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/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
vop_eopnotsupp(struct vop_generic_args * ap)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
vop_ebadf(struct vop_generic_args * ap)158 vop_ebadf(struct vop_generic_args *ap)
159 {
160
161 return (EBADF);
162 }
163
164 int
vop_enotty(struct vop_generic_args * ap)165 vop_enotty(struct vop_generic_args *ap)
166 {
167
168 return (ENOTTY);
169 }
170
171 int
vop_einval(struct vop_generic_args * ap)172 vop_einval(struct vop_generic_args *ap)
173 {
174
175 return (EINVAL);
176 }
177
178 int
vop_enoent(struct vop_generic_args * ap)179 vop_enoent(struct vop_generic_args *ap)
180 {
181
182 return (ENOENT);
183 }
184
185 int
vop_null(struct vop_generic_args * ap)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
vop_panic(struct vop_generic_args * ap)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
vop_nolookup(ap)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
vop_norename(struct vop_rename_args * ap)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
vop_nostrategy(struct vop_strategy_args * ap)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
get_next_dirent(struct vnode * vp,struct dirent ** dpp,char * dirbuf,int dirbuflen,off_t * off,char ** cpos,int * len,int * eofflag,struct thread * td)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
dirent_exists(struct vnode * vp,const char * dirname,struct thread * td)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
vop_stdaccess(struct vop_access_args * ap)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
vop_stdaccessx(struct vop_accessx_args * ap)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
vop_stdadvlock(struct vop_advlock_args * ap)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
vop_stdadvlockasync(struct vop_advlockasync_args * ap)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
vop_stdadvlockpurge(struct vop_advlockpurge_args * ap)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
vop_stdpathconf(ap)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
vop_stdlock(ap)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
vop_stdunlock(ap)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
vop_stdislocked(ap)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
vop_nopoll(ap)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
vop_stdpoll(ap)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
vop_stdgetwritemount(ap)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
vop_stdbmap(ap)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
vop_stdfsync(ap)635 vop_stdfsync(ap)
636 struct vop_fsync_args /* {
637 struct vnode *a_vp;
638 struct ucred *a_cred;
639 int a_waitfor;
640 struct thread *a_td;
641 } */ *ap;
642 {
643 struct vnode *vp = ap->a_vp;
644 struct buf *bp;
645 struct bufobj *bo;
646 struct buf *nbp;
647 int error = 0;
648 int maxretry = 1000; /* large, arbitrarily chosen */
649
650 bo = &vp->v_bufobj;
651 BO_LOCK(bo);
652 loop1:
653 /*
654 * MARK/SCAN initialization to avoid infinite loops.
655 */
656 TAILQ_FOREACH(bp, &bo->bo_dirty.bv_hd, b_bobufs) {
657 bp->b_vflags &= ~BV_SCANNED;
658 bp->b_error = 0;
659 }
660
661 /*
662 * Flush all dirty buffers associated with a vnode.
663 */
664 loop2:
665 TAILQ_FOREACH_SAFE(bp, &bo->bo_dirty.bv_hd, b_bobufs, nbp) {
666 if ((bp->b_vflags & BV_SCANNED) != 0)
667 continue;
668 bp->b_vflags |= BV_SCANNED;
669 if (BUF_LOCK(bp, LK_EXCLUSIVE | LK_NOWAIT, NULL)) {
670 if (ap->a_waitfor != MNT_WAIT)
671 continue;
672 if (BUF_LOCK(bp,
673 LK_EXCLUSIVE | LK_INTERLOCK | LK_SLEEPFAIL,
674 BO_LOCKPTR(bo)) != 0) {
675 BO_LOCK(bo);
676 goto loop1;
677 }
678 BO_LOCK(bo);
679 }
680 BO_UNLOCK(bo);
681 KASSERT(bp->b_bufobj == bo,
682 ("bp %p wrong b_bufobj %p should be %p",
683 bp, bp->b_bufobj, bo));
684 if ((bp->b_flags & B_DELWRI) == 0)
685 panic("fsync: not dirty");
686 if ((vp->v_object != NULL) && (bp->b_flags & B_CLUSTEROK)) {
687 vfs_bio_awrite(bp);
688 } else {
689 bremfree(bp);
690 bawrite(bp);
691 }
692 BO_LOCK(bo);
693 goto loop2;
694 }
695
696 /*
697 * If synchronous the caller expects us to completely resolve all
698 * dirty buffers in the system. Wait for in-progress I/O to
699 * complete (which could include background bitmap writes), then
700 * retry if dirty blocks still exist.
