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("$FreeBSD$");
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/namei.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 /*
82 * This vnode table stores what we want to do if the filesystem doesn't
83 * implement a particular VOP.
84 *
85 * If there is no specific entry here, we will return EOPNOTSUPP.
86 *
87 * Note that every filesystem has to implement either vop_access
88 * or vop_accessx; failing to do so will result in immediate crash
89 * due to stack overflow, as vop_stdaccess() calls vop_stdaccessx(),
90 * which calls vop_stdaccess() etc.
91 */
92
93 struct vop_vector default_vnodeops = {
94 .vop_default = NULL,
95 .vop_bypass = VOP_EOPNOTSUPP,
96
97 .vop_access = vop_stdaccess,
98 .vop_accessx = vop_stdaccessx,
99 .vop_advise = vop_stdadvise,
100 .vop_advlock = vop_stdadvlock,
101 .vop_advlockasync = vop_stdadvlockasync,
102 .vop_allocate = vop_stdallocate,
103 .vop_bmap = vop_stdbmap,
104 .vop_close = VOP_NULL,
105 .vop_fsync = VOP_NULL,
106 .vop_getpages = vop_stdgetpages,
107 .vop_getwritemount = vop_stdgetwritemount,
108 .vop_inactive = VOP_NULL,
109 .vop_ioctl = VOP_ENOTTY,
110 .vop_kqfilter = vop_stdkqfilter,
111 .vop_islocked = vop_stdislocked,
112 .vop_lock1 = vop_stdlock,
113 .vop_lookup = vop_nolookup,
114 .vop_open = VOP_NULL,
115 .vop_pathconf = VOP_EINVAL,
116 .vop_poll = vop_nopoll,
117 .vop_putpages = vop_stdputpages,
118 .vop_readlink = VOP_EINVAL,
119 .vop_rename = vop_norename,
120 .vop_revoke = VOP_PANIC,
121 .vop_strategy = vop_nostrategy,
122 .vop_unlock = vop_stdunlock,
123 .vop_vptocnp = vop_stdvptocnp,
124 .vop_vptofh = vop_stdvptofh,
125 };
126
127 /*
128 * Series of placeholder functions for various error returns for
129 * VOPs.
130 */
131
132 int
133 vop_eopnotsupp(struct vop_generic_args *ap)
134 {
135 /*
136 printf("vop_notsupp[%s]\n", ap->a_desc->vdesc_name);
137 */
138
139 return (EOPNOTSUPP);
140 }
141
142 int
143 vop_ebadf(struct vop_generic_args *ap)
144 {
145
146 return (EBADF);
147 }
148
149 int
150 vop_enotty(struct vop_generic_args *ap)
151 {
152
153 return (ENOTTY);
154 }
155
156 int
157 vop_einval(struct vop_generic_args *ap)
158 {
159
160 return (EINVAL);
161 }
162
163 int
164 vop_enoent(struct vop_generic_args *ap)
165 {
166
167 return (ENOENT);
168 }
169
170 int
171 vop_null(struct vop_generic_args *ap)
172 {
173
174 return (0);
175 }
176
177 /*
178 * Helper function to panic on some bad VOPs in some filesystems.
179 */
180 int
181 vop_panic(struct vop_generic_args *ap)
182 {
183
184 panic("filesystem goof: vop_panic[%s]", ap->a_desc->vdesc_name);
185 }
186
187 /*
188 * vop_std<something> and vop_no<something> are default functions for use by
189 * filesystems that need the "default reasonable" implementation for a
190 * particular operation.
191 *
192 * The documentation for the operations they implement exists (if it exists)
193 * in the VOP_<SOMETHING>(9) manpage (all uppercase).
194 */
195
196 /*
197 * Default vop for filesystems that do not support name lookup
198 */
199 static int
200 vop_nolookup(ap)
201 struct vop_lookup_args /* {
202 struct vnode *a_dvp;
203 struct vnode **a_vpp;
204 struct componentname *a_cnp;
205 } */ *ap;
206 {
207
208 *ap->a_vpp = NULL;
209 return (ENOTDIR);
210 }
211
212 /*
213 * vop_norename:
214 *
215 * Handle unlock and reference counting for arguments of vop_rename
216 * for filesystems that do not implement rename operation.
217 */
218 static int
219 vop_norename(struct vop_rename_args *ap)
220 {
221
222 vop_rename_fail(ap);
223 return (EOPNOTSUPP);
224 }
225
226 /*
227 * vop_nostrategy:
228 *
229 * Strategy routine for VFS devices that have none.
