct vnode *vp = ap->a_vp;
516
517 return (_lockmgr_args(vp->v_vnlock, ap->a_flags, VI_MTX(vp),
518 LK_WMESG_DEFAULT, LK_PRIO_DEFAULT, LK_TIMO_DEFAULT, ap->a_file,
519 ap->a_line));
520 }
521
522 /* See above. */
523 int
524 vop_stdunlock(ap)
525 struct vop_unlock_args /* {
526 struct vnode *a_vp;
527 int a_flags;
528 } */ *ap;
529 {
530 struct vnode *vp = ap->a_vp;
531
532 return (lockmgr(vp->v_vnlock, ap->a_flags | LK_RELEASE, VI_MTX(vp)));
533 }
534
535 /* See above. */
536 int
537 vop_stdislocked(ap)
538 struct vop_islocked_args /* {
539 struct vnode *a_vp;
540 } */ *ap;
541 {
542
543 return (lockstatus(ap->a_vp->v_vnlock));
544 }
545
546 /*
547 * Return true for select/poll.
548 */
549 int
550 vop_nopoll(ap)
551 struct vop_poll_args /* {
552 struct vnode *a_vp;
553 int a_events;
554 struct ucred *a_cred;
555 struct thread *a_td;
556 } */ *ap;
557 {
558
559 return (poll_no_poll(ap->a_events));
560 }
561
562 /*
563 * Implement poll for local filesystems that support it.
564 */
565 int
566 vop_stdpoll(ap)
567 struct vop_poll_args /* {
568 struct vnode *a_vp;
569 int a_events;
570 struct ucred *a_cred;
571 struct thread *a_td;
572 } */ *ap;
573 {
574 if (ap->a_events & ~POLLSTANDARD)
575 return (vn_pollrecord(ap->a_vp, ap->a_td, ap->a_events));
576 return (ap->a_events & (POLLIN | POLLOUT | POLLRDNORM | POLLWRNORM));
577 }
578
579 /*
580 * Return our mount point, as we will take charge of the writes.
581 */
582 int
583 vop_stdgetwritemount(ap)
584 struct vop_getwritemount_args /* {
585 struct vnode *a_vp;
586 struct mount **a_mpp;
587 } */ *ap;
588 {
589 struct mount *mp;
590
591 /*
592 * XXX Since this is called unlocked we may be recycled while
593 * attempting to ref the mount. If this is the case or mountpoint
594 * will be set to NULL. We only have to prevent this call from
595 * returning with a ref to an incorrect mountpoint. It is not
596 * harmful to return with a ref to our previous mountpoint.
597 */
598 mp = ap->a_vp->v_mount;
599 if (mp != NULL) {
600 vfs_ref(mp);
601 if (mp != ap->a_vp->v_mount) {
602 vfs_rel(mp);
603 mp = NULL;
604 }
605 }
606 *(ap->a_mpp) = mp;
607 return (0);
608 }
609
610 /* XXX Needs good comment and VOP_BMAP(9) manpage */
611 int
612 vop_stdbmap(ap)
613 struct vop_bmap_args /* {
614 struct vnode *a_vp;
615 daddr_t a_bn;
616 struct bufobj **a_bop;
617 daddr_t *a_bnp;
618 int *a_runp;
619 int *a_runb;
620 } */ *ap;
621 {
622
623 if (ap->a_bop != NULL)
624 *ap->a_bop = &ap->a_vp->v_bufobj;
625 if (ap->a_bnp != NULL)
626 *ap->a_bnp = ap->a_bn * btodb(ap->a_vp->v_mount->mnt_stat.f_iosize);
627 if (ap->a_runp != NULL)
628 *ap->a_runp = 0;
629 if (ap->a_runb != NULL)
630 *ap->a_runb = 0;
631 return (0);
632 }
633
634 int
635 vop_stdfsync(ap)
636 struct vop_fsync_args /* {
637 struct vnode *a_vp;
638 int a_waitfor;
639 struct thread *a_td;
640 } */ *ap;
641 {
642 struct vnode *vp = ap->a_vp;
643 struct buf *bp;
644 struct bufobj *bo;
645 struct buf *nbp;
646 int error = 0;
647 int maxretry = 1000; /* large, arbitrarily chosen */
648
649 bo = &vp->v_bufobj;
650 BO_LOCK(bo);
651 loop1:
652 /*
653 * MARK/SCAN initialization to avoid infinite loops.