701 */
702 if (ap->a_waitfor == MNT_WAIT) {
703 bufobj_wwait(bo, 0, 0);
704 if (bo->bo_dirty.bv_cnt > 0) {
705 /*
706 * If we are unable to write any of these buffers
707 * then we fail now rather than trying endlessly
708 * to write them out.
709 */
710 TAILQ_FOREACH(bp, &bo->bo_dirty.bv_hd, b_bobufs)
711 if ((error = bp->b_error) == 0)
712 continue;
713 if (error == 0 && --maxretry >= 0)
714 goto loop1;
715 error = EAGAIN;
716 }
717 }
718 BO_UNLOCK(bo);
719 if (error == EAGAIN)
720 vprint("fsync: giving up on dirty", vp);
721
722 return (error);
723 }
724
725 /* XXX Needs good comment and more info in the manpage (VOP_GETPAGES(9)). */
726 int
vop_stdgetpages(ap)727 vop_stdgetpages(ap)
728 struct vop_getpages_args /* {
729 struct vnode *a_vp;
730 vm_page_t *a_m;
731 int a_count;
732 int a_reqpage;
733 vm_ooffset_t a_offset;
734 } */ *ap;
735 {
736
737 return vnode_pager_generic_getpages(ap->a_vp, ap->a_m,
738 ap->a_count, ap->a_reqpage);
739 }
740
741 int
vop_stdkqfilter(struct vop_kqfilter_args * ap)742 vop_stdkqfilter(struct vop_kqfilter_args *ap)
743 {
744 return vfs_kqfilter(ap);
745 }
746
747 /* XXX Needs good comment and more info in the manpage (VOP_PUTPAGES(9)). */
748 int
vop_stdputpages(ap)749 vop_stdputpages(ap)
750 struct vop_putpages_args /* {
751 struct vnode *a_vp;
752 vm_page_t *a_m;
753 int a_count;
754 int a_sync;
755 int *a_rtvals;
756 vm_ooffset_t a_offset;
757 } */ *ap;
758 {
759
760 return vnode_pager_generic_putpages(ap->a_vp, ap->a_m, ap->a_count,
761 ap->a_sync, ap->a_rtvals);
762 }
763
764 int
vop_stdvptofh(struct vop_vptofh_args * ap)765 vop_stdvptofh(struct vop_vptofh_args *ap)
766 {
767 return (EOPNOTSUPP);
768 }
769
770 int
vop_stdvptocnp(struct vop_vptocnp_args * ap)771 vop_stdvptocnp(struct vop_vptocnp_args *ap)
772 {
773 struct vnode *vp = ap->a_vp;
774 struct vnode **dvp = ap->a_vpp;
775 struct ucred *cred = ap->a_cred;
776 char *buf = ap->a_buf;
777 int *buflen = ap->a_buflen;
778 char *dirbuf, *cpos;
779 int i, error, eofflag, dirbuflen, flags, locked, len, covered;
780 off_t off;
781 ino_t fileno;
782 struct vattr va;
783 struct nameidata nd;
784 struct thread *td;
785 struct dirent *dp;
786 struct vnode *mvp;
787
788 i = *buflen;
789 error = 0;
790 covered = 0;
791 td = curthread;
792
793 if (vp->v_type != VDIR)
794 return (ENOENT);
795
796 error = VOP_GETATTR(vp, &va, cred);
797 if (error)
798 return (error);
799
800 VREF(vp);
801 locked = VOP_ISLOCKED(vp);
802 VOP_UNLOCK(vp, 0);
803 NDINIT_ATVP(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE,
804 "..", vp, td);
805 flags = FREAD;
806 error = vn_open_cred(&nd, &flags, 0, VN_OPEN_NOAUDIT, cred, NULL);
807 if (error) {
808 vn_lock(vp, locked | LK_RETRY);
809 return (error);
810 }
811 NDFREE(&nd, NDF_ONLY_PNBUF);
