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/11.1/sys/kern/vfs_default.c 315475 2017-03-18 05:53:09Z alc $");
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 static int vop_stdfdatasync(struct vop_fdatasync_args *ap);
87 static int vop_stdgetpages_async(struct vop_getpages_async_args *ap);
88
89 /*
90 * This vnode table stores what we want to do if the filesystem doesn't
91 * implement a particular VOP.
92 *
93 * If there is no specific entry here, we will return EOPNOTSUPP.
94 *
95 * Note that every filesystem has to implement either vop_access
96 * or vop_accessx; failing to do so will result in immediate crash
97 * due to stack overflow, as vop_stdaccess() calls vop_stdaccessx(),
98 * which calls vop_stdaccess() etc.
99 */
100
101 struct vop_vector default_vnodeops = {
102 .vop_default = NULL,
103 .vop_bypass = VOP_EOPNOTSUPP,
104
105 .vop_access = vop_stdaccess,
106 .vop_accessx = vop_stdaccessx,
107 .vop_advise = vop_stdadvise,
108 .vop_advlock = vop_stdadvlock,
109 .vop_advlockasync = vop_stdadvlockasync,
110 .vop_advlockpurge = vop_stdadvlockpurge,
111 .vop_allocate = vop_stdallocate,
112 .vop_bmap = vop_stdbmap,
113 .vop_close = VOP_NULL,
114 .vop_fsync = VOP_NULL,
115 .vop_fdatasync = vop_stdfdatasync,
116 .vop_getpages = vop_stdgetpages,
117 .vop_getpages_async = vop_stdgetpages_async,
118 .vop_getwritemount = vop_stdgetwritemount,
119 .vop_inactive = VOP_NULL,
120 .vop_ioctl = VOP_ENOTTY,
121 .vop_kqfilter = vop_stdkqfilter,
122 .vop_islocked = vop_stdislocked,
123 .vop_lock1 = vop_stdlock,
124 .vop_lookup = vop_nolookup,
125 .vop_open = VOP_NULL,
126 .vop_pathconf = VOP_EINVAL,
127 .vop_poll = vop_nopoll,
128 .vop_putpages = vop_stdputpages,
129 .vop_readlink = VOP_EINVAL,
130 .vop_rename = vop_norename,
131 .vop_revoke = VOP_PANIC,
132 .vop_strategy = vop_nostrategy,
133 .vop_unlock = vop_stdunlock,
134 .vop_vptocnp = vop_stdvptocnp,
135 .vop_vptofh = vop_stdvptofh,
136 .vop_unp_bind = vop_stdunp_bind,
137 .vop_unp_connect = vop_stdunp_connect,
138 .vop_unp_detach = vop_stdunp_detach,
139 .vop_is_text = vop_stdis_text,
140 .vop_set_text = vop_stdset_text,
141 .vop_unset_text = vop_stdunset_text,
142 .vop_get_writecount = vop_stdget_writecount,
143 .vop_add_writecount = vop_stdadd_writecount,
144 };
145
146 /*
147 * Series of placeholder functions for various error returns for
148 * VOPs.
149 */
150
151 int
152 vop_eopnotsupp(struct vop_generic_args *ap)
153 {
154 /*
155 printf("vop_notsupp[%s]\n", ap->a_desc->vdesc_name);
156 */
157
158 return (EOPNOTSUPP);
159 }
160
161 int
162 vop_ebadf(struct vop_generic_args *ap)
163 {
164
165 return (EBADF);
166 }
167
168 int
169 vop_enotty(struct vop_generic_args *ap)
170 {
171
172 return (ENOTTY);
173 }
174
175 int
176 vop_einval(struct vop_generic_args *ap)
177 {
178
179 return (EINVAL);
180 }
181
182 int
183 vop_enoent(struct vop_generic_args *ap)
184 {
185
186 return (ENOENT);
187 }
188
189 int
190 vop_null(struct vop_generic_args *ap)
191 {
192
193 return (0);
194 }
195
196 /*
197 * Helper function to panic on some bad VOPs in some filesystems.
198 */
199 int
200 vop_panic(struct vop_generic_args *ap)
201 {
202
203 panic("filesystem goof: vop_panic[%s]", ap->a_desc->vdesc_name);
204 }
205
206 /*
207 * vop_std<something> and vop_no<something> are default functions for use by
208 * filesystems that need the "default reasonable" implementation for a
209 * particular operation.
210 *
211 * The documentation for the operations they implement exists (if it exists)
212 * in the VOP_<SOMETHING>(9) manpage (all uppercase).
213 */
214
215 /*
216 * Default vop for filesystems that do not support name lookup
217 */
218 static int
219 vop_nolookup(ap)
220 struct vop_lookup_args /* {
221 struct vnode *a_dvp;
222 struct vnode **a_vpp;
223 struct componentname *a_cnp;
224 } */ *ap;
225 {
226
227 *ap->a_vpp = NULL;
228 return (ENOTDIR);
229 }
230
231 /*
232 * vop_norename:
233 *
234 * Handle unlock and reference counting for arguments of vop_rename
235 * for filesystems that do not implement rename operation.