230 *
231 * BIO_ERROR and B_INVAL must be cleared prior to calling any strategy
232 * routine. Typically this is done for a BIO_READ strategy call.
233 * Typically B_INVAL is assumed to already be clear prior to a write
234 * and should not be cleared manually unless you just made the buffer
235 * invalid. BIO_ERROR should be cleared either way.
236 */
237
238 static int
239 vop_nostrategy (struct vop_strategy_args *ap)
240 {
241 printf("No strategy for buffer at %p\n", ap->a_bp);
242 vprint("vnode", ap->a_vp);
243 ap->a_bp->b_ioflags |= BIO_ERROR;
244 ap->a_bp->b_error = EOPNOTSUPP;
245 bufdone(ap->a_bp);
246 return (EOPNOTSUPP);
247 }
248
249 static int
250 get_next_dirent(struct vnode *vp, struct dirent **dpp, char *dirbuf,
251 int dirbuflen, off_t *off, char **cpos, int *len,
252 int *eofflag, struct thread *td)
253 {
254 int error, reclen;
255 struct uio uio;
256 struct iovec iov;
257 struct dirent *dp;
258
259 KASSERT(VOP_ISLOCKED(vp), ("vp %p is not locked", vp));
260 KASSERT(vp->v_type == VDIR, ("vp %p is not a directory", vp));
261
262 if (*len == 0) {
263 iov.iov_base = dirbuf;
264 iov.iov_len = dirbuflen;
265
266 uio.uio_iov = &iov;
267 uio.uio_iovcnt = 1;
268 uio.uio_offset = *off;
269 uio.uio_resid = dirbuflen;
270 uio.uio_segflg = UIO_SYSSPACE;
271 uio.uio_rw = UIO_READ;
272 uio.uio_td = td;
273
274 *eofflag = 0;
275
276 #ifdef MAC
277 error = mac_vnode_check_readdir(td->td_ucred, vp);
278 if (error == 0)
279 #endif
280 error = VOP_READDIR(vp, &uio, td->td_ucred, eofflag,
281 NULL, NULL);
282 if (error)
283 return (error);
284
285 *off = uio.uio_offset;
286
287 *cpos = dirbuf;
288 *len = (dirbuflen - uio.uio_resid);
289
290 if (*len == 0)
291 return (ENOENT);
292 }
293
294 dp = (struct dirent *)(*cpos);
295 reclen = dp->d_reclen;
296 *dpp = dp;
297
298 /* check for malformed directory.. */
299 if (reclen < DIRENT_MINSIZE)
300 return (EINVAL);
301
302 *cpos += reclen;
303 *len -= reclen;
304
305 return (0);
306 }
307
308 /*
309 * Check if a named file exists in a given directory vnode.
310 */
311 static int
312 dirent_exists(struct vnode *vp, const char *dirname, struct thread *td)
313 {
314 char *dirbuf, *cpos;
315 int error, eofflag, dirbuflen, len, found;
316 off_t off;
317 struct dirent *dp;
318 struct vattr va;
319
320 KASSERT(VOP_ISLOCKED(vp), ("vp %p is not locked", vp));
321 KASSERT(vp->v_type == VDIR, ("vp %p is not a directory", vp));
322
323 found = 0;
324
325 error = VOP_GETATTR(vp, &va, td->td_ucred);
326 if (error)
327 return (found);
328
329 dirbuflen = DEV_BSIZE;
330 if (dirbuflen < va.va_blocksize)
331 dirbuflen = va.va_blocksize;
332 dirbuf = (char *)malloc(dirbuflen, M_TEMP, M_WAITOK);
333
334 off = 0;
335 len = 0;
336 do {
337 error = get_next_dirent(vp, &dp, dirbuf, dirbuflen, &off,
338 &cpos, &len, &eofflag, td);
339 if (error)
340 goto out;
341
342 if ((dp->d_type != DT_WHT) &&
343 !strcmp(dp->d_name, dirname)) {
344 found = 1;
345 goto out;
346 }
347 } while (len > 0 || !eofflag);
348
349 out:
350 free(dirbuf, M_TEMP);
351 return (found);
352 }
353
354 int
355 vop_stdaccess(struct vop_access_args *ap)
356 {
357
358 KASSERT((ap->a_accmode & ~(VEXEC | VWRITE | VREAD | VADMIN |
359 VAPPEND)) == 0, ("invalid bit in accmode"));
360