812
813 mvp = *dvp = nd.ni_vp;
814
815 if (vp->v_mount != (*dvp)->v_mount &&
816 ((*dvp)->v_vflag & VV_ROOT) &&
817 ((*dvp)->v_mount->mnt_flag & MNT_UNION)) {
818 *dvp = (*dvp)->v_mount->mnt_vnodecovered;
819 VREF(mvp);
820 VOP_UNLOCK(mvp, 0);
821 vn_close(mvp, FREAD, cred, td);
822 VREF(*dvp);
823 vn_lock(*dvp, LK_EXCLUSIVE | LK_RETRY);
824 covered = 1;
825 }
826
827 fileno = va.va_fileid;
828
829 dirbuflen = DEV_BSIZE;
830 if (dirbuflen < va.va_blocksize)
831 dirbuflen = va.va_blocksize;
832 dirbuf = (char *)malloc(dirbuflen, M_TEMP, M_WAITOK);
833
834 if ((*dvp)->v_type != VDIR) {
835 error = ENOENT;
836 goto out;
837 }
838
839 off = 0;
840 len = 0;
841 do {
842 /* call VOP_READDIR of parent */
843 error = get_next_dirent(*dvp, &dp, dirbuf, dirbuflen, &off,
844 &cpos, &len, &eofflag, td);
845 if (error)
846 goto out;
847
848 if ((dp->d_type != DT_WHT) &&
849 (dp->d_fileno == fileno)) {
850 if (covered) {
851 VOP_UNLOCK(*dvp, 0);
852 vn_lock(mvp, LK_EXCLUSIVE | LK_RETRY);
853 if (dirent_exists(mvp, dp->d_name, td)) {
854 error = ENOENT;
855 VOP_UNLOCK(mvp, 0);
856 vn_lock(*dvp, LK_EXCLUSIVE | LK_RETRY);
857 goto out;
858 }
859 VOP_UNLOCK(mvp, 0);
860 vn_lock(*dvp, LK_EXCLUSIVE | LK_RETRY);
861 }
862 i -= dp->d_namlen;
863
864 if (i < 0) {
865 error = ENOMEM;
866 goto out;
867 }
868 if (dp->d_namlen == 1 && dp->d_name[0] == '.') {
869 error = ENOENT;
870 } else {
871 bcopy(dp->d_name, buf + i, dp->d_namlen);
872 error = 0;
873 }
874 goto out;
875 }
876 } while (len > 0 || !eofflag);
877 error = ENOENT;
878
879 out:
880 free(dirbuf, M_TEMP);
881 if (!error) {
882 *buflen = i;
883 vref(*dvp);
884 }
885 if (covered) {
886 vput(*dvp);
887 vrele(mvp);
888 } else {
889 VOP_UNLOCK(mvp, 0);
890 vn_close(mvp, FREAD, cred, td);
891 }
892 vn_lock(vp, locked | LK_RETRY);
893 return (error);
894 }
895
896 int
vop_stdallocate(struct vop_allocate_args * ap)897 vop_stdallocate(struct vop_allocate_args *ap)
898 {
899 #ifdef __notyet__
900 struct statfs sfs;
901 #endif
902 struct iovec aiov;
903 struct vattr vattr, *vap;
904 struct uio auio;
905 off_t fsize, len, cur, offset;
906 uint8_t *buf;
907 struct thread *td;
908 struct vnode *vp;
909 size_t iosize;
910 int error;
911
912 buf = NULL;
913 error = 0;
914 td = curthread;
915 vap = &vattr;
916 vp = ap->a_vp;
917 len = *ap->a_len;
918 offset = *ap->a_offset;
919
920 error = VOP_GETATTR(vp, vap, td->td_ucred);
921 if (error != 0)
922 goto out;
923 fsize = vap->va_size;
924 iosize = vap->va_blocksize;
925 if (iosize == 0)
926 iosize = BLKDEV_IOSIZE;
927 if (iosize > MAXPHYS)
928 iosize = MAXPHYS;
929 buf = malloc(iosize, M_TEMP, M_WAITOK);
930
931 #ifdef __notyet__
932 /*
933 * Check if the filesystem sets f_maxfilesize; if not use
934 * VOP_SETATTR to perform the check.