236 */
237 static int
238 vop_norename(struct vop_rename_args *ap)
239 {
240
241 vop_rename_fail(ap);
242 return (EOPNOTSUPP);
243 }
244
245 /*
246 * vop_nostrategy:
247 *
248 * Strategy routine for VFS devices that have none.
249 *
250 * BIO_ERROR and B_INVAL must be cleared prior to calling any strategy
251 * routine. Typically this is done for a BIO_READ strategy call.
252 * Typically B_INVAL is assumed to already be clear prior to a write
253 * and should not be cleared manually unless you just made the buffer
254 * invalid. BIO_ERROR should be cleared either way.
255 */
256
257 static int
258 vop_nostrategy (struct vop_strategy_args *ap)
259 {
260 printf("No strategy for buffer at %p\n", ap->a_bp);
261 vn_printf(ap->a_vp, "vnode ");
262 ap->a_bp->b_ioflags |= BIO_ERROR;
263 ap->a_bp->b_error = EOPNOTSUPP;
264 bufdone(ap->a_bp);
265 return (EOPNOTSUPP);
266 }
267
268 static int
269 get_next_dirent(struct vnode *vp, struct dirent **dpp, char *dirbuf,
270 int dirbuflen, off_t *off, char **cpos, int *len,
271 int *eofflag, struct thread *td)
272 {
273 int error, reclen;
274 struct uio uio;
275 struct iovec iov;
276 struct dirent *dp;
277
278 KASSERT(VOP_ISLOCKED(vp), ("vp %p is not locked", vp));
279 KASSERT(vp->v_type == VDIR, ("vp %p is not a directory", vp));
280
281 if (*len == 0) {
282 iov.iov_base = dirbuf;
283 iov.iov_len = dirbuflen;
284
285 uio.uio_iov = &iov;
286 uio.uio_iovcnt = 1;
287 uio.uio_offset = *off;
288 uio.uio_resid = dirbuflen;
289 uio.uio_segflg = UIO_SYSSPACE;
290 uio.uio_rw = UIO_READ;
291 uio.uio_td = td;
292
293 *eofflag = 0;
294
295 #ifdef MAC
296 error = mac_vnode_check_readdir(td->td_ucred, vp);
297 if (error == 0)
298 #endif
299 error = VOP_READDIR(vp, &uio, td->td_ucred, eofflag,
300 NULL, NULL);
301 if (error)
302 return (error);
303
304 *off = uio.uio_offset;
305
306 *cpos = dirbuf;
307 *len = (dirbuflen - uio.uio_resid);
308
309 if (*len == 0)
310 return (ENOENT);
311 }
312
313 dp = (struct dirent *)(*cpos);
314 reclen = dp->d_reclen;
315 *dpp = dp;
316
317 /* check for malformed directory.. */
318 if (reclen < DIRENT_MINSIZE)
319 return (EINVAL);
320
321 *cpos += reclen;
322 *len -= reclen;
323
324 return (0);
325 }
326
327 /*
328 * Check if a named file exists in a given directory vnode.
329 */
330 static int
331 dirent_exists(struct vnode *vp, const char *dirname, struct thread *td)
332 {
333 char *dirbuf, *cpos;
334 int error, eofflag, dirbuflen, len, found;
335 off_t off;
336 struct dirent *dp;
337 struct vattr va;
338
339 KASSERT(VOP_ISLOCKED(vp), ("vp %p is not locked", vp));
340 KASSERT(vp->v_type == VDIR, ("vp %p is not a directory", vp));
341
342 found = 0;
343
344 error = VOP_GETATTR(vp, &va, td->td_ucred);
345 if (error)
346 return (found);
347
348 dirbuflen = DEV_BSIZE;
349 if (dirbuflen < va.va_blocksize)
350 dirbuflen = va.va_blocksize;
351 dirbuf = (char *)malloc(dirbuflen, M_TEMP, M_WAITOK);
352
353 off = 0;
354 len = 0;
355 do {
356 error = get_next_dirent(vp, &dp, dirbuf, dirbuflen, &off,
357 &cpos, &len, &eofflag, td);
358 if (error)
359 goto out;
360
361 if (dp->d_type != DT_WHT && dp->d_fileno != 0 &&
362 strcmp(dp->d_name, dirname) == 0) {
363 found = 1;
364 goto out;
365 }
366 } while (len > 0 || !eofflag);
367
368 out:
369 free(dirbuf, M_TEMP);
370 return (found);
371 }
372
373 int
374 vop_stdaccess(struct vop_access_args *ap)
375 {
376
377 KASSERT((ap->a_accmode & ~(VEXEC | VWRITE | VREAD | VADMIN |
378 VAPPEND)) == 0, ("invalid bit in accmode"));
379
380 return (VOP_ACCESSX(ap->a_vp, ap->a_accmode, ap->a_cred, ap->a_td));
381 }
382
383 int
384 vop_stdaccessx(struct vop_accessx_args *ap)
385 {
386 int error;
387 accmode_t accmode = ap->a_accmode;
388
389 error = vfs_unixify_accmode(&accmode);
390 if (error != 0)
391 return (error);
392
393 if (accmode == 0)
394 return (0);
395
396 return (VOP_ACCESS(ap->a_vp, accmode, ap->a_cred, ap->a_td));
397 }
398
399 /*
400 * Advisory record locking support
401 */
402 int
403 vop_stdadvlock(struct vop_advlock_args *ap)
404 {
405 struct vnode *vp;
406 struct vattr vattr;
407 int error;
408
409 vp = ap->a_vp;
410 if (ap->a_fl->l_whence == SEEK_END) {
411 /*
412 * The NFSv4 server must avoid doing a vn_lock() here, since it
413 * can deadlock the nfsd threads, due to a LOR. Fortunately
414 * the NFSv4 server always uses SEEK_SET and this code is
415 * only required for the SEEK_END case.