361 return (VOP_ACCESSX(ap->a_vp, ap->a_accmode, ap->a_cred, ap->a_td));
362 }
363
364 int
365 vop_stdaccessx(struct vop_accessx_args *ap)
366 {
367 int error;
368 accmode_t accmode = ap->a_accmode;
369
370 error = vfs_unixify_accmode(&accmode);
371 if (error != 0)
372 return (error);
373
374 if (accmode == 0)
375 return (0);
376
377 return (VOP_ACCESS(ap->a_vp, accmode, ap->a_cred, ap->a_td));
378 }
379
380 /*
381 * Advisory record locking support
382 */
383 int
384 vop_stdadvlock(struct vop_advlock_args *ap)
385 {
386 struct vnode *vp;
387 struct ucred *cred;
388 struct vattr vattr;
389 int error;
390
391 vp = ap->a_vp;
392 cred = curthread->td_ucred;
393 vn_lock(vp, LK_SHARED | LK_RETRY);
394 error = VOP_GETATTR(vp, &vattr, cred);
395 VOP_UNLOCK(vp, 0);
396 if (error)
397 return (error);
398
399 return (lf_advlock(ap, &(vp->v_lockf), vattr.va_size));
400 }
401
402 int
403 vop_stdadvlockasync(struct vop_advlockasync_args *ap)
404 {
405 struct vnode *vp;
406 struct ucred *cred;
407 struct vattr vattr;
408 int error;
409
410 vp = ap->a_vp;
411 cred = curthread->td_ucred;
412 vn_lock(vp, LK_SHARED | LK_RETRY);
413 error = VOP_GETATTR(vp, &vattr, cred);
414 VOP_UNLOCK(vp, 0);
415 if (error)
416 return (error);
417
418 return (lf_advlockasync(ap, &(vp->v_lockf), vattr.va_size));
419 }
420
421 /*
422 * vop_stdpathconf:
423 *
424 * Standard implementation of POSIX pathconf, to get information about limits
425 * for a filesystem.
426 * Override per filesystem for the case where the filesystem has smaller
427 * limits.
428 */
429 int
430 vop_stdpathconf(ap)
431 struct vop_pathconf_args /* {
432 struct vnode *a_vp;
433 int a_name;
434 int *a_retval;
435 } */ *ap;
436 {
437
438 switch (ap->a_name) {
439 case _PC_NAME_MAX:
440 *ap->a_retval = NAME_MAX;
441 return (0);
442 case _PC_PATH_MAX:
443 *ap->a_retval = PATH_MAX;
444 return (0);
445 case _PC_LINK_MAX:
446 *ap->a_retval = LINK_MAX;
447 return (0);
448 case _PC_MAX_CANON:
449 *ap->a_retval = MAX_CANON;
450 return (0);
451 case _PC_MAX_INPUT:
452 *ap->a_retval = MAX_INPUT;
453 return (0);
454 case _PC_PIPE_BUF:
455 *ap->a_retval = PIPE_BUF;
456 return (0);
457 case _PC_CHOWN_RESTRICTED:
458 *ap->a_retval = 1;
459 return (0);
460 case _PC_VDISABLE:
461 *ap->a_retval = _POSIX_VDISABLE;
462 return (0);
463 default:
464 return (EINVAL);
465 }
466 /* NOTREACHED */
467 }
468
469 /*
470 * Standard lock, unlock and islocked functions.
471 */
472 int
473 vop_stdlock(ap)
474 struct vop_lock1_args /* {
475 struct vnode *a_vp;
476 int a_flags;
477 char *file;
478 int line;
479 } */ *ap;
480 {
481 struct vnode *vp = ap->a_vp;
482
483 return (_lockmgr_args(vp->v_vnlock, ap->a_flags, VI_MTX(vp),
484 LK_WMESG_DEFAULT, LK_PRIO_DEFAULT, LK_TIMO_DEFAULT, ap->a_file,
485 ap->a_line));
486 }
487
488 /* See above. */
489 int
490 vop_stdunlock(ap)
491 struct vop_unlock_args /* {
492 struct vnode *a_vp;
493 int a_flags;
494 } */ *ap;
495 {
496 struct vnode *vp = ap->a_vp;
497
498 return (lockmgr(vp->v_vnlock, ap->a_flags | LK_RELEASE, VI_MTX(vp)));
499 }
500
501 /* See above. */
502 int
503 vop_stdislocked(ap)
504 struct vop_islocked_args /* {
505 struct vnode *a_vp;
506 } */ *ap;
507 {
508
509 return (lockstatus(ap->a_vp->v_vnlock));
510 }
511
512 /*
513 * Return true for select/poll.
