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/7.4/sys/kern/vfs_default.c 189766 2009-03-13 10:52:22Z 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/mutex.h>
51 #include <sys/unistd.h>
52 #include <sys/vnode.h>
53 #include <sys/poll.h>
54
55 #include <vm/vm.h>
56 #include <vm/vm_object.h>
57 #include <vm/vm_extern.h>
58 #include <vm/pmap.h>
59 #include <vm/vm_map.h>
60 #include <vm/vm_page.h>
61 #include <vm/vm_pager.h>
62 #include <vm/vnode_pager.h>
63
64 static int vop_nolookup(struct vop_lookup_args *);
65 static int vop_nostrategy(struct vop_strategy_args *);
66
67 /*
68 * This vnode table stores what we want to do if the filesystem doesn't
69 * implement a particular VOP.
70 *
71 * If there is no specific entry here, we will return EOPNOTSUPP.
72 *
73 */
74
75 struct vop_vector default_vnodeops = {
76 .vop_default = NULL,
77 .vop_bypass = VOP_EOPNOTSUPP,
78
79 .vop_advlock = vop_stdadvlock,
80 .vop_advlockasync = vop_stdadvlockasync,
81 .vop_bmap = vop_stdbmap,
82 .vop_close = VOP_NULL,
83 .vop_fsync = VOP_NULL,
84 .vop_getpages = vop_stdgetpages,
85 .vop_getwritemount = vop_stdgetwritemount,
86 .vop_inactive = VOP_NULL,
87 .vop_ioctl = VOP_ENOTTY,
88 .vop_kqfilter = vop_stdkqfilter,
89 .vop_islocked = vop_stdislocked,
90 .vop_lease = VOP_NULL,
91 .vop_lock1 = vop_stdlock,
92 .vop_lookup = vop_nolookup,
93 .vop_open = VOP_NULL,
94 .vop_pathconf = VOP_EINVAL,
95 .vop_poll = vop_nopoll,
96 .vop_putpages = vop_stdputpages,
97 .vop_readlink = VOP_EINVAL,
98 .vop_revoke = VOP_PANIC,
99 .vop_strategy = vop_nostrategy,
100 .vop_unlock = vop_stdunlock,
101 .vop_vptofh = vop_stdvptofh,
102 };
103
104 /*
105 * Series of placeholder functions for various error returns for
106 * VOPs.
107 */
108
109 int
110 vop_eopnotsupp(struct vop_generic_args *ap)
111 {
112 /*
113 printf("vop_notsupp[%s]\n", ap->a_desc->vdesc_name);
114 */
115
116 return (EOPNOTSUPP);
117 }
118
119 int
120 vop_ebadf(struct vop_generic_args *ap)
121 {
122
123 return (EBADF);
124 }
125
126 int
127 vop_enotty(struct vop_generic_args *ap)
128 {
129
130 return (ENOTTY);
131 }
132
133 int
134 vop_einval(struct vop_generic_args *ap)
135 {
136
137 return (EINVAL);
138 }
139
140 int
141 vop_null(struct vop_generic_args *ap)
142 {
143
144 return (0);
145 }
146
147 /*
148 * Helper function to panic on some bad VOPs in some filesystems.
149 */
150 int
151 vop_panic(struct vop_generic_args *ap)
152 {
153
154 panic("filesystem goof: vop_panic[%s]", ap->a_desc->vdesc_name);
155 }
156
157 /*
158 * vop_std<something> and vop_no<something> are default functions for use by
159 * filesystems that need the "default reasonable" implementation for a
160 * particular operation.
161 *
162 * The documentation for the operations they implement exists (if it exists)
163 * in the VOP_<SOMETHING>(9) manpage (all uppercase).
164 */
165
166 /*
167 * Default vop for filesystems that do not support name lookup
168 */
169 static int
170 vop_nolookup(ap)
171 struct vop_lookup_args /* {
172 struct vnode *a_dvp;
173 struct vnode **a_vpp;
174 struct componentname *a_cnp;
175 } */ *ap;
176 {
177
178 *ap->a_vpp = NULL;
179 return (ENOTDIR);
180 }
181
182 /*
183 * vop_nostrategy:
184 *
185 * Strategy routine for VFS devices that have none.