416 */
417 vn_lock(vp, LK_SHARED | LK_RETRY);
418 error = VOP_GETATTR(vp, &vattr, curthread->td_ucred);
419 VOP_UNLOCK(vp, 0);
420 if (error)
421 return (error);
422 } else
423 vattr.va_size = 0;
424
425 return (lf_advlock(ap, &(vp->v_lockf), vattr.va_size));
426 }
427
428 int
429 vop_stdadvlockasync(struct vop_advlockasync_args *ap)
430 {
431 struct vnode *vp;
432 struct vattr vattr;
433 int error;
434
435 vp = ap->a_vp;
436 if (ap->a_fl->l_whence == SEEK_END) {
437 /* The size argument is only needed for SEEK_END. */
438 vn_lock(vp, LK_SHARED | LK_RETRY);
439 error = VOP_GETATTR(vp, &vattr, curthread->td_ucred);
440 VOP_UNLOCK(vp, 0);
441 if (error)
442 return (error);
443 } else
444 vattr.va_size = 0;
445
446 return (lf_advlockasync(ap, &(vp->v_lockf), vattr.va_size));
447 }
448
449 int
450 vop_stdadvlockpurge(struct vop_advlockpurge_args *ap)
451 {
452 struct vnode *vp;
453
454 vp = ap->a_vp;
455 lf_purgelocks(vp, &vp->v_lockf);
456 return (0);
457 }
458
459 /*
460 * vop_stdpathconf:
461 *
462 * Standard implementation of POSIX pathconf, to get information about limits
463 * for a filesystem.
464 * Override per filesystem for the case where the filesystem has smaller
465 * limits.
466 */
467 int
468 vop_stdpathconf(ap)
469 struct vop_pathconf_args /* {
470 struct vnode *a_vp;
471 int a_name;
472 int *a_retval;
473 } */ *ap;
474 {
475
476 switch (ap->a_name) {
477 case _PC_ASYNC_IO:
478 *ap->a_retval = _POSIX_ASYNCHRONOUS_IO;
479 return (0);
480 case _PC_NAME_MAX:
481 *ap->a_retval = NAME_MAX;
482 return (0);
483 case _PC_PATH_MAX:
484 *ap->a_retval = PATH_MAX;
485 return (0);
486 case _PC_LINK_MAX:
487 *ap->a_retval = LINK_MAX;
488 return (0);
489 case _PC_MAX_CANON:
490 *ap->a_retval = MAX_CANON;
491 return (0);
492 case _PC_MAX_INPUT:
493 *ap->a_retval = MAX_INPUT;
494 return (0);
495 case _PC_PIPE_BUF:
496 *ap->a_retval = PIPE_BUF;
497 return (0);
498 case _PC_CHOWN_RESTRICTED:
499 *ap->a_retval = 1;
500 return (0);
501 case _PC_VDISABLE:
502 *ap->a_retval = _POSIX_VDISABLE;
503 return (0);
504 default:
505 return (EINVAL);
506 }
507 /* NOTREACHED */
508 }
509
510 /*
511 * Standard lock, unlock and islocked functions.
512 */
513 int
514 vop_stdlock(ap)
515 struct vop_lock1_args /* {
516 struct vnode *a_vp;
517 int a_flags;
518 char *file;
519 int line;
520 } */ *ap;
521 {
522 struct vnode *vp = ap->a_vp;
523 struct mtx *ilk;
524
525 ilk = VI_MTX(vp);
526 return (lockmgr_lock_fast_path(vp->v_vnlock, ap->a_flags,
527 (ilk != NULL) ? &ilk->lock_object : NULL, ap->a_file, ap->a_line));
528 }
529
530 /* See above. */
531 int
532 vop_stdunlock(ap)
533 struct vop_unlock_args /* {
534 struct vnode *a_vp;
535 int a_flags;
536 } */ *ap;
537 {
538 struct vnode *vp = ap->a_vp;
539 struct mtx *ilk;
540
541 ilk = VI_MTX(vp);
542 return (lockmgr_unlock_fast_path(vp->v_vnlock, ap->a_flags,
543 (ilk != NULL) ? &ilk->lock_object : NULL));
544 }
545
546 /* See above. */
547 int
548 vop_stdislocked(ap)
549 struct vop_islocked_args /* {
550 struct vnode *a_vp;
551 } */ *ap;
552 {
553
554 return (lockstatus(ap->a_vp->v_vnlock));
555 }
556
557 /*
558 * Return true for select/poll.