514 */
515 int
516 vop_nopoll(ap)
517 struct vop_poll_args /* {
518 struct vnode *a_vp;
519 int a_events;
520 struct ucred *a_cred;
521 struct thread *a_td;
522 } */ *ap;
523 {
524
525 return (poll_no_poll(ap->a_events));
526 }
527
528 /*
529 * Implement poll for local filesystems that support it.
530 */
531 int
532 vop_stdpoll(ap)
533 struct vop_poll_args /* {
534 struct vnode *a_vp;
535 int a_events;
536 struct ucred *a_cred;
537 struct thread *a_td;
538 } */ *ap;
539 {
540 if (ap->a_events & ~POLLSTANDARD)
541 return (vn_pollrecord(ap->a_vp, ap->a_td, ap->a_events));
542 return (ap->a_events & (POLLIN | POLLOUT | POLLRDNORM | POLLWRNORM));
543 }
544
545 /*
546 * Return our mount point, as we will take charge of the writes.
547 */
548 int
549 vop_stdgetwritemount(ap)
550 struct vop_getwritemount_args /* {
551 struct vnode *a_vp;
552 struct mount **a_mpp;
553 } */ *ap;
554 {
555 struct mount *mp;
556
557 /*
558 * XXX Since this is called unlocked we may be recycled while
559 * attempting to ref the mount. If this is the case or mountpoint
560 * will be set to NULL. We only have to prevent this call from
561 * returning with a ref to an incorrect mountpoint. It is not
562 * harmful to return with a ref to our previous mountpoint.
563 */
564 mp = ap->a_vp->v_mount;
565 if (mp != NULL) {
566 vfs_ref(mp);
567 if (mp != ap->a_vp->v_mount) {
568 vfs_rel(mp);
569 mp = NULL;
570 }
571 }
572 *(ap->a_mpp) = mp;
573 return (0);
574 }
575
576 /* XXX Needs good comment and VOP_BMAP(9) manpage */
577 int
578 vop_stdbmap(ap)
579 struct vop_bmap_args /* {
580 struct vnode *a_vp;
581 daddr_t a_bn;
582 struct bufobj **a_bop;
583 daddr_t *a_bnp;
584 int *a_runp;
585 int *a_runb;
586 } */ *ap;
587 {
588
589 if (ap->a_bop != NULL)
590 *ap->a_bop = &ap->a_vp->v_bufobj;
591 if (ap->a_bnp != NULL)
592 *ap->a_bnp = ap->a_bn * btodb(ap->a_vp->v_mount->mnt_stat.f_iosize);
593 if (ap->a_runp != NULL)
594 *ap->a_runp = 0;
595 if (ap->a_runb != NULL)
596 *ap->a_runb = 0;
597 return (0);
598 }
599
600 int
601 vop_stdfsync(ap)
602 struct vop_fsync_args /* {
603 struct vnode *a_vp;
604 struct ucred *a_cred;
605 int a_waitfor;
606 struct thread *a_td;
607 } */ *ap;
608 {
609 struct vnode *vp = ap->a_vp;
610 struct buf *bp;
611 struct bufobj *bo;
612 struct buf *nbp;
613 int error = 0;
614 int maxretry = 1000; /* large, arbitrarily chosen */
615
616 bo = &vp->v_bufobj;
617 BO_LOCK(bo);
618 loop1:
619 /*
620 * MARK/SCAN initialization to avoid infinite loops.
621 */
622 TAILQ_FOREACH(bp, &bo->bo_dirty.bv_hd, b_bobufs) {
623 bp->b_vflags &= ~BV_SCANNED;
624 bp->b_error = 0;
625 }
626
627 /*
628 * Flush all dirty buffers associated with a vnode.