654 */
655 TAILQ_FOREACH(bp, &bo->bo_dirty.bv_hd, b_bobufs) {
656 bp->b_vflags &= ~BV_SCANNED;
657 bp->b_error = 0;
658 }
659
660 /*
661 * Flush all dirty buffers associated with a vnode.
662 */
663 loop2:
664 TAILQ_FOREACH_SAFE(bp, &bo->bo_dirty.bv_hd, b_bobufs, nbp) {
665 if ((bp->b_vflags & BV_SCANNED) != 0)
666 continue;
667 bp->b_vflags |= BV_SCANNED;
668 if (BUF_LOCK(bp, LK_EXCLUSIVE | LK_NOWAIT, NULL)) {
669 if (ap->a_waitfor != MNT_WAIT)
670 continue;
671 if (BUF_LOCK(bp,
672 LK_EXCLUSIVE | LK_INTERLOCK | LK_SLEEPFAIL,
673 BO_LOCKPTR(bo)) != 0) {
674 BO_LOCK(bo);
675 goto loop1;
676 }
677 BO_LOCK(bo);
678 }
679 BO_UNLOCK(bo);
680 KASSERT(bp->b_bufobj == bo,
681 ("bp %p wrong b_bufobj %p should be %p",
682 bp, bp->b_bufobj, bo));
683 if ((bp->b_flags & B_DELWRI) == 0)
684 panic("fsync: not dirty");
685 if ((vp->v_object != NULL) && (bp->b_flags & B_CLUSTEROK)) {
686 vfs_bio_awrite(bp);
687 } else {
688 bremfree(bp);
689 bawrite(bp);
690 }
691 BO_LOCK(bo);
692 goto loop2;
693 }
694
695 /*
696 * If synchronous the caller expects us to completely resolve all
697 * dirty buffers in the system. Wait for in-progress I/O to
698 * complete (which could include background bitmap writes), then
699 * retry if dirty blocks still exist.
700 */
701 if (ap->a_waitfor == MNT_WAIT) {
702 bufobj_wwait(bo, 0, 0);
703 if (bo->bo_dirty.bv_cnt > 0) {
704 /*
705 * If we are unable to write any of these buffers
706 * then we fail now rather than trying endlessly
707 * to write them out.
708 */
709 TAILQ_FOREACH(bp, &bo->bo_dirty.bv_hd, b_bobufs)
710 if ((error = bp->b_error) == 0)
711 continue;
712 if (error == 0 && --maxretry >= 0)
713 goto loop1;
714 error = EAGAIN;
715 }
716 }
717 BO_UNLOCK(bo);
718 if (error == EAGAIN)
719 vprint("fsync: giving up on dirty", vp);
720
721 return (error);
722 }
723
724 /* XXX Needs good comment and more info in the manpage (VOP_GETPAGES(9)). */
725 int
726 vop_stdgetpages(ap)
727 struct vop_getpages_args /* {
728 struct vnode *a_vp;
729 vm_page_t *a_m;
730 int a_count;
731 int a_reqpage;
732 vm_ooffset_t a_offset;
733 } */ *ap;
734 {
735
736 return vnode_pager_generic_getpages(ap->a_vp, ap->a_m,
737 ap->a_count, ap->a_reqpage);
738 }
739
740 int
741 vop_stdkqfilter(struct vop_kqfilter_args *ap)
742 {
743 return vfs_kqfilter(ap);
744 }
745
746 /* XXX Needs good comment and more info in the manpage (VOP_PUTPAGES(9)). */
747 int
748 vop_stdputpages(ap)
749 struct vop_putpages_args /* {
750 struct vnode *a_vp;