186 *
187 * BIO_ERROR and B_INVAL must be cleared prior to calling any strategy
188 * routine. Typically this is done for a BIO_READ strategy call.
189 * Typically B_INVAL is assumed to already be clear prior to a write
190 * and should not be cleared manually unless you just made the buffer
191 * invalid. BIO_ERROR should be cleared either way.
192 */
193
194 static int
195 vop_nostrategy (struct vop_strategy_args *ap)
196 {
197 printf("No strategy for buffer at %p\n", ap->a_bp);
198 vprint("vnode", ap->a_vp);
199 ap->a_bp->b_ioflags |= BIO_ERROR;
200 ap->a_bp->b_error = EOPNOTSUPP;
201 bufdone(ap->a_bp);
202 return (EOPNOTSUPP);
203 }
204
205 /*
206 * Advisory record locking support
207 */
208 int
209 vop_stdadvlock(struct vop_advlock_args *ap)
210 {
211 struct vnode *vp = ap->a_vp;
212 struct thread *td = curthread;
213 struct vattr vattr;
214 int error;
215
216 vn_lock(vp, LK_SHARED | LK_RETRY, td);
217 error = VOP_GETATTR(vp, &vattr, td->td_ucred, td);
218 VOP_UNLOCK(vp, 0, td);
219 if (error)
220 return (error);
221
222 return (lf_advlock(ap, &(vp->v_lockf), vattr.va_size));
223 }
224
225 int
226 vop_stdadvlockasync(struct vop_advlockasync_args *ap)
227 {
228 struct vnode *vp = ap->a_vp;
229 struct thread *td = curthread;
230 struct vattr vattr;
231 int error;
232
233 vn_lock(vp, LK_SHARED | LK_RETRY, td);
234 error = VOP_GETATTR(vp, &vattr, td->td_ucred, td);
235 VOP_UNLOCK(vp, 0, td);
236 if (error)
237 return (error);
238
239 return (lf_advlockasync(ap, &(vp->v_lockf), vattr.va_size));
240 }
241
242 /*
243 * vop_stdpathconf:
244 *
245 * Standard implementation of POSIX pathconf, to get information about limits
246 * for a filesystem.
247 * Override per filesystem for the case where the filesystem has smaller
248 * limits.
249 */
250 int
251 vop_stdpathconf(ap)
252 struct vop_pathconf_args /* {
253 struct vnode *a_vp;
254 int a_name;
255 int *a_retval;
256 } */ *ap;
257 {
258
259 switch (ap->a_name) {
260 case _PC_NAME_MAX:
261 *ap->a_retval = NAME_MAX;
262 return (0);
263 case _PC_PATH_MAX:
264 *ap->a_retval = PATH_MAX;
265 return (0);
266 case _PC_LINK_MAX:
267 *ap->a_retval = LINK_MAX;
268 return (0);
269 case _PC_MAX_CANON:
270 *ap->a_retval = MAX_CANON;
271 return (0);
272 case _PC_MAX_INPUT:
273 *ap->a_retval = MAX_INPUT;
274 return (0);
275 case _PC_PIPE_BUF:
276 *ap->a_retval = PIPE_BUF;
277 return (0);
278 case _PC_CHOWN_RESTRICTED:
279 *ap->a_retval = 1;
280 return (0);
281 case _PC_VDISABLE:
282 *ap->a_retval = _POSIX_VDISABLE;
283 return (0);
284 default:
285 return (EINVAL);
286 }
287 /* NOTREACHED */
288 }
289
290 /*
291 * Standard lock, unlock and islocked functions.