559 */
560 int
561 vop_nopoll(ap)
562 struct vop_poll_args /* {
563 struct vnode *a_vp;
564 int a_events;
565 struct ucred *a_cred;
566 struct thread *a_td;
567 } */ *ap;
568 {
569
570 return (poll_no_poll(ap->a_events));
571 }
572
573 /*
574 * Implement poll for local filesystems that support it.
575 */
576 int
577 vop_stdpoll(ap)
578 struct vop_poll_args /* {
579 struct vnode *a_vp;
580 int a_events;
581 struct ucred *a_cred;
582 struct thread *a_td;
583 } */ *ap;
584 {
585 if (ap->a_events & ~POLLSTANDARD)
586 return (vn_pollrecord(ap->a_vp, ap->a_td, ap->a_events));
587 return (ap->a_events & (POLLIN | POLLOUT | POLLRDNORM | POLLWRNORM));
588 }
589
590 /*
591 * Return our mount point, as we will take charge of the writes.
592 */
593 int
594 vop_stdgetwritemount(ap)
595 struct vop_getwritemount_args /* {
596 struct vnode *a_vp;
597 struct mount **a_mpp;
598 } */ *ap;
599 {
600 struct mount *mp;
601
602 /*
603 * XXX Since this is called unlocked we may be recycled while
604 * attempting to ref the mount. If this is the case or mountpoint
605 * will be set to NULL. We only have to prevent this call from
606 * returning with a ref to an incorrect mountpoint. It is not
607 * harmful to return with a ref to our previous mountpoint.
608 */
609 mp = ap->a_vp->v_mount;
610 if (mp != NULL) {
611 vfs_ref(mp);
612 if (mp != ap->a_vp->v_mount) {
613 vfs_rel(mp);
614 mp = NULL;
615 }
616 }
617 *(ap->a_mpp) = mp;
618 return (0);
619 }
620
621 /* XXX Needs good comment and VOP_BMAP(9) manpage */
622 int
623 vop_stdbmap(ap)
624 struct vop_bmap_args /* {
625 struct vnode *a_vp;
626 daddr_t a_bn;
627 struct bufobj **a_bop;
628 daddr_t *a_bnp;
629 int *a_runp;
630 int *a_runb;
631 } */ *ap;
632 {
633
634 if (ap->a_bop != NULL)
635 *ap->a_bop = &ap->a_vp->v_bufobj;
636 if (ap->a_bnp != NULL)
637 *ap->a_bnp = ap->a_bn * btodb(ap->a_vp->v_mount->mnt_stat.f_iosize);
638 if (ap->a_runp != NULL)
639 *ap->a_runp = 0;
640 if (ap->a_runb != NULL)
641 *ap->a_runb = 0;
642 return (0);
643 }
644
645 int
646 vop_stdfsync(ap)
647 struct vop_fsync_args /* {
648 struct vnode *a_vp;
649 int a_waitfor;
650 struct thread *a_td;
651 } */ *ap;
652 {
653 struct vnode *vp = ap->a_vp;
654 struct buf *bp;
655 struct bufobj *bo;
656 struct buf *nbp;
657 int error = 0;
658 int maxretry = 1000; /* large, arbitrarily chosen */
659
660 bo = &vp->v_bufobj;
661 BO_LOCK(bo);
662 loop1:
663 /*
664 * MARK/SCAN initialization to avoid infinite loops.
665 */
666 TAILQ_FOREACH(bp, &bo->bo_dirty.bv_hd, b_bobufs) {
667 bp->b_vflags &= ~BV_SCANNED;
668 bp->b_error = 0;
669 }
670
671 /*
672 * Flush all dirty buffers associated with a vnode.
673 */
674 loop2:
675 TAILQ_FOREACH_SAFE(bp, &bo->bo_dirty.bv_hd, b_bobufs, nbp) {
676 if ((bp->b_vflags & BV_SCANNED) != 0)
677 continue;
678 bp->b_vflags |= BV_SCANNED;
679 if (BUF_LOCK(bp, LK_EXCLUSIVE | LK_NOWAIT, NULL)) {
680 if (ap->a_waitfor != MNT_WAIT)
681 continue;
682 if (BUF_LOCK(bp,
683 LK_EXCLUSIVE | LK_INTERLOCK | LK_SLEEPFAIL,
684 BO_LOCKPTR(bo)) != 0) {
685 BO_LOCK(bo);
686 goto loop1;
687 }
688 BO_LOCK(bo);
689 }
690 BO_UNLOCK(bo);
691 KASSERT(bp->b_bufobj == bo,
692 ("bp %p wrong b_bufobj %p should be %p",
693 bp, bp->b_bufobj, bo));
694 if ((bp->b_flags & B_DELWRI) == 0)
695 panic("fsync: not dirty");
696 if ((vp->v_object != NULL) && (bp->b_flags & B_CLUSTEROK)) {
697 vfs_bio_awrite(bp);
698 } else {
699 bremfree(bp);
700 bawrite(bp);
701 }
702 BO_LOCK(bo);
703 goto loop2;
704 }
705
706 /*
707 * If synchronous the caller expects us to completely resolve all
708 * dirty buffers in the system. Wait for in-progress I/O to
709 * complete (which could include background bitmap writes), then
710 * retry if dirty blocks still exist.