935 */
936 error = VFS_STATFS(vp->v_mount, &sfs, td);
937 if (error != 0)
938 goto out;
939 if (sfs.f_maxfilesize) {
940 if (offset > sfs.f_maxfilesize || len > sfs.f_maxfilesize ||
941 offset + len > sfs.f_maxfilesize) {
942 error = EFBIG;
943 goto out;
944 }
945 } else
946 #endif
947 if (offset + len > vap->va_size) {
948 /*
949 * Test offset + len against the filesystem's maxfilesize.
950 */
951 VATTR_NULL(vap);
952 vap->va_size = offset + len;
953 error = VOP_SETATTR(vp, vap, td->td_ucred);
954 if (error != 0)
955 goto out;
956 VATTR_NULL(vap);
957 vap->va_size = fsize;
958 error = VOP_SETATTR(vp, vap, td->td_ucred);
959 if (error != 0)
960 goto out;
961 }
962
963 for (;;) {
964 /*
965 * Read and write back anything below the nominal file
966 * size. There's currently no way outside the filesystem
967 * to know whether this area is sparse or not.
968 */
969 cur = iosize;
970 if ((offset % iosize) != 0)
971 cur -= (offset % iosize);
972 if (cur > len)
973 cur = len;
974 if (offset < fsize) {
975 aiov.iov_base = buf;
976 aiov.iov_len = cur;
977 auio.uio_iov = &aiov;
978 auio.uio_iovcnt = 1;
979 auio.uio_offset = offset;
980 auio.uio_resid = cur;
981 auio.uio_segflg = UIO_SYSSPACE;
982 auio.uio_rw = UIO_READ;
983 auio.uio_td = td;
984 error = VOP_READ(vp, &auio, 0, td->td_ucred);
985 if (error != 0)
986 break;
987 if (auio.uio_resid > 0) {
988 bzero(buf + cur - auio.uio_resid,
989 auio.uio_resid);
990 }
991 } else {
992 bzero(buf, cur);
993 }
994
995 aiov.iov_base = buf;
996 aiov.iov_len = cur;
997 auio.uio_iov = &aiov;
998 auio.uio_iovcnt = 1;
999 auio.uio_offset = offset;
1000 auio.uio_resid = cur;
1001 auio.uio_segflg = UIO_SYSSPACE;
1002 auio.uio_rw = UIO_WRITE;
1003 auio.uio_td = td;
1004
1005 error = VOP_WRITE(vp, &auio, 0, td->td_ucred);
1006 if (error != 0)
1007 break;
1008
1009 len -= cur;
1010 offset += cur;
1011 if (len == 0)
1012 break;
1013 if (should_yield())
1014 break;
1015 }
1016
1017 out:
1018 *ap->a_len = len;
1019 *ap->a_offset = offset;
1020 free(buf, M_TEMP);
1021 return (error);
1022 }
1023
1024 int
vop_stdadvise(struct vop_advise_args * ap)1025 vop_stdadvise(struct vop_advise_args *ap)
1026 {
1027 struct vnode *vp;
1028 off_t start, end;
1029 int error;
1030
1031 vp = ap->a_vp;
1032 switch (ap->a_advice) {
1033 case POSIX_FADV_WILLNEED:
1034 /*
1035 * Do nothing for now. Filesystems should provide a
1036 * custom method which starts an asynchronous read of
1037 * the requested region.