292 */
293 int
294 vop_stdlock(ap)
295 struct vop_lock1_args /* {
296 struct vnode *a_vp;
297 int a_flags;
298 struct thread *a_td;
299 char *file;
300 int line;
301 } */ *ap;
302 {
303 struct vnode *vp = ap->a_vp;
304
305 return (_lockmgr(vp->v_vnlock, ap->a_flags, VI_MTX(vp), ap->a_td, ap->a_file, ap->a_line));
306 }
307
308 /* See above. */
309 int
310 vop_stdunlock(ap)
311 struct vop_unlock_args /* {
312 struct vnode *a_vp;
313 int a_flags;
314 struct thread *a_td;
315 } */ *ap;
316 {
317 struct vnode *vp = ap->a_vp;
318
319 return (lockmgr(vp->v_vnlock, ap->a_flags | LK_RELEASE, VI_MTX(vp),
320 ap->a_td));
321 }
322
323 /* See above. */
324 int
325 vop_stdislocked(ap)
326 struct vop_islocked_args /* {
327 struct vnode *a_vp;
328 struct thread *a_td;
329 } */ *ap;
330 {
331
332 return (lockstatus(ap->a_vp->v_vnlock, ap->a_td));
333 }
334
335 /*
336 * Return true for select/poll.
337 */
338 int
339 vop_nopoll(ap)
340 struct vop_poll_args /* {
341 struct vnode *a_vp;
342 int a_events;
343 struct ucred *a_cred;
344 struct thread *a_td;
345 } */ *ap;
346 {
347
348 return (poll_no_poll(ap->a_events));
349 }
350
351 /*
352 * Implement poll for local filesystems that support it.
353 */
354 int
355 vop_stdpoll(ap)
356 struct vop_poll_args /* {
357 struct vnode *a_vp;
358 int a_events;
359 struct ucred *a_cred;
360 struct thread *a_td;
361 } */ *ap;
362 {
363 if (ap->a_events & ~POLLSTANDARD)
364 return (vn_pollrecord(ap->a_vp, ap->a_td, ap->a_events));
365 return (ap->a_events & (POLLIN | POLLOUT | POLLRDNORM | POLLWRNORM));
366 }
367
368 /*
369 * Return our mount point, as we will take charge of the writes.
370 */
371 int
372 vop_stdgetwritemount(ap)
373 struct vop_getwritemount_args /* {
374 struct vnode *a_vp;
375 struct mount **a_mpp;
376 } */ *ap;
377 {
378 struct mount *mp;
379
380 /*
381 * XXX Since this is called unlocked we may be recycled while
382 * attempting to ref the mount. If this is the case or mountpoint
383 * will be set to NULL. We only have to prevent this call from
384 * returning with a ref to an incorrect mountpoint. It is not
385 * harmful to return with a ref to our previous mountpoint.
386 */
387 mp = ap->a_vp->v_mount;
388 if (mp != NULL) {
389 vfs_ref(mp);
390 if (mp != ap->a_vp->v_mount) {
391 vfs_rel(mp);
392 mp = NULL;
393 }
394 }
395 *(ap->a_mpp) = mp;
396 return (0);
397 }
398
399 /* XXX Needs good comment and VOP_BMAP(9) manpage */
400 int
401 vop_stdbmap(ap)
402 struct vop_bmap_args /* {
403 struct vnode *a_vp;
404 daddr_t a_bn;
405 struct bufobj **a_bop;
406 daddr_t *a_bnp;
407 int *a_runp;
408 int *a_runb;
409 } */ *ap;
410 {
411
412 if (ap->a_bop != NULL)
413 *ap->a_bop = &ap->a_vp->v_bufobj;
414 if (ap->a_bnp != NULL)
415 *ap->a_bnp = ap->a_bn * btodb(ap->a_vp->v_mount->mnt_stat.f_iosize);
416 if (ap->a_runp != NULL)
417 *ap->a_runp = 0;
418 if (ap->a_runb != NULL)
419 *ap->a_runb = 0;
420 return (0);
421 }
422
423 int
424 vop_stdfsync(ap)
425 struct vop_fsync_args /* {
426 struct vnode *a_vp;
427 struct ucred *a_cred;
428 int a_waitfor;
429 struct thread *a_td;
430 } */ *ap;
431 {
432 struct vnode *vp = ap->a_vp;
433 struct buf *bp;
434 struct bufobj *bo;
435 struct buf *nbp;
436 int error = 0;
437 int maxretry = 1000; /* large, arbitrarily chosen */
438
439 VI_LOCK(vp);
440 loop1:
441 /*
442 * MARK/SCAN initialization to avoid infinite loops.