711 */
712 if (ap->a_waitfor == MNT_WAIT) {
713 bufobj_wwait(bo, 0, 0);
714 if (bo->bo_dirty.bv_cnt > 0) {
715 /*
716 * If we are unable to write any of these buffers
717 * then we fail now rather than trying endlessly
718 * to write them out.
719 */
720 TAILQ_FOREACH(bp, &bo->bo_dirty.bv_hd, b_bobufs)
721 if ((error = bp->b_error) == 0)
722 continue;
723 if (error == 0 && --maxretry >= 0)
724 goto loop1;
725 error = EAGAIN;
726 }
727 }
728 BO_UNLOCK(bo);
729 if (error == EAGAIN)
730 vn_printf(vp, "fsync: giving up on dirty ");
731
732 return (error);
733 }
734
735 static int
736 vop_stdfdatasync(struct vop_fdatasync_args *ap)
737 {
738
739 return (VOP_FSYNC(ap->a_vp, MNT_WAIT, ap->a_td));
740 }
741
742 int
743 vop_stdfdatasync_buf(struct vop_fdatasync_args *ap)
744 {
745 struct vop_fsync_args apf;
746
747 apf.a_vp = ap->a_vp;
748 apf.a_waitfor = MNT_WAIT;
749 apf.a_td = ap->a_td;
750 return (vop_stdfsync(&apf));
751 }
752
753 /* XXX Needs good comment and more info in the manpage (VOP_GETPAGES(9)). */
754 int
755 vop_stdgetpages(ap)
756 struct vop_getpages_args /* {
757 struct vnode *a_vp;
758 vm_page_t *a_m;
759 int a_count;
760 int *a_rbehind;
761 int *a_rahead;
762 } */ *ap;
763 {
764
765 return vnode_pager_generic_getpages(ap->a_vp, ap->a_m,
766 ap->a_count, ap->a_rbehind, ap->a_rahead, NULL, NULL);
767 }
768
769 static int
770 vop_stdgetpages_async(struct vop_getpages_async_args *ap)
771 {
772 int error;
773
774 error = VOP_GETPAGES(ap->a_vp, ap->a_m, ap->a_count, ap->a_rbehind,
775 ap->a_rahead);
776 ap->a_iodone(ap->a_arg, ap->a_m, ap->a_count, error);
777 return (error);
778 }
779
780 int
781 vop_stdkqfilter(struct vop_kqfilter_args *ap)
782 {
783 return vfs_kqfilter(ap);
784 }
785
786 /* XXX Needs good comment and more info in the manpage (VOP_PUTPAGES(9)). */
787 int
788 vop_stdputpages(ap)
789 struct vop_putpages_args /* {
790 struct vnode *a_vp;
791 vm_page_t *a_m;
792 int a_count;
793 int a_sync;
794 int *a_rtvals;
795 } */ *ap;
796 {
797
798 return vnode_pager_generic_putpages(ap->a_vp, ap->a_m, ap->a_count,
799 ap->a_sync, ap->a_rtvals);
800 }
801
802 int
803 vop_stdvptofh(struct vop_vptofh_args *ap)
804 {
805 return (EOPNOTSUPP);
806 }
807
808 int
809 vop_stdvptocnp(struct vop_vptocnp_args *ap)
810 {
811 struct vnode *vp = ap->a_vp;
812 struct vnode **dvp = ap->a_vpp;
813 struct ucred *cred = ap->a_cred;
814 char *buf = ap->a_buf;
815 int *buflen = ap->a_buflen;
816 char *dirbuf, *cpos;
817 int i, error, eofflag, dirbuflen, flags, locked, len, covered;
818 off_t off;
819 ino_t fileno;
820 struct vattr va;
821 struct nameidata nd;
822 struct thread *td;
823 struct dirent *dp;
824 struct vnode *mvp;
825
826 i = *buflen;
827 error = 0;
828 covered = 0;
829 td = curthread;
830
831 if (vp->v_type != VDIR)
832 return (ENOENT);
833
834 error = VOP_GETATTR(vp, &va, cred);
835 if (error)
836 return (error);
837
838 VREF(vp);
839 locked = VOP_ISLOCKED(vp);
840 VOP_UNLOCK(vp, 0);
841 NDINIT_ATVP(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF, UIO_SYSSPACE,
842 "..", vp, td);
843 flags = FREAD;
844 error = vn_open_cred(&nd, &flags, 0, VN_OPEN_NOAUDIT, cred, NULL);
845 if (error) {
846 vn_lock(vp, locked | LK_RETRY);
847 return (error);
848 }
849 NDFREE(&nd, NDF_ONLY_PNBUF);