751 vm_page_t *a_m;
752 int a_count;
753 int a_sync;
754 int *a_rtvals;
755 vm_ooffset_t a_offset;
756 } */ *ap;
757 {
758
759 return vnode_pager_generic_putpages(ap->a_vp, ap->a_m, ap->a_count,
760 ap->a_sync, ap->a_rtvals);
761 }
762
763 int
764 vop_stdvptofh(struct vop_vptofh_args *ap)
765 {
766 return (EOPNOTSUPP);
767 }
768
769 int
770 vop_stdvptocnp(struct vop_vptocnp_args *ap)
771 {
772 struct vnode *vp = ap->a_vp;
773 struct vnode **dvp = ap->a_vpp;
774 struct ucred *cred = ap->a_cred;
775 char *buf = ap->a_buf;
776 int *buflen = ap->a_buflen;
777 char *dirbuf, *cpos;
778 int i, error, eofflag, dirbuflen, flags, locked, len, covered;
779 off_t off;
780 ino_t fileno;
781 struct vattr va;
782 struct nameidata nd;
783 struct thread *td;
784 struct dirent *dp;
785 struct vnode *mvp;
786
787 i = *buflen;
788 error = 0;
789 covered = 0;
790 td = curthread;
791
792 if (vp->v_type != VDIR)
793 return (ENOENT);
794
795 error = VOP_GETATTR(vp, &va, cred);
796 if (error)
797 return (error);
798
799 VREF(vp);
800 locked = VOP_ISLOCKED(vp);
801 VOP_UNLOCK(vp, 0);
802 NDINIT_ATVP(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE,
803 "..", vp, td);
804 flags = FREAD;
805 error = vn_open_cred(&nd, &flags, 0, VN_OPEN_NOAUDIT, cred, NULL);
806 if (error) {
807 vn_lock(vp, locked | LK_RETRY);
808 return (error);
809 }
810 NDFREE(&nd, NDF_ONLY_PNBUF);
811
812 mvp = *dvp = nd.ni_vp;
813
814 if (vp->v_mount != (*dvp)->v_mount &&
815 ((*dvp)->v_vflag & VV_ROOT) &&
816 ((*dvp)->v_mount->mnt_flag & MNT_UNION)) {
817 *dvp = (*dvp)->v_mount->mnt_vnodecovered;
818 VREF(mvp);
819 VOP_UNLOCK(mvp, 0);
820 vn_close(mvp, FREAD, cred, td);
821 VREF(*dvp);
822 vn_lock(*dvp, LK_EXCLUSIVE | LK_RETRY);
823 covered = 1;
824 }
825
826 fileno = va.va_fileid;
827
828 dirbuflen = DEV_BSIZE;
829 if (dirbuflen < va.va_blocksize)
830 dirbuflen = va.va_blocksize;
831 dirbuf = (char *)malloc(dirbuflen, M_TEMP, M_WAITOK);
832
833 if ((*dvp)->v_type != VDIR) {
834 error = ENOENT;
835 goto out;
836 }
837
838 off = 0;
839 len = 0;
840 do {
841 /* call VOP_READDIR of parent */
842 error = get_next_dirent(*dvp, &dp, dirbuf, dirbuflen, &off,
843 &cpos, &len, &eofflag, td);
844 if (error)
845 goto out;
846
847 if ((dp->d_type != DT_WHT) &&
848 (dp->d_fileno == fileno)) {
849 if (covered) {
850 VOP_UNLOCK(*dvp, 0);
851 vn_lock(mvp, LK_EXCLUSIVE | LK_RETRY);
852 if (dirent_exists(mvp, dp->d_name, td)) {
853 error = ENOENT;
854 VOP_UNLOCK(mvp, 0);
855 vn_lock(*dvp, LK_EXCLUSIVE | LK_RETRY);