629 */
630 loop2:
631 TAILQ_FOREACH_SAFE(bp, &bo->bo_dirty.bv_hd, b_bobufs, nbp) {
632 if ((bp->b_vflags & BV_SCANNED) != 0)
633 continue;
634 bp->b_vflags |= BV_SCANNED;
635 if (BUF_LOCK(bp, LK_EXCLUSIVE | LK_NOWAIT, NULL))
636 continue;
637 BO_UNLOCK(bo);
638 KASSERT(bp->b_bufobj == bo,
639 ("bp %p wrong b_bufobj %p should be %p",
640 bp, bp->b_bufobj, bo));
641 if ((bp->b_flags & B_DELWRI) == 0)
642 panic("fsync: not dirty");
643 if ((vp->v_object != NULL) && (bp->b_flags & B_CLUSTEROK)) {
644 vfs_bio_awrite(bp);
645 } else {
646 bremfree(bp);
647 bawrite(bp);
648 }
649 BO_LOCK(bo);
650 goto loop2;
651 }
652
653 /*
654 * If synchronous the caller expects us to completely resolve all
655 * dirty buffers in the system. Wait for in-progress I/O to
656 * complete (which could include background bitmap writes), then
657 * retry if dirty blocks still exist.
658 */
659 if (ap->a_waitfor == MNT_WAIT) {
660 bufobj_wwait(bo, 0, 0);
661 if (bo->bo_dirty.bv_cnt > 0) {
662 /*
663 * If we are unable to write any of these buffers
664 * then we fail now rather than trying endlessly
665 * to write them out.
666 */
667 TAILQ_FOREACH(bp, &bo->bo_dirty.bv_hd, b_bobufs)
668 if ((error = bp->b_error) == 0)
669 continue;
670 if (error == 0 && --maxretry >= 0)
671 goto loop1;
672 error = EAGAIN;
673 }
674 }
675 BO_UNLOCK(bo);
676 if (error == EAGAIN)
677 vprint("fsync: giving up on dirty", vp);
678
679 return (error);
680 }
681
682 /* XXX Needs good comment and more info in the manpage (VOP_GETPAGES(9)). */
683 int
684 vop_stdgetpages(ap)
685 struct vop_getpages_args /* {
686 struct vnode *a_vp;
687 vm_page_t *a_m;
688 int a_count;
689 int a_reqpage;
690 vm_ooffset_t a_offset;
691 } */ *ap;
692 {
693
694 return vnode_pager_generic_getpages(ap->a_vp, ap->a_m,
695 ap->a_count, ap->a_reqpage);
696 }
697
698 int
699 vop_stdkqfilter(struct vop_kqfilter_args *ap)
700 {
701 return vfs_kqfilter(ap);
702 }
703
704 /* XXX Needs good comment and more info in the manpage (VOP_PUTPAGES(9)). */
705 int
706 vop_stdputpages(ap)
707 struct vop_putpages_args /* {
708 struct vnode *a_vp;
709 vm_page_t *a_m;
710 int a_count;
711 int a_sync;
712 int *a_rtvals;
713 vm_ooffset_t a_offset;
714 } */ *ap;
715 {
716
717 return vnode_pager_generic_putpages(ap->a_vp, ap->a_m, ap->a_count,
718 ap->a_sync, ap->a_rtvals);
719 }
720
721 int
722 vop_stdvptofh(struct vop_vptofh_args *ap)
723 {
724 return (EOPNOTSUPP);
725 }
726
727 int
728 vop_stdvptocnp(struct vop_vptocnp_args *ap)
729 {
730 struct vnode *vp = ap->a_vp;
731 struct vnode **dvp = ap->a_vpp;
732 struct ucred *cred = ap->a_cred;
733 char *buf = ap->a_buf;
734 int *buflen = ap->a_buflen;
735 char *dirbuf, *cpos;
736 int i, error, eofflag, dirbuflen, flags, locked, len, covered;
737 off_t off;
738 ino_t fileno;
739 struct vattr va;
740 struct nameidata nd;
741 struct thread *td;
742 struct dirent *dp;
743 struct vnode *mvp;
744
745 i = *buflen;
746 error = 0;
747 covered = 0;
748 td = curthread;
749
750 if (vp->v_type != VDIR)
751 return (ENOENT);
752
753 error = VOP_GETATTR(vp, &va, cred);
754 if (error)
755 return (error);