850
851 mvp = *dvp = nd.ni_vp;
852
853 if (vp->v_mount != (*dvp)->v_mount &&
854 ((*dvp)->v_vflag & VV_ROOT) &&
855 ((*dvp)->v_mount->mnt_flag & MNT_UNION)) {
856 *dvp = (*dvp)->v_mount->mnt_vnodecovered;
857 VREF(mvp);
858 VOP_UNLOCK(mvp, 0);
859 vn_close(mvp, FREAD, cred, td);
860 VREF(*dvp);
861 vn_lock(*dvp, LK_SHARED | LK_RETRY);
862 covered = 1;
863 }
864
865 fileno = va.va_fileid;
866
867 dirbuflen = DEV_BSIZE;
868 if (dirbuflen < va.va_blocksize)
869 dirbuflen = va.va_blocksize;
870 dirbuf = (char *)malloc(dirbuflen, M_TEMP, M_WAITOK);
871
872 if ((*dvp)->v_type != VDIR) {
873 error = ENOENT;
874 goto out;
875 }
876
877 off = 0;
878 len = 0;
879 do {
880 /* call VOP_READDIR of parent */
881 error = get_next_dirent(*dvp, &dp, dirbuf, dirbuflen, &off,
882 &cpos, &len, &eofflag, td);
883 if (error)
884 goto out;
885
886 if ((dp->d_type != DT_WHT) &&
887 (dp->d_fileno == fileno)) {
888 if (covered) {
889 VOP_UNLOCK(*dvp, 0);
890 vn_lock(mvp, LK_SHARED | LK_RETRY);
891 if (dirent_exists(mvp, dp->d_name, td)) {
892 error = ENOENT;
893 VOP_UNLOCK(mvp, 0);
894 vn_lock(*dvp, LK_SHARED | LK_RETRY);
895 goto out;
896 }
897 VOP_UNLOCK(mvp, 0);
898 vn_lock(*dvp, LK_SHARED | LK_RETRY);
899 }
900 i -= dp->d_namlen;
901
902 if (i < 0) {
903 error = ENOMEM;
904 goto out;
905 }
906 if (dp->d_namlen == 1 && dp->d_name[0] == '.') {
907 error = ENOENT;
908 } else {
909 bcopy(dp->d_name, buf + i, dp->d_namlen);
910 error = 0;
911 }
912 goto out;
913 }
914 } while (len > 0 || !eofflag);
915 error = ENOENT;
916
917 out:
918 free(dirbuf, M_TEMP);
919 if (!error) {
920 *buflen = i;
921 vref(*dvp);
922 }
923 if (covered) {
924 vput(*dvp);
925 vrele(mvp);
926 } else {
927 VOP_UNLOCK(mvp, 0);
928 vn_close(mvp, FREAD, cred, td);
929 }
930 vn_lock(vp, locked | LK_RETRY);
931 return (error);
932 }
933
934 int
935 vop_stdallocate(struct vop_allocate_args *ap)
936 {
937 #ifdef __notyet__
938 struct statfs *sfs;
939 off_t maxfilesize = 0;
940 #endif
941 struct iovec aiov;
942 struct vattr vattr, *vap;
943 struct uio auio;
944 off_t fsize, len, cur, offset;
945 uint8_t *buf;
946 struct thread *td;
947 struct vnode *vp;
948 size_t iosize;
949 int error;
950
951 buf = NULL;
952 error = 0;
953 td = curthread;
954 vap = &vattr;
955 vp = ap->a_vp;
956 len = *ap->a_len;
957 offset = *ap->a_offset;
958
959 error = VOP_GETATTR(vp, vap, td->td_ucred);
960 if (error != 0)
961 goto out;
962 fsize = vap->va_size;
963 iosize = vap->va_blocksize;
964 if (iosize == 0)
965 iosize = BLKDEV_IOSIZE;
966 if (iosize > MAXPHYS)
967 iosize = MAXPHYS;
968 buf = malloc(iosize, M_TEMP, M_WAITOK);
969
970 #ifdef __notyet__
971 /*
972 * Check if the filesystem sets f_maxfilesize; if not use
973 * VOP_SETATTR to perform the check.
974 */
975 sfs = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK);
976 error = VFS_STATFS(vp->v_mount, sfs, td);
977 if (error == 0)
978 maxfilesize = sfs->f_maxfilesize;
979 free(sfs, M_STATFS);
980 if (error != 0)
981 goto out;
982 if (maxfilesize) {
983 if (offset > maxfilesize || len > maxfilesize ||
984 offset + len > maxfilesize) {
985 error = EFBIG;
986 goto out;
987 }
988 } else
989 #endif
990 if (offset + len > vap->va_size) {
991 /*
992 * Test offset + len against the filesystem's maxfilesize.