856 goto out;
857 }
858 VOP_UNLOCK(mvp, 0);
859 vn_lock(*dvp, LK_EXCLUSIVE | LK_RETRY);
860 }
861 i -= dp->d_namlen;
862
863 if (i < 0) {
864 error = ENOMEM;
865 goto out;
866 }
867 if (dp->d_namlen == 1 && dp->d_name[0] == '.') {
868 error = ENOENT;
869 } else {
870 bcopy(dp->d_name, buf + i, dp->d_namlen);
871 error = 0;
872 }
873 goto out;
874 }
875 } while (len > 0 || !eofflag);
876 error = ENOENT;
877
878 out:
879 free(dirbuf, M_TEMP);
880 if (!error) {
881 *buflen = i;
882 vref(*dvp);
883 }
884 if (covered) {
885 vput(*dvp);
886 vrele(mvp);
887 } else {
888 VOP_UNLOCK(mvp, 0);
889 vn_close(mvp, FREAD, cred, td);
890 }
891 vn_lock(vp, locked | LK_RETRY);
892 return (error);
893 }
894
895 int
896 vop_stdallocate(struct vop_allocate_args *ap)
897 {
898 #ifdef __notyet__
899 struct statfs sfs;
900 #endif
901 struct iovec aiov;
902 struct vattr vattr, *vap;
903 struct uio auio;
904 off_t fsize, len, cur, offset;
905 uint8_t *buf;
906 struct thread *td;
907 struct vnode *vp;
908 size_t iosize;
909 int error;
910
911 buf = NULL;
912 error = 0;
913 td = curthread;
914 vap = &vattr;
915 vp = ap->a_vp;
916 len = *ap->a_len;
917 offset = *ap->a_offset;
918
919 error = VOP_GETATTR(vp, vap, td->td_ucred);
920 if (error != 0)
921 goto out;
922 fsize = vap->va_size;
923 iosize = vap->va_blocksize;
924 if (iosize == 0)
925 iosize = BLKDEV_IOSIZE;
926 if (iosize > MAXPHYS)
927 iosize = MAXPHYS;
928 buf = malloc(iosize, M_TEMP, M_WAITOK);
929
930 #ifdef __notyet__
931 /*
932 * Check if the filesystem sets f_maxfilesize; if not use
933 * VOP_SETATTR to perform the check.
934 */
935 error = VFS_STATFS(vp->v_mount, &sfs, td);
936 if (error != 0)
937 goto out;
938 if (sfs.f_maxfilesize) {
939 if (offset > sfs.f_maxfilesize || len > sfs.f_maxfilesize ||
940 offset + len > sfs.f_maxfilesize) {
941 error = EFBIG;
942 goto out;
943 }
944 } else
945 #endif
946 if (offset + len > vap->va_size) {
947 /*
948 * Test offset + len against the filesystem's maxfilesize.
949 */
950 VATTR_NULL(vap);
951 vap->va_size = offset + len;
952 error = VOP_SETATTR(vp, vap, td->td_ucred);
953 if (error != 0)
954 goto out;
955 VATTR_NULL(vap);
956 vap->va_size = fsize;
957 error = VOP_SETATTR(vp, vap, td->td_ucred);
958 if (error != 0)
959 goto out;
960 }
961
962 for (;;) {
963 /*
964 * Read and write back anything below the nominal file
965 * size. There's currently no way outside the filesystem
966 * to know whether this area is sparse or not.