756
757 VREF(vp);
758 locked = VOP_ISLOCKED(vp);
759 VOP_UNLOCK(vp, 0);
760 NDINIT_ATVP(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE,
761 "..", vp, td);
762 flags = FREAD;
763 error = vn_open_cred(&nd, &flags, 0, VN_OPEN_NOAUDIT, cred, NULL);
764 if (error) {
765 vn_lock(vp, locked | LK_RETRY);
766 return (error);
767 }
768 NDFREE(&nd, NDF_ONLY_PNBUF);
769
770 mvp = *dvp = nd.ni_vp;
771
772 if (vp->v_mount != (*dvp)->v_mount &&
773 ((*dvp)->v_vflag & VV_ROOT) &&
774 ((*dvp)->v_mount->mnt_flag & MNT_UNION)) {
775 *dvp = (*dvp)->v_mount->mnt_vnodecovered;
776 VREF(mvp);
777 VOP_UNLOCK(mvp, 0);
778 vn_close(mvp, FREAD, cred, td);
779 VREF(*dvp);
780 vn_lock(*dvp, LK_EXCLUSIVE | LK_RETRY);
781 covered = 1;
782 }
783
784 fileno = va.va_fileid;
785
786 dirbuflen = DEV_BSIZE;
787 if (dirbuflen < va.va_blocksize)
788 dirbuflen = va.va_blocksize;
789 dirbuf = (char *)malloc(dirbuflen, M_TEMP, M_WAITOK);
790
791 if ((*dvp)->v_type != VDIR) {
792 error = ENOENT;
793 goto out;
794 }
795
796 off = 0;
797 len = 0;
798 do {
799 /* call VOP_READDIR of parent */
800 error = get_next_dirent(*dvp, &dp, dirbuf, dirbuflen, &off,
801 &cpos, &len, &eofflag, td);
802 if (error)
803 goto out;
804
805 if ((dp->d_type != DT_WHT) &&
806 (dp->d_fileno == fileno)) {
807 if (covered) {
808 VOP_UNLOCK(*dvp, 0);
809 vn_lock(mvp, LK_EXCLUSIVE | LK_RETRY);
810 if (dirent_exists(mvp, dp->d_name, td)) {
811 error = ENOENT;
812 VOP_UNLOCK(mvp, 0);
813 vn_lock(*dvp, LK_EXCLUSIVE | LK_RETRY);
814 goto out;
815 }
816 VOP_UNLOCK(mvp, 0);
817 vn_lock(*dvp, LK_EXCLUSIVE | LK_RETRY);
818 }
819 i -= dp->d_namlen;
820
821 if (i < 0) {
822 error = ENOMEM;
823 goto out;
824 }
825 if (dp->d_namlen == 1 && dp->d_name[0] == '.') {
826 error = ENOENT;
827 } else {
828 bcopy(dp->d_name, buf + i, dp->d_namlen);
829 error = 0;
830 }
831 goto out;
832 }
833 } while (len > 0 || !eofflag);
834 error = ENOENT;
835
836 out:
837 free(dirbuf, M_TEMP);
838 if (!error) {
839 *buflen = i;
840 vhold(*dvp);
841 }
842 if (covered) {
843 vput(*dvp);
844 vrele(mvp);
845 } else {
846 VOP_UNLOCK(mvp, 0);
847 vn_close(mvp, FREAD, cred, td);
848 }
849 vn_lock(vp, locked | LK_RETRY);
850 return (error);
851 }
852
853 int
854 vop_stdallocate(struct vop_allocate_args *ap)
855 {
856 #ifdef __notyet__
857 struct statfs sfs;
858 #endif
859 struct iovec aiov;
860 struct vattr vattr, *vap;
861 struct uio auio;
862 off_t fsize, len, cur, offset;
863 uint8_t *buf;
864 struct thread *td;
865 struct vnode *vp;
866 size_t iosize;
867 int error;
868
869 buf = NULL;
870 error = 0;
871 td = curthread;
872 vap = &vattr;
873 vp = ap->a_vp;
874 len = *ap->a_len;
875 offset = *ap->a_offset;
876
877 error = VOP_GETATTR(vp, vap, td->td_ucred);
878 if (error != 0)
879 goto out;
880 fsize = vap->va_size;
881 iosize = vap->va_blocksize;
882 if (iosize == 0)
883 iosize = BLKDEV_IOSIZE;
884 if (iosize > MAXPHYS)
885 iosize = MAXPHYS;
886 buf = malloc(iosize, M_TEMP, M_WAITOK);
887
888 #ifdef __notyet__
889 /*
890 * Check if the filesystem sets f_maxfilesize; if not use
891 * VOP_SETATTR to perform the check.