993 */
994 VATTR_NULL(vap);
995 vap->va_size = offset + len;
996 error = VOP_SETATTR(vp, vap, td->td_ucred);
997 if (error != 0)
998 goto out;
999 VATTR_NULL(vap);
1000 vap->va_size = fsize;
1001 error = VOP_SETATTR(vp, vap, td->td_ucred);
1002 if (error != 0)
1003 goto out;
1004 }
1005
1006 for (;;) {
1007 /*
1008 * Read and write back anything below the nominal file
1009 * size. There's currently no way outside the filesystem
1010 * to know whether this area is sparse or not.
1011 */
1012 cur = iosize;
1013 if ((offset % iosize) != 0)
1014 cur -= (offset % iosize);
1015 if (cur > len)
1016 cur = len;
1017 if (offset < fsize) {
1018 aiov.iov_base = buf;
1019 aiov.iov_len = cur;
1020 auio.uio_iov = &aiov;
1021 auio.uio_iovcnt = 1;
1022 auio.uio_offset = offset;
1023 auio.uio_resid = cur;
1024 auio.uio_segflg = UIO_SYSSPACE;
1025 auio.uio_rw = UIO_READ;
1026 auio.uio_td = td;
1027 error = VOP_READ(vp, &auio, 0, td->td_ucred);
1028 if (error != 0)
1029 break;
1030 if (auio.uio_resid > 0) {
1031 bzero(buf + cur - auio.uio_resid,
1032 auio.uio_resid);
1033 }
1034 } else {
1035 bzero(buf, cur);
1036 }
1037
1038 aiov.iov_base = buf;
1039 aiov.iov_len = cur;
1040 auio.uio_iov = &aiov;
1041 auio.uio_iovcnt = 1;
1042 auio.uio_offset = offset;
1043 auio.uio_resid = cur;
1044 auio.uio_segflg = UIO_SYSSPACE;
1045 auio.uio_rw = UIO_WRITE;
1046 auio.uio_td = td;
1047
1048 error = VOP_WRITE(vp, &auio, 0, td->td_ucred);
1049 if (error != 0)
1050 break;
1051
1052 len -= cur;
1053 offset += cur;
1054 if (len == 0)
1055 break;
1056 if (should_yield())
1057 break;
1058 }
1059
1060 out:
1061 *ap->a_len = len;
1062 *ap->a_offset = offset;
1063 free(buf, M_TEMP);
1064 return (error);
1065 }
1066
1067 int
1068 vop_stdadvise(struct vop_advise_args *ap)
1069 {
1070 struct vnode *vp;
1071 struct bufobj *bo;
1072 daddr_t startn, endn;
1073 off_t start, end;
1074 int bsize, error;
1075
1076 vp = ap->a_vp;
1077 switch (ap->a_advice) {
1078 case POSIX_FADV_WILLNEED:
1079 /*
1080 * Do nothing for now. Filesystems should provide a
1081 * custom method which starts an asynchronous read of
1082 * the requested region.
1083 */
1084 error = 0;
1085 break;
1086 case POSIX_FADV_DONTNEED:
1087 error = 0;
1088 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1089 if (vp->v_iflag & VI_DOOMED) {
1090 VOP_UNLOCK(vp, 0);
1091 break;
1092 }
1093
1094 /*
1095 * Deactivate pages in the specified range from the backing VM
1096 * object. Pages that are resident in the buffer cache will
1097 * remain wired until their corresponding buffers are released
1098 * below.
1099 */
1100 if (vp->v_object != NULL) {
1101 start = trunc_page(ap->a_start);
1102 end = round_page(ap->a_end);
1103 VM_OBJECT_RLOCK(vp->v_object);
1104 vm_object_page_noreuse(vp->v_object, OFF_TO_IDX(start),
1105 OFF_TO_IDX(end));
1106 VM_OBJECT_RUNLOCK(vp->v_object);
1107 }
1108
1109 bo = &vp->v_bufobj;
1110 BO_RLOCK(bo);
1111 bsize = vp->v_bufobj.bo_bsize;
1112 startn = ap->a_start / bsize;
1113 endn = ap->a_end / bsize;
1114 error = bnoreuselist(&bo->bo_clean, bo, startn, endn);
1115 if (error == 0)
1116 error = bnoreuselist(&bo->bo_dirty, bo, startn, endn);
1117 BO_RUNLOCK(bo);
1118 VOP_UNLOCK(vp, 0);
1119 break;
1120 default:
1121 error = EINVAL;
1122 break;
1123 }
1124 return (error);
1125 }
1126
1127 int
1128 vop_stdunp_bind(struct vop_unp_bind_args *ap)
1129 {
1130
1131 ap->a_vp->v_socket = ap->a_socket;
1132 return (0);
1133 }
1134
1135 int
1136 vop_stdunp_connect(struct vop_unp_connect_args *ap)
1137 {
1138
1139 *ap->a_socket = ap->a_vp->v_socket;
1140 return (0);
1141 }
1142
1143 int
1144 vop_stdunp_detach(struct vop_unp_detach_args *ap)
1145 {
1146
1147 ap->a_vp->v_socket = NULL;
1148 return (0);
1149 }
1150
1151 static int
1152 vop_stdis_text(struct vop_is_text_args *ap)
1153 {
1154
1155 return ((ap->a_vp->v_vflag & VV_TEXT) != 0);
1156 }
1157
1158 static int
1159 vop_stdset_text(struct vop_set_text_args *ap)
1160 {
1161
1162 ap->a_vp->v_vflag |= VV_TEXT;
1163 return (0);
1164 }
1165
1166 static int
1167 vop_stdunset_text(struct vop_unset_text_args *ap)
1168 {
1169
1170 ap->a_vp->v_vflag &= ~VV_TEXT;
1171 return (0);
1172 }
1173
1174 static int
1175 vop_stdget_writecount(struct vop_get_writecount_args *ap)
1176 {
1177
1178 *ap->a_writecount = ap->a_vp->v_writecount;
1179 return (0);
1180 }
1181
1182 static int
1183 vop_stdadd_writecount(struct vop_add_writecount_args *ap)
1184 {
1185
1186 ap->a_vp->v_writecount += ap->a_inc;
1187 return (0);
1188 }
1189
1190 /*
1191 * vfs default ops
1192 * used to fill the vfs function table to get reasonable default return values.