967 */
968 cur = iosize;
969 if ((offset % iosize) != 0)
970 cur -= (offset % iosize);
971 if (cur > len)
972 cur = len;
973 if (offset < fsize) {
974 aiov.iov_base = buf;
975 aiov.iov_len = cur;
976 auio.uio_iov = &aiov;
977 auio.uio_iovcnt = 1;
978 auio.uio_offset = offset;
979 auio.uio_resid = cur;
980 auio.uio_segflg = UIO_SYSSPACE;
981 auio.uio_rw = UIO_READ;
982 auio.uio_td = td;
983 error = VOP_READ(vp, &auio, 0, td->td_ucred);
984 if (error != 0)
985 break;
986 if (auio.uio_resid > 0) {
987 bzero(buf + cur - auio.uio_resid,
988 auio.uio_resid);
989 }
990 } else {
991 bzero(buf, cur);
992 }
993
994 aiov.iov_base = buf;
995 aiov.iov_len = cur;
996 auio.uio_iov = &aiov;
997 auio.uio_iovcnt = 1;
998 auio.uio_offset = offset;
999 auio.uio_resid = cur;
1000 auio.uio_segflg = UIO_SYSSPACE;
1001 auio.uio_rw = UIO_WRITE;
1002 auio.uio_td = td;
1003
1004 error = VOP_WRITE(vp, &auio, 0, td->td_ucred);
1005 if (error != 0)
1006 break;
1007
1008 len -= cur;
1009 offset += cur;
1010 if (len == 0)
1011 break;
1012 if (should_yield())
1013 break;
1014 }
1015
1016 out:
1017 *ap->a_len = len;
1018 *ap->a_offset = offset;
1019 free(buf, M_TEMP);
1020 return (error);
1021 }
1022
1023 int
1024 vop_stdadvise(struct vop_advise_args *ap)
1025 {
1026 struct vnode *vp;
1027 off_t start, end;
1028 int error;
1029
1030 vp = ap->a_vp;
1031 switch (ap->a_advice) {
1032 case POSIX_FADV_WILLNEED:
1033 /*
1034 * Do nothing for now. Filesystems should provide a
1035 * custom method which starts an asynchronous read of
1036 * the requested region.
1037 */
1038 error = 0;
1039 break;
1040 case POSIX_FADV_DONTNEED:
1041 /*
1042 * Flush any open FS buffers and then remove pages
1043 * from the backing VM object. Using vinvalbuf() here
1044 * is a bit heavy-handed as it flushes all buffers for
1045 * the given vnode, not just the buffers covering the
1046 * requested range.
1047 */
1048 error = 0;
1049 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1050 if (vp->v_iflag & VI_DOOMED) {
1051 VOP_UNLOCK(vp, 0);
1052 break;
1053 }
1054 vinvalbuf(vp, V_CLEANONLY, 0, 0);
1055 if (vp->v_object != NULL) {
1056 start = trunc_page(ap->a_start);
1057 end = round_page(ap->a_end);
1058 VM_OBJECT_WLOCK(vp->v_object);
1059 vm_object_page_cache(vp->v_object, OFF_TO_IDX(start),
1060 OFF_TO_IDX(end));
1061 VM_OBJECT_WUNLOCK(vp->v_object);
1062 }
1063 VOP_UNLOCK(vp, 0);
1064 break;
1065 default:
1066 error = EINVAL;
1067 break;
1068 }
1069 return (error);
1070 }
1071
1072 int
1073 vop_stdunp_bind(struct vop_unp_bind_args *ap)
1074 {
1075
1076 ap->a_vp->v_socket = ap->a_socket;
1077 return (0);
1078 }
1079
1080 int
1081 vop_stdunp_connect(struct vop_unp_connect_args *ap)
1082 {
1083
1084 *ap->a_socket = ap->a_vp->v_socket;
1085 return (0);
1086 }
1087
1088 int
1089 vop_stdunp_detach(struct vop_unp_detach_args *ap)
1090 {
1091
1092 ap->a_vp->v_socket = NULL;
1093 return (0);
1094 }
1095
1096 static int
1097 vop_stdis_text(struct vop_is_text_args *ap)
1098 {
1099
1100 return ((ap->a_vp->v_vflag & VV_TEXT) != 0);
1101 }
1102
1103 static int
1104 vop_stdset_text(struct vop_set_text_args *ap)
1105 {
1106
1107 ap->a_vp->v_vflag |= VV_TEXT;
1108 return (0);
1109 }
1110
1111 static int
1112 vop_stdunset_text(struct vop_unset_text_args *ap)
1113 {
1114
1115 ap->a_vp->v_vflag &= ~VV_TEXT;
1116 return (0);
1117 }
1118
1119 static int
1120 vop_stdget_writecount(struct vop_get_writecount_args *ap)
1121 {
1122
1123 *ap->a_writecount = ap->a_vp->v_writecount;
1124 return (0);
1125 }
1126
1127 static int
1128 vop_stdadd_writecount(struct vop_add_writecount_args *ap)
1129 {
1130
1131 ap->a_vp->v_writecount += ap->a_inc;
1132 return (0);
1133 }
1134
1135 /*
1136 * vfs default ops
1137 * used to fill the vfs function table to get reasonable default return values.