892 */
893 error = VFS_STATFS(vp->v_mount, &sfs, td);
894 if (error != 0)
895 goto out;
896 if (sfs.f_maxfilesize) {
897 if (offset > sfs.f_maxfilesize || len > sfs.f_maxfilesize ||
898 offset + len > sfs.f_maxfilesize) {
899 error = EFBIG;
900 goto out;
901 }
902 } else
903 #endif
904 if (offset + len > vap->va_size) {
905 /*
906 * Test offset + len against the filesystem's maxfilesize.
907 */
908 VATTR_NULL(vap);
909 vap->va_size = offset + len;
910 error = VOP_SETATTR(vp, vap, td->td_ucred);
911 if (error != 0)
912 goto out;
913 VATTR_NULL(vap);
914 vap->va_size = fsize;
915 error = VOP_SETATTR(vp, vap, td->td_ucred);
916 if (error != 0)
917 goto out;
918 }
919
920 for (;;) {
921 /*
922 * Read and write back anything below the nominal file
923 * size. There's currently no way outside the filesystem
924 * to know whether this area is sparse or not.
925 */
926 cur = iosize;
927 if ((offset % iosize) != 0)
928 cur -= (offset % iosize);
929 if (cur > len)
930 cur = len;
931 if (offset < fsize) {
932 aiov.iov_base = buf;
933 aiov.iov_len = cur;
934 auio.uio_iov = &aiov;
935 auio.uio_iovcnt = 1;
936 auio.uio_offset = offset;
937 auio.uio_resid = cur;
938 auio.uio_segflg = UIO_SYSSPACE;
939 auio.uio_rw = UIO_READ;
940 auio.uio_td = td;
941 error = VOP_READ(vp, &auio, 0, td->td_ucred);
942 if (error != 0)
943 break;
944 if (auio.uio_resid > 0) {
945 bzero(buf + cur - auio.uio_resid,
946 auio.uio_resid);
947 }
948 } else {
949 bzero(buf, cur);
950 }
951
952 aiov.iov_base = buf;
953 aiov.iov_len = cur;
954 auio.uio_iov = &aiov;
955 auio.uio_iovcnt = 1;
956 auio.uio_offset = offset;
957 auio.uio_resid = cur;
958 auio.uio_segflg = UIO_SYSSPACE;
959 auio.uio_rw = UIO_WRITE;
960 auio.uio_td = td;
961
962 error = VOP_WRITE(vp, &auio, 0, td->td_ucred);
963 if (error != 0)
964 break;
965
966 len -= cur;
967 offset += cur;
968 if (len == 0)
969 break;
970 if (should_yield())
971 break;
972 }
973
974 out:
975 *ap->a_len = len;
976 *ap->a_offset = offset;
977 free(buf, M_TEMP);
978 return (error);
979 }
980
981 int
982 vop_stdadvise(struct vop_advise_args *ap)
983 {
984 struct vnode *vp;
985 off_t start, end;
986 int error, vfslocked;
987
988 vp = ap->a_vp;
989 switch (ap->a_advice) {
990 case POSIX_FADV_WILLNEED:
991 /*
992 * Do nothing for now. Filesystems should provide a
993 * custom method which starts an asynchronous read of
994 * the requested region.
995 */
996 error = 0;
997 break;
998 case POSIX_FADV_DONTNEED:
999 /*
1000 * Flush any open FS buffers and then remove pages
1001 * from the backing VM object. Using vinvalbuf() here
1002 * is a bit heavy-handed as it flushes all buffers for
1003 * the given vnode, not just the buffers covering the
1004 * requested range.
1005 */
1006 error = 0;
1007 vfslocked = VFS_LOCK_GIANT(vp->v_mount);
1008 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1009 if (vp->v_iflag & VI_DOOMED) {
1010 VOP_UNLOCK(vp, 0);
1011 VFS_UNLOCK_GIANT(vfslocked);
1012 break;
1013 }
1014 vinvalbuf(vp, V_CLEANONLY, 0, 0);
1015 if (vp->v_object != NULL) {
1016 start = trunc_page(ap->a_start);
1017 end = round_page(ap->a_end);
1018 VM_OBJECT_LOCK(vp->v_object);
1019 vm_object_page_cache(vp->v_object, OFF_TO_IDX(start),
1020 OFF_TO_IDX(end));
1021 VM_OBJECT_UNLOCK(vp->v_object);
1022 }
1023 VOP_UNLOCK(vp, 0);
1024 VFS_UNLOCK_GIANT(vfslocked);
1025 break;
1026 default:
1027 error = EINVAL;
1028 break;
1029 }
1030 return (error);
1031 }
1032
1033 /*
1034 * vfs default ops
1035 * used to fill the vfs function table to get reasonable default return values.