443 */
444 TAILQ_FOREACH(bp, &vp->v_bufobj.bo_dirty.bv_hd, b_bobufs) {
445 bp->b_vflags &= ~BV_SCANNED;
446 bp->b_error = 0;
447 }
448
449 /*
450 * Flush all dirty buffers associated with a vnode.
451 */
452 loop2:
453 TAILQ_FOREACH_SAFE(bp, &vp->v_bufobj.bo_dirty.bv_hd, b_bobufs, nbp) {
454 if ((bp->b_vflags & BV_SCANNED) != 0)
455 continue;
456 bp->b_vflags |= BV_SCANNED;
457 if (BUF_LOCK(bp, LK_EXCLUSIVE | LK_NOWAIT, NULL))
458 continue;
459 VI_UNLOCK(vp);
460 KASSERT(bp->b_bufobj == &vp->v_bufobj,
461 ("bp %p wrong b_bufobj %p should be %p",
462 bp, bp->b_bufobj, &vp->v_bufobj));
463 if ((bp->b_flags & B_DELWRI) == 0)
464 panic("fsync: not dirty");
465 if ((vp->v_object != NULL) && (bp->b_flags & B_CLUSTEROK)) {
466 vfs_bio_awrite(bp);
467 } else {
468 bremfree(bp);
469 bawrite(bp);
470 }
471 VI_LOCK(vp);
472 goto loop2;
473 }
474
475 /*
476 * If synchronous the caller expects us to completely resolve all
477 * dirty buffers in the system. Wait for in-progress I/O to
478 * complete (which could include background bitmap writes), then
479 * retry if dirty blocks still exist.
480 */
481 if (ap->a_waitfor == MNT_WAIT) {
482 bo = &vp->v_bufobj;
483 bufobj_wwait(bo, 0, 0);
484 if (bo->bo_dirty.bv_cnt > 0) {
485 /*
486 * If we are unable to write any of these buffers
487 * then we fail now rather than trying endlessly
488 * to write them out.
489 */
490 TAILQ_FOREACH(bp, &bo->bo_dirty.bv_hd, b_bobufs)
491 if ((error = bp->b_error) == 0)
492 continue;
493 if (error == 0 && --maxretry >= 0)
494 goto loop1;
495 error = EAGAIN;
496 }
497 }
498 VI_UNLOCK(vp);
499 if (error == EAGAIN)
500 vprint("fsync: giving up on dirty", vp);
501
502 return (error);
503 }
504
505 /* XXX Needs good comment and more info in the manpage (VOP_GETPAGES(9)). */
506 int
507 vop_stdgetpages(ap)
508 struct vop_getpages_args /* {
509 struct vnode *a_vp;
510 vm_page_t *a_m;
511 int a_count;
512 int a_reqpage;
513 vm_ooffset_t a_offset;
514 } */ *ap;
515 {
516
517 return vnode_pager_generic_getpages(ap->a_vp, ap->a_m,
518 ap->a_count, ap->a_reqpage);
519 }
520
521 int
522 vop_stdkqfilter(struct vop_kqfilter_args *ap)
523 {
524 return vfs_kqfilter(ap);
525 }
526
527 /* XXX Needs good comment and more info in the manpage (VOP_PUTPAGES(9)). */
528 int
529 vop_stdputpages(ap)
530 struct vop_putpages_args /* {
531 struct vnode *a_vp;
532 vm_page_t *a_m;
533 int a_count;
534 int a_sync;
535 int *a_rtvals;
536 vm_ooffset_t a_offset;
537 } */ *ap;
538 {
539
540 return vnode_pager_generic_putpages(ap->a_vp, ap->a_m, ap->a_count,
541 ap->a_sync, ap->a_rtvals);
542 }
543
544 int
545 vop_stdvptofh(struct vop_vptofh_args *ap)
546 {
547 return (EOPNOTSUPP);
548 }
549
550 /*
551 * vfs default ops
552 * used to fill the vfs function table to get reasonable default return values.