1193 */
1194 int
1195 vfs_stdroot (mp, flags, vpp)
1196 struct mount *mp;
1197 int flags;
1198 struct vnode **vpp;
1199 {
1200
1201 return (EOPNOTSUPP);
1202 }
1203
1204 int
1205 vfs_stdstatfs (mp, sbp)
1206 struct mount *mp;
1207 struct statfs *sbp;
1208 {
1209
1210 return (EOPNOTSUPP);
1211 }
1212
1213 int
1214 vfs_stdquotactl (mp, cmds, uid, arg)
1215 struct mount *mp;
1216 int cmds;
1217 uid_t uid;
1218 void *arg;
1219 {
1220
1221 return (EOPNOTSUPP);
1222 }
1223
1224 int
1225 vfs_stdsync(mp, waitfor)
1226 struct mount *mp;
1227 int waitfor;
1228 {
1229 struct vnode *vp, *mvp;
1230 struct thread *td;
1231 int error, lockreq, allerror = 0;
1232
1233 td = curthread;
1234 lockreq = LK_EXCLUSIVE | LK_INTERLOCK;
1235 if (waitfor != MNT_WAIT)
1236 lockreq |= LK_NOWAIT;
1237 /*
1238 * Force stale buffer cache information to be flushed.
1239 */
1240 loop:
1241 MNT_VNODE_FOREACH_ALL(vp, mp, mvp) {
1242 if (vp->v_bufobj.bo_dirty.bv_cnt == 0) {
1243 VI_UNLOCK(vp);
1244 continue;
1245 }
1246 if ((error = vget(vp, lockreq, td)) != 0) {
1247 if (error == ENOENT) {
1248 MNT_VNODE_FOREACH_ALL_ABORT(mp, mvp);
1249 goto loop;
1250 }
1251 continue;
1252 }
1253 error = VOP_FSYNC(vp, waitfor, td);
1254 if (error)
1255 allerror = error;
1256 vput(vp);
1257 }
1258 return (allerror);
1259 }
1260
1261 int
1262 vfs_stdnosync (mp, waitfor)
1263 struct mount *mp;
1264 int waitfor;
1265 {
1266
1267 return (0);
1268 }
1269
1270 int
1271 vfs_stdvget (mp, ino, flags, vpp)
1272 struct mount *mp;
1273 ino_t ino;
1274 int flags;
1275 struct vnode **vpp;
1276 {
1277
1278 return (EOPNOTSUPP);
1279 }
1280
1281 int
1282 vfs_stdfhtovp (mp, fhp, flags, vpp)
1283 struct mount *mp;
1284 struct fid *fhp;
1285 int flags;
1286 struct vnode **vpp;
1287 {
1288
1289 return (EOPNOTSUPP);
1290 }
1291
1292 int
1293 vfs_stdinit (vfsp)
1294 struct vfsconf *vfsp;
1295 {
1296
1297 return (0);
1298 }
1299
1300 int
1301 vfs_stduninit (vfsp)
1302 struct vfsconf *vfsp;
1303 {
1304
1305 return(0);
1306 }
1307
1308 int
1309 vfs_stdextattrctl(mp, cmd, filename_vp, attrnamespace, attrname)
1310 struct mount *mp;
1311 int cmd;
1312 struct vnode *filename_vp;
1313 int attrnamespace;
1314 const char *attrname;
1315 {
1316
1317 if (filename_vp != NULL)
1318 VOP_UNLOCK(filename_vp, 0);
1319 return (EOPNOTSUPP);
1320 }
1321
1322 int
1323 vfs_stdsysctl(mp, op, req)
1324 struct mount *mp;
1325 fsctlop_t op;
1326 struct sysctl_req *req;
1327 {
1328
1329 return (EOPNOTSUPP);
1330 }
1331
1332 /* end of vfs default ops */
Cache object: a2f7003adc4dfb9b1c0e855e39938bbb
|