1036 */
1037 int
1038 vfs_stdroot (mp, flags, vpp)
1039 struct mount *mp;
1040 int flags;
1041 struct vnode **vpp;
1042 {
1043
1044 return (EOPNOTSUPP);
1045 }
1046
1047 int
1048 vfs_stdstatfs (mp, sbp)
1049 struct mount *mp;
1050 struct statfs *sbp;
1051 {
1052
1053 return (EOPNOTSUPP);
1054 }
1055
1056 int
1057 vfs_stdquotactl (mp, cmds, uid, arg)
1058 struct mount *mp;
1059 int cmds;
1060 uid_t uid;
1061 void *arg;
1062 {
1063
1064 return (EOPNOTSUPP);
1065 }
1066
1067 int
1068 vfs_stdsync(mp, waitfor)
1069 struct mount *mp;
1070 int waitfor;
1071 {
1072 struct vnode *vp, *mvp;
1073 struct thread *td;
1074 int error, lockreq, allerror = 0;
1075
1076 td = curthread;
1077 lockreq = LK_EXCLUSIVE | LK_INTERLOCK;
1078 if (waitfor != MNT_WAIT)
1079 lockreq |= LK_NOWAIT;
1080 /*
1081 * Force stale buffer cache information to be flushed.
1082 */
1083 MNT_ILOCK(mp);
1084 loop:
1085 MNT_VNODE_FOREACH(vp, mp, mvp) {
1086 /* bv_cnt is an acceptable race here. */
1087 if (vp->v_bufobj.bo_dirty.bv_cnt == 0)
1088 continue;
1089 VI_LOCK(vp);
1090 MNT_IUNLOCK(mp);
1091 if ((error = vget(vp, lockreq, td)) != 0) {
1092 MNT_ILOCK(mp);
1093 if (error == ENOENT) {
1094 MNT_VNODE_FOREACH_ABORT_ILOCKED(mp, mvp);
1095 goto loop;
1096 }
1097 continue;
1098 }
1099 error = VOP_FSYNC(vp, waitfor, td);
1100 if (error)
1101 allerror = error;
1102
1103 /* Do not turn this into vput. td is not always curthread. */
1104 VOP_UNLOCK(vp, 0);
1105 vrele(vp);
1106 MNT_ILOCK(mp);
1107 }
1108 MNT_IUNLOCK(mp);
1109 return (allerror);
1110 }
1111
1112 int
1113 vfs_stdnosync (mp, waitfor)
1114 struct mount *mp;
1115 int waitfor;
1116 {
1117
1118 return (0);
1119 }
1120
1121 int
1122 vfs_stdvget (mp, ino, flags, vpp)
1123 struct mount *mp;
1124 ino_t ino;
1125 int flags;
1126 struct vnode **vpp;
1127 {
1128
1129 return (EOPNOTSUPP);
1130 }
1131
1132 int
1133 vfs_stdfhtovp (mp, fhp, vpp)
1134 struct mount *mp;
1135 struct fid *fhp;
1136 struct vnode **vpp;
1137 {
1138
1139 return (EOPNOTSUPP);
1140 }
1141
1142 int
1143 vfs_stdinit (vfsp)
1144 struct vfsconf *vfsp;
1145 {
1146
1147 return (0);
1148 }
1149
1150 int
1151 vfs_stduninit (vfsp)
1152 struct vfsconf *vfsp;
1153 {
1154
1155 return(0);
1156 }
1157
1158 int
1159 vfs_stdextattrctl(mp, cmd, filename_vp, attrnamespace, attrname)
1160 struct mount *mp;
1161 int cmd;
1162 struct vnode *filename_vp;
1163 int attrnamespace;
1164 const char *attrname;
1165 {
1166
1167 if (filename_vp != NULL)
1168 VOP_UNLOCK(filename_vp, 0);
1169 return (EOPNOTSUPP);
1170 }
1171
1172 int
1173 vfs_stdsysctl(mp, op, req)
1174 struct mount *mp;
1175 fsctlop_t op;
1176 struct sysctl_req *req;
1177 {
1178
1179 return (EOPNOTSUPP);
1180 }
1181
1182 /* end of vfs default ops */
Cache object: bba47a9ef9a69ded074ec932fd7a64ca
|