1038 */
1039 error = 0;
1040 break;
1041 case POSIX_FADV_DONTNEED:
1042 /*
1043 * Flush any open FS buffers and then remove pages
1044 * from the backing VM object. Using vinvalbuf() here
1045 * is a bit heavy-handed as it flushes all buffers for
1046 * the given vnode, not just the buffers covering the
1047 * requested range.
1048 */
1049 error = 0;
1050 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1051 if (vp->v_iflag & VI_DOOMED) {
1052 VOP_UNLOCK(vp, 0);
1053 break;
1054 }
1055 vinvalbuf(vp, V_CLEANONLY, 0, 0);
1056 if (vp->v_object != NULL) {
1057 start = trunc_page(ap->a_start);
1058 end = round_page(ap->a_end);
1059 VM_OBJECT_WLOCK(vp->v_object);
1060 vm_object_page_cache(vp->v_object, OFF_TO_IDX(start),
1061 OFF_TO_IDX(end));
1062 VM_OBJECT_WUNLOCK(vp->v_object);
1063 }
1064 VOP_UNLOCK(vp, 0);
1065 break;
1066 default:
1067 error = EINVAL;
1068 break;
1069 }
1070 return (error);
1071 }
1072
1073 int
vop_stdunp_bind(struct vop_unp_bind_args * ap)1074 vop_stdunp_bind(struct vop_unp_bind_args *ap)
1075 {
1076
1077 ap->a_vp->v_socket = ap->a_socket;
1078 return (0);
1079 }
1080
1081 int
vop_stdunp_connect(struct vop_unp_connect_args * ap)1082 vop_stdunp_connect(struct vop_unp_connect_args *ap)
1083 {
1084
1085 *ap->a_socket = ap->a_vp->v_socket;
1086 return (0);
1087 }
1088
1089 int
vop_stdunp_detach(struct vop_unp_detach_args * ap)1090 vop_stdunp_detach(struct vop_unp_detach_args *ap)
1091 {
1092
1093 ap->a_vp->v_socket = NULL;
1094 return (0);
1095 }
1096
1097 static int
vop_stdis_text(struct vop_is_text_args * ap)1098 vop_stdis_text(struct vop_is_text_args *ap)
1099 {
1100
1101 return ((ap->a_vp->v_vflag & VV_TEXT) != 0);
1102 }
1103
1104 static int
vop_stdset_text(struct vop_set_text_args * ap)1105 vop_stdset_text(struct vop_set_text_args *ap)
1106 {
1107
1108 ap->a_vp->v_vflag |= VV_TEXT;
1109 return (0);
1110 }
1111
1112 static int
vop_stdunset_text(struct vop_unset_text_args * ap)1113 vop_stdunset_text(struct vop_unset_text_args *ap)
1114 {
1115
1116 ap->a_vp->v_vflag &= ~VV_TEXT;
1117 return (0);
1118 }
1119
1120 static int
vop_stdget_writecount(struct vop_get_writecount_args * ap)1121 vop_stdget_writecount(struct vop_get_writecount_args *ap)
1122 {
1123
1124 *ap->a_writecount = ap->a_vp->v_writecount;
1125 return (0);
1126 }
1127
1128 static int
vop_stdadd_writecount(struct vop_add_writecount_args * ap)1129 vop_stdadd_writecount(struct vop_add_writecount_args *ap)
1130 {
1131
1132 ap->a_vp->v_writecount += ap->a_inc;
1133 return (0);
1134 }
1135
1136 /*
1137 * vfs default ops
1138 * used to fill the vfs function table to get reasonable default return values.