553 */
554 int
555 vfs_stdroot (mp, flags, vpp, td)
556 struct mount *mp;
557 int flags;
558 struct vnode **vpp;
559 struct thread *td;
560 {
561
562 return (EOPNOTSUPP);
563 }
564
565 int
566 vfs_stdstatfs (mp, sbp, td)
567 struct mount *mp;
568 struct statfs *sbp;
569 struct thread *td;
570 {
571
572 return (EOPNOTSUPP);
573 }
574
575 int
576 vfs_stdquotactl (mp, cmds, uid, arg, td)
577 struct mount *mp;
578 int cmds;
579 uid_t uid;
580 void *arg;
581 struct thread *td;
582 {
583
584 return (EOPNOTSUPP);
585 }
586
587 int
588 vfs_stdsync(mp, waitfor, td)
589 struct mount *mp;
590 int waitfor;
591 struct thread *td;
592 {
593 struct vnode *vp, *mvp;
594 int error, lockreq, allerror = 0;
595
596 lockreq = LK_EXCLUSIVE | LK_INTERLOCK;
597 if (waitfor != MNT_WAIT)
598 lockreq |= LK_NOWAIT;
599 /*
600 * Force stale buffer cache information to be flushed.
601 */
602 MNT_ILOCK(mp);
603 loop:
604 MNT_VNODE_FOREACH(vp, mp, mvp) {
605
606 VI_LOCK(vp);
607 if (vp->v_bufobj.bo_dirty.bv_cnt == 0) {
608 VI_UNLOCK(vp);
609 continue;
610 }
611 MNT_IUNLOCK(mp);
612
613 if ((error = vget(vp, lockreq, td)) != 0) {
614 MNT_ILOCK(mp);
615 if (error == ENOENT) {
616 MNT_VNODE_FOREACH_ABORT_ILOCKED(mp, mvp);
617 goto loop;
618 }
619 continue;
620 }
621 error = VOP_FSYNC(vp, waitfor, td);
622 if (error)
623 allerror = error;
624
625 /* Do not turn this into vput. td is not always curthread. */
626 VOP_UNLOCK(vp, 0, td);
627 vrele(vp);
628 MNT_ILOCK(mp);
629 }
630 MNT_IUNLOCK(mp);
631 return (allerror);
632 }
633
634 int
635 vfs_stdnosync (mp, waitfor, td)
636 struct mount *mp;
637 int waitfor;
638 struct thread *td;
639 {
640
641 return (0);
642 }
643
644 int
645 vfs_stdvget (mp, ino, flags, vpp)
646 struct mount *mp;
647 ino_t ino;
648 int flags;
649 struct vnode **vpp;
650 {
651
652 return (EOPNOTSUPP);
653 }
654
655 int
656 vfs_stdfhtovp (mp, fhp, vpp)
657 struct mount *mp;
658 struct fid *fhp;
659 struct vnode **vpp;
660 {
661
662 return (EOPNOTSUPP);
663 }
664
665 int
666 vfs_stdinit (vfsp)
667 struct vfsconf *vfsp;
668 {
669
670 return (0);
671 }
672
673 int
674 vfs_stduninit (vfsp)
675 struct vfsconf *vfsp;
676 {
677
678 return(0);
679 }
680
681 int
682 vfs_stdextattrctl(mp, cmd, filename_vp, attrnamespace, attrname, td)
683 struct mount *mp;
684 int cmd;
685 struct vnode *filename_vp;
686 int attrnamespace;
687 const char *attrname;
688 struct thread *td;
689 {
690
691 if (filename_vp != NULL)
692 VOP_UNLOCK(filename_vp, 0, td);
693 return (EOPNOTSUPP);
694 }
695
696 int
697 vfs_stdsysctl(mp, op, req)
698 struct mount *mp;
699 fsctlop_t op;
700 struct sysctl_req *req;
701 {
702
703 return (EOPNOTSUPP);
704 }
705
706 /* end of vfs default ops */
Cache object: 9cd79f84413023359fa4bca58d53c4e8
|