1138 */
1139 int
1140 vfs_stdroot (mp, flags, vpp)
1141 struct mount *mp;
1142 int flags;
1143 struct vnode **vpp;
1144 {
1145
1146 return (EOPNOTSUPP);
1147 }
1148
1149 int
1150 vfs_stdstatfs (mp, sbp)
1151 struct mount *mp;
1152 struct statfs *sbp;
1153 {
1154
1155 return (EOPNOTSUPP);
1156 }
1157
1158 int
1159 vfs_stdquotactl (mp, cmds, uid, arg)
1160 struct mount *mp;
1161 int cmds;
1162 uid_t uid;
1163 void *arg;
1164 {
1165
1166 return (EOPNOTSUPP);
1167 }
1168
1169 int
1170 vfs_stdsync(mp, waitfor)
1171 struct mount *mp;
1172 int waitfor;
1173 {
1174 struct vnode *vp, *mvp;
1175 struct thread *td;
1176 int error, lockreq, allerror = 0;
1177
1178 td = curthread;
1179 lockreq = LK_EXCLUSIVE | LK_INTERLOCK;
1180 if (waitfor != MNT_WAIT)
1181 lockreq |= LK_NOWAIT;
1182 /*
1183 * Force stale buffer cache information to be flushed.
1184 */
1185 loop:
1186 MNT_VNODE_FOREACH_ALL(vp, mp, mvp) {
1187 if (vp->v_bufobj.bo_dirty.bv_cnt == 0) {
1188 VI_UNLOCK(vp);
1189 continue;
1190 }
1191 if ((error = vget(vp, lockreq, td)) != 0) {
1192 if (error == ENOENT) {
1193 MNT_VNODE_FOREACH_ALL_ABORT(mp, mvp);
1194 goto loop;
1195 }
1196 continue;
1197 }
1198 error = VOP_FSYNC(vp, waitfor, td);
1199 if (error)
1200 allerror = error;
1201 vput(vp);
1202 }
1203 return (allerror);
1204 }
1205
1206 int
1207 vfs_stdnosync (mp, waitfor)
1208 struct mount *mp;
1209 int waitfor;
1210 {
1211
1212 return (0);
1213 }
1214
1215 int
1216 vfs_stdvget (mp, ino, flags, vpp)
1217 struct mount *mp;
1218 ino_t ino;
1219 int flags;
1220 struct vnode **vpp;
1221 {
1222
1223 return (EOPNOTSUPP);
1224 }
1225
1226 int
1227 vfs_stdfhtovp (mp, fhp, flags, vpp)
1228 struct mount *mp;
1229 struct fid *fhp;
1230 int flags;
1231 struct vnode **vpp;
1232 {
1233
1234 return (EOPNOTSUPP);
1235 }
1236
1237 int
1238 vfs_stdinit (vfsp)
1239 struct vfsconf *vfsp;
1240 {
1241
1242 return (0);
1243 }
1244
1245 int
1246 vfs_stduninit (vfsp)
1247 struct vfsconf *vfsp;
1248 {
1249
1250 return(0);
1251 }
1252
1253 int
1254 vfs_stdextattrctl(mp, cmd, filename_vp, attrnamespace, attrname)
1255 struct mount *mp;
1256 int cmd;
1257 struct vnode *filename_vp;
1258 int attrnamespace;
1259 const char *attrname;
1260 {
1261
1262 if (filename_vp != NULL)
1263 VOP_UNLOCK(filename_vp, 0);
1264 return (EOPNOTSUPP);
1265 }
1266
1267 int
1268 vfs_stdsysctl(mp, op, req)
1269 struct mount *mp;
1270 fsctlop_t op;
1271 struct sysctl_req *req;
1272 {
1273
1274 return (EOPNOTSUPP);
1275 }
1276
1277 /* end of vfs default ops */
Cache object: cfaa1228c4672e63df10b30146ab5aaf
|