1139 */
1140 int
vfs_stdroot(mp,flags,vpp)1141 vfs_stdroot (mp, flags, vpp)
1142 struct mount *mp;
1143 int flags;
1144 struct vnode **vpp;
1145 {
1146
1147 return (EOPNOTSUPP);
1148 }
1149
1150 int
vfs_stdstatfs(mp,sbp)1151 vfs_stdstatfs (mp, sbp)
1152 struct mount *mp;
1153 struct statfs *sbp;
1154 {
1155
1156 return (EOPNOTSUPP);
1157 }
1158
1159 int
vfs_stdquotactl(mp,cmds,uid,arg)1160 vfs_stdquotactl (mp, cmds, uid, arg)
1161 struct mount *mp;
1162 int cmds;
1163 uid_t uid;
1164 void *arg;
1165 {
1166
1167 return (EOPNOTSUPP);
1168 }
1169
1170 int
vfs_stdsync(mp,waitfor)1171 vfs_stdsync(mp, waitfor)
1172 struct mount *mp;
1173 int waitfor;
1174 {
1175 struct vnode *vp, *mvp;
1176 struct thread *td;
1177 int error, lockreq, allerror = 0;
1178
1179 td = curthread;
1180 lockreq = LK_EXCLUSIVE | LK_INTERLOCK;
1181 if (waitfor != MNT_WAIT)
1182 lockreq |= LK_NOWAIT;
1183 /*
1184 * Force stale buffer cache information to be flushed.
1185 */
1186 loop:
1187 MNT_VNODE_FOREACH_ALL(vp, mp, mvp) {
1188 if (vp->v_bufobj.bo_dirty.bv_cnt == 0) {
1189 VI_UNLOCK(vp);
1190 continue;
1191 }
1192 if ((error = vget(vp, lockreq, td)) != 0) {
1193 if (error == ENOENT) {
1194 MNT_VNODE_FOREACH_ALL_ABORT(mp, mvp);
1195 goto loop;
1196 }
1197 continue;
1198 }
1199 error = VOP_FSYNC(vp, waitfor, td);
1200 if (error)
1201 allerror = error;
1202 vput(vp);
1203 }
1204 return (allerror);
1205 }
1206
1207 int
vfs_stdnosync(mp,waitfor)1208 vfs_stdnosync (mp, waitfor)
1209 struct mount *mp;
1210 int waitfor;
1211 {
1212
1213 return (0);
1214 }
1215
1216 int
vfs_stdvget(mp,ino,flags,vpp)1217 vfs_stdvget (mp, ino, flags, vpp)
1218 struct mount *mp;
1219 ino_t ino;
1220 int flags;
1221 struct vnode **vpp;
1222 {
1223
1224 return (EOPNOTSUPP);
1225 }
1226
1227 int
vfs_stdfhtovp(mp,fhp,flags,vpp)1228 vfs_stdfhtovp (mp, fhp, flags, vpp)
1229 struct mount *mp;
1230 struct fid *fhp;
1231 int flags;
1232 struct vnode **vpp;
1233 {
1234
1235 return (EOPNOTSUPP);
1236 }
1237
1238 int
vfs_stdinit(vfsp)1239 vfs_stdinit (vfsp)
1240 struct vfsconf *vfsp;
1241 {
1242
1243 return (0);
1244 }
1245
1246 int
vfs_stduninit(vfsp)1247 vfs_stduninit (vfsp)
1248 struct vfsconf *vfsp;
1249 {
1250
1251 return(0);
1252 }
1253
1254 int
vfs_stdextattrctl(mp,cmd,filename_vp,attrnamespace,attrname)1255 vfs_stdextattrctl(mp, cmd, filename_vp, attrnamespace, attrname)
1256 struct mount *mp;
1257 int cmd;
1258 struct vnode *filename_vp;
1259 int attrnamespace;
1260 const char *attrname;
1261 {
1262
1263 if (filename_vp != NULL)
1264 VOP_UNLOCK(filename_vp, 0);
1265 return (EOPNOTSUPP);
1266 }
1267
1268 int
vfs_stdsysctl(mp,op,req)1269 vfs_stdsysctl(mp, op, req)
1270 struct mount *mp;
1271 fsctlop_t op;
1272 struct sysctl_req *req;
1273 {
1274
1275 return (EOPNOTSUPP);
1276 }
1277
1278 /* end of vfs default ops */
1279